[MDEV-6477] Server crashes in in sf_malloc_usable_size on PPC64 Created: 2014-07-23  Updated: 2014-09-03  Resolved: 2014-09-03

Status: Closed
Project: MariaDB Server
Component/s: None
Affects Version/s: None
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Elena Stepanova Assignee: Elena Stepanova
Resolution: Cannot Reproduce Votes: 0
Labels: powerpc
Environment:

PPC64 # RHEL 6.5 # export PATH=/opt/at7.0/bin:$PATH && cmake -DCMAKE_BUILD_TYPE=Debug && make -j3


Issue Links:
PartOf
is part of MDEV-6478 MariaDB on Power8 Closed

 Description   

#2  0x000000001050aa5c in handle_fatal_signal (sig=11) at 10.0/sql/signal_handler.cc:262
#3  <signal handler called>
#4  0x0000000010d25aa0 in sf_malloc_usable_size (ptr=0x1, is_thread_specific=0xfff78164b38 "") at 10.0/mysys/safemalloc.c:212
#5  0x0000000010d0da84 in my_free (ptr=0x1) at 10.0/mysys/my_malloc.c:214
#6  0x00000000101ce2b0 in Log_event::free_temp_buf (this=0xfff0c07e070) at 10.0/sql/log_event.h:1306
#7  0x00000000101ce1d0 in Log_event::~Log_event (this=0xfff0c07e070, __in_chrg=<optimized out>) at 10.0/sql/log_event.h:1295
#8  0x00000000106806ec in Rows_log_event::~Rows_log_event (this=0xfff0c07e070, __in_chrg=<optimized out>) at 10.0/sql/log_event.cc:9427
#9  0x000000001068c55c in Write_rows_log_event::~Write_rows_log_event (this=0xfff0c07e070, __in_chrg=<optimized out>) at 10.0/sql/log_event.h:4433
#10 0x000000001068c5bc in Write_rows_log_event::~Write_rows_log_event (this=0xfff0c07e070, __in_chrg=<optimized out>) at 10.0/sql/log_event.h:4433
#11 0x0000000010650e20 in MYSQL_BIN_LOG::flush_and_set_pending_rows_event (this=0x11489d90 <mysql_bin_log>, thd=0x100404fc9c0, event=0x0, is_transactional=false) at 10.0/sql/log.cc:5533
#12 0x0000000010242604 in THD::binlog_flush_pending_rows_event (this=0x100404fc9c0, stmt_end=true, is_transactional=false) at 10.0/sql/sql_class.cc:5938
#13 0x0000000010243024 in THD::binlog_query (this=0x100404fc9c0, qtype=THD::ROW_QUERY_TYPE, query_arg=0xfff0c0052b8 "INSERT INTO `view_DD` SELECT * FROM `tmp`", query_len=41, is_trans=false, direct=false, suppress_use=false, errcode=1062) at 10.0/sql/sql_class.cc:6198
#14 0x0000000010267d90 in select_insert::abort_result_set (this=0xfff0c075bc8) at 10.0/sql/sql_insert.cc:3770
#15 0x00000000102c7fcc in handle_select (thd=0x100404fc9c0, lex=0x10040500610, result=0xfff0c075bc8, setup_tables_done_option=1073741824) at 10.0/sql/sql_select.cc:378
#16 0x00000000102889cc in mysql_execute_command (thd=0x100404fc9c0) at 10.0/sql/sql_parse.cc:3522
#17 0x00000000102910e0 in mysql_parse (thd=0x100404fc9c0, rawbuf=0xfff0c0052b8 "INSERT INTO `view_DD` SELECT * FROM `tmp`", length=41, parser_state=0xfff78165f28) at 10.0/sql/sql_parse.cc:6409
#18 0x0000000010282be4 in dispatch_command (command=COM_QUERY, thd=0x100404fc9c0, packet=0x10040502751 " INSERT INTO `view_DD` SELECT * FROM `tmp`", packet_length=42) at 10.0/sql/sql_parse.cc:1309
#19 0x0000000010281db4 in do_command (thd=0x100404fc9c0) at 10.0/sql/sql_parse.cc:1006
#20 0x00000000103eef48 in do_handle_one_connection (thd_arg=0x100404fc9c0) at 10.0/sql/sql_connect.cc:1379
#21 0x00000000103eebac in handle_one_connection (arg=0x100404fc9c0) at 10.0/sql/sql_connect.cc:1293
#22 0x00000fffa19fc548 in start_thread (arg=0xfff78167190) at pthread_create.c:311
#23 0x00000fffa12d3c90 in .__clone () at ../sysdeps/unix/sysv/linux/powerpc/powerpc64/clone.S:106

Stack trace from

revision-id: holyfoot@askmonty.org-20140722102815-y27a18djzjbpnchf
revno: 4301
branch-nick: 10.0-for-stress-tests

RQG command line that triggered the bug (it does not do it every time, so it's just information, not a way to reproduce):

perl ./runall-new.pl \
--queries=100000000 \
--no-mask \
--seed=1406056211 \
--threads=8 \
--duration=600 \
--queries=100M \
--reporters=QueryTimeout,Backtrace,ErrorLog,Deadlock \
--validators=TransformerNoComparator \
--transformers=ExecuteAsDeleteReturning,ExecuteAsPreparedTwice,DisableOptimizations,EnableOptimizations,OrderBy \
--redefine=conf/mariadb/general-workarounds.yy \
--redefine=conf/mariadb/10.0-features-redefine.yy \
--mysqld=--log_output=FILE \
--mysqld=--query_cache_size=64M \
--mysqld=--slow_query_log \
--mysqld=--long_query_time=0.000001 \
--mysqld=--log_bin_trust_function_creators=1 \
--views \
--grammar=conf/runtime/information_schema.yy \
--engine=Aria \
--rpl_mode=mixed \
--mysqld=--slave-skip-errors=1049,1305,1539,1505,1317 \
--use-gtid=current_pos \
--mysqld=--optimizer_switch=extended_keys=on,exists_to_in=on \
--mysqld=--use_stat_tables=PREFERABLY \
--mysqld=--optimizer_selectivity_sampling_limit=100 \
--mysqld=--optimizer_use_condition_selectivity=5 \
--mysqld=--histogram_size=100 \
--mysqld=--histogram_type=DOUBLE_PREC_HB \
--mysqld=--log_slow_verbosity=query_plan,explain \
--mysqld=--slave_parallel_threads=8 \
--mtr-build-thread=300 \
--mask=20623 \
--basedir1=<basedir> --vardir1=<vardir>

randgen/mariadb-patches:

revision-id: elenst@montyprogram.com-20140716103317-2my8cfdk323poca3
revno: 945
branch-nick: randgen-mariadb-patches



 Comments   
Comment by Michael Widenius [ 2014-08-05 ]

This is a similar memory corruption as we have with hash_delete and hash_insert.

I would need a smaller or repeatable test case to be able to find what's going on.

Comment by Elena Stepanova [ 2014-09-03 ]

Couldn't reproduce it any longer after the merge 5.5=>10.0 that included some Monty's fixes, even though the fixes were not supposed to be related to it.

Generated at Thu Feb 08 07:12:12 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.