X.org Lockups (part 2)

x86usr · n00b Joined: 06 Sep 2005 Posts: 2

I think the freezes with ati-drivers or nv-drivers are caused by different bugs, which just produce similar things. Nothing is tweaked in my system and nv never crashed, nvidia at least one time per day. The most answers in this thread (except other hardware than nvidia) match my experiences on several machines. I'm convinced it must be a bug in the nvidia 32bit drivers (probably made and tested in a 64bit environment).

kmare · Posted: Wed Sep 07, 2005 10:37 pm Post subject:

my system is still up... no freezes since I switched to 2.6.13 with preemption... I changed nothing else.
Also I tested everything from 3d games like doom3, world of warcraft, glest, whatever for many hours.. I played many dvd movies and videos, sometimes I also enabled shadows under composite... and the system is still up

Gergan Penkov · Posted: Fri Sep 09, 2005 2:10 am Post subject:

kmare, are you using the renderaccel-option?
And I don't think that this has sth to do with the drivers, because I have tested it exhaustively at the beginning (3-4 months ago) :

) if I do not start anything mozilla- and kde/qt-based-applications (no evolution and so on, anythink which uses and links against mozilla), anything others than that would not lock the server (I have had uptimes of two weeks) - opera, playing castle wolfenstein and so on.
And now with the nvidia-driver, but without renderaccel I could see the mozilla mini-freezes sometimes, but somehow it passes through and the system could recover itself. It is extremely easy to see this on gnome start-up I use galeon and have 5-6 windows with probably 40-tabs, if I use renderaccel this locks the system. kde does not start and locks the system with renderaccel, although runs fine without it. Probably it has sth to do with the network, who knows (that was the first thing of which I thought at that time, because of the inclusion of the ipv6 use flag in 2005.0).
But I hope that someone will be lucky enough and will find the source of this nasty bug ::

))
_________________
"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

kmare · Posted: Fri Sep 09, 2005 9:46 am Post subject:

yup.. I'm using renderaccel. It's too slow without it. Also I only use kde as my main DE and with kernel 2.6.13 my system is up for more than a week.. that's a first

maybe I'm just lucky for now... But I agree with you..it's not a problem with the driver itself.. maybe the driver triggers something...if only we could say what it is exactly...

PS: I'm not even getting Xid errors like before while playing doom3..

kmare · Posted: Fri Sep 09, 2005 5:53 pm Post subject:

great... after more than a week of uptime I got a freeze while playing glest.. I officially give up!

beppu · n00b Joined: 17 Mar 2003 Posts: 9 Location: SoCal

JohnY · Guru Joined: 26 Jul 2002 Posts: 305

joryndo · n00b Joined: 12 Sep 2005 Posts: 4

Hi. Don't know if any of this could be related...

This morning my system totally locked, and kept locking whenever I tried to run any X application. I've rebuilt the kernel without ACPI and the CPU frequency scaling. Now the system doesn't lock, but certain systems keep dying. For a while, if I tried to play any audio, my USB mouse would stop. I swapped it for a PS2 mouse, and that no longer happens.

Just now I reconnected it to the network, figuring now its stable enough at least to use. Except as soon as I try to play any music, the sound system dies, and all network traffic stops.

update: i disabled APIC in the BIOS, and ALSA has stopped killing itself and taking the network with it. So I'm pretty much back to normal. Just got to hope it stays this way.

Meypqr · n00b Joined: 03 Apr 2005 Posts: 5

In on of the first posts in the old Thread there's this "badness in pci_find_subsys" thing. I'm getting this error right before X goes crazy.

Am I the "only" one, or is this common?

Note from the kernel-source "/drivers/pci/search.c", for "pci_find_subsys":
* NOTE: Do not use this function anymore, use pci_get_subsys() instead, as
* the pci device returned by this function can disappear at any moment in
* time.

Volt3r · Posted: Thu Sep 15, 2005 6:01 am Post subject:

I have the same problem. Then i saw this thread i`m totally scared :/ I can only post here that the same lockups with mouse movement i had in both: 32bit gentoo with ati card (ati-drivers) and amd64 full 64bit system with nvidia-kernel driver :/ and of course i cant find out whats the problem...

tdm · Posted: Thu Sep 15, 2005 6:00 pm Post subject:

I think it solved!

I just upgraded to Xorg v.6.8.2-r3 and I'm running in X now for about an hour or so with out lockup.
I'll keep testing and keep my status up-to-date.

(Finally I can use the desktop normally again

)

Edit:
Some extra info, I'm running:
AMD XP
Kernel 2.6.12-gentoo-r10
with an NVidia GF4 ti4600

spitefulcrow · n00b Joined: 18 Jul 2004 Posts: 44

Right now I'm at school, but from re-reading some bug reports and earlier posts here, it seems like this may be an issue created by an update some number of kernel versions ago - everyone seems to be using recent kernels.
When I get home, I'm gonna go back and try it with some ridiculously early version like 2.6.3 without Gentoo patches. I seem to remember that when I first set up my system (back when the kernel version was somewhere around 2.6.5) it didn't have this issue until I updated to 2.6.8 or so. Worth a try, especially because I saw five crashes in the space of 30 minutes last night.
Does anyone have thoughts as to what specific part of the kernel tree seems to be causing this so we can start looking through changelogs?
_________________
Athlon XP 2600+ (1.9gHz)/1024MB DDR333/Asus A7V880/Samsung 160GB SATA HDD/Chaintech GeForce FX5500 256MB

Revoluzzer · n00b Joined: 17 May 2005 Posts: 12

Hm.

One Question: Why does everybody ignore Meypqr's post about the "badness in pci_find_subsys" thing?

Im my opinion it's a hint which should be followed. Other post's are mainly "me too's", which don't help to find the bug.
As spitefulcrow noted, it may have something to do with the kernel version. And the "badness in pci_find_subsys" problem is kernel-related, too. :!:

tdm · Posted: Fri Sep 16, 2005 3:04 pm Post subject:

hielvc · Posted: Fri Sep 16, 2005 3:43 pm Post subject:

Scince moving to the 2.6.13 kernel my frezees are gone, but I dont play games. One box is a sempron 2600 SIS chipset with geforce 4 useng fbdev in xorg, and fb tng in kernel. My other box is a VIA C3 800M with SIS chipset SIS 513PRO on board viedo with SIS support in the kernel & in xorg, no frezees so far. Aint it great that everyone has the same results, :wink:

Im going to try NV suport in the Geforce box latter to today but ifn I remmeber right that caused frezzeing no matter how I setup xorg.
_________________
An A-Z Index of the Linux BASH command line

tdm · Posted: Fri Sep 16, 2005 8:28 pm Post subject:

After running about a day now with the new Xorg, the lockup stiked again :'(

I've used a hell lot of applications the last day, and just before the lockdown I started vi in the console and BANG,
total lockup.

I've checked my log, and this was the last line before the lockup.

spitefulcrow · n00b Joined: 18 Jul 2004 Posts: 44

Well. I got home today and built a 2.6.5 kernel to start checking for where this bug was introduced. And of course, init dies with "Kernel too old!"
It doesn't take too long for me to build a kernel, so I'll see if it'll boot with anything prior to 2.6.10 or so, which is where I seem to remember first seeing this whole issue.
_________________
Athlon XP 2600+ (1.9gHz)/1024MB DDR333/Asus A7V880/Samsung 160GB SATA HDD/Chaintech GeForce FX5500 256MB

kmare · Posted: Fri Sep 16, 2005 10:22 pm Post subject:

Meypqr · n00b Joined: 03 Apr 2005 Posts: 5

I'm running kernel v2.6.5 and even with something like 2.6.2 those lockups happened. I've got another machine with v2.4.? (cpu is broken, so i can't check) and there are no lockups.

spitefulcrow · n00b Joined: 18 Jul 2004 Posts: 44

Well, scratch that plan. I just got a crash with 2.6.9 unpatched.
And if Myepqr is reporting that this exists even with 2.6.2, we can either assume that it's a change introduced with 2.6 in general that has been backported to 2.4 series (I saw reports of these crashes on late-version 2.4 machines somewhere) or that the bug lies elsewhere, either in X or a major library like glibc.
We CAN rule out the possibility of a specific graphics driver causing the issue because I've seen it on my machine with both nvidia and the nv driver without AGP support even loaded into the kernel, and people are seeing this with ati drivers (proprietary and open-source) too.
My system seems to stay up longer if I don't use applications that cause constant redraws of windows - I got to 4 days uptime last week by not running gkrellm2, but then I started it again and it crashed within a day. Obviously, the monitors of CPU activity and network/disk I/O cause redraws on the order of every half second, if not more frequently. Since people report crashes under heavy scrolling in Firefox, and I've seen crashes when I start running emerges in an Eterm, I think that it's safe to say that lots of screen refreshing puts some sort of strain on the X server that it can't deal with due to this bug, triggering the crash.
It also appears that, since people see two different types of crashes (ones that you can recover with SSH or magic SysRq, and ones that deadlock everything, requiring a hard reset), hardware and kernel configuration can exacerbate this problem.
If anyone is still watching this thread, how about we get out of the habit of just posting confirmations and actually do some thinking on this so we can track it down and get it FIXED. I love Linux and Gentoo, but I'm getting tired of living with daily system crashes, and it'd be nice to get this fixed so I don't have to go learn a BSD or something.
_________________
Athlon XP 2600+ (1.9gHz)/1024MB DDR333/Asus A7V880/Samsung 160GB SATA HDD/Chaintech GeForce FX5500 256MB

tdm · Posted: Sun Sep 18, 2005 10:06 pm Post subject:

I do not have had a crash since I upgraded Xorg,
but I'm willing to do some tests on my machine.
But I do need to get some commands of what to do, because I do use linux but I'm not more than a user trying to learn and use linux for beginners cross-platform C++ programming.

I know how to configure and compile a kernel, and I have some knowledge of how the OS works inside..
but it all stays at the level of "some" knowledge.

So, you're call is my command (or something like that).

ps. It justs takes some time for answering, because it's 00:04 hour now here at this side of 'the Tauri' aka earth :lol:

kmare · Posted: Mon Sep 19, 2005 1:02 am Post subject:

what version did you upgrade xorg to? and try different 3d apps and different web browsers...

PS: also try playing glest for more than an hour...

tdm · Posted: Mon Sep 19, 2005 6:12 am Post subject:

I upgraded to Xorg v.6.8.2-r3

I'll do some tests with OpenGL/3D apps/games and webbrowsers today

paulbarker · Posted: Mon Sep 19, 2005 7:39 pm Post subject:

Ok, I'm jumping back in after a long absence. A summary of what I know / have experienced:

- Type 1 crashes, restartable using sysrq's.
- My hunch is this is a mobo issue. My mobo is known to be dodgy (at least 1 dead PCI slot).
- Not tried Gentoo on the affected computer in about a year, i think.

My idea is to build the latest version of xorg with debugging support, setup ssh properly and try to connect a debugger to the broken X process via ssh (from my laptop). Then I'm going to (try to) find where in the code the problem actually is, i expect a tight loop over some kernel call, possibly related to the PCI/AGP busses.

I'm also having a slightly less-nasty version of this bug in Mandriva, on the same box, the only difference is the frequency of crashes (Mandriva crashes ~1x a day, Gentoo within seconds of logging in).

I'm going to do a fresh Gentoo install, stage 3 and a GRP version of X first, just to see if I can sort it without doing the above. I'm expecting not, I think I bashed all the config options to death trying to fix this one. I'll post results as soon as I get them, I expect it to take about a week because of other commitments.

Wish me luck,
Paul Barker.

tdm · Posted: Mon Sep 19, 2005 7:54 pm Post subject:

I've been running some 3D apps/game like Blender, SuperTux, Enemy Territory and such some hours today but no lockups at all.

A few minutes ago I started unmerging the Mozilla Suite and DANG, total lockup
(unmerging was withing a Gnome Terminal).

Via SSH I found the same error as I had the other times X locked up: