hang on BOGUS LENGTH in write keyboard desc

[Matt Drake]

Binary package hint: xorg

IMPACT: Under certain circumstances, the X server hangs causing the entire system to be non-responsive (including sysrq)

DEVELOPMENT: Karmic picked up the upstream fix in xserver-xorg-core 2:1.6.3-1ubuntu1

PATCH: Attached at <http://launchpadlibrarian.net/36321173/xorg-server_1.6.0-0ubuntu14.1.debdiff>. Test builds available in my PPA (https://launchpad.net/~broder/+archive/ubuntu-tests)

INSTRUCTIONS: There hasn't been a consistent way to reproduce the bug found, but the hang seems to trigger more frequently under VMWare.

REGRESSION: This is a patch taken directly from upstream. It only affects a handful of lines.

==
Original bug description:

*apport seems broken right now, so let me know if i need to upload anything else*

on the xserver in jaunty, i get random hangs where the entire system is unresponsive, meaning that i can't escape to a tty, restart the computer, or even do a sysreq-REISUB. this has happened twice, both times while using my bluetooth mouse, so that might be the cause of it. in both logs there's a line that seems important:

[xkb] BOGUS LENGTH in write keyboard desc, expected 5976, got 5992

thanks,
matt drake

[Timo Aaltonen (tjaalton)]

You should try getting a proper backtrace:

https://wiki.ubuntu.com/X/Backtracing

[Matt Drake (mattduckman)]

I'm having trouble reproducing this bug again. If I can get it to occur again, I'll reopen this bug with a backtrace.

[psyray (psyray)]

Hi

Same problem
gdm.log -> BOGUS LENGTH in write keyboard desc, expected 6156, got 6392
syslog -> Jan 31 14:25:40 raynald-laptop gdm[5928]: WARNING: gdm_slave_xioerror_handler : erreur X fatale - Redémarrage de :0

X automatically shutdown & restart.

This occurs randomly when i pressed a shortcut key on my MS Digital media keyboard (like calc, volume...)

[psyray (psyray)]

thanks
psyray

[lnx (nwegener)]

With gdb attached no "hang on BOGUS LENGth has been written.
Nevertheless everything hang.

[Thomas Morin (tmmorin-orange)]

I've been experiencing what look like the same bug on Debian, and I've filled-in a bug report in X.org bugzilla, including two clean stack traces.

https://bugs.freedesktop.org/show_bug.cgi?id=21464

[Alexis de Lattre (alexis-via)]

I have the same bug here on two different laptop at my office : one with Intrepid and one with Jaunty. The error message in the X server log file after the crash is exactly the same : "BOGUS LENGTH in write keyboard desc". On these two laptops, I had changed some default settings in System > Preferences > Keyboard > Layout > Layout options. Here is an example of the options I had set :

gconftool-2 -R /desktop/gnome/peripherals/keyboard/kbd
 layouts = [fr]
 options = [caps caps:shiftlock,grp grp:alts_toggle,kpdl kpdl:comma]
 model = latitude

[Alexis de Lattre (alexis-via)]

Here are the logs of the X server for the same crash on an Intrepid system.

[Alexis de Lattre (alexis-via)]

I have removed the special options for the keyboard in System > Preferences > Keyboard > Layout > Layout options. But it has crashed again on the Intrepid laptop. The only particularity of this laptop is that it has a USB keypad.

[Alexis de Lattre (alexis-via)]

Now that a fix is released upstream, let's fix it in Ubuntu ! When do you guys plan to upload a package with the fix ? We are all waiting for it because having an X server which crashes regularly is very annoying !

[Philip Aston (philipa)]

I agree - this bug is a major stability issue for Ubuntu, and affects me on a near daily basis. The importance of this bug should be changed from Undecided to High.

[Valdemar Lemche (atterdag)]

I only experience this issue when running Vmware Workstation 6.5.2, Lotus Notes 8.5, and Eclispse, but since this are the three programs I use the most (apart from firefox), then these resets occurs maybe 3-5 times a day.

So please, please upload the fix!

[sperlyjinx (jgbaum)]

I have run into this bug on jaunty running 3 monitors on 2 nvidia quadro 290 GPUs with the nvidia driver (version 180). I've attached the GDM log from the time of the crash.

[sperlyjinx (jgbaum)]

Here's my xorg.conf. BTW, what does it mean that a fix has been released? If a fix was released, am I experiencing a different bug, or was the fix not incorporated into the latest Ubuntu package?

[Alexis de Lattre (alexis-via)]

The "Fix-released" statuts means that the bug has been fixed upstream, but the fix is not in the Ubuntu packages yet (the name of this status is confusing many people).

The problem is that the fix has been released upstream for two month now, but nothing has arrived in Ubuntu yet, so we are still experiencing our daily X server crashes, which is not fun !

[Alexis de Lattre (alexis-via)]

According to my research (but I may be wrong, I am not an expert), the patch in not in Karmic Koala either. In current karmic koala, the package xserver-xorg-core is version 1.6.1.901, and when I look at the source packages :

ubuntu/pool/main/x/xorg-server/xorg-server_1.6.1.901.orig.tar.gz
ubuntu/pool/main/x/xorg-server/xorg-server_1.6.1.901-2ubuntu2.diff.gz

and I check the source code of xkb/xkb.c, I see that the fix (https://bugs.freedesktop.org/attachment.cgi?id=26038) is not present.

Could it be possible to apply this small patch (5 lines of changes) in Karmic Koala at least ? I am using Karmic Koala and I would be able to give some feedback.

[Alexis de Lattre (alexis-via)]

The bug is now fixed in Debian Sid, cf http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=529625
The new package is xserver-xorg-core version 1.6.2.901-1.
Now that the fix is in Debian, let's have it in Ubuntu !

[Loïc Minier (lool)]

Fixed upstream in 1.6.2 (which is in Debian sid), we only have 1.6.1.901 in karmic ATM and multiple people have this issue in Ubuntu => Confirmed.

[Loïc Minier (lool)]

Actually has all relevant developer information => Triaged

[Timo Aaltonen (tjaalton)]

1.6.2.901 has been uploaded already, and will build as soon as audit has been moved to main (libaudit-dev is a build-dep of xorg-server).

[Alexis de Lattre (alexis-via)]

The new package for xorg-server is now karmic : http://packages.ubuntu.com/karmic/xserver-xorg-core (version 2:1.6.3-1ubuntu1)

The changelog says "Fixes xkb bogus length in write keyboard desc", with a reference to the Debian bug #529625.

I'll upgrade my karmic soon to test it.

[Florian Schröck (mael-reverted)]

This bug occurs often when using VMware Remote Console (nvidia gfx with TwinView). Is there a chance that this will be fixed for jaunty, too? or can i install the karmic package on jaunty?

thanks!

[Arie Skliarouk (skliarie)]

About a week ago I did exactly that (Installed xserver-xorg-core package from karmic on jaunty) and I had no failures since then.

In my case failures also happened when I was using VMWare player, but I am not sure it did not happen when I was not.

[Florian Schröck (mael-reverted)]

Thanks for this information, I'm installing the karmic package right now

I'll report back in a couple of weeks

[Florian Schröck (mael-reverted)]

The package from karmic works like a charm! No problems since the update! :)

[Ingo Karkat (inkarkat)]

Here, no more crashes after the update, too. I manually installed the package plus required newer dependencies:
- xserver-org-core_1.6.3-1ubuntu6_i386.deb
- libgcrypt11_1.4.4-2ubuntu2_i386.deb
- libgpgerror0_1.6-1ubuntu1_i386.deb

Later, I noticed that printing wasn't possible any more, because the CUPS daemon aborts on startup. This is what I found in /var/log/syslog:

Sep 27 12:44:14 meisterin cupsd: Libgcrypt error: reading `/proc/sys/crypto/fips_enabled' failed: Permission denied - abort
Sep 27 12:44:14 meisterin kernel: [ 4532.516155] type=1503 audit(1254048254.089:11): operation="sysctl" requested_mask="r::" denied_mask="r::" fsuid=0 name="/proc/sys/crypto/fips_enabled" pid=7711 profile="/usr/sbin/cupsd"

It seems that the updated libgcrypt required by xserver-org-core doesn't correctly integrate into Jaunty's AppArmor profile for cupsd. After disabling AppArmor enforcement for cupsd, the daemon started up again:

sudo aa-complain /usr/sbin/cupsd

(This setting persists even after reboots.) A better way would be to update the AppArmor profile, but I didn't want to delve into that. The whole mismatch will resolve itself soon anyway, when the system is upgraded to Karmic.

-- ingo

[Arie Skliarouk (skliarie)]

The libgcrypt's complaint is related to bug in apparmor, which already was fixed:
https://bugs.launchpad.net/ubuntu/+source/cups/+bug/392337

[Bryce Harrington (bryce)]

[According to discussion, this is fixed in Karmic]

[Evan Broder (broder)]

I've pulled the fix from bugs.freedesktop.org into a debdiff against the xorg-server currently in Ubuntu Jaunty. I have a test build in my PPA (https://launchpad.net/~broder/+archive/ubuntu-tests) that contains the patch.

[Martin Pitt (pitti)]

Accepted into jaunty-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Martin Pitt (pitti)]

Anyone who can test the proposed package? If not, the proposed update will be removed again. Thanks!

[Kees Cook (kees)]

This package has been superseded by a security update. Please update the merge and re-upload to -proposed.

[Martin Pitt (pitti)]

There has been no activity on this SRU for more than 4 months, and at this point of Jaunty's release cycle this doesn't seem interesting enough any more to SRU.