The mailing list for listmasters using Sympa

[Erik Olson <address@concealed>]

Follow-up.  Figured out proper incantation to get core dumps from apache 
cgi processes...  again, apologies, I'm learning as I go. Here's the start 
of the backtrace.  Interestingly, all four cases (I outlined in the 
original post) die on this same line 232 of mg.c



[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
Core was generated by `/usr/bin/perl -U -d:Trace 
/home/sympa/bin/wwsympa.fcgi'.
Program terminated with signal 11, Segmentation fault.
#0  0x00007fe01a823821 in Perl_mg_get (my_perl=0xbd3010, sv=0x61ac888) at 
mg.c:232
232             if (!(mg->mg_flags & MGf_GSKIP) && vtbl && vtbl->svt_get) {
(gdb) bt
#0  0x00007fe01a823821 in Perl_mg_get (my_perl=0xbd3010, sv=0x61ac888) at 
mg.c:232
#1  0x00007fe01a8467e1 in Perl_sv_setsv_flags (my_perl=0xbd3010, 
dstr=0x68ebe08, sstr=0x61ac888, flags=18) at sv.c:4097
#2  0x00007fe01a84754b in Perl_newSVsv (my_perl=0xbd3010, old=<optimized 
out>) at sv.c:8733
#3  0x00007fe01a7fd3d6 in S_pad_findlex (my_perl=0xbd3010, name=<optimized out>, 
cv=<optimized out>, seq=<optimized out>, warn=1, out_capture=0x0,
[...]

(gdb) print mg
$1 = (MAGIC *) 0xbd6d28
(gdb) print mg->mg_flags
$2 = 0 '\000'
(gdb) print vtbl
$3 = (const MGVTBL * const) 0x2200000c00000003
(gdb) print vtbl->svt_get
Cannot access memory at address 0x2200000c00000003
(gdb) print mg->mg_virtual
$4 = (MGVTBL *) 0x2200000c00000003

So looks like it's accessing freed or bogus memory.

  - Erik

On Sun, 1 Jan 2012, Erik Olson wrote:

> Yes, this is also a change to perl 5.14.2 from some previous version.
>
> I started adding debug print messages to the perl modules themselves, and 
> have confirmed that the line-by-line traces I sent in my original mail are 
> probably accurate.  In both the post case and the subscribe case, it's on a 
> call to MIME::Parser->parse_data and I don't see anything obviously wrong 
> with the data... and the actual blow up occurs several levels deeper, on 
> IO::File -- perhaps attempting to open a file it shouldn't?
>
> Thanks for the confirmation!
>
>  - Erik
>
> On Sun, 1 Jan 2012, Davy Kager wrote:
>
> > Hello,
> >
> > I am experiencing what I believe to be an identical issue, see 
> > https://listes.cru.fr/sympa/arc/sympa-users/2011-12/msg00022.html
> >
> > For me this came about when upgrading from openSUSE 11.4 to openSUSE 12.1 
> > (x86_64).  This upgrade also carried an upgrade to Perl, from version 
> > 5.12.x to 5.14.2.  I would say that is the problem (or possibly one of the 
> > upgraded modules?).
> >
> > Kind regards,
> > Davy
> >
> > -----Original Message-----
> > From: address@concealed [mailto:address@concealed] On 
> > Behalf Of Erik Olson
> > Sent: Sunday, January 01, 2012 06:02
> > To: address@concealed
> > Subject: [sympa-users] perl crashing on wwsympa.fcgi
> >
> > I am wondering if anyone can help me here.
> >
> > I had been running Sympa 6.1.6 on a 64-bit Fedora 12 Linux box.
> >
> > This week, I transferred all the files from ~sympa and appropriate /etc 
> > directories to a brand new rebuilt 64-bit Fedora 16 box.  Restored the 
> > database on the new box, and I also upgraded to Sympa 6.1.7 in the process.
> >
> > Things are mostly working, but wwsympa.cgi is now page faulting in 
> > libperl.so when certain functions are accessed:
> >
> >   1) Posting to a list via the web
> >
> >   2) Subscribing a user to a list via the web (but only when Quiet is
> >      unchecked... if Quiet is checked, it works)
> >
> >   3) Attempting to approve a moderated message, whether it be via a ticket
> >      link, or clicking on "Moderate -> Message"
> >
> >   4) User who has been subscribed but never logged in attempts to set up a
> >         password by clicking on the "First Login" link.  wwsympa.fcgi
> > 	blows up when they enter their e-mail address and click "Request 
> > First
> >         Password"
> >
> > I am guessing the common thread here is that all these functions involve 
> > sending an e-mail directly from wwsympa.fcgi, or maybe creating/dealing 
> > with tickets?
> >
> > Here is the relevant line from syslog:
> >
> > [...]
> > Dec 31 20:39:59 pulcher wwsympa[6729]: info [robot gsas.org] [session 
> > 11550608881860] [client 192.168.1.2] main::do_requestpasswd() 
> > do_requestpasswd(address@concealed) Dec 31 20:39:59 pulcher wwsympa[6729]: info 
> > Auth::create_one_time_ticket() 
> > Auth::create_one_time_ticket(address@concealed,gsas.org,choosepasswd,192.168.1.2) 
> > value = 67959831025349 Dec 31 20:39:59 pulcher kernel: [  587.774025] 
> > wwsympa.fcgi[6729] general protection ip:7f22b4fd5821 sp:7fff4074bc60 
> > error:0 in libperl.so[7f22b4f3f000+171000]
> >
> >
> > I am not a perl expert, nor familiar with how to get any form of a 
> > backtrace when perl itself crashes.  I did try adding -d Devel::Trace to 
> > the execution of wwsympa.fcgi, and examined the line-by-line traces of each 
> > module.  I've enclosed partial traces from each of these four cases right 
> > up to the crashes.  I don't see a lot of commonality other than the use of 
> > mail.pm
> >
> > Can anyone help?
> >
> > Thanks in advance,
> >
> >   - Erik Olson
> >
> >
> > --
> > Erik Olson
> > Sent from my spiffy new Linux box

--
Erik Olson
Sent from my spiffy new Linux box