FS#16952 - [nvidia/xorg-server] segmentation faults

Attached to Project: Arch Linux
Opened by Dmitry (rmammoth) - Sunday, 01 November 2009, 07:44 GMT
Last edited by Jan de Groot (JGC) - Saturday, 21 November 2009, 11:27 GMT
Task Type Bug Report
Category Upstream Bugs
Status Closed
Assigned To Tobias Powalowski (tpowa)
Pierre Schmitz (Pierre)
Thomas Bächler (brain0)
Architecture x86_64
Severity Medium
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 9
Private No

Details

Description: In KDE 4.3.2 is installed GIMP. Sometimes, when I did attempt to add guideline to image, KDE session crashes and login screen appears. This trouble happens not each time, but approx. in 2/3 of times.


Additional info:
* package version(s)
local/kdebase-lib 4.3.2-2
KDE libraries for the basic desktop applications
local/kdebase-plasma 4.3.2-2 (kde kdebase)
Display the contents of folders (User´s home folder as default)
local/kdebase-runtime 4.3.2-1
KDE Base Runtime Environment
local/kdebase-workspace 4.3.2-2 (kde)
KDE Base Workspace
local/xorg-server 1.7.1-1 (xorg)
X.Org X servers
local/xorg-server-utils 7.5-2 (xorg)
X.Org utilities required by xorg-server
local/gimp 2.6.6-3
GNU Image Manipulation Program
local/nvidia 190.42-1
NVIDIA drivers for kernel26.
local/nvidia-utils 190.42-1
NVIDIA drivers utilities and libraries.

* config and/or log files etc.
GeForce9800GT Phenom II 810 4GB RAM

Steps to reproduce:
1. Start KDE
2. Start GIMP
3. Open image (in my case with was RGB PNG image with size approx. 1,5K x 2K)
4. Drag'n'drop guideline from side rule in GIMP window
5. Session crashes
   xorg.conf (1.4 KiB)
   xorg.0.log (13.8 KiB)
This task depends upon

Closed by  Jan de Groot (JGC)
Saturday, 21 November 2009, 11:27 GMT
Reason for closing:  Fixed
Additional comments about closing:  Fixed in 1.7.1.902-1.
Comment by Jan de Groot (JGC) - Sunday, 01 November 2009, 14:28 GMT
This is a bug in the nvidia driver.
Comment by Laszlo Papp (djszapi) - Sunday, 01 November 2009, 17:30 GMT
Did you try to downgrade the nvidia driver to see whether it works so ?
Comment by Gerardo Exequiel Pozzi (djgera) - Monday, 02 November 2009, 02:26 GMT
I try some times addind many an many guides from side rules both horizontal and vertical, with different png images sizes, 3300x3600 and 1024x600, etc.

No problems (or at least for now) here on both i686 and x86_64. But I am using a custom kernel 2.6.30.9 and nvidia 192.42 (with MSI-X ints enabled) with a 7300GT.

Latest packages from repos.
Comment by Dmitry (rmammoth) - Monday, 02 November 2009, 11:31 GMT
Video with this bug
Comment by Dmitry (rmammoth) - Monday, 02 November 2009, 11:36 GMT
Sorry, was above 2 mb. Same video with smaller size
   bug.avi (1.03 MiB)
Comment by Gerardo Exequiel Pozzi (djgera) - Monday, 02 November 2009, 12:27 GMT
Yes, I do the same steps but nothing.

Can you try to rebuild in your system nvidia pkg, then nvidia-util ? Only for test.
Comment by Vadim Naboikin (clearwater67) - Tuesday, 03 November 2009, 10:25 GMT
Yes, I confirm this bug on x86_64.
I have 190.42-1 nvidia driver with GeForce 9400 GT
Today I'll try to rebuild driver on my computer, then will tell you about the results
Comment by Cedric Bellegarde (gnumdk) - Thursday, 05 November 2009, 21:26 GMT
Don't think it's an nvidia bug, http://bugs.archlinux.org/task/17013

I'm running 190.42 with fedora patch on all my computers and no crash anymore... Look at nvidia forums, patch fix bug for others users.
   xserver-1.7.1-window-pictures... (3.4 KiB)
Comment by Jan de Groot (JGC) - Thursday, 05 November 2009, 21:33 GMT
That patch has been replaced by a new version, but regressions have been reported. I'll wait a bit before applying the patch to the xorg-server package.
Comment by Panos Filip (Flamelab) - Saturday, 07 November 2009, 15:49 GMT
With the latest Xorg server (1.7.1.901), but without the patch, I confirm that there is such an error: I tried to open multiple images with Gimp, and there was a crash (back to KDM).

I haven't used the latest one (1.7.1.901) with the new fedora patch, since with that one, there is a huge lag in KDE.
Comment by Thomas Bächler (brain0) - Saturday, 07 November 2009, 15:54 GMT
Yes, the version from extra is unusable with nvidia. Every time I open a window X hangs for about 10 seconds.
Comment by Alex Parker (AlexP) - Saturday, 07 November 2009, 23:31 GMT
Also here with Geforce G105M and kde 4.3.3 - with autohiding panels, the panels take ~10 seconds to appear and during that time the system is not useable. The previous version segfaulted when I opened thunderbird (which no longer happens).
Comment by solsTiCe (zebul666) - Sunday, 08 November 2009, 15:20 GMT
i confirm this bug. Xorg seg fault with this erros in Xorg.0.log
Backtrace:
0: /usr/bin/Xorg (xorg_backtrace+0x3b) [0x809fb4b]
1: /usr/bin/Xorg (0x8048000+0x54b35) [0x809cb35]
2: (vdso) (__kernel_rt_sigreturn+0x0) [0xb782040c]
Segmentation fault at address 0xaddf0b1c

Fatal server error:
Caught signal 11 (Segmentation fault). Server aborting

with xorg-server 1.7.1.901-1 nvidia 190.42-1 on a Geforce 8300GS

i was trying to resize a big image in gimp when it crashes
Comment by jib (jib2) - Monday, 09 November 2009, 07:16 GMT
Same bug with new xorg-server 1.7.1.901-2
[Nvidia 7300- driver nvidia 190.42-1 - Gnome]

Segfault when launching rhythmbox (GTK) or VLC (qt).
Comment by jib (jib2) - Monday, 09 November 2009, 09:08 GMT
Additional info:

Errors from Xorg.0.log :

Backtrace:
0: /usr/bin/X (xorg_backtrace+0x28) [0x45b7d8]
1: /usr/bin/X (0x400000+0x62689) [0x462689]
2: /lib/libpthread.so.0 (0x7fab10387000+0xe7e0) [0x7fab103957e0]
3: /usr/lib/xorg/modules/drivers/nvidia_drv.so (0x7fab0c135000+0xa4aa2) [0x7fab0c1d9aa2]
4: /usr/lib/xorg/modules/libwfb.so (wfbPadPixmap+0xbb) [0x7fab0bd05f5b]
5: /usr/lib/xorg/modules/libwfb.so (wfbValidateGC+0x301) [0x7fab0bd062d1]
6: /usr/lib/xorg/modules/drivers/nvidia_drv.so (0x7fab0c135000+0x349ad1) [0x7fab0c47ead1]
7: /usr/bin/X (0x400000+0xaa2ba) [0x4aa2ba]
8: /usr/bin/X (ValidateGC+0x24) [0x422334]
9: /usr/bin/X (0x400000+0x43ab8) [0x443ab8]
10: /usr/bin/X (0x400000+0x46674) [0x446674]
11: /usr/bin/X (0x400000+0x2205c) [0x42205c]
12: /lib/libc.so.6 (__libc_start_main+0xfd) [0x7fab0f4319ed]
13: /usr/bin/X (0x400000+0x21c09) [0x421c09]
Segmentation fault at address 0x7faafbd3c200

Fatal server error:
Caught signal 11 (Segmentation fault). Server aborting
Comment by Maximilien Cuony (the_glu) - Thursday, 12 November 2009, 22:22 GMT
Same problem here, see  FS#17112 
Comment by svnieuwe (svnieuwe) - Monday, 16 November 2009, 00:24 GMT
The fix_pixmap_nvidia_issue.patch from the following PKGBUILD seems to solve the issue: http://gitorious.org/arch-linux-greece/arch-linux-greece/trees/master/list-of-sources/adslgr64/xorg-server
X does not crash anymore and KDE works fine (nvidia 190.42-1 on a Quadro NVS 140M)
Comment by Jan de Groot (JGC) - Monday, 16 November 2009, 07:35 GMT
That patch is an older patch that gives regressions on all other drivers. The segfault patch from fedora that causes the slowdowns is the right patch which will get applied to our package. It fixes issues with Intel graphics also. The fact that nvidia freaks out with the new patch is a problem with the nvidia driver, Fedora already forwarded the issue to Nvidia who will look into it for the next driver version.
Comment by Panos Filip (Flamelab) - Monday, 16 November 2009, 13:52 GMT
@JGC: Are you sure that it is the one that causes the segfaults with the current xorg-server* ? I have that package uploaded on adslgr64, and from the feedback I get from many users, there has been no problem.

*I mean, does it cause segfaults on previous xorg-server versions ?

Comment by solsTiCe (zebul666) - Tuesday, 17 November 2009, 10:16 GMT
a fix for that is really needed. asap. you can not do anything in GIMP whitout it crashes sooner or later !
Comment by Jan de Groot (JGC) - Tuesday, 17 November 2009, 10:27 GMT
I will apply the latest version of the fedora patch soon. The slowdowns in KDE caused by applying this patch when using the nvidia driver is a side effect that should be fixed by nvidia. This is already reported to nvidia by Fedora, but as Nvidia doesn't open their drivers, we can't and won't fix that issue until Nvidia releases a new driver.
Comment by Pierre Schmitz (Pierre) - Tuesday, 17 November 2009, 11:08 GMT
Could you keep the patched xorg in testing then? It seems like only newer nvidia crds are affected by the crash mentioned here while everyone is affected by the massive slowdown.
Comment by Jan de Groot (JGC) - Tuesday, 17 November 2009, 11:29 GMT
The patch also fixes dualscreen issues with Intel drivers and fixes crashes with Radeon, so the patch is needed. Besides fixing crashes, it also fixes things like notification areas that don't appear with some configurations. I can't and won't support a binary blob that is unfixable. People who are affected by the slowdowns can build their own xorg-server if they really have to run the nvidia drivers with KDE.
Comment by Pierre Schmitz (Pierre) - Tuesday, 17 November 2009, 11:45 GMT
OK, I did not know that this fixes other issues in xorg-server as well. In that case we have no choice.
Comment by Daniel Riedemann (darie) - Tuesday, 17 November 2009, 13:02 GMT
Hello all, I just wanted to leave the link to the xorg bug report here. The bug report was opened by nvidia:
https://bugs.freedesktop.org/show_bug.cgi?id=25136

So it's not nvidias fault?!
Comment by Thomas Bächler (brain0) - Tuesday, 17 November 2009, 13:34 GMT
From the description in that last report, it seems nvidia that the problem is in fact in Xorg, not their driver. This bug report has only been opened yesterday, so we might want to wait a few days and see how the Xorg people respond to nvidia's arguments here.
Comment by Panos Filip (Flamelab) - Tuesday, 17 November 2009, 14:03 GMT
@JGC: Please, if you release a new xorg-server with the new patch, please, move it at [testing] first, so that Nvidia users will test it.
Comment by (N/A) (wantilles) - Wednesday, 18 November 2009, 07:24 GMT
@JGC

"...binary blob..."

This excellent nvidia "binary blob", is one the major reasons why tens of thousands of people have been using Linux for the last few years.

With all due respect:

Is this an "ideology" issue?

Because silly "ideology", silly "political" decisions, and "extremities", was the reason I stopped using Debian 3 years ago and started using Arch.
Comment by Yaro Kasear (Yaro) - Wednesday, 18 November 2009, 20:52 GMT
@wantilles - I don't think its an ideology issue. And I to get sick and tired of the GNU half of the Linux community whining about proprietary software.

The nv driver is far from feature complete and likely never will be. I want to get the most of my hardware, and I am sorry to say that may require the "evil" binary blobs.

And frankly, these people like to hop on the "blame the bug on the binary blob" wagon every time there's a bug related to them. I did digging myself, including the bug report nVidia filed with the actual Xorg developers. It's become pretty clear to me this is a bug in xorg-server, not in nvidia.

I use Linux because it's great and it works, not because its FOSS. That's just a perk to me.

Now, as for this bug, I can confirm it, and its not just for that specific action in GIMP. It's ranged for me from applying a gradient, to filling a solid color, to even just starting a new project.
Comment by Yaro Kasear (Yaro) - Wednesday, 18 November 2009, 23:21 GMT
@wantilles - I don't think its an ideology issue. And I to get sick and tired of the GNU half of the Linux community whining about proprietary software.

The nv driver is far from feature complete and likely never will be. I want to get the most of my hardware, and I am sorry to say that may require the "evil" binary blobs.

And frankly, these people like to hop on the "blame the bug on the binary blob" wagon every time there's a bug related to them. I did digging myself, including the bug report nVidia filed with the actual Xorg developers. It's become pretty clear to me this is a bug in xorg-server, not in nvidia.

I use Linux because it's great and it works, not because its FOSS. That's just a perk to me.

Now, as for this bug, I can confirm it, and its not just for that specific action in GIMP. It's ranged for me from applying a gradient, to filling a solid color, to even just starting a new project.
Comment by Yaro Kasear (Yaro) - Wednesday, 18 November 2009, 23:22 GMT
Oh crumbs. How do I remove comments?
Comment by Kevin DIxon (soggydoggy) - Thursday, 19 November 2009, 20:55 GMT
I don't think this is just GIMP, or KDE related.
I have the same X crash back to GDM when I use the built-in browser in Songbird.
The error reported is exactly the same as you have here "Fatal server error: Caught signal 11 (Segmentation fault). Server aborting".
I have a BFG Nvidia 9800GTX+ OC so use the latest Nvidia drivers from the repos and the latest X from the repos.
Looks like it is an X bug indeed, but I'm confident that it will get sorted asap as does everything in Arch <3
Comment by svnieuwe (svnieuwe) - Friday, 20 November 2009, 22:13 GMT
Proposed patch on the xorg-devel mailinglist: http://lists.x.org/archives/xorg-devel/2009-November/0035
Comment by Yaro Kasear (Yaro) - Saturday, 21 November 2009, 02:43 GMT
@Stijn - That link is no good.
Comment by solsTiCe (zebul666) - Saturday, 21 November 2009, 08:56 GMT
may be http://lists.x.org/archives/xorg-devel/2009-November/003569.html ?
it is a fix for the kde slow down. the fix for crashes comes from fedora
Comment by svnieuwe (svnieuwe) - Saturday, 21 November 2009, 09:21 GMT
Something went wrong with the previous link I posted. The link posted by solsTiCe should contain a patch that fixes the performance issues of the Fedora patch. The patch is provided by a NVIDIA employee.
Comment by svnieuwe (svnieuwe) - Saturday, 21 November 2009, 10:08 GMT
Added quick and dirty patch and PKGBUILD.

The above patch seems to solve the issue for me.
   PKGBUILD (4.2 KiB)
   xserver-1.7.1-nvidia-performa... (2 KiB)

Loading...