Just throwing it out there because I caught it on a forum - but if you have a BIOS cpu setting called something similar to “Execute Disable Function”, see if it's *enabled*. Some people have experienced XOrg hangs on 64bit with SUSE 10.1 when that is *disabled*.

It's a long shot, but you never know. You might want to poke around other BIOS settings too, to see if something relevant with your mboard and this config stands out, since Mark is very close to your setup but can't repro.

Cheers,

J.C.

JColdrick
if you have a BIOS cpu setting called something similar to “Execute Disable Function”

Nope, I have no such feature in BIOS

JColdrick
Some people have experienced XOrg hangs on 64bit with SUSE 10.1 when that is *disabled*.

I have problems only with mplay, all other apps, including these that intensively uses OpenGL such as Maya, Nuke and so on works well..

JColdrick
It's a long shot, but you never know. You might want to poke around other BIOS settings too, to see if something relevant with your mboard and this config stands out, since Mark is very close to your setup but can't repro.

I saw such problem on different PCs, so I would agree with Mark it probably some combination of kernel/driver/xorg features reveal that problem…

Here is my xdpyinfo (enabled X extensions are listed there) and glx and glew infos (glx related stuff)
Maybe that would help somehow, try to compare with your's… especially for X extensions and XOrg version…

I still have problems with build 8.1.683

On resolution 1024@768 (or less) mplay hungs, untill I begin to resize it`s window, when I press “f” key.
On 1280@1024 X hungs immediately (no need to press “f”).
(this is my monitor resolution at home)

(All images were opened from disk)

TimurH
On 1280@1024 X hungs immediately (no need to press “f”).
(this is my monitor resolution at home)

I have 1920x1200 desktop resolution, but anyway exactly 1280x1024 resolution hangs X process when Houdini is opened in background.
One more observation:
If I open that image using mplay <image>, it hangs immidiately (if houdini isn't in the background, it may not hang X, only mplay hangs for a few seconds, but image is shown incorrectly).
If I open it from mplay window using open command, it shows ok, until I press “f” key.

I've found that each small hang from “f” key, when X is still alive, is accompanied by the following log in dmesg | tail:
NVRM: Xid (0001:00): 13, 0001 bfef000c 0000009f 00000300 0000ff00 00000002
NVRM: Xid (0001:00): 28, L1 -> L0
Definately this is something related to nVidia driver functionality but I can't understand why the same driver works ok for twod :shock:
I've tried all the advices I found on nvidia forum, like acpi=off, noapic, disabling NvAGP - nothing helps.
And why it reveals only here and nowhere else if it's the driver issue?

Are you running by any chance the Xgl and compiz extensions?

Nope.
Tried once, just for fun, but houdini doesn't work with them for me, all fonts in houdini start to look strange with Xgl, and it crashes almost on every step

Well, that bit of NV dump looks about as useful to me as it likely does to anyone else, I'm afraid.

The window manager also uses the video card to do various things in 2D. I'm running sawfish with Gnome, and although MPlay occasionally locks up on exit, I've never had it lock up on resize. So it's entirely possible that the window manager is changing the conditions of the test, and thus giving us different results.

Do you 1) have a dual monitor setup? 2) work with multiple workspaces? Just trying to think of other possible differences.

However, this doesn't look too promising:
http://www.nvnews.net/vbulletin/showthread.php?t=71750 [nvnews.net]

But the 100% CPU timeouts & hangs sound very similar to the problem you're describing. You could try disabling irqbalance, as stated in this post (second post):

http://www.nvnews.net/vbulletin/showthread.php?t=74547 [nvnews.net]

twod
Well, that bit of NV dump looks about as useful to me as it likely does to anyone else, I'm afraid.

I felt it

twod
The window manager also uses the video card to do various things in 2D. I'm running sawfish with Gnome, and although MPlay occasionally locks up on exit, I've never had it lock up on resize. So it's entirely possible that the window manager is changing the conditions of the test, and thus giving us different results.

I also thought about that, and therefore tried it in TWM (very very simple window manager), i've just have no gnome installed. And, in TWM there was exactly the same thing I had in KDE

twod
Do you 1) have a dual monitor setup? 2) work with multiple workspaces? Just trying to think of other possible differences.

1) No.
2) What do you mean “workspaces”? I have 2 desktops in KDE, but Im afraid it doesn't matter because in TWM there was only one possible

twod
However, this doesn't look too promising:
http://www.nvnews.net/vbulletin/showthread.php?t=71750 [nvnews.net]

But the 100% CPU timeouts & hangs sound very similar to the problem you're describing. You could try disabling irqbalance, as stated in this post (second post):

http://www.nvnews.net/vbulletin/showthread.php?t=74547 [nvnews.net]

I've read almost all posts on nVidia forum that may relate to my problem, but I've noticed almost all of similar posts are describing problems everywhere, in almost any OpenGL intensive application.
In my case, however, the problem arise only in mplay and nowhere else.
The first time I saw that I thought something wrong with my hardware, but after I saw it on other PC's with completely different configs then my, I think it's related to some combination of driver version and xorg version (and maybe configuration) and to kernel of course..
And maybe only some unique sequence of commands used in mplay arise that bug from such combination
Twod, can you say exact version of xorg you're running, and your PC config? I'll try to compare it to mine.
The other thing that we may try (don't know if it's possible) If you make some special build of mplay with printf's there and there, to help you get an idea about the piece of code where the bug arises…

twod
You could try disabling irqbalance, as stated in …

I've tried, no effect.
BTW, installed Gnome and tried - still the same thing.

Hallelujah!
http://www.nvnews.net/vbulletin/showthread.php?p=963486#post963486 [nvnews.net]

Interestingly enough, I found an internal web page of ours, with scores of images on it (some large), will also produce exactly the same result very repeatably. This is with Firefox, of all things. It's hung my X session once and booted me out to the login twice (I stopped trying to look at that page after that ). However, I don't get any messages in /var/log/messages.

This is with NV driver 87.62 & a QuadroFX 1000, using Xorg 7.0 under Debian Linux (kernel 2.6.11) with 2 AthlonMP 2800s & a Tyan Tiger MB, 2GB RAM.

wow! the same issue on quadro.. It's really something wrong with that drivers..
As for other hand, now you know how to reproduce it so you may isolate sequence of procedure calls that cause that hang, to help nvidia find a problem

BTW, 8774 drivers are available at nvidia, bug still there for me, however.

mlesin,

we have also been getting similar crashes at realise. with a combination of machines and OS's.

-paul

Good news.
For my config, it seems like NVidia fixed this bug in their BETA drivers:
http://www.nzone.com/object/nzone_downloads_rel70betadriver.html [nzone.com]

Phew.

How nice of them