Re: [exim] postfix ssl_accept error and exim remote_smtp def…

Top Page
Delete this message
Reply to this message
Author: Jelle de Jong
Date:  
To: exim-users
Subject: Re: [exim] postfix ssl_accept error and exim remote_smtp defer terminated by signal 11
Phil Pennock wrote:
> On 2009-06-19 at 19:09 +0200, Jelle de Jong wrote:
>> I would also like to point that signall 11 is SIGSEGV Core Invalid
>> memory reference aka segmentation fault that happens on the exim side,
>> that does not sound good, but I don't know why it is happening, I hope
>> somebody can help.
>
> This is not normally a sign of a problem in Exim itself, which has a
> mature codebase, but a sign of something like a library conflict, with
> Exim built against one version of a library and then another version
> being used.
>
> The next version of Exim will hopefully report conflicts like this; see:
> http://bugs.exim.org/show_bug.cgi?id=745
>
>> The postfix server has no other ssl_accept errors and is able to receive
>> tls secure mail from other servers. I just don't see what goes wrong.
>
> nb: -qff runs the entire queue, you want -M <message id> instead.
>
> exim-server-a-information-discreet.txt shows successful delivery over
> SSL, with the message accepted by the remote side. Twice.
>
> I'm not seeing here evidence of the problem which you're reporting.
>
> I don't read postfix logs well, but well enough to see that you didn't
> censor the logs quite well enough to avoid showing via /etc/mailname
> which serverb is. Note though that censoring data from the people
> you're asking for help is somewhat frowned upon.
>
> In this case, you seem to be showing logs from a different connection,
> since postfix is complaining:
> Jun 16 10:42:04 emily postfix/smtpd[25488]: SSL_accept:before/accept initialization
> Jun 16 10:42:04 emily postfix/smtpd[25488]: read from B88E0F90 [B88EAC50] (11 bytes => -1 (0xFFFFFFFF))
> Jun 16 10:42:04 emily postfix/smtpd[25488]: SSL_accept error from alphaip2.servera.nl[12.34.56.789]: -1
>
> and since server-a showed successful delivery, with the message accepted
> and logging the recipient's transaction message id (BAB1C5E883 and then
> later 4D1135E883), these clearly don't line up.
>
> -Phil
>


Hi Phil thank you for taking the time to look at the problem. I will ask
for a -M <message id> log. And sorry for not making clear what test were
done for the log files so I rearranged some logs and made a readme for
information explaining the logs. I am also really sorry for the
censoring of the data I totally understand that this is harder to read.

The hosting company of server-a (exim) asked with strong emphasis to
filter all ip addresses and domains directing to there company. I am all
for transparency and easy to read logs, but I will respect there request
for privacy.

server-a is the exim server sending the mail via some webform system

server-b is the postfix server trying to receive the mail

server-a is of an other hosting company outside my scope of control but
i can sent mail to them asking for information and help.

server-b is is my own controlled server, I can generate every logs you
want no censoring needed.

terms that are used for a replacement:
- alpha
- servera
- golf
- 12.34.56.789

exim-server-a-information.txt
this is the information I received from the hosting company it appears
that when the server tries to send mail via a webform there processes
generates signal 11 segmentation fault. probably a library mismatch with
ssl systems or some resource limitation. What commands could be helpful
to receive more information.

openssl-server-a-check.txt
this is a separate test to see if server-a and openssl is not complete
broken this is an incoming test and the problem is outgoing but this way
i can test some certificates. The certificates are full with localhost
and some... there is nothing indicating the relation of the certificated
with the server I don't know if this could cause any problems but I
don't think this is the problem in this case.

openssl-helmwijk-check.txt
this test has nothing to do with server-a it is an test from other
systems making tls connection with the postfix server to test the
delivery of mails. no filters here since it has nothing to do with the
server-a hosting company.

postconf-n-helmwijk.txt
my postfix configuration i also use smtpd -v in the master file, nothing
filtered since it has nothing to do with the server-a

ssldump-smtpd-v-helmwijk-server-a-fail.txt
this is important information from the postfix server that tries to
receive mail from the exim server systems. you can see that when my
server wants to talk starttls the exim server just hangs up the line.
(segmentation fault on exim side) this log is filtered.

I hope this provide some better information about what happens.

Thanks in advance,

Best regards,

Jelle de Jong