Postfix 2.9.1 Crashing with Signal 6 - Postfix 2.8.5-2 works perfectly

[joni]

Just upgraded from 11.10 to 12.04 and sadly noticed that Postfix 2.9.1 breaks my virtual mail setup. It worked perfectly in 11.10.
My setup is using courier and postfix with mysql, Saslauthdm (for auth) and amavis (which I disabled for bug researching).

Postfix refuses any action, sending and receiving, in 12.04 with the following log:
postfix/master[19710]: warning: process /usr/lib/postfix/cleanup pid 19734 killed by signal 6

I have not found any information how to fix this, so i thought I better report it before 12.04 is out. The interesting thing is that if a install postfix 2.8.5-1 from oneiric everything works again, so this has to be a bug according to postfix 2.9.1.

[Scott Kitterman (kitterman)]

What comes right after:

warning: /usr/lib/postfix/smtpd: bad command startup -- throttling

[joni (joni93)]

There is nothing. The next log is a minute later and it is a IMAP logon. Then there is exactly the same log, ending with
Apr 1 18:23:33 home postfix/master[17125]: warning: process /usr/lib/postfix/smtpd pid 17170 killed by signal 6
Apr 1 18:23:33 home postfix/master[17125]: warning: /usr/lib/postfix/smtpd: bad command startup -- throttling.

If there is any other information that you need, tell me and i will deliver.

[joni (joni93)]

Even with -vvv logging of postfix this is the last message. :(

[M4dMike (m4dmike)]

I got exactly the same Error, after upgrading my Ubuntu 11.04 to 12.04.

[LaMont Jones (lamont)]

I don't suppose that /var/spool/postfix has a core file in it? if so, that would be wonderful to have

[joni (joni93)]

One interesting thing i discovered, everything works if postfix does not run in chroot.

[joni (joni93)]

I have to correct myself, it just works partially.

If postfix (2.9.1) runs in non chroot, everything works (i think) but mails that come from fetchmail trigger the same error.
Postfix (2.8.5) in non chroot works like in chroot, without any problems.

[Milan (public-mjh)]

I observe the very same behaviour on 12.04: The processes cleanup, rewrite and smtpd crash with signal 6 if they are within a chroot. (In the default configuration, chroots are used for these processes and it works here e.g. on Ubuntu 10.04).

Sample log message:

Apr 8 15:57:29 two postfix/master[7938]: warning: process /usr/lib/postfix/smtpd pid 7946 killed by signal 6
Apr 8 15:57:29 two postfix/master[7938]: warning: /usr/lib/postfix/smtpd: bad command startup -- throttling

I did an "strace -s 4096 -f" of the postfix master process with smtpd et al. in chroot, and I believe that these are the relevant parts:

(process 7938 is the postfix master process)

7938 clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7f71f2a439d0) = 7946
...
7946 execve("/usr/lib/postfix/smtpd", ["smtpd", "-n", "127.0.0.1:25", "-t", "inet", "-u", "-c", "-o", "smtpd_tls_auth_only=no"], [/* 4 vars */]) = 0
...
7946 chroot("/var/spool/postfix") = 0
...
7946 clone(child_stack=0x7f95d51cbff0, flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID, parent_tidptr=0x7f95d51cc9d0, tls=0x7f95d51cc700, child_tidptr=0x7f95d51cc9d0) = 7954
7954 set_robust_list(0x7f95d51cc9e0, 0x18) = 0
7954 open("/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)
...
7954 open("/dev/tty", O_RDWR|O_NOCTTY|O_NONBLOCK) = -1 ENOENT (No such file or directory)
7954 writev(2, [{"libgcc_s.so.1 must be installed for pthread_cancel to work\n", 59}], 1) = 59
7954 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f95d88ca000
7954 rt_sigprocmask(SIG_UNBLOCK, [ABRT], NULL, 8) = 0
7954 tgkill(7946, 7954, SIGABRT) = 0
7954 --- SIGABRT (Aborted) @ 0 (0) ---
...
7946 +++ killed by SIGABRT +++
...
7938 wait4(-1, [{WIFSIGNALED(s) && WTERMSIG(s) == SIGABRT}], WNOHANG, NULL) = 7946
7938 sendto(3, "<20>Apr 8 15:57:29 postfix/master[7938]: warning: process /usr/lib/postfix/smtpd pid 7946 killed by signal 6", 109, MSG_NOSIGNAL, NULL, 0) = 109
7938 sendto(3, "<20>Apr 8 15:57:29 postfix/master[7938]: warning: /usr/lib/postfix/smtpd: bad command startup -- throttling", 108, MSG_NOSIGNAL, NULL, 0) = 108

However, I haven't investigated what to do about that.

[LaMont Jones (lamont)]

I am unable to reproduce this, but I also lack the mysql setup to really test this well.

Try this:
# cp /lib/x86_64-linux-gnu/libgcc_s.so.1 /var/spool/postfix/lib/x86_64-linux-gnu/libgcc_s.so.1

and see if the problem goes away, (you may need to repeat that if strace gives you yet a different library that must be installed...

If that doesn't fix it, perhaps you can toss me a more complete sample to reproduce the issue.

thanks,
lamont

[joni (joni93)]

You can Take a Look at this
http://www.howtoforge.com/virtual-users-and-domains-with-postfix-courier-mysql-and-squirrelmail-ubuntu-11.10

This is basically my setup, mine is just a little bit different.
If you follow the tutorial you should be able to reproduce the error.

Thanks,
Joni

[Scott Kitterman (kitterman)]

This is the final freeze day, so we're in a bit of a crunch with all the
developers trying to get as much fixed as they can. It would help hugely if
you could test the proposed package on your setup as it's unlikely any of us
are going to have the time to set up and tear down something that complex
today.

[Milan (public-mjh)]

Your suggestion,

# cp /lib/x86_64-linux-gnu/libgcc_s.so.1 /var/spool/postfix/lib/x86_64-linux-gnu/libgcc_s.so.1

fixes the problem here. I also use postfix-mysql for mail aliases etc. I didn't even know that Postfix needs so many files in its chroot.

Thanks for your help. I assume this will be fixed in the Postfix packages?

Regards,
Milan

[Scott Kitterman (kitterman)]

On Thursday, April 12, 2012 01:25:42 PM you wrote:
> I assume this will be fixed in the Postfix packages?

Yes. Lamont is working on an update now.

[LaMont Jones (lamont)]

I will be in -3, with a changelog entry that is a bit vague about this bug.

[LaMont Jones (lamont)]

fixed in 2.9.1-3

[joni (joni93)]

Fixed for me as well. Really nice support, thank you.

:)