Mysterious Freeze

Message

jannis · Post by **jannis** » Thu Jan 06, 2005 11:25 pm

korngerd your system keeps crashing? what about disabling ACPI and enabling APM? (just a try).

bunder · Post by **bunder** » Fri Jan 07, 2005 12:44 am

Hey guys, is it possible for someone to test this combination (or most of it)?

2.6.9-gentoo-r6 config
xorg-x11-6.8.0-r3 config
nvidia-kernel-1.0.6629-r1
nvidia-glx-1.0.6629-r1

It's been running solid for the past 3 weeks or so for me...

korngerd · Post by **korngerd** » Fri Jan 07, 2005 2:13 pm

jannis wrote:korngerd your system keeps crashing? what about disabling ACPI and enabling APM? (just a try).

Yup, well it crashes after a few hours (as described by pretty much all users here - mouse moves, no other activities possible, X uses 100% CPU cycles, ssh'ing in and killing X is the solution, blah blah). I just recompiled my kernel without ACPI and enabled APM instead. I'll post back on the status of the system in a bit.

korngerd · Post by **korngerd** » Fri Jan 07, 2005 2:52 pm

OK, finally crashed. The kernel log gave me the exact same error I got when using the ACPI kernel:

Code: Select all

Jan  7 09:51:01 [kernel] NVRM: Xid: 6, PE0000 030c 00b8bbc6 00000000 00b8bbc6 00b8bbc6

So, it's probably not ACPI/APM related.

korngerd · Post by **korngerd** » Fri Jan 07, 2005 3:06 pm

Just something new - when X crashed, I tried an strace on the X process. It wasn't doing any syscalls, so I tried moving the mouse in it. The mouse movement was actually generating some strange syscalls:

Code: Select all

--- SIGIO (I/O possible) @ 0 (0) ---
select(13, [12], NULL, NULL, {0, 0})    = 1 (in [12], left {0, 0})
rt_sigprocmask(SIG_BLOCK, [IO], [IO], 8) = 0
read(12, "\30\377\0\0", 64)             = 4
gettimeofday({1105109891, 958570}, NULL) = 0
rt_sigprocmask(SIG_BLOCK, [], [IO], 8)  = 0
select(1024, [12], NULL, NULL, {0, 0})  = 0 (Timeout)
sigreturn()                             = ? (mask now [])

Obviously, I also tried doing some keyboard input, but that didn't generate any traces. I'm wondering what kind of syscall is being made (if it is) when X crashes... I might try it later today.

korngerd · Post by **korngerd** » Sat Jan 08, 2005 3:08 am

beugh wrote:Hey guys, is it possible for someone to test this combination (or most of it)?

I edited my kernel config to mimic yours. I haven't checked my xorg config against yours, but just wanted to do it step-by-step. If this works well, it might be something in our kernel configs. If not, I'll check my xorg config against yours

*hopes this all works...*

yaneurabeya · Post by **yaneurabeya** » Sat Jan 08, 2005 6:57 am

korngerd wrote:
yaneurabeya wrote:I'm seriously curious what the Nvidia chipset users have for their power settings. Please list them and let's compare this.
I'm on a full ACPI kernel:
Code: Select all
CONFIG_PM=y
CONFIG_SOFTWARE_SUSPEND=y
CONFIG_PM_STD_PARTITION=""
CONFIG_ACPI=y
CONFIG_ACPI_BOOT=y
CONFIG_ACPI_INTERPRETER=y
CONFIG_ACPI_SLEEP=y
CONFIG_ACPI_SLEEP_PROC_FS=y
CONFIG_ACPI_BUTTON=y
CONFIG_ACPI_PROCESSOR=y
CONFIG_ACPI_THERMAL=y
CONFIG_ACPI_BLACKLIST_YEAR=0
CONFIG_ACPI_BUS=y
CONFIG_ACPI_EC=y
CONFIG_ACPI_POWER=y
CONFIG_ACPI_PCI=y
CONFIG_ACPI_SYSTEM=y
yaneurabeya wrote: I don't think that anything weird with cold/hotplug was occurring but just as a point of question, how many people are still running devfs, what kernel version, and who's running udev?
I'm on a udev system running:
Code: Select all
$ uname -a
Linux nyamochan 2.6.9-gentoo-r13 #3 Wed Dec 29 21:00:04 EST 2004 i686 AMD Athlon(tm) Processor AuthenticAMD GNU/Linux

Ok, cool. What I'm just doing is trying to solve this using a process of elimination. It might be chipset specific for some, but I think that there is an overall issue that needs to be addressed that-at least in my case-is power management related and it appears that ACPI doesn't affect my Dell as much as APM does, and for some this is different.

As a couple of sidenotes if I haven't mentioned them already:
1. I did not install anything apmd related on my pc after I added the APM elements into the kernel, but rather just left APM to be managed by the kernel itself.
2. I haven't tweaked with hdparm any bit since my last reinstall although I did get quirky errors when I used it back in the day =\...
3. If memory served me correctly I didn't have issues with X.org/power support at or before the gentoo-dev-sources-2.6.9-r4.

Maybe we should submit something like this to the kernel maintainers if we determine that power management is at fault.

korngerd · Post by **korngerd** » Sat Jan 08, 2005 2:01 pm

beugh wrote:Hey guys, is it possible for someone to test this combination (or most of it)?

2.6.9-gentoo-r6 config

The kernel config alone didn't fix it. It crashed after about 2 hours again. I'll try to see how different my xorg config is from yours, and try it out with both kernel + xorg config changes.

korngerd · Post by **korngerd** » Sat Jan 08, 2005 2:23 pm

beugh wrote:Hey guys, is it possible for someone to test this combination (or most of it)?

2.6.9-gentoo-r6 config
xorg-x11-6.8.0-r3 config
nvidia-kernel-1.0.6629-r1
nvidia-glx-1.0.6629-r1

It's been running solid for the past 3 weeks or so for me...

Nope, died again. This time it was less than 20 minutes... So, it's probably not the kernel/xorg config (at least not for me).

bunder · Post by **bunder** » Sat Jan 08, 2005 2:32 pm

Dang. So I wonder what it is then...

yaneurabeya · Post by **yaneurabeya** » Sun Jan 09, 2005 3:56 am

Can anyone possibly take a look at their /var/log/messages file prior to the crash? I will be sure to do that as well once I return to school.

gspr · Post by **gspr** » Sun Jan 09, 2005 2:21 pm

yaneurabeya wrote:Can anyone possibly take a look at their /var/log/messages file prior to the crash? I will be sure to do that as well once I return to school.

Will check as soon as I return to my desktop computer (in a few days). I'm quite sure it was one of the infamous NVRM messages.

Has anybody tried with an 6.8.1.x version of X.org?

iarwain · Post by **iarwain** » Sun Jan 09, 2005 8:32 pm

yaneurabeya wrote: Has anybody tried with an 6.8.1.x version of X.org?

I'm testing in another computer with Fedora Core 3 and X.org 6.8.1.
2 weeks with ~16h uptime/day, and no crashes. Rock solid.

Specs:

ASUS P4C800-E Deluxe
Pentium IV HT
XFX Geforce 5900 XT
AGPGART enabled 8X
SBA ON and FatWrites OFF (It was like that by default)
Kernel 2.6.9-smp
NVIDIA drivers 6629

I haven't touched anything in the kernel, I have only disabled riva-fb. ACPI and IO-APIC enabled. APM too.

piizba · Post by **piizba** » Sun Jan 09, 2005 10:06 pm

You can add me to the list...
I installed gentoo last year and had no problems.
But now i get freezes if i use xmms,gaim or firefox,
i dont get freezes on GL screensavers hmmm.

Card:
Hercules Radeon 9000Pro 128Mb

gspr · Post by **gspr** » Mon Jan 10, 2005 5:37 pm

I will try with X.org 6.8.1.xxx as soon as the new freetype is out (the current freetype on which the latest X.org depends breaks Firefox etc., and I can't live without firefox).

meff · Post by **meff** » Mon Jan 10, 2005 10:50 pm

Hey guys.

For those of you getting a hard-lock with a VIA chipset and an nVidia card (possibly even others) .. Check your BIOS for an "AGP Drive Strength" setting, and play around with the hex. Before I set this manually to "CA" my box would die randomly within 1-5 days on average using my nVidia GF4 MX440 64M.

There is a little note about this setting for VIA chipsets in the nVidia readme, and you can also find a bit more info here: http://www.rojakpot.com/showFreeBOG.asp ... 0&bogno=34

HTH
-r

pmaciolek · Post by **pmaciolek** » Tue Jan 11, 2005 5:49 pm

I had similar problems with my X-s (both xfree 4.3, xorg 6.8.0, xorg 6.8.1). 1-10 crashes every day... (nothing interesting in /var/log/messages, X-s were usually saying "Caught signal 11" or "*** glibc detected corrupted double-linked list: 0xxxxxxxx ***" and sometimes nothing but 100% CPU usage).

Now I have 48h+ crashfree at my Xorg 6.8.1. My solution was to switch off all acpi, apm, power management support and even CPU frequency options in kernel (basically all in "Power management options" is turned off). Now I will try to switch APM and cpufreq back, maybe problems are just ACPI related...

My system specification is MSI K8N, Athlon64 3200+, 1x Twinmos 512MB CL2.5, Leadtek FX5700 128MB, Matrox SATA 160GB. I'm using kernel 2.6.9-gentoo-r1.

rshadow · Post by **rshadow** » Wed Jan 12, 2005 7:18 am

Ok, this is absolutly driving me crazy! I beleive I'm having similar problems to everybody else. Here is my situation.

Sparaticly (as far as I can tell) my computer just freezes (partialy). Keyboard and Mouse die, and PAM or agetty appears to die. This happens in X (xorg) and just at the console without a X running.

If I am local at the console out input/output just stops. All daemons are responsive a little. for example

I can SSH to the box type in a username and password, but after typing the password it freezes.. I never get a login prompt.

No relevant information for /var/log/anylogfile

At first I thought this was a screen (the program) problem, because it happened whenever I tried to SSH into the box.. I could login 1 or 2 times and then it would freeze. Then I thought it was system load because it was only happening when I would emerge sync or try to emerge a large application like open office or mysql.

So I figure I have hardware problems (bad memory / hard disk).. nope.. all checks show my hardware to be just fine (btw: I have had gentoo running on this same box for over a year.. only happened after the fresh install that I did)

So my next though is that it is gcc 3.4 (I know it's not NPTL becaue I've used NPTL in other builds just fine (on this same system) ). So basicly I'm going to wipe everything and start over withouth gcc 3.4 and see how it goes...

HHansson · Post by **HHansson** » Wed Jan 12, 2005 2:09 pm

About a month ago, I experienced the same behaviour (X freexes, 100% CPU util. etc.) and I had to do a hard reboot to get back. I could never log on remotely. Prior to that I had no problems, and it was not triggered by any software updates. Once it started, it happened after 10 minutes -5 hours randomly.

I first suspected some hw problem and installed the lm sensors package but everything looked normal when freezing. I tried just about everything suggested here and elsewhere, upgrading, downgrading, kernel tweaking, you name it. I started with Fedora Core 1 but now I am on Core 3 with kernel 2.6.10.

Two days ago I went to buy a new power supply and it has been running fine since then. I cannot say that it will solve other people's problem but it solved my problem (so far). And perhaps it will point others in the right direction (power management issues).

korngerd · Post by **korngerd** » Wed Jan 12, 2005 7:34 pm

yaneurabeya wrote:Can anyone possibly take a look at their /var/log/messages file prior to the crash? I will be sure to do that as well once I return to school.

Running xorg-x11-6.8.0-r3 and nvidia-kernel-1.0.6629-r1 with nvidia-glx-1.0.6629-r1, I get the following RIGHT before it crashes:

Code: Select all

NVRM: Xid: 13, 0000 02005f00 0000009f 00000300 0010c300 00000002

It's some variant of the above, anyway (different memory addresses, I believe).

gabio · Post by **gabio** » Thu Jan 13, 2005 1:02 am

Hi, I get weird freeze here with firefox.
when i'm browsing. sometime it freeze. but not the mouse, though the pointer keep the same shape. Keyboard is dead. But some time listen to some music and it still play. like firefox would freeze the input devise like mouse and keybord?

VortexKI · Post by **VortexKI** » Thu Jan 13, 2005 3:07 pm

Hi,
I'm using Debian Sid with kernel 2.6.10 and nvidia 66.29.
ACPI and APM enabled.

Interesting, the same behaviour here:
If X crashes: mouse moves and xmms runs fine in the background, but other activities are not possible.

Especially mozilla and firefox make my system crash.
Is it possible, that the flash-plugin is partial responibel ?

Doesn't seem to be a gentoo specific problem.

davecs · Post by **davecs** » Thu Jan 13, 2005 8:48 pm

Just discovered this thread, which is odd, because I've been involved with several others on the same subject. My hardware is listed below.

At present, I have a triple-booting system using Gentoo, Win98SE and PClinuxOS. Under PCLOS I've had a couple of freezes today, but under strain. The one where the mouse moves but nothing else works.

One of the things I have done after the freeze is to type ALT-SYSRQ-K and ALT-SYSRQ-R. I have then pressed CTRL-ALT-F2 and typed root, and the password, though I have to do it blind. After that, sometimes pressing and holding the enter key causes the screen to rotate horizontally, though pressing Scroll Lock stops this. Pressing Scroll Lock again makes all the scrolls take place at once.

That is very relevant, because if you log into a terminal (or even an X terminal) pressing scroll lock suppresses output onto the terminal which re-appears when you press it again!

This suggests that the computer is still working after a fashion. I have even, on occasion, been able to type "reboot" to do a clean restart.

I don't get the problem that often, but I have had it on three different computers with three different types of Linux. All were set up for GL with nvidia cards. Both 2.4 and 2.6 kernels were involved. From what others have said, it must be a problem with accelerated 3D graphics. The fact that you are not at that point using them is not relevant, they are enabled and the code can be run, even if by accident.

Someone on another site somewhere said it may be a consequence of the memory timing involved with the graphics, and suggested I tried less aggressive timings, or upping the RAM voltage a point. I have done the latter for now, as the RAM timings are the manufacturer's recommendation. Let's see how we get on.

davecs · Post by **davecs** » Thu Jan 13, 2005 11:22 pm

I have managed to freeze the system playing lbreakout2, but I could not see whether the mouse was moving because there is no mouse pointer in this program.

However, I tried again, changing the setting of the Graphics to limit Frames per second to 100 and it seems OK. Am I getting somewhere?

yaneurabeya · Post by **yaneurabeya** » Fri Jan 14, 2005 9:27 pm

Sadly enough, it crashed again for me ;(. This is not doing wonders for my poor partitions, but at least I'm just unmounting everything but my / partition.

I've basically chased down the issue to being power management, since I am using 2.6.10-r4 now and it crashed once I compiled in APM. ACPI won't work because it still causes issues with my Realtek card. Using nv versus nvidia does NOT make a difference on crashing, conflicting with some of the bug people at Gentoo. I was going to try switching back to xorg-x11, but considering that console blanking still crashes the system, it's most likely APM related, and possibly dealing with udev (since I'm running a straight udev system without the devfs tarball or support). I should try devfs sometime, but seeing as how I don't have any issues if I don't compile in APM, I'll leave my setup as-is until the kernel people fix the issue.

Which makes me wonder... is it somehow related to the fact that Dell power supplies are non-standard?

Mysterious Freeze

Problem solved for me