Random error with sasl+pam+mysql

[Dan White]

On 10/11/12 13:17 +0200, Tobias Kirchhofer wrote:
>Hello list,
>
>my setup is Ubuntu 12.04, postfix, cyrus, cyrus-sasl and pam-mysql 
>using standard dist-packages. MySQL is on another server on the same 
>machine in a openvz container.
>
>After around 3 days sasl starts with erratic behaviour. Sometimes 
>authentification works, sometimes not. All of a sudden. I tried 
>tessaslauthd from console with success, next time with an error.
>
>root at mail:~# testsaslauthd -u user at example.com -p XXX -f 
>/var/spool/postfix/var/run/saslauthd/mux
>0: NO "authentication failed"
>root at mail:~# testsaslauthd -u user at example.com -p XXX -f 
>/var/spool/postfix/var/run/saslauthd/mux
>0: OK "Success."
>
>Restart of saslauthd daemon fixes everything until around 3 days when 
>the misbehaviour starts again.

That could be a file descriptor (open socket) leak. Compare the output
of 'netstat -anp' between the two situations.

>I observed everything carefully with all relevant debugging options 
>but no real insights. Below you find some strace from the saslauthd. 
>First part is not working, second part works.
>
>Does someone has an idea to solve or give me a hint what else i could 
>do to trace down the problem?

Try starting saslauthd with '-n 0' which will cause saslauthd to fork a
new process for each connection. You might also be able to mask the problem
to some extent by enabling caching with '-c'.

Check the upstream changelog from the pam-mysql maintainer to see if this
is a known issue.

>26722 sendto(8, "\"0\1\0\0\1\0\0\0\0\0\0\5mysql\f<cut>\3"..., 

You might have revealed a substring of your password to the mailing list
here.

-- 
Dan White