diff options
Diffstat (limited to 'storage/innodb_plugin/row')
| -rw-r--r-- | storage/innodb_plugin/row/row0ins.c | 92 | ||||
| -rw-r--r-- | storage/innodb_plugin/row/row0merge.c | 35 | ||||
| -rw-r--r-- | storage/innodb_plugin/row/row0mysql.c | 14 | ||||
| -rw-r--r-- | storage/innodb_plugin/row/row0row.c | 33 | ||||
| -rw-r--r-- | storage/innodb_plugin/row/row0sel.c | 4 | ||||
| -rw-r--r-- | storage/innodb_plugin/row/row0upd.c | 73 |
6 files changed, 202 insertions, 49 deletions
diff --git a/storage/innodb_plugin/row/row0ins.c b/storage/innodb_plugin/row/row0ins.c index 2cbe1e13edc..9a603d5690f 100644 --- a/storage/innodb_plugin/row/row0ins.c +++ b/storage/innodb_plugin/row/row0ins.c @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2011, Oracle and/or its affiliates. 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 @@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with -this program; if not, write to the Free Software Foundation, Inc., 59 Temple -Place, Suite 330, Boston, MA 02111-1307 USA +this program; if not, write to the Free Software Foundation, Inc., +51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA *****************************************************************************/ @@ -23,6 +23,15 @@ Insert into a table Created 4/20/1996 Heikki Tuuri *******************************************************/ +#ifdef __WIN__ +/* error LNK2001: unresolved external symbol _debug_sync_C_callback_ptr */ +# define DEBUG_SYNC_C(dummy) ((void) 0) +#else +# include "my_global.h" /* HAVE_* */ +# include "m_string.h" /* for my_sys.h */ +# include "my_sys.h" /* DEBUG_SYNC_C */ +#endif + #include "row0ins.h" #ifdef UNIV_NONINL @@ -345,9 +354,9 @@ row_ins_clust_index_entry_by_modify( return(DB_LOCK_TABLE_FULL); } - err = btr_cur_pessimistic_update(0, cursor, - heap, big_rec, update, - 0, thr, mtr); + err = btr_cur_pessimistic_update( + BTR_KEEP_POS_FLAG, cursor, heap, big_rec, update, + 0, thr, mtr); } return(err); @@ -1989,6 +1998,7 @@ row_ins_index_entry_low( ulint modify = 0; /* remove warning */ rec_t* insert_rec; rec_t* rec; + ulint* offsets; ulint err; ulint n_unique; big_rec_t* big_rec = NULL; @@ -2092,6 +2102,64 @@ row_ins_index_entry_low( err = row_ins_clust_index_entry_by_modify( mode, &cursor, &heap, &big_rec, entry, thr, &mtr); + + if (big_rec) { + ut_a(err == DB_SUCCESS); + /* Write out the externally stored + columns while still x-latching + index->lock and block->lock. Allocate + pages for big_rec in the mtr that + modified the B-tree, but be sure to skip + any pages that were freed in mtr. We will + write out the big_rec pages before + committing the B-tree mini-transaction. If + the system crashes so that crash recovery + will not replay the mtr_commit(&mtr), the + big_rec pages will be left orphaned until + the pages are allocated for something else. + + TODO: If the allocation extends the + tablespace, it will not be redo + logged, in either mini-transaction. + Tablespace extension should be + redo-logged in the big_rec + mini-transaction, so that recovery + will not fail when the big_rec was + written to the extended portion of the + file, in case the file was somehow + truncated in the crash. */ + + rec = btr_cur_get_rec(&cursor); + offsets = rec_get_offsets( + rec, index, NULL, + ULINT_UNDEFINED, &heap); + + DEBUG_SYNC_C("before_row_ins_upd_extern"); + err = btr_store_big_rec_extern_fields( + index, btr_cur_get_block(&cursor), + rec, offsets, big_rec, &mtr, + BTR_STORE_INSERT_UPDATE); + DEBUG_SYNC_C("after_row_ins_upd_extern"); + /* If writing big_rec fails (for + example, because of DB_OUT_OF_FILE_SPACE), + the record will be corrupted. Even if + we did not update any externally + stored columns, our update could cause + the record to grow so that a + non-updated column was selected for + external storage. This non-update + would not have been written to the + undo log, and thus the record cannot + be rolled back. + + However, because we have not executed + mtr_commit(mtr) yet, the update will + not be replayed in crash recovery, and + the following assertion failure will + effectively "roll back" the operation. */ + ut_a(err == DB_SUCCESS); + goto stored_big_rec; + } } else { ut_ad(!n_ext); err = row_ins_sec_index_entry_by_modify( @@ -2120,10 +2188,13 @@ function_exit: mtr_commit(&mtr); if (UNIV_LIKELY_NULL(big_rec)) { - rec_t* rec; - ulint* offsets; + DBUG_EXECUTE_IF( + "row_ins_extern_checkpoint", + log_make_checkpoint_at(IB_ULONGLONG_MAX, TRUE);); + mtr_start(&mtr); + DEBUG_SYNC_C("before_row_ins_extern_latch"); btr_cur_search_to_nth_level(index, 0, entry, PAGE_CUR_LE, BTR_MODIFY_TREE, &cursor, 0, __FILE__, __LINE__, &mtr); @@ -2131,10 +2202,13 @@ function_exit: offsets = rec_get_offsets(rec, index, NULL, ULINT_UNDEFINED, &heap); + DEBUG_SYNC_C("before_row_ins_extern"); err = btr_store_big_rec_extern_fields( index, btr_cur_get_block(&cursor), - rec, offsets, &mtr, FALSE, big_rec); + rec, offsets, big_rec, &mtr, BTR_STORE_INSERT); + DEBUG_SYNC_C("after_row_ins_extern"); +stored_big_rec: if (modify) { dtuple_big_rec_free(big_rec); } else { diff --git a/storage/innodb_plugin/row/row0merge.c b/storage/innodb_plugin/row/row0merge.c index 647d0031635..7f59d7cf9e9 100644 --- a/storage/innodb_plugin/row/row0merge.c +++ b/storage/innodb_plugin/row/row0merge.c @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2005, 2010, Innobase Oy. All Rights Reserved. +Copyright (c) 1995, 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 @@ -576,7 +576,7 @@ row_merge_buf_write( REC_STATUS_ORDINARY, entry, n_fields, &extra_size); - ut_ad(size > 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; @@ -2012,7 +2012,7 @@ row_merge_drop_index( tables in Innobase. Deleting a row from SYS_INDEXES table also frees the file segments of the B-tree associated with the index. */ - static const char str1[] = + static const char sql[] = "PROCEDURE DROP_INDEX_PROC () IS\n" "BEGIN\n" /* Rename the index, so that it will be dropped by @@ -2036,9 +2036,19 @@ row_merge_drop_index( ut_a(trx->dict_operation_lock_mode == RW_X_LATCH); - err = que_eval_sql(info, str1, FALSE, trx); + err = que_eval_sql(info, sql, FALSE, trx); - ut_a(err == DB_SUCCESS); + + if (err != DB_SUCCESS) { + /* Even though we ensure that DDL transactions are WAIT + and DEADLOCK free, we could encounter other errors e.g., + DB_TOO_MANY_TRANSACTIONS. */ + trx->error_state = DB_SUCCESS; + + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: Error: row_merge_drop_index failed " + "with error code: %lu.\n", (ulint) err); + } /* Replace this index with another equivalent index for all foreign key constraints on this table where this index is used */ @@ -2290,7 +2300,7 @@ row_merge_rename_indexes( /* We use the private SQL parser of Innobase to generate the query graphs needed in renaming indexes. */ - static const char rename_indexes[] = + static const char sql[] = "PROCEDURE RENAME_INDEXES_PROC () IS\n" "BEGIN\n" "UPDATE SYS_INDEXES SET NAME=SUBSTR(NAME,1,LENGTH(NAME)-1)\n" @@ -2306,7 +2316,7 @@ row_merge_rename_indexes( pars_info_add_dulint_literal(info, "tableid", table->id); - err = que_eval_sql(info, rename_indexes, FALSE, trx); + err = que_eval_sql(info, sql, FALSE, trx); if (err == DB_SUCCESS) { dict_index_t* index = dict_table_get_first_index(table); @@ -2316,6 +2326,15 @@ row_merge_rename_indexes( } index = dict_table_get_next_index(index); } while (index); + } else { + /* Even though we ensure that DDL transactions are WAIT + and DEADLOCK free, we could encounter other errors e.g., + DB_TOO_MANY_TRANSACTIONS. */ + trx->error_state = DB_SUCCESS; + + ut_print_timestamp(stderr); + fprintf(stderr, " InnoDB: Error: row_merge_rename_indexes " + "failed with error code: %lu.\n", (ulint) err); } trx->op_info = ""; @@ -2354,7 +2373,7 @@ row_merge_rename_tables( memcpy(old_name, old_table->name, strlen(old_table->name) + 1); } else { ut_print_timestamp(stderr); - fprintf(stderr, "InnoDB: too long table name: '%s', " + fprintf(stderr, " InnoDB: too long table name: '%s', " "max length is %d\n", old_table->name, MAX_FULL_NAME_LEN); ut_error; diff --git a/storage/innodb_plugin/row/row0mysql.c b/storage/innodb_plugin/row/row0mysql.c index 6f689f9909d..36aad4f8ca2 100644 --- a/storage/innodb_plugin/row/row0mysql.c +++ b/storage/innodb_plugin/row/row0mysql.c @@ -1900,6 +1900,20 @@ err_exit: } break; + case DB_TOO_MANY_CONCURRENT_TRXS: + /* We already have .ibd file here. it should be deleted. */ + + if (table->space && !fil_delete_tablespace(table->space)) { + ut_print_timestamp(stderr); + fprintf(stderr, + " InnoDB: Error: not able to" + " delete tablespace %lu of table ", + (ulong) table->space); + ut_print_name(stderr, trx, TRUE, table->name); + fputs("!\n", stderr); + } + /* fall through */ + case DB_DUPLICATE_KEY: default: /* We may also get err == DB_ERROR if the .ibd file for the diff --git a/storage/innodb_plugin/row/row0row.c b/storage/innodb_plugin/row/row0row.c index 9cdbbe76e04..7ec05f01821 100644 --- a/storage/innodb_plugin/row/row0row.c +++ b/storage/innodb_plugin/row/row0row.c @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2011, Oracle and/or its affiliates. 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 @@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with -this program; if not, write to the Free Software Foundation, Inc., 59 Temple -Place, Suite 330, Boston, MA 02111-1307 USA +this program; if not, write to the Free Software Foundation, Inc., +51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA *****************************************************************************/ @@ -243,19 +243,20 @@ row_build( } #if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG - /* This condition can occur during crash recovery before - trx_rollback_active() has completed execution. - - This condition is possible if the server crashed - during an insert or update before - btr_store_big_rec_extern_fields() did mtr_commit() all - BLOB pointers to the clustered index record. - - If the record contains a null BLOB pointer, look up the - transaction that holds the implicit lock on this record, and - assert that it was recovered (and will soon be rolled back). */ - ut_a(!rec_offs_any_null_extern(rec, offsets) - || trx_assert_recovered(row_get_rec_trx_id(rec, index, offsets))); + if (rec_offs_any_null_extern(rec, offsets)) { + /* This condition can occur during crash recovery + before trx_rollback_active() has completed execution. + + This condition is possible if the server crashed + during an insert or update-by-delete-and-insert before + btr_store_big_rec_extern_fields() did mtr_commit() all + BLOB pointers to the freshly inserted clustered index + record. */ + ut_a(trx_assert_recovered( + row_get_rec_trx_id(rec, index, offsets))); + ut_a(trx_undo_roll_ptr_is_insert( + row_get_rec_roll_ptr(rec, index, offsets))); + } #endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */ if (type != ROW_COPY_POINTERS) { diff --git a/storage/innodb_plugin/row/row0sel.c b/storage/innodb_plugin/row/row0sel.c index 32f21dbe198..54172e71a47 100644 --- a/storage/innodb_plugin/row/row0sel.c +++ b/storage/innodb_plugin/row/row0sel.c @@ -4362,7 +4362,9 @@ no_gap_lock: applicable to unique secondary indexes. Current behaviour is to widen the scope of a lock on an already delete marked record if the same record is deleted twice by the same transaction */ - if (index == clust_index && unique_search) { + if (index == clust_index && unique_search + && !prebuilt->used_in_HANDLER) { + err = DB_RECORD_NOT_FOUND; goto normal_return; diff --git a/storage/innodb_plugin/row/row0upd.c b/storage/innodb_plugin/row/row0upd.c index aec3dc19b86..5d2d114161c 100644 --- a/storage/innodb_plugin/row/row0upd.c +++ b/storage/innodb_plugin/row/row0upd.c @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2011, Oracle and/or its affiliates. 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 @@ -11,8 +11,8 @@ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with -this program; if not, write to the Free Software Foundation, Inc., 59 Temple -Place, Suite 330, Boston, MA 02111-1307 USA +this program; if not, write to the Free Software Foundation, Inc., +51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA *****************************************************************************/ @@ -23,6 +23,15 @@ Update of a row Created 12/27/1996 Heikki Tuuri *******************************************************/ +#ifdef __WIN__ +/* error LNK2001: unresolved external symbol _debug_sync_C_callback_ptr */ +# define DEBUG_SYNC_C(dummy) ((void) 0) +#else +# include "my_global.h" /* HAVE_* */ +# include "m_string.h" /* for my_sys.h */ +# include "my_sys.h" /* DEBUG_SYNC_C */ +#endif + #include "row0upd.h" #ifdef UNIV_NONINL @@ -1969,28 +1978,62 @@ row_upd_clust_rec( ut_ad(!rec_get_deleted_flag(btr_pcur_get_rec(pcur), dict_table_is_comp(index->table))); - err = btr_cur_pessimistic_update(BTR_NO_LOCKING_FLAG, btr_cur, - &heap, &big_rec, node->update, - node->cmpl_info, thr, mtr); - mtr_commit(mtr); - - if (err == DB_SUCCESS && big_rec) { - ulint offsets_[REC_OFFS_NORMAL_SIZE]; - rec_t* rec; + err = btr_cur_pessimistic_update( + BTR_NO_LOCKING_FLAG | BTR_KEEP_POS_FLAG, btr_cur, + &heap, &big_rec, node->update, node->cmpl_info, thr, mtr); + if (big_rec) { + ulint offsets_[REC_OFFS_NORMAL_SIZE]; + rec_t* rec; rec_offs_init(offsets_); - mtr_start(mtr); + ut_a(err == DB_SUCCESS); + /* Write out the externally stored + columns while still x-latching + index->lock and block->lock. Allocate + pages for big_rec in the mtr that + modified the B-tree, but be sure to skip + any pages that were freed in mtr. We will + write out the big_rec pages before + committing the B-tree mini-transaction. If + the system crashes so that crash recovery + will not replay the mtr_commit(&mtr), the + big_rec pages will be left orphaned until + the pages are allocated for something else. + + TODO: If the allocation extends the tablespace, it + will not be redo logged, in either mini-transaction. + Tablespace extension should be redo-logged in the + big_rec mini-transaction, so that recovery will not + fail when the big_rec was written to the extended + portion of the file, in case the file was somehow + truncated in the crash. */ - ut_a(btr_pcur_restore_position(BTR_MODIFY_TREE, pcur, mtr)); rec = btr_cur_get_rec(btr_cur); + DEBUG_SYNC_C("before_row_upd_extern"); err = btr_store_big_rec_extern_fields( index, btr_cur_get_block(btr_cur), rec, rec_get_offsets(rec, index, offsets_, ULINT_UNDEFINED, &heap), - mtr, TRUE, big_rec); - mtr_commit(mtr); + big_rec, mtr, BTR_STORE_UPDATE); + DEBUG_SYNC_C("after_row_upd_extern"); + /* If writing big_rec fails (for example, because of + DB_OUT_OF_FILE_SPACE), the record will be corrupted. + Even if we did not update any externally stored + columns, our update could cause the record to grow so + that a non-updated column was selected for external + storage. This non-update would not have been written + to the undo log, and thus the record cannot be rolled + back. + + However, because we have not executed mtr_commit(mtr) + yet, the update will not be replayed in crash + recovery, and the following assertion failure will + effectively "roll back" the operation. */ + ut_a(err == DB_SUCCESS); } + mtr_commit(mtr); + if (UNIV_LIKELY_NULL(heap)) { mem_heap_free(heap); } |