Mysterious Freeze

Message

caslca · Post by **caslca** » Sun Feb 20, 2005 1:28 pm

yaneurabeya wrote:
Get rid of your power management in the kernel altogether, or look into possible IRQ mapping issues and how to fix them (there are patches for some systems, ie Dells out there...).
Also, be sure to submit a bug report to the kernel.org people.

Sorry, I don't mean to sound rude, but this hardly seems like a solution at all. Maybe I'll just not turn on my laptop at all then.
Disabling power management means not being able to use any battery/AC management, Hyperthreading, frequency scaling, hotplugging, sleeping, hybernating, etc. What the hell is the point of having a laptop then !?!?
Now, I've made my peace with not being able to suspend/sleep the computer (due to nvidia issues) but that's where I draw the line.
I'm sure there must be a better solution short of writing my own kernel. I bet the developers will fix that issue.

As far as IRQ conflicts, I agree - that seems to be the issue at hand here. Unfortunately, I have no idea what is conflicting with what, and this being a laptop system, I have no way of unplugging hardware to see what causes problems.

As far as bug reports, I have nothing to send really, since there are no abnormal entries in any of the logs.

For now sadly, I've decided to think of these random freezes as Gentoo's BSOD(black screen of death). I can live with that for a bit, but I surely hope it gets fixed.

davecs · Post by **davecs** » Sun Feb 20, 2005 5:31 pm

My main frustration here is not that it happens. I mean compared to the regularity of problems you get under Windoze, my occasional lockups, with a built-in key sequence to close files cleanly before rebooting, is relatively minor.

But of course it shouldn't happen at all, and no-one seems to know why it does happen! And the really annoying thing is that none of the people who are involved in the programming side of things seem to know, accept it could be their part of the code, or even seem to have an idea.

Possibly the bug is very well hidden, so maybe we need some kind of escape route which leaves the Window Manager and programs running, maybe autodetecting the malfunction by some kind of system call once every second and forcing an unfreeze. Could such a thing be done?

yaneurabeya · Post by **yaneurabeya** » Mon Feb 21, 2005 3:18 am

As mentioned before, there are several different issues at hand with the lockups. People have noted the lockups occurring with 2.4 kernels and other with 2.6 kernels. It occurs on both XFree and Xorg. It can also be mistaken as an X issue when it is in fact a kernel issue (my case). Most of the time the problem seems to root from CPU usage and possibly OpenGL based programs in XFree/Xorg.

ultraViolet · Post by **ultraViolet** » Mon Feb 21, 2005 11:54 am

I was having the same problem with my computer. Since I have disabled Fast Writes in my bios all is working fine ... Hope it can help ...

raf · Post by **raf** » Wed Feb 23, 2005 6:59 am

Hi,

Does everyone here use NPTL? I have recently re-compiled my glibc without nptl and so far no crashes. Can anyone verify this? Just add -nptl in your make.conf and re-emerge glibc. I haven't been able to crash my system since.

Raf

massimo · Post by **massimo** » Wed Feb 23, 2005 8:23 am

I tried it with and without NPTL, both didn't work for me. I want to emphasize something I already posted here: the lockups occur when I use my MSI Geforce4 Ti 4400 VIVO, they do not happen when I use my Gainward Geforce4 Ti 4200.

skorec · Post by **skorec** » Wed Feb 23, 2005 8:00 pm

I have had problems with random segfaults, which were mainly happening when both hard disk and CPU load was high.
Using nv drivers instead of nvidia drivers only lowered the probability of these segfaults. I have both tested the RAM and changed for another one, but it didn't solve the problem, either. Disabling fastwrite wasn't of any help.
Finally, underclocking the RAM to 333MHz has helped.
I suppose the problem occurs at 266 FSB and 400MHz RAM when the traffic CPU/RAM, and the HDD/RAM or GPU/RAM is high.

vola · Post by **vola** » Thu Feb 24, 2005 12:55 am

raf wrote:Hi,

Does everyone here use NPTL? I have recently re-compiled my glibc without nptl and so far no crashes. Can anyone verify this? Just add -nptl in your make.conf and re-emerge glibc. I haven't been able to crash my system since.

Raf

No.
I had (better have) the freezes with and without nptl.

Regards

joehni · Post by **joehni** » Thu Feb 24, 2005 6:39 pm

ikaro wrote:
A A A A RRRR R RR GGGGGGG !!!!!!!!

same here, again and again

davecs · Post by **davecs** » Thu Feb 24, 2005 8:09 pm

Tuttle wrote:Hi !

I've follow this thread some weeks ago after have same symptoms and so many crashes.
I have no more crash actually but cannot be sure why because i installed so many different
kernel since. But i think that have stopped since i've tweak my /etc/X11/xorg.conf.
So there is the relevant part of my xorg.conf.

I actually have a Geforce 4 5700. gcc 3.4. xorg 6.8.2 but it's really recent and last nvidia driver (unmasked ones)
I remember when i play with transset and xcompmgr, that i could not launch an open GL API a long time ago.
So i think this problem is something between xorg-x11 with Composite extension and GLX API...
Code: Select all
Section "Module"
	Load  "record"
	Load  "extmod"
	Load  "dbe"
	Load  "dri"
	Load  "glx"
	Load  "xtrap"
	Load  "type1"
	Load  "freetype"
EndSection

Section "Extensions"
	Option  "RENDER"			"Enable"
EndSection

Section "Device"
        ### Available Driver options are:-
        ### Values: <i>: integer, <f>: float, <bool>: "True"/"False",
        ### <string>: "String", <freq>: "<f> Hz/kHz/MHz"
        ### [arg]: arg optional
	Option	"AllowGLXWithComposite"		"true"
        Option     "HWcursor"           	"True"
        Option     "NvAGP"              "2"
        Option     "RenderAccel"        	"True"
        Option     "CursorShadow"       	"True"
        Option     "CursorShadowXOffset" 	"3"
        Option     "CursorShadowYOffset" 	"3"
        Option     "NoBandWidthTest"    	"True"
	Option     "DPMS"
	Identifier  "FX5700" #or whatever yours is
	Driver      "nvidia"
	VendorName  ""
	BoardName   ""
EndSection
Hoping that could help...

Quote above has had hashed out lines removed. (It was too long) I used these settings in my xorg.conf file and since then I have not had a lockup. To be honest it was a weekly event, but now it's been about 10 days. Still I'm not convinced until three months or so is up, but have others tried this and had a change in their luck?

lookinin · Post by **lookinin** » Thu Feb 24, 2005 8:13 pm

Which part of the above do you think would be solving it? the nvidia sections or the module sections? Cause some of us are using ati & having gl freezes.

And as a side note, I haven't checked into this thread in awhile, but am still having problems...

ikaro · Post by **ikaro** » Thu Feb 24, 2005 10:44 pm

everyone who can, please upgrade to 2.6.11-rc5 if that wont bring you problems.

running 100% cpu now, and browsing at the same time, this was something that caused hardlocks instantly on my box just
2 weeks ago.

please try the kernel and post the results.

lookinin · Post by **lookinin** » Thu Feb 24, 2005 10:50 pm

I am using gentoo-dev-sources... and can find no such kernel

sorry for the spam - what would you suggest?

Code: Select all

# find . -name *2.6.11*
./development-sources/files/digest-development-sources-2.6.11_rc1
./development-sources/files/digest-development-sources-2.6.11_rc2
./development-sources/files/digest-development-sources-2.6.11_rc3
./development-sources/files/digest-development-sources-2.6.11_rc4
./development-sources/development-sources-2.6.11_rc1.ebuild
./development-sources/development-sources-2.6.11_rc2.ebuild
./development-sources/development-sources-2.6.11_rc3.ebuild
./development-sources/development-sources-2.6.11_rc4.ebuild
./mm-sources/files/digest-mm-sources-2.6.11_rc1-r1
./mm-sources/files/digest-mm-sources-2.6.11_rc2-r1
./mm-sources/files/digest-mm-sources-2.6.11_rc2-r2
./mm-sources/mm-sources-2.6.11_rc1-r1.ebuild
./mm-sources/mm-sources-2.6.11_rc2-r1.ebuild
./mm-sources/mm-sources-2.6.11_rc2-r2.ebuild
./vanilla-sources/files/digest-vanilla-sources-2.6.11_rc1
./vanilla-sources/files/digest-vanilla-sources-2.6.11_rc2
./vanilla-sources/files/digest-vanilla-sources-2.6.11_rc3
./vanilla-sources/files/digest-vanilla-sources-2.6.11_rc4
./vanilla-sources/vanilla-sources-2.6.11_rc1.ebuild
./vanilla-sources/vanilla-sources-2.6.11_rc2.ebuild
./vanilla-sources/vanilla-sources-2.6.11_rc3.ebuild
./vanilla-sources/vanilla-sources-2.6.11_rc4.ebuild

ikaro · Post by **ikaro** » Thu Feb 24, 2005 11:05 pm

its on kernel.org

weyhan · Post by **weyhan** » Fri Feb 25, 2005 2:17 pm

ikaro wrote:everyone who can, please upgrade to 2.6.11-rc5 if that wont bring you problems.

running 100% cpu now, and browsing at the same time, this was something that caused hardlocks instantly on my box just
2 weeks ago.

please try the kernel and post the results.

Tried it with development-sources-2.6.11-rc5. Still freeze, no mouse, no keyboard, can't ssh in.

davecs · Post by **davecs** » Fri Feb 25, 2005 6:17 pm

lookinin wrote:Which part of the above do you think would be solving it? the nvidia sections or the module sections? Cause some of us are using ati & having gl freezes.

I don't really know. Put it this way, none of this adversely affected performance, so I see no need to try to whittle it down.

weyhan wrote:
ikaro wrote:everyone who can, please upgrade to 2.6.11-rc5 if that wont bring you problems.

running 100% cpu now, and browsing at the same time, this was something that caused hardlocks instantly on my box just
2 weeks ago.

please try the kernel and post the results.
Tried it with development-sources-2.6.11-rc5. Still freeze, no mouse, no keyboard, can't ssh in.

Weyhan: Before update, was your freeze the one where only the mouse keeps working? There is more than one type. If it is not, your solution may be different.

Ikaro: Is the latest kernel recommended to fix the problem, or just because there's no Gentoo patches in it?

weyhan · Post by **weyhan** » Sat Feb 26, 2005 6:17 am

davecs wrote:Weyhan: Before update, was your freeze the one where only the mouse keeps working? There is more than one type. If it is not, your solution may be different.

Sorry for not being clear. It was late and frustrating that none of the solution solves the problem. However, I did mention, no mouse, no keyboard, can't ssh in when it freeze.

So I have the variant that everything freezes tight. Nothing generates a response except for the power button and the reset button. Because of this problem, I have yet to install other packages I normally use to keep the troubleshooting to a smaller set of possibilities. At the moment, I only use mozilla, gaim, and xterm when I am on Gentoo.

The PC is used as a desktop dual boot with Gentoo and WinXP. It is not on all the time. I use it everyday and boot it up every time when I need to use it. I'm on the WinXP side more because of the freeze. They all happened during 1 of the 2 scenario.

1) An interaction with mozilla that cause mozilla to load a page, like clicking on a link, clicking on the 'back' button, or hitting 'Enter' on a textbox (e.g. logging in to my.yahoo.com). The freeze happens at random and when it freezes it is immediate. My mouse button is still down most of the time. However, not before the logo/progress indicator at the top right of the mozilla window started its animation. I can see it start to move and stop almost immediately. Nothing else will cause the freeze in mozilla though (scrolling, etc).

2) Freeze at random when the session is left open with the screensaver running. No telling how long it will take to freeze and which screensaver it is running at that time. This is rare though. 9 out of 10 times the PC freeze in scenario 1.

In my daily usage I have not seen it crash free for more then 2 session of my usage and it can freeze anytime from the first button/link click on mozilla right after it was brought up or hours of usage before it freezes. There are days in between when I complete my session and shutdown normally.

I have tried most of all the suggestion found on this forums and nothing have made a significant change except when I turn off AGP fast write, there is a 2 second lag between the click on mozilla and the system freezing. Looks like more troubleshooting. If anyone wants more information, it is available upon request. At the moment, that's the only think I could think of that is related to the freezing problem. (AAAHHHHH... wish I can turn up the heater and the freezing will go away!!!)

If anyone is interested, below are the specs.

Hardware:

Asus P4S533-E mb
P4 1.80GHz
256Mb RAM
ATI Radeon VE (OEM, those that is not made by ATI except for the chipset)
Adaptec AHA-2940UW SCSI controller (HD on IDE though)

OS - stage1 install no bin package (and settings I remember):

gentoo-dev-sources-2.6.10-r6 (initial install 2.6.9-rx. same problem)
APIC
APCI
vesafb-ng (tried removing. same result)
agp
dvi
xorg-6.8.0-r4
gnome-2.8.1-r1
mozilla-1.7.5

ikaro · Post by **ikaro** » Sat Feb 26, 2005 3:02 pm

davecs wrote: Ikaro: Is the latest kernel recommended to fix the problem, or just because there's no Gentoo patches in it?

All i can say if that using the vanilla 2.6.11-rc5 kernel Im not having hardlocks or freezes anymore, until further notice.

The development sources from gentoo have patches in it .. and so have other sources......
I'd recomend using a Vanilla kernel w/out any external patches until you find something that works....
That way you can exclude the possibility that some patch its causing the trouble.

joehni · Post by **joehni** » Sat Feb 26, 2005 11:49 pm

Today I had a new variant. After the usual freeze I sshed into my frozen system and X took nearly 99% CPU as usual. I waited with killing X and a reboot though, since I had started a background task I wanted to complete. Suddenly after ~20 mins the screen refreshed and the system was back to normal. Happened to anyone else?

BTW: Disabling fast AGP writes does not help for my system. Next will be Tuttle's settings ...

ikaro · Post by **ikaro** » Sun Feb 27, 2005 12:29 am

ive been trying kde tonite and whenever I opened konqueror the machine hardlocked instantly.

but using fvwm, using firefox no freezes or hardlocks whatsoever until now.

So anyone who still has problems is using KDE ?

joehni · Post by **joehni** » Sun Feb 27, 2005 11:30 am

ikaro wrote:So anyone who still has problems is using KDE ?

For me it's definately X. Although I normally work with KDE, I do my OpenGL tests that provoke the freeze within twm.

gspr · Post by **gspr** » Sun Feb 27, 2005 1:36 pm

ikaro wrote:
davecs wrote: Ikaro: Is the latest kernel recommended to fix the problem, or just because there's no Gentoo patches in it?
All i can say if that using the vanilla 2.6.11-rc5 kernel Im not having hardlocks or freezes anymore, until further notice.

The development sources from gentoo have patches in it .. and so have other sources......
I'd recomend using a Vanilla kernel w/out any external patches until you find something that works....
That way you can exclude the possibility that some patch its causing the trouble.

No dice with 2.6.11-rc5 vanilla here. However, I did find the most efficient way for me to test if lockups occur - Enemy Territory always seems to cause the issue within 5 minutes

As for the issue being related to KDE: One of the first things I did when I discovered this many many months ago, was to try to run games in Fluxbox or without a WM. It sadly had no effect.

paulbarker · Post by **paulbarker** » Sun Feb 27, 2005 2:03 pm

Hi all,

I'm having the same problem, type 1 freezes where i must use SysRq keys to shutdown the computer safely. Having read thru this thread I am going to try the nv drivers instead of nvidia and check whether GLX is enabled. These crashes happen randomly, not dependent on whether an opengl,etc. app is running. I know that my make.conf is pretty standard with flags of "-O2 -fomit-frame-pointer".

I will keep you posted on how things go.

paulbarker · Post by **paulbarker** » Sun Feb 27, 2005 2:26 pm

Ok, slight problem

find / -iname xorg.conf

gives me a blank. I managed to dig up /etc/X11/xorg.conf.example but havn't a clue where to start with this file. It doesn't even look like the defaults that are loaded.

The question is, if I write an xorg.conf, do I just put it at /etc/X11/xorg.conf, and if I just include the changes that I want to make to the defaults, will it still work or will it expect me to write values for all the settings??

Cheers,
Paul

paulbarker · Post by **paulbarker** » Sun Feb 27, 2005 3:19 pm

Posted too quickly, I've now ran xorgconfig and got the thing setup properly. Using nv as the driver doesn't solve the problem, so I'll try again in a few hours.

My analysis of the problem after reading the whole thread is this:

X-Windows system (any version) causes a freeze. Main offenders are firefox and GLX accelerated programs. It seems that the crash is in the kernel graphics driver, rather than X since after X is killed, the display remains corrupted (not able to enter text-mode properly). The rest of the kernel seems fine, input thru keyboard works once X has been killed. If the problem is hardware related it may be due to either AGP support in linux or a feature common between ATI and NVidia graphics cards. The problem does not seem to be related to power supply or overheating (intense compilation under Windows 98/Cygwin works fine on my comp, no crashes in text-mode so far), and is not restricted to Gentoo linux.

My config, for those interested is:

Duron 1.2Ghz, Aopen Motherboard (AK-77 333), AOpen NVidia graphics, 256M ram @ 333Mhz.
Kernel version gentoo-dev-sources-2.6.10-r5, using genkernel and nvidia-kernel
Latest non-masked version of xorg, gnome and kde (problem occurs on both kde and gnome)
emerge --sync and emerge -u world ran yesterday so everything else is up to date.

HTH,
Paul

Mysterious Freeze

NPTL ?

Re: NPTL ?

Re: No more freezes since a while...