On Fri, Dec 16, 2011 at 01:51:09PM -0600, Shaun Meyer
wrote:> freebsd-stable@,
>
> I'm building Samba using net/samba36 with the following config:
> LDAP=on "With LDAP support"
> ADS=on "With Active Directory support"
> CUPS=off "With CUPS printing support"
> WINBIND=on "With WinBIND support"
> SWAT=off "With SWAT WebGUI"
> ACL_SUPPORT=on "With ACL support"
> AIO_SUPPORT=off "With Asyncronous IO support"
> FAM_SUPPORT=off "With File Alteration Monitor"
> SYSLOG=off "With Syslog support"
> QUOTAS=on "With Disk quota support"
> UTMP=off "With UTMP accounting support"
> PAM_SMBPASS=off "With PAM authentication vs passdb backends"
> DNSUPDATE=off "With dynamic DNS update(require ADS)"
> AVAHI=off "With Bonjour service discovery support"
> EXP_MODULES=on "With experimental modules"
> POPT=on "With system-wide POPT library"
> IPV6=on "With IPv6 support"
> MAX_DEBUG=off "With maximum debugging"
> SMBTORTURE=off "With smbtorture"
>
> I am able to join winbind to our 2003 domain, no problem. I'm using
> idmap_rid(8) backend and am using nss_winbind to enumerate uid/gid's.
>
> This set up runs fine for a few minutes (10?) until winbind begins
> dumping core when I do a group look up something like `getent group`
> or `id`.
>
> Before the core dumping begins: `wbinfo -g` returns all groups,
> `getent passwd` returns all users, `getent group` returns all groups,
> and `id [user]` returns all information as it should.
>
> After the core dumping begins: `wbinfo -g` returns all groups, `getent
> passwd` returns all users, `getent group` or `id [user]` causes a core
> dump in winbindd. The daemon continues running however.
>
> After the core dumping routine starts, I have to do some gyrations of
> `rm -rf /var/db/samba /usr/local/etc/samba`, restarting services, and
> joining/leaving the domain to get another ~10 minutes of functioning
> winbind.
>
> With MAX_DEBUG=on, I tried to use gdb on the winbindd.core file which
> shows lines like "#1627 0x0000000000000000 in ?? ()" and similar.
I'm
> guessing this is to do with other system libs that are not compiled
> with -g, but I'm over my head here on FreeBSD system specifics.
>
> So I have two questions:
> 1) Will a buildworld with debugging symbols get me more meaningful
> output from gdb `backtrace full` and is this the best way?
>
> 2) Has anyone else experienced similar issues with net/samba36? I can
> reproduce this bug on both 9.0-RC2 and 8.2-RELEASE-p3.
>
> Your advice is appreciated,
Regarding MAX_DEBUG and trying to get a binary that's built with
debugging support -- you are correct. Samba also has dependencies, as
you're familiar with, and all of those would need to be built with
debugging support as well. Usually when there's a core with a calling
stack that's very long and all zeros, somewhere down at the bottom there
is something remotely coherent. Keep in mind, however, that a software
crash can sometimes completely smash the contents of the calling stack,
which makes after-the-fact debugging very difficult.
I would strongly advise opening up a ticket with the Samba folks, as
well as posting this to freebsd-ports instead of freebsd-stable. The
processes which are coring are purely Samba-specific. Remember, ports
are third-party software, so starting there (with the Samba folks)
would be good.
--
| Jeremy Chadwick jdc at parodius.com |
| Parodius Networking http://www.parodius.com/ |
| UNIX Systems Administrator Mountain View, CA, US |
| Making life hard for others since 1977. PGP 4BD6C0CB |