diff options
author | Georgi Kodinov <Georgi.Kodinov@Oracle.com> | 2011-01-10 15:09:57 +0200 |
---|---|---|
committer | Georgi Kodinov <Georgi.Kodinov@Oracle.com> | 2011-01-10 15:09:57 +0200 |
commit | 1ecfdbf2bc99b340f3caea30228cd562133c50ae (patch) | |
tree | c07fefa8d085a12ec5bb331be01664c0e3431eb7 /storage | |
parent | b0306dc50687a872e6b0791f6780827fee760aae (diff) | |
parent | 90650edf697d498d0969f4bd9b81301ffc633dfa (diff) | |
download | mariadb-git-1ecfdbf2bc99b340f3caea30228cd562133c50ae.tar.gz |
merge
Diffstat (limited to 'storage')
42 files changed, 1559 insertions, 737 deletions
diff --git a/storage/innobase/CMakeLists.txt b/storage/innobase/CMakeLists.txt index fa45cc96687..4bbda0d2477 100644 --- a/storage/innobase/CMakeLists.txt +++ b/storage/innobase/CMakeLists.txt @@ -40,10 +40,9 @@ IF(UNIX) ENDIF() ENDIF() -# Enable InnoDB's UNIV_DEBUG if MySQL's WITH_DEBUG is defined -IF(WITH_DEBUG) - ADD_DEFINITIONS("-DUNIV_DEBUG") -ENDIF() +# Enable InnoDB's UNIV_DEBUG for debug builds +SET(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} -DUNIV_DEBUG") +SET(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -DUNIV_DEBUG") IF(NOT MSVC) # either define HAVE_IB_GCC_ATOMIC_BUILTINS or not @@ -190,7 +189,7 @@ IF(SIZEOF_PTHREAD_T) ENDIF() IF(MSVC) - ADD_DEFINITIONS(-DHAVE_WINDOWS_ATOMICS -DHAVE_IB_PAUSE_INSTRUCTION) + ADD_DEFINITIONS(-DHAVE_WINDOWS_ATOMICS) ENDIF() diff --git a/storage/innobase/btr/btr0btr.c b/storage/innobase/btr/btr0btr.c index 06cfdf5a798..ff82372b52b 100644 --- a/storage/innobase/btr/btr0btr.c +++ b/storage/innobase/btr/btr0btr.c @@ -664,7 +664,7 @@ btr_page_get_father_node_ptr_func( " to fix the\n" "InnoDB: corruption. If the crash happens at " "the database startup, see\n" - "InnoDB: " REFMAN "forcing-recovery.html about\n" + "InnoDB: " REFMAN "forcing-innodb-recovery.html about\n" "InnoDB: forcing recovery. " "Then dump + drop + reimport.\n", stderr); diff --git a/storage/innobase/btr/btr0cur.c b/storage/innobase/btr/btr0cur.c index 50489ae4cc6..a5856479147 100644 --- a/storage/innobase/btr/btr0cur.c +++ b/storage/innobase/btr/btr0cur.c @@ -1874,7 +1874,8 @@ btr_cur_update_in_place( NOT call it if index is secondary */ if (!dict_index_is_clust(index) - || row_upd_changes_ord_field_binary(NULL, index, update)) { + || row_upd_changes_ord_field_binary(NULL, NULL, + index, update)) { /* Remove possible hash index pointer to this record */ btr_search_update_hash_on_delete(cursor); @@ -2626,27 +2627,24 @@ ulint btr_cur_del_mark_set_clust_rec( /*===========================*/ ulint flags, /*!< in: undo logging and locking flags */ - btr_cur_t* cursor, /*!< in: cursor */ + buf_block_t* block, /*!< in/out: buffer block of the record */ + rec_t* rec, /*!< in/out: record */ + dict_index_t* index, /*!< in: clustered index of the record */ + const ulint* offsets,/*!< in: rec_get_offsets(rec) */ ibool val, /*!< in: value to set */ que_thr_t* thr, /*!< in: query thread */ mtr_t* mtr) /*!< in: mtr */ { - dict_index_t* index; - buf_block_t* block; roll_ptr_t roll_ptr; ulint err; - rec_t* rec; page_zip_des_t* page_zip; trx_t* trx; - mem_heap_t* heap = NULL; - ulint offsets_[REC_OFFS_NORMAL_SIZE]; - ulint* offsets = offsets_; - rec_offs_init(offsets_); - rec = btr_cur_get_rec(cursor); - index = cursor->index; + ut_ad(dict_index_is_clust(index)); + ut_ad(rec_offs_validate(rec, index, offsets)); ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table)); - offsets = rec_get_offsets(rec, index, offsets, ULINT_UNDEFINED, &heap); + ut_ad(buf_block_get_frame(block) == page_align(rec)); + ut_ad(page_is_leaf(page_align(rec))); #ifdef UNIV_DEBUG if (btr_cur_print_record_ops && thr) { @@ -2658,13 +2656,12 @@ btr_cur_del_mark_set_clust_rec( ut_ad(dict_index_is_clust(index)); ut_ad(!rec_get_deleted_flag(rec, rec_offs_comp(offsets))); - err = lock_clust_rec_modify_check_and_lock(flags, - btr_cur_get_block(cursor), + err = lock_clust_rec_modify_check_and_lock(flags, block, rec, index, offsets, thr); if (err != DB_SUCCESS) { - goto func_exit; + return(err); } err = trx_undo_report_row_operation(flags, TRX_UNDO_MODIFY_OP, thr, @@ -2672,11 +2669,9 @@ btr_cur_del_mark_set_clust_rec( &roll_ptr); if (err != DB_SUCCESS) { - goto func_exit; + return(err); } - block = btr_cur_get_block(cursor); - if (block->is_hashed) { rw_lock_x_lock(&btr_search_latch); } @@ -2699,10 +2694,6 @@ btr_cur_del_mark_set_clust_rec( btr_cur_del_mark_set_clust_rec_log(flags, rec, index, val, trx, roll_ptr, mtr); -func_exit: - if (UNIV_LIKELY_NULL(heap)) { - mem_heap_free(heap); - } return(err); } @@ -3754,107 +3745,35 @@ btr_cur_set_ownership_of_extern_field( } /*******************************************************************//** -Marks not updated extern fields as not-owned by this record. The ownership -is transferred to the updated record which is inserted elsewhere in the +Marks non-updated off-page fields as disowned by this record. The ownership +must be transferred to the updated record which is inserted elsewhere in the index tree. In purge only the owner of externally stored field is allowed -to free the field. -@return TRUE if BLOB ownership was transferred */ +to free the field. */ UNIV_INTERN -ibool -btr_cur_mark_extern_inherited_fields( -/*=================================*/ +void +btr_cur_disown_inherited_fields( +/*============================*/ page_zip_des_t* page_zip,/*!< in/out: compressed page whose uncompressed part will be updated, or NULL */ rec_t* rec, /*!< in/out: record in a clustered index */ dict_index_t* index, /*!< in: index of the page */ const ulint* offsets,/*!< in: array returned by rec_get_offsets() */ const upd_t* update, /*!< in: update vector */ - mtr_t* mtr) /*!< in: mtr, or NULL if not logged */ + mtr_t* mtr) /*!< in/out: mini-transaction */ { - ulint n; - ulint j; ulint i; - ibool change_ownership = FALSE; - ut_ad(rec_offs_validate(rec, NULL, offsets)); + ut_ad(rec_offs_validate(rec, index, offsets)); ut_ad(!rec_offs_comp(offsets) || !rec_get_node_ptr_flag(rec)); + ut_ad(rec_offs_any_extern(offsets)); + ut_ad(mtr); - if (!rec_offs_any_extern(offsets)) { - - return(FALSE); - } - - n = rec_offs_n_fields(offsets); - - for (i = 0; i < n; i++) { - if (rec_offs_nth_extern(offsets, i)) { - - /* Check it is not in updated fields */ - - if (update) { - for (j = 0; j < upd_get_n_fields(update); - j++) { - if (upd_get_nth_field(update, j) - ->field_no == i) { - - goto updated; - } - } - } - + for (i = 0; i < rec_offs_n_fields(offsets); i++) { + if (rec_offs_nth_extern(offsets, i) + && !upd_get_field_by_field_no(update, i)) { btr_cur_set_ownership_of_extern_field( page_zip, rec, index, offsets, i, FALSE, mtr); - - change_ownership = TRUE; -updated: - ; - } - } - - return(change_ownership); -} - -/*******************************************************************//** -The complement of the previous function: in an update entry may inherit -some externally stored fields from a record. We must mark them as inherited -in entry, so that they are not freed in a rollback. */ -UNIV_INTERN -void -btr_cur_mark_dtuple_inherited_extern( -/*=================================*/ - dtuple_t* entry, /*!< in/out: updated entry to be - inserted to clustered index */ - const upd_t* update) /*!< in: update vector */ -{ - ulint i; - - for (i = 0; i < dtuple_get_n_fields(entry); i++) { - - dfield_t* dfield = dtuple_get_nth_field(entry, i); - byte* data; - ulint len; - ulint j; - - if (!dfield_is_ext(dfield)) { - continue; - } - - /* Check if it is in updated fields */ - - for (j = 0; j < upd_get_n_fields(update); j++) { - if (upd_get_nth_field(update, j)->field_no == i) { - - goto is_updated; - } } - - data = dfield_get_data(dfield); - len = dfield_get_len(dfield); - data[len - BTR_EXTERN_FIELD_REF_SIZE + BTR_EXTERN_LEN] - |= BTR_EXTERN_INHERITED_FLAG; - -is_updated: - ; } } @@ -3894,29 +3813,6 @@ btr_cur_unmark_extern_fields( } /*******************************************************************//** -Marks all extern fields in a dtuple as owned by the record. */ -UNIV_INTERN -void -btr_cur_unmark_dtuple_extern_fields( -/*================================*/ - dtuple_t* entry) /*!< in/out: clustered index entry */ -{ - ulint i; - - for (i = 0; i < dtuple_get_n_fields(entry); i++) { - dfield_t* dfield = dtuple_get_nth_field(entry, i); - - if (dfield_is_ext(dfield)) { - byte* data = dfield_get_data(dfield); - ulint len = dfield_get_len(dfield); - - data[len - BTR_EXTERN_FIELD_REF_SIZE + BTR_EXTERN_LEN] - &= ~BTR_EXTERN_OWNER_FLAG; - } - } -} - -/*******************************************************************//** Flags the data tuple fields that are marked as extern storage in the update vector. We use this function to remember which fields we must mark as extern storage in a record inserted for an update. diff --git a/storage/innobase/buf/buf0buf.c b/storage/innobase/buf/buf0buf.c index b79b5634957..3ff69d8fd1b 100644 --- a/storage/innobase/buf/buf0buf.c +++ b/storage/innobase/buf/buf0buf.c @@ -536,7 +536,7 @@ buf_page_is_corrupted( "you may have copied the InnoDB\n" "InnoDB: tablespace but not the InnoDB " "log files. See\n" - "InnoDB: " REFMAN "forcing-recovery.html\n" + "InnoDB: " REFMAN "forcing-innodb-recovery.html\n" "InnoDB: for more information.\n", (ulong) mach_read_from_4(read_buf + FIL_PAGE_OFFSET), @@ -2745,7 +2745,8 @@ buf_page_get_gen( ut_ad(zip_size == fil_space_get_zip_size(space)); ut_ad(ut_is_2pow(zip_size)); #ifndef UNIV_LOG_DEBUG - ut_ad(!ibuf_inside() || ibuf_page(space, zip_size, offset, NULL)); + ut_ad(!ibuf_inside() || ibuf_page_low(space, zip_size, offset, + FALSE, file, line, NULL)); #endif buf_pool->stat.n_page_gets++; fold = buf_page_address_fold(space, offset); @@ -4038,7 +4039,7 @@ corrupt: "InnoDB: TABLE to scan your" " table for corruption.\n" "InnoDB: See also " - REFMAN "forcing-recovery.html\n" + REFMAN "forcing-innodb-recovery.html\n" "InnoDB: about forcing recovery.\n", stderr); if (srv_force_recovery < SRV_FORCE_IGNORE_CORRUPT) { @@ -4793,23 +4794,203 @@ buf_get_modified_ratio_pct(void) return(ratio); } +/*******************************************************************//** +Aggregates a pool stats information with the total buffer pool stats */ +static +void +buf_stats_aggregate_pool_info( +/*==========================*/ + buf_pool_info_t* total_info, /*!< in/out: the buffer pool + info to store aggregated + result */ + const buf_pool_info_t* pool_info) /*!< in: individual buffer pool + stats info */ +{ + ut_a(total_info && pool_info); + + /* Nothing to copy if total_info is the same as pool_info */ + if (total_info == pool_info) { + return; + } + + total_info->pool_size += pool_info->pool_size; + total_info->lru_len += pool_info->lru_len; + total_info->old_lru_len += pool_info->old_lru_len; + total_info->free_list_len += pool_info->free_list_len; + total_info->flush_list_len += pool_info->flush_list_len; + total_info->n_pend_unzip += pool_info->n_pend_unzip; + total_info->n_pend_reads += pool_info->n_pend_reads; + total_info->n_pending_flush_lru += pool_info->n_pending_flush_lru; + total_info->n_pending_flush_list += pool_info->n_pending_flush_list; + total_info->n_pending_flush_single_page += + pool_info->n_pending_flush_single_page; + total_info->n_pages_made_young += pool_info->n_pages_made_young; + total_info->n_pages_not_made_young += pool_info->n_pages_not_made_young; + total_info->n_pages_read += pool_info->n_pages_read; + total_info->n_pages_created += pool_info->n_pages_created; + total_info->n_pages_written += pool_info->n_pages_written; + total_info->n_page_gets += pool_info->n_page_gets; + total_info->n_ra_pages_read += pool_info->n_ra_pages_read; + total_info->n_ra_pages_evicted += pool_info->n_ra_pages_evicted; + total_info->page_made_young_rate += pool_info->page_made_young_rate; + total_info->page_not_made_young_rate += + pool_info->page_not_made_young_rate; + total_info->pages_read_rate += pool_info->pages_read_rate; + total_info->pages_created_rate += pool_info->pages_created_rate; + total_info->pages_written_rate += pool_info->pages_written_rate; + total_info->n_page_get_delta += pool_info->n_page_get_delta; + total_info->page_read_delta += pool_info->page_read_delta; + total_info->young_making_delta += pool_info->young_making_delta; + total_info->not_young_making_delta += pool_info->not_young_making_delta; + total_info->pages_readahead_rate += pool_info->pages_readahead_rate; + total_info->pages_evicted_rate += pool_info->pages_evicted_rate; + total_info->unzip_lru_len += pool_info->unzip_lru_len; + total_info->io_sum += pool_info->io_sum; + total_info->io_cur += pool_info->io_cur; + total_info->unzip_sum += pool_info->unzip_sum; + total_info->unzip_cur += pool_info->unzip_cur; +} +/*******************************************************************//** +Collect buffer pool stats information for a buffer pool. Also +record aggregated stats if there are more than one buffer pool +in the server */ +static +void +buf_stats_get_pool_info( +/*====================*/ + buf_pool_t* buf_pool, /*!< in: buffer pool */ + ulint pool_id, /*!< in: buffer pool ID */ + buf_pool_info_t* all_pool_info) /*!< in/out: buffer pool info + to fill */ +{ + buf_pool_info_t* pool_info; + time_t current_time; + double time_elapsed; + + /* Find appropriate pool_info to store stats for this buffer pool */ + pool_info = &all_pool_info[pool_id]; + + buf_pool_mutex_enter(buf_pool); + buf_flush_list_mutex_enter(buf_pool); + + pool_info->pool_unique_id = pool_id; + + pool_info->pool_size = buf_pool->curr_size; + + pool_info->lru_len = UT_LIST_GET_LEN(buf_pool->LRU); + + pool_info->old_lru_len = buf_pool->LRU_old_len; + + pool_info->free_list_len = UT_LIST_GET_LEN(buf_pool->free); + + pool_info->flush_list_len = UT_LIST_GET_LEN(buf_pool->flush_list); + + pool_info->n_pend_unzip = UT_LIST_GET_LEN(buf_pool->unzip_LRU); + + pool_info->n_pend_reads = buf_pool->n_pend_reads; + + pool_info->n_pending_flush_lru = + (buf_pool->n_flush[BUF_FLUSH_LRU] + + buf_pool->init_flush[BUF_FLUSH_LRU]); + + pool_info->n_pending_flush_list = + (buf_pool->n_flush[BUF_FLUSH_LIST] + + buf_pool->init_flush[BUF_FLUSH_LIST]); + + pool_info->n_pending_flush_single_page = + buf_pool->n_flush[BUF_FLUSH_SINGLE_PAGE]; + + buf_flush_list_mutex_exit(buf_pool); + + current_time = time(NULL); + time_elapsed = 0.001 + difftime(current_time, + buf_pool->last_printout_time); + + pool_info->n_pages_made_young = buf_pool->stat.n_pages_made_young; + + pool_info->n_pages_not_made_young = + buf_pool->stat.n_pages_not_made_young; + + pool_info->n_pages_read = buf_pool->stat.n_pages_read; + + pool_info->n_pages_created = buf_pool->stat.n_pages_created; + + pool_info->n_pages_written = buf_pool->stat.n_pages_written; + + pool_info->n_page_gets = buf_pool->stat.n_page_gets; + + pool_info->n_ra_pages_read = buf_pool->stat.n_ra_pages_read; + + pool_info->n_ra_pages_evicted = buf_pool->stat.n_ra_pages_evicted; + + pool_info->page_made_young_rate = + (buf_pool->stat.n_pages_made_young + - buf_pool->old_stat.n_pages_made_young) / time_elapsed; + + pool_info->page_not_made_young_rate = + (buf_pool->stat.n_pages_not_made_young + - buf_pool->old_stat.n_pages_not_made_young) / time_elapsed; + + pool_info->pages_read_rate = + (buf_pool->stat.n_pages_read + - buf_pool->old_stat.n_pages_read) / time_elapsed; + + pool_info->pages_created_rate = + (buf_pool->stat.n_pages_created + - buf_pool->old_stat.n_pages_created) / time_elapsed; + + pool_info->pages_written_rate = + (buf_pool->stat.n_pages_written + - buf_pool->old_stat.n_pages_written) / time_elapsed; + + pool_info->n_page_get_delta = buf_pool->stat.n_page_gets + - buf_pool->old_stat.n_page_gets; + + if (pool_info->n_page_get_delta) { + pool_info->page_read_delta = buf_pool->stat.n_pages_read + - buf_pool->old_stat.n_pages_read; + + pool_info->young_making_delta = + buf_pool->stat.n_pages_made_young + - buf_pool->old_stat.n_pages_made_young; + + pool_info->not_young_making_delta = + buf_pool->stat.n_pages_not_made_young + - buf_pool->old_stat.n_pages_not_made_young; + } + + pool_info->pages_readahead_rate = + (buf_pool->stat.n_ra_pages_read + - buf_pool->old_stat.n_ra_pages_read) / time_elapsed; + + pool_info->pages_evicted_rate = + (buf_pool->stat.n_ra_pages_evicted + - buf_pool->old_stat.n_ra_pages_evicted) / time_elapsed; + + pool_info->unzip_lru_len = UT_LIST_GET_LEN(buf_pool->unzip_LRU); + + pool_info->io_sum = buf_LRU_stat_sum.io; + + pool_info->io_cur = buf_LRU_stat_cur.io; + + pool_info->unzip_sum = buf_LRU_stat_sum.unzip; + + pool_info->unzip_cur = buf_LRU_stat_cur.unzip; + + buf_refresh_io_stats(buf_pool); + buf_pool_mutex_exit(buf_pool); +} + /*********************************************************************//** Prints info of the buffer i/o. */ UNIV_INTERN void buf_print_io_instance( /*==================*/ - buf_pool_t* buf_pool, /*!< in: buffer pool instance */ + buf_pool_info_t*pool_info, /*!< in: buffer pool info */ FILE* file) /*!< in/out: buffer where to print */ { - time_t current_time; - double time_elapsed; - ulint n_gets_diff; - - ut_ad(buf_pool); - - buf_pool_mutex_enter(buf_pool); - buf_flush_list_mutex_enter(buf_pool); + ut_ad(pool_info); fprintf(file, "Buffer pool size %lu\n" @@ -4819,70 +5000,42 @@ buf_print_io_instance( "Modified db pages %lu\n" "Pending reads %lu\n" "Pending writes: LRU %lu, flush list %lu, single page %lu\n", - (ulong) buf_pool->curr_size, - (ulong) UT_LIST_GET_LEN(buf_pool->free), - (ulong) UT_LIST_GET_LEN(buf_pool->LRU), - (ulong) buf_pool->LRU_old_len, - (ulong) UT_LIST_GET_LEN(buf_pool->flush_list), - (ulong) buf_pool->n_pend_reads, - (ulong) buf_pool->n_flush[BUF_FLUSH_LRU] - + buf_pool->init_flush[BUF_FLUSH_LRU], - (ulong) buf_pool->n_flush[BUF_FLUSH_LIST] - + buf_pool->init_flush[BUF_FLUSH_LIST], - (ulong) buf_pool->n_flush[BUF_FLUSH_SINGLE_PAGE]); - - buf_flush_list_mutex_exit(buf_pool); - - current_time = time(NULL); - time_elapsed = 0.001 + difftime(current_time, - buf_pool->last_printout_time); + pool_info->pool_size, + pool_info->free_list_len, + pool_info->lru_len, + pool_info->old_lru_len, + pool_info->flush_list_len, + pool_info->n_pend_reads, + pool_info->n_pending_flush_lru, + pool_info->n_pending_flush_list, + pool_info->n_pending_flush_single_page); fprintf(file, "Pages made young %lu, not young %lu\n" "%.2f youngs/s, %.2f non-youngs/s\n" "Pages read %lu, created %lu, written %lu\n" "%.2f reads/s, %.2f creates/s, %.2f writes/s\n", - (ulong) buf_pool->stat.n_pages_made_young, - (ulong) buf_pool->stat.n_pages_not_made_young, - (buf_pool->stat.n_pages_made_young - - buf_pool->old_stat.n_pages_made_young) - / time_elapsed, - (buf_pool->stat.n_pages_not_made_young - - buf_pool->old_stat.n_pages_not_made_young) - / time_elapsed, - (ulong) buf_pool->stat.n_pages_read, - (ulong) buf_pool->stat.n_pages_created, - (ulong) buf_pool->stat.n_pages_written, - (buf_pool->stat.n_pages_read - - buf_pool->old_stat.n_pages_read) - / time_elapsed, - (buf_pool->stat.n_pages_created - - buf_pool->old_stat.n_pages_created) - / time_elapsed, - (buf_pool->stat.n_pages_written - - buf_pool->old_stat.n_pages_written) - / time_elapsed); - - n_gets_diff = buf_pool->stat.n_page_gets - - buf_pool->old_stat.n_page_gets; - - if (n_gets_diff) { + pool_info->n_pages_made_young, + pool_info->n_pages_not_made_young, + pool_info->page_made_young_rate, + pool_info->page_not_made_young_rate, + pool_info->n_pages_read, + pool_info->n_pages_created, + pool_info->n_pages_written, + pool_info->pages_read_rate, + pool_info->pages_created_rate, + pool_info->pages_written_rate); + + if (pool_info->n_page_get_delta) { fprintf(file, "Buffer pool hit rate %lu / 1000," " young-making rate %lu / 1000 not %lu / 1000\n", - (ulong) - (1000 - ((1000 * (buf_pool->stat.n_pages_read - - buf_pool->old_stat.n_pages_read)) - / (buf_pool->stat.n_page_gets - - buf_pool->old_stat.n_page_gets))), - (ulong) - (1000 * (buf_pool->stat.n_pages_made_young - - buf_pool->old_stat.n_pages_made_young) - / n_gets_diff), - (ulong) - (1000 * (buf_pool->stat.n_pages_not_made_young - - buf_pool->old_stat.n_pages_not_made_young) - / n_gets_diff)); + (ulong) (1000 - (1000 * pool_info->page_read_delta + / pool_info->n_page_get_delta)), + (ulong) (1000 * pool_info->young_making_delta + / pool_info->n_page_get_delta), + (ulong) (1000 * pool_info->not_young_making_delta + / pool_info->n_page_get_delta)); } else { fputs("No buffer pool page gets since the last printout\n", file); @@ -4891,25 +5044,17 @@ buf_print_io_instance( /* Statistics about read ahead algorithm */ fprintf(file, "Pages read ahead %.2f/s," " evicted without access %.2f/s\n", - (buf_pool->stat.n_ra_pages_read - - buf_pool->old_stat.n_ra_pages_read) - / time_elapsed, - (buf_pool->stat.n_ra_pages_evicted - - buf_pool->old_stat.n_ra_pages_evicted) - / time_elapsed); + pool_info->pages_readahead_rate, + pool_info->pages_evicted_rate); /* Print some values to help us with visualizing what is happening with LRU eviction. */ fprintf(file, "LRU len: %lu, unzip_LRU len: %lu\n" "I/O sum[%lu]:cur[%lu], unzip sum[%lu]:cur[%lu]\n", - UT_LIST_GET_LEN(buf_pool->LRU), - UT_LIST_GET_LEN(buf_pool->unzip_LRU), - buf_LRU_stat_sum.io, buf_LRU_stat_cur.io, - buf_LRU_stat_sum.unzip, buf_LRU_stat_cur.unzip); - - buf_refresh_io_stats(buf_pool); - buf_pool_mutex_exit(buf_pool); + pool_info->lru_len, pool_info->unzip_lru_len, + pool_info->io_sum, pool_info->io_cur, + pool_info->unzip_sum, pool_info->unzip_cur); } /*********************************************************************//** @@ -4920,14 +5065,58 @@ buf_print_io( /*=========*/ FILE* file) /*!< in/out: buffer where to print */ { - ulint i; + ulint i; + buf_pool_info_t* pool_info; + buf_pool_info_t* pool_info_total; + + /* If srv_buf_pool_instances is greater than 1, allocate + one extra buf_pool_info_t, the last one stores + aggregated/total values from all pools */ + if (srv_buf_pool_instances > 1) { + pool_info = (buf_pool_info_t*) mem_zalloc(( + srv_buf_pool_instances + 1) * sizeof *pool_info); + + pool_info_total = &pool_info[srv_buf_pool_instances]; + } else { + ut_a(srv_buf_pool_instances == 1); + pool_info_total = pool_info = (buf_pool_info_t*) mem_zalloc( + sizeof *pool_info) + } for (i = 0; i < srv_buf_pool_instances; i++) { buf_pool_t* buf_pool; buf_pool = buf_pool_from_array(i); - buf_print_io_instance(buf_pool, file); + + /* Fetch individual buffer pool info and calculate + aggregated stats along the way */ + buf_stats_get_pool_info(buf_pool, i, pool_info); + + /* If we have more than one buffer pool, store + the aggregated stats */ + if (srv_buf_pool_instances > 1) { + buf_stats_aggregate_pool_info(pool_info_total, + &pool_info[i]); + } } + + /* Print the aggreate buffer pool info */ + buf_print_io_instance(pool_info_total, file); + + /* If there are more than one buffer pool, print each individual pool + info */ + if (srv_buf_pool_instances > 1) { + fputs("----------------------\n" + "INDIVIDUAL BUFFER POOL INFO\n" + "----------------------\n", file); + + for (i = 0; i < srv_buf_pool_instances; i++) { + fprintf(file, "---BUFFER POOL %lu\n", i); + buf_print_io_instance(&pool_info[i], file); + } + } + + mem_free(pool_info); } /**********************************************************************//** diff --git a/storage/innobase/buf/buf0flu.c b/storage/innobase/buf/buf0flu.c index 5b0617f17b1..07a32e55f97 100644 --- a/storage/innobase/buf/buf0flu.c +++ b/storage/innobase/buf/buf0flu.c @@ -88,6 +88,34 @@ ibool buf_flush_validate_low( /*===================*/ buf_pool_t* buf_pool); /*!< in: Buffer pool instance */ + +/******************************************************************//** +Validates the flush list some of the time. +@return TRUE if ok or the check was skipped */ +static +ibool +buf_flush_validate_skip( +/*====================*/ + buf_pool_t* buf_pool) /*!< in: Buffer pool instance */ +{ +/** Try buf_flush_validate_low() every this many times */ +# define BUF_FLUSH_VALIDATE_SKIP 23 + + /** The buf_flush_validate_low() call skip counter. + Use a signed type because of the race condition below. */ + static int buf_flush_validate_count = BUF_FLUSH_VALIDATE_SKIP; + + /* There is a race condition below, but it does not matter, + because this call is only for heuristic purposes. We want to + reduce the call frequency of the costly buf_flush_validate_low() + check in debug builds. */ + if (--buf_flush_validate_count > 0) { + return(TRUE); + } + + buf_flush_validate_count = BUF_FLUSH_VALIDATE_SKIP; + return(buf_flush_validate_low(buf_pool)); +} #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */ /******************************************************************//** @@ -293,7 +321,7 @@ buf_flush_insert_into_flush_list( } #endif /* UNIV_DEBUG_VALGRIND */ #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG - ut_a(buf_flush_validate_low(buf_pool)); + ut_a(buf_flush_validate_skip(buf_pool)); #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */ buf_flush_list_mutex_exit(buf_pool); @@ -515,7 +543,7 @@ buf_flush_remove( bpage->oldest_modification = 0; #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG - ut_a(buf_flush_validate_low(buf_pool)); + ut_a(buf_flush_validate_skip(buf_pool)); #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */ buf_flush_list_mutex_exit(buf_pool); diff --git a/storage/innobase/buf/buf0lru.c b/storage/innobase/buf/buf0lru.c index 8ad5f2dad83..d819782ac9d 100644 --- a/storage/innobase/buf/buf0lru.c +++ b/storage/innobase/buf/buf0lru.c @@ -2076,6 +2076,7 @@ buf_LRU_stat_update(void) buf_LRU_stat_t* item; buf_pool_t* buf_pool; ibool evict_started = FALSE; + buf_LRU_stat_t cur_stat; /* If we haven't started eviction yet then don't update stats. */ for (i = 0; i < srv_buf_pool_instances; i++) { @@ -2097,12 +2098,19 @@ buf_LRU_stat_update(void) buf_LRU_stat_arr_ind++; buf_LRU_stat_arr_ind %= BUF_LRU_STAT_N_INTERVAL; - /* Add the current value and subtract the obsolete entry. */ - buf_LRU_stat_sum.io += buf_LRU_stat_cur.io - item->io; - buf_LRU_stat_sum.unzip += buf_LRU_stat_cur.unzip - item->unzip; + /* Add the current value and subtract the obsolete entry. + Since buf_LRU_stat_cur is not protected by any mutex, + it can be changing between adding to buf_LRU_stat_sum + and copying to item. Assign it to local variables to make + sure the same value assign to the buf_LRU_stat_sum + and item */ + cur_stat = buf_LRU_stat_cur; + + buf_LRU_stat_sum.io += cur_stat.io - item->io; + buf_LRU_stat_sum.unzip += cur_stat.unzip - item->unzip; /* Put current entry in the array. */ - memcpy(item, &buf_LRU_stat_cur, sizeof *item); + memcpy(item, &cur_stat, sizeof *item); func_exit: /* Clear the current entry. */ diff --git a/storage/innobase/dict/dict0dict.c b/storage/innobase/dict/dict0dict.c index a4da4772526..4b21a51ded9 100644 --- a/storage/innobase/dict/dict0dict.c +++ b/storage/innobase/dict/dict0dict.c @@ -52,8 +52,9 @@ UNIV_INTERN dict_index_t* dict_ind_compact; #include "que0que.h" #include "rem0cmp.h" #include "row0merge.h" +#include "srv0srv.h" /* srv_lower_case_table_names */ #include "m_ctype.h" /* my_isspace() */ -#include "ha_prototypes.h" /* innobase_strcasecmp() */ +#include "ha_prototypes.h" /* innobase_strcasecmp(), innobase_casedn_str()*/ #include <ctype.h> @@ -74,6 +75,7 @@ UNIV_INTERN rw_lock_t dict_operation_lock; #ifdef UNIV_PFS_RWLOCK UNIV_INTERN mysql_pfs_key_t dict_operation_lock_key; UNIV_INTERN mysql_pfs_key_t index_tree_rw_lock_key; +UNIV_INTERN mysql_pfs_key_t dict_table_stats_latch_key; #endif /* UNIV_PFS_RWLOCK */ #ifdef UNIV_PFS_MUTEX @@ -714,7 +716,7 @@ dict_init(void) &dict_foreign_err_mutex, SYNC_ANY_LATCH); for (i = 0; i < DICT_TABLE_STATS_LATCHES_SIZE; i++) { - rw_lock_create(PFS_NOT_INSTRUMENTED, + rw_lock_create(dict_table_stats_latch_key, &dict_table_stats_latches[i], SYNC_INDEX_TREE); } } @@ -1080,13 +1082,13 @@ dict_table_rename_in_cache( /* Allocate a longer name buffer; TODO: store buf len to save memory */ - foreign->foreign_table_name - = mem_heap_alloc(foreign->heap, - ut_strlen(table->name) + 1); + foreign->foreign_table_name = mem_heap_strdup( + foreign->heap, table->name); + dict_mem_foreign_table_name_lookup_set(foreign, TRUE); + } else { + strcpy(foreign->foreign_table_name, table->name); + dict_mem_foreign_table_name_lookup_set(foreign, FALSE); } - - strcpy(foreign->foreign_table_name, table->name); - if (strchr(foreign->id, '/')) { ulint db_len; char* old_id; @@ -1152,12 +1154,14 @@ dict_table_rename_in_cache( /* Allocate a longer name buffer; TODO: store buf len to save memory */ - foreign->referenced_table_name = mem_heap_alloc( - foreign->heap, strlen(table->name) + 1); + foreign->referenced_table_name = mem_heap_strdup( + foreign->heap, table->name); + dict_mem_referenced_table_name_lookup_set(foreign, TRUE); + } else { + /* Use the same buffer */ + strcpy(foreign->referenced_table_name, table->name); + dict_mem_referenced_table_name_lookup_set(foreign, FALSE); } - - strcpy(foreign->referenced_table_name, table->name); - foreign = UT_LIST_GET_NEXT(referenced_list, foreign); } @@ -2583,10 +2587,10 @@ dict_foreign_add_to_cache( ut_ad(mutex_own(&(dict_sys->mutex))); for_table = dict_table_check_if_in_cache_low( - foreign->foreign_table_name); + foreign->foreign_table_name_lookup); ref_table = dict_table_check_if_in_cache_low( - foreign->referenced_table_name); + foreign->referenced_table_name_lookup); ut_a(for_table || ref_table); if (for_table) { @@ -2703,7 +2707,7 @@ dict_scan_to( quote = '\0'; } else if (quote) { /* Within quotes: do nothing. */ - } else if (*ptr == '`' || *ptr == '"') { + } else if (*ptr == '`' || *ptr == '"' || *ptr == '\'') { /* Starting quote: remember the quote character. */ quote = *ptr; } else { @@ -3015,19 +3019,25 @@ dict_scan_table_name( memcpy(ref, database_name, database_name_len); ref[database_name_len] = '/'; memcpy(ref + database_name_len + 1, table_name, table_name_len + 1); -#ifndef __WIN__ - if (srv_lower_case_table_names) { -#endif /* !__WIN__ */ - /* The table name is always put to lower case on Windows. */ + + /* Values; 0 = Store and compare as given; case sensitive + 1 = Store and compare in lower; case insensitive + 2 = Store as given, compare in lower; case semi-sensitive */ + if (srv_lower_case_table_names == 2) { innobase_casedn_str(ref); -#ifndef __WIN__ + *table = dict_table_get_low(ref); + memcpy(ref, database_name, database_name_len); + ref[database_name_len] = '/'; + memcpy(ref + database_name_len + 1, table_name, table_name_len + 1); + } else { + if (srv_lower_case_table_names == 1) { + innobase_casedn_str(ref); + } + *table = dict_table_get_low(ref); } -#endif /* !__WIN__ */ *success = TRUE; *ref_name = ref; - *table = dict_table_get_low(ref); - return(ptr); } @@ -3516,8 +3526,10 @@ col_loop1: } foreign->foreign_table = table; - foreign->foreign_table_name = mem_heap_strdup(foreign->heap, - table->name); + foreign->foreign_table_name = mem_heap_strdup( + foreign->heap, table->name); + dict_mem_foreign_table_name_lookup_set(foreign, TRUE); + foreign->foreign_index = index; foreign->n_fields = (unsigned int) i; foreign->foreign_col_names = mem_heap_alloc(foreign->heap, @@ -3774,8 +3786,9 @@ try_find_index: foreign->referenced_index = index; foreign->referenced_table = referenced_table; - foreign->referenced_table_name - = mem_heap_strdup(foreign->heap, referenced_table_name); + foreign->referenced_table_name = mem_heap_strdup( + foreign->heap, referenced_table_name); + dict_mem_referenced_table_name_lookup_set(foreign, TRUE); foreign->referenced_col_names = mem_heap_alloc(foreign->heap, i * sizeof(void*)); @@ -4586,8 +4599,8 @@ dict_print_info_on_foreign_key_in_create_format( fputs(") REFERENCES ", file); - if (dict_tables_have_same_db(foreign->foreign_table_name, - foreign->referenced_table_name)) { + if (dict_tables_have_same_db(foreign->foreign_table_name_lookup, + foreign->referenced_table_name_lookup)) { /* Do not print the database name of the referenced table */ ut_print_name(file, trx, TRUE, dict_remove_db_name( diff --git a/storage/innobase/dict/dict0load.c b/storage/innobase/dict/dict0load.c index 5b01669af29..3252e85b358 100644 --- a/storage/innobase/dict/dict0load.c +++ b/storage/innobase/dict/dict0load.c @@ -40,6 +40,7 @@ Created 4/24/1996 Heikki Tuuri #include "rem0cmp.h" #include "srv0start.h" #include "srv0srv.h" +#include "ha_prototypes.h" /* innobase_casedn_str() */ /** Following are six InnoDB system tables */ @@ -435,10 +436,12 @@ dict_process_sys_fields_rec( return(err_msg); } + +#ifdef FOREIGN_NOT_USED /********************************************************************//** This function parses a SYS_FOREIGN record and populate a dict_foreign_t structure with the information from the record. For detail information -about SYS_FOREIGN fields, please refer to dict_load_foreign() function +about SYS_FOREIGN fields, please refer to dict_load_foreign() function. @return error message, or NULL on success */ UNIV_INTERN const char* @@ -466,6 +469,11 @@ dict_process_sys_foreign_rec( err_len: return("incorrect column length in SYS_FOREIGN"); } + + /* This recieves a dict_foreign_t* that points to a stack variable. + So mem_heap_free(foreign->heap) is not used as elsewhere. + Since the heap used here is freed elsewhere, foreign->heap + is not assigned. */ foreign->id = mem_heap_strdupl(heap, (const char*) field, len); rec_get_nth_field_offs_old(rec, 1/*DB_TRX_ID*/, &len); @@ -477,6 +485,9 @@ err_len: goto err_len; } + /* The _lookup versions of the referenced and foreign table names + are not assigned since they are not used in this dict_foreign_t */ + field = rec_get_nth_field_old(rec, 3/*FOR_NAME*/, &len); if (UNIV_UNLIKELY(len < 1 || len == UNIV_SQL_NULL)) { goto err_len; @@ -502,6 +513,9 @@ err_len: return(NULL); } +#endif /* FOREIGN_NOT_USED */ + +#ifdef FOREIGN_NOT_USED /********************************************************************//** This function parses a SYS_FOREIGN_COLS record and extract necessary information from the record and return to caller. @@ -565,6 +579,8 @@ err_len: return(NULL); } +#endif /* FOREIGN_NOT_USED */ + /********************************************************************//** Determine the flags of a table described in SYS_TABLES. @return compressed page size in kilobytes; or 0 if the tablespace is @@ -2057,12 +2073,15 @@ dict_load_foreign( foreign->id = mem_heap_strdup(foreign->heap, id); field = rec_get_nth_field_old(rec, 3, &len); + foreign->foreign_table_name = mem_heap_strdupl( foreign->heap, (char*) field, len); + dict_mem_foreign_table_name_lookup_set(foreign, TRUE); field = rec_get_nth_field_old(rec, 4, &len); foreign->referenced_table_name = mem_heap_strdupl( foreign->heap, (char*) field, len); + dict_mem_referenced_table_name_lookup_set(foreign, TRUE); btr_pcur_close(&pcur); mtr_commit(&mtr); @@ -2070,7 +2089,7 @@ dict_load_foreign( dict_load_foreign_cols(id, foreign); ref_table = dict_table_check_if_in_cache_low( - foreign->referenced_table_name); + foreign->referenced_table_name_lookup); /* We could possibly wind up in a deep recursive calls if we call dict_table_get_low() again here if there @@ -2103,7 +2122,7 @@ dict_load_foreign( have to load it so that we are able to make type comparisons in the next function call. */ - for_table = dict_table_get_low(foreign->foreign_table_name); + for_table = dict_table_get_low(foreign->foreign_table_name_lookup); if (for_table && ref_table && check_recursive) { /* This is to record the longest chain of ancesters diff --git a/storage/innobase/dict/dict0mem.c b/storage/innobase/dict/dict0mem.c index bbb8f810f44..8d3d78f3900 100644 --- a/storage/innobase/dict/dict0mem.c +++ b/storage/innobase/dict/dict0mem.c @@ -33,6 +33,8 @@ Created 1/8/1996 Heikki Tuuri #include "data0type.h" #include "mach0data.h" #include "dict0dict.h" +#include "srv0srv.h" /* srv_lower_case_table_names */ +#include "ha_prototypes.h" /* innobase_casedn_str()*/ #ifndef UNIV_HOTBACKUP # include "lock0lock.h" #endif /* !UNIV_HOTBACKUP */ @@ -288,6 +290,60 @@ dict_mem_foreign_create(void) } /**********************************************************************//** +Sets the foreign_table_name_lookup pointer based on the value of +srv_lower_case_table_names. If that is 0 or 1, foreign_table_name_lookup +will point to foreign_table_name. If 2, then another string is allocated +of the heap and set to lower case. */ +UNIV_INTERN +void +dict_mem_foreign_table_name_lookup_set( +/*===================================*/ + dict_foreign_t* foreign, /*!< in/out: foreign struct */ + ibool do_alloc) /*!< in: is an alloc needed */ +{ + if (srv_lower_case_table_names == 2) { + if (do_alloc) { + foreign->foreign_table_name_lookup = mem_heap_alloc( + foreign->heap, + strlen(foreign->foreign_table_name) + 1); + } + strcpy(foreign->foreign_table_name_lookup, + foreign->foreign_table_name); + innobase_casedn_str(foreign->foreign_table_name_lookup); + } else { + foreign->foreign_table_name_lookup + = foreign->foreign_table_name; + } +} + +/**********************************************************************//** +Sets the referenced_table_name_lookup pointer based on the value of +srv_lower_case_table_names. If that is 0 or 1, +referenced_table_name_lookup will point to referenced_table_name. If 2, +then another string is allocated of the heap and set to lower case. */ +UNIV_INTERN +void +dict_mem_referenced_table_name_lookup_set( +/*======================================*/ + dict_foreign_t* foreign, /*!< in/out: foreign struct */ + ibool do_alloc) /*!< in: is an alloc needed */ +{ + if (srv_lower_case_table_names == 2) { + if (do_alloc) { + foreign->referenced_table_name_lookup = mem_heap_alloc( + foreign->heap, + strlen(foreign->referenced_table_name) + 1); + } + strcpy(foreign->referenced_table_name_lookup, + foreign->referenced_table_name); + innobase_casedn_str(foreign->referenced_table_name_lookup); + } else { + foreign->referenced_table_name_lookup + = foreign->referenced_table_name; + } +} + +/**********************************************************************//** Adds a field definition to an index. NOTE: does not take a copy of the column name if the field is a column. The memory occupied by the column name may be released only after publishing the index. */ diff --git a/storage/innobase/fil/fil0fil.c b/storage/innobase/fil/fil0fil.c index cb4ccc005b5..d940bc70609 100644 --- a/storage/innobase/fil/fil0fil.c +++ b/storage/innobase/fil/fil0fil.c @@ -299,6 +299,34 @@ struct fil_system_struct { initialized. */ static fil_system_t* fil_system = NULL; +#ifdef UNIV_DEBUG +/** Try fil_validate() every this many times */ +# define FIL_VALIDATE_SKIP 17 + +/******************************************************************//** +Checks the consistency of the tablespace cache some of the time. +@return TRUE if ok or the check was skipped */ +static +ibool +fil_validate_skip(void) +/*===================*/ +{ + /** The fil_validate() call skip counter. Use a signed type + because of the race condition below. */ + static int fil_validate_count = FIL_VALIDATE_SKIP; + + /* There is a race condition below, but it does not matter, + because this call is only for heuristic purposes. We want to + reduce the call frequency of the costly fil_validate() check + in debug builds. */ + if (--fil_validate_count > 0) { + return(TRUE); + } + + fil_validate_count = FIL_VALIDATE_SKIP; + return(fil_validate()); +} +#endif /* UNIV_DEBUG */ /********************************************************************//** NOTE: you must call fil_mutex_enter_and_prepare_for_io() first! @@ -4307,7 +4335,7 @@ fil_io( #if (1 << UNIV_PAGE_SIZE_SHIFT) != UNIV_PAGE_SIZE # error "(1 << UNIV_PAGE_SIZE_SHIFT) != UNIV_PAGE_SIZE" #endif - ut_ad(fil_validate()); + ut_ad(fil_validate_skip()); #ifndef UNIV_HOTBACKUP # ifndef UNIV_LOG_DEBUG /* ibuf bitmap pages must be read in the sync aio mode: */ @@ -4466,7 +4494,7 @@ fil_io( mutex_exit(&fil_system->mutex); - ut_ad(fil_validate()); + ut_ad(fil_validate_skip()); } return(DB_SUCCESS); @@ -4490,7 +4518,7 @@ fil_aio_wait( void* message; ulint type; - ut_ad(fil_validate()); + ut_ad(fil_validate_skip()); if (srv_use_native_aio) { srv_set_io_thread_op_info(segment, "native aio handle"); @@ -4521,7 +4549,7 @@ fil_aio_wait( mutex_exit(&fil_system->mutex); - ut_ad(fil_validate()); + ut_ad(fil_validate_skip()); /* Do the i/o handling */ /* IMPORTANT: since i/o handling for reads will read also the insert diff --git a/storage/innobase/fsp/fsp0fsp.c b/storage/innobase/fsp/fsp0fsp.c index ffe2b21b80c..5bf5411bc6a 100644 --- a/storage/innobase/fsp/fsp0fsp.c +++ b/storage/innobase/fsp/fsp0fsp.c @@ -3321,7 +3321,7 @@ fseg_free_page_low( "InnoDB: database!\n", (ulong) page); crash: fputs("InnoDB: Please refer to\n" - "InnoDB: " REFMAN "forcing-recovery.html\n" + "InnoDB: " REFMAN "forcing-innodb-recovery.html\n" "InnoDB: about forcing recovery.\n", stderr); ut_error; } diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc index a431c15bc84..7635ea3af07 100644 --- a/storage/innobase/handler/ha_innodb.cc +++ b/storage/innobase/handler/ha_innodb.cc @@ -285,7 +285,8 @@ static PSI_rwlock_info all_innodb_rwlocks[] = { {&checkpoint_lock_key, "checkpoint_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_tree_rw_lock_key, "index_tree_rw_lock", 0}, + {&dict_table_stats_latch_key, "dict_table_stats", 0} }; # endif /* UNIV_PFS_RWLOCK */ @@ -2164,13 +2165,13 @@ innobase_init( ut_a(DATA_MYSQL_TRUE_VARCHAR == (ulint)MYSQL_TYPE_VARCHAR); -#ifdef UNIV_DEBUG +#ifndef DBUG_OFF static const char test_filename[] = "-@"; char test_tablename[sizeof test_filename + sizeof srv_mysql50_table_name_prefix]; if ((sizeof test_tablename) - 1 != filename_to_tablename(test_filename, test_tablename, - sizeof test_tablename) + sizeof test_tablename, true) || strncmp(test_tablename, srv_mysql50_table_name_prefix, sizeof srv_mysql50_table_name_prefix) @@ -2180,7 +2181,7 @@ innobase_init( sql_print_error("tablename encoding has been changed"); goto error; } -#endif /* UNIV_DEBUG */ +#endif /* DBUG_OFF */ /* Check that values don't overflow on 32-bit systems. */ if (sizeof(ulint) == 4) { @@ -3639,6 +3640,7 @@ ha_innobase::open( UT_NOT_USED(test_if_locked); thd = ha_thd(); + srv_lower_case_table_names = lower_case_table_names; /* Under some cases MySQL seems to call this function while holding btr_search_latch. This breaks the latching order as @@ -6509,10 +6511,11 @@ create_clustered_index_when_no_primary( /*****************************************************************//** Return a display name for the row format @return row format name */ - -const char *get_row_format_name( -/*============================*/ -enum row_type row_format) /*!< in: Row Format */ +UNIV_INTERN +const char* +get_row_format_name( +/*================*/ + enum row_type row_format) /*!< in: Row Format */ { switch (row_format) { case ROW_TYPE_COMPACT: @@ -6527,12 +6530,38 @@ enum row_type row_format) /*!< in: Row Format */ return("DEFAULT"); case ROW_TYPE_FIXED: return("FIXED"); - default: + case ROW_TYPE_PAGE: + case ROW_TYPE_NOT_USED: break; } return("NOT USED"); } +/** If file-per-table is missing, issue warning and set ret false */ +#define CHECK_ERROR_ROW_TYPE_NEEDS_FILE_PER_TABLE \ + if (!srv_file_per_table) { \ + push_warning_printf( \ + thd, MYSQL_ERROR::WARN_LEVEL_WARN, \ + ER_ILLEGAL_HA_CREATE_OPTION, \ + "InnoDB: ROW_FORMAT=%s requires" \ + " innodb_file_per_table.", \ + get_row_format_name(row_format)); \ + ret = FALSE; \ + } + +/** If file-format is Antelope, issue warning and set ret false */ +#define CHECK_ERROR_ROW_TYPE_NEEDS_GT_ANTELOPE \ + if (srv_file_format < DICT_TF_FORMAT_ZIP) { \ + push_warning_printf( \ + thd, MYSQL_ERROR::WARN_LEVEL_WARN, \ + ER_ILLEGAL_HA_CREATE_OPTION, \ + "InnoDB: ROW_FORMAT=%s requires" \ + " innodb_file_format > Antelope.", \ + get_row_format_name(row_format)); \ + ret = FALSE; \ + } + + /*****************************************************************//** Validates the create options. We may build on this function in future. For now, it checks two specifiers: @@ -6550,7 +6579,7 @@ create_options_are_valid( { ibool kbs_specified = FALSE; ibool ret = TRUE; - enum row_type row_type = form->s->row_type; + enum row_type row_format = form->s->row_type; ut_ad(thd != NULL); @@ -6559,23 +6588,6 @@ create_options_are_valid( return(TRUE); } - /* Check for a valid Innodb ROW_FORMAT specifier. For example, - ROW_TYPE_FIXED can be sent to Innodb */ - switch (row_type) { - case ROW_TYPE_COMPACT: - case ROW_TYPE_COMPRESSED: - case ROW_TYPE_DYNAMIC: - case ROW_TYPE_REDUNDANT: - case ROW_TYPE_DEFAULT: - break; - default: - push_warning( - thd, MYSQL_ERROR::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: invalid ROW_FORMAT specifier."); - ret = FALSE; - } - ut_ad(form != NULL); ut_ad(create_info != NULL); @@ -6588,7 +6600,23 @@ create_options_are_valid( case 4: case 8: case 16: - /* Valid value. */ + /* Valid KEY_BLOCK_SIZE, check its dependencies. */ + if (!srv_file_per_table) { + push_warning( + thd, MYSQL_ERROR::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: KEY_BLOCK_SIZE requires" + " innodb_file_per_table."); + ret = FALSE; + } + if (srv_file_format < DICT_TF_FORMAT_ZIP) { + push_warning( + thd, MYSQL_ERROR::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, + "InnoDB: KEY_BLOCK_SIZE requires" + " innodb_file_format > Antelope."); + ret = FALSE; + } break; default: push_warning_printf( @@ -6598,72 +6626,43 @@ create_options_are_valid( " Valid values are [1, 2, 4, 8, 16]", create_info->key_block_size); ret = FALSE; + break; } } - /* If KEY_BLOCK_SIZE was specified, check for its - dependencies. */ - if (kbs_specified && !srv_file_per_table) { - push_warning( - thd, MYSQL_ERROR::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: KEY_BLOCK_SIZE requires" - " innodb_file_per_table."); - ret = FALSE; - } - - if (kbs_specified && srv_file_format < DICT_TF_FORMAT_ZIP) { - push_warning( - thd, MYSQL_ERROR::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: KEY_BLOCK_SIZE requires" - " innodb_file_format > Antelope."); - ret = FALSE; - } - - switch (row_type) { + /* Check for a valid Innodb ROW_FORMAT specifier and + other incompatibilities. */ + switch (row_format) { case ROW_TYPE_COMPRESSED: - case ROW_TYPE_DYNAMIC: - /* These two ROW_FORMATs require srv_file_per_table - and srv_file_format > Antelope */ - if (!srv_file_per_table) { - push_warning_printf( - thd, MYSQL_ERROR::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: ROW_FORMAT=%s requires" - " innodb_file_per_table.", - get_row_format_name(row_type)); - ret = FALSE; - } - - if (srv_file_format < DICT_TF_FORMAT_ZIP) { - push_warning_printf( - thd, MYSQL_ERROR::WARN_LEVEL_WARN, - ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: ROW_FORMAT=%s requires" - " innodb_file_format > Antelope.", - get_row_format_name(row_type)); - ret = FALSE; - } - default: + CHECK_ERROR_ROW_TYPE_NEEDS_FILE_PER_TABLE; + CHECK_ERROR_ROW_TYPE_NEEDS_GT_ANTELOPE; break; - } - - switch (row_type) { - case ROW_TYPE_REDUNDANT: - case ROW_TYPE_COMPACT: case ROW_TYPE_DYNAMIC: - /* KEY_BLOCK_SIZE is only allowed with Compressed or Default */ + CHECK_ERROR_ROW_TYPE_NEEDS_FILE_PER_TABLE; + CHECK_ERROR_ROW_TYPE_NEEDS_GT_ANTELOPE; + /* fall through since dynamic also shuns KBS */ + case ROW_TYPE_COMPACT: + case ROW_TYPE_REDUNDANT: if (kbs_specified) { push_warning_printf( thd, MYSQL_ERROR::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: cannot specify ROW_FORMAT = %s" " with KEY_BLOCK_SIZE.", - get_row_format_name(row_type)); - ret = FALSE; + get_row_format_name(row_format)); + ret = FALSE; } - default: + break; + case ROW_TYPE_DEFAULT: + break; + case ROW_TYPE_FIXED: + case ROW_TYPE_PAGE: + case ROW_TYPE_NOT_USED: + push_warning( + thd, MYSQL_ERROR::WARN_LEVEL_WARN, + ER_ILLEGAL_HA_CREATE_OPTION, \ + "InnoDB: invalid ROW_FORMAT specifier."); + ret = FALSE; break; } @@ -6714,7 +6713,7 @@ ha_innobase::create( const ulint file_format = srv_file_format; const char* stmt; size_t stmt_len; - enum row_type row_type; + enum row_type row_format; DBUG_ENTER("ha_innobase::create"); @@ -6764,11 +6763,7 @@ ha_innobase::create( trx = innobase_trx_allocate(thd); - if (lower_case_table_names) { - srv_lower_case_table_names = TRUE; - } else { - srv_lower_case_table_names = FALSE; - } + srv_lower_case_table_names = lower_case_table_names; strcpy(name2, name); @@ -6832,20 +6827,19 @@ ha_innobase::create( push_warning_printf( thd, MYSQL_ERROR::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, - "InnoDB: ignoring" - " KEY_BLOCK_SIZE=%lu.", + "InnoDB: ignoring KEY_BLOCK_SIZE=%lu.", create_info->key_block_size); } } - row_type = form->s->row_type; + row_format = form->s->row_type; if (flags) { /* if ROW_FORMAT is set to default, automatically change it to COMPRESSED.*/ - if (row_type == ROW_TYPE_DEFAULT) { - row_type = ROW_TYPE_COMPRESSED; - } else if (row_type != ROW_TYPE_COMPRESSED) { + if (row_format == ROW_TYPE_DEFAULT) { + row_format = ROW_TYPE_COMPRESSED; + } else if (row_format != ROW_TYPE_COMPRESSED) { /* ROW_FORMAT other than COMPRESSED ignores KEY_BLOCK_SIZE. It does not make sense to reject conflicting @@ -6862,7 +6856,7 @@ ha_innobase::create( } } else { /* flags == 0 means no KEY_BLOCK_SIZE.*/ - if (row_type == ROW_TYPE_COMPRESSED) { + if (row_format == ROW_TYPE_COMPRESSED) { /* ROW_FORMAT=COMPRESSED without KEY_BLOCK_SIZE implies half the maximum KEY_BLOCK_SIZE. */ @@ -6877,7 +6871,7 @@ ha_innobase::create( } } - switch (row_type) { + switch (row_format) { case ROW_TYPE_REDUNDANT: break; case ROW_TYPE_COMPRESSED: @@ -6888,25 +6882,25 @@ ha_innobase::create( ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: ROW_FORMAT=%s requires" " innodb_file_per_table.", - get_row_format_name(row_type)); + get_row_format_name(row_format)); } else if (file_format < DICT_TF_FORMAT_ZIP) { push_warning_printf( thd, MYSQL_ERROR::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, "InnoDB: ROW_FORMAT=%s requires" " innodb_file_format > Antelope.", - get_row_format_name(row_type)); + get_row_format_name(row_format)); } else { flags |= DICT_TF_COMPACT - | (DICT_TF_FORMAT_ZIP - << DICT_TF_FORMAT_SHIFT); + | (DICT_TF_FORMAT_ZIP + << DICT_TF_FORMAT_SHIFT); break; } /* fall through */ case ROW_TYPE_NOT_USED: case ROW_TYPE_FIXED: - default: + case ROW_TYPE_PAGE: push_warning( thd, MYSQL_ERROR::WARN_LEVEL_WARN, ER_ILLEGAL_HA_CREATE_OPTION, @@ -7047,23 +7041,25 @@ ha_innobase::create( 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 TABLE or CREATE INDEX because CREATE INDEX - does a table copy too. */ + this is an ALTER|OPTIMIZE TABLE or CREATE INDEX because CREATE INDEX + does a table copy too. If query was one of : + + CREATE TABLE ...AUTO_INCREMENT = x; or + ALTER TABLE...AUTO_INCREMENT = x; or + OPTIMIZE TABLE t; or + CREATE INDEX x on t(...); + + 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 (((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) { - /* Query was one of : - CREATE TABLE ...AUTO_INCREMENT = x; or - ALTER TABLE...AUTO_INCREMENT = x; or - CREATE INDEX x on t(...); - 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. */ - auto_inc_value = create_info->auto_increment_value; dict_table_autoinc_lock(innobase_table); @@ -7192,11 +7188,7 @@ ha_innobase::delete_table( trx = innobase_trx_allocate(thd); - if (lower_case_table_names) { - srv_lower_case_table_names = TRUE; - } else { - srv_lower_case_table_names = FALSE; - } + srv_lower_case_table_names = lower_case_table_names; name_len = strlen(name); @@ -7319,11 +7311,7 @@ innobase_rename_table( char* norm_to; char* norm_from; - if (lower_case_table_names) { - srv_lower_case_table_names = TRUE; - } else { - srv_lower_case_table_names = FALSE; - } + srv_lower_case_table_names = lower_case_table_names; // Magic number 64 arbitrary norm_to = (char*) my_malloc(strlen(to) + 64, MYF(0)); @@ -7860,14 +7848,14 @@ ha_innobase::info_low( are asked by MySQL to avoid locking. Another reason to avoid the call is that it uses quite a lot of CPU. See Bug#38185. */ - if (flag & HA_STATUS_NO_LOCK || - !(flag & HA_STATUS_VARIABLE_EXTRA)) { + if (flag & HA_STATUS_NO_LOCK + || !(flag & HA_STATUS_VARIABLE_EXTRA)) { /* We do not update delete_length if no locking is requested so the "old" value can remain. delete_length is initialized to 0 in the ha_statistics' constructor. Also we only - need delete_length to be set when - HA_STATUS_VARIABLE_EXTRA is set */ + 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)) { /* Avoid accessing the tablespace if diff --git a/storage/innobase/handler/i_s.cc b/storage/innobase/handler/i_s.cc index 3bcc399a070..00aced180ce 100644 --- a/storage/innobase/handler/i_s.cc +++ b/storage/innobase/handler/i_s.cc @@ -495,8 +495,16 @@ fill_innodb_trx_from_cache( row->trx_mysql_thread_id)); /* trx_query */ - OK(field_store_string(fields[IDX_TRX_QUERY], - row->trx_query)); + if (row->trx_query) { + /* store will do appropriate character set + conversion check */ + fields[IDX_TRX_QUERY]->store( + row->trx_query, strlen(row->trx_query), + row->trx_query_cs); + fields[IDX_TRX_QUERY]->set_notnull(); + } else { + fields[IDX_TRX_QUERY]->set_null(); + } /* trx_operation_state */ OK(field_store_string(fields[IDX_TRX_OPERATION_STATE], diff --git a/storage/innobase/ibuf/ibuf0ibuf.c b/storage/innobase/ibuf/ibuf0ibuf.c index ab42f1ad4f3..0cec0318bf4 100644 --- a/storage/innobase/ibuf/ibuf0ibuf.c +++ b/storage/innobase/ibuf/ibuf0ibuf.c @@ -656,22 +656,49 @@ ibuf_parse_bitmap_init( return(ptr); } #ifndef UNIV_HOTBACKUP +# ifdef UNIV_DEBUG +/** Gets the desired bits for a given page from a bitmap page. +@param page in: bitmap page +@param offset in: page whose bits to get +@param zs in: compressed page size in bytes; 0 for uncompressed pages +@param bit in: IBUF_BITMAP_FREE, IBUF_BITMAP_BUFFERED, ... +@param mtr in: mini-transaction holding an x-latch on the bitmap page +@return value of bits */ +# define ibuf_bitmap_page_get_bits(page, offset, zs, bit, mtr) \ + ibuf_bitmap_page_get_bits_low(page, offset, zs, \ + MTR_MEMO_PAGE_X_FIX, mtr, bit) +# else /* UNIV_DEBUG */ +/** Gets the desired bits for a given page from a bitmap page. +@param page in: bitmap page +@param offset in: page whose bits to get +@param zs in: compressed page size in bytes; 0 for uncompressed pages +@param bit in: IBUF_BITMAP_FREE, IBUF_BITMAP_BUFFERED, ... +@param mtr in: mini-transaction holding an x-latch on the bitmap page +@return value of bits */ +# define ibuf_bitmap_page_get_bits(page, offset, zs, bit, mtr) \ + ibuf_bitmap_page_get_bits_low(page, offset, zs, bit) +# endif /* UNIV_DEBUG */ + /********************************************************************//** Gets the desired bits for a given page from a bitmap page. @return value of bits */ UNIV_INLINE ulint -ibuf_bitmap_page_get_bits( -/*======================*/ +ibuf_bitmap_page_get_bits_low( +/*==========================*/ const page_t* page, /*!< in: bitmap page */ ulint page_no,/*!< in: page whose bits to get */ ulint zip_size,/*!< in: compressed page size in bytes; 0 for uncompressed pages */ - ulint bit, /*!< in: IBUF_BITMAP_FREE, +#ifdef UNIV_DEBUG + ulint latch_type, + /*!< in: MTR_MEMO_PAGE_X_FIX, + MTR_MEMO_BUF_FIX, ... */ + mtr_t* mtr, /*!< in: mini-transaction holding latch_type + on the bitmap page */ +#endif /* UNIV_DEBUG */ + ulint bit) /*!< in: IBUF_BITMAP_FREE, IBUF_BITMAP_BUFFERED, ... */ - mtr_t* mtr __attribute__((unused))) - /*!< in: mtr containing an - x-latch to the bitmap page */ { ulint byte_offset; ulint bit_offset; @@ -683,7 +710,7 @@ ibuf_bitmap_page_get_bits( # error "IBUF_BITS_PER_PAGE % 2 != 0" #endif ut_ad(ut_is_2pow(zip_size)); - ut_ad(mtr_memo_contains_page(mtr, page, MTR_MEMO_PAGE_X_FIX)); + ut_ad(mtr_memo_contains_page(mtr, page, latch_type)); if (!zip_size) { bit_offset = (page_no % UNIV_PAGE_SIZE) * IBUF_BITS_PER_PAGE @@ -1109,21 +1136,29 @@ Must not be called when recv_no_ibuf_operations==TRUE. @return TRUE if level 2 or level 3 page */ UNIV_INTERN ibool -ibuf_page( -/*======*/ - ulint space, /*!< in: space id */ - ulint zip_size,/*!< in: compressed page size in bytes, or 0 */ - ulint page_no,/*!< in: page number */ - mtr_t* mtr) /*!< in: mtr which will contain an x-latch to the - bitmap page if the page is not one of the fixed - address ibuf pages, or NULL, in which case a new - transaction is created. */ +ibuf_page_low( +/*==========*/ + ulint space, /*!< in: space id */ + ulint zip_size,/*!< in: compressed page size in bytes, or 0 */ + ulint page_no,/*!< in: page number */ +#ifdef UNIV_DEBUG + ibool x_latch,/*!< in: FALSE if relaxed check + (avoid latching the bitmap page) */ +#endif /* UNIV_DEBUG */ + const char* file, /*!< in: file name */ + ulint line, /*!< in: line where called */ + mtr_t* mtr) /*!< in: mtr which will contain an + x-latch to the bitmap page if the page + is not one of the fixed address ibuf + pages, or NULL, in which case a new + transaction is created. */ { ibool ret; mtr_t local_mtr; page_t* bitmap_page; ut_ad(!recv_no_ibuf_operations); + ut_ad(x_latch || mtr == NULL); if (ibuf_fixed_addr_page(space, zip_size, page_no)) { @@ -1135,12 +1170,55 @@ ibuf_page( ut_ad(fil_space_get_type(IBUF_SPACE_ID) == FIL_TABLESPACE); +#ifdef UNIV_DEBUG + if (!x_latch) { + mtr_start(&local_mtr); + + /* Get the bitmap page without a page latch, so that + we will not be violating the latching order when + another bitmap page has already been latched by this + thread. The page will be buffer-fixed, and thus it + cannot be removed or relocated while we are looking at + it. The contents of the page could change, but the + IBUF_BITMAP_IBUF bit that we are interested in should + not be modified by any other thread. Nobody should be + calling ibuf_add_free_page() or ibuf_remove_free_page() + while the page is linked to the insert buffer b-tree. */ + + bitmap_page = buf_block_get_frame( + buf_page_get_gen( + space, zip_size, + ibuf_bitmap_page_no_calc(zip_size, page_no), + RW_NO_LATCH, NULL, BUF_GET_NO_LATCH, + file, line, &local_mtr)); +# ifdef UNIV_SYNC_DEBUG + /* This is for tracking Bug #58212. This check and message can + be removed once it has been established that our assumptions + about this condition are correct. The bug was only a one-time + occurrence, unable to repeat since then. */ + void* latch = sync_thread_levels_contains(SYNC_IBUF_BITMAP); + if (latch) { + fprintf(stderr, "Bug#58212 UNIV_SYNC_DEBUG" + " levels %p (%u,%u)\n", + latch, (unsigned) space, (unsigned) page_no); + } +# endif /* UNIV_SYNC_DEBUG */ + ret = ibuf_bitmap_page_get_bits_low( + bitmap_page, page_no, zip_size, + MTR_MEMO_BUF_FIX, &local_mtr, IBUF_BITMAP_IBUF); + + mtr_commit(&local_mtr); + return(ret); + } +#endif /* UNIV_DEBUG */ + if (mtr == NULL) { mtr = &local_mtr; mtr_start(mtr); } - bitmap_page = ibuf_bitmap_get_map_page(space, page_no, zip_size, mtr); + bitmap_page = ibuf_bitmap_get_map_page_func(space, page_no, zip_size, + file, line, mtr); ret = ibuf_bitmap_page_get_bits(bitmap_page, page_no, zip_size, IBUF_BITMAP_IBUF, mtr); diff --git a/storage/innobase/include/btr0cur.h b/storage/innobase/include/btr0cur.h index 9865ebfbfc2..cb65c6bfab7 100644 --- a/storage/innobase/include/btr0cur.h +++ b/storage/innobase/include/btr0cur.h @@ -333,10 +333,14 @@ ulint btr_cur_del_mark_set_clust_rec( /*===========================*/ ulint flags, /*!< in: undo logging and locking flags */ - btr_cur_t* cursor, /*!< in: cursor */ + buf_block_t* block, /*!< in/out: buffer block of the record */ + rec_t* rec, /*!< in/out: record */ + dict_index_t* index, /*!< in: clustered index of the record */ + const ulint* offsets,/*!< in: rec_get_offsets(rec) */ ibool val, /*!< in: value to set */ que_thr_t* thr, /*!< in: query thread */ - mtr_t* mtr); /*!< in: mtr */ + mtr_t* mtr) /*!< in: mtr */ + __attribute__((nonnull)); /***********************************************************//** Sets a secondary index record delete mark to TRUE or FALSE. @return DB_SUCCESS, DB_LOCK_WAIT, or error number */ @@ -469,40 +473,22 @@ btr_estimate_number_of_different_key_vals( /*======================================*/ dict_index_t* index); /*!< in: index */ /*******************************************************************//** -Marks not updated extern fields as not-owned by this record. The ownership -is transferred to the updated record which is inserted elsewhere in the +Marks non-updated off-page fields as disowned by this record. The ownership +must be transferred to the updated record which is inserted elsewhere in the index tree. In purge only the owner of externally stored field is allowed -to free the field. -@return TRUE if BLOB ownership was transferred */ +to free the field. */ UNIV_INTERN -ibool -btr_cur_mark_extern_inherited_fields( -/*=================================*/ +void +btr_cur_disown_inherited_fields( +/*============================*/ page_zip_des_t* page_zip,/*!< in/out: compressed page whose uncompressed part will be updated, or NULL */ rec_t* rec, /*!< in/out: record in a clustered index */ dict_index_t* index, /*!< in: index of the page */ const ulint* offsets,/*!< in: array returned by rec_get_offsets() */ const upd_t* update, /*!< in: update vector */ - mtr_t* mtr); /*!< in: mtr, or NULL if not logged */ -/*******************************************************************//** -The complement of the previous function: in an update entry may inherit -some externally stored fields from a record. We must mark them as inherited -in entry, so that they are not freed in a rollback. */ -UNIV_INTERN -void -btr_cur_mark_dtuple_inherited_extern( -/*=================================*/ - dtuple_t* entry, /*!< in/out: updated entry to be - inserted to clustered index */ - const upd_t* update); /*!< in: update vector */ -/*******************************************************************//** -Marks all extern fields in a dtuple as owned by the record. */ -UNIV_INTERN -void -btr_cur_unmark_dtuple_extern_fields( -/*================================*/ - dtuple_t* entry); /*!< in/out: clustered index entry */ + mtr_t* mtr) /*!< in/out: mini-transaction */ + __attribute__((nonnull(2,3,4,5,6))); /*******************************************************************//** Stores the fields in big_rec_vec to the tablespace and puts pointers to them in rec. The extern flags in rec will have to be set beforehand. diff --git a/storage/innobase/include/buf0buf.h b/storage/innobase/include/buf0buf.h index a42eba57fd2..46cae25f0f5 100644 --- a/storage/innobase/include/buf0buf.h +++ b/storage/innobase/include/buf0buf.h @@ -116,6 +116,76 @@ enum buf_page_state { before putting to the free list */ }; + +/** This structure defines information we will fetch from each buffer pool. It +will be used to print table IO stats */ +struct buf_pool_info_struct{ + /* General buffer pool info */ + ulint pool_unique_id; /*!< Buffer Pool ID */ + ulint pool_size; /*!< Buffer Pool size in pages */ + ulint lru_len; /*!< Length of buf_pool->LRU */ + ulint old_lru_len; /*!< buf_pool->LRU_old_len */ + ulint free_list_len; /*!< Length of buf_pool->free list */ + ulint flush_list_len; /*!< Length of buf_pool->flush_list */ + ulint n_pend_unzip; /*!< buf_pool->n_pend_unzip, pages + pending decompress */ + ulint n_pend_reads; /*!< buf_pool->n_pend_reads, pages + pending read */ + ulint n_pending_flush_lru; /*!< Pages pending flush in LRU */ + ulint n_pending_flush_list; /*!< Pages pending flush in FLUSH + LIST */ + ulint n_pending_flush_single_page;/*!< Pages pending flush in + BUF_FLUSH_SINGLE_PAGE list */ + ulint n_pages_made_young; /*!< number of pages made young */ + ulint n_pages_not_made_young; /*!< number of pages not made young */ + ulint n_pages_read; /*!< buf_pool->n_pages_read */ + ulint n_pages_created; /*!< buf_pool->n_pages_created */ + ulint n_pages_written; /*!< buf_pool->n_pages_written */ + ulint n_page_gets; /*!< buf_pool->n_page_gets */ + ulint n_ra_pages_read; /*!< buf_pool->n_ra_pages_read, number + of pages readahead */ + ulint n_ra_pages_evicted; /*!< buf_pool->n_ra_pages_evicted, + number of readahead pages evicted + without access */ + ulint n_page_get_delta; /*!< num of buffer pool page gets since + last printout */ + + /* Buffer pool access stats */ + double page_made_young_rate; /*!< page made young rate in pages + per second */ + double page_not_made_young_rate;/*!< page not made young rate + in pages per second */ + double pages_read_rate; /*!< num of pages read per second */ + double pages_created_rate; /*!< num of pages create per second */ + double pages_written_rate; /*!< num of pages written per second */ + ulint page_read_delta; /*!< num of pages read since last + printout */ + ulint young_making_delta; /*!< num of pages made young since + last printout */ + ulint not_young_making_delta; /*!< num of pages not make young since + last printout */ + + /* Statistics about read ahead algorithm. */ + double pages_readahead_rate; /*!< readahead rate in pages per + second */ + double pages_evicted_rate; /*!< rate of readahead page evicted + without access, in pages per second */ + + /* Stats about LRU eviction */ + ulint unzip_lru_len; /*!< length of buf_pool->unzip_LRU + list */ + /* Counters for LRU policy */ + ulint io_sum; /*!< buf_LRU_stat_sum.io */ + ulint io_cur; /*!< buf_LRU_stat_cur.io, num of IO + for current interval */ + ulint unzip_sum; /*!< buf_LRU_stat_sum.unzip */ + ulint unzip_cur; /*!< buf_LRU_stat_cur.unzip, num + pages decompressed in current + interval */ +}; + +typedef struct buf_pool_info_struct buf_pool_info_t; + #ifndef UNIV_HOTBACKUP /********************************************************************//** Acquire mutex on all buffer pool instances */ diff --git a/storage/innobase/include/data0data.h b/storage/innobase/include/data0data.h index 11e4027777a..f7bdd29ed90 100644 --- a/storage/innobase/include/data0data.h +++ b/storage/innobase/include/data0data.h @@ -154,14 +154,19 @@ dfield_dup( dfield_t* field, /*!< in/out: data field */ mem_heap_t* heap); /*!< in: memory heap where allocated */ /*********************************************************************//** -Tests if data length and content is equal for two dfields. -@return TRUE if equal */ +Tests if two data fields are equal. +If len==0, tests the data length and content for equality. +If len>0, tests the first len bytes of the content for equality. +@return TRUE if both fields are NULL or if they are equal */ UNIV_INLINE ibool dfield_datas_are_binary_equal( /*==========================*/ const dfield_t* field1, /*!< in: field */ - const dfield_t* field2);/*!< in: field */ + const dfield_t* field2, /*!< in: field */ + ulint len) /*!< in: maximum prefix to compare, + or 0 to compare the whole field length */ + __attribute__((nonnull, warn_unused_result)); /*********************************************************************//** Tests if dfield data length and content is equal to the given. @return TRUE if equal */ diff --git a/storage/innobase/include/data0data.ic b/storage/innobase/include/data0data.ic index 2e3adf4b707..5c0f8039c80 100644 --- a/storage/innobase/include/data0data.ic +++ b/storage/innobase/include/data0data.ic @@ -229,20 +229,30 @@ dfield_dup( } /*********************************************************************//** -Tests if data length and content is equal for two dfields. -@return TRUE if equal */ +Tests if two data fields are equal. +If len==0, tests the data length and content for equality. +If len>0, tests the first len bytes of the content for equality. +@return TRUE if both fields are NULL or if they are equal */ UNIV_INLINE ibool dfield_datas_are_binary_equal( /*==========================*/ const dfield_t* field1, /*!< in: field */ - const dfield_t* field2) /*!< in: field */ + const dfield_t* field2, /*!< in: field */ + ulint len) /*!< in: maximum prefix to compare, + or 0 to compare the whole field length */ { - ulint len; + ulint len2 = len; - len = field1->len; + if (field1->len == UNIV_SQL_NULL || len == 0 || field1->len < len) { + len = field1->len; + } + + if (field2->len == UNIV_SQL_NULL || len2 == 0 || field2->len < len2) { + len2 = field2->len; + } - return(len == field2->len + return(len == len2 && (len == UNIV_SQL_NULL || !memcmp(field1->data, field2->data, len))); } diff --git a/storage/innobase/include/dict0mem.h b/storage/innobase/include/dict0mem.h index d448e57e395..7526e4dc8dc 100644 --- a/storage/innobase/include/dict0mem.h +++ b/storage/innobase/include/dict0mem.h @@ -238,6 +238,26 @@ dict_foreign_t* dict_mem_foreign_create(void); /*=========================*/ +/**********************************************************************//** +Sets the foreign_table_name_lookup pointer based on the value of +srv_lower_case_table_names. */ +UNIV_INTERN +void +dict_mem_foreign_table_name_lookup_set( +/*===================================*/ + dict_foreign_t* foreign, /*!< in/out: foreign struct */ + ibool do_alloc); /*!< in: is an alloc needed */ + +/**********************************************************************//** +Sets the reference_table_name_lookup pointer based on the value of +srv_lower_case_table_names. */ +UNIV_INTERN +void +dict_mem_referenced_table_name_lookup_set( +/*======================================*/ + dict_foreign_t* foreign, /*!< in/out: foreign struct */ + ibool do_alloc); /*!< in: is an alloc needed */ + /** Data structure for a column in a table */ struct dict_col_struct{ /*----------------------*/ @@ -393,10 +413,14 @@ struct dict_foreign_struct{ unsigned type:6; /*!< 0 or DICT_FOREIGN_ON_DELETE_CASCADE or DICT_FOREIGN_ON_DELETE_SET_NULL */ char* foreign_table_name;/*!< foreign table name */ + char* foreign_table_name_lookup; + /*!< foreign table name used for dict lookup */ dict_table_t* foreign_table; /*!< table where the foreign key is */ const char** foreign_col_names;/*!< names of the columns in the foreign key */ char* referenced_table_name;/*!< referenced table name */ + char* referenced_table_name_lookup; + /*!< referenced table name for dict lookup*/ dict_table_t* referenced_table;/*!< table where the referenced key is */ const char** referenced_col_names;/*!< names of the referenced diff --git a/storage/innobase/include/ibuf0ibuf.h b/storage/innobase/include/ibuf0ibuf.h index dd05bcb0608..330efae780c 100644 --- a/storage/innobase/include/ibuf0ibuf.h +++ b/storage/innobase/include/ibuf0ibuf.h @@ -244,15 +244,44 @@ Must not be called when recv_no_ibuf_operations==TRUE. @return TRUE if level 2 or level 3 page */ UNIV_INTERN ibool -ibuf_page( -/*======*/ - ulint space, /*!< in: space id */ - ulint zip_size,/*!< in: compressed page size in bytes, or 0 */ - ulint page_no,/*!< in: page number */ - mtr_t* mtr); /*!< in: mtr which will contain an x-latch to the - bitmap page if the page is not one of the fixed - address ibuf pages, or NULL, in which case a new - transaction is created. */ +ibuf_page_low( +/*==========*/ + ulint space, /*!< in: space id */ + ulint zip_size,/*!< in: compressed page size in bytes, or 0 */ + ulint page_no,/*!< in: page number */ +#ifdef UNIV_DEBUG + ibool x_latch,/*!< in: FALSE if relaxed check + (avoid latching the bitmap page) */ +#endif /* UNIV_DEBUG */ + const char* file, /*!< in: file name */ + ulint line, /*!< in: line where called */ + mtr_t* mtr) /*!< in: mtr which will contain an + x-latch to the bitmap page if the page + is not one of the fixed address ibuf + pages, or NULL, in which case a new + transaction is created. */ + __attribute__((warn_unused_result)); +#ifdef UNIV_DEBUG +/** Checks if a page is a level 2 or 3 page in the ibuf hierarchy of +pages. Must not be called when recv_no_ibuf_operations==TRUE. +@param space tablespace identifier +@param zip_size compressed page size in bytes, or 0 +@param page_no page number +@param mtr mini-transaction or NULL +@return TRUE if level 2 or level 3 page */ +# define ibuf_page(space, zip_size, page_no, mtr) \ + ibuf_page_low(space, zip_size, page_no, TRUE, __FILE__, __LINE__, mtr) +#else /* UVIV_DEBUG */ +/** Checks if a page is a level 2 or 3 page in the ibuf hierarchy of +pages. Must not be called when recv_no_ibuf_operations==TRUE. +@param space tablespace identifier +@param zip_size compressed page size in bytes, or 0 +@param page_no page number +@param mtr mini-transaction or NULL +@return TRUE if level 2 or level 3 page */ +# define ibuf_page(space, zip_size, page_no, mtr) \ + ibuf_page_low(space, zip_size, page_no, __FILE__, __LINE__, mtr) +#endif /* UVIV_DEBUG */ /***********************************************************************//** Frees excess pages from the ibuf free list. This function is called when an OS thread calls fsp services to allocate a new file segment, or a new page to a diff --git a/storage/innobase/include/mem0mem.ic b/storage/innobase/include/mem0mem.ic index cbce2edc661..d214c3fe6c9 100644 --- a/storage/innobase/include/mem0mem.ic +++ b/storage/innobase/include/mem0mem.ic @@ -350,27 +350,27 @@ mem_heap_get_top( ulint n) /*!< in: size of the topmost element */ { mem_block_t* block; - void* buf; + byte* buf; ut_ad(mem_heap_check(heap)); block = UT_LIST_GET_LAST(heap->base); - buf = (byte*)block + mem_block_get_free(block) - MEM_SPACE_NEEDED(n); + buf = (byte*) block + mem_block_get_free(block) - MEM_SPACE_NEEDED(n); #ifdef UNIV_MEM_DEBUG - ut_ad(mem_block_get_start(block) <=(ulint)((byte*)buf - (byte*)block)); + ut_ad(mem_block_get_start(block) <= (ulint) (buf - (byte*) block)); /* In the debug version, advance buf to point at the storage which was given to the caller in the allocation*/ - buf = (byte*)buf + MEM_FIELD_HEADER_SIZE; + buf += MEM_FIELD_HEADER_SIZE; /* Check that the field lengths agree */ - ut_ad(n == (ulint)mem_field_header_get_len(buf)); + ut_ad(n == mem_field_header_get_len(buf)); #endif - return(buf); + return((void*) buf); } /*****************************************************************//** diff --git a/storage/innobase/include/mtr0mtr.ic b/storage/innobase/include/mtr0mtr.ic index 18f8e87b3cf..55f3cf7f147 100644 --- a/storage/innobase/include/mtr0mtr.ic +++ b/storage/innobase/include/mtr0mtr.ic @@ -160,7 +160,7 @@ mtr_memo_contains( while (offset > 0) { offset -= sizeof(mtr_memo_slot_t); - slot = dyn_array_get_element(memo, offset); + slot = (mtr_memo_slot_t*) dyn_array_get_element(memo, offset); if ((object == slot->object) && (type == slot->type)) { diff --git a/storage/innobase/include/os0file.h b/storage/innobase/include/os0file.h index 6d95b280330..fb13120a481 100644 --- a/storage/innobase/include/os0file.h +++ b/storage/innobase/include/os0file.h @@ -373,6 +373,7 @@ typedef HANDLE os_file_dir_t; /*!< directory stream */ typedef DIR* os_file_dir_t; /*!< directory stream */ #endif +#ifdef __WIN__ /***********************************************************************//** Gets the operating system version. Currently works only on Windows. @return OS_WIN95, OS_WIN31, OS_WINNT, OS_WIN2000, OS_WINXP, OS_WINVISTA, @@ -381,6 +382,7 @@ UNIV_INTERN ulint os_get_os_version(void); /*===================*/ +#endif /* __WIN__ */ #ifndef UNIV_HOTBACKUP /****************************************************************//** Creates the seek mutexes used in positioned reads and writes. */ diff --git a/storage/innobase/include/row0upd.h b/storage/innobase/include/row0upd.h index 051beb09277..9c2e003c609 100644 --- a/storage/innobase/include/row0upd.h +++ b/storage/innobase/include/row0upd.h @@ -286,10 +286,13 @@ row_upd_changes_ord_field_binary( row and the data values in update are not known when this function is called, e.g., at compile time */ + const row_ext_t*ext, /*!< NULL, or prefixes of the externally + stored columns in the old row */ dict_index_t* index, /*!< in: index of the record */ - const upd_t* update);/*!< in: update vector for the row; NOTE: the + const upd_t* update) /*!< in: update vector for the row; NOTE: the field numbers in this MUST be clustered index positions! */ + __attribute__((nonnull(3,4), warn_unused_result)); /***********************************************************//** Checks if an update vector changes an ordering field of an index record. This function is fast if the update vector is short or the number of ordering @@ -462,11 +465,16 @@ struct upd_node_struct{ #define UPD_NODE_INSERT_CLUSTERED 3 /* clustered index record should be inserted, old record is already delete marked */ -#define UPD_NODE_UPDATE_ALL_SEC 4 /* an ordering field of the clustered +#define UPD_NODE_INSERT_BLOB 4 /* clustered index record should be + inserted, old record is already + delete-marked; non-updated BLOBs + should be inherited by the new record + and disowned by the old record */ +#define UPD_NODE_UPDATE_ALL_SEC 5 /* an ordering field of the clustered index record was changed, or this is a delete operation: should update all the secondary index records */ -#define UPD_NODE_UPDATE_SOME_SEC 5 /* secondary index entries should be +#define UPD_NODE_UPDATE_SOME_SEC 6 /* secondary index entries should be looked at and updated if an ordering field changed */ diff --git a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h index 98b07f5e893..612ff4c6e54 100644 --- a/storage/innobase/include/srv0srv.h +++ b/storage/innobase/include/srv0srv.h @@ -71,8 +71,8 @@ at a time */ #define SRV_AUTO_EXTEND_INCREMENT \ (srv_auto_extend_increment * ((1024 * 1024) / UNIV_PAGE_SIZE)) -/* This is set to TRUE if the MySQL user has set it in MySQL */ -extern ibool srv_lower_case_table_names; +/* This is set to the MySQL server value for this variable. */ +extern uint srv_lower_case_table_names; /* Mutex for locking srv_monitor_file */ extern mutex_t srv_monitor_file_mutex; @@ -476,6 +476,14 @@ enum srv_thread_type srv_get_thread_type(void); /*=====================*/ /*********************************************************************//** +Check whether thread type has reserved a slot. +@return slot number or UNDEFINED if not found*/ +UNIV_INTERN +ulint +srv_thread_has_reserved_slot( +/*=========================*/ + enum srv_thread_type type); /*!< in: thread type to check */ +/*********************************************************************//** Sets the info describing an i/o thread current state. */ UNIV_INTERN void diff --git a/storage/innobase/include/sync0rw.h b/storage/innobase/include/sync0rw.h index 70471186f6d..5988fdfb382 100644 --- a/storage/innobase/include/sync0rw.h +++ b/storage/innobase/include/sync0rw.h @@ -122,6 +122,7 @@ extern mysql_pfs_key_t checkpoint_lock_key; extern mysql_pfs_key_t trx_i_s_cache_lock_key; extern mysql_pfs_key_t trx_purge_latch_key; extern mysql_pfs_key_t index_tree_rw_lock_key; +extern mysql_pfs_key_t dict_table_stats_latch_key; #endif /* UNIV_PFS_RWLOCK */ diff --git a/storage/innobase/include/trx0i_s.h b/storage/innobase/include/trx0i_s.h index 8f8b7367fb0..73896a3cb76 100644 --- a/storage/innobase/include/trx0i_s.h +++ b/storage/innobase/include/trx0i_s.h @@ -140,6 +140,9 @@ struct i_s_trx_row_struct { ulint trx_mysql_thread_id; /*!< thd_get_thread_id() */ const char* trx_query; /*!< MySQL statement being executed in the transaction */ + struct charset_info_st* trx_query_cs; + /*!< charset encode the MySQL + statement */ const char* trx_operation_state; /*!< trx_struct::op_info */ ulint trx_tables_in_use;/*!< n_mysql_tables_in_use in trx_struct */ diff --git a/storage/innobase/include/univ.i b/storage/innobase/include/univ.i index 058277e6e61..4ca02d30db1 100644 --- a/storage/innobase/include/univ.i +++ b/storage/innobase/include/univ.i @@ -44,9 +44,14 @@ Created 1/20/1994 Heikki Tuuri #include "hb_univ.i" #endif /* UNIV_HOTBACKUP */ +/* aux macros to convert M into "123" (string) if M is defined like +#define M 123 */ +#define _IB_TO_STR(s) #s +#define IB_TO_STR(s) _IB_TO_STR(s) + #define INNODB_VERSION_MAJOR 1 #define INNODB_VERSION_MINOR 1 -#define INNODB_VERSION_BUGFIX 4 +#define INNODB_VERSION_BUGFIX 5 /* The following is the InnoDB version as shown in SELECT plugin_version FROM information_schema.plugins; @@ -57,16 +62,14 @@ component, i.e. we show M.N.P as M.N */ #define INNODB_VERSION_SHORT \ (INNODB_VERSION_MAJOR << 8 | INNODB_VERSION_MINOR) -/* auxiliary macros to help creating the version as string */ -#define __INNODB_VERSION(a, b, c) (#a "." #b "." #c) -#define _INNODB_VERSION(a, b, c) __INNODB_VERSION(a, b, c) - #define INNODB_VERSION_STR \ - _INNODB_VERSION(INNODB_VERSION_MAJOR, \ - INNODB_VERSION_MINOR, \ - INNODB_VERSION_BUGFIX) + IB_TO_STR(INNODB_VERSION_MAJOR) "." \ + IB_TO_STR(INNODB_VERSION_MINOR) "." \ + IB_TO_STR(INNODB_VERSION_BUGFIX) -#define REFMAN "http://dev.mysql.com/doc/refman/5.1/en/" +#define REFMAN "http://dev.mysql.com/doc/refman/" \ + IB_TO_STR(MYSQL_MAJOR_VERSION) "." \ + IB_TO_STR(MYSQL_MINOR_VERSION) "/en/" #ifdef MYSQL_DYNAMIC_PLUGIN /* In the dynamic plugin, redefine some externally visible symbols @@ -250,7 +253,7 @@ easy way to get it to work. See http://bugs.mysql.com/bug.php?id=52263. */ # define UNIV_INTERN #endif -#if (!defined(UNIV_DEBUG) && !defined(UNIV_MUST_NOT_INLINE)) +#ifndef UNIV_MUST_NOT_INLINE /* Definition for inline version */ #ifdef __WIN__ diff --git a/storage/innobase/include/ut0ut.h b/storage/innobase/include/ut0ut.h index dd59b3eba46..cd5c7ca99f1 100644 --- a/storage/innobase/include/ut0ut.h +++ b/storage/innobase/include/ut0ut.h @@ -55,24 +55,24 @@ Created 1/20/1994 Heikki Tuuri typedef time_t ib_time_t; #ifndef UNIV_HOTBACKUP -#if defined(HAVE_IB_PAUSE_INSTRUCTION) -# ifdef WIN32 - /* In the Win32 API, the x86 PAUSE instruction is executed by calling - the YieldProcessor macro defined in WinNT.h. It is a CPU architecture- - independent way by using YieldProcessor.*/ -# define UT_RELAX_CPU() YieldProcessor() -# else - /* According to the gcc info page, asm volatile means that the - instruction has important side-effects and must not be removed. - Also asm volatile may trigger a memory barrier (spilling all registers - to memory). */ -# define UT_RELAX_CPU() __asm__ __volatile__ ("pause") -# endif +#if defined(HAVE_PAUSE_INSTRUCTION) + /* According to the gcc info page, asm volatile means that the + instruction has important side-effects and must not be removed. + Also asm volatile may trigger a memory barrier (spilling all registers + to memory). */ +# define UT_RELAX_CPU() __asm__ __volatile__ ("pause") +#elif defined(HAVE_FAKE_PAUSE_INSTRUCTION) +# define UT_RELAX_CPU() __asm__ __volatile__ ("rep; nop") #elif defined(HAVE_ATOMIC_BUILTINS) # define UT_RELAX_CPU() do { \ volatile lint volatile_var; \ os_compare_and_swap_lint(&volatile_var, 0, 1); \ } while (0) +#elif defined(HAVE_WINDOWS_ATOMICS) + /* In the Win32 API, the x86 PAUSE instruction is executed by calling + the YieldProcessor macro defined in WinNT.h. It is a CPU architecture- + independent way by using YieldProcessor. */ +# define UT_RELAX_CPU() YieldProcessor() #else # define UT_RELAX_CPU() ((void)0) /* avoid warning for an empty statement */ #endif diff --git a/storage/innobase/log/log0recv.c b/storage/innobase/log/log0recv.c index 5d5caf1b48a..f9e0fecb6c6 100644 --- a/storage/innobase/log/log0recv.c +++ b/storage/innobase/log/log0recv.c @@ -2199,7 +2199,7 @@ recv_report_corrupt_log( "InnoDB: far enough in recovery! Please run CHECK TABLE\n" "InnoDB: on your InnoDB tables to check that they are ok!\n" "InnoDB: If mysqld crashes after this recovery, look at\n" - "InnoDB: " REFMAN "forcing-recovery.html\n" + "InnoDB: " REFMAN "forcing-innodb-recovery.html\n" "InnoDB: about forcing recovery.\n", stderr); fflush(stderr); diff --git a/storage/innobase/os/os0file.c b/storage/innobase/os/os0file.c index 93d2f72746d..74dbac3bc96 100644 --- a/storage/innobase/os/os0file.c +++ b/storage/innobase/os/os0file.c @@ -302,6 +302,36 @@ UNIV_INTERN ulint os_n_pending_writes = 0; /** Number of pending read operations */ UNIV_INTERN ulint os_n_pending_reads = 0; +#ifdef UNIV_DEBUG +/**********************************************************************//** +Validates the consistency the aio system some of the time. +@return TRUE if ok or the check was skipped */ +UNIV_INTERN +ibool +os_aio_validate_skip(void) +/*======================*/ +{ +/** Try os_aio_validate() every this many times */ +# define OS_AIO_VALIDATE_SKIP 13 + + /** The os_aio_validate() call skip counter. + Use a signed type because of the race condition below. */ + static int os_aio_validate_count = OS_AIO_VALIDATE_SKIP; + + /* There is a race condition below, but it does not matter, + because this call is only for heuristic purposes. We want to + reduce the call frequency of the costly os_aio_validate() + check in debug builds. */ + if (--os_aio_validate_count > 0) { + return(TRUE); + } + + os_aio_validate_count = OS_AIO_VALIDATE_SKIP; + return(os_aio_validate()); +} +#endif /* UNIV_DEBUG */ + +#ifdef __WIN__ /***********************************************************************//** Gets the operating system version. Currently works only on Windows. @return OS_WIN95, OS_WIN31, OS_WINNT, OS_WIN2000, OS_WINXP, OS_WINVISTA, @@ -311,7 +341,6 @@ ulint os_get_os_version(void) /*===================*/ { -#ifdef __WIN__ OSVERSIONINFO os_info; os_info.dwOSVersionInfoSize = sizeof(OSVERSIONINFO); @@ -340,12 +369,8 @@ os_get_os_version(void) ut_error; return(0); } -#else - ut_error; - - return(0); -#endif } +#endif /* __WIN__ */ /***********************************************************************//** Retrieves the last error number if an error occurs in a file io function. @@ -4008,7 +4033,7 @@ os_aio_func( ut_ad(n > 0); ut_ad(n % OS_FILE_LOG_BLOCK_SIZE == 0); ut_ad(offset % OS_FILE_LOG_BLOCK_SIZE == 0); - ut_ad(os_aio_validate()); + ut_ad(os_aio_validate_skip()); #ifdef WIN_ASYNC_IO ut_ad((n & 0xFFFFFFFFUL) == n); #endif @@ -4210,7 +4235,7 @@ os_aio_windows_handle( /* NOTE! We only access constant fields in os_aio_array. Therefore we do not have to acquire the protecting mutex yet */ - ut_ad(os_aio_validate()); + ut_ad(os_aio_validate_skip()); ut_ad(segment < array->n_segments); n = array->n_slots / array->n_segments; @@ -4630,7 +4655,7 @@ restart: srv_set_io_thread_op_info(global_segment, "looking for i/o requests (a)"); - ut_ad(os_aio_validate()); + ut_ad(os_aio_validate_skip()); ut_ad(segment < array->n_segments); n = array->n_slots / array->n_segments; diff --git a/storage/innobase/row/row0ins.c b/storage/innobase/row/row0ins.c index 09ae80adff4..dbfba358f9f 100644 --- a/storage/innobase/row/row0ins.c +++ b/storage/innobase/row/row0ins.c @@ -1525,7 +1525,7 @@ row_ins_check_foreign_constraints( if (foreign->foreign_index == index) { if (foreign->referenced_table == NULL) { - dict_table_get(foreign->referenced_table_name, + dict_table_get(foreign->referenced_table_name_lookup, FALSE); } diff --git a/storage/innobase/row/row0mysql.c b/storage/innobase/row/row0mysql.c index 709750fc4e7..63f30276df6 100644 --- a/storage/innobase/row/row0mysql.c +++ b/storage/innobase/row/row0mysql.c @@ -632,7 +632,7 @@ handle_new_error: "InnoDB: If the mysqld server crashes" " after the startup or when\n" "InnoDB: you dump the tables, look at\n" - "InnoDB: " REFMAN "forcing-recovery.html" + "InnoDB: " REFMAN "forcing-innodb-recovery.html" " for help.\n", stderr); break; case DB_FOREIGN_EXCEED_MAX_CASCADE: @@ -3163,7 +3163,7 @@ check_next_foreign: if (foreign && trx->check_foreigns && !(drop_db && dict_tables_have_same_db( - name, foreign->foreign_table_name))) { + name, foreign->foreign_table_name_lookup))) { FILE* ef = dict_foreign_err_file; /* We only allow dropping a referenced table if diff --git a/storage/innobase/row/row0purge.c b/storage/innobase/row/row0purge.c index 031b5258e98..08abc8a3e59 100644 --- a/storage/innobase/row/row0purge.c +++ b/storage/innobase/row/row0purge.c @@ -515,7 +515,7 @@ row_purge_upd_exist_or_extern( while (node->index != NULL) { index = node->index; - if (row_upd_changes_ord_field_binary(NULL, node->index, + if (row_upd_changes_ord_field_binary(NULL, NULL, node->index, node->update)) { /* Build the older version of the index entry */ entry = row_build_index_entry(node->row, NULL, diff --git a/storage/innobase/row/row0umod.c b/storage/innobase/row/row0umod.c index aef653b3150..0d4c42b958c 100644 --- a/storage/innobase/row/row0umod.c +++ b/storage/innobase/row/row0umod.c @@ -693,19 +693,18 @@ row_undo_mod_upd_exist_sec( while (node->index != NULL) { index = node->index; - if (row_upd_changes_ord_field_binary(node->row, node->index, - node->update)) { + if (row_upd_changes_ord_field_binary( + node->row, node->ext, node->index, node->update)) { /* Build the newest version of the index entry */ entry = row_build_index_entry(node->row, node->ext, index, heap); if (UNIV_UNLIKELY(!entry)) { /* The server must have crashed in - row_upd_clust_rec_by_insert(), in - row_ins_index_entry_low() before - btr_store_big_rec_extern_fields() - has written the externally stored columns - (BLOBs) of the new clustered index entry. */ + row_upd_clust_rec_by_insert() before + the updated externally stored columns (BLOBs) + of the new clustered index entry were + written. */ /* The table must be in DYNAMIC or COMPRESSED format. REDUNDANT and COMPACT formats diff --git a/storage/innobase/row/row0upd.c b/storage/innobase/row/row0upd.c index 81750938f58..03ed40c6647 100644 --- a/storage/innobase/row/row0upd.c +++ b/storage/innobase/row/row0upd.c @@ -238,7 +238,7 @@ row_upd_check_references_constraints( foreign->n_fields))) { if (foreign->foreign_table == NULL) { - dict_table_get(foreign->foreign_table_name, + dict_table_get(foreign->foreign_table_name_lookup, FALSE); } @@ -1198,20 +1198,21 @@ row_upd_changes_ord_field_binary( row and the data values in update are not known when this function is called, e.g., at compile time */ + const row_ext_t*ext, /*!< NULL, or prefixes of the externally + stored columns in the old row */ dict_index_t* index, /*!< in: index of the record */ const upd_t* update) /*!< in: update vector for the row; NOTE: the field numbers in this MUST be clustered index positions! */ { - ulint n_unique; - ulint n_upd_fields; - ulint i, j; - dict_index_t* clust_index; + ulint n_unique; + ulint i; + const dict_index_t* clust_index; - ut_ad(update && index); + ut_ad(update); + ut_ad(index); n_unique = dict_index_get_n_unique(index); - n_upd_fields = upd_get_n_fields(update); clust_index = dict_table_get_first_index(index->table); @@ -1219,33 +1220,72 @@ row_upd_changes_ord_field_binary( const dict_field_t* ind_field; const dict_col_t* col; - ulint col_pos; ulint col_no; + const upd_field_t* upd_field; + const dfield_t* dfield; + dfield_t dfield_ext; + ulint dfield_len; + const byte* buf; ind_field = dict_index_get_nth_field(index, i); col = dict_field_get_col(ind_field); - col_pos = dict_col_get_clust_pos(col, clust_index); col_no = dict_col_get_no(col); - for (j = 0; j < n_upd_fields; j++) { + upd_field = upd_get_field_by_field_no( + update, dict_col_get_clust_pos(col, clust_index)); - const upd_field_t* upd_field - = upd_get_nth_field(update, j); + if (upd_field == NULL) { + continue; + } - /* Note that if the index field is a column prefix - then it may be that row does not contain an externally - stored part of the column value, and we cannot compare - the datas */ + if (row == NULL) { + ut_ad(ext == NULL); + return(TRUE); + } - if (col_pos == upd_field->field_no - && (row == NULL - || ind_field->prefix_len > 0 - || !dfield_datas_are_binary_equal( - dtuple_get_nth_field(row, col_no), - &(upd_field->new_val)))) { + dfield = dtuple_get_nth_field(row, col_no); - return(TRUE); + /* This treatment of column prefix indexes is loosely + based on row_build_index_entry(). */ + + if (UNIV_LIKELY(ind_field->prefix_len == 0) + || dfield_is_null(dfield)) { + /* do nothing special */ + } else if (UNIV_LIKELY_NULL(ext)) { + /* See if the column is stored externally. */ + buf = row_ext_lookup(ext, col_no, &dfield_len); + + ut_ad(col->ord_part); + + if (UNIV_LIKELY_NULL(buf)) { + if (UNIV_UNLIKELY(buf == field_ref_zero)) { + /* This should never happen, but + we try to fail safe here. */ + ut_ad(0); + return(TRUE); + } + + goto copy_dfield; } + } else if (dfield_is_ext(dfield)) { + dfield_len = dfield_get_len(dfield); + ut_a(dfield_len > BTR_EXTERN_FIELD_REF_SIZE); + dfield_len -= BTR_EXTERN_FIELD_REF_SIZE; + ut_a(dict_index_is_clust(index) + || ind_field->prefix_len <= dfield_len); + buf = dfield_get_data(dfield); +copy_dfield: + ut_a(dfield_len > 0); + dfield_copy(&dfield_ext, dfield); + dfield_set_data(&dfield_ext, buf, dfield_len); + dfield = &dfield_ext; + } + + if (!dfield_datas_are_binary_equal( + dfield, &upd_field->new_val, + ind_field->prefix_len)) { + + return(TRUE); } } @@ -1329,7 +1369,7 @@ row_upd_changes_first_fields_binary( if (col_pos == upd_field->field_no && !dfield_datas_are_binary_equal( dtuple_get_nth_field(entry, i), - &(upd_field->new_val))) { + &upd_field->new_val, 0)) { return(TRUE); } @@ -1596,14 +1636,99 @@ row_upd_sec_step( ut_ad(!dict_index_is_clust(node->index)); if (node->state == UPD_NODE_UPDATE_ALL_SEC - || row_upd_changes_ord_field_binary(node->row, node->index, - node->update)) { + || row_upd_changes_ord_field_binary(node->row, node->ext, + node->index, node->update)) { return(row_upd_sec_index_entry(node, thr)); } return(DB_SUCCESS); } +#ifdef UNIV_DEBUG +# define row_upd_clust_rec_by_insert_inherit(rec,offsets,entry,update) \ + row_upd_clust_rec_by_insert_inherit_func(rec,offsets,entry,update) +#else /* UNIV_DEBUG */ +# define row_upd_clust_rec_by_insert_inherit(rec,offsets,entry,update) \ + row_upd_clust_rec_by_insert_inherit_func(entry,update) +#endif /* UNIV_DEBUG */ +/*******************************************************************//** +Mark non-updated off-page columns inherited when the primary key is +updated. We must mark them as inherited in entry, so that they are not +freed in a rollback. A limited version of this function used to be +called btr_cur_mark_dtuple_inherited_extern(). +@return TRUE if any columns were inherited */ +static __attribute__((warn_unused_result)) +ibool +row_upd_clust_rec_by_insert_inherit_func( +/*=====================================*/ +#ifdef UNIV_DEBUG + const rec_t* rec, /*!< in: old record, or NULL */ + const ulint* offsets,/*!< in: rec_get_offsets(rec), or NULL */ +#endif /* UNIV_DEBUG */ + dtuple_t* entry, /*!< in/out: updated entry to be + inserted into the clustered index */ + const upd_t* update) /*!< in: update vector */ +{ + ibool inherit = FALSE; + ulint i; + + ut_ad(!rec == !offsets); + ut_ad(!rec || rec_offs_any_extern(offsets)); + + for (i = 0; i < dtuple_get_n_fields(entry); i++) { + dfield_t* dfield = dtuple_get_nth_field(entry, i); + byte* data; + ulint len; + + ut_ad(!offsets + || !rec_offs_nth_extern(offsets, i) + == !dfield_is_ext(dfield) + || upd_get_field_by_field_no(update, i)); + if (!dfield_is_ext(dfield) + || upd_get_field_by_field_no(update, i)) { + continue; + } + +#ifdef UNIV_DEBUG + if (UNIV_LIKELY(rec != NULL)) { + const byte* rec_data + = rec_get_nth_field(rec, offsets, i, &len); + ut_ad(len == dfield_get_len(dfield)); + ut_ad(len != UNIV_SQL_NULL); + ut_ad(len >= BTR_EXTERN_FIELD_REF_SIZE); + + rec_data += len - BTR_EXTERN_FIELD_REF_SIZE; + + /* The pointer must not be zero. */ + ut_ad(memcmp(rec_data, field_ref_zero, + BTR_EXTERN_FIELD_REF_SIZE)); + /* The BLOB must be owned. */ + ut_ad(!(rec_data[BTR_EXTERN_LEN] + & BTR_EXTERN_OWNER_FLAG)); + } +#endif /* UNIV_DEBUG */ + + len = dfield_get_len(dfield); + ut_a(len != UNIV_SQL_NULL); + ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE); + data = dfield_get_data(dfield); + data += len - BTR_EXTERN_FIELD_REF_SIZE; + /* The pointer must not be zero. */ + ut_a(memcmp(data, field_ref_zero, BTR_EXTERN_FIELD_REF_SIZE)); + /* The BLOB must be owned. */ + ut_a(!(data[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG)); + + data[BTR_EXTERN_LEN] |= BTR_EXTERN_INHERITED_FLAG; + /* The BTR_EXTERN_INHERITED_FLAG only matters in + rollback. Purge will always free the extern fields of + a delete-marked row. */ + + inherit = TRUE; + } + + return(inherit); +} + /***********************************************************//** Marks the clustered index record deleted and inserts the updated version of the record to the index. This function should be used when the ordering @@ -1622,14 +1747,16 @@ row_upd_clust_rec_by_insert( a foreign key constraint */ mtr_t* mtr) /*!< in/out: mtr; gets committed here */ { - mem_heap_t* heap = NULL; + mem_heap_t* heap; btr_pcur_t* pcur; btr_cur_t* btr_cur; trx_t* trx; dict_table_t* table; dtuple_t* entry; ulint err; - ibool change_ownership = FALSE; + ibool change_ownership = FALSE; + rec_t* rec; + ulint* offsets = NULL; ut_ad(node); ut_ad(dict_index_is_clust(index)); @@ -1639,34 +1766,64 @@ row_upd_clust_rec_by_insert( pcur = node->pcur; btr_cur = btr_pcur_get_btr_cur(pcur); - if (node->state != UPD_NODE_INSERT_CLUSTERED) { - rec_t* rec; - dict_index_t* index; - ulint offsets_[REC_OFFS_NORMAL_SIZE]; - ulint* offsets; - rec_offs_init(offsets_); + heap = mem_heap_create(1000); + + entry = row_build_index_entry(node->upd_row, node->upd_ext, + index, heap); + ut_a(entry); - err = btr_cur_del_mark_set_clust_rec(BTR_NO_LOCKING_FLAG, - btr_cur, TRUE, thr, mtr); + row_upd_index_entry_sys_field(entry, index, DATA_TRX_ID, trx->id); + + switch (node->state) { + default: + ut_error; + case UPD_NODE_INSERT_BLOB: + /* A lock wait occurred in row_ins_index_entry() in + the previous invocation of this function. Mark the + off-page columns in the entry inherited. */ + + change_ownership = row_upd_clust_rec_by_insert_inherit( + NULL, NULL, entry, node->update); + ut_a(change_ownership); + /* fall through */ + case UPD_NODE_INSERT_CLUSTERED: + /* A lock wait occurred in row_ins_index_entry() in + the previous invocation of this function. */ + break; + case UPD_NODE_UPDATE_CLUSTERED: + /* This is the first invocation of the function where + we update the primary key. Delete-mark the old record + in the clustered index and prepare to insert a new entry. */ + rec = btr_cur_get_rec(btr_cur); + offsets = rec_get_offsets(rec, index, NULL, + ULINT_UNDEFINED, &heap); + ut_ad(page_rec_is_user_rec(rec)); + + err = btr_cur_del_mark_set_clust_rec( + BTR_NO_LOCKING_FLAG, btr_cur_get_block(btr_cur), + rec, index, offsets, TRUE, thr, mtr); if (err != DB_SUCCESS) { +err_exit: mtr_commit(mtr); + mem_heap_free(heap); return(err); } - /* Mark as not-owned the externally stored fields which the new - row inherits from the delete marked record: purge should not - free those externally stored fields even if the delete marked - record is removed from the index tree, or updated. */ + /* If the the new row inherits externally stored + fields (off-page columns a.k.a. BLOBs) from the + delete-marked old record, mark them disowned by the + old record and owned by the new entry. */ - rec = btr_cur_get_rec(btr_cur); - index = dict_table_get_first_index(table); - offsets = rec_get_offsets(rec, index, offsets_, - ULINT_UNDEFINED, &heap); - change_ownership = btr_cur_mark_extern_inherited_fields( - btr_cur_get_page_zip(btr_cur), - rec, index, offsets, node->update, mtr); - if (referenced) { + if (rec_offs_any_extern(offsets)) { + change_ownership = row_upd_clust_rec_by_insert_inherit( + rec, offsets, entry, node->update); + + if (change_ownership) { + btr_pcur_store_position(pcur, mtr); + } + } + if (referenced) { /* NOTE that the following call loses the position of pcur ! */ @@ -1674,48 +1831,49 @@ row_upd_clust_rec_by_insert( node, pcur, table, index, offsets, thr, mtr); if (err != DB_SUCCESS) { - - mtr_commit(mtr); - - if (UNIV_LIKELY_NULL(heap)) { - mem_heap_free(heap); - } - - return(err); + goto err_exit; } } } mtr_commit(mtr); - if (!heap) { - heap = mem_heap_create(500); - } - node->state = UPD_NODE_INSERT_CLUSTERED; + err = row_ins_index_entry(index, entry, + node->upd_ext ? node->upd_ext->n_ext : 0, + TRUE, thr); + node->state = change_ownership + ? UPD_NODE_INSERT_BLOB + : UPD_NODE_INSERT_CLUSTERED; - entry = row_build_index_entry(node->upd_row, node->upd_ext, - index, heap); - ut_a(entry); + if (err == DB_SUCCESS && change_ownership) { + /* Mark the non-updated fields disowned by the old record. */ - row_upd_index_entry_sys_field(entry, index, DATA_TRX_ID, trx->id); + /* NOTE: this transaction has an x-lock on the record + and therefore other transactions cannot modify the + record when we have no latch on the page. In addition, + we assume that other query threads of the same + transaction do not modify the record in the meantime. + Therefore we can assert that the restoration of the + cursor succeeds. */ - if (change_ownership) { - /* If we return from a lock wait, for example, we may have - extern fields marked as not-owned in entry (marked in the - if-branch above). We must unmark them, take the ownership - back. */ + mtr_start(mtr); + + if (!btr_pcur_restore_position(BTR_MODIFY_LEAF, pcur, mtr)) { + ut_error; + } - btr_cur_unmark_dtuple_extern_fields(entry); + rec = btr_cur_get_rec(btr_cur); + offsets = rec_get_offsets(rec, index, offsets, + ULINT_UNDEFINED, &heap); + ut_ad(page_rec_is_user_rec(rec)); - /* We must mark non-updated extern fields in entry as - inherited, so that a possible rollback will not free them. */ + btr_cur_disown_inherited_fields( + btr_cur_get_page_zip(btr_cur), + rec, index, offsets, node->update, mtr); - btr_cur_mark_dtuple_inherited_extern(entry, node->update); + mtr_commit(mtr); } - err = row_ins_index_entry(index, entry, - node->upd_ext ? node->upd_ext->n_ext : 0, - TRUE, thr); mem_heap_free(heap); return(err); @@ -1860,8 +2018,9 @@ row_upd_del_mark_clust_rec( /* Mark the clustered index record deleted; we do not have to check locks, because we assume that we have an x-lock on the record */ - err = btr_cur_del_mark_set_clust_rec(BTR_NO_LOCKING_FLAG, - btr_cur, TRUE, thr, mtr); + err = btr_cur_del_mark_set_clust_rec( + BTR_NO_LOCKING_FLAG, btr_cur_get_block(btr_cur), + btr_cur_get_rec(btr_cur), index, offsets, TRUE, thr, mtr); if (err == DB_SUCCESS && referenced) { /* NOTE that the following call loses the position of pcur ! */ @@ -2006,7 +2165,8 @@ exit_func: row_upd_store_row(node); - if (row_upd_changes_ord_field_binary(node->row, index, node->update)) { + if (row_upd_changes_ord_field_binary(node->row, node->ext, index, + node->update)) { /* Update causes an ordering field (ordering fields within the B-tree) of the clustered index record to change: perform @@ -2076,7 +2236,8 @@ row_upd( } if (node->state == UPD_NODE_UPDATE_CLUSTERED - || node->state == UPD_NODE_INSERT_CLUSTERED) { + || node->state == UPD_NODE_INSERT_CLUSTERED + || node->state == UPD_NODE_INSERT_BLOB) { log_free_check(); err = row_upd_clust_step(node, thr); diff --git a/storage/innobase/srv/srv0srv.c b/storage/innobase/srv/srv0srv.c index 2ac5c8ce4fc..1c57d6960d0 100644 --- a/storage/innobase/srv/srv0srv.c +++ b/storage/innobase/srv/srv0srv.c @@ -87,9 +87,11 @@ Created 10/8/1995 Heikki Tuuri #include "mysql/plugin.h" #include "mysql/service_thd_wait.h" -/* This is set to TRUE if the MySQL user has set it in MySQL; currently -affects only FOREIGN KEY definition parsing */ -UNIV_INTERN ibool srv_lower_case_table_names = FALSE; +/* This is set to the MySQL server value for this variable. It is only +needed for FOREIGN KEY definition parsing since FOREIGN KEY names are not +stored in the server metadata. The server stores and enforces it for +regular database and table names.*/ +UNIV_INTERN uint srv_lower_case_table_names = 0; /* The following counter is incremented whenever there is some user activity in the server */ @@ -831,6 +833,7 @@ srv_table_reserve_slot( ut_a(type > 0); ut_a(type <= SRV_MASTER); + ut_ad(mutex_own(&kernel_mutex)); i = 0; slot = srv_table_get_nth_slot(i); @@ -980,6 +983,37 @@ srv_get_thread_type(void) } /*********************************************************************//** +Check whether thread type has reserved a slot. Return the first slot that +is found. This works because we currently have only 1 thread of each type. +@return slot number or ULINT_UNDEFINED if not found*/ +UNIV_INTERN +ulint +srv_thread_has_reserved_slot( +/*=========================*/ + enum srv_thread_type type) /*!< in: thread type to check */ +{ + ulint i; + ulint slot_no = ULINT_UNDEFINED; + + mutex_enter(&kernel_mutex); + + for (i = 0; i < OS_THREAD_MAX_N; i++) { + srv_slot_t* slot; + + slot = srv_table_get_nth_slot(i); + + if (slot->in_use && slot->type == type) { + slot_no = i; + break; + } + } + + mutex_exit(&kernel_mutex); + + return(slot_no); +} + +/*********************************************************************//** Initializes the server. */ UNIV_INTERN void @@ -2627,10 +2661,10 @@ srv_master_thread( srv_main_thread_process_no = os_proc_get_number(); srv_main_thread_id = os_thread_pf(os_thread_get_curr_id()); - srv_table_reserve_slot(SRV_MASTER); - mutex_enter(&kernel_mutex); + srv_table_reserve_slot(SRV_MASTER); + srv_n_threads_active[SRV_MASTER]++; mutex_exit(&kernel_mutex); diff --git a/storage/innobase/srv/srv0start.c b/storage/innobase/srv/srv0start.c index d7ea3d5d3da..acb448a1a67 100644 --- a/storage/innobase/srv/srv0start.c +++ b/storage/innobase/srv/srv0start.c @@ -1026,26 +1026,35 @@ innobase_start_or_create_for_mysql(void) on Mac OS X 10.3 or later. */ struct utsname utsname; if (uname(&utsname)) { - fputs("InnoDB: cannot determine Mac OS X version!\n", stderr); + ut_print_timestamp(stderr); + fputs(" InnoDB: cannot determine Mac OS X version!\n", stderr); } else { srv_have_fullfsync = strcmp(utsname.release, "7.") >= 0; } if (!srv_have_fullfsync) { - fputs("InnoDB: On Mac OS X, fsync() may be" - " broken on internal drives,\n" - "InnoDB: making transactions unsafe!\n", stderr); + ut_print_timestamp(stderr); + fputs(" InnoDB: On Mac OS X, fsync() may be " + "broken on internal drives,\n", stderr); + ut_print_timestamp(stderr); + fputs(" InnoDB: making transactions unsafe!\n", stderr); } # endif /* F_FULLFSYNC */ #endif /* HAVE_DARWIN_THREADS */ if (sizeof(ulint) != sizeof(void*)) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: size of InnoDB's ulint is %lu," - " but size of void* is %lu.\n" - "InnoDB: The sizes should be the same" - " so that on a 64-bit platform you can\n" - "InnoDB: allocate more than 4 GB of memory.", - (ulong)sizeof(ulint), (ulong)sizeof(void*)); + " InnoDB: Error: size of InnoDB's ulint is %lu, " + "but size of void*\n", (ulong) sizeof(ulint)); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: is %lu. The sizes should be the same " + "so that on a 64-bit\n", + (ulong) sizeof(void*)); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: platforms you can allocate more than 4 GB " + "of memory.\n"); } /* System tables are created in tablespace 0. Thus, we must @@ -1054,53 +1063,68 @@ innobase_start_or_create_for_mysql(void) innodb_file_per_table) until this function has returned. */ srv_file_per_table = FALSE; #ifdef UNIV_DEBUG + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!!!!!!! UNIV_DEBUG switched on !!!!!!!!!\n"); + " InnoDB: !!!!!!!! UNIV_DEBUG switched on !!!!!!!!!\n"); #endif #ifdef UNIV_IBUF_DEBUG + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!!!!!!! UNIV_IBUF_DEBUG switched on !!!!!!!!!\n" + " InnoDB: !!!!!!!! UNIV_IBUF_DEBUG switched on !!!!!!!!!\n"); # ifdef UNIV_IBUF_COUNT_DEBUG - "InnoDB: !!!!!!!! UNIV_IBUF_COUNT_DEBUG switched on !!!!!!!!!\n" - "InnoDB: Crash recovery will fail with UNIV_IBUF_COUNT_DEBUG\n" + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: !!!!!!!! UNIV_IBUF_COUNT_DEBUG switched on " + "!!!!!!!!!\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Crash recovery will fail with UNIV_IBUF_COUNT_DEBUG\n"); # endif - ); #endif #ifdef UNIV_SYNC_DEBUG + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!!!!!!! UNIV_SYNC_DEBUG switched on !!!!!!!!!\n"); + " InnoDB: !!!!!!!! UNIV_SYNC_DEBUG switched on !!!!!!!!!\n"); #endif #ifdef UNIV_SEARCH_DEBUG + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!!!!!!! UNIV_SEARCH_DEBUG switched on !!!!!!!!!\n"); + " InnoDB: !!!!!!!! UNIV_SEARCH_DEBUG switched on !!!!!!!!!\n"); #endif #ifdef UNIV_LOG_LSN_DEBUG + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!!!!!!! UNIV_LOG_LSN_DEBUG switched on !!!!!!!!!\n"); + " InnoDB: !!!!!!!! UNIV_LOG_LSN_DEBUG switched on !!!!!!!!!\n"); #endif /* UNIV_LOG_LSN_DEBUG */ #ifdef UNIV_MEM_DEBUG + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!!!!!!! UNIV_MEM_DEBUG switched on !!!!!!!!!\n"); + " InnoDB: !!!!!!!! UNIV_MEM_DEBUG switched on !!!!!!!!!\n"); #endif if (UNIV_LIKELY(srv_use_sys_malloc)) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: The InnoDB memory heap is disabled\n"); + " InnoDB: The InnoDB memory heap is disabled\n"); } - fputs("InnoDB: " IB_ATOMICS_STARTUP_MSG - "\nInnoDB: Compressed tables use zlib " ZLIB_VERSION + ut_print_timestamp(stderr); + fputs(" InnoDB: " IB_ATOMICS_STARTUP_MSG "\n", stderr); + + ut_print_timestamp(stderr); + fputs(" InnoDB: Compressed tables use zlib " ZLIB_VERSION #ifdef UNIV_ZIP_DEBUG " with validation" #endif /* UNIV_ZIP_DEBUG */ + "\n" , stderr); #ifdef UNIV_ZIP_COPY - " and extra copying" + ut_print_timestamp(stderr); + fputs(" InnoDB: and extra copying\n", stderr); #endif /* UNIV_ZIP_COPY */ - "\n" , stderr); /* Since InnoDB does not currently clean up all its internal data structures in MySQL Embedded Server Library server_end(), we @@ -1108,13 +1132,17 @@ innobase_start_or_create_for_mysql(void) second time during the process lifetime. */ if (srv_start_has_been_called) { - fprintf(stderr, - "InnoDB: Error: startup called second time" - " during the process lifetime.\n" - "InnoDB: In the MySQL Embedded Server Library" - " you cannot call server_init()\n" - "InnoDB: more than once during" - " the process lifetime.\n"); + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: Error: startup called second time " + "during the process\n"); + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: lifetime. In the MySQL Embedded " + "Server Library you\n"); + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: cannot call server_init() more " + "than once during the\n"); + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: process lifetime.\n"); } srv_start_has_been_called = TRUE; @@ -1158,7 +1186,7 @@ innobase_start_or_create_for_mysql(void) if (srv_use_native_aio) { ut_print_timestamp(stderr); fprintf(stderr, - " InnoDB: Using Linux native AIO\n"); + " InnoDB: Using Linux native AIO\n"); } #else /* Currently native AIO is supported only on windows and linux @@ -1203,8 +1231,9 @@ innobase_start_or_create_for_mysql(void) srv_win_file_flush_method = SRV_WIN_IO_UNBUFFERED; #endif } else { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Unrecognized value %s for" + " InnoDB: Unrecognized value %s for" " innodb_flush_method\n", srv_file_flush_method_str); return(DB_ERROR); @@ -1285,7 +1314,8 @@ innobase_start_or_create_for_mysql(void) we'll emit a message telling the user that this parameter is now deprecated. */ if (srv_n_file_io_threads != 4) { - fprintf(stderr, "InnoDB: Warning:" + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: Warning:" " innodb_file_io_threads is deprecated." " Please use innodb_read_io_threads and" " innodb_write_io_threads instead\n"); @@ -1316,7 +1346,7 @@ innobase_start_or_create_for_mysql(void) /* Print time to initialize the buffer pool */ ut_print_timestamp(stderr); fprintf(stderr, - " InnoDB: Initializing buffer pool, size ="); + " InnoDB: Initializing buffer pool, size ="); if (srv_buf_pool_size >= 1024 * 1024 * 1024) { fprintf(stderr, @@ -1332,11 +1362,12 @@ innobase_start_or_create_for_mysql(void) ut_print_timestamp(stderr); fprintf(stderr, - " InnoDB: Completed initialization of buffer pool\n"); + " InnoDB: Completed initialization of buffer pool\n"); if (err != DB_SUCCESS) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Fatal error: cannot allocate the memory" + " InnoDB: Fatal error: cannot allocate memory" " for the buffer pool\n"); return(DB_ERROR); @@ -1348,7 +1379,8 @@ innobase_start_or_create_for_mysql(void) if (srv_buf_pool_size <= 5 * 1024 * 1024) { - fprintf(stderr, "InnoDB: Warning: Small buffer pool size " + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: Warning: Small buffer pool size " "(%luM), the flst_validate() debug function " "can cause a deadlock if the buffer pool fills up.\n", srv_buf_pool_size / 1024 / 1024); @@ -1370,18 +1402,19 @@ innobase_start_or_create_for_mysql(void) #ifdef UNIV_LOG_ARCHIVE if (0 != ut_strcmp(srv_log_group_home_dirs[0], srv_arch_dir)) { - fprintf(stderr, - "InnoDB: Error: you must set the log group" - " home dir in my.cnf the\n" - "InnoDB: same as log arch dir.\n"); + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: Error: you must set the log group home dir in my.cnf\n"); + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: the same as log arch dir.\n"); return(DB_ERROR); } #endif /* UNIV_LOG_ARCHIVE */ if (srv_n_log_files * srv_log_file_size >= 262144) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: combined size of log files" + " InnoDB: Error: combined size of log files" " must be < 4 GB\n"); return(DB_ERROR); @@ -1392,10 +1425,13 @@ innobase_start_or_create_for_mysql(void) for (i = 0; i < srv_n_data_files; i++) { #ifndef __WIN__ if (sizeof(off_t) < 5 && srv_data_file_sizes[i] >= 262144) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Error: file size must be < 4 GB" + " with this MySQL binary\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: file size must be < 4 GB" - " with this MySQL binary\n" - "InnoDB: and operating system combination," + " InnoDB: and operating system combination," " in some OS's < 2 GB\n"); return(DB_ERROR); @@ -1405,8 +1441,9 @@ innobase_start_or_create_for_mysql(void) } if (sum_of_new_sizes < 10485760 / UNIV_PAGE_SIZE) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: tablespace size must be" + " InnoDB: Error: tablespace size must be" " at least 10 MB\n"); return(DB_ERROR); @@ -1419,19 +1456,32 @@ innobase_start_or_create_for_mysql(void) &min_flushed_lsn, &max_flushed_lsn, &sum_of_new_sizes); if (err != DB_SUCCESS) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Could not open or create data files.\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: If you tried to add new data files," + " and it failed here,\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: you should now edit innodb_data_file_path" + " in my.cnf back\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Could not open or create data files.\n" - "InnoDB: If you tried to add new data files," - " and it failed here,\n" - "InnoDB: you should now edit innodb_data_file_path" - " in my.cnf back\n" - "InnoDB: to what it was, and remove the" - " new ibdata files InnoDB created\n" - "InnoDB: in this failed attempt. InnoDB only wrote" - " those files full of\n" - "InnoDB: zeros, but did not yet use them in any way." - " But be careful: do not\n" - "InnoDB: remove old data files" + " InnoDB: to what it was, and remove the" + " new ibdata files InnoDB created\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: in this failed attempt. InnoDB only wrote" + " those files full of\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: zeros, but did not yet use them in any way." + " But be careful: do not\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: remove old data files" " which contain your precious data!\n"); return((int) err); @@ -1457,18 +1507,29 @@ innobase_start_or_create_for_mysql(void) } if ((log_opened && create_new_db) || (log_opened && log_created)) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Error: all log files must be" + " created at the same time.\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: All log files must be" + " created also in database creation.\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: If you want bigger or smaller" + " log files, shut down the\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: database and make sure there" + " were no errors in shutdown.\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Then delete the existing log files." + " Edit the .cnf file\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: all log files must be" - " created at the same time.\n" - "InnoDB: All log files must be" - " created also in database creation.\n" - "InnoDB: If you want bigger or smaller" - " log files, shut down the\n" - "InnoDB: database and make sure there" - " were no errors in shutdown.\n" - "InnoDB: Then delete the existing log files." - " Edit the .cnf file\n" - "InnoDB: and start the database again.\n"); + " InnoDB: and start the database again.\n"); return(DB_ERROR); } @@ -1489,27 +1550,41 @@ innobase_start_or_create_for_mysql(void) || max_arch_log_no != min_arch_log_no #endif /* UNIV_LOG_ARCHIVE */ ) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Cannot initialize created" + " log files because\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: data files were not in sync" + " with each other\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Cannot initialize created" - " log files because\n" - "InnoDB: data files were not in sync" - " with each other\n" - "InnoDB: or the data files are corrupt.\n"); + " InnoDB: or the data files are corrupt.\n"); return(DB_ERROR); } if (max_flushed_lsn < (ib_uint64_t) 1000) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Cannot initialize created" + " log files because\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Cannot initialize created" - " log files because\n" - "InnoDB: data files are corrupt," - " or new data files were\n" - "InnoDB: created when the database" - " was started previous\n" - "InnoDB: time but the database" - " was not shut down\n" - "InnoDB: normally after that.\n"); + " InnoDB: data files are corrupt," + " or new data files were\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: created when the database" + " was started previous\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: time but the database" + " was not shut down\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: normally after that.\n"); return(DB_ERROR); } @@ -1548,8 +1623,9 @@ innobase_start_or_create_for_mysql(void) #ifdef UNIV_LOG_ARCHIVE } else if (srv_archive_recovery) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Starting archive" + " InnoDB: Starting archive" " recovery from a backup...\n"); err = recv_recovery_from_archive_start( min_flushed_lsn, srv_archive_recovery_limit_lsn, @@ -1757,6 +1833,24 @@ innobase_start_or_create_for_mysql(void) os_thread_create(&srv_purge_thread, NULL, NULL); } + /* Wait for the purge and master thread to startup. */ + + while (srv_shutdown_state == SRV_SHUTDOWN_NONE) { + if (srv_thread_has_reserved_slot(SRV_MASTER) == ULINT_UNDEFINED + || (srv_n_purge_threads == 1 + && srv_thread_has_reserved_slot(SRV_WORKER) + == ULINT_UNDEFINED)) { + + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: " + "Waiting for the background threads to " + "start\n"); + os_thread_sleep(1000000); + } else { + break; + } + } + #ifdef UNIV_DEBUG /* buf_debug_prints = TRUE; */ #endif /* UNIV_DEBUG */ @@ -1771,11 +1865,14 @@ innobase_start_or_create_for_mysql(void) if (!srv_auto_extend_last_data_file && sum_of_data_file_sizes != tablespace_size_in_header) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Error: tablespace size" + " stored in header is %lu pages, but\n", + (ulong) tablespace_size_in_header); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: tablespace size" - " stored in header is %lu pages, but\n" "InnoDB: the sum of data file sizes is %lu pages\n", - (ulong) tablespace_size_in_header, (ulong) sum_of_data_file_sizes); if (srv_force_recovery == 0 @@ -1783,16 +1880,25 @@ innobase_start_or_create_for_mysql(void) /* This is a fatal error, the tail of a tablespace is missing */ + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Cannot start InnoDB." + " The tail of the system tablespace is\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: missing. Have you edited" + " innodb_data_file_path in my.cnf in an\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Cannot start InnoDB." - " The tail of the system tablespace is\n" - "InnoDB: missing. Have you edited" - " innodb_data_file_path in my.cnf in an\n" - "InnoDB: inappropriate way, removing" - " ibdata files from there?\n" - "InnoDB: You can set innodb_force_recovery=1" - " in my.cnf to force\n" - "InnoDB: a startup if you are trying" + " InnoDB: inappropriate way, removing" + " ibdata files from there?\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: You can set innodb_force_recovery=1" + " in my.cnf to force\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: a startup if you are trying" " to recover a badly corrupt database.\n"); return(DB_ERROR); @@ -1802,26 +1908,38 @@ innobase_start_or_create_for_mysql(void) if (srv_auto_extend_last_data_file && sum_of_data_file_sizes < tablespace_size_in_header) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: tablespace size stored in header" - " is %lu pages, but\n" - "InnoDB: the sum of data file sizes" + " InnoDB: Error: tablespace size stored in header" + " is %lu pages, but\n", + (ulong) tablespace_size_in_header); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: the sum of data file sizes" " is only %lu pages\n", - (ulong) tablespace_size_in_header, (ulong) sum_of_data_file_sizes); if (srv_force_recovery == 0) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Cannot start InnoDB. The tail of" + " the system tablespace is\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: missing. Have you edited" + " innodb_data_file_path in my.cnf in an\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: inappropriate way, removing" + " ibdata files from there?\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: You can set innodb_force_recovery=1" + " in my.cnf to force\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Cannot start InnoDB. The tail of" - " the system tablespace is\n" - "InnoDB: missing. Have you edited" - " innodb_data_file_path in my.cnf in an\n" - "InnoDB: inappropriate way, removing" - " ibdata files from there?\n" - "InnoDB: You can set innodb_force_recovery=1" - " in my.cnf to force\n" - "InnoDB: a startup if you are trying to" + " InnoDB: a startup if you are trying to" " recover a badly corrupt database.\n"); return(DB_ERROR); @@ -1832,10 +1950,13 @@ innobase_start_or_create_for_mysql(void) os_fast_mutex_init(&srv_os_test_mutex); if (0 != os_fast_mutex_trylock(&srv_os_test_mutex)) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Error: pthread_mutex_trylock returns" - " an unexpected value on\n" - "InnoDB: success! Cannot continue.\n"); + " InnoDB: Error: pthread_mutex_trylock returns" + " an unexpected value on\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: success! Cannot continue.\n"); exit(1); } @@ -1850,14 +1971,15 @@ innobase_start_or_create_for_mysql(void) if (srv_print_verbose_log) { ut_print_timestamp(stderr); fprintf(stderr, - " InnoDB: %s started; " + " InnoDB: %s started; " "log sequence number %llu\n", INNODB_VERSION_STR, srv_start_lsn); } if (srv_force_recovery > 0) { + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: !!! innodb_force_recovery" + " InnoDB: !!! innodb_force_recovery" " is set to %lu !!!\n", (ulong) srv_force_recovery); } @@ -1878,12 +2000,17 @@ innobase_start_or_create_for_mysql(void) 4.1.1. It is essential that the insert buffer is emptied here! */ + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: You are upgrading to an" + " InnoDB version which allows multiple\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: tablespaces. Wait that purge" + " and insert buffer merge run to\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: You are upgrading to an" - " InnoDB version which allows multiple\n" - "InnoDB: tablespaces. Wait that purge" - " and insert buffer merge run to\n" - "InnoDB: completion...\n"); + " InnoDB: completion...\n"); for (;;) { os_thread_sleep(1000000); @@ -1895,21 +2022,29 @@ innobase_start_or_create_for_mysql(void) break; } } + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: Full purge and insert buffer merge" + " InnoDB: Full purge and insert buffer merge" " completed.\n"); trx_sys_mark_upgraded_to_multiple_tablespaces(); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: You have now successfully upgraded" + " to the multiple tablespaces\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: format. You should NOT DOWNGRADE" + " to an earlier version of\n"); + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: InnoDB! But if you absolutely need to" + " downgrade, see\n"); + ut_print_timestamp(stderr); fprintf(stderr, - "InnoDB: You have now successfully upgraded" - " to the multiple tablespaces\n" - "InnoDB: format. You should NOT DOWNGRADE" - " to an earlier version of\n" - "InnoDB: InnoDB! But if you absolutely need to" - " downgrade, see\n" - "InnoDB: " REFMAN "multiple-tablespaces.html\n" - "InnoDB: for instructions.\n"); + " InnoDB: " REFMAN "multiple-tablespaces.html\n" + " InnoDB: for instructions.\n"); } if (srv_force_recovery == 0) { diff --git a/storage/innobase/sync/sync0arr.c b/storage/innobase/sync/sync0arr.c index 753ebd958ac..ce9d42c4921 100644 --- a/storage/innobase/sync/sync0arr.c +++ b/storage/innobase/sync/sync0arr.c @@ -590,9 +590,6 @@ sync_array_deadlock_step( ulint depth) /*!< in: recursion depth */ { sync_cell_t* new; - ibool ret; - - depth++; if (pass != 0) { /* If pass != 0, then we do not know which threads are @@ -604,7 +601,7 @@ sync_array_deadlock_step( new = sync_array_find_thread(arr, thread); - if (new == start) { + if (UNIV_UNLIKELY(new == start)) { /* Stop running of other threads */ ut_dbg_stop_threads = TRUE; @@ -616,11 +613,7 @@ sync_array_deadlock_step( return(TRUE); } else if (new) { - ret = sync_array_detect_deadlock(arr, start, new, depth); - - if (ret) { - return(TRUE); - } + return(sync_array_detect_deadlock(arr, start, new, depth + 1)); } return(FALSE); } @@ -930,12 +923,25 @@ sync_array_print_long_waits(void) ulint fatal_timeout = srv_fatal_semaphore_wait_threshold; ibool fatal = FALSE; +#ifdef UNIV_DEBUG_VALGRIND + /* Increase the timeouts if running under valgrind because it executes + extremely slowly. UNIV_DEBUG_VALGRIND does not necessary mean that + we are running under valgrind but we have no better way to tell. + See Bug#58432 innodb.innodb_bug56143 fails under valgrind + for an example */ +# define SYNC_ARRAY_TIMEOUT 2400 + fatal_timeout *= 10; +#else +# define SYNC_ARRAY_TIMEOUT 240 +#endif + for (i = 0; i < sync_primary_wait_array->n_cells; i++) { cell = sync_array_get_nth_cell(sync_primary_wait_array, i); if (cell->wait_object != NULL && cell->waiting - && difftime(time(NULL), cell->reservation_time) > 240) { + && difftime(time(NULL), cell->reservation_time) + > SYNC_ARRAY_TIMEOUT) { fputs("InnoDB: Warning: a long semaphore wait:\n", stderr); sync_array_cell_print(stderr, cell); @@ -977,6 +983,8 @@ sync_array_print_long_waits(void) " to the standard error stream\n"); } +#undef SYNC_ARRAY_TIMEOUT + return(fatal); } diff --git a/storage/innobase/trx/trx0i_s.c b/storage/innobase/trx/trx0i_s.c index 8b11e706180..aff49e20726 100644 --- a/storage/innobase/trx/trx0i_s.c +++ b/storage/innobase/trx/trx0i_s.c @@ -442,7 +442,7 @@ i_s_locks_row_validate( /* record lock */ ut_ad(!strcmp("RECORD", row->lock_type)); ut_ad(row->lock_index != NULL); - ut_ad(row->lock_data != NULL); + /* row->lock_data == NULL if buf_page_try_get() == NULL */ ut_ad(row->lock_page != ULINT_UNDEFINED); ut_ad(row->lock_rec != ULINT_UNDEFINED); } @@ -508,7 +508,6 @@ fill_trx_row( stmt = innobase_get_stmt(trx->mysql_thd, &stmt_len); if (stmt != NULL) { - char query[TRX_I_S_TRX_QUERY_MAX_LEN + 1]; if (stmt_len > TRX_I_S_TRX_QUERY_MAX_LEN) { @@ -522,6 +521,8 @@ fill_trx_row( cache->storage, stmt, stmt_len + 1, MAX_ALLOWED_FOR_STORAGE(cache)); + row->trx_query_cs = innobase_get_charset(trx->mysql_thd); + if (row->trx_query == NULL) { return(FALSE); diff --git a/storage/innobase/trx/trx0sys.c b/storage/innobase/trx/trx0sys.c index 26498a1b712..101f225a06f 100644 --- a/storage/innobase/trx/trx0sys.c +++ b/storage/innobase/trx/trx0sys.c @@ -1160,7 +1160,7 @@ trx_sys_file_format_max_check( ut_print_timestamp(stderr); fprintf(stderr, - " InnoDB: highest supported file format is %s.\n", + " InnoDB: highest supported file format is %s.\n", trx_sys_file_format_id_to_name(DICT_TF_FORMAT_MAX)); if (format_id > DICT_TF_FORMAT_MAX) { @@ -1169,7 +1169,7 @@ trx_sys_file_format_max_check( ut_print_timestamp(stderr); fprintf(stderr, - " InnoDB: %s: the system tablespace is in a file " + " InnoDB: %s: the system tablespace is in a file " "format that this version doesn't support - %s\n", ((max_format_id <= DICT_TF_FORMAT_MAX) ? "Error" : "Warning"), diff --git a/storage/innobase/ut/ut0dbg.c b/storage/innobase/ut/ut0dbg.c index e79217d5b86..42ad8c03762 100644 --- a/storage/innobase/ut/ut0dbg.c +++ b/storage/innobase/ut/ut0dbg.c @@ -74,7 +74,7 @@ ut_dbg_assertion_failed( " or crashes, even\n" "InnoDB: immediately after the mysqld startup, there may be\n" "InnoDB: corruption in the InnoDB tablespace. Please refer to\n" - "InnoDB: " REFMAN "forcing-recovery.html\n" + "InnoDB: " REFMAN "forcing-innodb-recovery.html\n" "InnoDB: about forcing recovery.\n", stderr); #if defined(UNIV_SYNC_DEBUG) || !defined(UT_DBG_USE_ABORT) ut_dbg_stop_threads = TRUE; |