MySQL Bugs: #63810: Failing assertion: block->page.file_page_was

Bug #63810	Failing assertion: block->page.file_page_was_freed == FALSE
Submitted:	20 Dec 2011 17:41	Modified:	14 Jan 2012 11:22
Reporter:	Davi Arnaut (OCA)	Email Updates:
Status:	Can't repeat	Impact on me:	None
Category:	MySQL Server: InnoDB storage engine	Severity:	S3 (Non-critical)
Version:	5.5.19	OS:	Any
Assigned to:		CPU Architecture:	Any

Description:
InnoDB assertion during a RQG run of the varchar scenario:

111220  6:21:19  InnoDB: Assertion failure in thread 1137543488 in file buf0buf.c line 2837
InnoDB: Failing assertion: block->page.file_page_was_freed == FALSE
InnoDB: We intentionally generate a memory trap.

Backtrace:

#0  0x00002b5e4de35d02 in pthread_kill () from /lib64/libpthread.so.0
#0  0x00002b5e4de35d02 in pthread_kill () from /lib64/libpthread.so.0
#1  0x00000000007a0c78 in my_write_core ()
#2  0x00000000004e9955 in handle_segfault ()
#3  <signal handler called>
#4  0x00002b5e4ed54265 in raise () from /lib64/libc.so.6
#5  0x00002b5e4ed55d10 in abort () from /lib64/libc.so.6
#6  0x000000000088ef98 in buf_page_get_known_nowait ()
#7  0x00000000008e8038 in ibuf_merge_or_delete_for_page ()
#8  0x00000000008897e9 in buf_page_io_complete ()
#9  0x00000000008c5bc6 in fil_aio_wait ()
#10 0x00000000008360ad in io_handler_thread ()
#11 0x00002b5e4de3073d in start_thread () from /lib64/libpthread.so.0
#12 0x00002b5e4edf7f6d in clone () from /lib64/libc.so.6

How to repeat:
RQG command line:

randgen-2.2-snapshot/gentest.pl \
--dsn=dbi:mysql:host=127.0.0.1:port=6400:user=root:database=test \
--logconf=/tmp/var-randgen/std_data/randgen/log.conf \
--gendata=randgen-2.2-snapshot/conf/engines/varchar.zz \ 
--grammar=randgen-2.2-snapshot/conf/engines/varchar.yy \
 --reporter=Backtrace,ErrorLog \
 --engine=InnoDB \
 --queries=1M \
 --rows=10000 \
 --seed=time \
 --duration=600 \

# 2011-12-20'T'06:10:06 Converting --seed=time to --seed=1324350606

With recent enough 5.5.20-debug on Mac I've got assertion failure in another place while running this test:

111217 20:35:07 [Note] /Users/openxs/dbs/5.5/bin/mysqld: ready for connections.
Version: '5.5.20-debug'  socket: '/tmp/mysql.sock'  port: 3306  Source distribution
111217 21:22:39 mysqld_safe A mysqld process already exists
Assertion failed: (min_key || max_key), function records_in_range, file /Users/openxs/bzr2/mysql-5.5-work/storage/innobase/handler/ha_innodb.cc, line 7530.
18:51:45 UTC - mysqld got signal 6 ;
...

Also we still have bug #58815 as verified.

Same assertion with 5.5.21-debug on Linux:

...
mysqld: /home/openxs/bzr/mysql-5.5-work/storage/innobase/handler/ha_innodb.cc:7529: virtual ha_rows ha_innobase::records_in_range(uint, key_range*, key_range*): Assertion `min_key || max_key' failed.
07:37:48 UTC - mysqld got signal 6 ;
...

The above assertion is probably Bug#63618. You would have to comment that out to reproduce this one.

Now I see why assertion looks familiar... Anyway, after commenting it out that test run passed OK with 5.5.20-debug on Mac OS X.

Thank you for the feedback.

Crash is not repeatable for me on Linux too. Looks like this bug solved by same code modification which fixed bug #58815