Bug#402164: it isn't about ldap, just Cyrus 2.1/2.2 and GSSAPI

[Russ Allbery]

Roberto C. Sánchez <roberto at connexer.com> writes:
> On Tue, Oct 07, 2008 at 07:46:45PM +1300, Dennis Vshivkov wrote:

>>     Program received signal SIGSEGV, Segmentation fault.
>>     [Switching to Thread -1212221216 (LWP 16246)]
>>     0x00000000 in ?? ()
>>     (gdb) bt
>>     #0  0x00000000 in ?? ()
>>     #1  0xb65ca1a8 in ?? () from /usr/lib/libgssapi_krb5.so.2
>>     #2  0xb65e1e88 in ?? () from /usr/lib/libgssapi_krb5.so.2
>>     #3  0xb65ca8f0 in ?? () from /usr/lib/libgssapi_krb5.so.2
>>     #4  0x00000000 in ?? ()
>>     (gdb) _

>> Maybe the whole problem belongs to the krb5 package instead?

> Sam, Russ, and Ben,

> Bug #402164 was reported against the cyrus-sasl2 package quite some time
> ago.  The initial report supposed that the problem was in libsasl2-2 or
> somewhere in NSS/LDAP, while the above follow-up indicated that perhaps
> the problem is in krb5.

> Do you mind taking a look to see and perhaps either reassigning the bug
> to krb5 or if something jumps out at you, perhaps making a suggestion on
> a solution?

There isn't really enough information in that bug to be able to figure out
what's going on.  If the analysis that a mutex is being set to an invalid
pointer is correct, that could well cause a segfault in the Kerberos
libraries without being a Kerberos problem.  But the only backtrace in
that bug report was made without debugging symbols, so it's pretty hard to
figure out what might be happening.

It's not even clear to me that the two reports in the bug (the one from
2006 and the one from 2008) are actually the same problem.  Just because
they both had segfaults doesn't mean they're the same issue.

This is the sort of bug where I'm dubious there's much benefit to holding
it open unless someone can reproduce it with the packages currently in the
archive, as unsatisfying as that approach is.  The last report was six
years ago, and it was a segfault in Cyrus imapd.  If Cyrus impad were
regularly segfaulting for lots of users, there would be a lot more bug
reports, so we know it's some sort of edge case.  If that edge case isn't
still happening, I'm not sure there's any real action that you can take at
this point.

-- 
Russ Allbery (rra at debian.org)               <http://www.eyrie.org/~eagle/>