| Bug #44670 | ALTER TABLE crashes in StorageTableShare::haveIndexes() | ||
|---|---|---|---|
| Submitted: | 5 May 2009 13:34 | Modified: | 26 May 2010 17:46 |
| Reporter: | Olav Sandstå | Email Updates: | |
| Status: | Unsupported | Impact on me: | |
| Category: | MySQL Server: Falcon storage engine | Severity: | S2 (Serious) |
| Version: | 6.0.12-alpha | OS: | Any |
| Assigned to: | Christopher Powers | CPU Architecture: | Any |
| Tags: | F_ONLINE ALTER | ||
[5 Aug 2009 3:10]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/80115 2764 Christopher Powers 2009-08-04 Bug #44670, "ALTER TABLE crashes in StorageTableShare::haveIndexes()" Disabled unnecessary and unprotected check of StorageTableShare::indexes. modified: storage/falcon/StorageTableShare.cpp
[5 Aug 2009 3:10]
Christopher Powers
StorageTableShare::haveIndexes() performs an unnecessary and unprotected check of the ::indexes linked list.
[5 Aug 2009 23:38]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/80230 2765 Christopher Powers 2009-08-05 Bug#44670, "ALTER TABLE crashes in StorageTableShare::haveIndexes()" Don't use StorageTableShare::haveIndexes() at all modified: storage/falcon/ha_falcon.cpp
[17 Aug 2009 18:32]
Kevin Lewis
Falcon cannot crash in StorageTableShare::haveIndexes() if it is not called. Without it, StorageInterface::remapIndexes() is just called more often. But StorageInterface::setIndexes() is only called during create() and open(), which is not often enough to worry about performance. Note; StorageTableShare::haveIndexes() should be deleted later as unused code.
[26 Aug 2009 2:52]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/81565 2770 Christopher Powers 2009-08-25 Bug#44670 "ALTER TABLE crashes in StorageTableShare::haveIndexes()" Restore call to modified StorageTableShare::haveIndexes() to avoid deadlock. modified: storage/falcon/StorageTableShare.cpp storage/falcon/StorageTableShare.h storage/falcon/ha_falcon.cpp
[14 Sep 2009 6:41]
Christopher Powers
[Assigned patch to wrong bug, Bug#46670] A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/81568 2771 Christopher Powers 2009-08-25 Bug#46670 "ALTER TABLE crashes in StorageTableShare::haveIndexes()" Rewrote StorageTableShare::haveIndexes() to more effectively check for the need to remap Falcon indexes modified: storage/falcon/StorageTableShare.cpp storage/falcon/StorageTableShare.h storage/falcon/ha_falcon.cpp
Description: A run of the falcon_online_alter test resulted in the following crash: # 12:15:02 Program terminated with signal 10, Bus error. # 12:15:02 #0 0xfef411e4 in _lwp_kill () from /lib/libc.so.1 # 12:15:02 #0 0xfef411e4 in _lwp_kill () from /lib/libc.so.1 # 12:15:02 #1 0x00a2b30c in my_write_core (sig=10) at stacktrace.c:309 # 12:15:02 #2 0x002a859c in handle_segfault (sig=10) at mysqld.cc:2710 # 12:15:02 #3 <signal handler called> # 12:15:02 #4 0x00644314 in StorageTableShare::haveIndexes (this=0x1e34838, # 12:15:02 indexCount=13) at StorageTableShare.cpp:574 # 12:15:02 #5 0x0061f0a4 in StorageInterface::setIndexes (this=0x782b918, # 12:15:02 srvTable=0x7875928) at ha_falcon.cpp:2720 # 12:15:02 #6 0x006258f0 in StorageInterface::open (this=0x782b918, # 12:15:02 name=0x7502398 "./test/B", mode=2, test_if_locked=2) at ha_falcon.cpp:590 # 12:15:02 #7 0x0045efa8 in handler::ha_open (this=0x782b918, table_arg=0x7875928, # 12:15:02 name=0x7502398 "./test/B", mode=2, test_if_locked=2) at handler.cc:2043 # 12:15:02 #8 0x0033bb20 in open_table_from_share (thd=0x7aa1908, share=0x7502128, # 12:15:02 alias=0x7885a58 "B", db_stat=39, prgflag=44, ha_open_flags=0, # 12:15:02 outparam=0x7875928, open_mode=OTM_OPEN) at table.cc:2025 # 12:15:02 #9 0x003340b4 in open_table (thd=0x7aa1908, table_list=0x7885a60, # 12:15:02 mem_root=0xfd4f02c8, action=0xfd4f0308, flags=32) at sql_base.cc:2777 # 12:15:02 #10 0x0033679c in open_tables (thd=0x7aa1908, start=0xfd4f0418, # 12:15:02 counter=0xfd4f03b8, flags=32) at sql_base.cc:3772 # 12:15:02 #11 0x00337204 in open_and_lock_tables_derived (thd=0x7aa1908, # 12:15:02 tables=0x7885a60, derived=false, flags=32) at sql_base.cc:4233 # 12:15:02 #12 0x003376f0 in open_n_lock_single_table (thd=0x7aa1908, table_l=0x7885a60, # 12:15:02 lock_type=TL_WRITE_ALLOW_READ, flags=32) at sql_base.cc:4083 # 12:15:02 #13 0x00495064 in mysql_alter_table (thd=0x7aa1908, new_db=0x7885c68 "test", # 12:15:02 new_name=0x0, create_info=0xfd4f1060, table_list=0x7885a60, # 12:15:02 alter_info=0xfd4f1560, order_num=0, order=0x0, ignore=false) # 12:15:02 at sql_table.cc:6768 # 12:15:02 #14 0x002c9c84 in mysql_execute_command (thd=0x7aa1908) at sql_parse.cc:2892 # 12:15:02 #15 0x002d20ac in mysql_parse (thd=0x7aa1908, # 12:15:02 inBuf=0x7885900 "ALTER TABLE `B` ADD PRIMARY KEY t ( `datetime_nokey` , `varchar_nokey` ) , ADD UNIQUE s ( `time_key` , `pk` )", length=110, # 12:15:02 found_semicolon=0xfd4f1dc8) at sql_parse.cc:5937 # 12:15:02 #16 0x002d2bd0 in dispatch_command (command=COM_QUERY, thd=0x7aa1908, # 12:15:02 packet=0x7b558f1 "ALTER TABLE `B` ADD PRIMARY KEY t ( `datetime_nokey` , `varchar_nokey` ) , ADD UNIQUE s ( `time_key` , `pk` )", packet_length=110) # 12:15:02 at sql_parse.cc:1049 # 12:15:02 #17 0x002d4368 in do_command (thd=0x7aa1908) at sql_parse.cc:731 # 12:15:02 #18 0x002bb7fc in handle_one_connection (arg=0x7aa1908) at sql_connect.cc:1146 # 12:15:02 #19 0xfef3fff8 in _lwp_start () from /lib/libc.so.1 # 12:15:02 #20 0xfef3fff8 in _lwp_start () from /lib/libc.so.1 The code where the crash occured: for (StorageIndexDesc *indexDesc = indexes; indexDesc; indexDesc = indexDesc->next, n++) { if (!indexDesc->index) <<<< line causing the SIGBUS error return false; } How to repeat: Seen once when running the RGQ test falcon_online_alter. Suggested fix: Don't miss the BUS.