diff options
Diffstat (limited to 'storage')
38 files changed, 1050 insertions, 441 deletions
diff --git a/storage/archive/ha_archive.cc b/storage/archive/ha_archive.cc index e705d1dd615..f5b96989811 100644 --- a/storage/archive/ha_archive.cc +++ b/storage/archive/ha_archive.cc @@ -909,10 +909,11 @@ int ha_archive::write_row(uchar *buf) table->timestamp_field->set_time(); mysql_mutex_lock(&share->mutex); - if (!share->archive_write_open) - if (init_archive_writer()) - DBUG_RETURN(errno); - + if (!share->archive_write_open && init_archive_writer()) + { + rc= errno; + goto error; + } if (table->next_number_field && record == table->record[0]) { @@ -992,7 +993,6 @@ int ha_archive::write_row(uchar *buf) error: mysql_mutex_unlock(&share->mutex); my_free(read_buf); - DBUG_RETURN(rc); } diff --git a/storage/heap/hp_delete.c b/storage/heap/hp_delete.c index d00ac94a918..a01db50a478 100644 --- a/storage/heap/hp_delete.c +++ b/storage/heap/hp_delete.c @@ -105,7 +105,7 @@ int hp_rb_delete_key(HP_INFO *info, register HP_KEYDEF *keyinfo, int hp_delete_key(HP_INFO *info, register HP_KEYDEF *keyinfo, const uchar *record, uchar *recpos, int flag) { - ulong blength,pos2,pos_hashnr,lastpos_hashnr; + ulong blength, pos2, pos_hashnr, lastpos_hashnr, key_pos; HASH_INFO *lastpos,*gpos,*pos,*pos3,*empty,*last_ptr; HP_SHARE *share=info->s; DBUG_ENTER("hp_delete_key"); @@ -117,9 +117,9 @@ int hp_delete_key(HP_INFO *info, register HP_KEYDEF *keyinfo, last_ptr=0; /* Search after record with key */ - pos= hp_find_hash(&keyinfo->block, - hp_mask(hp_rec_hashnr(keyinfo, record), blength, - share->records + 1)); + key_pos= hp_mask(hp_rec_hashnr(keyinfo, record), blength, share->records + 1); + pos= hp_find_hash(&keyinfo->block, key_pos); + gpos = pos3 = 0; while (pos->ptr_to_rec != recpos) diff --git a/storage/innobase/btr/btr0btr.c b/storage/innobase/btr/btr0btr.c index 1f4d8126be6..24ff78df7c0 100644 --- a/storage/innobase/btr/btr0btr.c +++ b/storage/innobase/btr/btr0btr.c @@ -3094,6 +3094,8 @@ btr_lift_page_up( buf_block_t* blocks[BTR_MAX_LEVELS]; ulint n_blocks; /*!< last used index in blocks[] */ ulint i; + ibool lift_father_up = FALSE; + buf_block_t* block_orig = block; ut_ad(btr_page_get_prev(page, mtr) == FIL_NULL); ut_ad(btr_page_get_next(page, mtr) == FIL_NULL); @@ -3104,11 +3106,13 @@ btr_lift_page_up( { btr_cur_t cursor; - mem_heap_t* heap = mem_heap_create(100); - ulint* offsets; + ulint* offsets = NULL; + mem_heap_t* heap = mem_heap_create( + sizeof(*offsets) + * (REC_OFFS_HEADER_SIZE + 1 + 1 + index->n_fields)); buf_block_t* b; - offsets = btr_page_get_father_block(NULL, heap, index, + offsets = btr_page_get_father_block(offsets, heap, index, block, mtr, &cursor); father_block = btr_cur_get_block(&cursor); father_page_zip = buf_block_get_page_zip(father_block); @@ -3132,6 +3136,29 @@ btr_lift_page_up( blocks[n_blocks++] = b = btr_cur_get_block(&cursor); } + if (n_blocks && page_level == 0) { + /* The father page also should be the only on its level (not + root). We should lift up the father page at first. + Because the leaf page should be lifted up only for root page. + The freeing page is based on page_level (==0 or !=0) + to choose segment. If the page_level is changed ==0 from !=0, + later freeing of the page doesn't find the page allocation + to be freed.*/ + + lift_father_up = TRUE; + block = father_block; + page = buf_block_get_frame(block); + page_level = btr_page_get_level(page, mtr); + + ut_ad(btr_page_get_prev(page, mtr) == FIL_NULL); + ut_ad(btr_page_get_next(page, mtr) == FIL_NULL); + ut_ad(mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX)); + + father_block = blocks[0]; + father_page_zip = buf_block_get_page_zip(father_block); + father_page = buf_block_get_frame(father_block); + } + mem_heap_free(heap); } @@ -3139,6 +3166,7 @@ btr_lift_page_up( /* Make the father empty */ btr_page_empty(father_block, father_page_zip, index, page_level, mtr); + page_level++; /* Copy the records to the father page one by one. */ if (0 @@ -3171,7 +3199,7 @@ btr_lift_page_up( lock_update_copy_and_discard(father_block, block); /* Go upward to root page, decrementing levels by one. */ - for (i = 0; i < n_blocks; i++, page_level++) { + for (i = lift_father_up ? 1 : 0; i < n_blocks; i++, page_level++) { page_t* page = buf_block_get_frame(blocks[i]); page_zip_des_t* page_zip= buf_block_get_page_zip(blocks[i]); @@ -3193,7 +3221,7 @@ btr_lift_page_up( ut_ad(page_validate(father_page, index)); ut_ad(btr_check_node_ptr(index, father_block, mtr)); - return(father_block); + return(lift_father_up ? block_orig : father_block); } /*************************************************************//** diff --git a/storage/innobase/btr/btr0cur.c b/storage/innobase/btr/btr0cur.c index 1f92de245b1..955be01f032 100644 --- a/storage/innobase/btr/btr0cur.c +++ b/storage/innobase/btr/btr0cur.c @@ -97,6 +97,11 @@ srv_refresh_innodb_monitor_stats(). Referenced by srv_printf_innodb_monitor(). */ UNIV_INTERN ulint btr_cur_n_sea_old = 0; +#ifdef UNIV_DEBUG +/* Flag to limit optimistic insert records */ +UNIV_INTERN uint btr_cur_limit_optimistic_insert_debug = 0; +#endif /* UNIV_DEBUG */ + /** In the optimistic insert, if the insert does not fit, but this much space can be released by page reorganize, then it is reorganized */ #define BTR_CUR_PAGE_REORGANIZE_LIMIT (UNIV_PAGE_SIZE / 32) @@ -1273,6 +1278,9 @@ btr_cur_optimistic_insert( } } + LIMIT_OPTIMISTIC_INSERT_DEBUG(page_get_n_recs(page), + goto fail); + /* If there have been many consecutive inserts, and we are on the leaf level, check if we have to split the page to reserve enough free space for future updates of records. */ diff --git a/storage/innobase/buf/buf0buf.c b/storage/innobase/buf/buf0buf.c index 1af019dfc45..2406980196e 100644 --- a/storage/innobase/buf/buf0buf.c +++ b/storage/innobase/buf/buf0buf.c @@ -370,6 +370,33 @@ buf_get_total_list_len( } /********************************************************************//** +Get total list size in bytes from all buffer pools. */ +UNIV_INTERN +void +buf_get_total_list_size_in_bytes( +/*=============================*/ + buf_pools_list_size_t* buf_pools_list_size) /*!< out: list sizes + in all buffer pools */ +{ + ulint i; + ut_ad(buf_pools_list_size); + memset(buf_pools_list_size, 0, sizeof(*buf_pools_list_size)); + + for (i = 0; i < srv_buf_pool_instances; i++) { + buf_pool_t* buf_pool; + + buf_pool = buf_pool_from_array(i); + /* We don't need mutex protection since this is + for statistics purpose */ + buf_pools_list_size->LRU_bytes += buf_pool->stat.LRU_bytes; + buf_pools_list_size->unzip_LRU_bytes += + UT_LIST_GET_LEN(buf_pool->unzip_LRU) * UNIV_PAGE_SIZE; + buf_pools_list_size->flush_list_bytes += + buf_pool->stat.flush_list_bytes; + } +} + +/********************************************************************//** Get total buffer pool statistics. */ UNIV_INTERN void @@ -1657,34 +1684,24 @@ buf_page_make_young( } /********************************************************************//** -Sets the time of the first access of a page and moves a page to the -start of the buffer pool LRU list if it is too old. This high-level -function can be used to prevent an important page from slipping -out of the buffer pool. */ +Moves a page to the start of the buffer pool LRU list if it is too old. +This high-level function can be used to prevent an important page from +slipping out of the buffer pool. */ static void -buf_page_set_accessed_make_young( -/*=============================*/ - buf_page_t* bpage, /*!< in/out: buffer block of a +buf_page_make_young_if_needed( +/*==========================*/ + buf_page_t* bpage) /*!< in/out: buffer block of a file page */ - unsigned access_time) /*!< in: bpage->access_time - read under mutex protection, - or 0 if unknown */ { +#ifdef UNIV_DEBUG buf_pool_t* buf_pool = buf_pool_from_bpage(bpage); - ut_ad(!buf_pool_mutex_own(buf_pool)); +#endif /* UNIV_DEBUG */ ut_a(buf_page_in_file(bpage)); if (buf_page_peek_if_too_old(bpage)) { - buf_pool_mutex_enter(buf_pool); - buf_LRU_make_block_young(bpage); - buf_pool_mutex_exit(buf_pool); - } else if (!access_time) { - ulint time_ms = ut_time_ms(); - buf_pool_mutex_enter(buf_pool); - buf_page_set_accessed(bpage, time_ms); - buf_pool_mutex_exit(buf_pool); + buf_page_make_young(bpage); } } @@ -1797,7 +1814,6 @@ buf_page_get_zip( buf_page_t* bpage; mutex_t* block_mutex; ibool must_read; - unsigned access_time; buf_pool_t* buf_pool = buf_pool_get(space, offset); buf_pool->stat.n_page_gets++; @@ -1865,13 +1881,14 @@ err_exit: got_block: must_read = buf_page_get_io_fix(bpage) == BUF_IO_READ; - access_time = buf_page_is_accessed(bpage); buf_pool_mutex_exit(buf_pool); + buf_page_set_accessed(bpage); + mutex_exit(block_mutex); - buf_page_set_accessed_make_young(bpage, access_time); + buf_page_make_young_if_needed(bpage); #if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG ut_a(!bpage->file_page_was_freed); @@ -2458,22 +2475,29 @@ wait_until_unfixed: UNIV_MEM_INVALID(bpage, sizeof *bpage); - mutex_exit(&block->mutex); - mutex_exit(&buf_pool->zip_mutex); buf_pool->n_pend_unzip++; - buf_pool_mutex_exit(buf_pool); + access_time = buf_page_is_accessed(&block->page); + mutex_exit(&block->mutex); + mutex_exit(&buf_pool->zip_mutex); + buf_page_free_descriptor(bpage); - /* Decompress the page and apply buffered operations - while not holding buf_pool->mutex or block->mutex. */ + /* Decompress the page while not holding + buf_pool->mutex or block->mutex. */ success = buf_zip_decompress(block, srv_use_checksums); ut_a(success); if (UNIV_LIKELY(!recv_no_ibuf_operations)) { - ibuf_merge_or_delete_for_page(block, space, offset, - zip_size, TRUE); + if (access_time) { +#ifdef UNIV_IBUF_COUNT_DEBUG + ut_a(ibuf_count_get(space, offset) == 0); +#endif /* UNIV_IBUF_COUNT_DEBUG */ + } else { + ibuf_merge_or_delete_for_page( + block, space, offset, zip_size, TRUE); + } } /* Unfix and unlatch the block. */ @@ -2550,16 +2574,17 @@ wait_until_unfixed: ut_a(mode == BUF_GET_POSSIBLY_FREED || !block->page.file_page_was_freed); #endif - mutex_exit(&block->mutex); + buf_pool_mutex_exit(buf_pool); /* Check if this is the first access to the page */ - access_time = buf_page_is_accessed(&block->page); - buf_pool_mutex_exit(buf_pool); + buf_page_set_accessed(&block->page); - if (UNIV_LIKELY(mode != BUF_PEEK_IF_IN_POOL)) { - buf_page_set_accessed_make_young(&block->page, access_time); + mutex_exit(&block->mutex); + + if (mode != BUF_PEEK_IF_IN_POOL) { + buf_page_make_young_if_needed(&block->page); } #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG @@ -2610,7 +2635,7 @@ wait_until_unfixed: mtr_memo_push(mtr, block, fix_type); - if (UNIV_LIKELY(mode != BUF_PEEK_IF_IN_POOL) && !access_time) { + if (mode != BUF_PEEK_IF_IN_POOL && !access_time) { /* In the case of a first access, try to apply linear read-ahead */ @@ -2662,15 +2687,13 @@ buf_page_optimistic_get( buf_block_buf_fix_inc(block, file, line); - mutex_exit(&block->mutex); + access_time = buf_page_is_accessed(&block->page); - /* Check if this is the first access to the page. - We do a dirty read on purpose, to avoid mutex contention. - This field is only used for heuristic purposes; it does not - affect correctness. */ + buf_page_set_accessed(&block->page); - access_time = buf_page_is_accessed(&block->page); - buf_page_set_accessed_make_young(&block->page, access_time); + mutex_exit(&block->mutex); + + buf_page_make_young_if_needed(&block->page); ut_ad(!ibuf_inside(mtr) || ibuf_page(buf_block_get_space(block), @@ -2722,7 +2745,7 @@ buf_page_optimistic_get( #if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG ut_a(block->page.file_page_was_freed == FALSE); #endif - if (UNIV_UNLIKELY(!access_time)) { + if (!access_time) { /* In the case of a first access, try to apply linear read-ahead */ @@ -2785,24 +2808,14 @@ buf_page_get_known_nowait( buf_block_buf_fix_inc(block, file, line); + buf_page_set_accessed(&block->page); + mutex_exit(&block->mutex); buf_pool = buf_pool_from_block(block); - if (mode == BUF_MAKE_YOUNG && buf_page_peek_if_too_old(&block->page)) { - buf_pool_mutex_enter(buf_pool); - buf_LRU_make_block_young(&block->page); - buf_pool_mutex_exit(buf_pool); - } else if (!buf_page_is_accessed(&block->page)) { - /* Above, we do a dirty read on purpose, to avoid - mutex contention. The field buf_page_t::access_time - is only used for heuristic purposes. Writes to the - field must be protected by mutex, however. */ - ulint time_ms = ut_time_ms(); - - buf_pool_mutex_enter(buf_pool); - buf_page_set_accessed(&block->page, time_ms); - buf_pool_mutex_exit(buf_pool); + if (mode == BUF_MAKE_YOUNG) { + buf_page_make_young_if_needed(&block->page); } ut_ad(!ibuf_inside(mtr) || mode == BUF_KEEP_OLD); @@ -2965,6 +2978,7 @@ buf_page_init( ulint offset, /*!< in: offset of the page within space in units of a page */ ulint fold, /*!< in: buf_page_address_fold(space,offset) */ + ulint zip_size,/*!< in: compressed page size, or 0 */ buf_block_t* block) /*!< in/out: block to init */ { buf_page_t* hash_page; @@ -3027,6 +3041,9 @@ buf_page_init( ut_d(block->page.in_page_hash = TRUE); HASH_INSERT(buf_page_t, hash, buf_pool->page_hash, fold, &block->page); + if (zip_size) { + page_zip_set_size(&block->page.zip, zip_size); + } } /********************************************************************//** @@ -3128,7 +3145,7 @@ err_exit: ut_ad(buf_pool_from_bpage(bpage) == buf_pool); - buf_page_init(buf_pool, space, offset, fold, block); + buf_page_init(buf_pool, space, offset, fold, zip_size, block); /* The block must be put to the LRU list, to the old blocks */ buf_LRU_add_block(bpage, TRUE/* to old blocks */); @@ -3146,8 +3163,6 @@ err_exit: buf_page_set_io_fix(bpage, BUF_IO_READ); if (UNIV_UNLIKELY(zip_size)) { - page_zip_set_size(&block->page.zip, zip_size); - /* buf_pool->mutex may be released and reacquired by buf_buddy_alloc(). Thus, we must release block->mutex in order not to @@ -3240,7 +3255,8 @@ err_exit: HASH_INSERT(buf_page_t, hash, buf_pool->page_hash, fold, bpage); - /* The block must be put to the LRU list, to the old blocks */ + /* The block must be put to the LRU list, to the old blocks + The zip_size is already set into the page zip */ buf_LRU_add_block(bpage, TRUE/* to old blocks */); #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG buf_LRU_insert_zip_clean(bpage); @@ -3284,7 +3300,6 @@ buf_page_create( buf_block_t* block; ulint fold; buf_block_t* free_block = NULL; - ulint time_ms = ut_time_ms(); buf_pool_t* buf_pool = buf_pool_get(space, offset); ut_ad(mtr); @@ -3332,7 +3347,7 @@ buf_page_create( mutex_enter(&block->mutex); - buf_page_init(buf_pool, space, offset, fold, block); + buf_page_init(buf_pool, space, offset, fold, zip_size, block); /* The block must be put to the LRU list */ buf_LRU_add_block(&block->page, FALSE); @@ -3350,8 +3365,6 @@ buf_page_create( buf_page_set_io_fix(&block->page, BUF_IO_READ); rw_lock_x_lock(&block->lock); - - page_zip_set_size(&block->page.zip, zip_size); mutex_exit(&block->mutex); /* buf_pool->mutex may be released and reacquired by buf_buddy_alloc(). Thus, we must release block->mutex @@ -3375,12 +3388,12 @@ buf_page_create( rw_lock_x_unlock(&block->lock); } - buf_page_set_accessed(&block->page, time_ms); - buf_pool_mutex_exit(buf_pool); mtr_memo_push(mtr, block, MTR_MEMO_BUF_FIX); + buf_page_set_accessed(&block->page); + mutex_exit(&block->mutex); /* Delete possible entries for the page from the insert buffer: diff --git a/storage/innobase/buf/buf0flu.c b/storage/innobase/buf/buf0flu.c index 7cd09d6675e..c385dc793ab 100644 --- a/storage/innobase/buf/buf0flu.c +++ b/storage/innobase/buf/buf0flu.c @@ -79,6 +79,23 @@ static ulint buf_lru_flush_page_count = 0; /* @} */ +/******************************************************************//** +Increases flush_list size in bytes with zip_size for compressed page, +UNIV_PAGE_SIZE for uncompressed page in inline function */ +static inline +void +incr_flush_list_size_in_bytes( +/*==========================*/ + buf_block_t* block, /*!< in: control block */ + buf_pool_t* buf_pool) /*!< in: buffer pool instance */ +{ + ulint zip_size; + ut_ad(buf_flush_list_mutex_own(buf_pool)); + zip_size = page_zip_get_size(&block->page.zip); + buf_pool->stat.flush_list_bytes += zip_size ? zip_size : UNIV_PAGE_SIZE; + ut_ad(buf_pool->stat.flush_list_bytes <= buf_pool->curr_pool_size); +} + #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG /******************************************************************//** Validates the flush list. @@ -308,6 +325,7 @@ buf_flush_insert_into_flush_list( ut_d(block->page.in_flush_list = TRUE); block->page.oldest_modification = lsn; UT_LIST_ADD_FIRST(list, buf_pool->flush_list, &block->page); + incr_flush_list_size_in_bytes(block, buf_pool); #ifdef UNIV_DEBUG_VALGRIND { @@ -412,6 +430,8 @@ buf_flush_insert_sorted_into_flush_list( prev_b, &block->page); } + incr_flush_list_size_in_bytes(block, buf_pool); + #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG ut_a(buf_flush_validate_low(buf_pool)); #endif /* UNIV_DEBUG || UNIV_BUF_DEBUG */ @@ -504,6 +524,7 @@ buf_flush_remove( buf_page_t* bpage) /*!< in: pointer to the block in question */ { buf_pool_t* buf_pool = buf_pool_from_bpage(bpage); + ulint zip_size; ut_ad(buf_pool_mutex_own(buf_pool)); ut_ad(mutex_own(buf_page_get_mutex(bpage))); @@ -542,6 +563,9 @@ buf_flush_remove( because we assert on in_flush_list in comparison function. */ ut_d(bpage->in_flush_list = FALSE); + zip_size = page_zip_get_size(&bpage->zip); + buf_pool->stat.flush_list_bytes -= zip_size ? zip_size : UNIV_PAGE_SIZE; + bpage->oldest_modification = 0; #if defined UNIV_DEBUG || defined UNIV_BUF_DEBUG diff --git a/storage/innobase/buf/buf0lru.c b/storage/innobase/buf/buf0lru.c index 60894cae6ad..b3e282d24b2 100644 --- a/storage/innobase/buf/buf0lru.c +++ b/storage/innobase/buf/buf0lru.c @@ -151,6 +151,23 @@ buf_LRU_block_free_hashed_page( be in a state where it can be freed */ /******************************************************************//** +Increases LRU size in bytes with zip_size for compressed page, +UNIV_PAGE_SIZE for uncompressed page in inline function */ +static inline +void +incr_LRU_size_in_bytes( +/*===================*/ + buf_page_t* bpage, /*!< in: control block */ + buf_pool_t* buf_pool) /*!< in: buffer pool instance */ +{ + ulint zip_size; + ut_ad(buf_pool_mutex_own(buf_pool)); + zip_size = page_zip_get_size(&bpage->zip); + buf_pool->stat.LRU_bytes += zip_size ? zip_size : UNIV_PAGE_SIZE; + ut_ad(buf_pool->stat.LRU_bytes <= buf_pool->curr_pool_size); +} + +/******************************************************************//** Determines if the unzip_LRU list should be used for evicting a victim instead of the general LRU list. @return TRUE if should use unzip_LRU */ @@ -1356,6 +1373,7 @@ buf_LRU_remove_block( buf_page_t* bpage) /*!< in: control block */ { buf_pool_t* buf_pool = buf_pool_from_bpage(bpage); + ulint zip_size; ut_ad(buf_pool); ut_ad(bpage); @@ -1391,6 +1409,9 @@ buf_LRU_remove_block( UT_LIST_REMOVE(LRU, buf_pool->LRU, bpage); ut_d(bpage->in_LRU_list = FALSE); + zip_size = page_zip_get_size(&bpage->zip); + buf_pool->stat.LRU_bytes -= zip_size ? zip_size : UNIV_PAGE_SIZE; + buf_unzip_LRU_remove_block_if_needed(bpage); /* If the LRU list is so short that LRU_old is not defined, @@ -1451,7 +1472,10 @@ buf_unzip_LRU_add_block( } /******************************************************************//** -Adds a block to the LRU list end. */ +Adds a block to the LRU list end. Please make sure that the zip_size is +already set into the page zip when invoking the function, so that we +can get correct zip_size from the buffer page when adding a block +into LRU */ UNIV_INLINE void buf_LRU_add_block_to_end_low( @@ -1470,6 +1494,8 @@ buf_LRU_add_block_to_end_low( UT_LIST_ADD_LAST(LRU, buf_pool->LRU, bpage); ut_d(bpage->in_LRU_list = TRUE); + incr_LRU_size_in_bytes(bpage, buf_pool); + if (UT_LIST_GET_LEN(buf_pool->LRU) > BUF_LRU_OLD_MIN_LEN) { ut_ad(buf_pool->LRU_old); @@ -1498,7 +1524,10 @@ buf_LRU_add_block_to_end_low( } /******************************************************************//** -Adds a block to the LRU list. */ +Adds a block to the LRU list. Please make sure that the zip_size is +already set into the page zip when invoking the function, so that we +can get correct zip_size from the buffer page when adding a block +into LRU */ UNIV_INLINE void buf_LRU_add_block_low( @@ -1540,6 +1569,8 @@ buf_LRU_add_block_low( ut_d(bpage->in_LRU_list = TRUE); + incr_LRU_size_in_bytes(bpage, buf_pool); + if (UT_LIST_GET_LEN(buf_pool->LRU) > BUF_LRU_OLD_MIN_LEN) { ut_ad(buf_pool->LRU_old); @@ -1567,7 +1598,10 @@ buf_LRU_add_block_low( } /******************************************************************//** -Adds a block to the LRU list. */ +Adds a block to the LRU list. Please make sure that the zip_size is +already set into the page zip when invoking the function, so that we +can get correct zip_size from the buffer page when adding a block +into LRU */ UNIV_INTERN void buf_LRU_add_block( @@ -1753,6 +1787,8 @@ alloc: UT_LIST_INSERT_AFTER(LRU, buf_pool->LRU, prev_b, b); + incr_LRU_size_in_bytes(b, buf_pool); + if (buf_page_is_old(b)) { buf_pool->LRU_old_len++; if (UNIV_UNLIKELY diff --git a/storage/innobase/dict/dict0load.c b/storage/innobase/dict/dict0load.c index 22f0d4456d5..208d013e82d 100644 --- a/storage/innobase/dict/dict0load.c +++ b/storage/innobase/dict/dict0load.c @@ -2328,7 +2328,8 @@ dict_load_foreigns( ibool check_charsets) /*!< in: TRUE=check charset compatibility */ { - char tuple_buf[DTUPLE_EST_ALLOC(1)]; + ulint tuple_buf[(DTUPLE_EST_ALLOC(1) + sizeof(ulint) - 1) + / sizeof(ulint)]; btr_pcur_t pcur; dtuple_t* tuple; dfield_t* dfield; diff --git a/storage/innobase/fil/fil0fil.c b/storage/innobase/fil/fil0fil.c index 8fe33459994..2f875663039 100644 --- a/storage/innobase/fil/fil0fil.c +++ b/storage/innobase/fil/fil0fil.c @@ -1466,7 +1466,7 @@ fil_space_get_size( ut_ad(fil_system); - fil_mutex_enter_and_prepare_for_io(id); + mutex_enter(&fil_system->mutex); space = fil_space_get_by_id(id); @@ -1481,6 +1481,23 @@ fil_space_get_size( ut_a(1 == UT_LIST_GET_LEN(space->chain)); + mutex_exit(&fil_system->mutex); + + /* It is possible that the space gets evicted at this point + before the fil_mutex_enter_and_prepare_for_io() acquires + the fil_system->mutex. Check for this after completing the + call to fil_mutex_enter_and_prepare_for_io(). */ + fil_mutex_enter_and_prepare_for_io(id); + + /* We are still holding the fil_system->mutex. Check if + the space is still in memory cache. */ + space = fil_space_get_by_id(id); + + if (space == NULL) { + mutex_exit(&fil_system->mutex); + return(0); + } + node = UT_LIST_GET_FIRST(space->chain); /* It must be a single-table tablespace and we have not opened @@ -1518,7 +1535,7 @@ fil_space_get_flags( return(0); } - fil_mutex_enter_and_prepare_for_io(id); + mutex_enter(&fil_system->mutex); space = fil_space_get_by_id(id); @@ -1533,6 +1550,23 @@ fil_space_get_flags( ut_a(1 == UT_LIST_GET_LEN(space->chain)); + mutex_exit(&fil_system->mutex); + + /* It is possible that the space gets evicted at this point + before the fil_mutex_enter_and_prepare_for_io() acquires + the fil_system->mutex. Check for this after completing the + call to fil_mutex_enter_and_prepare_for_io(). */ + fil_mutex_enter_and_prepare_for_io(id); + + /* We are still holding the fil_system->mutex. Check if + the space is still in memory cache. */ + space = fil_space_get_by_id(id); + + if (space == NULL) { + mutex_exit(&fil_system->mutex); + return(0); + } + node = UT_LIST_GET_FIRST(space->chain); /* It must be a single-table tablespace and we have not opened @@ -2702,7 +2736,7 @@ retry: mutex_exit(&fil_system->mutex); #ifndef UNIV_HOTBACKUP - if (success) { + if (success && !recv_recovery_on) { mtr_t mtr; mtr_start(&mtr); @@ -4947,3 +4981,28 @@ fil_close(void) fil_system = NULL; } + +/****************************************************************//** +Generate redo logs for swapping two .ibd files */ +UNIV_INTERN +void +fil_mtr_rename_log( +/*===============*/ + ulint old_space_id, /*!< in: tablespace id of the old + table. */ + const char* old_name, /*!< in: old table name */ + ulint new_space_id, /*!< in: tablespace id of the new + table */ + const char* new_name, /*!< in: new table name */ + const char* tmp_name) /*!< in: temp table name used while + swapping */ +{ + mtr_t mtr; + mtr_start(&mtr); + fil_op_write_log(MLOG_FILE_RENAME, old_space_id, + 0, 0, old_name, tmp_name, &mtr); + fil_op_write_log(MLOG_FILE_RENAME, new_space_id, + 0, 0, new_name, old_name, &mtr); + mtr_commit(&mtr); +} + diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc index 7c036d3b601..8d433f68d78 100644 --- a/storage/innobase/handler/ha_innodb.cc +++ b/storage/innobase/handler/ha_innodb.cc @@ -603,8 +603,12 @@ innobase_commit_low( static SHOW_VAR innodb_status_variables[]= { {"buffer_pool_pages_data", (char*) &export_vars.innodb_buffer_pool_pages_data, SHOW_LONG}, + {"buffer_pool_bytes_data", + (char*) &export_vars.innodb_buffer_pool_bytes_data, SHOW_LONG}, {"buffer_pool_pages_dirty", (char*) &export_vars.innodb_buffer_pool_pages_dirty, SHOW_LONG}, + {"buffer_pool_bytes_dirty", + (char*) &export_vars.innodb_buffer_pool_bytes_dirty, SHOW_LONG}, {"buffer_pool_pages_flushed", (char*) &export_vars.innodb_buffer_pool_pages_flushed, SHOW_LONG}, {"buffer_pool_pages_free", @@ -695,6 +699,12 @@ static SHOW_VAR innodb_status_variables[]= { (char*) &export_vars.innodb_rows_updated, SHOW_LONG}, {"truncated_status_writes", (char*) &export_vars.innodb_truncated_status_writes, SHOW_LONG}, +#ifdef UNIV_DEBUG + {"purge_trx_id_age", + (char*) &export_vars.innodb_purge_trx_id_age, SHOW_LONG}, + {"purge_view_trx_id_age", + (char*) &export_vars.innodb_purge_view_trx_id_age, SHOW_LONG}, +#endif /* UNIV_DEBUG */ {NullS, NullS, SHOW_LONG} }; @@ -1064,6 +1074,8 @@ convert_error_code_to_mysql( return(HA_ERR_INDEX_CORRUPT); case DB_UNDO_RECORD_TOO_BIG: return(HA_ERR_UNDO_REC_TOO_BIG); + case DB_OUT_OF_MEMORY: + return(HA_ERR_OUT_OF_MEM); } } @@ -1241,16 +1253,6 @@ innobase_get_lower_case_table_names(void) return(lower_case_table_names); } -#if defined (__WIN__) && defined (MYSQL_DYNAMIC_PLUGIN) -extern MYSQL_PLUGIN_IMPORT MY_TMPDIR mysql_tmpdir_list; -/*******************************************************************//** -Map an OS error to an errno value. The OS error number is stored in -_doserrno and the mapped value is stored in errno) */ -extern "C" -void __cdecl -_dosmaperr( - unsigned long); /*!< in: OS error value */ - /*********************************************************************//** Creates a temporary file. @return temporary file descriptor, or < 0 on error */ @@ -1259,92 +1261,16 @@ int innobase_mysql_tmpfile(void) /*========================*/ { - int fd; /* handle of opened file */ - HANDLE osfh; /* OS handle of opened file */ - char* tmpdir; /* point to the directory - where to create file */ - TCHAR path_buf[MAX_PATH - 14]; /* buffer for tmp file path. - The length cannot be longer - than MAX_PATH - 14, or - GetTempFileName will fail. */ - char filename[MAX_PATH]; /* name of the tmpfile */ - DWORD fileaccess = GENERIC_READ /* OS file access */ - | GENERIC_WRITE - | DELETE; - DWORD fileshare = FILE_SHARE_READ /* OS file sharing mode */ - | FILE_SHARE_WRITE - | FILE_SHARE_DELETE; - DWORD filecreate = CREATE_ALWAYS; /* OS method of open/create */ - DWORD fileattrib = /* OS file attribute flags */ - FILE_ATTRIBUTE_NORMAL - | FILE_FLAG_DELETE_ON_CLOSE - | FILE_ATTRIBUTE_TEMPORARY - | FILE_FLAG_SEQUENTIAL_SCAN; - - DBUG_ENTER("innobase_mysql_tmpfile"); - - tmpdir = my_tmpdir(&mysql_tmpdir_list); - - /* The tmpdir parameter can not be NULL for GetTempFileName. */ - if (!tmpdir) { - uint ret; - - /* Use GetTempPath to determine path for temporary files. */ - ret = GetTempPath(sizeof(path_buf), path_buf); - if (ret > sizeof(path_buf) || (ret == 0)) { - - _dosmaperr(GetLastError()); /* map error */ - DBUG_RETURN(-1); - } - - tmpdir = path_buf; - } - - /* Use GetTempFileName to generate a unique filename. */ - if (!GetTempFileName(tmpdir, "ib", 0, filename)) { - - _dosmaperr(GetLastError()); /* map error */ - DBUG_RETURN(-1); - } - - DBUG_PRINT("info", ("filename: %s", filename)); - - /* Open/Create the file. */ - osfh = CreateFile(filename, fileaccess, fileshare, NULL, - filecreate, fileattrib, NULL); - if (osfh == INVALID_HANDLE_VALUE) { - - /* open/create file failed! */ - _dosmaperr(GetLastError()); /* map error */ - DBUG_RETURN(-1); - } - - do { - /* Associates a CRT file descriptor with the OS file handle. */ - fd = _open_osfhandle((intptr_t) osfh, 0); - } while (fd == -1 && errno == EINTR); + int fd2 = -1; + File fd; - if (fd == -1) { - /* Open failed, close the file handle. */ + DBUG_EXECUTE_IF( + "innobase_tmpfile_creation_failure", + return(-1); + ); - _dosmaperr(GetLastError()); /* map error */ - CloseHandle(osfh); /* no need to check if - CloseHandle fails */ - } + fd = mysql_tmpfile("ib"); - DBUG_RETURN(fd); -} -#else -/*********************************************************************//** -Creates a temporary file. -@return temporary file descriptor, or < 0 on error */ -extern "C" UNIV_INTERN -int -innobase_mysql_tmpfile(void) -/*========================*/ -{ - int fd2 = -1; - File fd = mysql_tmpfile("ib"); if (fd >= 0) { /* Copy the file descriptor, so that the additional resources allocated by create_temp_file() can be freed by invoking @@ -1388,7 +1314,6 @@ innobase_mysql_tmpfile(void) } return(fd2); } -#endif /* defined (__WIN__) && defined (MYSQL_DYNAMIC_PLUGIN) */ /*********************************************************************//** Wrapper around MySQL's copy_and_convert function. @@ -11773,13 +11698,30 @@ static MYSQL_SYSVAR_ULONG(read_ahead_threshold, srv_read_ahead_threshold, "trigger a readahead.", NULL, NULL, 56, 0, 64, 0); -#ifdef UNIV_DEBUG_never +#ifdef UNIV_DEBUG static MYSQL_SYSVAR_UINT(trx_rseg_n_slots_debug, trx_rseg_n_slots_debug, - PLUGIN_VAR_RQCMDARG, + PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_NOCMDOPT, "Debug flags for InnoDB to limit TRX_RSEG_N_SLOTS for trx_rsegf_undo_find_free()", NULL, NULL, 0, 0, 1024, 0); + +static MYSQL_SYSVAR_UINT(limit_optimistic_insert_debug, + btr_cur_limit_optimistic_insert_debug, PLUGIN_VAR_RQCMDARG, + "Artificially limit the number of records per B-tree page (0=unlimited).", + NULL, NULL, 0, 0, UINT_MAX32, 0); + +static MYSQL_SYSVAR_BOOL(trx_purge_view_update_only_debug, + srv_purge_view_update_only_debug, PLUGIN_VAR_NOCMDOPT, + "Pause actual purging any delete-marked records, but merely update the purge view. " + "It is to create artificially the situation the purge view have been updated " + "but the each purges were not done yet.", + NULL, NULL, FALSE); #endif /* UNIV_DEBUG */ +static MYSQL_SYSVAR_BOOL(print_all_deadlocks, srv_print_all_deadlocks, + PLUGIN_VAR_OPCMDARG, + "Print all deadlocks to MySQL error log (off by default)", + NULL, NULL, FALSE); + static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(additional_mem_pool_size), MYSQL_SYSVAR(autoextend_increment), @@ -11849,9 +11791,12 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(purge_threads), MYSQL_SYSVAR(purge_batch_size), MYSQL_SYSVAR(rollback_segments), -#ifdef UNIV_DEBUG_never /* disable this flag. --innodb-trx becomes ambiguous */ +#ifdef UNIV_DEBUG MYSQL_SYSVAR(trx_rseg_n_slots_debug), + MYSQL_SYSVAR(limit_optimistic_insert_debug), + MYSQL_SYSVAR(trx_purge_view_update_only_debug), #endif /* UNIV_DEBUG */ + MYSQL_SYSVAR(print_all_deadlocks), NULL }; diff --git a/storage/innobase/handler/handler0alter.cc b/storage/innobase/handler/handler0alter.cc index 451e37f2136..cdd5078c262 100644 --- a/storage/innobase/handler/handler0alter.cc +++ b/storage/innobase/handler/handler0alter.cc @@ -101,8 +101,6 @@ innobase_col_to_mysql( ut_ad(flen >= len); ut_ad(DATA_MBMAXLEN(col->mbminmaxlen) >= DATA_MBMINLEN(col->mbminmaxlen)); - ut_ad(DATA_MBMAXLEN(col->mbminmaxlen) - > DATA_MBMINLEN(col->mbminmaxlen) || flen == len); memcpy(dest, data, len); break; diff --git a/storage/innobase/include/btr0cur.h b/storage/innobase/include/btr0cur.h index 6819f2bc2c5..283a6eec852 100644 --- a/storage/innobase/include/btr0cur.h +++ b/storage/innobase/include/btr0cur.h @@ -805,6 +805,11 @@ srv_printf_innodb_monitor(). */ extern ulint btr_cur_n_sea_old; #endif /* !UNIV_HOTBACKUP */ +#ifdef UNIV_DEBUG +/* Flag to limit optimistic insert records */ +extern uint btr_cur_limit_optimistic_insert_debug; +#endif /* UNIV_DEBUG */ + #ifndef UNIV_NONINL #include "btr0cur.ic" #endif diff --git a/storage/innobase/include/btr0cur.ic b/storage/innobase/include/btr0cur.ic index e31f77c77eb..5fc4651ca13 100644 --- a/storage/innobase/include/btr0cur.ic +++ b/storage/innobase/include/btr0cur.ic @@ -27,6 +27,16 @@ Created 10/16/1994 Heikki Tuuri #include "btr0btr.h" #ifdef UNIV_DEBUG +# define LIMIT_OPTIMISTIC_INSERT_DEBUG(NREC, CODE)\ +if (btr_cur_limit_optimistic_insert_debug\ + && (NREC) >= (ulint)btr_cur_limit_optimistic_insert_debug) {\ + CODE;\ +} +#else +# define LIMIT_OPTIMISTIC_INSERT_DEBUG(NREC, CODE) +#endif /* UNIV_DEBUG */ + +#ifdef UNIV_DEBUG /*********************************************************//** Returns the page cursor component of a tree cursor. @return pointer to page cursor component */ @@ -146,6 +156,9 @@ btr_cur_compress_recommendation( page = btr_cur_get_page(cursor); + LIMIT_OPTIMISTIC_INSERT_DEBUG(page_get_n_recs(page) * 2, + return(FALSE)); + if ((page_get_data_size(page) < BTR_CUR_PAGE_COMPRESS_LIMIT) || ((btr_page_get_next(page, mtr) == FIL_NULL) && (btr_page_get_prev(page, mtr) == FIL_NULL))) { diff --git a/storage/innobase/include/buf0buf.h b/storage/innobase/include/buf0buf.h index fd8620eab3b..b7f6eae3f09 100644 --- a/storage/innobase/include/buf0buf.h +++ b/storage/innobase/include/buf0buf.h @@ -198,6 +198,15 @@ struct buf_pool_info_struct{ typedef struct buf_pool_info_struct buf_pool_info_t; +/** The occupied bytes of lists in all buffer pools */ +struct buf_pools_list_size_struct { + ulint LRU_bytes; /*!< LRU size in bytes */ + ulint unzip_LRU_bytes; /*!< unzip_LRU size in bytes */ + ulint flush_list_bytes; /*!< flush_list size in bytes */ +}; + +typedef struct buf_pools_list_size_struct buf_pools_list_size_t; + #ifndef UNIV_HOTBACKUP /********************************************************************//** Acquire mutex on all buffer pool instances */ @@ -1024,8 +1033,7 @@ UNIV_INLINE void buf_page_set_accessed( /*==================*/ - buf_page_t* bpage, /*!< in/out: control block */ - ulint time_ms) /*!< in: ut_time_ms() */ + buf_page_t* bpage) /*!< in/out: control block */ __attribute__((nonnull)); /*********************************************************************//** Gets the buf_block_t handle of a buffered file block if an uncompressed @@ -1334,6 +1342,14 @@ buf_get_total_list_len( ulint* free_len, /*!< out: length of all free lists */ ulint* flush_list_len);/*!< out: length of all flush lists */ /********************************************************************//** +Get total list size in bytes from all buffer pools. */ +UNIV_INTERN +void +buf_get_total_list_size_in_bytes( +/*=============================*/ + buf_pools_list_size_t* buf_pools_list_size); /*!< out: list sizes + in all buffer pools */ +/********************************************************************//** Get total buffer pool statistics. */ UNIV_INTERN void @@ -1504,10 +1520,11 @@ struct buf_page_struct{ to read this for heuristic purposes without holding any mutex or latch */ - unsigned access_time:32; /*!< time of first access, or - 0 if the block was never accessed - in the buffer pool */ /* @} */ + unsigned access_time; /*!< time of first access, or + 0 if the block was never accessed + in the buffer pool. Protected by + block mutex */ # if defined UNIV_DEBUG_FILE_ACCESSES || defined UNIV_DEBUG ibool file_page_was_freed; /*!< this is set to TRUE when fsp @@ -1685,6 +1702,8 @@ struct buf_pool_stat_struct{ young because the first access was not long enough ago, in buf_page_peek_if_too_old() */ + ulint LRU_bytes; /*!< LRU size in bytes */ + ulint flush_list_bytes;/*!< flush_list size in bytes */ }; /** Statistics of buddy blocks of a given size. */ diff --git a/storage/innobase/include/buf0buf.ic b/storage/innobase/include/buf0buf.ic index 84c0ab9d610..bcbe454b037 100644 --- a/storage/innobase/include/buf0buf.ic +++ b/storage/innobase/include/buf0buf.ic @@ -614,18 +614,18 @@ UNIV_INLINE void buf_page_set_accessed( /*==================*/ - buf_page_t* bpage, /*!< in/out: control block */ - ulint time_ms) /*!< in: ut_time_ms() */ + buf_page_t* bpage) /*!< in/out: control block */ { #ifdef UNIV_DEBUG buf_pool_t* buf_pool = buf_pool_from_bpage(bpage); - ut_ad(buf_pool_mutex_own(buf_pool)); + ut_ad(!buf_pool_mutex_own(buf_pool)); + ut_ad(mutex_own(buf_page_get_mutex(bpage))); #endif ut_a(buf_page_in_file(bpage)); if (!bpage->access_time) { /* Make this the time of the first access. */ - bpage->access_time = time_ms; + bpage->access_time = ut_time_ms(); } } diff --git a/storage/innobase/include/buf0lru.h b/storage/innobase/include/buf0lru.h index 9ecb9de2afe..0e08f455800 100644 --- a/storage/innobase/include/buf0lru.h +++ b/storage/innobase/include/buf0lru.h @@ -149,7 +149,10 @@ buf_LRU_block_free_non_file_page( /*=============================*/ buf_block_t* block); /*!< in: block, must not contain a file page */ /******************************************************************//** -Adds a block to the LRU list. */ +Adds a block to the LRU list. Please make sure that the zip_size is +already set into the page zip when invoking the function, so that we +can get correct zip_size from the buffer page when adding a block +into LRU */ UNIV_INTERN void buf_LRU_add_block( diff --git a/storage/innobase/include/data0type.ic b/storage/innobase/include/data0type.ic index 757dd815c5e..7ec2cb6cf36 100644 --- a/storage/innobase/include/data0type.ic +++ b/storage/innobase/include/data0type.ic @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2010, Innobase Oy. All Rights Reserved. +Copyright (c) 1996, 2012, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -459,36 +459,18 @@ dtype_get_fixed_size_low( } else if (!comp) { return(len); } else { - /* We play it safe here and ask MySQL for - mbminlen and mbmaxlen. Although - mbminlen and mbmaxlen are - initialized if and only if prtype - is (in one of the 3 functions in this file), - it could be that none of these functions - has been called. */ - +#ifdef UNIV_DEBUG ulint i_mbminlen, i_mbmaxlen; innobase_get_cset_width( dtype_get_charset_coll(prtype), &i_mbminlen, &i_mbmaxlen); - if (UNIV_UNLIKELY - (DATA_MBMINMAXLEN(i_mbminlen, i_mbmaxlen) - != mbminmaxlen)) { - - ut_print_timestamp(stderr); - fprintf(stderr, " InnoDB: " - "mbminlen=%lu, " - "mbmaxlen=%lu, " - "type->mbminlen=%lu, " - "type->mbmaxlen=%lu\n", - (ulong) i_mbminlen, - (ulong) i_mbmaxlen, - (ulong) DATA_MBMINLEN(mbminmaxlen), - (ulong) DATA_MBMAXLEN(mbminmaxlen)); - } - if (i_mbminlen == i_mbmaxlen) { + ut_ad(DATA_MBMINMAXLEN(i_mbminlen, i_mbmaxlen) + == mbminmaxlen); +#endif /* UNIV_DEBUG */ + if (DATA_MBMINLEN(mbminmaxlen) + == DATA_MBMAXLEN(mbminmaxlen)) { return(len); } } diff --git a/storage/innobase/include/fil0fil.h b/storage/innobase/include/fil0fil.h index ed47ccff052..a804c261447 100644 --- a/storage/innobase/include/fil0fil.h +++ b/storage/innobase/include/fil0fil.h @@ -734,6 +734,21 @@ fil_tablespace_is_being_deleted( /*============================*/ ulint id); /*!< in: space id */ +/****************************************************************//** +Generate redo logs for swapping two .ibd files */ +UNIV_INTERN +void +fil_mtr_rename_log( +/*===============*/ + ulint old_space_id, /*!< in: tablespace id of the old + table. */ + const char* old_name, /*!< in: old table name */ + ulint new_space_id, /*!< in: tablespace id of the new + table */ + const char* new_name, /*!< in: new table name */ + const char* tmp_name); /*!< in: temp table name used while + swapping */ + typedef struct fil_space_struct fil_space_t; #endif diff --git a/storage/innobase/include/lock0lock.h b/storage/innobase/include/lock0lock.h index 29fdc3bbe97..b2a622296e0 100644 --- a/storage/innobase/include/lock0lock.h +++ b/storage/innobase/include/lock0lock.h @@ -797,14 +797,22 @@ lock_rec_get_page_no( remains set when the waiting lock is granted, or if the lock is inherited to a neighboring record */ -#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION)&LOCK_MODE_MASK +#define LOCK_CONV_BY_OTHER 4096 /*!< this bit is set when the lock is created + by other transaction */ +#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION|LOCK_CONV_BY_OTHER)&LOCK_MODE_MASK # error #endif -#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION)&LOCK_TYPE_MASK +#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION|LOCK_CONV_BY_OTHER)&LOCK_TYPE_MASK # error #endif /* @} */ +/** Checks if this is a waiting lock created by lock->trx itself. +@param type_mode lock->type_mode +@return whether it is a waiting lock belonging to lock->trx */ +#define lock_is_wait_not_by_other(type_mode) \ + ((type_mode & (LOCK_CONV_BY_OTHER | LOCK_WAIT)) == LOCK_WAIT) + /** Lock operation struct */ typedef struct lock_op_struct lock_op_t; /** Lock operation struct */ diff --git a/storage/innobase/include/rem0rec.h b/storage/innobase/include/rem0rec.h index 10b74d18c13..98bf889b996 100644 --- a/storage/innobase/include/rem0rec.h +++ b/storage/innobase/include/rem0rec.h @@ -362,24 +362,6 @@ rec_get_offsets_func( rec_get_offsets_func(rec,index,offsets,n,heap,__FILE__,__LINE__) /******************************************************//** -Determine the offset to each field in a leaf-page record -in ROW_FORMAT=COMPACT. This is a special case of -rec_init_offsets() and rec_get_offsets_func(). */ -UNIV_INTERN -void -rec_init_offsets_comp_ordinary( -/*===========================*/ - const rec_t* rec, /*!< in: physical record in - ROW_FORMAT=COMPACT */ - ulint extra, /*!< in: number of bytes to reserve - between the record header and - the data payload - (usually REC_N_NEW_EXTRA_BYTES) */ - const dict_index_t* index, /*!< in: record descriptor */ - ulint* offsets);/*!< in/out: array of offsets; - in: n=rec_offs_n_fields(offsets) */ - -/******************************************************//** The following function determines the offsets to each field in the record. It can reuse a previously allocated array. */ UNIV_INTERN @@ -644,8 +626,48 @@ rec_copy( /*=====*/ void* buf, /*!< in: buffer */ const rec_t* rec, /*!< in: physical record */ - const ulint* offsets);/*!< in: array returned by rec_get_offsets() */ + const ulint* offsets)/*!< in: array returned by rec_get_offsets() */ + __attribute__((nonnull)); #ifndef UNIV_HOTBACKUP +/**********************************************************//** +Determines the size of a data tuple prefix in a temporary file. +@return total size */ +UNIV_INTERN +ulint +rec_get_converted_size_temp( +/*========================*/ + const dict_index_t* index, /*!< in: record descriptor */ + const dfield_t* fields, /*!< in: array of data fields */ + ulint n_fields,/*!< in: number of data fields */ + ulint* extra) /*!< out: extra size */ + __attribute__((warn_unused_result, nonnull)); + +/******************************************************//** +Determine the offset to each field in temporary file. +@see rec_convert_dtuple_to_temp() */ +UNIV_INTERN +void +rec_init_offsets_temp( +/*==================*/ + const rec_t* rec, /*!< in: temporary file record */ + const dict_index_t* index, /*!< in: record descriptor */ + ulint* offsets)/*!< in/out: array of offsets; + in: n=rec_offs_n_fields(offsets) */ + __attribute__((nonnull)); + +/*********************************************************//** +Builds a temporary file record out of a data tuple. +@see rec_init_offsets_temp() */ +UNIV_INTERN +void +rec_convert_dtuple_to_temp( +/*=======================*/ + rec_t* rec, /*!< out: record */ + const dict_index_t* index, /*!< in: record descriptor */ + const dfield_t* fields, /*!< in: array of data fields */ + ulint n_fields) /*!< in: number of fields */ + __attribute__((nonnull)); + /**************************************************************//** Copies the first n fields of a physical record to a new physical record in a buffer. @@ -680,21 +702,6 @@ rec_fold( __attribute__((pure)); #endif /* !UNIV_HOTBACKUP */ /*********************************************************//** -Builds a ROW_FORMAT=COMPACT record out of a data tuple. */ -UNIV_INTERN -void -rec_convert_dtuple_to_rec_comp( -/*===========================*/ - rec_t* rec, /*!< in: origin of record */ - ulint extra, /*!< in: number of bytes to - reserve between the record - header and the data payload - (normally REC_N_NEW_EXTRA_BYTES) */ - const dict_index_t* index, /*!< in: record descriptor */ - ulint status, /*!< in: status bits of the record */ - const dfield_t* fields, /*!< in: array of data fields */ - ulint n_fields);/*!< in: number of data fields */ -/*********************************************************//** Builds a physical record out of a data tuple and stores it into the given buffer. @return pointer to the origin of physical record */ @@ -727,10 +734,7 @@ UNIV_INTERN ulint rec_get_converted_size_comp_prefix( /*===============================*/ - const dict_index_t* index, /*!< in: record descriptor; - dict_table_is_comp() is - assumed to hold, even if - it does not */ + const dict_index_t* index, /*!< in: record descriptor */ const dfield_t* fields, /*!< in: array of data fields */ ulint n_fields,/*!< in: number of data fields */ ulint* extra); /*!< out: extra size */ diff --git a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h index 5290ef5fe11..e2d061969f3 100644 --- a/storage/innobase/include/srv0srv.h +++ b/storage/innobase/include/srv0srv.h @@ -265,6 +265,10 @@ extern ulint srv_fatal_semaphore_wait_threshold; #define SRV_SEMAPHORE_WAIT_EXTENSION 7200 extern ulint srv_dml_needed_delay; +#ifdef UNIV_DEBUG +extern my_bool srv_purge_view_update_only_debug; +#endif /* UNIV_DEBUG */ + extern mutex_t* kernel_mutex_temp;/* mutex protecting the server, trx structs, query threads, and lock table: we allocate it from dynamic memory to get it to the @@ -333,6 +337,9 @@ extern ulint srv_buf_pool_flushed; reading of a disk page */ extern ulint srv_buf_pool_reads; +/** print all user-level transactions deadlocks to mysqld stderr */ +extern my_bool srv_print_all_deadlocks; + /** Status variables to be passed to MySQL */ typedef struct export_var_struct export_struc; @@ -700,7 +707,9 @@ struct export_var_struct{ ulint innodb_data_reads; /*!< I/O read requests */ ulint innodb_buffer_pool_pages_total; /*!< Buffer pool size */ ulint innodb_buffer_pool_pages_data; /*!< Data pages */ + ulint innodb_buffer_pool_bytes_data; /*!< File bytes used */ ulint innodb_buffer_pool_pages_dirty; /*!< Dirty data pages */ + ulint innodb_buffer_pool_bytes_dirty; /*!< File bytes modified */ ulint innodb_buffer_pool_pages_misc; /*!< Miscellanous pages */ ulint innodb_buffer_pool_pages_free; /*!< Free pages */ #ifdef UNIV_DEBUG @@ -742,6 +751,11 @@ struct export_var_struct{ ulint innodb_rows_updated; /*!< srv_n_rows_updated */ ulint innodb_rows_deleted; /*!< srv_n_rows_deleted */ ulint innodb_truncated_status_writes; /*!< srv_truncated_status_writes */ +#ifdef UNIV_DEBUG + ulint innodb_purge_trx_id_age; /*!< max_trx_id - purged trx_id */ + ulint innodb_purge_view_trx_id_age; /*!< rw_max_trx_id + - purged view's min trx_id */ +#endif /* UNIV_DEBUG */ }; /** Thread slot in the thread table */ diff --git a/storage/innobase/include/trx0purge.h b/storage/innobase/include/trx0purge.h index 2bd9e64476b..1e8acd65e01 100644 --- a/storage/innobase/include/trx0purge.h +++ b/storage/innobase/include/trx0purge.h @@ -156,6 +156,10 @@ struct trx_purge_struct{ than this */ undo_no_t purge_undo_no; /*!< Purge has advanced past all records whose undo number is less than this */ +#ifdef UNIV_DEBUG + trx_id_t done_trx_no; /* Indicate 'purge pointer' which have + purged already accurately. */ +#endif /* UNIV_DEBUG */ /*-----------------------------*/ ibool next_stored; /*!< TRUE if the info of the next record to purge is stored below: if yes, then diff --git a/storage/innobase/include/univ.i b/storage/innobase/include/univ.i index 0b0b4b8d04c..800f4359c8a 100644 --- a/storage/innobase/include/univ.i +++ b/storage/innobase/include/univ.i @@ -49,9 +49,10 @@ Created 1/20/1994 Heikki Tuuri #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 8 +#include <mysql_version.h> + +#define INNODB_VERSION_MAJOR MYSQL_MAJOR_VERSION +#define INNODB_VERSION_MINOR MYSQL_MINOR_VERSION /* The following is the InnoDB version as shown in SELECT plugin_version FROM information_schema.plugins; @@ -62,10 +63,7 @@ component, i.e. we show M.N.P as M.N */ #define INNODB_VERSION_SHORT \ (INNODB_VERSION_MAJOR << 8 | INNODB_VERSION_MINOR) -#define INNODB_VERSION_STR \ - IB_TO_STR(INNODB_VERSION_MAJOR) "." \ - IB_TO_STR(INNODB_VERSION_MINOR) "." \ - IB_TO_STR(INNODB_VERSION_BUGFIX) +#define INNODB_VERSION_STR MYSQL_SERVER_VERSION #define REFMAN "http://dev.mysql.com/doc/refman/" \ IB_TO_STR(MYSQL_MAJOR_VERSION) "." \ diff --git a/storage/innobase/lock/lock0lock.c b/storage/innobase/lock/lock0lock.c index 66ec18aeee7..a72be540c1b 100644 --- a/storage/innobase/lock/lock0lock.c +++ b/storage/innobase/lock/lock0lock.c @@ -789,12 +789,16 @@ lock_reset_lock_and_trx_wait( /*=========================*/ lock_t* lock) /*!< in: record lock */ { - ut_ad((lock->trx)->wait_lock == lock); ut_ad(lock_get_wait(lock)); /* Reset the back pointer in trx to this waiting lock request */ - (lock->trx)->wait_lock = NULL; + if (!(lock->type_mode & LOCK_CONV_BY_OTHER)) { + ut_ad((lock->trx)->wait_lock == lock); + (lock->trx)->wait_lock = NULL; + } else { + ut_ad(lock_get_type_low(lock) == LOCK_REC); + } lock->type_mode &= ~LOCK_WAIT; } @@ -1430,9 +1434,9 @@ lock_rec_has_expl( while (lock) { if (lock->trx == trx + && !lock_is_wait_not_by_other(lock->type_mode) && lock_mode_stronger_or_eq(lock_get_mode(lock), precise_mode & LOCK_MODE_MASK) - && !lock_get_wait(lock) && (!lock_rec_get_rec_not_gap(lock) || (precise_mode & LOCK_REC_NOT_GAP) || heap_no == PAGE_HEAP_NO_SUPREMUM) @@ -1730,7 +1734,7 @@ lock_rec_create( HASH_INSERT(lock_t, hash, lock_sys->rec_hash, lock_rec_fold(space, page_no), lock); - if (UNIV_UNLIKELY(type_mode & LOCK_WAIT)) { + if (lock_is_wait_not_by_other(type_mode)) { lock_set_lock_and_trx_wait(lock, trx); } @@ -1761,10 +1765,11 @@ lock_rec_enqueue_waiting( const buf_block_t* block, /*!< in: buffer block containing the record */ ulint heap_no,/*!< in: heap number of the record */ + lock_t* lock, /*!< in: lock object; NULL if a new + one should be created. */ dict_index_t* index, /*!< in: index of record */ que_thr_t* thr) /*!< in: query thread */ { - lock_t* lock; trx_t* trx; ut_ad(mutex_own(&kernel_mutex)); @@ -1799,9 +1804,17 @@ lock_rec_enqueue_waiting( ut_ad(0); } - /* Enqueue the lock request that will wait to be granted */ - lock = lock_rec_create(type_mode | LOCK_WAIT, - block, heap_no, index, trx); + if (lock == NULL) { + /* Enqueue the lock request that will wait to be granted */ + lock = lock_rec_create(type_mode | LOCK_WAIT, + block, heap_no, index, trx); + } else { + ut_ad(lock->type_mode & LOCK_WAIT); + ut_ad(lock->type_mode & LOCK_CONV_BY_OTHER); + + lock->type_mode &= ~LOCK_CONV_BY_OTHER; + lock_set_lock_and_trx_wait(lock, trx); + } /* Check if a deadlock occurs: if yes, remove the lock request and return an error code */ @@ -2046,6 +2059,7 @@ lock_rec_lock_slow( que_thr_t* thr) /*!< in: query thread */ { trx_t* trx; + lock_t* lock; ut_ad(mutex_own(&kernel_mutex)); ut_ad((LOCK_MODE_MASK & mode) != LOCK_S @@ -2060,7 +2074,27 @@ lock_rec_lock_slow( trx = thr_get_trx(thr); - if (lock_rec_has_expl(mode, block, heap_no, trx)) { + lock = lock_rec_has_expl(mode, block, heap_no, trx); + if (lock) { + if (lock->type_mode & LOCK_CONV_BY_OTHER) { + /* This lock or lock waiting was created by the other + transaction, not by the transaction (trx) itself. + So, the transaction (trx) should treat it collectly + according as whether granted or not. */ + + if (lock->type_mode & LOCK_WAIT) { + /* This lock request was not granted yet. + Should wait for granted. */ + + goto enqueue_waiting; + } else { + /* This lock request was already granted. + Just clearing the flag. */ + + lock->type_mode &= ~LOCK_CONV_BY_OTHER; + } + } + /* The trx already has a strong enough lock on rec: do nothing */ @@ -2070,8 +2104,10 @@ lock_rec_lock_slow( the queue, as this transaction does not have a lock strong enough already granted on the record, we have to wait. */ + ut_ad(lock == NULL); +enqueue_waiting: return(lock_rec_enqueue_waiting(mode, block, heap_no, - index, thr)); + lock, index, thr)); } else if (!impl) { /* Set the requested lock on the record */ @@ -2213,7 +2249,8 @@ lock_grant( TRX_QUE_LOCK_WAIT state, and there is no need to end the lock wait for it */ - if (lock->trx->que_state == TRX_QUE_LOCK_WAIT) { + if (!(lock->type_mode & LOCK_CONV_BY_OTHER) + && lock->trx->que_state == TRX_QUE_LOCK_WAIT) { trx_end_lock_wait(lock->trx); } } @@ -2230,6 +2267,7 @@ lock_rec_cancel( { ut_ad(mutex_own(&kernel_mutex)); ut_ad(lock_get_type_low(lock) == LOCK_REC); + ut_ad(!(lock->type_mode & LOCK_CONV_BY_OTHER)); /* Reset the bit (there can be only one set bit) in the lock bitmap */ lock_rec_reset_nth_bit(lock, lock_rec_find_set_bit(lock)); @@ -2372,8 +2410,12 @@ lock_rec_reset_and_release_wait( lock = lock_rec_get_first(block, heap_no); while (lock != NULL) { - if (lock_get_wait(lock)) { + if (lock_is_wait_not_by_other(lock->type_mode)) { lock_rec_cancel(lock); + } else if (lock_get_wait(lock)) { + /* just reset LOCK_WAIT */ + lock_rec_reset_nth_bit(lock, heap_no); + lock_reset_lock_and_trx_wait(lock); } else { lock_rec_reset_nth_bit(lock, heap_no); } @@ -3261,6 +3303,80 @@ lock_rec_restore_from_page_infimum( /*=========== DEADLOCK CHECKING ======================================*/ +/*********************************************************************//** +rewind(3) the file used for storing the latest detected deadlock and +print a heading message to stderr if printing of all deadlocks to stderr +is enabled. */ +UNIV_INLINE +void +lock_deadlock_start_print() +/*=======================*/ +{ + rewind(lock_latest_err_file); + ut_print_timestamp(lock_latest_err_file); + + if (srv_print_all_deadlocks) { + fprintf(stderr, "InnoDB: transactions deadlock detected, " + "dumping detailed information.\n"); + ut_print_timestamp(stderr); + } +} + +/*********************************************************************//** +Print a message to the deadlock file and possibly to stderr. */ +UNIV_INLINE +void +lock_deadlock_fputs( +/*================*/ + const char* msg) /*!< in: message to print */ +{ + fputs(msg, lock_latest_err_file); + + if (srv_print_all_deadlocks) { + fputs(msg, stderr); + } +} + +/*********************************************************************//** +Print transaction data to the deadlock file and possibly to stderr. */ +UNIV_INLINE +void +lock_deadlock_trx_print( +/*====================*/ + trx_t* trx, /*!< in: transaction */ + ulint max_query_len) /*!< in: max query length to print, or 0 to + use the default max length */ +{ + trx_print(lock_latest_err_file, trx, max_query_len); + + if (srv_print_all_deadlocks) { + trx_print(stderr, trx, max_query_len); + } +} + +/*********************************************************************//** +Print lock data to the deadlock file and possibly to stderr. */ +UNIV_INLINE +void +lock_deadlock_lock_print( +/*=====================*/ + const lock_t* lock) /*!< in: record or table type lock */ +{ + if (lock_get_type_low(lock) == LOCK_REC) { + lock_rec_print(lock_latest_err_file, lock); + + if (srv_print_all_deadlocks) { + lock_rec_print(stderr, lock); + } + } else { + lock_table_print(lock_latest_err_file, lock); + + if (srv_print_all_deadlocks) { + lock_table_print(stderr, lock); + } + } +} + /********************************************************************//** Checks if a lock request results in a deadlock. @return TRUE if a deadlock was detected and we chose trx as a victim; @@ -3304,30 +3420,25 @@ retry: /* If the lock search exceeds the max step or the max depth, the current trx will be the victim. Print its information. */ - rewind(lock_latest_err_file); - ut_print_timestamp(lock_latest_err_file); + lock_deadlock_start_print(); - fputs("TOO DEEP OR LONG SEARCH IN THE LOCK TABLE" - " WAITS-FOR GRAPH, WE WILL ROLL BACK" - " FOLLOWING TRANSACTION \n", - lock_latest_err_file); + lock_deadlock_fputs( + "TOO DEEP OR LONG SEARCH IN THE LOCK TABLE" + " WAITS-FOR GRAPH, WE WILL ROLL BACK" + " FOLLOWING TRANSACTION \n\n" + "*** TRANSACTION:\n"); - fputs("\n*** TRANSACTION:\n", lock_latest_err_file); - trx_print(lock_latest_err_file, trx, 3000); + lock_deadlock_trx_print(trx, 3000); - fputs("*** WAITING FOR THIS LOCK TO BE GRANTED:\n", - lock_latest_err_file); + lock_deadlock_fputs( + "*** WAITING FOR THIS LOCK TO BE GRANTED:\n"); + + lock_deadlock_lock_print(lock); - if (lock_get_type(lock) == LOCK_REC) { - lock_rec_print(lock_latest_err_file, lock); - } else { - lock_table_print(lock_latest_err_file, lock); - } break; case LOCK_VICTIM_IS_START: - fputs("*** WE ROLL BACK TRANSACTION (2)\n", - lock_latest_err_file); + lock_deadlock_fputs("*** WE ROLL BACK TRANSACTION (2)\n"); break; default: @@ -3442,45 +3553,33 @@ lock_deadlock_recursive( point: a deadlock detected; or we have searched the waits-for graph too long */ - FILE* ef = lock_latest_err_file; + lock_deadlock_start_print(); - rewind(ef); - ut_print_timestamp(ef); + lock_deadlock_fputs("\n*** (1) TRANSACTION:\n"); - fputs("\n*** (1) TRANSACTION:\n", ef); + lock_deadlock_trx_print(wait_lock->trx, 3000); - trx_print(ef, wait_lock->trx, 3000); + lock_deadlock_fputs( + "*** (1) WAITING FOR THIS LOCK" + " TO BE GRANTED:\n"); - fputs("*** (1) WAITING FOR THIS LOCK" - " TO BE GRANTED:\n", ef); + lock_deadlock_lock_print(wait_lock); - if (lock_get_type_low(wait_lock) == LOCK_REC) { - lock_rec_print(ef, wait_lock); - } else { - lock_table_print(ef, wait_lock); - } + lock_deadlock_fputs("*** (2) TRANSACTION:\n"); - fputs("*** (2) TRANSACTION:\n", ef); + lock_deadlock_trx_print(lock->trx, 3000); - trx_print(ef, lock->trx, 3000); + lock_deadlock_fputs( + "*** (2) HOLDS THE LOCK(S):\n"); - fputs("*** (2) HOLDS THE LOCK(S):\n", ef); + lock_deadlock_lock_print(lock); - if (lock_get_type_low(lock) == LOCK_REC) { - lock_rec_print(ef, lock); - } else { - lock_table_print(ef, lock); - } + lock_deadlock_fputs( + "*** (2) WAITING FOR THIS LOCK" + " TO BE GRANTED:\n"); - fputs("*** (2) WAITING FOR THIS LOCK" - " TO BE GRANTED:\n", ef); + lock_deadlock_lock_print(start->wait_lock); - if (lock_get_type_low(start->wait_lock) - == LOCK_REC) { - lock_rec_print(ef, start->wait_lock); - } else { - lock_table_print(ef, start->wait_lock); - } #ifdef UNIV_DEBUG if (lock_print_waits) { fputs("Deadlock detected\n", @@ -3503,8 +3602,8 @@ lock_deadlock_recursive( as a victim to try to avoid deadlocking our recursion starting point transaction */ - fputs("*** WE ROLL BACK TRANSACTION (1)\n", - ef); + lock_deadlock_fputs( + "*** WE ROLL BACK TRANSACTION (1)\n"); wait_lock->trx->was_chosen_as_deadlock_victim = TRUE; @@ -3589,6 +3688,7 @@ lock_table_create( ut_ad(table && trx); ut_ad(mutex_own(&kernel_mutex)); + ut_ad(!(type_mode & LOCK_CONV_BY_OTHER)); if ((type_mode & LOCK_MODE_MASK) == LOCK_AUTO_INC) { ++table->n_waiting_or_granted_auto_inc_locks; @@ -4142,6 +4242,7 @@ lock_cancel_waiting_and_release( lock_t* lock) /*!< in: waiting lock request */ { ut_ad(mutex_own(&kernel_mutex)); + ut_ad(!(lock->type_mode & LOCK_CONV_BY_OTHER)); if (lock_get_type_low(lock) == LOCK_REC) { @@ -5183,7 +5284,7 @@ lock_rec_insert_check_and_lock( err = lock_rec_enqueue_waiting(LOCK_X | LOCK_GAP | LOCK_INSERT_INTENTION, block, next_rec_heap_no, - index, thr); + NULL, index, thr); } else { err = DB_SUCCESS; } @@ -5259,10 +5360,23 @@ lock_rec_convert_impl_to_expl( if (!lock_rec_has_expl(LOCK_X | LOCK_REC_NOT_GAP, block, heap_no, impl_trx)) { + ulint type_mode = (LOCK_REC | LOCK_X + | LOCK_REC_NOT_GAP); + + /* If the delete-marked record was locked already, + we should reserve lock waiting for impl_trx as + implicit lock. Because cannot lock at this moment.*/ + + if (rec_get_deleted_flag(rec, rec_offs_comp(offsets)) + && lock_rec_other_has_conflicting( + LOCK_X | LOCK_REC_NOT_GAP, block, + heap_no, impl_trx)) { + + type_mode |= (LOCK_WAIT | LOCK_CONV_BY_OTHER); + } lock_rec_add_to_queue( - LOCK_REC | LOCK_X | LOCK_REC_NOT_GAP, - block, heap_no, index, impl_trx); + type_mode, block, heap_no, index, impl_trx); } } } diff --git a/storage/innobase/log/log0recv.c b/storage/innobase/log/log0recv.c index 03ba4d12cd0..c4ad94a4906 100644 --- a/storage/innobase/log/log0recv.c +++ b/storage/innobase/log/log0recv.c @@ -965,8 +965,11 @@ recv_parse_or_apply_log_rec_body( not NULL, then the log record is applied to the page, and the log record should be complete then */ - mtr_t* mtr) /*!< in: mtr or NULL; should be non-NULL + mtr_t* mtr, /*!< in: mtr or NULL; should be non-NULL if and only if block is non-NULL */ + ulint space_id) + /*!< in: tablespace id obtained by + parsing initial log record */ { dict_index_t* index = NULL; page_t* page; @@ -1238,8 +1241,11 @@ recv_parse_or_apply_log_rec_body( ut_ad(!page || page_type != FIL_PAGE_TYPE_ALLOCATED); ptr = mlog_parse_string(ptr, end_ptr, page, page_zip); break; - case MLOG_FILE_CREATE: case MLOG_FILE_RENAME: + ptr = fil_op_log_parse_or_replay(ptr, end_ptr, type, + space_id, 0); + break; + case MLOG_FILE_CREATE: case MLOG_FILE_DELETE: case MLOG_FILE_CREATE2: ptr = fil_op_log_parse_or_replay(ptr, end_ptr, type, 0, 0); @@ -1611,7 +1617,8 @@ recv_recover_page_func( recv_parse_or_apply_log_rec_body(recv->type, buf, buf + recv->len, - block, &mtr); + block, &mtr, + recv_addr->space); end_lsn = recv->start_lsn + recv->len; mach_write_to_8(FIL_PAGE_LSN + page, end_lsn); @@ -2078,7 +2085,7 @@ recv_parse_log_rec( #endif /* UNIV_LOG_LSN_DEBUG */ new_ptr = recv_parse_or_apply_log_rec_body(*type, new_ptr, end_ptr, - NULL, NULL); + NULL, NULL, *space); if (UNIV_UNLIKELY(new_ptr == NULL)) { return(0); diff --git a/storage/innobase/rem/rem0rec.c b/storage/innobase/rem/rem0rec.c index 30fc28561fa..6bd40c54a0c 100644 --- a/storage/innobase/rem/rem0rec.c +++ b/storage/innobase/rem/rem0rec.c @@ -167,7 +167,6 @@ rec_get_n_extern_new( { const byte* nulls; const byte* lens; - dict_field_t* field; ulint null_mask; ulint n_extern; ulint i; @@ -188,10 +187,13 @@ rec_get_n_extern_new( /* read the lengths of fields 0..n */ do { - ulint len; + const dict_field_t* field + = dict_index_get_nth_field(index, i); + const dict_col_t* col + = dict_field_get_col(field); + ulint len; - field = dict_index_get_nth_field(index, i); - if (!(dict_field_get_col(field)->prtype & DATA_NOT_NULL)) { + if (!(col->prtype & DATA_NOT_NULL)) { /* nullable field => read the null flag */ if (UNIV_UNLIKELY(!(byte) null_mask)) { @@ -209,8 +211,6 @@ rec_get_n_extern_new( if (UNIV_UNLIKELY(!field->fixed_len)) { /* Variable-length field: read the length */ - const dict_col_t* col - = dict_field_get_col(field); len = *lens--; /* If the maximum length of the field is up to 255 bytes, the actual length is always @@ -239,16 +239,15 @@ rec_get_n_extern_new( Determine the offset to each field in a leaf-page record in ROW_FORMAT=COMPACT. This is a special case of rec_init_offsets() and rec_get_offsets_func(). */ -UNIV_INTERN +UNIV_INLINE __attribute__((nonnull)) void rec_init_offsets_comp_ordinary( /*===========================*/ const rec_t* rec, /*!< in: physical record in ROW_FORMAT=COMPACT */ - ulint extra, /*!< in: number of bytes to reserve - between the record header and - the data payload - (usually REC_N_NEW_EXTRA_BYTES) */ + ibool temp, /*!< in: whether to use the + format for temporary files in + index creation */ const dict_index_t* index, /*!< in: record descriptor */ ulint* offsets)/*!< in/out: array of offsets; in: n=rec_offs_n_fields(offsets) */ @@ -256,27 +255,38 @@ rec_init_offsets_comp_ordinary( ulint i = 0; ulint offs = 0; ulint any_ext = 0; - const byte* nulls = rec - (extra + 1); + const byte* nulls = temp + ? rec - 1 + : rec - (1 + REC_N_NEW_EXTRA_BYTES); const byte* lens = nulls - UT_BITS_IN_BYTES(index->n_nullable); - dict_field_t* field; ulint null_mask = 1; #ifdef UNIV_DEBUG - /* We cannot invoke rec_offs_make_valid() here, because it can hold - that extra != REC_N_NEW_EXTRA_BYTES. Similarly, rec_offs_validate() - will fail in that case, because it invokes rec_get_status(). */ + /* We cannot invoke rec_offs_make_valid() here if temp=TRUE. + Similarly, rec_offs_validate() will fail in that case, because + it invokes rec_get_status(). */ offsets[2] = (ulint) rec; offsets[3] = (ulint) index; #endif /* UNIV_DEBUG */ + ut_ad(temp || dict_table_is_comp(index->table)); + + if (temp && dict_table_is_comp(index->table)) { + /* No need to do adjust fixed_len=0. We only need to + adjust it for ROW_FORMAT=REDUNDANT. */ + temp = FALSE; + } + /* read the lengths of fields 0..n */ do { - ulint len; + const dict_field_t* field + = dict_index_get_nth_field(index, i); + const dict_col_t* col + = dict_field_get_col(field); + ulint len; - field = dict_index_get_nth_field(index, i); - if (!(dict_field_get_col(field)->prtype - & DATA_NOT_NULL)) { + if (!(col->prtype & DATA_NOT_NULL)) { /* nullable field => read the null flag */ if (UNIV_UNLIKELY(!(byte) null_mask)) { @@ -296,10 +306,9 @@ rec_init_offsets_comp_ordinary( null_mask <<= 1; } - if (UNIV_UNLIKELY(!field->fixed_len)) { + if (!field->fixed_len + || (temp && !dict_col_get_fixed_size(col, temp))) { /* Variable-length field: read the length */ - const dict_col_t* col - = dict_field_get_col(field); len = *lens--; /* If the maximum length of the field is up to 255 bytes, the actual length is always @@ -393,9 +402,8 @@ rec_init_offsets( = dict_index_get_n_unique_in_tree(index); break; case REC_STATUS_ORDINARY: - rec_init_offsets_comp_ordinary(rec, - REC_N_NEW_EXTRA_BYTES, - index, offsets); + rec_init_offsets_comp_ordinary( + rec, FALSE, index, offsets); return; } @@ -766,17 +774,19 @@ rec_get_nth_field_offs_old( /**********************************************************//** Determines the size of a data tuple prefix in ROW_FORMAT=COMPACT. @return total size */ -UNIV_INTERN +UNIV_INLINE __attribute__((warn_unused_result, nonnull(1,2))) ulint -rec_get_converted_size_comp_prefix( -/*===============================*/ +rec_get_converted_size_comp_prefix_low( +/*===================================*/ const dict_index_t* index, /*!< in: record descriptor; dict_table_is_comp() is assumed to hold, even if it does not */ const dfield_t* fields, /*!< in: array of data fields */ ulint n_fields,/*!< in: number of data fields */ - ulint* extra) /*!< out: extra size */ + ulint* extra, /*!< out: extra size */ + ibool temp) /*!< in: whether this is a + temporary file record */ { ulint extra_size; ulint data_size; @@ -785,15 +795,25 @@ rec_get_converted_size_comp_prefix( ut_ad(fields); ut_ad(n_fields > 0); ut_ad(n_fields <= dict_index_get_n_fields(index)); + ut_ad(!temp || extra); - extra_size = REC_N_NEW_EXTRA_BYTES + extra_size = temp + ? UT_BITS_IN_BYTES(index->n_nullable) + : REC_N_NEW_EXTRA_BYTES + UT_BITS_IN_BYTES(index->n_nullable); data_size = 0; + if (temp && dict_table_is_comp(index->table)) { + /* No need to do adjust fixed_len=0. We only need to + adjust it for ROW_FORMAT=REDUNDANT. */ + temp = FALSE; + } + /* read the lengths of fields 0..n */ for (i = 0; i < n_fields; i++) { const dict_field_t* field; ulint len; + ulint fixed_len; const dict_col_t* col; field = dict_index_get_nth_field(index, i); @@ -809,8 +829,14 @@ rec_get_converted_size_comp_prefix( continue; } - ut_ad(len <= col->len || col->mtype == DATA_BLOB); + ut_ad(len <= col->len || col->mtype == DATA_BLOB + || (col->len == 0 && col->mtype == DATA_VARCHAR)); + fixed_len = field->fixed_len; + if (temp && fixed_len + && !dict_col_get_fixed_size(col, temp)) { + fixed_len = 0; + } /* If the maximum length of a variable-length field is up to 255 bytes, the actual length is always stored in one byte. If the maximum length is more than 255 @@ -818,11 +844,20 @@ rec_get_converted_size_comp_prefix( 0..127. The length will be encoded in two bytes when it is 128 or more, or when the field is stored externally. */ - if (field->fixed_len) { - ut_ad(len == field->fixed_len); + if (fixed_len) { +#ifdef UNIV_DEBUG + ulint mbminlen = DATA_MBMINLEN(col->mbminmaxlen); + ulint mbmaxlen = DATA_MBMAXLEN(col->mbminmaxlen); + + ut_ad(len <= fixed_len); + + ut_ad(!mbmaxlen || len >= mbminlen + * (fixed_len / mbmaxlen)); + /* dict_index_add_col() should guarantee this */ ut_ad(!field->prefix_len - || field->fixed_len == field->prefix_len); + || fixed_len == field->prefix_len); +#endif /* UNIV_DEBUG */ } else if (dfield_is_ext(&fields[i])) { ut_ad(col->len >= 256 || col->mtype == DATA_BLOB); extra_size += 2; @@ -839,7 +874,7 @@ rec_get_converted_size_comp_prefix( data_size += len; } - if (UNIV_LIKELY_NULL(extra)) { + if (extra) { *extra = extra_size; } @@ -847,6 +882,23 @@ rec_get_converted_size_comp_prefix( } /**********************************************************//** +Determines the size of a data tuple prefix in ROW_FORMAT=COMPACT. +@return total size */ +UNIV_INTERN +ulint +rec_get_converted_size_comp_prefix( +/*===============================*/ + const dict_index_t* index, /*!< in: record descriptor */ + const dfield_t* fields, /*!< in: array of data fields */ + ulint n_fields,/*!< in: number of data fields */ + ulint* extra) /*!< out: extra size */ +{ + ut_ad(dict_table_is_comp(index->table)); + return(rec_get_converted_size_comp_prefix_low( + index, fields, n_fields, extra, FALSE)); +} + +/**********************************************************//** Determines the size of a data tuple in ROW_FORMAT=COMPACT. @return total size */ UNIV_INTERN @@ -890,8 +942,8 @@ rec_get_converted_size_comp( return(ULINT_UNDEFINED); } - return(size + rec_get_converted_size_comp_prefix(index, fields, - n_fields, extra)); + return(size + rec_get_converted_size_comp_prefix_low( + index, fields, n_fields, extra, FALSE)); } /***********************************************************//** @@ -1068,19 +1120,18 @@ rec_convert_dtuple_to_rec_old( /*********************************************************//** Builds a ROW_FORMAT=COMPACT record out of a data tuple. */ -UNIV_INTERN +UNIV_INLINE __attribute__((nonnull)) void rec_convert_dtuple_to_rec_comp( /*===========================*/ rec_t* rec, /*!< in: origin of record */ - ulint extra, /*!< in: number of bytes to - reserve between the record - header and the data payload - (normally REC_N_NEW_EXTRA_BYTES) */ const dict_index_t* index, /*!< in: record descriptor */ - ulint status, /*!< in: status bits of the record */ const dfield_t* fields, /*!< in: array of data fields */ - ulint n_fields)/*!< in: number of data fields */ + ulint n_fields,/*!< in: number of data fields */ + ulint status, /*!< in: status bits of the record */ + ibool temp) /*!< in: whether to use the + format for temporary files in + index creation */ { const dfield_t* field; const dtype_t* type; @@ -1092,31 +1143,44 @@ rec_convert_dtuple_to_rec_comp( ulint n_node_ptr_field; ulint fixed_len; ulint null_mask = 1; - ut_ad(extra == 0 || dict_table_is_comp(index->table)); - ut_ad(extra == 0 || extra == REC_N_NEW_EXTRA_BYTES); + ut_ad(temp || dict_table_is_comp(index->table)); ut_ad(n_fields > 0); - switch (UNIV_EXPECT(status, REC_STATUS_ORDINARY)) { - case REC_STATUS_ORDINARY: + if (temp) { + ut_ad(status == REC_STATUS_ORDINARY); ut_ad(n_fields <= dict_index_get_n_fields(index)); n_node_ptr_field = ULINT_UNDEFINED; - break; - case REC_STATUS_NODE_PTR: - ut_ad(n_fields == dict_index_get_n_unique_in_tree(index) + 1); - n_node_ptr_field = n_fields - 1; - break; - case REC_STATUS_INFIMUM: - case REC_STATUS_SUPREMUM: - ut_ad(n_fields == 1); - n_node_ptr_field = ULINT_UNDEFINED; - break; - default: - ut_error; - return; + nulls = rec - 1; + if (dict_table_is_comp(index->table)) { + /* No need to do adjust fixed_len=0. We only + need to adjust it for ROW_FORMAT=REDUNDANT. */ + temp = FALSE; + } + } else { + nulls = rec - (REC_N_NEW_EXTRA_BYTES + 1); + + switch (UNIV_EXPECT(status, REC_STATUS_ORDINARY)) { + case REC_STATUS_ORDINARY: + ut_ad(n_fields <= dict_index_get_n_fields(index)); + n_node_ptr_field = ULINT_UNDEFINED; + break; + case REC_STATUS_NODE_PTR: + ut_ad(n_fields + == dict_index_get_n_unique_in_tree(index) + 1); + n_node_ptr_field = n_fields - 1; + break; + case REC_STATUS_INFIMUM: + case REC_STATUS_SUPREMUM: + ut_ad(n_fields == 1); + n_node_ptr_field = ULINT_UNDEFINED; + break; + default: + ut_error; + return; + } } end = rec; - nulls = rec - (extra + 1); lens = nulls - UT_BITS_IN_BYTES(index->n_nullable); /* clear the SQL-null flags */ memset(lens + 1, 0, nulls - lens); @@ -1162,6 +1226,10 @@ rec_convert_dtuple_to_rec_comp( ifield = dict_index_get_nth_field(index, i); fixed_len = ifield->fixed_len; + if (temp && fixed_len + && !dict_col_get_fixed_size(ifield->col, temp)) { + fixed_len = 0; + } /* If the maximum length of a variable-length field is up to 255 bytes, the actual length is always stored in one byte. If the maximum length is more than 255 @@ -1169,8 +1237,17 @@ rec_convert_dtuple_to_rec_comp( 0..127. The length will be encoded in two bytes when it is 128 or more, or when the field is stored externally. */ if (fixed_len) { - ut_ad(len == fixed_len); +#ifdef UNIV_DEBUG + ulint mbminlen = DATA_MBMINLEN( + ifield->col->mbminmaxlen); + ulint mbmaxlen = DATA_MBMAXLEN( + ifield->col->mbminmaxlen); + + ut_ad(len <= fixed_len); + ut_ad(!mbmaxlen || len >= mbminlen + * (fixed_len / mbmaxlen)); ut_ad(!dfield_is_ext(field)); +#endif /* UNIV_DEBUG */ } else if (dfield_is_ext(field)) { ut_ad(ifield->col->len >= 256 || ifield->col->mtype == DATA_BLOB); @@ -1222,8 +1299,7 @@ rec_convert_dtuple_to_rec_new( rec = buf + extra_size; rec_convert_dtuple_to_rec_comp( - rec, REC_N_NEW_EXTRA_BYTES, index, status, - dtuple->fields, dtuple->n_fields); + rec, index, dtuple->fields, dtuple->n_fields, status, FALSE); /* Set the info bits of the record */ rec_set_info_and_status_bits(rec, dtuple_get_info_bits(dtuple)); @@ -1285,6 +1361,54 @@ rec_convert_dtuple_to_rec( return(rec); } +#ifndef UNIV_HOTBACKUP +/**********************************************************//** +Determines the size of a data tuple prefix in ROW_FORMAT=COMPACT. +@return total size */ +UNIV_INTERN +ulint +rec_get_converted_size_temp( +/*========================*/ + const dict_index_t* index, /*!< in: record descriptor */ + const dfield_t* fields, /*!< in: array of data fields */ + ulint n_fields,/*!< in: number of data fields */ + ulint* extra) /*!< out: extra size */ +{ + return(rec_get_converted_size_comp_prefix_low( + index, fields, n_fields, extra, TRUE)); +} + +/******************************************************//** +Determine the offset to each field in temporary file. +@see rec_convert_dtuple_to_temp() */ +UNIV_INTERN +void +rec_init_offsets_temp( +/*==================*/ + const rec_t* rec, /*!< in: temporary file record */ + const dict_index_t* index, /*!< in: record descriptor */ + ulint* offsets)/*!< in/out: array of offsets; + in: n=rec_offs_n_fields(offsets) */ +{ + rec_init_offsets_comp_ordinary(rec, TRUE, index, offsets); +} + +/*********************************************************//** +Builds a temporary file record out of a data tuple. +@see rec_init_offsets_temp() */ +UNIV_INTERN +void +rec_convert_dtuple_to_temp( +/*=======================*/ + rec_t* rec, /*!< out: record */ + const dict_index_t* index, /*!< in: record descriptor */ + const dfield_t* fields, /*!< in: array of data fields */ + ulint n_fields) /*!< in: number of fields */ +{ + rec_convert_dtuple_to_rec_comp(rec, index, fields, n_fields, + REC_STATUS_ORDINARY, TRUE); +} + /**************************************************************//** Copies the first n fields of a physical record to a data tuple. The fields are copied to the memory heap. */ @@ -1495,6 +1619,7 @@ rec_copy_prefix_to_buf( return(*buf + (rec - (lens + 1))); } +#endif /* UNIV_HOTBACKUP */ /***************************************************************//** Validates the consistency of an old-style physical record. diff --git a/storage/innobase/row/row0ins.c b/storage/innobase/row/row0ins.c index 18486514267..21afa9eff0d 100644 --- a/storage/innobase/row/row0ins.c +++ b/storage/innobase/row/row0ins.c @@ -2254,7 +2254,10 @@ row_ins_index_entry( err = row_ins_index_entry_low(BTR_MODIFY_LEAF, index, entry, n_ext, thr); if (err != DB_FAIL) { - + if (index == dict_table_get_first_index(index->table) + && thr_get_trx(thr)->mysql_thd != 0) { + DEBUG_SYNC_C("row_ins_clust_index_entry_leaf_after"); + } return(err); } diff --git a/storage/innobase/row/row0merge.c b/storage/innobase/row/row0merge.c index 7c9fb8812e6..d34ad467588 100644 --- a/storage/innobase/row/row0merge.c +++ b/storage/innobase/row/row0merge.c @@ -298,6 +298,7 @@ row_merge_buf_add( for (i = 0; i < n_fields; i++, field++, ifield++) { const dict_col_t* col; ulint col_no; + ulint fixed_len; const dfield_t* row_field; ulint len; @@ -346,9 +347,30 @@ row_merge_buf_add( ut_ad(len <= col->len || col->mtype == DATA_BLOB); - if (ifield->fixed_len) { - ut_ad(len == ifield->fixed_len); + fixed_len = ifield->fixed_len; + if (fixed_len && !dict_table_is_comp(index->table) + && DATA_MBMINLEN(col->mbminmaxlen) + != DATA_MBMAXLEN(col->mbminmaxlen)) { + /* CHAR in ROW_FORMAT=REDUNDANT is always + fixed-length, but in the temporary file it is + variable-length for variable-length character + sets. */ + fixed_len = 0; + } + + if (fixed_len) { +#ifdef UNIV_DEBUG + ulint mbminlen = DATA_MBMINLEN(col->mbminmaxlen); + ulint mbmaxlen = DATA_MBMAXLEN(col->mbminmaxlen); + + /* len should be between size calcualted base on + mbmaxlen and mbminlen */ + ut_ad(len <= fixed_len); + ut_ad(!mbmaxlen || len >= mbminlen + * (fixed_len / mbmaxlen)); + ut_ad(!dfield_is_ext(field)); +#endif /* UNIV_DEBUG */ } else if (dfield_is_ext(field)) { extra_size += 2; } else if (len < 128 @@ -369,12 +391,11 @@ row_merge_buf_add( ulint size; ulint extra; - size = rec_get_converted_size_comp(index, - REC_STATUS_ORDINARY, - entry, n_fields, &extra); + size = rec_get_converted_size_temp( + index, entry, n_fields, &extra); - ut_ad(data_size + extra_size + REC_N_NEW_EXTRA_BYTES == size); - ut_ad(extra_size + REC_N_NEW_EXTRA_BYTES == extra); + ut_ad(data_size + extra_size == size); + ut_ad(extra_size == extra); } #endif /* UNIV_DEBUG */ @@ -578,14 +599,9 @@ row_merge_buf_write( ulint extra_size; const dfield_t* entry = buf->tuples[i]; - size = rec_get_converted_size_comp(index, - REC_STATUS_ORDINARY, - entry, n_fields, - &extra_size); + size = rec_get_converted_size_temp( + index, entry, n_fields, &extra_size); ut_ad(size >= extra_size); - ut_ad(extra_size >= REC_N_NEW_EXTRA_BYTES); - extra_size -= REC_N_NEW_EXTRA_BYTES; - size -= REC_N_NEW_EXTRA_BYTES; /* Encode extra_size + 1 */ if (extra_size + 1 < 0x80) { @@ -598,9 +614,8 @@ row_merge_buf_write( ut_ad(b + size < block[1]); - rec_convert_dtuple_to_rec_comp(b + extra_size, 0, index, - REC_STATUS_ORDINARY, - entry, n_fields); + rec_convert_dtuple_to_temp(b + extra_size, index, + entry, n_fields); b += size; @@ -704,6 +719,8 @@ row_merge_read( ib_uint64_t ofs = ((ib_uint64_t) offset) * sizeof *buf; ibool success; + DBUG_EXECUTE_IF("row_merge_read_failure", return(FALSE);); + #ifdef UNIV_DEBUG if (row_merge_print_block_read) { fprintf(stderr, "row_merge_read fd=%d ofs=%lu\n", @@ -750,6 +767,8 @@ row_merge_write( (ulint) (ofs >> 32), buf_len); + DBUG_EXECUTE_IF("row_merge_write_failure", return(FALSE);); + #ifdef UNIV_DEBUG if (row_merge_print_block_write) { fprintf(stderr, "row_merge_write fd=%d ofs=%lu\n", @@ -862,7 +881,7 @@ err_exit: *mrec = *buf + extra_size; - rec_init_offsets_comp_ordinary(*mrec, 0, index, offsets); + rec_init_offsets_temp(*mrec, index, offsets); data_size = rec_offs_data_size(offsets); @@ -881,7 +900,7 @@ err_exit: *mrec = b + extra_size; - rec_init_offsets_comp_ordinary(*mrec, 0, index, offsets); + rec_init_offsets_temp(*mrec, index, offsets); data_size = rec_offs_data_size(offsets); ut_ad(extra_size + data_size < sizeof *buf); @@ -2214,7 +2233,7 @@ row_merge_drop_temp_indexes(void) /*********************************************************************//** Creates temperary merge files, and if UNIV_PFS_IO defined, register the file descriptor with Performance Schema. -@return File descriptor */ +@return file descriptor, or -1 on failure */ UNIV_INLINE int row_merge_file_create_low(void) @@ -2236,12 +2255,19 @@ row_merge_file_create_low(void) #ifdef UNIV_PFS_IO register_pfs_file_open_end(locker, fd); #endif + if (fd < 0) { + fprintf(stderr, + "InnoDB: Error: Cannot create temporary merge file\n"); + return(-1); + } return(fd); } + /*********************************************************************//** -Create a merge file. */ -static -void +Create a merge file. +@return file descriptor, or -1 on failure */ +static __attribute__((nonnull, warn_unused_result)) +int row_merge_file_create( /*==================*/ merge_file_t* merge_file) /*!< out: merge file structure */ @@ -2249,6 +2275,7 @@ row_merge_file_create( merge_file->fd = row_merge_file_create_low(); merge_file->offset = 0; merge_file->n_rec = 0; + return(merge_file->fd); } /*********************************************************************//** @@ -2494,6 +2521,28 @@ row_merge_rename_tables( goto err_exit; } + /* Generate the redo logs for file operations */ + fil_mtr_rename_log(old_table->space, old_name, + new_table->space, new_table->name, tmp_name); + + /* What if the redo logs are flushed to disk here? This is + tested with following crash point */ + DBUG_EXECUTE_IF("bug14669848_precommit", log_buffer_flush_to_disk(); + DBUG_SUICIDE();); + + /* File operations cannot be rolled back. So, before proceeding + with file operations, commit the dictionary changes.*/ + trx_commit_for_mysql(trx); + + /* If server crashes here, the dictionary in InnoDB and MySQL + will differ. The .ibd files and the .frm files must be swapped + manually by the administrator. No loss of data. */ + DBUG_EXECUTE_IF("bug14669848", DBUG_SUICIDE();); + + /* Ensure that the redo logs are flushed to disk. The config + innodb_flush_log_at_trx_commit must not affect this. */ + log_buffer_flush_to_disk(); + /* The following calls will also rename the .ibd data files if the tables are stored in a single-table tablespace */ @@ -2665,7 +2714,7 @@ row_merge_build_indexes( ulint block_size; ulint i; ulint error; - int tmpfd; + int tmpfd = -1; ut_ad(trx); ut_ad(old_table); @@ -2682,13 +2731,31 @@ row_merge_build_indexes( block_size = 3 * sizeof *block; block = os_mem_alloc_large(&block_size); + /* Initialize all the merge file descriptors, so that we + don't call row_merge_file_destroy() on uninitialized + merge file descriptor */ + + for (i = 0; i < n_indexes; i++) { + merge_files[i].fd = -1; + } + for (i = 0; i < n_indexes; i++) { - row_merge_file_create(&merge_files[i]); + if (row_merge_file_create(&merge_files[i]) < 0) + { + error = DB_OUT_OF_MEMORY; + goto func_exit; + } } tmpfd = row_merge_file_create_low(); + if (tmpfd < 0) + { + error = DB_OUT_OF_MEMORY; + goto func_exit; + } + /* Reset the MySQL row buffer that is used when reporting duplicate keys. */ innobase_rec_reset(table); diff --git a/storage/innobase/row/row0mysql.c b/storage/innobase/row/row0mysql.c index 7e1d18d41b6..73da6215f0d 100644 --- a/storage/innobase/row/row0mysql.c +++ b/storage/innobase/row/row0mysql.c @@ -4140,6 +4140,13 @@ end: trx->error_state = DB_SUCCESS; trx_general_rollback_for_mysql(trx, NULL); trx->error_state = DB_SUCCESS; + } else { + if (old_is_tmp && !new_is_tmp) { + /* After ALTER TABLE the table statistics + needs to be rebuilt. It will be rebuilt + when the table is loaded again. */ + table->stat_initialized = FALSE; + } } } diff --git a/storage/innobase/row/row0umod.c b/storage/innobase/row/row0umod.c index 9e5fb8686c6..3c933c87b27 100644 --- a/storage/innobase/row/row0umod.c +++ b/storage/innobase/row/row0umod.c @@ -519,6 +519,7 @@ row_undo_mod_upd_del_sec( ulint err = DB_SUCCESS; ut_ad(node->rec_type == TRX_UNDO_UPD_DEL_REC); + ut_ad(!node->undo_row); heap = mem_heap_create(1024); while (node->index != NULL) { @@ -580,6 +581,8 @@ row_undo_mod_del_mark_sec( dict_index_t* index; ulint err; + ut_ad(!node->undo_row); + heap = mem_heap_create(1024); while (node->index != NULL) { diff --git a/storage/innobase/row/row0undo.c b/storage/innobase/row/row0undo.c index a1c1d72f8c6..74fc1baf1d2 100644 --- a/storage/innobase/row/row0undo.c +++ b/storage/innobase/row/row0undo.c @@ -216,7 +216,7 @@ row_undo_search_clust_to_pcur( node->row = row_build(ROW_COPY_DATA, clust_index, rec, offsets, NULL, ext, node->heap); - if (node->update) { + if (node->rec_type == TRX_UNDO_UPD_EXIST_REC) { node->undo_row = dtuple_copy(node->row, node->heap); row_upd_replace(node->undo_row, &node->undo_ext, clust_index, node->update, node->heap); diff --git a/storage/innobase/srv/srv0srv.c b/storage/innobase/srv/srv0srv.c index df89156baae..5872ff2c437 100644 --- a/storage/innobase/srv/srv0srv.c +++ b/storage/innobase/srv/srv0srv.c @@ -58,6 +58,8 @@ Created 10/8/1995 Heikki Tuuri *******************************************************/ /* Dummy comment */ +#include "m_string.h" /* for my_sys.h */ +#include "my_sys.h" /* DEBUG_SYNC_C */ #include "srv0srv.h" #include "ut0mem.h" @@ -83,6 +85,7 @@ Created 10/8/1995 Heikki Tuuri #include "ha_prototypes.h" #include "trx0i_s.h" #include "os0sync.h" /* for HAVE_ATOMIC_BUILTINS */ +#include "read0read.h" #include "mysql/plugin.h" #include "mysql/service_thd_wait.h" @@ -356,6 +359,9 @@ UNIV_INTERN lint srv_conc_n_threads = 0; InnoDB */ UNIV_INTERN ulint srv_conc_n_waiting_threads = 0; +/* print all user-level transactions deadlocks to mysqld stderr */ +UNIV_INTERN my_bool srv_print_all_deadlocks = FALSE; + typedef struct srv_conc_slot_struct srv_conc_slot_t; struct srv_conc_slot_struct{ os_event_t event; /*!< event to wait */ @@ -1568,6 +1574,10 @@ srv_suspend_mysql_thread( trx = thr_get_trx(thr); + if (trx->mysql_thd != 0) { + DEBUG_SYNC_C("srv_suspend_mysql_thread_enter"); + } + os_event_set(srv_lock_timeout_thread_event); mutex_enter(&kernel_mutex); @@ -2006,13 +2016,15 @@ void srv_export_innodb_status(void) /*==========================*/ { - buf_pool_stat_t stat; - ulint LRU_len; - ulint free_len; - ulint flush_list_len; + buf_pool_stat_t stat; + buf_pools_list_size_t buf_pools_list_size; + ulint LRU_len; + ulint free_len; + ulint flush_list_len; buf_get_total_stat(&stat); buf_get_total_list_len(&LRU_len, &free_len, &flush_list_len); + buf_get_total_list_size_in_bytes(&buf_pools_list_size); mutex_enter(&srv_innodb_monitor_mutex); @@ -2041,7 +2053,12 @@ srv_export_innodb_status(void) export_vars.innodb_buffer_pool_read_ahead_evicted = stat.n_ra_pages_evicted; export_vars.innodb_buffer_pool_pages_data = LRU_len; + export_vars.innodb_buffer_pool_bytes_data = + buf_pools_list_size.LRU_bytes + + buf_pools_list_size.unzip_LRU_bytes; export_vars.innodb_buffer_pool_pages_dirty = flush_list_len; + export_vars.innodb_buffer_pool_bytes_dirty = + buf_pools_list_size.flush_list_bytes; export_vars.innodb_buffer_pool_pages_free = free_len; #ifdef UNIV_DEBUG export_vars.innodb_buffer_pool_pages_latched @@ -2087,6 +2104,23 @@ srv_export_innodb_status(void) export_vars.innodb_rows_deleted = srv_n_rows_deleted; export_vars.innodb_truncated_status_writes = srv_truncated_status_writes; +#ifdef UNIV_DEBUG + if (trx_sys->max_trx_id < purge_sys->done_trx_no) { + export_vars.innodb_purge_trx_id_age = 0; + } else { + export_vars.innodb_purge_trx_id_age = + trx_sys->max_trx_id - purge_sys->done_trx_no; + } + + if (!purge_sys->view + || trx_sys->max_trx_id < purge_sys->view->up_limit_id) { + export_vars.innodb_purge_view_trx_id_age = 0; + } else { + export_vars.innodb_purge_view_trx_id_age = + trx_sys->max_trx_id - purge_sys->view->up_limit_id; + } +#endif /* UNIV_DEBUG */ + mutex_exit(&srv_innodb_monitor_mutex); } @@ -2773,6 +2807,26 @@ loop: for (i = 0; i < 10; i++) { ulint cur_time = ut_time_ms(); +#ifdef UNIV_DEBUG + if (btr_cur_limit_optimistic_insert_debug + && srv_n_purge_threads == 0) { + /* If btr_cur_limit_optimistic_insert_debug is enabled + and no purge_threads, purge opportunity is increased + by x100 (1purge/100msec), to speed up debug scripts + which should wait for purged. */ + next_itr_time -= 900; + + srv_main_thread_op_info = "master purging"; + + srv_master_do_purge(); + + if (srv_fast_shutdown && srv_shutdown_state > 0) { + + goto background_loop; + } + } +#endif /* UNIV_DEBUG */ + /* ALTER TABLE in MySQL requires on Unix that the table handler can drop tables lazily after there no longer are SELECT queries to them. */ diff --git a/storage/innobase/trx/trx0purge.c b/storage/innobase/trx/trx0purge.c index 96f01ea81b5..719851218c3 100644 --- a/storage/innobase/trx/trx0purge.c +++ b/storage/innobase/trx/trx0purge.c @@ -61,6 +61,10 @@ UNIV_INTERN mysql_pfs_key_t trx_purge_latch_key; UNIV_INTERN mysql_pfs_key_t purge_sys_bh_mutex_key; #endif /* UNIV_PFS_MUTEX */ +#ifdef UNIV_DEBUG +UNIV_INTERN my_bool srv_purge_view_update_only_debug; +#endif /* UNIV_DEBUG */ + /*****************************************************************//** Checks if trx_id is >= purge_view: then it is guaranteed that its update undo log still exists in the system. @@ -236,6 +240,7 @@ trx_purge_sys_create( purge_sys->purge_trx_no = 0; purge_sys->purge_undo_no = 0; purge_sys->next_stored = FALSE; + ut_d(purge_sys->done_trx_no = 0); rw_lock_create(trx_purge_latch_key, &purge_sys->latch, SYNC_PURGE_LATCH); @@ -656,6 +661,12 @@ trx_purge_truncate_if_arr_empty(void) { static ulint count; +#ifdef UNIV_DEBUG + if (purge_sys->arr->n_used == 0) { + purge_sys->done_trx_no = purge_sys->purge_trx_no; + } +#endif /* UNIV_DEBUG */ + if (!(++count % TRX_SYS_N_RSEGS) && purge_sys->arr->n_used == 0) { trx_purge_truncate_history(); @@ -1173,6 +1184,12 @@ trx_purge( rw_lock_x_unlock(&(purge_sys->latch)); +#ifdef UNIV_DEBUG + if (srv_purge_view_update_only_debug) { + return(0); + } +#endif + purge_sys->state = TRX_PURGE_ON; purge_sys->handle_limit = purge_sys->n_pages_handled + limit; diff --git a/storage/innobase/trx/trx0rec.c b/storage/innobase/trx/trx0rec.c index 2f1389ae263..f80f58493aa 100644 --- a/storage/innobase/trx/trx0rec.c +++ b/storage/innobase/trx/trx0rec.c @@ -36,6 +36,7 @@ Created 3/26/1996 Heikki Tuuri #ifndef UNIV_HOTBACKUP #include "dict0dict.h" #include "ut0mem.h" +#include "read0read.h" #include "row0ext.h" #include "row0upd.h" #include "que0que.h" @@ -1634,6 +1635,25 @@ trx_undo_prev_version_build( if (row_upd_changes_field_size_or_external(index, offsets, update)) { ulint n_ext; + /* We should confirm the existence of disowned external data, + if the previous version record is delete marked. If the trx_id + of the previous record is seen by purge view, we should treat + it as missing history, because the disowned external data + might be purged already. + + The inherited external data (BLOBs) can be freed (purged) + after trx_id was committed, provided that no view was started + before trx_id. If the purge view can see the committed + delete-marked record by trx_id, no transactions need to access + the BLOB. */ + + if ((update->info_bits & REC_INFO_DELETED_FLAG) + && read_view_sees_trx_id(purge_sys->view, trx_id)) { + /* treat as a fresh insert, not to + cause assertion error at the caller. */ + return(DB_SUCCESS); + } + /* We have to set the appropriate extern storage bits in the old version of the record: the extern bits in rec for those fields that update does NOT update, as well as the bits for diff --git a/storage/myisam/mi_open.c b/storage/myisam/mi_open.c index 438057e22df..22225303bae 100644 --- a/storage/myisam/mi_open.c +++ b/storage/myisam/mi_open.c @@ -349,6 +349,12 @@ MI_INFO *mi_open(const char *name, int mode, uint open_flags) } else if (pos->type == HA_KEYTYPE_BINARY) pos->charset= &my_charset_bin; + if (!(share->keyinfo[i].flag & HA_SPATIAL) && + pos->start > share->base.reclength) + { + my_errno= HA_ERR_CRASHED; + goto err; + } } if (share->keyinfo[i].flag & HA_SPATIAL) { diff --git a/storage/myisam/mi_search.c b/storage/myisam/mi_search.c index 968cb9624a6..01fa10de7a3 100644 --- a/storage/myisam/mi_search.c +++ b/storage/myisam/mi_search.c @@ -949,9 +949,7 @@ uint _mi_get_binary_pack_key(register MI_KEYDEF *keyinfo, uint nod_flag, ("Found too long binary packed key: %u of %u at 0x%lx", length, keyinfo->maxlength, (long) *page_pos)); DBUG_DUMP("key", *page_pos, 16); - mi_print_error(keyinfo->share, HA_ERR_CRASHED); - my_errno=HA_ERR_CRASHED; - DBUG_RETURN(0); /* Wrong key */ + goto crashed; /* Wrong key */ } /* Key is packed against prev key, take prefix from prev key. */ from= key; @@ -994,6 +992,8 @@ uint _mi_get_binary_pack_key(register MI_KEYDEF *keyinfo, uint nod_flag, if (from == from_end) { from=page; from_end=page_end; } length+= (uint) ((*key++ = *from++)); } + if (length > keyseg->length) + goto crashed; } else length=keyseg->length; @@ -1033,15 +1033,18 @@ uint _mi_get_binary_pack_key(register MI_KEYDEF *keyinfo, uint nod_flag, if (from_end != page_end) { DBUG_PRINT("error",("Error when unpacking key")); - mi_print_error(keyinfo->share, HA_ERR_CRASHED); - my_errno=HA_ERR_CRASHED; - DBUG_RETURN(0); /* Error */ + goto crashed; /* Error */ } /* Copy data pointer and, if appropriate, key block pointer. */ memcpy((uchar*) key,(uchar*) from,(size_t) length); *page_pos= from+length; } DBUG_RETURN((uint) (key-start_key)+keyseg->length); + + crashed: + mi_print_error(keyinfo->share, HA_ERR_CRASHED); + my_errno= HA_ERR_CRASHED; + DBUG_RETURN(0); } diff --git a/storage/perfschema/pfs_events_waits.cc b/storage/perfschema/pfs_events_waits.cc index b6cadf9e61c..a941f815dca 100644 --- a/storage/perfschema/pfs_events_waits.cc +++ b/storage/perfschema/pfs_events_waits.cc @@ -93,6 +93,9 @@ static inline void copy_events_waits(PFS_events_waits *dest, */ void insert_events_waits_history(PFS_thread *thread, PFS_events_waits *wait) { + if (unlikely(events_waits_history_per_thread == 0)) + return; + uint index= thread->m_waits_history_index; /* @@ -120,6 +123,9 @@ void insert_events_waits_history(PFS_thread *thread, PFS_events_waits *wait) */ void insert_events_waits_history_long(PFS_events_waits *wait) { + if (unlikely(events_waits_history_long_size == 0)) + return; + uint index= PFS_atomic::add_u32(&events_waits_history_long_index, 1); index= index % events_waits_history_long_size; diff --git a/storage/xtradb/handler/ha_innodb.cc b/storage/xtradb/handler/ha_innodb.cc index 269bf8436a0..3bc49cfa751 100644 --- a/storage/xtradb/handler/ha_innodb.cc +++ b/storage/xtradb/handler/ha_innodb.cc @@ -13043,9 +13043,9 @@ static MYSQL_SYSVAR_ULONG(read_ahead_threshold, srv_read_ahead_threshold, "trigger a readahead.", NULL, NULL, 56, 0, 64, 0); -#ifdef UNIV_DEBUG_never +#ifdef UNIV_DEBUG static MYSQL_SYSVAR_UINT(trx_rseg_n_slots_debug, trx_rseg_n_slots_debug, - PLUGIN_VAR_RQCMDARG, + PLUGIN_VAR_RQCMDARG | PLUGIN_VAR_NOCMDOPT, "Debug flags for InnoDB to limit TRX_RSEG_N_SLOTS for trx_rsegf_undo_find_free()", NULL, NULL, 0, 0, 1024, 0); #endif /* UNIV_DEBUG */ @@ -13331,7 +13331,7 @@ static struct st_mysql_sys_var* innobase_system_variables[]= { MYSQL_SYSVAR(purge_threads), MYSQL_SYSVAR(purge_batch_size), MYSQL_SYSVAR(rollback_segments), -#ifdef UNIV_DEBUG_never /* disable this flag. --innodb-trx becomes ambiguous */ +#ifdef UNIV_DEBUG MYSQL_SYSVAR(trx_rseg_n_slots_debug), #endif /* UNIV_DEBUG */ MYSQL_SYSVAR(corrupt_table_action), |