diff options
Diffstat (limited to 'sql/ha_innodb.cc')
| -rw-r--r-- | sql/ha_innodb.cc | 81 |
1 files changed, 70 insertions, 11 deletions
diff --git a/sql/ha_innodb.cc b/sql/ha_innodb.cc index 870ad1af52b..e29560f15a5 100644 --- a/sql/ha_innodb.cc +++ b/sql/ha_innodb.cc @@ -45,7 +45,8 @@ have disables the InnoDB inlining in this file. */ #include "ha_innodb.h" -pthread_mutex_t innobase_mutex; +pthread_mutex_t innobase_share_mutex, // to protect innobase_open_files + prepare_commit_mutex; // to force correct commit order in binlog bool innodb_inited= 0; /* Store MySQL definition of 'byte': in Linux it is char while InnoDB @@ -1268,7 +1269,8 @@ innobase_init(void) (void) hash_init(&innobase_open_tables,system_charset_info, 32, 0, 0, (hash_get_key) innobase_get_key, 0, 0); - pthread_mutex_init(&innobase_mutex, MY_MUTEX_INIT_FAST); + pthread_mutex_init(&innobase_share_mutex, MY_MUTEX_INIT_FAST); + pthread_mutex_init(&prepare_commit_mutex, MY_MUTEX_INIT_FAST); innodb_inited= 1; /* If this is a replication slave and we needed to do a crash recovery, @@ -1307,6 +1309,9 @@ innobase_end(void) } #endif if (innodb_inited) { + +#ifndef __NETWARE__ /* NetWare can't close unclosed files, kill remaining + threads, etc, so we disable the very fast shutdown */ if (innobase_very_fast_shutdown) { srv_very_fast_shutdown = TRUE; fprintf(stderr, @@ -1314,6 +1319,7 @@ innobase_end(void) "InnoDB: the InnoDB buffer pool to data files. At the next mysqld startup\n" "InnoDB: InnoDB will do a crash recovery!\n"); } +#endif innodb_inited = 0; if (innobase_shutdown_for_mysql() != DB_SUCCESS) { @@ -1322,7 +1328,8 @@ innobase_end(void) hash_free(&innobase_open_tables); my_free(internal_innobase_data_file_path, MYF(MY_ALLOW_ZERO_PTR)); - pthread_mutex_destroy(&innobase_mutex); + pthread_mutex_destroy(&innobase_share_mutex); + pthread_mutex_destroy(&prepare_commit_mutex); } DBUG_RETURN(err); @@ -1480,9 +1487,20 @@ innobase_commit( /* We were instructed to commit the whole transaction, or this is an SQL statement end and autocommit is on */ + /* We need current binlog position for ibbackup to work. + Note, the position is current because of prepare_commit_mutex */ + trx->mysql_log_file_name = mysql_bin_log.get_log_fname(); + trx->mysql_log_offset = + (ib_longlong)mysql_bin_log.get_log_file()->pos_in_file; + innobase_commit_low(trx); + if (trx->active_trans == 2) { + + pthread_mutex_unlock(&prepare_commit_mutex); + } trx->active_trans = 0; + } else { /* We just mark the SQL statement ended and do not do a transaction commit */ @@ -2435,7 +2453,14 @@ ha_innobase::store_key_val_for_row( (byte*) (record + (ulint)get_field_offset(table, field)), lenlen); + + /* In a column prefix index, we may need to truncate + the stored value: */ + if (len > key_part->length) { + len = key_part->length; + } + /* The length in a key value is always stored in 2 bytes */ @@ -2472,6 +2497,11 @@ ha_innobase::store_key_val_for_row( ut_a(get_field_offset(table, field) == key_part->offset); + + /* All indexes on BLOB and TEXT are column prefix + indexes, and we may need to truncate the data to be + stored in the kay value: */ + if (blob_len > key_part->length) { blob_len = key_part->length; } @@ -2490,11 +2520,17 @@ ha_innobase::store_key_val_for_row( buff += key_part->length; } else { + /* Here we handle all other data types except the + true VARCHAR, BLOB and TEXT. Note that the column + value we store may be also in a column prefix + index. */ + if (is_null) { buff += key_part->length; continue; } + memcpy(buff, record + key_part->offset, key_part->length); buff += key_part->length; @@ -5953,7 +5989,7 @@ static mysql_byte* innobase_get_key(INNOBASE_SHARE *share,uint *length, static INNOBASE_SHARE *get_share(const char *table_name) { INNOBASE_SHARE *share; - pthread_mutex_lock(&innobase_mutex); + pthread_mutex_lock(&innobase_share_mutex); uint length=(uint) strlen(table_name); if (!(share=(INNOBASE_SHARE*) hash_search(&innobase_open_tables, (mysql_byte*) table_name, @@ -5967,7 +6003,7 @@ static INNOBASE_SHARE *get_share(const char *table_name) strmov(share->table_name,table_name); if (my_hash_insert(&innobase_open_tables, (mysql_byte*) share)) { - pthread_mutex_unlock(&innobase_mutex); + pthread_mutex_unlock(&innobase_share_mutex); my_free((gptr) share,0); return 0; } @@ -5976,13 +6012,13 @@ static INNOBASE_SHARE *get_share(const char *table_name) } } share->use_count++; - pthread_mutex_unlock(&innobase_mutex); + pthread_mutex_unlock(&innobase_share_mutex); return share; } static void free_share(INNOBASE_SHARE *share) { - pthread_mutex_lock(&innobase_mutex); + pthread_mutex_lock(&innobase_share_mutex); if (!--share->use_count) { hash_delete(&innobase_open_tables, (mysql_byte*) share); @@ -5990,7 +6026,7 @@ static void free_share(INNOBASE_SHARE *share) pthread_mutex_destroy(&share->mutex); my_free((gptr) share, MYF(0)); } - pthread_mutex_unlock(&innobase_mutex); + pthread_mutex_unlock(&innobase_share_mutex); } /********************************************************************* @@ -6454,15 +6490,38 @@ innobase_xa_prepare( FALSE - the current SQL statement ended */ { int error = 0; - trx_t* trx; + trx_t* trx = check_trx_exists(thd); + + if (thd->lex->sql_command != SQLCOM_XA_PREPARE) { + + /* For ibbackup to work the order of transactions in binlog + and InnoDB must be the same. Consider the situation + + thread1> prepare; write to binlog; ... + <context switch> + thread2> prepare; write to binlog; commit + thread1> ... commit + + To ensure this will not happen we're taking the mutex on + prepare, and releasing it on commit. + + Note: only do it for normal commits, done via ha_commit_trans. + If 2pc protocol is executed by external transaction + coordinator, it will be just a regular MySQL client + executing XA PREPARE and XA COMMIT commands. + In this case we cannot know how many minutes or hours + will be between XA PREPARE and XA COMMIT, and we don't want + to block for undefined period of time. + */ + pthread_mutex_lock(&prepare_commit_mutex); + trx->active_trans = 2; + } if (!thd->variables.innodb_support_xa) { return(0); } - trx = check_trx_exists(thd); - trx->xid=thd->transaction.xid; /* Release a possible FIFO ticket and search latch. Since we will |