X crashes watching videos

[berto81]

Hi.
Sometimes, when I open a video in mplayer or xine, I got this problem:
For just a moment, the screen becomes black, with some horizontal coloured lines. Then the xserver restarts (I've kdm service running).
This has never happened with older versions of drivers.
Has anyone experienced the same problem?
Thanks.

[bwkaz]

I haven't seen it, but there are a couple of things to try.

Is this only when playing a DVD? Or only playing an AVI? Or only playing something else? If it only happens when trying to play one media or format type, try something else and see if it still happens (fyi: mplayer can play pretty much ... anything. ;) )

If it happens with a lot of formats, turn off xdm (by booting to a text login), startx, and get it to happen. Then, copy your /var/log/XFree86.0.log to somewhere other than /var/log (like your home directory) and post it here.

[berto81]

It has happened with mpg, wmv and divx files.
I will disable kdm to see the xfree log.
Thanks.

[a_j_j]

I've been having exactly the same problem for ages, but nobody seems to have been able to help. I've submitted a bug report to NVidia, but to no avail.

I seem to get an X server crash about 50% of the time when starting any app which uses the Xv (Xvideo) extensions - e.g. xine, mplayer, xawtv, etc.

It's gratifying to know that I'm not going totally insane...

[berto81]

no, you aren't insane... : )
I also use the xv output.
My xserver doesn't crash the 50% of times, but quite rarely.
It never crashed after I disabled kdm.
I will post my log if it happens again.

Maybe this problem depends by videocard or agp chipset?
I have a gf2 mx400 (asus v7100 pro) with tv-out (I enable it by nvtv when needed, not by XF86Config).
My mobo is an Asus via694X. I use agpgart with agp_try_unsupported option. This is the only way I can get agp working on my mobo.
Is your configuration similar?

[a_j_j]

I'm using a GF4 Ti4400 on an Epox 8KHA+ mobo, Athlon XP 1800+. Kernel is 2.4.20, using agpgart (although I get the same problem using NVidia's AGP code), but TV output not enabled. I'm using gdm as my login manager.

I've managed to extract core files from X (using the -core option), but they're not really a lot of use, since gdb won't backtrace through the helpfully symbol-free binary libraries...

[elanthis]

Setup your login manager (gdm, kdm, whatever) to restart the X server after logout. I've found that with me, the problem, if it happens, always happens after a fresh restart of X - i.e., if xvideo doesn't crash the first time I use it, it never will there-after, until I log out and back in. I have GDM setup to kill/restart the X server on every logout, and the problem is basically "fixed".

[a_j_j]

Thanks for the suggestion - I've already got gdm set up this way, and certainly you can get to watch videos in the end... but that's not really the issue. It'd really be best if it didn't crash at all - mostly because it's very annoying when, in the middle of a session, you try to watch TV and it takes all your work with it. (Yes, I've learned to save beforehand ;)

Since this appears to be a problem which several people have experienced, does anyone from NVidia lurk here and know anything about it?

[elanthis]

Odd - after setting up my GDM this way, X never crashes anymore. X only crashed on Xvideo for me if I "soft-restarted" X (i.e., when X does the reinitialization I think is when the bug is manifesting itself). Setting GDM to fully restart the X server, boom, no more problem.

Of course, the desktop is unbelievably slow with the new drivers' crappy Render bug...

[a_j_j]

Hmm. Maybe my GDM config is wrong, then - what option should I check for? I can't see anything obvious in gdm.conf or find any command-line switches I should use...

Thanks.

[elanthis]

In the [daemon] section, make sure you have

AlwaysRestartServer=true

be sure there isn't a conflicting statement, and that it is uncommented, and all that good stuff.

[elanthis]

I should probably note, this is with GDM2 (i.e., the GTK2/GNOM2 version). I don't use the deprecated GNOME1.x stuff at all.

[a_j_j]

Ah, right. Since I'm using Slackware 8.1, which has Gnome 1.something, and since I'm behind a dialup link and therefore don't have a whole lifetime to download Gnome 2, I've improvised a solution which appears to work...

Since I'm only running X on display :0, in /etc/opt/gnome/gdm/PostSession, I copied the Default script to :0 and added the following lines to the end:

# Force X server restart
if [ -f /tmp/.X0-lock ]; then
kill $(cat /tmp/.X0-lock)
fi

This seems to have done the trick, despite being a hideous kludgy hack. :)

Thanks very much to elanthis for the hints. Hopefully this could solve the issue for everyone else as well.

[elanthis]

I'm fairly sure GDM1 had the option; a search thru the docs might bring up what it is (if GDM even has docs... ~,^ )

Otherwise, a post to gnome-list@gnome.org might return some useful answers.

[a_j_j]

Well, gdm does have some HTML man pages squirrelled away somewhere in /opt/gnome/share and they don't mention the option at all. I added it to the gdm.conf file anyway before trying anything else, but it seemed not to do anything.

For reference, I think my gdm is version 2.2.1 (going by the Slackware package info)

[berto81]

yeah, X never crashed on my machine before I disabled kdm. I would try to do as you suggested with kdm.
So, how to setup kdm to restart xserver every logout?
I edited /usr/share/config/kdm/kdmrc, but haven't found an option that sounds to help.
Any idea?
I use latest kde version.

p.s.: concluding, it's a problem of nvidia driver or of XFree?

[berto81]

/etc/X11/xdm/TakeConsole
I added a_j_j script lines and it works.
Thanks!

[lcarey]

Did I miss something or has no one come up with a real solution? The horizontal line X crashes occur for me (RedHat 8.0, k2.4.18-7.80, nvidia driver 1.0.4191, Quadro2 Pro) perhaps once a month on one of 4 identical computers. I'm not sure if its video playing related, as i'm not always the one using the computer when X goes down. Changing to init 3 is not an option, as these are lab computers. I've very happy with the 3d acceleration of the nvidia drivers, both for tuxracer and for real work, however this crashing thing is really annoying. Its certainly window-manager nonspecific, as it happens for people using kde, enlightenment, and fvwm2. The machines use gdm as the login manager, and its does a pretty good job of restarting itself, but you still loose all your work and open windows. Is there a way to get any crash information that would help debug this problem? It appears to be fairly common, and really the only black spot on an otherwise great driver.
-Lucas

[berto81]

yes, those are just workarounds...

the real solution can only be provided by nvidia on next version of drivers. I hope they won't ignore this problem, after they saw so much people experienced it.

I will try to not restart xserver every logout, and make a copy of xfree log after the crash (and before the x restart!), modifying some scripts. maybe we will get information for debugging.
suggestions?

[lcarey]

There's always the -core option on Xserver, but the guy a couple posts back reported no luck with that. I imagine that the /var/log/XFree86.0.log would have something usefull in it if you turn of [gxk]dm so that it doesn't get overwritten. I can't do that unfortunately. I've been having this crash problem for several revisions now, so i'm not hedging my bets as to a fix.
Can you make it crash consistently? I can't, though its seems like playing videos and logging out of gnome or kde are more likely to make in crash than, say, opening up a couple of xterms.
-Lucas

[berto81]

This is the ending part of my XFree86.0.log after a crash:

(II) Loading extension NV-GLX
(II) [GLX]: Initializing GLX extension
(II) Open APM successful
(II) NVIDIA(0): Setting mode "1280x1024"
GetModeLine - scrn: 0 clock: 135000
GetModeLine - hdsp: 1280 hbeg: 1296 hend: 1440 httl: 1688
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1066 flags: 5

Fatal server error:
Caught signal 11. Server aborting

Really useful... :-(
I found nothing of strange in the rest of file, only some warnings about unsopported modes (shouldn't care...).

In XFree86.9.log I found a "warning" output:

(WW) NVIDIA(0): Failed to verify AGP usage

but looking in /proc/driver/nvidia/agp/status, it seems agp is working.

Status: Enabled
Driver: AGPGART
AGP Rate: 4x
Fast Writes: Disabled
SBA: Disabled

I attached the logs.

[lcarey]

SIGSEGV (11) usually leaves a useful core dump. It might help to look at that if debugging symbols were compiled into the xserver. Thats something someone at nvidia should look into. I have no idea (nor do I care) how to begin to debug an xserver. I've got the agp warning as well, and 16 of the GetModeLine entry pairs at the bottom of my log for a working xserver--been up a couple of days now. It's a bit of a shame that noone from nvidia has responded to any of this. We don't pay for the drivers, but we pay loads of money for the graphics card. I get better support from gpl software.
-Lucas

[Argh!]

I use KDM (KDE 3.1 RC6), and after looking around on the net, I found that adding TerminateServer=true to my /opt/kde/share/config/kdm/kdmrc file under the [X-*-Core] section forces a X server restart after logout. The effect appears to be the same as adding killall X to TakeConsole. The one difference I noticed is that with automatic login turned on, KDM logs me right back in with the TakeConsole approach, while if I use the kdmrc approach, KDM stays at the login screen.

[Ryo-Ohki]

Well, I just got xine installed. It works great with DVDs and misc. video formats (the binary codecs help of course), BUT... the X server restart issue came out of nowhere after two days of normal operation. I think it has to do with the fact I rebooted my box inbetween.

:(

The problem is now 100% consistent. Log off current session, log in again as root or my regular user account, start xine... BAM, server crash. Once X has restarted, I can log in again and use xine normally though, it won't crash anymore.

Switching to runlevel 3 (text mode) for system start-up and lauching X manually circumvent the issue. Back to runlevel 4 I used the kdm autorestart trick above, looks like things are going well for now...

The system is 100% stable otherwise, wether in games (TuxRacer, UT & UT2003) or anything... To say I'm disappointed would be an understatement!

[gsbarry]

Dear all,

My problem is quite similar.

My situationis :
I played some muisc videos one by one (several minutes for one video) using mplayer. After 3-4 hours, the video card signal will disappear.


I use RedHat 7.3, KDE and the 1.0.4363 nvidia driver.


When I view the detail of post here, I found that people usually fact this problem when they logout and re-login to the X-win. But, for my situation, I login only once a day, and then play the videos for the whole day.

Therefore, the "restart X " solution is not suitable for me.
I also read the message in /var/log/XFree86.0.log. But I can't found any message about server crashing.

Dear all, do you have any suggestion to me, I am very annoying by this problem.