On 23 May 2003, Philip Hazel wrote:
On 23 May 2003, Nigel Metheringham wrote:
> > The old behaviour of exim (pre 4.14) on new linux kernels, where
> > reports of the form:-
> > May 23 04:37:38 pillarbox kernel: application bug: exim(1877) has SIGCHLD set to SIG_IGN but calls wait().
> > May 23 04:37:38 pillarbox kernel: (see the NOTES section of 'man 2 wait'). Workaround activated.
> >
> > are seen on all deliveries, appears to be reoccurring in 4.20 - looks
> > like the fix got backed out or worked round.
>
> You need to have SIG_IGN_WORKS defined to skip the code that causes
> this. It should be defined in OS/os.h-Linux.
I'm currently seeing the same problem using the pam_exim helper program
from
http://www.e-admin.de/pam_exim/. That fix uses a setuid PAM helper
executable, which presumably is the subprocess involved.
Could this error be causing pam_exim to fail? The exim_chkpwd binary is
logging success when passed the right credentials via AUTH BASIC, but PAM
is logging a failure and exim returns a 535.
Jun 3 06:29:40 mail exim(pam_exim)[1312]: args to give gowen
Jun 3 06:29:40 mail kernel: application bug: exim(1311) has SIGCHLD set to SIG_IGN but calls wait().
Jun 3 06:29:40 mail kernel: (see the NOTES section of 'man 2 wait'). Workaround activated.
Jun 3 06:29:40 mail exim_chkpwd[1312]: pass okay - returnig UNIX_PASSED
Jun 3 06:29:40 mail exim(pam_exim)[1311]: authentication failure; logname= uid=95 euid=95 tty= ruser= rhost= user=gowen
I realize that the pam_exim isn't part of the exim distribution, but I
thought I'd post this to this list because I saw the previous post about
SIGCHLD and SIG_IGN and thought this might be relevant; apologies if not.
--
gowen -- Greg Owen -- gowen@???
79A7 4063 96B6 9974 86CA 3BEF 521C 860F 5A93 D66D
