X.org Lockups (part 2)

schmobag · Posted: Thu Aug 04, 2005 11:13 pm Post subject:

I'm not sure if I'm really understanding how the "Magic Sysrq" thing is supposed to work. According to this, I should be able to do Alt-SysRq to have the kernel put the keyboard in raw mode, followed by Alt-Ctrl-Backspace to kill X. I can't get the second command to work though. Maybe it's because that doc was written for Mandrake. On the other hand, Alt-SysRq followed by Ctrl-Alt-Delete will reboot my computer gracefully.

Does anyone know a way to use SysRq to kill X or switch to a console where X can be killed and restarted?

pinr · Posted: Fri Aug 05, 2005 12:35 am Post subject:

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

spitefulcrow · n00b Joined: 18 Jul 2004 Posts: 44

Still experiencing lockups even after trying nvidia-{glx,kernel}-1.0.6629 as suggested by other posters instead of the latest 1.0.7667, and then switching to the Xorg nv driver instead of nvidia.
I just disabled the GLX extension to see if that eases the load on the driver/GPU (I really don't use OpenGL that much at all, haven't noticed a performance hit after switching to nv). If I still get a lockup, my next step is to reduce AGP to 2x or disable it entirely.
From reading that Xorg mailing-list thread someone linked above, it seems like X.Org is either unwilling or unable to do anything about this. I wouldn't put my money on nVidia and ATI fixing their drivers anytime soon, so it looks like we're screwed, guys -- either live with daily crashes, or disable all graphics performance options and use 2D only. Some choice that is. :roll:

_________________
Athlon XP 2600+ (1.9gHz)/1024MB DDR333/Asus A7V880/Samsung 160GB SATA HDD/Chaintech GeForce FX5500 256MB

schmobag · Posted: Fri Aug 05, 2005 3:49 pm Post subject:

pinr:

The problem that most people are writing about in this thread has the following symptoms:
1 - Computer freezes (often while doing something graphically intensive), but the mouse still moves.
2 - Keyboard doesn't respond to anything, but you can bypass that by going straight to the kernel with the sysrq thing.
3 - Though the display is frozen, you can ssh into the machine and restart X (I haven't tried this yet, but it's been widely reported).

If you're seeing total lockups, no mouse, no sysrq, then your problem is probably unrelated to the one most of us are talking about. It might be X, it might be a misconfigured kernel, it might be too-aggressive optimizations, anything. Switching to another distro might solve your problem, because other distros don't give you as much rope to hang yourself with.

On the other hand, Lasker is right, the X freeze we're discussing is not just a Gentoo problem. If you google for it, you'll find reports of this problem everywhere: gentoo, ubuntu, nvidia discussion boards, linuxformat.co.uk, everywhere. One thread I saw called it "Linux's BSOD." That hurt.

What's needed is more incentives for xorg developers to come up with a way for X to gracefully deal with a hung driver. Does anyone know a place where you can advertise a software development bounty? There are so many people seeing this problem, we should be able to get like 100 people to contribute $10 each to motivate an xorg dev to do the work in return for a $1000 bounty.

deadman_kein · n00b Joined: 20 Apr 2005 Posts: 5 Location: Poland

Experienced the lockup just as I was writing a reply. And it was just like schmobag described it. I was able to ssh onto locked machine, but was unable to kill X (in a way to use the machine again). I've been experiencing theese lockups from the beginning. No matter which drivers I used (tried all of them from 6629 and up), tried kernels 2.6.11 and 2.6.12. Freezes are totally random, sometimes when X starts, sometimes couple times a day, sometimes no lockups for few days. Usually mplayer can make X lockup, but sometimes they occur when nothing happens. And my system is AMD64, nForce4 mobo, Geforce 6600 PCI Express, SATA drive.

So, to sum it up:
- it's not nvidia olny
- it's not linux only
- 32 and 64 bit systems are both affected

My questions are:
- Is XFree86 also affected?
- Will going back to older distro (like Slack 9.1 with 2.4 kernel - used that on old machine) fix the problem? I can live without bleeding edge features, I need a stable system. Thought I really like gentoo...

I'm really tired of this bug...

schmobag · Posted: Fri Aug 05, 2005 5:04 pm Post subject:

Xfree86 is also affected. From what I can glean from the xorg mailing list thread I mentioned several posts ago, the problem is caused when the video driver hangs (lots of things could cause this) and X doesn't know what to do in response. It doesn't matter whether you're using xorg or xfree. There's also no reason to think that switching distros or kernels will solve it. There is a chance that you can eliminate or reduce freezes by switching drivers, as some are more stable than others, and therefore less likely to cause a freeze.

You're right that this is not an nvidia-specific problem. I see it a couple times a week on my ATI Radeon Mobility 7500, using the radeon driver that comes with xorg.

EDIT: just noticed that you tried all the nvidia drivers you can get your hands on. Some people have increased their stability by turning down their AGP settings, though I'm not really sure how that's done (and it might not even apply to you, since you have a PCI express card).

Bobbie · Tux's lil' helper Joined: 21 Jul 2002 Posts: 77

This has to be one of the worst bug ever still not fixed...and driving me insane (XP is rock stable compared to this ...sad but true)

My report here:

A recent sony VAIO runing PCLinuxOS P9.1 with customized kernel 2.6.11 and 2.6.12
Video card is ATI X600 running the open source Xorg driver.

X hard freeze (no mouse, no ping, no rebootable) and I can kind of reproduce it in limited time (< 3h).

If I let nicotine (a p2p client) which do constant R/W access to my USB hard drive X will surely hang after a few hours.

I had the system not crashing for two days without nicotine (and not much disk activity).

Other lockup using Ubuntu hoary (custom kernel 2.6.10): nicotine here never triggered a crash. But lockup randomly occurs when umounting (manually or
automatically) drives whenever USB or my SATA DVD drive.

My feeling of this is this could be a kernel bug.

Something is badly broken in recent distros...

kmare · Posted: Sat Aug 06, 2005 1:26 pm Post subject:

I believe that it is something in the system libraries (like eg. glibc). I have those lockups with my nvidia card, but I heard that ati drivers have this problem too... so we can (almost) conclude that it's not a nvidia/ati driver problem. Also I read somewhere (I think in the nvidia linux forums) that some ppl with freebsd had Xid/locks up problems too... so I've been thinking that the only thing that it's shared between those OSs/setups are the gnu subsystems like glibc, binutils or even gcc.. on the other hand I really don't understand how this bug is not triggered on every linux setup.
anyway.. nvidia will release a new driver set soon... let's see if they fixed anything (I doubt it, since I don't think it's their problem)

btw.. what bothers me is that nobody found where the problem really is, and as such nobody will fix it anytime soon...

eor · n00b Joined: 28 Apr 2004 Posts: 13

As lasker said, random lockups are often an interrupt problem.

noapic doesn't help though.

anybody looked at the scheduler in the kernel code? Has it changed recently?

Next guess would be bad ram.

And third guess would be insufficient/spikey power supply.

then heat.

For what it's worth, I was looking for a way to combine nvidia framebuffer support in the kernel, with nvidia's kernel module for X.
nvidia frame buffer support prevents the nvidia kernel module from loading.

Now you guys got me freaked out about stabilty.

Have we compared hardware? There's got to be a common point of failure. Rather than stating what it isn't, perhaps state what it is.
Provide hardware specs and installed software. Find what you all have in common.

If not video card specific, how about chipsets? via? intel? nvidia?

John-Boy · Posted: Sat Aug 06, 2005 5:15 pm Post subject:

Lasker · Guru Joined: 17 Jul 2002 Posts: 445

What I really hate to admit: I never saw those lock ups under XP (dual boot system).
So searching for hardware issues seems a waste of time, at least in my case.
Though currently I'm still 'lock up free'...

pankkake · n00b Joined: 27 Apr 2005 Posts: 24 Location: France

Bobbie · Tux's lil' helper Joined: 21 Jul 2002 Posts: 77

I reverted from PCLinuxOS to Ubuntu Hoary and as I thought no more lockup running nicotine (while in PCLOS it would freeze the system in less than 2H).

Both distro use Xorg 6.8.2. PCLOS use KDE 3.4.2 while Ubuntu uses gnome 2.10 (probably not related to the crash).
Ubutu use an older heavily patched kernel (2.6.10) while PCLOS use newer kernels (2.6.11 and 2.6.12).

So as far as I'm concerned I think it's a kernel (driver) problem maybe related to USB (disk acess) or maybe some setup problem
as both distro have a quite different init system.

RaZoR1394 · Guru Joined: 09 Jan 2005 Posts: 356

I've had hardlocks with X.org since the start but not until now I've figured out the cause. The lockups happen once in a while when surfing, playing videos and gaming. What's interesting is that these lockups never happen when firefox ain't running. And when I'm only browsing it could be sufficient with pressing the back button to reload the previous page or clicking a link to load another page. It seems that other users just like me suspect firefox.

I'm using ~amd64 with Firefox 1.0.6-r4, Xorg 6.8.2-r2 and kde 3.4.2.

spitefulcrow · n00b Joined: 18 Jul 2004 Posts: 44

I have a feeling that there's something wrong with the kernel and/or major libraries like glibc that's underlying everything else people are reporting.
I typically get crashes if I do heavy X usage. However, I think I remember having a crash or two in console mode while running a big set of emerges (a -uD world or something, most likely) and I had left the vt emerge was running on as the active console. Emerge obviously produces a lot of scrolling activity, and it seems like a common thread to a lot of the X locks being reported is high refresh rates caused by scrolling or dynamically generated Web content. So maybe, just maybe, there's an issue involving kernel drivers and constant refreshing and calls to the graphics bus. Any thoughts?
_________________
Athlon XP 2600+ (1.9gHz)/1024MB DDR333/Asus A7V880/Samsung 160GB SATA HDD/Chaintech GeForce FX5500 256MB

RaZoR1394 · Guru Joined: 09 Jan 2005 Posts: 356

So much for the firefox blame, hehe. It locked up some times without it running now when I was playing gta:sa. What I've noticed is that Cedega games lockup the comp for me very easy apart from Doom3 and Ut2004 which are running for god only knows how long, and they have never locked up on me. I also noticed that GTA:sa often locks up EXACTLY when big fonts appear for ex "Mission failed", "Los santos" etc.

Some weeks ago I also remember that the comp locked up right when starting Steam with Cedega and certain win apps with Wine. Howewer that problem disappeared shortly after. I don't know what's causing this; kernel, ati-drivers, xorg, no idea... So for me, nearly all lockups happens together with Firefox, Cedega and Wine.

I've checked my logs and there's nothing interesting in them except some warnings about IRQ and USB.
_________________
http://en.wikipedia.org/wiki/Cell_(microprocessor)

deadman_kein · n00b Joined: 20 Apr 2005 Posts: 5 Location: Poland

Well, I switched to nv driver, and it helped a little. I can watch movies now, and it's stable while I write code, surf web, etc. It did lock up today while I was gone (azureus was the only running app), and it seems that didn't change. So when I leave my comp running by itself it will lock up sooner or later.

So I think that toomorow I'll leave it in console and see what happens.

And IMHO it's not a hardware problem for sure, too many reports with wildly different configurations. At the begginnning I also thought it's an interrupt problem, or something with powersaving mode etc., but it's not.

curmudgeon · Veteran Joined: 08 Aug 2003 Posts: 1741

To add my two electrons:

I have experienced numerous COMPLETE lockups
(no keyboard, no mouse, no ssh) on one of my
machines (but not the other). This happens ALWAYS
(meaning the machine eventually WILL lock up)
with any 2.6.12 kernel, and NEVER with any 2.6.11
kernel.

I have an ATI Radeon on the motherboard, and
use only the kernel and x.org drivers (not the
ati-drivers package).

deadman_kein · n00b Joined: 20 Apr 2005 Posts: 5 Location: Poland

I left my machine running all day without X. Only the console with bittorrent client. When I came back it was 'dead'. That is, only black screen, monitor had a green light on, so it got signal from gfx card, but no other signs of activity. Unfortunately, I hadn't started sshd, so I couldn't determine if it was a complete lockup. Think I'll try that tomorrow. And I'll deffinitefy try installing some old distro on spare partition and check how it behaves.

cayenne · l33t Joined: 17 Oct 2002 Posts: 945 Location: New Orleans

kmare · Posted: Tue Aug 09, 2005 7:22 pm Post subject:

well.. it could also be the kernel itself (as someone already mentioned).. I'll switch to an older kernel and report back how things are going...

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

So many different issues; so many prognosises. Like before, I haven't noticed any FreeBSD issues with the latest copy of Xorg, so the problem has to lie with the actual OS internals that tie into the Linux kernel I've deduced.

I think that the issue actually lies with the MesaGL/OpenGL implementation as this issue only occurs with OpenGL enabled cards for some odd reason from what I've been seeing.

No issues with Apple's X11 server drivers as well, but most of that stuff is submitted to the Xorg/XF86 people by Apple I think (I'm using a mobile ATI Radeon driver I think in my iBook). It's also a FBSD userland/kernel backend with the Mach kernel interface too, so it's not exactly comparing apples to oranges.

The issue is with the Linux kernel though and OpenGL junk. I swear.

sted · n00b Joined: 24 Nov 2004 Posts: 23 Location: l-wo, poland

my computer also lockups.

my software specs are as follows:

linux-2.6.11-gentoo-r9
xorg-x11-6.8.2-r2
gcc version 3.3.5-20050130
glibc-2.3.4
nvidia driver (lockups with any version)
majority of lockups happen when i browse web with Opera (all versions).
i never got any single lockup when playing WoW (with opengl) on wine for many hours.

hardware specs:
amd athlonXP 2500+
geforce 6600GT
asus mobo with via chipset

the last time i remember no lockups was with kernel 2.4.
i think that it's a matter of some bug in the linux kernel..

have anyone had a lockup on the 2.4 kernel?

kmare · Posted: Thu Aug 11, 2005 12:45 pm Post subject:

I don't have lockups playing wow (wine with opengl) for hours too...