X.org Lockups

[frilled]

I have an old Dell Optiplex G110 that I've reinstalled Gentoo on two days ago. I emerged gnome and xorg-x11 and of course got xorg-x11-6.8.0-r1. I could start X and work for about 2 minutes and then the mouse and keyboard would lock up and I'd get colored vertical lines across my screen. The entire system was frozen. It didn't matter what application I was using. So I did an emerge unmerge xorg-X11 and emerge xorg-x11-6.7.0-r2. All seems to be well know.

What gfx card are you using? The onboard chip? I don't recall what's on a G110 - is it an Intel 8x0, too?

I hated that driver. The recent version seems to be a lot better, though. But at least it was stable.

Apart from that I had my box crash on xfree 4.3, xorg 6.7 and 6.8 with no apparent differences. And, as I can only say again, this box used to work *fine* for months. I gave it away to a friend who installed windoze on it, so I had to reinstall Gentoo.

What I want to express here is that all my "older" installations seem to work fine, but this new one is b0rked. I even re-installed it once more, with the same result.

[flyfishin]

I´m using the onboard chip which is an Intel. Here´s the output from lspci:

Intel Corp. 82810E DC-133 GMCH [Graphics Memory Controller Hub] (rev 03)


I´ve been working on the machine for over an hour now with no problems using the old xorg-x11. I´m going emerge xorg-X11-6.8.0-r1 again and see what happens.

[Nate_S]

It's not really that the agp has anything to do with gaming so much, but rather that certian agp cards combined with certian via chipsets cause much instability. I believet that readme file I refrenced said something like the chipset wasn't putting out enough {power, something? i forget} to drive the card at agp 4x. Oddly enough, I've never gotten it to freeze in an opengl game, just in firefox and a few other things.

As far as the performance hit in games, that's with ut2004, on some pretty high settings. I'd wager that the 128mb on my gfx card isn't enough to store all the textures for a map, and thus the bus is needed heavily. Note that a radeon 9000 with the fglrx drivers handeled the exact same settings with agp 4x just fine.

also, this happens with all versions I tried (4.x, 5.x, 6.x) of nvidia drivers, and with xorg 6.7, xfree 4.3 and 4.4. (6.8 was not out at the time I tried this.)


-Nate

[frilled]

It's not really that the agp has anything to do with gaming so much, but rather that certian agp cards combined with certian via chipsets cause much instability. [...]

Yes. The more I was surprised that the MX440 I plugged in today immediately and without evil sideeffects ran in AGP 8x mode. Not that I care, actually I'd rather get my Radeon-7200-based system working stable again :-/

Now I lost track of the nvidida-FAQ somehow <g>, but I still use the kernel's agpgart. But why worry as long as it works?

[frilled]

Well, it is definitely userland-related. My personal list of "bad apps" (in this box only) is now:

- Firefox (1.0PR, the worst release ever)
- rdesktop (latest stable [like 1.3.1-somewhat?])

xchat is fine now, for whatever reason. gkrellm2 has left the lobby, too.

[jedimark]

I am running vesa without lockups, but I seriously miss being able to play tuxracer..

Are you guys using any USB input devices by any chance? Keyboard or mice.. or using the linux input subsystem? (/dev/input/mice or /dev/input/event)

I've experienced this problem with my logitec cordless elite kb+mouse combo, and a small usb keyboard. What input devices are you guys using?

Compiles i've left running in an xterm seemed to keep going when my box was 'locked up' (not that the screen was updating...I could tell from my harddrive lights and then shelling in..)

If in some cases X isn't running 100% then the only thing I can think of that could cause it is an input subsystem screwup.

I've still found it responds 99% of the time when I alt-sysRq-K (SAK) my session.

Both the Nvidia and nv drivers use hardware mouse accelleration, which goes regardless of the X.org input system... thats probably why the mouse cursor still appears to work, but you can't click on anything - it's seperated a bit from the input queue. (I've had glitches where i've SAK'ed back to the framebuffer console, and still had the graphical mouse cursor cruising on the screen. (i'm not running gpm either))

UPDATE:
------------
If i start x in twm, with nvidia drivers, everything runs fine... if i change wm to kwin (kde's window manager), and go nuts with the mouse, it locks in a few seconds flat.. with just a couple of xterms and xclock running.

Some new X.org feature that kwin (and probably others) 'supports' is causing problems with certain accelerated drivers. It's either the X feature is buggy, it's not being initialized, etc properly by the WM, or both are screwed up...

UPDATE2:
-------------
I compared the logs between starting with nvidia drivers and fbdev... One thing I noticed was this in the nvidia side of things...
(II) XINPUT: Adding extended input device "NVIDIA Event Handler" (type: Other)

Maybe a bug in the way external event handlers are called?

[frilled]

Hm, I have USB mice on most of the machines, and some Logitech, too, but on the doomed box it's a Micro$oft mouse. PS/2 keybords everywhere except for one (stable) box with an additional USB keyboard.

I had some troubles with a USB mouse (suddenly disconnected and was no longer visible under /proc/bus/usb), but that was some weeks ago and strangely enough solved by plugging it into another port.

I have been thinking in the same direction as you, since the underlying OS keeps running (well, sort of since X hogs all the CPU cycles...), but then again, it should happen with all applications (or the WM itself as you suggest), but I definitely have "evil apps" without which the system keeps running however hard I rock the mouse.

[flyfishin]

Okay. I reemerged xorg-x11-6.8.0-r1. I opened up Mozilla and clicked on the scrollbar to move down the page and the system locked and the screen created a lovely pattern of blue,red, and green vertical lines. I rebooted and decided to try some other applications. I was able to move the scrollbar up and down in some of the preference windows for Gnome. I was then trying to set the preferences for my mouse when things went south again. So for me it doesn't appear to be related to only Mozilla. I'm going back to xorg-x11-6.7.0-r2. That one worked for me with no issues.

[frilled]

I'm going back to xorg-x11-6.7.0-r2. That one worked for me with no issues.

That's good for you - made no difference for me

[flyfishin]

Here is another thread with some more potential help. Looks like the i810 driver I use had issues with certain settings in 6.8.0. It looks like the issue might also apply to some other cards as well.

http://forums.gentoo.org/viewtopic.php?t=242785

[frilled]

Here is another thread with some more potential help. Looks like the i810 driver I use had issues with certain settings in 6.8.0. It looks like the issue might also apply to some other cards as well.

http://forums.gentoo.org/viewtopic.php?t=242785

Thanks, but (un)fortunately I'm not using i810 And I'm always on depth 24. BTW: I had issues with i810+24 bpp, which went away in 16 bpp

[jludwig]

Anyone check to see if this is a 4K versus 8K stack issue with the 2.6 kernels (its some config option)? I've seen this param botch some things up, just a shot in the dark... Is everyone having this problem with 2.6 kernels?

Jeff

[frilled]

Anyone check to see if this is a 4K versus 8K stack issue with the 2.6 kernels (its some config option)? I've seen this param botch some things up, just a shot in the dark... Is everyone having this problem with 2.6 kernels?

Jeff

Actually I started compiling an 8k-Stack kernel just 10 minutes ago since I remembered the latest stable nvidia binary driver will not work with 4k stacks. Since this machine is a little slow I'll let you know if it gets better.

I don't really have high hopes, though, since I copied the kernel config from a machine that works without problems - but then again, it's using the ATI closed source drivers. You never know. I'll post the results.

[frilled]

Just booted up the 8k stack version. Now I'm going to open rdesktop and Firefox with flurry (http://forums.gentoo.org/viewtopic.php?t=243753) and scroll like mad...

[frilled]

Man...

I am really afraid of jumping to conclusions here, but ...

... I can no longer bring the box down. So I'm either lucky and got a "good day" or you just hit the nail on the head, jludwig!

Darn. Unbelievable.

Now would someone please tell me why exactly this one box out of 8 gentoos chokes on this??

Unbelievable.

I scrolled like mad. I did everything that's evil on this box, and it still purrs like a kitten.

Unbelievable.

Even if it crashes RIGHT NOW you still made my day, guy

New let's face the facts. I copied a kernel config over from another machine. The only thing I changed was:

CPU type: from Athlon XP to P3

The rest remained the same.

This box uses the X "radeon" driver while the box I got the kernel config from uses ATI binaries.

Then it crashed. Now I changed back from 4k to 8k stacks and it works (prove me wrong bitch, and you'll have to endure windoze again ).

Hell.

[Nate_S]

That's good that works for you. In my case 4k versus 8k made no difference.

Also, BTW, nvidia drivers since 6101 have been ok with 4k stacks.

-Nate

[frilled]

Yes, but I'm talking about xorg|free's radeon driver in this specific case. I do have an nvidia machine, but that one is rock solid with 8k stacks, too (since I use the latest driver marked as stable [5336] I needed to go back to 8k).

[flyfishin]

1. My 2.6 kernel was already compiled for 8 stacks so that wasn't an issue with me.
2. I cannot get 6.8.0 to work no matter how much I tinker with xorg.conf
3. I really, really wish I had read about quickpkg 3 days ago.
4. I'm back on 6.7.0-r2

[frilled]

1. My 2.6 kernel was already compiled for 8 stacks so that wasn't an issue with me.
2. I cannot get 6.8.0 to work no matter how much I tinker with xorg.conf

That's with the i810 driver, still? Did you try to go without DRI and GLX?

3. I really, really wish I had read about quickpkg 3 days ago.

Yes, I know what you mean ... but then again, always discovering new stuff, no matter how long you've been in the business...

4. I'm back on 6.7.0-r2

Does it help? I'm so happy now with 6.8 (not that I use any of its features)...

[flyfishin]

That's with the i810 driver, still? Did you try to go without DRI and GLX?

i810 driver with glx and dri disabled. Works perfectly now, won't work under 6.8.0.

Luckily this is just a play machine that I try out different distros on. So for me running 6.7.0 really doesn't have a major impact. Well, the major impact is that it works for me.

[jedimark]

Are you guys building with distcc by any chance? I'm shooting arrows into the air here.. but the possibility of a build fault causing this is probably worth investigating.

I've been running distcc 2.16 & 2.18, but that's only on for the end part of the build, and I have to switch it off for building most of KDE. X turns it off for most of it.

I find it slightly annoying that I can pick any one of my athlon / athlon-xp boxes with nvidia cards (3 k7s5a's (sis), 1 k7som+(sis) and 1 epox 8kmm+ (via) install gentoo starting at stage one from scratch with "-O3 -march=athlon-xp -fomit-frame-pointer -pipe", download the drivers from nvidia's site, and have it lock.

I'm slowly going nuts.. But thankfully being a coder I don't need no stinkin 3d to make me happy.

[frilled]

Are you guys building with distcc by any chance? I'm shooting arrows into the air here.. but the possibility of a build fault causing this is probably worth investigating.

I usually used distcc because this machine isn't too powerful, but when I rebuilt it this time, I didn't.

I actually had little problems with distcc during the last year. Some builds not using it, some builds breaking with error messages, but that's about it.

I find it slightly annoying that I can pick any one of my athlon / athlon-xp boxes with nvidia cards (3 k7s5a's (sis), 1 k7som+(sis) and 1 epox 8kmm+ (via) install gentoo starting at stage one from scratch with "-O3 -march=athlon-xp -fomit-frame-pointer -pipe", download the drivers from nvidia's site, and have it lock.

What I actually find annoying is having to use closed source drivers at all. But sorry, can't comment on this combination since the only nvidia card I use is plugged in a Dell and runs fine.

I'm slowly going nuts.. But thankfully being a coder I don't need no stinkin 3d to make me happy.

Well, even if I don't *need* it a little eyecandy doesn't hurt from time to time...

[Nate_S]

Update: I popped the radeon card back in again a few days ago, and realized I never did change it back to AGP 4x, but it's still perfectly fast (plays ut2004 just fine,) so my guess was that the slowness was unrealted.

If you have a via chipset (and I think it's not just limited to nvidia cards, nor any specific driver) try turning your agp settings down as far as you can (even off,) and see if the problem persists.

-Nate

[StifflerStealth]

I did a fresh install of Gentoo and I started to get those crashes everytime that I tried to compile something in Konsole (I use KDE 3.3.1). I recompiled my kernel after I took out IO_ACPI and I changed a few settings related to Hard Drive controllers. Then I edited the xorg.conf where I commented out the lines "Load dri" and "Load freetype" and saved that. I shutdown the computer and then booted into the new kernel and new xorg settings, and I have not had one single hard lock up yet. I am using the Nvidia driver and the Nvidia OpenGL interface. I remember reading somewhere on this forum that there have been issues with freetype, so that is why I commented it out.

I hope this helps.

[jedimark]

crap.. StifflerStealth you may just have solved it! (for me at least anyway

my box normally crashes within a minute or two of using it... or sooner if i dare to move my mouse in away that doesn't please the forces of X.

I just disabled freetype then, and renabled my nvidia drivers... logged in i've gone nuts with the mouse and still no lockups...

I have all the ACPI and APIC stuff enabled though...

Could it have been freetype all along???

I bet all those problem apps use freetype... I know firefox does..

And my fonts dont look like crap without it either...