You are not logged in.
The 3.9 kernel upgrade and/or the nVidia 319 upgrade seems to have broken Xinerama support in xorg. With those upgrades in place and Xinerama enabled I can not start X because any X application will segfault in libX11:
xterm[1378]: segfault at 8 ip 00007fb1c0e09080 sp 00007fff3f994c08 error 4 in libX11.so.6.3.0[7fb1c0dd8000+133000]
If I disable Xinerama then it works.
If I downgrade to kernel 3.8 and nVidia 313 then Xinerama works.
Ideas? Maybe just libX11 needs to be recompiled against the newer libraries or something, I don't know. Is anyone successfully using Xinerama + nVidia 319 + kernel 3.9?
Last edited by oz (2013-05-14 13:53:28)
Offline
Thanks for posting, I encountered the same problem today and downgrading to kernel 3.8 and nvidia 313 fixes it for now.
Offline
Seems that I've encountered the same problem, also going back to 3.8/313 for now.
Offline
I don't know if it's the same issue. GDM(3.8) couldn't work even downgraded kernel/nvidia to version 3.8/313.
Seems that I've encountered the same problem, also going back to 3.8/313 for now.
Last edited by ijk (2013-05-17 14:52:18)
Offline
Same problem here.
Will it ever be fixed or if the statement here is true https://bugs.archlinux.org/task/35361 has it been reported to the nvidia devs?
Offline
I had to downgrade a whole bunch of packages when I did it:
nvidia-313.30-5-x86_64.pkg.tar.xz
nvidia-utils-313.30-2-x86_64.pkg.tar.xz
nvidia-libgl-313.30-2-x86_64.pkg.tar.xz
opencl-nvidia-313.30-2-x86_64.pkg.tar.xz
linux-headers-3.8.11-1-x86_64.pkg.tar.xz
linux-3.8.11-1-x86_64.pkg.tar.xz
virtualbox-host-modules-4.2.12-3-x86_64.pkg.tar.xz
lib32-nvidia-libgl-313.30-2-x86_64.pkg.tar.xz
lib32-nvidia-utils-313.30-2-x86_64.pkg.tar.xz
Even now it's not perfect, maybe unrelated but I get segfaults in some Gtk and Qt applications (eg. VMware, VLC, and others). I'm not sure if this could also be related to the X.org libraries or not. Since the upgrade/downgrade the whole system has been in a somewhat broken state.
Offline
Same here. Currently using linux 3.8.7, nvidia 313.30-3, and had to downgrade a similar set of packages as oz.
Are we sure this is an xinerama thing? I would like to file a bug report or whatever needs to be done, but I don't have enough information. I got a similar line to oz in my /var/log/everything.log, but I don't know what other information might be helpful.
Here is my experience:
I use i3 window manager, and start X from a login shell. i3 starts fine, but when I press Super+<Return> (to bring up a urxvt terminal), nothing happens. The only other way I have of launching programs is by using dmenu, and that also doesn't work. The only indication of failure is that segfault error message in my everything.log.
I tried getting a coredump, but I'm not sure how useful this is (certainly I don't know how to use it). Does anybody have any ideas about who this should be reported to?
Also, are people finding that downgrading just nvidia and linux is okay, i.e. doesn't break anything else? I downgraded my whole system to the last working state just to be safe. But I would prefer to keep everything apart from nvidia and linux up-to-date if that is possible...
Offline
It is a xinerama thing for sure just disable it within the xorg.conf or 10-monitor.conf (depending on the kind of configuration file you're using for setup) and see how it's working like a charm but on seperate X screens.
I'm not sure whether the kernel is causing the trouble or the nvidia blob or both of them. You probably won't be able to track it without looking into both so I suggest reporting it to the nvidia devs and the kernel devs.
Downgrading to the appropriate nvidia and linux packages doesn't break anything if the packages are being maintained like they should. If a package needs a higher version of those it should be one of it's dependencies.
I'm using a setup with xinerama to create 1 big X screen using 2 different cards are you guys using 2 cards either? I don't know if the bug appears while only using 1 card and prefering xinerama over TwinView, too.
Last edited by Gwenonkled (2013-05-22 12:26:12)
Offline
I'm using 3 cards, 5 screens monitors, so I need Xinerama (can't do without).
I have submitted the following report to nVidia. The issue reference number is #130522-000018.
The binary nVidia 319 driver, together with Linux kernel 3.9, is causing segmentation faults on Xorg using Xinerama.
A forum thread on the Arch linux forums describing this issue is here:
https://bbs.archlinux.org/viewtopic.php?id=163319Briefly, users running xinerama are unable to start many applications after upgrading to 319. Attempting to start a program results in the message
xterm[1378]: segfault at 8 ip 00007fb1c0e09080 sp 00007fff3f994c08 error 4 in libX11.so.6.3.0[7fb1c0dd8000+133000]
being recorded to the syslog, so there is a problem with libX11. One user reports that turning off Xinerama resolves this issue.
The users on the forum thread use a variety of hardware. I personally use 3 cards (GeForce GTX 550Ti, GeForce GT 440, GeForce 210); some other users are using two or one cards. The unifying issue seems to be that we are using xinerama.
nVidia will review this issue, and may post it on http://nvidia-submit.custhelp.com/app/answers/list. (It needs to be moderated first, so might take a while). As you can see, I have included a link to this thread in the bug report, so if anybody has further information, please post here so that the nVidia folks can see!
Last edited by ElKatrina (2013-05-22 12:50:09)
Offline
I'm using a setup with xinerama to create 1 big X screen using 2 different cards are you guys using 2 cards either? I don't know if the bug appears while only using 1 card and prefering xinerama over TwinView, too.
I'm using 2 video cards, 4 monitors, TwinView, and Xinerama. TwinView on each card then Xinerama to link the two cards to one desktop.
Offline
Update: I noticed that the bug tracker has been inactive since 2011. So I used the "Live chat" feature on nVidia's site to talk to a human.
He said that he will forward the case to the Linux Support team, who will contact me for more information. So I'll keep you folks updated.
Offline
Debian just had quite a number of security updates related to X.org. I wonder if any of that could fix these issues. Or maybe Arch is already running equivalent newer versions.
Last edited by oz (2013-05-24 18:09:09)
Offline
So it's been a while, and I've had no luck with contacting nVidia.
After posting a bug report, I spoke to somebody using the live chat feature. He then sent the bug report to somebody, who then told me to email it to linux-bugs@nvidia.com. I mailed this address (twice) and have got no response.
This is obviously very annoying, but not sure what else to do.
I think nVidia and/or linux have had some minor updates since I last posted, as well as libxinerama. Has anybody tried upgrading yet, or you all waiting for the next major version?
Offline
I tried updating last week or maybe it was the week before and it was still broken.
Offline
If someone is feeling productive it might be useful to try to either run the latest nVidia on the older 3.8 kernel or use the 3.9 kernel with the older nVidia. That might narrow down where the problem is. This is something I have been wanting to try but have not had time.
Offline
If someone is feeling productive it might be useful to try to either run the latest nVidia on the older 3.8 kernel or use the 3.9 kernel with the older nVidia. That might narrow down where the problem is. This is something I have been wanting to try but have not had time.
The problem still persists when running the new nvidia driver (dkms) with 3.8 kernel. Do we have any news from nvidia regarding this issue?
Offline
Thanks for testing that! I noticed there have been several updates in the recent nVidia drivers related to segfaults in X apps. I wonder if this could be related to that and maybe they never tested with Xinerama enabled.
Offline
Still broken with latest updates but there has been no new version of the nVidia driver.
I can confirm that it seems to be the nVidia driver alone that is the problem. I upgraded everything and Xinerama was broken. I then built and installed the 313.30 nVidia driver and Xinerama now works normally. So I'm fully upgraded except the nVidia driver.
Last edited by oz (2013-06-27 13:14:20)
Offline
any news from Nvidia?
should we contact them again?
i'm trying to contact any devs on IRC but still not luck.
Offline
I just tested the nvidia-beta-dkms 319.32-1 package from AUR together with the newest kernel (3.9.7)
The error still persists with that configuration.
Offline
I just tried the 325.08 driver, didn't fix it. I didn't see anything related to Xinerama in the change log so I wasn't expecting much.
The situation is getting really bad. The 313 driver will not compile with the 3.10 kernel. Downgrading the kernel is a pain due to complications with virtualbox and anything else that uses the kernel.
Not sure what to do at this point and I'm not even sure nvidia is going to fix this. This seems like a major issue to go on this long...
Offline
I have only one nVidia card, but I can use mulitple monitors as one X screen even if Xinerama is disabled by using metamodes (see the Screen section):
Section "ServerLayout"
Identifier "Layout0"
Screen 0 "Screen0" 0 0
EndSection
Section "Monitor"
Identifier "Monitor0"
VendorName "Acer"
ModelName "Acer X213H"
HorizSync 31.0 - 83.0
VertRefresh 56.0 - 76.0
Option "DPMS"
EndSection
Section "Device"
Identifier "Device0"
Driver "nvidia"
VendorName "NVIDIA Corporation"
BoardName "GeForce 9600 GT"
Option "NoLogo" "True"
EndSection
Section "Screen"
Identifier "Screen0"
Device "Device0"
Monitor "Monitor0"
DefaultDepth 24
Option "metamodes" "DFP-0: nvidia-auto-select +0+0, DFP-1: nvidia-auto-select +1920+0"
SubSection "Display"
Depth 24
EndSubSection
EndSection
Could something like this help any of you?
Offline
Unfortunately, no. When using multiple cards, each with multiple monitors, you have use either separate screens or Xinerama. No other way unfortunately unless you are using SLI (I believe SLI can stitch 4 monitors together without Xinerama, I'm not sure). It would be great if we didn't have to use Xinerama because it breaks all sorts of things (composite, etc) but Linux seems to be far behind Windows and OSX in its ability to handle multiple monitors.
Last edited by oz (2013-08-01 16:01:13)
Offline
The situation is getting really bad. The 313 driver will not compile with the 3.10 kernel. Downgrading the kernel is a pain due to complications with virtualbox and anything else that uses the kernel.
Not sure what to do at this point and I'm not even sure nvidia is going to fix this. This seems like a major issue to go on this long...
That sure are some bad news. My way of dealing with the problem was sticking with the old 3.8 kernel and the 313 driver until it would be fixed. I tried different kernel versions (3.8 and 3.9) which worked perfectly with the 313 nvidia-dkms module. I don't know if that would help with the 3.10 kernel.
Another Question: Has the problem already reached Debian/Red Hat/Ubuntu/SuSe and their siblings? Hopefully we won't be the only ones using a multihead setup with 2 different cards and nvidia will find motivation to fix it when these distros upgrade their nvidia packages for it'll break the support on these "mainstream" distros, too.
Come on nvidia! There has to be a way to get, at least, an answer from a driver maintainer. If they refuse to fix it they should consider that they are knowingly shipping broken products since this is a major flaw. If they don't want to maintain a proper linux driver any longer they should make it open source and let the community take over. Ignoring problems is no way of dealing with them.
Offline
this is getting bleak.
No one has managed to get any contact from Nvidia?
I dont like where this is going.
What are our options on pushing/managing a fix on this?
Offline