On Sat, February 25, 2006 10:59 am, Aaron Stone wrote:
> Ok, let's see about this one. One of the processes in here gets itself
> into some major trouble!

I belive it is the imap daemon from what I saw with that idle connection
but i havn't had time to test it.

>
> Also, in build_args_array_ext in dbmail-imapsession.c, it looks like
> memory is being allocated into a global array named the_args, but it's
> already allocated on the process stack. That's not so good I don't think.

Yeah thus the comment in my patch around it :) I didn't see an easy fix
otherwise I would have given one. It can leak up to about 1k per
connection from what I have seen.

-leif

>
> Aaron
>
>
> On Fri, 2006-02-24 at 20:01 -0500, Leif Jackson wrote:
>
>> Aaron,
>>
>>
>> I saw a way to repoduce the SIGALRM, when running the daemons under
>> valgrind with the tracechildren option I mentioned in my submission
>> (with
>> dbmail.supp) you can connect to any of the daemons from the network and
>> then just let the connection sit open till the OS cleans up the idle
>> connection (I saw this from someone scanning my network a day or so
>> ago.), this will send the signal SIGALRM to the daemon, as far as I can
>> tell on my Linux FC2/ CentOS x86_64 4.2 machines. Or could you use kill
>> -s SIGALRM
>> <pid of -n running dameon> ?
>>
>>
>>
>> Just my .02
>>
>>
>> -leif
>>
>>
>> On Fri, February 24, 2006 6:44 pm, Aaron Stone wrote:
>>
>>> Oleg, Paul and I have started seeing weird segfaults in the daemons.
>>> Possibly starting from about two weeks ago, although we're not totally
>>>  sure. I have some evidence (posted earlier this week) that it's in
>>> the SIGALRM handler (maybe all signal handlers?), and that's what's
>>> causing the "glib: corrupted double linked list" error messages.
>>>
>>> What we need now are some good trace logs, gdb logs, valgrind logs,
>>> to help narrow down when things go crazy and what leads up to it. Time
>>> for more eyeballs to make this bug shallow!
>>>
>>> DBMail 2.1.4 will be released as soon as we figure this out!
>>>
>>>
>>>
>>> Aaron
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Dbmail-dev mailing list
>>> [email protected]
>>> http://twister.fastxs.net/mailman/listinfo/dbmail-dev
>>>
>>>
>>>
>>
>>
>> _______________________________________________
>> Dbmail-dev mailing list
>> [email protected]
>> http://twister.fastxs.net/mailman/listinfo/dbmail-dev
>>
> _______________________________________________
> Dbmail-dev mailing list
> [email protected]
> http://twister.fastxs.net/mailman/listinfo/dbmail-dev
>
>


Reply via email to