[Pgpool-hackers] [Pgpool-committers] pgpool - pgpool-II: Fix child process crashes by accessing illegal

Tatsuo Ishii ishii at sraoss.co.jp
Mon Oct 31 05:06:22 UTC 2011


>> Toshihiro,
>> 
>> I think you need to explain why you committed this.
>> 
>> It seems you patches just postpone the problem we currently have.
> 
> Yes, this is the temporary fix until we fix a fundamental problem
> which causes segfault at memory context.
> 
>> In your patches if we have problem with pool_memory_free(), you just do nothing and instead do:
>> 
>>    pool_debug("An adress \"%p\" does not exist in memory pool.", chunk);
>> 
>> Because this is just a debug statement and debug mode never turned on
>> in the production environment, I think this makes things even
>> worse. Because the patches silently ignore the problem currently we
>> have, it is possible that we later on have even worse problem.
>> 
>> I think at least we should change above to pool_error().
> 
> This fix is based on the core file which a user offered.

I would say it's not a fix.

> I could not reproduce the problem and find the location which cause
> the breakage of memory context. However the user said it reappears
> by all means when I operate application with a certain hardware.
> 
> I am concerned about this fix having possibilities to emit a large
> quantity of error logs by an unknown certain operation.

No way. Please change pool_debug() to pool_error().

>> From: kitagawa at pgfoundry.org (User Kitagawa)
>> Subject: [Pgpool-committers] pgpool - pgpool-II: Fix child process crashes by accessing illegal
>> Date: Mon, 31 Oct 2011 02:35:33 +0000 (UTC)
>> Message-ID: <20111031023533.702FA532C80E at pgfoundry.org>
>> 
>> > Log Message:
>> > -----------
>> > Fix child process crashes by accessing illegal memory context.
>> > 
>> > Tags:
>> > ----
>> > V3_0_STABLE
>> > 
>> > Modified Files:
>> > --------------
>> >     pgpool-II/parser:
>> >         pool_memory.c (r1.10 -> r1.10.8.1)
>> >         (http://cvs.pgfoundry.org/cgi-bin/cvsweb.cgi/pgpool/pgpool-II/parser/pool_memory.c?r1=1.10&r2=1.10.8.1)
>> > _______________________________________________
>> > Pgpool-committers mailing list
>> > Pgpool-committers at pgfoundry.org
>> > http://pgfoundry.org/mailman/listinfo/pgpool-committers
>> 
> 


More information about the Pgpool-hackers mailing list