[nvidia-new, hardy] certain websites in firefox causes X restart due to lack of wfb symlink

I can confirm this bug on my 64-bit Hardy box, while the page loads fine on my 32-bit laptop.

Crashes X here as well. 32bit though. Using the Nvidia driver with a GeForce 8400M GS

I am using 64-bit hardy with nvidia 7900gt and i CANNOT replicate this bug.

I found this digging through the logs

Apr 6 00:10:18 HardyHeron-laptop kernel: [ 313.847284] Xorg[3687]: segfault at 0000ff10 eip b6f96ba0 esp bfa65240 error 4

It looks to me as though this is a video driver bug, possibly in the proprietary nvidia drivers. Nik, what drivers are you using?

Crashes my X also. Using 32 bit hardy with nvidia 169.12 driver (geforce 8800gts 320).

My Xorg.0.log.old finishes with this entry;

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x7e) [0x80c780e]
1: [0xb7f27420]

Fatal server error:
Caught signal 11. Server aborting

This definitely seems to be something with the Nvidia driver. I installed the Beta driver and was not able to reproduce this.

crashes here too.

dell inspiron 1420

nvidia 8400M GS

please attach your xorg.conf and your complete Xorg log taken _after_ reproducing the crash.

Update:

opening the site ubuntuguide.org gives the same result

http://ubuntuguide.org/wiki/Ubuntu:Gutsy

This only happens in Firefox3 beta 5 and epiphany

it DOES NOT happen in firefox 2 in hardy

where is my xorg log so i can upload it?

http://ubuntuguide.org/wiki/Ubuntu:Gutsy also gives me the same result.

Attached is my Xorg.0.log.

Also, deleting my .mozilla folder does not fix this issue for me.

Here are my x-files after visit (crash) to http://ubuntuguide.org/wiki/Ubuntu:Gutsy

/Sune

Ups..forgot this

Backtrace generated from /etc/X11/core

Confirmed, 64bit 8.04, nvidia 8600gt, 169.12 drivers.

Closing superfluous xorg entry.

There was a workaround for a Nvidia driver issue and Compiz posted here
https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/194851/comments/37
Interestingly it also fixes this problem as well. It seems to be just a workaround though. I think the underlying problem has to do with the packaging of the driver since this bug and the bug with Compiz only seem to affect the 8 series cards.

On my system it sometimes crashes X, other times it forces a hard restart.

Can confirm with 64-bit and 8400 GS (it seems to be platform independent, 8 series only).

From Xorg.0.log.old:

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x6a) [0x48402a]
1: /lib/libc.so.6 [0x7fa58dfaf100]
2: /usr/lib/xorg/modules//libwfb.so(wfbRasterizeTrapezoid+0x18b) [0x7fa58b02d6cb]
3: /usr/lib/xorg/modules/drivers//nvidia_drv.so(_nv000745X+0x12f) [0x7fa58ba3e38f]

Fatal server error:
Caught signal 11. Server aborting

firefox -safe-mode did not help. I tried nohup firefox and got this from nohup.out:

debugger.onModuleDeactivate Attempt to deactive context that is not active http://www.google.com/search?q=ubuntuguide.org+restarts+X&ie=utf-8&oe=utf-8&aq=t&rls=com.ubuntu:en-US:unofficial&client=firefox-a
debugger.onModuleDeactivate Attempt to deactive context that is not active http://www.google.com/search?q=ubuntuguide.org+restarts+X&ie=utf-8&oe=utf-8&aq=t&rls=com.ubuntu:en-US:unofficial&client=firefox-a
debugger.onModuleDeactivate Attempt to deactive context that is not active http://ubuntuguide.org/wiki/Ubuntu:Gutsy
** Message: GetValue variable 1 (1)
** Message: GetValue variable 2 (2)
** Message: GetValue variable 1 (1)
** Message: GetValue variable 2 (2)
** Message: GetValue variable 1 (1)
** Message: GetValue variable 2 (2)
** Message: GetValue variable 1 (1)
** Message: GetValue variable 2 (2)
** Message: GetValue variable 1 (1)
** Message: GetValue variable 2 (2)
** Message: GetValue variable 1 (1)
** Message: GetValue variable 2 (2)
firefox: Fatal IO error 104 (Connection reset by peer) on X server :0.0.

...don't know if it's relevant. I propose an importance raise - having a web site able to crash Xorg might have security implications, right?

Hi,

THis bug is truly disgusting,

sorry i forgot to mention, i am running hardy 64bit, firefox 3.0 beta 5, 8800 gts (g92 i think) nvidia. feel free to ask for any more info.

Does somebody want to try this?
sudo ln -sf /usr/lib/nvidia/libwfb.so.xserver-xorg-core /usr/lib/xorg/modules/libwfb.so
Restart X and see if you can reproduce the crash.
Someone had a very similar issue here
http://www.nvnews.net/vbulletin/showthread.php?t=112098
This is the explanation according to Nvidia
http://www.nvnews.net/vbulletin/showpost.php?p=1639245&postcount=13

13 crashes later (about 4 of which were complete system freezes), I have managed to reproduce the bug with a locally saved and reduced version of ubuntuguide.org, using a minimal HTML (only an inclusion of /skins/monobook/main.css and a bodyContent of height 200000px). imho, this is almost certainly related to an on-screen buffer size: Firefox does not crash when loading this page in a tab in the background, only when displaying it.

I will now try to reduce the required CSS and see if Compiz makes any difference. I will try to post the reduced test case afterwards.

Another 12 crashes later, I have managed to write a reduced HTML file that reproduces the bug. It doesn't even deserve its own attachment, being a 1-liner.

To reproduce the bug, gedit killer.html and paste the following line:

<div style="border: 1px solid #000; border-left: none; height: 200000px;"></div>

Save, close, and open with Firefox. WARNING: This will probably crash X for users with proprietary Nvidia drivers and 8xxx series cards (at least it crashes for me).

Compiz does not appear to make a difference.

All style declarations seem to be necessary (the border color shouldn't matter, and probably neither should the border size, just its presence).

I didn't pinpoint an exact value for the height yet - I only know that it's greater than 100k px. I will try to find a more precise value now (at least for my hardware and/or settings).

On my system, Firefox crashes with heights starting from 131063px (it works for 131062px). This happens at any screen resolution (up to 1280x800), at any Firefox window size (up to maximized), and independent of Compiz being enabled or disabled. I'm using an 8400GS with 256Mb RAM.

Updated test case (WARNING - this can crash your X server or freeze your entire computer):
<div style="border: 1px solid #000; border-left: none; height: 131063px;"></div>

I can confirm that Silviu Julean's latest test case did work for me.

I can also confirm that Nick B.'s suggestion solved it! Both the test case,
and the 'http://www.themareks.com/xf/conspiracy.shtml' link go through just
fine.

Please fix soon, as this isn't nvidia problem, and it's best fixed before it
hits the media.

I'm on Hardy 8.04 AMD64 I cannot reproduce the bug, I'm using Intel GMA 3100 builtin VGA.

b33r, your inability to reproduce the bug is expected. This problem seems to occur only with Nvidia 8xxx (8-series) cards, and only for users of the proprietary drivers.

As a side note, I have successfully used the fix proposed by Nick B. (I don't know what libwfb does - or should do - but creating the symlink to the Xorg version apparently works with no side effects, so this might indeed be a packaging issue.)

I confirm that Nick B.' workaround works.
Hardy 8.04 AMD64
GF 8600GT

I confirm the bug and the Nick B solution.
Kubuntu hardy heron
8800GtS

İ have the same problem with ubuntuguide.org . my xorg.conf is attached

I have been having the same problem at ubuntuguide.org, and I can confirm that this occurs at the website mentioned in the bug. I have an Nvidia 8400m GS, running with proprietery drivers on Hardy 64bit. I can confirm that Nick B.'s workaround solved the problem for me.

Confirm the bug on Nvidia 8600, Ubuntu 8.10.
Also Confirm the fix by Nick B. (thanks)
"sudo ln -sf /usr/lib/nvidia/libwfb.so.xserver-xorg-core /usr/lib/xorg/modules/libwfb.so"

Good luck.

Ok, so the libwfb links were basically my fault, since I added them soon after Hardy was released, damn...

Fixing this should also fix the pink shadows bug (bug 186382).

I did the following and installed apparently the 'fixed' 173.08 drivers. now everything works so far.
however i had to reboot esc and ask ubuntu to fix the xorg for me during boot up

[Update]
Nvidia reports
 (http://www.nvnews.net/vbulletin/showpost.php?p=1639245&postcount=13):
 Re: This web page crashex 169.12 and 173.08 but not 169.04
 It should be using the one from the server. That looks like a bug in
 the package.

 The .run installer will always create a symlink from
 libnvidia-wfb.so.1 to libnvidia-wfb.so.1.x.y.z, and then will create a symlink from
 libwfb.so to libnvidia-wfb.so.1 only if an existing libwfb.so doesn't
 already exist. The idea is that if the server installed one, it'll use that,
 and if someone used a distribution package and then installed the .run
 file over it, the distro package would have already created a
 libwfb.so -> libnvidia-wfb.so.1 symlink. In practice, I don't think that was
 communicated clearly enough to the distributions.

[Testcase]
To reproduce the bug, gedit killer.html and paste the following line:

<div style="border: 1px solid #000; border-left: none; height:
 200000px;"></div>

Save, close, and open with Firefox. WARNING: This will probably crash X
 for users with proprietary Nvidia drivers and 8xxx series cards (at
 least it crashes for me).

[Workaround]
sudo ln -sf /usr/lib/nvidia/libwfb.so.xserver-xorg-core
 /usr/lib/xorg/modules/libwfb.so

Or alternately,

To temporarily fix the problem (if you're running nvidia-glx-new)
 download the updated libwfb and then copy it over to your
 /usr/lib/xorg/modules directory, and then create a new symbolic link to libwfb:

$ wget
 http://people.freedesktop.org/~aplattner/wfb/Linux-x86_64/libnvidia-wfb.so.173.08
$ mv libnvidia-wfb.so.173.08 /usr/lib/xorg/modules/
$ sudo chown root /usr/lib/xorg/modules/libnvidia-wfb.so.173.08
$ sudo ln -sf /usr/lib/xorg/modules/libnvidia-wfb.so.173.08
 /usr/lib/xorg/modules/libwfb.so

Restart xorg and everything should function normally. Enjoy!

To restore back to the original, drop back to the X terminal
 (ctrl+alt+f1) and then link back the original library.

$ sudo ln -sf /usr/lib/xorg/modules/libnvidia-wfb.so.169.12
 /usr/lib/xorg/modules/libwfb.so

instead of libnvidia-wfb.so.173.08

And your X should function as normal without having to reinstall the
 nvidia drivers.

without having to install it manually
how do i install the 169.04 drivers? i remember having that in hardy. i've got a 8800gts (g92) thanks

I'll upload the fix to hardy-proposed.

Here is the debdiff, contains another fix too.

Accepted into hardy-proposed, please test and give feedback here.

Whoops, sorry. I accepted linux-restricted-modules.2-6.24, not envy (yet).

I can confirm the driver from Hardy Proposed fixes this problem. I reverted the workaround before updating the driver to make sure it would work for someone that didn't do the workaround. And it's working fine now, no more crashes.

:~$ apt-cache policy nvidia-glx-new
nvidia-glx-new:
  Installed: 169.12+2.6.24.12-17.36
  Candidate: 169.12+2.6.24.12-17.36
  Version table:
 *** 169.12+2.6.24.12-17.36 0
        500 http://us.archive.ubuntu.com hardy-proposed/restricted Packages
        100 /var/lib/dpkg/status
     169.12+2.6.24.12-16.34 0
        500 http://us.archive.ubuntu.com hardy/restricted Packages

Without having read the whole bug report, I want to make sure we are getting this right. If I don't tell anything useful, just ignore it.

This bug is not yet fixed with Version: 169.12+2.6.24.500-500.24~envy of nvidia-glx-new.

(I just read about the bug yesterday, confirmed that my X crashed when opening that site, got all available updates from proposed and rebooted. X still crashing. But that may be due to the fact that I use the envy version.)

dennda [2008-05-08 7:36 -0000]:
> This bug is not yet fixed with Version: 169.12+2.6.24.500-500.24~envy of
> nvidia-glx-new.

Right. A new -envy l-r-m package was just uploaded which has the same
bug fix as the standard l-r-m package. It should become available in a
few hours.

I can confirm that upgrading to the package in hardy-proposed worked. Before this page (http://developer.pidgin.im/ticket/4986) crashed, now it doesn't. Thanks!

I can confirm too that upgrading using the package in hardy-proposed fixes the bug 221906 (which is a dupe of this one...). Thanks !

l-r-m-envy moved to hardy-updates and intrepid, see bug 221304.

http://ubuntuguide.org/wiki/Ubuntu:Gutsy#disablectrlaltconsole does that too.

I'm running 32 bit version of Ubuntu 8.04, upgraded from 7.10. For some reason it doesn't happen if I use firefox 2 instead of the 3 beta, so I'm using that as my main web browser while waiting for a bug fix.

linux-restricted-modules 2.6.24-17.36 copied to hardy-updates.

Thank you so much to everyone who helped fix this!

On Mon, May 26, 2008 at 4:18 AM, Martin Pitt <email address hidden> wrote:

> linux-restricted-modules 2.6.24-17.36 copied to hardy-updates.
>
> ** Changed in: linux-restricted-modules-2.6.24 (Ubuntu Hardy)
> Status: Fix Committed => Fix Released
>
> --
> [nvidia-new, hardy] certain websites in firefox causes X restart due to
> lack of wfb symlink
> https://bugs.launchpad.net/bugs/212648
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>

Shouldn't this bug be marked as "fix released" then?

right, closing as fixed.

Im still getting bugs on this i just told him to enable all repos and see if it upgrades for him. This is from bug #229816

Firefox 3 on ubuntu 7.04 crashes the X server on the following url:

http://el.wikisource.org/wiki/%CE%A3%CF%8D%CE%BD%CF%84%CE%B1%CE%B3%CE%BC%CE%B1_%CF%84%CE%B7%CF%82_%CE%95%CE%BB%CE%BB%CE%AC%CE%B4%CE%B1%CF%82

It is totally repeatable.

The last part of the url is in greek. However eg. visiting the home page of the above site that also has greek characters in the url, doesn't create a problem. There is no pausing, cpu spark or disk thrashing.

var/log/messages , dmesg and the X log file, don't show anything.

from xdpyinfo:
name of display: :0.0
version number: 11.0
vendor string: The X.Org Foundation
vendor release number: 70200000
X.Org version: 7.2.0

I am pretty sure that the computer *doesn't* use a nvidia card, but I forgot to look, and it is at work. I will confirm tomorrow.

I tried the above url with 8.04 , ff 3, and a 6600 nvidia card at home, and there is no problem.

Above link doesn't crash on nvidia 169.12, 8600gt, ff3, ubuntu 8.04

waterpie: This bug report describes a problem where Firefox can crash Xorg due to an issue with the packaging of the NVIDIA binary drivers on Hardy. If your computer doesn't have a NVIDIA card or it's not running Hardy, then you have a different problem (unrelated to this one) and should open a new bug report.

Chris Coulson wrote:
> waterpie: This bug report describes a problem where Firefox can crash
> Xorg due to an issue with the packaging of the NVIDIA binary drivers on
> Hardy. If your computer doesn't have a NVIDIA card or it's not running
> Hardy, then you have a different problem (unrelated to this one) and
> should open a new bug report.
>
>
If you look at the master bug its been fixed that is why i told the user
to make sure he is up to date. Not all users will see this as some users
use upstream drivers using evny or whatever they use. and even some of
the users with our drivers dont have an issue.

--
Sincerely Yours,
    John Vivirito

https://launchpad.net/~gnomefreak
https://wiki.ubuntu.com/JohnVivirito
Linux User# 414246

Continuing my post above, the computer that has the problem doesn't have a nvidia card. It has a:
ati rage xl

Should I open a new bug report? Or should I add the info to the "master bug"? (If yes, which is the "master bug"? )

waterpie: Yes, you should open a new bug report. Your problem is completely unrelated to this bug, which is now fixed. You might find these helpful to read when writing your bug report:

https://wiki.ubuntu.com/DebuggingXorg
http://www.chiark.greenend.org.uk/~sgtatham/bugs.html