diff options
Diffstat (limited to 'storage/xtradb/include')
| -rw-r--r-- | storage/xtradb/include/btr0btr.ic | 2 | ||||
| -rw-r--r-- | storage/xtradb/include/buf0buf.h | 51 | ||||
| -rw-r--r-- | storage/xtradb/include/buf0rea.h | 38 | ||||
| -rw-r--r-- | storage/xtradb/include/db0err.h | 2 | ||||
| -rw-r--r-- | storage/xtradb/include/dict0dict.h | 27 | ||||
| -rw-r--r-- | storage/xtradb/include/dict0dict.ic | 17 | ||||
| -rw-r--r-- | storage/xtradb/include/dict0mem.h | 40 | ||||
| -rw-r--r-- | storage/xtradb/include/fil0crypt.h | 13 | ||||
| -rw-r--r-- | storage/xtradb/include/fil0fil.h | 171 | ||||
| -rw-r--r-- | storage/xtradb/include/os0sync.h | 57 | ||||
| -rw-r--r-- | storage/xtradb/include/srv0srv.h | 28 | ||||
| -rw-r--r-- | storage/xtradb/include/trx0sys.h | 5 | ||||
| -rw-r--r-- | storage/xtradb/include/univ.i | 2 | ||||
| -rw-r--r-- | storage/xtradb/include/ut0counter.h | 90 |
14 files changed, 303 insertions, 240 deletions
diff --git a/storage/xtradb/include/btr0btr.ic b/storage/xtradb/include/btr0btr.ic index 62a24873482..0f5f025d6a3 100644 --- a/storage/xtradb/include/btr0btr.ic +++ b/storage/xtradb/include/btr0btr.ic @@ -61,7 +61,7 @@ btr_block_get_func( if (err == DB_DECRYPTION_FAILED) { if (index && index->table) { - index->table->is_encrypted = true; + index->table->file_unreadable = true; } } diff --git a/storage/xtradb/include/buf0buf.h b/storage/xtradb/include/buf0buf.h index 4a632e2345f..9b4276efaa8 100644 --- a/storage/xtradb/include/buf0buf.h +++ b/storage/xtradb/include/buf0buf.h @@ -1,7 +1,7 @@ /***************************************************************************** Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. -Copyright (c) 2013, 2017, MariaDB Corporation. All Rights Reserved. +Copyright (c) 2013, 2017, MariaDB Corporation. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -675,13 +675,13 @@ buf_page_is_checksum_valid_none( ulint checksum_field2) MY_ATTRIBUTE((warn_unused_result)); -/********************************************************************//** -Checks if a page is corrupt. +/** Check if a page is corrupt. @param[in] check_lsn true if LSN should be checked @param[in] read_buf Page to be checked @param[in] zip_size compressed size or 0 @param[in] space Pointer to tablespace @return true if corrupted, false if not */ +UNIV_INTERN bool buf_page_is_corrupted( bool check_lsn, @@ -689,26 +689,13 @@ buf_page_is_corrupted( ulint zip_size, const fil_space_t* space) MY_ATTRIBUTE((warn_unused_result)); -/********************************************************************//** -Check if page is maybe compressed, encrypted or both when we encounter -corrupted page. Note that we can't be 100% sure if page is corrupted -or decrypt/decompress just failed. -@param[in] bpage Page -@return true if page corrupted, false if not */ -bool -buf_page_check_corrupt( - buf_page_t* bpage) /*!< in/out: buffer page read from disk */ - MY_ATTRIBUTE(( warn_unused_result)); - -/********************************************************************//** -Checks if a page is all zeroes. -@return TRUE if the page is all zeroes */ +/** Check if a page is all zeroes. +@param[in] read_buf database page +@param[in] zip_size ROW_FORMAT=COMPRESSED page size, or 0 +@return whether the page is all zeroes */ +UNIV_INTERN bool -buf_page_is_zeroes( -/*===============*/ - const byte* read_buf, /*!< in: a database page */ - const ulint zip_size); /*!< in: size of compressed page; - 0 for uncompressed pages */ +buf_page_is_zeroes(const byte* read_buf, ulint zip_size); #ifndef UNIV_HOTBACKUP /**********************************************************************//** Gets the space id, page offset, and byte offset within page of a @@ -1270,15 +1257,18 @@ buf_page_init_for_read( version of the tablespace in case we have done DISCARD + IMPORT */ ulint offset);/*!< in: page number */ -/********************************************************************//** -Completes an asynchronous read or write request of a file page to or from -the buffer pool. -@return true if successful */ +/** Complete a read or write request of a file page to or from the buffer pool. +@param[in,out] bpage Page to complete +@return whether the operation succeeded +@retval DB_SUCCESS always when writing, or if a read page was OK +@retval DB_PAGE_CORRUPTED if the checksum fails on a page read +@retval DB_DECRYPTION_FAILED if page post encryption checksum matches but + after decryption normal page checksum does + not match */ UNIV_INTERN -bool -buf_page_io_complete( -/*=================*/ - buf_page_t* bpage); /*!< in: pointer to the block in question */ +dberr_t +buf_page_io_complete(buf_page_t* bpage) + MY_ATTRIBUTE((nonnull)); /********************************************************************//** Calculates a folded value of a file page address to use in the page hash table. @@ -1680,7 +1670,6 @@ struct buf_page_t{ if written again we check is TRIM operation needed. */ - unsigned key_version; /*!< key version for this block */ bool encrypted; /*!< page is still encrypted */ ulint real_size; /*!< Real size of the page diff --git a/storage/xtradb/include/buf0rea.h b/storage/xtradb/include/buf0rea.h index f0652b5d2cd..ab73108a71e 100644 --- a/storage/xtradb/include/buf0rea.h +++ b/storage/xtradb/include/buf0rea.h @@ -1,7 +1,7 @@ /***************************************************************************** Copyright (c) 1995, 2013, Oracle and/or its affiliates. All Rights Reserved. -Copyright (c) 2015, MariaDB Corporation. +Copyright (c) 2015, 2017, MariaDB Corporation. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -35,29 +35,37 @@ High-level function which reads a page asynchronously from a file to the buffer buf_pool if it is not already there. Sets the io_fix flag and sets an exclusive lock on the buffer frame. The flag is cleared and the x-lock released by the i/o-handler thread. -@return TRUE if page has been read in, FALSE in case of failure */ + +@param[in] space space_id +@param[in] zip_size compressed page size in bytes, or 0 +@param[in] offset page number +@param[in] trx transaction +@return DB_SUCCESS if page has been read and is not corrupted, +@retval DB_PAGE_CORRUPTED if page based on checksum check is corrupted, +@retval DB_DECRYPTION_FAILED if page post encryption checksum matches but +after decryption normal page checksum does not match. +@retval DB_TABLESPACE_DELETED if tablespace .ibd file is missing */ UNIV_INTERN -ibool +dberr_t buf_read_page( -/*==========*/ - ulint space, /*!< in: space id */ - ulint zip_size, /*!< in: compressed page size in bytes, or 0 */ - ulint offset, /*!< in: page number */ - trx_t* trx, /*!< in: trx */ - buf_page_t** bpage /*!< out: page */ -); + ulint space, + ulint zip_size, + ulint offset, + trx_t* trx); + /********************************************************************//** High-level function which reads a page asynchronously from a file to the buffer buf_pool if it is not already there. Sets the io_fix flag and sets an exclusive lock on the buffer frame. The flag is cleared and the x-lock released by the i/o-handler thread. -@return TRUE if page has been read in, FALSE in case of failure */ +@param[in] space Tablespace id +@param[in] offset Page number */ UNIV_INTERN -ibool +void buf_read_page_async( -/*================*/ - ulint space, /*!< in: space id */ - ulint offset);/*!< in: page number */ + ulint space, + ulint offset); + /********************************************************************//** Applies a random read-ahead in buf_pool if there are at least a threshold value of accessed pages from the random read-ahead area. Does not read any diff --git a/storage/xtradb/include/db0err.h b/storage/xtradb/include/db0err.h index b11c5f4ea1a..8bd3beda110 100644 --- a/storage/xtradb/include/db0err.h +++ b/storage/xtradb/include/db0err.h @@ -138,6 +138,8 @@ enum dberr_t { of missing key management plugin, or missing or incorrect key or incorret AES method or algorithm. */ + DB_PAGE_CORRUPTED, /* Page read from tablespace is + corrupted. */ /* The following are partial failure codes */ DB_FAIL = 1000, DB_OVERFLOW, diff --git a/storage/xtradb/include/dict0dict.h b/storage/xtradb/include/dict0dict.h index 1622b927a76..6da8eb892d9 100644 --- a/storage/xtradb/include/dict0dict.h +++ b/storage/xtradb/include/dict0dict.h @@ -1767,16 +1767,6 @@ dict_close(void); /*============*/ #ifndef UNIV_HOTBACKUP /**********************************************************************//** -Check whether the table is corrupted. -@return nonzero for corrupted table, zero for valid tables */ -UNIV_INLINE -ulint -dict_table_is_corrupted( -/*====================*/ - const dict_table_t* table) /*!< in: table */ - MY_ATTRIBUTE((nonnull, warn_unused_result)); - -/**********************************************************************//** Check whether the index is corrupted. @return nonzero for corrupted index, zero for valid indexes */ UNIV_INLINE @@ -1820,6 +1810,15 @@ dict_set_corrupted_by_space( /*========================*/ ulint space_id); /*!< in: space ID */ +/**********************************************************************//** +Flags a table with specified space_id encrypted in the data dictionary +cache +@param[in] space_id Tablespace id */ +UNIV_INTERN +void +dict_set_encrypted_by_space( + ulint space_id); + /********************************************************************//** Validate the table flags. @return true if valid. */ @@ -1900,14 +1899,6 @@ dict_table_get_index_on_first_col( in table */ #endif /* !UNIV_HOTBACKUP */ -/************************************************************************* -set is_corrupt flag by space_id*/ - -void -dict_table_set_corrupt_by_space( -/*============================*/ - ulint space_id, - ibool need_mutex); #ifndef UNIV_NONINL #include "dict0dict.ic" diff --git a/storage/xtradb/include/dict0dict.ic b/storage/xtradb/include/dict0dict.ic index e3de7a33123..4ed1afc8094 100644 --- a/storage/xtradb/include/dict0dict.ic +++ b/storage/xtradb/include/dict0dict.ic @@ -1538,21 +1538,6 @@ dict_max_field_len_store_undo( } /********************************************************************//** -Check whether the table is corrupted. -@return nonzero for corrupted table, zero for valid tables */ -UNIV_INLINE -ulint -dict_table_is_corrupted( -/*====================*/ - const dict_table_t* table) /*!< in: table */ -{ - ut_ad(table); - ut_ad(table->magic_n == DICT_TABLE_MAGIC_N); - - return(table->corrupted); -} - -/********************************************************************//** Check whether the index is corrupted. @return nonzero for corrupted index, zero for valid indexes */ UNIV_INLINE @@ -1564,7 +1549,7 @@ dict_index_is_corrupted( ut_ad(index->magic_n == DICT_INDEX_MAGIC_N); return((index->type & DICT_CORRUPT) - || (index->table && index->table->corrupted)); + || (index->table && index->table->corrupted)); } /********************************************************************//** diff --git a/storage/xtradb/include/dict0mem.h b/storage/xtradb/include/dict0mem.h index 96c85cd8a99..a32581a0e90 100644 --- a/storage/xtradb/include/dict0mem.h +++ b/storage/xtradb/include/dict0mem.h @@ -2,7 +2,7 @@ Copyright (c) 1996, 2016, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2012, Facebook Inc. -Copyright (c) 2013, 2016, MariaDB Corporation. +Copyright (c) 2013, 2017, MariaDB Corporation. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -795,6 +795,9 @@ struct dict_index_t{ to first_blob_page_no; protected by blobs_mutex; @see btr_blob_dbg_t */ #endif /* UNIV_BLOB_DEBUG */ + + bool is_readable() const; + #ifdef UNIV_DEBUG ulint magic_n;/*!< magic number */ /** Value of dict_index_t::magic_n */ @@ -1062,11 +1065,13 @@ struct dict_table_t{ table is placed */ unsigned flags:DICT_TF_BITS; /*!< DICT_TF_... */ unsigned flags2:DICT_TF2_BITS; /*!< DICT_TF2_... */ - unsigned ibd_file_missing:1; - /*!< TRUE if this is in a single-table - tablespace and the .ibd file is missing; then - we must return in ha_innodb.cc an error if the - user tries to query such an orphaned table */ + unsigned file_unreadable:1; + /*!< true if this is in a single-table + tablespace and the .ibd file is missing + or page decryption failed and page is + corrupted; then we must return in + ha_innodb.cc an error if the + user tries to query such table */ unsigned cached:1;/*!< TRUE if the table object has been added to the dictionary cache */ unsigned to_be_dropped:1; @@ -1364,10 +1369,19 @@ struct dict_table_t{ UT_LIST_BASE_NODE_T(lock_t) locks; /*!< list of locks on the table; protected by lock_sys->mutex */ - ibool is_corrupt; - ibool is_encrypted; + #endif /* !UNIV_HOTBACKUP */ + /* Returns true if this is a single-table tablespace + and the .ibd file is missing or page decryption failed + and/or page is corrupted. + @return true if table is readable + @retval false if table is not readable */ + inline bool is_readable() const + { + return(UNIV_LIKELY(!file_unreadable)); + } + #ifdef UNIV_DEBUG ulint magic_n;/*!< magic number */ /** Value of dict_table_t::magic_n */ @@ -1375,6 +1389,16 @@ struct dict_table_t{ #endif /* UNIV_DEBUG */ }; +/* Returns true if this is a single-table tablespace +and the .ibd file is missing or page decryption failed +and/or page is corrupted. +@return true if table is readable +@retval false if table is not readable */ +inline bool dict_index_t::is_readable() const +{ + return(UNIV_LIKELY(!table->file_unreadable)); +} + /** A function object to add the foreign key constraint to the referenced set of the referenced table, if it exists in the dictionary cache. */ struct dict_foreign_add_to_referenced_table { diff --git a/storage/xtradb/include/fil0crypt.h b/storage/xtradb/include/fil0crypt.h index e7e9676aa3a..228dfb895fe 100644 --- a/storage/xtradb/include/fil0crypt.h +++ b/storage/xtradb/include/fil0crypt.h @@ -117,8 +117,7 @@ struct fil_space_crypt_t : st_encryption_scheme min_key_version(new_min_key_version), page0_offset(0), encryption(new_encryption), - mutex(), - key_found(new_min_key_version), + key_found(0), rotate_state() { key_id = new_key_id; @@ -136,6 +135,8 @@ struct fil_space_crypt_t : st_encryption_scheme type = CRYPT_SCHEME_1; min_key_version = key_get_latest_version(); } + + key_found = min_key_version; } /** Destructor */ @@ -298,13 +299,15 @@ Parse a MLOG_FILE_WRITE_CRYPT_DATA log entry @param[in] ptr Log entry start @param[in] end_ptr Log entry end @param[in] block buffer block +@param[out] err DB_SUCCESS or DB_DECRYPTION_FAILED @return position on log buffer */ UNIV_INTERN -const byte* +byte* fil_parse_write_crypt_data( - const byte* ptr, + byte* ptr, const byte* end_ptr, - const buf_block_t* block) + const buf_block_t* block, + dberr_t* err) MY_ATTRIBUTE((warn_unused_result)); /****************************************************************** diff --git a/storage/xtradb/include/fil0fil.h b/storage/xtradb/include/fil0fil.h index d73a68d9d34..2f03d2aa0f5 100644 --- a/storage/xtradb/include/fil0fil.h +++ b/storage/xtradb/include/fil0fil.h @@ -136,6 +136,7 @@ extern fil_addr_t fil_addr_null; #define FIL_PAGE_DATA 38 /*!< start of the data on the page */ /* Following are used when page compression is used */ + #define FIL_PAGE_COMPRESSED_SIZE 2 /*!< Number of bytes used to store actual payload data size on compressed pages. */ @@ -313,13 +314,21 @@ struct fil_space_t { ulint n_pending_flushes; /*!< this is positive when flushing the tablespace to disk; dropping of the tablespace is forbidden if this is positive */ - ulint n_pending_ops;/*!< this is positive when we - have pending operations against this - tablespace. The pending operations can - be ibuf merges or lock validation code - trying to read a block. - Dropping of the tablespace is forbidden - if this is positive */ + /** Number of pending buffer pool operations accessing the tablespace + without holding a table lock or dict_operation_lock S-latch + that would prevent the table (and tablespace) from being + dropped. An example is change buffer merge. + The tablespace cannot be dropped while this is nonzero, + or while fil_node_t::n_pending is nonzero. + Protected by fil_system->mutex. */ + ulint n_pending_ops; + /** Number of pending block read or write operations + (when a write is imminent or a read has recently completed). + The tablespace object cannot be freed while this is nonzero, + but it can be detached from fil_system. + Note that fil_node_t::n_pending tracks actual pending I/O requests. + Protected by fil_system->mutex. */ + ulint n_pending_ios; hash_node_t hash; /*!< hash chain node */ hash_node_t name_hash;/*!< hash chain the name_hash table */ #ifndef UNIV_HOTBACKUP @@ -650,13 +659,28 @@ fil_write_flushed_lsn_to_data_files( Used by background threads that do not necessarily hold proper locks for concurrency control. @param[in] id tablespace ID +@param[in] silent whether to silently ignore missing tablespaces +@return the tablespace +@retval NULL if missing or being deleted or truncated */ +UNIV_INTERN +fil_space_t* +fil_space_acquire_low(ulint id, bool silent) + MY_ATTRIBUTE((warn_unused_result)); + +/** Acquire a tablespace when it could be dropped concurrently. +Used by background threads that do not necessarily hold proper locks +for concurrency control. +@param[in] id tablespace ID @param[in] for_io whether to look up the tablespace while performing I/O (possibly executing TRUNCATE) @return the tablespace @retval NULL if missing or being deleted or truncated */ +inline fil_space_t* -fil_space_acquire(ulint id, bool for_io = false) - MY_ATTRIBUTE((warn_unused_result)); +fil_space_acquire(ulint id) +{ + return(fil_space_acquire_low(id, false)); +} /** Acquire a tablespace that may not exist. Used by background threads that do not necessarily hold proper locks @@ -664,15 +688,34 @@ for concurrency control. @param[in] id tablespace ID @return the tablespace @retval NULL if missing or being deleted */ +inline fil_space_t* fil_space_acquire_silent(ulint id) - MY_ATTRIBUTE((warn_unused_result)); +{ + return(fil_space_acquire_low(id, true)); +} /** Release a tablespace acquired with fil_space_acquire(). @param[in,out] space tablespace to release */ +UNIV_INTERN void fil_space_release(fil_space_t* space); +/** Acquire a tablespace for reading or writing a block, +when it could be dropped concurrently. +@param[in] id tablespace ID +@return the tablespace +@retval NULL if missing */ +UNIV_INTERN +fil_space_t* +fil_space_acquire_for_io(ulint id); + +/** Release a tablespace acquired with fil_space_acquire_for_io(). +@param[in,out] space tablespace to release */ +UNIV_INTERN +void +fil_space_release_for_io(fil_space_t* space); + /** Return the next fil_space_t. Once started, the caller must keep calling this until it returns NULL. fil_space_acquire() and fil_space_release() are invoked here which @@ -681,6 +724,7 @@ blocks a concurrent operation from dropping the tablespace. If NULL, use the first fil_space_t on fil_system->space_list. @return pointer to the next fil_space_t. @retval NULL if this was the last */ +UNIV_INTERN fil_space_t* fil_space_next( fil_space_t* prev_space) @@ -694,11 +738,67 @@ blocks a concurrent operation from dropping the tablespace. If NULL, use the first fil_space_t on fil_system->space_list. @return pointer to the next fil_space_t. @retval NULL if this was the last*/ +UNIV_INTERN fil_space_t* fil_space_keyrotate_next( fil_space_t* prev_space) MY_ATTRIBUTE((warn_unused_result)); +/** Wrapper with reference-counting for a fil_space_t. */ +class FilSpace +{ +public: + /** Default constructor: Use this when reference counting + is done outside this wrapper. */ + FilSpace() : m_space(NULL) {} + + /** Constructor: Look up the tablespace and increment the + reference count if found. + @param[in] space_id tablespace ID + @param[in] silent whether not to print any errors */ + explicit FilSpace(ulint space_id, bool silent = false) + : m_space(fil_space_acquire_low(space_id, silent)) {} + + /** Assignment operator: This assumes that fil_space_acquire() + has already been done for the fil_space_t. The caller must + assign NULL if it calls fil_space_release(). + @param[in] space tablespace to assign */ + class FilSpace& operator=(fil_space_t* space) + { + /* fil_space_acquire() must have been invoked. */ + ut_ad(space == NULL || space->n_pending_ops > 0); + m_space = space; + return(*this); + } + + /** Destructor - Decrement the reference count if a fil_space_t + is still assigned. */ + ~FilSpace() + { + if (m_space != NULL) { + fil_space_release(m_space); + } + } + + /** Implicit type conversion + @return the wrapped object */ + operator const fil_space_t*() const + { + return(m_space); + } + + /** Explicit type conversion + @return the wrapped object */ + const fil_space_t* operator()() const + { + return(m_space); + } + +private: + /** The wrapped pointer */ + fil_space_t* m_space; +}; + /*******************************************************************//** Reads the flushed lsn, arch no, and tablespace flag fields from a data file at database startup. @@ -721,22 +821,6 @@ fil_read_first_page( fil_space_crypt_t** crypt_data) /*!< out: crypt data */ __attribute__((warn_unused_result)); -/*******************************************************************//** -Increments the count of pending operation, if space is not being deleted. -@return TRUE if being deleted, and operation should be skipped */ -UNIV_INTERN -ibool -fil_inc_pending_ops( -/*================*/ - ulint id, /*!< in: space id */ - ibool print_err); /*!< in: need to print error or not */ -/*******************************************************************//** -Decrements the count of pending operations. */ -UNIV_INTERN -void -fil_decr_pending_ops( -/*=================*/ - ulint id); /*!< in: space id */ #endif /* !UNIV_HOTBACKUP */ /*******************************************************************//** Parses the body of a log record written about an .ibd file operation. That is, @@ -978,7 +1062,7 @@ space id is != 0. @return DB_SUCCESS or error number */ UNIV_INTERN dberr_t -fil_load_single_table_tablespaces(void); +fil_load_single_table_tablespaces(ibool (*pred)(const char*, const char*)=0); /*===================================*/ /*******************************************************************//** Returns TRUE if a single-table tablespace does not exist in the memory cache, @@ -1017,6 +1101,9 @@ fil_space_for_table_exists_in_mem( information to the .err log if a matching tablespace is not found from memory */ + bool remove_from_data_dict_if_does_not_exist, + /*!< in: remove from the data dictionary + if tablespace does not exist */ bool adjust_space, /*!< in: whether to adjust space id when find table space mismatch */ mem_heap_t* heap, /*!< in: heap memory */ @@ -1428,36 +1515,6 @@ fil_space_get_by_id( /*================*/ ulint id); /*!< in: space id */ -/****************************************************************** -Get id of first tablespace or ULINT_UNDEFINED if none */ -UNIV_INTERN -ulint -fil_get_first_space(); -/*=================*/ - -/****************************************************************** -Get id of next tablespace or ULINT_UNDEFINED if none */ -UNIV_INTERN -ulint -fil_get_next_space( -/*===============*/ - ulint id); /*!< in: space id */ - -/****************************************************************** -Get id of first tablespace that has node or ULINT_UNDEFINED if none */ -UNIV_INTERN -ulint -fil_get_first_space_safe(); -/*======================*/ - -/****************************************************************** -Get id of next tablespace that has node or ULINT_UNDEFINED if none */ -UNIV_INTERN -ulint -fil_get_next_space_safe( -/*====================*/ - ulint id); /*!< in: previous space id */ - #endif /* UNIV_INNOCHECKSUM */ /****************************************************************//** diff --git a/storage/xtradb/include/os0sync.h b/storage/xtradb/include/os0sync.h index 48c56a73369..62f651413e1 100644 --- a/storage/xtradb/include/os0sync.h +++ b/storage/xtradb/include/os0sync.h @@ -2,6 +2,7 @@ Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2008, Google Inc. +Copyright (c) 2017, MariaDB Corporation. Portions of this file contain modifications contributed and copyrighted by Google, Inc. Those modifications are gratefully acknowledged and are described @@ -38,12 +39,11 @@ Created 9/6/1995 Heikki Tuuri #include "ut0lst.h" #include "sync0types.h" -#if defined __i386__ || defined __x86_64__ || defined _M_IX86 \ - || defined _M_X64 || defined __WIN__ - -#define IB_STRONG_MEMORY_MODEL - -#endif /* __i386__ || __x86_64__ || _M_IX86 || _M_X64 || __WIN__ */ +#ifdef CPU_LEVEL1_DCACHE_LINESIZE +# define CACHE_LINE_SIZE CPU_LEVEL1_DCACHE_LINESIZE +#else +# error CPU_LEVEL1_DCACHE_LINESIZE is undefined +#endif /* CPU_LEVEL1_DCACHE_LINESIZE */ #ifdef HAVE_WINDOWS_ATOMICS typedef LONG lock_word_t; /*!< On Windows, InterlockedExchange operates @@ -940,6 +940,51 @@ for synchronization */ "Memory barrier is not used" #endif + +/** Simple counter aligned to CACHE_LINE_SIZE +@tparam Type the integer type of the counter +@tparam atomic whether to use atomic memory access */ +template <typename Type = ulint, bool atomic = false> +struct MY_ALIGNED(CACHE_LINE_SIZE) simple_counter +{ + /** Increment the counter */ + Type inc() { return add(1); } + /** Decrement the counter */ + Type dec() { return sub(1); } + + /** Add to the counter + @param[in] i amount to be added + @return the value of the counter after adding */ + Type add(Type i) + { + compile_time_assert(!atomic || sizeof(Type) == sizeof(ulint)); + if (atomic) { + return os_atomic_increment_ulint(&m_counter, i); + } else { + return m_counter += i; + } + } + /** Subtract from the counter + @param[in] i amount to be subtracted + @return the value of the counter after adding */ + Type sub(Type i) + { + compile_time_assert(!atomic || sizeof(Type) == sizeof(ulint)); + if (atomic) { + return os_atomic_decrement_ulint(&m_counter, i); + } else { + return m_counter -= i; + } + } + + /** @return the value of the counter (non-atomic access)! */ + operator Type() const { return m_counter; } + +private: + /** The counter */ + Type m_counter; +}; + #ifndef UNIV_NONINL #include "os0sync.ic" #endif diff --git a/storage/xtradb/include/srv0srv.h b/storage/xtradb/include/srv0srv.h index c18bc7c1fc3..cf7824d91e7 100644 --- a/storage/xtradb/include/srv0srv.h +++ b/storage/xtradb/include/srv0srv.h @@ -3,7 +3,7 @@ Copyright (c) 1995, 2016, Oracle and/or its affiliates. All rights reserved. Copyright (c) 2008, 2009, Google Inc. Copyright (c) 2009, Percona Inc. -Copyright (c) 2013, 2017, MariaDB Corporation Ab. All Rights Reserved. +Copyright (c) 2013, 2017, MariaDB Corporation. Portions of this file contain modifications contributed and copyrighted by Google, Inc. Those modifications are gratefully acknowledged and are described @@ -55,11 +55,10 @@ Created 10/10/1995 Heikki Tuuri /* Global counters used inside InnoDB. */ struct srv_stats_t { - typedef ib_counter_t<lsn_t, 1, single_indexer_t> lsn_ctr_1_t; - typedef ib_counter_t<ulint, 1, single_indexer_t> ulint_ctr_1_t; - typedef ib_counter_t<lint, 1, single_indexer_t> lint_ctr_1_t; typedef ib_counter_t<ulint, 64> ulint_ctr_64_t; - typedef ib_counter_t<ib_int64_t, 1, single_indexer_t> ib_int64_ctr_1_t; + typedef simple_counter<lsn_t> lsn_ctr_1_t; + typedef simple_counter<ulint> ulint_ctr_1_t; + typedef simple_counter<ib_int64_t> ib_int64_ctr_1_t; /** Count the amount of data written in total (in bytes) */ ulint_ctr_1_t data_written; @@ -73,8 +72,9 @@ struct srv_stats_t { /** Amount of data written to the log files in bytes */ lsn_ctr_1_t os_log_written; - /** Number of writes being done to the log files */ - lint_ctr_1_t os_log_pending_writes; + /** Number of writes being done to the log files. + Protected by log_sys->write_mutex. */ + ulint_ctr_1_t os_log_pending_writes; /** We increase this counter, when we don't have enough space in the log buffer and have to flush it */ @@ -148,7 +148,7 @@ struct srv_stats_t { ulint_ctr_1_t n_lock_wait_count; /** Number of threads currently waiting on database locks */ - lint_ctr_1_t n_lock_wait_current_count; + simple_counter<ulint, true> n_lock_wait_current_count; /** Number of rows read. */ ulint_ctr_64_t n_rows_read; @@ -496,7 +496,9 @@ as enum type because the configure option takes unsigned integer type. */ extern ulong srv_innodb_stats_method; #ifdef UNIV_LOG_ARCHIVE -extern ibool srv_log_archive_on; +extern bool srv_log_archive_on; +extern bool srv_archive_recovery; +extern ib_uint64_t srv_archive_recovery_limit_lsn; #endif /* UNIV_LOG_ARCHIVE */ extern char* srv_file_flush_method_str; @@ -547,6 +549,14 @@ extern ulong srv_pass_corrupt_table; extern ulong srv_log_checksum_algorithm; +extern bool srv_apply_log_only; + +extern bool srv_backup_mode; +extern bool srv_close_files; +extern bool srv_xtrabackup; + +#define IS_XTRABACKUP() (srv_xtrabackup) + extern my_bool srv_force_primary_key; /* Helper macro to support srv_pass_corrupt_table checks. If 'cond' is FALSE, diff --git a/storage/xtradb/include/trx0sys.h b/storage/xtradb/include/trx0sys.h index 0c18b657fd7..9bfffd09532 100644 --- a/storage/xtradb/include/trx0sys.h +++ b/storage/xtradb/include/trx0sys.h @@ -336,8 +336,9 @@ trx_sys_update_wsrep_checkpoint( trx_sysf_t* sys_header, /*!< in: sys_header */ mtr_t* mtr); /*!< in: mtr */ -void -/** Read WSREP checkpoint XID from sys header. */ +/** Read WSREP checkpoint XID from sys header. +@return true on success, false on error. */ +bool trx_sys_read_wsrep_checkpoint( XID* xid); /*!< out: WSREP XID */ #endif /* WITH_WSREP */ diff --git a/storage/xtradb/include/univ.i b/storage/xtradb/include/univ.i index e698f08f15b..310053b9145 100644 --- a/storage/xtradb/include/univ.i +++ b/storage/xtradb/include/univ.i @@ -635,7 +635,7 @@ functions. */ #ifdef __WIN__ #define usleep(a) Sleep((a)/1000) -typedef ulint os_thread_ret_t; +typedef DWORD os_thread_ret_t; #define OS_THREAD_DUMMY_RETURN return(0) #else typedef void* os_thread_ret_t; diff --git a/storage/xtradb/include/ut0counter.h b/storage/xtradb/include/ut0counter.h index 447484ba985..4f736428a17 100644 --- a/storage/xtradb/include/ut0counter.h +++ b/storage/xtradb/include/ut0counter.h @@ -1,6 +1,7 @@ /***************************************************************************** Copyright (c) 2012, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2017, MariaDB Corporation. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -48,8 +49,6 @@ Created 2012/04/12 by Sunny Bains /** Get the offset into the counter array. */ template <typename Type, int N> struct generic_indexer_t { - /** Default constructor/destructor should be OK. */ - /** @return offset within m_counter */ size_t offset(size_t index) const UNIV_NOTHROW { return(((index % N) + 1) * (CACHE_LINE_SIZE / sizeof(Type))); @@ -62,8 +61,6 @@ struct generic_indexer_t { use the thread id. */ template <typename Type, int N> struct get_sched_indexer_t : public generic_indexer_t<Type, N> { - /** Default constructor/destructor should be OK. */ - /* @return result from sched_getcpu(), the thread id if it fails. */ size_t get_rnd_index() const UNIV_NOTHROW { @@ -80,31 +77,17 @@ struct get_sched_indexer_t : public generic_indexer_t<Type, N> { /** Use the thread id to index into the counter array. */ template <typename Type, int N> struct thread_id_indexer_t : public generic_indexer_t<Type, N> { - /** Default constructor/destructor should are OK. */ - /* @return a random number, currently we use the thread id. Where thread id is represented as a pointer, it may not work as effectively. */ size_t get_rnd_index() const UNIV_NOTHROW { return((lint) os_thread_get_curr_id()); } -}; - -/** For counters wher N=1 */ -template <typename Type, int N=1> -struct single_indexer_t { - /** Default constructor/destructor should are OK. */ - - /** @return offset within m_counter */ - size_t offset(size_t index) const UNIV_NOTHROW { - ut_ad(N == 1); - return((CACHE_LINE_SIZE / sizeof(Type))); - } - /* @return 1 */ - size_t get_rnd_index() const UNIV_NOTHROW { - ut_ad(N == 1); - return(1); + /** @return a random offset to the array */ + size_t get_rnd_offset() const UNIV_NOTHROW + { + return(generic_indexer_t<Type, N>::offset(get_rnd_index())); } }; @@ -116,17 +99,11 @@ template < typename Type, int N = IB_N_SLOTS, template<typename, int> class Indexer = thread_id_indexer_t> -class ib_counter_t { -public: - ib_counter_t() { memset(m_counter, 0x0, sizeof(m_counter)); } - +struct MY_ALIGNED(CACHE_LINE_SIZE) ib_counter_t +{ +#ifdef UNIV_DEBUG ~ib_counter_t() { - ut_ad(validate()); - } - - bool validate() UNIV_NOTHROW { -#ifdef UNIV_DEBUG size_t n = (CACHE_LINE_SIZE / sizeof(Type)); /* Check that we aren't writing outside our defined bounds. */ @@ -135,27 +112,23 @@ public: ut_ad(m_counter[i + j] == 0); } } -#endif /* UNIV_DEBUG */ - return(true); } +#endif /* UNIV_DEBUG */ - /** If you can't use a good index id. Increment by 1. */ + /** Increment the counter by 1. */ void inc() UNIV_NOTHROW { add(1); } - /** If you can't use a good index id. - * @param n - is the amount to increment */ - void add(Type n) UNIV_NOTHROW { - size_t i = m_policy.offset(m_policy.get_rnd_index()); - - ut_ad(i < UT_ARR_SIZE(m_counter)); + /** Increment the counter by 1. + @param[in] index a reasonably thread-unique identifier */ + void inc(size_t index) UNIV_NOTHROW { add(index, 1); } - m_counter[i] += n; - } + /** Add to the counter. + @param[in] n amount to be added */ + void add(Type n) UNIV_NOTHROW { add(m_policy.get_rnd_offset(), n); } - /** Use this if you can use a unique indentifier, saves a - call to get_rnd_index(). - @param i - index into a slot - @param n - amount to increment */ + /** Add to the counter. + @param[in] index a reasonably thread-unique identifier + @param[in] n amount to be added */ void add(size_t index, Type n) UNIV_NOTHROW { size_t i = m_policy.offset(index); @@ -164,31 +137,6 @@ public: m_counter[i] += n; } - /** If you can't use a good index id. Decrement by 1. */ - void dec() UNIV_NOTHROW { sub(1); } - - /** If you can't use a good index id. - * @param - n is the amount to decrement */ - void sub(Type n) UNIV_NOTHROW { - size_t i = m_policy.offset(m_policy.get_rnd_index()); - - ut_ad(i < UT_ARR_SIZE(m_counter)); - - m_counter[i] -= n; - } - - /** Use this if you can use a unique indentifier, saves a - call to get_rnd_index(). - @param i - index into a slot - @param n - amount to decrement */ - void sub(size_t index, Type n) UNIV_NOTHROW { - size_t i = m_policy.offset(index); - - ut_ad(i < UT_ARR_SIZE(m_counter)); - - m_counter[i] -= n; - } - /* @return total value - not 100% accurate, since it is not atomic. */ operator Type() const UNIV_NOTHROW { Type total = 0; |