Bug 140081

Created attachment 107066 [details]
strace while running gnome-volume-manager

I'm seeing both gnome-volume-manager and eggcups crash after applying
the latest round of FC3 updates. I don't know what the actual cause
is, but it's not the kernel -- rebooting into the old kernel (667)
does not eliminate the problem.

Oh, right, I totally forgot about eggcups. The same here. I tried
around with the session and so the eggcups only occurred once.

Ok, I can reproduce this.  Tim can you post your /var/log/yum.log file
here.  Thanks.

Please run 
/usr/sbin/prelink /usr/lib/libdbus*so*
and see if this fixes the problem.

Tim, I would still like your yum.log so I can track down what is
causing the dbus prelink stuff to get out of sync.  Thanks.

*** Bug 139682 has been marked as a duplicate of this bug. ***

I'm getting similar problems after applying the latest round of
updates for FC3 (11/19) Additionally, nautilus-cd-burner seems to
crash.  Running /usr/lib/libdbus*so* gives me a segfault.  :(

here are the updates that lead to the problem:

[Fri Nov 19 12:43:18 2004] up2date installing packages:
['kernel-2.6.9-1.678_FC3', 'rhgb-0.15.1-1.FC3',
'sound-juicer-0.5.14-1.FC3.0', 'system-config-users-1.2.28-0.fc3.1',
'udev-039-10.FC3.2', 'xorg-x11-6.8.1-12.FC3.1',
'xorg-x11-Mesa-libGL-6.8.1-12.FC3.1',
'xorg-x11-Mesa-libGLU-6.8.1-12.FC3.1',
'xorg-x11-deprecated-libs-6.8.1-12.FC3.1',
'xorg-x11-deprecated-libs-devel-6.8.1-12.FC3.1',
'xorg-x11-devel-6.8.1-12.FC3.1', 'xorg-x11-font-utils-6.8.1-12.FC3.1',
'xorg-x11-libs-6.8.1-12.FC3.1', 'xorg-x11-tools-6.8.1-12.FC3.1',
'xorg-x11-twm-6.8.1-12.FC3.1', 'xorg-x11-xauth-6.8.1-12.FC3.1',
'xorg-x11-xfs-6.8.1-12.FC3.1']

OK, the problems have gone away after running /etc/cron.daily/prelink

Weird.

Re: comment #5

Yes, running "/usr/sbin/prelink /usr/lib/libdbus*so*" made the problem
go away. Now, I'm going to try /etc/cron.daily/prelink and see if that
makes it come back or not.

/etc/cron.daily/prelink did not make the problem reappear.

(BTW, I swear I ran /etc/cron.daily/prelink after the latest round of updates
but before rebooting. It seems weird to me that I had to run prelink a second
time in order to not have a crash...)

Running /etc/cron.daily/prelink did fix the problem for me! Here is
the list of packages that yum installed:

Nov 19 14:01:56 Installed: kernel.i686 2.6.9-1.678_FC3
Nov 19 14:02:02 Updated: udev.i386 039-10.FC3.2
Nov 19 14:02:03 Updated: xorg-x11-deprecated-libs-devel.i386
6.8.1-12.FC3.1
Nov 19 14:02:07 Updated: system-config-users.noarch 1.2.28-0.fc3.1
Nov 19 14:02:11 Updated: xorg-x11-libs.i386 6.8.1-12.FC3.1
Nov 19 14:02:12 Updated: xorg-x11-Mesa-libGL.i386 6.8.1-12.FC3.1
Nov 19 14:02:12 Updated: xorg-x11-font-utils.i386 6.8.1-12.FC3.1
Nov 19 14:02:16 Updated: xorg-x11-Mesa-libGLU.i386 6.8.1-12.FC3.1
Nov 19 14:02:16 Updated: xorg-x11-xauth.i386 6.8.1-12.FC3.1
Nov 19 14:02:22 Updated: xorg-x11-xfs.i386 6.8.1-12.FC3.1
Nov 19 14:02:29 Updated: xorg-x11.i386 6.8.1-12.FC3.1
Nov 19 14:02:37 Updated: xorg-x11-devel.i386 6.8.1-12.FC3.1
Nov 19 14:02:41 Updated: gaim.i386 1:1.0.3-0.FC3
Nov 19 14:02:41 Updated: xorg-x11-deprecated-libs.i386 6.8.1-12.FC3.1
Nov 19 14:02:45 Updated: planner.i386 0.12.1-1.1.fc3.rf
Nov 19 14:02:46 Updated: imlib2.i386 1.1.2-2.1.fc3.rf
Nov 19 14:02:48 Updated: xorg-x11-twm.i386 6.8.1-12.FC3.1
Nov 19 14:02:50 Updated: xorg-x11-tools.i386 6.8.1-12.FC3.1
Nov 19 14:02:51 Updated: rhgb.i386 0.15.1-1.FC3
Nov 19 14:02:54 Updated: sound-juicer.i386 0.5.14-1.FC3.0

I removed a bunch of third-party RPMs that have been installed that I
think that are unrelated to the problem.

That reminds me, here's the list of packages that up2date installed
for me (before this I had no problem, but after this I had the problem):

[Thu Nov 18 15:44:14 2004] up2date installing packages:
['kernel-2.6.9-1.678_FC3', 'kernel-doc-2.6.9-1.678_FC3',
'libxml-1.8.17-12', 'libxml-devel-1.8.17-12', 'rhgb-0.15.1-1.FC3',
'sound-juicer-0.5.14-1.FC3.0', 'system-config-users-1.2.28-0.fc3.1',
'udev-039-10.FC3.2', 'xorg-x11-6.8.1-12.FC3.1',
'xorg-x11-Mesa-libGL-6.8.1-12.FC3.1',
'xorg-x11-Mesa-libGLU-6.8.1-12.FC3.1', 'xorg-x11-Xdmx-6.8.1-12.FC3.1',
'xorg-x11-Xnest-6.8.1-12.FC3.1', 'xorg-x11-Xvfb-6.8.1-12.FC3.1',
'xorg-x11-deprecated-libs-6.8.1-12.FC3.1',
'xorg-x11-deprecated-libs-devel-6.8.1-12.FC3.1',
'xorg-x11-devel-6.8.1-12.FC3.1', 'xorg-x11-doc-6.8.1-12.FC3.1',
'xorg-x11-font-utils-6.8.1-12.FC3.1', 'xorg-x11-libs-6.8.1-12.FC3.1',
'xorg-x11-sdk-6.8.1-12.FC3.1', 'xorg-x11-tools-6.8.1-12.FC3.1',
'xorg-x11-twm-6.8.1-12.FC3.1', 'xorg-x11-xauth-6.8.1-12.FC3.1',
'xorg-x11-xdm-6.8.1-12.FC3.1', 'xorg-x11-xfs-6.8.1-12.FC3.1']

On this system I have been using up2date rather than yum (I don't
remember if I have a particular reason why; if I do, then I don't
remember the reason right now).

Created attachment 107108 [details]
tar archive of crashing /usr/lib/libdbus*so*

On another system, the problem happened for me again. I'm not completely sure
the package list is relevant this time -- this was in some sense an upgrade
from a late October/early November rawhide snapshot to FC3 final + updates,
although some of the updates had been applied previously.

This time, I !!KNOW!! I ran /etc/cron.daily/prelink before rebooting into the
updated system!!

In any case, here's a tar archive of the failing /usr/lib/libdbus*so* files. Is
this what is needed? If not, let me know and I'll try to provide whatever is
needed. (Like my first experience, this is gnome-volume-manager and eggcups
failing.)

No, this is not sufficient.
For a program that segfaults when prelinked and doesn't when not prelinked,
I need:
1) description how to reproduce the problem (what options use, what actions to do
2) say /usr/bin/gnome-foo is failing, then I need
tar chjf gnome-foo.tar.bz2 `LD_TRACE_PRELINKING=1 LD_WARN= /usr/bin/gnome-foo | awk '{print $3}'`
tarball at the point when you CAN reproduce it.  After you tar it up, do
for i in `LD_TRACE_PRELINKING=1 LD_WARN= /usr/bin/gnome-foo |__
   awk '{print $3}'`; do prelink -u $i; done
and verify it doesn't fail.

You have packed up only some libraries, which is not useful.  Prelinking info
is used only if the binary and all its libraries are currently prelinked.

*** Bug 140158 has been marked as a duplicate of this bug. ***

gnome-volume-manager is failing for me too after updates. Here is the
tarfile with the libraries as asked for:

http://www.infostrategique.com/linuxrpms/marc/gnome-volume-manager.tar.bz2

Doing the "for i in..." command to prelink the libraries resolved the
problem for me.

I found a workaround, but it is a temporary fix...

 After finding such pretties, as this in my log, with the
same problem:

Nov 21 10:49:39 smoker kernel: hald[3648]: segfault at
0000003985010f30 rip 0000003985010f30 rsp 0000007fbffff738 error 14
Nov 21 10:49:39 smoker haldaemon: haldaemon startup succeeded
Nov 21 10:49:45 smoker haldaemon: haldaemon shutdown failed
Nov 21 10:49:50 smoker kernel: hald[3664]: segfault at
0000003985010f30 rip 0000003985010f30 rsp 0000007fbffff738 error 14
Nov 21 10:49:50 smoker haldaemon: haldaemon startup succeeded
Nov 21 10:51:04 smoker kernel: gnome-vfs-daemo[3676]: segfault at
0000003985010f30 rip 0000003985010f30 rsp 0000007fbfffec68 error 14
Nov 21 10:53:29 smoker kernel: gnome-vfs-daemo[3748]: segfault at
0000003985010f30 rip 0000003985010f30 rsp 0000007fbfffec68 error 14

  I decided to do a haldaemon upgrade reversal...

And dropped a great big rpm -i --force on the following files:

dbus-0.22-10.x86_64.rpm         gnome-volume-manager-1.1.0-5.x86_64.rpm
dbus-devel-0.22-10.x86_64.rpm   hal-0.4.0-10.x86_64.rpm
dbus-glib-0.22-10.x86_64.rpm    hal-cups-utils-0.5.2-8.x86_64.rpm
dbus-python-0.22-10.x86_64.rpm  hal-devel-0.4.0-10.x86_64.rpm
dbus-x11-0.22-10.x86_64.rpm     hal-gnome-0.4.0-10.x86_64.rpm

  Yeah, I know it was atom bombing the solution, but it 
works. BTW: The platform is X_86, not i386.... at least in my case.


  Good luck finding the root cause.

Looking at the #17 tarball, this is a prelink bug.
From what I can see:
- most of the packages were prelinked on 2004-11-16
- then I guess xorg-x11-libs (not sure if something else too) got upgraded
- prelink was run again on 2004-11-20:
  - libICE.so.6 libXrender.so.1 libXrandr.so.2 libXi.so.6 libXinerama.so.1 libXft.so.2 libXext.so.6 libXcursor.so.1 libX11.so.6 libSM.so.6
  libraries were prelinked because they changed (by the xorg-x11-libs upgrade)
  - libpangoxft-1.0.so.0 libpangox-1.0.so.0 libbonoboui-2.so.0 libgdk-x11-2.0.so.0 libgnomecanvas-2.so.0 libgnomeui-2.so.0 libgtk-x11-2.0.so.0
  libraries were re-prelinked because they depend on the libraries in the first
  set
  - now, for some not yet known reason (most probably because the layout code
  saw some library grew up and occupied some memory slot that was previously
  used by something else) prelink decided to re-prelink libdbus-1.so.0 too
  (it doesn't depend on anything that changed)
  - but libhal.so.0 that depends on libdbus-1.so.0 was not re-prelinked
  - gnome-volume-manager was re-prelinked (because several of its dependencies
  changed)

I'll now try to figure out why prelink did not re-prelink libhal.so.0 and worse
the consistency checks before re-prelinking a binary did not catch that libhal.so.0 has not been re-prelinked.

*** Bug 140357 has been marked as a duplicate of this bug. ***

I too am seeing these issues with eggcups and gnome-volume-manager.
Same as everyone else, goes away when I manually prelink. But, as an
extra twist, and maybe (un)related, after the round of updates that
triggered this issue, I also see that on shutdown, the hal daemon
fails to shutdown. 

/etc/cron.daily/prelink cured the problem for me too.
6 day old new install, new hardware;
one set of yum updates.

then yum update of kernel2.6.9-1.681_fc3
reboot. Problem persists.

prelink, reboot, OK.
Changing from gnome to kde seems to bypass the problem.

Until prelink is fixed, I guess a change like this (commenting out
PRELINK_FULL_TIME_INTERVAL) should keep the problem from reoccurring
in the future.

--- /etc/sysconfig/prelink.previous     2004-11-22 12:54:51.165557993
-0800
+++ /etc/sysconfig/prelink      2004-11-22 12:55:07.536232188 -0800
@@ -18,4 +18,4 @@
 # $PRELINK_FULL_TIME_INTERVAL days it will be run
 # in normal mode.  Comment it out if it should be run
 # in normal mode always.
-PRELINK_FULL_TIME_INTERVAL=14
+#PRELINK_FULL_TIME_INTERVAL=14

Anyone who can reproduce this, can you figure out what packages (ENVR+arch)
you had installed before the up2date that brought in xorg-x11-*, after which
another prelink run failed to prelink libhal.so.0?
If I figure out what happened just from source analysis, I'd still like to
verify, if I don't, I'd like to certainly see it under the debugger.
If I have a list of packages before, I can install them in a chroot, prelink,
update xorg-x11 (and other packages that you updated during that update) and
debug the re-prelink.

Thanks.

Prelinking failed again for me this morning. I sat in front of my
computer that was already logged in. Clicked on the evolution icon, I
get a segfault, I clicked on gnome-terminal, another segfault. Pretty
much everything was broke. I hit Ctrl-Alt-F1, logged in as root and
did the following:

prelink -ua
added --no-exec-shield to the /etc/sysconfig/prelink file
prelink -am --no-exec-shield
sysctl -w kernel.exec-shield-randomize=0
sysctl -w kernel.exec-shield=0

Switched back to Ctrl-Alt-F7, did a Ctrl-Backspace. Logged in again
and everything was back to normal.

Something is very wrong in prelink land :)

  FWIW, the downgrade using the RPMs I mentioned, above,
has proven stable and reliable..... so far.
 
 No more Segv's... no issue's, at all.

(Actually, the system -smokes-   :)

 It would seem implied that culprit sits within that
grouping of RPM's.... and their upgrades.

*** Bug 140498 has been marked as a duplicate of this bug. ***

Ok, can you try ftp://people.redhat.com/jakub/prelink/prelink-0.3.3-1.*.rpm ?
Would be thankful if somebody tried it as well on a FC3 box without the most
recent updates (from last 4 days or so), then up2date it and rerun
/etc/cron.daily/prelink.
If all is fine, I'll move this to FC3 updates/testing soon.

*** Bug 140103 has been marked as a duplicate of this bug. ***

I had the same problem after updating the kernel with yum. Removing
desktop-printing with 'yum remove desktop-printing' made it stop.
Reinstalling after logging out and back in again did not cause any
problems.

prelink-0.3.3-0.fc3 is now in updates/testing/3.

I followed the step:

1.  ftp://people.redhat.com/jakub/prelink/prelink-0.3.3-1.*.rpm
2.  up2date prelink
3.  /etc/cron.daily/prelink

step 3 runs forever. It seems to run in an infinite loop.

Forever means you killed it after how long exactly?
prelink upgrade means everything is re-prelinked, that can very well take
half an hour.
Watch /var/log/prelink.log if there is any progress.
If prelink really run into an infinite loop (haven't seen that ever), as
in /var/log/prelink.log wouldn't keep moving and say /proc/`pidof prelink`/fd/
wouldn't show any progress, then I need you to install
prelink-debuginfo* (look into updates/testing/3/i386/debug/ on Fedora mirrors)
and while it is still running attach debugger to it and see what is it doing.

I had this problem also. First boot after running the updates I
swapped out the DVD burner in my laptop. Shutting done I noticed the
HAL daemon failing to stop. Thinking it was a UDEV problem I started
to follow the steps on the Fedora UDEV page. Ran start_udev. I got the
kernel source to have a go at run mkinitrd but accidentally started
the kernel compilation as well. Let it run thinking it would do the
job and it did.
Does a kernel compile run the prelinker?

Same problem that others are experiencing, gnome-volume-manager and
eggcups crashing on startup.  This is a new install completed with
updates 26 Nov 04.  I did not have any issues until I configured
LM-Sensors and rebooted?  Removing the configuration changes for
LM-Sensors and running "/etc/cron.daily/prelink" and rebooting fixed
the problem.

Hardware:
Shuttle MK40VN
XP 1600+
512 MB (2 x 256)

I just tested the updates/testing/3 prelink package. It fixes the
problem (i.e. installing FC3, prelinking, installing updates, and
prelinking again causes crashes for me if I use FC3's prelink, but not
if I use the new prelink package).

Can I get answer for my #34 question, please?
Multiple people reported the same as #37, so I'd like to release normal updates
(not testing) RSN.

*** Bug 141166 has been marked as a duplicate of this bug. ***

Well with me happen in this situation
I never switch to xfce, after enter in xfce and back to default
session happen:

reinitialize eggcups
reinitialize gnome-volume-manager 
(dialog boxes)

http://listman.redhat.com/archives/fedora-announce-list/2004-November/msg00082.html

I just tested prelink-0.3.3-1. Running /etc/cron.daily/prelink didn't
do the trick. Hald keeps on failing to start. Hmmmmm ...