X.org Lockups

Message

frilled · Post by **frilled** » Mon Nov 29, 2004 5:30 am

headgap wrote:i'm going to go out on a limb here, and suggest, that there's a config do-hickey in nvidia code that's not getting set right for SMP kernels: if you insist on running the driver, the cpu goes off into space on a bad pointer.

what's not clear is why some folks have no problems at all. perhaps they have a combination of SMP with an actual 2nd processor ? or AMD ? or something else?

Well, I do have a machine running an 2.6.8/2.6.9 SMP kernel on a single Hyperthreading CPU with the nvidia driver, and no problems at all. The chipset is an Intel 865.

juniper · Post by **juniper** » Mon Nov 29, 2004 6:07 am

i have been having a pretty serious xorg lockup. I recently upgraded to xorg-6.8.0 and am running a 2.6.7 kernel. also, i have a dreaded ati radeon 9000 card. AFAIK, i did not have this problem in xorg-6.7.0. I am using the native xorg radeon drivers and i DO have opengl working (on ati 9000, the native xorg drivers give 3D).

my problem is with xorg. I posted this problem in

http://forums.gentoo.org/viewtopic.php? ... 647e74e1dc

and i got no reply. anyways, to summarize, xorg lockups hard (i mean really hard) when i play tuxracer. it is sort of random (it does it sometimes) but it only does it during tuxracer. anyways, the lock is so hard that i cannot there is no keyboard response or anything. i am able to login from another machine and top tells me that X is taking 99% of my cpu, also, i am not even able to kill X (i send a kill -9 signal and it seems to try and kill it but it does not work). anyways, the only way out of this is to reboot, and if i do reboot, it takes about 2 mins to shutdown.

so, my questions are

1) anyone else having this problem? solutions?
2) when X does lock up, can i send it a stronger signal than -9? is there some "kill --slaughter X" option or anything? i hate to have to reboot.

thanks.

PS. i did look in the log files, but they didn't tell me anything. however, i don't really know where i should look. i looked in /var/log/Xorg.log.0.old and ~/.xsession-errors and they didn't tell me anything useful. Is there somewhere else i can look?

thanks

j

j

cayenne · Post by **cayenne** » Mon Nov 29, 2004 4:28 pm

headgap wrote:oh, and...

i was looking back through archived boot partitions, and found the
last known-good nvidia driver install on a 2.4.20 kernel with nvidia
1.0-4191. probably not worth much for debugging, sorry.

But....this is not just an NVIDIA problem....Xorg is flaking out on Radeon's....and not just x86...also blows on ppc.

I'm guessing that's why it seems so hard to track down, not easy to see a pattern to why it blows, it isn't consistant with settings, kernel versions or hardware.

C

Nate_S · Post by **Nate_S** » Mon Nov 29, 2004 8:15 pm

The only thing I know of "stronger" than a kill -9 is the magic sysreq key, killing all processes. However, if you search, I've seen a couple threads in the past about killing something that won't die with kill -9

-Nate

juniper · Post by **juniper** » Mon Nov 29, 2004 11:48 pm

[Quote]
The only thing I know of "stronger" than a kill -9 is the magic sysreq key, killing all processes. However, if you search, I've seen a couple threads in the past about killing something that won't die with kill -9
[/Quote]

how do you use 'sysreq key'? also, if it kills all processes, won't it kill init? won't that just shut everything down?

j

Dolio · Post by **Dolio** » Tue Nov 30, 2004 1:21 am

To use sysrq, you need to compile your kernel with the magic sysrq key enabled (it's in the kernel hacking section, I think).

There's a doc file in the kernel tree that describes all the things you can do with the key. Basically, you press alt-sysrq-?, where ? is some letter. For a relatively save shutdown, you do:

Code: Select all

alt-sysrq-s: sync discs
alt-sysrq-u: remount read-only
alt-sysrq-b: reboot

When X freezes, I usually do

Code: Select all

alt-sysrq-k

That should kill all programs on the current virtual console, which will let you alt-F? to a different virtual console and log into a console (lately when I've done it, the display stays garbled after X dies, so I actually need to start a new X session and exit before I can see anything on a regular console).

Curiously, this doesn't seem to really kill X, probably since it wasn't actually started from vc7. However, it does disrupt its hold on the virtual console, so I can log into a shell and kill the X process, which fixes everything.

mmontg1 · Post by **mmontg1** » Wed Dec 01, 2004 6:58 am

, which fixes everything.

LOL, GOD I wish...

cayenne · Post by **cayenne** » Fri Dec 03, 2004 8:30 pm

Bump...

Any updates? My system, and I assume other's..is pretty much useless with this problem. Is there a bug registered with this problem?

I wanted to use my laptop that is affected by this...to do some business demos with, but, not like this...

C

frilled · Post by **frilled** » Fri Dec 03, 2004 10:29 pm

Strangely enough, mine has been stable since xorg 6.8.0-r3 and kernel 2.6.9-r4 came out (iirc, currently not in front of the box). This does not mean the problem is really gone (I already had a week without problems and then two lockups during the course of an hour), but up to now it looks pretty good. Don't know when those came out, like two weeks ago? I'm almost happy. Almost, because I fear the problem will resurface :-/

sog · Post by **sog** » Sat Dec 04, 2004 3:44 am

i've been having the same problem of late. X locks up, but my mouse goes with it too. no keyboard + mouse = trouble. haven't tried to SSH into it yet, but i expect that i'd see the same thing other folks have - X chewing up my processor.

didn't have the problem prior to x.org 6.8, but i also have upgraded firefox, installed APM, and a host of other changes so that's not necessarily it. occurence is typically on mouse movement in firefox, but that is what i'm doing most.

it's not NVIDIA, b/c i'm on the ATI generic drivers.

would love to get this fixed.

frilled · Post by **frilled** » Sat Dec 04, 2004 8:23 am

sog wrote:it's not NVIDIA, b/c i'm on the ATI generic drivers.

It's neither xorg [6.8] since I had trouble with 6.7 and even XFree. (ATI open source drivers here, BTW)

BlinkEye · Post by **BlinkEye** » Sat Dec 04, 2004 7:22 pm

which kernel-headers are you guys using? at the moment i'm pretty sure it is the linux26-headers. i just emerged xfree and had a freeze too so at the moment my laptop is compiling xorg again (btw, xorg-6.8.0-r4 freezes too). having done that i'll try emerging good old linux-headers (2.4) and hope the problem will be gone

ekinskofer · Post by **ekinskofer** » Sat Dec 04, 2004 7:37 pm

hey guys..

I'm still having this problem as well.. it's killing me. I'm sorta new to linux and emerged xorg, etc. I'm wondering.. does xfree have the same problems? I haven't really seen anything like this or heard any of my pals that have had problems with it. Also is there any documentation on emerging/installing any of the other x managers?

sog · Post by **sog** » Sat Dec 04, 2004 8:57 pm

i am on the 26 headers. i don't think i can go back, however, as i needed them for some package or another. be interested to know if that's the problem though.

frilled · Post by **frilled** » Sat Dec 04, 2004 9:37 pm

Yeah, 2.6. Just got a new version today. But I switched all other boxes to 2.6, too (just to use NPTL). In IIRC I already had the problems with 2.4 headers (I rebuilt the box after I couldn't stand it anymore)... Might be worth investigating further, though!

BlinkEye · Post by **BlinkEye** » Sat Dec 04, 2004 9:46 pm

i'd investigate further but i found my bug: iptables. i droped every input along with the loopback interface and somehow after some time it locks kde up (but in fact its just slowing it down immensly because connection to the loopback interface can't be done anymore). so, for me the problem seems to be solved. i hope i haven't confused anyone

wel · Post by **wel** » Sun Dec 05, 2004 6:20 pm

Well, i've solved my problem. I just disabled Nvidia's driver AGP support and all works fine. At least in the past week i haven't had any lock-ups.

At the moment, this is the only thing that works for me. 3D performance is affected, obviously, but what the hell... i need a stable work environment, not a bleeding edge gaming machine.

frilled · Post by **frilled** » Sun Dec 05, 2004 9:10 pm

wel wrote:3D performance is affected, obviously, but what the hell...

Right, what the hell ... I never understood why GLX refuses to load without AGP support. As far as I can understand there is no logical reason. Whatever. Glad to hear your problem is solved.

kip · Post by **kip** » Sun Dec 05, 2004 11:24 pm

i went back to 2.6.4 today whick improved the system-performance a lot. Firefox still produces high cpu loads, but the x-freezes a gone so far ...

wel · Post by **wel** » Mon Dec 06, 2004 11:28 am

I forgot to say that there's a thread in the nvnews linux forum where some other people discuss about the same problem.

SnEptUne · Post by **SnEptUne** » Thu Dec 09, 2004 12:46 am

I have read thru this whole thread. I have the same problem. XOrg always locks up when I run glxgears with my mouse moving. It run fines if I don't move my mouse; however, if I move it in circle for around 5 second during the benchmark, my system would lock up completely. My mouse halted, keyboard doesn't response, and even Alt-Sys Rq-K would not kill X. I am using Xorg 6.8 with hardened-dev-gentoo 2.6.7-r16 using Radeon driver from the kernel (I have a Radeon 7000 VE). I tried to comment out freetype, drm, dri, and other options beside glx (since I need it for games); nevertheless, it still locks up if I move my mouse during glxgears. Is there anything else I could have done? From this thread, I don't believe recompling my entire system would help. Beside, I need my system to do serious works.

Eskarel · Post by **Eskarel** » Thu Dec 09, 2004 2:48 am

Well I just finished an emerge -e world and while I didn't have X up the whole time, it's the first time I've managed to keep my PC up for that long(almost 48 hours) in a long while. I will see how things go running X full time again and let you know, though my problem seems more that my PC goes to sleep and won't wake up depsite turning off both apm and acpi in my grub.conf

frilled · Post by **frilled** » Thu Dec 09, 2004 3:21 pm

Well, I thought I'd heard the last of it, but after ~3 stable weeks, it suddenly barfed again yesterday while switching virtual desktops (XFCE4) with the mouse wheel. How I hate this

frilled · Post by **frilled** » Thu Dec 09, 2004 3:25 pm

SnEptUne wrote:From this thread, I don't believe recompling my entire system would help. Beside, I need my system to do serious works.

You're probably right. I actually rebuilt my messed up box, with no difference whatsoever...

ktk · Post by **ktk** » Thu Dec 09, 2004 4:42 pm

since monday I'm struggling with this problem as well, I've discovered this thread quite early so I tested about everything that was listed in here including:

- Option "NoMTRR" in xorg.conf
- disabling AGP in xorg.conf (can't disable it in the BIOS, no such option)
- switching from xorg 6.8 to 6.7 (and back cause it didn't help)
- changing from DVI input to Analog (I have a Dell 2001FP Display which seems to have problems with nvidia drivers cause they try to (write)access some stuff in the screen firmware which is read only)
- changing to nv driver
- changing to Scitech SNAP driver (commercial)
- recompiling all nvidia driver stuff & the kernel
- Gnome & OpenBox WM, both lock

BTW first Xorg worked perfectly well but after some weeks I got another strange problem, the TT font rendering was completely broken, see my thread here: http://forums.gentoo.org/viewtopic.php?t=260102

Starting with that I did disable RenderAccel which actually did help resolving that problem. BTW I do get the lockup as well if I reenable it.

so, I'm really pissed by X once again. Like most in this thread I guess

Will check the bug reportings soon, I really wonder why that happens.

Anyone knows how I can completely disable truetype? If I simply uncomment the TrueType option in the config my box stops as soon as I try to enter the username in GDM

BTW my gnome locks almost immediately after I did a login. Sometimes it can't even load all the gnome stuff correctly before it locks up.

In the beginning of the problem earlier this week it took at least a few minutes to lock.

X.org Lockups

Re: nVidia woes (a different take)

insert humor here...

Bump..

Damn!

same game for me...