diff options
Diffstat (limited to 'storage/innobase/handler/ha_innodb.cc')
| -rw-r--r-- | storage/innobase/handler/ha_innodb.cc | 14601 |
1 files changed, 8001 insertions, 6600 deletions
diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc index a0e6a385b37..638b8d73770 100644 --- a/storage/innobase/handler/ha_innodb.cc +++ b/storage/innobase/handler/ha_innodb.cc @@ -33,95 +33,118 @@ this program; if not, write to the Free Software Foundation, Inc., *****************************************************************************/ -#define lower_case_file_system lower_case_file_system_server -#define mysql_unpacked_real_data_home mysql_unpacked_real_data_home_server -#include <sql_table.h> // explain_filename, nz2, EXPLAIN_PARTITIONS_AS_COMMENT, - // EXPLAIN_FILENAME_MAX_EXTRA_LENGTH - -#include <sql_acl.h> // PROCESS_ACL -#include <debug_sync.h> // DEBUG_SYNC -#include <my_base.h> // HA_OPTION_* +/** @file ha_innodb.cc */ + +#include "univ.i" + +/* Include necessary SQL headers */ +#include "ha_prototypes.h" +#include <debug_sync.h> +#include <gstream.h> +#include <log.h> #include <mysys_err.h> #include <innodb_priv.h> +#include <strfunc.h> +#include <sql_acl.h> +#include <sql_class.h> +#include <sql_show.h> +#include <sql_table.h> #include <table_cache.h> #include <my_check_opt.h> +#include <my_bitmap.h> +#include <mysql/service_thd_alloc.h> +#include <mysql/service_thd_wait.h> -#undef lower_case_file_system -#undef mysql_unpacked_real_data_home -MYSQL_PLUGIN_IMPORT extern my_bool lower_case_file_system; -MYSQL_PLUGIN_IMPORT extern char mysql_unpacked_real_data_home[]; - -#ifdef _WIN32 -#include <io.h> -#endif +// MYSQL_PLUGIN_IMPORT extern my_bool lower_case_file_system; +// MYSQL_PLUGIN_IMPORT extern char mysql_unpacked_real_data_home[]; #include <my_service_manager.h> - -/** @file ha_innodb.cc */ +#include <key.h> /* Include necessary InnoDB headers */ -#include "univ.i" +#include "btr0btr.h" +#include "btr0cur.h" +#include "btr0bulk.h" +#include "btr0sea.h" +#include "buf0dblwr.h" #include "buf0dump.h" -#include "buf0lru.h" #include "buf0flu.h" -#include "buf0dblwr.h" -#include "btr0sea.h" +#include "buf0lru.h" +#include "dict0boot.h" #include "btr0defragment.h" +#include "dict0crea.h" +#include "dict0dict.h" +#include "dict0stats.h" +#include "dict0stats_bg.h" +#include "fil0fil.h" +#include "fsp0fsp.h" +#include "fsp0space.h" +#include "fsp0sysspace.h" +#include "fts0fts.h" +#include "fts0plugin.h" +#include "fts0priv.h" +#include "fts0types.h" +#include "ibuf0ibuf.h" +#include "lock0lock.h" +#include "log0crypt.h" +#include "mem0mem.h" +#include "mtr0mtr.h" #include "os0file.h" #include "os0thread.h" -#include "srv0start.h" -#include "srv0srv.h" -#include "trx0roll.h" -#include "trx0trx.h" -#include "trx0sys.h" +#include "page0zip.h" +#include "pars0pars.h" #include "rem0types.h" +#include "row0import.h" #include "row0ins.h" +#include "row0merge.h" #include "row0mysql.h" +#include "row0quiesce.h" #include "row0sel.h" +#include "row0trunc.h" #include "row0upd.h" -#include "log0log.h" -#include "lock0lock.h" -#include "dict0crea.h" -#include "btr0cur.h" -#include "btr0btr.h" -#include "fsp0fsp.h" -#include "sync0sync.h" -#include "fil0fil.h" #include "fil0crypt.h" -#include "trx0xa.h" -#include "row0merge.h" -#include "dict0boot.h" -#include "dict0stats.h" -#include "dict0stats_bg.h" -#include "ha_prototypes.h" -#include "ut0mem.h" #include "ut0timer.h" -#include "ibuf0ibuf.h" -#include "dict0dict.h" #include "srv0mon.h" -#include "api0api.h" -#include "api0misc.h" -#include "pars0pars.h" -#include "fts0fts.h" -#include "fts0types.h" -#include "row0import.h" -#include "row0quiesce.h" +#include "srv0srv.h" +#include "srv0start.h" #ifdef UNIV_DEBUG #include "trx0purge.h" #endif /* UNIV_DEBUG */ -#include "fts0priv.h" -#include "page0zip.h" +#include "trx0roll.h" +#include "trx0sys.h" +#include "trx0trx.h" #include "fil0pagecompress.h" +#include "trx0xa.h" +#include "ut0mem.h" +#include "row0ext.h" #define thd_get_trx_isolation(X) ((enum_tx_isolation)thd_tx_isolation(X)) +extern "C" void thd_mark_transaction_to_rollback(MYSQL_THD thd, bool all); +unsigned long long thd_get_query_id(const MYSQL_THD thd); +TABLE *find_fk_open_table(THD *thd, const char *db, size_t db_len, + const char *table, size_t table_len); +MYSQL_THD create_thd(); +void destroy_thd(MYSQL_THD thd); +void reset_thd(MYSQL_THD thd); +TABLE *open_purge_table(THD *thd, const char *db, size_t dblen, + const char *tb, size_t tblen); +void close_thread_tables(THD* thd); + +/** Check if user has used xtradb extended system variable that +is not currently supported by innodb or marked as deprecated. */ +static +void +innodb_check_deprecated(void); + #ifdef MYSQL_DYNAMIC_PLUGIN -#define tc_size 2000 +#define tc_size 400 #define tdc_size 400 #endif #include "ha_innodb.h" #include "i_s.h" +#include "sync0sync.h" #include <string> #include <sstream> @@ -129,18 +152,11 @@ MYSQL_PLUGIN_IMPORT extern char mysql_unpacked_real_data_home[]; #include <mysql/plugin.h> #include <mysql/service_wsrep.h> -# ifndef MYSQL_PLUGIN_IMPORT -# define MYSQL_PLUGIN_IMPORT /* nothing */ -# endif /* MYSQL_PLUGIN_IMPORT */ - #ifdef WITH_WSREP #include "dict0priv.h" -#include "../storage/innobase/include/ut0byte.h" +#include "ut0byte.h" #include <mysql/service_md5.h> -class binlog_trx_data; -extern handlerton *binlog_hton; - extern MYSQL_PLUGIN_IMPORT MYSQL_BIN_LOG mysql_bin_log; static inline wsrep_ws_handle_t* @@ -160,90 +176,78 @@ static int innobase_wsrep_set_checkpoint(handlerton* hton, const XID* xid); static int innobase_wsrep_get_checkpoint(handlerton* hton, XID* xid); #endif /* WITH_WSREP */ -/** to protect innobase_open_files */ -static mysql_mutex_t innobase_share_mutex; /** to force correct commit order in binlog */ static ulong commit_threads = 0; static mysql_cond_t commit_cond; static mysql_mutex_t commit_cond_m; static mysql_mutex_t pending_checkpoint_mutex; -static bool innodb_inited = 0; #define INSIDE_HA_INNOBASE_CC #define EQ_CURRENT_THD(thd) ((thd) == current_thd) -static struct handlerton* innodb_hton_ptr; +struct handlerton* innodb_hton_ptr; static const long AUTOINC_OLD_STYLE_LOCKING = 0; static const long AUTOINC_NEW_STYLE_LOCKING = 1; static const long AUTOINC_NO_LOCKING = 2; -static long innobase_mirrored_log_groups; static long innobase_log_buffer_size; -static long innobase_additional_mem_pool_size; -static long innobase_file_io_threads; -static long innobase_open_files; +static long innobase_open_files=0; static long innobase_autoinc_lock_mode; static ulong innobase_commit_concurrency = 0; static ulong innobase_read_io_threads; static ulong innobase_write_io_threads; -static long innobase_buffer_pool_instances = 1; -static long long innobase_buffer_pool_size, innobase_log_file_size; +static long long innobase_buffer_pool_size; /** Percentage of the buffer pool to reserve for 'old' blocks. Connected to buf_LRU_old_ratio. */ static uint innobase_old_blocks_pct; -/** Maximum on-disk size of change buffer in terms of percentage -of the buffer pool. */ -static uint innobase_change_buffer_max_size = CHANGE_BUFFER_DEFAULT_SIZE; - /* The default values for the following char* start-up parameters are determined in innobase_init below: */ -static char* innobase_data_home_dir = NULL; -static char* innobase_data_file_path = NULL; -static char* innobase_file_format_name = NULL; -static char* innobase_change_buffering = NULL; -static char* innobase_enable_monitor_counter = NULL; -static char* innobase_disable_monitor_counter = NULL; -static char* innobase_reset_monitor_counter = NULL; -static char* innobase_reset_all_monitor_counter = NULL; +static char* innobase_data_home_dir; +static char* innobase_data_file_path; +static char* innobase_temp_data_file_path; +static char* innobase_file_format_name; +static char* innobase_change_buffering; +static char* innobase_enable_monitor_counter; +static char* innobase_disable_monitor_counter; +static char* innobase_reset_monitor_counter; +static char* innobase_reset_all_monitor_counter; /* The highest file format being used in the database. The value can be set by user, however, it will be adjusted to the newer file format if a table of such format is created/opened. */ -static char* innobase_file_format_max = NULL; +char* innobase_file_format_max; + +/** Default value of innodb_file_format */ +static const char* innodb_file_format_default = "Barracuda"; +/** Default value of innodb_file_format_max */ +static const char* innodb_file_format_max_default = "Antelope"; -static char* innobase_file_flush_method = NULL; +static char* innobase_file_flush_method; /* This variable can be set in the server configure file, specifying stopword table to be used */ -static char* innobase_server_stopword_table = NULL; +static char* innobase_server_stopword_table; /* Below we have boolean-valued start-up parameters, and their default values */ -static ulong innobase_fast_shutdown = 1; -static my_bool innobase_file_format_check = TRUE; -#ifdef UNIV_LOG_ARCHIVE -static my_bool innobase_log_archive = FALSE; -static char* innobase_log_arch_dir = NULL; -#endif /* UNIV_LOG_ARCHIVE */ -static my_bool innobase_use_atomic_writes = FALSE; -static my_bool innobase_use_fallocate = TRUE; -static my_bool innobase_use_doublewrite = TRUE; -static my_bool innobase_use_checksums = TRUE; -static my_bool innobase_locks_unsafe_for_binlog = FALSE; -static my_bool innobase_rollback_on_timeout = FALSE; -static my_bool innobase_create_status_file = FALSE; -static my_bool innobase_stats_on_metadata = TRUE; -static my_bool innobase_large_prefix = FALSE; -static my_bool innodb_optimize_fulltext_only = FALSE; - -static char* internal_innobase_data_file_path = NULL; +static my_bool innobase_file_format_check; +static my_bool innobase_use_atomic_writes; +static my_bool innobase_use_fallocate; +static my_bool innobase_use_doublewrite; +static my_bool innobase_use_checksums; +static my_bool innobase_locks_unsafe_for_binlog; +static my_bool innobase_rollback_on_timeout; +static my_bool innobase_create_status_file; +my_bool innobase_stats_on_metadata; +static my_bool innobase_large_prefix; +static my_bool innodb_optimize_fulltext_only; static char* innodb_version_str = (char*) INNODB_VERSION_STR; @@ -259,6 +263,123 @@ extern uint srv_background_scrub_data_check_interval; extern my_bool srv_scrub_force_testing; #endif +/** Note we cannot use rec_format_enum because we do not allow +COMPRESSED row format for innodb_default_row_format option. */ +enum default_row_format_enum { + DEFAULT_ROW_FORMAT_REDUNDANT = 0, + DEFAULT_ROW_FORMAT_COMPACT = 1, + DEFAULT_ROW_FORMAT_DYNAMIC = 2, +}; + +static +void set_my_errno(int err) +{ + errno = err; +} + +static uint mysql_fields(const TABLE *table) +{ + return table->s->frm_version < FRM_VER_EXPRESSSIONS + ? table->s->stored_fields : table->s->fields; +} + +/** Checks whether the file name belongs to a partition of a table. +@param[in] file_name file name +@return pointer to the end of the table name part of the file name, or NULL */ +static +char* +is_partition( +/*=========*/ + char* file_name) +{ + /* We look for pattern #P# to see if the table is partitioned + MariaDB table. */ + return strstr(file_name, table_name_t::part_suffix); +} + +/** Signal to shut down InnoDB (NULL if shutdown was signaled, or if +running in innodb_read_only mode, srv_read_only_mode) */ +st_my_thread_var *srv_running; +/** Service thread that waits for the server shutdown and stops purge threads. +Purge workers have THDs that are needed to calculate virtual columns. +This THDs must be destroyed rather early in the server shutdown sequence. +This service thread creates a THD and idly waits for it to get a signal to +die. Then it notifies all purge workers to shutdown. +*/ +static pthread_t thd_destructor_thread; + +pthread_handler_t +thd_destructor_proxy(void *) +{ + mysql_mutex_t thd_destructor_mutex; + mysql_cond_t thd_destructor_cond; + + my_thread_init(); + mysql_mutex_init(PSI_NOT_INSTRUMENTED, &thd_destructor_mutex, 0); + mysql_cond_init(PSI_NOT_INSTRUMENTED, &thd_destructor_cond, 0); + + st_my_thread_var *myvar= _my_thread_var(); + THD *thd= create_thd(); + thd_proc_info(thd, "InnoDB shutdown handler"); + + myvar->current_mutex = &thd_destructor_mutex; + myvar->current_cond = &thd_destructor_cond; + + mysql_mutex_lock(&thd_destructor_mutex); + my_atomic_storeptr_explicit(reinterpret_cast<void**>(&srv_running), + myvar, + MY_MEMORY_ORDER_RELAXED); + /* wait until the server wakes the THD to abort and die */ + while (!srv_running->abort) + mysql_cond_wait(&thd_destructor_cond, &thd_destructor_mutex); + mysql_mutex_unlock(&thd_destructor_mutex); + my_atomic_storeptr_explicit(reinterpret_cast<void**>(&srv_running), + NULL, + MY_MEMORY_ORDER_RELAXED); + + while (srv_fast_shutdown == 0 && + (trx_sys_any_active_transactions() || + (uint)thread_count > srv_n_purge_threads + 1)) { + thd_proc_info(thd, "InnoDB slow shutdown wait"); + os_thread_sleep(1000); + } + + /* Some background threads might generate undo pages that will + need to be purged, so they have to be shut down before purge + threads if slow shutdown is requested. */ + srv_shutdown_bg_undo_sources(); + srv_purge_shutdown(); + + destroy_thd(thd); + mysql_cond_destroy(&thd_destructor_cond); + mysql_mutex_destroy(&thd_destructor_mutex); + my_thread_end(); + return 0; +} + +/** Return the InnoDB ROW_FORMAT enum value +@param[in] row_format row_format from "innodb_default_row_format" +@return InnoDB ROW_FORMAT value from rec_format_t enum. */ +static +rec_format_t +get_row_format( + ulong row_format) +{ + switch(row_format) { + case DEFAULT_ROW_FORMAT_REDUNDANT: + return(REC_FORMAT_REDUNDANT); + case DEFAULT_ROW_FORMAT_COMPACT: + return(REC_FORMAT_COMPACT); + case DEFAULT_ROW_FORMAT_DYNAMIC: + return(REC_FORMAT_DYNAMIC); + default: + ut_ad(0); + return(REC_FORMAT_DYNAMIC); + } +} + +static ulong innodb_default_row_format = DEFAULT_ROW_FORMAT_DYNAMIC; + /** Possible values for system variable "innodb_stats_method". The values are defined the same as its corresponding MyISAM system variable "myisam_stats_method"(see "myisam_stats_method_names"), for better usability */ @@ -278,42 +399,26 @@ static TYPELIB innodb_stats_method_typelib = { NULL }; -/** Possible values for system variable "innodb_checksum_algorithm". */ -static const char* innodb_checksum_algorithm_names[] = { - "CRC32", - "STRICT_CRC32", - "INNODB", - "STRICT_INNODB", - "NONE", - "STRICT_NONE", +/** Possible values of the parameter innodb_checksum_algorithm */ +const char* innodb_checksum_algorithm_names[] = { + "crc32", + "strict_crc32", + "innodb", + "strict_innodb", + "none", + "strict_none", NullS }; /** Used to define an enumerate type of the system variable innodb_checksum_algorithm. */ -static TYPELIB innodb_checksum_algorithm_typelib = { +TYPELIB innodb_checksum_algorithm_typelib = { array_elements(innodb_checksum_algorithm_names) - 1, "innodb_checksum_algorithm_typelib", innodb_checksum_algorithm_names, NULL }; -/** Possible values of the parameter innodb_lock_schedule_algorithm */ -static const char* innodb_lock_schedule_algorithm_names[] = { - "fcfs", - "vats", - NullS -}; - -/** Used to define an enumerate type of the system variable -innodb_lock_schedule_algorithm. */ -static TYPELIB innodb_lock_schedule_algorithm_typelib = { - array_elements(innodb_lock_schedule_algorithm_names) - 1, - "innodb_lock_schedule_algorithm_typelib", - innodb_lock_schedule_algorithm_names, - NULL -}; - /** Possible values for system variable "innodb_default_row_format". */ static const char* innodb_default_row_format_names[] = { "redundant", @@ -331,6 +436,22 @@ static TYPELIB innodb_default_row_format_typelib = { NULL }; +/** Possible values of the parameter innodb_lock_schedule_algorithm */ +static const char* innodb_lock_schedule_algorithm_names[] = { + "fcfs", + "vats", + NullS +}; + +/** Used to define an enumerate type of the system variable +innodb_lock_schedule_algorithm. */ +static TYPELIB innodb_lock_schedule_algorithm_typelib = { + array_elements(innodb_lock_schedule_algorithm_names) - 1, + "innodb_lock_schedule_algorithm_typelib", + innodb_lock_schedule_algorithm_names, + NULL +}; + /* The following counter is used to convey information to InnoDB about server activity: in case of normal DML ops it is not sensible to call srv_active_wake_master_thread after each @@ -339,8 +460,6 @@ operation, we only do it every INNOBASE_WAKE_INTERVAL'th step. */ #define INNOBASE_WAKE_INTERVAL 32 static ulong innobase_active_counter = 0; -static hash_table_t* innobase_open_tables; - /** Allowed values of innodb_change_buffering */ static const char* innobase_change_buffering_values[IBUF_USE_COUNT] = { "none", /* IBUF_USE_NONE */ @@ -351,34 +470,30 @@ static const char* innobase_change_buffering_values[IBUF_USE_COUNT] = { "all" /* IBUF_USE_ALL */ }; - -/** Note we cannot use rec_format_enum because we do not allow -COMPRESSED row format for innodb_default_row_format option. */ -enum default_row_format_enum { - DEFAULT_ROW_FORMAT_REDUNDANT = 0, - DEFAULT_ROW_FORMAT_COMPACT = 1, - DEFAULT_ROW_FORMAT_DYNAMIC = 2 -}; - -/** Convert an InnoDB ROW_FORMAT value. -@param[in] row_format row_format from "innodb_default_row_format" -@return converted ROW_FORMAT */ -static rec_format_t get_row_format(ulong row_format) -{ - switch (row_format) { - case DEFAULT_ROW_FORMAT_REDUNDANT: - return REC_FORMAT_REDUNDANT; - case DEFAULT_ROW_FORMAT_COMPACT: - return REC_FORMAT_COMPACT; - case DEFAULT_ROW_FORMAT_DYNAMIC: - return REC_FORMAT_DYNAMIC; - default: - ut_ad(0); - return REC_FORMAT_COMPACT; - } -} - -static ulong innodb_default_row_format; +/** Retrieve the FTS Relevance Ranking result for doc with doc_id +of m_prebuilt->fts_doc_id +@param[in,out] fts_hdl FTS handler +@return the relevance ranking value */ +static +float +innobase_fts_retrieve_ranking( + FT_INFO* fts_hdl); +/** Free the memory for the FTS handler +@param[in,out] fts_hdl FTS handler */ +static +void +innobase_fts_close_ranking( + FT_INFO* fts_hdl); +/** Find and Retrieve the FTS Relevance Ranking result for doc with doc_id +of m_prebuilt->fts_doc_id +@param[in,out] fts_hdl FTS handler +@return the relevance ranking value */ +static +float +innobase_fts_find_ranking( + FT_INFO* fts_hdl, + uchar*, + uint); /* Call back function array defined by MySQL and used to retrieve FTS results. */ @@ -388,26 +503,74 @@ const struct _ft_vft ft_vft_result = {NULL, innobase_fts_retrieve_ranking, NULL}; +/** @return version of the extended FTS API */ +static +uint +innobase_fts_get_version() +{ + /* Currently this doesn't make much sense as returning + HA_CAN_FULLTEXT_EXT automatically mean this version is supported. + This supposed to ease future extensions. */ + return(2); +} + +/** @return Which part of the extended FTS API is supported */ +static +ulonglong +innobase_fts_flags() +{ + return(FTS_ORDERED_RESULT | FTS_DOCID_IN_RESULT); +} + +/** Find and Retrieve the FTS doc_id for the current result row +@param[in,out] fts_hdl FTS handler +@return the document ID */ +static +ulonglong +innobase_fts_retrieve_docid( + FT_INFO_EXT* fts_hdl); + +/** Find and retrieve the size of the current result +@param[in,out] fts_hdl FTS handler +@return number of matching rows */ +static +ulonglong +innobase_fts_count_matches( + FT_INFO_EXT* fts_hdl) /*!< in: FTS handler */ +{ + NEW_FT_INFO* handle = reinterpret_cast<NEW_FT_INFO*>(fts_hdl); + + if (handle->ft_result->rankings_by_id != NULL) { + return(rbt_size(handle->ft_result->rankings_by_id)); + } else { + return(0); + } +} + const struct _ft_vft_ext ft_vft_ext_result = {innobase_fts_get_version, innobase_fts_flags, innobase_fts_retrieve_docid, innobase_fts_count_matches}; #ifdef HAVE_PSI_INTERFACE +# define PSI_KEY(n) {&n##_key, #n, 0} +/* All RWLOCK used in Innodb are SX-locks */ +# define PSI_RWLOCK_KEY(n) {&n##_key, #n, PSI_RWLOCK_FLAG_SX} + /* Keys to register pthread mutexes/cond in the current file with performance schema */ -static mysql_pfs_key_t innobase_share_mutex_key; static mysql_pfs_key_t commit_cond_mutex_key; static mysql_pfs_key_t commit_cond_key; static mysql_pfs_key_t pending_checkpoint_mutex_key; +static mysql_pfs_key_t thd_destructor_thread_key; static PSI_mutex_info all_pthread_mutexes[] = { - {&commit_cond_mutex_key, "commit_cond_mutex", 0}, - {&innobase_share_mutex_key, "innobase_share_mutex", 0} + PSI_KEY(commit_cond_mutex), + PSI_KEY(pending_checkpoint_mutex), }; static PSI_cond_info all_innodb_conds[] = { - {&commit_cond_key, "commit_cond", 0} + PSI_KEY(commit_cond) }; # ifdef UNIV_PFS_MUTEX @@ -415,76 +578,63 @@ static PSI_cond_info all_innodb_conds[] = { performance schema instrumented if "UNIV_PFS_MUTEX" is defined */ static PSI_mutex_info all_innodb_mutexes[] = { - {&autoinc_mutex_key, "autoinc_mutex", 0}, + PSI_KEY(autoinc_mutex), # ifndef PFS_SKIP_BUFFER_MUTEX_RWLOCK - {&buffer_block_mutex_key, "buffer_block_mutex", 0}, + PSI_KEY(buffer_block_mutex), # endif /* !PFS_SKIP_BUFFER_MUTEX_RWLOCK */ - {&buf_pool_mutex_key, "buf_pool_mutex", 0}, - {&buf_pool_zip_mutex_key, "buf_pool_zip_mutex", 0}, - {&cache_last_read_mutex_key, "cache_last_read_mutex", 0}, - {&dict_foreign_err_mutex_key, "dict_foreign_err_mutex", 0}, - {&dict_sys_mutex_key, "dict_sys_mutex", 0}, - {&file_format_max_mutex_key, "file_format_max_mutex", 0}, - {&fil_system_mutex_key, "fil_system_mutex", 0}, - {&flush_list_mutex_key, "flush_list_mutex", 0}, - {&fts_bg_threads_mutex_key, "fts_bg_threads_mutex", 0}, - {&fts_delete_mutex_key, "fts_delete_mutex", 0}, - {&fts_optimize_mutex_key, "fts_optimize_mutex", 0}, - {&fts_doc_id_mutex_key, "fts_doc_id_mutex", 0}, - {&fts_pll_tokenize_mutex_key, "fts_pll_tokenize_mutex", 0}, - {&log_flush_order_mutex_key, "log_flush_order_mutex", 0}, - {&hash_table_mutex_key, "hash_table_mutex", 0}, - {&ibuf_bitmap_mutex_key, "ibuf_bitmap_mutex", 0}, - {&ibuf_mutex_key, "ibuf_mutex", 0}, - {&ibuf_pessimistic_insert_mutex_key, - "ibuf_pessimistic_insert_mutex", 0}, -# ifndef HAVE_ATOMIC_BUILTINS - {&server_mutex_key, "server_mutex", 0}, -# endif /* !HAVE_ATOMIC_BUILTINS */ - {&log_sys_mutex_key, "log_sys_mutex", 0}, -# ifdef UNIV_MEM_DEBUG - {&mem_hash_mutex_key, "mem_hash_mutex", 0}, -# endif /* UNIV_MEM_DEBUG */ - {&mem_pool_mutex_key, "mem_pool_mutex", 0}, - {&mutex_list_mutex_key, "mutex_list_mutex", 0}, - {&page_zip_stat_per_index_mutex_key, "page_zip_stat_per_index_mutex", 0}, - {&purge_sys_bh_mutex_key, "purge_sys_bh_mutex", 0}, - {&recv_sys_mutex_key, "recv_sys_mutex", 0}, - {&recv_writer_mutex_key, "recv_writer_mutex", 0}, - {&rseg_mutex_key, "rseg_mutex", 0}, -# ifdef UNIV_SYNC_DEBUG - {&rw_lock_debug_mutex_key, "rw_lock_debug_mutex", 0}, -# endif /* UNIV_SYNC_DEBUG */ - {&rw_lock_list_mutex_key, "rw_lock_list_mutex", 0}, - {&rw_lock_mutex_key, "rw_lock_mutex", 0}, - {&srv_dict_tmpfile_mutex_key, "srv_dict_tmpfile_mutex", 0}, - {&srv_innodb_monitor_mutex_key, "srv_innodb_monitor_mutex", 0}, - {&srv_misc_tmpfile_mutex_key, "srv_misc_tmpfile_mutex", 0}, - {&srv_monitor_file_mutex_key, "srv_monitor_file_mutex", 0}, -# ifdef UNIV_SYNC_DEBUG - {&sync_thread_mutex_key, "sync_thread_mutex", 0}, -# endif /* UNIV_SYNC_DEBUG */ - {&buf_dblwr_mutex_key, "buf_dblwr_mutex", 0}, - {&trx_undo_mutex_key, "trx_undo_mutex", 0}, - {&srv_sys_mutex_key, "srv_sys_mutex", 0}, - {&lock_sys_mutex_key, "lock_mutex", 0}, - {&lock_sys_wait_mutex_key, "lock_wait_mutex", 0}, - {&trx_mutex_key, "trx_mutex", 0}, - {&srv_sys_tasks_mutex_key, "srv_threads_mutex", 0}, - /* mutex with os_fast_mutex_ interfaces */ + PSI_KEY(buf_pool_mutex), + PSI_KEY(buf_pool_zip_mutex), + PSI_KEY(cache_last_read_mutex), + PSI_KEY(dict_foreign_err_mutex), + PSI_KEY(dict_sys_mutex), + PSI_KEY(recalc_pool_mutex), + PSI_KEY(file_format_max_mutex), + PSI_KEY(fil_system_mutex), + PSI_KEY(flush_list_mutex), + PSI_KEY(fts_bg_threads_mutex), + PSI_KEY(fts_delete_mutex), + PSI_KEY(fts_optimize_mutex), + PSI_KEY(fts_doc_id_mutex), + PSI_KEY(log_flush_order_mutex), + PSI_KEY(hash_table_mutex), + PSI_KEY(ibuf_bitmap_mutex), + PSI_KEY(ibuf_mutex), + PSI_KEY(ibuf_pessimistic_insert_mutex), + PSI_KEY(log_sys_mutex), + PSI_KEY(log_sys_write_mutex), + PSI_KEY(mutex_list_mutex), + PSI_KEY(page_zip_stat_per_index_mutex), + PSI_KEY(purge_sys_pq_mutex), + PSI_KEY(recv_sys_mutex), + PSI_KEY(recv_writer_mutex), + PSI_KEY(redo_rseg_mutex), + PSI_KEY(noredo_rseg_mutex), +# ifdef UNIV_DEBUG + PSI_KEY(rw_lock_debug_mutex), +# endif /* UNIV_DEBUG */ + PSI_KEY(rw_lock_list_mutex), + PSI_KEY(rw_lock_mutex), + PSI_KEY(srv_innodb_monitor_mutex), + PSI_KEY(srv_misc_tmpfile_mutex), + PSI_KEY(srv_monitor_file_mutex), + PSI_KEY(buf_dblwr_mutex), + PSI_KEY(trx_undo_mutex), + PSI_KEY(trx_pool_mutex), + PSI_KEY(trx_pool_manager_mutex), + PSI_KEY(srv_sys_mutex), + PSI_KEY(lock_mutex), + PSI_KEY(lock_wait_mutex), + PSI_KEY(trx_mutex), + PSI_KEY(srv_threads_mutex), # ifndef PFS_SKIP_EVENT_MUTEX - {&event_os_mutex_key, "event_os_mutex", 0}, + PSI_KEY(event_mutex), # endif /* PFS_SKIP_EVENT_MUTEX */ - {&os_mutex_key, "os_mutex", 0}, -#ifndef HAVE_ATOMIC_BUILTINS - {&srv_conc_mutex_key, "srv_conc_mutex", 0}, -#endif /* !HAVE_ATOMIC_BUILTINS */ -#ifndef HAVE_ATOMIC_BUILTINS_64 - {&monitor_mutex_key, "monitor_mutex", 0}, -#endif /* !HAVE_ATOMIC_BUILTINS_64 */ - {&ut_list_mutex_key, "ut_list_mutex", 0}, - {&trx_sys_mutex_key, "trx_sys_mutex", 0}, - {&zip_pad_mutex_key, "zip_pad_mutex", 0}, + PSI_KEY(rtr_active_mutex), + PSI_KEY(rtr_match_mutex), + PSI_KEY(rtr_path_mutex), + PSI_KEY(rtr_ssn_mutex), + PSI_KEY(trx_sys_mutex), + PSI_KEY(zip_pad_mutex) }; # endif /* UNIV_PFS_MUTEX */ @@ -493,27 +643,24 @@ static PSI_mutex_info all_innodb_mutexes[] = { performance schema instrumented if "UNIV_PFS_RWLOCK" is defined */ static PSI_rwlock_info all_innodb_rwlocks[] = { -# ifdef UNIV_LOG_ARCHIVE - {&archive_lock_key, "archive_lock", 0}, -# endif /* UNIV_LOG_ARCHIVE */ - {&btr_search_latch_key, "btr_search_latch", 0}, + PSI_RWLOCK_KEY(btr_search_latch), # ifndef PFS_SKIP_BUFFER_MUTEX_RWLOCK - {&buf_block_lock_key, "buf_block_lock", 0}, + PSI_RWLOCK_KEY(buf_block_lock), # endif /* !PFS_SKIP_BUFFER_MUTEX_RWLOCK */ -# ifdef UNIV_SYNC_DEBUG - {&buf_block_debug_latch_key, "buf_block_debug_latch", 0}, -# endif /* UNIV_SYNC_DEBUG */ - {&dict_operation_lock_key, "dict_operation_lock", 0}, - {&fil_space_latch_key, "fil_space_latch", 0}, - {&checkpoint_lock_key, "checkpoint_lock", 0}, - {&fts_cache_rw_lock_key, "fts_cache_rw_lock", 0}, - {&fts_cache_init_rw_lock_key, "fts_cache_init_rw_lock", 0}, - {&trx_i_s_cache_lock_key, "trx_i_s_cache_lock", 0}, - {&trx_purge_latch_key, "trx_purge_latch", 0}, - {&index_tree_rw_lock_key, "index_tree_rw_lock", 0}, - {&index_online_log_key, "index_online_log", 0}, - {&dict_table_stats_key, "dict_table_stats", 0}, - {&hash_table_rw_lock_key, "hash_table_locks", 0} +# ifdef UNIV_DEBUG + PSI_RWLOCK_KEY(buf_block_debug_latch), +# endif /* UNIV_DEBUG */ + PSI_RWLOCK_KEY(dict_operation_lock), + PSI_RWLOCK_KEY(fil_space_latch), + PSI_RWLOCK_KEY(checkpoint_lock), + PSI_RWLOCK_KEY(fts_cache_rw_lock), + PSI_RWLOCK_KEY(fts_cache_init_rw_lock), + PSI_RWLOCK_KEY(trx_i_s_cache_lock), + PSI_RWLOCK_KEY(trx_purge_latch), + PSI_RWLOCK_KEY(index_tree_rw_lock), + PSI_RWLOCK_KEY(index_online_log), + PSI_RWLOCK_KEY(dict_table_stats), + PSI_RWLOCK_KEY(hash_table_locks) }; # endif /* UNIV_PFS_RWLOCK */ @@ -522,15 +669,23 @@ static PSI_rwlock_info all_innodb_rwlocks[] = { performance schema instrumented if "UNIV_PFS_THREAD" is defined */ static PSI_thread_info all_innodb_threads[] = { - {&trx_rollback_clean_thread_key, "trx_rollback_clean_thread", 0}, - {&io_handler_thread_key, "io_handler_thread", 0}, - {&srv_lock_timeout_thread_key, "srv_lock_timeout_thread", 0}, - {&srv_error_monitor_thread_key, "srv_error_monitor_thread", 0}, - {&srv_monitor_thread_key, "srv_monitor_thread", 0}, - {&srv_master_thread_key, "srv_master_thread", 0}, - {&srv_purge_thread_key, "srv_purge_thread", 0}, - {&buf_page_cleaner_thread_key, "page_cleaner_thread", 0}, - {&recv_writer_thread_key, "recv_writer_thread", 0} + PSI_KEY(buf_dump_thread), + PSI_KEY(dict_stats_thread), + PSI_KEY(io_handler_thread), + PSI_KEY(io_ibuf_thread), + PSI_KEY(io_log_thread), + PSI_KEY(io_read_thread), + PSI_KEY(io_write_thread), + PSI_KEY(page_cleaner_thread), + PSI_KEY(recv_writer_thread), + PSI_KEY(srv_error_monitor_thread), + PSI_KEY(srv_lock_timeout_thread), + PSI_KEY(srv_master_thread), + PSI_KEY(srv_monitor_thread), + PSI_KEY(srv_purge_thread), + PSI_KEY(srv_worker_thread), + PSI_KEY(trx_rollback_clean_thread), + PSI_KEY(thd_destructor_thread), }; # endif /* UNIV_PFS_THREAD */ @@ -538,152 +693,21 @@ static PSI_thread_info all_innodb_threads[] = { /* all_innodb_files array contains the type of files that are performance schema instrumented if "UNIV_PFS_IO" is defined */ static PSI_file_info all_innodb_files[] = { - {&innodb_file_data_key, "innodb_data_file", 0}, - {&innodb_file_log_key, "innodb_log_file", 0}, - {&innodb_file_temp_key, "innodb_temp_file", 0} + PSI_KEY(innodb_data_file), + PSI_KEY(innodb_log_file), + PSI_KEY(innodb_temp_file) }; # endif /* UNIV_PFS_IO */ #endif /* HAVE_PSI_INTERFACE */ -/** Always normalize table name to lower case on Windows */ -#ifdef __WIN__ -#define normalize_table_name(norm_name, name) \ - normalize_table_name_low(norm_name, name, TRUE) -#else -#define normalize_table_name(norm_name, name) \ - normalize_table_name_low(norm_name, name, FALSE) -#endif /* __WIN__ */ - -/** Set up InnoDB API callback function array */ -ib_cb_t innodb_api_cb[] = { - (ib_cb_t) ib_cursor_open_table, - (ib_cb_t) ib_cursor_read_row, - (ib_cb_t) ib_cursor_insert_row, - (ib_cb_t) ib_cursor_delete_row, - (ib_cb_t) ib_cursor_update_row, - (ib_cb_t) ib_cursor_moveto, - (ib_cb_t) ib_cursor_first, - (ib_cb_t) ib_cursor_next, - (ib_cb_t) ib_cursor_last, - (ib_cb_t) ib_cursor_set_match_mode, - (ib_cb_t) ib_sec_search_tuple_create, - (ib_cb_t) ib_clust_read_tuple_create, - (ib_cb_t) ib_tuple_delete, - (ib_cb_t) ib_tuple_copy, - (ib_cb_t) ib_tuple_read_u8, - (ib_cb_t) ib_tuple_write_u8, - (ib_cb_t) ib_tuple_read_u16, - (ib_cb_t) ib_tuple_write_u16, - (ib_cb_t) ib_tuple_read_u32, - (ib_cb_t) ib_tuple_write_u32, - (ib_cb_t) ib_tuple_read_u64, - (ib_cb_t) ib_tuple_write_u64, - (ib_cb_t) ib_tuple_read_i8, - (ib_cb_t) ib_tuple_write_i8, - (ib_cb_t) ib_tuple_read_i16, - (ib_cb_t) ib_tuple_write_i16, - (ib_cb_t) ib_tuple_read_i32, - (ib_cb_t) ib_tuple_write_i32, - (ib_cb_t) ib_tuple_read_i64, - (ib_cb_t) ib_tuple_write_i64, - (ib_cb_t) ib_tuple_get_n_cols, - (ib_cb_t) ib_col_set_value, - (ib_cb_t) ib_col_get_value, - (ib_cb_t) ib_col_get_meta, - (ib_cb_t) ib_trx_begin, - (ib_cb_t) ib_trx_commit, - (ib_cb_t) ib_trx_rollback, - (ib_cb_t) ib_trx_start, - (ib_cb_t) ib_trx_release, - (ib_cb_t) ib_trx_state, - (ib_cb_t) ib_cursor_lock, - (ib_cb_t) ib_cursor_close, - (ib_cb_t) ib_cursor_new_trx, - (ib_cb_t) ib_cursor_reset, - (ib_cb_t) ib_open_table_by_name, - (ib_cb_t) ib_col_get_name, - (ib_cb_t) ib_table_truncate, - (ib_cb_t) ib_cursor_open_index_using_name, - (ib_cb_t) ib_close_thd, - (ib_cb_t) ib_cfg_get_cfg, - (ib_cb_t) ib_cursor_set_memcached_sync, - (ib_cb_t) ib_cursor_set_cluster_access, - (ib_cb_t) ib_cursor_commit_trx, - (ib_cb_t) ib_cfg_trx_level, - (ib_cb_t) ib_tuple_get_n_user_cols, - (ib_cb_t) ib_cursor_set_lock_mode, - (ib_cb_t) ib_cursor_clear_trx, - (ib_cb_t) ib_get_idx_field_name, - (ib_cb_t) ib_trx_get_start_time, - (ib_cb_t) ib_cfg_bk_commit_interval, - (ib_cb_t) ib_cursor_stmt_begin, - (ib_cb_t) ib_trx_read_only -}; - -/** - Test a file path whether it is same as mysql data directory path. - - @param path null terminated character string - - @return - @retval TRUE The path is different from mysql data directory. - @retval FALSE The path is same as mysql data directory. -*/ -static bool is_mysql_datadir_path(const char *path) -{ - if (path == NULL) - return false; - - char mysql_data_dir[FN_REFLEN], path_dir[FN_REFLEN]; - convert_dirname(path_dir, path, NullS); - convert_dirname(mysql_data_dir, mysql_unpacked_real_data_home, NullS); - size_t mysql_data_home_len= dirname_length(mysql_data_dir); - size_t path_len = dirname_length(path_dir); - - if (path_len < mysql_data_home_len) - return true; - - if (!lower_case_file_system) - return(memcmp(mysql_data_dir, path_dir, mysql_data_home_len)); - - return(files_charset_info->coll->strnncoll(files_charset_info, - (uchar *) path_dir, path_len, - (uchar *) mysql_data_dir, - mysql_data_home_len, - TRUE)); - -} - - -static int mysql_tmpfile_path(const char *path, const char *prefix) -{ - DBUG_ASSERT(path != NULL); - DBUG_ASSERT((strlen(path) + strlen(prefix)) <= FN_REFLEN); - - char filename[FN_REFLEN]; - File fd = create_temp_file(filename, path, prefix, -#ifdef __WIN__ - O_BINARY | O_TRUNC | O_SEQUENTIAL | - O_SHORT_LIVED | -#endif /* __WIN__ */ - O_CREAT | O_EXCL | O_RDWR | O_TEMPORARY, - MYF(MY_WME)); - if (fd >= 0) { -#ifndef __WIN__ - /* - This can be removed once the following bug is fixed: - Bug #28903 create_temp_file() doesn't honor O_TEMPORARY option - (file not removed) (Unix) - */ - unlink(filename); -#endif /* !__WIN__ */ - } - - return fd; -} - static void innodb_remember_check_sysvar_funcs(); mysql_var_check_func check_sysvar_enum; +mysql_var_check_func check_sysvar_int; + +// should page compression be used by default for new tables +static MYSQL_THDVAR_BOOL(compression_default, PLUGIN_VAR_OPCMDARG, + "Is compression the default for new tables", + NULL, NULL, FALSE); static MYSQL_THDVAR_UINT(default_encryption_key_id, PLUGIN_VAR_RQCMDARG, "Default encryption key id used for table encryption.", @@ -702,12 +726,10 @@ ha_create_table_option innodb_table_option_list[]= { /* With this option user can enable page compression feature for the table */ - HA_TOPTION_BOOL("PAGE_COMPRESSED", page_compressed, 0), + HA_TOPTION_SYSVAR("PAGE_COMPRESSED", page_compressed, compression_default), /* With this option user can set zip compression level for page compression for this table*/ HA_TOPTION_NUMBER("PAGE_COMPRESSION_LEVEL", page_compression_level, 0, 1, 9, 1), - /* With this option user can enable atomic writes feature for this table */ - HA_TOPTION_ENUM("ATOMIC_WRITES", atomic_writes, "DEFAULT,ON,OFF", 0), /* With this option the user can enable encryption for the table */ HA_TOPTION_ENUM("ENCRYPTED", encryption, "DEFAULT,YES,NO", 0), /* With this option the user defines the key identifier using for the encryption */ @@ -731,6 +753,8 @@ innodb_stopword_table_validate( for update function */ struct st_mysql_value* value); /*!< in: incoming string */ +static bool is_mysql_datadir_path(const char *path); + /** Validate passed-in "value" is a valid directory name. This function is registered as a callback with MySQL. @param[in,out] thd thread handle @@ -781,6 +805,7 @@ innodb_tmpdir_validate( return(1); } + os_normalize_path(alter_tmp_dir); my_realpath(tmp_abs_path, alter_tmp_dir, 0); size_t tmp_abs_len = strlen(tmp_abs_path); @@ -833,10 +858,6 @@ innodb_tmpdir_validate( return(0); } -/** "GEN_CLUST_INDEX" is the name reserved for InnoDB default -system clustered index when there is no primary key. */ -const char innobase_index_reserve_name[] = "GEN_CLUST_INDEX"; - /******************************************************************//** Maps a MySQL trx isolation level code to the InnoDB isolation level code @return InnoDB isolation level */ @@ -846,15 +867,19 @@ innobase_map_isolation_level( /*=========================*/ enum_tx_isolation iso); /*!< in: MySQL isolation level code */ -/******************************************************************//** -Maps a MySQL trx isolation level code to the InnoDB isolation level code -@return InnoDB isolation level */ +/** Gets field offset for a field in a table. +@param[in] table MySQL table object +@param[in] field MySQL field object (from table->field array) +@return offset */ static inline -ulint -innobase_map_isolation_level( -/*=========================*/ - enum_tx_isolation iso); /*!< in: MySQL isolation level code - */ +uint +get_field_offset( + const TABLE* table, + const Field* field) +{ + return field->offset(table->record[0]); +} + /*************************************************************//** Check for a valid value of innobase_compression_algorithm. @@ -875,6 +900,7 @@ static ibool innodb_have_lz4=IF_LZ4(1, 0); static ibool innodb_have_lzma=IF_LZMA(1, 0); static ibool innodb_have_bzip2=IF_BZIP2(1, 0); static ibool innodb_have_snappy=IF_SNAPPY(1, 0); +static ibool innodb_have_punch_hole=IF_PUNCH_HOLE(1, 0); static int @@ -889,9 +915,40 @@ innodb_encrypt_tables_validate( static const char innobase_hton_name[]= "InnoDB"; +static const char* deprecated_innodb_support_xa + = "Using innodb_support_xa is deprecated and the" + " parameter may be removed in future releases."; + +static const char* deprecated_innodb_support_xa_off + = "Using innodb_support_xa is deprecated and the" + " parameter may be removed in future releases." + " Only innodb_support_xa=ON is allowed."; + +/** Update the session variable innodb_support_xa. +@param[in] thd current session +@param[in] var the system variable innodb_support_xa +@param[in,out] var_ptr the contents of the variable +@param[in] save the to-be-updated value */ +static +void +innodb_support_xa_update( + THD* thd, + struct st_mysql_sys_var* var, + void* var_ptr, + const void* save) +{ + my_bool innodb_support_xa = *static_cast<const my_bool*>(save); + + push_warning(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_WRONG_COMMAND, + innodb_support_xa + ? deprecated_innodb_support_xa + : deprecated_innodb_support_xa_off); +} + static MYSQL_THDVAR_BOOL(support_xa, PLUGIN_VAR_OPCMDARG, "Enable InnoDB support for the XA two-phase commit", - /* check_func */ NULL, /* update_func */ NULL, + /* check_func */ NULL, innodb_support_xa_update, /* default */ TRUE); static MYSQL_THDVAR_BOOL(table_locks, PLUGIN_VAR_OPCMDARG, @@ -901,7 +958,7 @@ static MYSQL_THDVAR_BOOL(table_locks, PLUGIN_VAR_OPCMDARG, static MYSQL_THDVAR_BOOL(strict_mode, PLUGIN_VAR_OPCMDARG, "Use strict mode when evaluating create options.", - NULL, NULL, FALSE); + NULL, NULL, TRUE); static MYSQL_THDVAR_BOOL(ft_enable_stopword, PLUGIN_VAR_OPCMDARG, "Create FTS index with stopword.", @@ -927,6 +984,8 @@ static SHOW_VAR innodb_status_variables[]= { (char*) &export_vars.innodb_buffer_pool_dump_status, SHOW_CHAR}, {"buffer_pool_load_status", (char*) &export_vars.innodb_buffer_pool_load_status, SHOW_CHAR}, + {"buffer_pool_resize_status", + (char*) &export_vars.innodb_buffer_pool_resize_status, SHOW_CHAR}, {"buffer_pool_pages_data", (char*) &export_vars.innodb_buffer_pool_pages_data, SHOW_LONG}, {"buffer_pool_bytes_data", @@ -981,8 +1040,6 @@ static SHOW_VAR innodb_status_variables[]= { (char*) &export_vars.innodb_dblwr_pages_written, SHOW_LONG}, {"dblwr_writes", (char*) &export_vars.innodb_dblwr_writes, SHOW_LONG}, - {"have_atomic_builtins", - (char*) &export_vars.innodb_have_atomic_builtins, SHOW_BOOL}, {"log_waits", (char*) &export_vars.innodb_log_waits, SHOW_LONG}, {"log_write_requests", @@ -1039,29 +1096,12 @@ static SHOW_VAR innodb_status_variables[]= { (char*) &export_vars.innodb_truncated_status_writes, SHOW_LONG}, {"available_undo_logs", (char*) &export_vars.innodb_available_undo_logs, SHOW_LONG}, -#ifdef UNIV_DEBUG - {"purge_trx_id_age", - (char*) &export_vars.innodb_purge_trx_id_age, SHOW_LONG}, - {"purge_view_trx_id_age", - (char*) &export_vars.innodb_purge_view_trx_id_age, SHOW_LONG}, -#endif /* UNIV_DEBUG */ + {"undo_truncations", + (char*) &export_vars.innodb_undo_truncations, SHOW_LONG}, + /* Status variables for page compression */ {"page_compression_saved", (char*) &export_vars.innodb_page_compression_saved, SHOW_LONGLONG}, - {"page_compression_trim_sect512", - (char*) &export_vars.innodb_page_compression_trim_sect512, SHOW_LONGLONG}, - {"page_compression_trim_sect1024", - (char*) &export_vars.innodb_page_compression_trim_sect1024, SHOW_LONGLONG}, - {"page_compression_trim_sect2048", - (char*) &export_vars.innodb_page_compression_trim_sect2048, SHOW_LONGLONG}, - {"page_compression_trim_sect4096", - (char*) &export_vars.innodb_page_compression_trim_sect4096, SHOW_LONGLONG}, - {"page_compression_trim_sect8192", - (char*) &export_vars.innodb_page_compression_trim_sect8192, SHOW_LONGLONG}, - {"page_compression_trim_sect16384", - (char*) &export_vars.innodb_page_compression_trim_sect16384, SHOW_LONGLONG}, - {"page_compression_trim_sect32768", - (char*) &export_vars.innodb_page_compression_trim_sect32768, SHOW_LONGLONG}, {"num_index_pages_written", (char*) &export_vars.innodb_index_pages_written, SHOW_LONGLONG}, {"num_non_index_pages_written", @@ -1070,8 +1110,6 @@ static SHOW_VAR innodb_status_variables[]= { (char*) &export_vars.innodb_pages_page_compressed, SHOW_LONGLONG}, {"num_page_compressed_trim_op", (char*) &export_vars.innodb_page_compressed_trim_op, SHOW_LONGLONG}, - {"num_page_compressed_trim_op_saved", - (char*) &export_vars.innodb_page_compressed_trim_op_saved, SHOW_LONGLONG}, {"num_pages_page_decompressed", (char*) &export_vars.innodb_pages_page_decompressed, SHOW_LONGLONG}, {"num_pages_page_compression_error", @@ -1090,6 +1128,8 @@ static SHOW_VAR innodb_status_variables[]= { (char*) &innodb_have_bzip2, SHOW_BOOL}, {"have_snappy", (char*) &innodb_have_snappy, SHOW_BOOL}, + {"have_punch_hole", + (char*) &innodb_have_punch_hole, SHOW_BOOL}, /* Defragmentation */ {"defragment_compression_failures", @@ -1165,37 +1205,23 @@ static SHOW_VAR innodb_status_variables[]= { {"scrub_background_page_split_failures_unknown", (char*) &export_vars.innodb_scrub_page_split_failures_unknown, SHOW_LONG}, + {"scrub_log", + (char*) &export_vars.innodb_scrub_log, + SHOW_LONGLONG}, {"encryption_num_key_requests", (char*) &export_vars.innodb_encryption_key_requests, SHOW_LONGLONG}, {NullS, NullS, SHOW_LONG} }; -/************************************************************************//** -Handling the shared INNOBASE_SHARE structure that is needed to provide table -locking. Register the table name if it doesn't exist in the hash table. */ -static -INNOBASE_SHARE* -get_share( -/*======*/ - const char* table_name); /*!< in: table to lookup */ - -/************************************************************************//** -Free the shared object that was registered with get_share(). */ -static -void -free_share( -/*=======*/ - INNOBASE_SHARE* share); /*!< in/own: share to free */ - /*****************************************************************//** Frees a possible InnoDB trx object associated with the current THD. -@return 0 or error number */ +@return 0 or error number */ static int innobase_close_connection( /*======================*/ - handlerton* hton, /*!< in/out: Innodb handlerton */ + handlerton* hton, /*!< in/out: InnoDB handlerton */ THD* thd); /*!< in: MySQL thread handle for which to close the connection */ @@ -1207,12 +1233,12 @@ static void innobase_commit_ordered(handlerton *hton, THD* thd, bool all); /*****************************************************************//** Commits a transaction in an InnoDB database or marks an SQL statement ended. -@return 0 */ +@return 0 */ static int innobase_commit( /*============*/ - handlerton* hton, /*!< in/out: Innodb handlerton */ + handlerton* hton, /*!< in/out: InnoDB handlerton */ THD* thd, /*!< in: MySQL thread handle of the user for whom the transaction should be committed */ @@ -1228,7 +1254,7 @@ static int innobase_rollback( /*==============*/ - handlerton* hton, /*!< in/out: Innodb handlerton */ + handlerton* hton, /*!< in/out: InnoDB handlerton */ THD* thd, /*!< in: handle to the MySQL thread of the user whose transaction should be rolled back */ @@ -1265,7 +1291,7 @@ innobase_rollback_to_savepoint_can_release_mdl( /*****************************************************************//** Sets a transaction savepoint. -@return always 0, that is, always succeeds */ +@return always 0, that is, always succeeds */ static int innobase_savepoint( @@ -1284,7 +1310,7 @@ static int innobase_release_savepoint( /*=======================*/ - handlerton* hton, /*!< in/out: handlerton for Innodb */ + handlerton* hton, /*!< in/out: handlerton for InnoDB */ THD* thd, /*!< in: handle to the MySQL thread of the user whose transaction's savepoint should be released */ @@ -1292,16 +1318,6 @@ innobase_release_savepoint( static void innobase_checkpoint_request(handlerton *hton, void *cookie); -/************************************************************************//** -Function for constructing an InnoDB table handler instance. */ -static -handler* -innobase_create_handler( -/*====================*/ - handlerton* hton, /*!< in/out: handlerton for Innodb */ - TABLE_SHARE* table, - MEM_ROOT* mem_root); - /** @brief Initialize the default value of innodb_commit_concurrency. Once InnoDB is running, the innodb_commit_concurrency must not change @@ -1316,19 +1332,15 @@ void innobase_commit_concurrency_init_default(); /*=======================================*/ -/** @brief Initialize the default and max value of innodb_undo_logs. - -Once InnoDB is running, the default value and the max value of -innodb_undo_logs must be equal to the available undo logs, -given by srv_available_undo_logs. */ +/** @brief Adjust some InnoDB startup parameters based on file contents +or innodb_page_size. */ static void -innobase_undo_logs_init_default_max(); -/*==================================*/ +innodb_params_adjust(); /************************************************************//** Validate the file format name and return its corresponding id. -@return valid file format id */ +@return valid file format id */ static uint innobase_file_format_name_lookup( @@ -1338,7 +1350,7 @@ innobase_file_format_name_lookup( /************************************************************//** Validate the file format check config parameters, as a side effect it sets the srv_max_file_format_at_startup variable. -@return the format_id if valid config value, otherwise, return -1 */ +@return the format_id if valid config value, otherwise, return -1 */ static int innobase_file_format_validate_and_set( @@ -1347,7 +1359,7 @@ innobase_file_format_validate_and_set( /*******************************************************************//** This function is used to prepare an X/Open XA distributed transaction. -@return 0 or error number */ +@return 0 or error number */ static int innobase_xa_prepare( @@ -1361,7 +1373,7 @@ innobase_xa_prepare( ended */ /*******************************************************************//** This function is used to recover X/Open XA distributed transactions. -@return number of prepared transactions stored in xid_list */ +@return number of prepared transactions stored in xid_list */ static int innobase_xa_recover( @@ -1372,7 +1384,7 @@ innobase_xa_recover( /*******************************************************************//** This function is used to commit one X/Open XA distributed transaction which is in the prepared state -@return 0 or error number */ +@return 0 or error number */ static int innobase_commit_by_xid( @@ -1383,7 +1395,7 @@ innobase_commit_by_xid( /*******************************************************************//** This function is used to rollback one X/Open XA distributed transaction which is in the prepared state -@return 0 or error number */ +@return 0 or error number */ static int innobase_rollback_by_xid( @@ -1391,55 +1403,18 @@ innobase_rollback_by_xid( handlerton* hton, /*!< in: InnoDB handlerton */ XID* xid); /*!< in: X/Open XA transaction identification */ -/*******************************************************************//** -Create a consistent view for a cursor based on current transaction -which is created if the corresponding MySQL thread still lacks one. -This consistent view is then used inside of MySQL when accessing records -using a cursor. -@return pointer to cursor view or NULL */ -static -void* -innobase_create_cursor_view( -/*========================*/ - handlerton* hton, /*!< in: innobase hton */ - THD* thd); /*!< in: user thread handle */ -/*******************************************************************//** -Set the given consistent cursor view to a transaction which is created -if the corresponding MySQL thread still lacks one. If the given -consistent cursor view is NULL global read view of a transaction is -restored to a transaction read view. */ -static -void -innobase_set_cursor_view( -/*=====================*/ - handlerton* hton, /*!< in: handlerton of Innodb */ - THD* thd, /*!< in: user thread handle */ - void* curview); /*!< in: Consistent cursor view to - be set */ -/*******************************************************************//** -Close the given consistent cursor view of a transaction and restore -global read view to a transaction read view. Transaction is created if the -corresponding MySQL thread still lacks one. */ -static -void -innobase_close_cursor_view( -/*=======================*/ - handlerton* hton, /*!< in: handlerton of Innodb */ - THD* thd, /*!< in: user thread handle */ - void* curview); /*!< in: Consistent read view to be - closed */ -/*****************************************************************//** -Removes all tables in the named database inside InnoDB. */ + +/** Remove all tables in the named database inside InnoDB. +@param[in] hton handlerton from InnoDB +@param[in] path Database path; Inside InnoDB the name of the last +directory in the path is used as the database name. +For example, in 'mysql/data/test' the database name is 'test'. */ static void innobase_drop_database( -/*===================*/ - handlerton* hton, /*!< in: handlerton of Innodb */ - char* path); /*!< in: database path; inside InnoDB - the name of the last directory in - the path is used as the database name: - for example, in 'mysql/data/test' the - database name is 'test' */ + handlerton* hton, + char* path); + /** Shut down the InnoDB storage engine. @return 0 */ static @@ -1451,24 +1426,66 @@ Creates an InnoDB transaction struct for the thd if it does not yet have one. Starts a new InnoDB transaction if a transaction is not yet started. And assigns a new snapshot for a consistent read if the transaction does not yet have one. -@return 0 */ +@return 0 */ static int innobase_start_trx_and_assign_read_view( /*====================================*/ - handlerton* hton, /* in: Innodb handlerton */ + handlerton* hton, /* in: InnoDB handlerton */ THD* thd); /* in: MySQL thread handle of the user for whom the transaction should be committed */ -/****************************************************************//** -Flushes InnoDB logs to disk and makes a checkpoint. Really, a commit flushes -the logs, and the name of this function should be innobase_checkpoint. -@return TRUE if error */ + +/** Flush InnoDB redo logs to the file system. +@param[in] hton InnoDB handlerton +@param[in] binlog_group_flush true if we got invoked by binlog +group commit during flush stage, false in other cases. +@return false */ static bool innobase_flush_logs( -/*================*/ - handlerton* hton); /*!< in: InnoDB handlerton */ + handlerton* hton, + bool binlog_group_flush) +{ + DBUG_ENTER("innobase_flush_logs"); + DBUG_ASSERT(hton == innodb_hton_ptr); + + if (srv_read_only_mode) { + DBUG_RETURN(false); + } + + /* If !binlog_group_flush, we got invoked by FLUSH LOGS or similar. + Else, we got invoked by binlog group commit during flush stage. */ + + if (binlog_group_flush && srv_flush_log_at_trx_commit == 0) { + /* innodb_flush_log_at_trx_commit=0 + (write and sync once per second). + Do not flush the redo log during binlog group commit. */ + DBUG_RETURN(false); + } + + /* Flush the redo log buffer to the redo log file. + Sync it to disc if we are in FLUSH LOGS, or if + innodb_flush_log_at_trx_commit=1 + (write and sync at each commit). */ + log_buffer_flush_to_disk(!binlog_group_flush + || srv_flush_log_at_trx_commit == 1); + + DBUG_RETURN(false); +} + +/** Flush InnoDB redo logs to the file system. +@param[in] hton InnoDB handlerton +@param[in] binlog_group_flush true if we got invoked by binlog +group commit during flush stage, false in other cases. +@return false */ +static +bool +innobase_flush_logs( + handlerton* hton) +{ + return innobase_flush_logs(hton, true); +} /************************************************************************//** Implements the SHOW ENGINE INNODB STATUS command. Sends the output of the @@ -1495,14 +1512,6 @@ innobase_show_status( stat_print_fn* stat_print, enum ha_stat_type stat_type); -/*****************************************************************//** -Commits a transaction in an InnoDB database. */ -static -void -innobase_commit_low( -/*================*/ - trx_t* trx); /*!< in: transaction handle */ - /****************************************************************//** Parse and enable InnoDB monitor counters during server startup. User can enable monitor counters/groups by specifying @@ -1514,23 +1523,28 @@ innodb_enable_monitor_at_startup( /*=============================*/ char* str); /*!< in: monitor counter enable list */ -/********************************************************************* -Normalizes a table name string. A normalized name consists of the -database name catenated to '/' and table name. An example: -test/mytable. On Windows normalization puts both the database name and the -table name always to lower case if "set_lower_case" is set to TRUE. */ +#ifdef MYSQL_STORE_FTS_DOC_ID +/** Store doc_id value into FTS_DOC_ID field +@param[in,out] tbl table containing FULLTEXT index +@param[in] doc_id FTS_DOC_ID value */ +static void -normalize_table_name_low( -/*=====================*/ - char* norm_name, /* out: normalized name as a - null-terminated string */ - const char* name, /* in: table name string */ - ibool set_lower_case); /* in: TRUE if we want to set - name to lower case */ +innobase_fts_store_docid( + TABLE* tbl, + ulonglong doc_id) +{ + my_bitmap_map* old_map + = dbug_tmp_use_all_columns(tbl, tbl->write_set); + + tbl->fts_doc_id_field->store(static_cast<longlong>(doc_id), true); + + dbug_tmp_restore_column_map(tbl->write_set, old_map); +} +#endif /*************************************************************//** Check for a valid value of innobase_commit_concurrency. -@return 0 for valid innodb_commit_concurrency */ +@return 0 for valid innodb_commit_concurrency */ static int innobase_commit_concurrency_validate( @@ -1575,15 +1589,15 @@ innobase_create_handler( /* General functions */ -/*************************************************************//** -Check that a page_size is correct for InnoDB. If correct, set the -associated page_size_shift which is the power of 2 for this page size. -@return an associated page_size_shift if valid, 0 if invalid. */ +/** Check that a page_size is correct for InnoDB. +If correct, set the associated page_size_shift which is the power of 2 +for this page size. +@param[in] page_size Page Size to evaluate +@return an associated page_size_shift if valid, 0 if invalid. */ inline -int +ulong innodb_page_size_validate( -/*======================*/ - ulong page_size) /*!< in: Page Size to evaluate */ + ulong page_size) { ulong n; @@ -1592,7 +1606,7 @@ innodb_page_size_validate( for (n = UNIV_PAGE_SIZE_SHIFT_MIN; n <= UNIV_PAGE_SIZE_SHIFT_MAX; n++) { - if (page_size == (ulong) (1 << n)) { + if (page_size == static_cast<ulong>(1 << n)) { DBUG_RETURN(n); } } @@ -1606,8 +1620,7 @@ server. Used in srv_conc_enter_innodb() to determine if the thread should be allowed to enter InnoDB - the replication thread is treated differently than other threads. Also used in srv_conc_force_exit_innodb(). -@return true if thd is the replication thread */ -UNIV_INTERN +@return true if thd is the replication thread */ ibool thd_is_replication_slave_thread( /*============================*/ @@ -1618,9 +1631,8 @@ thd_is_replication_slave_thread( /******************************************************************//** Returns true if transaction should be flagged as read-only. -@return true if the thd is marked as read-only */ -UNIV_INTERN -ibool +@return true if the thd is marked as read-only */ +bool thd_trx_is_read_only( /*=================*/ THD* thd) /*!< in: thread handle */ @@ -1628,11 +1640,62 @@ thd_trx_is_read_only( return(thd != 0 && thd_tx_is_read_only(thd)); } +static MYSQL_THDVAR_BOOL(background_thread, + PLUGIN_VAR_NOCMDOPT | PLUGIN_VAR_NOSYSVAR, + "Internal (not user visible) flag to mark " + "background purge threads", NULL, NULL, 0); + +/** Create a MYSQL_THD for a background thread and mark it as such. +@param name thread info for SHOW PROCESSLIST +@return new MYSQL_THD */ +MYSQL_THD +innobase_create_background_thd(const char* name) +/*============================*/ +{ + MYSQL_THD thd= create_thd(); + thd_proc_info(thd, name); + THDVAR(thd, background_thread) = true; + return thd; +} + + +/** Destroy a background purge thread THD. +@param[in] thd MYSQL_THD to destroy */ +void +innobase_destroy_background_thd( +/*============================*/ + MYSQL_THD thd) +{ + /* need to close the connection explicitly, the server won't do it + if innodb is in the PLUGIN_IS_DYING state */ + innobase_close_connection(innodb_hton_ptr, thd); + thd_set_ha_data(thd, innodb_hton_ptr, NULL); + destroy_thd(thd); +} + +/** Close opened tables, free memory, delete items for a MYSQL_THD. +@param[in] thd MYSQL_THD to reset */ +void +innobase_reset_background_thd(MYSQL_THD thd) +{ + if (!thd) { + thd = current_thd; + } + + ut_ad(thd); + ut_ad(THDVAR(thd, background_thread)); + + /* background purge thread */ + const char *proc_info= thd_proc_info(thd, "reset"); + reset_thd(thd); + thd_proc_info(thd, proc_info); +} + + /******************************************************************//** Check if the transaction is an auto-commit transaction. TRUE also implies that it is a SELECT (read-only) transaction. -@return true if the transaction is an auto commit read-only transaction. */ -UNIV_INTERN +@return true if the transaction is an auto commit read-only transaction. */ ibool thd_trx_is_auto_commit( /*===================*/ @@ -1645,19 +1708,38 @@ thd_trx_is_auto_commit( && thd_is_select(thd)); } +extern "C" time_t thd_start_time(const THD* thd); + /******************************************************************//** -Save some CPU by testing the value of srv_thread_concurrency in inline -functions. */ +Get the thread start time. +@return the thread start time in seconds since the epoch. */ +ulint +thd_start_time_in_secs( +/*===================*/ + THD* thd) /*!< in: thread handle, or NULL */ +{ + // FIXME: This function should be added to the server code. + //return(thd_start_time(thd)); + return(ulint(ut_time())); +} + +/** Enter InnoDB engine after checking the max number of user threads +allowed, else the thread is put into sleep. +@param[in,out] prebuilt row prebuilt handler */ static inline void innobase_srv_conc_enter_innodb( -/*===========================*/ - trx_t* trx) /*!< in: transaction handle */ + row_prebuilt_t* prebuilt) { #ifdef WITH_WSREP - if (wsrep_on(trx->mysql_thd) && - wsrep_thd_is_BF(trx->mysql_thd, FALSE)) return; + if (wsrep_on(prebuilt->trx->mysql_thd) && + wsrep_thd_is_BF(prebuilt->trx->mysql_thd, FALSE)) { + return; + } #endif /* WITH_WSREP */ + + trx_t* trx = prebuilt->trx; + if (srv_thread_concurrency) { if (trx->n_tickets_to_enter_innodb > 0) { @@ -1674,29 +1756,31 @@ innobase_srv_conc_enter_innodb( < srv_thread_concurrency, srv_replication_delay * 1000); - } else { - srv_conc_enter_innodb(trx); + } else { + srv_conc_enter_innodb(prebuilt); } } } -/******************************************************************//** -Note that the thread wants to leave InnoDB only if it doesn't have -any spare tickets. */ +/** Note that the thread wants to leave InnoDB only if it doesn't have +any spare tickets. +@param[in,out] m_prebuilt row prebuilt handler */ static inline void innobase_srv_conc_exit_innodb( -/*==========================*/ - trx_t* trx) /*!< in: transaction handle */ + row_prebuilt_t* prebuilt) { -#ifdef UNIV_SYNC_DEBUG - ut_ad(!sync_thread_levels_nonempty_trx(trx->has_search_latch)); -#endif /* UNIV_SYNC_DEBUG */ + ut_ad(!sync_check_iterate(sync_check())); + #ifdef WITH_WSREP - if (wsrep_on(trx->mysql_thd) && - wsrep_thd_is_BF(trx->mysql_thd, FALSE)) return; + if (wsrep_on(prebuilt->trx->mysql_thd) && + wsrep_thd_is_BF(prebuilt->trx->mysql_thd, FALSE)) { + return; + } #endif /* WITH_WSREP */ + trx_t* trx = prebuilt->trx; + /* This is to avoid making an unnecessary function call. */ if (trx->declared_to_be_inside_innodb && trx->n_tickets_to_enter_innodb == 0) { @@ -1713,9 +1797,7 @@ innobase_srv_conc_force_exit_innodb( /*================================*/ trx_t* trx) /*!< in: transaction handle */ { -#ifdef UNIV_SYNC_DEBUG - ut_ad(!sync_thread_levels_nonempty_trx(trx->has_search_latch)); -#endif /* UNIV_SYNC_DEBUG */ + ut_ad(!sync_check_iterate(sync_check())); /* This is to avoid making an unnecessary function call. */ if (trx->declared_to_be_inside_innodb) { @@ -1725,8 +1807,7 @@ innobase_srv_conc_force_exit_innodb( /******************************************************************//** Returns the NUL terminated value of glob_hostname. -@return pointer to glob_hostname. */ -UNIV_INTERN +@return pointer to glob_hostname. */ const char* server_get_hostname() /*=================*/ @@ -1739,8 +1820,7 @@ Returns true if the transaction this thread is processing has edited non-transactional tables. Used by the deadlock detector when deciding which transaction to rollback in case of a deadlock - we try to avoid rolling back transactions that have edited non-transactional tables. -@return true if non-transactional tables have been edited */ -UNIV_INTERN +@return true if non-transactional tables have been edited */ ibool thd_has_edited_nontrans_tables( /*===========================*/ @@ -1760,8 +1840,7 @@ thd_query_start_micro( /******************************************************************//** Returns true if the thread is executing a SELECT statement. -@return true if thd is executing SELECT */ -UNIV_INTERN +@return true if thd is executing SELECT */ ibool thd_is_select( /*==========*/ @@ -1771,43 +1850,8 @@ thd_is_select( } /******************************************************************//** -Returns true if the thread supports XA, -global value of innodb_supports_xa if thd is NULL. -@return true if thd has XA support */ -UNIV_INTERN -ibool -thd_supports_xa( -/*============*/ - THD* thd) /*!< in: thread handle, or NULL to query - the global innodb_supports_xa */ -{ - return(THDVAR(thd, support_xa)); -} - -/** Get the value of innodb_tmpdir. -@param[in] thd thread handle, or NULL to query - the global innodb_tmpdir. -@retval NULL if innodb_tmpdir="" */ -UNIV_INTERN -const char* -thd_innodb_tmpdir( - THD* thd) -{ -#ifdef UNIV_SYNC_DEBUG - ut_ad(!sync_thread_levels_nonempty_trx(false)); -#endif /* UNIV_SYNC_DEBUG */ - - const char* tmp_dir = THDVAR(thd, tmpdir); - if (tmp_dir != NULL && *tmp_dir == '\0') { - tmp_dir = NULL; - } - - return(tmp_dir); -} -/******************************************************************//** Returns the lock wait timeout for the current connection. -@return the lock wait timeout, in seconds */ -UNIV_INTERN +@return the lock wait timeout, in seconds */ ulong thd_lock_wait_timeout( /*==================*/ @@ -1821,7 +1865,6 @@ thd_lock_wait_timeout( /******************************************************************//** Set the time waited for the lock for the current query. */ -UNIV_INTERN void thd_set_lock_wait_time( /*===================*/ @@ -1833,22 +1876,40 @@ thd_set_lock_wait_time( } } -/********************************************************************//** -Obtain the InnoDB transaction of a MySQL thread. -@return reference to transaction pointer */ -MY_ATTRIBUTE((warn_unused_result, nonnull)) -static inline -trx_t*& -thd_to_trx( -/*=======*/ - THD* thd) /*!< in: MySQL thread */ +/** Get the value of innodb_tmpdir. +@param[in] thd thread handle, or NULL to query + the global innodb_tmpdir. +@retval NULL if innodb_tmpdir="" */ +const char* +thd_innodb_tmpdir( + THD* thd) +{ + ut_ad(!sync_check_iterate(sync_check())); + + const char* tmp_dir = THDVAR(thd, tmpdir); + + if (tmp_dir != NULL && *tmp_dir == '\0') { + tmp_dir = NULL; + } + + return(tmp_dir); +} + +/** Obtain the InnoDB transaction of a MySQL thread. +@param[in,out] thd thread handle +@return reference to transaction pointer */ +static trx_t* thd_to_trx(THD* thd) { - return(*(trx_t**) thd_ha_data(thd, innodb_hton_ptr)); + return *reinterpret_cast<trx_t**>(thd_ha_data(thd, innodb_hton_ptr)); } + #ifdef WITH_WSREP +/********************************************************************//** +Obtain the InnoDB transaction id of a MySQL thread. +@return transaction id */ +__attribute__((warn_unused_result, nonnull)) ulonglong thd_to_trx_id( -/*=======*/ THD* thd) /*!< in: MySQL thread */ { return(thd_to_trx(thd)->id); @@ -1856,39 +1917,11 @@ thd_to_trx_id( #endif /* WITH_WSREP */ /********************************************************************//** -Call this function when mysqld passes control to the client. That is to -avoid deadlocks on the adaptive hash S-latch possibly held by thd. For more -documentation, see handler.cc. -@return 0 */ -static -int -innobase_release_temporary_latches( -/*===============================*/ - handlerton* hton, /*!< in: handlerton */ - THD* thd) /*!< in: MySQL thread */ -{ - DBUG_ASSERT(hton == innodb_hton_ptr); - - if (!innodb_inited) { - - return(0); - } - - trx_t* trx = thd_to_trx(thd); - - if (trx != NULL) { - trx_search_latch_release_if_reserved(trx); - } - - return(0); -} - -/********************************************************************//** Increments innobase_active_counter and every INNOBASE_WAKE_INTERVALth time calls srv_active_wake_master_thread. This function should be used when a single database operation may introduce a small need for server utility activity, like checkpointing. */ -static inline +inline void innobase_active_small(void) /*=======================*/ @@ -1904,8 +1937,7 @@ innobase_active_small(void) Converts an InnoDB error code to a MySQL error code and also tells to MySQL about a possible transaction rollback inside InnoDB caused by a lock wait timeout or a deadlock. -@return MySQL error code */ -static +@return MySQL error code */ int convert_error_code_to_mysql( /*========================*/ @@ -1918,7 +1950,7 @@ convert_error_code_to_mysql( return(0); case DB_INTERRUPTED: - return(HA_ERR_ABORTED_BY_USER); + return(HA_ERR_ABORTED_BY_USER); case DB_FOREIGN_EXCEED_MAX_CASCADE: ut_ad(thd); @@ -1930,12 +1962,15 @@ convert_error_code_to_mysql( "depth of %d. Please " "drop extra constraints and try " "again", DICT_FK_MAX_RECURSIVE_LOAD); + return(HA_ERR_FK_DEPTH_EXCEEDED); - /* fall through */ + case DB_CANT_CREATE_GEOMETRY_OBJECT: + my_error(ER_CANT_CREATE_GEOMETRY_OBJECT, MYF(0)); + return(HA_ERR_NULL_IN_SPATIAL); case DB_ERROR: default: - return(-1); /* unspecified error */ + return(HA_ERR_GENERIC); /* unspecified error */ case DB_DUPLICATE_KEY: /* Be cautious with returning this error, since @@ -1963,8 +1998,8 @@ convert_error_code_to_mysql( tell it also to MySQL so that MySQL knows to empty the cached binlog for this transaction */ - if (thd) { - thd_mark_transaction_to_rollback(thd, TRUE); + if (thd != NULL) { + thd_mark_transaction_to_rollback(thd, 1); } return(HA_ERR_LOCK_DEADLOCK); @@ -1987,6 +2022,7 @@ convert_error_code_to_mysql( case DB_ROW_IS_REFERENCED: return(HA_ERR_ROW_IS_REFERENCED); + case DB_NO_FK_ON_S_BASE_COL: case DB_CANNOT_ADD_CONSTRAINT: case DB_CHILD_NO_INDEX: case DB_PARENT_NO_INDEX: @@ -2004,10 +2040,10 @@ convert_error_code_to_mysql( case DB_OUT_OF_FILE_SPACE: return(HA_ERR_RECORD_FILE_FULL); - case DB_TEMP_FILE_WRITE_FAILURE: + case DB_TEMP_FILE_WRITE_FAIL: my_error(ER_GET_ERRMSG, MYF(0), - DB_TEMP_FILE_WRITE_FAILURE, - ut_strerr(DB_TEMP_FILE_WRITE_FAILURE), + DB_TEMP_FILE_WRITE_FAIL, + ut_strerr(DB_TEMP_FILE_WRITE_FAIL), "InnoDB"); return(HA_ERR_INTERNAL_ERROR); @@ -2017,7 +2053,6 @@ convert_error_code_to_mysql( case DB_TABLE_IS_BEING_USED: return(HA_ERR_WRONG_COMMAND); - case DB_TABLESPACE_DELETED: case DB_TABLE_NOT_FOUND: return(HA_ERR_NO_SUCH_TABLE); @@ -2025,19 +2060,17 @@ convert_error_code_to_mysql( return(HA_ERR_DECRYPTION_FAILED); case DB_TABLESPACE_NOT_FOUND: - return(HA_ERR_NO_SUCH_TABLE); + return(HA_ERR_TABLESPACE_MISSING); case DB_TOO_BIG_RECORD: { /* If prefix is true then a 768-byte prefix is stored locally for BLOB fields. Refer to dict_table_get_format(). - Note that in page0zip.ic page_zip_rec_needs_ext() rec_size - is limited to COMPRESSED_REC_MAX_DATA_SIZE (16K) or - REDUNDANT_REC_MAX_DATA_SIZE (16K-1). */ + We limit max record size to 16k for 64k page size. */ bool prefix = (dict_tf_get_format(flags) == UNIV_FORMAT_A); bool comp = !!(flags & DICT_TF_COMPACT); ulint free_space = page_get_free_space_of_empty(comp) / 2; - if (free_space >= (comp ? COMPRESSED_REC_MAX_DATA_SIZE : + if (free_space >= ulint(comp ? COMPRESSED_REC_MAX_DATA_SIZE : REDUNDANT_REC_MAX_DATA_SIZE)) { free_space = (comp ? COMPRESSED_REC_MAX_DATA_SIZE : REDUNDANT_REC_MAX_DATA_SIZE) - 1; @@ -2059,18 +2092,9 @@ convert_error_code_to_mysql( return(HA_ERR_TO_BIG_ROW); } - - case DB_TOO_BIG_FOR_REDO: - my_printf_error(ER_TOO_BIG_ROWSIZE, "%s" , MYF(0), - "The size of BLOB/TEXT data inserted" - " in one transaction is greater than" - " 10% of redo log size. Increase the" - " redo log size using innodb_log_file_size."); - return(HA_ERR_TO_BIG_ROW); - case DB_TOO_BIG_INDEX_COL: my_error(ER_INDEX_COLUMN_TOO_LONG, MYF(0), - DICT_MAX_FIELD_LEN_BY_FORMAT_FLAG(flags)); + (ulong) DICT_MAX_FIELD_LEN_BY_FORMAT_FLAG(flags)); return(HA_ERR_INDEX_COL_TOO_LONG); case DB_NO_SAVEPOINT: @@ -2082,7 +2106,7 @@ convert_error_code_to_mysql( cached binlog for this transaction */ if (thd) { - thd_mark_transaction_to_rollback(thd, TRUE); + thd_mark_transaction_to_rollback(thd, 1); } return(HA_ERR_LOCK_TABLE_FULL); @@ -2103,16 +2127,21 @@ convert_error_code_to_mysql( return(HA_ERR_OUT_OF_MEM); case DB_TABLESPACE_EXISTS: return(HA_ERR_TABLESPACE_EXISTS); + case DB_TABLESPACE_DELETED: + return(HA_ERR_TABLESPACE_MISSING); case DB_IDENTIFIER_TOO_LONG: return(HA_ERR_INTERNAL_ERROR); + case DB_TABLE_CORRUPT: + return(HA_ERR_TABLE_CORRUPT); case DB_FTS_TOO_MANY_WORDS_IN_PHRASE: return(HA_ERR_FTS_TOO_MANY_WORDS_IN_PHRASE); + case DB_COMPUTE_VALUE_FAILED: + return(HA_ERR_GENERIC); // impossible } } /*************************************************************//** Prints info of a THD object (== user session thread) to the given file. */ -UNIV_INTERN void innobase_mysql_print_thd( /*=====================*/ @@ -2131,7 +2160,6 @@ innobase_mysql_print_thd( /******************************************************************//** Get the error message format string. @return the format string or 0 if not found. */ -UNIV_INTERN const char* innobase_get_err_msg( /*=================*/ @@ -2142,7 +2170,6 @@ innobase_get_err_msg( /******************************************************************//** Get the variable length bounds of the given character set. */ -UNIV_INTERN void innobase_get_cset_width( /*====================*/ @@ -2168,11 +2195,11 @@ innobase_get_cset_width( /* Fix bug#46256: allow tables to be dropped if the collation is not found, but issue a warning. */ - if ((global_system_variables.log_warnings) - && (cset != 0)){ + if (cset != 0) { sql_print_warning( - "Unknown collation #%lu.", cset); + "Unknown collation #" ULINTPF ".", + cset); } } else { @@ -2185,14 +2212,13 @@ innobase_get_cset_width( /******************************************************************//** Converts an identifier to a table name. */ -UNIV_INTERN void innobase_convert_from_table_id( /*===========================*/ - struct charset_info_st* cs, /*!< in: the 'from' character set */ - char* to, /*!< out: converted identifier */ - const char* from, /*!< in: identifier to convert */ - ulint len) /*!< in: length of 'to', in bytes */ + CHARSET_INFO* cs, /*!< in: the 'from' character set */ + char* to, /*!< out: converted identifier */ + const char* from, /*!< in: identifier to convert */ + ulint len) /*!< in: length of 'to', in bytes */ { uint errors; @@ -2202,7 +2228,6 @@ innobase_convert_from_table_id( /********************************************************************** Check if the length of the identifier exceeds the maximum allowed. return true when length of identifier is too long. */ -UNIV_INTERN my_bool innobase_check_identifier_length( /*=============================*/ @@ -2213,7 +2238,7 @@ innobase_check_identifier_length( CHARSET_INFO *cs = system_charset_info; DBUG_ENTER("innobase_check_identifier_length"); - size_t len = cs->cset->well_formed_len( + size_t len = my_well_formed_length( cs, id, id + strlen(id), NAME_CHAR_LEN, &well_formed_error); @@ -2226,14 +2251,13 @@ innobase_check_identifier_length( /******************************************************************//** Converts an identifier to UTF-8. */ -UNIV_INTERN void innobase_convert_from_id( /*=====================*/ - struct charset_info_st* cs, /*!< in: the 'from' character set */ - char* to, /*!< out: converted identifier */ - const char* from, /*!< in: identifier to convert */ - ulint len) /*!< in: length of 'to', in bytes */ + CHARSET_INFO* cs, /*!< in: the 'from' character set */ + char* to, /*!< out: converted identifier */ + const char* from, /*!< in: identifier to convert */ + ulint len) /*!< in: length of 'to', in bytes */ { uint errors; @@ -2242,8 +2266,7 @@ innobase_convert_from_id( /******************************************************************//** Compares NUL-terminated UTF-8 strings case insensitively. -@return 0 if a=b, <0 if a<b, >1 if a>b */ -UNIV_INTERN +@return 0 if a=b, <0 if a<b, >1 if a>b */ int innobase_strcasecmp( /*================*/ @@ -2267,7 +2290,7 @@ innobase_strcasecmp( Compares NUL-terminated UTF-8 strings case insensitively. The second string contains wildcards. @return 0 if a match is found, 1 if not */ -UNIV_INTERN +static int innobase_wildcasecmp( /*=================*/ @@ -2277,14 +2300,12 @@ innobase_wildcasecmp( return(wild_case_compare(system_charset_info, a, b)); } -/******************************************************************//** -Strip dir name from a full path name and return only the file name +/** Strip dir name from a full path name and return only the file name +@param[in] path_name full path name @return file name or "null" if no file name */ -UNIV_INTERN const char* innobase_basename( -/*==============*/ - const char* path_name) /*!< in: full path name */ + const char* path_name) { const char* name = base_name(path_name); @@ -2293,7 +2314,6 @@ innobase_basename( /******************************************************************//** Makes all characters in a NUL-terminated UTF-8 string lower case. */ -UNIV_INTERN void innobase_casedn_str( /*================*/ @@ -2304,9 +2324,8 @@ innobase_casedn_str( /**********************************************************************//** Determines the connection character set. -@return connection character set */ -UNIV_INTERN -struct charset_info_st* +@return connection character set */ +CHARSET_INFO* innobase_get_charset( /*=================*/ THD* mysql_thd) /*!< in: MySQL thread handle */ @@ -2314,29 +2333,46 @@ innobase_get_charset( return(thd_charset(mysql_thd)); } -/**********************************************************************//** -Determines the current SQL statement. -@return SQL statement string */ -UNIV_INTERN +/** Determines the current SQL statement. +Thread unsafe, can only be called from the thread owning the THD. +@param[in] thd MySQL thread handle +@param[out] length Length of the SQL statement +@return SQL statement string */ const char* -innobase_get_stmt( -/*==============*/ - THD* thd, /*!< in: MySQL thread handle */ - size_t* length) /*!< out: length of the SQL statement */ +innobase_get_stmt_unsafe( + THD* thd, + size_t* length) { if (const LEX_STRING *stmt = thd_query_string(thd)) { *length = stmt->length; return stmt->str; } + + *length = 0; return NULL; } +/** Determines the current SQL statement. +Thread safe, can be called from any thread as the string is copied +into the provided buffer. +@param[in] thd MySQL thread handle +@param[out] buf Buffer containing SQL statement +@param[in] buflen Length of provided buffer +@return Length of the SQL statement */ +size_t +innobase_get_stmt_safe( + THD* thd, + char* buf, + size_t buflen) +{ + return thd_query_safe(thd, buf, buflen); +} + /**********************************************************************//** Get the current setting of the tdc_size global parameter. We do a dirty read because for one there is no synchronization object and secondly there is little harm in doing so even if we get a torn read. @return value of tdc_size */ -UNIV_INTERN ulint innobase_get_table_cache_size(void) /*===============================*/ @@ -2349,8 +2385,7 @@ Get the current setting of the lower_case_table_names global parameter from mysqld.cc. We do a dirty read because for one there is no synchronization object and secondly there is little harm in doing so even if we get a torn read. -@return value of lower_case_table_names */ -UNIV_INTERN +@return value of lower_case_table_names */ ulint innobase_get_lower_case_table_names(void) /*=====================================*/ @@ -2358,11 +2393,70 @@ innobase_get_lower_case_table_names(void) return(lower_case_table_names); } -/** Create a temporary file in the location specified by the parameter -path. If the path is null, then it will be created in tmpdir. +/** + Test a file path whether it is same as mysql data directory path. + + @param path null terminated character string + + @return + @retval TRUE The path is different from mysql data directory. + @retval FALSE The path is same as mysql data directory. +*/ +static bool is_mysql_datadir_path(const char *path) +{ + if (path == NULL) + return false; + + char mysql_data_dir[FN_REFLEN], path_dir[FN_REFLEN]; + convert_dirname(path_dir, path, NullS); + convert_dirname(mysql_data_dir, mysql_unpacked_real_data_home, NullS); + size_t mysql_data_home_len= dirname_length(mysql_data_dir); + size_t path_len = dirname_length(path_dir); + + if (path_len < mysql_data_home_len) + return true; + + if (!lower_case_file_system) + return(memcmp(mysql_data_dir, path_dir, mysql_data_home_len)); + + return(files_charset_info->coll->strnncoll(files_charset_info, + (uchar *) path_dir, path_len, + (uchar *) mysql_data_dir, + mysql_data_home_len, + TRUE)); +} + +static int mysql_tmpfile_path(const char *path, const char *prefix) +{ + DBUG_ASSERT(path != NULL); + DBUG_ASSERT((strlen(path) + strlen(prefix)) <= FN_REFLEN); + + char filename[FN_REFLEN]; + File fd = create_temp_file(filename, path, prefix, +#ifdef __WIN__ + O_BINARY | O_TRUNC | O_SEQUENTIAL | + O_SHORT_LIVED | +#endif /* __WIN__ */ + O_CREAT | O_EXCL | O_RDWR | O_TEMPORARY, + MYF(MY_WME)); + if (fd >= 0) { +#ifndef __WIN__ + /* + This can be removed once the following bug is fixed: + Bug #28903 create_temp_file() doesn't honor O_TEMPORARY option + (file not removed) (Unix) + */ + unlink(filename); +#endif /* !__WIN__ */ + } + + return fd; +} + +/** Creates a temporary file in the location specified by the parameter +path. If the path is NULL, then it will be created in tmpdir. @param[in] path location for creating temporary file -@return temporary file descriptor, or < 0 on error */ -UNIV_INTERN +@return temporary file descriptor, or < 0 on error */ int innobase_mysql_tmpfile( const char* path) @@ -2421,11 +2515,13 @@ innobase_mysql_tmpfile( #endif #endif if (fd2 < 0) { + char errbuf[MYSYS_STRERROR_SIZE]; DBUG_PRINT("error",("Got error %d on dup",fd2)); - my_errno=errno; + set_my_errno(errno); my_error(EE_OUT_OF_FILERESOURCES, - MYF(ME_BELL+ME_WAITTANG), - "ib*", my_errno); + MYF(0), + "ib*", errno, + my_strerror(errbuf, sizeof(errbuf), errno)); } my_close(fd, MYF(MY_WME)); } @@ -2434,8 +2530,8 @@ innobase_mysql_tmpfile( /*********************************************************************//** Wrapper around MySQL's copy_and_convert function. -@return number of bytes copied to 'to' */ -UNIV_INTERN +@return number of bytes copied to 'to' */ +static ulint innobase_convert_string( /*====================*/ @@ -2464,8 +2560,7 @@ Not more than "buf_size" bytes are written to "buf". The result is always NUL-terminated (provided buf_size > 0) and the number of bytes that were written to "buf" is returned (including the terminating NUL). -@return number of bytes that were written */ -UNIV_INTERN +@return number of bytes that were written */ ulint innobase_raw_format( /*================*/ @@ -2511,8 +2606,7 @@ values we want to reserve for multi-value inserts e.g., innobase_next_autoinc() will be called with increment set to 3 where autoinc_lock_mode != TRADITIONAL because we want to reserve 3 values for the multi-value INSERT above. -@return the next value */ -UNIV_INTERN +@return the next value */ ulonglong innobase_next_autoinc( /*==================*/ @@ -2605,6 +2699,64 @@ innobase_next_autoinc( return(next_value); } +/********************************************************************//** +Reset the autoinc value in the table. +@return DB_SUCCESS if all went well else error code */ +UNIV_INTERN +dberr_t +ha_innobase::innobase_reset_autoinc( +/*================================*/ + ulonglong autoinc) /*!< in: value to store */ +{ + dberr_t error; + + error = innobase_lock_autoinc(); + + if (error == DB_SUCCESS) { + + dict_table_autoinc_initialize(m_prebuilt->table, autoinc); + + dict_table_autoinc_unlock(m_prebuilt->table); + } + + return(error); +} + +/*******************************************************************//** +Reset the auto-increment counter to the given value, i.e. the next row +inserted will get the given value. This is called e.g. after TRUNCATE +is emulated by doing a 'DELETE FROM t'. HA_ERR_WRONG_COMMAND is +returned by storage engines that don't support this operation. +@return 0 or error code */ +UNIV_INTERN +int +ha_innobase::reset_auto_increment( +/*==============================*/ + ulonglong value) /*!< in: new value for table autoinc */ +{ + DBUG_ENTER("ha_innobase::reset_auto_increment"); + + dberr_t error; + + update_thd(ha_thd()); + + error = row_lock_table_autoinc_for_mysql(m_prebuilt); + + if (error != DB_SUCCESS) { + DBUG_RETURN(convert_error_code_to_mysql( + error, m_prebuilt->table->flags, m_user_thd)); + } + + /* The next value can never be 0. */ + if (value == 0) { + value = 1; + } + + innobase_reset_autoinc(value); + + DBUG_RETURN(0); +} + /*********************************************************************//** Initializes some fields in an InnoDB transaction object. */ static @@ -2628,8 +2780,7 @@ innobase_trx_init( /*********************************************************************//** Allocates an InnoDB transaction for a MySQL handler object for DML. -@return InnoDB transaction handle */ -UNIV_INTERN +@return InnoDB transaction handle */ trx_t* innobase_trx_allocate( /*==================*/ @@ -2654,26 +2805,22 @@ innobase_trx_allocate( Gets the InnoDB transaction handle for a MySQL handler object, creates an InnoDB transaction struct if the corresponding MySQL thread struct still lacks one. -@return InnoDB transaction handle */ +@return InnoDB transaction handle */ static inline trx_t* check_trx_exists( /*=============*/ THD* thd) /*!< in: user thread handle */ { - trx_t*& trx = thd_to_trx(thd); - - if (trx == NULL) { + if (trx_t* trx = thd_to_trx(thd)) { + ut_a(trx->magic_n == TRX_MAGIC_N); + innobase_trx_init(thd, trx); + return trx; + } else { trx = innobase_trx_allocate(thd); thd_set_ha_data(thd, innodb_hton_ptr, trx); - } else if (UNIV_UNLIKELY(trx->magic_n != TRX_MAGIC_N)) { - mem_analyze_corruption(trx); - ut_error; + return trx; } - - innobase_trx_init(thd, trx); - - return(trx); } /************************************************************************* @@ -2683,8 +2830,7 @@ innobase_get_trx() { THD *thd=current_thd; if (likely(thd != 0)) { - trx_t*& trx = thd_to_trx(thd); - return(trx); + return thd_to_trx(thd); } else { return(NULL); } @@ -2696,7 +2842,7 @@ Note that a transaction has been registered with MySQL. static inline bool trx_is_registered_for_2pc( -/*=========================*/ +/*======================*/ const trx_t* trx) /* in: transaction */ { return(trx->is_registered == 1); @@ -2707,7 +2853,7 @@ Note that innobase_commit_ordered() was run. */ static inline void trx_set_active_commit_ordered( -/*==============================*/ +/*==========================*/ trx_t* trx) /* in: transaction */ { ut_a(trx_is_registered_for_2pc(trx)); @@ -2735,7 +2881,7 @@ trx_deregister_from_2pc( trx_t* trx) /* in: transaction */ { trx->is_registered = 0; - trx->active_commit_ordered = 0; + trx->active_commit_ordered = 0; } /*********************************************************************//** @@ -2750,23 +2896,11 @@ trx_is_active_commit_ordered( } /*********************************************************************//** -Check if transaction is started. -@reutrn true if transaction is in state started */ -static -bool -trx_is_started( -/*===========*/ - trx_t* trx) /* in: transaction */ -{ - return(trx->state != TRX_STATE_NOT_STARTED); -} - -/*********************************************************************//** Copy table flags from MySQL's HA_CREATE_INFO into an InnoDB table object. Those flags are stored in .frm file and end up in the MySQL table object, but are frequently used inside InnoDB so we keep their copies into the InnoDB table object. */ -UNIV_INTERN +static void innobase_copy_frm_flags_from_create_info( /*=====================================*/ @@ -2802,7 +2936,6 @@ Copy table flags from MySQL's TABLE_SHARE into an InnoDB table object. Those flags are stored in .frm file and end up in the MySQL table object, but are frequently used inside InnoDB so we keep their copies into the InnoDB table object. */ -UNIV_INTERN void innobase_copy_frm_flags_from_table_share( /*=====================================*/ @@ -2835,30 +2968,44 @@ innobase_copy_frm_flags_from_table_share( /*********************************************************************//** Construct ha_innobase handler. */ -UNIV_INTERN + ha_innobase::ha_innobase( /*=====================*/ handlerton* hton, TABLE_SHARE* table_arg) :handler(hton, table_arg), - int_table_flags(HA_REC_NOT_IN_SEQ | - HA_NULL_IN_KEY | HA_CAN_VIRTUAL_COLUMNS | - HA_CAN_INDEX_BLOBS | HA_CONCURRENT_OPTIMIZE | - HA_CAN_SQL_HANDLER | - HA_PRIMARY_KEY_REQUIRED_FOR_POSITION | - HA_PRIMARY_KEY_IN_READ_INDEX | - HA_BINLOG_ROW_CAPABLE | - HA_CAN_GEOMETRY | HA_PARTIAL_COLUMN_READ | - HA_TABLE_SCAN_ON_INDEX | HA_CAN_FULLTEXT | - (srv_force_primary_key ? HA_REQUIRE_PRIMARY_KEY : 0 ) | - HA_CAN_FULLTEXT_EXT | HA_CAN_EXPORT), - start_of_scan(0), - num_write_row(0) + m_prebuilt(), + m_prebuilt_ptr(&m_prebuilt), + m_user_thd(), + m_int_table_flags(HA_REC_NOT_IN_SEQ + | HA_NULL_IN_KEY + | HA_CAN_VIRTUAL_COLUMNS + | HA_CAN_INDEX_BLOBS + | HA_CAN_SQL_HANDLER + | HA_REQUIRES_KEY_COLUMNS_FOR_DELETE + | HA_PRIMARY_KEY_REQUIRED_FOR_POSITION + | HA_PRIMARY_KEY_IN_READ_INDEX + | HA_BINLOG_ROW_CAPABLE + | HA_CAN_GEOMETRY + | HA_PARTIAL_COLUMN_READ + | HA_TABLE_SCAN_ON_INDEX + | HA_CAN_FULLTEXT + | HA_CAN_FULLTEXT_EXT + /* JAN: TODO: MySQL 5.7 + | HA_CAN_FULLTEXT_HINTS + */ + | HA_CAN_EXPORT + | HA_CAN_RTREEKEYS + | HA_CONCURRENT_OPTIMIZE + | (srv_force_primary_key ? HA_REQUIRE_PRIMARY_KEY : 0) + ), + m_start_of_scan(), + m_mysql_has_locked() {} /*********************************************************************//** Destruct ha_innobase handler. */ -UNIV_INTERN + ha_innobase::~ha_innobase() /*======================*/ { @@ -2867,38 +3014,42 @@ ha_innobase::~ha_innobase() /*********************************************************************//** Updates the user_thd field in a handle and also allocates a new InnoDB transaction handle if needed, and updates the transaction fields in the -prebuilt struct. */ -UNIV_INTERN inline +m_prebuilt struct. */ void ha_innobase::update_thd( /*====================*/ THD* thd) /*!< in: thd to use the handle */ { - trx_t* trx; - DBUG_ENTER("ha_innobase::update_thd"); DBUG_PRINT("ha_innobase::update_thd", ("user_thd: %p -> %p", - user_thd, thd)); + m_user_thd, thd)); /* The table should have been opened in ha_innobase::open(). */ - DBUG_ASSERT(prebuilt->table->n_ref_count > 0); + DBUG_ASSERT(m_prebuilt->table->get_ref_count() > 0); - trx = check_trx_exists(thd); + trx_t* trx = check_trx_exists(thd); - if (prebuilt->trx != trx) { + ut_ad(trx->dict_operation_lock_mode == 0); + ut_ad(trx->dict_operation == TRX_DICT_OP_NONE); - row_update_prebuilt_trx(prebuilt, trx); + if (m_prebuilt->trx != trx) { + + row_update_prebuilt_trx(m_prebuilt, trx); } - user_thd = thd; + m_user_thd = thd; + + DBUG_ASSERT(m_prebuilt->trx->magic_n == TRX_MAGIC_N); + DBUG_ASSERT(m_prebuilt->trx == thd_to_trx(m_user_thd)); + DBUG_VOID_RETURN; } /*********************************************************************//** Updates the user_thd field in a handle and also allocates a new InnoDB transaction handle if needed, and updates the transaction fields in the -prebuilt struct. */ -UNIV_INTERN +m_prebuilt struct. */ + void ha_innobase::update_thd() /*=====================*/ @@ -2924,11 +3075,18 @@ innobase_register_trx( THD* thd, /* in: MySQL thd (connection) object */ trx_t* trx) /* in: transaction to register */ { + /* JAN: TODO: MySQL 5.7 PSI + const ulonglong trx_id = static_cast<const ulonglong>( + trx_get_id_for_print(trx)); + + trans_register_ha(thd, FALSE, hton, &trx_id); + */ trans_register_ha(thd, FALSE, hton); if (!trx_is_registered_for_2pc(trx) && thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { + //trans_register_ha(thd, TRUE, hton, &trx_id); trans_register_ha(thd, TRUE, hton); } @@ -2980,6 +3138,83 @@ AUTOCOMMIT==0 or we are inside BEGIN ... COMMIT. Thus transactions no longer put restrictions on the use of the query cache. */ +/** Check if mysql can allow the transaction to read from/store to +the query cache. +@param[in] table table object +@param[in] trx transaction object +@return whether the storing or retrieving from the query cache is permitted */ +static bool innobase_query_caching_table_check_low( + const dict_table_t* table, + trx_t* trx) +{ + /* The following conditions will decide the query cache + retrieval or storing into: + + (1) There should not be any locks on the table. + (2) Someother trx shouldn't invalidate the cache before this + transaction started. + (3) Read view shouldn't exist. If exists then the view + low_limit_id should be greater than or equal to the transaction that + invalidates the cache for the particular table. + + For read-only transaction: should satisfy (1) and (3) + For read-write transaction: should satisfy (1), (2), (3) */ + + if (lock_table_get_n_locks(table)) { + return false; + } + + if (trx->id && trx->id < table->query_cache_inv_trx_id) { + return false; + } + + return !MVCC::is_view_active(trx->read_view) + || trx->read_view->low_limit_id() + >= table->query_cache_inv_trx_id; +} + +/** Checks if MySQL at the moment is allowed for this table to retrieve a +consistent read result, or store it to the query cache. +@param[in,out] trx transaction +@param[in] norm_name concatenation of database name, + '/' char, table name +@return whether storing or retrieving from the query cache is permitted */ +static bool innobase_query_caching_table_check( + trx_t* trx, + const char* norm_name) +{ + dict_table_t* table = dict_table_open_on_name( + norm_name, FALSE, FALSE, DICT_ERR_IGNORE_NONE); + + if (table == NULL) { + return false; + } + + /* Start the transaction if it is not started yet */ + trx_start_if_not_started(trx, false); + + bool allow = innobase_query_caching_table_check_low(table, trx); + + dict_table_close(table, FALSE, FALSE); + + if (allow) { + /* If the isolation level is high, assign a read view for the + transaction if it does not yet have one */ + + if (trx->isolation_level >= TRX_ISO_REPEATABLE_READ + && !srv_read_only_mode + && !MVCC::is_view_active(trx->read_view)) { + + /* Start the transaction if it is not started yet */ + trx_start_if_not_started(trx, false); + + trx_sys->mvcc->view_open(trx->read_view, trx); + } + } + + return allow; +} + /******************************************************************//** The MySQL query cache uses this to check from InnoDB if the query cache at the moment is allowed to operate on an InnoDB table. The SQL query must @@ -2996,7 +3231,7 @@ Why a deadlock of threads is not possible: the query cache calls this function at the start of a SELECT processing. Then the calling thread cannot be holding any InnoDB semaphores. The calling thread is holding the query cache mutex, and this function will reserve the InnoDB trx_sys->mutex. -Thus, the 'rank' in sync0sync.h of the MySQL query cache mutex is above +Thus, the 'rank' in sync0mutex.h of the MySQL query cache mutex is above the InnoDB trx_sys->mutex. @return TRUE if permitted, FALSE if not; note that the value FALSE does not mean we should invalidate the query cache: invalidation is @@ -3010,44 +3245,25 @@ innobase_query_caching_of_table_permitted( retrieve it */ char* full_name, /*!< in: normalized path to the table */ uint full_name_len, /*!< in: length of the normalized path - to the table */ + to the table */ ulonglong *unused) /*!< unused for this engine */ { - ibool is_autocommit; - trx_t* trx; char norm_name[1000]; + trx_t* trx = check_trx_exists(thd); ut_a(full_name_len < 999); - trx = check_trx_exists(thd); - if (trx->isolation_level == TRX_ISO_SERIALIZABLE) { /* In the SERIALIZABLE mode we add LOCK IN SHARE MODE to every plain SELECT if AUTOCOMMIT is not on. */ - return((my_bool)FALSE); - } - - if (UNIV_UNLIKELY(trx->has_search_latch)) { - sql_print_error("The calling thread is holding the adaptive " - "search, latch though calling " - "innobase_query_caching_of_table_permitted."); - trx_print(stderr, trx, 1024); + return(false); } - trx_search_latch_release_if_reserved(trx); - innobase_srv_conc_force_exit_innodb(trx); - if (!thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { - - is_autocommit = TRUE; - } else { - is_autocommit = FALSE; - - } - - if (is_autocommit && trx->n_mysql_tables_in_use == 0) { + if (!thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN) + && trx->n_mysql_tables_in_use == 0) { /* We are going to retrieve the query result from the query cache. This cannot be a store operation to the query cache because then MySQL would have locks on tables already. @@ -3066,7 +3282,7 @@ innobase_query_caching_of_table_permitted( then trx2 would have already invalidated the cache. Thus we can trust the result in the cache is ok for this query. */ - return((my_bool)TRUE); + return(true); } /* Normalize the table name to InnoDB format */ @@ -3074,219 +3290,203 @@ innobase_query_caching_of_table_permitted( innobase_register_trx(innodb_hton_ptr, thd, trx); - if (row_search_check_if_query_cache_permitted(trx, norm_name)) { - - /* printf("Query cache for %s permitted\n", norm_name); */ - - return((my_bool)TRUE); - } - - /* printf("Query cache for %s NOT permitted\n", norm_name); */ - - return((my_bool)FALSE); + return innobase_query_caching_table_check(trx, norm_name); } /*****************************************************************//** Invalidates the MySQL query cache for the table. */ -UNIV_INTERN void innobase_invalidate_query_cache( /*============================*/ trx_t* trx, /*!< in: transaction which modifies the table */ const char* full_name, /*!< in: concatenation of - database name, null char NUL, + database name, path separator, table name, null char NUL; NOTE that in Windows this is always in LOWER CASE! */ ulint full_name_len) /*!< in: full name length where also the null chars count */ { - /* Note that the sync0sync.h rank of the query cache mutex is just + /* Note that the sync0mutex.h rank of the query cache mutex is just above the InnoDB trx_sys_t->lock. The caller of this function must not have latches of a lower rank. */ #ifdef HAVE_QUERY_CACHE - char qcache_key_name[2 * (NAME_LEN + 1)]; - size_t tabname_len; - size_t dbname_len; - - /* Construct the key("db-name\0table$name\0") for the query cache using - the path name("db@002dname\0table@0024name\0") of the table in its + char qcache_key_name[2 * (NAME_LEN + 1)]; + char db_name[NAME_CHAR_LEN * MY_CS_MBMAXLEN + 1]; + const char *key_ptr; + size_t tabname_len; + size_t dbname_len; + + // Extract the database name. + key_ptr= strchr(full_name, '/'); + DBUG_ASSERT(key_ptr != NULL); // Database name should be present + memcpy(db_name, full_name, (dbname_len= (key_ptr - full_name))); + db_name[dbname_len]= '\0'; + + /* Construct the key("db-name\0table$name\0") for the query cache using + the path name("db@002dname\0table@0024name\0") of the table in its canonical form. */ - dbname_len = filename_to_tablename(full_name, qcache_key_name, - sizeof(qcache_key_name)); - tabname_len = filename_to_tablename(full_name + strlen(full_name) + 1, - qcache_key_name + dbname_len + 1, - sizeof(qcache_key_name) - - dbname_len - 1); - - /* Argument TRUE below means we are using transactions */ - mysql_query_cache_invalidate4(trx->mysql_thd, - qcache_key_name, - (dbname_len + tabname_len + 2), - TRUE); + dbname_len = filename_to_tablename(db_name, qcache_key_name, + sizeof(qcache_key_name)); + tabname_len = filename_to_tablename(++key_ptr, + (qcache_key_name + dbname_len + 1), + sizeof(qcache_key_name) - + dbname_len - 1); + + /* Argument TRUE below means we are using transactions */ + mysql_query_cache_invalidate4(trx->mysql_thd, + qcache_key_name, + (dbname_len + tabname_len + 2), + TRUE); #endif } -/*****************************************************************//** -Convert an SQL identifier to the MySQL system_charset_info (UTF-8) -and quote it if needed. -@return pointer to the end of buf */ -static -char* -innobase_convert_identifier( -/*========================*/ - char* buf, /*!< out: buffer for converted identifier */ - ulint buflen, /*!< in: length of buf, in bytes */ - const char* id, /*!< in: identifier to convert */ - ulint idlen, /*!< in: length of id, in bytes */ - THD* thd, /*!< in: MySQL connection thread, or NULL */ - ibool file_id)/*!< in: TRUE=id is a table or database name; - FALSE=id is an UTF-8 string */ +/** Quote a standard SQL identifier like index or column name. +@param[in] file output stream +@param[in] trx InnoDB transaction, or NULL +@param[in] id identifier to quote */ +void +innobase_quote_identifier( + FILE* file, + trx_t* trx, + const char* id) { - char nz2[MAX_TABLE_NAME_LEN + 1]; - const char* s = id; - int q; + const int q = trx != NULL && trx->mysql_thd != NULL + ? get_quote_char_for_identifier(trx->mysql_thd, id, strlen(id)) + : '`'; - if (file_id) { - - char nz[MAX_TABLE_NAME_LEN + 1]; + if (q == EOF) { + fputs(id, file); + } else { + putc(q, file); - /* Decode the table name. The MySQL function expects - a NUL-terminated string. The input and output strings - buffers must not be shared. */ - ut_a(idlen <= MAX_TABLE_NAME_LEN); - memcpy(nz, id, idlen); - nz[idlen] = 0; + while (int c = *id++) { + if (c == q) { + putc(c, file); + } + putc(c, file); + } - s = nz2; - idlen = explain_filename(thd, nz, nz2, sizeof nz2, - EXPLAIN_PARTITIONS_AS_COMMENT); - goto no_quote; + putc(q, file); } +} - /* See if the identifier needs to be quoted. */ - if (UNIV_UNLIKELY(!thd)) { - q = '"'; - } else { - q = get_quote_char_for_identifier(thd, s, (int) idlen); - } +/** Quote a standard SQL identifier like tablespace, index or column name. +@param[in] trx InnoDB transaction, or NULL +@param[in] id identifier to quote +@return quoted identifier */ +std::string +innobase_quote_identifier( +/*======================*/ + trx_t* trx, + const char* id) +{ + std::string quoted_identifier; + const int q = trx != NULL && trx->mysql_thd != NULL + ? get_quote_char_for_identifier(trx->mysql_thd, id, strlen(id)) + : '`'; if (q == EOF) { -no_quote: - if (UNIV_UNLIKELY(idlen > buflen)) { - idlen = buflen; - } - memcpy(buf, s, idlen); - return(buf + idlen); + quoted_identifier.append(id); + } else { + quoted_identifier += (unsigned char)q; + quoted_identifier.append(id); + quoted_identifier += (unsigned char)q; } - /* Quote the identifier. */ - if (buflen < 2) { - return(buf); - } + return (quoted_identifier); +} - *buf++ = q; - buflen--; +/** Convert a table name to the MySQL system_charset_info (UTF-8) +and quote it. +@param[out] buf buffer for converted identifier +@param[in] buflen length of buf, in bytes +@param[in] id identifier to convert +@param[in] idlen length of id, in bytes +@param[in] thd MySQL connection thread, or NULL +@return pointer to the end of buf */ +static +char* +innobase_convert_identifier( + char* buf, + ulint buflen, + const char* id, + ulint idlen, + THD* thd) +{ + const char* s = id; - for (; idlen; idlen--) { - int c = *s++; - if (UNIV_UNLIKELY(c == q)) { - if (UNIV_UNLIKELY(buflen < 3)) { - break; - } + char nz[MAX_TABLE_NAME_LEN + 1]; + char nz2[MAX_TABLE_NAME_LEN + 1]; - *buf++ = c; - *buf++ = c; - buflen -= 2; - } else { - if (UNIV_UNLIKELY(buflen < 2)) { - break; - } + /* Decode the table name. The MySQL function expects + a NUL-terminated string. The input and output strings + buffers must not be shared. */ + ut_a(idlen <= MAX_TABLE_NAME_LEN); + memcpy(nz, id, idlen); + nz[idlen] = 0; - *buf++ = c; - buflen--; - } + s = nz2; + idlen = explain_filename(thd, nz, nz2, sizeof nz2, + EXPLAIN_PARTITIONS_AS_COMMENT); + if (idlen > buflen) { + idlen = buflen; } - - *buf++ = q; - return(buf); + memcpy(buf, s, idlen); + return(buf + idlen); } /*****************************************************************//** -Convert a table or index name to the MySQL system_charset_info (UTF-8) -and quote it if needed. -@return pointer to the end of buf */ -UNIV_INTERN +Convert a table name to the MySQL system_charset_info (UTF-8). +@return pointer to the end of buf */ char* innobase_convert_name( /*==================*/ char* buf, /*!< out: buffer for converted identifier */ ulint buflen, /*!< in: length of buf, in bytes */ - const char* id, /*!< in: identifier to convert */ + const char* id, /*!< in: table name to convert */ ulint idlen, /*!< in: length of id, in bytes */ - THD* thd, /*!< in: MySQL connection thread, or NULL */ - ibool table_id)/*!< in: TRUE=id is a table or database name; - FALSE=id is an index name */ + THD* thd) /*!< in: MySQL connection thread, or NULL */ { char* s = buf; const char* bufend = buf + buflen; - if (table_id) { - const char* slash = (const char*) memchr(id, '/', idlen); - if (!slash) { - - goto no_db_name; - } + const char* slash = (const char*) memchr(id, '/', idlen); - /* Print the database name and table name separately. */ - s = innobase_convert_identifier(s, bufend - s, id, slash - id, - thd, TRUE); - if (UNIV_LIKELY(s < bufend)) { - *s++ = '.'; - s = innobase_convert_identifier(s, bufend - s, - slash + 1, idlen - - (slash - id) - 1, - thd, TRUE); - } - } else if (UNIV_UNLIKELY(*id == TEMP_INDEX_PREFIX)) { - /* Temporary index name (smart ALTER TABLE) */ - const char temp_index_suffix[]= "--temporary--"; + if (slash == NULL) { + return(innobase_convert_identifier( + buf, buflen, id, idlen, thd)); + } - s = innobase_convert_identifier(buf, buflen, id + 1, idlen - 1, - thd, FALSE); - if (s - buf + (sizeof temp_index_suffix - 1) < buflen) { - memcpy(s, temp_index_suffix, - sizeof temp_index_suffix - 1); - s += sizeof temp_index_suffix - 1; - } - } else { -no_db_name: - s = innobase_convert_identifier(buf, buflen, id, idlen, - thd, table_id); + /* Print the database name and table name separately. */ + s = innobase_convert_identifier(s, bufend - s, id, slash - id, thd); + if (s < bufend) { + *s++ = '.'; + s = innobase_convert_identifier(s, bufend - s, + slash + 1, idlen + - (slash - id) - 1, + thd); } return(s); } /*****************************************************************//** -A wrapper function of innobase_convert_name(), convert a table or -index name to the MySQL system_charset_info (UTF-8) and quote it if needed. -@return pointer to the end of buf */ -UNIV_INTERN +A wrapper function of innobase_convert_name(), convert a table name +to the MySQL system_charset_info (UTF-8) and quote it if needed. +@return pointer to the end of buf */ void innobase_format_name( /*==================*/ char* buf, /*!< out: buffer for converted identifier */ ulint buflen, /*!< in: length of buf, in bytes */ - const char* name, /*!< in: index or table name to format */ - ibool is_index_name) /*!< in: index name */ + const char* name) /*!< in: table name to format */ { const char* bufend; - bufend = innobase_convert_name(buf, buflen, name, strlen(name), - NULL, !is_index_name); + bufend = innobase_convert_name(buf, buflen, name, strlen(name), NULL); ut_ad((ulint) (bufend - buf) < buflen); @@ -3295,20 +3495,18 @@ innobase_format_name( /**********************************************************************//** Determines if the currently running transaction has been interrupted. -@return TRUE if interrupted */ -UNIV_INTERN +@return TRUE if interrupted */ ibool trx_is_interrupted( /*===============*/ const trx_t* trx) /*!< in: transaction */ { - return(trx && trx->mysql_thd && thd_kill_level(trx->mysql_thd)); + return(trx && trx->mysql_thd && thd_kill_level(trx->mysql_thd)); } /**********************************************************************//** Determines if the currently running transaction is in strict mode. -@return TRUE if strict */ -UNIV_INTERN +@return TRUE if strict */ ibool trx_is_strict( /*==========*/ @@ -3318,104 +3516,232 @@ trx_is_strict( } /**************************************************************//** -Resets some fields of a prebuilt struct. The template is used in fast +Resets some fields of a m_prebuilt struct. The template is used in fast retrieval of just those column values MySQL needs in its processing. */ -inline void ha_innobase::reset_template(void) /*=============================*/ { - ut_ad(prebuilt->magic_n == ROW_PREBUILT_ALLOCATED); - ut_ad(prebuilt->magic_n2 == prebuilt->magic_n); + ut_ad(m_prebuilt->magic_n == ROW_PREBUILT_ALLOCATED); + ut_ad(m_prebuilt->magic_n2 == m_prebuilt->magic_n); /* Force table to be freed in close_thread_table(). */ DBUG_EXECUTE_IF("free_table_in_fts_query", - if (prebuilt->in_fts_query) { + if (m_prebuilt->in_fts_query) { table->m_needs_reopen = true; } ); - prebuilt->keep_other_fields_on_keyread = 0; - prebuilt->read_just_key = 0; - prebuilt->in_fts_query = 0; + m_prebuilt->keep_other_fields_on_keyread = 0; + m_prebuilt->read_just_key = 0; + m_prebuilt->in_fts_query = 0; + /* Reset index condition pushdown state. */ - if (prebuilt->idx_cond) { - prebuilt->idx_cond = NULL; - prebuilt->idx_cond_n_cols = 0; - /* Invalidate prebuilt->mysql_template + if (m_prebuilt->idx_cond) { + m_prebuilt->idx_cond = NULL; + m_prebuilt->idx_cond_n_cols = 0; + /* Invalidate m_prebuilt->mysql_template in ha_innobase::write_row(). */ - prebuilt->template_type = ROW_MYSQL_NO_TEMPLATE; + m_prebuilt->template_type = ROW_MYSQL_NO_TEMPLATE; } } /*****************************************************************//** Call this when you have opened a new table handle in HANDLER, before you -call index_read_idx() etc. Actually, we can let the cursor stay open even +call index_read_map() etc. Actually, we can let the cursor stay open even over a transaction commit! Then you should call this before every operation, fetch next etc. This function inits the necessary things even after a transaction commit. */ -UNIV_INTERN + void ha_innobase::init_table_handle_for_HANDLER(void) /*============================================*/ { /* If current thd does not yet have a trx struct, create one. - If the current handle does not yet have a prebuilt struct, create - one. Update the trx pointers in the prebuilt struct. Normally + If the current handle does not yet have a m_prebuilt struct, create + one. Update the trx pointers in the m_prebuilt struct. Normally this operation is done in external_lock. */ update_thd(ha_thd()); - /* Initialize the prebuilt struct much like it would be inited in + /* Initialize the m_prebuilt struct much like it would be inited in external_lock */ - trx_search_latch_release_if_reserved(prebuilt->trx); - - innobase_srv_conc_force_exit_innodb(prebuilt->trx); + innobase_srv_conc_force_exit_innodb(m_prebuilt->trx); /* If the transaction is not started yet, start it */ - trx_start_if_not_started_xa(prebuilt->trx); + trx_start_if_not_started_xa(m_prebuilt->trx, false); /* Assign a read view if the transaction does not have it yet */ - trx_assign_read_view(prebuilt->trx); + trx_assign_read_view(m_prebuilt->trx); - innobase_register_trx(ht, user_thd, prebuilt->trx); + innobase_register_trx(ht, m_user_thd, m_prebuilt->trx); /* We did the necessary inits in this function, no need to repeat them in row_search_for_mysql */ - prebuilt->sql_stat_start = FALSE; + m_prebuilt->sql_stat_start = FALSE; /* We let HANDLER always to do the reads as consistent reads, even if the trx isolation level would have been specified as SERIALIZABLE */ - prebuilt->select_lock_type = LOCK_NONE; - prebuilt->stored_select_lock_type = LOCK_NONE; + m_prebuilt->select_lock_type = LOCK_NONE; + m_prebuilt->stored_select_lock_type = LOCK_NONE; /* Always fetch all columns in the index record */ - prebuilt->hint_need_to_fetch_extra_cols = ROW_RETRIEVE_ALL_COLS; + m_prebuilt->hint_need_to_fetch_extra_cols = ROW_RETRIEVE_ALL_COLS; /* We want always to fetch all columns in the whole row? Or do we???? */ - prebuilt->used_in_HANDLER = TRUE; + m_prebuilt->used_in_HANDLER = TRUE; + reset_template(); } +/*********************************************************************//** +Free tablespace resources allocated. */ +static +void +innobase_space_shutdown() +/*=====================*/ +{ + DBUG_ENTER("innobase_space_shutdown"); + + srv_sys_space.shutdown(); + if (srv_tmp_space.get_sanity_check_status()) { + fil_space_close(srv_tmp_space.name()); + srv_tmp_space.delete_files(); + } + srv_tmp_space.shutdown(); + +#ifdef WITH_INNODB_DISALLOW_WRITES + os_event_destroy(srv_allow_writes_event); +#endif /* WITH_INNODB_DISALLOW_WRITES */ + + DBUG_VOID_RETURN; +} + +/*********************************************************************//** +Free any resources that were allocated and return failure. +@return always return 1 */ +static +int +innobase_init_abort() +/*=================*/ +{ + DBUG_ENTER("innobase_init_abort"); + innobase_space_shutdown(); + DBUG_RETURN(1); +} + +/** Return partitioning flags. */ +static uint innobase_partition_flags() +{ + /* JAN: TODO: MYSQL 5.7 + return(HA_CAN_EXCHANGE_PARTITION | HA_CANNOT_PARTITION_FK); + */ + return (0); +} + +/** Deprecation message about InnoDB file format related parameters */ +#define DEPRECATED_FORMAT_PARAMETER(x) \ + "Using " x " is deprecated and the parameter" \ + " may be removed in future releases." \ + " See https://mariadb.com/kb/en/library/xtradbinnodb-file-format/" + +/** Deprecation message about innodb_file_format */ +static const char* deprecated_file_format + = DEPRECATED_FORMAT_PARAMETER("innodb_file_format"); + +/** Deprecation message about innodb_large_prefix */ +static const char* deprecated_large_prefix + = DEPRECATED_FORMAT_PARAMETER("innodb_large_prefix"); + +/** Deprecation message about innodb_file_format_check */ +static const char* deprecated_file_format_check + = DEPRECATED_FORMAT_PARAMETER("innodb_file_format_check"); + +/** Deprecation message about innodb_file_format_max */ +static const char* deprecated_file_format_max + = DEPRECATED_FORMAT_PARAMETER("innodb_file_format_max"); + +/** Deprecation message about innodb_use_trim */ +static const char* deprecated_use_trim + = "Using innodb_use_trim is deprecated" + " and the parameter will be removed in MariaDB 10.3."; + +/** Deprecation message about innodb_instrument_semaphores */ +static const char* deprecated_instrument_semaphores + = "Using innodb_instrument_semaphores is deprecated" + " and the parameter will be removed in MariaDB 10.3."; + +static const char* deprecated_use_mtflush + = "Using innodb_use_mtflush is deprecated" + " and the parameter will be removed in MariaDB 10.3." + " Use innodb-page-cleaners instead. "; + +static const char* deprecated_mtflush_threads + = "Using innodb_mtflush_threads is deprecated" + " and the parameter will be removed in MariaDB 10.3." + " Use innodb-page-cleaners instead. "; + +static my_bool innodb_instrument_semaphores; + +/** Update log_checksum_algorithm_ptr with a pointer to the function +corresponding to whether checksums are enabled. +@param[in,out] thd client session, or NULL if at startup +@param[in] check whether redo log block checksums are enabled +@return whether redo log block checksums are enabled */ +static inline +bool +innodb_log_checksums_func_update(THD* thd, bool check) +{ + static const char msg[] = "innodb_encrypt_log implies" + " innodb_log_checksums"; + + ut_ad(!thd == !srv_was_started); + + if (!check) { + check = srv_encrypt_log; + if (!check) { + } else if (thd) { + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_UNSUPPORTED, msg); + } else { + sql_print_warning(msg); + } + } + + if (thd) { + log_mutex_enter(); + log_checksum_algorithm_ptr = check + ? log_block_calc_checksum_crc32 + : log_block_calc_checksum_none; + log_mutex_exit(); + } else { + log_checksum_algorithm_ptr = check + ? log_block_calc_checksum_crc32 + : log_block_calc_checksum_none; + } + + return(check); +} + /****************************************************************//** Gives the file extension of an InnoDB single-table tablespace. */ static const char* ha_innobase_exts[] = { - ".ibd", - ".isl", - NullS + dot_ext[IBD], + dot_ext[ISL], + NullS }; /*********************************************************************//** Opens an InnoDB database. -@return 0 on success, error code on failure */ +@return 0 on success, 1 on failure */ static int innobase_init( @@ -3424,23 +3750,25 @@ innobase_init( { static char current_dir[3]; /*!< Set if using current lib */ int err; - bool ret; char *default_path; uint format_id; ulong num_pll_degree; DBUG_ENTER("innobase_init"); - handlerton *innobase_hton= (handlerton*) p; + handlerton* innobase_hton= (handlerton*) p; innodb_hton_ptr = innobase_hton; innobase_hton->state = SHOW_OPTION_YES; - innobase_hton->db_type= DB_TYPE_INNODB; + innobase_hton->db_type = DB_TYPE_INNODB; innobase_hton->savepoint_offset = sizeof(trx_named_savept_t); innobase_hton->close_connection = innobase_close_connection; + innobase_hton->kill_query = innobase_kill_query; innobase_hton->savepoint_set = innobase_savepoint; innobase_hton->savepoint_rollback = innobase_rollback_to_savepoint; + innobase_hton->savepoint_rollback_can_release_mdl = innobase_rollback_to_savepoint_can_release_mdl; + innobase_hton->savepoint_release = innobase_release_savepoint; innobase_hton->prepare_ordered= NULL; innobase_hton->commit_ordered= innobase_commit_ordered; @@ -3451,12 +3779,11 @@ innobase_init( innobase_hton->commit_by_xid = innobase_commit_by_xid; innobase_hton->rollback_by_xid = innobase_rollback_by_xid; innobase_hton->commit_checkpoint_request=innobase_checkpoint_request; - innobase_hton->create_cursor_read_view = innobase_create_cursor_view; - innobase_hton->set_cursor_read_view = innobase_set_cursor_view; - innobase_hton->close_cursor_read_view = innobase_close_cursor_view; innobase_hton->create = innobase_create_handler; + innobase_hton->drop_database = innobase_drop_database; innobase_hton->panic = innobase_end; + innobase_hton->partition_flags= innobase_partition_flags; innobase_hton->start_consistent_snapshot = innobase_start_trx_and_assign_read_view; @@ -3466,19 +3793,14 @@ innobase_init( innobase_hton->flags = HTON_SUPPORTS_EXTENDED_KEYS | HTON_SUPPORTS_FOREIGN_KEYS; - innobase_hton->release_temporary_latches = - innobase_release_temporary_latches; #ifdef WITH_WSREP innobase_hton->abort_transaction=wsrep_abort_transaction; innobase_hton->set_checkpoint=innobase_wsrep_set_checkpoint; innobase_hton->get_checkpoint=innobase_wsrep_get_checkpoint; innobase_hton->fake_trx_id=wsrep_fake_trx_id; #endif /* WITH_WSREP */ - innobase_hton->kill_query = innobase_kill_query; - - if (srv_file_per_table) - innobase_hton->tablefile_extensions = ha_innobase_exts; + innobase_hton->tablefile_extensions = ha_innobase_exts; innobase_hton->table_options = innodb_table_option_list; innodb_remember_check_sysvar_funcs(); @@ -3501,8 +3823,7 @@ innobase_init( test_filename)) { sql_print_error("tablename encoding has been changed"); - - goto error; + DBUG_RETURN(innobase_init_abort()); } #endif /* DBUG_OFF */ @@ -3510,22 +3831,27 @@ innobase_init( if (sizeof(ulint) == 4) { if (innobase_buffer_pool_size > UINT_MAX32) { sql_print_error( - "innobase_buffer_pool_size can't be over 4GB" + "innodb_buffer_pool_size can't be over 4GB" " on 32-bit systems"); - goto error; + DBUG_RETURN(innobase_init_abort()); } } + os_file_set_umask(my_umask); + + /* Setup the memory alloc/free tracing mechanisms before calling + any functions that could possibly allocate memory. */ + ut_new_boot(); + /* The buffer pool needs to be able to accommodate enough many pages, even for larger pages */ if (UNIV_PAGE_SIZE > UNIV_PAGE_SIZE_DEF && innobase_buffer_pool_size < (24 * 1024 * 1024)) { - ib_logf(IB_LOG_LEVEL_INFO, - "innodb_page_size= " ULINTPF " requires " - "innodb_buffer_pool_size > 24M current %lld. ", - UNIV_PAGE_SIZE, - innobase_buffer_pool_size); + ib::info() << "innodb_page_size=" + << UNIV_PAGE_SIZE << " requires " + << "innodb_buffer_pool_size > 24M current " + << innobase_buffer_pool_size; goto error; } @@ -3533,10 +3859,8 @@ innobase_init( /* Currently, Galera does not support VATS lock schedule algorithm. */ if (innodb_lock_schedule_algorithm == INNODB_LOCK_SCHEDULE_ALGORITHM_VATS && global_system_variables.wsrep_on) { - /* Do not allow InnoDB startup with VATS and Galera */ - sql_print_error("In Galera, innodb_lock_schedule_algorithm=vats" - " is not supported."); - goto error; + ib::info() << "For Galera, using innodb_lock_schedule_algorithm=fcfs"; + innodb_lock_schedule_algorithm = INNODB_LOCK_SCHEDULE_ALGORITHM_FCFS; } #endif /* WITH_WSREP */ @@ -3592,7 +3916,8 @@ innobase_init( goto error; } - os_innodb_umask = (ulint) my_umask; + innodb_check_deprecated(); + /* First calculate the default path for innodb_data_home_dir etc., in case the user has not given any value. @@ -3602,7 +3927,6 @@ innobase_init( if (mysqld_embedded) { default_path = mysql_real_data_home; - fil_path_to_mysql_datadir = mysql_real_data_home; } else { /* It's better to use current lib, to keep paths short */ current_dir[0] = FN_CURLIB; @@ -3613,46 +3937,94 @@ innobase_init( ut_a(default_path); + fil_path_to_mysql_datadir = default_path; + /* Set InnoDB initialization parameters according to the values read from MySQL .cnf file */ - /*--------------- Data files -------------------------*/ - /* The default dir for data files is the datadir of MySQL */ - srv_data_home = (innobase_data_home_dir ? innobase_data_home_dir : - default_path); - + srv_data_home = innobase_data_home_dir + ? innobase_data_home_dir : default_path; #ifdef WITH_WSREP /* If we use the wsrep API, then we need to tell the server the path to the data files (for passing it to the SST scripts): */ - wsrep_set_data_home_dir(innobase_data_home_dir); + wsrep_set_data_home_dir(srv_data_home); #endif /* WITH_WSREP */ - /* Set default InnoDB data file size to 12 MB and let it be - auto-extending. Thus users can use InnoDB in >= 4.0 without having - to specify any startup options. */ + /*--------------- Shared tablespaces -------------------------*/ + + /* Check that the value of system variable innodb_page_size was + set correctly. Its value was put into srv_page_size. If valid, + return the associated srv_page_size_shift. */ + srv_page_size_shift = innodb_page_size_validate(srv_page_size); + if (!srv_page_size_shift) { + sql_print_error("InnoDB: Invalid page size=%lu.\n", + srv_page_size); + DBUG_RETURN(innobase_init_abort()); + } + + /* Set default InnoDB temp data file size to 12 MB and let it be + auto-extending. */ if (!innobase_data_file_path) { innobase_data_file_path = (char*) "ibdata1:12M:autoextend"; } - /* Since InnoDB edits the argument in the next call, we make another - copy of it: */ + /* This is the first time univ_page_size is used. + It was initialized to 16k pages before srv_page_size was set */ + univ_page_size.copy_from( + page_size_t(srv_page_size, srv_page_size, false)); - internal_innobase_data_file_path = my_strdup(innobase_data_file_path, - MYF(MY_FAE)); + srv_sys_space.set_space_id(TRX_SYS_SPACE); + srv_sys_space.set_flags(FSP_FLAGS_PAGE_SSIZE()); + srv_sys_space.set_name("innodb_system"); + srv_sys_space.set_path(srv_data_home); - ret = (bool) srv_parse_data_file_paths_and_sizes( - internal_innobase_data_file_path); - if (ret == FALSE) { - sql_print_error( - "InnoDB: syntax error in innodb_data_file_path" - " or size specified is less than 1 megabyte"); -mem_free_and_error: - srv_free_paths_and_sizes(); - my_free(internal_innobase_data_file_path); - goto error; + /* Supports raw devices */ + if (!srv_sys_space.parse_params(innobase_data_file_path, true)) { + ib::error() << "Unable to parse innodb_data_file_path=" + << innobase_data_file_path; + DBUG_RETURN(innobase_init_abort()); + } + + /* Set default InnoDB temp data file size to 12 MB and let it be + auto-extending. */ + + if (!innobase_temp_data_file_path) { + innobase_temp_data_file_path = (char*) "ibtmp1:12M:autoextend"; + } + + /* We set the temporary tablspace id later, after recovery. + The temp tablespace doesn't support raw devices. + Set the name and path. */ + srv_tmp_space.set_name("innodb_temporary"); + srv_tmp_space.set_path(srv_data_home); + srv_tmp_space.set_flags(FSP_FLAGS_PAGE_SSIZE()); + + if (!srv_tmp_space.parse_params(innobase_temp_data_file_path, false)) { + ib::error() << "Unable to parse innodb_temp_data_file_path=" + << innobase_temp_data_file_path; + DBUG_RETURN(innobase_init_abort()); + } + + /* Perform all sanity check before we take action of deleting files*/ + if (srv_sys_space.intersection(&srv_tmp_space)) { + sql_print_error("%s and %s file names seem to be the same.", + srv_tmp_space.name(), srv_sys_space.name()); + DBUG_RETURN(innobase_init_abort()); + } + + /* ------------ UNDO tablespaces files ---------------------*/ + if (!srv_undo_dir) { + srv_undo_dir = default_path; + } + + os_normalize_path(srv_undo_dir); + + if (strchr(srv_undo_dir, ';')) { + sql_print_error("syntax error in innodb_undo_directory"); + DBUG_RETURN(innobase_init_abort()); } /* -------------- All log files ---------------------------*/ @@ -3663,43 +4035,36 @@ mem_free_and_error: srv_log_group_home_dir = default_path; } -#ifdef UNIV_LOG_ARCHIVE - /* Since innodb_log_arch_dir has no relevance under MySQL, - starting from 4.0.6 we always set it the same as - innodb_log_group_home_dir: */ + os_normalize_path(srv_log_group_home_dir); - innobase_log_arch_dir = innobase_log_group_home_dir; + if (strchr(srv_log_group_home_dir, ';')) { + sql_print_error("syntax error in innodb_log_group_home_dir"); + DBUG_RETURN(innobase_init_abort()); + } - srv_arch_dir = innobase_log_arch_dir; -#endif /* UNIG_LOG_ARCHIVE */ + if (!innobase_large_prefix) { + ib::warn() << deprecated_large_prefix; + } - srv_normalize_path_for_win(srv_log_group_home_dir); + if (!THDVAR(NULL, support_xa)) { + ib::warn() << deprecated_innodb_support_xa_off; + THDVAR(NULL, support_xa) = TRUE; + } - if (strchr(srv_log_group_home_dir, ';')) { - sql_print_error("syntax error in innodb_log_group_home_dir"); - goto mem_free_and_error; + if (innobase_file_format_name != innodb_file_format_default) { + ib::warn() << deprecated_file_format; } - if (innobase_mirrored_log_groups == 1) { - sql_print_warning( - "innodb_mirrored_log_groups is an unimplemented " - "feature and the variable will be completely " - "removed in a future version."); + if (innodb_instrument_semaphores) { + ib::warn() << deprecated_instrument_semaphores; } - if (innobase_mirrored_log_groups > 1) { - sql_print_error( - "innodb_mirrored_log_groups is an unimplemented feature and " - "the variable will be completely removed in a future version. " - "Using values other than 1 is not supported."); - goto mem_free_and_error; + if (srv_use_mtflush) { + ib::warn() << deprecated_use_mtflush; } - if (innobase_mirrored_log_groups == 0) { - /* To throw a deprecation warning message when the option is - passed, the default was changed to '0' (as a workaround). Since - the only value accepted for this option is '1', reset it to 1 */ - innobase_mirrored_log_groups = 1; + if (srv_use_mtflush && srv_mtflush_threads != MTFLUSH_DEFAULT_WORKER) { + ib::warn() << deprecated_mtflush_threads; } /* Validate the file format by animal name */ @@ -3712,7 +4077,7 @@ mem_free_and_error: sql_print_error("InnoDB: wrong innodb_file_format."); - goto mem_free_and_error; + DBUG_RETURN(innobase_init_abort()); } } else { /* Set it to the default file format id. Though this @@ -3732,6 +4097,7 @@ mem_free_and_error: /* Check innobase_file_format_check variable */ if (!innobase_file_format_check) { + ib::warn() << deprecated_file_format_check; /* Set the value to disable checking. */ srv_max_file_format_at_startup = UNIV_FORMAT_MAX + 1; @@ -3742,20 +4108,24 @@ mem_free_and_error: srv_max_file_format_at_startup = UNIV_FORMAT_MIN; } + if (innobase_file_format_max != innodb_file_format_max_default) { + ib::warn() << deprecated_file_format_max; + } + /* Did the user specify a format name that we support? As a side effect it will update the variable srv_max_file_format_at_startup */ if (innobase_file_format_validate_and_set( innobase_file_format_max) < 0) { - sql_print_error("InnoDB: invalid " - "innodb_file_format_max value: " - "should be any value up to %s or its " - "equivalent numeric id", + sql_print_error("InnoDB: invalid" + " innodb_file_format_max value:" + " should be any value up to %s or its" + " equivalent numeric id", trx_sys_file_format_id_to_name( UNIV_FORMAT_MAX)); - goto mem_free_and_error; + DBUG_RETURN(innobase_init_abort()); } if (innobase_change_buffering) { @@ -3772,10 +4142,10 @@ mem_free_and_error: } } - sql_print_error("InnoDB: invalid value " - "innodb_change_buffering=%s", + sql_print_error("InnoDB: invalid value" + " innodb_change_buffering=%s", innobase_change_buffering); - goto mem_free_and_error; + DBUG_RETURN(innobase_init_abort()); } innobase_change_buffering_inited_ok: @@ -3803,16 +4173,15 @@ innobase_change_buffering_inited_ok: } else { /* The user has not set the value. We should set it based on innodb_io_capacity. */ - srv_max_io_capacity = static_cast<ulong>( - ut_max(2 * srv_io_capacity, 2000)); + srv_max_io_capacity = + ut_max(2 * srv_io_capacity, 2000UL); } } else if (srv_max_io_capacity < srv_io_capacity) { sql_print_warning("InnoDB: innodb_io_capacity" " cannot be set higher than" - " innodb_io_capacity_max.\n" - "InnoDB: Setting" - " innodb_io_capacity to %lu\n", + " innodb_io_capacity_max." + "Setting innodb_io_capacity=%lu", srv_max_io_capacity); srv_io_capacity = srv_max_io_capacity; @@ -3822,98 +4191,61 @@ innobase_change_buffering_inited_ok: strlen(srv_buf_dump_filename), FALSE)) { sql_print_error("InnoDB: innodb_buffer_pool_filename" " cannot have colon (:) in the file name."); - goto mem_free_and_error; + DBUG_RETURN(innobase_init_abort()); } /* --------------------------------------------------*/ srv_file_flush_method_str = innobase_file_flush_method; - srv_log_file_size = (ib_uint64_t) innobase_log_file_size; - -#ifdef UNIV_LOG_ARCHIVE - srv_log_archive_on = (ulint) innobase_log_archive; -#endif /* UNIV_LOG_ARCHIVE */ - - /* Check that the value of system variable innodb_page_size was - set correctly. Its value was put into srv_page_size. If valid, - return the associated srv_page_size_shift.*/ - srv_page_size_shift = innodb_page_size_validate(srv_page_size); - if (!srv_page_size_shift) { - sql_print_error("InnoDB: Invalid page size=%lu.\n", - srv_page_size); - goto mem_free_and_error; - } - if (UNIV_PAGE_SIZE_DEF != srv_page_size) { - ib_logf(IB_LOG_LEVEL_INFO, - " innodb-page-size has been changed" - " from the default value %d to " ULINTPF ".", - UNIV_PAGE_SIZE_DEF, srv_page_size); - } + ib::info() << "innodb_page_size=" << srv_page_size; - srv_log_buffer_size = (ulint) innobase_log_buffer_size; + srv_max_undo_log_size = std::max( + srv_max_undo_log_size, + ulonglong(SRV_UNDO_TABLESPACE_SIZE_IN_PAGES) + * srv_page_size); + } - if (innobase_buffer_pool_instances == 0) { - innobase_buffer_pool_instances = 8; + if (srv_log_write_ahead_size > srv_page_size) { + srv_log_write_ahead_size = srv_page_size; + } else { + ulong srv_log_write_ahead_size_tmp = OS_FILE_LOG_BLOCK_SIZE; -#if defined(__WIN__) && !defined(_WIN64) - if (innobase_buffer_pool_size > 1331 * 1024 * 1024) { - innobase_buffer_pool_instances - = ut_min(MAX_BUFFER_POOLS, - (long) (innobase_buffer_pool_size - / (128 * 1024 * 1024))); + while (srv_log_write_ahead_size_tmp + < srv_log_write_ahead_size) { + srv_log_write_ahead_size_tmp + = srv_log_write_ahead_size_tmp * 2; + } + if (srv_log_write_ahead_size_tmp + != srv_log_write_ahead_size) { + srv_log_write_ahead_size + = srv_log_write_ahead_size_tmp / 2; } -#endif /* defined(__WIN__) && !defined(_WIN64) */ } - srv_buf_pool_size = (ulint) innobase_buffer_pool_size; - srv_buf_pool_instances = (ulint) innobase_buffer_pool_instances; - - srv_mem_pool_size = (ulint) innobase_additional_mem_pool_size; - - if (innobase_additional_mem_pool_size - != 8*1024*1024L /* the default */ ) { - ut_print_timestamp(stderr); - fprintf(stderr, - " InnoDB: Warning: Using " - "innodb_additional_mem_pool_size is DEPRECATED. " - "This option may be removed in future releases, " - "together with the option innodb_use_sys_malloc " - "and with the InnoDB's internal memory " - "allocator.\n"); - } + srv_log_buffer_size = (ulint) innobase_log_buffer_size; - if (!srv_use_sys_malloc ) { - ut_print_timestamp(stderr); - fprintf(stderr, - " InnoDB: Warning: Setting " - "innodb_use_sys_malloc to FALSE is DEPRECATED. " - "This option may be removed in future releases, " - "together with the InnoDB's internal memory " - "allocator.\n"); - } + srv_buf_pool_size = (ulint) innobase_buffer_pool_size; - srv_n_file_io_threads = (ulint) innobase_file_io_threads; srv_n_read_io_threads = (ulint) innobase_read_io_threads; srv_n_write_io_threads = (ulint) innobase_write_io_threads; srv_use_doublewrite_buf = (ibool) innobase_use_doublewrite; if (!innobase_use_checksums) { - ut_print_timestamp(stderr); - fprintf(stderr, - " InnoDB: Warning: Setting " - "innodb_checksums to OFF is DEPRECATED. " - "This option may be removed in future releases. " - "You should set innodb_checksum_algorithm=NONE " - "instead.\n"); + ib::warn() << "Setting innodb_checksums to OFF is DEPRECATED." + " This option may be removed in future releases. You" + " should set innodb_checksum_algorithm=NONE instead."; srv_checksum_algorithm = SRV_CHECKSUM_ALGORITHM_NONE; } -#ifdef HAVE_LARGE_PAGES - if ((os_use_large_pages = (ibool) my_use_large_pages)) { - os_large_page_size = (ulint) opt_large_page_size; + innodb_log_checksums = innodb_log_checksums_func_update( + NULL, innodb_log_checksums); + +#ifdef HAVE_LINUX_LARGE_PAGES + if ((os_use_large_pages = my_use_large_pages)) { + os_large_page_size = opt_large_page_size; } #endif @@ -3921,27 +4253,26 @@ innobase_change_buffering_inited_ok: srv_locks_unsafe_for_binlog = (ibool) innobase_locks_unsafe_for_binlog; if (innobase_locks_unsafe_for_binlog) { - ut_print_timestamp(stderr); - fprintf(stderr, - " InnoDB: Warning: Using " - "innodb_locks_unsafe_for_binlog is DEPRECATED. " - "This option may be removed in future releases. " - "Please use READ COMMITTED transaction isolation " - "level instead, see " REFMAN "set-transaction.html.\n"); + ib::warn() << "Using innodb_locks_unsafe_for_binlog is" + " DEPRECATED. This option may be removed in future" + " releases. Please use READ COMMITTED transaction" + " isolation level instead; " << SET_TRANSACTION_MSG; } if (innobase_open_files < 10) { innobase_open_files = 300; - if (srv_file_per_table && tc_size > 300) { + if (srv_file_per_table && tc_size > 300 && tc_size < open_files_limit) { innobase_open_files = tc_size; } } - if (innobase_open_files > (long) tc_size) { - fprintf(stderr, - "innodb_open_files should not be greater" - " than the open_files_limit.\n"); - innobase_open_files = tc_size; + if (innobase_open_files > (long) open_files_limit) { + ib::warn() << "innodb_open_files " << innobase_open_files + << " should not be greater" + << "than the open_files_limit " << open_files_limit; + if (innobase_open_files > (long) tc_size) { + innobase_open_files = tc_size; + } } srv_max_n_open_files = (ulint) innobase_open_files; @@ -3964,50 +4295,36 @@ innobase_change_buffering_inited_ok: data_mysql_default_charset_coll = (ulint) default_charset_info->number; - ut_a(DATA_MYSQL_LATIN1_SWEDISH_CHARSET_COLL == - my_charset_latin1.number); - ut_a(DATA_MYSQL_BINARY_CHARSET_COLL == my_charset_bin.number); - - /* Store the latin1_swedish_ci character ordering table to InnoDB. For - non-latin1_swedish_ci charsets we use the MySQL comparison functions, - and consequently we do not need to know the ordering internally in - InnoDB. */ - - srv_latin1_ordering = my_charset_latin1.sort_order; - innobase_commit_concurrency_init_default(); -#ifdef HAVE_POSIX_FALLOCATE - srv_use_posix_fallocate = (ibool) innobase_use_fallocate; -#endif - srv_use_atomic_writes = (ibool) innobase_use_atomic_writes; - - if (innobase_use_atomic_writes) { - fprintf(stderr, "InnoDB: using atomic writes.\n"); + if (innobase_use_fallocate) { + ib::warn() << "innodb_use_fallocate is DEPRECATED" + " and has no effect in MariaDB 10.2." + " It will be removed in MariaDB 10.3."; + } - /* Force doublewrite buffer off, atomic writes replace it. */ - if (srv_use_doublewrite_buf) { - fprintf(stderr, "InnoDB: Switching off doublewrite buffer " - "because of atomic writes.\n"); - innobase_use_doublewrite = srv_use_doublewrite_buf = FALSE; - } + srv_use_atomic_writes + = innobase_use_atomic_writes && my_may_have_atomic_write; + if (srv_use_atomic_writes && !srv_file_per_table) + { + fprintf(stderr, "InnoDB: Disabling atomic_writes as file_per_table is not used.\n"); + srv_use_atomic_writes= 0; + } - /* Force O_DIRECT on Unixes (on Windows writes are always unbuffered)*/ + if (srv_use_atomic_writes) { + fprintf(stderr, "InnoDB: using atomic writes.\n"); + /* + Force O_DIRECT on Unixes (on Windows writes are always + unbuffered) + */ #ifndef _WIN32 - if(!innobase_file_flush_method || + if (!innobase_file_flush_method || !strstr(innobase_file_flush_method, "O_DIRECT")) { innobase_file_flush_method = srv_file_flush_method_str = (char*)"O_DIRECT"; fprintf(stderr, "InnoDB: using O_DIRECT due to atomic writes.\n"); } #endif -#ifdef HAVE_POSIX_FALLOCATE - /* Due to a bug in directFS, using atomics needs - * posix_fallocate to extend the file - * pwrite() past end of the file won't work - */ - srv_use_posix_fallocate = TRUE; -#endif } #ifdef HAVE_PSI_INTERFACE @@ -4015,7 +4332,7 @@ innobase_change_buffering_inited_ok: int count; count = array_elements(all_pthread_mutexes); - mysql_mutex_register("innodb", all_pthread_mutexes, count); + mysql_mutex_register("innodb", all_pthread_mutexes, count); # ifdef UNIV_PFS_MUTEX count = array_elements(all_innodb_mutexes); @@ -4041,37 +4358,37 @@ innobase_change_buffering_inited_ok: mysql_cond_register("innodb", all_innodb_conds, count); #endif /* HAVE_PSI_INTERFACE */ - /* Since we in this module access directly the fields of a trx - struct, and due to different headers and flags it might happen that - ib_mutex_t has a different size in this module and in InnoDB - modules, we check at run time that the size is the same in - these compilation modules. */ - err = innobase_start_or_create_for_mysql(); + innobase_buffer_pool_size = static_cast<long long>(srv_buf_pool_size); + if (err != DB_SUCCESS) { - goto mem_free_and_error; + innodb_shutdown(); + DBUG_RETURN(innobase_init_abort()); + } else if (!srv_read_only_mode) { + mysql_thread_create(thd_destructor_thread_key, + &thd_destructor_thread, + NULL, thd_destructor_proxy, NULL); + while (!my_atomic_loadptr_explicit(reinterpret_cast<void**> + (&srv_running), + MY_MEMORY_ORDER_RELAXED)) + os_thread_sleep(20); } - /* Adjust the innodb_undo_logs config object */ - innobase_undo_logs_init_default_max(); + srv_was_started = true; + innodb_params_adjust(); innobase_old_blocks_pct = static_cast<uint>( buf_LRU_old_ratio_update(innobase_old_blocks_pct, TRUE)); - ibuf_max_size_update(innobase_change_buffer_max_size); + ibuf_max_size_update(srv_change_buffer_max_size); - innobase_open_tables = hash_create(200); - mysql_mutex_init(innobase_share_mutex_key, - &innobase_share_mutex, - MY_MUTEX_INIT_FAST); mysql_mutex_init(commit_cond_mutex_key, &commit_cond_m, MY_MUTEX_INIT_FAST); - mysql_cond_init(commit_cond_key, &commit_cond, NULL); + mysql_cond_init(commit_cond_key, &commit_cond, 0); mysql_mutex_init(pending_checkpoint_mutex_key, &pending_checkpoint_mutex, MY_MUTEX_INIT_FAST); - innodb_inited= 1; #ifdef MYSQL_DYNAMIC_PLUGIN if (innobase_hton != p) { innobase_hton = reinterpret_cast<handlerton*>(p); @@ -4098,9 +4415,30 @@ innobase_change_buffering_inited_ok: /* Turn on monitor counters that are default on */ srv_mon_default_on(); - DBUG_RETURN(FALSE); + + /* Unit Tests */ +#ifdef UNIV_ENABLE_UNIT_TEST_GET_PARENT_DIR + unit_test_os_file_get_parent_dir(); +#endif /* UNIV_ENABLE_UNIT_TEST_GET_PARENT_DIR */ + +#ifdef UNIV_ENABLE_UNIT_TEST_MAKE_FILEPATH + test_make_filepath(); +#endif /*UNIV_ENABLE_UNIT_TEST_MAKE_FILEPATH */ + +#ifdef UNIV_ENABLE_DICT_STATS_TEST + test_dict_stats_all(); +#endif /*UNIV_ENABLE_DICT_STATS_TEST */ + +#ifdef UNIV_ENABLE_UNIT_TEST_ROW_RAW_FORMAT_INT +# ifdef HAVE_UT_CHRONO_T + test_row_raw_format_int(); +# endif /* HAVE_UT_CHRONO_T */ +#endif /* UNIV_ENABLE_UNIT_TEST_ROW_RAW_FORMAT_INT */ + + DBUG_RETURN(0); + error: - DBUG_RETURN(TRUE); + DBUG_RETURN(1); } /** Shut down the InnoDB storage engine. @@ -4111,8 +4449,7 @@ innobase_end(handlerton*, ha_panic_function) { DBUG_ENTER("innobase_end"); - if (innodb_inited) { - + if (srv_was_started) { THD *thd= current_thd; if (thd) { // may be UNINSTALL PLUGIN statement trx_t* trx = thd_to_trx(thd); @@ -4121,15 +4458,23 @@ innobase_end(handlerton*, ha_panic_function) } } - srv_fast_shutdown = (ulint) innobase_fast_shutdown; + st_my_thread_var* running = reinterpret_cast<st_my_thread_var*>( + my_atomic_loadptr_explicit( + reinterpret_cast<void**>(&srv_running), + MY_MEMORY_ORDER_RELAXED)); + if (!abort_loop && running) { + // may be UNINSTALL PLUGIN statement + running->abort = 1; + mysql_cond_broadcast(running->current_cond); + } + + if (!srv_read_only_mode) { + pthread_join(thd_destructor_thread, NULL); + } - innodb_inited = 0; - hash_table_free(innobase_open_tables); - innobase_open_tables = NULL; innodb_shutdown(); - srv_free_paths_and_sizes(); - my_free(internal_innobase_data_file_path); - mysql_mutex_destroy(&innobase_share_mutex); + innobase_space_shutdown(); + mysql_mutex_destroy(&commit_cond_m); mysql_cond_destroy(&commit_cond); mysql_mutex_destroy(&pending_checkpoint_mutex); @@ -4138,31 +4483,8 @@ innobase_end(handlerton*, ha_panic_function) DBUG_RETURN(0); } -/****************************************************************//** -Flushes InnoDB logs to disk and makes a checkpoint. Really, a commit flushes -the logs, and the name of this function should be innobase_checkpoint. -@return TRUE if error */ -static -bool -innobase_flush_logs( -/*================*/ - handlerton* hton) /*!< in/out: InnoDB handlerton */ -{ - bool result = 0; - - DBUG_ENTER("innobase_flush_logs"); - DBUG_ASSERT(hton == innodb_hton_ptr); - - if (!srv_read_only_mode) { - log_buffer_flush_to_disk(); - } - - DBUG_RETURN(result); -} - /*****************************************************************//** Commits a transaction in an InnoDB database. */ -static void innobase_commit_low( /*================*/ @@ -4189,6 +4511,7 @@ innobase_commit_low( trx_commit_for_mysql(trx); } + trx->will_lock = 0; #ifdef WITH_WSREP if (thd && wsrep_on(thd)) { thd_proc_info(thd, tmp); } #endif /* WITH_WSREP */ @@ -4199,36 +4522,29 @@ Creates an InnoDB transaction struct for the thd if it does not yet have one. Starts a new InnoDB transaction if a transaction is not yet started. And assigns a new snapshot for a consistent read if the transaction does not yet have one. -@return 0 */ +@return 0 */ static int innobase_start_trx_and_assign_read_view( /*====================================*/ - handlerton* hton, /*!< in: Innodb handlerton */ + handlerton* hton, /*!< in: InnoDB handlerton */ THD* thd) /*!< in: MySQL thread handle of the user for whom the transaction should be committed */ { - trx_t* trx; - DBUG_ENTER("innobase_start_trx_and_assign_read_view"); DBUG_ASSERT(hton == innodb_hton_ptr); /* Create a new trx struct for thd, if it does not yet have one */ - trx = check_trx_exists(thd); - - /* This is just to play safe: release a possible FIFO ticket and - search latch. Since we can potentially reserve the trx_sys->mutex, - we have to release the search system latch first to obey the latching - order. */ - - trx_search_latch_release_if_reserved(trx); + trx_t* trx = check_trx_exists(thd); innobase_srv_conc_force_exit_innodb(trx); - /* If the transaction is not started yet, start it */ + /* The transaction should not be active yet, start it */ - trx_start_if_not_started_xa(trx); + ut_ad(!trx_is_started(trx)); + + trx_start_if_not_started_xa(trx, false); /* Assign a read view if the transaction does not have it yet. Do this only if transaction is using REPEATABLE READ isolation @@ -4241,10 +4557,10 @@ innobase_start_trx_and_assign_read_view( } else { push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_UNSUPPORTED, - "InnoDB: WITH CONSISTENT SNAPSHOT " - "was ignored because this phrase " - "can only be used with " - "REPEATABLE READ isolation level."); + "InnoDB: WITH CONSISTENT SNAPSHOT" + " was ignored because this phrase" + " can only be used with" + " REPEATABLE READ isolation level."); } /* Set the MySQL flag to mark that there is an active transaction */ @@ -4257,48 +4573,75 @@ innobase_start_trx_and_assign_read_view( static void innobase_commit_ordered_2( -/*============*/ +/*======================*/ trx_t* trx, /*!< in: Innodb transaction */ THD* thd) /*!< in: MySQL thread handle */ { DBUG_ENTER("innobase_commit_ordered_2"); - /* We need current binlog position for mysqlbackup to work. - Note, the position is current because commit_ordered is guaranteed - to be called in same sequenece as writing to binlog. */ + bool read_only = trx->read_only || trx->id == 0; -retry: - if (innobase_commit_concurrency > 0) { - mysql_mutex_lock(&commit_cond_m); - commit_threads++; + if (!read_only) { + + while (innobase_commit_concurrency > 0) { + + mysql_mutex_lock(&commit_cond_m); + + ++commit_threads; + + if (commit_threads + <= innobase_commit_concurrency) { + + mysql_mutex_unlock(&commit_cond_m); + break; + } + + --commit_threads; + + mysql_cond_wait(&commit_cond, &commit_cond_m); - if (commit_threads > innobase_commit_concurrency) { - commit_threads--; - mysql_cond_wait(&commit_cond, - &commit_cond_m); - mysql_mutex_unlock(&commit_cond_m); - goto retry; - } - else { mysql_mutex_unlock(&commit_cond_m); } + + /* The following call reads the binary log position of + the transaction being committed. + + Binary logging of other engines is not relevant to + InnoDB as all InnoDB requires is that committing + InnoDB transactions appear in the same order in the + MySQL binary log as they appear in InnoDB logs, which + is guaranteed by the server. + + If the binary log is not enabled, or the transaction + is not written to the binary log, the file name will + be a NULL pointer. */ + ulonglong pos; + + thd_binlog_pos(thd, &trx->mysql_log_file_name, &pos); + + trx->mysql_log_offset = static_cast<int64_t>(pos); + + /* Don't do write + flush right now. For group commit + to work we want to do the flush later. */ + trx->flush_log_later = true; } - unsigned long long pos; - thd_binlog_pos(thd, &trx->mysql_log_file_name, &pos); - trx->mysql_log_offset= static_cast<ib_int64_t>(pos); - /* Don't do write + flush right now. For group commit - to work we want to do the flush in the innobase_commit() - method, which runs without holding any locks. */ - trx->flush_log_later = TRUE; innobase_commit_low(trx); - trx->flush_log_later = FALSE; - if (innobase_commit_concurrency > 0) { - mysql_mutex_lock(&commit_cond_m); - commit_threads--; - mysql_cond_signal(&commit_cond); - mysql_mutex_unlock(&commit_cond_m); + if (!read_only) { + trx->flush_log_later = false; + + if (innobase_commit_concurrency > 0) { + + mysql_mutex_lock(&commit_cond_m); + + ut_ad(commit_threads > 0); + --commit_threads; + + mysql_cond_signal(&commit_cond); + + mysql_mutex_unlock(&commit_cond_m); + } } DBUG_VOID_RETURN; @@ -4319,7 +4662,7 @@ the one handling the rest of the transaction. */ static void innobase_commit_ordered( -/*============*/ +/*====================*/ handlerton *hton, /*!< in: Innodb handlerton */ THD* thd, /*!< in: MySQL thread handle of the user for whom the transaction should be committed */ @@ -4332,12 +4675,6 @@ innobase_commit_ordered( trx = check_trx_exists(thd); - /* Since we will reserve the kernel mutex, we must not be holding the - search system latch, or we will disobey the latching order. But we - already released it in innobase_xa_prepare() (if not before), so just - have an assert here.*/ - ut_ad(!trx->has_search_latch); - if (!trx_is_registered_for_2pc(trx) && trx_is_started(trx)) { /* We cannot throw error here; instead we will catch this error again in innobase_commit() and report it from there. */ @@ -4351,7 +4688,7 @@ innobase_commit_ordered( innobase_commit_ordered_2(trx, thd); - trx_set_active_commit_ordered(trx); + trx_set_active_commit_ordered(trx); DBUG_VOID_RETURN; } @@ -4359,12 +4696,13 @@ innobase_commit_ordered( /*****************************************************************//** Commits a transaction in an InnoDB database or marks an SQL statement ended. -@return 0 */ +@return 0 or deadlock error if the transaction was aborted by another + higher priority transaction. */ static int innobase_commit( /*============*/ - handlerton* hton, /*!< in: Innodb handlerton */ + handlerton* hton, /*!< in: InnoDB handlerton */ THD* thd, /*!< in: MySQL thread handle of the user for whom the transaction should be committed */ @@ -4372,20 +4710,15 @@ innobase_commit( false - the current SQL statement ended */ { - trx_t* trx; - DBUG_ENTER("innobase_commit"); + DBUG_PRINT("enter", ("commit_trx: %d", commit_trx)); DBUG_ASSERT(hton == innodb_hton_ptr); DBUG_PRINT("trans", ("ending transaction")); - trx = check_trx_exists(thd); + trx_t* trx = check_trx_exists(thd); - /* Since we will reserve the trx_sys->mutex, we have to release - the search system latch first to obey the latching order. */ - - if (trx->has_search_latch && !trx_is_active_commit_ordered(trx)) { - trx_search_latch_release_if_reserved(trx); - } + ut_ad(trx->dict_operation_lock_mode == 0); + ut_ad(trx->dict_operation == TRX_DICT_OP_NONE); /* Transaction is deregistered only in a commit or a rollback. If it is deregistered we know there cannot be resources to be freed @@ -4394,16 +4727,23 @@ innobase_commit( if (!trx_is_registered_for_2pc(trx) && trx_is_started(trx)) { - sql_print_error("Transaction not registered for MySQL 2PC, " - "but transaction is active"); + sql_print_error("Transaction not registered for MariaDB 2PC," + " but transaction is active"); } + bool read_only = trx->read_only || trx->id == 0; + DBUG_PRINT("info", ("readonly: %d", read_only)); + if (commit_trx || (!thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN))) { + DBUG_EXECUTE_IF("crash_innodb_before_commit", + DBUG_SUICIDE();); + /* Run the fast part of commit if we did not already. */ if (!trx_is_active_commit_ordered(trx)) { innobase_commit_ordered_2(trx, thd); + } /* We were instructed to commit the whole transaction, or @@ -4414,10 +4754,10 @@ innobase_commit( this one, to allow then to group commit with us. */ thd_wakeup_subsequent_commits(thd, 0); - /* We did the first part already in innobase_commit_ordered(), - Now finish by doing a write + flush of logs. */ + /* Now do a write + flush of logs. */ trx_commit_complete_for_mysql(trx); - trx_deregister_from_2pc(trx); + + trx_deregister_from_2pc(trx); } else { /* We just mark the SQL statement ended and do not do a transaction commit */ @@ -4425,7 +4765,9 @@ innobase_commit( /* If we had reserved the auto-inc lock for some table in this SQL statement we release it now */ - lock_unlock_table_autoinc(trx); + if (!read_only) { + lock_unlock_table_autoinc(trx); + } /* Store the current undo_no of the transaction so that we know where to roll back if we have to roll back the next @@ -4434,7 +4776,8 @@ innobase_commit( trx_mark_sql_stat_end(trx); } - trx->n_autoinc_rows = 0; /* Reset the number AUTO-INC rows required */ + /* Reset the number AUTO-INC rows required */ + trx->n_autoinc_rows = 0; /* This is a statement level variable. */ trx->fts_next_doc_id = 0; @@ -4446,12 +4789,12 @@ innobase_commit( /*****************************************************************//** Rolls back a transaction or the latest SQL statement. -@return 0 or error number */ +@return 0 or error number */ static int innobase_rollback( /*==============*/ - handlerton* hton, /*!< in: Innodb handlerton */ + handlerton* hton, /*!< in: InnoDB handlerton */ THD* thd, /*!< in: handle to the MySQL thread of the user whose transaction should be rolled back */ @@ -4459,79 +4802,72 @@ innobase_rollback( transaction FALSE - rollback the current statement only */ { - dberr_t error; - trx_t* trx; - DBUG_ENTER("innobase_rollback"); DBUG_ASSERT(hton == innodb_hton_ptr); DBUG_PRINT("trans", ("aborting transaction")); - trx = check_trx_exists(thd); - - /* Release a possible FIFO ticket and search latch. Since we will - reserve the trx_sys->mutex, we have to release the search system - latch first to obey the latching order. */ + trx_t* trx = check_trx_exists(thd); - trx_search_latch_release_if_reserved(trx); + ut_ad(trx->dict_operation_lock_mode == 0); + ut_ad(trx->dict_operation == TRX_DICT_OP_NONE); innobase_srv_conc_force_exit_innodb(trx); - trx->n_autoinc_rows = 0; /* Reset the number AUTO-INC rows required */ + /* Reset the number AUTO-INC rows required */ + + trx->n_autoinc_rows = 0; /* If we had reserved the auto-inc lock for some table (if we come here to roll back the latest SQL statement) we release it now before a possibly lengthy rollback */ - lock_unlock_table_autoinc(trx); /* This is a statement level variable. */ + trx->fts_next_doc_id = 0; + dberr_t error; + if (rollback_trx || !thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { error = trx_rollback_for_mysql(trx); + trx_deregister_from_2pc(trx); } else { + error = trx_rollback_last_sql_stat_for_mysql(trx); } - DBUG_RETURN(convert_error_code_to_mysql(error, 0, NULL)); + DBUG_RETURN(convert_error_code_to_mysql(error, 0, trx->mysql_thd)); } /*****************************************************************//** Rolls back a transaction -@return 0 or error number */ +@return 0 or error number */ static int innobase_rollback_trx( /*==================*/ trx_t* trx) /*!< in: transaction */ { - dberr_t error = DB_SUCCESS; - DBUG_ENTER("innobase_rollback_trx"); DBUG_PRINT("trans", ("aborting transaction")); - /* Release a possible FIFO ticket and search latch. Since we will - reserve the trx_sys->mutex, we have to release the search system - latch first to obey the latching order. */ - - trx_search_latch_release_if_reserved(trx); - innobase_srv_conc_force_exit_innodb(trx); /* If we had reserved the auto-inc lock for some table (if we come here to roll back the latest SQL statement) we release it now before a possibly lengthy rollback */ - lock_unlock_table_autoinc(trx); - if (!trx->read_only) { - error = trx_rollback_for_mysql(trx); + if (!trx->has_logged()) { + trx->will_lock = 0; + DBUG_RETURN(0); } - DBUG_RETURN(convert_error_code_to_mysql(error, 0, NULL)); + DBUG_RETURN(convert_error_code_to_mysql(trx_rollback_for_mysql(trx), + 0, trx->mysql_thd)); } @@ -4610,7 +4946,7 @@ checkpoint when necessary.*/ UNIV_INTERN void innobase_mysql_log_notify( -/*===============*/ +/*======================*/ ib_uint64_t write_lsn, /*!< in: LSN written to log file */ ib_uint64_t flush_lsn) /*!< in: LSN flushed to disk */ { @@ -4683,35 +5019,29 @@ static int innobase_rollback_to_savepoint( /*===========================*/ - handlerton* hton, /*!< in: Innodb handlerton */ + handlerton* hton, /*!< in: InnoDB handlerton */ THD* thd, /*!< in: handle to the MySQL thread of the user whose transaction should be rolled back to savepoint */ void* savepoint) /*!< in: savepoint data */ { - ib_int64_t mysql_binlog_cache_pos; - dberr_t error; - trx_t* trx; - char name[64]; DBUG_ENTER("innobase_rollback_to_savepoint"); DBUG_ASSERT(hton == innodb_hton_ptr); - trx = check_trx_exists(thd); - - /* Release a possible FIFO ticket and search latch. Since we will - reserve the trx_sys->mutex, we have to release the search system - latch first to obey the latching order. */ - - trx_search_latch_release_if_reserved(trx); + trx_t* trx = check_trx_exists(thd); innobase_srv_conc_force_exit_innodb(trx); /* TODO: use provided savepoint data area to store savepoint data */ + char name[64]; + longlong2str((ulint) savepoint, name, 36); - error = trx_rollback_to_savepoint_for_mysql( + int64_t mysql_binlog_cache_pos; + + dberr_t error = trx_rollback_to_savepoint_for_mysql( trx, name, &mysql_binlog_cache_pos); if (error == DB_SUCCESS && trx->fts_trx != NULL) { @@ -4736,17 +5066,15 @@ innobase_rollback_to_savepoint_can_release_mdl( of the user whose transaction should be rolled back to savepoint */ { - trx_t* trx; - DBUG_ENTER("innobase_rollback_to_savepoint_can_release_mdl"); DBUG_ASSERT(hton == innodb_hton_ptr); - trx = check_trx_exists(thd); - ut_ad(trx); + trx_t* trx = check_trx_exists(thd); + + /* If transaction has not acquired any locks then it is safe + to release MDL after rollback to savepoint */ + if (UT_LIST_GET_LEN(trx->lock.trx_locks) == 0) { - /* If transaction has not acquired any locks then it is safe - to release MDL after rollback to savepoint */ - if (!(UT_LIST_GET_LEN(trx->lock.trx_locks))) { DBUG_RETURN(true); } @@ -4761,7 +5089,7 @@ static int innobase_release_savepoint( /*=======================*/ - handlerton* hton, /*!< in: handlerton for Innodb */ + handlerton* hton, /*!< in: handlerton for InnoDB */ THD* thd, /*!< in: handle to the MySQL thread of the user whose transaction's savepoint should be released */ @@ -4776,10 +5104,6 @@ innobase_release_savepoint( trx = check_trx_exists(thd); - if (trx->state == TRX_STATE_NOT_STARTED) { - trx_start_if_not_started(trx); - } - /* TODO: use provided savepoint data area to store savepoint data */ longlong2str((ulint) savepoint, name, 36); @@ -4795,18 +5119,15 @@ innobase_release_savepoint( /*****************************************************************//** Sets a transaction savepoint. -@return always 0, that is, always succeeds */ +@return always 0, that is, always succeeds */ static int innobase_savepoint( /*===============*/ - handlerton* hton, /*!< in: handle to the Innodb handlerton */ - THD* thd, /*!< in: handle to the MySQL thread */ - void* savepoint) /*!< in: savepoint data */ + handlerton* hton, /*!< in: handle to the InnoDB handlerton */ + THD* thd, /*!< in: handle to the MySQL thread */ + void* savepoint)/*!< in: savepoint data */ { - dberr_t error; - trx_t* trx; - DBUG_ENTER("innobase_savepoint"); DBUG_ASSERT(hton == innodb_hton_ptr); @@ -4814,13 +5135,7 @@ innobase_savepoint( (unless we are in sub-statement), so SQL layer ensures that this method is never called in such situation. */ - trx = check_trx_exists(thd); - - /* Release a possible FIFO ticket and search latch. Since we will - reserve the trx_sys->mutex, we have to release the search system - latch first to obey the latching order. */ - - trx_search_latch_release_if_reserved(trx); + trx_t* trx = check_trx_exists(thd); innobase_srv_conc_force_exit_innodb(trx); @@ -4828,10 +5143,11 @@ innobase_savepoint( DBUG_ASSERT(trx_is_registered_for_2pc(trx)); /* TODO: use provided savepoint data area to store savepoint data */ - char name[64]; + char name[64]; + longlong2str((ulint) savepoint,name,36); - error = trx_savepoint_for_mysql(trx, name, (ib_int64_t)0); + dberr_t error = trx_savepoint_for_mysql(trx, name, 0); if (error == DB_SUCCESS && trx->fts_trx != NULL) { fts_savepoint_take(trx, trx->fts_trx, name); @@ -4842,7 +5158,7 @@ innobase_savepoint( /*****************************************************************//** Frees a possible InnoDB trx object associated with the current THD. -@return 0 or error number */ +@return 0 or error number */ static int innobase_close_connection( @@ -4851,53 +5167,63 @@ innobase_close_connection( THD* thd) /*!< in: handle to the MySQL thread of the user whose resources should be free'd */ { - trx_t* trx; DBUG_ENTER("innobase_close_connection"); DBUG_ASSERT(hton == innodb_hton_ptr); - trx = thd_to_trx(thd); - ut_a(trx); - - if (!trx_is_registered_for_2pc(trx) && trx_is_started(trx)) { + trx_t* trx = thd_to_trx(thd); - sql_print_error("Transaction not registered for MySQL 2PC, " - "but transaction is active"); - } + /* During server initialization MySQL layer will try to open + some of the master-slave tables those residing in InnoDB. + After MySQL layer is done with needed checks these tables + are closed followed by invocation of close_connection on the + associated thd. - if (trx_is_started(trx) && global_system_variables.log_warnings) { + close_connection rolls back the trx and then frees it. + Once trx is freed thd should avoid maintaining reference to + it else it can be classified as stale reference. - sql_print_warning( - "MySQL is closing a connection that has an active " - "InnoDB transaction. " TRX_ID_FMT " row modifications " - "will roll back.", - trx->undo_no); - } + Re-invocation of innodb_close_connection on same thd should + get trx as NULL. */ - innobase_rollback_trx(trx); + if (trx) { - trx_free_for_mysql(trx); + if (!trx_is_registered_for_2pc(trx) && trx_is_started(trx)) { - DBUG_RETURN(0); -} - -/*****************************************************************//** -Frees a possible InnoDB trx object associated with the current THD. -@return 0 or error number */ -UNIV_INTERN -int -innobase_close_thd( -/*===============*/ - THD* thd) /*!< in: handle to the MySQL thread of the user - whose resources should be free'd */ -{ - trx_t* trx = thd_to_trx(thd); + sql_print_error("Transaction not registered for MariaDB 2PC, " + "but transaction is active"); + } - if (!trx) { - return(0); + /* Disconnect causes rollback in the following cases: + - trx is not started, or + - trx is in *not* in PREPARED state, or + - trx has not updated any persistent data. + TODO/FIXME: it does not make sense to initiate rollback + in the 1st and 3rd case. */ + if (trx_is_started(trx)) { + if (trx_state_eq(trx, TRX_STATE_PREPARED)) { + if (trx->has_logged_persistent()) { + trx_disconnect_prepared(trx); + } else { + trx_deregister_from_2pc(trx); + goto rollback_and_free; + } + } else { + sql_print_warning( + "MariaDB is closing a connection that has an active " + "InnoDB transaction. " TRX_ID_FMT " row modifications " + "will roll back.", + trx->undo_no); + goto rollback_and_free; + } + } else { +rollback_and_free: + innobase_rollback_trx(trx); + trx_free_for_mysql(trx); + } } - return(innobase_close_connection(innodb_hton_ptr, thd)); + DBUG_RETURN(0); } UNIV_INTERN void lock_cancel_waiting_and_release(lock_t* lock); @@ -4921,35 +5247,8 @@ static void innobase_kill_query(handlerton*, THD* thd, enum thd_kill_levels) if (trx_t* trx = thd_to_trx(thd)) { ut_ad(trx->mysql_thd == thd); - - switch (trx->abort_type) { -#ifdef WITH_WSREP - case TRX_WSREP_ABORT: - break; -#endif - case TRX_SERVER_ABORT: - if (!wsrep_thd_is_BF(trx->mysql_thd, FALSE)) { - lock_mutex_enter(); - } - /* fall through */ - case TRX_REPLICATION_ABORT: - trx_mutex_enter(trx); - } /* Cancel a pending lock request if there are any */ lock_trx_handle_wait(trx); - switch (trx->abort_type) { -#ifdef WITH_WSREP - case TRX_WSREP_ABORT: - break; -#endif - case TRX_SERVER_ABORT: - if (!wsrep_thd_is_BF(trx->mysql_thd, FALSE)) { - lock_mutex_exit(); - } - /* fall through */ - case TRX_REPLICATION_ABORT: - trx_mutex_exit(trx); - } } DBUG_VOID_RETURN; @@ -4960,17 +5259,15 @@ static void innobase_kill_query(handlerton*, THD* thd, enum thd_kill_levels) ** InnoDB database tables *****************************************************************************/ -/****************************************************************//** -Get the record format from the data dictionary. +/** Get the record format from the data dictionary. @return one of ROW_TYPE_REDUNDANT, ROW_TYPE_COMPACT, ROW_TYPE_COMPRESSED, ROW_TYPE_DYNAMIC */ -UNIV_INTERN + enum row_type ha_innobase::get_row_type() const -/*=============================*/ { - if (prebuilt && prebuilt->table) { - const ulint flags = prebuilt->table->flags; + if (m_prebuilt && m_prebuilt->table) { + const ulint flags = m_prebuilt->table->flags; switch (dict_tf_get_rec_format(flags)) { case REC_FORMAT_REDUNDANT: @@ -4987,31 +5284,33 @@ ha_innobase::get_row_type() const return(ROW_TYPE_NOT_USED); } - - /****************************************************************//** Get the table flags to use for the statement. -@return table flags */ -UNIV_INTERN +@return table flags */ + handler::Table_flags ha_innobase::table_flags() const /*============================*/ { + THD* thd = ha_thd(); + handler::Table_flags flags = m_int_table_flags; + /* Need to use tx_isolation here since table flags is (also) called before prebuilt is inited. */ - ulong const tx_isolation = thd_tx_isolation(ha_thd()); + + ulong const tx_isolation = thd_tx_isolation(thd); if (tx_isolation <= ISO_READ_COMMITTED) { - return(int_table_flags); + return(flags); } - return(int_table_flags | HA_BINLOG_STMT_CAPABLE); + return(flags | HA_BINLOG_STMT_CAPABLE); } /****************************************************************//** Returns the table type (storage engine name). -@return table type */ -UNIV_INTERN +@return table type */ + const char* ha_innobase::table_type() const /*===========================*/ @@ -5022,7 +5321,7 @@ ha_innobase::table_type() const /****************************************************************//** Returns the index type. @return index type */ -UNIV_INTERN + const char* ha_innobase::index_type( /*====================*/ @@ -5032,6 +5331,8 @@ ha_innobase::index_type( if (index && index->type & DICT_FTS) { return("FULLTEXT"); + } else if (dict_index_is_spatial(index)) { + return("SPATIAL"); } else { return("BTREE"); } @@ -5039,8 +5340,8 @@ ha_innobase::index_type( /****************************************************************//** Returns the table file name extension. -@return file extension string */ -UNIV_INTERN +@return file extension string */ + const char** ha_innobase::bas_ext() const /*========================*/ @@ -5050,8 +5351,8 @@ ha_innobase::bas_ext() const /****************************************************************//** Returns the operations supported for indexes. -@return flags of supported operations */ -UNIV_INTERN +@return flags of supported operations */ + ulong ha_innobase::index_flags( /*=====================*/ @@ -5059,20 +5360,35 @@ ha_innobase::index_flags( uint, bool) const { - ulong extra_flag= 0; - if (table && key == table->s->primary_key) - extra_flag= HA_CLUSTERED_INDEX; - return((table_share->key_info[key].algorithm == HA_KEY_ALG_FULLTEXT) - ? 0 - : (HA_READ_NEXT | HA_READ_PREV | HA_READ_ORDER - | HA_READ_RANGE | HA_KEYREAD_ONLY | extra_flag - | HA_DO_INDEX_COND_PUSHDOWN)); + if (table_share->key_info[key].algorithm == HA_KEY_ALG_FULLTEXT) { + return(0); + } + + ulong extra_flag= 0; + + if (table && key == table->s->primary_key) { + extra_flag= HA_CLUSTERED_INDEX; + } + + ulong flags = HA_READ_NEXT | HA_READ_PREV | HA_READ_ORDER + | HA_READ_RANGE | HA_KEYREAD_ONLY + | extra_flag + | HA_DO_INDEX_COND_PUSHDOWN; + + /* For spatial index, we don't support descending scan + and ICP so far. */ + if (table_share->key_info[key].flags & HA_SPATIAL) { + flags = HA_READ_NEXT | HA_READ_ORDER| HA_READ_RANGE + | HA_KEYREAD_ONLY | HA_KEY_SCAN_NOT_ROR; + } + + return(flags); } /****************************************************************//** Returns the maximum number of keys. -@return MAX_KEY */ -UNIV_INTERN +@return MAX_KEY */ + uint ha_innobase::max_supported_keys() const /*===================================*/ @@ -5082,8 +5398,8 @@ ha_innobase::max_supported_keys() const /****************************************************************//** Returns the maximum key length. -@return maximum supported key length, in bytes */ -UNIV_INTERN +@return maximum supported key length, in bytes */ + uint ha_innobase::max_supported_key_length() const /*=========================================*/ @@ -5105,22 +5421,23 @@ ha_innobase::max_supported_key_length() const switch (UNIV_PAGE_SIZE) { case 4096: - return(768); + /* Hack: allow mysql.innodb_index_stats to be created. */ + /* FIXME: rewrite this API, and in sql_table.cc consider + that in index-organized tables (such as InnoDB), secondary + index records will be padded with the PRIMARY KEY, instead + of some short ROWID or record heap address. */ + return(1173); case 8192: return(1536); default: -#ifdef WITH_WSREP return(3500); -#else - return(3500); -#endif } } /****************************************************************//** Returns the key map of keys that are usable for scanning. -@return key_map_full */ -UNIV_INTERN +@return key_map_full */ + const key_map* ha_innobase::keys_to_use_for_scanning() /*===================================*/ @@ -5129,9 +5446,39 @@ ha_innobase::keys_to_use_for_scanning() } /****************************************************************//** +Ensures that if there's a concurrent inplace ADD INDEX, being-indexed virtual +columns are computed. They are not marked as indexed in the old table, so the +server won't add them to the vcol_set automatically */ +void +ha_innobase::column_bitmaps_signal() +/*================================*/ +{ + if (!table->vfield || table->current_lock != F_WRLCK) { + return; + } + + dict_index_t* clust_index = dict_table_get_first_index(m_prebuilt->table); + uint num_v = 0; + for (uint j = 0; j < table->s->virtual_fields; j++) { + if (table->vfield[j]->stored_in_db()) { + continue; + } + + dict_col_t* col = &m_prebuilt->table->v_cols[num_v].m_col; + if (col->ord_part || + (dict_index_is_online_ddl(clust_index) && + row_log_col_is_indexed(clust_index, num_v))) { + table->mark_virtual_col(table->vfield[j]); + } + num_v++; + } +} + + +/****************************************************************//** Determines if table caching is supported. -@return HA_CACHE_TBL_ASKTRANSACT */ -UNIV_INTERN +@return HA_CACHE_TBL_ASKTRANSACT */ + uint8 ha_innobase::table_cache_type() /*===========================*/ @@ -5141,8 +5488,8 @@ ha_innobase::table_cache_type() /****************************************************************//** Determines if the primary key is clustered index. -@return true */ -UNIV_INTERN +@return true */ + bool ha_innobase::primary_key_is_clustered() /*===================================*/ @@ -5150,19 +5497,22 @@ ha_innobase::primary_key_is_clustered() return(true); } -/*****************************************************************//** -Normalizes a table name string. A normalized name consists of the -database name catenated to '/' and table name. Example: test/mytable. -On Windows normalization puts both the database name and the -table name always to lower case if "set_lower_case" is set to TRUE. */ +/** Normalizes a table name string. +A normalized name consists of the database name catenated to '/' +and table name. For example: test/mytable. +On Windows, normalization puts both the database name and the +table name always to lower case if "set_lower_case" is set to TRUE. +@param[out] norm_name Normalized name, null-terminated. +@param[in] name Name to normalize. +@param[in] set_lower_case True if we also should fold to lower case. */ void -normalize_table_name_low( -/*=====================*/ - char* norm_name, /*!< out: normalized name as a +normalize_table_name_c_low( +/*=======================*/ + char* norm_name, /* out: normalized name as a null-terminated string */ - const char* name, /*!< in: table name string */ - ibool set_lower_case) /*!< in: TRUE if we want to set name - to lower case */ + const char* name, /* in: table name string */ + ibool set_lower_case) /* in: TRUE if we want to set + name to lower case */ { char* name_ptr; ulint name_len; @@ -5215,6 +5565,23 @@ normalize_table_name_low( } } +/** Normalizes a table name string. +A normalized name consists of the database name catenated to '/' +and table name. For example: test/mytable. +On Windows, normalization puts both the database name and the +table name always to lower case if "set_lower_case" is set to TRUE. +@param[out] norm_name Normalized name, null-terminated. +@param[in] name Name to normalize. +@param[in] set_lower_case True if we also should fold to lower case. */ +void +create_table_info_t::normalize_table_name_low( + char* norm_name, + const char* name, + ibool set_lower_case) +{ + normalize_table_name_c_low(norm_name, name, set_lower_case); +} + #if !defined(DBUG_OFF) /********************************************************************* Test normalize_table_name_low(). */ @@ -5265,11 +5632,12 @@ test_normalize_table_name_low() }; for (size_t i = 0; i < UT_ARR_SIZE(test_data); i++) { - printf("test_normalize_table_name_low(): " - "testing \"%s\", expected \"%s\"... ", + printf("test_normalize_table_name_low():" + " testing \"%s\", expected \"%s\"... ", test_data[i][0], test_data[i][1]); - normalize_table_name_low(norm_name, test_data[i][0], FALSE); + create_table_info_t::normalize_table_name_low( + norm_name, test_data[i][0], FALSE); if (strcmp(norm_name, test_data[i][1]) == 0) { printf("ok\n"); @@ -5291,30 +5659,27 @@ test_ut_format_name() struct { const char* name; - ibool is_table; ulint buf_size; const char* expected; } test_data[] = { - {"test/t1", TRUE, sizeof(buf), "\"test\".\"t1\""}, - {"test/t1", TRUE, 12, "\"test\".\"t1\""}, - {"test/t1", TRUE, 11, "\"test\".\"t1"}, - {"test/t1", TRUE, 10, "\"test\".\"t"}, - {"test/t1", TRUE, 9, "\"test\".\""}, - {"test/t1", TRUE, 8, "\"test\"."}, - {"test/t1", TRUE, 7, "\"test\""}, - {"test/t1", TRUE, 6, "\"test"}, - {"test/t1", TRUE, 5, "\"tes"}, - {"test/t1", TRUE, 4, "\"te"}, - {"test/t1", TRUE, 3, "\"t"}, - {"test/t1", TRUE, 2, "\""}, - {"test/t1", TRUE, 1, ""}, - {"test/t1", TRUE, 0, "BUF_NOT_CHANGED"}, - {"table", TRUE, sizeof(buf), "\"table\""}, - {"ta'le", TRUE, sizeof(buf), "\"ta'le\""}, - {"ta\"le", TRUE, sizeof(buf), "\"ta\"\"le\""}, - {"ta`le", TRUE, sizeof(buf), "\"ta`le\""}, - {"index", FALSE, sizeof(buf), "\"index\""}, - {"ind/ex", FALSE, sizeof(buf), "\"ind/ex\""}, + {"test/t1", sizeof(buf), "`test`.`t1`"}, + {"test/t1", 12, "`test`.`t1`"}, + {"test/t1", 11, "`test`.`t1"}, + {"test/t1", 10, "`test`.`t"}, + {"test/t1", 9, "`test`.`"}, + {"test/t1", 8, "`test`."}, + {"test/t1", 7, "`test`"}, + {"test/t1", 6, "`test"}, + {"test/t1", 5, "`tes"}, + {"test/t1", 4, "`te"}, + {"test/t1", 3, "`t"}, + {"test/t1", 2, "`"}, + {"test/t1", 1, ""}, + {"test/t1", 0, "BUF_NOT_CHANGED"}, + {"table", sizeof(buf), "`table`"}, + {"ta'le", sizeof(buf), "`ta'le`"}, + {"ta\"le", sizeof(buf), "`ta\"le`"}, + {"ta`le", sizeof(buf), "`ta``le`"}, }; for (size_t i = 0; i < UT_ARR_SIZE(test_data); i++) { @@ -5324,109 +5689,38 @@ test_ut_format_name() char* ret; ret = ut_format_name(test_data[i].name, - test_data[i].is_table, buf, test_data[i].buf_size); ut_a(ret == buf); if (strcmp(buf, test_data[i].expected) == 0) { - fprintf(stderr, - "ut_format_name(%s, %s, buf, %lu), " - "expected %s, OK\n", - test_data[i].name, - test_data[i].is_table ? "TRUE" : "FALSE", - test_data[i].buf_size, - test_data[i].expected); + ib::info() << "ut_format_name(" << test_data[i].name + << ", buf, " << test_data[i].buf_size << ")," + " expected " << test_data[i].expected + << ", OK"; } else { - fprintf(stderr, - "ut_format_name(%s, %s, buf, %lu), " - "expected %s, ERROR: got %s\n", - test_data[i].name, - test_data[i].is_table ? "TRUE" : "FALSE", - test_data[i].buf_size, - test_data[i].expected, - buf); + ib::error() << "ut_format_name(" << test_data[i].name + << ", buf, " << test_data[i].buf_size << ")," + " expected " << test_data[i].expected + << ", ERROR: got " << buf; ut_error; } } } #endif /* !DBUG_OFF */ -/********************************************************************//** -Get the upper limit of the MySQL integral and floating-point type. -@return maximum allowed value for the field */ -UNIV_INTERN -ulonglong -innobase_get_int_col_max_value( -/*===========================*/ - const Field* field) /*!< in: MySQL field */ -{ - ulonglong max_value = 0; - - switch (field->key_type()) { - /* TINY */ - case HA_KEYTYPE_BINARY: - max_value = 0xFFULL; - break; - case HA_KEYTYPE_INT8: - max_value = 0x7FULL; - break; - /* SHORT */ - case HA_KEYTYPE_USHORT_INT: - max_value = 0xFFFFULL; - break; - case HA_KEYTYPE_SHORT_INT: - max_value = 0x7FFFULL; - break; - /* MEDIUM */ - case HA_KEYTYPE_UINT24: - max_value = 0xFFFFFFULL; - break; - case HA_KEYTYPE_INT24: - max_value = 0x7FFFFFULL; - break; - /* LONG */ - case HA_KEYTYPE_ULONG_INT: - max_value = 0xFFFFFFFFULL; - break; - case HA_KEYTYPE_LONG_INT: - max_value = 0x7FFFFFFFULL; - break; - /* BIG */ - case HA_KEYTYPE_ULONGLONG: - max_value = 0xFFFFFFFFFFFFFFFFULL; - break; - case HA_KEYTYPE_LONGLONG: - max_value = 0x7FFFFFFFFFFFFFFFULL; - break; - case HA_KEYTYPE_FLOAT: - /* We use the maximum as per IEEE754-2008 standard, 2^24 */ - max_value = 0x1000000ULL; - break; - case HA_KEYTYPE_DOUBLE: - /* We use the maximum as per IEEE754-2008 standard, 2^53 */ - max_value = 0x20000000000000ULL; - break; - default: - ut_error; - } - - return(max_value); -} - -/*******************************************************************//** +/** Match index columns between MySQL and InnoDB. This function checks whether the index column information is consistent between KEY info from mysql and that from innodb index. -@return TRUE if all column types match. */ +@param[in] key_info Index info from mysql +@param[in] index_info Index info from InnoDB +@return true if all column types match. */ static -ibool +bool innobase_match_index_columns( -/*=========================*/ - const KEY* key_info, /*!< in: Index info - from mysql */ - const dict_index_t* index_info) /*!< in: Index info - from Innodb */ + const KEY* key_info, + const dict_index_t* index_info) { const KEY_PART_INFO* key_part; const KEY_PART_INFO* key_end; @@ -5451,7 +5745,7 @@ innobase_match_index_columns( column name got modified in mysql but such change does not propagate to InnoDB. One hidden assumption here is that the index column sequences - are matched up between those in mysql and Innodb. */ + are matched up between those in mysql and InnoDB. */ for (; key_part != key_end; ++key_part) { ulint col_type; ibool is_unsigned; @@ -5459,10 +5753,10 @@ innobase_match_index_columns( /* Need to translate to InnoDB column type before comparison. */ - col_type = get_innobase_type_from_mysql_type(&is_unsigned, - key_part->field); + col_type = get_innobase_type_from_mysql_type( + &is_unsigned, key_part->field); - /* Ignore Innodb specific system columns. */ + /* Ignore InnoDB specific system columns. */ while (mtype == DATA_SYS) { innodb_idx_fld++; @@ -5471,15 +5765,31 @@ innobase_match_index_columns( } } - // MariaDB-5.5 compatibility - if ((key_part->field->real_type() == MYSQL_TYPE_ENUM || - key_part->field->real_type() == MYSQL_TYPE_SET) && - mtype == DATA_FIXBINARY) - col_type= DATA_FIXBINARY; + /* MariaDB-5.5 compatibility */ + if ((key_part->field->real_type() == MYSQL_TYPE_ENUM || + key_part->field->real_type() == MYSQL_TYPE_SET) && + mtype == DATA_FIXBINARY) { + col_type= DATA_FIXBINARY; + } if (col_type != mtype) { - /* Column Type mismatches */ - DBUG_RETURN(FALSE); + /* If the col_type we get from mysql type is a geometry + data type, we should check if mtype is a legacy type + from 5.6, either upgraded to DATA_GEOMETRY or not. + This is indeed not an accurate check, but should be + safe, since DATA_BLOB would be upgraded once we create + spatial index on it and we intend to use DATA_GEOMETRY + for legacy GIS data types which are of var-length. */ + switch (col_type) { + case DATA_GEOMETRY: + if (mtype == DATA_BLOB) { + break; + } + /* Fall through */ + default: + /* Column type mismatches */ + DBUG_RETURN(false); + } } innodb_idx_fld++; @@ -5488,80 +5798,230 @@ innobase_match_index_columns( DBUG_RETURN(TRUE); } -/*******************************************************************//** -This function builds a translation table in INNOBASE_SHARE -structure for fast index location with mysql array number from its -table->key_info structure. This also provides the necessary translation -between the key order in mysql key_info and Innodb ib_table->indexes if -they are not fully matched with each other. -Note we do not have any mutex protecting the translation table -building based on the assumption that there is no concurrent -index creation/drop and DMLs that requires index lookup. All table -handle will be closed before the index creation/drop. -@return TRUE if index translation table built successfully */ +/** Build a template for a base column for a virtual column +@param[in] table MySQL TABLE +@param[in] clust_index InnoDB clustered index +@param[in] field field in MySQL table +@param[in] col InnoDB column +@param[in,out] templ template to fill +@param[in] col_no field index for virtual col +*/ static -ibool -innobase_build_index_translation( -/*=============================*/ - const TABLE* table, /*!< in: table in MySQL data - dictionary */ - dict_table_t* ib_table,/*!< in: table in Innodb data - dictionary */ - INNOBASE_SHARE* share) /*!< in/out: share structure - where index translation table - will be constructed in. */ +void +innobase_vcol_build_templ( + const TABLE* table, + dict_index_t* clust_index, + Field* field, + const dict_col_t* col, + mysql_row_templ_t* templ, + ulint col_no) +{ + if (dict_col_is_virtual(col)) { + templ->is_virtual = true; + templ->col_no = col_no; + templ->clust_rec_field_no = ULINT_UNDEFINED; + templ->rec_field_no = col->ind; + } else { + templ->is_virtual = false; + templ->col_no = col_no; + templ->clust_rec_field_no = dict_col_get_clust_pos( + col, clust_index); + ut_a(templ->clust_rec_field_no != ULINT_UNDEFINED); + + templ->rec_field_no = templ->clust_rec_field_no; + } + + if (field->real_maybe_null()) { + templ->mysql_null_byte_offset = + field->null_offset(); + + templ->mysql_null_bit_mask = (ulint) field->null_bit; + } else { + templ->mysql_null_bit_mask = 0; + } + + templ->mysql_col_offset = static_cast<ulint>( + get_field_offset(table, field)); + templ->mysql_col_len = static_cast<ulint>(field->pack_length()); + templ->type = col->mtype; + templ->mysql_type = static_cast<ulint>(field->type()); + + if (templ->mysql_type == DATA_MYSQL_TRUE_VARCHAR) { + templ->mysql_length_bytes = static_cast<ulint>( + ((Field_varstring*) field)->length_bytes); + } + + templ->charset = dtype_get_charset_coll(col->prtype); + templ->mbminlen = dict_col_get_mbminlen(col); + templ->mbmaxlen = dict_col_get_mbmaxlen(col); + templ->is_unsigned = col->prtype & DATA_UNSIGNED; +} + +/** Build template for the virtual columns and their base columns. This +is done when the table first opened. +@param[in] table MySQL TABLE +@param[in] ib_table InnoDB dict_table_t +@param[in,out] s_templ InnoDB template structure +@param[in] add_v new virtual columns added along with + add index call +@param[in] locked true if dict_sys mutex is held */ +void +innobase_build_v_templ( + const TABLE* table, + const dict_table_t* ib_table, + dict_vcol_templ_t* s_templ, + const dict_add_v_col_t* add_v, + bool locked) { - ulint mysql_num_index; - ulint ib_num_index; - dict_index_t** index_mapping; - ibool ret = TRUE; + ulint ncol = ib_table->n_cols - DATA_N_SYS_COLS; + ulint n_v_col = ib_table->n_v_cols; + bool marker[REC_MAX_N_FIELDS]; - DBUG_ENTER("innobase_build_index_translation"); + ut_ad(ncol < REC_MAX_N_FIELDS); - mutex_enter(&dict_sys->mutex); + if (add_v != NULL) { + n_v_col += add_v->n_v_col; + } - mysql_num_index = table->s->keys; - ib_num_index = UT_LIST_GET_LEN(ib_table->indexes); + ut_ad(n_v_col > 0); - index_mapping = share->idx_trans_tbl.index_mapping; + if (!locked) { + mutex_enter(&dict_sys->mutex); + } - /* If there exists inconsistency between MySQL and InnoDB dictionary - (metadata) information, the number of index defined in MySQL - could exceed that in InnoDB, do not build index translation - table in such case */ - if (UNIV_UNLIKELY(ib_num_index < mysql_num_index)) { - ret = FALSE; - goto func_exit; + if (s_templ->vtempl) { + if (!locked) { + mutex_exit(&dict_sys->mutex); + } + return; } - /* If index entry count is non-zero, nothing has - changed since last update, directly return TRUE */ - if (share->idx_trans_tbl.index_count) { - /* Index entry count should still match mysql_num_index */ - ut_a(share->idx_trans_tbl.index_count == mysql_num_index); - goto func_exit; + memset(marker, 0, sizeof(bool) * ncol); + + s_templ->vtempl = static_cast<mysql_row_templ_t**>( + ut_zalloc_nokey((ncol + n_v_col) + * sizeof *s_templ->vtempl)); + s_templ->n_col = ncol; + s_templ->n_v_col = n_v_col; + s_templ->rec_len = table->s->reclength; + s_templ->default_rec = table->s->default_values; + + /* Mark those columns could be base columns */ + for (ulint i = 0; i < ib_table->n_v_cols; i++) { + const dict_v_col_t* vcol = dict_table_get_nth_v_col( + ib_table, i); + + for (ulint j = 0; j < vcol->num_base; j++) { + ulint col_no = vcol->base_col[j]->ind; + marker[col_no] = true; + } } - /* The number of index increased, rebuild the mapping table */ - if (mysql_num_index > share->idx_trans_tbl.array_size) { - index_mapping = (dict_index_t**) my_realloc(index_mapping, - mysql_num_index * - sizeof(*index_mapping), - MYF(MY_ALLOW_ZERO_PTR)); - - if (!index_mapping) { - /* Report an error if index_mapping continues to be - NULL and mysql_num_index is a non-zero value */ - sql_print_error("InnoDB: fail to allocate memory for " - "index translation table. Number of " - "Index:%lu, array size:%lu", - mysql_num_index, - share->idx_trans_tbl.array_size); - ret = FALSE; - goto func_exit; + if (add_v) { + for (ulint i = 0; i < add_v->n_v_col; i++) { + const dict_v_col_t* vcol = &add_v->v_col[i]; + + for (ulint j = 0; j < vcol->num_base; j++) { + ulint col_no = vcol->base_col[j]->ind; + marker[col_no] = true; + } + } + } + + ulint j = 0; + ulint z = 0; + + dict_index_t* clust_index = dict_table_get_first_index(ib_table); + + for (ulint i = 0; i < table->s->fields; i++) { + Field* field = table->field[i]; + + /* Build template for virtual columns */ + if (innobase_is_v_fld(field)) { +#ifdef UNIV_DEBUG + const char* name; + + if (z >= ib_table->n_v_def) { + name = add_v->v_col_name[z - ib_table->n_v_def]; + } else { + name = dict_table_get_v_col_name(ib_table, z); + } + + ut_ad(!my_strcasecmp(system_charset_info, name, + field->field_name)); +#endif + const dict_v_col_t* vcol; + + if (z >= ib_table->n_v_def) { + vcol = &add_v->v_col[z - ib_table->n_v_def]; + } else { + vcol = dict_table_get_nth_v_col(ib_table, z); + } + + s_templ->vtempl[z + s_templ->n_col] + = static_cast<mysql_row_templ_t*>( + ut_malloc_nokey( + sizeof *s_templ->vtempl[j])); + + innobase_vcol_build_templ( + table, clust_index, field, + &vcol->m_col, + s_templ->vtempl[z + s_templ->n_col], + z); + z++; + continue; + } + + ut_ad(j < ncol); + + /* Build template for base columns */ + if (marker[j]) { + dict_col_t* col = dict_table_get_nth_col( + ib_table, j); + + ut_ad(!my_strcasecmp(system_charset_info, + dict_table_get_col_name( + ib_table, j), + field->field_name)); + + s_templ->vtempl[j] = static_cast< + mysql_row_templ_t*>( + ut_malloc_nokey( + sizeof *s_templ->vtempl[j])); + + innobase_vcol_build_templ( + table, clust_index, field, col, + s_templ->vtempl[j], j); } - share->idx_trans_tbl.array_size = mysql_num_index; + j++; + } + + if (!locked) { + mutex_exit(&dict_sys->mutex); + } + + s_templ->db_name = table->s->db.str; + s_templ->tb_name = table->s->table_name.str; +} + +/** Check consistency between .frm indexes and InnoDB indexes. +@param[in] table table object formed from .frm +@param[in] ib_table InnoDB table definition +@retval true if not errors were found */ +static bool +check_index_consistency(const TABLE* table, const dict_table_t* ib_table) +{ + ulint mysql_num_index = table->s->keys; + ulint ib_num_index = UT_LIST_GET_LEN(ib_table->indexes); + bool ret = true; + + /* If there exists inconsistency between MySQL and InnoDB dictionary + (metadata) information, the number of index defined in MySQL + could exceed that in InnoDB, return error */ + if (ib_num_index < mysql_num_index) { + ret = false; + goto func_exit; } /* For each index in the mysql key_info array, fetch its @@ -5569,102 +6029,123 @@ innobase_build_index_translation( array. */ for (ulint count = 0; count < mysql_num_index; count++) { - /* Fetch index pointers into index_mapping according to mysql - index sequence */ - index_mapping[count] = dict_table_get_index_on_name( - ib_table, table->key_info[count].name); + const dict_index_t* index = dict_table_get_index_on_name( + ib_table, table->key_info[count].name); - if (!index_mapping[count]) { - sql_print_error("Cannot find index %s in InnoDB " - "index dictionary.", + if (index == NULL) { + sql_print_error("Cannot find index %s in InnoDB" + " index dictionary.", table->key_info[count].name); - ret = FALSE; + ret = false; goto func_exit; } /* Double check fetched index has the same column info as those in mysql key_info. */ if (!innobase_match_index_columns(&table->key_info[count], - index_mapping[count])) { - sql_print_error("Found index %s whose column info " - "does not match that of MySQL.", + index)) { + sql_print_error("Found index %s whose column info" + " does not match that of MariaDB.", table->key_info[count].name); - ret = FALSE; + ret = false; goto func_exit; } } - /* Successfully built the translation table */ - share->idx_trans_tbl.index_count = mysql_num_index; - func_exit: - if (!ret) { - /* Build translation table failed. */ - my_free(index_mapping); - - share->idx_trans_tbl.array_size = 0; - share->idx_trans_tbl.index_count = 0; - index_mapping = NULL; - } - - share->idx_trans_tbl.index_mapping = index_mapping; - - mutex_exit(&dict_sys->mutex); - - DBUG_RETURN(ret); + return ret; } -/*******************************************************************//** -This function uses index translation table to quickly locate the -requested index structure. -Note we do not have mutex protection for the index translatoin table -access, it is based on the assumption that there is no concurrent -translation table rebuild (fter create/drop index) and DMLs that -require index lookup. -@return dict_index_t structure for requested index. NULL if -fail to locate the index structure. */ -static -dict_index_t* -innobase_index_lookup( -/*==================*/ - INNOBASE_SHARE* share, /*!< in: share structure for index - translation table. */ - uint keynr) /*!< in: index number for the requested - index */ +/********************************************************************//** +Get the upper limit of the MySQL integral and floating-point type. +@return maximum allowed value for the field */ +UNIV_INTERN +ulonglong +innobase_get_int_col_max_value( +/*===========================*/ + const Field* field) /*!< in: MySQL field */ { - if (!share->idx_trans_tbl.index_mapping - || keynr >= share->idx_trans_tbl.index_count) { - return(NULL); + ulonglong max_value = 0; + + switch (field->key_type()) { + /* TINY */ + case HA_KEYTYPE_BINARY: + max_value = 0xFFULL; + break; + case HA_KEYTYPE_INT8: + max_value = 0x7FULL; + break; + /* SHORT */ + case HA_KEYTYPE_USHORT_INT: + max_value = 0xFFFFULL; + break; + case HA_KEYTYPE_SHORT_INT: + max_value = 0x7FFFULL; + break; + /* MEDIUM */ + case HA_KEYTYPE_UINT24: + max_value = 0xFFFFFFULL; + break; + case HA_KEYTYPE_INT24: + max_value = 0x7FFFFFULL; + break; + /* LONG */ + case HA_KEYTYPE_ULONG_INT: + max_value = 0xFFFFFFFFULL; + break; + case HA_KEYTYPE_LONG_INT: + max_value = 0x7FFFFFFFULL; + break; + /* BIG */ + case HA_KEYTYPE_ULONGLONG: + max_value = 0xFFFFFFFFFFFFFFFFULL; + break; + case HA_KEYTYPE_LONGLONG: + max_value = 0x7FFFFFFFFFFFFFFFULL; + break; + case HA_KEYTYPE_FLOAT: + /* We use the maximum as per IEEE754-2008 standard, 2^24 */ + max_value = 0x1000000ULL; + break; + case HA_KEYTYPE_DOUBLE: + /* We use the maximum as per IEEE754-2008 standard, 2^53 */ + max_value = 0x20000000000000ULL; + break; + default: + ut_error; } - return(share->idx_trans_tbl.index_mapping[keynr]); + return(max_value); } -/************************************************************************ -Set the autoinc column max value. This should only be called once from -ha_innobase::open(). Therefore there's no need for a covering lock. */ -UNIV_INTERN +/** Initialize the AUTO_INCREMENT column metadata. + +Since a partial table definition for a persistent table can already be +present in the InnoDB dict_sys cache before it is accessed from SQL, +we have to initialize the AUTO_INCREMENT counter on the first +ha_innobase::open(). + +@param[in,out] table persistent table +@param[in] field the AUTO_INCREMENT column */ +static void -ha_innobase::innobase_initialize_autoinc() -/*======================================*/ +initialize_auto_increment(dict_table_t* table, const Field* field) { - ulonglong auto_inc; - const Field* field = table->found_next_number_field; + ut_ad(!dict_table_is_temporary(table)); - if (field != NULL) { - auto_inc = innobase_get_int_col_max_value(field); - } else { - /* We have no idea what's been passed in to us as the - autoinc column. We set it to the 0, effectively disabling - updates to the table. */ - auto_inc = 0; + const unsigned col_no = innodb_col_no(field); - ut_print_timestamp(stderr); - fprintf(stderr, " InnoDB: Unable to determine the AUTOINC " - "column name\n"); - } + dict_table_autoinc_lock(table); - if (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE) { + table->persistent_autoinc = 1 + + dict_table_get_nth_col_pos(table, col_no, NULL); + + if (table->autoinc) { + /* Already initialized. Our caller checked + table->persistent_autoinc without + dict_table_autoinc_lock(), and there might be multiple + ha_innobase::open() executing concurrently. */ + } else if (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE) { /* If the recovery level is set so high that writes are disabled we force the AUTOINC counter to 0 value effectively disabling writes to the table. @@ -5675,132 +6156,41 @@ ha_innobase::innobase_initialize_autoinc() tables can be dumped with minimal hassle. If an error were returned in this case, the first attempt to read the table would fail and subsequent SELECTs would succeed. */ - auto_inc = 0; - } else if (field == NULL) { - /* This is a far more serious error, best to avoid - opening the table and return failure. */ - my_error(ER_AUTOINC_READ_FAILED, MYF(0)); - } else { - dict_index_t* index; - const char* col_name; - ib_uint64_t read_auto_inc; - ulint err; - - update_thd(ha_thd()); - - ut_a(prebuilt->trx == thd_to_trx(user_thd)); - - col_name = field->field_name; - index = innobase_get_index(table->s->next_number_index); - - /* Execute SELECT MAX(col_name) FROM TABLE; */ - err = row_search_max_autoinc(index, col_name, &read_auto_inc); - - switch (err) { - case DB_SUCCESS: { - ulonglong col_max_value; - - col_max_value = innobase_get_int_col_max_value(field); - - /* At the this stage we do not know the increment - nor the offset, so use a default increment of 1. */ - - auto_inc = innobase_next_autoinc( - read_auto_inc, 1, 1, 0, col_max_value); - - break; - } - case DB_RECORD_NOT_FOUND: - char buf[MAX_FULL_NAME_LEN]; - ut_format_name(index->table->name, TRUE, buf, sizeof(buf)); - - ib_logf(IB_LOG_LEVEL_ERROR, - "MySQL and InnoDB data " - "dictionaries are out of sync." - " Unable to find the AUTOINC column " - " %s in the InnoDB table %s." - " We set the next AUTOINC column " - "value to 0" - " in effect disabling the AUTOINC " - "next value generation." - " You can either set the next " - "AUTOINC value explicitly using ALTER TABLE " - " or fix the data dictionary by " - "recreating the table.", - col_name, buf); - - /* This will disable the AUTOINC generation. */ - auto_inc = 0; - - /* We want the open to succeed, so that the user can - take corrective action. ie. reads should succeed but - updates should fail. */ - err = DB_SUCCESS; - break; - default: - /* row_search_max_autoinc() should only return - one of DB_SUCCESS or DB_RECORD_NOT_FOUND. */ - ut_error; - } + } else if (table->persistent_autoinc) { + table->autoinc = innobase_next_autoinc( + btr_read_autoinc_with_fallback(table, col_no), + 1 /* need */, + 1 /* auto_increment_increment */, + 0 /* auto_increment_offset */, + innobase_get_int_col_max_value(field)); } - dict_table_autoinc_initialize(prebuilt->table, auto_inc); + dict_table_autoinc_unlock(table); } -/*****************************************************************//** -Creates and opens a handle to a table which already exists in an InnoDB -database. -@return 1 if error, 0 if success */ -UNIV_INTERN +/** Open an InnoDB table +@param[in] name table name +@return error code +@retval 0 on success */ int -ha_innobase::open( -/*==============*/ - const char* name, /*!< in: table name */ - int mode, /*!< in: not used */ - uint test_if_locked) /*!< in: not used */ +ha_innobase::open(const char* name, int, uint) { dict_table_t* ib_table; char norm_name[FN_REFLEN]; - THD* thd; - char* is_part = NULL; - ibool par_case_name_set = FALSE; - char par_case_name[FN_REFLEN]; dict_err_ignore_t ignore_err = DICT_ERR_IGNORE_NONE; DBUG_ENTER("ha_innobase::open"); - UT_NOT_USED(mode); - UT_NOT_USED(test_if_locked); - - thd = ha_thd(); - - /* Under some cases MySQL seems to call this function while - holding btr_search_latch. This breaks the latching order as - we acquire dict_sys->mutex below and leads to a deadlock. */ - if (thd != NULL) { - innobase_release_temporary_latches(ht, thd); - } - normalize_table_name(norm_name, name); - user_thd = NULL; - - if (!(share=get_share(name))) { - - DBUG_RETURN(1); - } + m_user_thd = NULL; /* Will be allocated if it is needed in ::update_row() */ - upd_buf = NULL; - upd_buf_size = 0; + m_upd_buf = NULL; + m_upd_buf_size = 0; - /* We look for pattern #P# to see if the table is partitioned - MySQL table. */ -#ifdef __WIN__ - is_part = strstr(norm_name, "#p#"); -#else - is_part = strstr(norm_name, "#P#"); -#endif /* __WIN__ */ + char* is_part = is_partition(norm_name); + THD* thd = ha_thd(); /* Check whether FOREIGN_KEY_CHECKS is set to 0. If so, the table can be opened even if some FK indexes are missing. If not, the table @@ -5809,121 +6199,41 @@ ha_innobase::open( ignore_err = DICT_ERR_IGNORE_FK_NOKEY; } - /* Get pointer to a table object in InnoDB dictionary cache */ - ib_table = dict_table_open_on_name(norm_name, FALSE, TRUE, ignore_err); - - if (ib_table - && ((!DICT_TF2_FLAG_IS_SET(ib_table, DICT_TF2_FTS_HAS_DOC_ID) - && table->s->stored_fields != dict_table_get_n_user_cols(ib_table)) - || (DICT_TF2_FLAG_IS_SET(ib_table, DICT_TF2_FTS_HAS_DOC_ID) - && (table->s->fields - != dict_table_get_n_user_cols(ib_table) - 1)))) { - ib_logf(IB_LOG_LEVEL_WARN, - "table %s contains " ULINTPF " user defined columns " - "in InnoDB, but %u columns in MySQL. Please " - "check INFORMATION_SCHEMA.INNODB_SYS_COLUMNS and " - REFMAN "innodb-troubleshooting.html " - "for how to resolve it", - norm_name, dict_table_get_n_user_cols(ib_table), - table->s->fields); - - /* Mark this table as corrupted, so the drop table - or force recovery can still use it, but not others. */ - ib_table->file_unreadable = true; - ib_table->corrupted = true; - dict_table_close(ib_table, FALSE, FALSE); - ib_table = NULL; - is_part = NULL; - } + ib_table = open_dict_table(name, norm_name, is_part, ignore_err); if (NULL == ib_table) { - if (is_part) { - /* MySQL partition engine hard codes the file name - separator as "#P#". The text case is fixed even if - lower_case_table_names is set to 1 or 2. This is true - for sub-partition names as well. InnoDB always - normalises file names to lower case on Windows, this - can potentially cause problems when copying/moving - tables between platforms. - - 1) If boot against an installation from Windows - platform, then its partition table name could - be in lower case in system tables. So we will - need to check lower case name when load table. - - 2) If we boot an installation from other case - sensitive platform in Windows, we might need to - check the existence of table name without lower - case in the system table. */ - if (innobase_get_lower_case_table_names() == 1) { - - if (!par_case_name_set) { -#ifndef __WIN__ - /* Check for the table using lower - case name, including the partition - separator "P" */ - strcpy(par_case_name, norm_name); - innobase_casedn_str(par_case_name); -#else - /* On Windows platfrom, check - whether there exists table name in - system table whose name is - not being normalized to lower case */ - normalize_table_name_low( - par_case_name, name, FALSE); -#endif - par_case_name_set = TRUE; - } - - ib_table = dict_table_open_on_name( - par_case_name, FALSE, TRUE, - ignore_err); - } - - if (ib_table) { -#ifndef __WIN__ - sql_print_warning("Partition table %s opened " - "after converting to lower " - "case. The table may have " - "been moved from a case " - "in-sensitive file system. " - "Please recreate table in " - "the current file system\n", - norm_name); -#else - sql_print_warning("Partition table %s opened " - "after skipping the step to " - "lower case the table name. " - "The table may have been " - "moved from a case sensitive " - "file system. Please " - "recreate table in the " - "current file system\n", - norm_name); -#endif - goto table_opened; - } - } if (is_part) { sql_print_error("Failed to open table %s.\n", norm_name); } - - ib_logf(IB_LOG_LEVEL_WARN, - "Cannot open table %s from the internal data " - "dictionary of InnoDB though the .frm file " - "for the table exists. See " - REFMAN "innodb-troubleshooting.html for how " - "you can resolve the problem.", norm_name); - - free_share(share); - my_errno = ENOENT; +no_such_table: + set_my_errno(ENOENT); DBUG_RETURN(HA_ERR_NO_SUCH_TABLE); } -table_opened: + uint n_fields = mysql_fields(table); + uint n_cols = dict_table_get_n_user_cols(ib_table) + + dict_table_get_n_v_cols(ib_table) + - !!DICT_TF2_FLAG_IS_SET(ib_table, DICT_TF2_FTS_HAS_DOC_ID); + + if (n_cols != n_fields) { + ib::warn() << "Table " << norm_name << " contains " + << n_cols << " user" + " defined columns in InnoDB, but " << n_fields + << " columns in MariaDB. Please check" + " INFORMATION_SCHEMA.INNODB_SYS_COLUMNS and " REFMAN + "innodb-troubleshooting.html for how to resolve the" + " issue."; + + /* Mark this table as corrupted, so the drop table + or force recovery can still use it, but not others. */ + ib_table->file_unreadable = true; + ib_table->corrupted = true; + dict_table_close(ib_table, FALSE, FALSE); + goto no_such_table; + } innobase_copy_frm_flags_from_table_share(ib_table, table->s); @@ -5980,8 +6290,7 @@ table_opened: } if (!thd_tablespace_op(thd) && no_tablespace) { - free_share(share); - my_errno = ENOENT; + set_my_errno(ENOENT); int ret_err = HA_ERR_NO_SUCH_TABLE; /* If table has no talespace but it has crypt data, check @@ -5990,16 +6299,15 @@ table_opened: if (encrypted) { bool warning_pushed = false; - char buf[MAX_FULL_NAME_LEN]; - ut_format_name(ib_table->name, TRUE, buf, sizeof(buf)); - if (!encryption_key_id_exists(space()->crypt_data->key_id)) { - push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + push_warning_printf( + thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_DECRYPTION_FAILED, "Table %s in file %s is encrypted but encryption service or" " used key_id %u is not available. " " Can't continue reading table.", - buf, space()->chain.start->name, + table_share->table_name.str, + space()->chain.start->name, space()->crypt_data->key_id); ret_err = HA_ERR_DECRYPTION_FAILED; warning_pushed = true; @@ -6009,33 +6317,58 @@ table_opened: warning if it has not been already done as used key_id might be found but it is incorrect. */ if (!warning_pushed) { - push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + push_warning_printf( + thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_DECRYPTION_FAILED, "Table %s in file %s is encrypted but encryption service or" " used key_id is not available. " " Can't continue reading table.", - buf, space()->chain.start->name); + table_share->table_name.str, + space()->chain.start->name); ret_err = HA_ERR_DECRYPTION_FAILED; } } dict_table_close(ib_table, FALSE, FALSE); - DBUG_RETURN(ret_err); } - prebuilt = row_create_prebuilt(ib_table, table->s->stored_rec_length); + m_prebuilt = row_create_prebuilt(ib_table, table->s->reclength); + + m_prebuilt->default_rec = table->s->default_values; + ut_ad(m_prebuilt->default_rec); - prebuilt->default_rec = table->s->default_values; - ut_ad(prebuilt->default_rec); + m_prebuilt->m_mysql_table = table; /* Looks like MySQL-3.23 sometimes has primary key number != 0 */ - primary_key = table->s->primary_key; - key_used_on_scan = primary_key; + m_primary_key = table->s->primary_key; + + key_used_on_scan = m_primary_key; + + if (ib_table->n_v_cols) { + mutex_enter(&dict_sys->mutex); + if (ib_table->vc_templ == NULL) { + ib_table->vc_templ = UT_NEW_NOKEY(dict_vcol_templ_t()); + } else if (ib_table->get_ref_count() == 1) { + /* Clean and refresh the template if no one else + get hold on it */ + dict_free_vc_templ(ib_table->vc_templ); + ib_table->vc_templ->vtempl = NULL; + } + + if (ib_table->vc_templ->vtempl == NULL) { + innobase_build_v_templ( + table, ib_table, ib_table->vc_templ, NULL, + true); + } + + mutex_exit(&dict_sys->mutex); + } - if (!innobase_build_index_translation(table, ib_table, share)) { - sql_print_error("Build InnoDB index translation table for" - " Table %s failed", name); + if (!check_index_consistency(table, ib_table)) { + sql_print_error("InnoDB indexes are inconsistent with what " + "defined in .frm for table %s", + name); } /* Allocate a buffer for a 'row reference'. A row reference is @@ -6043,12 +6376,9 @@ table_opened: a row in our table. Note that MySQL may also compare two row references for equality by doing a simple memcmp on the strings of length ref_length! */ - - if (!row_table_got_default_clust_index(ib_table)) { - - prebuilt->clust_index_was_generated = FALSE; - - if (UNIV_UNLIKELY(primary_key >= MAX_KEY)) { + if (!(m_prebuilt->clust_index_was_generated + = dict_index_is_auto_gen_clust(ib_table->indexes.start))) { + if (m_primary_key >= MAX_KEY) { ib_table->dict_frm_mismatch = DICT_FRM_NO_PK; /* This mismatch could cause further problems @@ -6058,7 +6388,7 @@ table_opened: ib_push_frm_error(thd, ib_table, table, 0, true); - /* If primary_key >= MAX_KEY, its (primary_key) + /* If m_primary_key >= MAX_KEY, its (m_primary_key) value could be out of bound if continue to index into key_info[] array. Find InnoDB primary index, and assign its key_length to ref_length. @@ -6099,10 +6429,10 @@ table_opened: save space, because all row reference buffers are allocated based on ref_length. */ - ref_length = table->key_info[primary_key].key_length; + ref_length = table->key_info[m_primary_key].key_length; } } else { - if (primary_key != MAX_KEY) { + if (m_primary_key != MAX_KEY) { ib_table->dict_frm_mismatch = DICT_NO_PK_FRM_HAS; @@ -6113,8 +6443,6 @@ table_opened: ib_push_frm_error(thd, ib_table, table, 0, true); } - prebuilt->clust_index_was_generated = TRUE; - ref_length = DATA_ROW_ID_LEN; /* If we automatically created the clustered index, then @@ -6127,74 +6455,180 @@ table_opened: if (key_used_on_scan != MAX_KEY) { sql_print_warning( - "Table %s key_used_on_scan is %lu even " + "Table %s key_used_on_scan is %u even " "though there is no primary key inside " - "InnoDB.", name, (ulong) key_used_on_scan); + "InnoDB.", name, key_used_on_scan); } } /* Index block size in InnoDB: used by MySQL in query optimization */ stats.block_size = UNIV_PAGE_SIZE; - /* Init table lock structure */ - thr_lock_data_init(&share->lock,&lock,(void*) 0); - - if (prebuilt->table) { + if (m_prebuilt->table != NULL) { /* We update the highest file format in the system table space, if this table has higher file format setting. */ trx_sys_file_format_max_upgrade( (const char**) &innobase_file_format_max, - dict_table_get_format(prebuilt->table)); + dict_table_get_format(m_prebuilt->table)); } - /* Only if the table has an AUTOINC column. */ - if (prebuilt->table != NULL - && prebuilt->table->is_readable() - && table->found_next_number_field != NULL) { - dict_table_autoinc_lock(prebuilt->table); + if (m_prebuilt->table == NULL + || dict_table_is_temporary(m_prebuilt->table) + || m_prebuilt->table->persistent_autoinc + || !m_prebuilt->table->is_readable()) { + } else if (const Field* ai = table->found_next_number_field) { + initialize_auto_increment(m_prebuilt->table, ai); + } - /* Since a table can already be "open" in InnoDB's internal - data dictionary, we only init the autoinc counter once, the - first time the table is loaded. We can safely reuse the - autoinc value from a previous MySQL open. */ - if (dict_table_autoinc_read(prebuilt->table) == 0) { + /* Set plugin parser for fulltext index */ + for (uint i = 0; i < table->s->keys; i++) { + if (table->key_info[i].flags & HA_USES_PARSER) { + dict_index_t* index = innobase_get_index(i); + plugin_ref parser = table->key_info[i].parser; - innobase_initialize_autoinc(); - } + ut_ad(index->type & DICT_FTS); + index->parser = + static_cast<st_mysql_ftparser *>( + plugin_decl(parser)->info); - dict_table_autoinc_unlock(prebuilt->table); + DBUG_EXECUTE_IF("fts_instrument_use_default_parser", + index->parser = &fts_default_parser;); + } } info(HA_STATUS_NO_LOCK | HA_STATUS_VARIABLE | HA_STATUS_CONST); - DBUG_RETURN(0); } -UNIV_INTERN +/** Convert MySQL column number to dict_table_t::cols[] offset. +@param[in] field non-virtual column +@return column number relative to dict_table_t::cols[] */ +unsigned +innodb_col_no(const Field* field) +{ + ut_ad(!innobase_is_s_fld(field)); + const TABLE* table = field->table; + unsigned col_no = 0; + ut_ad(field == table->field[field->field_index]); + for (unsigned i = 0; i < field->field_index; i++) { + if (table->field[i]->stored_in_db()) { + col_no++; + } + } + return(col_no); +} + +/** Opens dictionary table object using table name. For partition, we need to +try alternative lower/upper case names to support moving data files across +platforms. +@param[in] table_name name of the table/partition +@param[in] norm_name normalized name of the table/partition +@param[in] is_partition if this is a partition of a table +@param[in] ignore_err error to ignore for loading dictionary object +@return dictionary table object or NULL if not found */ +dict_table_t* +ha_innobase::open_dict_table( + const char* table_name, + const char* norm_name, + bool is_partition, + dict_err_ignore_t ignore_err) +{ + DBUG_ENTER("ha_innobase::open_dict_table"); + dict_table_t* ib_table = dict_table_open_on_name(norm_name, FALSE, + TRUE, ignore_err); + + if (NULL == ib_table && is_partition) { + /* MySQL partition engine hard codes the file name + separator as "#P#". The text case is fixed even if + lower_case_table_names is set to 1 or 2. This is true + for sub-partition names as well. InnoDB always + normalises file names to lower case on Windows, this + can potentially cause problems when copying/moving + tables between platforms. + + 1) If boot against an installation from Windows + platform, then its partition table name could + be in lower case in system tables. So we will + need to check lower case name when load table. + + 2) If we boot an installation from other case + sensitive platform in Windows, we might need to + check the existence of table name without lower + case in the system table. */ + if (innobase_get_lower_case_table_names() == 1) { + char par_case_name[FN_REFLEN]; + +#ifndef _WIN32 + /* Check for the table using lower + case name, including the partition + separator "P" */ + strcpy(par_case_name, norm_name); + innobase_casedn_str(par_case_name); +#else + /* On Windows platfrom, check + whether there exists table name in + system table whose name is + not being normalized to lower case */ + create_table_info_t:: + normalize_table_name_low( + par_case_name, + table_name, FALSE); +#endif + ib_table = dict_table_open_on_name( + par_case_name, FALSE, TRUE, + ignore_err); + } + + if (ib_table != NULL) { +#ifndef _WIN32 + sql_print_warning("Partition table %s opened" + " after converting to lower" + " case. The table may have" + " been moved from a case" + " in-sensitive file system." + " Please recreate table in" + " the current file system\n", + norm_name); +#else + sql_print_warning("Partition table %s opened" + " after skipping the step to" + " lower case the table name." + " The table may have been" + " moved from a case sensitive" + " file system. Please" + " recreate table in the" + " current file system\n", + norm_name); +#endif + } + } + + DBUG_RETURN(ib_table); +} + handler* ha_innobase::clone( /*===============*/ const char* name, /*!< in: table name */ MEM_ROOT* mem_root) /*!< in: memory context */ { - ha_innobase* new_handler; - DBUG_ENTER("ha_innobase::clone"); - new_handler = static_cast<ha_innobase*>(handler::clone(name, - mem_root)); - if (new_handler) { - DBUG_ASSERT(new_handler->prebuilt != NULL); + ha_innobase* new_handler = static_cast<ha_innobase*>( + handler::clone(name, mem_root)); - new_handler->prebuilt->select_lock_type - = prebuilt->select_lock_type; + if (new_handler != NULL) { + DBUG_ASSERT(new_handler->m_prebuilt != NULL); + + new_handler->m_prebuilt->select_lock_type + = m_prebuilt->select_lock_type; } DBUG_RETURN(new_handler); } -UNIV_INTERN + uint ha_innobase::max_supported_key_part_length() const /*==============================================*/ @@ -6208,32 +6642,23 @@ ha_innobase::max_supported_key_part_length() const /******************************************************************//** Closes a handle to an InnoDB table. -@return 0 */ -UNIV_INTERN +@return 0 */ + int ha_innobase::close() /*================*/ { - THD* thd; - DBUG_ENTER("ha_innobase::close"); - thd = ha_thd(); - if (thd != NULL) { - innobase_release_temporary_latches(ht, thd); - } + row_prebuilt_free(m_prebuilt, FALSE); - row_prebuilt_free(prebuilt, FALSE); - - if (upd_buf != NULL) { - ut_ad(upd_buf_size != 0); - my_free(upd_buf); - upd_buf = NULL; - upd_buf_size = 0; + if (m_upd_buf != NULL) { + ut_ad(m_upd_buf_size != 0); + my_free(m_upd_buf); + m_upd_buf = NULL; + m_upd_buf_size = 0; } - free_share(share); - MONITOR_INC(MONITOR_TABLE_CLOSE); /* Tell InnoDB server that there might be work for @@ -6246,24 +6671,11 @@ ha_innobase::close() /* The following accessor functions should really be inside MySQL code! */ -/**************************************************************//** -Gets field offset for a field in a table. -@return offset */ -static inline -uint -get_field_offset( -/*=============*/ - const TABLE* table, /*!< in: MySQL table object */ - const Field* field) /*!< in: MySQL field object */ -{ - return((uint) (field->ptr - table->record[0])); -} - #ifdef WITH_WSREP UNIV_INTERN int wsrep_innobase_mysql_sort( -/*===============*/ +/*======================*/ /* out: str contains sort string */ int mysql_type, /* in: MySQL type */ uint charset_number, /* in: number of the charset */ @@ -6309,7 +6721,7 @@ wsrep_innobase_mysql_sort( if (charset == NULL) { sql_print_error("InnoDB needs charset %lu for doing " - "a comparison, but MySQL cannot " + "a comparison, but MariaDB cannot " "find that charset.", (ulong) charset_number); ut_a(0); @@ -6320,8 +6732,8 @@ wsrep_innobase_mysql_sort( memcpy(tmp_str, str, str_length); tmp_length = charset->coll->strnxfrm(charset, str, str_length, - str_length, tmp_str, - tmp_length, 0); + str_length, tmp_str, + tmp_length, 0); DBUG_ASSERT(tmp_length <= str_length); if (wsrep_protocol_version < 3) { tmp_length = charset->coll->strnxfrm( @@ -6339,7 +6751,7 @@ wsrep_innobase_mysql_sort( DBUG_ASSERT(tmp_length <= buf_length); ret_length = tmp_length; } - + break; } case MYSQL_TYPE_DECIMAL : @@ -6370,172 +6782,8 @@ wsrep_innobase_mysql_sort( } #endif /* WITH_WSREP */ -/*************************************************************//** -InnoDB uses this function to compare two data fields for which the data type -is such that we must use MySQL code to compare them. NOTE that the prototype -of this function is in rem0cmp.cc in InnoDB source code! If you change this -function, remember to update the prototype there! -@return 1, 0, -1, if a is greater, equal, less than b, respectively */ -UNIV_INTERN -int -innobase_mysql_cmp( -/*===============*/ - int mysql_type, /*!< in: MySQL type */ - uint charset_number, /*!< in: number of the charset */ - const unsigned char* a, /*!< in: data field */ - unsigned int a_length, /*!< in: data field length, - not UNIV_SQL_NULL */ - const unsigned char* b, /*!< in: data field */ - unsigned int b_length) /*!< in: data field length, - not UNIV_SQL_NULL */ -{ - CHARSET_INFO* charset; - enum_field_types mysql_tp; - int ret; - - DBUG_ASSERT(a_length != UNIV_SQL_NULL); - DBUG_ASSERT(b_length != UNIV_SQL_NULL); - - mysql_tp = (enum_field_types) mysql_type; - - switch (mysql_tp) { - - case MYSQL_TYPE_BIT: - case MYSQL_TYPE_STRING: - case MYSQL_TYPE_VAR_STRING: - case MYSQL_TYPE_TINY_BLOB: - case MYSQL_TYPE_MEDIUM_BLOB: - case MYSQL_TYPE_BLOB: - case MYSQL_TYPE_LONG_BLOB: - case MYSQL_TYPE_VARCHAR: - /* Use the charset number to pick the right charset struct for - the comparison. Since the MySQL function get_charset may be - slow before Bar removes the mutex operation there, we first - look at 2 common charsets directly. */ - - if (charset_number == default_charset_info->number) { - charset = default_charset_info; - } else if (charset_number == my_charset_latin1.number) { - charset = &my_charset_latin1; - } else { - charset = get_charset(charset_number, MYF(MY_WME)); - - if (charset == NULL) { - sql_print_error("InnoDB needs charset %lu for doing " - "a comparison, but MySQL cannot " - "find that charset.", - (ulong) charset_number); - ut_a(0); - } - } - - /* Starting from 4.1.3, we use strnncollsp() in comparisons of - non-latin1_swedish_ci strings. NOTE that the collation order - changes then: 'b\0\0...' is ordered BEFORE 'b ...'. Users - having indexes on such data need to rebuild their tables! */ - - ret = charset->coll->strnncollsp( - charset, a, a_length, b, b_length, 0); - - if (ret < 0) { - return(-1); - } else if (ret > 0) { - return(1); - } else { - return(0); - } - default: - ut_error; - } - - return(0); -} - - -/*************************************************************//** -Get the next token from the given string and store it in *token. */ -UNIV_INTERN -CHARSET_INFO* -innobase_get_fts_charset( -/*=====================*/ - int mysql_type, /*!< in: MySQL type */ - uint charset_number) /*!< in: number of the charset */ -{ - enum_field_types mysql_tp; - CHARSET_INFO* charset; - - mysql_tp = (enum_field_types) mysql_type; - - switch (mysql_tp) { - - case MYSQL_TYPE_BIT: - case MYSQL_TYPE_STRING: - case MYSQL_TYPE_VAR_STRING: - case MYSQL_TYPE_TINY_BLOB: - case MYSQL_TYPE_MEDIUM_BLOB: - case MYSQL_TYPE_BLOB: - case MYSQL_TYPE_LONG_BLOB: - case MYSQL_TYPE_VARCHAR: - /* Use the charset number to pick the right charset struct for - the comparison. Since the MySQL function get_charset may be - slow before Bar removes the mutex operation there, we first - look at 2 common charsets directly. */ - - if (charset_number == default_charset_info->number) { - charset = default_charset_info; - } else if (charset_number == my_charset_latin1.number) { - charset = &my_charset_latin1; - } else { - charset = get_charset(charset_number, MYF(MY_WME)); - - if (charset == NULL) { - sql_print_error("InnoDB needs charset %lu for doing " - "a comparison, but MySQL cannot " - "find that charset.", - (ulong) charset_number); - ut_a(0); - } - } - break; - default: - ut_error; - } - - return(charset); -} - -/*************************************************************//** -InnoDB uses this function to compare two data fields for which the data type -is such that we must use MySQL code to compare them. NOTE that the prototype -of this function is in rem0cmp.c in InnoDB source code! If you change this -function, remember to update the prototype there! -@return 1, 0, -1, if a is greater, equal, less than b, respectively */ -UNIV_INTERN -int -innobase_mysql_cmp_prefix( -/*======================*/ - int mysql_type, /*!< in: MySQL type */ - uint charset_number, /*!< in: number of the charset */ - const unsigned char* a, /*!< in: data field */ - unsigned int a_length, /*!< in: data field length, - not UNIV_SQL_NULL */ - const unsigned char* b, /*!< in: data field */ - unsigned int b_length) /*!< in: data field length, - not UNIV_SQL_NULL */ -{ - CHARSET_INFO* charset; - int result; - - charset = innobase_get_fts_charset(mysql_type, charset_number); - - result = ha_compare_text(charset, (uchar*) a, a_length, - (uchar*) b, b_length, 1, 0); - - return(result); -} /******************************************************************//** compare two character string according to their charset. */ -UNIV_INTERN int innobase_fts_text_cmp( /*==================*/ @@ -6549,11 +6797,11 @@ innobase_fts_text_cmp( return(ha_compare_text( charset, s1->f_str, static_cast<uint>(s1->f_len), - s2->f_str, static_cast<uint>(s2->f_len), 0, 0)); + s2->f_str, static_cast<uint>(s2->f_len), 0)); } + /******************************************************************//** compare two character string case insensitively according to their charset. */ -UNIV_INTERN int innobase_fts_text_case_cmp( /*=======================*/ @@ -6572,11 +6820,11 @@ innobase_fts_text_case_cmp( return(ha_compare_text( charset, s1->f_str, static_cast<uint>(s1->f_len), - s2->f_str, static_cast<uint>(newlen), 0, 0)); + s2->f_str, static_cast<uint>(newlen), 0)); } + /******************************************************************//** Get the first character's code position for FTS index partition. */ -UNIV_INTERN ulint innobase_strnxfrm( /*==============*/ @@ -6605,7 +6853,6 @@ innobase_strnxfrm( /******************************************************************//** compare two character string according to their charset. */ -UNIV_INTERN int innobase_fts_text_cmp_prefix( /*=========================*/ @@ -6620,7 +6867,7 @@ innobase_fts_text_cmp_prefix( result = ha_compare_text( charset, s2->f_str, static_cast<uint>(s2->f_len), - s1->f_str, static_cast<uint>(s1->f_len), 1, 0); + s1->f_str, static_cast<uint>(s1->f_len), 1); /* We switched s1, s2 position in ha_compare_text. So we need to negate the result */ @@ -6629,7 +6876,6 @@ innobase_fts_text_cmp_prefix( /******************************************************************//** Makes all characters in a string lower case. */ -UNIV_INTERN size_t innobase_fts_casedn_str( /*====================*/ @@ -6658,7 +6904,6 @@ innobase_fts_casedn_str( Get the next token from the given string and store it in *token. It is mostly copied from MyISAM's doc parsing function ft_simple_get_word() @return length of string processed */ -UNIV_INTERN ulint innobase_mysql_fts_get_token( /*=========================*/ @@ -6666,10 +6911,7 @@ innobase_mysql_fts_get_token( const byte* start, /*!< in: start of text */ const byte* end, /*!< in: one character past end of text */ - fts_string_t* token, /*!< out: token's text */ - ulint* offset) /*!< out: offset to token, - measured as characters from - 'start' */ + fts_string_t* token) /*!< out: token's text */ { int mbl; const uchar* doc = start; @@ -6727,21 +6969,17 @@ innobase_mysql_fts_get_token( return(doc - start); } -/**************************************************************//** -Converts a MySQL type to an InnoDB type. Note that this function returns +/** Converts a MySQL type to an InnoDB type. Note that this function returns the 'mtype' of InnoDB. InnoDB differentiates between MySQL's old <= 4.1 VARCHAR and the new true VARCHAR in >= 5.0.3 by the 'prtype'. -@return DATA_BINARY, DATA_VARCHAR, ... */ -UNIV_INTERN +@param[out] unsigned_flag DATA_UNSIGNED if an 'unsigned type'; at least +ENUM and SET, and unsigned integer types are 'unsigned types' +@param[in] f MySQL Field +@return DATA_BINARY, DATA_VARCHAR, ... */ ulint get_innobase_type_from_mysql_type( -/*==============================*/ - ulint* unsigned_flag, /*!< out: DATA_UNSIGNED if an - 'unsigned type'; - at least ENUM and SET, - and unsigned integer - types are 'unsigned types' */ - const void* f) /*!< in: MySQL Field */ + ulint* unsigned_flag, + const void* f) { const class Field* field = reinterpret_cast<const class Field*>(f); @@ -6808,13 +7046,14 @@ get_innobase_type_from_mysql_type( case MYSQL_TYPE_YEAR: case MYSQL_TYPE_NEWDATE: return(DATA_INT); - case MYSQL_TYPE_TIMESTAMP: case MYSQL_TYPE_TIME: case MYSQL_TYPE_DATETIME: - if (field->key_type() == HA_KEYTYPE_BINARY) + case MYSQL_TYPE_TIMESTAMP: + if (field->key_type() == HA_KEYTYPE_BINARY) { return(DATA_FIXBINARY); - else + } else { return(DATA_INT); + } case MYSQL_TYPE_FLOAT: return(DATA_FLOAT); case MYSQL_TYPE_DOUBLE: @@ -6822,6 +7061,7 @@ get_innobase_type_from_mysql_type( case MYSQL_TYPE_DECIMAL: return(DATA_DECIMAL); case MYSQL_TYPE_GEOMETRY: + return(DATA_GEOMETRY); case MYSQL_TYPE_TINY_BLOB: case MYSQL_TYPE_MEDIUM_BLOB: case MYSQL_TYPE_BLOB: @@ -6840,25 +7080,9 @@ get_innobase_type_from_mysql_type( } /*******************************************************************//** -Writes an unsigned integer value < 64k to 2 bytes, in the little-endian -storage format. */ -static inline -void -innobase_write_to_2_little_endian( -/*==============================*/ - byte* buf, /*!< in: where to store */ - ulint val) /*!< in: value to write, must be < 64k */ -{ - ut_a(val < 256 * 256); - - buf[0] = (byte)(val & 0xFF); - buf[1] = (byte)(val / 256); -} - -/*******************************************************************//** Reads an unsigned integer value < 64k from 2 bytes, in the little-endian storage format. -@return value */ +@return value */ static inline uint innobase_read_from_2_little_endian( @@ -6875,7 +7099,7 @@ Stores a key value for a row to a buffer. UNIV_INTERN uint wsrep_store_key_val_for_row( -/*===============================*/ +/*=========================*/ THD* thd, TABLE* table, uint keynr, /*!< in: key number */ @@ -6905,7 +7129,7 @@ wsrep_store_key_val_for_row( if (key_part->null_bit) { if (buff_space > 0) { - if (record[key_part->null_offset] + if (record[key_part->null_offset] & key_part->null_bit) { *buff = 1; part_is_null = TRUE; @@ -6964,11 +7188,11 @@ wsrep_store_key_val_for_row( the true length of the key */ if (len > 0 && cs->mbmaxlen > 1) { - true_len = (ulint) cs->cset->well_formed_len(cs, + true_len = (ulint) my_well_formed_length(cs, (const char *) data, (const char *) data + len, - (uint) (key_len / - cs->mbmaxlen), + (uint) (key_len / + cs->mbmaxlen), &error); } @@ -6981,15 +7205,14 @@ wsrep_store_key_val_for_row( memcpy(sorted, data, true_len); true_len = wsrep_innobase_mysql_sort( - mysql_type, cs->number, sorted, true_len, + mysql_type, cs->number, sorted, true_len, REC_VERSION_56_MAX_INDEX_COL_LEN); - if (wsrep_protocol_version > 1) { - /* Note that we always reserve the maximum possible - length of the true VARCHAR in the key value, though - only len first bytes after the 2 length bytes contain - actual data. The rest of the space was reset to zero - in the bzero() call above. */ + /* Note that we always reserve the maximum possible + length of the true VARCHAR in the key value, though + only len first bytes after the 2 length bytes contain + actual data. The rest of the space was reset to zero + in the bzero() call above. */ if (true_len > buff_space) { fprintf (stderr, "WSREP: key truncated: %s\n", @@ -6997,11 +7220,11 @@ wsrep_store_key_val_for_row( true_len = buff_space; } memcpy(buff, sorted, true_len); - buff += true_len; + buff += true_len; buff_space -= true_len; - } else { - buff += key_len; - } + } else { + buff += key_len; + } } else if (mysql_type == MYSQL_TYPE_TINY_BLOB || mysql_type == MYSQL_TYPE_MEDIUM_BLOB || mysql_type == MYSQL_TYPE_BLOB @@ -7051,12 +7274,12 @@ wsrep_store_key_val_for_row( the true length of the key */ if (blob_len > 0 && cs->mbmaxlen > 1) { - true_len = (ulint) cs->cset->well_formed_len(cs, + true_len = (ulint) my_well_formed_length(cs, (const char *) blob_data, (const char *) blob_data + blob_len, - (uint) (key_len / - cs->mbmaxlen), + (uint) (key_len / + cs->mbmaxlen), &error); } @@ -7076,7 +7299,7 @@ wsrep_store_key_val_for_row( /* Note that we always reserve the maximum possible length of the BLOB prefix in the key value. */ - if (wsrep_protocol_version > 1) { + if (wsrep_protocol_version > 1) { if (true_len > buff_space) { fprintf (stderr, "WSREP: key truncated: %s\n", @@ -7140,12 +7363,12 @@ wsrep_store_key_val_for_row( if (key_len > 0 && cs->mbmaxlen > 1) { true_len = (ulint) - cs->cset->well_formed_len(cs, + my_well_formed_length(cs, (const char *)src_start, (const char *)src_start + key_len, - (uint) (key_len / - cs->mbmaxlen), + (uint) (key_len / + cs->mbmaxlen), &error); } memcpy(sorted, src_start, true_len); @@ -7173,285 +7396,8 @@ wsrep_store_key_val_for_row( DBUG_RETURN((uint)(buff - buff_start)); } #endif /* WITH_WSREP */ - -/*******************************************************************//** -Stores a key value for a row to a buffer. -@return key value length as stored in buff */ -UNIV_INTERN -uint -ha_innobase::store_key_val_for_row( -/*===============================*/ - uint keynr, /*!< in: key number */ - char* buff, /*!< in/out: buffer for the key value (in MySQL - format) */ - uint buff_len,/*!< in: buffer length */ - const uchar* record)/*!< in: row in MySQL format */ -{ - KEY* key_info = table->key_info + keynr; - KEY_PART_INFO* key_part = key_info->key_part; - KEY_PART_INFO* end = - key_part + key_info->user_defined_key_parts; - char* buff_start = buff; - enum_field_types mysql_type; - Field* field; - ibool is_null; - - DBUG_ENTER("store_key_val_for_row"); - - /* The format for storing a key field in MySQL is the following: - - 1. If the column can be NULL, then in the first byte we put 1 if the - field value is NULL, 0 otherwise. - - 2. If the column is of a BLOB type (it must be a column prefix field - in this case), then we put the length of the data in the field to the - next 2 bytes, in the little-endian format. If the field is SQL NULL, - then these 2 bytes are set to 0. Note that the length of data in the - field is <= column prefix length. - - 3. In a column prefix field, prefix_len next bytes are reserved for - data. In a normal field the max field length next bytes are reserved - for data. For a VARCHAR(n) the max field length is n. If the stored - value is the SQL NULL then these data bytes are set to 0. - - 4. We always use a 2 byte length for a true >= 5.0.3 VARCHAR. Note that - in the MySQL row format, the length is stored in 1 or 2 bytes, - depending on the maximum allowed length. But in the MySQL key value - format, the length always takes 2 bytes. - - We have to zero-fill the buffer so that MySQL is able to use a - simple memcmp to compare two key values to determine if they are - equal. MySQL does this to compare contents of two 'ref' values. */ - - memset(buff, 0, buff_len); - - for (; key_part != end; key_part++) { - is_null = FALSE; - - if (key_part->null_bit) { - if (record[key_part->null_offset] - & key_part->null_bit) { - *buff = 1; - is_null = TRUE; - } else { - *buff = 0; - } - buff++; - } - - field = key_part->field; - mysql_type = field->type(); - - if (mysql_type == MYSQL_TYPE_VARCHAR) { - /* >= 5.0.3 true VARCHAR */ - ulint lenlen; - ulint len; - const byte* data; - ulint key_len; - ulint true_len; - const CHARSET_INFO* cs; - int error=0; - - key_len = key_part->length; - - if (is_null) { - buff += key_len + 2; - - continue; - } - cs = field->charset(); - - lenlen = (ulint) - (((Field_varstring*) field)->length_bytes); - - data = row_mysql_read_true_varchar(&len, - (byte*) (record - + (ulint) get_field_offset(table, field)), - lenlen); - - true_len = len; - - /* For multi byte character sets we need to calculate - the true length of the key */ - - if (len > 0 && cs->mbmaxlen > 1) { - true_len = (ulint) cs->cset->well_formed_len(cs, - (const char*) data, - (const char*) data + len, - (uint) (key_len / cs->mbmaxlen), - &error); - } - - /* In a column prefix index, we may need to truncate - the stored value: */ - - if (true_len > key_len) { - true_len = key_len; - } - - /* The length in a key value is always stored in 2 - bytes */ - - row_mysql_store_true_var_len((byte*) buff, true_len, 2); - buff += 2; - - memcpy(buff, data, true_len); - - /* Note that we always reserve the maximum possible - length of the true VARCHAR in the key value, though - only len first bytes after the 2 length bytes contain - actual data. The rest of the space was reset to zero - in the memset() call above. */ - - buff += key_len; - - } else if (mysql_type == MYSQL_TYPE_TINY_BLOB - || mysql_type == MYSQL_TYPE_MEDIUM_BLOB - || mysql_type == MYSQL_TYPE_BLOB - || mysql_type == MYSQL_TYPE_LONG_BLOB - /* MYSQL_TYPE_GEOMETRY data is treated - as BLOB data in innodb. */ - || mysql_type == MYSQL_TYPE_GEOMETRY) { - - const CHARSET_INFO* cs; - ulint key_len; - ulint true_len; - int error=0; - ulint blob_len; - const byte* blob_data; - - ut_a(key_part->key_part_flag & HA_PART_KEY_SEG); - - key_len = key_part->length; - - if (is_null) { - buff += key_len + 2; - - continue; - } - - cs = field->charset(); - - blob_data = row_mysql_read_blob_ref(&blob_len, - (byte*) (record - + (ulint) get_field_offset(table, field)), - (ulint) field->pack_length()); - - true_len = blob_len; - - ut_a(get_field_offset(table, field) - == key_part->offset); - - /* For multi byte character sets we need to calculate - the true length of the key */ - - if (blob_len > 0 && cs->mbmaxlen > 1) { - true_len = (ulint) cs->cset->well_formed_len(cs, - (const char*) blob_data, - (const char*) blob_data - + blob_len, - (uint) (key_len / cs->mbmaxlen), - &error); - } - - /* All indexes on BLOB and TEXT are column prefix - indexes, and we may need to truncate the data to be - stored in the key value: */ - - if (true_len > key_len) { - true_len = key_len; - } - - /* MySQL reserves 2 bytes for the length and the - storage of the number is little-endian */ - - innobase_write_to_2_little_endian( - (byte*) buff, true_len); - buff += 2; - - memcpy(buff, blob_data, true_len); - - /* Note that we always reserve the maximum possible - length of the BLOB prefix in the key value. */ - - buff += key_len; - } else { - /* Here we handle all other data types except the - true VARCHAR, BLOB and TEXT. Note that the column - value we store may be also in a column prefix - index. */ - - const CHARSET_INFO* cs = NULL; - ulint true_len; - ulint key_len; - const uchar* src_start; - int error=0; - enum_field_types real_type; - - key_len = key_part->length; - - if (is_null) { - buff += key_len; - - continue; - } - - src_start = record + key_part->offset; - real_type = field->real_type(); - true_len = key_len; - - /* Character set for the field is defined only - to fields whose type is string and real field - type is not enum or set. For these fields check - if character set is multi byte. */ - - if (real_type != MYSQL_TYPE_ENUM - && real_type != MYSQL_TYPE_SET - && ( mysql_type == MYSQL_TYPE_VAR_STRING - || mysql_type == MYSQL_TYPE_STRING)) { - - cs = field->charset(); - - /* For multi byte character sets we need to - calculate the true length of the key */ - - if (key_len > 0 && cs->mbmaxlen > 1) { - - true_len = (ulint) - cs->cset->well_formed_len(cs, - (const char*) src_start, - (const char*) src_start - + key_len, - (uint) (key_len - / cs->mbmaxlen), - &error); - } - } - - memcpy(buff, src_start, true_len); - buff += true_len; - - /* Pad the unused space with spaces. */ - - if (true_len < key_len) { - ulint pad_len = key_len - true_len; - ut_a(cs != NULL); - ut_a(!(pad_len % cs->mbminlen)); - - cs->cset->fill(cs, buff, pad_len, - 0x20 /* space */); - buff += pad_len; - } - } - } - - ut_a(buff <= buff_start + buff_len); - - DBUG_RETURN((uint)(buff - buff_start)); -} - /**************************************************************//** -Determines if a field is needed in a prebuilt struct 'template'. +Determines if a field is needed in a m_prebuilt struct 'template'. @return field to use, or NULL if the field is not needed */ static const Field* @@ -7475,11 +7421,9 @@ build_template_needs_field( dict_index_t* index, /*!< in: InnoDB index to use */ const TABLE* table, /*!< in: MySQL table object */ ulint i, /*!< in: field index in InnoDB table */ - ulint sql_idx) /*!< in: field index in SQL table */ + ulint num_v) /*!< in: num virtual column so far */ { - const Field* field = table->field[sql_idx]; - - ut_ad(index_contains == dict_index_contains_col_or_prefix(index, i)); + const Field* field = table->field[i]; if (!index_contains) { if (read_just_key) { @@ -7494,15 +7438,16 @@ build_template_needs_field( return(field); } - if (bitmap_is_set(table->read_set, static_cast<uint>(sql_idx)) - || bitmap_is_set(table->write_set, static_cast<uint>(sql_idx))) { + if (bitmap_is_set(table->read_set, static_cast<uint>(i)) + || bitmap_is_set(table->write_set, static_cast<uint>(i))) { /* This field is needed in the query */ return(field); } + ut_ad(i >= num_v); if (fetch_primary_key_cols - && dict_table_col_in_clustered_key(index->table, i)) { + && dict_table_col_in_clustered_key(index->table, i - num_v)) { /* This field is needed in the query */ return(field); @@ -7514,7 +7459,7 @@ build_template_needs_field( } /**************************************************************//** -Determines if a field is needed in a prebuilt struct 'template'. +Determines if a field is needed in a m_prebuilt struct 'template'. @return whether the field is needed for index condition pushdown */ inline bool @@ -7524,17 +7469,19 @@ build_template_needs_field_in_icp( const row_prebuilt_t* prebuilt,/*!< in: row fetch template */ bool contains,/*!< in: whether the index contains column i */ - ulint i) /*!< in: column number */ + ulint i, /*!< in: column number */ + bool is_virtual) + /*!< in: a virtual column or not */ { - ut_ad(contains == dict_index_contains_col_or_prefix(index, i)); + ut_ad(contains == dict_index_contains_col_or_prefix(index, i, is_virtual)); return(index == prebuilt->index ? contains - : dict_index_contains_col_or_prefix(prebuilt->index, i)); + : dict_index_contains_col_or_prefix(prebuilt->index, i, is_virtual)); } /**************************************************************//** -Adds a field to a prebuilt struct 'template'. +Adds a field to a m_prebuilt struct 'template'. @return the field template */ static mysql_row_templ_t* @@ -7545,94 +7492,113 @@ build_template_field( dict_index_t* index, /*!< in: InnoDB index to use */ TABLE* table, /*!< in: MySQL table object */ const Field* field, /*!< in: field in MySQL table */ - ulint i) /*!< in: field index in InnoDB table */ + ulint i, /*!< in: field index in InnoDB table */ + ulint v_no) /*!< in: field index for virtual col */ { mysql_row_templ_t* templ; const dict_col_t* col; - //ut_ad(field == table->field[i]); ut_ad(clust_index->table == index->table); - col = dict_table_get_nth_col(index->table, i); - templ = prebuilt->mysql_template + prebuilt->n_template++; UNIV_MEM_INVALID(templ, sizeof *templ); - templ->col_no = i; - templ->clust_rec_field_no = dict_col_get_clust_pos(col, clust_index); - - /* If clustered index record field is not found, lets print out - field names and all the rest to understand why field is not found. */ - if (templ->clust_rec_field_no == ULINT_UNDEFINED) { - const char* tb_col_name = dict_table_get_col_name(clust_index->table, i); - dict_field_t* field=NULL; - size_t size = 0; - - for(ulint j=0; j < clust_index->n_user_defined_cols; j++) { - dict_field_t* ifield = &(clust_index->fields[j]); - if (ifield && !memcmp(tb_col_name, ifield->name, - strlen(tb_col_name))) { - field = ifield; - break; + + if (innobase_is_v_fld(field)) { + templ->is_virtual = true; + col = &dict_table_get_nth_v_col(index->table, v_no)->m_col; + } else { + templ->is_virtual = false; + col = dict_table_get_nth_col(index->table, i); + } + + if (!templ->is_virtual) { + templ->col_no = i; + templ->clust_rec_field_no = dict_col_get_clust_pos( + col, clust_index); + /* If clustered index record field is not found, lets print out + field names and all the rest to understand why field is not found. */ + if (templ->clust_rec_field_no == ULINT_UNDEFINED) { + const char* tb_col_name = dict_table_get_col_name(clust_index->table, i); + dict_field_t* field=NULL; + size_t size = 0; + + for(ulint j=0; j < clust_index->n_user_defined_cols; j++) { + dict_field_t* ifield = &(clust_index->fields[j]); + if (ifield && !memcmp(tb_col_name, ifield->name, + strlen(tb_col_name))) { + field = ifield; + break; + } } - } - ib_logf(IB_LOG_LEVEL_INFO, - "Looking for field %lu name %s from table %s", - i, - (tb_col_name ? tb_col_name : "NULL"), - clust_index->table->name); - - - for(ulint j=0; j < clust_index->n_user_defined_cols; j++) { - dict_field_t* ifield = &(clust_index->fields[j]); - ib_logf(IB_LOG_LEVEL_INFO, - "InnoDB Table %s field %lu name %s", - clust_index->table->name, - j, - (ifield ? ifield->name : "NULL")); - } - - for(ulint j=0; j < table->s->stored_fields; j++) { - ib_logf(IB_LOG_LEVEL_INFO, - "MySQL table %s field %lu name %s", - table->s->table_name.str, - j, - table->field[j]->field_name); - } - - ib_logf(IB_LOG_LEVEL_ERROR, - "Clustered record field for column %lu" - " not found table n_user_defined %d" - " index n_user_defined %d" - " InnoDB table %s field name %s" - " MySQL table %s field name %s n_fields %d" - " query %s", - i, - clust_index->n_user_defined_cols, - clust_index->table->n_cols - DATA_N_SYS_COLS, - clust_index->table->name, - (field ? field->name : "NULL"), - table->s->table_name.str, - (tb_col_name ? tb_col_name : "NULL"), - table->s->stored_fields, - innobase_get_stmt(current_thd, &size)); + ib::info() << "Looking for field " << i << " name " + << (tb_col_name ? tb_col_name : "NULL") + << " from table " << clust_index->table->name; - ut_a(templ->clust_rec_field_no != ULINT_UNDEFINED); - } - templ->rec_field_is_prefix = FALSE; - if (dict_index_is_clust(index)) { - templ->rec_field_no = templ->clust_rec_field_no; + for(ulint j=0; j < clust_index->n_user_defined_cols; j++) { + dict_field_t* ifield = &(clust_index->fields[j]); + ib::info() << "InnoDB Table " + << clust_index->table->name + << "field " << j << " name " + << (ifield ? ifield->name() : "NULL"); + } + + for(ulint j=0; j < table->s->stored_fields; j++) { + ib::info() << "MySQL table " + << table->s->table_name.str + << " field " << j << " name " + << table->field[j]->field_name; + } + + ib::error() << "Clustered record field for column " << i + << " not found table n_user_defined " + << clust_index->n_user_defined_cols + << " index n_user_defined " + << clust_index->table->n_cols - DATA_N_SYS_COLS + << " InnoDB table " + << clust_index->table->name + << " field name " + << (field ? field->name() : "NULL") + << " MySQL table " + << table->s->table_name.str + << " field name " + << (tb_col_name ? tb_col_name : "NULL") + << " n_fields " + << table->s->stored_fields + << " query " + << innobase_get_stmt_unsafe(current_thd, &size); + + ut_a(templ->clust_rec_field_no != ULINT_UNDEFINED); + } + templ->rec_field_is_prefix = FALSE; templ->rec_prefix_field_no = ULINT_UNDEFINED; - } else { - /* If we're in a secondary index, keep track - * of the original index position even if this - * is just a prefix index; we will use this - * later to avoid a cluster index lookup in - * some cases.*/ - templ->rec_field_no = dict_index_get_nth_col_pos(index, i, + if (dict_index_is_clust(index)) { + templ->rec_field_no = templ->clust_rec_field_no; + } else { + /* If we're in a secondary index, keep track + * of the original index position even if this + * is just a prefix index; we will use this + * later to avoid a cluster index lookup in + * some cases.*/ + + templ->rec_field_no = dict_index_get_nth_col_pos(index, i, &templ->rec_prefix_field_no); + } + } else { + templ->clust_rec_field_no = v_no; + templ->rec_prefix_field_no = ULINT_UNDEFINED; + + if (dict_index_is_clust(index)) { + templ->rec_field_no = templ->clust_rec_field_no; + } else { + templ->rec_field_no + = dict_index_get_nth_col_or_prefix_pos( + index, v_no, FALSE, true, + &templ->rec_prefix_field_no); + } + templ->icp_rec_field_no = ULINT_UNDEFINED; } if (field->real_maybe_null()) { @@ -7644,8 +7610,8 @@ build_template_field( templ->mysql_null_bit_mask = 0; } - templ->mysql_col_offset = (ulint) get_field_offset(table, field); + templ->mysql_col_offset = (ulint) get_field_offset(table, field); templ->mysql_col_len = (ulint) field->pack_length(); templ->type = col->mtype; templ->mysql_type = (ulint) field->type(); @@ -7653,6 +7619,8 @@ build_template_field( if (templ->mysql_type == DATA_MYSQL_TRUE_VARCHAR) { templ->mysql_length_bytes = (ulint) (((Field_varstring*) field)->length_bytes); + } else { + templ->mysql_length_bytes = 0; } templ->charset = dtype_get_charset_coll(col->prtype); @@ -7672,13 +7640,18 @@ build_template_field( } } + /* For spatial index, we need to access cluster index. */ + if (dict_index_is_spatial(index)) { + prebuilt->need_to_access_clustered = TRUE; + } + if (prebuilt->mysql_prefix_len < templ->mysql_col_offset + templ->mysql_col_len) { prebuilt->mysql_prefix_len = templ->mysql_col_offset + templ->mysql_col_len; } - if (templ->type == DATA_BLOB) { + if (DATA_LARGE_MTYPE(templ->type)) { prebuilt->templ_contains_blob = TRUE; } @@ -7686,9 +7659,9 @@ build_template_field( } /**************************************************************//** -Builds a 'template' to the prebuilt struct. The template is used in fast +Builds a 'template' to the m_prebuilt struct. The template is used in fast retrieval of just those column values MySQL needs in its processing. */ -UNIV_INTERN + void ha_innobase::build_template( /*========================*/ @@ -7697,25 +7670,25 @@ ha_innobase::build_template( { dict_index_t* index; dict_index_t* clust_index; - ulint n_stored_fields; + ulint n_fields; ibool fetch_all_in_key = FALSE; ibool fetch_primary_key_cols = FALSE; - ulint i, sql_idx; + ulint i; - if (prebuilt->select_lock_type == LOCK_X) { + if (m_prebuilt->select_lock_type == LOCK_X) { /* We always retrieve the whole clustered index record if we use exclusive row level locks, for example, if the read is done in an UPDATE statement. */ whole_row = true; } else if (!whole_row) { - if (prebuilt->hint_need_to_fetch_extra_cols + if (m_prebuilt->hint_need_to_fetch_extra_cols == ROW_RETRIEVE_ALL_COLS) { /* We know we must at least fetch all columns in the key, or all columns in the table */ - if (prebuilt->read_just_key) { + if (m_prebuilt->read_just_key) { /* MySQL has instructed us that it is enough to fetch the columns in the key; looks like MySQL can set this flag also when there is @@ -7727,68 +7700,79 @@ ha_innobase::build_template( } else { whole_row = true; } - } else if (prebuilt->hint_need_to_fetch_extra_cols + } else if (m_prebuilt->hint_need_to_fetch_extra_cols == ROW_RETRIEVE_PRIMARY_KEY) { /* We must at least fetch all primary key cols. Note that if the clustered index was internally generated by InnoDB on the row id (no primary key was defined), then row_search_for_mysql() will always retrieve the row id to a special buffer in the - prebuilt struct. */ + m_prebuilt struct. */ fetch_primary_key_cols = TRUE; } } - clust_index = dict_table_get_first_index(prebuilt->table); + clust_index = dict_table_get_first_index(m_prebuilt->table); - index = whole_row ? clust_index : prebuilt->index; + index = whole_row ? clust_index : m_prebuilt->index; - prebuilt->need_to_access_clustered = (index == clust_index); + m_prebuilt->need_to_access_clustered = (index == clust_index); - /* Either prebuilt->index should be a secondary index, or it + /* Either m_prebuilt->index should be a secondary index, or it should be the clustered index. */ ut_ad(dict_index_is_clust(index) == (index == clust_index)); /* Below we check column by column if we need to access the clustered index. */ - n_stored_fields= (ulint)table->s->stored_fields; /* number of stored columns */ + n_fields = (ulint) mysql_fields(table); - if (!prebuilt->mysql_template) { - prebuilt->mysql_template = (mysql_row_templ_t*) - mem_alloc(n_stored_fields * sizeof(mysql_row_templ_t)); + if (!m_prebuilt->mysql_template) { + m_prebuilt->mysql_template = (mysql_row_templ_t*) + ut_malloc_nokey(n_fields * sizeof(mysql_row_templ_t)); } - prebuilt->template_type = whole_row + m_prebuilt->template_type = whole_row ? ROW_MYSQL_WHOLE_ROW : ROW_MYSQL_REC_FIELDS; - prebuilt->null_bitmap_len = table->s->null_bytes; + m_prebuilt->null_bitmap_len = table->s->null_bytes; - /* Prepare to build prebuilt->mysql_template[]. */ - prebuilt->templ_contains_blob = FALSE; - prebuilt->mysql_prefix_len = 0; - prebuilt->n_template = 0; - prebuilt->idx_cond_n_cols = 0; + /* Prepare to build m_prebuilt->mysql_template[]. */ + m_prebuilt->templ_contains_blob = FALSE; + m_prebuilt->mysql_prefix_len = 0; + m_prebuilt->n_template = 0; + m_prebuilt->idx_cond_n_cols = 0; /* Note that in InnoDB, i is the column number in the table. MySQL calls columns 'fields'. */ - if (active_index != MAX_KEY && active_index == pushed_idx_cond_keyno) { - /* Push down an index condition or an end_range check. */ - for (i = 0, sql_idx = 0; i < n_stored_fields; i++, sql_idx++) { + if (active_index != MAX_KEY + && active_index == pushed_idx_cond_keyno) { + ulint num_v = 0; - while (!table->field[sql_idx]->stored_in_db) { - sql_idx++; - } - - const ibool index_contains - = dict_index_contains_col_or_prefix(index, i); + /* Push down an index condition or an end_range check. */ + for (i = 0; i < n_fields; i++) { + ibool index_contains; + + if (innobase_is_v_fld(table->field[i])) { + index_contains = dict_index_contains_col_or_prefix( + index, num_v, true); + if (index_contains) + { + m_prebuilt->n_template = 0; + goto no_icp; + } + } else { + index_contains = dict_index_contains_col_or_prefix( + index, i - num_v, false); + } /* Test if an end_range or an index condition refers to the field. Note that "index" and "index_contains" may refer to the clustered index. - Index condition pushdown is relative to prebuilt->index - (the index that is being looked up first). */ + Index condition pushdown is relative to + m_prebuilt->index (the index that is being + looked up first). */ /* When join_read_always_key() invokes this code via handler::ha_index_init() and @@ -7798,44 +7782,54 @@ ha_innobase::build_template( the subset field->part_of_key.is_set(active_index) which would be acceptable if end_range==NULL. */ + bool is_v = innobase_is_v_fld(table->field[i]); if (build_template_needs_field_in_icp( - index, prebuilt, index_contains, i)) { + index, m_prebuilt, index_contains, + is_v ? num_v : i - num_v, is_v)) { /* Needed in ICP */ const Field* field; mysql_row_templ_t* templ; if (whole_row) { - field = table->field[sql_idx]; + field = table->field[i]; } else { field = build_template_needs_field( index_contains, - prebuilt->read_just_key, + m_prebuilt->read_just_key, fetch_all_in_key, fetch_primary_key_cols, - index, table, i, sql_idx); + index, table, i, num_v); if (!field) { + if (innobase_is_v_fld( + table->field[i])) { + num_v++; + } continue; } } templ = build_template_field( - prebuilt, clust_index, index, - table, field, i); - prebuilt->idx_cond_n_cols++; - ut_ad(prebuilt->idx_cond_n_cols - == prebuilt->n_template); + m_prebuilt, clust_index, index, + table, field, i - num_v, 0); + + ut_ad(!templ->is_virtual); - if (index == prebuilt->index) { + m_prebuilt->idx_cond_n_cols++; + ut_ad(m_prebuilt->idx_cond_n_cols + == m_prebuilt->n_template); + + if (index == m_prebuilt->index) { templ->icp_rec_field_no = templ->rec_field_no; } else { templ->icp_rec_field_no = dict_index_get_nth_col_pos( - prebuilt->index, i, - NULL); + m_prebuilt->index, + i - num_v, + &templ->rec_prefix_field_no); } - if (dict_index_is_clust(prebuilt->index)) { + if (dict_index_is_clust(m_prebuilt->index)) { ut_ad(templ->icp_rec_field_no != ULINT_UNDEFINED); /* If the primary key includes @@ -7846,7 +7840,7 @@ ha_innobase::build_template( off-page (externally stored) columns. */ if (templ->icp_rec_field_no - < prebuilt->index->n_uniq) { + < m_prebuilt->index->n_uniq) { /* This is a key column; all set. */ continue; @@ -7862,7 +7856,9 @@ ha_innobase::build_template( templ->icp_rec_field_no = dict_index_get_nth_col_or_prefix_pos( - prebuilt->index, i, TRUE, NULL); + m_prebuilt->index, i - num_v, + true, false, + &templ->rec_prefix_field_no); ut_ad(templ->icp_rec_field_no != ULINT_UNDEFINED); @@ -7882,91 +7878,146 @@ ha_innobase::build_template( we were unable to use an accurate condition for end_range in the "if" condition above, and the following assertion would fail. - ut_ad(!dict_index_is_clust(prebuilt->index) + ut_ad(!dict_index_is_clust(m_prebuilt->index) || templ->rec_field_no - < prebuilt->index->n_uniq); + < m_prebuilt->index->n_uniq); */ } + if (innobase_is_v_fld(table->field[i])) { + num_v++; + } } - ut_ad(prebuilt->idx_cond_n_cols > 0); - ut_ad(prebuilt->idx_cond_n_cols == prebuilt->n_template); + ut_ad(m_prebuilt->idx_cond_n_cols > 0); + ut_ad(m_prebuilt->idx_cond_n_cols == m_prebuilt->n_template); + + num_v = 0; /* Include the fields that are not needed in index condition pushdown. */ - for (i = 0, sql_idx = 0; i < n_stored_fields; i++, sql_idx++) { + for (i = 0; i < n_fields; i++) { + mysql_row_templ_t* templ; + ibool index_contains; - while (!table->field[sql_idx]->stored_in_db) { - sql_idx++; - } + if (innobase_is_v_fld(table->field[i])) { + index_contains = dict_index_contains_col_or_prefix( + index, num_v, true); + } else { + index_contains = dict_index_contains_col_or_prefix( + index, i - num_v, false); + } - const ibool index_contains - = dict_index_contains_col_or_prefix(index, i); + bool is_v = innobase_is_v_fld(table->field[i]); if (!build_template_needs_field_in_icp( - index, prebuilt, index_contains, i)) { + index, m_prebuilt, index_contains, + is_v ? num_v : i - num_v, is_v)) { /* Not needed in ICP */ const Field* field; if (whole_row) { - field = table->field[sql_idx]; + field = table->field[i]; } else { field = build_template_needs_field( index_contains, - prebuilt->read_just_key, + m_prebuilt->read_just_key, fetch_all_in_key, fetch_primary_key_cols, - index, table, i, sql_idx); + index, table, i, num_v); if (!field) { + if (innobase_is_v_fld(table->field[i])) { + num_v++; + } continue; } } - build_template_field(prebuilt, - clust_index, index, - table, field, i); + templ = build_template_field( + m_prebuilt, clust_index, index, + table, field, i - num_v, num_v); + + if (templ->is_virtual) { + num_v++; + } } } - prebuilt->idx_cond = this; + m_prebuilt->idx_cond = this; } else { +no_icp: + mysql_row_templ_t* templ; + ulint num_v = 0; /* No index condition pushdown */ - prebuilt->idx_cond = NULL; + m_prebuilt->idx_cond = NULL; - for (i = 0, sql_idx = 0; i < n_stored_fields; i++, sql_idx++) { + for (i = 0; i < n_fields; i++) { const Field* field; - while (!table->field[sql_idx]->stored_in_db) { - sql_idx++; - } - if (whole_row) { - field = table->field[sql_idx]; + /* Even this is whole_row, if the seach is + on a virtual column, and read_just_key is + set, and field is not in this index, we + will not try to fill the value since they + are not stored in such index nor in the + cluster index. */ + if (innobase_is_v_fld(table->field[i]) + && m_prebuilt->read_just_key + && !dict_index_contains_col_or_prefix( + m_prebuilt->index, num_v, true)) + { + /* Turn off ROW_MYSQL_WHOLE_ROW */ + m_prebuilt->template_type = + ROW_MYSQL_REC_FIELDS; + num_v++; + continue; + } + + field = table->field[i]; } else { + ibool contain; + + if (!innobase_is_v_fld(table->field[i])) { + contain = dict_index_contains_col_or_prefix( + index, i - num_v, + false); + } else if (dict_index_is_clust(index)) { + num_v++; + continue; + } else { + contain = dict_index_contains_col_or_prefix( + index, num_v, true); + } + field = build_template_needs_field( - dict_index_contains_col_or_prefix( - index, i), - prebuilt->read_just_key, + contain, + m_prebuilt->read_just_key, fetch_all_in_key, fetch_primary_key_cols, - index, table, i, sql_idx); + index, table, i, num_v); if (!field) { + if (innobase_is_v_fld(table->field[i])) { + num_v++; + } continue; } } - build_template_field(prebuilt, clust_index, index, - table, field, i); + templ = build_template_field( + m_prebuilt, clust_index, index, + table, field, i - num_v, num_v); + if (templ->is_virtual) { + num_v++; + } } } - if (index != clust_index && prebuilt->need_to_access_clustered) { + if (index != clust_index && m_prebuilt->need_to_access_clustered) { /* Change rec_field_no's to correspond to the clustered index record */ - for (i = 0; i < prebuilt->n_template; i++) { + for (i = 0; i < m_prebuilt->n_template; i++) { mysql_row_templ_t* templ - = &prebuilt->mysql_template[i]; + = &m_prebuilt->mysql_template[i]; templ->rec_field_no = templ->clust_rec_field_no; } @@ -7979,8 +8030,8 @@ binlogging. We need to eliminate the non-determinism that will arise in INSERT ... SELECT type of statements, since MySQL binlog only stores the min value of the autoinc interval. Once that is fixed we can get rid of the special lock handling. -@return DB_SUCCESS if all OK else error code */ -UNIV_INTERN +@return DB_SUCCESS if all OK else error code */ + dberr_t ha_innobase::innobase_lock_autoinc(void) /*====================================*/ @@ -7993,7 +8044,7 @@ ha_innobase::innobase_lock_autoinc(void) switch (innobase_autoinc_lock_mode) { case AUTOINC_NO_LOCKING: /* Acquire only the AUTOINC mutex. */ - dict_table_autoinc_lock(prebuilt->table); + dict_table_autoinc_lock(m_prebuilt->table); break; case AUTOINC_NEW_STYLE_LOCKING: @@ -8002,21 +8053,20 @@ ha_innobase::innobase_lock_autoinc(void) transaction has already acquired the AUTOINC lock on behalf of a LOAD FILE or INSERT ... SELECT etc. type of statement. */ - if (thd_sql_command(user_thd) == SQLCOM_INSERT - || thd_sql_command(user_thd) == SQLCOM_REPLACE - || thd_sql_command(user_thd) == SQLCOM_END // RBR event + if (thd_sql_command(m_user_thd) == SQLCOM_INSERT + || thd_sql_command(m_user_thd) == SQLCOM_REPLACE + || thd_sql_command(m_user_thd) == SQLCOM_END // RBR event ) { - dict_table_t* ib_table = prebuilt->table; /* Acquire the AUTOINC mutex. */ - dict_table_autoinc_lock(ib_table); + dict_table_autoinc_lock(m_prebuilt->table); /* We need to check that another transaction isn't already holding the AUTOINC lock on the table. */ - if (ib_table->n_waiting_or_granted_auto_inc_locks) { + if (m_prebuilt->table->n_waiting_or_granted_auto_inc_locks) { /* Release the mutex to avoid deadlocks and fall back to old style locking. */ - dict_table_autoinc_unlock(ib_table); + dict_table_autoinc_unlock(m_prebuilt->table); } else { /* Do not fall back to old style locking. */ break; @@ -8027,12 +8077,12 @@ ha_innobase::innobase_lock_autoinc(void) case AUTOINC_OLD_STYLE_LOCKING: DBUG_EXECUTE_IF("die_if_autoinc_old_lock_style_used", ut_ad(0);); - error = row_lock_table_autoinc_for_mysql(prebuilt); + error = row_lock_table_autoinc_for_mysql(m_prebuilt); if (error == DB_SUCCESS) { /* Acquire the AUTOINC mutex. */ - dict_table_autoinc_lock(prebuilt->table); + dict_table_autoinc_lock(m_prebuilt->table); } break; @@ -8044,33 +8094,10 @@ ha_innobase::innobase_lock_autoinc(void) } /********************************************************************//** -Reset the autoinc value in the table. -@return DB_SUCCESS if all went well else error code */ -UNIV_INTERN -dberr_t -ha_innobase::innobase_reset_autoinc( -/*================================*/ - ulonglong autoinc) /*!< in: value to store */ -{ - dberr_t error; - - error = innobase_lock_autoinc(); - - if (error == DB_SUCCESS) { - - dict_table_autoinc_initialize(prebuilt->table, autoinc); - - dict_table_autoinc_unlock(prebuilt->table); - } - - return(error); -} - -/********************************************************************//** Store the autoinc value in the table. The autoinc value is only set if it's greater than the existing autoinc value in the table. -@return DB_SUCCESS if all went well else error code */ -UNIV_INTERN +@return DB_SUCCESS if all went well else error code */ + dberr_t ha_innobase::innobase_set_max_autoinc( /*==================================*/ @@ -8082,9 +8109,9 @@ ha_innobase::innobase_set_max_autoinc( if (error == DB_SUCCESS) { - dict_table_autoinc_update_if_greater(prebuilt->table, auto_inc); + dict_table_autoinc_update_if_greater(m_prebuilt->table, auto_inc); - dict_table_autoinc_unlock(prebuilt->table); + dict_table_autoinc_unlock(m_prebuilt->table); } return(error); @@ -8093,38 +8120,38 @@ ha_innobase::innobase_set_max_autoinc( /********************************************************************//** Stores a row in an InnoDB database, to the table specified in this handle. -@return error code */ -UNIV_INTERN +@return error code */ + int ha_innobase::write_row( /*===================*/ uchar* record) /*!< in: a row in MySQL format */ { dberr_t error; - int error_result= 0; - ibool auto_inc_used= FALSE; #ifdef WITH_WSREP - ibool auto_inc_inserted= FALSE; /* if NULL was inserted */ + ibool auto_inc_inserted= FALSE; /* if NULL was inserted */ #endif - ulint sql_command; - trx_t* trx = thd_to_trx(user_thd); + int error_result = 0; + bool auto_inc_used = false; DBUG_ENTER("ha_innobase::write_row"); + trx_t* trx = thd_to_trx(m_user_thd); + + /* Validation checks before we commence write_row operation. */ if (high_level_read_only) { ib_senderrf(ha_thd(), IB_LOG_LEVEL_WARN, ER_READ_ONLY_MODE); DBUG_RETURN(HA_ERR_TABLE_READONLY); - } else if (prebuilt->trx != trx) { - sql_print_error("The transaction object for the table handle " - "is at %p, but for the current thread it is at " - "%p", - (const void*) prebuilt->trx, (const void*) trx); - - fputs("InnoDB: Dump of 200 bytes around prebuilt: ", stderr); - ut_print_buf(stderr, ((const byte*) prebuilt) - 100, 200); - fputs("\n" - "InnoDB: Dump of 200 bytes around ha_data: ", - stderr); + } else if (m_prebuilt->trx != trx) { + + ib::error() << "The transaction object for the table handle is" + " at " << static_cast<const void*>(m_prebuilt->trx) + << ", but for the current thread it is at " + << static_cast<const void*>(trx); + + fputs("InnoDB: Dump of 200 bytes around m_prebuilt: ", stderr); + ut_print_buf(stderr, ((const byte*) m_prebuilt) - 100, 200); + fputs("\nInnoDB: Dump of 200 bytes around ha_data: ", stderr); ut_print_buf(stderr, ((const byte*) trx) - 100, 200); putc('\n', stderr); ut_error; @@ -8132,142 +8159,12 @@ ha_innobase::write_row( ++trx->will_lock; } - ha_statistic_increment(&SSV::ha_write_count); - - sql_command = thd_sql_command(user_thd); - - if ((sql_command == SQLCOM_ALTER_TABLE - || sql_command == SQLCOM_OPTIMIZE - || sql_command == SQLCOM_CREATE_INDEX -#ifdef WITH_WSREP - || (wsrep_on(user_thd) && wsrep_load_data_splitting && - sql_command == SQLCOM_LOAD && - !thd_test_options( - user_thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) -#endif /* WITH_WSREP */ - || sql_command == SQLCOM_DROP_INDEX) - && num_write_row >= 10000) { -#ifdef WITH_WSREP - if (wsrep_on(user_thd) && sql_command == SQLCOM_LOAD) { - WSREP_DEBUG("forced trx split for LOAD: %s", - wsrep_thd_query(user_thd)); - } -#endif /* WITH_WSREP */ - /* ALTER TABLE is COMMITted at every 10000 copied rows. - The IX table lock for the original table has to be re-issued. - As this method will be called on a temporary table where the - contents of the original table is being copied to, it is - a bit tricky to determine the source table. The cursor - position in the source table need not be adjusted after the - intermediate COMMIT, since writes by other transactions are - being blocked by a MySQL table lock TL_WRITE_ALLOW_READ. */ - - dict_table_t* src_table; - enum lock_mode mode; - - num_write_row = 0; - - /* Commit the transaction. This will release the table - locks, so they have to be acquired again. */ - - /* Altering an InnoDB table */ - /* Get the source table. */ - src_table = lock_get_src_table( - prebuilt->trx, prebuilt->table, &mode); - if (!src_table) { -no_commit: - /* Unknown situation: do not commit */ - /* - ut_print_timestamp(stderr); - fprintf(stderr, - " InnoDB: ALTER TABLE is holding lock" - " on %lu tables!\n", - prebuilt->trx->mysql_n_tables_locked); - */ - ; - } else if (src_table == prebuilt->table) { -#ifdef WITH_WSREP - if (wsrep_on(user_thd) && - wsrep_load_data_splitting && - sql_command == SQLCOM_LOAD && - !thd_test_options(user_thd, - OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) - { - switch (wsrep_run_wsrep_commit(user_thd, 1)) - { - case WSREP_TRX_OK: - break; - case WSREP_TRX_SIZE_EXCEEDED: - case WSREP_TRX_CERT_FAIL: - case WSREP_TRX_ERROR: - DBUG_RETURN(1); - } - - if (binlog_hton->commit(binlog_hton, user_thd, 1)) - DBUG_RETURN(1); - wsrep_post_commit(user_thd, TRUE); - } -#endif /* WITH_WSREP */ - /* Source table is not in InnoDB format: - no need to re-acquire locks on it. */ - - /* Altering to InnoDB format */ - innobase_commit(ht, user_thd, 1); - /* Note that this transaction is still active. */ - trx_register_for_2pc(prebuilt->trx); - /* We will need an IX lock on the destination table. */ - prebuilt->sql_stat_start = TRUE; - } else { -#ifdef WITH_WSREP - if (wsrep_on(user_thd) && - wsrep_load_data_splitting && - sql_command == SQLCOM_LOAD && - !thd_test_options(user_thd, - OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) - { - switch (wsrep_run_wsrep_commit(user_thd, 1)) - { - case WSREP_TRX_OK: - break; - case WSREP_TRX_SIZE_EXCEEDED: - case WSREP_TRX_CERT_FAIL: - case WSREP_TRX_ERROR: - DBUG_RETURN(1); - } - - if (binlog_hton->commit(binlog_hton, user_thd, 1)) - DBUG_RETURN(1); - wsrep_post_commit(user_thd, TRUE); - } -#endif /* WITH_WSREP */ - /* Ensure that there are no other table locks than - LOCK_IX and LOCK_AUTO_INC on the destination table. */ - - if (!lock_is_table_exclusive(prebuilt->table, - prebuilt->trx)) { - goto no_commit; - } - - /* Commit the transaction. This will release the table - locks, so they have to be acquired again. */ - innobase_commit(ht, user_thd, 1); - /* Note that this transaction is still active. */ - trx_register_for_2pc(prebuilt->trx); - /* Re-acquire the table lock on the source table. */ - row_lock_table_for_mysql(prebuilt, src_table, mode); - /* We will need an IX lock on the destination table. */ - prebuilt->sql_stat_start = TRUE; - } - } - - num_write_row++; - - /* This is the case where the table has an auto-increment column */ + /* Handling of Auto-Increment Columns. */ if (table->next_number_field && record == table->record[0]) { /* Reset the error code before calling innobase_get_auto_increment(). */ - prebuilt->autoinc_error = DB_SUCCESS; + m_prebuilt->autoinc_error = DB_SUCCESS; #ifdef WITH_WSREP auto_inc_inserted= (table->next_number_field->val_int() == 0); @@ -8278,13 +8175,13 @@ no_commit: /* Handle the case where the AUTOINC sub-system failed during initialization. */ - if (prebuilt->autoinc_error == DB_UNSUPPORTED) { + if (m_prebuilt->autoinc_error == DB_UNSUPPORTED) { error_result = ER_AUTOINC_READ_FAILED; /* Set the error message to report too. */ my_error(ER_AUTOINC_READ_FAILED, MYF(0)); goto func_exit; - } else if (prebuilt->autoinc_error != DB_SUCCESS) { - error = prebuilt->autoinc_error; + } else if (m_prebuilt->autoinc_error != DB_SUCCESS) { + error = m_prebuilt->autoinc_error; goto report_error; } @@ -8292,11 +8189,13 @@ no_commit: goto func_exit; } - auto_inc_used = TRUE; + auto_inc_used = true; } - if (prebuilt->mysql_template == NULL - || prebuilt->template_type != ROW_MYSQL_WHOLE_ROW) { + /* Prepare INSERT graph that will be executed for actual INSERT + (This is a one time operation) */ + if (m_prebuilt->mysql_template == NULL + || m_prebuilt->template_type != ROW_MYSQL_WHOLE_ROW) { /* Build the template used in converting quickly between the two database formats */ @@ -8304,12 +8203,14 @@ no_commit: build_template(true); } - innobase_srv_conc_enter_innodb(prebuilt->trx); + innobase_srv_conc_enter_innodb(m_prebuilt); + + /* Execute insert graph that will result in actual insert. */ + error = row_insert_for_mysql((byte*) record, m_prebuilt); - error = row_insert_for_mysql((byte*) record, prebuilt); - DEBUG_SYNC(user_thd, "ib_after_row_insert"); + DEBUG_SYNC(m_user_thd, "ib_after_row_insert"); - /* Handle duplicate key errors */ + /* Handling of errors related to auto-increment. */ if (auto_inc_used) { ulonglong auto_inc; ulonglong col_max_value; @@ -8317,7 +8218,7 @@ no_commit: /* Note the number of rows processed for this statement, used by get_auto_increment() to determine the number of AUTO-INC values to reserve. This is only useful for a mult-value INSERT - and is a statement level counter.*/ + and is a statement level counter. */ if (trx->n_autoinc_rows > 0) { --trx->n_autoinc_rows; } @@ -8338,13 +8239,11 @@ no_commit: must update the autoinc counter if we are performing those statements. */ - switch (sql_command) { + switch (thd_sql_command(m_user_thd)) { case SQLCOM_LOAD: - if (trx->duplicates) { - - goto set_max_autoinc; + if (!trx->duplicates) { + break; } - break; case SQLCOM_REPLACE: case SQLCOM_INSERT_SELECT: @@ -8357,27 +8256,27 @@ no_commit: WSREP_DEBUG("DUPKEY error for autoinc\n" "THD %ld, value %llu, off %llu inc %llu", - thd_get_thread_id(current_thd), + thd_get_thread_id(m_user_thd), auto_inc, - prebuilt->autoinc_offset, - prebuilt->autoinc_increment); + m_prebuilt->autoinc_offset, + m_prebuilt->autoinc_increment); - if (wsrep_on(current_thd) && + if (wsrep_on(m_user_thd) && auto_inc_inserted && wsrep_drupal_282555_workaround && - wsrep_thd_retry_counter(current_thd) == 0 && - !thd_test_options(current_thd, - OPTION_NOT_AUTOCOMMIT | + wsrep_thd_retry_counter(m_user_thd) == 0 && + !thd_test_options(m_user_thd, + OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { WSREP_DEBUG( "retrying insert: %s", - (*wsrep_thd_query(current_thd)) ? - wsrep_thd_query(current_thd) : + (*wsrep_thd_query(m_user_thd)) ? + wsrep_thd_query(m_user_thd) : (char *)"void"); error= DB_SUCCESS; wsrep_thd_set_conflict_state( - current_thd, MUST_ABORT); - innobase_srv_conc_exit_innodb(prebuilt->trx); + m_user_thd, MUST_ABORT); + innobase_srv_conc_exit_innodb(m_prebuilt); /* jump straight to func exit over * later wsrep hooks */ goto func_exit; @@ -8395,21 +8294,21 @@ no_commit: /* If the actual value inserted is greater than the upper limit of the interval, then we try and update the table upper limit. Note: last_value - will be 0 if get_auto_increment() was not called.*/ + will be 0 if get_auto_increment() was not called. */ - if (auto_inc >= prebuilt->autoinc_last_value) { + if (auto_inc >= m_prebuilt->autoinc_last_value) { set_max_autoinc: /* This should filter out the negative values set explicitly by the user. */ if (auto_inc <= col_max_value) { - ut_a(prebuilt->autoinc_increment > 0); + ut_a(m_prebuilt->autoinc_increment > 0); ulonglong offset; ulonglong increment; dberr_t err; - offset = prebuilt->autoinc_offset; - increment = prebuilt->autoinc_increment; + offset = m_prebuilt->autoinc_offset; + increment = m_prebuilt->autoinc_increment; auto_inc = innobase_next_autoinc( auto_inc, @@ -8430,9 +8329,10 @@ set_max_autoinc: } } - innobase_srv_conc_exit_innodb(prebuilt->trx); + innobase_srv_conc_exit_innodb(m_prebuilt); report_error: + /* Cleanup and exit. */ if (error == DB_TABLESPACE_DELETED) { ib_senderrf( trx->mysql_thd, IB_LOG_LEVEL_ERROR, @@ -8440,17 +8340,16 @@ report_error: table->s->table_name.str); } - error_result = convert_error_code_to_mysql(error, - prebuilt->table->flags, - user_thd); + error_result = convert_error_code_to_mysql( + error, m_prebuilt->table->flags, m_user_thd); #ifdef WITH_WSREP if (!error_result - && wsrep_on(user_thd) - && wsrep_thd_exec_mode(user_thd) == LOCAL_STATE - && !wsrep_consistency_check(user_thd) - && !wsrep_thd_ignore_table(user_thd)) { - if (wsrep_append_keys(user_thd, false, record, NULL)) { + && wsrep_on(m_user_thd) + && wsrep_thd_exec_mode(m_user_thd) == LOCAL_STATE + && !wsrep_consistency_check(m_user_thd) + && !wsrep_thd_ignore_table(m_user_thd)) { + if (wsrep_append_keys(m_user_thd, false, record, NULL)) { DBUG_PRINT("wsrep", ("row key failed")); error_result = HA_ERR_INTERNAL_ERROR; goto wsrep_error; @@ -8469,32 +8368,78 @@ func_exit: DBUG_RETURN(error_result); } -/**********************************************************************//** -Checks which fields have changed in a row and stores information -of them to an update vector. -@return DB_SUCCESS or error code */ +/** Fill the update vector's "old_vrow" field for those non-updated, +but indexed columns. Such columns could stil present in the virtual +index rec fields even if they are not updated (some other fields updated), +so needs to be logged. +@param[in] prebuilt InnoDB prebuilt struct +@param[in,out] vfield field to filled +@param[in] o_len actual column length +@param[in,out] col column to be filled +@param[in] old_mysql_row_col MySQL old field ptr +@param[in] col_pack_len MySQL field col length +@param[in,out] buf buffer for a converted integer value +@return used buffer ptr from row_mysql_store_col_in_innobase_format() */ +static +byte* +innodb_fill_old_vcol_val( + row_prebuilt_t* prebuilt, + dfield_t* vfield, + ulint o_len, + dict_col_t* col, + const byte* old_mysql_row_col, + ulint col_pack_len, + byte* buf) +{ + dict_col_copy_type( + col, dfield_get_type(vfield)); + if (o_len != UNIV_SQL_NULL) { + + buf = row_mysql_store_col_in_innobase_format( + vfield, + buf, + TRUE, + old_mysql_row_col, + col_pack_len, + dict_table_is_comp(prebuilt->table)); + } else { + dfield_set_null(vfield); + } + + return(buf); +} + +/** Calculate an update vector corresponding to the changes +between old_row and new_row. +@param[out] uvect update vector +@param[in] old_row current row in MySQL format +@param[in] new_row intended updated row in MySQL format +@param[in] table MySQL table handle +@param[in,out] upd_buff buffer to use for converted values +@param[in] buff_len length of upd_buff +@param[in,out] prebuilt InnoDB execution context +@param[out] auto_inc updated AUTO_INCREMENT value, or 0 if none +@return DB_SUCCESS or error code */ static dberr_t calc_row_difference( -/*================*/ - upd_t* uvect, /*!< in/out: update vector */ - uchar* old_row, /*!< in: old row in MySQL format */ - uchar* new_row, /*!< in: new row in MySQL format */ - TABLE* table, /*!< in: table in MySQL data - dictionary */ - uchar* upd_buff, /*!< in: buffer to use */ - ulint buff_len, /*!< in: buffer length */ - row_prebuilt_t* prebuilt, /*!< in: InnoDB prebuilt struct */ - THD* thd) /*!< in: user thread */ + upd_t* uvect, + const uchar* old_row, + uchar* new_row, + TABLE* table, + uchar* upd_buff, + ulint buff_len, + row_prebuilt_t* prebuilt, + ib_uint64_t& auto_inc) { uchar* original_upd_buff = upd_buff; Field* field; enum_field_types field_mysql_type; - uint n_fields; ulint o_len; ulint n_len; ulint col_pack_len; const byte* new_mysql_row_col; + const byte* old_mysql_row_col; const byte* o_ptr; const byte* n_ptr; byte* buf; @@ -8503,24 +8448,32 @@ calc_row_difference( ulint n_changed = 0; dfield_t dfield; dict_index_t* clust_index; - uint sql_idx, innodb_idx= 0; + uint i; ibool changes_fts_column = FALSE; ibool changes_fts_doc_col = FALSE; - trx_t* trx = thd_to_trx(thd); + trx_t* const trx = prebuilt->trx; doc_id_t doc_id = FTS_NULL_DOC_ID; + ulint num_v = 0; + uint n_fields = mysql_fields(table); ut_ad(!srv_read_only_mode); - n_fields = table->s->fields; clust_index = dict_table_get_first_index(prebuilt->table); + auto_inc = 0; /* We use upd_buff to convert changed fields */ buf = (byte*) upd_buff; - for (sql_idx = 0; sql_idx < n_fields; sql_idx++) { - field = table->field[sql_idx]; - if (!field->stored_in_db) - continue; + for (i = 0; i < n_fields; i++) { + field = table->field[i]; + bool is_virtual = innobase_is_v_fld(field); + dict_col_t* col; + + if (is_virtual) { + col = &prebuilt->table->v_cols[num_v].m_col; + } else { + col = &prebuilt->table->cols[i - num_v]; + } o_ptr = (const byte*) old_row + get_field_offset(table, field); n_ptr = (const byte*) new_row + get_field_offset(table, field); @@ -8528,6 +8481,7 @@ calc_row_difference( /* Use new_mysql_row_col and col_pack_len save the values */ new_mysql_row_col = n_ptr; + old_mysql_row_col = o_ptr; col_pack_len = field->pack_length(); o_len = col_pack_len; @@ -8538,11 +8492,12 @@ calc_row_difference( field_mysql_type = field->type(); - col_type = prebuilt->table->cols[innodb_idx].mtype; + col_type = col->mtype; switch (col_type) { case DATA_BLOB: + case DATA_GEOMETRY: o_ptr = row_mysql_read_blob_ref(&o_len, o_ptr, o_len); n_ptr = row_mysql_read_blob_ref(&n_len, n_ptr, n_len); @@ -8583,7 +8538,6 @@ calc_row_difference( } } - if (field->real_maybe_null()) { if (field->is_null_in_record(old_row)) { o_len = UNIV_SQL_NULL; @@ -8594,6 +8548,64 @@ calc_row_difference( } } +#ifdef UNIV_DEBUG + bool online_ord_part = false; +#endif + + if (is_virtual) { + /* If the virtual column is not indexed, + we shall ignore it for update */ + if (!col->ord_part) { + /* Check whether there is a table-rebuilding + online ALTER TABLE in progress, and this + virtual column could be newly indexed, thus + it will be materialized. Then we will have + to log its update. + Note, we do not support online dropping virtual + column while adding new index, nor with + online alter column order while adding index, + so the virtual column sequence must not change + if it is online operation */ + if (dict_index_is_online_ddl(clust_index) + && row_log_col_is_indexed(clust_index, + num_v)) { +#ifdef UNIV_DEBUG + online_ord_part = true; +#endif + } else { + num_v++; + continue; + } + } + + if (!uvect->old_vrow) { + uvect->old_vrow = dtuple_create_with_vcol( + uvect->heap, 0, prebuilt->table->n_v_cols); + } + + ulint max_field_len = DICT_MAX_FIELD_LEN_BY_FORMAT( + prebuilt->table); + + /* for virtual columns, we only materialize + its index, and index field length would not + exceed max_field_len. So continue if the + first max_field_len bytes are matched up */ + if (o_len != UNIV_SQL_NULL + && n_len != UNIV_SQL_NULL + && o_len >= max_field_len + && n_len >= max_field_len + && memcmp(o_ptr, n_ptr, max_field_len) == 0) { + dfield_t* vfield = dtuple_get_nth_v_field( + uvect->old_vrow, num_v); + buf = innodb_fill_old_vcol_val( + prebuilt, vfield, o_len, + col, old_mysql_row_col, + col_pack_len, buf); + num_v++; + continue; + } + } + if (o_len != n_len || (o_len != 0 && o_len != UNIV_SQL_NULL && 0 != memcmp(o_ptr, n_ptr, o_len))) { /* The field has changed */ @@ -8604,10 +8616,18 @@ calc_row_difference( /* Let us use a dummy dfield to make the conversion from the MySQL column format to the InnoDB format */ + + /* If the length of new geometry object is 0, means + this object is invalid geometry object, we need + to block it. */ + if (DATA_GEOMETRY_MTYPE(col_type) + && o_len != 0 && n_len == 0) { + return(DB_CANT_CREATE_GEOMETRY_OBJECT); + } + if (n_len != UNIV_SQL_NULL) { - dict_col_copy_type(prebuilt->table->cols + - innodb_idx, - dfield_get_type(&dfield)); + dict_col_copy_type( + col, dfield_get_type(&dfield)); buf = row_mysql_store_col_in_innobase_format( &dfield, @@ -8618,13 +8638,63 @@ calc_row_difference( dict_table_is_comp(prebuilt->table)); dfield_copy(&ufield->new_val, &dfield); } else { + dict_col_copy_type( + col, dfield_get_type(&ufield->new_val)); dfield_set_null(&ufield->new_val); } ufield->exp = NULL; ufield->orig_len = 0; - ufield->field_no = dict_col_get_clust_pos( - &prebuilt->table->cols[innodb_idx], clust_index); + if (is_virtual) { + dfield_t* vfield = dtuple_get_nth_v_field( + uvect->old_vrow, num_v); + upd_fld_set_virtual_col(ufield); + ufield->field_no = num_v; + + ut_ad(col->ord_part || online_ord_part); + ufield->old_v_val = static_cast<dfield_t*>( + mem_heap_alloc( + uvect->heap, + sizeof *ufield->old_v_val)); + + if (!field->is_null_in_record(old_row)) { + if (n_len == UNIV_SQL_NULL) { + dict_col_copy_type( + col, dfield_get_type( + &dfield)); + } + + buf = row_mysql_store_col_in_innobase_format( + &dfield, + (byte*) buf, + TRUE, + old_mysql_row_col, + col_pack_len, + dict_table_is_comp( + prebuilt->table)); + dfield_copy(ufield->old_v_val, + &dfield); + dfield_copy(vfield, &dfield); + } else { + dict_col_copy_type( + col, dfield_get_type( + ufield->old_v_val)); + dfield_set_null(ufield->old_v_val); + dfield_set_null(vfield); + } + num_v++; + ut_ad(field != table->found_next_number_field); + } else { + ufield->field_no = dict_col_get_clust_pos( + &prebuilt->table->cols[i - num_v], + clust_index); + ufield->old_v_val = NULL; + if (field != table->found_next_number_field + || dfield_is_null(&ufield->new_val)) { + } else { + auto_inc = field->val_uint(); + } + } n_changed++; /* If an FTS indexed column was changed by this @@ -8636,8 +8706,8 @@ calc_row_difference( checking only once here. Later we will need to note which columns have been updated and do selective processing. */ - if (prebuilt->table->fts != NULL) { - ulint offset; + if (prebuilt->table->fts != NULL && !is_virtual) { + ulint offset; dict_table_t* innodb_table; innodb_table = prebuilt->table; @@ -8657,9 +8727,16 @@ calc_row_difference( innodb_table, ufield); } } + } else if (is_virtual) { + dfield_t* vfield = dtuple_get_nth_v_field( + uvect->old_vrow, num_v); + buf = innodb_fill_old_vcol_val( + prebuilt, vfield, o_len, + col, old_mysql_row_col, + col_pack_len, buf); + ut_ad(col->ord_part || online_ord_part); + num_v++; } - if (field->stored_in_db) - innodb_idx++; } /* If the update changes a column with an FTS index on it, we @@ -8667,7 +8744,7 @@ calc_row_difference( other changes. We piggy back our changes on the normal UPDATE to reduce processing and IO overhead. */ if (!prebuilt->table->fts) { - trx->fts_next_doc_id = 0; + trx->fts_next_doc_id = 0; } else if (changes_fts_column || changes_fts_doc_col) { dict_table_t* innodb_table = prebuilt->table; @@ -8681,37 +8758,31 @@ calc_row_difference( Doc ID must also be updated. Otherwise, return error */ if (changes_fts_column && !changes_fts_doc_col) { - ut_print_timestamp(stderr); - fprintf(stderr, " InnoDB: A new Doc ID" - " must be supplied while updating" - " FTS indexed columns.\n"); + ib::warn() << "A new Doc ID must be supplied" + " while updating FTS indexed columns."; return(DB_FTS_INVALID_DOCID); } /* Doc ID must monotonically increase */ ut_ad(innodb_table->fts->cache); if (doc_id < prebuilt->table->fts->cache->next_doc_id) { - fprintf(stderr, - "InnoDB: FTS Doc ID must be larger than" - " " IB_ID_FMT " for table", - innodb_table->fts->cache->next_doc_id - - 1); - ut_print_name(stderr, trx, - TRUE, innodb_table->name); - putc('\n', stderr); + + ib::warn() << "FTS Doc ID must be larger than " + << innodb_table->fts->cache->next_doc_id + - 1 << " for table " + << innodb_table->name; return(DB_FTS_INVALID_DOCID); } else if ((doc_id - prebuilt->table->fts->cache->next_doc_id) >= FTS_DOC_ID_MAX_STEP) { - fprintf(stderr, - "InnoDB: Doc ID " UINT64PF " is too" + + ib::warn() << "Doc ID " << doc_id << " is too" " big. Its difference with largest" - " Doc ID used " UINT64PF " cannot" - " exceed or equal to %d\n", - doc_id, - prebuilt->table->fts->cache->next_doc_id - 1, - FTS_DOC_ID_MAX_STEP); + " Doc ID used " << prebuilt->table->fts + ->cache->next_doc_id - 1 + << " cannot exceed or equal to " + << FTS_DOC_ID_MAX_STEP; } @@ -8743,6 +8814,7 @@ calc_row_difference( ut_a(buf <= (byte*) original_upd_buff + buff_len); + ut_ad(uvect->validate()); return(DB_SUCCESS); } @@ -8767,9 +8839,9 @@ wsrep_calc_row_hash( uint i; void *ctx = alloca(my_md5_context_size()); - my_md5_init(ctx); + my_md5_init(ctx); - n_fields = table->s->fields; + n_fields = mysql_fields(table); for (i = 0; i < n_fields; i++) { byte null_byte=0; @@ -8828,28 +8900,31 @@ wsrep_calc_row_hash( return(0); } #endif /* WITH_WSREP */ -/**********************************************************************//** + +/** Updates a row given as a parameter to a new value. Note that we are given whole rows, not just the fields which are updated: this incurs some overhead for CPU when we check which fields are actually updated. TODO: currently InnoDB does not prevent the 'Halloween problem': in a searched update a single row can get updated several times if its index columns are updated! -@return error number or 0 */ -UNIV_INTERN +@param[in] old_row Old row contents in MySQL format +@param[out] new_row Updated row contents in MySQL format +@return error number or 0 */ + int ha_innobase::update_row( -/*====================*/ - const uchar* old_row, /*!< in: old row in MySQL format */ - uchar* new_row) /*!< in: new row in MySQL format */ + const uchar* old_row, + uchar* new_row) { - upd_t* uvect; + int err; + dberr_t error; - trx_t* trx = thd_to_trx(user_thd); + trx_t* trx = thd_to_trx(m_user_thd); DBUG_ENTER("ha_innobase::update_row"); - ut_a(prebuilt->trx == trx); + ut_a(m_prebuilt->trx == trx); if (high_level_read_only) { ib_senderrf(ha_thd(), IB_LOG_LEVEL_WARN, ER_READ_ONLY_MODE); @@ -8858,106 +8933,92 @@ ha_innobase::update_row( ++trx->will_lock; } - if (upd_buf == NULL) { - ut_ad(upd_buf_size == 0); + if (m_upd_buf == NULL) { + ut_ad(m_upd_buf_size == 0); + + /* Create a buffer for packing the fields of a record. Why + table->reclength did not work here? Obviously, because char + fields when packed actually became 1 byte longer, when we also + stored the string length as the first byte. */ + + m_upd_buf_size = table->s->reclength + table->s->max_key_length + + MAX_REF_PARTS * 3; - /* Create a buffer for packing the fields of a record. Why - table->stored_rec_length did not work here? Obviously, - because char fields when packed actually became 1 byte - longer, when we also stored the string length as the first - byte. */ + m_upd_buf = reinterpret_cast<uchar*>( + my_malloc(//PSI_INSTRUMENT_ME, + m_upd_buf_size, + MYF(MY_WME))); - upd_buf_size = table->s->stored_rec_length + - table->s->max_key_length + MAX_REF_PARTS * 3; - upd_buf = (uchar*) my_malloc(upd_buf_size, MYF(MY_WME)); - if (upd_buf == NULL) { - upd_buf_size = 0; + if (m_upd_buf == NULL) { + m_upd_buf_size = 0; DBUG_RETURN(HA_ERR_OUT_OF_MEM); } } - ha_statistic_increment(&SSV::ha_update_count); - - if (prebuilt->upd_node) { - uvect = prebuilt->upd_node->update; - } else { - uvect = row_get_prebuilt_update_vector(prebuilt); - } + upd_t* uvect = row_get_prebuilt_update_vector(m_prebuilt); + ib_uint64_t autoinc; /* Build an update vector from the modified fields in the rows - (uses upd_buf of the handle) */ + (uses m_upd_buf of the handle) */ - error = calc_row_difference(uvect, (uchar*) old_row, new_row, table, - upd_buf, upd_buf_size, prebuilt, user_thd); + error = calc_row_difference( + uvect, old_row, new_row, table, m_upd_buf, m_upd_buf_size, + m_prebuilt, autoinc); if (error != DB_SUCCESS) { goto func_exit; } - /* This is not a delete */ - prebuilt->upd_node->is_delete = FALSE; - - ut_a(prebuilt->template_type == ROW_MYSQL_WHOLE_ROW); - - innobase_srv_conc_enter_innodb(trx); - - error = row_update_for_mysql((byte*) old_row, prebuilt); - - /* We need to do some special AUTOINC handling for the following case: - - INSERT INTO t (c1,c2) VALUES(x,y) ON DUPLICATE KEY UPDATE ... - - We need to use the AUTOINC counter that was actually used by - MySQL in the UPDATE statement, which can be different from the - value used in the INSERT statement.*/ - - if (error == DB_SUCCESS - && table->next_number_field - && new_row == table->record[0] - && thd_sql_command(user_thd) == SQLCOM_INSERT - && trx->duplicates) { - - ulonglong auto_inc; - ulonglong col_max_value; + if (!uvect->n_fields) { + /* This is the same as success, but instructs + MySQL that the row is not really updated and it + should not increase the count of updated rows. + This is fix for http://bugs.mysql.com/29157 */ + DBUG_RETURN(HA_ERR_RECORD_IS_THE_SAME); + } - auto_inc = table->next_number_field->val_uint(); + /* This is not a delete */ + m_prebuilt->upd_node->is_delete = FALSE; - /* We need the upper limit of the col type to check for - whether we update the table autoinc counter or not. */ - col_max_value = innobase_get_int_col_max_value( - table->next_number_field); + innobase_srv_conc_enter_innodb(m_prebuilt); - if (auto_inc <= col_max_value && auto_inc != 0) { + error = row_update_for_mysql(m_prebuilt); - ulonglong offset; - ulonglong increment; + if (error == DB_SUCCESS && autoinc) { + /* A value for an AUTO_INCREMENT column + was specified in the UPDATE statement. */ - offset = prebuilt->autoinc_offset; - increment = prebuilt->autoinc_increment; + autoinc = innobase_next_autoinc( + autoinc, 1, + m_prebuilt->autoinc_increment, + m_prebuilt->autoinc_offset, + innobase_get_int_col_max_value( + table->found_next_number_field)); - auto_inc = innobase_next_autoinc( - auto_inc, 1, increment, offset, col_max_value); + error = innobase_set_max_autoinc(autoinc); - error = innobase_set_max_autoinc(auto_inc); + if (m_prebuilt->table->persistent_autoinc) { + /* Update the PAGE_ROOT_AUTO_INC. Yes, we do + this even if dict_table_t::autoinc already was + greater than autoinc, because we cannot know + if any INSERT actually used (and wrote to + PAGE_ROOT_AUTO_INC) a value bigger than our + autoinc. */ + btr_write_autoinc(dict_table_get_first_index( + m_prebuilt->table), + autoinc); } } - innobase_srv_conc_exit_innodb(trx); + innobase_srv_conc_exit_innodb(m_prebuilt); func_exit: - int err = convert_error_code_to_mysql(error, - prebuilt->table->flags, user_thd); - - /* If success and no columns were updated. */ - if (err == 0 && uvect->n_fields == 0) { - - /* This is the same as success, but instructs - MySQL that the row is not really updated and it - should not increase the count of updated rows. - This is fix for http://bugs.mysql.com/29157 */ - err = HA_ERR_RECORD_IS_THE_SAME; - } else if (err == HA_FTS_INVALID_DOCID) { + if (error == DB_FTS_INVALID_DOCID) { + err = HA_FTS_INVALID_DOCID; my_error(HA_FTS_INVALID_DOCID, MYF(0)); + } else { + err = convert_error_code_to_mysql( + error, m_prebuilt->table->flags, m_user_thd); } /* Tell InnoDB server that there might be work for @@ -8966,14 +9027,13 @@ func_exit: innobase_active_small(); #ifdef WITH_WSREP - if (error == DB_SUCCESS && - wsrep_thd_exec_mode(user_thd) == LOCAL_STATE && - wsrep_on(user_thd) && - !wsrep_thd_ignore_table(user_thd)) - { + if (error == DB_SUCCESS && + wsrep_thd_exec_mode(m_user_thd) == LOCAL_STATE && + wsrep_on(m_user_thd) && + !wsrep_thd_ignore_table(m_user_thd)) { DBUG_PRINT("wsrep", ("update row key")); - if (wsrep_append_keys(user_thd, false, old_row, new_row)) { + if (wsrep_append_keys(m_user_thd, false, old_row, new_row)) { WSREP_DEBUG("WSREP: UPDATE_ROW_KEY FAILED"); DBUG_PRINT("wsrep", ("row key failed")); err = HA_ERR_INTERNAL_ERROR; @@ -8989,19 +9049,19 @@ wsrep_error: /**********************************************************************//** Deletes a row given as the parameter. -@return error number or 0 */ -UNIV_INTERN +@return error number or 0 */ + int ha_innobase::delete_row( /*====================*/ const uchar* record) /*!< in: a row in MySQL format */ { dberr_t error; - trx_t* trx = thd_to_trx(user_thd); + trx_t* trx = thd_to_trx(m_user_thd); DBUG_ENTER("ha_innobase::delete_row"); - ut_a(prebuilt->trx == trx); + ut_a(m_prebuilt->trx == trx); if (high_level_read_only) { ib_senderrf(ha_thd(), IB_LOG_LEVEL_WARN, ER_READ_ONLY_MODE); @@ -9010,21 +9070,19 @@ ha_innobase::delete_row( ++trx->will_lock; } - ha_statistic_increment(&SSV::ha_delete_count); - - if (!prebuilt->upd_node) { - row_get_prebuilt_update_vector(prebuilt); + if (!m_prebuilt->upd_node) { + row_get_prebuilt_update_vector(m_prebuilt); } /* This is a delete */ - prebuilt->upd_node->is_delete = TRUE; + m_prebuilt->upd_node->is_delete = TRUE; - innobase_srv_conc_enter_innodb(trx); + innobase_srv_conc_enter_innodb(m_prebuilt); - error = row_update_for_mysql((byte*) record, prebuilt); + error = row_update_for_mysql(m_prebuilt); - innobase_srv_conc_exit_innodb(trx); + innobase_srv_conc_exit_innodb(m_prebuilt); /* Tell the InnoDB server that there might be work for utility threads: */ @@ -9032,12 +9090,11 @@ ha_innobase::delete_row( innobase_active_small(); #ifdef WITH_WSREP - if (error == DB_SUCCESS && - wsrep_thd_exec_mode(user_thd) == LOCAL_STATE && - wsrep_on(user_thd) && - !wsrep_thd_ignore_table(user_thd)) - { - if (wsrep_append_keys(user_thd, false, record, NULL)) { + if (error == DB_SUCCESS && + wsrep_thd_exec_mode(m_user_thd) == LOCAL_STATE && + wsrep_on(m_user_thd) && + !wsrep_thd_ignore_table(m_user_thd)) { + if (wsrep_append_keys(m_user_thd, false, record, NULL)) { DBUG_PRINT("wsrep", ("delete fail")); error = (dberr_t) HA_ERR_INTERNAL_ERROR; goto wsrep_error; @@ -9046,46 +9103,49 @@ ha_innobase::delete_row( wsrep_error: #endif /* WITH_WSREP */ DBUG_RETURN(convert_error_code_to_mysql( - error, prebuilt->table->flags, user_thd)); + error, m_prebuilt->table->flags, m_user_thd)); +} + +/** Delete all rows from the table. +@return error number or 0 */ + +int +ha_innobase::delete_all_rows() +{ + DBUG_ENTER("ha_innobase::delete_all_rows"); + DBUG_RETURN(HA_ERR_WRONG_COMMAND); } /**********************************************************************//** Removes a new lock set on a row, if it was not read optimistically. This can be called after a row has been read in the processing of an UPDATE or a DELETE query, if the option innodb_locks_unsafe_for_binlog is set. */ -UNIV_INTERN + void ha_innobase::unlock_row(void) /*=========================*/ { DBUG_ENTER("ha_innobase::unlock_row"); - /* Consistent read does not take any locks, thus there is - nothing to unlock. */ - - if (prebuilt->select_lock_type == LOCK_NONE) { + if (m_prebuilt->select_lock_type == LOCK_NONE) { DBUG_VOID_RETURN; } - /* Ideally, this assert must be in the beginning of the function. - But there are some calls to this function from the SQL layer when the - transaction is in state TRX_STATE_NOT_STARTED. The check on - prebuilt->select_lock_type above gets around this issue. */ - ut_ad(trx_state_eq(prebuilt->trx, TRX_STATE_ACTIVE, true)); + ut_ad(trx_state_eq(m_prebuilt->trx, TRX_STATE_ACTIVE, true)); - switch (prebuilt->row_read_type) { + switch (m_prebuilt->row_read_type) { case ROW_READ_WITH_LOCKS: if (!srv_locks_unsafe_for_binlog - && prebuilt->trx->isolation_level + && m_prebuilt->trx->isolation_level > TRX_ISO_READ_COMMITTED) { break; } /* fall through */ case ROW_READ_TRY_SEMI_CONSISTENT: - row_unlock_for_mysql(prebuilt, FALSE); + row_unlock_for_mysql(m_prebuilt, FALSE); break; case ROW_READ_DID_SEMI_CONSISTENT: - prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT; + m_prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT; break; } @@ -9093,21 +9153,21 @@ ha_innobase::unlock_row(void) } /* See handler.h and row0mysql.h for docs on this function. */ -UNIV_INTERN + bool ha_innobase::was_semi_consistent_read(void) /*=======================================*/ { - return(prebuilt->row_read_type == ROW_READ_DID_SEMI_CONSISTENT); + return(m_prebuilt->row_read_type == ROW_READ_DID_SEMI_CONSISTENT); } /* See handler.h and row0mysql.h for docs on this function. */ -UNIV_INTERN + void ha_innobase::try_semi_consistent_read(bool yes) /*===========================================*/ { - ut_a(prebuilt->trx == thd_to_trx(ha_thd())); + ut_a(m_prebuilt->trx == thd_to_trx(ha_thd())); /* Row read type is set to semi consistent read if this was requested by the MySQL and either innodb_locks_unsafe_for_binlog @@ -9116,22 +9176,26 @@ ha_innobase::try_semi_consistent_read(bool yes) if (yes && (srv_locks_unsafe_for_binlog - || prebuilt->trx->isolation_level <= TRX_ISO_READ_COMMITTED)) { - prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT; + || m_prebuilt->trx->isolation_level + <= TRX_ISO_READ_COMMITTED)) { + + m_prebuilt->row_read_type = ROW_READ_TRY_SEMI_CONSISTENT; + } else { - prebuilt->row_read_type = ROW_READ_WITH_LOCKS; + m_prebuilt->row_read_type = ROW_READ_WITH_LOCKS; } } /******************************************************************//** Initializes a handle to use an index. -@return 0 or error number */ -UNIV_INTERN +@return 0 or error number */ + int ha_innobase::index_init( /*====================*/ - uint keynr, /*!< in: key (index) number */ - bool sorted) /*!< in: 1 if result MUST be sorted according to index */ + uint keynr, /*!< in: key (index) number */ + bool sorted) /*!< in: 1 if result MUST be sorted + according to index */ { DBUG_ENTER("index_init"); @@ -9140,65 +9204,55 @@ ha_innobase::index_init( /******************************************************************//** Currently does nothing. -@return 0 */ -UNIV_INTERN +@return 0 */ + int ha_innobase::index_end(void) /*========================*/ { - int error = 0; DBUG_ENTER("index_end"); + active_index = MAX_KEY; + in_range_check_pushed_down = FALSE; - ds_mrr.dsmrr_close(); - DBUG_RETURN(error); + + m_ds_mrr.dsmrr_close(); + + DBUG_RETURN(0); } /*********************************************************************//** Converts a search mode flag understood by MySQL to a flag understood by InnoDB. */ -static inline -ulint +page_cur_mode_t convert_search_mode_to_innobase( /*============================*/ - enum ha_rkey_function find_flag) + ha_rkey_function find_flag) { switch (find_flag) { case HA_READ_KEY_EXACT: /* this does not require the index to be UNIQUE */ - return(PAGE_CUR_GE); case HA_READ_KEY_OR_NEXT: return(PAGE_CUR_GE); - case HA_READ_KEY_OR_PREV: - return(PAGE_CUR_LE); case HA_READ_AFTER_KEY: return(PAGE_CUR_G); case HA_READ_BEFORE_KEY: return(PAGE_CUR_L); - case HA_READ_PREFIX: - return(PAGE_CUR_GE); + case HA_READ_KEY_OR_PREV: case HA_READ_PREFIX_LAST: - return(PAGE_CUR_LE); case HA_READ_PREFIX_LAST_OR_PREV: return(PAGE_CUR_LE); - /* In MySQL-4.0 HA_READ_PREFIX and HA_READ_PREFIX_LAST always - pass a complete-field prefix of a key value as the search - tuple. I.e., it is not allowed that the last field would - just contain n first bytes of the full field value. - MySQL uses a 'padding' trick to convert LIKE 'abc%' - type queries so that it can use as a search tuple - a complete-field-prefix of a key value. Thus, the InnoDB - search mode PAGE_CUR_LE_OR_EXTENDS is never used. - TODO: when/if MySQL starts to use also partial-field - prefixes, we have to deal with stripping of spaces - and comparison of non-latin1 char type fields in - innobase_mysql_cmp() to get PAGE_CUR_LE_OR_EXTENDS to - work correctly. */ case HA_READ_MBR_CONTAIN: + return(PAGE_CUR_CONTAIN); case HA_READ_MBR_INTERSECT: + return(PAGE_CUR_INTERSECT); case HA_READ_MBR_WITHIN: + return(PAGE_CUR_WITHIN); case HA_READ_MBR_DISJOINT: + return(PAGE_CUR_DISJOINT); case HA_READ_MBR_EQUAL: + return(PAGE_CUR_MBR_EQUAL); + case HA_READ_PREFIX: return(PAGE_CUR_UNSUPP); /* do not use "default:" in order to produce a gcc warning: enumeration value '...' not handled in switch @@ -9218,22 +9272,22 @@ the start of a new SQL statement, and what is associated with it. For each table in the database the MySQL interpreter may have several table handle instances in use, also in a single SQL query. For each table -handle instance there is an InnoDB 'prebuilt' struct which contains most +handle instance there is an InnoDB 'm_prebuilt' struct which contains most of the InnoDB data associated with this table handle instance. A) if the user has not explicitly set any MySQL table level locks: 1) MySQL calls ::external_lock to set an 'intention' table level lock on the table of the handle instance. There we set -prebuilt->sql_stat_start = TRUE. The flag sql_stat_start should be set +m_prebuilt->sql_stat_start = TRUE. The flag sql_stat_start should be set true if we are taking this table handle instance to use in a new SQL statement issued by the user. We also increment trx->n_mysql_tables_in_use. - 2) If prebuilt->sql_stat_start == TRUE we 'pre-compile' the MySQL search -instructions to prebuilt->template of the table handle instance in + 2) If m_prebuilt->sql_stat_start == TRUE we 'pre-compile' the MySQL search +instructions to m_prebuilt->template of the table handle instance in ::index_read. The template is used to save CPU time in large joins. - 3) In row_search_for_mysql, if prebuilt->sql_stat_start is true, we + 3) In row_search_for_mysql, if m_prebuilt->sql_stat_start is true, we allocate a new consistent read view for the trx if it does not yet have one, or in the case of a locking read, set an InnoDB 'intention' table level lock on the table. @@ -9262,8 +9316,8 @@ start of a new SQL statement. */ /**********************************************************************//** Positions an index cursor to the index specified in the handle. Fetches the row if any. -@return 0, HA_ERR_KEY_NOT_FOUND, or error number */ -UNIV_INTERN +@return 0, HA_ERR_KEY_NOT_FOUND, or error number */ + int ha_innobase::index_read( /*====================*/ @@ -9281,28 +9335,21 @@ ha_innobase::index_read( uint key_len,/*!< in: key value length */ enum ha_rkey_function find_flag)/*!< in: search flags from my_base.h */ { - ulint mode; - dict_index_t* index; - ulint match_mode = 0; - int error; - dberr_t ret; - DBUG_ENTER("index_read"); DEBUG_SYNC_C("ha_innobase_index_read_begin"); - ut_a(prebuilt->trx == thd_to_trx(user_thd)); + ut_a(m_prebuilt->trx == thd_to_trx(m_user_thd)); ut_ad(key_len != 0 || find_flag != HA_READ_KEY_EXACT); - ha_statistic_increment(&SSV::ha_read_key_count); - - index = prebuilt->index; + dict_index_t* index = m_prebuilt->index; - if (UNIV_UNLIKELY(index == NULL) || dict_index_is_corrupted(index)) { - prebuilt->index_usable = FALSE; + if (index == NULL || index->is_corrupted()) { + m_prebuilt->index_usable = FALSE; DBUG_RETURN(HA_ERR_CRASHED); } - if (UNIV_UNLIKELY(!prebuilt->index_usable)) { - DBUG_RETURN(dict_index_is_corrupted(index) + + if (!m_prebuilt->index_usable) { + DBUG_RETURN(index->is_corrupted() ? HA_ERR_INDEX_CORRUPT : HA_ERR_TABLE_DEF_CHANGED); } @@ -9311,105 +9358,122 @@ ha_innobase::index_read( DBUG_RETURN(HA_ERR_KEY_NOT_FOUND); } + /* For R-Tree index, we will always place the page lock to + pages being searched */ + if (dict_index_is_spatial(index)) { + ++m_prebuilt->trx->will_lock; + } + /* Note that if the index for which the search template is built is not - necessarily prebuilt->index, but can also be the clustered index */ + necessarily m_prebuilt->index, but can also be the clustered index */ - if (prebuilt->sql_stat_start) { + if (m_prebuilt->sql_stat_start) { build_template(false); } - if (key_ptr) { + if (key_ptr != NULL) { /* Convert the search key value to InnoDB format into - prebuilt->search_tuple */ + m_prebuilt->search_tuple */ row_sel_convert_mysql_key_to_innobase( - prebuilt->search_tuple, - prebuilt->srch_key_val1, - prebuilt->srch_key_val_len, + m_prebuilt->search_tuple, + m_prebuilt->srch_key_val1, + m_prebuilt->srch_key_val_len, index, (byte*) key_ptr, (ulint) key_len, - prebuilt->trx); - DBUG_ASSERT(prebuilt->search_tuple->n_fields > 0); + m_prebuilt->trx); + + DBUG_ASSERT(m_prebuilt->search_tuple->n_fields > 0); } else { /* We position the cursor to the last or the first entry in the index */ - dtuple_set_n_fields(prebuilt->search_tuple, 0); + dtuple_set_n_fields(m_prebuilt->search_tuple, 0); } - mode = convert_search_mode_to_innobase(find_flag); + page_cur_mode_t mode = convert_search_mode_to_innobase(find_flag); - match_mode = 0; + ulint match_mode = 0; if (find_flag == HA_READ_KEY_EXACT) { match_mode = ROW_SEL_EXACT; - } else if (find_flag == HA_READ_PREFIX - || find_flag == HA_READ_PREFIX_LAST) { + } else if (find_flag == HA_READ_PREFIX_LAST) { match_mode = ROW_SEL_EXACT_PREFIX; } - last_match_mode = (uint) match_mode; + m_last_match_mode = (uint) match_mode; + + dberr_t ret; if (mode != PAGE_CUR_UNSUPP) { - innobase_srv_conc_enter_innodb(prebuilt->trx); + innobase_srv_conc_enter_innodb(m_prebuilt); - ret = row_search_for_mysql((byte*) buf, mode, prebuilt, - match_mode, 0); + ret = row_search_mvcc( + buf, mode, m_prebuilt, match_mode, 0); - innobase_srv_conc_exit_innodb(prebuilt->trx); + innobase_srv_conc_exit_innodb(m_prebuilt); } else { ret = DB_UNSUPPORTED; } + DBUG_EXECUTE_IF("ib_select_query_failure", ret = DB_ERROR;); + + int error; + switch (ret) { case DB_SUCCESS: error = 0; table->status = 0; - if (prebuilt->table->is_system_db) { + if (m_prebuilt->table->is_system_db) { srv_stats.n_system_rows_read.add( - (size_t) prebuilt->trx->id, 1); + thd_get_thread_id(m_prebuilt->trx->mysql_thd), 1); } else { srv_stats.n_rows_read.add( - (size_t) prebuilt->trx->id, 1); + thd_get_thread_id(m_prebuilt->trx->mysql_thd), 1); } break; + case DB_RECORD_NOT_FOUND: error = HA_ERR_KEY_NOT_FOUND; table->status = STATUS_NOT_FOUND; break; + case DB_END_OF_INDEX: error = HA_ERR_KEY_NOT_FOUND; table->status = STATUS_NOT_FOUND; break; - case DB_TABLESPACE_DELETED: + case DB_TABLESPACE_DELETED: ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, ER_TABLESPACE_DISCARDED, table->s->table_name.str); table->status = STATUS_NOT_FOUND; error = HA_ERR_NO_SUCH_TABLE; break; + case DB_TABLESPACE_NOT_FOUND: ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, - ER_TABLESPACE_MISSING, MYF(0), + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + ER_TABLESPACE_MISSING, table->s->table_name.str); table->status = STATUS_NOT_FOUND; - error = HA_ERR_NO_SUCH_TABLE; + //error = HA_ERR_TABLESPACE_MISSING; + error = HA_ERR_NO_SUCH_TABLE; break; + default: error = convert_error_code_to_mysql( - ret, prebuilt->table->flags, user_thd); + ret, m_prebuilt->table->flags, m_user_thd); table->status = STATUS_NOT_FOUND; break; @@ -9421,8 +9485,8 @@ ha_innobase::index_read( /*******************************************************************//** The following functions works like index_read, but it find the last row with the current key value or prefix. -@return 0, HA_ERR_KEY_NOT_FOUND, or an error code */ -UNIV_INTERN +@return 0, HA_ERR_KEY_NOT_FOUND, or an error code */ + int ha_innobase::index_read_last( /*=========================*/ @@ -9437,8 +9501,8 @@ ha_innobase::index_read_last( /********************************************************************//** Get the index for a handle. Does not change active index. -@return NULL or index instance. */ -UNIV_INTERN +@return NULL or index instance. */ + dict_index_t* ha_innobase::innobase_get_index( /*============================*/ @@ -9446,68 +9510,26 @@ ha_innobase::innobase_get_index( clustered index, even if it was internally generated by InnoDB */ { - KEY* key = 0; - dict_index_t* index = 0; + KEY* key = NULL; + dict_table_t* ib_table = m_prebuilt->table; + dict_index_t* index; DBUG_ENTER("innobase_get_index"); if (keynr != MAX_KEY && table->s->keys > 0) { - key = table->key_info + keynr; - - index = innobase_index_lookup(share, keynr); - - if (index) { - if (!key || ut_strcmp(index->name, key->name) != 0) { - fprintf(stderr, "InnoDB: [Error] Index for key no %u" - " mysql name %s , InnoDB name %s for table %s\n", - keynr, key ? key->name : "NULL", - index->name, - prebuilt->table->name); - - for(ulint i=0; i < table->s->keys; i++) { - index = innobase_index_lookup(share, i); - key = table->key_info + keynr; - - if (index) { - - fprintf(stderr, "InnoDB: [Note] Index for key no %u" - " mysql name %s , InnoDB name %s for table %s\n", - keynr, key ? key->name : "NULL", - index->name, - prebuilt->table->name); - } - } - - } - - ut_a(ut_strcmp(index->name, key->name) == 0); - } else { - /* Can't find index with keynr in the translation - table. Only print message if the index translation - table exists */ - if (share->idx_trans_tbl.index_mapping) { - sql_print_warning("InnoDB could not find " - "index %s key no %u for " - "table %s through its " - "index translation table", - key ? key->name : "NULL", - keynr, - prebuilt->table->name); - } - - index = dict_table_get_index_on_name(prebuilt->table, - key->name); - } + key = &table->key_info[keynr]; + index = dict_table_get_index_on_name(ib_table, key->name); + ut_ad(index); } else { - index = dict_table_get_first_index(prebuilt->table); + index = dict_table_get_first_index(ib_table); } - if (!index) { + if (index == NULL) { sql_print_error( - "Innodb could not find key n:o %u with name %s " - "from dict cache for table %s", + "InnoDB could not find key no %u with name %s" + " from dict cache for table %s", keynr, key ? key->name : "NULL", - prebuilt->table->name); + ib_table->name.m_name); } DBUG_RETURN(index); @@ -9515,8 +9537,8 @@ ha_innobase::innobase_get_index( /********************************************************************//** Changes the active index of a handle. -@return 0 or error code */ -UNIV_INTERN +@return 0 or error code */ + int ha_innobase::change_active_index( /*=============================*/ @@ -9526,46 +9548,52 @@ ha_innobase::change_active_index( { DBUG_ENTER("change_active_index"); - ut_ad(user_thd == ha_thd()); - ut_a(prebuilt->trx == thd_to_trx(user_thd)); + ut_ad(m_user_thd == ha_thd()); + ut_a(m_prebuilt->trx == thd_to_trx(m_user_thd)); active_index = keynr; - prebuilt->index = innobase_get_index(keynr); + m_prebuilt->index = innobase_get_index(keynr); - if (UNIV_UNLIKELY(!prebuilt->index)) { + if (m_prebuilt->index == NULL) { sql_print_warning("InnoDB: change_active_index(%u) failed", keynr); - prebuilt->index_usable = FALSE; + m_prebuilt->index_usable = FALSE; DBUG_RETURN(1); } - prebuilt->index_usable = row_merge_is_index_usable(prebuilt->trx, - prebuilt->index); + m_prebuilt->index_usable = row_merge_is_index_usable( + m_prebuilt->trx, m_prebuilt->index); - if (UNIV_UNLIKELY(!prebuilt->index_usable)) { - if (dict_index_is_corrupted(prebuilt->index)) { - char index_name[MAX_FULL_NAME_LEN + 1]; - char table_name[MAX_FULL_NAME_LEN + 1]; - - innobase_format_name( - index_name, sizeof index_name, - prebuilt->index->name, TRUE); + if (!m_prebuilt->index_usable) { + if (m_prebuilt->index->is_corrupted()) { + char table_name[MAX_FULL_NAME_LEN + 1]; innobase_format_name( table_name, sizeof table_name, - prebuilt->index->table->name, FALSE); + m_prebuilt->index->table->name.m_name); - push_warning_printf( - user_thd, Sql_condition::WARN_LEVEL_WARN, - HA_ERR_INDEX_CORRUPT, - "InnoDB: Index %s for table %s is" - " marked as corrupted", - index_name, table_name); - DBUG_RETURN(HA_ERR_INDEX_CORRUPT); + if (m_prebuilt->index->is_primary()) { + ut_ad(m_prebuilt->index->table->corrupted); + push_warning_printf( + m_user_thd, Sql_condition::WARN_LEVEL_WARN, + ER_TABLE_CORRUPT, + "InnoDB: Table %s is corrupted.", + table_name); + DBUG_RETURN(ER_TABLE_CORRUPT); + } else { + push_warning_printf( + m_user_thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_INDEX_CORRUPT, + "InnoDB: Index %s for table %s is" + " marked as corrupted", + m_prebuilt->index->name(), + table_name); + DBUG_RETURN(HA_ERR_INDEX_CORRUPT); + } } else { push_warning_printf( - user_thd, Sql_condition::WARN_LEVEL_WARN, + m_user_thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_TABLE_DEF_CHANGED, "InnoDB: insufficient history for index %u", keynr); @@ -9574,15 +9602,41 @@ ha_innobase::change_active_index( /* The caller seems to ignore this. Thus, we must check this again in row_search_for_mysql(). */ DBUG_RETURN(convert_error_code_to_mysql(DB_MISSING_HISTORY, - 0, NULL)); + 0, NULL)); } - ut_a(prebuilt->search_tuple != 0); + ut_a(m_prebuilt->search_tuple != 0); - dtuple_set_n_fields(prebuilt->search_tuple, prebuilt->index->n_fields); + /* Initialization of search_tuple is not needed for FT index + since FT search returns rank only. In addition engine should + be able to retrieve FTS_DOC_ID column value if necessary. */ + if ((m_prebuilt->index->type & DICT_FTS)) { +#ifdef MYSQL_STORE_FTS_DOC_ID + if (table->fts_doc_id_field + && bitmap_is_set(table->read_set, + table->fts_doc_id_field->field_index + && m_prebuilt->read_just_key)) { + m_prebuilt->fts_doc_id_in_read_set = 1; + } +#endif + } else { + dtuple_set_n_fields(m_prebuilt->search_tuple, + m_prebuilt->index->n_fields); + + dict_index_copy_types( + m_prebuilt->search_tuple, m_prebuilt->index, + m_prebuilt->index->n_fields); - dict_index_copy_types(prebuilt->search_tuple, prebuilt->index, - prebuilt->index->n_fields); + /* If it's FTS query and FTS_DOC_ID exists FTS_DOC_ID field is + always added to read_set. */ + +#ifdef MYSQL_STORE_FTS_DOC_ID + m_prebuilt->fts_doc_id_in_read_set = + (m_prebuilt->read_just_key && table->fts_doc_id_field + && m_prebuilt->in_fts_query); +#endif + + } /* MySQL changes the active index for a handle also during some queries, for example SELECT MAX(a), SUM(a) first retrieves the MAX() @@ -9595,37 +9649,11 @@ ha_innobase::change_active_index( DBUG_RETURN(0); } -/**********************************************************************//** -Positions an index cursor to the index specified in keynr. Fetches the -row if any. -??? This is only used to read whole keys ??? -@return error number or 0 */ -UNIV_INTERN -int -ha_innobase::index_read_idx( -/*========================*/ - uchar* buf, /*!< in/out: buffer for the returned - row */ - uint keynr, /*!< in: use this index */ - const uchar* key, /*!< in: key value; if this is NULL - we position the cursor at the - start or end of index */ - uint key_len, /*!< in: key value length */ - enum ha_rkey_function find_flag)/*!< in: search flags from my_base.h */ -{ - if (change_active_index(keynr)) { - - return(1); - } - - return(index_read(buf, key, key_len, find_flag)); -} - /***********************************************************************//** Reads the next or previous row from a cursor, which must have previously been positioned using index_read. -@return 0, HA_ERR_END_OF_FILE, or error number */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error number */ + int ha_innobase::general_fetch( /*=======================*/ @@ -9635,50 +9663,42 @@ ha_innobase::general_fetch( uint match_mode) /*!< in: 0, ROW_SEL_EXACT, or ROW_SEL_EXACT_PREFIX */ { - dberr_t ret; - int error; - DBUG_ENTER("general_fetch"); - /* If transaction is not startted do not continue, instead return a error code. */ - if(!(prebuilt->sql_stat_start || (prebuilt->trx && prebuilt->trx->state == 1))) { - DBUG_RETURN(HA_ERR_END_OF_FILE); - } + const trx_t* trx = m_prebuilt->trx; - ut_a(prebuilt->trx == thd_to_trx(user_thd)); + ut_ad(trx == thd_to_trx(m_user_thd)); - if (prebuilt->table->is_readable()) { + if (m_prebuilt->table->is_readable()) { + } else if (m_prebuilt->table->corrupted) { + DBUG_RETURN(HA_ERR_CRASHED); } else { - if (prebuilt->table->corrupted) { - DBUG_RETURN(HA_ERR_CRASHED); - } else { - FilSpace space(prebuilt->table->space, true); + FilSpace space(m_prebuilt->table->space, true); - if (space()) { - DBUG_RETURN(HA_ERR_DECRYPTION_FAILED); - } else { - DBUG_RETURN(HA_ERR_NO_SUCH_TABLE); - } - } + DBUG_RETURN(space() + ? HA_ERR_DECRYPTION_FAILED + : HA_ERR_NO_SUCH_TABLE); } - innobase_srv_conc_enter_innodb(prebuilt->trx); + innobase_srv_conc_enter_innodb(m_prebuilt); - ret = row_search_for_mysql( - (byte*) buf, 0, prebuilt, match_mode, direction); + dberr_t ret = row_search_mvcc( + buf, PAGE_CUR_UNSUPP, m_prebuilt, match_mode, direction); - innobase_srv_conc_exit_innodb(prebuilt->trx); + innobase_srv_conc_exit_innodb(m_prebuilt); + + int error; switch (ret) { case DB_SUCCESS: error = 0; table->status = 0; - if (prebuilt->table->is_system_db) { + if (m_prebuilt->table->is_system_db) { srv_stats.n_system_rows_read.add( - (size_t) prebuilt->trx->id, 1); + thd_get_thread_id(trx->mysql_thd), 1); } else { srv_stats.n_rows_read.add( - (size_t) prebuilt->trx->id, 1); + thd_get_thread_id(trx->mysql_thd), 1); } break; case DB_RECORD_NOT_FOUND: @@ -9690,9 +9710,8 @@ ha_innobase::general_fetch( table->status = STATUS_NOT_FOUND; break; case DB_TABLESPACE_DELETED: - ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + trx->mysql_thd, IB_LOG_LEVEL_ERROR, ER_TABLESPACE_DISCARDED, table->s->table_name.str); @@ -9702,16 +9721,16 @@ ha_innobase::general_fetch( case DB_TABLESPACE_NOT_FOUND: ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + trx->mysql_thd, IB_LOG_LEVEL_ERROR, ER_TABLESPACE_MISSING, table->s->table_name.str); table->status = STATUS_NOT_FOUND; - error = HA_ERR_NO_SUCH_TABLE; + error = HA_ERR_TABLESPACE_MISSING; break; default: error = convert_error_code_to_mysql( - ret, prebuilt->table->flags, user_thd); + ret, m_prebuilt->table->flags, m_user_thd); table->status = STATUS_NOT_FOUND; break; @@ -9723,23 +9742,21 @@ ha_innobase::general_fetch( /***********************************************************************//** Reads the next row from a cursor, which must have previously been positioned using index_read. -@return 0, HA_ERR_END_OF_FILE, or error number */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error number */ + int ha_innobase::index_next( /*====================*/ uchar* buf) /*!< in/out: buffer for next row in MySQL format */ { - ha_statistic_increment(&SSV::ha_read_next_count); - return(general_fetch(buf, ROW_SEL_NEXT, 0)); } /*******************************************************************//** Reads the next row matching to the key value given as the parameter. -@return 0, HA_ERR_END_OF_FILE, or error number */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error number */ + int ha_innobase::index_next_same( /*=========================*/ @@ -9747,42 +9764,35 @@ ha_innobase::index_next_same( const uchar* key, /*!< in: key value */ uint keylen) /*!< in: key value length */ { - ha_statistic_increment(&SSV::ha_read_next_count); - - return(general_fetch(buf, ROW_SEL_NEXT, last_match_mode)); + return(general_fetch(buf, ROW_SEL_NEXT, m_last_match_mode)); } /***********************************************************************//** Reads the previous row from a cursor, which must have previously been positioned using index_read. -@return 0, HA_ERR_END_OF_FILE, or error number */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error number */ + int ha_innobase::index_prev( /*====================*/ uchar* buf) /*!< in/out: buffer for previous row in MySQL format */ { - ha_statistic_increment(&SSV::ha_read_prev_count); - return(general_fetch(buf, ROW_SEL_PREV, 0)); } /********************************************************************//** Positions a cursor on the first record in an index and reads the corresponding row to buf. -@return 0, HA_ERR_END_OF_FILE, or error code */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error code */ + int ha_innobase::index_first( /*=====================*/ uchar* buf) /*!< in/out: buffer for the row */ { - int error; - DBUG_ENTER("index_first"); - ha_statistic_increment(&SSV::ha_read_first_count); - error = index_read(buf, NULL, 0, HA_READ_AFTER_KEY); + int error = index_read(buf, NULL, 0, HA_READ_AFTER_KEY); /* MySQL does not seem to allow this to return HA_ERR_KEY_NOT_FOUND */ @@ -9796,19 +9806,16 @@ ha_innobase::index_first( /********************************************************************//** Positions a cursor on the last record in an index and reads the corresponding row to buf. -@return 0, HA_ERR_END_OF_FILE, or error code */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error code */ + int ha_innobase::index_last( /*====================*/ uchar* buf) /*!< in/out: buffer for the row */ { - int error; - DBUG_ENTER("index_last"); - ha_statistic_increment(&SSV::ha_read_last_count); - error = index_read(buf, NULL, 0, HA_READ_BEFORE_KEY); + int error = index_read(buf, NULL, 0, HA_READ_BEFORE_KEY); /* MySQL does not seem to allow this to return HA_ERR_KEY_NOT_FOUND */ @@ -9821,22 +9828,22 @@ ha_innobase::index_last( /****************************************************************//** Initialize a table scan. -@return 0 or error number */ -UNIV_INTERN +@return 0 or error number */ + int ha_innobase::rnd_init( /*==================*/ - bool scan) /*!< in: TRUE if table/index scan FALSE otherwise */ + bool scan) /*!< in: true if table/index scan FALSE otherwise */ { - int err; + int err; /* Store the active index value so that we can restore the original value after a scan */ - if (prebuilt->clust_index_was_generated) { + if (m_prebuilt->clust_index_was_generated) { err = change_active_index(MAX_KEY); } else { - err = change_active_index(primary_key); + err = change_active_index(m_primary_key); } /* Don't use semi-consistent read in random row reads (by position). @@ -9846,15 +9853,15 @@ ha_innobase::rnd_init( try_semi_consistent_read(0); } - start_of_scan = 1; + m_start_of_scan = true; return(err); } /*****************************************************************//** Ends a table scan. -@return 0 or error number */ -UNIV_INTERN +@return 0 or error number */ + int ha_innobase::rnd_end(void) /*======================*/ @@ -9865,8 +9872,8 @@ ha_innobase::rnd_end(void) /*****************************************************************//** Reads the next row in a table scan (also used to read the FIRST row in a table scan). -@return 0, HA_ERR_END_OF_FILE, or error number */ -UNIV_INTERN +@return 0, HA_ERR_END_OF_FILE, or error number */ + int ha_innobase::rnd_next( /*==================*/ @@ -9876,16 +9883,15 @@ ha_innobase::rnd_next( int error; DBUG_ENTER("rnd_next"); - ha_statistic_increment(&SSV::ha_read_rnd_next_count); - if (start_of_scan) { + if (m_start_of_scan) { error = index_first(buf); if (error == HA_ERR_KEY_NOT_FOUND) { error = HA_ERR_END_OF_FILE; } - start_of_scan = 0; + m_start_of_scan = false; } else { error = general_fetch(buf, ROW_SEL_NEXT, 0); } @@ -9895,8 +9901,8 @@ ha_innobase::rnd_next( /**********************************************************************//** Fetches a row from the table based on a row reference. -@return 0, HA_ERR_KEY_NOT_FOUND, or error code */ -UNIV_INTERN +@return 0, HA_ERR_KEY_NOT_FOUND, or error code */ + int ha_innobase::rnd_pos( /*=================*/ @@ -9906,20 +9912,17 @@ ha_innobase::rnd_pos( index was internally generated by InnoDB; the length of data in pos has to be ref_length */ { - int error; DBUG_ENTER("rnd_pos"); DBUG_DUMP("key", pos, ref_length); - ha_statistic_increment(&SSV::ha_read_rnd_count); - - ut_a(prebuilt->trx == thd_to_trx(ha_thd())); + ut_a(m_prebuilt->trx == thd_to_trx(ha_thd())); /* Note that we assume the length of the row reference is fixed for the table, and it is == ref_length */ - error = index_read(buf, pos, ref_length, HA_READ_KEY_EXACT); + int error = index_read(buf, pos, ref_length, HA_READ_KEY_EXACT); - if (error) { + if (error != 0) { DBUG_PRINT("error", ("Got error: %d", error)); } @@ -9929,7 +9932,7 @@ ha_innobase::rnd_pos( /**********************************************************************//** Initialize FT index scan @return 0 or error number */ -UNIV_INTERN + int ha_innobase::ft_init() /*==================*/ @@ -9953,7 +9956,7 @@ ha_innobase::ft_init() /**********************************************************************//** Initialize FT index scan @return FT_INFO structure if successful or NULL */ -UNIV_INTERN + FT_INFO* ha_innobase::ft_init_ext( /*=====================*/ @@ -9961,47 +9964,48 @@ ha_innobase::ft_init_ext( uint keynr, /* in: */ String* key) /* in: */ { - trx_t* trx; - dict_table_t* ft_table; - dberr_t error; - byte* query = (byte*) key->ptr(); - ulint query_len = key->length(); - const CHARSET_INFO* char_set = key->charset(); NEW_FT_INFO* fts_hdl = NULL; dict_index_t* index; fts_result_t* result; char buf_tmp[8192]; ulint buf_tmp_used; uint num_errors; + ulint query_len = key->length(); + const CHARSET_INFO* char_set = key->charset(); + const char* query = key->ptr(); if (fts_enable_diag_print) { - fprintf(stderr, "keynr=%u, '%.*s'\n", - keynr, (int) key->length(), (byte*) key->ptr()); + { + ib::info out; + out << "keynr=" << keynr << ", '"; + out.write(key->ptr(), key->length()); + } if (flags & FT_BOOL) { - fprintf(stderr, "BOOL search\n"); + ib::info() << "BOOL search"; } else { - fprintf(stderr, "NL search\n"); + ib::info() << "NL search"; } } /* FIXME: utf32 and utf16 are not compatible with some string function used. So to convert them to uft8 before - proceed. */ + we proceed. */ if (strcmp(char_set->csname, "utf32") == 0 || strcmp(char_set->csname, "utf16") == 0) { + buf_tmp_used = innobase_convert_string( buf_tmp, sizeof(buf_tmp) - 1, &my_charset_utf8_general_ci, query, query_len, (CHARSET_INFO*) char_set, &num_errors); - query = (byte*) buf_tmp; + buf_tmp[buf_tmp_used] = 0; + query = buf_tmp; query_len = buf_tmp_used; - query[query_len] = 0; } - trx = prebuilt->trx; + trx_t* trx = m_prebuilt->trx; /* FTS queries are not treated as autocommit non-locking selects. This is because the FTS implementation can acquire locks behind @@ -10012,7 +10016,7 @@ ha_innobase::ft_init_ext( ++trx->will_lock; } - ft_table = prebuilt->table; + dict_table_t* ft_table = m_prebuilt->table; /* Table does not have an FTS index */ if (!ft_table->fts || ib_vector_is_empty(ft_table->fts->indexes)) { @@ -10029,12 +10033,13 @@ ha_innobase::ft_init_ext( if (keynr == NO_SUCH_KEY) { /* FIXME: Investigate the NO_SUCH_KEY usage */ - index = (dict_index_t*) ib_vector_getp(ft_table->fts->indexes, 0); + index = reinterpret_cast<dict_index_t*> + (ib_vector_getp(ft_table->fts->indexes, 0)); } else { index = innobase_get_index(keynr); } - if (!index || index->type != DICT_FTS) { + if (index == NULL || index->type != DICT_FTS) { my_error(ER_TABLE_HAS_NO_FT, MYF(0)); return(NULL); } @@ -10045,51 +10050,32 @@ ha_innobase::ft_init_ext( ft_table->fts->fts_status |= ADDED_TABLE_SYNCED; } - error = fts_query(trx, index, flags, query, query_len, &result); + const byte* q = reinterpret_cast<const byte*>( + const_cast<char*>(query)); + + // FIXME: support ft_init_ext_with_hints(), pass LIMIT + dberr_t error = fts_query(trx, index, flags, q, query_len, &result); if (error != DB_SUCCESS) { - my_error(convert_error_code_to_mysql(error, 0, NULL), - MYF(0)); + my_error(convert_error_code_to_mysql(error, 0, NULL), MYF(0)); return(NULL); } /* Allocate FTS handler, and instantiate it before return */ - fts_hdl = static_cast<NEW_FT_INFO*>(my_malloc(sizeof(NEW_FT_INFO), - MYF(0))); + fts_hdl = reinterpret_cast<NEW_FT_INFO*>( + my_malloc(/*PSI_INSTRUMENT_ME,*/ sizeof(NEW_FT_INFO), MYF(0))); fts_hdl->please = const_cast<_ft_vft*>(&ft_vft_result); fts_hdl->could_you = const_cast<_ft_vft_ext*>(&ft_vft_ext_result); - fts_hdl->ft_prebuilt = prebuilt; + fts_hdl->ft_prebuilt = m_prebuilt; fts_hdl->ft_result = result; - /* FIXME: Re-evluate the condition when Bug 14469540 - is resolved */ - prebuilt->in_fts_query = true; - - return((FT_INFO*) fts_hdl); -} + /* FIXME: Re-evaluate the condition when Bug 14469540 is resolved */ + m_prebuilt->in_fts_query = true; -/*****************************************************************//** -Copy a cached MySQL row. -If requested, also avoids overwriting non-read columns. -@param[out] buf Row in MySQL format. -@param[in] cached_row Which row to copy. -@param[in] rec_len Record length. */ -void -ha_innobase::copy_cached_row( - uchar* buf, - const uchar* cached_row, - uint rec_len) -{ - if (prebuilt->keep_other_fields_on_keyread) { - row_sel_copy_cached_fields_for_mysql(buf, cached_row, - prebuilt); - } else { - memcpy(buf, cached_row, rec_len); - } + return(reinterpret_cast<FT_INFO*>(fts_hdl)); } - /*****************************************************************//** Set up search tuple for a query through FTS_DOC_ID_INDEX on supplied Doc ID. This is used by MySQL to retrieve the documents @@ -10098,7 +10084,7 @@ static void innobase_fts_create_doc_id_key( /*===========================*/ - dtuple_t* tuple, /* in/out: prebuilt->search_tuple */ + dtuple_t* tuple, /* in/out: m_prebuilt->search_tuple */ const dict_index_t* index, /* in: index (FTS_DOC_ID_INDEX) */ doc_id_t* doc_id) /* in/out: doc id to search, value @@ -10118,7 +10104,7 @@ innobase_fts_create_doc_id_key( dict_field_t* field = dict_index_get_nth_field(index, 0); ut_a(field->col->mtype == DATA_INT); ut_ad(sizeof(*doc_id) == field->fixed_len); - ut_ad(innobase_strcasecmp(index->name, FTS_DOC_ID_INDEX_NAME) == 0); + ut_ad(!strcmp(index->name, FTS_DOC_ID_INDEX_NAME)); #endif /* UNIV_DEBUG */ /* Convert to storage byte order */ @@ -10137,21 +10123,21 @@ innobase_fts_create_doc_id_key( /**********************************************************************//** Fetch next result from the FT result set @return error code */ -UNIV_INTERN + int ha_innobase::ft_read( /*=================*/ uchar* buf) /*!< in/out: buf contain result row */ { - fts_result_t* result; - int error; row_prebuilt_t* ft_prebuilt; - ft_prebuilt = ((NEW_FT_INFO*) ft_handler)->ft_prebuilt; + ft_prebuilt = reinterpret_cast<NEW_FT_INFO*>(ft_handler)->ft_prebuilt; - ut_a(ft_prebuilt == prebuilt); + ut_a(ft_prebuilt == m_prebuilt); + + fts_result_t* result; - result = ((NEW_FT_INFO*) ft_handler)->ft_result; + result = reinterpret_cast<NEW_FT_INFO*>(ft_handler)->ft_result; if (result->current == NULL) { /* This is the case where the FTS query did not @@ -10176,25 +10162,34 @@ ha_innobase::ft_read( next_record: if (result->current != NULL) { - dict_index_t* index; - dtuple_t* tuple = prebuilt->search_tuple; doc_id_t search_doc_id; + dtuple_t* tuple = m_prebuilt->search_tuple; /* If we only need information from result we can return without fetching the table row */ if (ft_prebuilt->read_just_key) { +#ifdef MYSQL_STORE_FTS_DOC_ID + if (m_prebuilt->fts_doc_id_in_read_set) { + fts_ranking_t* ranking; + ranking = rbt_value(fts_ranking_t, + result->current); + innobase_fts_store_docid( + table, ranking->doc_id); + } +#endif table->status= 0; return(0); } - index = dict_table_get_index_on_name( - prebuilt->table, FTS_DOC_ID_INDEX_NAME); + dict_index_t* index; + + index = m_prebuilt->table->fts_doc_id_index; /* Must find the index */ - ut_a(index); + ut_a(index != NULL); /* Switch to the FTS doc id index */ - prebuilt->index = index; + m_prebuilt->index = index; fts_ranking_t* ranking = rbt_value( fts_ranking_t, result->current); @@ -10206,12 +10201,14 @@ next_record: tuple. */ innobase_fts_create_doc_id_key(tuple, index, &search_doc_id); - innobase_srv_conc_enter_innodb(prebuilt->trx); + innobase_srv_conc_enter_innodb(m_prebuilt); dberr_t ret = row_search_for_mysql( - (byte*) buf, PAGE_CUR_GE, prebuilt, ROW_SEL_EXACT, 0); + (byte*) buf, PAGE_CUR_GE, m_prebuilt, ROW_SEL_EXACT, 0); - innobase_srv_conc_exit_innodb(prebuilt->trx); + innobase_srv_conc_exit_innodb(m_prebuilt); + + int error; switch (ret) { case DB_SUCCESS: @@ -10241,7 +10238,7 @@ next_record: case DB_TABLESPACE_DELETED: ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, ER_TABLESPACE_DISCARDED, table->s->table_name.str); @@ -10251,16 +10248,16 @@ next_record: case DB_TABLESPACE_NOT_FOUND: ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, ER_TABLESPACE_MISSING, table->s->table_name.str); table->status = STATUS_NOT_FOUND; - error = HA_ERR_NO_SUCH_TABLE; + error = HA_ERR_TABLESPACE_MISSING; break; default: error = convert_error_code_to_mysql( - ret, 0, user_thd); + ret, 0, m_user_thd); table->status = STATUS_NOT_FOUND; break; @@ -10275,6 +10272,7 @@ next_record: #ifdef WITH_WSREP extern dict_index_t* wsrep_dict_foreign_find_index( +/*==========================*/ dict_table_t* table, const char** col_names, const char** columns, @@ -10298,16 +10296,16 @@ wsrep_append_foreign_key( THD* thd = (THD*)trx->mysql_thd; ulint rcode = DB_SUCCESS; char cache_key[513] = {'\0'}; - int cache_key_len; - bool const copy = true; + int cache_key_len=0; + bool const copy = true; if (!wsrep_on(trx->mysql_thd) || - wsrep_thd_exec_mode(thd) != LOCAL_STATE) + wsrep_thd_exec_mode(thd) != LOCAL_STATE) { return DB_SUCCESS; + } if (!thd || !foreign || - (!foreign->referenced_table && !foreign->foreign_table)) - { + (!foreign->referenced_table && !foreign->foreign_table)) { WSREP_INFO("FK: %s missing in: %s", (!thd) ? "thread" : ((!foreign) ? "constraint" : @@ -10319,40 +10317,36 @@ wsrep_append_foreign_key( } if ( !((referenced) ? - foreign->referenced_table : foreign->foreign_table)) - { + foreign->referenced_table : foreign->foreign_table)) { WSREP_DEBUG("pulling %s table into cache", (referenced) ? "referenced" : "foreign"); mutex_enter(&(dict_sys->mutex)); - if (referenced) - { + + if (referenced) { foreign->referenced_table = dict_table_get_low( foreign->referenced_table_name_lookup); - if (foreign->referenced_table) - { + if (foreign->referenced_table) { foreign->referenced_index = wsrep_dict_foreign_find_index( foreign->referenced_table, NULL, foreign->referenced_col_names, - foreign->n_fields, + foreign->n_fields, foreign->foreign_index, TRUE, FALSE); } - } - else - { + } else { foreign->foreign_table = dict_table_get_low( foreign->foreign_table_name_lookup); - if (foreign->foreign_table) - { + + if (foreign->foreign_table) { foreign->foreign_index = wsrep_dict_foreign_find_index( foreign->foreign_table, NULL, foreign->foreign_col_names, foreign->n_fields, - foreign->referenced_index, + foreign->referenced_index, TRUE, FALSE); } } @@ -10360,8 +10354,7 @@ wsrep_append_foreign_key( } if ( !((referenced) ? - foreign->referenced_table : foreign->foreign_table)) - { + foreign->referenced_table : foreign->foreign_table)) { WSREP_WARN("FK: %s missing in query: %s", (!foreign->referenced_table) ? "referenced table" : "foreign table", @@ -10369,6 +10362,7 @@ wsrep_append_foreign_key( wsrep_thd_query(thd) : "void"); return DB_ERROR; } + byte key[WSREP_MAX_SUPPORTED_KEY_LENGTH+1] = {'\0'}; ulint len = WSREP_MAX_SUPPORTED_KEY_LENGTH; @@ -10378,36 +10372,41 @@ wsrep_append_foreign_key( UT_LIST_GET_FIRST(foreign->referenced_table->indexes) : UT_LIST_GET_FIRST(foreign->foreign_table->indexes); int i = 0; + while (idx != NULL && idx != idx_target) { if (innobase_strcasecmp (idx->name, innobase_index_reserve_name) != 0) { i++; } idx = UT_LIST_GET_NEXT(indexes, idx); } + ut_a(idx); key[0] = (char)i; rcode = wsrep_rec_get_foreign_key( &key[1], &len, rec, index, idx, wsrep_protocol_version > 1); + if (rcode != DB_SUCCESS) { WSREP_ERROR( - "FK key set failed: %lu (%lu %lu), index: %s %s, %s", + "FK key set failed: " ULINTPF + " (" ULINTPF " " ULINTPF "), index: %s %s, %s", rcode, referenced, shared, - (index && index->name) ? index->name : - "void index", - (index && index->table_name) ? index->table_name : + (index) ? index->name() : "void index", + (index && index->table) ? index->table->name.m_name : "void table", wsrep_thd_query(thd)); return DB_ERROR; } + strncpy(cache_key, (wsrep_protocol_version > 1) ? ((referenced) ? - foreign->referenced_table->name : - foreign->foreign_table->name) : - foreign->foreign_table->name, sizeof(cache_key) - 1); + foreign->referenced_table->name.m_name : + foreign->foreign_table->name.m_name) : + foreign->foreign_table->name.m_name, sizeof(cache_key) - 1); cache_key_len = strlen(cache_key); + #ifdef WSREP_DEBUG_PRINT ulint j; fprintf(stderr, "FK parent key, table: %s %s len: %lu ", @@ -10418,16 +10417,18 @@ wsrep_append_foreign_key( fprintf(stderr, "\n"); #endif char *p = strchr(cache_key, '/'); + if (p) { *p = '\0'; } else { WSREP_WARN("unexpected foreign key table %s %s", - foreign->referenced_table->name, - foreign->foreign_table->name); + foreign->referenced_table->name.m_name, + foreign->foreign_table->name.m_name); } wsrep_buf_t wkey_part[3]; wsrep_key_t wkey = {wkey_part, 3}; + if (!wsrep_prepare_key( (const uchar*)cache_key, cache_key_len + 1, @@ -10439,7 +10440,9 @@ wsrep_append_foreign_key( wsrep_thd_query(thd) : "void"); return DB_ERROR; } + wsrep_t *wsrep= get_wsrep(); + rcode = (int)wsrep->append_key( wsrep, wsrep_ws_handle(thd, trx), @@ -10447,9 +10450,11 @@ wsrep_append_foreign_key( 1, shared ? WSREP_KEY_SHARED : WSREP_KEY_EXCLUSIVE, copy); + if (rcode) { - DBUG_PRINT("wsrep", ("row key failed: %zu", rcode)); - WSREP_ERROR("Appending cascaded fk row key failed: %s, %lu", + DBUG_PRINT("wsrep", ("row key failed: " ULINTPF, rcode)); + WSREP_ERROR("Appending cascaded fk row key failed: %s, " + ULINTPF, (wsrep_thd_query(thd)) ? wsrep_thd_query(thd) : "void", rcode); return DB_ERROR; @@ -10460,7 +10465,7 @@ wsrep_append_foreign_key( static int wsrep_append_key( -/*==================*/ +/*=============*/ THD *thd, trx_t *trx, TABLE_SHARE *table_share, @@ -10484,6 +10489,7 @@ wsrep_append_key( #endif wsrep_buf_t wkey_part[3]; wsrep_key_t wkey = {wkey_part, 3}; + if (!wsrep_prepare_key( (const uchar*)table_share->table_cache_key.str, table_share->table_cache_key.length, @@ -10497,6 +10503,7 @@ wsrep_append_key( } wsrep_t *wsrep= get_wsrep(); + int rcode = (int)wsrep->append_key( wsrep, wsrep_ws_handle(thd, trx), @@ -10511,21 +10518,26 @@ wsrep_append_key( wsrep_thd_query(thd) : "void", rcode); DBUG_RETURN(HA_ERR_INTERNAL_ERROR); } + DBUG_RETURN(0); } static bool -referenced_by_foreign_key2(dict_table_t* table, - dict_index_t* index) { - ut_ad(table != NULL); - ut_ad(index != NULL); +referenced_by_foreign_key2( +/*=======================*/ + dict_table_t* table, + dict_index_t* index) +{ + ut_ad(table != NULL); + ut_ad(index != NULL); + + const dict_foreign_set* fks = &table->referenced_set; - const dict_foreign_set* fks = &table->referenced_set; - for (dict_foreign_set::const_iterator it = fks->begin(); + for (dict_foreign_set::const_iterator it = fks->begin(); it != fks->end(); - ++it) - { + ++it) { dict_foreign_t* foreign = *it; + if (foreign->referenced_index != index) { continue; } @@ -10537,7 +10549,7 @@ referenced_by_foreign_key2(dict_table_t* table, int ha_innobase::wsrep_append_keys( -/*==================*/ +/*===========================*/ THD *thd, bool shared, const uchar* record0, /* in: row in MySQL format */ @@ -10550,10 +10562,10 @@ ha_innobase::wsrep_append_keys( trx_t *trx = thd_to_trx(thd); if (table_share && table_share->tmp_table != NO_TMP_TABLE) { - WSREP_DEBUG("skipping tmp table DML: THD: %lu tmp: %d SQL: %s", + WSREP_DEBUG("skipping tmp table DML: THD: %lu tmp: %d SQL: %s", thd_get_thread_id(thd), table_share->tmp_table, - (wsrep_thd_query(thd)) ? + (wsrep_thd_query(thd)) ? wsrep_thd_query(thd) : "void"); DBUG_RETURN(0); } @@ -10570,13 +10582,14 @@ ha_innobase::wsrep_append_keys( if (!is_null) { rcode = wsrep_append_key( - thd, trx, table_share, table, keyval, + thd, trx, table_share, table, keyval, len, shared); - if (rcode) DBUG_RETURN(rcode); - } - else - { - WSREP_DEBUG("NULL key skipped (proto 0): %s", + + if (rcode) { + DBUG_RETURN(rcode); + } + } else { + WSREP_DEBUG("NULL key skipped (proto 0): %s", wsrep_thd_query(thd)); } } else { @@ -10607,8 +10620,8 @@ ha_innobase::wsrep_append_keys( keyval1[0] = (char)i; if (!tab) { - WSREP_WARN("MySQL-InnoDB key mismatch %s %s", - table->s->table_name.str, + WSREP_WARN("MariaDB-InnoDB key mismatch %s %s", + table->s->table_name.str, key_info->name); } /* !hasPK == table with no PK, must append all non-unique keys */ @@ -10618,32 +10631,35 @@ ha_innobase::wsrep_append_keys( (!tab && referenced_by_foreign_key()))) { len = wsrep_store_key_val_for_row( - thd, table, i, key0, - WSREP_MAX_SUPPORTED_KEY_LENGTH, + thd, table, i, key0, + WSREP_MAX_SUPPORTED_KEY_LENGTH, record0, &is_null); if (!is_null) { rcode = wsrep_append_key( - thd, trx, table_share, table, + thd, trx, table_share, table, keyval0, len+1, shared); - if (rcode) DBUG_RETURN(rcode); + + if (rcode) { + DBUG_RETURN(rcode); + } if (key_info->flags & HA_NOSAME || shared) key_appended = true; - } - else - { - WSREP_DEBUG("NULL key skipped: %s", + } else { + WSREP_DEBUG("NULL key skipped: %s", wsrep_thd_query(thd)); } + if (record1) { len = wsrep_store_key_val_for_row( - thd, table, i, key1, + thd, table, i, key1, WSREP_MAX_SUPPORTED_KEY_LENGTH, record1, &is_null); + if (!is_null && memcmp(key0, key1, len)) { rcode = wsrep_append_key( - thd, trx, table_share, - table, + thd, trx, table_share, + table, keyval1, len+1, shared); if (rcode) DBUG_RETURN(rcode); } @@ -10657,19 +10673,20 @@ ha_innobase::wsrep_append_keys( uchar digest[16]; int rcode; - wsrep_calc_row_hash(digest, record0, table, prebuilt, thd); - if ((rcode = wsrep_append_key(thd, trx, table_share, table, - (const char*) digest, 16, + wsrep_calc_row_hash(digest, record0, table, m_prebuilt, thd); + + if ((rcode = wsrep_append_key(thd, trx, table_share, table, + (const char*) digest, 16, shared))) { DBUG_RETURN(rcode); } if (record1) { wsrep_calc_row_hash( - digest, record1, table, prebuilt, thd); - if ((rcode = wsrep_append_key(thd, trx, table_share, + digest, record1, table, m_prebuilt, thd); + if ((rcode = wsrep_append_key(thd, trx, table_share, table, - (const char*) digest, + (const char*) digest, 16, shared))) { DBUG_RETURN(rcode); } @@ -10689,7 +10706,7 @@ is the current 'position' of the handle, because if row ref is actually the row id internally generated in InnoDB, then 'record' does not contain it. We just guess that the row id must be for the record where the handle was positioned the last time. */ -UNIV_INTERN + void ha_innobase::position( /*==================*/ @@ -10697,9 +10714,9 @@ ha_innobase::position( { uint len; - ut_a(prebuilt->trx == thd_to_trx(ha_thd())); + ut_a(m_prebuilt->trx == thd_to_trx(ha_thd())); - if (prebuilt->clust_index_was_generated) { + if (m_prebuilt->clust_index_was_generated) { /* No primary key was defined for the table and we generated the clustered index from row id: the row reference will be the row id, not any key value @@ -10707,19 +10724,16 @@ ha_innobase::position( len = DATA_ROW_ID_LEN; - memcpy(ref, prebuilt->row_id, len); + memcpy(ref, m_prebuilt->row_id, len); } else { - len = store_key_val_for_row(primary_key, (char*) ref, - ref_length, record); - } - - /* We assume that the 'ref' value len is always fixed for the same - table. */ - if (len != ref_length) { - sql_print_error("Stored ref len is %lu, but table ref len is " - "%lu", (ulong) len, (ulong) ref_length); + /* Copy primary key as the row reference */ + KEY* key_info = table->key_info + m_primary_key; + key_copy(ref, (uchar*)record, key_info, key_info->key_length); + len = key_info->key_length; } + + ut_ad(len == ref_length); } /*****************************************************************//** @@ -10738,16 +10752,21 @@ create_table_check_doc_id_col( ULINT_UNDEFINED if column is of the wrong type/name/size */ { - for (ulint i = 0; i < form->s->fields; i++) { + uint n_fields = mysql_fields(form); + + for (ulint i = 0; i < n_fields; i++) { const Field* field; ulint col_type; ulint col_len; ulint unsigned_type; field = form->field[i]; + if (!field->stored_in_db()) { + continue; + } - col_type = get_innobase_type_from_mysql_type(&unsigned_type, - field); + col_type = get_innobase_type_from_mysql_type( + &unsigned_type, field); col_len = field->pack_length(); @@ -10767,9 +10786,9 @@ create_table_check_doc_id_col( trx->mysql_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: FTS_DOC_ID column must be " - "of BIGINT NOT NULL type, and named " - "in all capitalized characters"); + "InnoDB: FTS_DOC_ID column must be" + " of BIGINT NOT NULL type, and named" + " in all capitalized characters"); my_error(ER_WRONG_COLUMN_NAME, MYF(0), field->field_name); *doc_id_col = ULINT_UNDEFINED; @@ -10782,34 +10801,133 @@ create_table_check_doc_id_col( return(false); } -/*****************************************************************//** -Creates a table definition to an InnoDB database. */ -static MY_ATTRIBUTE((nonnull, warn_unused_result)) + +/** Finds all base columns needed to compute a given generated column. +This is returned as a bitmap, in field->table->tmp_set. +Works for both dict_v_col_t and dict_s_col_t columns. +@param[in] table InnoDB table +@param[in] field MySQL field +@param[in,out] col virtual or stored column */ +template <typename T> +void +prepare_vcol_for_base_setup( +/*========================*/ + const dict_table_t* table, + const Field* field, + T* col) +{ + ut_ad(col->num_base == 0); + ut_ad(col->base_col == NULL); + + MY_BITMAP *old_read_set = field->table->read_set; + MY_BITMAP *old_vcol_set = field->table->vcol_set; + + field->table->read_set = field->table->vcol_set = &field->table->tmp_set; + + bitmap_clear_all(&field->table->tmp_set); + field->vcol_info->expr->walk( + &Item::register_field_in_read_map, 1, field->table); + col->num_base= bitmap_bits_set(&field->table->tmp_set); + if (col->num_base != 0) { + col->base_col = static_cast<dict_col_t**>(mem_heap_zalloc( + table->heap, col->num_base * sizeof( + * col->base_col))); + } + field->table->read_set= old_read_set; + field->table->vcol_set= old_vcol_set; +} + + +/** Set up base columns for virtual column +@param[in] table InnoDB table +@param[in] field MySQL field +@param[in,out] v_col virtual column */ +void +innodb_base_col_setup( + dict_table_t* table, + const Field* field, + dict_v_col_t* v_col) +{ + int n = 0; + + prepare_vcol_for_base_setup(table, field, v_col); + + for (uint i= 0; i < field->table->s->fields; ++i) { + const Field* base_field = field->table->field[i]; + if (base_field->stored_in_db() + && bitmap_is_set(&field->table->tmp_set, i)) { + ulint z; + + for (z = 0; z < table->n_cols; z++) { + const char* name = dict_table_get_col_name(table, z); + if (!innobase_strcasecmp(name, + base_field->field_name)) { + break; + } + } + + ut_ad(z != table->n_cols); + + v_col->base_col[n] = dict_table_get_nth_col(table, z); + ut_ad(v_col->base_col[n]->ind == z); + n++; + } + } + v_col->num_base= n; +} + +/** Set up base columns for stored column +@param[in] table InnoDB table +@param[in] field MySQL field +@param[in,out] s_col stored column */ +void +innodb_base_col_setup_for_stored( + const dict_table_t* table, + const Field* field, + dict_s_col_t* s_col) +{ + ulint n = 0; + + prepare_vcol_for_base_setup(table, field, s_col); + + for (uint i= 0; i < field->table->s->fields; ++i) { + const Field* base_field = field->table->field[i]; + + if (!innobase_is_s_fld(base_field) + && !innobase_is_v_fld(base_field) + && bitmap_is_set(&field->table->tmp_set, + i)) { + ulint z; + for (z = 0; z < table->n_cols; z++) { + const char* name = dict_table_get_col_name( + table, z); + if (!innobase_strcasecmp( + name, base_field->field_name)) { + break; + } + } + + ut_ad(z != table->n_cols); + + s_col->base_col[n] = dict_table_get_nth_col(table, z); + n++; + + if (n == s_col->num_base) { + break; + } + } + } + s_col->num_base= n; +} + +/** Create a table definition to an InnoDB database. +@return ER_* level error */ +inline MY_ATTRIBUTE((warn_unused_result)) int -create_table_def( -/*=============*/ - trx_t* trx, /*!< in: InnoDB transaction handle */ - const TABLE* form, /*!< in: information on table - columns and indexes */ - const char* table_name, /*!< in: table name */ - const char* temp_path, /*!< in: if this is a table explicitly - created by the user with the - TEMPORARY keyword, then this - parameter is the dir path where the - table should be placed if we create - an .ibd file for it (no .ibd extension - in the path, though). Otherwise this - is a zero length-string */ - const char* remote_path, /*!< in: Remote path or zero length-string */ - ulint flags, /*!< in: table flags */ - ulint flags2, /*!< in: table flags2 */ - fil_encryption_t mode, /*!< in: encryption mode */ - ulint key_id) /*!< in: encryption key_id */ -{ - THD* thd = trx->mysql_thd; +create_table_info_t::create_table_def() +{ dict_table_t* table; ulint n_cols; - dberr_t err; ulint col_type; ulint col_len; ulint nulls_allowed; @@ -10818,44 +10936,58 @@ create_table_def( ulint long_true_varchar; ulint charset_no; ulint i; + ulint j = 0; ulint doc_id_col = 0; ibool has_doc_id_col = FALSE; mem_heap_t* heap; + ulint num_v = 0; + ulint space_id = 0; + ulint actual_n_cols; + ha_table_option_struct *options= m_form->s->option_struct; + dberr_t err = DB_SUCCESS; DBUG_ENTER("create_table_def"); - DBUG_PRINT("enter", ("table_name: %s", table_name)); + DBUG_PRINT("enter", ("table_name: %s", m_table_name)); - DBUG_ASSERT(thd != NULL); + DBUG_ASSERT(m_trx->mysql_thd == m_thd); /* MySQL does the name length check. But we do additional check on the name length here */ - const size_t table_name_len = strlen(table_name); + const size_t table_name_len = strlen(m_table_name); if (table_name_len > MAX_FULL_NAME_LEN) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_TABLE_NAME, "InnoDB: Table Name or Database Name is too long"); DBUG_RETURN(ER_TABLE_NAME); } - if (table_name[table_name_len - 1] == '/') { + if (m_table_name[table_name_len - 1] == '/') { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_TABLE_NAME, "InnoDB: Table name is empty"); DBUG_RETURN(ER_WRONG_TABLE_NAME); } - n_cols = form->s->fields; + n_cols = m_form->s->fields; + + /* Find out any virtual column */ + for (i = 0; i < n_cols; i++) { + Field* field = m_form->field[i]; + + if (innobase_is_v_fld(field)) { + num_v++; + } + } /* Check whether there already exists a FTS_DOC_ID column */ - if (create_table_check_doc_id_col(trx, form, &doc_id_col)){ + if (create_table_check_doc_id_col(m_trx, m_form, &doc_id_col)){ /* Raise error if the Doc ID column is of wrong type or name */ if (doc_id_col == ULINT_UNDEFINED) { - trx_commit_for_mysql(trx); err = DB_ERROR; goto error_ret; @@ -10864,59 +10996,55 @@ create_table_def( } } - /* We pass 0 as the space id, and determine at a lower level the space - id where to store the table */ + /* For single-table tablespaces, we pass 0 as the space id, and then + determine the actual space id when the tablespace is created. */ - if (flags2 & DICT_TF2_FTS) { - /* Adjust for the FTS hidden field */ - if (!has_doc_id_col) { - table = dict_mem_table_create(table_name, 0, form->s->stored_fields + 1, - flags, flags2); - - /* Set the hidden doc_id column. */ - table->fts->doc_col = form->s->stored_fields; - } else { - table = dict_mem_table_create(table_name, 0, form->s->stored_fields, - flags, flags2); - table->fts->doc_col = doc_id_col; - } - } else { - table = dict_mem_table_create(table_name, 0, form->s->stored_fields, - flags, flags2); + /* Adjust the number of columns for the FTS hidden field */ + actual_n_cols = n_cols; + if (m_flags2 & DICT_TF2_FTS && !has_doc_id_col) { + actual_n_cols += 1; } - if (flags2 & DICT_TF2_TEMPORARY) { - ut_a(strlen(temp_path)); - table->dir_path_of_temp_table = - mem_heap_strdup(table->heap, temp_path); + table = dict_mem_table_create(m_table_name, space_id, + actual_n_cols, num_v, m_flags, m_flags2); + + /* Set the hidden doc_id column. */ + if (m_flags2 & DICT_TF2_FTS) { + table->fts->doc_col = has_doc_id_col + ? doc_id_col : n_cols - num_v; } - if (DICT_TF_HAS_DATA_DIR(flags)) { - ut_a(strlen(remote_path)); - table->data_dir_path = mem_heap_strdup(table->heap, remote_path); + if (DICT_TF_HAS_DATA_DIR(m_flags)) { + ut_a(strlen(m_remote_path)); + + table->data_dir_path = mem_heap_strdup( + table->heap, m_remote_path); + } else { table->data_dir_path = NULL; } + heap = mem_heap_create(1000); for (i = 0; i < n_cols; i++) { - Field* field = form->field[i]; - if (!field->stored_in_db) - continue; + ulint is_virtual; + bool is_stored = false; + + Field* field = m_form->field[i]; - col_type = get_innobase_type_from_mysql_type(&unsigned_type, - field); + col_type = get_innobase_type_from_mysql_type( + &unsigned_type, field); if (!col_type) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_CANT_CREATE_TABLE, - "Error creating table '%s' with " - "column '%s'. Please check its " - "column type and try to re-create " - "the table with an appropriate " - "column type.", - table->name, field->field_name); + "Error creating table '%s' with" + " column '%s'. Please check its" + " column type and try to re-create" + " the table with an appropriate" + " column type.", + table->name.m_name, field->field_name); goto err_col; } @@ -10929,24 +11057,27 @@ create_table_def( charset_no = (ulint) field->charset()->number; - if (UNIV_UNLIKELY(charset_no > MAX_CHAR_COLL_NUM)) { + DBUG_EXECUTE_IF("simulate_max_char_col", + charset_no = MAX_CHAR_COLL_NUM + 1; + ); + + if (charset_no > MAX_CHAR_COLL_NUM) { /* in data0type.h we assume that the number fits in one byte in prtype */ push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_CANT_CREATE_TABLE, "In InnoDB, charset-collation codes" " must be below 256." - " Unsupported code %lu.", - (ulong) charset_no); + " Unsupported code " ULINTPF ".", + charset_no); mem_heap_free(heap); + dict_mem_table_free(table); + DBUG_RETURN(ER_CANT_CREATE_TABLE); } } - /* we assume in dtype_form_prtype() that this fits in - two bytes */ - ut_a(static_cast<uint>(field->type()) <= MAX_CHAR_COLL_NUM); col_len = field->pack_length(); /* The MySQL pack length contains 1 or 2 bytes length field @@ -10964,6 +11095,9 @@ create_table_def( } } + is_virtual = (innobase_is_v_fld(field)) ? DATA_VIRTUAL : 0; + is_stored = innobase_is_s_fld(field); + /* First check whether the column to be added has a system reserved name. */ if (dict_col_name_is_reserved(field->field_name)){ @@ -10972,29 +11106,117 @@ create_table_def( err_col: dict_mem_table_free(table); mem_heap_free(heap); - trx_commit_for_mysql(trx); err = DB_ERROR; goto error_ret; } - dict_mem_table_add_col(table, heap, - field->field_name, - col_type, - dtype_form_prtype( - (ulint) field->type() - | nulls_allowed | unsigned_type - | binary_type | long_true_varchar, - charset_no), - col_len); + if (!is_virtual) { + dict_mem_table_add_col(table, heap, + field->field_name, col_type, + dtype_form_prtype( + (ulint) field->type() + | nulls_allowed | unsigned_type + | binary_type | long_true_varchar, + charset_no), + col_len); + } else { + dict_mem_table_add_v_col(table, heap, + field->field_name, col_type, + dtype_form_prtype( + (ulint) field->type() + | nulls_allowed | unsigned_type + | binary_type | long_true_varchar + | is_virtual, + charset_no), + col_len, i, 0); + } + + if (is_stored) { + ut_ad(!is_virtual); + /* Added stored column in m_s_cols list. */ + dict_mem_table_add_s_col( + table, 0); + } + } + + if (num_v) { + for (i = 0; i < n_cols; i++) { + dict_v_col_t* v_col; + + Field* field = m_form->field[i]; + + if (!innobase_is_v_fld(field)) { + continue; + } + + v_col = dict_table_get_nth_v_col(table, j); + + j++; + + innodb_base_col_setup(table, field, v_col); + } + } + + /** Fill base columns for the stored column present in the list. */ + if (table->s_cols && table->s_cols->size()) { + + for (i = 0; i < n_cols; i++) { + Field* field = m_form->field[i]; + + if (!innobase_is_s_fld(field)) { + continue; + } + + dict_s_col_list::iterator it; + for (it = table->s_cols->begin(); + it != table->s_cols->end(); ++it) { + dict_s_col_t s_col = *it; + + if (s_col.s_pos == i) { + innodb_base_col_setup_for_stored( + table, field, &s_col); + break; + } + } + } } /* Add the FTS doc_id hidden column. */ - if (flags2 & DICT_TF2_FTS && !has_doc_id_col) { + if (m_flags2 & DICT_TF2_FTS && !has_doc_id_col) { fts_add_doc_id_column(table, heap); } - err = row_create_table_for_mysql(table, trx, false, mode, key_id); + /* If temp table, then we avoid creation of entries in SYSTEM TABLES. + Given that temp table lifetime is limited to connection/server lifetime + on re-start we don't need to restore temp-table and so no entry is + needed in SYSTEM tables. */ + if (dict_table_is_temporary(table)) { + /* Get a new table ID */ + dict_table_assign_new_id(table, m_trx); + table->space = SRV_TMP_SPACE_ID; + + /* Temp-table are maintained in memory and so + can_be_evicted is FALSE. */ + mem_heap_t* temp_table_heap = mem_heap_create(256); + + dict_table_add_to_cache(table, FALSE, temp_table_heap); + + DBUG_EXECUTE_IF("ib_ddl_crash_during_create2", + DBUG_SUICIDE();); + + mem_heap_free(temp_table_heap); + } else { + if (err == DB_SUCCESS) { + err = row_create_table_for_mysql( + table, m_trx, + (fil_encryption_t)options->encryption, + (uint32_t)options->encryption_key_id); + } + + DBUG_EXECUTE_IF("ib_crash_during_create_for_encryption", + DBUG_SUICIDE();); + } mem_heap_free(heap); @@ -11005,8 +11227,8 @@ err_col: char display_name[FN_REFLEN]; char* buf_end = innobase_convert_identifier( display_name, sizeof(display_name) - 1, - table_name, strlen(table_name), - thd, TRUE); + m_table_name, strlen(m_table_name), + m_thd); *buf_end = '\0'; @@ -11015,17 +11237,17 @@ err_col: : ER_TABLESPACE_EXISTS, MYF(0), display_name); } - if (err == DB_SUCCESS && (flags2 & DICT_TF2_FTS)) { + if (err == DB_SUCCESS && (m_flags2 & DICT_TF2_FTS)) { fts_optimize_add_table(table); } error_ret: - DBUG_RETURN(convert_error_code_to_mysql(err, flags, thd)); + DBUG_RETURN(convert_error_code_to_mysql(err, m_flags, m_thd)); } /*****************************************************************//** Creates an index in an InnoDB database. */ -static +inline int create_index( /*=========*/ @@ -11049,13 +11271,29 @@ create_index( /* Assert that "GEN_CLUST_INDEX" cannot be used as non-primary index */ ut_a(innobase_strcasecmp(key->name, innobase_index_reserve_name) != 0); - if (key->flags & HA_FULLTEXT) { + ind_type = 0; + if (key->flags & HA_SPATIAL) { + ind_type = DICT_SPATIAL; + } else if (key->flags & HA_FULLTEXT) { + ind_type = DICT_FTS; + } + + if (ind_type != 0) + { index = dict_mem_index_create(table_name, key->name, 0, - DICT_FTS, + ind_type, key->user_defined_key_parts); for (ulint i = 0; i < key->user_defined_key_parts; i++) { KEY_PART_INFO* key_part = key->key_part + i; + + /* We do not support special (Fulltext or Spatial) + index on virtual columns */ + if (innobase_is_v_fld(key_part->field)) { + ut_ad(0); + DBUG_RETURN(HA_ERR_UNSUPPORTED); + } + dict_mem_index_add_field( index, key_part->field->field_name, 0); } @@ -11077,7 +11315,7 @@ create_index( ind_type |= DICT_UNIQUE; } - field_lengths = (ulint*) my_malloc( + field_lengths = (ulint*) my_malloc(//PSI_INSTRUMENT_ME, key->user_defined_key_parts * sizeof * field_lengths, MYF(MY_FAE)); @@ -11099,29 +11337,22 @@ create_index( specified number of first bytes of the column to the index field.) The flag does not seem to be properly set by MySQL. Let us fall back on testing - the length of the key part versus the column. */ + the length of the key part versus the column. + We first reach to the table's column; if the index is on a + prefix, key_part->field is not the table's column (it's a + "fake" field forged in open_table_from_share() with length + equal to the length of the prefix); so we have to go to + form->fied. */ + Field* field= form->field[key_part->field->field_index]; + if (field == NULL) + ut_error; + + const char* field_name = key_part->field->field_name; - Field* field = NULL; - - for (ulint j = 0; j < form->s->fields; j++) { - - field = form->field[j]; - - if (0 == innobase_strcasecmp( - field->field_name, - key_part->field->field_name)) { - /* Found the corresponding column */ - - goto found; - } - } - - ut_error; -found: col_type = get_innobase_type_from_mysql_type( &is_unsigned, key_part->field); - if (DATA_BLOB == col_type + if (DATA_LARGE_MTYPE(col_type) || (key_part->length < field->pack_length() && field->type() != MYSQL_TYPE_VARCHAR) || (field->type() == MYSQL_TYPE_VARCHAR @@ -11137,10 +11368,10 @@ found: case DATA_DOUBLE: case DATA_DECIMAL: sql_print_error( - "MySQL is trying to create a column " - "prefix index field, on an " - "inappropriate data type. Table " - "name %s, column name %s.", + "MariaDB is trying to create a column" + " prefix index field, on an" + " inappropriate data type. Table" + " name %s, column name %s.", table_name, key_part->field->field_name); @@ -11152,8 +11383,11 @@ found: field_lengths[i] = key_part->length; - dict_mem_index_add_field( - index, key_part->field->field_name, prefix_len); + if (innobase_is_v_fld(key_part->field)) { + index->type |= DICT_VIRTUAL; + } + + dict_mem_index_add_field(index, field_name, prefix_len); } ut_ad(key->flags & HA_FULLTEXT || !(index->type & DICT_FTS)); @@ -11174,7 +11408,7 @@ found: /*****************************************************************//** Creates an index to an InnoDB table when the user has defined no primary index. */ -static +inline int create_clustered_index_when_no_primary( /*===================================*/ @@ -11196,14 +11430,13 @@ create_clustered_index_when_no_primary( return(convert_error_code_to_mysql(error, flags, NULL)); } -/*****************************************************************//** -Return a display name for the row format +/** Return a display name for the row format +@param[in] row_format Row Format @return row format name */ -UNIV_INTERN +static const char* get_row_format_name( -/*================*/ - enum row_type row_format) /*!< in: Row Format */ + enum row_type row_format) { switch (row_format) { case ROW_TYPE_COMPACT: @@ -11225,75 +11458,103 @@ get_row_format_name( return("NOT USED"); } -/** If file-per-table is missing, issue warning and set ret false */ -#define CHECK_ERROR_ROW_TYPE_NEEDS_FILE_PER_TABLE(use_tablespace)\ - if (!use_tablespace) { \ - push_warning_printf( \ - thd, Sql_condition::WARN_LEVEL_WARN, \ - ER_ILLEGAL_HA_CREATE_OPTION, \ - "InnoDB: ROW_FORMAT=%s requires" \ - " innodb_file_per_table.", \ - get_row_format_name(row_format)); \ - ret = "ROW_FORMAT"; \ +/** Validate DATA DIRECTORY option. +@return true if valid, false if not. */ +bool +create_table_info_t::create_option_data_directory_is_valid() +{ + bool is_valid = true; + + ut_ad(m_create_info->data_file_name + && m_create_info->data_file_name[0] != '\0'); + + /* Use DATA DIRECTORY only with file-per-table. */ + if (!m_allow_file_per_table) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: DATA DIRECTORY requires" + " innodb_file_per_table."); + is_valid = false; } -/** If file-format is Antelope, issue warning and set ret false */ -#define CHECK_ERROR_ROW_TYPE_NEEDS_GT_ANTELOPE \ - if (srv_file_format < UNIV_FORMAT_B) { \ - push_warning_printf( \ - thd, Sql_condition::WARN_LEVEL_WARN, \ - ER_ILLEGAL_HA_CREATE_OPTION, \ - "InnoDB: ROW_FORMAT=%s requires" \ - " innodb_file_format > Antelope.", \ - get_row_format_name(row_format)); \ - ret = "ROW_FORMAT"; \ + /* Do not use DATA DIRECTORY with TEMPORARY TABLE. */ + if (m_create_info->options & HA_LEX_CREATE_TMP_TABLE) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: DATA DIRECTORY cannot be used" + " for TEMPORARY tables."); + is_valid = false; } + /* We check for a DATA DIRECTORY mixed with TABLESPACE in + create_option_tablespace_is_valid(), no need to here. */ -/*****************************************************************//** -Validates the create options. We may build on this function -in future. For now, it checks two specifiers: -KEY_BLOCK_SIZE and ROW_FORMAT -If innodb_strict_mode is not set then this function is a no-op -@return NULL if valid, string if not. */ -UNIV_INTERN + return(is_valid); +} + +/** Validate the create options. Check that the options KEY_BLOCK_SIZE, +ROW_FORMAT, DATA DIRECTORY, TEMPORARY are compatible with +each other and other settings. These CREATE OPTIONS are not validated +here unless innodb_strict_mode is on. With strict mode, this function +will report each problem it finds using a custom message with error +code ER_ILLEGAL_HA_CREATE_OPTION, not its built-in message. +@return NULL if valid, string name of bad option if not. */ const char* -create_options_are_invalid( -/*=======================*/ - THD* thd, /*!< in: connection thread. */ - TABLE* form, /*!< in: information on table - columns and indexes */ - HA_CREATE_INFO* create_info, /*!< in: create info. */ - bool use_tablespace) /*!< in: srv_file_per_table */ +create_table_info_t::create_options_are_invalid() { - ibool kbs_specified = FALSE; - const char* ret = NULL; - enum row_type row_format = form->s->row_type; + bool has_key_block_size = (m_create_info->key_block_size != 0); - ut_ad(thd != NULL); + const char* ret = NULL; + enum row_type row_format = m_create_info->row_type; + const bool is_temp + = m_create_info->options & HA_LEX_CREATE_TMP_TABLE; + + ut_ad(m_thd != NULL); - /* If innodb_strict_mode is not set don't do any validation. */ - if (!(THDVAR(thd, strict_mode))) { + /* If innodb_strict_mode is not set don't do any more validation. */ + if (!THDVAR(m_thd, strict_mode)) { return(NULL); } - ut_ad(form != NULL); - ut_ad(create_info != NULL); + /* Check if a non-zero KEY_BLOCK_SIZE was specified. */ + if (has_key_block_size) { + if (is_temp) { + my_error(ER_UNSUPPORT_COMPRESSED_TEMPORARY_TABLE, + MYF(0)); + return("KEY_BLOCK_SIZE"); + } - /* First check if a non-zero KEY_BLOCK_SIZE was specified. */ - if (create_info->key_block_size) { - kbs_specified = TRUE; - switch (create_info->key_block_size) { + switch (m_create_info->key_block_size) { ulint kbs_max; case 1: case 2: case 4: case 8: case 16: + /* The maximum KEY_BLOCK_SIZE (KBS) is + UNIV_PAGE_SIZE_MAX. But if UNIV_PAGE_SIZE is + smaller than UNIV_PAGE_SIZE_MAX, the maximum + KBS is also smaller. */ + kbs_max = ut_min( + 1 << (UNIV_PAGE_SSIZE_MAX - 1), + 1 << (PAGE_ZIP_SSIZE_MAX - 1)); + if (m_create_info->key_block_size > kbs_max) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: KEY_BLOCK_SIZE=%ld" + " cannot be larger than %ld.", + m_create_info->key_block_size, + kbs_max); + ret = "KEY_BLOCK_SIZE"; + } + /* Valid KEY_BLOCK_SIZE, check its dependencies. */ - if (!use_tablespace) { + if (!m_allow_file_per_table) { push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: KEY_BLOCK_SIZE requires" " innodb_file_per_table."); @@ -11301,57 +11562,70 @@ create_options_are_invalid( } if (srv_file_format < UNIV_FORMAT_B) { push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: KEY_BLOCK_SIZE requires" " innodb_file_format > Antelope."); ret = "KEY_BLOCK_SIZE"; } - - /* The maximum KEY_BLOCK_SIZE (KBS) is 16. But if - UNIV_PAGE_SIZE is smaller than 16k, the maximum - KBS is also smaller. */ - kbs_max = ut_min( - 1 << (UNIV_PAGE_SSIZE_MAX - 1), - 1 << (PAGE_ZIP_SSIZE_MAX - 1)); - if (create_info->key_block_size > kbs_max) { - push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: KEY_BLOCK_SIZE=%ld" - " cannot be larger than %ld.", - create_info->key_block_size, - kbs_max); - ret = "KEY_BLOCK_SIZE"; - } break; default: push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: invalid KEY_BLOCK_SIZE = %lu." + "InnoDB: invalid KEY_BLOCK_SIZE = %u." " Valid values are [1, 2, 4, 8, 16]", - create_info->key_block_size); + (uint) m_create_info->key_block_size); ret = "KEY_BLOCK_SIZE"; break; } } - /* Check for a valid Innodb ROW_FORMAT specifier and + /* Check for a valid InnoDB ROW_FORMAT specifier and other incompatibilities. */ switch (row_format) { case ROW_TYPE_COMPRESSED: - CHECK_ERROR_ROW_TYPE_NEEDS_FILE_PER_TABLE(use_tablespace); - CHECK_ERROR_ROW_TYPE_NEEDS_GT_ANTELOPE; + if (is_temp) { + my_error(ER_UNSUPPORT_COMPRESSED_TEMPORARY_TABLE, + MYF(0)); + return("ROW_FORMAT"); + } + if (!m_allow_file_per_table) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: ROW_FORMAT=%s requires" + " innodb_file_per_table.", + get_row_format_name(row_format)); + ret = "ROW_FORMAT"; + } + if (srv_file_format < UNIV_FORMAT_B) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: ROW_FORMAT=%s requires" + " innodb_file_format > Antelope.", + get_row_format_name(row_format)); + ret = "ROW_FORMAT"; + } break; case ROW_TYPE_DYNAMIC: + if (!is_temp && srv_file_format < UNIV_FORMAT_B) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: ROW_FORMAT=%s requires" + " innodb_file_format > Antelope.", + get_row_format_name(row_format)); + ret = "ROW_FORMAT"; + } /* ROW_FORMAT=DYNAMIC also shuns KEY_BLOCK_SIZE */ /* fall through */ case ROW_TYPE_COMPACT: case ROW_TYPE_REDUNDANT: - if (kbs_specified) { + if (has_key_block_size) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: cannot specify ROW_FORMAT = %s" " with KEY_BLOCK_SIZE.", @@ -11365,52 +11639,37 @@ create_options_are_invalid( case ROW_TYPE_PAGE: case ROW_TYPE_NOT_USED: push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, \ + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: invalid ROW_FORMAT specifier."); ret = "ROW_TYPE"; break; } - /* Use DATA DIRECTORY only with file-per-table. */ - if (create_info->data_file_name && !use_tablespace) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: DATA DIRECTORY requires" - " innodb_file_per_table."); - ret = "DATA DIRECTORY"; - } - - /* Do not use DATA DIRECTORY with TEMPORARY TABLE. */ - if (create_info->data_file_name - && create_info->options & HA_LEX_CREATE_TMP_TABLE) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: DATA DIRECTORY cannot be used" - " for TEMPORARY tables."); + if (m_create_info->data_file_name + && m_create_info->data_file_name[0] != '\0' + && !create_option_data_directory_is_valid()) { ret = "DATA DIRECTORY"; } /* Do not allow INDEX_DIRECTORY */ - if (create_info->index_file_name) { + if (m_create_info->index_file_name) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: INDEX DIRECTORY is not supported"); ret = "INDEX DIRECTORY"; } - if ((kbs_specified || row_format == ROW_TYPE_COMPRESSED) - && UNIV_PAGE_SIZE > (1<<14)) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: Cannot create a COMPRESSED table" - " when innodb_page_size > 16k."); - - if (kbs_specified) { + /* Don't support compressed table when page size > 16k. */ + if ((has_key_block_size || row_format == ROW_TYPE_COMPRESSED) + && UNIV_PAGE_SIZE > UNIV_PAGE_SIZE_DEF) { + push_warning(m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: Cannot create a COMPRESSED table" + " when innodb_page_size > 16k."); + + if (has_key_block_size) { ret = "KEY_BLOCK_SIZE"; } else { ret = "ROW_TYPE"; @@ -11421,30 +11680,189 @@ create_options_are_invalid( } /*****************************************************************//** +Check engine specific table options not handled by SQL-parser. +@return NULL if valid, string if not */ +const char* +create_table_info_t::check_table_options() +{ + enum row_type row_format = m_form->s->row_type; + ha_table_option_struct *options= m_form->s->option_struct; + fil_encryption_t encrypt = (fil_encryption_t)options->encryption; + bool should_encrypt = (encrypt == FIL_ENCRYPTION_ON); + + /* Currently we do not support encryption for + spatial indexes thus do not allow creating table with forced + encryption */ + for(ulint i = 0; i < m_form->s->keys; i++) { + const KEY* key = m_form->key_info + i; + if (key->flags & HA_SPATIAL && should_encrypt) { + push_warning_printf(m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_UNSUPPORTED, + "InnoDB: ENCRYPTED=ON not supported for table because " + "it contains spatial index."); + return "ENCRYPTED"; + } + } + + if (encrypt != FIL_ENCRYPTION_DEFAULT && !m_allow_file_per_table) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: ENCRYPTED requires innodb_file_per_table"); + return "ENCRYPTED"; + } + + if (encrypt == FIL_ENCRYPTION_OFF && srv_encrypt_tables == 2) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: ENCRYPTED=OFF cannot be used when innodb_encrypt_tables=FORCE"); + return "ENCRYPTED"; + } + + /* Check page compression requirements */ + if (options->page_compressed) { + + if (row_format == ROW_TYPE_COMPRESSED) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: PAGE_COMPRESSED table can't have" + " ROW_TYPE=COMPRESSED"); + return "PAGE_COMPRESSED"; + } + + if (row_format == ROW_TYPE_REDUNDANT) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: PAGE_COMPRESSED table can't have" + " ROW_TYPE=REDUNDANT"); + return "PAGE_COMPRESSED"; + } + + if (!m_allow_file_per_table) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: PAGE_COMPRESSED requires" + " innodb_file_per_table."); + return "PAGE_COMPRESSED"; + } + + if (srv_file_format < UNIV_FORMAT_B) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: PAGE_COMPRESSED requires" + " innodb_file_format > Antelope."); + return "PAGE_COMPRESSED"; + } + + if (m_create_info->key_block_size) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: PAGE_COMPRESSED table can't have" + " key_block_size"); + return "PAGE_COMPRESSED"; + } + } + + /* Check page compression level requirements, some of them are + already checked above */ + if (options->page_compression_level != 0) { + if (options->page_compressed == false) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: PAGE_COMPRESSION_LEVEL requires" + " PAGE_COMPRESSED"); + return "PAGE_COMPRESSION_LEVEL"; + } + + if (options->page_compression_level < 1 || options->page_compression_level > 9) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: invalid PAGE_COMPRESSION_LEVEL = %lu." + " Valid values are [1, 2, 3, 4, 5, 6, 7, 8, 9]", + options->page_compression_level); + return "PAGE_COMPRESSION_LEVEL"; + } + } + + /* If encryption is set up make sure that used key_id is found */ + if (encrypt == FIL_ENCRYPTION_ON || + (encrypt == FIL_ENCRYPTION_DEFAULT && srv_encrypt_tables)) { + if (!encryption_key_id_exists((unsigned int)options->encryption_key_id)) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: ENCRYPTION_KEY_ID %u not available", + (uint)options->encryption_key_id + ); + return "ENCRYPTION_KEY_ID"; + } + } + + /* Ignore nondefault key_id if encryption is set off */ + if (encrypt == FIL_ENCRYPTION_OFF && + options->encryption_key_id != THDVAR(m_thd, default_encryption_key_id)) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: Ignored ENCRYPTION_KEY_ID %u when encryption is disabled", + (uint)options->encryption_key_id + ); + options->encryption_key_id = FIL_DEFAULT_ENCRYPTION_KEY; + } + + /* If default encryption is used and encryption is disabled, you may + not use nondefault encryption_key_id as it is not stored anywhere. */ + if (encrypt == FIL_ENCRYPTION_DEFAULT + && !srv_encrypt_tables + && options->encryption_key_id != FIL_DEFAULT_ENCRYPTION_KEY) { + compile_time_assert(FIL_DEFAULT_ENCRYPTION_KEY == 1); + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_WRONG_CREATE_OPTION, + "InnoDB: innodb_encrypt_tables=OFF only allows ENCRYPTION_KEY_ID=1" + ); + return "ENCRYPTION_KEY_ID"; + } + + return NULL; +} + +/*****************************************************************//** Update create_info. Used in SHOW CREATE TABLE et al. */ -UNIV_INTERN + void ha_innobase::update_create_info( /*============================*/ HA_CREATE_INFO* create_info) /*!< in/out: create info */ { if (!(create_info->used_fields & HA_CREATE_USED_AUTO)) { - ha_innobase::info(HA_STATUS_AUTO); + info(HA_STATUS_AUTO); create_info->auto_increment_value = stats.auto_increment_value; } + if (dict_table_is_temporary(m_prebuilt->table)) { + return; + } + /* Update the DATA DIRECTORY name from SYS_DATAFILES. */ - dict_get_and_save_data_dir_path(prebuilt->table, false); + dict_get_and_save_data_dir_path(m_prebuilt->table, false); - if (prebuilt->table->data_dir_path) { - create_info->data_file_name = prebuilt->table->data_dir_path; + if (m_prebuilt->table->data_dir_path) { + create_info->data_file_name = m_prebuilt->table->data_dir_path; } } /*****************************************************************//** Initialize the table FTS stopword list @return TRUE if success */ -UNIV_INTERN ibool innobase_fts_load_stopword( /*=======================*/ @@ -11458,29 +11876,16 @@ innobase_fts_load_stopword( THDVAR(thd, ft_enable_stopword), FALSE)); } -/*****************************************************************//** -Parses the table name into normal name and either temp path or remote path -if needed. -@return 0 if successful, otherwise, error number */ -UNIV_INTERN +/** Parse the table name into normal name and remote path if needed. +@param[in] name Table name (db/table or full path). +@return 0 if successful, otherwise, error number */ int -ha_innobase::parse_table_name( -/*==========================*/ - const char* name, /*!< in/out: table name provided*/ - HA_CREATE_INFO* create_info, /*!< in: more information of the - created table, contains also the - create statement string */ - ulint flags, /*!< in: flags*/ - ulint flags2, /*!< in: flags2*/ - char* norm_name, /*!< out: normalized table name */ - char* temp_path, /*!< out: absolute path of table */ - char* remote_path) /*!< out: remote path of table */ -{ - THD* thd = ha_thd(); - bool use_tablespace = flags2 & DICT_TF2_USE_TABLESPACE; - DBUG_ENTER("ha_innobase::parse_table_name"); +create_table_info_t::parse_table_name( + const char* name) +{ + DBUG_ENTER("parse_table_name"); -#ifdef __WIN__ +#ifdef _WIN32 /* Names passed in from server are in two formats: 1. <database_name>/<table_name>: for normal table creation 2. full path: for temp table creation, or DATA DIRECTORY. @@ -11492,9 +11897,9 @@ ha_innobase::parse_table_name( returns error if it is in full path format, but not creating a temp. table. Currently InnoDB does not support symbolic link on Windows. */ - if (use_tablespace + if (m_innodb_file_per_table && !mysqld_embedded - && !(create_info->options & HA_LEX_CREATE_TMP_TABLE)) { + && !(m_create_info->options & HA_LEX_CREATE_TMP_TABLE)) { if ((name[1] == ':') || (name[0] == '\\' && name[1] == '\\')) { @@ -11504,51 +11909,31 @@ ha_innobase::parse_table_name( } #endif - normalize_table_name(norm_name, name); - temp_path[0] = '\0'; - remote_path[0] = '\0'; + m_remote_path[0] = '\0'; - /* A full path is used for TEMPORARY TABLE and DATA DIRECTORY. - In the case of; + /* Make sure DATA DIRECTORY is compatible with other options + and set the remote path. In the case of either; CREATE TEMPORARY TABLE ... DATA DIRECTORY={path} ... ; - We ignore the DATA DIRECTORY. */ - if (create_info->options & HA_LEX_CREATE_TMP_TABLE) { - strncpy(temp_path, name, FN_REFLEN - 1); - } - - if (create_info->data_file_name) { - bool ignore = false; - - /* Use DATA DIRECTORY only with file-per-table. */ - if (!use_tablespace) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: DATA DIRECTORY requires" - " innodb_file_per_table."); - ignore = true; - } - - /* Do not use DATA DIRECTORY with TEMPORARY TABLE. */ - if (create_info->options & HA_LEX_CREATE_TMP_TABLE) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: DATA DIRECTORY cannot be" - " used for TEMPORARY tables."); - ignore = true; - } - - if (ignore) { - my_error(WARN_OPTION_IGNORED, ME_JUST_WARNING, + CREATE TABLE ... DATA DIRECTORY={path} TABLESPACE={name}... ; + we ignore the DATA DIRECTORY. */ + if (m_create_info->data_file_name + && m_create_info->data_file_name[0] != '\0') { + if (!create_option_data_directory_is_valid()) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + WARN_OPTION_IGNORED, + ER_DEFAULT(WARN_OPTION_IGNORED), "DATA DIRECTORY"); + + m_flags &= ~DICT_TF_MASK_DATA_DIR; } else { - strncpy(remote_path, create_info->data_file_name, + strncpy(m_remote_path, + m_create_info->data_file_name, FN_REFLEN - 1); } } - if (create_info->index_file_name) { + if (m_create_info->index_file_name) { my_error(WARN_OPTION_IGNORED, ME_JUST_WARNING, "INDEX DIRECTORY"); } @@ -11556,32 +11941,27 @@ ha_innobase::parse_table_name( DBUG_RETURN(0); } -/*****************************************************************//** -Determines InnoDB table flags. +/** Determine InnoDB table flags. +If strict_mode=OFF, this will adjust the flags to what should be assumed. @retval true if successful, false if error */ -UNIV_INTERN bool -innobase_table_flags( -/*=================*/ - const TABLE* form, /*!< in: table */ - const HA_CREATE_INFO* create_info, /*!< in: information - on table columns and indexes */ - THD* thd, /*!< in: connection */ - bool use_tablespace, /*!< in: whether to create - outside system tablespace */ - ulint* flags, /*!< out: DICT_TF flags */ - ulint* flags2) /*!< out: DICT_TF2 flags */ +create_table_info_t::innobase_table_flags() { DBUG_ENTER("innobase_table_flags"); const char* fts_doc_id_index_bad = NULL; - bool zip_allowed = true; ulint zip_ssize = 0; - enum row_type row_format; + enum row_type row_type; rec_format_t innodb_row_format = get_row_format(innodb_default_row_format); - bool use_data_dir; - ha_table_option_struct *options= form->s->option_struct; + const bool is_temp + = m_create_info->options & HA_LEX_CREATE_TMP_TABLE; + bool zip_allowed + = !is_temp; + + const ulint zip_ssize_max = + ut_min(static_cast<ulint>(UNIV_PAGE_SSIZE_MAX), + static_cast<ulint>(PAGE_ZIP_SSIZE_MAX)); /* Cache the value of innodb_file_format, in case it is modified by another thread while the table is being created. */ @@ -11591,34 +11971,25 @@ innobase_table_flags( modified by another thread while the table is being created. */ const ulint default_compression_level = page_zip_level; - *flags = 0; - *flags2 = 0; + ha_table_option_struct *options= m_form->s->option_struct; + + m_flags = 0; + m_flags2 = 0; /* Check if there are any FTS indexes defined on this table. */ - for (uint i = 0; i < form->s->keys; i++) { - const KEY* key = &form->key_info[i]; + for (uint i = 0; i < m_form->s->keys; i++) { + const KEY* key = &m_form->key_info[i]; if (key->flags & HA_FULLTEXT) { - *flags2 |= DICT_TF2_FTS; + m_flags2 |= DICT_TF2_FTS; /* We don't support FTS indexes in temporary tables. */ - if (create_info->options & HA_LEX_CREATE_TMP_TABLE) { - + if (is_temp) { my_error(ER_INNODB_NO_FT_TEMP_TABLE, MYF(0)); DBUG_RETURN(false); } - if (key->flags & HA_USES_PARSER) { - my_error(ER_INNODB_NO_FT_USES_PARSER, MYF(0)); - DBUG_RETURN(false); - } - - if (key->flags & HA_USES_PARSER) { - my_error(ER_INNODB_NO_FT_USES_PARSER, MYF(0)); - DBUG_RETURN(false); - } - if (fts_doc_id_index_bad) { goto index_bad; } @@ -11636,7 +12007,7 @@ innobase_table_flags( fts_doc_id_index_bad = key->name; } - if (fts_doc_id_index_bad && (*flags2 & DICT_TF2_FTS)) { + if (fts_doc_id_index_bad && (m_flags2 & DICT_TF2_FTS)) { index_bad: my_error(ER_INNODB_FT_WRONG_DOCID_INDEX, MYF(0), fts_doc_id_index_bad); @@ -11644,61 +12015,66 @@ index_bad: } } - row_format = form->s->row_type; - - if (create_info->key_block_size) { + if (m_create_info->key_block_size > 0) { /* The requested compressed page size (key_block_size) is given in kilobytes. If it is a valid number, store that value as the number of log2 shifts from 512 in zip_ssize. Zero means it is not compressed. */ - ulint zssize; /* Zip Shift Size */ - ulint kbsize; /* Key Block Size */ + ulint zssize; /* Zip Shift Size */ + ulint kbsize; /* Key Block Size */ for (zssize = kbsize = 1; - zssize <= ut_min(UNIV_PAGE_SSIZE_MAX, - PAGE_ZIP_SSIZE_MAX); + zssize <= zip_ssize_max; zssize++, kbsize <<= 1) { - if (kbsize == create_info->key_block_size) { + if (kbsize == m_create_info->key_block_size) { zip_ssize = zssize; break; } } /* Make sure compressed row format is allowed. */ - if (!use_tablespace) { + if (is_temp) { push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: KEY_BLOCK_SIZE is ignored" + " for TEMPORARY TABLE."); + zip_allowed = false; + } else if (!m_allow_file_per_table) { + push_warning( + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: KEY_BLOCK_SIZE requires" " innodb_file_per_table."); - zip_allowed = FALSE; + zip_allowed = false; } if (file_format_allowed < UNIV_FORMAT_B) { push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: KEY_BLOCK_SIZE requires" " innodb_file_format > Antelope."); - zip_allowed = FALSE; + zip_allowed = false; } if (!zip_allowed - || zssize > ut_min(UNIV_PAGE_SSIZE_MAX, - PAGE_ZIP_SSIZE_MAX)) { + || zssize > zip_ssize_max) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: ignoring KEY_BLOCK_SIZE=%lu.", - create_info->key_block_size); + "InnoDB: ignoring KEY_BLOCK_SIZE=%u.", + (uint) m_create_info->key_block_size); } } + row_type = m_form->s->row_type; + if (zip_ssize && zip_allowed) { /* if ROW_FORMAT is set to default, - automatically change it to COMPRESSED.*/ - if (row_format == ROW_TYPE_DEFAULT) { - row_format = ROW_TYPE_COMPRESSED; - } else if (row_format != ROW_TYPE_COMPRESSED) { + automatically change it to COMPRESSED. */ + if (row_type == ROW_TYPE_DEFAULT) { + row_type = ROW_TYPE_COMPRESSED; + } else if (row_type != ROW_TYPE_COMPRESSED) { /* ROW_FORMAT other than COMPRESSED ignores KEY_BLOCK_SIZE. It does not make sense to reject conflicting @@ -11706,460 +12082,479 @@ index_bad: such combinations can be obtained with ALTER TABLE anyway. */ push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: ignoring KEY_BLOCK_SIZE=%lu" + "InnoDB: ignoring KEY_BLOCK_SIZE=%u" " unless ROW_FORMAT=COMPRESSED.", - create_info->key_block_size); - zip_allowed = FALSE; + (uint) m_create_info->key_block_size); + zip_allowed = false; } } else { - /* zip_ssize == 0 means no KEY_BLOCK_SIZE.*/ - if (row_format == ROW_TYPE_COMPRESSED && zip_allowed) { + /* zip_ssize == 0 means no KEY_BLOCK_SIZE. */ + if (row_type == ROW_TYPE_COMPRESSED && zip_allowed) { /* ROW_FORMAT=COMPRESSED without KEY_BLOCK_SIZE implies half the maximum KEY_BLOCK_SIZE(*1k) or UNIV_PAGE_SIZE, whichever is less. */ - zip_ssize = ut_min(UNIV_PAGE_SSIZE_MAX, - PAGE_ZIP_SSIZE_MAX) - 1; + zip_ssize = zip_ssize_max - 1; } } /* Validate the row format. Correct it if necessary */ - switch (row_format) { - case ROW_TYPE_DEFAULT: - break; + + switch (row_type) { case ROW_TYPE_REDUNDANT: innodb_row_format = REC_FORMAT_REDUNDANT; break; - case ROW_TYPE_DYNAMIC: - innodb_row_format = REC_FORMAT_DYNAMIC; + case ROW_TYPE_COMPACT: + innodb_row_format = REC_FORMAT_COMPACT; break; case ROW_TYPE_COMPRESSED: - if (!use_tablespace) { + if (is_temp) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: ROW_FORMAT=%s is ignored for" + " TEMPORARY TABLE.", + get_row_format_name(row_type)); + } else if (!m_allow_file_per_table) { + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: ROW_FORMAT=COMPRESSED requires" " innodb_file_per_table."); + } else if (file_format_allowed == UNIV_FORMAT_A) { push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: ROW_FORMAT=COMPRESSED requires" " innodb_file_format > Antelope."); } else { innodb_row_format = REC_FORMAT_COMPRESSED; - break; /* Correct row_format */ + break; } - zip_allowed = FALSE; + zip_allowed = false; /* Set ROW_FORMAT = COMPACT */ /* fall through */ case ROW_TYPE_NOT_USED: case ROW_TYPE_FIXED: case ROW_TYPE_PAGE: push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: assuming ROW_FORMAT=COMPACT."); + "InnoDB: assuming ROW_FORMAT=DYNAMIC."); /* fall through */ - case ROW_TYPE_COMPACT: - innodb_row_format = REC_FORMAT_COMPACT; + case ROW_TYPE_DYNAMIC: + innodb_row_format = REC_FORMAT_DYNAMIC; break; + case ROW_TYPE_DEFAULT: + ; } /* Don't support compressed table when page size > 16k. */ if (zip_allowed && zip_ssize && UNIV_PAGE_SIZE > UNIV_PAGE_SIZE_DEF) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: Cannot create a COMPRESSED table" - " when innodb_page_size > 16k." - " Assuming ROW_FORMAT=COMPACT."); - zip_allowed = FALSE; + push_warning(m_thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: Cannot create a COMPRESSED table" + " when innodb_page_size > 16k." + " Assuming ROW_FORMAT=DYNAMIC."); + zip_allowed = false; } + ut_ad(!is_temp || !zip_allowed); + ut_ad(!is_temp || innodb_row_format != REC_FORMAT_COMPRESSED); + /* Set the table flags */ if (!zip_allowed) { zip_ssize = 0; } - use_data_dir = use_tablespace - && ((create_info->data_file_name != NULL) - && !(create_info->options & HA_LEX_CREATE_TMP_TABLE)); - - /* Set up table dictionary flags */ - dict_tf_set(flags, - innodb_row_format, - zip_ssize, - use_data_dir, - options->page_compressed, - options->page_compression_level == 0 ? - default_compression_level : options->page_compression_level, - options->atomic_writes); - - if (create_info->options & HA_LEX_CREATE_TMP_TABLE) { - *flags2 |= DICT_TF2_TEMPORARY; + if (is_temp) { + m_flags2 |= DICT_TF2_TEMPORARY; + } else if (m_use_file_per_table) { + m_flags2 |= DICT_TF2_USE_FILE_PER_TABLE; } - if (use_tablespace) { - *flags2 |= DICT_TF2_USE_TABLESPACE; - } + /* Set the table flags */ + dict_tf_set(&m_flags, innodb_row_format, zip_ssize, + m_use_data_dir, + options->page_compressed, + options->page_compression_level == 0 ? + default_compression_level : static_cast<ulint>(options->page_compression_level), + 0); /* Set the flags2 when create table or alter tables */ - *flags2 |= DICT_TF2_FTS_AUX_HEX_NAME; + m_flags2 |= DICT_TF2_FTS_AUX_HEX_NAME; DBUG_EXECUTE_IF("innodb_test_wrong_fts_aux_table_name", - *flags2 &= ~DICT_TF2_FTS_AUX_HEX_NAME;); + m_flags2 &= ~DICT_TF2_FTS_AUX_HEX_NAME;); DBUG_RETURN(true); } +/** Parse MERGE_THRESHOLD value from the string. +@param[in] thd connection +@param[in] str string which might include 'MERGE_THRESHOLD=' +@return value parsed. 0 means not found or invalid value. */ +static +ulint +innobase_parse_merge_threshold( + THD* thd, + const char* str) +{ + static const char* label = "MERGE_THRESHOLD="; + static const size_t label_len = strlen(label); + const char* pos = str; -/*****************************************************************//** -Check engine specific table options not handled by SQL-parser. -@return NULL if valid, string if not */ -UNIV_INTERN -const char* -ha_innobase::check_table_options( - THD *thd, /*!< in: thread handle */ - TABLE* table, /*!< in: information on table - columns and indexes */ - HA_CREATE_INFO* create_info, /*!< in: more information of the - created table, contains also the - create statement string */ - const bool use_tablespace, /*!< in: use file par table */ - const ulint file_format) -{ - enum row_type row_format = table->s->row_type; - ha_table_option_struct *options= table->s->option_struct; - atomic_writes_t awrites = (atomic_writes_t)options->atomic_writes; - fil_encryption_t encrypt = (fil_encryption_t)options->encryption; + pos = strstr(str, label); - if (encrypt != FIL_ENCRYPTION_DEFAULT && !use_tablespace) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: ENCRYPTED requires innodb_file_per_table"); - return "ENCRYPTED"; - } + if (pos == NULL) { + return(0); + } - if (encrypt == FIL_ENCRYPTION_OFF && srv_encrypt_tables == 2) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: ENCRYPTED=OFF cannot be used when innodb_encrypt_tables=FORCE"); - return "ENCRYPTED"; + pos += label_len; + + lint ret = atoi(pos); + + if (ret > 0 && ret <= 50) { + return(static_cast<ulint>(ret)); } - /* Check page compression requirements */ - if (options->page_compressed) { + push_warning_printf( + thd, Sql_condition::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: Invalid value for MERGE_THRESHOLD in the CREATE TABLE" + " statement. The value is ignored."); - if (row_format == ROW_TYPE_COMPRESSED) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: PAGE_COMPRESSED table can't have" - " ROW_TYPE=COMPRESSED"); - return "PAGE_COMPRESSED"; - } + return(0); +} - if (row_format == ROW_TYPE_REDUNDANT) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: PAGE_COMPRESSED table can't have" - " ROW_TYPE=REDUNDANT"); - return "PAGE_COMPRESSED"; - } +/** Parse hint for table and its indexes, and update the information +in dictionary. +@param[in] thd connection +@param[in,out] table target table +@param[in] table_share table definition */ +void +innobase_parse_hint_from_comment( + THD* thd, + dict_table_t* table, + const TABLE_SHARE* table_share) +{ + ulint merge_threshold_table; + ulint merge_threshold_index[MAX_KEY]; + bool is_found[MAX_KEY]; + + if (table_share->comment.str != NULL) { + merge_threshold_table + = innobase_parse_merge_threshold( + thd, table_share->comment.str); + } else { + merge_threshold_table = DICT_INDEX_MERGE_THRESHOLD_DEFAULT; + } - if (!use_tablespace) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: PAGE_COMPRESSED requires" - " innodb_file_per_table."); - return "PAGE_COMPRESSED"; - } + if (merge_threshold_table == 0) { + merge_threshold_table = DICT_INDEX_MERGE_THRESHOLD_DEFAULT; + } - if (file_format < UNIV_FORMAT_B) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: PAGE_COMPRESSED requires" - " innodb_file_format > Antelope."); - return "PAGE_COMPRESSED"; + for (uint i = 0; i < table_share->keys; i++) { + KEY* key_info = &table_share->key_info[i]; + + ut_ad(i < sizeof(merge_threshold_index) + / sizeof(merge_threshold_index[0])); + + if (key_info->flags & HA_USES_COMMENT + && key_info->comment.str != NULL) { + merge_threshold_index[i] + = innobase_parse_merge_threshold( + thd, key_info->comment.str); + } else { + merge_threshold_index[i] = merge_threshold_table; } - if (create_info->key_block_size) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: PAGE_COMPRESSED table can't have" - " key_block_size"); - return "PAGE_COMPRESSED"; + if (merge_threshold_index[i] == 0) { + merge_threshold_index[i] = merge_threshold_table; } } - /* Check page compression level requirements, some of them are - already checked above */ - if (options->page_compression_level != 0) { - if (options->page_compressed == false) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: PAGE_COMPRESSION_LEVEL requires" - " PAGE_COMPRESSED"); - return "PAGE_COMPRESSION_LEVEL"; + /* update SYS_INDEX table */ + if (!dict_table_is_temporary(table)) { + for (uint i = 0; i < table_share->keys; i++) { + is_found[i] = false; } - if (options->page_compression_level < 1 || options->page_compression_level > 9) { - push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: invalid PAGE_COMPRESSION_LEVEL = %lu." - " Valid values are [1, 2, 3, 4, 5, 6, 7, 8, 9]", - options->page_compression_level); - return "PAGE_COMPRESSION_LEVEL"; - } - } + for (dict_index_t* index = UT_LIST_GET_FIRST(table->indexes); + index != NULL; + index = UT_LIST_GET_NEXT(indexes, index)) { - /* If encryption is set up make sure that used key_id is found */ - if (encrypt == FIL_ENCRYPTION_ON || - (encrypt == FIL_ENCRYPTION_DEFAULT && srv_encrypt_tables)) { - if (!encryption_key_id_exists((unsigned int)options->encryption_key_id)) { - push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: ENCRYPTION_KEY_ID %u not available", - (uint)options->encryption_key_id - ); - return "ENCRYPTION_KEY_ID"; + if (dict_index_is_auto_gen_clust(index)) { + + /* GEN_CLUST_INDEX should use + merge_threshold_table */ + dict_index_set_merge_threshold( + index, merge_threshold_table); + continue; + } + + for (uint i = 0; i < table_share->keys; i++) { + if (is_found[i]) { + continue; + } + + KEY* key_info = &table_share->key_info[i]; + + if (innobase_strcasecmp( + index->name, key_info->name) == 0) { + + dict_index_set_merge_threshold( + index, + merge_threshold_index[i]); + is_found[i] = true; + break; + } + } } } - /* Ignore nondefault key_id if encryption is set off */ - if (encrypt == FIL_ENCRYPTION_OFF && - options->encryption_key_id != THDVAR(thd, default_encryption_key_id)) { - push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: Ignored ENCRYPTION_KEY_ID %u when encryption is disabled", - (uint)options->encryption_key_id - ); - options->encryption_key_id = FIL_DEFAULT_ENCRYPTION_KEY; + for (uint i = 0; i < table_share->keys; i++) { + is_found[i] = false; } - /* If default encryption is used and encryption is disabled, you may - not use nondefault encryption_key_id as it is not stored anywhere. */ - if (encrypt == FIL_ENCRYPTION_DEFAULT - && !srv_encrypt_tables - && options->encryption_key_id != FIL_DEFAULT_ENCRYPTION_KEY) { - compile_time_assert(FIL_DEFAULT_ENCRYPTION_KEY == 1); - push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: innodb_encrypt_tables=OFF only allows ENCRYPTION_KEY_ID=1" - ); - return "ENCRYPTION_KEY_ID"; - } + /* update in memory */ + for (dict_index_t* index = UT_LIST_GET_FIRST(table->indexes); + index != NULL; + index = UT_LIST_GET_NEXT(indexes, index)) { - /* Check atomic writes requirements */ - if (awrites == ATOMIC_WRITES_ON || - (awrites == ATOMIC_WRITES_DEFAULT && srv_use_atomic_writes)) { - if (!use_tablespace) { - push_warning( - thd, Sql_condition::WARN_LEVEL_WARN, - HA_WRONG_CREATE_OPTION, - "InnoDB: ATOMIC_WRITES requires" - " innodb_file_per_table."); - return "ATOMIC_WRITES"; + if (dict_index_is_auto_gen_clust(index)) { + + /* GEN_CLUST_INDEX should use merge_threshold_table */ + + /* x-lock index is needed to exclude concurrent + pessimistic tree operations */ + rw_lock_x_lock(dict_index_get_lock(index)); + index->merge_threshold = merge_threshold_table; + rw_lock_x_unlock(dict_index_get_lock(index)); + + continue; } - } - return 0; + for (uint i = 0; i < table_share->keys; i++) { + if (is_found[i]) { + continue; + } + + KEY* key_info = &table_share->key_info[i]; + + if (innobase_strcasecmp( + index->name, key_info->name) == 0) { + + /* x-lock index is needed to exclude concurrent + pessimistic tree operations */ + rw_lock_x_lock(dict_index_get_lock(index)); + index->merge_threshold + = merge_threshold_index[i]; + rw_lock_x_unlock(dict_index_get_lock(index)); + is_found[i] = true; + + break; + } + } + } } -/*****************************************************************//** -Creates a new table to an InnoDB database. -@return error number */ -UNIV_INTERN +/** Set m_use_* flags. */ +void +create_table_info_t::set_tablespace_type( + bool table_being_altered_is_file_per_table) +{ + /** Allow file_per_table for this table either because: + 1) the setting innodb_file_per_table=on, + 2) the table being altered is currently file_per_table */ + m_allow_file_per_table = + m_innodb_file_per_table + || table_being_altered_is_file_per_table; + + /* Ignore the current innodb-file-per-table setting if we are + creating a temporary table. */ + m_use_file_per_table = + m_allow_file_per_table + && !(m_create_info->options & HA_LEX_CREATE_TMP_TABLE); + + /* DATA DIRECTORY must have m_use_file_per_table but cannot be + used with TEMPORARY tables. */ + m_use_data_dir = + m_use_file_per_table + && (m_create_info->data_file_name != NULL) + && (m_create_info->data_file_name[0] != '\0'); +} + +/** Initialize the create_table_info_t object. +@return error number */ int -ha_innobase::create( -/*================*/ - const char* name, /*!< in: table name */ - TABLE* form, /*!< in: information on table - columns and indexes */ - HA_CREATE_INFO* create_info) /*!< in: more information of the - created table, contains also the - create statement string */ +create_table_info_t::initialize() { - int error; - trx_t* parent_trx; - trx_t* trx; - int primary_key_no; - uint i; - char norm_name[FN_REFLEN]; /* {database}/{tablename} */ - char temp_path[FN_REFLEN]; /* absolute path of temp frm */ - char remote_path[FN_REFLEN]; /* absolute path of table */ - THD* thd = ha_thd(); - ib_int64_t auto_inc_value; - - /* Cache the global variable "srv_file_per_table" to a local - variable before using it. Note that "srv_file_per_table" - is not under dict_sys mutex protection, and could be changed - while creating the table. So we read the current value here - and make all further decisions based on this. */ - bool use_tablespace = srv_file_per_table; - const ulint file_format = srv_file_format; - - /* Zip Shift Size - log2 - 9 of compressed page size, - zero for uncompressed */ - ulint flags; - ulint flags2; - dict_table_t* innobase_table = NULL; + DBUG_ENTER("create_table_info_t::initialize"); - const char* stmt; - size_t stmt_len; - /* Cache table options */ - ha_table_option_struct *options= form->s->option_struct; - fil_encryption_t encrypt = (fil_encryption_t)options->encryption; - uint key_id = (uint)options->encryption_key_id; + ut_ad(m_thd != NULL); + ut_ad(m_create_info != NULL); - DBUG_ENTER("ha_innobase::create"); + if (m_form->s->fields > REC_MAX_N_USER_FIELDS) { + DBUG_RETURN(HA_ERR_TOO_MANY_FIELDS); + } - DBUG_ASSERT(thd != NULL); - DBUG_ASSERT(create_info != NULL); + /* Check for name conflicts (with reserved name) for + any user indices to be created. */ + if (innobase_index_name_is_reserved(m_thd, m_form->key_info, + m_form->s->keys)) { + DBUG_RETURN(HA_ERR_WRONG_INDEX); + } + + ut_ad(m_form->s->row_type == m_create_info->row_type); + + /* Get the transaction associated with the current thd, or create one + if not yet created */ + + check_trx_exists(m_thd); + + DBUG_RETURN(0); +} + + +/** Check if a virtual column is part of a fulltext or spatial index. */ +bool +create_table_info_t::gcols_in_fulltext_or_spatial() +{ + for (ulint i = 0; i < m_form->s->keys; i++) { + const KEY* key = m_form->key_info + i; + if (key->flags & (HA_SPATIAL | HA_FULLTEXT)) { + for (ulint j = 0; j < key->user_defined_key_parts; j++) { + const KEY_PART_INFO* key_part = key->key_part + j; + + /* We do not support special (Fulltext or + Spatial) index on virtual columns */ + if (innobase_is_v_fld(key_part->field)) { + my_error(ER_UNSUPPORTED_ACTION_ON_GENERATED_COLUMN, MYF(0)); + return true; + } + } + } - if (form->s->stored_fields > REC_MAX_N_USER_FIELDS) { - DBUG_RETURN(HA_ERR_TOO_MANY_FIELDS); - } else if (high_level_read_only) { - DBUG_RETURN(HA_ERR_TABLE_READONLY); } + return false; +} + + +/** Prepare to create a new table to an InnoDB database. +@param[in] name Table name +@return error number */ +int create_table_info_t::prepare_create_table(const char* name, bool strict) +{ + DBUG_ENTER("prepare_create_table"); + + ut_ad(m_thd != NULL); + ut_ad(m_create_info != NULL); + + ut_ad(m_form->s->row_type == m_create_info->row_type); - /* Create the table definition in InnoDB */ + set_tablespace_type(false); + + normalize_table_name(m_table_name, name); /* Validate table options not handled by the SQL-parser */ - if(check_table_options(thd, form, create_info, use_tablespace, - file_format)) { + if (check_table_options()) { DBUG_RETURN(HA_WRONG_CREATE_OPTION); } - /* Validate create options if innodb_strict_mode is set. */ - if (create_options_are_invalid( - thd, form, create_info, use_tablespace)) { + /* Validate the create options if innodb_strict_mode is set. + Do not use the regular message for ER_ILLEGAL_HA_CREATE_OPTION + because InnoDB might actually support the option, but not under + the current conditions. The messages revealing the specific + problems are reported inside this function. */ + if (strict && create_options_are_invalid()) { DBUG_RETURN(HA_WRONG_CREATE_OPTION); } - if (!innobase_table_flags(form, create_info, - thd, use_tablespace, - &flags, &flags2)) { - DBUG_RETURN(-1); - } - - error = parse_table_name(name, create_info, flags, flags2, - norm_name, temp_path, remote_path); - if (error) { - DBUG_RETURN(error); + /* Create the table flags and flags2 */ + if (!innobase_table_flags()) { + DBUG_RETURN(HA_WRONG_CREATE_OPTION); } - /* Look for a primary key */ - primary_key_no = (form->s->primary_key != MAX_KEY ? - (int) form->s->primary_key : - -1); - - /* Our function innobase_get_mysql_key_number_for_index assumes - the primary key is always number 0, if it exists */ - ut_a(primary_key_no == -1 || primary_key_no == 0); - - /* Check for name conflicts (with reserved name) for - any user indices to be created. */ - if (innobase_index_name_is_reserved(thd, form->key_info, - form->s->keys)) { - DBUG_RETURN(-1); + if (high_level_read_only) { + DBUG_RETURN(HA_ERR_TABLE_READONLY); } - if (row_is_magic_monitor_table(norm_name)) { - push_warning_printf(thd, - Sql_condition::WARN_LEVEL_WARN, - HA_ERR_WRONG_COMMAND, - "Using the table name %s to enable " - "diagnostic output is deprecated " - "and may be removed in future releases. " - "Use INFORMATION_SCHEMA or " - "PERFORMANCE_SCHEMA tables or " - "SET GLOBAL innodb_status_output=ON.", - dict_remove_db_name(norm_name)); - - /* Limit innodb monitor access to users with PROCESS privilege. - See http://bugs.mysql.com/32710 why we chose PROCESS. */ - if (check_global_access(thd, PROCESS_ACL)) { - DBUG_RETURN(HA_ERR_GENERIC); - } + if (gcols_in_fulltext_or_spatial()) { + DBUG_RETURN(HA_ERR_UNSUPPORTED); } - /* Get the transaction associated with the current thd, or create one - if not yet created */ - - parent_trx = check_trx_exists(thd); + DBUG_RETURN(parse_table_name(name)); +} - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ +/** Create the internal innodb table. +@param create_fk whether to add FOREIGN KEY constraints */ +int create_table_info_t::create_table(bool create_fk) +{ + int error; + int primary_key_no; + uint i; + dict_table_t* innobase_table = NULL; + const char* stmt; + size_t stmt_len; - trx_search_latch_release_if_reserved(parent_trx); + DBUG_ENTER("create_table"); - trx = innobase_trx_allocate(thd); + /* Look for a primary key */ + primary_key_no = (m_form->s->primary_key != MAX_KEY ? + (int) m_form->s->primary_key : -1); - /* Latch the InnoDB data dictionary exclusively so that no deadlocks - or lock waits can happen in it during a table create operation. - Drop table etc. do this latching in row0mysql.cc. */ + /* Our function innobase_get_mysql_key_number_for_index assumes + the primary key is always number 0, if it exists */ + ut_a(primary_key_no == -1 || primary_key_no == 0); - row_mysql_lock_data_dictionary(trx); + error = create_table_def(); - error = create_table_def(trx, form, norm_name, temp_path, - remote_path, flags, flags2, encrypt, key_id); if (error) { - goto cleanup; + DBUG_RETURN(error); } /* Create the keys */ - if (form->s->keys == 0 || primary_key_no == -1) { + if (m_form->s->keys == 0 || primary_key_no == -1) { /* Create an index which is used as the clustered index; order the rows by their row id which is internally generated by InnoDB */ error = create_clustered_index_when_no_primary( - trx, flags, norm_name); + m_trx, m_flags, m_table_name); if (error) { - goto cleanup; + DBUG_RETURN(error); } } if (primary_key_no != -1) { /* In InnoDB the clustered index must always be created first */ - if ((error = create_index(trx, form, flags, norm_name, + if ((error = create_index(m_trx, m_form, m_flags, m_table_name, (uint) primary_key_no))) { - goto cleanup; + DBUG_RETURN(error); } } /* Create the ancillary tables that are common to all FTS indexes on this table. */ - if (flags2 & DICT_TF2_FTS) { - enum fts_doc_id_index_enum ret; + if (m_flags2 & DICT_TF2_FTS) { + fts_doc_id_index_enum ret; innobase_table = dict_table_open_on_name( - norm_name, TRUE, FALSE, DICT_ERR_IGNORE_NONE); + m_table_name, TRUE, FALSE, DICT_ERR_IGNORE_NONE); ut_a(innobase_table); /* Check whether there already exists FTS_DOC_ID_INDEX */ ret = innobase_fts_check_doc_id_index_in_def( - form->s->keys, form->key_info); + m_form->s->keys, m_form->key_info); switch (ret) { case FTS_INCORRECT_DOC_ID_INDEX: - push_warning_printf(thd, + push_warning_printf(m_thd, Sql_condition::WARN_LEVEL_WARN, ER_WRONG_NAME_FOR_INDEX, " InnoDB: Index name %s is reserved" @@ -12171,7 +12566,7 @@ ha_innobase::create( " make sure it is of correct" " type\n", FTS_DOC_ID_INDEX_NAME, - innobase_table->name); + innobase_table->name.m_name); if (innobase_table->fts) { fts_free(innobase_table); @@ -12181,14 +12576,14 @@ ha_innobase::create( my_error(ER_WRONG_NAME_FOR_INDEX, MYF(0), FTS_DOC_ID_INDEX_NAME); error = -1; - goto cleanup; + DBUG_RETURN(error); case FTS_EXIST_DOC_ID_INDEX: case FTS_NOT_EXIST_DOC_ID_INDEX: break; } dberr_t err = fts_create_common_tables( - trx, innobase_table, norm_name, + m_trx, innobase_table, m_table_name, (ret == FTS_EXIST_DOC_ID_INDEX)); error = convert_error_code_to_mysql(err, 0, NULL); @@ -12196,24 +12591,31 @@ ha_innobase::create( dict_table_close(innobase_table, TRUE, FALSE); if (error) { - goto cleanup; + trx_rollback_to_savepoint(m_trx, NULL); + m_trx->error_state = DB_SUCCESS; + + row_drop_table_for_mysql(m_table_name, m_trx, + SQLCOM_DROP_DB); + + m_trx->error_state = DB_SUCCESS; + DBUG_RETURN(error); } } - for (i = 0; i < form->s->keys; i++) { + for (i = 0; i < m_form->s->keys; i++) { if (i != static_cast<uint>(primary_key_no)) { - if ((error = create_index(trx, form, flags, - norm_name, i))) { - goto cleanup; + if ((error = create_index(m_trx, m_form, m_flags, + m_table_name, i))) { + DBUG_RETURN(error); } } } /* Cache all the FTS indexes on this table in the FTS specific structure. They are used for FTS indexed column update handling. */ - if (flags2 & DICT_TF2_FTS) { + if (m_flags2 & DICT_TF2_FTS) { fts_t* fts = innobase_table->fts; ut_a(fts != NULL); @@ -12221,61 +12623,95 @@ ha_innobase::create( dict_table_get_all_fts_indexes(innobase_table, fts->indexes); } - stmt = innobase_get_stmt(thd, &stmt_len); + stmt = innobase_get_stmt_unsafe(m_thd, &stmt_len); if (stmt) { dberr_t err = row_table_add_foreign_constraints( - trx, stmt, stmt_len, norm_name, - create_info->options & HA_LEX_CREATE_TMP_TABLE); + create_fk ? m_trx : NULL, stmt, stmt_len, m_table_name, + m_create_info->options & HA_LEX_CREATE_TMP_TABLE); switch (err) { case DB_PARENT_NO_INDEX: push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_CANNOT_ADD_FOREIGN, "Create table '%s' with foreign key constraint" " failed. There is no index in the referenced" " table where the referenced columns appear" - " as the first columns.\n", norm_name); + " as the first columns.\n", m_table_name); break; case DB_CHILD_NO_INDEX: push_warning_printf( - thd, Sql_condition::WARN_LEVEL_WARN, + m_thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_CANNOT_ADD_FOREIGN, "Create table '%s' with foreign key constraint" " failed. There is no index in the referencing" " table where referencing columns appear" - " as the first columns.\n", norm_name); + " as the first columns.\n", m_table_name); break; + case DB_NO_FK_ON_S_BASE_COL: + push_warning_printf( + m_thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_CANNOT_ADD_FOREIGN, + "Create table '%s' with foreign key constraint" + " failed. Cannot add foreign key constraint" + " placed on the base column of stored" + " column. \n", + m_table_name); default: break; } - error = convert_error_code_to_mysql(err, flags, NULL); + error = convert_error_code_to_mysql(err, m_flags, NULL); if (error) { - goto cleanup; + DBUG_RETURN(error); } } - innobase_commit_low(trx); + innobase_table = dict_table_open_on_name( + m_table_name, TRUE, FALSE, DICT_ERR_IGNORE_NONE); - row_mysql_unlock_data_dictionary(trx); + if (innobase_table != NULL) { + dict_table_close(innobase_table, TRUE, FALSE); + } - /* Flush the log to reduce probability that the .frm files and - the InnoDB data dictionary get out-of-sync if the user runs - with innodb_flush_log_at_trx_commit = 0 */ + DBUG_RETURN(0); +} - log_buffer_flush_to_disk(); +/** Update a new table in an InnoDB database. +@return error number */ +int +create_table_info_t::create_table_update_dict() +{ + dict_table_t* innobase_table; + + DBUG_ENTER("create_table_update_dict"); innobase_table = dict_table_open_on_name( - norm_name, FALSE, FALSE, DICT_ERR_IGNORE_NONE); + m_table_name, FALSE, FALSE, DICT_ERR_IGNORE_NONE); DBUG_ASSERT(innobase_table != 0); + if (innobase_table->fts != NULL) { + if (innobase_table->fts_doc_id_index == NULL) { + innobase_table->fts_doc_id_index + = dict_table_get_index_on_name( + innobase_table, FTS_DOC_ID_INDEX_NAME); + DBUG_ASSERT(innobase_table->fts_doc_id_index != NULL); + } else { + DBUG_ASSERT(innobase_table->fts_doc_id_index + == dict_table_get_index_on_name( + innobase_table, + FTS_DOC_ID_INDEX_NAME)); + } + } - innobase_copy_frm_flags_from_create_info(innobase_table, create_info); + DBUG_ASSERT((innobase_table->fts == NULL) + == (innobase_table->fts_doc_id_index == NULL)); + + innobase_copy_frm_flags_from_create_info(innobase_table, m_create_info); dict_stats_update(innobase_table, DICT_STATS_EMPTY_TABLE); @@ -12289,109 +12725,209 @@ ha_innobase::create( } /* Load server stopword into FTS cache */ - if (flags2 & DICT_TF2_FTS) { - if (!innobase_fts_load_stopword(innobase_table, NULL, thd)) { + if (m_flags2 & DICT_TF2_FTS) { + if (!innobase_fts_load_stopword(innobase_table, NULL, m_thd)) { dict_table_close(innobase_table, FALSE, FALSE); srv_active_wake_master_thread(); - trx_free_for_mysql(trx); + trx_free_for_mysql(m_trx); DBUG_RETURN(-1); } } - /* Note: We can't call update_thd() as prebuilt will not be - setup at this stage and so we use thd. */ - - /* We need to copy the AUTOINC value from the old table if - this is an ALTER|OPTIMIZE TABLE or CREATE INDEX because CREATE INDEX - does a table copy too. If query was one of : + if (const Field* ai = m_form->found_next_number_field) { + ut_ad(!innobase_is_v_fld(ai)); - CREATE TABLE ...AUTO_INCREMENT = x; or - ALTER TABLE...AUTO_INCREMENT = x; or - OPTIMIZE TABLE t; or - CREATE INDEX x on t(...); + ib_uint64_t autoinc = m_create_info->auto_increment_value; - Find out a table definition from the dictionary and get - the current value of the auto increment field. Set a new - value to the auto increment field if the value is greater - than the maximum value in the column. */ + if (autoinc == 0) { + autoinc = 1; + } - if (((create_info->used_fields & HA_CREATE_USED_AUTO) - || thd_sql_command(thd) == SQLCOM_ALTER_TABLE - || thd_sql_command(thd) == SQLCOM_OPTIMIZE - || thd_sql_command(thd) == SQLCOM_CREATE_INDEX) - && create_info->auto_increment_value > 0) { + dict_table_autoinc_lock(innobase_table); + dict_table_autoinc_initialize(innobase_table, autoinc); - auto_inc_value = create_info->auto_increment_value; + if (dict_table_is_temporary(innobase_table)) { + /* AUTO_INCREMENT is not persistent for + TEMPORARY TABLE. Temporary tables are never + evicted. Keep the counter in memory only. */ + } else { + const unsigned col_no = innodb_col_no(ai); + + innobase_table->persistent_autoinc = 1 + + dict_table_get_nth_col_pos( + innobase_table, col_no, NULL); + + /* Persist the "last used" value, which + typically is AUTO_INCREMENT - 1. + In btr_create(), the value 0 was already written. */ + if (--autoinc) { + btr_write_autoinc( + dict_table_get_first_index( + innobase_table), + autoinc); + } + } - dict_table_autoinc_lock(innobase_table); - dict_table_autoinc_initialize(innobase_table, auto_inc_value); dict_table_autoinc_unlock(innobase_table); } + innobase_parse_hint_from_comment(m_thd, innobase_table, m_form->s); + dict_table_close(innobase_table, FALSE, FALSE); + DBUG_RETURN(0); +} - /* Tell the InnoDB server that there might be work for - utility threads: */ +/** Allocate a new trx. */ +void +create_table_info_t::allocate_trx() +{ + m_trx = innobase_trx_allocate(m_thd); - srv_active_wake_master_thread(); + m_trx->will_lock++; + m_trx->ddl = true; +} - trx_free_for_mysql(trx); +/** Create a new table to an InnoDB database. +@param[in] name Table name, format: "db/table_name". +@param[in] form Table format; columns and index information. +@param[in] create_info Create info (including create statement string). +@param[in] file_per_table whether to create .ibd file +@param[in,out] trx dictionary transaction, or NULL to create new +@return 0 if success else error number. */ +inline int +ha_innobase::create( + const char* name, + TABLE* form, + HA_CREATE_INFO* create_info, + bool file_per_table, + trx_t* trx) +{ + int error; + char norm_name[FN_REFLEN]; /* {database}/{tablename} */ + char remote_path[FN_REFLEN]; /* Absolute path of table */ - DBUG_RETURN(0); + DBUG_ENTER("ha_innobase::create"); + + create_table_info_t info(ha_thd(), + form, + create_info, + norm_name, + remote_path, + file_per_table, trx); + + if ((error = info.initialize()) + || (error = info.prepare_create_table(name, !trx))) { + if (trx) { + trx_rollback_for_mysql(trx); + row_mysql_unlock_data_dictionary(trx); + } + DBUG_RETURN(error); + } + + const bool own_trx = !trx; + + if (own_trx) { + info.allocate_trx(); + trx = info.trx(); + /* Latch the InnoDB data dictionary exclusively so that no deadlocks + or lock waits can happen in it during a table create operation. + Drop table etc. do this latching in row0mysql.cc. */ + row_mysql_lock_data_dictionary(trx); + DBUG_ASSERT(trx_state_eq(trx, TRX_STATE_NOT_STARTED)); + } -cleanup: - trx_rollback_for_mysql(trx); + if ((error = info.create_table(own_trx))) { + trx_rollback_for_mysql(trx); + row_mysql_unlock_data_dictionary(trx); + if (own_trx) { + trx_free_for_mysql(trx); + DBUG_RETURN(error); + } + } + innobase_commit_low(trx); row_mysql_unlock_data_dictionary(trx); - trx_free_for_mysql(trx); + if (own_trx) { + trx_free_for_mysql(trx); + } + + /* Flush the log to reduce probability that the .frm files and + the InnoDB data dictionary get out-of-sync if the user runs + with innodb_flush_log_at_trx_commit = 0 */ + log_buffer_flush_to_disk(); + + ut_ad(!srv_read_only_mode); + + error = info.create_table_update_dict(); + + /* Tell the InnoDB server that there might be work for + utility threads: */ + + srv_active_wake_master_thread(); DBUG_RETURN(error); } +/** Create a new table to an InnoDB database. +@param[in] name Table name, format: "db/table_name". +@param[in] form Table format; columns and index information. +@param[in] create_info Create info (including create statement string). +@return 0 if success else error number. */ +int +ha_innobase::create( + const char* name, + TABLE* form, + HA_CREATE_INFO* create_info) +{ + return create(name, form, create_info, srv_file_per_table); +} + /*****************************************************************//** Discards or imports an InnoDB tablespace. -@return 0 == success, -1 == error */ -UNIV_INTERN +@return 0 == success, -1 == error */ + int ha_innobase::discard_or_import_tablespace( /*======================================*/ - my_bool discard) /*!< in: TRUE if discard, else import */ + my_bool discard) /*!< in: TRUE if discard, else import */ { - dberr_t err; - dict_table_t* dict_table; DBUG_ENTER("ha_innobase::discard_or_import_tablespace"); - ut_a(prebuilt->trx); - ut_a(prebuilt->trx->magic_n == TRX_MAGIC_N); - ut_a(prebuilt->trx == thd_to_trx(ha_thd())); + ut_a(m_prebuilt->trx != NULL); + ut_a(m_prebuilt->trx->magic_n == TRX_MAGIC_N); + ut_a(m_prebuilt->trx == thd_to_trx(ha_thd())); if (high_level_read_only) { DBUG_RETURN(HA_ERR_TABLE_READONLY); } - dict_table = prebuilt->table; + dict_table_t* dict_table = m_prebuilt->table; - if (dict_table->space == TRX_SYS_SPACE) { + if (dict_table_is_temporary(dict_table)) { ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, - ER_TABLE_IN_SYSTEM_TABLESPACE, - table->s->table_name.str); + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + ER_CANNOT_DISCARD_TEMPORARY_TABLE); DBUG_RETURN(HA_ERR_TABLE_NEEDS_UPGRADE); } - trx_start_if_not_started(prebuilt->trx); + if (dict_table->space == srv_sys_space.space_id()) { + ib_senderrf( + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + ER_TABLE_IN_SYSTEM_TABLESPACE, + dict_table->name.m_name); + + DBUG_RETURN(HA_ERR_TABLE_NEEDS_UPGRADE); + } - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads. */ - trx_search_latch_release_if_reserved(prebuilt->trx); + trx_start_if_not_started(m_prebuilt->trx, true); /* Obtain an exclusive lock on the table. */ - err = row_mysql_lock_table( - prebuilt->trx, dict_table, LOCK_X, + dberr_t err = row_mysql_lock_table( + m_prebuilt->trx, dict_table, LOCK_X, discard ? "setting table lock for DISCARD TABLESPACE" : "setting table lock for IMPORT TABLESPACE"); @@ -12406,35 +12942,33 @@ ha_innobase::discard_or_import_tablespace( if (!dict_table->is_readable()) { ib_senderrf( - prebuilt->trx->mysql_thd, + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_WARN, ER_TABLESPACE_MISSING, - table->s->table_name.str); + dict_table->name.m_name); } err = row_discard_tablespace_for_mysql( - dict_table->name, prebuilt->trx); + dict_table->name.m_name, m_prebuilt->trx); } else if (dict_table->is_readable()) { /* Commit the transaction in order to release the table lock. */ - trx_commit_for_mysql(prebuilt->trx); + trx_commit_for_mysql(m_prebuilt->trx); + ib::error() << "Unable to import tablespace " + << dict_table->name << " because it already" + " exists. Please DISCARD the tablespace" + " before IMPORT."; ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, - ER_TABLESPACE_EXISTS, table->s->table_name.str); + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + ER_TABLESPACE_EXISTS, dict_table->name.m_name); DBUG_RETURN(HA_ERR_TABLE_EXIST); } else { - err = row_import_for_mysql(dict_table, prebuilt); + err = row_import_for_mysql(dict_table, m_prebuilt); if (err == DB_SUCCESS) { - if (table->found_next_number_field) { - dict_table_autoinc_lock(dict_table); - innobase_initialize_autoinc(); - dict_table_autoinc_unlock(dict_table); - } - info(HA_STATUS_TIME | HA_STATUS_CONST | HA_STATUS_VARIABLE @@ -12445,7 +12979,7 @@ ha_innobase::discard_or_import_tablespace( } /* Commit the transaction in order to release the table lock. */ - trx_commit_for_mysql(prebuilt->trx); + trx_commit_for_mysql(m_prebuilt->trx); if (err == DB_SUCCESS && !discard && dict_stats_is_persistent_enabled(dict_table)) { @@ -12462,82 +12996,25 @@ ha_innobase::discard_or_import_tablespace( ER_ALTER_INFO, "Error updating stats for table '%s'" " after table rebuild: %s", - dict_table->name, ut_strerr(ret)); + dict_table->name.m_name, ut_strerr(ret)); } } DBUG_RETURN(convert_error_code_to_mysql(err, dict_table->flags, NULL)); } -/*****************************************************************//** -Deletes all rows of an InnoDB table. -@return error number */ -UNIV_INTERN -int -ha_innobase::truncate() -/*===================*/ -{ - dberr_t err; - int error; - - DBUG_ENTER("ha_innobase::truncate"); - - if (high_level_read_only) { - DBUG_RETURN(HA_ERR_TABLE_READONLY); - } - - /* Get the transaction associated with the current thd, or create one - if not yet created, and update prebuilt->trx */ - - update_thd(ha_thd()); - - if (!trx_is_started(prebuilt->trx)) { - ++prebuilt->trx->will_lock; - } - /* Truncate the table in InnoDB */ - - err = row_truncate_table_for_mysql(prebuilt->table, prebuilt->trx); - - switch (err) { - - case DB_TABLESPACE_DELETED: - case DB_TABLESPACE_NOT_FOUND: - ib_senderrf( - prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, - (err == DB_TABLESPACE_DELETED ? - ER_TABLESPACE_DISCARDED : ER_TABLESPACE_MISSING), - table->s->table_name.str); - table->status = STATUS_NOT_FOUND; - error = HA_ERR_NO_SUCH_TABLE; - break; - - default: - error = convert_error_code_to_mysql( - err, prebuilt->table->flags, - prebuilt->trx->mysql_thd); - table->status = STATUS_NOT_FOUND; - break; - } - DBUG_RETURN(error); -} - -/*****************************************************************//** +/** Drops a table from an InnoDB database. Before calling this function, MySQL calls innobase_commit to commit the transaction of the current user. Then the current user cannot have locks set on the table. Drop table operation inside InnoDB will remove all locks any user has on the table inside InnoDB. -@return error number */ -UNIV_INTERN -int -ha_innobase::delete_table( -/*======================*/ - const char* name) /*!< in: table name */ +@param[in] name table name +@param[in] sqlcom SQLCOM_DROP_DB, SQLCOM_TRUNCATE, ... +@return error number */ +inline int ha_innobase::delete_table(const char* name, enum_sql_command sqlcom) { - ulint name_len; dberr_t err; - trx_t* parent_trx; - trx_t* trx; THD* thd = ha_thd(); char norm_name[FN_REFLEN]; @@ -12559,21 +13036,33 @@ ha_innobase::delete_table( if (srv_read_only_mode || srv_force_recovery >= SRV_FORCE_NO_UNDO_LOG_SCAN) { DBUG_RETURN(HA_ERR_TABLE_READONLY); - } else if (row_is_magic_monitor_table(norm_name) - && check_global_access(thd, PROCESS_ACL)) { - DBUG_RETURN(HA_ERR_GENERIC); } - parent_trx = check_trx_exists(thd); + trx_t* parent_trx = check_trx_exists(thd); - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ + /* Remove the to-be-dropped table from the list of modified tables + by parent_trx. Otherwise we may end up with an orphaned pointer to + the table object from parent_trx::mod_tables. This could happen in: + SET AUTOCOMMIT=0; + CREATE TABLE t (PRIMARY KEY (a)) ENGINE=INNODB SELECT 1 AS a UNION + ALL SELECT 1 AS a; */ + trx_mod_tables_t::const_iterator iter; - trx_search_latch_release_if_reserved(parent_trx); + for (iter = parent_trx->mod_tables.begin(); + iter != parent_trx->mod_tables.end(); + ++iter) { - trx = innobase_trx_allocate(thd); + dict_table_t* table_to_drop = *iter; - name_len = strlen(name); + if (strcmp(norm_name, table_to_drop->name.m_name) == 0) { + parent_trx->mod_tables.erase(table_to_drop); + break; + } + } + + trx_t* trx = innobase_trx_allocate(thd); + + ulint name_len = strlen(name); ut_a(name_len < 1000); @@ -12584,22 +13073,14 @@ ha_innobase::delete_table( /* We are doing a DDL operation. */ ++trx->will_lock; - trx->ddl = true; /* Drop the table in InnoDB */ - err = row_drop_table_for_mysql( - norm_name, trx, thd_sql_command(thd) == SQLCOM_DROP_DB, - FALSE); + err = row_drop_table_for_mysql(norm_name, trx, sqlcom); if (err == DB_TABLE_NOT_FOUND && innobase_get_lower_case_table_names() == 1) { - char* is_part = NULL; -#ifdef __WIN__ - is_part = strstr(norm_name, "#p#"); -#else - is_part = strstr(norm_name, "#P#"); -#endif /* __WIN__ */ + char* is_part = is_partition(norm_name); if (is_part) { char par_case_name[FN_REFLEN]; @@ -12615,16 +13096,78 @@ ha_innobase::delete_table( whether there exists table name in system table whose name is not being normalized to lower case */ - normalize_table_name_low( + normalize_table_name_c_low( par_case_name, name, FALSE); #endif err = row_drop_table_for_mysql( - par_case_name, trx, - thd_sql_command(thd) == SQLCOM_DROP_DB, - FALSE); + par_case_name, trx, sqlcom); + } + } + + if (err == DB_TABLE_NOT_FOUND) { + /* Test to drop all tables which matches db/tablename + '#'. + Only partitions can have '#' as non-first character in + the table name! + + Temporary table names always start with '#', partitions are + the only 'tables' that can have '#' after the first character + and table name must have length > 0. User tables cannot have + '#' since it would be translated to @0023. Therefor this should + only match partitions. */ + uint len = (uint) strlen(norm_name); + ulint num_partitions; + ut_a(len < FN_REFLEN); + norm_name[len] = '#'; + norm_name[len + 1] = 0; + err = row_drop_database_for_mysql(norm_name, trx, + &num_partitions); + norm_name[len] = 0; + if (num_partitions == 0 + && !row_is_mysql_tmp_table_name(norm_name)) { + table_name_t tbl_name; + tbl_name.m_name = norm_name; + ib::error() << "Table " << tbl_name << + " does not exist in the InnoDB" + " internal data dictionary though MariaDB is" + " trying to drop it. Have you copied the .frm" + " file of the table to the MariaDB database" + " directory from another database? " + << TROUBLESHOOTING_MSG; + } + if (num_partitions == 0) { + err = DB_TABLE_NOT_FOUND; + } + } + + /* TODO: remove this when the conversion tool from ha_partition to + native innodb partitioning is completed */ + if (err == DB_TABLE_NOT_FOUND + && innobase_get_lower_case_table_names() == 1) { + char* is_part = is_partition(norm_name); + + if (is_part != NULL) { + char par_case_name[FN_REFLEN]; + +#ifndef _WIN32 + /* Check for the table using lower + case name, including the partition + separator "P" */ + strcpy(par_case_name, norm_name); + innobase_casedn_str(par_case_name); +#else + /* On Windows platfrom, check + whether there exists table name in + system table whose name is + not being normalized to lower case */ + create_table_info_t::normalize_table_name_low( + par_case_name, name, FALSE); +#endif /* _WIN32 */ + err = row_drop_table_for_mysql( + par_case_name, trx, sqlcom, true); } } + ut_ad(!srv_read_only_mode); /* Flush the log to reduce probability that the .frm files and the InnoDB data dictionary get out-of-sync if the user runs with innodb_flush_log_at_trx_commit = 0 */ @@ -12637,129 +13180,39 @@ ha_innobase::delete_table( DBUG_RETURN(convert_error_code_to_mysql(err, 0, NULL)); } -/*****************************************************************//** -Defragment table. -@return error number */ -UNIV_INTERN -int -ha_innobase::defragment_table( -/*==========================*/ - const char* name, /*!< in: table name */ - const char* index_name, /*!< in: index name */ - bool async) /*!< in: whether to wait until finish */ -{ - char norm_name[FN_REFLEN]; - dict_table_t* table = NULL; - dict_index_t* index = NULL; - ibool one_index = (index_name != 0); - int ret = 0; - dberr_t err = DB_SUCCESS; - - if (!srv_defragment) { - return ER_FEATURE_DISABLED; - } - - normalize_table_name(norm_name, name); - - table = dict_table_open_on_name(norm_name, FALSE, - FALSE, DICT_ERR_IGNORE_NONE); - - for (index = dict_table_get_first_index(table); index; - index = dict_table_get_next_index(index)) { - - if (dict_index_is_corrupted(index)) { - continue; - } - - if (index->page == FIL_NULL) { - /* Do not defragment auxiliary tables related - to FULLTEXT INDEX. */ - ut_ad(index->type & DICT_FTS); - continue; - } - - if (one_index && strcasecmp(index_name, index->name) != 0) { - continue; - } - - if (btr_defragment_find_index(index)) { - // We borrow this error code. When the same index is - // already in the defragmentation queue, issue another - // defragmentation only introduces overhead. We return - // an error here to let the user know this is not - // necessary. Note that this will fail a query that's - // trying to defragment a full table if one of the - // indicies in that table is already in defragmentation. - // We choose this behavior so user is aware of this - // rather than silently defragment other indicies of - // that table. - ret = ER_SP_ALREADY_EXISTS; - break; - } - - os_event_t event = btr_defragment_add_index(index, async, &err); - - if (err != DB_SUCCESS) { - push_warning_printf( - current_thd, - Sql_condition::WARN_LEVEL_WARN, - ER_NO_SUCH_TABLE, - "Table %s is encrypted but encryption service or" - " used key_id is not available. " - " Can't continue checking table.", - index->table->name); - - ret = convert_error_code_to_mysql(err, 0, current_thd); - break; - } - - if (!async && event) { - while(os_event_wait_time(event, 1000000)) { - if (thd_killed(current_thd)) { - btr_defragment_remove_index(index); - ret = ER_QUERY_INTERRUPTED; - break; - } - } - os_event_free(event); - } - - if (ret) { - break; - } - if (one_index) { - one_index = FALSE; - break; - } - } - - dict_table_close(table, FALSE, FALSE); +/** Drop an InnoDB table. +@param[in] name table name +@return error number */ +int ha_innobase::delete_table(const char* name) +{ + enum_sql_command sqlcom = enum_sql_command(thd_sql_command(ha_thd())); - if (ret == 0 && one_index) { - ret = ER_NO_SUCH_INDEX; - } + if (sqlcom == SQLCOM_TRUNCATE + && thd_killed(ha_thd()) + && (m_prebuilt == NULL + || dict_table_is_temporary(m_prebuilt->table))) { + sqlcom = SQLCOM_DROP_TABLE; + } - return ret; + /* SQLCOM_TRUNCATE will be passed via ha_innobase::truncate() only. */ + DBUG_ASSERT(sqlcom != SQLCOM_TRUNCATE); + return delete_table(name, sqlcom); } -/*****************************************************************//** -Removes all tables in the named database inside InnoDB. */ +/** Remove all tables in the named database inside InnoDB. +@param[in] hton handlerton from InnoDB +@param[in] path Database path; Inside InnoDB the name of the last +directory in the path is used as the database name. +For example, in 'mysql/data/test' the database name is 'test'. */ + static void innobase_drop_database( -/*===================*/ - handlerton* hton, /*!< in: handlerton of Innodb */ - char* path) /*!< in: database path; inside InnoDB the name - of the last directory in the path is used as - the database name: for example, in - 'mysql/data/test' the database name is 'test' */ + handlerton* hton, + char* path) { - ulint len = 0; - trx_t* trx; - char* ptr; char* namebuf; - THD* thd = current_thd; /* Get the transaction associated with the current thd, or create one if not yet created */ @@ -12770,18 +13223,10 @@ innobase_drop_database( return; } - /* In the Windows plugin, thd = current_thd is always NULL */ - if (thd) { - trx_t* parent_trx = check_trx_exists(thd); - - /* In case MySQL calls this in the middle of a SELECT - query, release possible adaptive hash latch to avoid - deadlocks of threads */ - - trx_search_latch_release_if_reserved(parent_trx); - } + THD* thd = current_thd; - ptr = strend(path) - 2; + ulint len = 0; + char* ptr = strend(path) - 2; while (ptr >= path && *ptr != '\\' && *ptr != '/') { ptr--; @@ -12789,15 +13234,17 @@ innobase_drop_database( } ptr++; - namebuf = (char*) my_malloc((uint) len + 2, MYF(0)); + namebuf = (char*) my_malloc(/*PSI_INSTRUMENT_ME,*/ (uint) len + 2, MYF(0)); memcpy(namebuf, ptr, len); namebuf[len] = '/'; namebuf[len + 1] = '\0'; -#ifdef __WIN__ + +#ifdef _WIN32 innobase_casedn_str(namebuf); -#endif - trx = innobase_trx_allocate(thd); +#endif /* _WIN32 */ + + trx_t* trx = innobase_trx_allocate(thd); /* Either the transaction is already flagged as a locking transaction or it hasn't been started yet. */ @@ -12807,7 +13254,9 @@ innobase_drop_database( /* We are doing a DDL operation. */ ++trx->will_lock; - row_drop_database_for_mysql(namebuf, trx); + ulint dummy; + + row_drop_database_for_mysql(namebuf, trx, &dummy); my_free(namebuf); @@ -12818,27 +13267,31 @@ innobase_drop_database( log_buffer_flush_to_disk(); innobase_commit_low(trx); + trx_free_for_mysql(trx); } -/*********************************************************************//** -Renames an InnoDB table. +/** Rename an InnoDB table. +@param[in,out] trx InnoDB data dictionary transaction +@param[in] from old table name +@param[in] to new table name +@param[in] commit whether to commit trx @return DB_SUCCESS or error code */ -static MY_ATTRIBUTE((nonnull, warn_unused_result)) +inline dberr_t innobase_rename_table( -/*==================*/ - THD* thd, /*!< Connection thread handle */ - trx_t* trx, /*!< in: transaction */ - const char* from, /*!< in: old name of the table */ - const char* to) /*!< in: new name of the table */ + trx_t* trx, + const char* from, + const char* to, + bool commit = true) { dberr_t error; char norm_to[FN_REFLEN]; char norm_from[FN_REFLEN]; DBUG_ENTER("innobase_rename_table"); - DBUG_ASSERT(trx_get_dict_operation(trx) == TRX_DICT_OP_INDEX); + DBUG_ASSERT(trx_get_dict_operation(trx) == TRX_DICT_OP_INDEX + || trx_get_dict_operation(trx) == TRX_DICT_OP_TABLE); ut_ad(!srv_read_only_mode); @@ -12847,19 +13300,21 @@ innobase_rename_table( DEBUG_SYNC_C("innodb_rename_table_ready"); - trx_start_if_not_started(trx); + trx_start_if_not_started(trx, true); + ut_ad(trx->will_lock > 0); - /* Serialize data dictionary operations with dictionary mutex: - no deadlocks can occur then in these operations. */ - - row_mysql_lock_data_dictionary(trx); + if (commit) { + /* Serialize data dictionary operations with dictionary mutex: + no deadlocks can occur then in these operations. */ + row_mysql_lock_data_dictionary(trx); + } dict_table_t* table = dict_table_open_on_name(norm_from, TRUE, FALSE, DICT_ERR_IGNORE_NONE); /* Since DICT_BG_YIELD has sleep for 250 milliseconds, Convert lock_wait_timeout unit from second to 250 milliseconds */ - long int lock_wait_timeout = thd_lock_wait_timeout(thd) * 4; + long int lock_wait_timeout = thd_lock_wait_timeout(trx->mysql_thd) * 4; if (table != NULL) { for (dict_index_t* index = dict_table_get_first_index(table); index != NULL; @@ -12880,8 +13335,7 @@ innobase_rename_table( /* FTS sync is in progress. We shall timeout this operation */ if (lock_wait_timeout < 0) { error = DB_LOCK_WAIT_TIMEOUT; - row_mysql_unlock_data_dictionary(trx); - DBUG_RETURN(error); + goto func_exit; } /* Transaction must be flagged as a locking transaction or it hasn't @@ -12889,22 +13343,16 @@ innobase_rename_table( ut_a(trx->will_lock > 0); - error = row_rename_table_for_mysql( - norm_from, norm_to, trx, TRUE); + error = row_rename_table_for_mysql(norm_from, norm_to, trx, commit); if (error != DB_SUCCESS) { if (error == DB_TABLE_NOT_FOUND && innobase_get_lower_case_table_names() == 1) { - char* is_part = NULL; -#ifdef __WIN__ - is_part = strstr(norm_from, "#p#"); -#else - is_part = strstr(norm_from, "#P#"); -#endif /* __WIN__ */ + char* is_part = is_partition(norm_from); if (is_part) { char par_case_name[FN_REFLEN]; -#ifndef __WIN__ +#ifndef _WIN32 /* Check for the table using lower case name, including the partition separator "P" */ @@ -12915,36 +13363,39 @@ innobase_rename_table( whether there exists table name in system table whose name is not being normalized to lower case */ - normalize_table_name_low( + create_table_info_t::normalize_table_name_low( par_case_name, from, FALSE); -#endif - trx_start_if_not_started(trx); +#endif /* _WIN32 */ + trx_start_if_not_started(trx, true); error = row_rename_table_for_mysql( par_case_name, norm_to, trx, TRUE); } } if (error == DB_SUCCESS) { -#ifndef __WIN__ - sql_print_warning("Rename partition table %s " - "succeeds after converting to lower " - "case. The table may have " - "been moved from a case " - "in-sensitive file system.\n", +#ifndef _WIN32 + sql_print_warning("Rename partition table %s" + " succeeds after converting to lower" + " case. The table may have" + " been moved from a case" + " in-sensitive file system.\n", norm_from); #else - sql_print_warning("Rename partition table %s " - "succeeds after skipping the step to " - "lower case the table name. " - "The table may have been " - "moved from a case sensitive " - "file system.\n", + sql_print_warning("Rename partition table %s" + " succeeds after skipping the step to" + " lower case the table name." + " The table may have been" + " moved from a case sensitive" + " file system.\n", norm_from); -#endif /* __WIN__ */ +#endif /* _WIN32 */ } } - row_mysql_unlock_data_dictionary(trx); +func_exit: + if (commit) { + row_mysql_unlock_data_dictionary(trx); + } /* Flush the log to reduce probability that the .frm files and the InnoDB data dictionary get out-of-sync @@ -12955,20 +13406,119 @@ innobase_rename_table( DBUG_RETURN(error); } +/** TRUNCATE TABLE +@return error code +@retval 0 on success */ +int ha_innobase::truncate() +{ + DBUG_ENTER("ha_innobase::truncate"); + + if (high_level_read_only) { + DBUG_RETURN(HA_ERR_TABLE_READONLY); + } + + update_thd(); + + if (!srv_safe_truncate) { + if (!trx_is_started(m_prebuilt->trx)) { + ++m_prebuilt->trx->will_lock; + } + + dberr_t err = row_truncate_table_for_mysql( + m_prebuilt->table, m_prebuilt->trx); + + int error; + + switch (err) { + case DB_TABLESPACE_DELETED: + case DB_TABLESPACE_NOT_FOUND: + ib_senderrf( + m_prebuilt->trx->mysql_thd, IB_LOG_LEVEL_ERROR, + err == DB_TABLESPACE_DELETED + ? ER_TABLESPACE_DISCARDED + : ER_TABLESPACE_MISSING, + table->s->table_name.str); + error = HA_ERR_TABLESPACE_MISSING; + break; + default: + error = convert_error_code_to_mysql( + err, m_prebuilt->table->flags, + m_prebuilt->trx->mysql_thd); + break; + } + table->status = STATUS_NOT_FOUND; + DBUG_RETURN(error); + } + + HA_CREATE_INFO info; + mem_heap_t* heap = mem_heap_create(1000); + dict_table_t* ib_table = m_prebuilt->table; + const time_t update_time = ib_table->update_time; + const ulint stored_lock = m_prebuilt->stored_select_lock_type; + memset(&info, 0, sizeof info); + update_create_info_from_table(&info, table); + + if (dict_table_is_temporary(ib_table)) { + info.options|= HA_LEX_CREATE_TMP_TABLE; + } else { + dict_get_and_save_data_dir_path(ib_table, false); + } + + char* data_file_name = ib_table->data_dir_path; + + if (data_file_name) { + info.data_file_name = data_file_name + = mem_heap_strdup(heap, data_file_name); + } + + const char* temp_name = dict_mem_create_temporary_tablename( + heap, ib_table->name.m_name, ib_table->id); + const char* name = mem_heap_strdup(heap, ib_table->name.m_name); + trx_t* trx = innobase_trx_allocate(m_user_thd); + + ++trx->will_lock; + trx_set_dict_operation(trx, TRX_DICT_OP_TABLE); + row_mysql_lock_data_dictionary(trx); + int err = convert_error_code_to_mysql( + innobase_rename_table(trx, ib_table->name.m_name, temp_name, false), + ib_table->flags, m_user_thd); + if (err) { + trx_rollback_for_mysql(trx); + row_mysql_unlock_data_dictionary(trx); + } else { + err = create(name, table, &info, + dict_table_is_file_per_table(ib_table), trx); + } + + trx_free_for_mysql(trx); + + if (!err) { + /* Reopen the newly created table, and drop the + original table that was renamed to temp_name. */ + close(); + err = open(name, 0, 0); + if (!err) { + m_prebuilt->stored_select_lock_type = stored_lock; + m_prebuilt->table->update_time = update_time; + delete_table(temp_name, SQLCOM_TRUNCATE); + } + } + + mem_heap_free(heap); + DBUG_RETURN(err); +} + /*********************************************************************//** Renames an InnoDB table. -@return 0 or error code */ -UNIV_INTERN +@return 0 or error code */ + int ha_innobase::rename_table( /*======================*/ const char* from, /*!< in: old name of the table */ const char* to) /*!< in: new name of the table */ { - trx_t* trx; - dberr_t error; - trx_t* parent_trx; - THD* thd = ha_thd(); + THD* thd = ha_thd(); DBUG_ENTER("ha_innobase::rename_table"); @@ -12977,27 +13527,18 @@ ha_innobase::rename_table( DBUG_RETURN(HA_ERR_TABLE_READONLY); } - /* Get the transaction associated with the current thd, or create one - if not yet created */ - - parent_trx = check_trx_exists(thd); - - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ - - trx_search_latch_release_if_reserved(parent_trx); - - trx = innobase_trx_allocate(thd); + trx_t* trx = innobase_trx_allocate(thd); /* We are doing a DDL operation. */ ++trx->will_lock; trx_set_dict_operation(trx, TRX_DICT_OP_INDEX); - error = innobase_rename_table(thd, trx, from, to); + dberr_t error = innobase_rename_table(trx, from, to); DEBUG_SYNC(thd, "after_innobase_rename_table"); innobase_commit_low(trx); + trx_free_for_mysql(trx); if (error == DB_SUCCESS) { @@ -13013,8 +13554,7 @@ ha_innobase::rename_table( errstr, sizeof(errstr)); if (ret != DB_SUCCESS) { - ut_print_timestamp(stderr); - fprintf(stderr, " InnoDB: %s\n", errstr); + ib::error() << errstr; push_warning(thd, Sql_condition::WARN_LEVEL_WARN, ER_LOCK_WAIT_TIMEOUT, errstr); @@ -13048,8 +13588,8 @@ ha_innobase::rename_table( /*********************************************************************//** Estimates the number of index records in a range. -@return estimated number of rows */ -UNIV_INTERN +@return estimated number of rows */ + ha_rows ha_innobase::records_in_range( /*==========================*/ @@ -13063,21 +13603,16 @@ ha_innobase::records_in_range( dict_index_t* index; dtuple_t* range_start; dtuple_t* range_end; - ib_int64_t n_rows; - ulint mode1; - ulint mode2; + int64_t n_rows; + page_cur_mode_t mode1; + page_cur_mode_t mode2; mem_heap_t* heap; DBUG_ENTER("records_in_range"); - ut_a(prebuilt->trx == thd_to_trx(ha_thd())); - - prebuilt->trx->op_info = (char*)"estimating records in index range"; + ut_a(m_prebuilt->trx == thd_to_trx(ha_thd())); - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ - - trx_search_latch_release_if_reserved(prebuilt->trx); + m_prebuilt->trx->op_info = "estimating records in index range"; active_index = keynr; @@ -13088,19 +13623,19 @@ ha_innobase::records_in_range( /* There exists possibility of not being able to find requested index due to inconsistency between MySQL and InoDB dictionary info. Necessary message should have been printed in innobase_get_index() */ - if (dict_table_is_discarded(prebuilt->table)) { + if (dict_table_is_discarded(m_prebuilt->table)) { n_rows = HA_POS_ERROR; goto func_exit; } - if (UNIV_UNLIKELY(!index)) { + if (!index) { n_rows = HA_POS_ERROR; goto func_exit; } - if (dict_index_is_corrupted(index)) { + if (index->is_corrupted()) { n_rows = HA_ERR_INDEX_CORRUPT; goto func_exit; } - if (UNIV_UNLIKELY(!row_merge_is_index_usable(prebuilt->trx, index))) { + if (!row_merge_is_index_usable(m_prebuilt->trx, index)) { n_rows = HA_ERR_TABLE_DEF_CHANGED; goto func_exit; } @@ -13115,41 +13650,47 @@ ha_innobase::records_in_range( dict_index_copy_types(range_end, index, key->ext_key_parts); row_sel_convert_mysql_key_to_innobase( - range_start, - prebuilt->srch_key_val1, - prebuilt->srch_key_val_len, - index, - (byte*) (min_key ? min_key->key : - (const uchar*) 0), - (ulint) (min_key ? min_key->length : 0), - prebuilt->trx); + range_start, + m_prebuilt->srch_key_val1, + m_prebuilt->srch_key_val_len, + index, + (byte*) (min_key ? min_key->key : (const uchar*) 0), + (ulint) (min_key ? min_key->length : 0), + m_prebuilt->trx); + DBUG_ASSERT(min_key ? range_start->n_fields > 0 : range_start->n_fields == 0); row_sel_convert_mysql_key_to_innobase( - range_end, - prebuilt->srch_key_val2, - prebuilt->srch_key_val_len, - index, - (byte*) (max_key ? max_key->key : - (const uchar*) 0), - (ulint) (max_key ? max_key->length : 0), - prebuilt->trx); + range_end, + m_prebuilt->srch_key_val2, + m_prebuilt->srch_key_val_len, + index, + (byte*) (max_key ? max_key->key : (const uchar*) 0), + (ulint) (max_key ? max_key->length : 0), + m_prebuilt->trx); + DBUG_ASSERT(max_key ? range_end->n_fields > 0 : range_end->n_fields == 0); - mode1 = convert_search_mode_to_innobase(min_key ? min_key->flag : - HA_READ_KEY_EXACT); - mode2 = convert_search_mode_to_innobase(max_key ? max_key->flag : - HA_READ_KEY_EXACT); + mode1 = convert_search_mode_to_innobase( + min_key ? min_key->flag : HA_READ_KEY_EXACT); + + mode2 = convert_search_mode_to_innobase( + max_key ? max_key->flag : HA_READ_KEY_EXACT); if (mode1 != PAGE_CUR_UNSUPP && mode2 != PAGE_CUR_UNSUPP) { - n_rows = btr_estimate_n_rows_in_range(index, range_start, - mode1, range_end, - mode2); + if (dict_index_is_spatial(index)) { + /*Only min_key used in spatial index. */ + n_rows = rtr_estimate_n_rows_in_range( + index, range_start, mode1); + } else { + n_rows = btr_estimate_n_rows_in_range( + index, range_start, mode1, range_end, mode2); + } } else { n_rows = HA_POS_ERROR; @@ -13157,9 +13698,18 @@ ha_innobase::records_in_range( mem_heap_free(heap); + DBUG_EXECUTE_IF( + "print_btr_estimate_n_rows_in_range_return_value", + push_warning_printf( + ha_thd(), Sql_condition::WARN_LEVEL_WARN, + ER_NO_DEFAULT, + "btr_estimate_n_rows_in_range(): %lld", + (longlong) n_rows); + ); + func_exit: - prebuilt->trx->op_info = (char*)""; + m_prebuilt->trx->op_info = (char*)""; /* The MySQL optimizer seems to believe an estimate of 0 rows is always accurate and may return the result 'Empty set' based on that. @@ -13177,8 +13727,8 @@ func_exit: /*********************************************************************//** Gives an UPPER BOUND to the number of rows in a table. This is used in filesort.cc. -@return upper bound of rows */ -UNIV_INTERN +@return upper bound of rows */ + ha_rows ha_innobase::estimate_rows_upper_bound() /*====================================*/ @@ -13186,7 +13736,6 @@ ha_innobase::estimate_rows_upper_bound() const dict_index_t* index; ulonglong estimate; ulonglong local_data_file_length; - ulint stat_n_leaf_pages; DBUG_ENTER("estimate_rows_upper_bound"); @@ -13196,16 +13745,11 @@ ha_innobase::estimate_rows_upper_bound() update_thd(ha_thd()); - prebuilt->trx->op_info = "calculating upper bound for table rows"; - - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ + m_prebuilt->trx->op_info = "calculating upper bound for table rows"; - trx_search_latch_release_if_reserved(prebuilt->trx); + index = dict_table_get_first_index(m_prebuilt->table); - index = dict_table_get_first_index(prebuilt->table); - - stat_n_leaf_pages = index->stat_n_leaf_pages; + ulint stat_n_leaf_pages = index->stat_n_leaf_pages; ut_a(stat_n_leaf_pages > 0); @@ -13220,7 +13764,7 @@ ha_innobase::estimate_rows_upper_bound() estimate = 2 * local_data_file_length / dict_index_calc_min_rec_len(index); - prebuilt->trx->op_info = ""; + m_prebuilt->trx->op_info = ""; /* Set num_rows less than MERGEBUFF to simulate the case where we do not have enough space to merge the externally sorted file blocks. */ @@ -13236,8 +13780,8 @@ ha_innobase::estimate_rows_upper_bound() How many seeks it will take to read through the table. This is to be comparable to the number returned by records_in_range so that we can decide if we should scan the table or use keys. -@return estimated time measured in disk seeks */ -UNIV_INTERN +@return estimated time measured in disk seeks */ + double ha_innobase::scan_time() /*====================*/ @@ -13251,26 +13795,22 @@ ha_innobase::scan_time() it we could end up returning uninitialized value to the caller, which in the worst case could make some query plan go bogus or issue a Valgrind warning. */ -#if 0 - /* avoid potential lock order violation with dict_table_stats_lock() - below */ - update_thd(ha_thd()); - trx_search_latch_release_if_reserved(prebuilt->trx); -#endif + if (m_prebuilt == NULL) { + /* In case of derived table, Optimizer will try to fetch stat + for table even before table is create or open. In such + cases return default value of 1. + TODO: This will be further improved to return some approximate + estimate but that would also needs pre-population of stats + structure. As of now approach is in sync with MyISAM. */ + return(ulonglong2double(stats.data_file_length) / IO_SIZE + 2); + } ulint stat_clustered_index_size; -#if 0 - dict_table_stats_lock(prebuilt->table, RW_S_LATCH); -#endif - - ut_a(prebuilt->table->stat_initialized); - - stat_clustered_index_size = prebuilt->table->stat_clustered_index_size; + ut_a(m_prebuilt->table->stat_initialized); -#if 0 - dict_table_stats_unlock(prebuilt->table, RW_S_LATCH); -#endif + stat_clustered_index_size = + m_prebuilt->table->stat_clustered_index_size; return((double) stat_clustered_index_size); } @@ -13278,8 +13818,8 @@ ha_innobase::scan_time() /******************************************************************//** Calculate the time it takes to read a set of ranges through an index This enables us to optimise reads for clustered indexes. -@return estimated time measured in disk seeks */ -UNIV_INTERN +@return estimated time measured in disk seeks */ + double ha_innobase::read_time( /*===================*/ @@ -13288,7 +13828,6 @@ ha_innobase::read_time( ha_rows rows) /*!< in: estimated number of rows in the ranges */ { ha_rows total_rows; - double time_for_scan; if (index != table->s->primary_key) { /* Not clustered */ @@ -13298,7 +13837,7 @@ ha_innobase::read_time( /* Assume that the read time is proportional to the scan time for all rows + at most one seek per range. */ - time_for_scan = scan_time(); + double time_for_scan = scan_time(); if ((total_rows = estimate_rows_upper_bound()) < rows) { @@ -13310,7 +13849,7 @@ ha_innobase::read_time( /******************************************************************//** Return the size of the InnoDB memory buffer. */ -UNIV_INTERN + longlong ha_innobase::get_memory_buffer_size() const /*=======================================*/ @@ -13318,32 +13857,31 @@ ha_innobase::get_memory_buffer_size() const return(innobase_buffer_pool_size); } +/** Update the system variable with the given value of the InnoDB +buffer pool size. +@param[in] buf_pool_size given value of buffer pool size.*/ +void +innodb_set_buf_pool_size(ulonglong buf_pool_size) +{ + innobase_buffer_pool_size = buf_pool_size; +} + /*********************************************************************//** -Calculates the key number used inside MySQL for an Innobase index. We will -first check the "index translation table" for a match of the index to get -the index number. If there does not exist an "index translation table", -or not able to find the index in the translation table, then we will fall back -to the traditional way of looping through dict_index_t list to find a -match. In this case, we have to take into account if we generated a -default clustered index for the table +Calculates the key number used inside MySQL for an Innobase index. @return the key number used inside MySQL */ static int innobase_get_mysql_key_number_for_index( /*====================================*/ - INNOBASE_SHARE* share, /*!< in: share structure for index - translation table. */ const TABLE* table, /*!< in: table in MySQL data dictionary */ - dict_table_t* ib_table,/*!< in: table in Innodb data + dict_table_t* ib_table,/*!< in: table in InnoDB data dictionary */ const dict_index_t* index) /*!< in: index */ { const dict_index_t* ind; unsigned int i; - ut_a(index); - /* If index does not belong to the table object of share structure (ib_table comes from the share structure) search the index->table object instead */ @@ -13356,7 +13894,8 @@ innobase_get_mysql_key_number_for_index( i++; } - if (row_table_got_default_clust_index(index->table)) { + if (dict_index_is_clust(index) + && dict_index_is_auto_gen_clust(index)) { ut_a(i > 0); i--; } @@ -13364,27 +13903,8 @@ innobase_get_mysql_key_number_for_index( return(i); } - /* If index translation table exists, we will first check - the index through index translation table for a match. */ - if (share->idx_trans_tbl.index_mapping) { - for (i = 0; i < share->idx_trans_tbl.index_count; i++) { - if (share->idx_trans_tbl.index_mapping[i] == index) { - return(i); - } - } - - /* Print an error message if we cannot find the index - in the "index translation table". */ - if (*index->name != TEMP_INDEX_PREFIX) { - sql_print_error("Cannot find index %s in InnoDB index " - "translation table.", index->name); - } - } - - /* If we do not have an "index translation table", or not able - to find the index in the translation table, we'll directly find - matching index with information from mysql TABLE structure and - InnoDB dict_index_t list */ + /* Directly find matching index with information from mysql TABLE + structure and InnoDB dict_index_t list */ for (i = 0; i < table->s->keys; i++) { ind = dict_table_get_index_on_name( ib_table, table->key_info[i].name); @@ -13402,12 +13922,13 @@ innobase_get_mysql_key_number_for_index( /* Temp index is internal to InnoDB, that is not present in the MySQL index list, so no need to print such mismatch warning. */ - if (*(index->name) != TEMP_INDEX_PREFIX) { + if (index->is_committed()) { sql_print_warning( - "Find index %s in InnoDB index list " - "but not its MySQL index number " - "It could be an InnoDB internal index.", - index->name); + "Found index %s in InnoDB index list" + " but not its MariaDB index number." + " It could be an InnoDB internal" + " index.", + index->name()); } return(-1); } @@ -13422,8 +13943,7 @@ innobase_get_mysql_key_number_for_index( Calculate Record Per Key value. Need to exclude the NULL value if innodb_stats_method is set to "nulls_ignored" @return estimated record per key value */ -static -ha_rows +rec_per_key_t innodb_rec_per_key( /*===============*/ dict_index_t* index, /*!< in: dict_index_t structure */ @@ -13431,18 +13951,25 @@ innodb_rec_per_key( calculating rec per key */ ha_rows records) /*!< in: estimated total records */ { - ha_rows rec_per_key; + rec_per_key_t rec_per_key; ib_uint64_t n_diff; ut_a(index->table->stat_initialized); ut_ad(i < dict_index_get_n_unique(index)); + ut_ad(!dict_index_is_spatial(index)); + + if (records == 0) { + /* "Records per key" is meaningless for empty tables. + Return 1.0 because that is most convenient to the Optimizer. */ + return(1.0); + } n_diff = index->stat_n_diff_key_vals[i]; if (n_diff == 0) { - rec_per_key = records; + rec_per_key = static_cast<rec_per_key_t>(records); } else if (srv_innodb_stats_method == SRV_STATS_NULLS_IGNORED) { ib_uint64_t n_null; ib_uint64_t n_non_null; @@ -13465,16 +13992,23 @@ innodb_rec_per_key( consider that the table consists mostly of NULL value. Set rec_per_key to 1. */ if (n_diff <= n_null) { - rec_per_key = 1; + rec_per_key = 1.0; } else { /* Need to exclude rows with NULL values from rec_per_key calculation */ - rec_per_key = (ha_rows) - ((records - n_null) / (n_diff - n_null)); + rec_per_key + = static_cast<rec_per_key_t>(records - n_null) + / (n_diff - n_null); } } else { DEBUG_SYNC_C("after_checking_for_0"); - rec_per_key = (ha_rows) (records / n_diff); + rec_per_key = static_cast<rec_per_key_t>(records) / n_diff; + } + + if (rec_per_key < 1.0) { + /* Values below 1.0 are meaningless and must be due to the + stats being imprecise. */ + rec_per_key = 1.0; } return(rec_per_key); @@ -13484,7 +14018,7 @@ innodb_rec_per_key( Returns statistics information of the table to the MySQL interpreter, in various fields of the handle object. @return HA_ERR_* error code or 0 */ -UNIV_INTERN + int ha_innobase::info_low( /*==================*/ @@ -13492,12 +14026,14 @@ ha_innobase::info_low( bool is_analyze) { dict_table_t* ib_table; - ha_rows rec_per_key; ib_uint64_t n_rows; + char path[FN_REFLEN]; os_file_stat_t stat_info; DBUG_ENTER("info"); + DEBUG_SYNC_C("ha_innobase_info_low"); + /* If we are forcing recovery at a high level, we will suppress statistics calculation on tables, because that may crash the server if an index is badly corrupted. */ @@ -13508,15 +14044,10 @@ ha_innobase::info_low( update_thd(ha_thd()); - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ - - prebuilt->trx->op_info = (char*)"returning various info to MySQL"; - - trx_search_latch_release_if_reserved(prebuilt->trx); + m_prebuilt->trx->op_info = "returning various info to MariaDB"; - ib_table = prebuilt->table; - DBUG_ASSERT(ib_table->n_ref_count > 0); + ib_table = m_prebuilt->table; + DBUG_ASSERT(ib_table->get_ref_count() > 0); if (flag & HA_STATUS_TIME) { if (is_analyze || innobase_stats_on_metadata) { @@ -13524,7 +14055,7 @@ ha_innobase::info_low( dict_stats_upd_option_t opt; dberr_t ret; - prebuilt->trx->op_info = "updating table statistics"; + m_prebuilt->trx->op_info = "updating table statistics"; if (dict_stats_is_persistent_enabled(ib_table)) { @@ -13543,19 +14074,20 @@ ha_innobase::info_low( ret = dict_stats_update(ib_table, opt); if (ret != DB_SUCCESS) { - prebuilt->trx->op_info = ""; + m_prebuilt->trx->op_info = ""; DBUG_RETURN(HA_ERR_GENERIC); } - prebuilt->trx->op_info = - "returning various info to MySQL"; + m_prebuilt->trx->op_info = + "returning various info to MariaDB"; } + + stats.update_time = (ulong) ib_table->update_time; } if (flag & HA_STATUS_VARIABLE) { - ulint page_size; ulint stat_clustered_index_size; ulint stat_sum_of_other_index_sizes; @@ -13595,32 +14127,30 @@ ha_innobase::info_low( n_rows can not be 0 unless the table is empty, set to 1 instead. The original problem of bug#29507 is actually fixed in the server code. */ - if (thd_sql_command(user_thd) == SQLCOM_TRUNCATE) { + if (thd_sql_command(m_user_thd) == SQLCOM_TRUNCATE) { n_rows = 1; - /* We need to reset the prebuilt value too, otherwise + /* We need to reset the m_prebuilt value too, otherwise checks for values greater than the last value written to the table will fail and the autoinc counter will not be updated. This will force write_row() into attempting an update of the table's AUTOINC counter. */ - prebuilt->autoinc_last_value = 0; + m_prebuilt->autoinc_last_value = 0; } - page_size = dict_table_zip_size(ib_table); - if (page_size == 0) { - page_size = UNIV_PAGE_SIZE; - } + const page_size_t& page_size + = dict_table_page_size(ib_table); stats.records = (ha_rows) n_rows; stats.deleted = 0; stats.data_file_length = ((ulonglong) stat_clustered_index_size) - * page_size; + * page_size.physical(); stats.index_file_length = ((ulonglong) stat_sum_of_other_index_sizes) - * page_size; + * page_size.physical(); /* Since fsp_get_available_space_in_free_extents() is acquiring latches inside InnoDB, we do not call it if we @@ -13635,19 +14165,19 @@ ha_innobase::info_low( the ha_statistics' constructor. Also we only need delete_length to be set when HA_STATUS_VARIABLE_EXTRA is set */ - } else if (UNIV_UNLIKELY - (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE)) { + } else if (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE) { /* Avoid accessing the tablespace if innodb_crash_recovery is set to a high value. */ stats.delete_length = 0; } else { - ullint avail_space; + uintmax_t avail_space; avail_space = fsp_get_available_space_in_free_extents( ib_table->space); - if (avail_space == ULLINT_UNDEFINED) { + if (avail_space == UINTMAX_MAX) { THD* thd; + char errbuf[MYSYS_STRERROR_SIZE]; thd = ha_thd(); @@ -13655,13 +14185,15 @@ ha_innobase::info_low( thd, Sql_condition::WARN_LEVEL_WARN, ER_CANT_GET_STAT, - "InnoDB: Trying to get the free " - "space for table %s but its " - "tablespace has been discarded or " - "the .ibd file is missing. Setting " - "the free space to zero. " - "(errno: %M)", - ib_table->name, errno); + "InnoDB: Trying to get the free" + " space for table %s but its" + " tablespace has been discarded or" + " the .ibd file is missing. Setting" + " the free space to zero." + " (errno: %d - %s)", + ib_table->name.m_name, errno, + my_strerror(errbuf, sizeof(errbuf), + errno)); stats.delete_length = 0; } else { @@ -13670,7 +14202,7 @@ ha_innobase::info_low( } stats.check_time = 0; - stats.mrr_length_per_rec= ref_length + 8; // 8 = max(sizeof(void *)); + stats.mrr_length_per_rec= ref_length + 8; // 8 = max(sizeof(void *)); if (stats.records == 0) { stats.mean_rec_length = 0; @@ -13682,12 +14214,11 @@ ha_innobase::info_low( if (flag & HA_STATUS_CONST) { ulong i; - char path[FN_REFLEN]; /* Verify the number of index in InnoDB and MySQL - matches up. If prebuilt->clust_index_was_generated + matches up. If m_prebuilt->clust_index_was_generated holds, InnoDB defines GEN_CLUST_INDEX internally */ ulint num_innodb_index = UT_LIST_GET_LEN(ib_table->indexes) - - prebuilt->clust_index_was_generated; + - m_prebuilt->clust_index_was_generated; if (table->s->keys < num_innodb_index) { /* If there are too many indexes defined inside InnoDB, ignore those that are being @@ -13707,7 +14238,7 @@ ha_innobase::info_low( time frame, dict_index_is_online_ddl() would not hold and the index would still not be included in TABLE_SHARE. */ - if (*index->name == TEMP_INDEX_PREFIX) { + if (!index->is_committed()) { num_innodb_index--; } } @@ -13722,7 +14253,7 @@ ha_innobase::info_low( if (table->s->keys != num_innodb_index) { ib_table->dict_frm_mismatch = DICT_FRM_INCONSISTENT_KEYS; - ib_push_frm_error(user_thd, ib_table, table, num_innodb_index, true); + ib_push_frm_error(m_user_thd, ib_table, table, num_innodb_index, true); } if (!(flag & HA_STATUS_NO_LOCK)) { @@ -13733,111 +14264,96 @@ ha_innobase::info_low( for (i = 0; i < table->s->keys; i++) { ulong j; - rec_per_key = 1; - /* We could get index quickly through internal - index mapping with the index translation table. - The identity of index (match up index name with - that of table->key_info[i]) is already verified in - innobase_get_index(). */ + dict_index_t* index = innobase_get_index(i); if (index == NULL) { ib_table->dict_frm_mismatch = DICT_FRM_INCONSISTENT_KEYS; - ib_push_frm_error(user_thd, ib_table, table, num_innodb_index, true); + ib_push_frm_error(m_user_thd, ib_table, table, num_innodb_index, true); break; } - for (j = 0; j < table->key_info[i].ext_key_parts; j++) { + KEY* key = &table->key_info[i]; + + for (j = 0; j < key->ext_key_parts; j++) { + + if ((key->flags & HA_FULLTEXT) + || (key->flags & HA_SPATIAL)) { - if (table->key_info[i].flags & HA_FULLTEXT) { - /* The whole concept has no validity - for FTS indexes. */ - table->key_info[i].rec_per_key[j] = 1; + /* The record per key does not apply to + FTS or Spatial indexes. */ + /* + key->rec_per_key[j] = 1; + key->set_records_per_key(j, 1.0); + */ continue; } if (j + 1 > index->n_uniq) { sql_print_error( - "Index %s of %s has %lu columns" + "Index %s of %s has %u columns" " unique inside InnoDB, but " "MySQL is asking statistics for" " %lu columns. Have you mixed " "up .frm files from different " - "installations? " - "See " REFMAN - "innodb-troubleshooting.html\n", - index->name, - ib_table->name, - (unsigned long) - index->n_uniq, j + 1); + " installations? %s", + index->name(), + ib_table->name.m_name, + index->n_uniq, j + 1, + TROUBLESHOOTING_MSG); break; } - DBUG_EXECUTE_IF("ib_ha_innodb_stat_not_initialized", - index->table->stat_initialized = FALSE;); - - if (!ib_table->stat_initialized || - (index->table != ib_table || - !index->table->stat_initialized)) { - fprintf(stderr, - "InnoDB: Warning: Index %s points to table %s" - " and ib_table %s statistics is initialized %d " - " but index table %s initialized %d " - " mysql table is %s. Have you mixed " - "up .frm files from different " - "installations? " - "See " REFMAN - "innodb-troubleshooting.html\n", - index->name, - index->table->name, - ib_table->name, - ib_table->stat_initialized, - index->table->name, - index->table->stat_initialized, - table->s->table_name.str - ); - - /* This is better than - assert on below function */ - dict_stats_init(index->table); - } - - rec_per_key = innodb_rec_per_key( - index, j, stats.records); + /* innodb_rec_per_key() will use + index->stat_n_diff_key_vals[] and the value we + pass index->table->stat_n_rows. Both are + calculated by ANALYZE and by the background + stats gathering thread (which kicks in when too + much of the table has been changed). In + addition table->stat_n_rows is adjusted with + each DML (e.g. ++ on row insert). Those + adjustments are not MVCC'ed and not even + reversed on rollback. So, + index->stat_n_diff_key_vals[] and + index->table->stat_n_rows could have been + calculated at different time. This is + acceptable. */ + + ulong rec_per_key_int = static_cast<ulong>( + innodb_rec_per_key(index, j, + stats.records)); /* Since MySQL seems to favor table scans too much over index searches, we pretend index selectivity is 2 times better than our estimate: */ - rec_per_key = rec_per_key / 2; + rec_per_key_int = rec_per_key_int / 2; - if (rec_per_key == 0) { - rec_per_key = 1; + if (rec_per_key_int == 0) { + rec_per_key_int = 1; } - table->key_info[i].rec_per_key[j] = - rec_per_key >= ~(ulong) 0 ? ~(ulong) 0 : - (ulong) rec_per_key; + key->rec_per_key[j] = rec_per_key_int; } - } if (!(flag & HA_STATUS_NO_LOCK)) { dict_table_stats_unlock(ib_table, RW_S_LATCH); } - my_snprintf(path, sizeof(path), "%s/%s%s", - mysql_data_home, - table->s->normalized_path.str, - reg_ext); + snprintf(path, sizeof(path), "%s/%s%s", + mysql_data_home, table->s->normalized_path.str, + reg_ext); unpack_filename(path,path); /* Note that we do not know the access time of the table, nor the CHECK TABLE time, nor the UPDATE or INSERT time. */ - if (os_file_get_status(path, &stat_info, false) == DB_SUCCESS) { + if (os_file_get_status( + path, &stat_info, false, + srv_read_only_mode) == DB_SUCCESS) { stats.create_time = (ulong) stat_info.ctime; } } @@ -13845,25 +14361,24 @@ ha_innobase::info_low( if (srv_force_recovery >= SRV_FORCE_NO_IBUF_MERGE) { goto func_exit; - } - if (flag & HA_STATUS_ERRKEY) { + } else if (flag & HA_STATUS_ERRKEY) { const dict_index_t* err_index; - ut_a(prebuilt->trx); - ut_a(prebuilt->trx->magic_n == TRX_MAGIC_N); + ut_a(m_prebuilt->trx); + ut_a(m_prebuilt->trx->magic_n == TRX_MAGIC_N); - err_index = trx_get_error_info(prebuilt->trx); + err_index = trx_get_error_info(m_prebuilt->trx); if (err_index) { errkey = innobase_get_mysql_key_number_for_index( - share, table, ib_table, err_index); + table, ib_table, err_index); } else { errkey = (unsigned int) ( - (prebuilt->trx->error_key_num + (m_prebuilt->trx->error_key_num == ULINT_UNDEFINED) ? ~0 - : prebuilt->trx->error_key_num); + : m_prebuilt->trx->error_key_num); } } @@ -13872,7 +14387,7 @@ ha_innobase::info_low( } func_exit: - prebuilt->trx->op_info = (char*)""; + m_prebuilt->trx->op_info = (char*)""; DBUG_RETURN(0); } @@ -13881,31 +14396,29 @@ func_exit: Returns statistics information of the table to the MySQL interpreter, in various fields of the handle object. @return HA_ERR_* error code or 0 */ -UNIV_INTERN + int ha_innobase::info( /*==============*/ uint flag) /*!< in: what information is requested */ { - return(this->info_low(flag, false /* not ANALYZE */)); + return(info_low(flag, false /* not ANALYZE */)); } -/**********************************************************************//** +/* Updates index cardinalities of the table, based on random dives into each index tree. This does NOT calculate exact statistics on the table. -@return HA_ADMIN_* error code or HA_ADMIN_OK */ -UNIV_INTERN +@return HA_ADMIN_* error code or HA_ADMIN_OK */ + int ha_innobase::analyze( /*=================*/ THD* thd, /*!< in: connection thread handle */ HA_CHECK_OPT* check_opt) /*!< in: currently ignored */ { - int ret; - - /* Simply call this->info_low() with all the flags + /* Simply call info_low() with all the flags and request recalculation of the statistics */ - ret = this->info_low( + int ret = info_low( HA_STATUS_TIME | HA_STATUS_CONST | HA_STATUS_VARIABLE, true /* this is ANALYZE */); @@ -13916,17 +14429,132 @@ ha_innobase::analyze( return(HA_ADMIN_OK); } +/*****************************************************************//** +Defragment table. +@return error number */ +UNIV_INTERN +int +ha_innobase::defragment_table( +/*==========================*/ + const char* name, /*!< in: table name */ + const char* index_name, /*!< in: index name */ + bool async) /*!< in: whether to wait until finish */ +{ + char norm_name[FN_REFLEN]; + dict_table_t* table = NULL; + dict_index_t* index = NULL; + ibool one_index = (index_name != 0); + int ret = 0; + dberr_t err = DB_SUCCESS; + + if (!srv_defragment) { + return ER_FEATURE_DISABLED; + } + + normalize_table_name(norm_name, name); + + table = dict_table_open_on_name(norm_name, FALSE, + FALSE, DICT_ERR_IGNORE_NONE); + + for (index = dict_table_get_first_index(table); index; + index = dict_table_get_next_index(index)) { + + if (index->is_corrupted()) { + continue; + } + + if (dict_index_is_spatial(index)) { + /* Do not try to defragment spatial indexes, + because doing it properly would require + appropriate logic around the SSN (split + sequence number). */ + continue; + } + + if (index->page == FIL_NULL) { + /* Do not defragment auxiliary tables related + to FULLTEXT INDEX. */ + ut_ad(index->type & DICT_FTS); + continue; + } + + if (one_index && strcasecmp(index_name, index->name) != 0) { + continue; + } + + if (btr_defragment_find_index(index)) { + // We borrow this error code. When the same index is + // already in the defragmentation queue, issue another + // defragmentation only introduces overhead. We return + // an error here to let the user know this is not + // necessary. Note that this will fail a query that's + // trying to defragment a full table if one of the + // indicies in that table is already in defragmentation. + // We choose this behavior so user is aware of this + // rather than silently defragment other indicies of + // that table. + ret = ER_SP_ALREADY_EXISTS; + break; + } + + os_event_t event = btr_defragment_add_index(index, async, &err); + + if (err != DB_SUCCESS) { + push_warning_printf( + current_thd, + Sql_condition::WARN_LEVEL_WARN, + ER_NO_SUCH_TABLE, + "Table %s is encrypted but encryption service or" + " used key_id is not available. " + " Can't continue checking table.", + index->table->name.m_name); + + ret = convert_error_code_to_mysql(err, 0, current_thd); + break; + } + + if (!async && event) { + while(os_event_wait_time(event, 1000000)) { + if (thd_killed(current_thd)) { + btr_defragment_remove_index(index); + ret = ER_QUERY_INTERRUPTED; + break; + } + } + os_event_destroy(event); + } + + if (ret) { + break; + } + + if (one_index) { + one_index = FALSE; + break; + } + } + + dict_table_close(table, FALSE, FALSE); + + if (ret == 0 && one_index) { + ret = ER_NO_SUCH_INDEX; + } + + return ret; +} + /**********************************************************************//** This is mapped to "ALTER TABLE tablename ENGINE=InnoDB", which rebuilds the table in MySQL. */ -UNIV_INTERN + int ha_innobase::optimize( /*==================*/ THD* thd, /*!< in: connection thread handle */ HA_CHECK_OPT* check_opt) /*!< in: currently ignored */ { - /*FTS-FIXME: Since MySQL doesn't support engine-specific commands, + + /* FTS-FIXME: Since MySQL doesn't support engine-specific commands, we have to hijack some existing command in order to be able to test the new admin commands added in InnoDB's FTS support. For now, we use MySQL's OPTIMIZE command, normally mapped to ALTER TABLE in @@ -13936,10 +14564,11 @@ ha_innobase::optimize( calls to OPTIMIZE, which is undesirable. */ bool try_alter = true; + /* TODO: Defragment is disabled for now */ if (srv_defragment) { int err; - err = defragment_table(prebuilt->table->name, NULL, false); + err = defragment_table(m_prebuilt->table->name.m_name, NULL, false); if (err == 0) { try_alter = false; @@ -13947,7 +14576,7 @@ ha_innobase::optimize( push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, err, "InnoDB: Cannot defragment table %s: returned error code %d\n", - prebuilt->table->name, err); + m_prebuilt->table->name, err); if(err == ER_SP_ALREADY_EXISTS) { try_alter = false; @@ -13956,10 +14585,10 @@ ha_innobase::optimize( } if (innodb_optimize_fulltext_only) { - if (prebuilt->table->fts && prebuilt->table->fts->cache - && !dict_table_is_discarded(prebuilt->table)) { - fts_sync_table(prebuilt->table, false, true, false); - fts_optimize_table(prebuilt->table); + if (m_prebuilt->table->fts && m_prebuilt->table->fts->cache + && !dict_table_is_discarded(m_prebuilt->table)) { + fts_sync_table(m_prebuilt->table, false, true, false); + fts_optimize_table(m_prebuilt->table); } try_alter = false; } @@ -13971,8 +14600,8 @@ ha_innobase::optimize( Tries to check that an InnoDB table is not corrupted. If corruption is noticed, prints to stderr information about it. In case of corruption may also assert a failure and crash the server. -@return HA_ADMIN_CORRUPT or HA_ADMIN_OK */ -UNIV_INTERN +@return HA_ADMIN_CORRUPT or HA_ADMIN_OK */ + int ha_innobase::check( /*===============*/ @@ -13984,22 +14613,21 @@ ha_innobase::check( ulint n_rows_in_table = ULINT_UNDEFINED; bool is_ok = true; ulint old_isolation_level; - ibool table_corrupted; + dberr_t ret; DBUG_ENTER("ha_innobase::check"); DBUG_ASSERT(thd == ha_thd()); - ut_a(prebuilt->trx); - ut_a(prebuilt->trx->magic_n == TRX_MAGIC_N); - ut_a(prebuilt->trx == thd_to_trx(thd)); + ut_a(m_prebuilt->trx->magic_n == TRX_MAGIC_N); + ut_a(m_prebuilt->trx == thd_to_trx(thd)); - if (prebuilt->mysql_template == NULL) { + if (m_prebuilt->mysql_template == NULL) { /* Build the template; we will use a dummy template in index scans done in checking */ build_template(true); } - if (dict_table_is_discarded(prebuilt->table)) { + if (dict_table_is_discarded(m_prebuilt->table)) { ib_senderrf( thd, @@ -14009,8 +14637,8 @@ ha_innobase::check( DBUG_RETURN(HA_ADMIN_CORRUPT); - } else if (prebuilt->table->file_unreadable && - fil_space_get(prebuilt->table->space) == NULL) { + } else if (!m_prebuilt->table->is_readable() && + !fil_space_get(m_prebuilt->table->space)) { ib_senderrf( thd, IB_LOG_LEVEL_ERROR, @@ -14020,87 +14648,71 @@ ha_innobase::check( DBUG_RETURN(HA_ADMIN_CORRUPT); } - if (prebuilt->table->corrupted) { - char index_name[MAX_FULL_NAME_LEN + 1]; + m_prebuilt->trx->op_info = "checking table"; + + if (m_prebuilt->table->corrupted) { /* If some previous operation has marked the table as corrupted in memory, and has not propagated such to clustered index, we will do so here */ - index = dict_table_get_first_index(prebuilt->table); + index = dict_table_get_first_index(m_prebuilt->table); - if (!dict_index_is_corrupted(index)) { - row_mysql_lock_data_dictionary(prebuilt->trx); - dict_set_corrupted(index, prebuilt->trx, "CHECK TABLE"); - row_mysql_unlock_data_dictionary(prebuilt->trx); + if (!index->is_corrupted()) { + dict_set_corrupted( + index, m_prebuilt->trx, "CHECK TABLE"); } - innobase_format_name(index_name, sizeof index_name, - index->name, TRUE); - - push_warning_printf(thd, + push_warning_printf(m_user_thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_INDEX_CORRUPT, "InnoDB: Index %s is marked as" - " corrupted", index_name); + " corrupted", + index->name()); /* Now that the table is already marked as corrupted, there is no need to check any index of this table */ - prebuilt->trx->op_info = ""; + m_prebuilt->trx->op_info = ""; DBUG_RETURN(HA_ADMIN_CORRUPT); } - prebuilt->trx->op_info = "checking table"; - - old_isolation_level = prebuilt->trx->isolation_level; + old_isolation_level = m_prebuilt->trx->isolation_level; /* We must run the index record counts at an isolation level >= READ COMMITTED, because a dirty read can see a wrong number of records in some index; to play safe, we use always REPEATABLE READ here */ + m_prebuilt->trx->isolation_level = TRX_ISO_REPEATABLE_READ; - prebuilt->trx->isolation_level = TRX_ISO_REPEATABLE_READ; - - /* Check whether the table is already marked as corrupted - before running the check table */ - table_corrupted = prebuilt->table->corrupted; + ut_ad(!m_prebuilt->table->corrupted); - /* Reset table->corrupted bit so that check table can proceed to - do additional check */ - prebuilt->table->corrupted = FALSE; - - for (index = dict_table_get_first_index(prebuilt->table); + for (index = dict_table_get_first_index(m_prebuilt->table); index != NULL; index = dict_table_get_next_index(index)) { - char index_name[MAX_FULL_NAME_LEN + 1]; - /* If this is an index being created or dropped, skip */ - if (*index->name == TEMP_INDEX_PREFIX) { + if (!index->is_committed()) { continue; } - if (!(check_opt->flags & T_QUICK)) { + if (!(check_opt->flags & T_QUICK) + && !index->is_corrupted()) { /* Enlarge the fatal lock wait timeout during CHECK TABLE. */ - os_increment_counter_by_amount( - server_mutex, - srv_fatal_semaphore_wait_threshold, + my_atomic_addlong( + &srv_fatal_semaphore_wait_threshold, SRV_SEMAPHORE_WAIT_EXTENSION); - dberr_t err = btr_validate_index(index, prebuilt->trx); + + dberr_t err = btr_validate_index( + index, m_prebuilt->trx, false); /* Restore the fatal lock wait timeout after CHECK TABLE. */ - os_decrement_counter_by_amount( - server_mutex, - srv_fatal_semaphore_wait_threshold, - SRV_SEMAPHORE_WAIT_EXTENSION); + my_atomic_addlong( + &srv_fatal_semaphore_wait_threshold, + -SRV_SEMAPHORE_WAIT_EXTENSION); if (err != DB_SUCCESS) { is_ok = false; - innobase_format_name( - index_name, sizeof index_name, - index->name, TRUE); - if (err == DB_DECRYPTION_FAILED) { push_warning_printf( thd, @@ -14109,7 +14721,7 @@ ha_innobase::check( "Table %s is encrypted but encryption service or" " used key_id is not available. " " Can't continue checking table.", - index->table->name); + index->table->name.m_name); } else { push_warning_printf( thd, @@ -14117,7 +14729,7 @@ ha_innobase::check( ER_NOT_KEYFILE, "InnoDB: The B-tree of" " index %s is corrupted.", - index_name); + index->name()); } continue; @@ -14127,131 +14739,117 @@ ha_innobase::check( /* Instead of invoking change_active_index(), set up a dummy template for non-locking reads, disabling access to the clustered index. */ - prebuilt->index = index; + m_prebuilt->index = index; - prebuilt->index_usable = row_merge_is_index_usable( - prebuilt->trx, prebuilt->index); + m_prebuilt->index_usable = row_merge_is_index_usable( + m_prebuilt->trx, m_prebuilt->index); DBUG_EXECUTE_IF( "dict_set_index_corrupted", - if (!dict_index_is_clust(index)) { - prebuilt->index_usable = FALSE; - row_mysql_lock_data_dictionary(prebuilt->trx); - dict_set_corrupted(index, prebuilt->trx, "dict_set_index_corrupted"); - row_mysql_unlock_data_dictionary(prebuilt->trx); + if (!index->is_primary()) { + m_prebuilt->index_usable = FALSE; + // row_mysql_lock_data_dictionary(m_prebuilt->trx); + dict_set_corrupted(index, m_prebuilt->trx, "dict_set_index_corrupted"); + // row_mysql_unlock_data_dictionary(m_prebuilt->trx); }); - if (UNIV_UNLIKELY(!prebuilt->index_usable)) { - innobase_format_name( - index_name, sizeof index_name, - prebuilt->index->name, TRUE); - - if (dict_index_is_corrupted(prebuilt->index)) { + if (UNIV_UNLIKELY(!m_prebuilt->index_usable)) { + if (index->is_corrupted()) { push_warning_printf( - user_thd, + m_user_thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_INDEX_CORRUPT, "InnoDB: Index %s is marked as" " corrupted", - index_name); + index->name()); is_ok = false; } else { push_warning_printf( - thd, + m_user_thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_TABLE_DEF_CHANGED, "InnoDB: Insufficient history for" " index %s", - index_name); + index->name()); } continue; } - prebuilt->sql_stat_start = TRUE; - prebuilt->template_type = ROW_MYSQL_DUMMY_TEMPLATE; - prebuilt->n_template = 0; - prebuilt->need_to_access_clustered = FALSE; + m_prebuilt->sql_stat_start = TRUE; + m_prebuilt->template_type = ROW_MYSQL_DUMMY_TEMPLATE; + m_prebuilt->n_template = 0; + m_prebuilt->need_to_access_clustered = FALSE; - dtuple_set_n_fields(prebuilt->search_tuple, 0); + dtuple_set_n_fields(m_prebuilt->search_tuple, 0); - prebuilt->select_lock_type = LOCK_NONE; + m_prebuilt->select_lock_type = LOCK_NONE; - if (!row_check_index_for_mysql(prebuilt, index, &n_rows)) { - innobase_format_name( - index_name, sizeof index_name, - index->name, TRUE); + /* Scan this index. */ + if (dict_index_is_spatial(index)) { + ret = row_count_rtree_recs(m_prebuilt, &n_rows); + } else { + ret = row_scan_index_for_mysql( + m_prebuilt, index, &n_rows); + } + + DBUG_EXECUTE_IF( + "dict_set_index_corrupted", + if (!index->is_primary()) { + ret = DB_CORRUPTION; + }); + if (ret == DB_INTERRUPTED || thd_killed(m_user_thd)) { + /* Do not report error since this could happen + during shutdown */ + break; + } + if (ret != DB_SUCCESS) { + /* Assume some kind of corruption. */ push_warning_printf( thd, Sql_condition::WARN_LEVEL_WARN, ER_NOT_KEYFILE, "InnoDB: The B-tree of" " index %s is corrupted.", - index_name); + index->name()); is_ok = false; dict_set_corrupted( - index, prebuilt->trx, "CHECK TABLE-check index"); + index, m_prebuilt->trx, "CHECK TABLE-check index"); } - if (thd_kill_level(user_thd)) { - break; - } - -#if 0 - fprintf(stderr, "%lu entries in index %s\n", n_rows, - index->name); -#endif - if (index == dict_table_get_first_index(prebuilt->table)) { + if (index == dict_table_get_first_index(m_prebuilt->table)) { n_rows_in_table = n_rows; } else if (!(index->type & DICT_FTS) && (n_rows != n_rows_in_table)) { push_warning_printf( thd, Sql_condition::WARN_LEVEL_WARN, ER_NOT_KEYFILE, - "InnoDB: Index '%-.200s' contains %lu" - " entries, should be %lu.", - index->name, - (ulong) n_rows, - (ulong) n_rows_in_table); + "InnoDB: Index '%-.200s' contains " ULINTPF + " entries, should be " ULINTPF ".", + index->name(), n_rows, n_rows_in_table); is_ok = false; dict_set_corrupted( - index, prebuilt->trx, + index, m_prebuilt->trx, "CHECK TABLE; Wrong count"); } } - if (table_corrupted) { - /* If some previous operation has marked the table as - corrupted in memory, and has not propagated such to - clustered index, we will do so here */ - index = dict_table_get_first_index(prebuilt->table); - - if (!dict_index_is_corrupted(index)) { - dict_set_corrupted( - index, prebuilt->trx, "CHECK TABLE"); - } - prebuilt->table->corrupted = TRUE; - } - /* Restore the original isolation level */ - prebuilt->trx->isolation_level = old_isolation_level; - + m_prebuilt->trx->isolation_level = old_isolation_level; +#ifdef BTR_CUR_HASH_ADAPT +# if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG /* We validate the whole adaptive hash index for all tables at every CHECK TABLE only when QUICK flag is not present. */ -#if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG if (!(check_opt->flags & T_QUICK) && !btr_search_validate()) { push_warning(thd, Sql_condition::WARN_LEVEL_WARN, ER_NOT_KEYFILE, "InnoDB: The adaptive hash index is corrupted."); is_ok = false; } -#endif /* defined UNIV_AHI_DEBUG || defined UNIV_DEBUG */ - - prebuilt->trx->op_info = ""; - if (thd_kill_level(user_thd)) { - my_error(ER_QUERY_INTERRUPTED, MYF(0)); - } +# endif /* defined UNIV_AHI_DEBUG || defined UNIV_DEBUG */ +#endif /* BTR_CUR_HASH_ADAPT */ + m_prebuilt->trx->op_info = ""; DBUG_RETURN(is_ok ? HA_ADMIN_OK : HA_ADMIN_CORRUPT); } @@ -14269,7 +14867,7 @@ ha_innobase::update_table_comment( { uint length = (uint) strlen(comment); char* str=0; - long flen; + size_t flen; std::string fk_str; /* We do not know if MySQL can call this function before calling @@ -14282,36 +14880,26 @@ ha_innobase::update_table_comment( update_thd(ha_thd()); - prebuilt->trx->op_info = (char*)"returning table comment"; - - /* In case MySQL calls this in the middle of a SELECT query, release - possible adaptive hash latch to avoid deadlocks of threads */ - - trx_search_latch_release_if_reserved(prebuilt->trx); + m_prebuilt->trx->op_info = "returning table comment"; #define SSTR( x ) reinterpret_cast< std::ostringstream & >( \ ( std::ostringstream() << std::dec << x ) ).str() fk_str.append("InnoDB free: "); fk_str.append(SSTR(fsp_get_available_space_in_free_extents( - prebuilt->table->space))); + m_prebuilt->table->space))); fk_str.append(dict_print_info_on_foreign_keys( - FALSE, prebuilt->trx, - prebuilt->table)); + FALSE, m_prebuilt->trx, + m_prebuilt->table)); flen = fk_str.length(); - if (flen < 0) { - flen = 0; - } else if (length + flen + 3 > 64000) { + if (length + flen + 3 > 64000) { flen = 64000 - 3 - length; } - /* allocate buffer for the full string */ - str = (char*) my_malloc(length + flen + 3, MYF(0)); - if (str) { char* pos = str + length; if (length) { @@ -14319,12 +14907,11 @@ ha_innobase::update_table_comment( *pos++ = ';'; *pos++ = ' '; } - memcpy(pos, fk_str.c_str(), flen); pos[flen] = 0; } - prebuilt->trx->op_info = (char*)""; + m_prebuilt->trx->op_info = (char*)""; return(str ? str : (char*) comment); } @@ -14334,14 +14921,12 @@ Gets the foreign key create info for a table stored in InnoDB. @return own: character string in the form which can be inserted to the CREATE TABLE statement, MUST be freed with ha_innobase::free_foreign_key_create_info */ -UNIV_INTERN + char* ha_innobase::get_foreign_key_create_info(void) /*==========================================*/ { - char* fk_str = 0; - - ut_a(prebuilt != NULL); + ut_a(m_prebuilt != NULL); /* We do not know if MySQL can call this function before calling external_lock(). To be safe, update the thd of the current table @@ -14349,23 +14934,24 @@ ha_innobase::get_foreign_key_create_info(void) update_thd(ha_thd()); - prebuilt->trx->op_info = (char*)"getting info on foreign keys"; + m_prebuilt->trx->op_info = "getting info on foreign keys"; - /* In case MySQL calls this in the middle of a SELECT query, - release possible adaptive hash latch to avoid - deadlocks of threads */ - - trx_search_latch_release_if_reserved(prebuilt->trx); - - /* Output the data to a temporary file */ + /* Output the data to a temporary string */ std::string str = dict_print_info_on_foreign_keys( - TRUE, prebuilt->trx, - prebuilt->table); + TRUE, m_prebuilt->trx, + m_prebuilt->table); - prebuilt->trx->op_info = (char*)""; + m_prebuilt->trx->op_info = ""; /* Allocate buffer for the string */ - fk_str = (char*) my_malloc(str.length() + 1, MYF(0)); + char* fk_str = (char*) my_malloc(str.length() + 1, MYF(0)); + + /* JAN: TODO: MySQL 5.7 + fk_str = reinterpret_cast<char*>( + my_malloc(PSI_INSTRUMENT_ME, str.length() + 1, MYF(0))); + */ + + if (fk_str) { memcpy(fk_str, str.c_str(), str.length()); @@ -14383,13 +14969,13 @@ static FOREIGN_KEY_INFO* get_foreign_key_info( /*=================*/ - THD* thd, /*!< in: user thread handle */ - dict_foreign_t* foreign) /*!< in: foreign key constraint */ + THD* thd, /*!< in: user thread handle */ + dict_foreign_t* foreign)/*!< in: foreign key constraint */ { FOREIGN_KEY_INFO f_key_info; FOREIGN_KEY_INFO* pf_key_info; uint i = 0; - ulint len; + size_t len; char tmp_buff[NAME_LEN+1]; char name_buff[NAME_LEN+1]; const char* ptr; @@ -14397,8 +14983,8 @@ get_foreign_key_info( LEX_STRING* name = NULL; ptr = dict_remove_db_name(foreign->id); - f_key_info.foreign_id = thd_make_lex_string(thd, 0, ptr, - (uint) strlen(ptr), 1); + f_key_info.foreign_id = thd_make_lex_string( + thd, 0, ptr, strlen(ptr), 1); /* Name format: database name, '/', table name, '\0' */ @@ -14410,13 +14996,13 @@ get_foreign_key_info( len = filename_to_tablename(tmp_buff, name_buff, sizeof(name_buff)); f_key_info.referenced_db = thd_make_lex_string( - thd, 0, name_buff, static_cast<unsigned int>(len), 1); + thd, 0, name_buff, len, 1); /* Referenced (parent) table name */ ptr = dict_remove_db_name(foreign->referenced_table_name); len = filename_to_tablename(ptr, name_buff, sizeof(name_buff), 1); f_key_info.referenced_table = thd_make_lex_string( - thd, 0, name_buff, static_cast<unsigned int>(len), 1); + thd, 0, name_buff, len, 1); /* Dependent (child) database name */ len = dict_get_db_name_len(foreign->foreign_table_name); @@ -14426,22 +15012,22 @@ get_foreign_key_info( len = filename_to_tablename(tmp_buff, name_buff, sizeof(name_buff)); f_key_info.foreign_db = thd_make_lex_string( - thd, 0, name_buff, static_cast<unsigned int>(len), 1); + thd, 0, name_buff, len, 1); /* Dependent (child) table name */ ptr = dict_remove_db_name(foreign->foreign_table_name); len = filename_to_tablename(ptr, name_buff, sizeof(name_buff), 1); f_key_info.foreign_table = thd_make_lex_string( - thd, 0, name_buff, static_cast<unsigned int>(len), 1); + thd, 0, name_buff, len, 1); do { ptr = foreign->foreign_col_names[i]; name = thd_make_lex_string(thd, name, ptr, - (uint) strlen(ptr), 1); + strlen(ptr), 1); f_key_info.foreign_fields.push_back(name); ptr = foreign->referenced_col_names[i]; name = thd_make_lex_string(thd, name, ptr, - (uint) strlen(ptr), 1); + strlen(ptr), 1); f_key_info.referenced_fields.push_back(name); } while (++i < foreign->n_fields); @@ -14455,6 +15041,7 @@ get_foreign_key_info( f_key_info.delete_method = FK_OPTION_RESTRICT; } + if (foreign->type & DICT_FOREIGN_ON_UPDATE_CASCADE) { f_key_info.update_method = FK_OPTION_CASCADE; } else if (foreign->type & DICT_FOREIGN_ON_UPDATE_SET_NULL) { @@ -14465,12 +15052,40 @@ get_foreign_key_info( f_key_info.update_method = FK_OPTION_RESTRICT; } - if (foreign->referenced_index && foreign->referenced_index->name) { - referenced_key_name = thd_make_lex_string(thd, - f_key_info.referenced_key_name, - foreign->referenced_index->name, - (uint) strlen(foreign->referenced_index->name), - 1); + /* Load referenced table to update FK referenced key name. */ + if (foreign->referenced_table == NULL) { + + dict_table_t* ref_table; + + ut_ad(mutex_own(&dict_sys->mutex)); + ref_table = dict_table_open_on_name( + foreign->referenced_table_name_lookup, + TRUE, FALSE, DICT_ERR_IGNORE_NONE); + + if (ref_table == NULL) { + + if (!thd_test_options( + thd, OPTION_NO_FOREIGN_KEY_CHECKS)) { + ib::info() + << "Foreign Key referenced table " + << foreign->referenced_table_name + << " not found for foreign table " + << foreign->foreign_table_name; + } + } else { + + dict_table_close(ref_table, TRUE, FALSE); + } + } + + if (foreign->referenced_index + && foreign->referenced_index->name != NULL) { + referenced_key_name = thd_make_lex_string( + thd, + f_key_info.referenced_key_name, + foreign->referenced_index->name, + strlen(foreign->referenced_index->name), + 1); } else { referenced_key_name = NULL; } @@ -14486,41 +15101,37 @@ get_foreign_key_info( /*******************************************************************//** Gets the list of foreign keys in this table. @return always 0, that is, always succeeds */ -UNIV_INTERN + int ha_innobase::get_foreign_key_list( /*==============================*/ THD* thd, /*!< in: user thread handle */ List<FOREIGN_KEY_INFO>* f_key_list) /*!< out: foreign key list */ { - FOREIGN_KEY_INFO* pf_key_info; - dict_foreign_t* foreign; - - ut_a(prebuilt != NULL); update_thd(ha_thd()); - prebuilt->trx->op_info = "getting list of foreign keys"; + m_prebuilt->trx->op_info = "getting list of foreign keys"; - trx_search_latch_release_if_reserved(prebuilt->trx); - - mutex_enter(&(dict_sys->mutex)); + mutex_enter(&dict_sys->mutex); for (dict_foreign_set::iterator it - = prebuilt->table->foreign_set.begin(); - it != prebuilt->table->foreign_set.end(); + = m_prebuilt->table->foreign_set.begin(); + it != m_prebuilt->table->foreign_set.end(); ++it) { - foreign = *it; + FOREIGN_KEY_INFO* pf_key_info; + dict_foreign_t* foreign = *it; pf_key_info = get_foreign_key_info(thd, foreign); - if (pf_key_info) { + + if (pf_key_info != NULL) { f_key_list->push_back(pf_key_info); } } - mutex_exit(&(dict_sys->mutex)); + mutex_exit(&dict_sys->mutex); - prebuilt->trx->op_info = ""; + m_prebuilt->trx->op_info = ""; return(0); } @@ -14528,41 +15139,202 @@ ha_innobase::get_foreign_key_list( /*******************************************************************//** Gets the set of foreign keys where this table is the referenced table. @return always 0, that is, always succeeds */ -UNIV_INTERN + int ha_innobase::get_parent_foreign_key_list( /*=====================================*/ THD* thd, /*!< in: user thread handle */ List<FOREIGN_KEY_INFO>* f_key_list) /*!< out: foreign key list */ { - FOREIGN_KEY_INFO* pf_key_info; - dict_foreign_t* foreign; - - ut_a(prebuilt != NULL); update_thd(ha_thd()); - prebuilt->trx->op_info = "getting list of referencing foreign keys"; + m_prebuilt->trx->op_info = "getting list of referencing foreign keys"; - trx_search_latch_release_if_reserved(prebuilt->trx); - - mutex_enter(&(dict_sys->mutex)); + mutex_enter(&dict_sys->mutex); for (dict_foreign_set::iterator it - = prebuilt->table->referenced_set.begin(); - it != prebuilt->table->referenced_set.end(); + = m_prebuilt->table->referenced_set.begin(); + it != m_prebuilt->table->referenced_set.end(); ++it) { - foreign = *it; + FOREIGN_KEY_INFO* pf_key_info; + dict_foreign_t* foreign = *it; pf_key_info = get_foreign_key_info(thd, foreign); - if (pf_key_info) { + + if (pf_key_info != NULL) { f_key_list->push_back(pf_key_info); } } - mutex_exit(&(dict_sys->mutex)); + mutex_exit(&dict_sys->mutex); + + m_prebuilt->trx->op_info = ""; + + return(0); +} + +/** Table list item structure is used to store only the table +and name. It is used by get_cascade_foreign_key_table_list to store +the intermediate result for fetching the table set. */ +struct table_list_item { + /** InnoDB table object */ + const dict_table_t* table; + /** Table name */ + const char* name; +}; + +/** Structure to compare two st_tablename objects using their +db and tablename. It is used in the ordering of cascade_fk_set. +It returns true if the first argument precedes the second argument +and false otherwise. */ +struct tablename_compare { + + bool operator()(const st_handler_tablename lhs, + const st_handler_tablename rhs) const + { + int cmp = strcmp(lhs.db, rhs.db); + if (cmp == 0) { + cmp = strcmp(lhs.tablename, rhs.tablename); + } + + return(cmp < 0); + } +}; + +/** Get the table name and database name for the given table. +@param[in,out] thd user thread handle +@param[out] f_key_info pointer to table_name_info object +@param[in] foreign foreign key constraint. */ +static +void +get_table_name_info( + THD* thd, + st_handler_tablename* f_key_info, + const dict_foreign_t* foreign) +{ +#define FILENAME_CHARSET_MBMAXLEN 5 + char tmp_buff[NAME_CHAR_LEN * FILENAME_CHARSET_MBMAXLEN + 1]; + char name_buff[NAME_CHAR_LEN * FILENAME_CHARSET_MBMAXLEN + 1]; + const char* ptr; + + size_t len = dict_get_db_name_len( + foreign->referenced_table_name_lookup); + ut_memcpy(tmp_buff, foreign->referenced_table_name_lookup, len); + tmp_buff[len] = 0; + + ut_ad(len < sizeof(tmp_buff)); + + len = filename_to_tablename(tmp_buff, name_buff, sizeof(name_buff)); + f_key_info->db = thd_strmake(thd, name_buff, len); + + ptr = dict_remove_db_name(foreign->referenced_table_name_lookup); + len = filename_to_tablename(ptr, name_buff, sizeof(name_buff)); + f_key_info->tablename = thd_strmake(thd, name_buff, len); +} + +/** Get the list of tables ordered by the dependency on the other tables using +the 'CASCADE' foreign key constraint. +@param[in,out] thd user thread handle +@param[out] fk_table_list set of tables name info for the + dependent table +@retval 0 for success. */ +int +ha_innobase::get_cascade_foreign_key_table_list( + THD* thd, + List<st_handler_tablename>* fk_table_list) +{ + m_prebuilt->trx->op_info = "getting cascading foreign keys"; + + std::list<table_list_item, ut_allocator<table_list_item> > table_list; + + typedef std::set<st_handler_tablename, tablename_compare, + ut_allocator<st_handler_tablename> > cascade_fk_set; + + cascade_fk_set fk_set; + + mutex_enter(&dict_sys->mutex); + + /* Initialize the table_list with prebuilt->table name. */ + struct table_list_item item = {m_prebuilt->table, + m_prebuilt->table->name.m_name}; + + table_list.push_back(item); + + /* Get the parent table, grand parent table info from the + table list by depth-first traversal. */ + do { + const dict_table_t* parent_table; + dict_table_t* parent = NULL; + std::pair<cascade_fk_set::iterator,bool> ret; + + item = table_list.back(); + table_list.pop_back(); + parent_table = item.table; + + if (parent_table == NULL) { + + ut_ad(item.name != NULL); + + parent_table = parent = dict_table_open_on_name( + item.name, TRUE, FALSE, + DICT_ERR_IGNORE_NONE); + + if (parent_table == NULL) { + /* foreign_key_checks is or was probably + disabled; ignore the constraint */ + continue; + } + } + + for (dict_foreign_set::const_iterator it = + parent_table->foreign_set.begin(); + it != parent_table->foreign_set.end(); ++it) { + + const dict_foreign_t* foreign = *it; + st_handler_tablename f1; - prebuilt->trx->op_info = ""; + /* Skip the table if there is no + cascading operation. */ + if (0 == (foreign->type + & ~(DICT_FOREIGN_ON_DELETE_NO_ACTION + | DICT_FOREIGN_ON_UPDATE_NO_ACTION))) { + continue; + } + + if (foreign->referenced_table_name_lookup != NULL) { + get_table_name_info(thd, &f1, foreign); + ret = fk_set.insert(f1); + + /* Ignore the table if it is already + in the set. */ + if (!ret.second) { + continue; + } + + struct table_list_item item1 = { + foreign->referenced_table, + foreign->referenced_table_name_lookup}; + + table_list.push_back(item1); + + st_handler_tablename* fk_table = + (st_handler_tablename*) thd_memdup( + thd, &f1, sizeof(*fk_table)); + + fk_table_list->push_back(fk_table); + } + } + + if (parent != NULL) { + dict_table_close(parent, true, false); + } + + } while(!table_list.empty()); + + mutex_exit(&dict_sys->mutex); + + m_prebuilt->trx->op_info = ""; return(0); } @@ -14571,26 +15343,26 @@ ha_innobase::get_parent_foreign_key_list( Checks if ALTER TABLE may change the storage engine of the table. Changing storage engines is not allowed for tables for which there are foreign key constraints (parent or child tables). -@return TRUE if can switch engines */ -UNIV_INTERN +@return TRUE if can switch engines */ + bool ha_innobase::can_switch_engines(void) /*=================================*/ { - bool can_switch; - DBUG_ENTER("ha_innobase::can_switch_engines"); + update_thd(); - prebuilt->trx->op_info = + m_prebuilt->trx->op_info = "determining if there are foreign key constraints"; - row_mysql_freeze_data_dictionary(prebuilt->trx); - can_switch = prebuilt->table->referenced_set.empty() - && prebuilt->table->foreign_set.empty(); + row_mysql_freeze_data_dictionary(m_prebuilt->trx); - row_mysql_unfreeze_data_dictionary(prebuilt->trx); - prebuilt->trx->op_info = ""; + bool can_switch = m_prebuilt->table->referenced_set.empty() + && m_prebuilt->table->foreign_set.empty(); + + row_mysql_unfreeze_data_dictionary(m_prebuilt->trx); + m_prebuilt->trx->op_info = ""; DBUG_RETURN(can_switch); } @@ -14600,13 +15372,13 @@ Checks if a table is referenced by a foreign key. The MySQL manual states that a REPLACE is either equivalent to an INSERT, or DELETE(s) + INSERT. Only a delete is then allowed internally to resolve a duplicate key conflict in REPLACE, not an update. -@return > 0 if referenced by a FOREIGN KEY */ -UNIV_INTERN +@return > 0 if referenced by a FOREIGN KEY */ + uint ha_innobase::referenced_by_foreign_key(void) /*========================================*/ { - if (dict_table_is_referenced_by_foreign_key(prebuilt->table)) { + if (dict_table_is_referenced_by_foreign_key(m_prebuilt->table)) { return(1); } @@ -14617,21 +15389,21 @@ ha_innobase::referenced_by_foreign_key(void) /*******************************************************************//** Frees the foreign key create info for a table stored in InnoDB, if it is non-NULL. */ -UNIV_INTERN + void ha_innobase::free_foreign_key_create_info( /*======================================*/ char* str) /*!< in, own: create info string to free */ { - if (str) { + if (str != NULL) { my_free(str); } } /*******************************************************************//** Tells something additional to the handler about how to do things. -@return 0 or error number */ -UNIV_INTERN +@return 0 or error number */ + int ha_innobase::extra( /*===============*/ @@ -14641,13 +15413,13 @@ ha_innobase::extra( check_trx_exists(ha_thd()); /* Warning: since it is not sure that MySQL calls external_lock - before calling this function, the trx field in prebuilt can be + before calling this function, the trx field in m_prebuilt can be obsolete! */ switch (operation) { case HA_EXTRA_FLUSH: - if (prebuilt->blob_heap) { - row_mysql_prebuilt_free_blob_heap(prebuilt); + if (m_prebuilt->blob_heap) { + row_mysql_prebuilt_free_blob_heap(m_prebuilt); } break; case HA_EXTRA_RESET_STATE: @@ -14655,16 +15427,16 @@ ha_innobase::extra( thd_to_trx(ha_thd())->duplicates = 0; break; case HA_EXTRA_NO_KEYREAD: - prebuilt->read_just_key = 0; + m_prebuilt->read_just_key = 0; break; case HA_EXTRA_KEYREAD: - prebuilt->read_just_key = 1; + m_prebuilt->read_just_key = 1; break; case HA_EXTRA_KEYREAD_PRESERVE_FIELDS: - prebuilt->keep_other_fields_on_keyread = 1; + m_prebuilt->keep_other_fields_on_keyread = 1; break; - /* IMPORTANT: prebuilt->trx can be obsolete in + /* IMPORTANT: m_prebuilt->trx can be obsolete in this method, because it is not sure that MySQL calls external_lock before this method with the parameters below. We must not invoke update_thd() @@ -14682,6 +15454,16 @@ ha_innobase::extra( case HA_EXTRA_WRITE_CANNOT_REPLACE: thd_to_trx(ha_thd())->duplicates &= ~TRX_DUP_REPLACE; break; + case HA_EXTRA_BEGIN_ALTER_COPY: + m_prebuilt->table->skip_alter_undo = 1; + break; + case HA_EXTRA_END_ALTER_COPY: + m_prebuilt->table->skip_alter_undo = 0; + break; + case HA_EXTRA_FAKE_START_STMT: + trx_register_for_2pc(m_prebuilt->trx); + m_prebuilt->sql_stat_start = true; + break; default:/* Do nothing */ ; } @@ -14689,29 +15471,40 @@ ha_innobase::extra( return(0); } -/******************************************************************//** -*/ -UNIV_INTERN +/** +MySQL calls this method at the end of each statement. This method +exists for readability only. ha_innobase::reset() doesn't give any +clue about the method. */ + int -ha_innobase::reset() -/*================*/ +ha_innobase::end_stmt() { - if (prebuilt->blob_heap) { - row_mysql_prebuilt_free_blob_heap(prebuilt); + if (m_prebuilt->blob_heap) { + row_mysql_prebuilt_free_blob_heap(m_prebuilt); } reset_template(); - ds_mrr.dsmrr_close(); + + m_ds_mrr.dsmrr_close(); /* TODO: This should really be reset in reset_template() but for now it's safer to do it explicitly here. */ /* This is a statement level counter. */ - prebuilt->autoinc_last_value = 0; + m_prebuilt->autoinc_last_value = 0; return(0); } +/** +MySQL calls this method at the end of each statement */ + +int +ha_innobase::reset() +{ + return(end_stmt()); +} + /******************************************************************//** MySQL calls this function at the start of each SQL statement inside LOCK TABLES. Inside LOCK TABLES the ::external_lock method does not work to @@ -14723,42 +15516,36 @@ procedure. To make the execution more deterministic for binlogging, MySQL-5.0 locks all tables involved in a stored procedure with full explicit table locks (thd_in_lock_tables(thd) holds in store_lock()) before executing the procedure. -@return 0 or error code */ -UNIV_INTERN +@return 0 or error code */ + int ha_innobase::start_stmt( /*====================*/ THD* thd, /*!< in: handle to the user thread */ thr_lock_type lock_type) { - trx_t* trx; + trx_t* trx = m_prebuilt->trx; + DBUG_ENTER("ha_innobase::start_stmt"); update_thd(thd); - trx = prebuilt->trx; - - /* Here we release the search latch and the InnoDB thread FIFO ticket - if they were reserved. They should have been released already at the - end of the previous statement, but because inside LOCK TABLES the - lock count method does not work to mark the end of a SELECT statement, - that may not be the case. We MUST release the search latch before an - INSERT, for example. */ + ut_ad(m_prebuilt->table != NULL); - trx_search_latch_release_if_reserved(trx); + trx = m_prebuilt->trx; innobase_srv_conc_force_exit_innodb(trx); /* Reset the AUTOINC statement level counter for multi-row INSERTs. */ trx->n_autoinc_rows = 0; - prebuilt->sql_stat_start = TRUE; - prebuilt->hint_need_to_fetch_extra_cols = 0; + m_prebuilt->sql_stat_start = TRUE; + m_prebuilt->hint_need_to_fetch_extra_cols = 0; reset_template(); - if (dict_table_is_temporary(prebuilt->table) - && prebuilt->mysql_has_locked - && prebuilt->select_lock_type == LOCK_NONE) { + if (dict_table_is_temporary(m_prebuilt->table) + && m_mysql_has_locked + && m_prebuilt->select_lock_type == LOCK_NONE) { dberr_t error; switch (thd_sql_command(thd)) { @@ -14767,12 +15554,12 @@ ha_innobase::start_stmt( case SQLCOM_DELETE: case SQLCOM_REPLACE: init_table_handle_for_HANDLER(); - prebuilt->select_lock_type = LOCK_X; - prebuilt->stored_select_lock_type = LOCK_X; - error = row_lock_table_for_mysql(prebuilt, NULL, 1); + m_prebuilt->select_lock_type = LOCK_X; + m_prebuilt->stored_select_lock_type = LOCK_X; + error = row_lock_table(m_prebuilt); if (error != DB_SUCCESS) { - int st = convert_error_code_to_mysql( + int st = convert_error_code_to_mysql( error, 0, thd); DBUG_RETURN(st); } @@ -14780,13 +15567,13 @@ ha_innobase::start_stmt( } } - if (!prebuilt->mysql_has_locked) { + if (!m_mysql_has_locked) { /* This handle is for a temporary table created inside this same LOCK TABLES; since MySQL does NOT call external_lock in this case, we must use x-row locks inside InnoDB to be prepared for an update of a row */ - prebuilt->select_lock_type = LOCK_X; + m_prebuilt->select_lock_type = LOCK_X; } else if (trx->isolation_level != TRX_ISO_SERIALIZABLE && thd_sql_command(thd) == SQLCOM_SELECT @@ -14795,18 +15582,19 @@ ha_innobase::start_stmt( /* For other than temporary tables, we obtain no lock for consistent read (plain SELECT). */ - prebuilt->select_lock_type = LOCK_NONE; + m_prebuilt->select_lock_type = LOCK_NONE; } else { /* Not a consistent read: restore the select_lock_type value. The value of stored_select_lock_type was decided in: 1) ::store_lock(), 2) ::external_lock(), - 3) ::init_table_handle_for_HANDLER(), and - 4) ::transactional_table_lock(). */ + 3) ::init_table_handle_for_HANDLER(). */ - ut_a(prebuilt->stored_select_lock_type != LOCK_NONE_UNSET); - prebuilt->select_lock_type = prebuilt->stored_select_lock_type; + ut_a(m_prebuilt->stored_select_lock_type != LOCK_NONE_UNSET); + + m_prebuilt->select_lock_type = + m_prebuilt->stored_select_lock_type; } *trx->detailed_error = 0; @@ -14822,7 +15610,7 @@ ha_innobase::start_stmt( /******************************************************************//** Maps a MySQL trx isolation level code to the InnoDB isolation level code -@return InnoDB isolation level */ +@return InnoDB isolation level */ static inline ulint innobase_map_isolation_level( @@ -14849,21 +15637,23 @@ the THD in the handle. We will also use this function to communicate to InnoDB that a new SQL statement has started and that we must store a savepoint to our transaction handle, so that we are able to roll back the SQL statement in case of an error. -@return 0 */ -UNIV_INTERN +@return 0 */ + int ha_innobase::external_lock( /*=======================*/ THD* thd, /*!< in: handle to the user thread */ int lock_type) /*!< in: lock type */ { - trx_t* trx; - DBUG_ENTER("ha_innobase::external_lock"); DBUG_PRINT("enter",("lock_type: %d", lock_type)); update_thd(thd); + trx_t* trx = m_prebuilt->trx; + + ut_ad(m_prebuilt->table); + /* Statement based binlogging does not work in isolation level READ UNCOMMITTED and READ COMMITTED since the necessary locks cannot be taken. In this case, we print an @@ -14876,18 +15666,22 @@ ha_innobase::external_lock( && thd_binlog_format(thd) == BINLOG_FORMAT_STMT && thd_binlog_filter_ok(thd) && thd_sqlcom_can_generate_row_events(thd)) { - bool skip = 0; + + bool skip = false; + /* used by test case */ DBUG_EXECUTE_IF("no_innodb_binlog_errors", skip = true;); + if (!skip) { #ifdef WITH_WSREP if (!wsrep_on(thd) || wsrep_thd_exec_mode(thd) == LOCAL_STATE) { #endif /* WITH_WSREP */ my_error(ER_BINLOG_STMT_MODE_AND_ROW_ENGINE, MYF(0), - " InnoDB is limited to row-logging when " - "transaction isolation level is " - "READ COMMITTED or READ UNCOMMITTED."); + " InnoDB is limited to row-logging when" + " transaction isolation level is" + " READ COMMITTED or READ UNCOMMITTED."); + DBUG_RETURN(HA_ERR_LOGGING_IMPOSSIBLE); #ifdef WITH_WSREP } @@ -14909,8 +15703,7 @@ ha_innobase::external_lock( || thd_sql_command(thd) == SQLCOM_DROP_INDEX || thd_sql_command(thd) == SQLCOM_DELETE)) { - if (thd_sql_command(thd) == SQLCOM_CREATE_TABLE) - { + if (thd_sql_command(thd) == SQLCOM_CREATE_TABLE) { ib_senderrf(thd, IB_LOG_LEVEL_WARN, ER_READ_ONLY_MODE); DBUG_RETURN(HA_ERR_TABLE_READONLY); @@ -14919,24 +15712,29 @@ ha_innobase::external_lock( ER_READ_ONLY_MODE); DBUG_RETURN(HA_ERR_TABLE_READONLY); } - } - trx = prebuilt->trx; - - prebuilt->sql_stat_start = TRUE; - prebuilt->hint_need_to_fetch_extra_cols = 0; + m_prebuilt->sql_stat_start = TRUE; + m_prebuilt->hint_need_to_fetch_extra_cols = 0; reset_template(); - switch (prebuilt->table->quiesce) { + switch (m_prebuilt->table->quiesce) { case QUIESCE_START: /* Check for FLUSH TABLE t WITH READ LOCK; */ if (!srv_read_only_mode && thd_sql_command(thd) == SQLCOM_FLUSH && lock_type == F_RDLCK) { - row_quiesce_table_start(prebuilt->table, trx); + if (dict_table_is_discarded(m_prebuilt->table)) { + ib_senderrf(trx->mysql_thd, IB_LOG_LEVEL_ERROR, + ER_TABLESPACE_DISCARDED, + table->s->table_name.str); + + DBUG_RETURN(HA_ERR_NO_SUCH_TABLE); + } + + row_quiesce_table_start(m_prebuilt->table, trx); /* Use the transaction instance to track UNLOCK TABLES. It can be done via START TRANSACTION; too @@ -14952,7 +15750,7 @@ ha_innobase::external_lock( if (trx->flush_tables > 0 && (lock_type == F_UNLCK || trx_is_interrupted(trx))) { - row_quiesce_table_complete(prebuilt->table, trx); + row_quiesce_table_complete(m_prebuilt->table, trx); ut_a(trx->flush_tables > 0); --trx->flush_tables; @@ -14968,8 +15766,8 @@ ha_innobase::external_lock( /* If this is a SELECT, then it is in UPDATE TABLE ... or SELECT ... FOR UPDATE */ - prebuilt->select_lock_type = LOCK_X; - prebuilt->stored_select_lock_type = LOCK_X; + m_prebuilt->select_lock_type = LOCK_X; + m_prebuilt->stored_select_lock_type = LOCK_X; } if (lock_type != F_UNLCK) { @@ -14980,7 +15778,7 @@ ha_innobase::external_lock( innobase_register_trx(ht, thd, trx); if (trx->isolation_level == TRX_ISO_SERIALIZABLE - && prebuilt->select_lock_type == LOCK_NONE + && m_prebuilt->select_lock_type == LOCK_NONE && thd_test_options( thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { @@ -14992,8 +15790,8 @@ ha_innobase::external_lock( can be serialized also if performed as consistent reads. */ - prebuilt->select_lock_type = LOCK_S; - prebuilt->stored_select_lock_type = LOCK_S; + m_prebuilt->select_lock_type = LOCK_S; + m_prebuilt->stored_select_lock_type = LOCK_S; } /* Starting from 4.1.9, no InnoDB table lock is taken in LOCK @@ -15007,17 +15805,17 @@ ha_innobase::external_lock( can hold in some cases, e.g., at the start of a stored procedure call (SQLCOM_CALL). */ - if (prebuilt->select_lock_type != LOCK_NONE) { + if (m_prebuilt->select_lock_type != LOCK_NONE) { if (thd_sql_command(thd) == SQLCOM_LOCK_TABLES && THDVAR(thd, table_locks) && thd_test_options(thd, OPTION_NOT_AUTOCOMMIT) && thd_in_lock_tables(thd)) { - dberr_t error = row_lock_table_for_mysql( - prebuilt, NULL, 0); + dberr_t error = row_lock_table(m_prebuilt); if (error != DB_SUCCESS) { + DBUG_RETURN( convert_error_code_to_mysql( error, 0, thd)); @@ -15028,28 +15826,24 @@ ha_innobase::external_lock( } trx->n_mysql_tables_in_use++; - prebuilt->mysql_has_locked = TRUE; + m_mysql_has_locked = true; if (!trx_is_started(trx) - && (prebuilt->select_lock_type != LOCK_NONE - || prebuilt->stored_select_lock_type != LOCK_NONE)) { + && (m_prebuilt->select_lock_type != LOCK_NONE + || m_prebuilt->stored_select_lock_type != LOCK_NONE)) { ++trx->will_lock; } DBUG_RETURN(0); + } else { + DEBUG_SYNC_C("ha_innobase_end_statement"); } /* MySQL is releasing a table lock */ trx->n_mysql_tables_in_use--; - prebuilt->mysql_has_locked = FALSE; - - /* Release a possible FIFO ticket and search latch. Since we - may reserve the trx_sys->mutex, we have to release the search - system latch first to obey the latching order. */ - - trx_search_latch_release_if_reserved(trx); + m_mysql_has_locked = false; innobase_srv_conc_force_exit_innodb(trx); @@ -15059,28 +15853,31 @@ ha_innobase::external_lock( if (trx->n_mysql_tables_in_use == 0) { trx->mysql_n_tables_locked = 0; - prebuilt->used_in_HANDLER = FALSE; + m_prebuilt->used_in_HANDLER = FALSE; if (!thd_test_options( thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { if (trx_is_started(trx)) { + innobase_commit(ht, thd, TRUE); } } else if (trx->isolation_level <= TRX_ISO_READ_COMMITTED - && trx->global_read_view) { + && MVCC::is_view_active(trx->read_view)) { - /* At low transaction isolation levels we let - each consistent read set its own snapshot */ + mutex_enter(&trx_sys->mutex); + + trx_sys->mvcc->view_close(trx->read_view, true); - read_view_close_for_mysql(trx); + mutex_exit(&trx_sys->mutex); } } if (!trx_is_started(trx) - && (prebuilt->select_lock_type != LOCK_NONE - || prebuilt->stored_select_lock_type != LOCK_NONE)) { + && lock_type != F_UNLCK + && (m_prebuilt->select_lock_type != LOCK_NONE + || m_prebuilt->stored_select_lock_type != LOCK_NONE)) { ++trx->will_lock; } @@ -15088,100 +15885,6 @@ ha_innobase::external_lock( DBUG_RETURN(0); } -/******************************************************************//** -With this function MySQL request a transactional lock to a table when -user issued query LOCK TABLES..WHERE ENGINE = InnoDB. -@return error code */ -UNIV_INTERN -int -ha_innobase::transactional_table_lock( -/*==================================*/ - THD* thd, /*!< in: handle to the user thread */ - int lock_type) /*!< in: lock type */ -{ - trx_t* trx; - - DBUG_ENTER("ha_innobase::transactional_table_lock"); - DBUG_PRINT("enter",("lock_type: %d", lock_type)); - - /* We do not know if MySQL can call this function before calling - external_lock(). To be safe, update the thd of the current table - handle. */ - - update_thd(thd); - - if (!thd_tablespace_op(thd)) { - - if (dict_table_is_discarded(prebuilt->table)) { - - ib_senderrf( - thd, IB_LOG_LEVEL_ERROR, - ER_TABLESPACE_DISCARDED, - table->s->table_name.str); - - } else if (!prebuilt->table->is_readable()) { - - ib_senderrf( - thd, IB_LOG_LEVEL_ERROR, - ER_TABLESPACE_MISSING, - table->s->table_name.str); - } - - DBUG_RETURN(HA_ERR_CRASHED); - } - - trx = prebuilt->trx; - - prebuilt->sql_stat_start = TRUE; - prebuilt->hint_need_to_fetch_extra_cols = 0; - - reset_template(); - - if (lock_type == F_WRLCK) { - prebuilt->select_lock_type = LOCK_X; - prebuilt->stored_select_lock_type = LOCK_X; - } else if (lock_type == F_RDLCK) { - prebuilt->select_lock_type = LOCK_S; - prebuilt->stored_select_lock_type = LOCK_S; - } else { - ib_logf(IB_LOG_LEVEL_ERROR, - "MySQL is trying to set transactional table lock " - "with corrupted lock type to table %s, lock type " - "%d does not exist.", - table->s->table_name.str, lock_type); - - DBUG_RETURN(HA_ERR_CRASHED); - } - - /* MySQL is setting a new transactional table lock */ - - innobase_register_trx(ht, thd, trx); - - if (THDVAR(thd, table_locks) && thd_in_lock_tables(thd)) { - dberr_t error; - - error = row_lock_table_for_mysql(prebuilt, NULL, 0); - - if (error != DB_SUCCESS) { - DBUG_RETURN( - convert_error_code_to_mysql( - error, prebuilt->table->flags, thd)); - } - - if (thd_test_options( - thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN)) { - - /* Store the current undo_no of the transaction - so that we know where to roll back if we have - to roll back the next SQL statement */ - - trx_mark_sql_stat_end(trx); - } - } - - DBUG_RETURN(0); -} - /************************************************************************//** Here we export InnoDB status variables to MySQL. */ static @@ -15189,7 +15892,7 @@ void innodb_export_status() /*==================*/ { - if (innodb_inited) { + if (srv_was_started) { srv_export_innodb_status(); } } @@ -15206,7 +15909,6 @@ innodb_show_status( THD* thd, /*!< in: the MySQL query thread of the caller */ stat_print_fn* stat_print) { - trx_t* trx; static const char truncated_msg[] = "... truncated...\n"; const long MAX_STATUS_SIZE = 1048576; ulint trx_list_start = ULINT_UNDEFINED; @@ -15223,9 +15925,9 @@ innodb_show_status( DBUG_RETURN(0); } - trx = check_trx_exists(thd); + srv_wake_purge_thread_if_not_active(); - trx_search_latch_release_if_reserved(trx); + trx_t* trx = check_trx_exists(thd); innobase_srv_conc_force_exit_innodb(trx); @@ -15233,7 +15935,7 @@ innodb_show_status( bytes of text. */ char* str; - ssize_t flen, usable_len; + ssize_t flen; mutex_enter(&srv_monitor_file_mutex); rewind(srv_monitor_file); @@ -15247,6 +15949,8 @@ innodb_show_status( flen = 0; } + ssize_t usable_len; + if (flen > MAX_STATUS_SIZE) { usable_len = MAX_STATUS_SIZE; srv_truncated_status_writes++; @@ -15257,7 +15961,8 @@ innodb_show_status( /* allocate buffer for the string, and read the contents of the temporary file */ - if (!(str = (char*) my_malloc(usable_len + 1, MYF(0)))) { + if (!(str = (char*) my_malloc(//PSI_INSTRUMENT_ME, + usable_len + 1, MYF(0)))) { mutex_exit(&srv_monitor_file_mutex); DBUG_RETURN(1); } @@ -15299,161 +16004,337 @@ innodb_show_status( DBUG_RETURN(ret_val); } -/************************************************************************//** -Implements the SHOW MUTEX STATUS command. -@return 0 on success. */ -static -int -innodb_mutex_show_status( -/*=====================*/ - handlerton* hton, /*!< in: the innodb handlerton */ - THD* thd, /*!< in: the MySQL query thread of the - caller */ - stat_print_fn* stat_print) /*!< in: function for printing - statistics */ -{ - char buf1[IO_SIZE]; - char buf2[IO_SIZE]; - ib_mutex_t* mutex; - rw_lock_t* lock; - ulint block_mutex_oswait_count = 0; - ulint block_lock_oswait_count = 0; - ib_mutex_t* block_mutex = NULL; - rw_lock_t* block_lock = NULL; -#ifdef UNIV_DEBUG - ulint rw_lock_count= 0; - ulint rw_lock_count_spin_loop= 0; - ulint rw_lock_count_spin_rounds= 0; - ulint rw_lock_count_os_wait= 0; - ulint rw_lock_count_os_yield= 0; - ulonglong rw_lock_wait_time= 0; -#endif /* UNIV_DEBUG */ - uint buf1len; - uint buf2len; - uint hton_name_len; +/** Callback for collecting mutex statistics */ +struct ShowStatus { + + /** For tracking the mutex metrics */ + struct Value { + + /** Constructor + @param[in] name Name of the mutex + @param[in] spins Number of spins + @param[in] os_waits OS waits so far + @param[in] calls Number of calls to enter() */ + Value(const char* name, + ulint spins, + uint64_t waits, + uint64_t calls) + : + m_name(name), + m_spins(spins), + m_waits(waits), + m_calls(calls) + { + /* No op */ + } - hton_name_len = (uint) strlen(innobase_hton_name); + /** Mutex name */ + std::string m_name; - DBUG_ENTER("innodb_mutex_show_status"); - DBUG_ASSERT(hton == innodb_hton_ptr); + /** Spins so far */ + ulint m_spins; - mutex_enter(&mutex_list_mutex); + /** Waits so far */ + uint64_t m_waits; - for (mutex = UT_LIST_GET_FIRST(mutex_list); mutex != NULL; - mutex = UT_LIST_GET_NEXT(list, mutex)) { - if (mutex->count_os_wait == 0) { - continue; + /** Number of calls so far */ + uint64_t m_calls; + }; + + /** Order by m_waits, in descending order. */ + struct OrderByWaits: public std::binary_function<Value, Value, bool> + { + /** @return true if rhs < lhs */ + bool operator()( + const Value& lhs, + const Value& rhs) const + UNIV_NOTHROW + { + return(rhs.m_waits < lhs.m_waits); } + }; - if (buf_pool_is_block_mutex(mutex)) { - block_mutex = mutex; - block_mutex_oswait_count += mutex->count_os_wait; - continue; + typedef std::vector<Value, ut_allocator<Value> > Values; + + /** Collect the individual latch counts */ + struct GetCount { + typedef latch_meta_t::CounterType::Count Count; + + /** Constructor + @param[in] name Latch name + @param[in,out] values Put the values here */ + GetCount( + const char* name, + Values* values) + UNIV_NOTHROW + : + m_name(name), + m_values(values) + { + /* No op */ } - buf1len= (uint) my_snprintf(buf1, sizeof(buf1), "%s:%lu", - innobase_basename(mutex->cfile_name), - (ulong) mutex->cline); - buf2len= (uint) my_snprintf(buf2, sizeof(buf2), "os_waits=%lu", - (ulong) mutex->count_os_wait); + /** Collect the latch metrics. Ignore entries where the + spins and waits are zero. + @param[in] count The latch metrics */ + void operator()(Count* count) + UNIV_NOTHROW + { + if (count->m_spins > 0 || count->m_waits > 0) { - if (stat_print(thd, innobase_hton_name, - hton_name_len, buf1, buf1len, - buf2, buf2len)) { - mutex_exit(&mutex_list_mutex); - DBUG_RETURN(1); + m_values->push_back(Value( + m_name, + count->m_spins, + count->m_waits, + count->m_calls)); + } } + + /** The latch name */ + const char* m_name; + + /** For collecting the active mutex stats. */ + Values* m_values; + }; + + /** Constructor */ + ShowStatus() { } + + /** Callback for collecting the stats + @param[in] latch_meta Latch meta data + @return always returns true */ + bool operator()(latch_meta_t& latch_meta) + UNIV_NOTHROW + { + latch_meta_t::CounterType* counter; + + counter = latch_meta.get_counter(); + + GetCount get_count(latch_meta.get_name(), &m_values); + + counter->iterate(get_count); + + return(true); } - if (block_mutex) { - buf1len = (uint) my_snprintf(buf1, sizeof buf1, - "combined %s:%lu", - innobase_basename( - block_mutex->cfile_name), - (ulong) block_mutex->cline); - buf2len = (uint) my_snprintf(buf2, sizeof buf2, - "os_waits=%lu", - (ulong) block_mutex_oswait_count); + /** Implements the SHOW MUTEX STATUS command, for mutexes. + The table structure is like so: Engine | Mutex Name | Status + We store the metrics in the "Status" column as: + + spins=N,waits=N,calls=N" + + The user has to parse the dataunfortunately + @param[in,out] hton the innodb handlerton + @param[in,out] thd the MySQL query thread of the caller + @param[in,out] stat_print function for printing statistics + @return true on success. */ + bool to_string( + handlerton* hton, + THD* thd, + stat_print_fn* stat_print) + UNIV_NOTHROW; + + /** For collecting the active mutex stats. */ + Values m_values; +}; + +/** Implements the SHOW MUTEX STATUS command, for mutexes. +The table structure is like so: Engine | Mutex Name | Status +We store the metrics in the "Status" column as: + + spins=N,waits=N,calls=N" + +The user has to parse the dataunfortunately +@param[in,out] hton the innodb handlerton +@param[in,out] thd the MySQL query thread of the caller +@param[in,out] stat_print function for printing statistics +@return true on success. */ +bool +ShowStatus::to_string( + handlerton* hton, + THD* thd, + stat_print_fn* stat_print) + UNIV_NOTHROW +{ + uint hton_name_len = (uint) strlen(innobase_hton_name); + + std::sort(m_values.begin(), m_values.end(), OrderByWaits()); + + Values::iterator end = m_values.end(); + + for (Values::iterator it = m_values.begin(); it != end; ++it) { + + int name_len; + char name_buf[IO_SIZE]; + + name_len = snprintf( + name_buf, sizeof(name_buf), "%s", it->m_name.c_str()); + + int status_len; + char status_buf[IO_SIZE]; + + status_len = snprintf( + status_buf, sizeof(status_buf), + "spins=%lu,waits=%lu,calls=%llu", + static_cast<ulong>(it->m_spins), + static_cast<long>(it->m_waits), + (ulonglong) it->m_calls); if (stat_print(thd, innobase_hton_name, - hton_name_len, buf1, buf1len, - buf2, buf2len)) { - mutex_exit(&mutex_list_mutex); - DBUG_RETURN(1); + hton_name_len, + name_buf, static_cast<uint>(name_len), + status_buf, static_cast<uint>(status_len))) { + + return(false); } } - mutex_exit(&mutex_list_mutex); + return(true); +} + +/** Implements the SHOW MUTEX STATUS command, for mutexes. +@param[in,out] hton the innodb handlerton +@param[in,out] thd the MySQL query thread of the caller +@param[in,out] stat_print function for printing statistics +@return 0 on success. */ +static +int +innodb_show_mutex_status( + handlerton* hton, + THD* thd, + stat_print_fn* stat_print) +{ + DBUG_ENTER("innodb_show_mutex_status"); + + ShowStatus collector; + + DBUG_ASSERT(hton == innodb_hton_ptr); + + mutex_monitor.iterate(collector); + + if (!collector.to_string(hton, thd, stat_print)) { + DBUG_RETURN(1); + } + + DBUG_RETURN(0); +} + +/** Implements the SHOW MUTEX STATUS command. +@param[in,out] hton the innodb handlerton +@param[in,out] thd the MySQL query thread of the caller +@param[in,out] stat_print function for printing statistics +@return 0 on success. */ +static +int +innodb_show_rwlock_status( + handlerton* hton, + THD* thd, + stat_print_fn* stat_print) +{ + DBUG_ENTER("innodb_show_rwlock_status"); + + rw_lock_t* block_rwlock = NULL; + ulint block_rwlock_oswait_count = 0; + uint hton_name_len = (uint) strlen(innobase_hton_name); + + DBUG_ASSERT(hton == innodb_hton_ptr); mutex_enter(&rw_lock_list_mutex); - for (lock = UT_LIST_GET_FIRST(rw_lock_list); lock != NULL; - lock = UT_LIST_GET_NEXT(list, lock)) { - if (lock->count_os_wait == 0) { + for (rw_lock_t* rw_lock = UT_LIST_GET_FIRST(rw_lock_list); + rw_lock != NULL; + rw_lock = UT_LIST_GET_NEXT(list, rw_lock)) { + + if (rw_lock->count_os_wait == 0) { continue; } - if (buf_pool_is_block_lock(lock)) { - block_lock = lock; - block_lock_oswait_count += lock->count_os_wait; + int buf1len; + char buf1[IO_SIZE]; + + if (rw_lock->is_block_lock) { + + block_rwlock = rw_lock; + block_rwlock_oswait_count += rw_lock->count_os_wait; + continue; } - buf1len = (uint) my_snprintf( - buf1, sizeof buf1, "%s:%lu", - innobase_basename(lock->cfile_name), - static_cast<ulong>(lock->cline)); - buf2len = (uint) my_snprintf( - buf2, sizeof buf2, "os_waits=%lu", - static_cast<ulong>(lock->count_os_wait)); + buf1len = snprintf( + buf1, sizeof buf1, "rwlock: %s:%u", + innobase_basename(rw_lock->cfile_name), + rw_lock->cline); + + int buf2len; + char buf2[IO_SIZE]; + + buf2len = snprintf( + buf2, sizeof buf2, "waits=%u", + rw_lock->count_os_wait); if (stat_print(thd, innobase_hton_name, - hton_name_len, buf1, buf1len, - buf2, buf2len)) { + hton_name_len, + buf1, static_cast<uint>(buf1len), + buf2, static_cast<uint>(buf2len))) { + mutex_exit(&rw_lock_list_mutex); + DBUG_RETURN(1); } } - if (block_lock) { - buf1len = (uint) my_snprintf(buf1, sizeof buf1, - "combined %s:%lu", - innobase_basename( - block_lock->cfile_name), - (ulong) block_lock->cline); - buf2len = (uint) my_snprintf(buf2, sizeof buf2, - "os_waits=%lu", - (ulong) block_lock_oswait_count); + if (block_rwlock != NULL) { + + int buf1len; + char buf1[IO_SIZE]; + + buf1len = snprintf( + buf1, sizeof buf1, "sum rwlock: %s:%u", + innobase_basename(block_rwlock->cfile_name), + block_rwlock->cline); + + int buf2len; + char buf2[IO_SIZE]; + + buf2len = snprintf( + buf2, sizeof buf2, "waits=" ULINTPF, + block_rwlock_oswait_count); if (stat_print(thd, innobase_hton_name, - hton_name_len, buf1, buf1len, - buf2, buf2len)) { + hton_name_len, + buf1, static_cast<uint>(buf1len), + buf2, static_cast<uint>(buf2len))) { + mutex_exit(&rw_lock_list_mutex); + DBUG_RETURN(1); } } mutex_exit(&rw_lock_list_mutex); -#ifdef UNIV_DEBUG - buf2len = static_cast<uint>(my_snprintf(buf2, sizeof buf2, - "count=%lu, spin_waits=%lu, spin_rounds=%lu, " - "os_waits=%lu, os_yields=%lu, os_wait_times=%lu", - (ulong) rw_lock_count, - (ulong) rw_lock_count_spin_loop, - (ulong) rw_lock_count_spin_rounds, - (ulong) rw_lock_count_os_wait, - (ulong) rw_lock_count_os_yield, - (ulong) (rw_lock_wait_time / 1000))); - - if (stat_print(thd, innobase_hton_name, hton_name_len, - STRING_WITH_LEN("rw_lock_mutexes"), buf2, buf2len)) { - DBUG_RETURN(1); + DBUG_RETURN(0); +} + +/** Implements the SHOW MUTEX STATUS command. +@param[in,out] hton the innodb handlerton +@param[in,out] thd the MySQL query thread of the caller +@param[in,out] stat_print function for printing statistics +@return 0 on success. */ +static +int +innodb_show_latch_status( + handlerton* hton, + THD* thd, + stat_print_fn* stat_print) +{ + int ret = innodb_show_mutex_status(hton, thd, stat_print); + + if (ret != 0) { + return(ret); } -#endif /* UNIV_DEBUG */ - /* Success */ - DBUG_RETURN(0); + return(innodb_show_rwlock_status(hton, thd, stat_print)); } /************************************************************************//** @@ -15477,8 +16358,7 @@ innobase_show_status( return(innodb_show_status(hton, thd, stat_print) != 0); case HA_ENGINE_MUTEX: - /* Non-zero return value means there was an error. */ - return(innodb_mutex_show_status(hton, thd, stat_print) != 0); + return(innodb_show_latch_status(hton, thd, stat_print) != 0); case HA_ENGINE_LOGS: /* Not handled */ @@ -15488,128 +16368,54 @@ innobase_show_status( /* Success */ return(false); } +/*********************************************************************//** +Returns number of THR_LOCK locks used for one instance of InnoDB table. +InnoDB no longer relies on THR_LOCK locks so 0 value is returned. +Instead of THR_LOCK locks InnoDB relies on combination of metadata locks +(e.g. for LOCK TABLES and DDL) and its own locking subsystem. +Note that even though this method returns 0, SQL-layer still calls +::store_lock(), ::start_stmt() and ::external_lock() methods for InnoDB +tables. */ -/************************************************************************//** -Handling the shared INNOBASE_SHARE structure that is needed to provide table -locking. Register the table name if it doesn't exist in the hash table. */ -static -INNOBASE_SHARE* -get_share( -/*======*/ - const char* table_name) -{ - INNOBASE_SHARE* share; - - mysql_mutex_lock(&innobase_share_mutex); - - ulint fold = ut_fold_string(table_name); - - HASH_SEARCH(table_name_hash, innobase_open_tables, fold, - INNOBASE_SHARE*, share, - ut_ad(share->use_count > 0), - !strcmp(share->table_name, table_name)); - - if (!share) { - - uint length = (uint) strlen(table_name); - - /* TODO: invoke HASH_MIGRATE if innobase_open_tables - grows too big */ - - share = (INNOBASE_SHARE*) my_malloc(sizeof(*share)+length+1, - MYF(MY_FAE | MY_ZEROFILL)); - - share->table_name = (char*) memcpy(share + 1, - table_name, length + 1); - - HASH_INSERT(INNOBASE_SHARE, table_name_hash, - innobase_open_tables, fold, share); - - thr_lock_init(&share->lock); - - /* Index translation table initialization */ - share->idx_trans_tbl.index_mapping = NULL; - share->idx_trans_tbl.index_count = 0; - share->idx_trans_tbl.array_size = 0; - } - - share->use_count++; - mysql_mutex_unlock(&innobase_share_mutex); - - return(share); -} - -/************************************************************************//** -Free the shared object that was registered with get_share(). */ -static -void -free_share( -/*=======*/ - INNOBASE_SHARE* share) /*!< in/own: table share to free */ +uint +ha_innobase::lock_count(void) const +/*===============================*/ { - mysql_mutex_lock(&innobase_share_mutex); - -#ifdef UNIV_DEBUG - INNOBASE_SHARE* share2; - ulint fold = ut_fold_string(share->table_name); - - HASH_SEARCH(table_name_hash, innobase_open_tables, fold, - INNOBASE_SHARE*, share2, - ut_ad(share->use_count > 0), - !strcmp(share->table_name, share2->table_name)); - - ut_a(share2 == share); -#endif /* UNIV_DEBUG */ - - if (!--share->use_count) { - ulint fold = ut_fold_string(share->table_name); - - HASH_DELETE(INNOBASE_SHARE, table_name_hash, - innobase_open_tables, fold, share); - thr_lock_delete(&share->lock); - - /* Free any memory from index translation table */ - my_free(share->idx_trans_tbl.index_mapping); - - my_free(share); - - /* TODO: invoke HASH_MIGRATE if innobase_open_tables - shrinks too much */ - } - - mysql_mutex_unlock(&innobase_share_mutex); + return 0; } /*****************************************************************//** -Converts a MySQL table lock stored in the 'lock' field of the handle to -a proper type before storing pointer to the lock into an array of pointers. +Supposed to convert a MySQL table lock stored in the 'lock' field of the +handle to a proper type before storing pointer to the lock into an array +of pointers. +In practice, since InnoDB no longer relies on THR_LOCK locks and its +lock_count() method returns 0 it just informs storage engine about type +of THR_LOCK which SQL-layer would have acquired for this specific statement +on this specific table. MySQL also calls this if it wants to reset some table locks to a not-locked state during the processing of an SQL query. An example is that during a SELECT the read lock is released early on the 'const' tables where we only fetch one row. MySQL does not call this when it releases all locks at the end of an SQL statement. -@return pointer to the next element in the 'to' array */ -UNIV_INTERN +@return pointer to the current element in the 'to' array. */ + THR_LOCK_DATA** ha_innobase::store_lock( /*====================*/ THD* thd, /*!< in: user thread handle */ - THR_LOCK_DATA** to, /*!< in: pointer to an array - of pointers to lock structs; - pointer to the 'lock' field - of current handle is stored - next to this array */ - enum thr_lock_type lock_type) /*!< in: lock type to store in + THR_LOCK_DATA** to, /*!< in: pointer to the current + element in an array of pointers + to lock structs; + only used as return value */ + thr_lock_type lock_type) /*!< in: lock type to store in 'lock'; this may also be TL_IGNORE */ { - trx_t* trx; - - /* Note that trx in this function is NOT necessarily prebuilt->trx + /* Note that trx in this function is NOT necessarily m_prebuilt->trx because we call update_thd() later, in ::external_lock()! Failure to understand this caused a serious memory corruption bug in 5.1.11. */ - trx = check_trx_exists(thd); + trx_t* trx = check_trx_exists(thd); /* NOTE: MySQL can call this function with lock 'type' TL_IGNORE! Be careful to ignore TL_IGNORE if we are going to do something with @@ -15624,12 +16430,16 @@ ha_innobase::store_lock( (enum_tx_isolation) thd_tx_isolation(thd)); if (trx->isolation_level <= TRX_ISO_READ_COMMITTED - && trx->global_read_view) { + && MVCC::is_view_active(trx->read_view)) { /* At low transaction isolation levels we let each consistent read set its own snapshot */ - read_view_close_for_mysql(trx); + mutex_enter(&trx_sys->mutex); + + trx_sys->mvcc->view_close(trx->read_view, true); + + mutex_exit(&trx_sys->mutex); } } @@ -15665,16 +16475,16 @@ ha_innobase::store_lock( detected in the function. */ dberr_t err = row_quiesce_set_state( - prebuilt->table, QUIESCE_START, trx); + m_prebuilt->table, QUIESCE_START, trx); ut_a(err == DB_SUCCESS || err == DB_UNSUPPORTED); if (trx->isolation_level == TRX_ISO_SERIALIZABLE) { - prebuilt->select_lock_type = LOCK_S; - prebuilt->stored_select_lock_type = LOCK_S; + m_prebuilt->select_lock_type = LOCK_S; + m_prebuilt->stored_select_lock_type = LOCK_S; } else { - prebuilt->select_lock_type = LOCK_NONE; - prebuilt->stored_select_lock_type = LOCK_NONE; + m_prebuilt->select_lock_type = LOCK_NONE; + m_prebuilt->stored_select_lock_type = LOCK_NONE; } /* Check for DROP TABLE */ @@ -15682,7 +16492,7 @@ ha_innobase::store_lock( /* MySQL calls this function in DROP TABLE though this table handle may belong to another thd that is running a query. Let - us in that case skip any changes to the prebuilt struct. */ + us in that case skip any changes to the m_prebuilt struct. */ /* Check for LOCK TABLE t1,...,tn WITH SHARED LOCKS */ } else if ((lock_type == TL_READ && in_lock_tables) @@ -15713,6 +16523,7 @@ ha_innobase::store_lock( /* Use consistent read for checksum table */ if (sql_command == SQLCOM_CHECKSUM + || (sql_command == SQLCOM_ANALYZE && lock_type == TL_READ) || ((srv_locks_unsafe_for_binlog || trx->isolation_level <= TRX_ISO_READ_COMMITTED) && trx->isolation_level != TRX_ISO_SERIALIZABLE @@ -15733,11 +16544,11 @@ ha_innobase::store_lock( MODE in select, then we use consistent read for select. */ - prebuilt->select_lock_type = LOCK_NONE; - prebuilt->stored_select_lock_type = LOCK_NONE; + m_prebuilt->select_lock_type = LOCK_NONE; + m_prebuilt->stored_select_lock_type = LOCK_NONE; } else { - prebuilt->select_lock_type = LOCK_S; - prebuilt->stored_select_lock_type = LOCK_S; + m_prebuilt->select_lock_type = LOCK_S; + m_prebuilt->stored_select_lock_type = LOCK_S; } } else if (lock_type != TL_IGNORE) { @@ -15745,82 +16556,13 @@ ha_innobase::store_lock( /* We set possible LOCK_X value in external_lock, not yet here even if this would be SELECT ... FOR UPDATE */ - prebuilt->select_lock_type = LOCK_NONE; - prebuilt->stored_select_lock_type = LOCK_NONE; + m_prebuilt->select_lock_type = LOCK_NONE; + m_prebuilt->stored_select_lock_type = LOCK_NONE; } - if (lock_type != TL_IGNORE && lock.type == TL_UNLOCK) { - - /* Starting from 5.0.7, we weaken also the table locks - set at the start of a MySQL stored procedure call, just like - we weaken the locks set at the start of an SQL statement. - MySQL does set in_lock_tables TRUE there, but in reality - we do not need table locks to make the execution of a - single transaction stored procedure call deterministic - (if it does not use a consistent read). */ - - if (lock_type == TL_READ - && sql_command == SQLCOM_LOCK_TABLES) { - /* We come here if MySQL is processing LOCK TABLES - ... READ LOCAL. MyISAM under that table lock type - reads the table as it was at the time the lock was - granted (new inserts are allowed, but not seen by the - reader). To get a similar effect on an InnoDB table, - we must use LOCK TABLES ... READ. We convert the lock - type here, so that for InnoDB, READ LOCAL is - equivalent to READ. This will change the InnoDB - behavior in mysqldump, so that dumps of InnoDB tables - are consistent with dumps of MyISAM tables. */ - - lock_type = TL_READ_NO_INSERT; - } - - /* If we are not doing a LOCK TABLE, DISCARD/IMPORT - TABLESPACE or TRUNCATE TABLE then allow multiple - writers. Note that ALTER TABLE uses a TL_WRITE_ALLOW_READ - < TL_WRITE_CONCURRENT_INSERT. - - We especially allow multiple writers if MySQL is at the - start of a stored procedure call (SQLCOM_CALL) or a - stored function call (MySQL does have in_lock_tables - TRUE there). */ - - if ((lock_type >= TL_WRITE_CONCURRENT_INSERT - && lock_type <= TL_WRITE) - && !(in_lock_tables - && sql_command == SQLCOM_LOCK_TABLES) - && !thd_tablespace_op(thd) - && sql_command != SQLCOM_TRUNCATE - && sql_command != SQLCOM_OPTIMIZE - && sql_command != SQLCOM_CREATE_TABLE) { - - lock_type = TL_WRITE_ALLOW_WRITE; - } - - /* In queries of type INSERT INTO t1 SELECT ... FROM t2 ... - MySQL would use the lock TL_READ_NO_INSERT on t2, and that - would conflict with TL_WRITE_ALLOW_WRITE, blocking all inserts - to t2. Convert the lock to a normal read lock to allow - concurrent inserts to t2. - - We especially allow concurrent inserts if MySQL is at the - start of a stored procedure call (SQLCOM_CALL) - (MySQL does have thd_in_lock_tables() TRUE there). */ - - if (lock_type == TL_READ_NO_INSERT - && sql_command != SQLCOM_LOCK_TABLES) { - - lock_type = TL_READ; - } - - lock.type = lock_type; - } - - *to++= &lock; - if (!trx_is_started(trx) - && (prebuilt->select_lock_type != LOCK_NONE - || prebuilt->stored_select_lock_type != LOCK_NONE)) { + && (m_prebuilt->select_lock_type != LOCK_NONE + || m_prebuilt->stored_select_lock_type != LOCK_NONE)) { ++trx->will_lock; } @@ -15832,8 +16574,8 @@ ha_innobase::store_lock( Read the next autoinc value. Acquire the relevant locks before reading the AUTOINC value. If SUCCESS then the table AUTOINC mutex will be locked on return and all relevant locks acquired. -@return DB_SUCCESS or error code */ -UNIV_INTERN +@return DB_SUCCESS or error code */ + dberr_t ha_innobase::innobase_get_autoinc( /*==============================*/ @@ -15841,28 +16583,28 @@ ha_innobase::innobase_get_autoinc( { *value = 0; - prebuilt->autoinc_error = innobase_lock_autoinc(); + m_prebuilt->autoinc_error = innobase_lock_autoinc(); - if (prebuilt->autoinc_error == DB_SUCCESS) { + if (m_prebuilt->autoinc_error == DB_SUCCESS) { /* Determine the first value of the interval */ - *value = dict_table_autoinc_read(prebuilt->table); + *value = dict_table_autoinc_read(m_prebuilt->table); /* It should have been initialized during open. */ if (*value == 0) { - prebuilt->autoinc_error = DB_UNSUPPORTED; - dict_table_autoinc_unlock(prebuilt->table); + m_prebuilt->autoinc_error = DB_UNSUPPORTED; + dict_table_autoinc_unlock(m_prebuilt->table); } } - return(prebuilt->autoinc_error); + return(m_prebuilt->autoinc_error); } /*******************************************************************//** This function reads the global auto-inc counter. It doesn't use the AUTOINC lock even if the lock mode is set to TRADITIONAL. -@return the autoinc value */ -UNIV_INTERN +@return the autoinc value */ + ulonglong ha_innobase::innobase_peek_autoinc(void) /*====================================*/ @@ -15870,19 +16612,18 @@ ha_innobase::innobase_peek_autoinc(void) ulonglong auto_inc; dict_table_t* innodb_table; - ut_a(prebuilt != NULL); - ut_a(prebuilt->table != NULL); + ut_a(m_prebuilt != NULL); + ut_a(m_prebuilt->table != NULL); - innodb_table = prebuilt->table; + innodb_table = m_prebuilt->table; dict_table_autoinc_lock(innodb_table); auto_inc = dict_table_autoinc_read(innodb_table); if (auto_inc == 0) { - ut_print_timestamp(stderr); - fprintf(stderr, " InnoDB: AUTOINC next value generation " - "is disabled for '%s'\n", innodb_table->name); + ib::info() << "AUTOINC next value generation is disabled for" + " '" << innodb_table->name << "'"; } dict_table_autoinc_unlock(innodb_table); @@ -15892,7 +16633,7 @@ ha_innobase::innobase_peek_autoinc(void) /*********************************************************************//** Returns the value of the auto-inc counter in *first_value and ~0 on failure. */ -UNIV_INTERN + void ha_innobase::get_auto_increment( /*============================*/ @@ -15909,7 +16650,7 @@ ha_innobase::get_auto_increment( dberr_t error; ulonglong autoinc = 0; - /* Prepare prebuilt->trx in the table handle */ + /* Prepare m_prebuilt->trx in the table handle */ update_thd(ha_thd()); error = innobase_get_autoinc(&autoinc); @@ -15927,7 +16668,7 @@ ha_innobase::get_auto_increment( called and count down from that as rows are written (see write_row()). */ - trx = prebuilt->trx; + trx = m_prebuilt->trx; /* Note: We can't rely on *first_value since some MySQL engines, in particular the partition engine, don't initialize it to 0 when @@ -15936,8 +16677,7 @@ ha_innobase::get_auto_increment( /* We need the upper limit of the col type to check for whether we update the table autoinc counter or not. */ - ulonglong col_max_value = innobase_get_int_col_max_value( - table->next_number_field); + ulonglong col_max_value = innobase_get_int_col_max_value(table->next_number_field); /** The following logic is needed to avoid duplicate key error for autoincrement column. @@ -15950,7 +16690,7 @@ ha_innobase::get_auto_increment( (3) It is restricted only for insert operations. */ - if (increment > 1 && thd_sql_command(user_thd) != SQLCOM_ALTER_TABLE + if (increment > 1 && thd_sql_command(m_user_thd) != SQLCOM_ALTER_TABLE && autoinc < col_max_value) { ulonglong prev_auto_inc = autoinc; @@ -15984,19 +16724,19 @@ ha_innobase::get_auto_increment( set_if_bigger(*first_value, autoinc); /* Not in the middle of a mult-row INSERT. */ - } else if (prebuilt->autoinc_last_value == 0) { + } else if (m_prebuilt->autoinc_last_value == 0) { set_if_bigger(*first_value, autoinc); } - if (*first_value > col_max_value) - { - /* Out of range number. Let handler::update_auto_increment() - take care of this */ - prebuilt->autoinc_last_value = 0; - dict_table_autoinc_unlock(prebuilt->table); - *nb_reserved_values= 0; - return; - } + if (*first_value > col_max_value) { + /* Out of range number. Let handler::update_auto_increment() + take care of this */ + m_prebuilt->autoinc_last_value = 0; + dict_table_autoinc_unlock(m_prebuilt->table); + *nb_reserved_values= 0; + return; + } + *nb_reserved_values = trx->n_autoinc_rows; /* With old style AUTOINC locking we only update the table's @@ -16007,23 +16747,24 @@ ha_innobase::get_auto_increment( current = *first_value; - if (prebuilt->autoinc_increment != increment) { + if (m_prebuilt->autoinc_increment != increment) { WSREP_DEBUG("autoinc decrease: %llu -> %llu\n" "THD: %ld, current: %llu, autoinc: %llu", - prebuilt->autoinc_increment, + m_prebuilt->autoinc_increment, increment, - thd_get_thread_id(ha_thd()), + thd_get_thread_id(m_user_thd), current, autoinc); - if (!wsrep_on(ha_thd())) - { - current = autoinc - prebuilt->autoinc_increment; + if (!wsrep_on(m_user_thd)) { + current = autoinc + - m_prebuilt->autoinc_increment; } current = innobase_next_autoinc( current, 1, increment, offset, col_max_value); - dict_table_autoinc_initialize(prebuilt->table, current); + dict_table_autoinc_initialize( + m_prebuilt->table, current); *first_value = current; } @@ -16033,69 +16774,35 @@ ha_innobase::get_auto_increment( current, *nb_reserved_values, increment, offset, col_max_value); - prebuilt->autoinc_last_value = next_value; + m_prebuilt->autoinc_last_value = next_value; - if (prebuilt->autoinc_last_value < *first_value) { + if (m_prebuilt->autoinc_last_value < *first_value) { *first_value = (~(ulonglong) 0); } else { /* Update the table autoinc variable */ dict_table_autoinc_update_if_greater( - prebuilt->table, prebuilt->autoinc_last_value); + m_prebuilt->table, + m_prebuilt->autoinc_last_value); } } else { /* This will force write_row() into attempting an update of the table's AUTOINC counter. */ - prebuilt->autoinc_last_value = 0; + m_prebuilt->autoinc_last_value = 0; } /* The increment to be used to increase the AUTOINC value, we use this in write_row() and update_row() to increase the autoinc counter for columns that are filled by the user. We need the offset and the increment. */ - prebuilt->autoinc_offset = offset; - prebuilt->autoinc_increment = increment; - - dict_table_autoinc_unlock(prebuilt->table); -} - -/*******************************************************************//** -Reset the auto-increment counter to the given value, i.e. the next row -inserted will get the given value. This is called e.g. after TRUNCATE -is emulated by doing a 'DELETE FROM t'. HA_ERR_WRONG_COMMAND is -returned by storage engines that don't support this operation. -@return 0 or error code */ -UNIV_INTERN -int -ha_innobase::reset_auto_increment( -/*==============================*/ - ulonglong value) /*!< in: new value for table autoinc */ -{ - DBUG_ENTER("ha_innobase::reset_auto_increment"); - - dberr_t error; - - update_thd(ha_thd()); - - error = row_lock_table_autoinc_for_mysql(prebuilt); - - if (error != DB_SUCCESS) { - DBUG_RETURN(convert_error_code_to_mysql( - error, prebuilt->table->flags, user_thd)); - } + m_prebuilt->autoinc_offset = offset; + m_prebuilt->autoinc_increment = increment; - /* The next value can never be 0. */ - if (value == 0) { - value = 1; - } - - innobase_reset_autoinc(value); - - DBUG_RETURN(0); + dict_table_autoinc_unlock(m_prebuilt->table); } /*******************************************************************//** See comment in handler.cc */ -UNIV_INTERN + bool ha_innobase::get_error_message( /*===========================*/ @@ -16115,24 +16822,21 @@ ha_innobase::get_error_message( return(FALSE); } -/*******************************************************************//** - Retrieves the names of the table and the key for which there was a - duplicate entry in the case of HA_ERR_FOREIGN_DUPLICATE_KEY. - - If any of the names is not available, then this method will return - false and will not change any of child_table_name or child_key_name. - - @param child_table_name[out] Table name - @param child_table_name_len[in] Table name buffer size - @param child_key_name[out] Key name - @param child_key_name_len[in] Key name buffer size - - @retval true table and key names were available - and were written into the corresponding - out parameters. - @retval false table and key names were not available, - the out parameters were not touched. -*/ +/** Retrieves the names of the table and the key for which there was a +duplicate entry in the case of HA_ERR_FOREIGN_DUPLICATE_KEY. + +If any of the names is not available, then this method will return +false and will not change any of child_table_name or child_key_name. + +@param[out] child_table_name Table name +@param[in] child_table_name_len Table name buffer size +@param[out] child_key_name Key name +@param[in] child_key_name_len Key name buffer size + +@retval true table and key names were available and were written into the +corresponding out parameters. +@retval false table and key names were not available, the out parameters +were not touched. */ bool ha_innobase::get_foreign_dup_key( /*=============================*/ @@ -16143,10 +16847,10 @@ ha_innobase::get_foreign_dup_key( { const dict_index_t* err_index; - ut_a(prebuilt->trx != NULL); - ut_a(prebuilt->trx->magic_n == TRX_MAGIC_N); + ut_a(m_prebuilt->trx != NULL); + ut_a(m_prebuilt->trx->magic_n == TRX_MAGIC_N); - err_index = trx_get_error_info(prebuilt->trx); + err_index = trx_get_error_info(m_prebuilt->trx); if (err_index == NULL) { return(false); @@ -16155,20 +16859,24 @@ ha_innobase::get_foreign_dup_key( /* copy table name (and convert from filename-safe encoding to system_charset_info) */ - char* p; - p = strchr(err_index->table->name, '/'); + char* p = strchr(err_index->table->name.m_name, '/'); + /* strip ".../" prefix if any */ if (p != NULL) { p++; } else { - p = err_index->table->name; + p = err_index->table->name.m_name; } - uint len; + + size_t len; + len = filename_to_tablename(p, child_table_name, child_table_name_len); + child_table_name[len] = '\0'; /* copy index name */ - ut_snprintf(child_key_name, child_key_name_len, "%s", err_index->name); + snprintf(child_key_name, child_key_name_len, "%s", + err_index->name()); return(true); } @@ -16177,8 +16885,8 @@ ha_innobase::get_foreign_dup_key( Compares two 'refs'. A 'ref' is the (internal) primary key value of the row. If there is no explicitly declared non-null unique key or a primary key, then InnoDB internally uses the row id as the primary key. -@return < 0 if ref1 < ref2, 0 if equal, else > 0 */ -UNIV_INTERN +@return < 0 if ref1 < ref2, 0 if equal, else > 0 */ + int ha_innobase::cmp_ref( /*=================*/ @@ -16195,7 +16903,7 @@ ha_innobase::cmp_ref( uint len2; int result; - if (prebuilt->clust_index_was_generated) { + if (m_prebuilt->clust_index_was_generated) { /* The 'ref' is an InnoDB row id */ return(memcmp(ref1, ref2, DATA_ROW_ID_LEN)); @@ -16207,7 +16915,7 @@ ha_innobase::cmp_ref( key_part = table->key_info[table->s->primary_key].key_part; key_part_end = key_part - + table->key_info[table->s->primary_key].user_defined_key_parts; + + table->key_info[table->s->primary_key].user_defined_key_parts; for (; key_part != key_part_end; ++key_part) { field = key_part->field; @@ -16244,8 +16952,8 @@ ha_innobase::cmp_ref( /*******************************************************************//** Ask InnoDB if a query to a table can be cached. -@return TRUE if query caching of the table is permitted */ -UNIV_INTERN +@return TRUE if query caching of the table is permitted */ + my_bool ha_innobase::register_query_cache_table( /*====================================*/ @@ -16260,34 +16968,13 @@ ha_innobase::register_query_cache_table( is permitted */ ulonglong *engine_data) /*!< in/out: data to call_back */ { - *call_back = innobase_query_caching_of_table_permitted; *engine_data = 0; - return(innobase_query_caching_of_table_permitted(thd, table_key, - key_length, - engine_data)); -} - -/*******************************************************************//** -Get the bin log name. */ -UNIV_INTERN -const char* -ha_innobase::get_mysql_bin_log_name() -/*=================================*/ -{ - return(trx_sys_mysql_bin_log_name); -} - -/*******************************************************************//** -Get the bin log offset (or file position). */ -UNIV_INTERN -ulonglong -ha_innobase::get_mysql_bin_log_pos() -/*================================*/ -{ - /* trx... is ib_int64_t, which is a typedef for a 64-bit integer - (__int64 or longlong) so it's ok to cast it to ulonglong. */ + *call_back = innobase_query_caching_of_table_permitted; - return(trx_sys_mysql_bin_log_pos); + return(innobase_query_caching_of_table_permitted( + thd, table_key, + static_cast<uint>(key_length), + engine_data)); } /******************************************************************//** @@ -16295,8 +16982,7 @@ This function is used to find the storage length in bytes of the first n characters for prefix indexes using a multibyte character set. The function finds charset information and returns length of prefix_len characters in the index field in bytes. -@return number of bytes occupied by the first n characters */ -UNIV_INTERN +@return number of bytes occupied by the first n characters */ ulint innobase_get_at_most_n_mbchars( /*===========================*/ @@ -16349,12 +17035,13 @@ innobase_get_at_most_n_mbchars( if (char_length > data_len) { char_length = data_len; } + } else if (data_len < prefix_len) { + + char_length = data_len; + } else { - if (data_len < prefix_len) { - char_length = data_len; - } else { - char_length = prefix_len; - } + + char_length = prefix_len; } return(char_length); @@ -16362,7 +17049,7 @@ innobase_get_at_most_n_mbchars( /*******************************************************************//** This function is used to prepare an X/Open XA distributed transaction. -@return 0 or error number */ +@return 0 or error number */ static int innobase_xa_prepare( @@ -16375,36 +17062,18 @@ innobase_xa_prepare( false - the current SQL statement ended */ { - int error = 0; trx_t* trx = check_trx_exists(thd); DBUG_ASSERT(hton == innodb_hton_ptr); - /* we use support_xa value as it was seen at transaction start - time, not the current session variable value. Any possible changes - to the session variable take effect only in the next transaction */ - if (!trx->support_xa) { - -#ifdef WITH_WSREP - thd_get_xid(thd, (MYSQL_XID*) &trx->xid); -#endif // WITH_WSREP - return(0); - } - - thd_get_xid(thd, (MYSQL_XID*) &trx->xid); - - /* Release a possible FIFO ticket and search latch. Since we will - reserve the trx_sys->mutex, we have to release the search system - latch first to obey the latching order. */ - - trx_search_latch_release_if_reserved(trx); + thd_get_xid(thd, (MYSQL_XID*) trx->xid); innobase_srv_conc_force_exit_innodb(trx); if (!trx_is_registered_for_2pc(trx) && trx_is_started(trx)) { - sql_print_error("Transaction not registered for MySQL 2PC, " - "but transaction is active"); + sql_print_error("Transaction not registered for MariaDB 2PC," + " but transaction is active"); } if (prepare_trx @@ -16416,8 +17085,6 @@ innobase_xa_prepare( ut_ad(trx_is_registered_for_2pc(trx)); trx_prepare_for_mysql(trx); - - error = 0; } else { /* We just mark the SQL statement ended and do not do a transaction prepare */ @@ -16434,12 +17101,30 @@ innobase_xa_prepare( trx_mark_sql_stat_end(trx); } - return(error); + if (thd_sql_command(thd) != SQLCOM_XA_PREPARE + && (prepare_trx + || !thd_test_options( + thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN))) { + + /* For mysqlbackup to work the order of transactions in binlog + and InnoDB must be the same. Consider the situation + + thread1> prepare; write to binlog; ... + <context switch> + thread2> prepare; write to binlog; commit + thread1> ... commit + + The server guarantees that writes to the binary log + and commits are in the same order, so we do not have + to handle this case. */ + } + + return(0); } /*******************************************************************//** This function is used to recover X/Open XA distributed transactions. -@return number of prepared transactions stored in xid_list */ +@return number of prepared transactions stored in xid_list */ static int innobase_xa_recover( @@ -16461,7 +17146,7 @@ innobase_xa_recover( /*******************************************************************//** This function is used to commit one X/Open XA distributed transaction which is in the prepared state -@return 0 or error number */ +@return 0 or error number */ static int innobase_commit_by_xid( @@ -16469,19 +17154,20 @@ innobase_commit_by_xid( handlerton* hton, XID* xid) /*!< in: X/Open XA transaction identification */ { - trx_t* trx; - DBUG_ASSERT(hton == innodb_hton_ptr); if (high_level_read_only) { return(XAER_RMFAIL); } - trx = trx_get_trx_by_xid(xid); - - if (trx) { + if (trx_t* trx = trx_get_trx_by_xid(xid)) { + /* use cases are: disconnected xa, slave xa, recovery */ innobase_commit_low(trx); + ut_ad(trx->mysql_thd == NULL); + trx_deregister_from_2pc(trx); + ut_ad(!trx->will_lock); /* trx cache requirement */ trx_free_for_background(trx); + return(XA_OK); } else { return(XAER_NOTA); @@ -16491,7 +17177,7 @@ innobase_commit_by_xid( /*******************************************************************//** This function is used to rollback one X/Open XA distributed transaction which is in the prepared state -@return 0 or error number */ +@return 0 or error number */ static int innobase_rollback_by_xid( @@ -16500,82 +17186,24 @@ innobase_rollback_by_xid( XID* xid) /*!< in: X/Open XA transaction identification */ { - trx_t* trx; - DBUG_ASSERT(hton == innodb_hton_ptr); if (high_level_read_only) { return(XAER_RMFAIL); } - trx = trx_get_trx_by_xid(xid); - if (trx) { - int ret = innobase_rollback_trx(trx); + if (trx_t* trx = trx_get_trx_by_xid(xid)) { + int ret = innobase_rollback_trx(trx); + trx_deregister_from_2pc(trx); + ut_ad(!trx->will_lock); trx_free_for_background(trx); + return(ret); } else { return(XAER_NOTA); } } -/*******************************************************************//** -Create a consistent view for a cursor based on current transaction -which is created if the corresponding MySQL thread still lacks one. -This consistent view is then used inside of MySQL when accessing records -using a cursor. -@return pointer to cursor view or NULL */ -static -void* -innobase_create_cursor_view( -/*========================*/ - handlerton* hton, /*!< in: innobase hton */ - THD* thd) /*!< in: user thread handle */ -{ - DBUG_ASSERT(hton == innodb_hton_ptr); - - return(read_cursor_view_create_for_mysql(check_trx_exists(thd))); -} - -/*******************************************************************//** -Close the given consistent cursor view of a transaction and restore -global read view to a transaction read view. Transaction is created if the -corresponding MySQL thread still lacks one. */ -static -void -innobase_close_cursor_view( -/*=======================*/ - handlerton* hton, /*!< in: innobase hton */ - THD* thd, /*!< in: user thread handle */ - void* curview)/*!< in: Consistent read view to be closed */ -{ - DBUG_ASSERT(hton == innodb_hton_ptr); - - read_cursor_view_close_for_mysql(check_trx_exists(thd), - (cursor_view_t*) curview); -} - -/*******************************************************************//** -Set the given consistent cursor view to a transaction which is created -if the corresponding MySQL thread still lacks one. If the given -consistent cursor view is NULL global read view of a transaction is -restored to a transaction read view. */ -static -void -innobase_set_cursor_view( -/*=====================*/ - handlerton* hton, /*!< in: innobase hton */ - THD* thd, /*!< in: user thread handle */ - void* curview)/*!< in: Consistent cursor view to be set */ -{ - DBUG_ASSERT(hton == innodb_hton_ptr); - - read_cursor_set_for_mysql(check_trx_exists(thd), - (cursor_view_t*) curview); -} - -/*******************************************************************//** -*/ -UNIV_INTERN bool ha_innobase::check_if_incompatible_data( /*====================================*/ @@ -16588,7 +17216,7 @@ ha_innobase::check_if_incompatible_data( param_new = info->option_struct; param_old = table->s->option_struct; - innobase_copy_frm_flags_from_create_info(prebuilt->table, info); + innobase_copy_frm_flags_from_create_info(m_prebuilt->table, info); if (table_changes != IS_EQUAL_YES) { @@ -16596,8 +17224,8 @@ ha_innobase::check_if_incompatible_data( } /* Check that auto_increment value was not changed */ - if ((info->used_fields & HA_CREATE_USED_AUTO) && - info->auto_increment_value != 0) { + if ((info->used_fields & HA_CREATE_USED_AUTO) + && info->auto_increment_value != 0) { return(COMPATIBLE_DATA_NO); } @@ -16616,8 +17244,8 @@ ha_innobase::check_if_incompatible_data( /* Changes on engine specific table options requests a rebuild of the table. */ if (param_new->page_compressed != param_old->page_compressed || - param_new->page_compression_level != param_old->page_compression_level || - param_new->atomic_writes != param_old->atomic_writes) { + param_new->page_compression_level != param_old->page_compression_level) + { return(COMPATIBLE_DATA_NO); } @@ -16640,6 +17268,7 @@ innodb_io_capacity_max_update( from check function */ { ulong in_val = *static_cast<const ulong*>(save); + if (in_val < srv_io_capacity) { push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WRONG_ARGUMENTS, @@ -16650,7 +17279,7 @@ innodb_io_capacity_max_update( srv_io_capacity = in_val; push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, - ER_WRONG_ARGUMENTS, + ER_WRONG_ARGUMENTS, "Setting innodb_io_capacity to %lu", srv_io_capacity); } @@ -16679,8 +17308,8 @@ innodb_io_capacity_update( push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WRONG_ARGUMENTS, "Setting innodb_io_capacity to %lu" - " higher than innodb_io_capacity_max %lu", - in_val, srv_max_io_capacity); + " higher than innodb_io_capacity_max %lu", + in_val, srv_max_io_capacity); srv_max_io_capacity = in_val * 2; @@ -16762,7 +17391,7 @@ innodb_max_dirty_pages_pct_lwm_update( /************************************************************//** Validate the file format name and return its corresponding id. -@return valid file format id */ +@return valid file format id */ static uint innobase_file_format_name_lookup( @@ -16829,7 +17458,7 @@ innobase_file_format_validate_and_set( /*************************************************************//** Check if it is a valid file format. This function is registered as a callback with MySQL. -@return 0 for valid file format */ +@return 0 for valid file format */ static int innodb_file_format_name_validate( @@ -16871,6 +17500,37 @@ innodb_file_format_name_validate( return(1); } +/*************************************************************//** +Don't allow to set innodb_fast_shutdown=0 if purge threads are +already down. +@return 0 if innodb_fast_shutdown can be set */ +static +int +fast_shutdown_validate( +/*=============================*/ + THD* thd, /*!< in: thread handle */ + struct st_mysql_sys_var* var, /*!< in: pointer to system + variable */ + void* save, /*!< out: immediate result + for update function */ + struct st_mysql_value* value) /*!< in: incoming string */ +{ + if (check_sysvar_int(thd, var, save, value)) { + return(1); + } + + uint new_val = *reinterpret_cast<uint*>(save); + + if (srv_fast_shutdown && !new_val + && !my_atomic_loadptr_explicit(reinterpret_cast<void**> + (&srv_running), + MY_MEMORY_ORDER_RELAXED)) { + return(1); + } + + return(0); +} + /****************************************************************//** Update the system variable innodb_file_format using the "saved" value. This function is registered as a callback with MySQL. */ @@ -16891,6 +17551,10 @@ innodb_file_format_name_update( ut_a(var_ptr != NULL); ut_a(save != NULL); + + push_warning(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_WRONG_COMMAND, deprecated_file_format); + format_name = *static_cast<const char*const*>(save); if (format_name) { @@ -16910,7 +17574,7 @@ innodb_file_format_name_update( /*************************************************************//** Check if valid argument to innodb_file_format_max. This function is registered as a callback with MySQL. -@return 0 for valid file format */ +@return 0 for valid file format */ static int innodb_file_format_max_validate( @@ -16950,9 +17614,9 @@ innodb_file_format_max_validate( push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WRONG_ARGUMENTS, - "InnoDB: invalid innodb_file_format_max " - "value; can be any format up to %s " - "or equivalent id of %d", + "InnoDB: invalid innodb_file_format_max" + " value; can be any format up to %s" + " or equivalent id of %d", trx_sys_file_format_id_to_name(UNIV_FORMAT_MAX), UNIV_FORMAT_MAX); } @@ -16984,6 +17648,10 @@ innodb_file_format_max_update( ut_a(save != NULL); ut_a(var_ptr != NULL); + + push_warning(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_WRONG_COMMAND, deprecated_file_format_max); + format_name_in = *static_cast<const char*const*>(save); if (!format_name_in) { @@ -17006,13 +17674,29 @@ innodb_file_format_max_update( /* Update the max format id in the system tablespace. */ if (trx_sys_file_format_max_set(format_id, format_name_out)) { - ut_print_timestamp(stderr); - fprintf(stderr, - " [Info] InnoDB: the file format in the system " - "tablespace is now set to %s.\n", *format_name_out); + ib::info() << "The file format in the system tablespace is now" + " set to " << *format_name_out << "."; } } +/** Update innodb_large_prefix. +@param[in,out] thd MySQL client connection +@param[out] var_ptr current value +@param[in] save to-be-assigned value */ +static +void +innodb_large_prefix_update( + THD* thd, + st_mysql_sys_var*, + void* var_ptr, + const void* save) +{ + push_warning(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_WRONG_COMMAND, deprecated_large_prefix); + + *static_cast<my_bool*>(var_ptr) = *static_cast<const my_bool*>(save); +} + /*************************************************************//** Check whether valid argument given to innobase_*_stopword_table. This function is registered as a callback with MySQL. @@ -17056,6 +17740,32 @@ innodb_stopword_table_validate( return(ret); } +/** Update the system variable innodb_buffer_pool_size using the "saved" +value. This function is registered as a callback with MySQL. +@param[in] thd thread handle +@param[in] var pointer to system variable +@param[out] var_ptr where the formal string goes +@param[in] save immediate result from check function */ +static +void +innodb_buffer_pool_size_update( + THD* thd, + struct st_mysql_sys_var* var, + void* var_ptr, + const void* save) +{ + longlong in_val = *static_cast<const longlong*>(save); + + snprintf(export_vars.innodb_buffer_pool_resize_status, + sizeof(export_vars.innodb_buffer_pool_resize_status), + "Requested to resize buffer pool."); + + os_event_set(srv_buf_resize_event); + + ib::info() << export_vars.innodb_buffer_pool_resize_status + << " (new size: " << in_val << " bytes)"; +} + /*************************************************************//** Check whether valid argument given to "innodb_fts_internal_tbl_name" This function is registered as a callback with MySQL. @@ -17132,16 +17842,8 @@ innodb_internal_table_update( table_name = *static_cast<const char*const*>(save); old = *(char**) var_ptr; - - if (table_name) { - *(char**) var_ptr = my_strdup(table_name, MYF(0)); - } else { - *(char**) var_ptr = NULL; - } - - if (old) { - my_free(old); - } + *(char**) var_ptr = table_name ? my_strdup(table_name, MYF(0)) : NULL; + my_free(old); fts_internal_tbl_name2 = *(char**) var_ptr; if (fts_internal_tbl_name2 == NULL) { @@ -17151,6 +17853,7 @@ innodb_internal_table_update( } } +#ifdef BTR_CUR_HASH_ADAPT /****************************************************************//** Update the system variable innodb_adaptive_hash_index using the "saved" value. This function is registered as a callback with MySQL. */ @@ -17169,9 +17872,10 @@ innodb_adaptive_hash_index_update( if (*(my_bool*) save) { btr_search_enable(); } else { - btr_search_disable(); + btr_search_disable(true); } } +#endif /* BTR_CUR_HASH_ADAPT */ /****************************************************************//** Update the system variable innodb_cmp_per_index using the "saved" @@ -17232,14 +17936,14 @@ innodb_change_buffer_max_size_update( const void* save) /*!< in: immediate result from check function */ { - innobase_change_buffer_max_size = + srv_change_buffer_max_size = (*static_cast<const uint*>(save)); - ibuf_max_size_update(innobase_change_buffer_max_size); + ibuf_max_size_update(srv_change_buffer_max_size); } #ifdef UNIV_DEBUG -ulong srv_fil_make_page_dirty_debug = 0; -ulong srv_saved_page_number_debug = 0; +static ulong srv_fil_make_page_dirty_debug = 0; +static ulong srv_saved_page_number_debug = 0; /****************************************************************//** Save an InnoDB page number. */ @@ -17257,9 +17961,8 @@ innodb_save_page_no( { srv_saved_page_number_debug = *static_cast<const ulong*>(save); - ib_logf(IB_LOG_LEVEL_INFO, - "Saving InnoDB page number: %lu", - srv_saved_page_number_debug); + ib::info() << "Saving InnoDB page number: " + << srv_saved_page_number_debug; } /****************************************************************//** @@ -17276,33 +17979,45 @@ innodb_make_page_dirty( const void* save) /*!< in: immediate result from check function */ { - mtr_t mtr; - ulong space_id = *static_cast<const ulong*>(save); + mtr_t mtr; + ulong space_id = *static_cast<const ulong*>(save); + fil_space_t* space = fil_space_acquire_silent(space_id); + + if (space == NULL) { + return; + } + + if (srv_saved_page_number_debug >= space->size) { + fil_space_release(space); + return; + } + + mtr.start(); + mtr.set_named_space(space); - mtr_start(&mtr); + buf_block_t* block = buf_page_get( + page_id_t(space_id, srv_saved_page_number_debug), + page_size_t(space->flags), RW_X_LATCH, &mtr); - buf_block_t* block = buf_page_get( - space_id, 0, srv_saved_page_number_debug, RW_X_LATCH, &mtr); + if (block != NULL) { + byte* page = block->frame; + + ib::info() << "Dirtying page: " << page_id_t( + page_get_space_id(page), page_get_page_no(page)); - if (block) { - byte* page = block->frame; - ib_logf(IB_LOG_LEVEL_INFO, - "Dirtying page:%lu of space:%lu", - page_get_page_no(page), - page_get_space_id(page)); mlog_write_ulint(page + FIL_PAGE_TYPE, fil_page_get_type(page), MLOG_2BYTES, &mtr); } - mtr_commit(&mtr); + mtr.commit(); + fil_space_release(space); } #endif // UNIV_DEBUG - /*************************************************************//** Find the corresponding ibuf_use_t value that indexes into innobase_change_buffering_values[] array for the input change buffering option name. -@return corresponding IBUF_USE_* value for the input variable +@return corresponding IBUF_USE_* value for the input variable name, or IBUF_USE_COUNT if not able to find a match */ static ibuf_use_t @@ -17311,14 +18026,14 @@ innodb_find_change_buffering_value( const char* input_name) /*!< in: input change buffering option name */ { - ulint use; + for (ulint i = 0; + i < UT_ARR_SIZE(innobase_change_buffering_values); + ++i) { - for (use = 0; use < UT_ARR_SIZE(innobase_change_buffering_values); - use++) { /* found a match */ if (!innobase_strcasecmp( - input_name, innobase_change_buffering_values[use])) { - return((ibuf_use_t) use); + input_name, innobase_change_buffering_values[i])) { + return(static_cast<ibuf_use_t>(i)); } } @@ -17329,7 +18044,7 @@ innodb_find_change_buffering_value( /*************************************************************//** Check if it is a valid value of innodb_change_buffering. This function is registered as a callback with MySQL. -@return 0 for valid innodb_change_buffering */ +@return 0 for valid innodb_change_buffering */ static int innodb_change_buffering_validate( @@ -17402,7 +18117,7 @@ innodb_change_buffering_update( /*************************************************************//** Just emit a warning that the usage of the variable is deprecated. -@return 0 */ +@return 0 */ static void innodb_stats_sample_pages_update( @@ -17415,19 +18130,16 @@ innodb_stats_sample_pages_update( const void* save) /*!< in: immediate result from check function */ { -#define STATS_SAMPLE_PAGES_DEPRECATED_MSG \ - "Using innodb_stats_sample_pages is deprecated and " \ - "the variable may be removed in future releases. " \ - "Please use innodb_stats_transient_sample_pages " \ - "instead." + + const char* STATS_SAMPLE_PAGES_DEPRECATED_MSG = + "Using innodb_stats_sample_pages is deprecated and" + " the variable may be removed in future releases." + " Please use innodb_stats_transient_sample_pages instead."; push_warning(thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_WRONG_COMMAND, STATS_SAMPLE_PAGES_DEPRECATED_MSG); - ut_print_timestamp(stderr); - fprintf(stderr, - " InnoDB: Warning: %s\n", - STATS_SAMPLE_PAGES_DEPRECATED_MSG); + ib::warn() << STATS_SAMPLE_PAGES_DEPRECATED_MSG; srv_stats_transient_sample_pages = *static_cast<const unsigned long long*>(save); @@ -17462,11 +18174,15 @@ innodb_monitor_set_option( exisitng monitor counter (status variable), make special processing to remember existing counter value. */ - if (monitor_info->monitor_type - & MONITOR_EXISTING) { + if (monitor_info->monitor_type & MONITOR_EXISTING) { srv_mon_process_existing_counter( monitor_id, MONITOR_TURN_ON); } + + if (MONITOR_IS_ON(MONITOR_LATCHES)) { + + mutex_monitor.enable(); + } break; case MONITOR_TURN_OFF: @@ -17477,14 +18193,25 @@ innodb_monitor_set_option( MONITOR_OFF(monitor_id); MONITOR_SET_OFF(monitor_id); + + if (!MONITOR_IS_ON(MONITOR_LATCHES)) { + + mutex_monitor.disable(); + } break; case MONITOR_RESET_VALUE: srv_mon_reset(monitor_id); + + if (monitor_id == (MONITOR_LATCHES)) { + + mutex_monitor.reset(); + } break; case MONITOR_RESET_ALL_VALUE: srv_mon_reset_all(monitor_id); + mutex_monitor.reset(); break; default: @@ -17557,7 +18284,7 @@ innodb_monitor_update_wildcard( /*************************************************************//** Given a configuration variable name, find corresponding monitor counter and return its monitor ID if found. -@return monitor ID if found, MONITOR_NO_MATCH if there is no match */ +@return monitor ID if found, MONITOR_NO_MATCH if there is no match */ static ulint innodb_monitor_id_by_name_get( @@ -17588,7 +18315,7 @@ innodb_monitor_id_by_name_get( /*************************************************************//** Validate that the passed in monitor name matches at least one monitor counter name with wildcard compare. -@return TRUE if at least one monitor name matches */ +@return TRUE if at least one monitor name matches */ static ibool innodb_monitor_validate_wildcard_name( @@ -17607,7 +18334,7 @@ innodb_monitor_validate_wildcard_name( /*************************************************************//** Validate the passed in monitor name, find and save the corresponding monitor name in the function parameter "save". -@return 0 if monitor name is valid */ +@return 0 if monitor name is valid */ static int innodb_monitor_valid_byname( @@ -17670,7 +18397,7 @@ innodb_monitor_valid_byname( /*************************************************************//** Validate passed-in "value" is a valid monitor counter name. This function is registered as a callback with MySQL. -@return 0 for valid name */ +@return 0 for valid name */ static int innodb_monitor_validate( @@ -17698,7 +18425,8 @@ innodb_monitor_validate( by InnoDB, so we can access it in another callback function innodb_monitor_update() and free it appropriately */ if (name) { - monitor_name = my_strdup(name, MYF(0)); + monitor_name = my_strdup(//PSI_INSTRUMENT_ME, + name, MYF(0)); } else { return(1); } @@ -17766,14 +18494,14 @@ innodb_monitor_update( push_warning_printf( thd, Sql_condition::WARN_LEVEL_WARN, ER_NO_DEFAULT, - "Default value is not defined for " - "this set option. Please specify " - "correct counter or module name."); + "Default value is not defined for" + " this set option. Please specify" + " correct counter or module name."); } else { sql_print_error( - "Default value is not defined for " - "this set option. Please specify " - "correct counter or module name.\n"); + "Default value is not defined for" + " this set option. Please specify" + " correct counter or module name.\n"); } if (var_ptr) { @@ -17816,7 +18544,7 @@ exit: been turned on, we will set err_monitor. Print related information */ if (err_monitor) { - sql_print_warning("Monitor %s is already enabled.", + sql_print_warning("InnoDB: Monitor %s is already enabled.", srv_mon_get_name((monitor_id_t) err_monitor)); } @@ -17827,13 +18555,13 @@ exit: return; } -#ifdef __WIN__ +#ifdef _WIN32 /*************************************************************//** Validate if passed-in "value" is a valid value for innodb_buffer_pool_filename. On Windows, file names with colon (:) are not allowed. -@return 0 for valid name */ +@return 0 for valid name */ static int innodb_srv_buf_dump_filename_validate( @@ -17845,16 +18573,15 @@ innodb_srv_buf_dump_filename_validate( for update function */ struct st_mysql_value* value) /*!< in: incoming string */ { - const char* buf_name; char buff[OS_FILE_MAX_PATH]; - int len= sizeof(buff); + int len = sizeof(buff); ut_a(save != NULL); ut_a(value != NULL); - buf_name = value->val_str(value, buff, &len); + const char* buf_name = value->val_str(value, buff, &len); - if (buf_name) { + if (buf_name != NULL) { if (is_filename_allowed(buf_name, len, FALSE)){ *static_cast<const char**>(save) = buf_name; return(0); @@ -17862,17 +18589,17 @@ innodb_srv_buf_dump_filename_validate( push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WRONG_ARGUMENTS, - "InnoDB: innodb_buffer_pool_filename " - "cannot have colon (:) in the file name."); + "InnoDB: innodb_buffer_pool_filename" + " cannot have colon (:) in the file name."); } } return(1); } -#else /* __WIN__ */ +#else /* _WIN32 */ # define innodb_srv_buf_dump_filename_validate NULL -#endif /* __WIN__ */ +#endif /* _WIN32 */ #ifdef UNIV_DEBUG static char* srv_buffer_pool_evict; @@ -18038,7 +18765,16 @@ innodb_defragment_frequency_update( { srv_defragment_frequency = (*static_cast<const uint*>(save)); srv_defragment_interval = ut_microseconds_to_timer( - 1000000.0 / srv_defragment_frequency); + (ulonglong) (1000000.0 / srv_defragment_frequency)); +} + +static inline char *my_strtok_r(char *str, const char *delim, char **saveptr) +{ +#if defined _WIN32 + return strtok_s(str, delim, saveptr); +#else + return strtok_r(str, delim, saveptr); +#endif } /****************************************************************//** @@ -18062,9 +18798,9 @@ innodb_enable_monitor_at_startup( and/or counter group name, and calling innodb_monitor_update() if successfully updated. Please note that the "str" would be changed by strtok_r() as it walks through it. */ - for (char* option = strtok_r(str, sep, &last); + for (char* option = my_strtok_r(str, sep, &last); option; - option = strtok_r(NULL, sep, &last)) { + option = my_strtok_r(NULL, sep, &last)) { ulint ret; char* option_name; @@ -18095,6 +18831,7 @@ show_innodb_vars( innodb_export_status(); var->type = SHOW_ARRAY; var->value = (char*) &innodb_status_variables; + //var->scope = SHOW_SCOPE_GLOBAL; return(0); } @@ -18105,7 +18842,6 @@ system default primary index name 'GEN_CLUST_INDEX'. If a name matches, this function pushes an warning message to the client, and returns true. @return true if the index name matches the reserved name */ -UNIV_INTERN bool innobase_index_name_is_reserved( /*============================*/ @@ -18126,10 +18862,10 @@ innobase_index_name_is_reserved( push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WRONG_NAME_FOR_INDEX, - "Cannot Create Index with name " - "'%s'. The name is reserved " - "for the system default primary " - "index.", + "Cannot Create Index with name" + " '%s'. The name is reserved" + " for the system default primary" + " index.", innobase_index_reserve_name); my_error(ER_WRONG_NAME_FOR_INDEX, MYF(0), @@ -18142,112 +18878,74 @@ innobase_index_name_is_reserved( return(false); } -/*********************************************************************** -Retrieve the FTS Relevance Ranking result for doc with doc_id -of prebuilt->fts_doc_id +/** Retrieve the FTS Relevance Ranking result for doc with doc_id +of m_prebuilt->fts_doc_id +@param[in,out] fts_hdl FTS handler @return the relevance ranking value */ -UNIV_INTERN +static float innobase_fts_retrieve_ranking( -/*============================*/ - FT_INFO * fts_hdl) /*!< in: FTS handler */ + FT_INFO* fts_hdl) { - row_prebuilt_t* ft_prebuilt; fts_result_t* result; + row_prebuilt_t* ft_prebuilt; - result = ((NEW_FT_INFO*) fts_hdl)->ft_result; + result = reinterpret_cast<NEW_FT_INFO*>(fts_hdl)->ft_result; - ft_prebuilt = ((NEW_FT_INFO*) fts_hdl)->ft_prebuilt; + ft_prebuilt = reinterpret_cast<NEW_FT_INFO*>(fts_hdl)->ft_prebuilt; - if (ft_prebuilt->read_just_key) { - fts_ranking_t* ranking = - rbt_value(fts_ranking_t, result->current); - return(ranking->rank); - } + fts_ranking_t* ranking = rbt_value(fts_ranking_t, result->current); + ft_prebuilt->fts_doc_id= ranking->doc_id; - /* Retrieve the ranking value for doc_id with value of - prebuilt->fts_doc_id */ - return(fts_retrieve_ranking(result, ft_prebuilt->fts_doc_id)); + return(ranking->rank); } -/*********************************************************************** -Free the memory for the FTS handler */ -UNIV_INTERN +/** Free the memory for the FTS handler +@param[in,out] fts_hdl FTS handler */ +static void innobase_fts_close_ranking( -/*=======================*/ - FT_INFO * fts_hdl) + FT_INFO* fts_hdl) { fts_result_t* result; - result = ((NEW_FT_INFO*) fts_hdl)->ft_result; + result = reinterpret_cast<NEW_FT_INFO*>(fts_hdl)->ft_result; fts_query_free_result(result); my_free((uchar*) fts_hdl); - - return; } -/*********************************************************************** -Find and Retrieve the FTS Relevance Ranking result for doc with doc_id -of prebuilt->fts_doc_id +/** Find and Retrieve the FTS Relevance Ranking result for doc with doc_id +of m_prebuilt->fts_doc_id +@param[in,out] fts_hdl FTS handler @return the relevance ranking value */ -UNIV_INTERN +static float -innobase_fts_find_ranking( -/*======================*/ - FT_INFO* fts_hdl, /*!< in: FTS handler */ - uchar* record, /*!< in: Unused */ - uint len) /*!< in: Unused */ +innobase_fts_find_ranking(FT_INFO* fts_hdl, uchar*, uint) { - row_prebuilt_t* ft_prebuilt; fts_result_t* result; + row_prebuilt_t* ft_prebuilt; - ft_prebuilt = ((NEW_FT_INFO*) fts_hdl)->ft_prebuilt; - result = ((NEW_FT_INFO*) fts_hdl)->ft_result; + ft_prebuilt = reinterpret_cast<NEW_FT_INFO*>(fts_hdl)->ft_prebuilt; + result = reinterpret_cast<NEW_FT_INFO*>(fts_hdl)->ft_result; /* Retrieve the ranking value for doc_id with value of - prebuilt->fts_doc_id */ + m_prebuilt->fts_doc_id */ return(fts_retrieve_ranking(result, ft_prebuilt->fts_doc_id)); } #ifdef UNIV_DEBUG -static my_bool innodb_purge_run_now = TRUE; -static my_bool innodb_purge_stop_now = TRUE; +static my_bool innodb_background_drop_list_empty = TRUE; static my_bool innodb_log_checkpoint_now = TRUE; static my_bool innodb_buf_flush_list_now = TRUE; +static uint innodb_merge_threshold_set_all_debug + = DICT_INDEX_MERGE_THRESHOLD_DEFAULT; -/****************************************************************//** -Set the purge state to RUN. If purge is disabled then it -is a no-op. This function is registered as a callback with MySQL. */ -static -void -purge_run_now_set( -/*==============*/ - THD* thd /*!< in: thread handle */ - MY_ATTRIBUTE((unused)), - struct st_mysql_sys_var* var /*!< in: pointer to system - variable */ - MY_ATTRIBUTE((unused)), - void* var_ptr /*!< out: where the formal - string goes */ - MY_ATTRIBUTE((unused)), - const void* save) /*!< in: immediate result from - check function */ -{ - if (*(my_bool*) save && trx_purge_state() != PURGE_STATE_DISABLED) { - trx_purge_run(); - } -} - -/****************************************************************//** -Set the purge state to STOP. If purge is disabled then it -is a no-op. This function is registered as a callback with MySQL. */ +/** Wait for the background drop list to become empty. */ static void -purge_stop_now_set( -/*===============*/ +wait_background_drop_list_empty( THD* thd /*!< in: thread handle */ MY_ATTRIBUTE((unused)), struct st_mysql_sys_var* var /*!< in: pointer to system @@ -18259,9 +18957,7 @@ purge_stop_now_set( const void* save) /*!< in: immediate result from check function */ { - if (*(my_bool*) save && trx_purge_state() != PURGE_STATE_DISABLED) { - trx_purge_stop(); - } + row_wait_for_background_drop_list_empty(); } /****************************************************************//** @@ -18282,18 +18978,19 @@ checkpoint_now_set( check function */ { if (*(my_bool*) save) { - while (log_sys->last_checkpoint_lsn < log_sys->lsn) { + while (log_sys->last_checkpoint_lsn + + SIZE_OF_MLOG_CHECKPOINT + + (log_sys->append_on_checkpoint != NULL + ? log_sys->append_on_checkpoint->size() : 0) + < log_sys->lsn) { log_make_checkpoint_at(LSN_MAX, TRUE); - fil_flush_file_spaces(FIL_LOG); + fil_flush_file_spaces(FIL_TYPE_LOG); } dberr_t err = fil_write_flushed_lsn(log_sys->lsn); if (err != DB_SUCCESS) { - ib_logf(IB_LOG_LEVEL_ERROR, - "Failed to write flush lsn to the " - "system tablespace at checkpoint err=%s", - ut_strerr(err)); + ib::warn() << "Checkpoint set failed " << err; } } } @@ -18316,74 +19013,56 @@ buf_flush_list_now_set( check function */ { if (*(my_bool*) save) { - buf_flush_list(ULINT_MAX, LSN_MAX, NULL); - buf_flush_wait_batch_end(NULL, BUF_FLUSH_LIST); + buf_flush_sync_all_buf_pools(); } } -#endif /* UNIV_DEBUG */ - -/*********************************************************************** -@return version of the extended FTS API */ -uint -innobase_fts_get_version() -/*======================*/ -{ - /* Currently this doesn't make much sense as returning - HA_CAN_FULLTEXT_EXT automatically mean this version is supported. - This supposed to ease future extensions. */ - return(2); -} -/*********************************************************************** -@return Which part of the extended FTS API is supported */ -ulonglong -innobase_fts_flags() -/*================*/ +/** Override current MERGE_THRESHOLD setting for all indexes at dictionary +now. +@param[in] thd thread handle +@param[in] var pointer to system variable +@param[out] var_ptr where the formal string goes +@param[in] save immediate result from check function */ +static +void +innodb_merge_threshold_set_all_debug_update( + THD* thd, + struct st_mysql_sys_var* var, + void* var_ptr, + const void* save) { - return(FTS_ORDERED_RESULT | FTS_DOCID_IN_RESULT); + innodb_merge_threshold_set_all_debug + = (*static_cast<const uint*>(save)); + dict_set_merge_threshold_all_debug( + innodb_merge_threshold_set_all_debug); } +#endif /* UNIV_DEBUG */ - -/*********************************************************************** -Find and Retrieve the FTS doc_id for the current result row +/** Find and Retrieve the FTS doc_id for the current result row +@param[in,out] fts_hdl FTS handler @return the document ID */ +static ulonglong innobase_fts_retrieve_docid( -/*========================*/ - FT_INFO_EXT * fts_hdl) /*!< in: FTS handler */ + FT_INFO_EXT* fts_hdl) { - row_prebuilt_t* ft_prebuilt; fts_result_t* result; + row_prebuilt_t* ft_prebuilt; - ft_prebuilt = ((NEW_FT_INFO *)fts_hdl)->ft_prebuilt; - result = ((NEW_FT_INFO *)fts_hdl)->ft_result; + ft_prebuilt = reinterpret_cast<NEW_FT_INFO *>(fts_hdl)->ft_prebuilt; + result = reinterpret_cast<NEW_FT_INFO *>(fts_hdl)->ft_result; if (ft_prebuilt->read_just_key) { + fts_ranking_t* ranking = rbt_value(fts_ranking_t, result->current); + return(ranking->doc_id); } return(ft_prebuilt->fts_doc_id); } -/*********************************************************************** -Find and retrieve the size of the current result -@return number of matching rows */ -ulonglong -innobase_fts_count_matches( -/*=======================*/ - FT_INFO_EXT* fts_hdl) /*!< in: FTS handler */ -{ - NEW_FT_INFO* handle = (NEW_FT_INFO *) fts_hdl; - - if (handle->ft_result->rankings_by_id != 0) { - return rbt_size(handle->ft_result->rankings_by_id); - } else { - return(0); - } -} - /* These variables are never read by InnoDB or changed. They are a kind of dummies that are needed by the MySQL infrastructure to call buffer_pool_dump_now(), buffer_pool_load_now() and buffer_pool_load_abort() @@ -18431,13 +19110,13 @@ void buffer_pool_load_now( /*=================*/ THD* thd /*!< in: thread handle */ - __attribute__((unused)), + MY_ATTRIBUTE((unused)), struct st_mysql_sys_var* var /*!< in: pointer to system variable */ - __attribute__((unused)), + MY_ATTRIBUTE((unused)), void* var_ptr /*!< out: where the formal string goes */ - __attribute__((unused)), + MY_ATTRIBUTE((unused)), const void* save) /*!< in: immediate result from check function */ { @@ -18454,13 +19133,13 @@ void buffer_pool_load_abort( /*===================*/ THD* thd /*!< in: thread handle */ - __attribute__((unused)), + MY_ATTRIBUTE((unused)), struct st_mysql_sys_var* var /*!< in: pointer to system variable */ - __attribute__((unused)), + MY_ATTRIBUTE((unused)), void* var_ptr /*!< out: where the formal string goes */ - __attribute__((unused)), + MY_ATTRIBUTE((unused)), const void* save) /*!< in: immediate result from check function */ { @@ -18469,10 +19148,56 @@ buffer_pool_load_abort( } } +/****************************************************************//** +Update the system variable innodb_log_write_ahead_size using the "saved" +value. This function is registered as a callback with MySQL. */ +static +void +innodb_log_write_ahead_size_update( +/*===============================*/ + THD* thd, /*!< in: thread handle */ + struct st_mysql_sys_var* var, /*!< in: pointer to + system variable */ + void* var_ptr,/*!< out: where the + formal string goes */ + const void* save) /*!< in: immediate result + from check function */ +{ + ulong val = OS_FILE_LOG_BLOCK_SIZE; + ulong in_val = *static_cast<const ulong*>(save); + + while (val < in_val) { + val = val * 2; + } + + if (val > UNIV_PAGE_SIZE) { + val = UNIV_PAGE_SIZE; + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "innodb_log_write_ahead_size cannot" + " be set higher than innodb_page_size."); + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "Setting innodb_log_write_ahead_size" + " to %lu", + UNIV_PAGE_SIZE); + } else if (val != in_val) { + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "innodb_log_write_ahead_size should be" + " set 2^n value and larger than 512."); + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "Setting innodb_log_write_ahead_size" + " to %lu", + val); + } + + srv_log_write_ahead_size = val; +} + /** Update innodb_status_output or innodb_status_output_locks, which control InnoDB "status monitor" output to the error log. -@param[in] thd thread handle -@param[in] var system variable @param[out] var_ptr current value @param[in] save to-be-assigned value */ static @@ -18557,6 +19282,23 @@ innodb_encrypt_tables_update( fil_crypt_set_encrypt_tables(*static_cast<const ulong*>(save)); } +/** Update the innodb_log_checksums parameter. +@param[in] thd thread handle +@param[in] var system variable +@param[out] var_ptr current value +@param[in] save immediate result from check function */ +static +void +innodb_log_checksums_update( + THD* thd, + struct st_mysql_sys_var* var, + void* var_ptr, + const void* save) +{ + *static_cast<my_bool*>(var_ptr) = innodb_log_checksums_func_update( + thd, *static_cast<const my_bool*>(save)); +} + static SHOW_VAR innodb_status_variables_export[]= { {"Innodb", (char*) &show_innodb_vars, SHOW_FUNC}, {NullS, NullS, SHOW_LONG} @@ -18567,7 +19309,10 @@ static struct st_mysql_storage_engine innobase_storage_engine= #ifdef WITH_WSREP void -wsrep_abort_slave_trx(wsrep_seqno_t bf_seqno, wsrep_seqno_t victim_seqno) +wsrep_abort_slave_trx( +/*==================*/ + wsrep_seqno_t bf_seqno, + wsrep_seqno_t victim_seqno) { WSREP_ERROR("Trx %lld tries to abort slave trx %lld. This could be " "caused by:\n\t" @@ -18583,6 +19328,7 @@ This function is used to kill one transaction in BF. */ UNIV_INTERN int wsrep_innobase_kill_one_trx( +/*========================*/ void * const bf_thd_ptr, const trx_t * const bf_trx, trx_t *victim_trx, @@ -18613,9 +19359,9 @@ wsrep_innobase_kill_one_trx( WSREP_LOG_CONFLICT(bf_thd, thd, TRUE); - WSREP_DEBUG("BF kill (%lu, seqno: %lld), victim: (%lu) trx: " - TRX_ID_FMT, - signal, (long long)bf_seqno, + WSREP_DEBUG("BF kill (" ULINTPF ", seqno: " INT64PF + "), victim: (%lu) trx: " TRX_ID_FMT, + signal, bf_seqno, thd_get_thread_id(thd), victim_trx->id); @@ -18642,7 +19388,7 @@ wsrep_innobase_kill_one_trx( DBUG_RETURN(0); } - if(wsrep_thd_exec_mode(thd) != LOCAL_STATE) { + if (wsrep_thd_exec_mode(thd) != LOCAL_STATE) { WSREP_DEBUG("withdraw for BF trx: " TRX_ID_FMT ", state: %d", victim_trx->id, wsrep_thd_get_conflict_state(thd)); @@ -18703,8 +19449,7 @@ wsrep_innobase_kill_one_trx( WSREP_ERROR( "cancel commit bad exit: %d " TRX_ID_FMT, - rcode, - victim_trx->id); + rcode, victim_trx->id); /* unable to interrupt, must abort */ /* note: kill_mysql() will block, if we cannot. * kill the lock holder first. @@ -18724,10 +19469,12 @@ wsrep_innobase_kill_one_trx( victim_trx->id); victim_trx->lock.was_chosen_as_deadlock_victim= TRUE; + if (victim_trx->lock.wait_lock) { WSREP_DEBUG("victim has wait flag: %ld", thd_get_thread_id(thd)); lock_t* wait_lock = victim_trx->lock.wait_lock; + if (wait_lock) { WSREP_DEBUG("canceling wait lock"); victim_trx->lock.was_chosen_as_deadlock_victim= TRUE; @@ -18773,7 +19520,7 @@ wsrep_innobase_kill_one_trx( if (wsrep_aborting_thd_contains(thd)) { WSREP_WARN("duplicate thd aborter %lu", - thd_get_thread_id(thd)); + (ulong) thd_get_thread_id(thd)); } else { wsrep_aborting_thd_enqueue(thd); DBUG_PRINT("wsrep",("enqueuing trx abort for %lu", @@ -18802,6 +19549,7 @@ wsrep_innobase_kill_one_trx( static int wsrep_abort_transaction( +/*====================*/ handlerton* hton, THD *bf_thd, THD *victim_thd, @@ -18820,12 +19568,10 @@ wsrep_abort_transaction( if (victim_trx) { lock_mutex_enter(); trx_mutex_enter(victim_trx); - victim_trx->abort_type = TRX_WSREP_ABORT; int rcode = wsrep_innobase_kill_one_trx(bf_thd, bf_trx, victim_trx, signal); - trx_mutex_exit(victim_trx); lock_mutex_exit(); - victim_trx->abort_type = TRX_SERVER_ABORT; + trx_mutex_exit(victim_trx); wsrep_srv_conc_cancel_wait(victim_trx); DBUG_RETURN(rcode); } else { @@ -18839,30 +19585,44 @@ wsrep_abort_transaction( DBUG_RETURN(-1); } -static int innobase_wsrep_set_checkpoint(handlerton* hton, const XID* xid) +static +int +innobase_wsrep_set_checkpoint( +/*==========================*/ + handlerton* hton, + const XID* xid) { DBUG_ASSERT(hton == innodb_hton_ptr); - if (wsrep_is_wsrep_xid(xid)) { - mtr_t mtr; - mtr_start(&mtr); - trx_sysf_t* sys_header = trx_sysf_get(&mtr); - trx_sys_update_wsrep_checkpoint(xid, sys_header, &mtr); - mtr_commit(&mtr); - innobase_flush_logs(hton); - return 0; - } else { - return 1; - } + + if (wsrep_is_wsrep_xid(xid)) { + mtr_t mtr; + mtr_start(&mtr); + trx_sysf_t* sys_header = trx_sysf_get(&mtr); + trx_sys_update_wsrep_checkpoint(xid, sys_header, &mtr); + mtr_commit(&mtr); + innobase_flush_logs(hton, false); + return 0; + } else { + return 1; + } } -static int innobase_wsrep_get_checkpoint(handlerton* hton, XID* xid) +static +int +innobase_wsrep_get_checkpoint( +/*==========================*/ + handlerton* hton, + XID* xid) { DBUG_ASSERT(hton == innodb_hton_ptr); trx_sys_read_wsrep_checkpoint(xid); return 0; } -static void wsrep_fake_trx_id( +static +void +wsrep_fake_trx_id( +/*==============*/ handlerton *hton, THD *thd) /*!< in: user thread handle */ { @@ -18876,38 +19636,81 @@ static void wsrep_fake_trx_id( #endif /* WITH_WSREP */ +/** Update the innodb_use_trim parameter. +@param[in] thd thread handle +@param[in] var system variable +@param[out] var_ptr current value +@param[in] save immediate result from check function */ +static +void +innodb_use_trim_update( + THD* thd, + struct st_mysql_sys_var* var, + void* var_ptr, + const void* save) +{ + srv_use_trim = *static_cast<const my_bool*>(save); + + push_warning(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_WRONG_COMMAND, deprecated_use_trim); +} + +/** Update the innodb_instrument_sempahores parameter. +@param[in] thd thread handle +@param[in] var system variable +@param[out] var_ptr current value +@param[in] save immediate result from check function */ +static +void +innodb_instrument_semaphores_update( + THD* thd, + struct st_mysql_sys_var* var, + void* var_ptr, + const void* save) +{ + innodb_instrument_semaphores = *static_cast<const my_bool*>(save); + + push_warning(thd, Sql_condition::WARN_LEVEL_WARN, + HA_ERR_WRONG_COMMAND, deprecated_instrument_semaphores); +} + /* plugin options */ static MYSQL_SYSVAR_ENUM(checksum_algorithm, srv_checksum_algorithm, PLUGIN_VAR_RQCMDARG, - "The algorithm InnoDB uses for page checksumming. Possible values are " - "CRC32 (hardware accelerated if the CPU supports it) " - "write crc32, allow any of the other checksums to match when reading; " - "STRICT_CRC32 " - "write crc32, do not allow other algorithms to match when reading; " - "INNODB " - "write a software calculated checksum, allow any other checksums " - "to match when reading; " - "STRICT_INNODB " - "write a software calculated checksum, do not allow other algorithms " - "to match when reading; " - "NONE " - "write a constant magic number, do not do any checksum verification " - "when reading (same as innodb_checksums=OFF); " - "STRICT_NONE " - "write a constant magic number, do not allow values other than that " - "magic number when reading; " - "Files updated when this option is set to crc32 or strict_crc32 will " - "not be readable by MySQL versions older than 5.6.3", - NULL, NULL, SRV_CHECKSUM_ALGORITHM_INNODB, + "The algorithm InnoDB uses for page checksumming. Possible values are" + " CRC32 (hardware accelerated if the CPU supports it)" + " write crc32, allow any of the other checksums to match when reading;" + " STRICT_CRC32" + " write crc32, do not allow other algorithms to match when reading;" + " INNODB" + " write a software calculated checksum, allow any other checksums" + " to match when reading;" + " STRICT_INNODB" + " write a software calculated checksum, do not allow other algorithms" + " to match when reading;" + " NONE" + " write a constant magic number, do not do any checksum verification" + " when reading (same as innodb_checksums=OFF);" + " STRICT_NONE" + " write a constant magic number, do not allow values other than that" + " magic number when reading;" + " Files updated when this option is set to crc32 or strict_crc32 will" + " not be readable by MariaDB versions older than 10.0.4", + NULL, NULL, SRV_CHECKSUM_ALGORITHM_CRC32, &innodb_checksum_algorithm_typelib); +static MYSQL_SYSVAR_BOOL(log_checksums, innodb_log_checksums, + PLUGIN_VAR_RQCMDARG, + "Whether to compute and require checksums for InnoDB redo log blocks", + NULL, innodb_log_checksums_update, TRUE); + static MYSQL_SYSVAR_BOOL(checksums, innobase_use_checksums, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "DEPRECATED. Use innodb_checksum_algorithm=NONE instead of setting " - "this to OFF. " - "Enable InnoDB checksums validation (enabled by default). " - "Disable with --skip-innodb-checksums.", + "DEPRECATED. Use innodb_checksum_algorithm=NONE instead of setting" + " this to OFF." + " Enable InnoDB checksums validation (enabled by default)." + " Disable with --skip-innodb-checksums.", NULL, NULL, TRUE); static MYSQL_SYSVAR_STR(data_home_dir, innobase_data_home_dir, @@ -18917,28 +19720,29 @@ static MYSQL_SYSVAR_STR(data_home_dir, innobase_data_home_dir, static MYSQL_SYSVAR_BOOL(doublewrite, innobase_use_doublewrite, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Enable InnoDB doublewrite buffer (enabled by default). " - "Disable with --skip-innodb-doublewrite.", + "Enable InnoDB doublewrite buffer (enabled by default)." + " Disable with --skip-innodb-doublewrite.", NULL, NULL, TRUE); static MYSQL_SYSVAR_BOOL(use_atomic_writes, innobase_use_atomic_writes, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Prevent partial page writes, via atomic writes." - "The option is used to prevent partial writes in case of a crash/poweroff, " - "as faster alternative to doublewrite buffer." - "Currently this option works only " - "on Linux only with FusionIO device, and directFS filesystem.", - NULL, NULL, FALSE); + "Enable atomic writes, instead of using the doublewrite buffer, for files " + "on devices that supports atomic writes. " + "To use this option one must use " + "file_per_table=1, flush_method=O_DIRECT and use_fallocate=1. " + "This option only works on Linux with either FusionIO cards using " + "the directFS filesystem or with Shannon cards using any file system.", + NULL, NULL, TRUE); static MYSQL_SYSVAR_BOOL(use_fallocate, innobase_use_fallocate, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Preallocate files fast, using operating system functionality. On POSIX systems, posix_fallocate system call is used.", + "Use posix_fallocate() to allocate files. DEPRECATED, has no effect.", NULL, NULL, FALSE); static MYSQL_SYSVAR_BOOL(stats_include_delete_marked, srv_stats_include_delete_marked, PLUGIN_VAR_OPCMDARG, - "Scan delete marked records for persistent stat", + "Include delete marked records when calculating persistent statistics", NULL, NULL, FALSE); static MYSQL_SYSVAR_ULONG(io_capacity, srv_io_capacity, @@ -18961,15 +19765,11 @@ static MYSQL_SYSVAR_ULONG(idle_flush_pct, NULL, NULL, 100, 0, 100, 0); #ifdef UNIV_DEBUG -static MYSQL_SYSVAR_BOOL(purge_run_now, innodb_purge_run_now, - PLUGIN_VAR_OPCMDARG, - "Set purge state to RUN", - NULL, purge_run_now_set, FALSE); - -static MYSQL_SYSVAR_BOOL(purge_stop_now, innodb_purge_stop_now, +static MYSQL_SYSVAR_BOOL(background_drop_list_empty, + innodb_background_drop_list_empty, PLUGIN_VAR_OPCMDARG, - "Set purge state to STOP", - NULL, purge_stop_now_set, FALSE); + "Wait for the background drop list to become empty", + NULL, wait_background_drop_list_empty, FALSE); static MYSQL_SYSVAR_BOOL(log_checkpoint_now, innodb_log_checkpoint_now, PLUGIN_VAR_OPCMDARG, @@ -18980,6 +19780,14 @@ static MYSQL_SYSVAR_BOOL(buf_flush_list_now, innodb_buf_flush_list_now, PLUGIN_VAR_OPCMDARG, "Force dirty page flush now", NULL, buf_flush_list_now_set, FALSE); + +static MYSQL_SYSVAR_UINT(merge_threshold_set_all_debug, + innodb_merge_threshold_set_all_debug, + PLUGIN_VAR_RQCMDARG, + "Override current MERGE_THRESHOLD setting for all indexes at dictionary" + " cache by the specified value dynamically, at the time.", + NULL, innodb_merge_threshold_set_all_debug_update, + DICT_INDEX_MERGE_THRESHOLD_DEFAULT, 1, 50, 0); #endif /* UNIV_DEBUG */ static MYSQL_SYSVAR_ULONG(purge_batch_size, srv_purge_batch_size, @@ -18992,9 +19800,9 @@ static MYSQL_SYSVAR_ULONG(purge_batch_size, srv_purge_batch_size, static MYSQL_SYSVAR_ULONG(purge_threads, srv_n_purge_threads, PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY, - "Purge threads can be from 1 to 32. Default is 1.", + "Purge threads can be from 1 to 32. Default is 4.", NULL, NULL, - 1, /* Default setting */ + 4, /* Default setting */ 1, /* Minimum value */ 32, 0); /* Maximum value */ @@ -19006,11 +19814,11 @@ static MYSQL_SYSVAR_ULONG(sync_array_size, srv_sync_array_size, 1, /* Minimum value */ 1024, 0); /* Maximum value */ -static MYSQL_SYSVAR_ULONG(fast_shutdown, innobase_fast_shutdown, +static MYSQL_SYSVAR_UINT(fast_shutdown, srv_fast_shutdown, PLUGIN_VAR_OPCMDARG, - "Speeds up the shutdown process of the InnoDB storage engine. Possible " - "values are 0, 1 (faster) or 2 (fastest - crash-like).", - NULL, NULL, 1, 0, 2, 0); + "Speeds up the shutdown process of the InnoDB storage engine. Possible" + " values are 0, 1 (faster) or 2 (fastest - crash-like).", + fast_shutdown_validate, NULL, 1, 0, 2, 0); static MYSQL_SYSVAR_BOOL(file_per_table, srv_file_per_table, PLUGIN_VAR_NOCMDARG, @@ -19021,7 +19829,7 @@ static MYSQL_SYSVAR_STR(file_format, innobase_file_format_name, PLUGIN_VAR_RQCMDARG, "File format to use for new tables in .ibd files.", innodb_file_format_name_validate, - innodb_file_format_name_update, "Antelope"); + innodb_file_format_name_update, innodb_file_format_default); /* "innobase_file_format_check" decides whether we would continue booting the server if the file format stamped on the system @@ -19042,7 +19850,7 @@ static MYSQL_SYSVAR_STR(file_format_max, innobase_file_format_max, PLUGIN_VAR_OPCMDARG, "The highest file format in the tablespace.", innodb_file_format_max_validate, - innodb_file_format_max_update, "Antelope"); + innodb_file_format_max_update, innodb_file_format_max_default); static MYSQL_SYSVAR_STR(ft_server_stopword_table, innobase_server_stopword_table, PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_MEMALLOC, @@ -19077,7 +19885,7 @@ static MYSQL_SYSVAR_STR(flush_method, innobase_file_flush_method, static MYSQL_SYSVAR_BOOL(large_prefix, innobase_large_prefix, PLUGIN_VAR_NOCMDARG, "Support large index prefix length of REC_VERSION_56_MAX_INDEX_COL_LEN (3072) bytes.", - NULL, NULL, FALSE); + NULL, innodb_large_prefix_update, TRUE); static MYSQL_SYSVAR_BOOL(force_load_corrupted, srv_load_corrupted, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, @@ -19086,35 +19894,30 @@ static MYSQL_SYSVAR_BOOL(force_load_corrupted, srv_load_corrupted, static MYSQL_SYSVAR_BOOL(locks_unsafe_for_binlog, innobase_locks_unsafe_for_binlog, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "DEPRECATED. This option may be removed in future releases. " - "Please use READ COMMITTED transaction isolation level instead. " - "Force InnoDB to not use next-key locking, to use only row-level locking.", + "DEPRECATED. This option may be removed in future releases." + " Please use READ COMMITTED transaction isolation level instead." + " Force InnoDB to not use next-key locking, to use only row-level locking.", NULL, NULL, FALSE); -#ifdef UNIV_LOG_ARCHIVE -static MYSQL_SYSVAR_STR(log_arch_dir, innobase_log_arch_dir, - PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, - "Where full logs should be archived.", NULL, NULL, NULL); - -static MYSQL_SYSVAR_BOOL(log_archive, innobase_log_archive, - PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY, - "Set to 1 if you want to have logs archived.", NULL, NULL, FALSE); -#endif /* UNIV_LOG_ARCHIVE */ - static MYSQL_SYSVAR_STR(log_group_home_dir, srv_log_group_home_dir, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, "Path to InnoDB log files.", NULL, NULL, NULL); +static MYSQL_SYSVAR_ULONG(page_cleaners, srv_n_page_cleaners, + PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY, + "Page cleaner threads can be from 1 to 64. Default is 4.", + NULL, NULL, 4, 1, 64, 0); + static MYSQL_SYSVAR_DOUBLE(max_dirty_pages_pct, srv_max_buf_pool_modified_pct, PLUGIN_VAR_RQCMDARG, "Percentage of dirty pages allowed in bufferpool.", - NULL, innodb_max_dirty_pages_pct_update, 75.0, 0.001, 99.999, 0); + NULL, innodb_max_dirty_pages_pct_update, 75.0, 0, 99.999, 0); static MYSQL_SYSVAR_DOUBLE(max_dirty_pages_pct_lwm, srv_max_dirty_pages_pct_lwm, PLUGIN_VAR_RQCMDARG, "Percentage of dirty pages at which flushing kicks in.", - NULL, innodb_max_dirty_pages_pct_lwm_update, 0.001, 0.000, 99.999, 0); + NULL, innodb_max_dirty_pages_pct_lwm_update, 0, 0, 99.999, 0); static MYSQL_SYSVAR_DOUBLE(adaptive_flushing_lwm, srv_adaptive_flushing_lwm, @@ -19127,6 +19930,11 @@ static MYSQL_SYSVAR_BOOL(adaptive_flushing, srv_adaptive_flushing, "Attempt flushing dirty pages to avoid IO bursts at checkpoints.", NULL, NULL, TRUE); +static MYSQL_SYSVAR_BOOL(flush_sync, srv_flush_sync, + PLUGIN_VAR_NOCMDARG, + "Allow IO bursts at the checkpoints ignoring io_capacity setting.", + NULL, NULL, TRUE); + static MYSQL_SYSVAR_ULONG(flushing_avg_loops, srv_flushing_avg_loops, PLUGIN_VAR_RQCMDARG, @@ -19158,8 +19966,8 @@ static MYSQL_SYSVAR_BOOL(status_file, innobase_create_status_file, static MYSQL_SYSVAR_BOOL(stats_on_metadata, innobase_stats_on_metadata, PLUGIN_VAR_OPCMDARG, - "Enable statistics gathering for metadata commands such as " - "SHOW TABLE STATUS for tables that use transient statistics (off by default)", + "Enable statistics gathering for metadata commands such as" + " SHOW TABLE STATUS for tables that use transient statistics (off by default)", NULL, NULL, FALSE); static MYSQL_SYSVAR_ULONGLONG(stats_sample_pages, srv_stats_transient_sample_pages, @@ -19170,29 +19978,29 @@ static MYSQL_SYSVAR_ULONGLONG(stats_sample_pages, srv_stats_transient_sample_pag static MYSQL_SYSVAR_ULONGLONG(stats_transient_sample_pages, srv_stats_transient_sample_pages, PLUGIN_VAR_RQCMDARG, - "The number of leaf index pages to sample when calculating transient " - "statistics (if persistent statistics are not used, default 8)", + "The number of leaf index pages to sample when calculating transient" + " statistics (if persistent statistics are not used, default 8)", NULL, NULL, 8, 1, ~0ULL, 0); static MYSQL_SYSVAR_BOOL(stats_persistent, srv_stats_persistent, PLUGIN_VAR_OPCMDARG, - "InnoDB persistent statistics enabled for all tables unless overridden " - "at table level", + "InnoDB persistent statistics enabled for all tables unless overridden" + " at table level", NULL, NULL, TRUE); static MYSQL_SYSVAR_BOOL(stats_auto_recalc, srv_stats_auto_recalc, PLUGIN_VAR_OPCMDARG, - "InnoDB automatic recalculation of persistent statistics enabled for all " - "tables unless overridden at table level (automatic recalculation is only " - "done when InnoDB decides that the table has changed too much and needs a " - "new statistics)", + "InnoDB automatic recalculation of persistent statistics enabled for all" + " tables unless overridden at table level (automatic recalculation is only" + " done when InnoDB decides that the table has changed too much and needs a" + " new statistics)", NULL, NULL, TRUE); static MYSQL_SYSVAR_ULONGLONG(stats_persistent_sample_pages, srv_stats_persistent_sample_pages, PLUGIN_VAR_RQCMDARG, - "The number of leaf index pages to sample when calculating persistent " - "statistics (by ANALYZE, default 20)", + "The number of leaf index pages to sample when calculating persistent" + " statistics (by ANALYZE, default 20)", NULL, NULL, 20, 1, ~0ULL, 0); static MYSQL_SYSVAR_ULONGLONG(stats_modified_counter, srv_stats_modified_counter, @@ -19205,16 +20013,26 @@ static MYSQL_SYSVAR_BOOL(stats_traditional, srv_stats_sample_traditional, "Enable traditional statistic calculation based on number of configured pages (default true)", NULL, NULL, TRUE); +#ifdef BTR_CUR_HASH_ADAPT static MYSQL_SYSVAR_BOOL(adaptive_hash_index, btr_search_enabled, PLUGIN_VAR_OPCMDARG, - "Enable InnoDB adaptive hash index (enabled by default). " - "Disable with --skip-innodb-adaptive-hash-index.", - NULL, innodb_adaptive_hash_index_update, TRUE); + "Enable InnoDB adaptive hash index (enabled by default). " + " Disable with --skip-innodb-adaptive-hash-index.", + NULL, innodb_adaptive_hash_index_update, true); + +/** Number of distinct partitions of AHI. +Each partition is protected by its own latch and so we have parts number +of latches protecting complete search system. */ +static MYSQL_SYSVAR_ULONG(adaptive_hash_index_parts, btr_ahi_parts, + PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY, + "Number of InnoDB Adaptive Hash Index Partitions (default 8)", + NULL, NULL, 8, 1, 512, 0); +#endif /* BTR_CUR_HASH_ADAPT */ static MYSQL_SYSVAR_ULONG(replication_delay, srv_replication_delay, PLUGIN_VAR_RQCMDARG, - "Replication thread delay (ms) on the slave server if " - "innodb_thread_concurrency is reached (0 by default)", + "Replication thread delay (ms) on the slave server if" + " innodb_thread_concurrency is reached (0 by default)", NULL, NULL, 0, 0, ~0UL, 0); static MYSQL_SYSVAR_UINT(compression_level, page_zip_level, @@ -19232,23 +20050,56 @@ static MYSQL_SYSVAR_BOOL(log_compressed_pages, page_zip_log_pages, " compression algorithm doesn't change.", NULL, NULL, TRUE); -static MYSQL_SYSVAR_LONG(additional_mem_pool_size, innobase_additional_mem_pool_size, - PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, - "DEPRECATED. This option may be removed in future releases, " - "together with the option innodb_use_sys_malloc and with the InnoDB's " - "internal memory allocator. " - "Size of a memory pool InnoDB uses to store data dictionary information and other internal data structures.", - NULL, NULL, 8*1024*1024L, 512*1024L, LONG_MAX, 1024); +static MYSQL_SYSVAR_BOOL(log_optimize_ddl, innodb_log_optimize_ddl, + PLUGIN_VAR_OPCMDARG, + "Reduce redo logging when natively creating indexes or rebuilding tables." + " Setting this OFF avoids delay due to page flushing and" + " allows concurrent backup.", + NULL, NULL, TRUE); -static MYSQL_SYSVAR_ULONG(autoextend_increment, srv_auto_extend_increment, +static MYSQL_SYSVAR_ULONG(autoextend_increment, + sys_tablespace_auto_extend_increment, PLUGIN_VAR_RQCMDARG, "Data file autoextend increment in megabytes", NULL, NULL, 64L, 1L, 1000L, 0); +/** Validate the requested buffer pool size. Also, reserve the necessary +memory needed for buffer pool resize. +@param[in] thd thread handle +@param[in] var pointer to system variable +@param[out] save immediate result for update function +@param[in] value incoming string +@return 0 on success, 1 on failure. +*/ +static +int +innodb_buffer_pool_size_validate( + THD* thd, + struct st_mysql_sys_var* var, + void* save, + struct st_mysql_value* value); + +/* If the default value of innodb_buffer_pool_size is increased to be more than +BUF_POOL_SIZE_THRESHOLD (srv/srv0start.cc), then srv_buf_pool_instances_default +can be removed and 8 used instead. The problem with the current setup is that +with 128MiB default buffer pool size and 8 instances by default we would emit +a warning when no options are specified. */ static MYSQL_SYSVAR_LONGLONG(buffer_pool_size, innobase_buffer_pool_size, - PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, + PLUGIN_VAR_RQCMDARG, "The size of the memory buffer InnoDB uses to cache data and indexes of its tables.", - NULL, NULL, 128*1024*1024L, 5*1024*1024L, LONGLONG_MAX, 1024*1024L); + innodb_buffer_pool_size_validate, + innodb_buffer_pool_size_update, + static_cast<longlong>(srv_buf_pool_def_size), + static_cast<longlong>(srv_buf_pool_min_size), + LLONG_MAX, 1024*1024L); + +static MYSQL_SYSVAR_ULONG(buffer_pool_chunk_size, srv_buf_pool_chunk_unit, + PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, + "Size of a single memory chunk within each buffer pool instance" + " for resizing buffer pool. Online buffer pool resizing happens" + " at this granularity. 0 means disable resizing buffer pool.", + NULL, NULL, + 128 * 1024 * 1024, 1024 * 1024, LONG_MAX, 1024 * 1024); #if defined UNIV_DEBUG || defined UNIV_PERF_DEBUG static MYSQL_SYSVAR_ULONG(page_hash_locks, srv_n_page_hash_locks, @@ -19267,17 +20118,17 @@ static MYSQL_SYSVAR_ENUM(lock_schedule_algorithm, innodb_lock_schedule_algorithm "The algorithm Innodb uses for deciding which locks to grant next when" " a lock is released. Possible values are" " FCFS" - " grant the locks in First-Come-First-Served order;" + " grant the locks in First-Come-First-Served order;" " VATS" - " use the Variance-Aware-Transaction-Scheduling algorithm, which" - " uses an Eldest-Transaction-First heuristic.", + " use the Variance-Aware-Transaction-Scheduling algorithm, which" + " uses an Eldest-Transaction-First heuristic.", NULL, NULL, INNODB_LOCK_SCHEDULE_ALGORITHM_FCFS, &innodb_lock_schedule_algorithm_typelib); -static MYSQL_SYSVAR_LONG(buffer_pool_instances, innobase_buffer_pool_instances, +static MYSQL_SYSVAR_ULONG(buffer_pool_instances, srv_buf_pool_instances, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, "Number of buffer pool instances, set to higher value on high-end machines to increase scalability", - NULL, NULL, 0L, 0L, MAX_BUFFER_POOLS, 1L); + NULL, NULL, srv_buf_pool_instances_default, 0, MAX_BUFFER_POOLS, 0); static MYSQL_SYSVAR_STR(buffer_pool_filename, srv_buf_dump_filename, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_MEMALLOC, @@ -19292,12 +20143,12 @@ static MYSQL_SYSVAR_BOOL(buffer_pool_dump_now, innodb_buffer_pool_dump_now, static MYSQL_SYSVAR_BOOL(buffer_pool_dump_at_shutdown, srv_buffer_pool_dump_at_shutdown, PLUGIN_VAR_RQCMDARG, "Dump the buffer pool into a file named @@innodb_buffer_pool_filename", - NULL, NULL, FALSE); + NULL, NULL, TRUE); static MYSQL_SYSVAR_ULONG(buffer_pool_dump_pct, srv_buf_pool_dump_pct, PLUGIN_VAR_RQCMDARG, - "Dump only the hottest N% of each buffer pool, defaults to 100", - NULL, NULL, 100, 1, 100, 0); + "Dump only the hottest N% of each buffer pool, defaults to 25", + NULL, NULL, 25, 1, 100, 0); #ifdef UNIV_DEBUG static MYSQL_SYSVAR_STR(buffer_pool_evict, srv_buffer_pool_evict, @@ -19320,7 +20171,7 @@ static MYSQL_SYSVAR_BOOL(buffer_pool_load_abort, innodb_buffer_pool_load_abort, static MYSQL_SYSVAR_BOOL(buffer_pool_load_at_startup, srv_buffer_pool_load_at_startup, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, "Load the buffer pool from a file named @@innodb_buffer_pool_filename", - NULL, NULL, FALSE); + NULL, NULL, TRUE); static MYSQL_SYSVAR_BOOL(defragment, srv_defragment, PLUGIN_VAR_RQCMDARG, @@ -19399,10 +20250,17 @@ static MYSQL_SYSVAR_ULONG(concurrency_tickets, srv_n_free_tickets_to_enter, "Number of times a thread is allowed to enter InnoDB within the same SQL query after it has once got the ticket", NULL, NULL, 5000L, 1L, ~0UL, 0); -static MYSQL_SYSVAR_LONG(file_io_threads, innobase_file_io_threads, - PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY | PLUGIN_VAR_NOSYSVAR, - "Number of file I/O threads in InnoDB.", - NULL, NULL, 4, 4, 64, 0); +static MYSQL_SYSVAR_BOOL(deadlock_detect, innobase_deadlock_detect, + PLUGIN_VAR_NOCMDARG, + "Enable/disable InnoDB deadlock detector (default ON)." + " if set to OFF, deadlock detection is skipped," + " and we rely on innodb_lock_wait_timeout in case of deadlock.", + NULL, NULL, TRUE); + +static MYSQL_SYSVAR_LONG(fill_factor, innobase_fill_factor, + PLUGIN_VAR_RQCMDARG, + "Percentage of B-tree page filled during bulk insert", + NULL, NULL, 100, 10, 100, 0); static MYSQL_SYSVAR_BOOL(ft_enable_diag_print, fts_enable_diag_print, PLUGIN_VAR_OPCMDARG, @@ -19415,7 +20273,7 @@ static MYSQL_SYSVAR_BOOL(disable_sort_file_cache, srv_disable_sort_file_cache, NULL, NULL, FALSE); static MYSQL_SYSVAR_STR(ft_aux_table, fts_internal_tbl_name2, - PLUGIN_VAR_NOCMDARG, + PLUGIN_VAR_RQCMDARG, "FTS internal auxiliary table to be checked", innodb_internal_table_validate, innodb_internal_table_update, NULL); @@ -19445,7 +20303,6 @@ static MYSQL_SYSVAR_ULONG(ft_max_token_size, fts_max_token_size, "InnoDB Fulltext search maximum token size in characters", NULL, NULL, FTS_MAX_WORD_LEN_IN_CHAR, 10, FTS_MAX_WORD_LEN_IN_CHAR, 0); - static MYSQL_SYSVAR_ULONG(ft_num_word_optimize, fts_num_word_optimize, PLUGIN_VAR_OPCMDARG, "InnoDB Fulltext search number of words to optimize for each optimize table call ", @@ -19497,22 +20354,24 @@ static MYSQL_SYSVAR_LONG(log_buffer_size, innobase_log_buffer_size, "The size of the buffer which InnoDB uses to write log to the log files on disk.", NULL, NULL, 16*1024*1024L, 256*1024L, LONG_MAX, 1024); -static MYSQL_SYSVAR_LONGLONG(log_file_size, innobase_log_file_size, +static MYSQL_SYSVAR_ULONGLONG(log_file_size, srv_log_file_size, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, "Size of each log file in a log group.", - NULL, NULL, 48*1024*1024L, 1*1024*1024L, LONGLONG_MAX, 1024*1024L); + NULL, NULL, 48 << 20, 1 << 20, 512ULL << 30, UNIV_PAGE_SIZE_MAX); +/* OS_FILE_LOG_BLOCK_SIZE would be more appropriate than UNIV_PAGE_SIZE_MAX, +but fil_space_t is being used for the redo log, and it uses data pages. */ static MYSQL_SYSVAR_ULONG(log_files_in_group, srv_n_log_files, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, "Number of log files in the log group. InnoDB writes to the files in a circular fashion.", - NULL, NULL, 2, 2, SRV_N_LOG_FILES_MAX, 0); + NULL, NULL, 2, 1, SRV_N_LOG_FILES_MAX, 0); -/* Note that the default and minimum values are set to 0 to -detect if the option is passed and print deprecation message */ -static MYSQL_SYSVAR_LONG(mirrored_log_groups, innobase_mirrored_log_groups, - PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, - "Number of identical copies of log groups we keep for the database. Currently this should be set to 1.", - NULL, NULL, 0, 0, 10, 0); +static MYSQL_SYSVAR_ULONG(log_write_ahead_size, srv_log_write_ahead_size, + PLUGIN_VAR_RQCMDARG, + "Redo log write ahead unit size to avoid read-on-write," + " it should match the OS cache block IO size", + NULL, innodb_log_write_ahead_size_update, + 8*1024L, OS_FILE_LOG_BLOCK_SIZE, UNIV_PAGE_SIZE_DEF, OS_FILE_LOG_BLOCK_SIZE); static MYSQL_SYSVAR_UINT(old_blocks_pct, innobase_old_blocks_pct, PLUGIN_VAR_RQCMDARG, @@ -19536,17 +20395,16 @@ static MYSQL_SYSVAR_ULONG(sync_spin_loops, srv_n_spin_wait_rounds, "Count of spin-loop rounds in InnoDB mutexes (30 by default)", NULL, NULL, 30L, 0L, ~0UL, 0); -static MYSQL_SYSVAR_ULONG(spin_wait_delay, srv_spin_wait_delay, +static MYSQL_SYSVAR_UINT(spin_wait_delay, srv_spin_wait_delay, PLUGIN_VAR_OPCMDARG, "Maximum delay between polling for a spin lock (6 by default)", - NULL, NULL, 6L, 0L, ~0UL, 0); + NULL, NULL, 6, 0, 6000, 0); static MYSQL_SYSVAR_ULONG(thread_concurrency, srv_thread_concurrency, PLUGIN_VAR_RQCMDARG, "Helps in performance tuning in heavily concurrent environments. Sets the maximum number of threads allowed inside InnoDB. Value 0 will disable the thread throttling.", NULL, NULL, 0, 0, 1000, 0); -#ifdef HAVE_ATOMIC_BUILTINS static MYSQL_SYSVAR_ULONG( adaptive_max_sleep_delay, srv_adaptive_max_sleep_delay, PLUGIN_VAR_RQCMDARG, @@ -19555,7 +20413,6 @@ static MYSQL_SYSVAR_ULONG( 150000, /* Default setting */ 0, /* Minimum value */ 1000000, 0); /* Maximum value */ -#endif /* HAVE_ATOMIC_BUILTINS */ static MYSQL_SYSVAR_BOOL(prefix_index_cluster_optimization, srv_prefix_index_cluster_optimization, @@ -19565,8 +20422,8 @@ static MYSQL_SYSVAR_BOOL(prefix_index_cluster_optimization, static MYSQL_SYSVAR_ULONG(thread_sleep_delay, srv_thread_sleep_delay, PLUGIN_VAR_RQCMDARG, - "Time of innodb thread sleeping before joining InnoDB queue (usec). " - "Value 0 disable a sleep", + "Time of innodb thread sleeping before joining InnoDB queue (usec)." + " Value 0 disable a sleep", NULL, NULL, 10000L, 0L, @@ -19577,18 +20434,23 @@ static MYSQL_SYSVAR_STR(data_file_path, innobase_data_file_path, "Path to individual files and their sizes.", NULL, NULL, NULL); +static MYSQL_SYSVAR_STR(temp_data_file_path, innobase_temp_data_file_path, + PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, + "Path to files and their sizes making temp-tablespace.", + NULL, NULL, NULL); + static MYSQL_SYSVAR_STR(undo_directory, srv_undo_dir, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, "Directory where undo tablespace files live, this path can be absolute.", - NULL, NULL, "."); + NULL, NULL, NULL); static MYSQL_SYSVAR_ULONG(undo_tablespaces, srv_undo_tablespaces, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, - "Number of undo tablespaces to use. ", + "Number of undo tablespaces to use.", NULL, NULL, 0L, /* Default seting */ 0L, /* Minimum value */ - 126L, 0); /* Maximum value */ + TRX_SYS_MAX_UNDO_SPACES, 0); /* Maximum value */ static MYSQL_SYSVAR_ULONG(undo_logs, srv_undo_logs, PLUGIN_VAR_OPCMDARG, @@ -19598,6 +20460,25 @@ static MYSQL_SYSVAR_ULONG(undo_logs, srv_undo_logs, 1, /* Minimum value */ TRX_SYS_N_RSEGS, 0); /* Maximum value */ +static MYSQL_SYSVAR_ULONGLONG(max_undo_log_size, srv_max_undo_log_size, + PLUGIN_VAR_OPCMDARG, + "Desired maximum UNDO tablespace size in bytes", + NULL, NULL, + 10 << 20, 10 << 20, + 1ULL << (32 + UNIV_PAGE_SIZE_SHIFT_MAX), 0); + +static MYSQL_SYSVAR_ULONG(purge_rseg_truncate_frequency, + srv_purge_rseg_truncate_frequency, + PLUGIN_VAR_OPCMDARG, + "Dictates rate at which UNDO records are purged. Value N means" + " purge rollback segment(s) on every Nth iteration of purge invocation", + NULL, NULL, 128, 1, 128, 0); + +static MYSQL_SYSVAR_BOOL(undo_log_truncate, srv_undo_log_truncate, + PLUGIN_VAR_OPCMDARG, + "Enable or Disable Truncate of UNDO tablespace.", + NULL, NULL, FALSE); + /* Alias for innodb_undo_logs, this config variable is deprecated. */ static MYSQL_SYSVAR_ULONG(rollback_segments, srv_undo_logs, PLUGIN_VAR_OPCMDARG, @@ -19609,11 +20490,10 @@ static MYSQL_SYSVAR_ULONG(rollback_segments, srv_undo_logs, static MYSQL_SYSVAR_LONG(autoinc_lock_mode, innobase_autoinc_lock_mode, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, - "The AUTOINC lock modes supported by InnoDB: " - "0 => Old style AUTOINC locking (for backward" - " compatibility) " - "1 => New style AUTOINC locking " - "2 => No AUTOINC locking (unsafe for SBR)", + "The AUTOINC lock modes supported by InnoDB:" + " 0 => Old style AUTOINC locking (for backward compatibility);" + " 1 => New style AUTOINC locking;" + " 2 => No AUTOINC locking (unsafe for SBR)", NULL, NULL, AUTOINC_NEW_STYLE_LOCKING, /* Default setting */ AUTOINC_OLD_STYLE_LOCKING, /* Minimum value */ @@ -19623,13 +20503,6 @@ static MYSQL_SYSVAR_STR(version, innodb_version_str, PLUGIN_VAR_NOCMDOPT | PLUGIN_VAR_READONLY, "InnoDB version", NULL, NULL, INNODB_VERSION_STR); -static MYSQL_SYSVAR_BOOL(use_sys_malloc, srv_use_sys_malloc, - PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "DEPRECATED. This option may be removed in future releases, " - "together with the InnoDB's internal memory allocator. " - "Use OS memory allocator instead of InnoDB's internal memory allocator", - NULL, NULL, TRUE); - static MYSQL_SYSVAR_BOOL(use_native_aio, srv_use_native_aio, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, "Use native AIO if supported on this platform.", @@ -19640,48 +20513,17 @@ static MYSQL_SYSVAR_BOOL(numa_interleave, srv_numa_interleave, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, "Use NUMA interleave memory policy to allocate InnoDB buffer pool.", NULL, NULL, FALSE); -#endif // HAVE_LIBNUMA - -static MYSQL_SYSVAR_BOOL(api_enable_binlog, ib_binlog_enabled, - PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Enable binlog for applications direct access InnoDB through InnoDB APIs", - NULL, NULL, FALSE); - -static MYSQL_SYSVAR_BOOL(api_enable_mdl, ib_mdl_enabled, - PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Enable MDL for applications direct access InnoDB through InnoDB APIs", - NULL, NULL, FALSE); - -static MYSQL_SYSVAR_BOOL(api_disable_rowlock, ib_disable_row_lock, - PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Disable row lock when direct access InnoDB through InnoDB APIs", - NULL, NULL, FALSE); - -static MYSQL_SYSVAR_ULONG(api_trx_level, ib_trx_level_setting, - PLUGIN_VAR_OPCMDARG, - "InnoDB API transaction isolation level", - NULL, NULL, - 0, /* Default setting */ - 0, /* Minimum value */ - 3, 0); /* Maximum value */ - -static MYSQL_SYSVAR_ULONG(api_bk_commit_interval, ib_bk_commit_interval, - PLUGIN_VAR_OPCMDARG, - "Background commit interval in seconds", - NULL, NULL, - 5, /* Default setting */ - 1, /* Minimum value */ - 1024 * 1024 * 1024, 0); /* Maximum value */ +#endif /* HAVE_LIBNUMA */ static MYSQL_SYSVAR_STR(change_buffering, innobase_change_buffering, PLUGIN_VAR_RQCMDARG, - "Buffer changes to reduce random access: " - "OFF, ON, inserting, deleting, changing, or purging.", + "Buffer changes to reduce random access:" + " OFF, ON, inserting, deleting, changing, or purging.", innodb_change_buffering_validate, innodb_change_buffering_update, "all"); static MYSQL_SYSVAR_UINT(change_buffer_max_size, - innobase_change_buffer_max_size, + srv_change_buffer_max_size, PLUGIN_VAR_RQCMDARG, "Maximum on-disk size of change buffer in terms of percentage" " of the buffer pool.", @@ -19690,9 +20532,9 @@ static MYSQL_SYSVAR_UINT(change_buffer_max_size, static MYSQL_SYSVAR_ENUM(stats_method, srv_innodb_stats_method, PLUGIN_VAR_RQCMDARG, - "Specifies how InnoDB index statistics collection code should " - "treat NULLs. Possible values are NULLS_EQUAL (default), " - "NULLS_UNEQUAL and NULLS_IGNORED", + "Specifies how InnoDB index statistics collection code should" + " treat NULLs. Possible values are NULLS_EQUAL (default)," + " NULLS_UNEQUAL and NULLS_IGNORED", NULL, NULL, SRV_STATS_NULLS_EQUAL, &innodb_stats_method_typelib); #if defined UNIV_DEBUG || defined UNIV_IBUF_DEBUG @@ -19739,10 +20581,11 @@ innobase_disallow_writes_update( { *(my_bool*)var_ptr = *(my_bool*)save; ut_a(srv_allow_writes_event); - if (*(my_bool*)var_ptr) + if (*(my_bool*)var_ptr) { os_event_reset(srv_allow_writes_event); - else + } else { os_event_set(srv_allow_writes_event); + } } static MYSQL_SYSVAR_BOOL(disallow_writes, innobase_disallow_writes, @@ -19750,6 +20593,7 @@ static MYSQL_SYSVAR_BOOL(disallow_writes, innobase_disallow_writes, "Tell InnoDB to stop any writes to disk", NULL, innobase_disallow_writes_update, FALSE); #endif /* WITH_INNODB_DISALLOW_WRITES */ + static MYSQL_SYSVAR_BOOL(random_read_ahead, srv_random_read_ahead, PLUGIN_VAR_NOCMDARG, "Whether to use read ahead for random access within an extent.", @@ -19757,8 +20601,8 @@ static MYSQL_SYSVAR_BOOL(random_read_ahead, srv_random_read_ahead, static MYSQL_SYSVAR_ULONG(read_ahead_threshold, srv_read_ahead_threshold, PLUGIN_VAR_RQCMDARG, - "Number of pages that must be accessed sequentially for InnoDB to " - "trigger a readahead.", + "Number of pages that must be accessed sequentially for InnoDB to" + " trigger a readahead.", NULL, NULL, 56, 0, 64, 0); static MYSQL_SYSVAR_STR(monitor_enable, innobase_enable_monitor_counter, @@ -19796,7 +20640,7 @@ static MYSQL_SYSVAR_BOOL(status_output_locks, srv_print_innodb_lock_monitor, static MYSQL_SYSVAR_BOOL(print_all_deadlocks, srv_print_all_deadlocks, PLUGIN_VAR_OPCMDARG, - "Print all deadlocks to MySQL error log (off by default)", + "Print all deadlocks to MariaDB error log (off by default)", NULL, NULL, FALSE); static MYSQL_SYSVAR_ULONG(compression_failure_threshold_pct, @@ -19817,10 +20661,15 @@ static MYSQL_SYSVAR_BOOL(read_only, srv_read_only_mode, "Start InnoDB in read only mode (off by default)", NULL, NULL, FALSE); +static MYSQL_SYSVAR_BOOL(safe_truncate, srv_safe_truncate, + PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY, + "Use backup-safe TRUNCATE TABLE and crash-safe RENAME (incompatible with older MariaDB 10.2; ON by default)", + NULL, NULL, TRUE); + static MYSQL_SYSVAR_BOOL(cmp_per_index_enabled, srv_cmp_per_index_enabled, PLUGIN_VAR_OPCMDARG, - "Enable INFORMATION_SCHEMA.innodb_cmp_per_index, " - "may have negative impact on performance (off by default)", + "Enable INFORMATION_SCHEMA.innodb_cmp_per_index," + " may have negative impact on performance (off by default)", NULL, innodb_cmp_per_index_update, FALSE); static MYSQL_SYSVAR_ENUM(default_row_format, innodb_default_row_format, @@ -19828,7 +20677,7 @@ static MYSQL_SYSVAR_ENUM(default_row_format, innodb_default_row_format, "The default ROW FORMAT for all innodb tables created without explicit" " ROW_FORMAT. Possible values are REDUNDANT, COMPACT, and DYNAMIC." " The ROW_FORMAT value COMPRESSED is not allowed", - NULL, NULL, DEFAULT_ROW_FORMAT_COMPACT, + NULL, NULL, DEFAULT_ROW_FORMAT_DYNAMIC, &innodb_default_row_format_typelib); #ifdef UNIV_DEBUG @@ -19844,9 +20693,9 @@ static MYSQL_SYSVAR_UINT(limit_optimistic_insert_debug, static MYSQL_SYSVAR_BOOL(trx_purge_view_update_only_debug, srv_purge_view_update_only_debug, PLUGIN_VAR_NOCMDOPT, - "Pause actual purging any delete-marked records, but merely update the purge view. " - "It is to create artificially the situation the purge view have been updated " - "but the each purges were not done yet.", + "Pause actual purging any delete-marked records, but merely update the purge view." + " It is to create artificially the situation the purge view have been updated" + " but the each purges were not done yet.", NULL, NULL, FALSE); static MYSQL_SYSVAR_UINT(data_file_size_debug, @@ -19864,12 +20713,40 @@ static MYSQL_SYSVAR_ULONG(saved_page_number_debug, srv_saved_page_number_debug, PLUGIN_VAR_OPCMDARG, "An InnoDB page number.", NULL, innodb_save_page_no, 0, 0, UINT_MAX32, 0); -#endif /* UNIV_DEBUG */ + +static MYSQL_SYSVAR_BOOL(disable_resize_buffer_pool_debug, + buf_disable_resize_buffer_pool_debug, PLUGIN_VAR_NOCMDARG, + "Disable resizing buffer pool to make assertion code not expensive.", + NULL, NULL, TRUE); + +static MYSQL_SYSVAR_BOOL(page_cleaner_disabled_debug, + innodb_page_cleaner_disabled_debug, + PLUGIN_VAR_OPCMDARG, + "Disable page cleaner", + NULL, buf_flush_page_cleaner_disabled_debug_update, FALSE); + +static MYSQL_SYSVAR_BOOL(sync_debug, srv_sync_debug, + PLUGIN_VAR_OPCMDARG | PLUGIN_VAR_READONLY, + "Enable the sync debug checks", + NULL, NULL, FALSE); + +static MYSQL_SYSVAR_BOOL(dict_stats_disabled_debug, + innodb_dict_stats_disabled_debug, + PLUGIN_VAR_OPCMDARG, + "Disable dict_stats thread", + NULL, dict_stats_disabled_debug_update, FALSE); + +static MYSQL_SYSVAR_BOOL(master_thread_disabled_debug, + srv_master_thread_disabled_debug, + PLUGIN_VAR_OPCMDARG, + "Disable master thread", + NULL, srv_master_thread_disabled_debug_update, FALSE); static MYSQL_SYSVAR_UINT(simulate_comp_failures, srv_simulate_comp_failures, PLUGIN_VAR_NOCMDARG, "Simulate compression failures.", NULL, NULL, 0, 0, 99, 0); +#endif /* UNIV_DEBUG */ static MYSQL_SYSVAR_BOOL(force_primary_key, srv_force_primary_key, @@ -19879,8 +20756,8 @@ static MYSQL_SYSVAR_BOOL(force_primary_key, static MYSQL_SYSVAR_BOOL(use_trim, srv_use_trim, PLUGIN_VAR_OPCMDARG, - "Use trim. Default FALSE.", - NULL, NULL, FALSE); + "Deallocate (punch_hole|trim) unused portions of the page compressed page (on by default)", + NULL, innodb_use_trim_update, TRUE); static const char *page_compression_algorithms[]= { "none", "zlib", "lz4", "lzo", "lzma", "bzip2", "snappy", 0 }; static TYPELIB page_compression_algorithms_typelib= @@ -19900,7 +20777,7 @@ static MYSQL_SYSVAR_ENUM(compression_algorithm, innodb_compression_algorithm, static MYSQL_SYSVAR_LONG(mtflush_threads, srv_mtflush_threads, PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_READONLY, - "Number of multi-threaded flush threads", + "DEPRECATED. Number of multi-threaded flush threads", NULL, NULL, MTFLUSH_DEFAULT_WORKER, /* Default setting */ 1, /* Minimum setting */ @@ -19909,7 +20786,7 @@ static MYSQL_SYSVAR_LONG(mtflush_threads, srv_mtflush_threads, static MYSQL_SYSVAR_BOOL(use_mtflush, srv_use_mtflush, PLUGIN_VAR_NOCMDARG | PLUGIN_VAR_READONLY, - "Use multi-threaded flush. Default FALSE.", + "DEPRECATED. Use multi-threaded flush. Default FALSE.", NULL, NULL, FALSE); static MYSQL_SYSVAR_ULONG(fatal_semaphore_wait_threshold, srv_fatal_semaphore_wait_threshold, @@ -20028,18 +20905,17 @@ static MYSQL_SYSVAR_BOOL(debug_force_scrubbing, NULL, NULL, FALSE); #endif /* UNIV_DEBUG */ -static MYSQL_SYSVAR_BOOL(instrument_semaphores, srv_instrument_semaphores, +static MYSQL_SYSVAR_BOOL(instrument_semaphores, innodb_instrument_semaphores, PLUGIN_VAR_OPCMDARG, - "Enable semaphore request instrumentation. This could have some effect on performance but allows better" - " information on long semaphore wait problems. (Default: not enabled)", - 0, 0, FALSE); + "DEPRECATED. This setting has no effect.", + NULL, innodb_instrument_semaphores_update, FALSE); + +#include "ha_xtradb.h" static struct st_mysql_sys_var* innobase_system_variables[]= { - MYSQL_SYSVAR(additional_mem_pool_size), - MYSQL_SYSVAR(api_trx_level), - MYSQL_SYSVAR(api_bk_commit_interval), MYSQL_SYSVAR(autoextend_increment), MYSQL_SYSVAR(buffer_pool_size), + MYSQL_SYSVAR(buffer_pool_chunk_size), MYSQL_SYSVAR(buffer_pool_instances), MYSQL_SYSVAR(buffer_pool_filename), MYSQL_SYSVAR(buffer_pool_dump_now), @@ -20060,21 +20936,19 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(lru_scan_depth), MYSQL_SYSVAR(flush_neighbors), MYSQL_SYSVAR(checksum_algorithm), + MYSQL_SYSVAR(log_checksums), MYSQL_SYSVAR(checksums), MYSQL_SYSVAR(commit_concurrency), MYSQL_SYSVAR(concurrency_tickets), MYSQL_SYSVAR(compression_level), MYSQL_SYSVAR(data_file_path), + MYSQL_SYSVAR(temp_data_file_path), MYSQL_SYSVAR(data_home_dir), MYSQL_SYSVAR(doublewrite), + MYSQL_SYSVAR(stats_include_delete_marked), MYSQL_SYSVAR(use_atomic_writes), MYSQL_SYSVAR(use_fallocate), - MYSQL_SYSVAR(stats_include_delete_marked), - MYSQL_SYSVAR(api_enable_binlog), - MYSQL_SYSVAR(api_enable_mdl), - MYSQL_SYSVAR(api_disable_rowlock), MYSQL_SYSVAR(fast_shutdown), - MYSQL_SYSVAR(file_io_threads), MYSQL_SYSVAR(read_io_threads), MYSQL_SYSVAR(write_io_threads), MYSQL_SYSVAR(file_per_table), @@ -20085,6 +20959,7 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(flush_log_at_trx_commit), MYSQL_SYSVAR(flush_method), MYSQL_SYSVAR(force_recovery), + MYSQL_SYSVAR(fill_factor), MYSQL_SYSVAR(ft_cache_size), MYSQL_SYSVAR(ft_total_cache_size), MYSQL_SYSVAR(ft_result_cache_limit), @@ -20098,24 +20973,23 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(lock_schedule_algorithm), MYSQL_SYSVAR(locks_unsafe_for_binlog), MYSQL_SYSVAR(lock_wait_timeout), -#ifdef UNIV_LOG_ARCHIVE - MYSQL_SYSVAR(log_arch_dir), - MYSQL_SYSVAR(log_archive), -#endif /* UNIV_LOG_ARCHIVE */ + MYSQL_SYSVAR(deadlock_detect), MYSQL_SYSVAR(page_size), MYSQL_SYSVAR(log_buffer_size), MYSQL_SYSVAR(log_file_size), MYSQL_SYSVAR(log_files_in_group), + MYSQL_SYSVAR(log_write_ahead_size), MYSQL_SYSVAR(log_group_home_dir), MYSQL_SYSVAR(log_compressed_pages), + MYSQL_SYSVAR(log_optimize_ddl), MYSQL_SYSVAR(max_dirty_pages_pct), MYSQL_SYSVAR(max_dirty_pages_pct_lwm), MYSQL_SYSVAR(adaptive_flushing_lwm), MYSQL_SYSVAR(adaptive_flushing), + MYSQL_SYSVAR(flush_sync), MYSQL_SYSVAR(flushing_avg_loops), MYSQL_SYSVAR(max_purge_lag), MYSQL_SYSVAR(max_purge_lag_delay), - MYSQL_SYSVAR(mirrored_log_groups), MYSQL_SYSVAR(old_blocks_pct), MYSQL_SYSVAR(old_blocks_time), MYSQL_SYSVAR(open_files), @@ -20134,7 +21008,10 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(stats_auto_recalc), MYSQL_SYSVAR(stats_modified_counter), MYSQL_SYSVAR(stats_traditional), +#ifdef BTR_CUR_HASH_ADAPT MYSQL_SYSVAR(adaptive_hash_index), + MYSQL_SYSVAR(adaptive_hash_index_parts), +#endif /* BTR_CUR_HASH_ADAPT */ MYSQL_SYSVAR(stats_method), MYSQL_SYSVAR(replication_delay), MYSQL_SYSVAR(status_file), @@ -20146,18 +21023,16 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(spin_wait_delay), MYSQL_SYSVAR(table_locks), MYSQL_SYSVAR(thread_concurrency), -#ifdef HAVE_ATOMIC_BUILTINS MYSQL_SYSVAR(adaptive_max_sleep_delay), -#endif /* HAVE_ATOMIC_BUILTINS */ MYSQL_SYSVAR(prefix_index_cluster_optimization), MYSQL_SYSVAR(thread_sleep_delay), + MYSQL_SYSVAR(tmpdir), MYSQL_SYSVAR(autoinc_lock_mode), MYSQL_SYSVAR(version), - MYSQL_SYSVAR(use_sys_malloc), MYSQL_SYSVAR(use_native_aio), #ifdef HAVE_LIBNUMA MYSQL_SYSVAR(numa_interleave), -#endif // HAVE_LIBNUMA +#endif /* HAVE_LIBNUMA */ MYSQL_SYSVAR(change_buffering), MYSQL_SYSVAR(change_buffer_max_size), #if defined UNIV_DEBUG || defined UNIV_IBUF_DEBUG @@ -20170,8 +21045,10 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(random_read_ahead), MYSQL_SYSVAR(read_ahead_threshold), MYSQL_SYSVAR(read_only), + MYSQL_SYSVAR(safe_truncate), MYSQL_SYSVAR(io_capacity), MYSQL_SYSVAR(io_capacity_max), + MYSQL_SYSVAR(page_cleaners), MYSQL_SYSVAR(idle_flush_pct), MYSQL_SYSVAR(monitor_enable), MYSQL_SYSVAR(monitor_disable), @@ -20180,10 +21057,10 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(purge_threads), MYSQL_SYSVAR(purge_batch_size), #ifdef UNIV_DEBUG - MYSQL_SYSVAR(purge_run_now), - MYSQL_SYSVAR(purge_stop_now), + MYSQL_SYSVAR(background_drop_list_empty), MYSQL_SYSVAR(log_checkpoint_now), MYSQL_SYSVAR(buf_flush_list_now), + MYSQL_SYSVAR(merge_threshold_set_all_debug), #endif /* UNIV_DEBUG */ #if defined UNIV_DEBUG || defined UNIV_PERF_DEBUG MYSQL_SYSVAR(page_hash_locks), @@ -20194,6 +21071,9 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(print_all_deadlocks), MYSQL_SYSVAR(cmp_per_index_enabled), MYSQL_SYSVAR(undo_logs), + MYSQL_SYSVAR(max_undo_log_size), + MYSQL_SYSVAR(purge_rseg_truncate_frequency), + MYSQL_SYSVAR(undo_log_truncate), MYSQL_SYSVAR(rollback_segments), MYSQL_SYSVAR(undo_directory), MYSQL_SYSVAR(undo_tablespaces), @@ -20201,20 +21081,25 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(compression_failure_threshold_pct), MYSQL_SYSVAR(compression_pad_pct_max), MYSQL_SYSVAR(default_row_format), - MYSQL_SYSVAR(simulate_comp_failures), #ifdef UNIV_DEBUG + MYSQL_SYSVAR(simulate_comp_failures), MYSQL_SYSVAR(trx_rseg_n_slots_debug), MYSQL_SYSVAR(limit_optimistic_insert_debug), MYSQL_SYSVAR(trx_purge_view_update_only_debug), MYSQL_SYSVAR(data_file_size_debug), MYSQL_SYSVAR(fil_make_page_dirty_debug), MYSQL_SYSVAR(saved_page_number_debug), + MYSQL_SYSVAR(disable_resize_buffer_pool_debug), + MYSQL_SYSVAR(page_cleaner_disabled_debug), + MYSQL_SYSVAR(dict_stats_disabled_debug), + MYSQL_SYSVAR(master_thread_disabled_debug), + MYSQL_SYSVAR(sync_debug), #endif /* UNIV_DEBUG */ - MYSQL_SYSVAR(tmpdir), MYSQL_SYSVAR(force_primary_key), MYSQL_SYSVAR(fatal_semaphore_wait_threshold), /* Table page compression feature */ MYSQL_SYSVAR(use_trim), + MYSQL_SYSVAR(compression_default), MYSQL_SYSVAR(compression_algorithm), MYSQL_SYSVAR(mtflush_threads), MYSQL_SYSVAR(use_mtflush), @@ -20238,6 +21123,12 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { #endif MYSQL_SYSVAR(instrument_semaphores), MYSQL_SYSVAR(buf_dump_status_frequency), + MYSQL_SYSVAR(background_thread), + + /* XtraDB compatibility system variables */ +#define HA_XTRADB_SYSVARS +#include "ha_xtradb.h" + NULL }; @@ -20285,6 +21176,7 @@ i_s_innodb_sys_foreign, i_s_innodb_sys_foreign_cols, i_s_innodb_sys_tablespaces, i_s_innodb_sys_datafiles, +i_s_innodb_sys_virtual, i_s_innodb_mutexes, i_s_innodb_sys_semaphore_waits, i_s_innodb_tablespaces_encryption, @@ -20309,219 +21201,626 @@ innobase_commit_concurrency_init_default() = innobase_commit_concurrency; } -/** @brief Initialize the default and max value of innodb_undo_logs. - -Once InnoDB is running, the default value and the max value of -innodb_undo_logs must be equal to the available undo logs, -given by srv_available_undo_logs. */ +/** @brief Adjust some InnoDB startup parameters based on file contents +or innodb_page_size. */ static void -innobase_undo_logs_init_default_max() -/*=================================*/ +innodb_params_adjust() { + /* The default value and the max value of + innodb_undo_logs must be equal to the available undo logs. */ MYSQL_SYSVAR_NAME(undo_logs).max_val = MYSQL_SYSVAR_NAME(undo_logs).def_val - = static_cast<unsigned long>(srv_available_undo_logs); + = srv_available_undo_logs; + MYSQL_SYSVAR_NAME(max_undo_log_size).max_val + = 1ULL << (32 + UNIV_PAGE_SIZE_SHIFT); + MYSQL_SYSVAR_NAME(max_undo_log_size).min_val + = MYSQL_SYSVAR_NAME(max_undo_log_size).def_val + = ulonglong(SRV_UNDO_TABLESPACE_SIZE_IN_PAGES) + * srv_page_size; + MYSQL_SYSVAR_NAME(max_undo_log_size).max_val + = 1ULL << (32 + UNIV_PAGE_SIZE_SHIFT); } -#ifdef UNIV_COMPILE_TEST_FUNCS +/**************************************************************************** + * DS-MRR implementation + ***************************************************************************/ -struct innobase_convert_name_test_t { - char* buf; - ulint buflen; - const char* id; - ulint idlen; - void* thd; - ibool file_id; +/** +Multi Range Read interface, DS-MRR calls */ +int +ha_innobase::multi_range_read_init( + RANGE_SEQ_IF* seq, + void* seq_init_param, + uint n_ranges, + uint mode, + HANDLER_BUFFER* buf) +{ + return(m_ds_mrr.dsmrr_init(this, seq, seq_init_param, + n_ranges, mode, buf)); +} - const char* expected; -}; +int +ha_innobase::multi_range_read_next( + range_id_t* range_info) +{ + return(m_ds_mrr.dsmrr_next(range_info)); +} -void -test_innobase_convert_name() -{ - char buf[1024]; - ulint i; - - innobase_convert_name_test_t test_input[] = { - {buf, sizeof(buf), "abcd", 4, NULL, TRUE, "\"abcd\""}, - {buf, 7, "abcd", 4, NULL, TRUE, "\"abcd\""}, - {buf, 6, "abcd", 4, NULL, TRUE, "\"abcd\""}, - {buf, 5, "abcd", 4, NULL, TRUE, "\"abc\""}, - {buf, 4, "abcd", 4, NULL, TRUE, "\"ab\""}, - - {buf, sizeof(buf), "ab@0060cd", 9, NULL, TRUE, "\"ab`cd\""}, - {buf, 9, "ab@0060cd", 9, NULL, TRUE, "\"ab`cd\""}, - {buf, 8, "ab@0060cd", 9, NULL, TRUE, "\"ab`cd\""}, - {buf, 7, "ab@0060cd", 9, NULL, TRUE, "\"ab`cd\""}, - {buf, 6, "ab@0060cd", 9, NULL, TRUE, "\"ab`c\""}, - {buf, 5, "ab@0060cd", 9, NULL, TRUE, "\"ab`\""}, - {buf, 4, "ab@0060cd", 9, NULL, TRUE, "\"ab\""}, - - {buf, sizeof(buf), "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#ab\"\"cd\""}, - {buf, 17, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#ab\"\"cd\""}, - {buf, 16, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#ab\"\"c\""}, - {buf, 15, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#ab\"\"\""}, - {buf, 14, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#ab\""}, - {buf, 13, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#ab\""}, - {buf, 12, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#a\""}, - {buf, 11, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50#\""}, - {buf, 10, "ab\"cd", 5, NULL, TRUE, - "\"#mysql50\""}, - - {buf, sizeof(buf), "ab/cd", 5, NULL, TRUE, "\"ab\".\"cd\""}, - {buf, 9, "ab/cd", 5, NULL, TRUE, "\"ab\".\"cd\""}, - {buf, 8, "ab/cd", 5, NULL, TRUE, "\"ab\".\"c\""}, - {buf, 7, "ab/cd", 5, NULL, TRUE, "\"ab\".\"\""}, - {buf, 6, "ab/cd", 5, NULL, TRUE, "\"ab\"."}, - {buf, 5, "ab/cd", 5, NULL, TRUE, "\"ab\"."}, - {buf, 4, "ab/cd", 5, NULL, TRUE, "\"ab\""}, - {buf, 3, "ab/cd", 5, NULL, TRUE, "\"a\""}, - {buf, 2, "ab/cd", 5, NULL, TRUE, "\"\""}, - /* XXX probably "" is a better result in this case - {buf, 1, "ab/cd", 5, NULL, TRUE, "."}, - */ - {buf, 0, "ab/cd", 5, NULL, TRUE, ""}, - }; +ha_rows +ha_innobase::multi_range_read_info_const( + uint keyno, + RANGE_SEQ_IF* seq, + void* seq_init_param, + uint n_ranges, + uint* bufsz, + uint* flags, + Cost_estimate* cost) +{ + /* See comments in ha_myisam::multi_range_read_info_const */ + m_ds_mrr.init(this, table); - for (i = 0; i < sizeof(test_input) / sizeof(test_input[0]); i++) { + if (m_prebuilt->select_lock_type != LOCK_NONE) { + *flags |= HA_MRR_USE_DEFAULT_IMPL; + } - char* end; - ibool ok = TRUE; - size_t res_len; + ha_rows res= m_ds_mrr.dsmrr_info_const(keyno, seq, seq_init_param, n_ranges, + bufsz, flags, cost); + return res; +} - fprintf(stderr, "TESTING %lu, %s, %lu, %s\n", - test_input[i].buflen, - test_input[i].id, - test_input[i].idlen, - test_input[i].expected); +ha_rows +ha_innobase::multi_range_read_info( + uint keyno, + uint n_ranges, + uint keys, + uint key_parts, + uint* bufsz, + uint* flags, + Cost_estimate* cost) +{ + m_ds_mrr.init(this, table); + ha_rows res= m_ds_mrr.dsmrr_info(keyno, n_ranges, keys, key_parts, bufsz, + flags, cost); + return res; +} - end = innobase_convert_name( - test_input[i].buf, - test_input[i].buflen, - test_input[i].id, - test_input[i].idlen, - test_input[i].thd, - test_input[i].file_id); +int +ha_innobase::multi_range_read_explain_info( + uint mrr_mode, + char *str, + size_t size) +{ + return m_ds_mrr.dsmrr_explain_info(mrr_mode, str, size); +} - res_len = (size_t) (end - test_input[i].buf); +/** +Index Condition Pushdown interface implementation */ - if (res_len != strlen(test_input[i].expected)) { +/*************************************************************//** +InnoDB index push-down condition check +@return ICP_NO_MATCH, ICP_MATCH, or ICP_OUT_OF_RANGE */ +ICP_RESULT +innobase_index_cond( +/*================*/ + void* file) /*!< in/out: pointer to ha_innobase */ +{ + return handler_index_cond_check(file); +} - fprintf(stderr, "unexpected len of the result: %u, " - "expected: %u\n", (unsigned) res_len, - (unsigned) strlen(test_input[i].expected)); - ok = FALSE; - } +/** Parse the table file name into table name and database name. +@param[in] tbl_name InnoDB table name +@param[out] dbname database name buffer (NAME_LEN + 1 bytes) +@param[out] tblname table name buffer (NAME_LEN + 1 bytes) +@param[out] dbnamelen database name length +@param[out] tblnamelen table name length +@return true if the table name is parsed properly. */ +static bool table_name_parse( + const table_name_t& tbl_name, + char* dbname, + char* tblname, + ulint& dbnamelen, + ulint& tblnamelen) +{ + dbnamelen = dict_get_db_name_len(tbl_name.m_name); + char db_buf[MAX_DATABASE_NAME_LEN + 1]; + char tbl_buf[MAX_TABLE_NAME_LEN + 1]; - if (memcmp(test_input[i].buf, - test_input[i].expected, - strlen(test_input[i].expected)) != 0 - || !ok) { + ut_ad(dbnamelen > 0); + ut_ad(dbnamelen <= MAX_DATABASE_NAME_LEN); - fprintf(stderr, "unexpected result: %.*s, " - "expected: %s\n", (int) res_len, - test_input[i].buf, - test_input[i].expected); - ok = FALSE; - } + memcpy(db_buf, tbl_name.m_name, dbnamelen); + db_buf[dbnamelen] = 0; - if (ok) { - fprintf(stderr, "OK: res: %.*s\n\n", (int) res_len, - buf); - } else { - fprintf(stderr, "FAILED\n\n"); - return; - } - } -} + tblnamelen = strlen(tbl_name.m_name + dbnamelen + 1); + memcpy(tbl_buf, tbl_name.m_name + dbnamelen + 1, tblnamelen); + tbl_buf[tblnamelen] = 0; -#endif /* UNIV_COMPILE_TEST_FUNCS */ + filename_to_tablename(db_buf, dbname, MAX_DATABASE_NAME_LEN + 1, true); -/**************************************************************************** - * DS-MRR implementation - ***************************************************************************/ + if (tblnamelen > TEMP_FILE_PREFIX_LENGTH + && !strncmp(tbl_buf, TEMP_FILE_PREFIX, TEMP_FILE_PREFIX_LENGTH)) { + return false; + } -/** - * Multi Range Read interface, DS-MRR calls - */ + if (char *is_part = strchr(tbl_buf, '#')) { + *is_part = '\0'; + tblnamelen = is_part - tbl_buf; + } -int ha_innobase::multi_range_read_init(RANGE_SEQ_IF *seq, void *seq_init_param, - uint n_ranges, uint mode, - HANDLER_BUFFER *buf) -{ - return ds_mrr.dsmrr_init(this, seq, seq_init_param, n_ranges, mode, buf); + filename_to_tablename(tbl_buf, tblname, MAX_TABLE_NAME_LEN + 1, true); + return true; } -int ha_innobase::multi_range_read_next(range_id_t *range_info) + +/** Acquire metadata lock and MariaDB table handle for an InnoDB table. +@param[in,out] thd thread handle +@param[in,out] table InnoDB table +@return MariaDB table handle +@retval NULL if the table does not exist, is unaccessible or corrupted. */ +static TABLE* innodb_acquire_mdl(THD* thd, dict_table_t* table) { - return ds_mrr.dsmrr_next(range_info); + char db_buf[NAME_LEN + 1], db_buf1[NAME_LEN + 1]; + char tbl_buf[NAME_LEN + 1], tbl_buf1[NAME_LEN + 1]; + ulint db_buf_len, db_buf1_len; + ulint tbl_buf_len, tbl_buf1_len; + + if (!table_name_parse(table->name, db_buf, tbl_buf, + db_buf_len, tbl_buf_len)) { + table->release(); + return NULL; + } + + const table_id_t table_id = table->id; +retry_mdl: + const bool unaccessible = !table->is_readable() || table->corrupted; + table->release(); + + if (unaccessible) { + return NULL; + } + + TABLE* mariadb_table = open_purge_table(thd, db_buf, db_buf_len, + tbl_buf, tbl_buf_len); + + table = dict_table_open_on_id(table_id, false, DICT_TABLE_OP_NORMAL); + + if (table == NULL) { + /* Table is dropped. */ + goto fail; + } + + if (!fil_table_accessible(table)) { +release_fail: + table->release(); +fail: + if (mariadb_table) { + close_thread_tables(thd); + } + + return NULL; + } + + if (!table_name_parse(table->name, db_buf1, tbl_buf1, + db_buf1_len, tbl_buf1_len)) { + goto release_fail; + } + + if (!mariadb_table) { + } else if (!strcmp(db_buf, db_buf1) && !strcmp(tbl_buf, tbl_buf1)) { + return mariadb_table; + } else { + /* Table is renamed. So release MDL for old name and try + to acquire the MDL for new table name. */ + close_thread_tables(thd); + } + + strcpy(tbl_buf, tbl_buf1); + strcpy(db_buf, db_buf1); + tbl_buf_len = tbl_buf1_len; + db_buf_len = db_buf1_len; + goto retry_mdl; } -ha_rows ha_innobase::multi_range_read_info_const(uint keyno, RANGE_SEQ_IF *seq, - void *seq_init_param, - uint n_ranges, uint *bufsz, - uint *flags, - Cost_estimate *cost) -{ - /* See comments in ha_myisam::multi_range_read_info_const */ - ds_mrr.init(this, table); +/** Find or open a table handle for the virtual column template +@param[in] thd thread handle +@param[in,out] table InnoDB table whose virtual column template + is to be updated +@return table handle +@retval NULL if the table is dropped, unaccessible or corrupted +for purge thread */ +static TABLE* innodb_find_table_for_vc(THD* thd, dict_table_t* table) +{ + if (THDVAR(thd, background_thread)) { + /* Purge thread acquires dict_operation_lock while + processing undo log record. Release the dict_operation_lock + before acquiring MDL on the table. */ + rw_lock_s_unlock(dict_operation_lock); + return innodb_acquire_mdl(thd, table); + } else { + if (table->vc_templ->mysql_table_query_id + == thd_get_query_id(thd)) { + return table->vc_templ->mysql_table; + } + } - if (prebuilt->select_lock_type != LOCK_NONE) - *flags |= HA_MRR_USE_DEFAULT_IMPL; + char db_buf[NAME_LEN + 1]; + char tbl_buf[NAME_LEN + 1]; + ulint db_buf_len, tbl_buf_len; + + if (!table_name_parse(table->name, db_buf, tbl_buf, + db_buf_len, tbl_buf_len)) { + ut_ad(!"invalid table name"); + return NULL; + } - ha_rows res= ds_mrr.dsmrr_info_const(keyno, seq, seq_init_param, n_ranges, - bufsz, flags, cost); - return res; + TABLE* mysql_table = find_fk_open_table(thd, db_buf, db_buf_len, + tbl_buf, tbl_buf_len); + + table->vc_templ->mysql_table = mysql_table; + table->vc_templ->mysql_table_query_id = thd_get_query_id(thd); + return mysql_table; } -ha_rows ha_innobase::multi_range_read_info(uint keyno, uint n_ranges, - uint keys, uint key_parts, - uint *bufsz, uint *flags, - Cost_estimate *cost) +/** Get the computed value by supplying the base column values. +@param[in,out] table table whose virtual column + template to be built */ +TABLE* innobase_init_vc_templ(dict_table_t* table) { - ds_mrr.init(this, table); - ha_rows res= ds_mrr.dsmrr_info(keyno, n_ranges, keys, key_parts, bufsz, - flags, cost); - return res; + if (table->vc_templ != NULL) { + return NULL; + } + + table->vc_templ = UT_NEW_NOKEY(dict_vcol_templ_t()); + + TABLE *mysql_table= innodb_find_table_for_vc(current_thd, table); + + ut_ad(mysql_table); + if (!mysql_table) { + return NULL; + } + + mutex_enter(&dict_sys->mutex); + innobase_build_v_templ(mysql_table, table, table->vc_templ, NULL, true); + mutex_exit(&dict_sys->mutex); + return mysql_table; } -int ha_innobase::multi_range_read_explain_info(uint mrr_mode, char *str, - size_t size) -{ - return ds_mrr.dsmrr_explain_info(mrr_mode, str, size); +/** Change dbname and table name in table->vc_templ. +@param[in,out] table the table whose virtual column template +dbname and tbname to be renamed. */ +void +innobase_rename_vc_templ( + dict_table_t* table) +{ + char dbname[MAX_DATABASE_NAME_LEN + 1]; + char tbname[MAX_DATABASE_NAME_LEN + 1]; + char* name = table->name.m_name; + ulint dbnamelen = dict_get_db_name_len(name); + ulint tbnamelen = strlen(name) - dbnamelen - 1; + char t_dbname[MAX_DATABASE_NAME_LEN + 1]; + char t_tbname[MAX_TABLE_NAME_LEN + 1]; + + strncpy(dbname, name, dbnamelen); + dbname[dbnamelen] = 0; + strncpy(tbname, name + dbnamelen + 1, tbnamelen); + tbname[tbnamelen] =0; + + /* For partition table, remove the partition name and use the + "main" table name to build the template */ + char* is_part = is_partition(tbname); + + if (is_part != NULL) { + *is_part = '\0'; + tbnamelen = is_part - tbname; + } + + dbnamelen = filename_to_tablename(dbname, t_dbname, + MAX_DATABASE_NAME_LEN + 1); + tbnamelen = filename_to_tablename(tbname, t_tbname, + MAX_TABLE_NAME_LEN + 1); + + table->vc_templ->db_name = t_dbname; + table->vc_templ->tb_name = t_tbname; +} + +/** Get the updated parent field value from the update vector for the +given col_no. +@param[in] foreign foreign key information +@param[in] update updated parent vector. +@param[in] col_no column position of the table +@return updated field from the parent update vector, else NULL */ +static +dfield_t* +innobase_get_field_from_update_vector( + dict_foreign_t* foreign, + upd_t* update, + ulint col_no) +{ + dict_table_t* parent_table = foreign->referenced_table; + dict_index_t* parent_index = foreign->referenced_index; + ulint parent_field_no; + ulint parent_col_no; + ulint prefix_col_no; + + for (ulint i = 0; i < foreign->n_fields; i++) { + + parent_col_no = dict_index_get_nth_col_no(parent_index, i); + parent_field_no = dict_table_get_nth_col_pos( + parent_table, parent_col_no, &prefix_col_no); + + for (ulint j = 0; j < update->n_fields; j++) { + upd_field_t* parent_ufield + = &update->fields[j]; + + if (parent_ufield->field_no == parent_field_no + && parent_col_no == col_no) { + return(&parent_ufield->new_val); + } + } + } + + return (NULL); } + /** - * Index Condition Pushdown interface implementation - */ + Allocate a heap and record for calculating virtual fields + Used mainly for virtual fields in indexes + +@param[in] thd MariaDB THD +@param[in] index Index in use +@param[out] heap Heap that holds temporary row +@param[in,out] table MariaDB table +@param[out] record Pointer to allocated MariaDB record +@param[out] storage Internal storage for blobs etc + +@retval false on success +@retval true on malloc failure or failed to open the maria table + for purge thread. +*/ -/*************************************************************//** -InnoDB index push-down condition check -@return ICP_NO_MATCH, ICP_MATCH, or ICP_OUT_OF_RANGE */ -UNIV_INTERN -enum icp_result -innobase_index_cond( -/*================*/ - void* file) /*!< in/out: pointer to ha_innobase */ -{ - return handler_index_cond_check(file); +bool innobase_allocate_row_for_vcol( + THD * thd, + dict_index_t* index, + mem_heap_t** heap, + TABLE** table, + byte** record, + VCOL_STORAGE** storage) +{ + TABLE *maria_table; + String *blob_value_storage; + if (!*table) + *table= innodb_find_table_for_vc(thd, index->table); + + /* For purge thread, there is a possiblity that table could have + dropped, corrupted or unaccessible. */ + if (!*table) + return true; + maria_table= *table; + if (!*heap && !(*heap= mem_heap_create(srv_page_size))) + { + *storage= 0; + return TRUE; + } + *record= static_cast<byte*>(mem_heap_alloc(*heap, + maria_table->s->reclength)); + *storage= static_cast<VCOL_STORAGE*> + (mem_heap_alloc(*heap, sizeof(**storage))); + blob_value_storage= static_cast<String*> + (mem_heap_alloc(*heap, + maria_table->s->virtual_not_stored_blob_fields * + sizeof(String))); + if (!*record || !*storage || !blob_value_storage) + { + *storage= 0; + return TRUE; + } + (*storage)->maria_table= maria_table; + (*storage)->innobase_record= *record; + (*storage)->maria_record= maria_table->field[0]->record_ptr(); + (*storage)->blob_value_storage= blob_value_storage; + + maria_table->move_fields(maria_table->field, *record, + (*storage)->maria_record); + maria_table->remember_blob_values(blob_value_storage); + + return FALSE; +} + + +/** Free memory allocated by innobase_allocate_row_for_vcol() */ + +void innobase_free_row_for_vcol(VCOL_STORAGE *storage) +{ + TABLE *maria_table= storage->maria_table; + maria_table->move_fields(maria_table->field, storage->maria_record, + storage->innobase_record); + maria_table->restore_blob_values(storage->blob_value_storage); +} + + +/** Get the computed value by supplying the base column values. +@param[in,out] row the data row +@param[in] col virtual column +@param[in] index index +@param[in,out] local_heap heap memory for processing large data etc. +@param[in,out] heap memory heap that copies the actual index row +@param[in] ifield index field +@param[in] thd MySQL thread handle +@param[in,out] mysql_table mysql table object +@param[in] old_table during ALTER TABLE, this is the old table + or NULL. +@param[in] parent_update update vector for the parent row +@param[in] foreign foreign key information +@return the field filled with computed value, or NULL if just want +to store the value in passed in "my_rec" */ +dfield_t* +innobase_get_computed_value( + const dtuple_t* row, + const dict_v_col_t* col, + const dict_index_t* index, + mem_heap_t** local_heap, + mem_heap_t* heap, + const dict_field_t* ifield, + THD* thd, + TABLE* mysql_table, + byte* mysql_rec, + const dict_table_t* old_table, + upd_t* parent_update, + dict_foreign_t* foreign) +{ + byte rec_buf2[REC_VERSION_56_MAX_INDEX_COL_LEN]; + byte* buf; + dfield_t* field; + ulint len; + + const page_size_t page_size = (old_table == NULL) + ? dict_table_page_size(index->table) + : dict_table_page_size(old_table); + + ulint ret = 0; + + ut_ad(index->table->vc_templ); + ut_ad(thd != NULL); + ut_ad(mysql_table); + + const mysql_row_templ_t* + vctempl = index->table->vc_templ->vtempl[ + index->table->vc_templ->n_col + col->v_pos]; + + if (!heap || index->table->vc_templ->rec_len + >= REC_VERSION_56_MAX_INDEX_COL_LEN) { + if (*local_heap == NULL) { + *local_heap = mem_heap_create(UNIV_PAGE_SIZE); + } + + buf = static_cast<byte*>(mem_heap_alloc( + *local_heap, index->table->vc_templ->rec_len)); + } else { + buf = rec_buf2; + } + + for (ulint i = 0; i < col->num_base; i++) { + dict_col_t* base_col = col->base_col[i]; + const dfield_t* row_field = NULL; + ulint col_no = base_col->ind; + const mysql_row_templ_t* templ + = index->table->vc_templ->vtempl[col_no]; + const byte* data; + + if (parent_update != NULL) { + /** Get the updated field from update vector + of the parent table. */ + row_field = innobase_get_field_from_update_vector( + foreign, parent_update, col_no); + } + + if (row_field == NULL) { + row_field = dtuple_get_nth_field(row, col_no); + } + + data = static_cast<const byte*>(row_field->data); + len = row_field->len; + + if (row_field->ext) { + if (*local_heap == NULL) { + *local_heap = mem_heap_create(UNIV_PAGE_SIZE); + } + + data = btr_copy_externally_stored_field( + &len, data, page_size, + dfield_get_len(row_field), *local_heap); + } + + if (len == UNIV_SQL_NULL) { + mysql_rec[templ->mysql_null_byte_offset] + |= (byte) templ->mysql_null_bit_mask; + memcpy(mysql_rec + templ->mysql_col_offset, + static_cast<const byte*>( + index->table->vc_templ->default_rec + + templ->mysql_col_offset), + templ->mysql_col_len); + } else { + + row_sel_field_store_in_mysql_format( + mysql_rec + templ->mysql_col_offset, + templ, index, templ->clust_rec_field_no, + (const byte*)data, len); + + if (templ->mysql_null_bit_mask) { + /* It is a nullable column with a + non-NULL value */ + mysql_rec[templ->mysql_null_byte_offset] + &= ~(byte) templ->mysql_null_bit_mask; + } + } + } + + field = dtuple_get_nth_v_field(row, col->v_pos); + + my_bitmap_map* old_write_set = dbug_tmp_use_all_columns(mysql_table, mysql_table->write_set); + my_bitmap_map* old_read_set = dbug_tmp_use_all_columns(mysql_table, mysql_table->read_set); + ret = mysql_table->update_virtual_field(mysql_table->field[col->m_col.ind]); + dbug_tmp_restore_column_map(mysql_table->read_set, old_read_set); + dbug_tmp_restore_column_map(mysql_table->write_set, old_write_set); + + if (ret != 0) { +#ifdef INNODB_VIRTUAL_DEBUG + ib::warn() << "Compute virtual column values failed "; + fputs("InnoDB: Cannot compute value for following record ", + stderr); + dtuple_print(stderr, row); +#endif /* INNODB_VIRTUAL_DEBUG */ + return(NULL); + } + + if (vctempl->mysql_null_bit_mask + && (mysql_rec[vctempl->mysql_null_byte_offset] + & vctempl->mysql_null_bit_mask)) { + dfield_set_null(field); + field->type.prtype |= DATA_VIRTUAL; + return(field); + } + + row_mysql_store_col_in_innobase_format( + field, buf, + TRUE, mysql_rec + vctempl->mysql_col_offset, + vctempl->mysql_col_len, dict_table_is_comp(index->table)); + field->type.prtype |= DATA_VIRTUAL; + + ulint max_prefix = col->m_col.max_prefix; + + if (max_prefix && ifield + && (ifield->prefix_len == 0 + || ifield->prefix_len > col->m_col.max_prefix)) { + max_prefix = ifield->prefix_len; + } + + /* If this is a prefix index, we only need a portion of the field */ + if (max_prefix) { + len = dtype_get_at_most_n_mbchars( + col->m_col.prtype, + col->m_col.mbminlen, col->m_col.mbmaxlen, + max_prefix, + field->len, + static_cast<char*>(dfield_get_data(field))); + dfield_set_len(field, len); + } + + if (heap) { + dfield_dup(field, heap); + } + + return(field); } + /** Attempt to push down an index condition. -* @param[in] keyno MySQL key number -* @param[in] idx_cond Index condition to be checked -* @return Part of idx_cond which the handler will not evaluate -*/ -UNIV_INTERN +@param[in] keyno MySQL key number +@param[in] idx_cond Index condition to be checked +@return Part of idx_cond which the handler will not evaluate */ + class Item* ha_innobase::idx_cond_push( uint keyno, @@ -20531,6 +21830,12 @@ ha_innobase::idx_cond_push( DBUG_ASSERT(keyno != MAX_KEY); DBUG_ASSERT(idx_cond != NULL); + /* We can only evaluate the condition if all columns are stored.*/ + dict_index_t* idx = innobase_get_index(keyno); + if (idx && dict_index_has_virtual(idx)) { + DBUG_RETURN(idx_cond); + } + pushed_idx_cond = idx_cond; pushed_idx_cond_keyno = keyno; in_range_check_pushed_down = TRUE; @@ -20545,10 +21850,9 @@ errmsg-utf8.txt directly as is. Push a warning message to the client, it is a wrapper around: void push_warning_printf( - THD *thd, Sql_condition::enum_warning_level level, + THD *thd, Sql_condition::enum_condition_level level, uint code, const char *format, ...); */ -UNIV_INTERN void ib_senderrf( /*========*/ @@ -20557,7 +21861,8 @@ ib_senderrf( ib_uint32_t code, /*!< MySQL error code */ ...) /*!< Args */ { - va_list args; + va_list args; + char* str = NULL; const char* format = innobase_get_err_msg(code); /* If the caller wants to push a message to the client then @@ -20570,27 +21875,32 @@ ib_senderrf( va_start(args, code); - myf l=0; + myf l = Sql_condition::WARN_LEVEL_NOTE; - switch(level) { + switch (level) { case IB_LOG_LEVEL_INFO: - l = ME_JUST_INFO; + l = ME_JUST_INFO; break; case IB_LOG_LEVEL_WARN: - l = ME_JUST_WARNING; + l = ME_JUST_WARNING; break; case IB_LOG_LEVEL_ERROR: + sd_notifyf(0, "STATUS=InnoDB: Error: %s", str); + l = 0; + break; case IB_LOG_LEVEL_FATAL: l = 0; + sd_notifyf(0, "STATUS=InnoDB: Fatal: %s", str); break; default: l = 0; break; } - my_printv_error(code, format, MYF(l), args); + my_printv_error(code, format, MYF(l), args); va_end(args); + free(str); if (level == IB_LOG_LEVEL_FATAL) { ut_error; @@ -20605,10 +21915,9 @@ must be: "Some string ... %s". Push a warning message to the client, it is a wrapper around: void push_warning_printf( - THD *thd, Sql_condition::enum_warning_level level, + THD *thd, Sql_condition::enum_condition_level level, uint code, const char *format, ...); */ -UNIV_INTERN void ib_errf( /*====*/ @@ -20618,7 +21927,7 @@ ib_errf( const char* format, /*!< printf format */ ...) /*!< Args */ { - char* str; + char* str = NULL; va_list args; /* If the caller wants to push a message to the client then @@ -20629,22 +21938,32 @@ ib_errf( va_start(args, format); -#ifdef __WIN__ +#ifdef _WIN32 int size = _vscprintf(format, args) + 1; - str = static_cast<char*>(malloc(size)); + if (size > 0) { + str = static_cast<char*>(malloc(size)); + } + if (str == NULL) { + va_end(args); + return; /* Watch for Out-Of-Memory */ + } str[size - 1] = 0x0; vsnprintf(str, size, format, args); #elif HAVE_VASPRINTF if (vasprintf(&str, format, args) == -1) { /* In case of failure use a fixed length string */ str = static_cast<char*>(malloc(BUFSIZ)); - my_vsnprintf(str, BUFSIZ, format, args); + vsnprintf(str, BUFSIZ, format, args); } #else /* Use a fixed length string. */ str = static_cast<char*>(malloc(BUFSIZ)); - my_vsnprintf(str, BUFSIZ, format, args); -#endif /* __WIN__ */ + if (str == NULL) { + va_end(args); + return; /* Watch for Out-Of-Memory */ + } + vsnprintf(str, BUFSIZ, format, args); +#endif /* _WIN32 */ ib_senderrf(thd, level, code, str); @@ -20652,62 +21971,40 @@ ib_errf( free(str); } -/******************************************************************//** -Write a message to the MySQL log, prefixed with "InnoDB: " */ -UNIV_INTERN -void -ib_logf( -/*====*/ - ib_log_level_t level, /*!< in: warning level */ - const char* format, /*!< printf format */ - ...) /*!< Args */ -{ - char* str; - va_list args; +/* Keep the first 16 characters as-is, since the url is sometimes used +as an offset from this.*/ +const char* TROUBLESHOOTING_MSG = + "Please refer to " REFMAN "innodb-troubleshooting.html" + " for how to resolve the issue."; - va_start(args, format); +const char* TROUBLESHOOT_DATADICT_MSG = + "Please refer to " REFMAN "innodb-troubleshooting-datadict.html" + " for how to resolve the issue."; -#ifdef __WIN__ - int size = _vscprintf(format, args) + 1; - str = static_cast<char*>(malloc(size)); - str[size - 1] = 0x0; - vsnprintf(str, size, format, args); -#elif HAVE_VASPRINTF - if (vasprintf(&str, format, args) == -1) { - /* In case of failure use a fixed length string */ - str = static_cast<char*>(malloc(BUFSIZ)); - my_vsnprintf(str, BUFSIZ, format, args); - } -#else - /* Use a fixed length string. */ - str = static_cast<char*>(malloc(BUFSIZ)); - my_vsnprintf(str, BUFSIZ, format, args); -#endif /* __WIN__ */ +const char* BUG_REPORT_MSG = + "Submit a detailed bug report to https://jira.mariadb.org/"; - switch(level) { - case IB_LOG_LEVEL_INFO: - sql_print_information("InnoDB: %s", str); - break; - case IB_LOG_LEVEL_WARN: - sql_print_warning("InnoDB: %s", str); - break; - case IB_LOG_LEVEL_ERROR: - sql_print_error("InnoDB: %s", str); - sd_notifyf(0, "STATUS=InnoDB: Error: %s", str); - break; - case IB_LOG_LEVEL_FATAL: - sql_print_error("InnoDB: %s", str); - sd_notifyf(0, "STATUS=InnoDB: Fatal: %s", str); - break; - } +const char* FORCE_RECOVERY_MSG = + "Please refer to " + "https://mariadb.com/kb/en/library/xtradbinnodb-recovery-modes/" + " for information about forcing recovery."; - va_end(args); - free(str); +const char* ERROR_CREATING_MSG = + "Please refer to " REFMAN "error-creating-innodb.html"; - if (level == IB_LOG_LEVEL_FATAL) { - ut_error; - } -} +const char* OPERATING_SYSTEM_ERROR_MSG = + "Some operating system error numbers are described at" + " https://mariadb.com/kb/en/library/operating-system-error-codes/"; + +const char* FOREIGN_KEY_CONSTRAINTS_MSG = + "Please refer to https://mariadb.com/kb/en/library/foreign-keys/" + " for correct foreign key definition."; + +const char* SET_TRANSACTION_MSG = + "Please refer to https://mariadb.com/kb/en/library/set-transaction/"; + +const char* INNODB_PARAMETERS_MSG = + "Please refer to https://mariadb.com/kb/en/library/xtradbinnodb-server-system-variables/"; /********************************************************************** Converts an identifier from my_charset_filename to UTF-8 charset. @@ -20723,8 +22020,9 @@ innobase_convert_to_filename_charset( CHARSET_INFO* cs_to = &my_charset_filename; CHARSET_INFO* cs_from = system_charset_info; - return(strconvert( cs_from, from, strlen(from), cs_to, to, - static_cast<uint>(len), &errors)); + return(static_cast<uint>(strconvert( + cs_from, from, strlen(from), + cs_to, to, static_cast<uint>(len), &errors))); } /********************************************************************** @@ -20741,13 +22039,13 @@ innobase_convert_to_system_charset( CHARSET_INFO* cs1 = &my_charset_filename; CHARSET_INFO* cs2 = system_charset_info; - return(strconvert(cs1, from, strlen(from), cs2, to, - static_cast<uint>(len), errors)); + return(static_cast<uint>(strconvert( + cs1, from, strlen(from), + cs2, to, static_cast<uint>(len), errors))); } /********************************************************************** Issue a warning that the row is too big. */ -UNIV_INTERN void ib_warn_row_too_big(const dict_table_t* table) { @@ -20761,13 +22059,10 @@ ib_warn_row_too_big(const dict_table_t* table) THD* thd = current_thd; - if (thd == NULL) { - return; - } - push_warning_printf( thd, Sql_condition::WARN_LEVEL_WARN, HA_ERR_TO_BIG_ROW, - "Row size too large (> %lu). Changing some columns to TEXT" + "Row size too large (> " ULINTPF ")." + " Changing some columns to TEXT" " or BLOB %smay help. In current row format, BLOB prefix of" " %d bytes is stored inline.", free_space , prefix ? "or using ROW_FORMAT=DYNAMIC or" @@ -20775,6 +22070,109 @@ ib_warn_row_too_big(const dict_table_t* table) , prefix ? DICT_MAX_FIXED_COL_LEN : 0); } +/** Validate the requested buffer pool size. Also, reserve the necessary +memory needed for buffer pool resize. +@param[in] thd thread handle +@param[in] var pointer to system variable +@param[out] save immediate result for update function +@param[in] value incoming string +@return 0 on success, 1 on failure. +*/ +static +int +innodb_buffer_pool_size_validate( + THD* thd, + struct st_mysql_sys_var* var, + void* save, + struct st_mysql_value* value) +{ + longlong intbuf; + + + value->val_int(value, &intbuf); + + if (!srv_was_started) { + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "Cannot update innodb_buffer_pool_size," + " because InnoDB is not started."); + return(1); + } + +#ifdef UNIV_DEBUG + if (buf_disable_resize_buffer_pool_debug == TRUE) { + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "Cannot update innodb_buffer_pool_size," + " because innodb_disable_resize_buffer_pool_debug" + " is set."); + ib::warn() << "Cannot update innodb_buffer_pool_size," + " because innodb_disable_resize_buffer_pool_debug" + " is set."; + return(1); + } +#endif /* UNIV_DEBUG */ + + + buf_pool_mutex_enter_all(); + + if (srv_buf_pool_old_size != srv_buf_pool_size) { + buf_pool_mutex_exit_all(); + my_printf_error(ER_WRONG_ARGUMENTS, + "Another buffer pool resize is already in progress.", MYF(0)); + return(1); + } + + if (srv_buf_pool_instances > 1 && intbuf < BUF_POOL_SIZE_THRESHOLD) { + buf_pool_mutex_exit_all(); + + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "Cannot update innodb_buffer_pool_size" + " to less than 1GB if" + " innodb_buffer_pool_instances > 1."); + return(1); + } + + ulint requested_buf_pool_size + = buf_pool_size_align(static_cast<ulint>(intbuf)); + + *static_cast<longlong*>(save) = requested_buf_pool_size; + + if (srv_buf_pool_size == static_cast<ulint>(intbuf)) { + buf_pool_mutex_exit_all(); + /* nothing to do */ + return(0); + } + + if (srv_buf_pool_size == requested_buf_pool_size) { + buf_pool_mutex_exit_all(); + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_WRONG_ARGUMENTS, + "innodb_buffer_pool_size must be at least" + " innodb_buffer_pool_chunk_size=%lu", + srv_buf_pool_chunk_unit); + /* nothing to do */ + return(0); + } + + srv_buf_pool_size = requested_buf_pool_size; + buf_pool_mutex_exit_all(); + + if (intbuf != static_cast<longlong>(requested_buf_pool_size)) { + char buf[64]; + int len = 64; + value->val_str(value, buf, &len); + push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, + ER_TRUNCATED_WRONG_VALUE, + "Truncated incorrect %-.32s value: '%-.128s'", + mysql_sysvar_buffer_pool_size.name, + value->val_str(value, buf, &len)); + } + + return(0); +} + /*************************************************************//** Check for a valid value of innobase_compression_algorithm. @return 0 for valid innodb_compression_algorithm. */ @@ -20900,6 +22298,9 @@ static void innodb_remember_check_sysvar_funcs() /* remember build-in sysvar check functions */ ut_ad((MYSQL_SYSVAR_NAME(checksum_algorithm).flags & 0x1FF) == PLUGIN_VAR_ENUM); check_sysvar_enum = MYSQL_SYSVAR_NAME(checksum_algorithm).check; + + ut_ad((MYSQL_SYSVAR_NAME(flush_log_at_timeout).flags & 15) == PLUGIN_VAR_INT); + check_sysvar_int = MYSQL_SYSVAR_NAME(flush_log_at_timeout).check; } /********************************************************************//** @@ -20979,13 +22380,13 @@ ib_push_frm_error( case DICT_FRM_NO_PK: sql_print_error("Table %s has a primary key in " "InnoDB data dictionary, but not " - "in MySQL!" + "in MariaDB!" " Have you mixed up " ".frm files from different " "installations? See " REFMAN "innodb-troubleshooting.html\n", - ib_table->name); + ib_table->name.m_name); if (push_warning) { push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, @@ -20993,20 +22394,20 @@ ib_push_frm_error( "InnoDB: Table %s has a " "primary key in InnoDB data " "dictionary, but not in " - "MySQL!", ib_table->name); + "MariaDB!", ib_table->name.m_name); } break; case DICT_NO_PK_FRM_HAS: sql_print_error( "Table %s has no primary key in InnoDB data " - "dictionary, but has one in MySQL! If you " - "created the table with a MySQL version < " + "dictionary, but has one in MariaDB! If you " + "created the table with a MariaDB version < " "3.23.54 and did not define a primary key, " "but defined a unique key with all non-NULL " - "columns, then MySQL internally treats that " + "columns, then MariaDB internally treats that " "key as the primary key. You can fix this " "error by dump + DROP + CREATE + reimport " - "of the table.", ib_table->name); + "of the table.", ib_table->name.m_name); if (push_warning) { push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, @@ -21014,32 +22415,32 @@ ib_push_frm_error( "InnoDB: Table %s has no " "primary key in InnoDB data " "dictionary, but has one in " - "MySQL!", - ib_table->name); + "MariaDB!", + ib_table->name.m_name); } break; case DICT_FRM_INCONSISTENT_KEYS: - sql_print_error("InnoDB: Table %s contains %lu " + sql_print_error("InnoDB: Table %s contains " ULINTPF " " "indexes inside InnoDB, which " "is different from the number of " - "indexes %u defined in the MySQL " + "indexes %u defined in the MariaDB " " Have you mixed up " ".frm files from different " "installations? See " REFMAN "innodb-troubleshooting.html\n", - ib_table->name, n_keys, + ib_table->name.m_name, n_keys, table->s->keys); if (push_warning) { push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_NO_SUCH_INDEX, - "InnoDB: Table %s contains %lu " + "InnoDB: Table %s contains " ULINTPF " " "indexes inside InnoDB, which " "is different from the number of " - "indexes %u defined in the MySQL ", - ib_table->name, n_keys, + "indexes %u defined in the MariaDB ", + ib_table->name.m_name, n_keys, table->s->keys); } break; @@ -21047,9 +22448,9 @@ ib_push_frm_error( case DICT_FRM_CONSISTENT: default: sql_print_error("InnoDB: Table %s is consistent " - "on InnoDB data dictionary and MySQL " + "on InnoDB data dictionary and MariaDB " " FRM file.", - ib_table->name); + ib_table->name.m_name); ut_error; break; } |