Netatalk shows kernel panic in syslog when trying to connect to server in OS X 10.6.8. Worked fine before upgrade to ocelot

Hi,

   Any update for the Netatalk afpd crash in Ocelot Oneiric?

-Wilson

Status changed to 'Confirmed' because the bug affects multiple users.

I'm having the same problem. Also running 64bit.

Oct 13 19:44:12 wagner afpd[18956]: ================================================== =============
Oct 13 19:44:12 wagner afpd[18956]: INTERNAL ERROR: Signal 11 in pid 18956 (2.2-beta4)
Oct 13 19:44:12 wagner afpd[18956]: ================================================== =============
Oct 13 19:44:12 wagner afpd[18956]: BACKTRACE: 3 stack frames:
Oct 13 19:44:12 wagner afpd[18956]: #0 /usr/sbin/afpd(netatalk_panic+0x1c) [0x7f885667d49c]
Oct 13 19:44:12 wagner afpd[18956]: #1 /usr/sbin/afpd(+0x4d59c) [0x7f885667d59c]
Oct 13 19:44:12 wagner afpd[18956]: #2 /lib/x86_64-linux-gnu/libc.so.6(+0x36420) [0x7f8855659420]

Weirdly compiling newer netatalk (2.2.1 atm) doesn't seem to have any affect. Same error arises. So probably a library issue somewhere...

I ran into this on my 32-bit system the moment I upgraded to Ocelot this morning. Like other commenters I tried to compile newer 2.2.1 to no avail.

I tried to understand the core of the issue but my debugging skills are not that good. I managed to see that afpd (actually libnetatalk) is registering a SIGSEGV handler and disabled it. That seems to point the problem to the uams_dhx2_pam.so module:

Oct 15 00:24:24 fermion kernel: [ 9989.023989] afpd[13284]: segfault at 0 ip (null) sp bf80020c error 4 in uams_dhx2_pam.so[110000+4000]
Oct 15 00:24:29 fermion kernel: [ 9993.743997] afpd[13285]: segfault at 0 ip (null) sp bf80020c error 4 in uams_dhx2_pam.so[110000+4000]

Being on Lion, I cannot use the uams_dhx UAM instead of uam_dhx2. On a hunch, I tried removing uam_dhx2 altogether and replaced it with uam_guest. This enabled me to connect to the server without crashing afpd but I couldn't connect to my volumes since now I was "nobody". Adding "allow:nobody" to the relevant volume configs in AppleVolumes.default was the final piece of the puzzle and the whole thing might be a suitable workaround for others until we have a resolution.

I know working this way is much less secure, but I can handle it for a short period of time since my server is behind a firewall that only responds to my LAN network and the volumes configured on my AppleVolumes.default have my LAN subnet in the "allowed_hosts" parameter.

Ufuk: thanks for the tip. Works for me too, altough like you said security isn't ideal. Theres something else also broken in there, even with guest I get randomly this:

Oct 15 12:18:57 wagner afpd[19207]: ===============================================================
Oct 15 12:18:57 wagner afpd[19207]: INTERNAL ERROR: Signal 11 in pid 19207 (2.2-beta4)
Oct 15 12:18:57 wagner afpd[19207]: ===============================================================
Oct 15 12:18:57 wagner afpd[19207]: BACKTRACE: 12 stack frames:
Oct 15 12:18:57 wagner afpd[19207]: #0 /usr/sbin/afpd(netatalk_panic+0x1c) [0x7f7601ee549c]
Oct 15 12:18:57 wagner afpd[19207]: #1 /usr/sbin/afpd(+0x4d59c) [0x7f7601ee559c]
Oct 15 12:18:57 wagner afpd[19207]: #2 /lib/x86_64-linux-gnu/libc.so.6(+0x36420) [0x7f7600ec1420]
Oct 15 12:18:57 wagner afpd[19207]: #3 /lib/x86_64-linux-gnu/libc.so.6(+0x82721) [0x7f7600f0d721]
Oct 15 12:18:57 wagner afpd[19207]: #4 /usr/sbin/afpd(ad_setname+0x1c) [0x7f7601ed79ac]
Oct 15 12:18:57 wagner afpd[19207]: #5 /usr/sbin/afpd(getdirparams+0x63b) [0x7f7601eb896b]
Oct 15 12:18:57 wagner afpd[19207]: #6 /usr/sbin/afpd(+0x23361) [0x7f7601ebb361]
Oct 15 12:18:57 wagner afpd[19207]: #7 /usr/sbin/afpd(afp_over_dsi+0x7a3) [0x7f7601eab743]
Oct 15 12:18:57 wagner afpd[19207]: #8 /usr/sbin/afpd(+0x122e6) [0x7f7601eaa2e6]
Oct 15 12:18:57 wagner afpd[19207]: #9 /usr/sbin/afpd(main+0x46b) [0x7f7601ea7c1b]
Oct 15 12:18:57 wagner afpd[19207]: #10 /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed) [0x7f7600eac30d]
Oct 15 12:18:57 wagner afpd[19207]: #11 /usr/sbin/afpd(+0x10005) [0x7f7601ea8005]

I can confirm that I have the same issue after upgrading to Oneiric today . The workaround to use uam_guest.so works but it is not feasible as I am in a multi user environment and need user permissions on the shares.

Before dist upgrading I was using a locally compiled 2.2.0-beta4 as I needed uam_dhx2 to communicate with OSX Lion machines.

Ok, this might have been my fix, at least no more errors. However, I have deleted contents of netatalk folder and reinstall does not put in the template files, so need someone else to confirm if this is a solution.

I went here http://icesquare.com/wordpress/how-to-remove-ds_store-and-appledouble/

but I also removed all .AppleDB .AppleDesktop .DS_Store and .AppleDouble.

However my shares settings files are gone, so I cant tell if its a fix.

Same bug. Mac OS 10.6.8, on 11.04 worked fine, on 11.10 this bug appeared:
Oct 17 19:52:50 mx afpd[8768]: ===============================================================
Oct 17 19:52:50 mx afpd[8768]: INTERNAL ERROR: Signal 11 in pid 8768 (2.2-beta4)
Oct 17 19:52:50 mx afpd[8768]: ===============================================================
Oct 17 19:52:50 mx afpd[8768]: BACKTRACE: 3 stack frames:
Oct 17 19:52:50 mx afpd[8768]: #0 /usr/sbin/afpd(netatalk_panic+0x1c) [0x7f3d1eec449c]
Oct 17 19:52:50 mx afpd[8768]: #1 /usr/sbin/afpd(+0x4d59c) [0x7f3d1eec459c]
Oct 17 19:52:50 mx afpd[8768]: #2 /lib/x86_64-linux-gnu/libc.so.6(+0x36420) [0x7f3d1dea0420]
Oct 17 19:52:53 mx afpd[8772]: ===============================================================
Oct 17 19:52:53 mx afpd[8772]: INTERNAL ERROR: Signal 11 in pid 8772 (2.2-beta4)
Oct 17 19:52:53 mx afpd[8772]: ===============================================================
Oct 17 19:52:53 mx afpd[8772]: BACKTRACE: 3 stack frames:
Oct 17 19:52:53 mx afpd[8772]: #0 /usr/sbin/afpd(netatalk_panic+0x1c) [0x7f3d1eec449c]
Oct 17 19:52:53 mx afpd[8772]: #1 /usr/sbin/afpd(+0x4d59c) [0x7f3d1eec459c]
Oct 17 19:52:53 mx afpd[8772]: #2 /lib/x86_64-linux-gnu/libc.so.6(+0x36420) [0x7f3d1dea0420]

Tried the above mentioned. Purged netatalk, deleted /etc/netatalk deleted all .Apple, .DS etc from my shares. rebooted. Installed netatalk added 1 share in AppleVolumes.default. with this line: /share/ "share" allow:user options:usedots. But still the same error...

Hey, I guess the actual workaround is even simpler:

I had previously written that I chased the problem to down to uams_dhx2_pam.so. Tonight I realized that there is also an "uams_dhx2_passwd.so" that gets built (at least with Netatalk 2.2.1). The default configuration is to symlink "uams_dhx2.so" to "uams_dhx2_pam.so" inside "/usr/lib/netatalk". Once I changed that symlink so that it points to "uams_dhx2_passwd.so", I got my old functionality back. At least it works for me. YMMV.

So:

===================
 BEGIN WORKAROUND
===================

# cd /usr/lib/netatalk
# mv uams_dhx2.so uams_dhx2.so.broken
# ln -s uams_dhx2_passwd.so uams_dhx2.so

===================
 END WORKAROUND
===================

PS: Do not also forget to undo any changes to your share configs if you previously made changes to use uams_guest.so

Thanks ufuk, works for me. I changed my /etc/netatalk/afpd.conf instead of the symlinks:

- -tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so

Yes, Luke. It should be working that way too. Thanks for pointing it out, it is actually much cleaner that way.

thanks guys for all the input, pointing me to a fix.
However the instructions here did not work fully for me, but were on the right track. Im not sure if im on the compiled newest version of netatalk or the one from package for latest ubuntu 11.10. The Ubuntu software center shows v. netatalk 2.2.0-0ppa2 is installed. When I do a "afpd -V" I show "afpd 2.2.1".

Ok, so when I follow the instructions you gave no changes were happening for me.

When I did a "ps -ef | grep afpd" I keep showing a "root 26886 1 28 16:44 ? 00:00:00 /usr/local/sbin/afpd -U uams_dhx2.so,uams_clrtxt.so -g nobody -c 50 -n bunbox" no matter where I changed the uams list.

But... when I edited the "/etc/default/netatalk" file, uncomenting and setting to "AFPD_UAMLIST="-U uams_dhx.so,uams_dhx2_passwd.so"" all worked, now I do a "ps -ef | grep afpd" and I see "bunbox 26985 26973 0 16:46 ? 00:00:00 /usr/local/sbin/afpd -U uams_dhx.so,uams_dhx2_passwd.so -g nobody -c 50 -n bunbox"

Restarting netatalk and avahi, all seems to be working again. Thanks so much guys for the help. Maybe my privileges have been altered and will not read the etc/netatalk files anymore, not sure. Now to get time machine working again.

Yup all is working again, with Ufuk and Lukes assistance, timemachine and regular mac osx sharing, bounjour. I tried an install of the latest netatalk, compiled and make, but since that didnt work I went to ubuntu software center and reinstalled that version too. So Im not sure whats really installed, if I knew how to purge the make thing I did, and delete everything then reinstall, so my system is messed up a bit, but its working.

I got my hint from this: "afpd -V" which showed me where the new/old config files are. at least editing them there, worked.
afpd 2.2.1 - Apple Filing Protocol (AFP) daemon of Netatalk

This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; either version 2 of the License, or (at your option) any later
version. Please see the file COPYING for further information and details.

afpd has been compiled with support for these features:

          AFP versions: 2.2 3.0 3.1 3.2 3.3
DDP(AppleTalk) Support: No
         CNID backends: dbd last tdb
           SLP support: No
      Zeroconf support: Yes
  TCP wrappers support: Yes
         Quota support: Yes
   Admin group support: Yes
    Valid shell checks: Yes
      cracklib support: No
        Dropbox kludge: No
  Force volume uid/gid: No
           ACL support: No
            EA support: ad | sys
          LDAP support: No

 afpd.conf: /usr/local/etc/netatalk/afpd.conf
 AppleVolumes.system: /usr/local/etc/netatalk/AppleVolumes.system
 AppleVolumes.default: /usr/local/etc/netatalk/AppleVolumes.default
 afp_signature.conf: /usr/local/etc/netatalk/afp_signature.conf
 afp_voluuid.conf: /usr/local/etc/netatalk/afp_voluuid.conf
 afp_ldap.conf: not supported
 UAM search path: /usr/local/etc/netatalk/uams/
 Server messages path: /usr/local/etc/netatalk/msg/
 lockfile: /var/lock/afpd

I've tested the new workaround mentioned above. Configs before workaround:

* AppleVolumes.default:
        /shared/timemachine/ "TimeMachine" allow:timemachine cnidscheme:dbd options:tm,upriv,usedots

afpd.conf:
        - -tcp -noddp -uamlist uams_dhx.so,uams_dhx2.so -nosavepassword -nozeroconf

/var/log/syslog (when trying to do a TimeMachine backup):

        Oct 18 09:33:37 rutger afpd[22363]: ===============================================================
        Oct 18 09:33:37 rutger afpd[22363]: INTERNAL ERROR: Signal 11 in pid 22363 (2.2-beta4)
        Oct 18 09:33:37 rutger afpd[22363]: ===============================================================
        Oct 18 09:33:37 rutger afpd[22363]: BACKTRACE: 3 stack frames:
        Oct 18 09:33:37 rutger afpd[22363]: #0 /usr/sbin/afpd(netatalk_panic+0x1c) [0x7f59d87e549c]
        Oct 18 09:33:37 rutger afpd[22363]: #1 /usr/sbin/afpd(+0x4d59c) [0x7f59d87e559c]
        Oct 18 09:33:37 rutger afpd[22363]: #2 /lib/x86_64-linux-gnu/libc.so.6(+0x36420) [0x7f59d77c1420]
        Oct 18 09:33:39 rutger afpd[22364]: AFP statistics: 0.52 KB read, 0.38 KB written

Applying workaround with uams_dhx2_passwd:

* afpd.conf:

        - -tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so -nosavepassword -nozeroconf

* /var/log/syslog (after restarting using "service netatalk restart" and starting a TimeMachine backup)

        Oct 18 10:41:31 rutger afpd[23362]: AFP/TCP started, advertising 172.22.22.150:548 (2.2-beta4)
        Oct 18 10:41:42 rutger afpd[23365]: AFP3.3 Login by timemachine
        Oct 18 10:41:57 rutger cnid_dbd[23367]: Set syslog logging to level: LOG_NOTE
        ...
        Oct 18 10:45:24 rutger afpd[23365]: AFP logout by timemachine
        Oct 18 10:45:24 rutger afpd[23365]: AFP statistics: 184393.60 KB read, 118304.24 KB written
        Oct 18 10:45:24 rutger afpd[23365]: done

Workaround confirmed and working!

root@rutger:/etc/netatalk# afpd -v
Oct 18 10:52:43 rutger [23500]: Set syslog logging to level: LOG_NOTE
afpd 2.2-beta4 - Apple Filing Protocol (AFP) daemon of Netatalk

This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; either version 2 of the License, or (at your option) any later
version. Please see the file COPYING for further information and details.

afpd has been compiled with support for these features:

        AFP3.x support: Yes
        TCP/IP Support: Yes
DDP(AppleTalk) Support: Yes
         CNID backends: dbd last tdb

             afpd.conf: /etc/netatalk/afpd.conf
   AppleVolumes.system: /etc/netatalk/AppleVolumes.system
  AppleVolumes.default: /etc/netatalk/AppleVolumes.default
    afp_signature.conf: /etc/netatalk/afp_signature.conf
      afp_voluuid.conf: /etc/netatalk/afp_voluuid.conf
         afp_ldap.conf: not supported
       UAM search path: /usr/lib/netatalk/
  Server messages path: /etc/netatalk/msg/
              lockfile: /var/run/afpd.pid

(sorry about the messed up formatting, tried to indent for readability)

For clarity I am using the Ubuntu netatalk 2.2beta4-1 package

All,

Followed the instructions and got my iMac backing-up again (finally, phew). I deleted the old backup to start with a fresh new one. iMac indicating that it had 171Gb to back up. Started nicely for the first couple of hours until it bailed out stating the backup device was no longer available (du -s of the Timemachine directory on my ubuntu box gives 151G stored so suspect it bailed out before the end). Nothing in the logs and afpd still running. Restarted all the services but still no joy: iMac does not see the backup device anymore.

Any thoughts on where to look for why it failed and won't find mu ubuntu box (i386 running the server kernel)?

One other thing, I installed netatalk from teh ubuntu repos. afpd -V gets me a version of 2.0.5, NOT 2.2 beta??

Appreciate the help.

Herman

Had another crack at this last night, and found that by removing old versions of libdb I was able to get it working again. Can't confirm if it's broken something else, but so far all seems to be OK.

Thanks Luke, I had exactly the same problem when upgrading to oneric.

The substitution uams_dhx2.so --> uams_dhx2_passwd.so in afpd.conf restored TimeMachine etc. to full working order.

David

Same here, but I simply changed the AFPD_UAMLIST variable in /etc/default/netatalk to look like:

AFPD_UAMLIST="-U uams_dhx.so,uams_dhx2_passwd.so"

After a restart of netatalk (2.2beta4) everything worked again on 11.10

Just if it helps, on powerpc platform (not supported by launchpad), the problem also happens and has the same workaround. I am attaching the stack trace in case it helps debugging.

Same issue here, about to test out the symlink workaround.

Running Ubuntu 11.10 64bit. Attaching the stack trace, because I can.

So I tried the symlink workaround and I was able to connect to the AFP server, but something else breaks upon attempting to access any shares.

Attaching the new stack trace...

Happily confirming that
- -tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so
worked for me.

This caused a new cnid fault, which was fixed by changing the cnidscheme to dbd in AppleVolumes.default

After months of successful operation under oneiric, serving several Lion clients, this problem hit me just now out of the blue, symptoms and log exactly as described above. The only thing I can think of that's changed in the timeframe since I last knew for sure it was working, is that I set up samba as well on the same server to serve one of the same shared volumes to a windows VM. No, I can't think of any reason why that should affect things, it's just all that's been changed on the server in the timeframe, so mentioning it in case anyone goes "aha!" :-)

The workaround in #12 (ie: changing to uams_dhx2_password.so in afpd.conf) seems to be working for me, at least for now... Didn't do any of the other stuff - removing .AppleDB .Appletalk etc. (though things are configured to try to minimize them anyway), or reinstalling netatalk from source.

NB: this is netatalk 2.2~beta4-1 on fully-updated oneiric 64-bit.

I have run into exactly the same issue, yet I have a different workaround. Since I am using centralized LDAP user database I had to use uams_dhx2_pam.so. I have figured out, that disabling pam module for Samba password sync fixed the issue. I have created copies of common-auth (common-auth-nosmb) and common-password (common-password-nosmb) that lack pam_smbpass.so lines and included them in /etc/pam.d/netatalk:

#%PAM-1.0

@include common-auth-nosmb
@include common-account
@include common-password-nosmb
@include common-session

This has fixed the problem.

I can confirm that disabling the samba PAM module is a workaround. Unfortunately I was using this for authenticating against an active directory server.

Is there any progress on fixing this bug? I've tried the latest 12.04 beta (all packages up-to-date), but the problem persists.

The problem with uams_dhx2_pam.so still exists in 12.04x64 server.
The workaround involving replacing the aforementioned module with uams_dhx2_password.so in afpd.conf works like a charm in my case.

My afpd.conf....

- -tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so -nosavepassword

and tried...

-tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so -nosavepassword

to no avail

I found out what was causing my netatalk to fail. When I was installing Ubuntu Linux Server, near the end of the install I would select the SSH, CUPS and SMB packages to install. I when back in reinstall only selecting the SSH package and netatalk worked like a charm. I then when back and installed Samba and netatalk still worked, however at this time I have been unable to get Samba working. And the problem lies withing authentication.

I was seeing this same crash on every attempt to connect via afp to the server. I stumbled across this post and figured out a workaround:

http://www.rudiswiki.de/wiki9/ServerUbuntu1204#Setup_netatalk

Using this line in /etc/netatalk/afpd.conf, the crash goes away. -mimicmodel RackMac seems to be the key. I don't know why...

- -tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so -nosavepassword -advertise_ssh -mimicmodel RackMac

Hi

I had a working configuration with Mac OSX Lion successfully backing up via TM onto Ubuntu 12.04 desktop. I then tried to get a Samsung Netbook (with Windows 7 Starter edition) to connect to Ubuntu using Samba for a backup. Tried everything but gave up in the end. During that process my TM backups started to fail.

I have just investigated that and have observed the same error reported here. I tried the -mimicmodel fix and this made no difference. I then found http://sourceforge.net/mailarchive/message.php?msg_id=29229826 and confirmed that during my Samba work the /etc/pam.d/common-auth file had indeed changed. I removed the 'pam_smbpass.so' (clearly related to Samba) reference and the problem has gone away. (Big thanks to HAT).

John

I just installed Ubuntu 12.10 server and getting the same problem

Linux mediaserver 3.5.0-17-generic #28-Ubuntu SMP Tue Oct 9 19:31:23 UTC 2012 x86_64 x86_64 x86_64 GNU/Linux

afpd 2.2.2

leslie@mediaserver:/etc/default$ dpkg -s netatalk
Package: netatalk
Status: install ok installed
Priority: extra
Section: net
Installed-Size: 3782
Maintainer: Ubuntu Developers <email address hidden>
Architecture: amd64
Version: 2.2.2-1ubuntu1

I can confirm that the fixes noted in activity 28 and 14 both resolved the issue

... I meant to say, either activity 28 or activity 14 will resolve the issue.

Same bug, found elsewhere that "- -tcp -noddp -uamlist uams_dhx.so,uams_dhx2_passwd.so" in afpd.conf suppressed the kernel panic, but not allow guest users, login of users, but works only for me with Time Machine.