X.org Lockups (part 2)

HermanR · Posted: Fri Jul 22, 2005 12:23 pm Post subject:

Okay, another user with lockups... :-(

I've been running my new AMD64-computer for about two weeks now. Up to yesterday, I had no problems with lockups or anything. Since then, my computer has lockup'ed up about 4-5 times. The lockups occurred while I was away from the computer and while I was working; in both cases nothing strained the computer to the max. One time the computer froze after just logging in to Kde; the other times after working for some time.

My gentoo systeem is stable x86_64, except for SuperKaramba 0.36. X is xorg-x11 on nVidia Geforce FX5200, nvidia-driver with RenderAccel on, AGP 8x.

I remember I had AGP-related lockups on my old Athlon 900MHz-system (Asus A7V), but they were different since I could get out of X and do Ctrl-Alt-Del. These issues stopped after upgrading to a new nvidia-driver.

I'll get back after doing some more testing.

Edit: Today two crashes while I was away; computer doing next to nothing, not even a screensaver. Tried to get in with ssh, but there was no connection; an open 'tail -f /var/log/messages /var/log/Xorg.0.log' didn't show anything worthwhile. The second lockup occurred with kernel option 'noapic'. Also tried RenderAccel off and updating nvidia-drivers; to no avail. We'll see what setting AGP to 4x and FastWrites off does... Nope, locking up after 55 minutes while writing a feature request to Macromedia (64-bits Flash player!).
_________________
Gentoo (64 bits) on AMD64 X2 4400+, MSI K8T Neo2-FIR, 2 Gb RAM, 320 + 160 Gb SATA, nVidia Geforce 6600

alex3683 · n00b Joined: 23 May 2005 Posts: 41 Location: Germany, NRW

I also had those lockups randomly using different applications or even doing nothing at all.

I could provoke this lockup manually after emerging the newest, stable kaffeine (0.6-r1) and trying to scan for dvb-s channels. After the scan was done, the X Server locked up when I tried to use the scrollbars or the mousewheel in this list.
So then I tried some of the things mentioned in this thread and finally those lockups where gone!

This is my old xorg.conf Device Section:

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

Solutions are all relative, so until someone makes a minute profile of each major set of packages related to X and hardware setups (ie video cards vs motherboards, etc), finding a generalized solution or making hind or tail out of this situation will be chaotic at best.

Fillepe The Mexican · n00b Joined: 12 Mar 2004 Posts: 71

Chalk me up on the scoreboard for this problem too. Using a Geforce2mx and Xorg 6.8.2-r2. Very frustrating although my crashes seem to happen much less often than your selves. Maybe once a week. Mostly when running firefox.

I also noticed that when I scroll in firefox my cursor briefly turns into to default cursor (think the one Twm uses). Does this happen to anybody else?

Catastrophy · n00b Joined: 24 Jul 2005 Posts: 1

fuck me and my lockups

[img]http://www.darkwars.org/index.php?a=click&id=80401[/img]

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

Downgrading xorg and my nvidia driver to stable did the trick for me. I also downgraded gnome-light to stable but that I think is another story, cause all these crashes also appeared in xfce4. Just an act of desperation...

The last days were getting worse and worse: I had hourly based lockups, sometimes even could not log into X. After the downgrade there was no single lockup anymore. I have to admit that some of those "lockups" actually were IO related problems: so either my USB or my PS2 port temporary died and switching from one to the other brought my keyboard back (sometimes). These oddities are also gone now.

Strange enough, like many others too, for some time I had firefox under suspicion since most of the lockups happened when using firefox. But I never downgraded firefox and meanwhile its 1.0.6-r3 (not the binary version) and stable now (or better: again).

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

Lasker:
Your problem sounds exactly what I used to have. Either stop using XFCE4.2 or use a *BSD. I can verify that FreeBSD works with the nVidia drivers, Firefox, and XFCE4.2 perfectly fine.

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

I believe you everything, but it seems you got me wrong. All my problems are over now! And as an owner of a pretty outdated system (Athlon t-bird 1Ghz, Ge-Force2 Gts 64Mb) I never could see any improvement by using bleeding edge nvidia drivers or xorg server anyway. Its still the same performance than before the massive downgrade, just stable now!

And I'm still using XFCE4 and gnome, btw... I didn't say I downgraded everything!

Fenster · Posted: Mon Jul 25, 2005 1:04 am Post subject:

I'm getting nearly hourly lockups on my laptop, using xorg 6.8.2 and the radeon driver on the Radeon Mobility 7500 (there's no ATI PPC Linux driver). I blamed xscreensaver at first, but I see how that actions that draw/redraw things rapidly are to blame, and a screensave does that.

What are the steps to emerging an older version of a program? I never had these problems under the older xorg on my last laptop...

yaneurabeya · Veteran Joined: 13 May 2004 Posts: 1754 Location: Seattle

If you have what I had, just wait until you go to some websites... so unless they fixed the issue with font sizes in the latest Xorg the issue will occur again in the future.

sbbeebe · Posted: Mon Jul 25, 2005 3:28 pm Post subject:

Lasker -

What version of Xorg did you downgrade to? Portage doesn't have anything older than 6.8.2-r1.

I assume nvidia was 6629.

thanks,
_________________
"smell the color 9"
steve

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

Xorg: latest stable 6.8.2-r2
Nvidia: yes, 6629
more specific:
nvidia-glx-1.0.6629-r6
nvidia-kernel-1.0.6629-r4

HermanR · Posted: Mon Jul 25, 2005 8:45 pm Post subject:

Well, maybe it's a little bit too early to get my hopes up, but just thought I should share. I replaced "stable" fam with "unstable" gamin yesterday, and up to now did not experience a single lockup.

It seems that this thread discusses a number of seemingly related, but distinct kinds of lockups. My system just freezes at some random moment, even when it's doing 'nothing'. When that happens, there is nothing I can do (not Ctrl-Al-Del, nor get in through ssh from another computer, etc.).

Can it be that fam is causing the troubles? Does anyone running gamin have these lockups?

Keeping my fingers crossed...

So much for that... :-(

After 6 hours and 49 minutes of happy computing, there was another lockup while I was browsing through the gentoo forums in Firefox and listening to music in Amarok. It must have been that I was playing a bit with my mouse.
_________________
Gentoo (64 bits) on AMD64 X2 4400+, MSI K8T Neo2-FIR, 2 Gb RAM, 320 + 160 Gb SATA, nVidia Geforce 6600

Fenster · Posted: Mon Jul 25, 2005 10:15 pm Post subject:

I think my problems have been solved-Gnome and its drawing of moving windows seems to have triggered a lot of my crashes. I moved to XFCE (I've been after an excuse to use it) and I've had the machine running for nearly 24 hours without a crash

unz · l33t Joined: 28 Jul 2004 Posts: 819 Location: Roma, Italia

i've got gamin and system lockup too ... i'm quite sure that all problems started with gamin-0.1.2, earlier versions "worked"
_________________
Ma che c'hai là? Sulla spalla!!!! http://lascimmia.it/

HermanR · Posted: Mon Jul 25, 2005 10:59 pm Post subject:

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

Although the main problems seem to be gone for me since my latest major downgrade, I still had some very strange issues like repeatedly ramdom keystrokes which I couldn't stop. Smells like an interrupt problem and since I recalled a certain comment from dmesg, I'm recently experimenting with this boot-option (from grub/menu.lst):

unz · l33t Joined: 28 Jul 2004 Posts: 819 Location: Roma, Italia

Have you tried using (or not) "PCI irq sharing"?
_________________
Ma che c'hai là? Sulla spalla!!!! http://lascimmia.it/

Gergan Penkov · Posted: Wed Jul 27, 2005 11:23 am Post subject:

I think gamin and fam were discussed in the previous thread, there is no considerable difference, probably only the frequency of the lockups is changing between the two

How yaneurabeya already pointed the solutions are all relative, and until someone with knowledge of it debugs it, we all will be victims of it in some way (I don't have lockups, but very strange behaviour sometimes).
I think that all this points to a race conditions somewhere it could be kernel or glibc.
btw yaneurabeya what are versions (xorg, glibc and so on), which you use on your freebsd?
_________________
"I knew when an angel whispered into my ear,
You gotta get him away, yeah
Hey little bitch!
Be glad you finally walked away or you may have not lived another day."
Godsmack

pinr · Posted: Sat Jul 30, 2005 11:32 pm Post subject:

Ok I've been having these X lockups on my machine since I bought it about 3 months ago. It is an AMD64 2800 with a FX5200 Nvidia card. At first the problem was quite bad with lockups occuring every one of two days. The symptons are complete system lockup mouse and keyboard do not work and the machine cannot be pinged from the network. The only solution is a hard reset. I followed some of the suggestions here on this thread and I thought I had solved the problem with an uptime of 260 hours! Today I went here:
http://news.bbc.co.uk/1/hi/england/merseyside/4730559.stm
and clicked on the video link "See where the attack took place" This causes firefox to open the realplayer plugin in order to play the video. The video starts to load into the buffer and then the machine locks. I have tried 3 times with the same result. Is anybody else getting the same on this link. If so it may assist in getting to the root of this problem.
UPDATE!!
I downgraded from nvidia-kernel-1.0.7664 to nvidia-kernel-1.0.6629-
r4 and nvidia-glx-1.0.7664 to nvidia-glx-1.0.6629-r6 and I'm no longer getting lockups playing that video. So hopefully this has solved the problem for me. Actually those two packages were masked anyway so maybe it's my own fault. Still won't know for sure untill I have an uptime of over 300 hours.

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

No. I get a plugin finder/installer which wants to download Real Player 10.5 (I have realplayer 10.0.5 installed). The install fails and I get an offer to manual install Real Player (which I didn't). No further problems, no crash. But then it seems I had solved the problem for me. From the moment on I use the boot option 'pci=routeirq' I had no hard crashes any more. But I still had issues with keyboard and mouse, which both died after some time. This even happened when both were plugged to PS2. I solved that too: I threw all USB related stuff out of the kernel, deactivated USB in BIOS and assured no USB related modul was loaded. Since then also mouse and keyboard are stable. The open question is still whether it's a hardware problem or something wrong with USB related code... where the latter I think is rather unlikely since I'm using an older kernel-2.6.10-gentoo-r6 (the problems should have occured much earlier then).
EDIT
And there it was again! But not soo bad this time, listen... After I thought I had solved the problem (and after being sure it only was an USB related problem), I got cooky and upgraded xorg-x11 to 6.8.99.15 again (and since apparently nvidia driver 1.0.6629 refused to work with this version, I also updated to nvidia 1.0.7174 and voila: after about 30 minutes X crashed again (kernel was still running though). So either xorg-x11-6.8.99.15 or nvidia-1.0.7174 can't be entirely innocent at those crashes. Back to xorg-x11-6.8.2-r2 and nvidia-1.0.7174 (w/o any performance loss, btw!) the system not only is stable again, my PC is also working clearly more quietly now (which indicates less stress -> less heat). I don't know where I should place the strange USB issues into this picture, maybe it just was a side-effect. But I'm really not going to revert my former USB situation, I'm not that crazy... :roll:

A stable system actually is worth to downgrade one or the other stuff!

theDOC_23 · Posted: Tue Aug 02, 2005 4:52 pm Post subject:

a few weeks ago, I wanted to play around with my nokia and bluetooth and stuff, so I had to change my system a bit, recompiled the kernel (a new one of course), so I needed to recompile the nvidia drivers, too. By that chance I switched to 1.0.7167 and also updated my xorg to 6.8.99-14.
Everything worked for about an hour, then X crashed. I could kill it with an ssh console from an other PC. After a few days and many random crashes I tried the changes back and forth and it came up, that the new nvidia drivers are the cause of my crashes. I changed them back to 1.0.6629-r6 and everything runs fine for about a week now. no crashes at all!

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

Meanwhile it looks like it actually was the nvidia driver. After using 1.0.7174 for a while I had still 1 crash. Yesterday I donwgraded to 1.0.6629-r6 again. One day w/o any crashes is of course too soon to say something definite. I'll post back after a weak, if the crashes have been gone by then...

schmobag · Posted: Thu Aug 04, 2005 10:04 pm Post subject:

I also, am seeing the problem where X freezes, but I can still move my cursor. It only happens to me once every couple of days, so I hadn't been motivated to do anything about it until today. I'm amazed by how long the threads on this have gone.

Unfortunately, I don't think upgrading, downgrading, switching software, or other similar solutions are actually going to solve the problem. They might reduce how often it happens though. I did some searching online, and found the following thread on the xorg mailing list:

http://lists.freedesktop.org/pipermail/xorg/2005-March/thread.html#7055

Notice that the thread is from March. It seems that they have been aware of the problem for a while, but don't know what to do about it. Apparently, it occurs when a driver gets into a bad state, and xorg (or xfree) is left waiting for a response that will never come. Some programs push video cards and their drivers harder than other ones do, so it makes sense that switching from kde to gnome or whatever has worked for some people; it just reduces or eliminates the condition that causes their particular card/driver to hang. It also makes sense that turning down AGP settings has worked for some; lower settings means lower demands on your card/driver, thus reducing the probability of a hang. In addition, this explanation jibes with the theory that a freeze is most likely to happen when there are a lot of things being drawn at once. There is talk in the mailing list of adding a watchdog program to xorg that can look for this problem and reset the driver somehow, but I don't know what will come of it. Maybe there's been something in the list more recently. I haven't looked carefully enough to know.

Here's what I think I do know:

here

sbbeebe · Posted: Thu Aug 04, 2005 10:34 pm Post subject:

Agreed with scmobag - there's no solution at the moment. I dug up some things that have helped me achieve some stability.

My hardware

Shuttle an35n ultra 400 nforce2 mobo
Athlon XP 1800+
512mb ram pc2700
eVGA Geforce4 mx440 64mb

I can run fine with the nv driver. OpenGL performance is very poor. That kept me trying things, seems like I played around endlessly...

No kernel boot params seemed to help. Leaving apci in or out doesn't help. No bios changes or updates seemed to help. I tried nvidia drivers from 6111 to 7667. I tried NvAGP 0,1,2. I tried loading the nvidia module passing NVreg_ReqAGP to limit it to 2x. I tried turning off hardware cursor, renderaccel, ... I tried kernels from 2.6.8 to 2.6.12 (not every one in between). I tried xorg from 6.8.0 to 6.8.99. No matter what I did when using the nvidia driver xorg would freeze (except the mouse).

I'm currently running 2.6.12-r6 with nvidia-kernel-1.0.7174 and xorg-x11-6.8.0-r5. I don't think the roll back on xorg has been any benefit. If I encouter issues after updating to 6.8.2, I'll post that.

After doing lots of reading, it's clear that the issue comes from the video drivers over time pushing the hardware more and more, and that seems to surface hardware (both mobo and video card) and bios issues (also - I have another machine with a different mobo and nvida graphics card that has no problem).

So, for some strange reason, I decided to emerge nvclock and slow down my graphics card (probably doesn't get used for that much) to see what would happen. Normally the eVGA mx440 runs with gpu at 275mhz and memory at 400 mhz. I started off with gpu at 200 and memory at 300. Still no luck with AGP (NvAPG 1 or 2). But, I was able to run xorg for 24 hours or so with no lockups - even doing things that would have caused a sure lock before. BUT, I was getting some screen artifacts - some of it pretty annoying. So I started playing with gpu / memory speeds and watching what happened.

I ended up leaving the gpu at 200 mhz and putting the memory speed back up to 400 mhz. Screen artifacts are gone and that gave me the best glxgears fps (well, I could 1 or 2 more fps with the gpu at 220 and 240, but didn't think it was worth it). Now I need to let it run and see how long it will go - all of the things that used to make it freeze, don't at this point, but only time will tell.

** disclaimer ** you could seriously screw up your video card using utilities like nvclock. if you chose to use nvclock, you do so at your own risk.
_________________
"smell the color 9"
steve