Mysterious Freeze

[gian]

(joehni)

I have exactly the same symptons... try restarting X from the other computer you ssh with... sometimes X restarts in a weird way (e.g. with no background...)

(korngerd)
I also believe that it is not kernel's fault, but even less nvidia drivers fault (as many have erroneously (I think) said) since I get exctly the same problem with Ati. I am not sure, but I think (must check) that a friend of mine has the same problems with a Debian (the unstable version) running Xfree (and not Xorg...) meaning that Xorg (that I though the most probable culprit, may be in some way innocent..... that would seem to leave xscreensaver itself... anyone gets frozen running other GL applications ??

[ikaro]

yep.

[bunder]

Today I had another freeze. Meanwhile I run sshd and I was able to connect to my "frozen" system. Interesting symptom calling top: X consumed more than 95% of the CPU! But even after killing X the keyboard and screen was still frozen (though I had now a weird screen output). At least I could shutdown the system from the remote console ...

Yeah, I've had that happen before too... I think the process gets stuck when being killed by init, so you end up having to reboot.

[bunder]

anyone gets frozen running other GL applications ??

All the time. (Although I haven't been able to reproduce it the past 2 weeks)

Enemy-Territory, Quake 3, cvscedega+counter-strike, gl screensavers to be a little more exact.

[ikaro]

since 2.6.10 the situation got worse .. .. !!!
The system lockup pretty much all the time, completly at random, which makes me more sure that this is a kernel issue.
with 2.6.10-rc3-bk14 at least I haven't experienced one single freeze.

my current sittuation is:
If I use the NVIDIA drivers and ANY kernel = freezes;
If I use the NV driver, and kernel > 2.6.10-rc3-bk14 = freezes;

[korngerd]

anyone gets frozen running other GL applications ??

I actually got freezes without running any GL apps (firefox, gkrell, and a few aterms). Seems like when I start running "intensive" applications (I guess firefox is somewhat... resource intensive), it freezes on me. Some of the xscreensavers kill it too. But, that's all only when I use the "nvidia" driver (not "nv"). But as you mentioned too, there are people here with ATI drivers, so I guess it can't really be the "nvidia" driver. I'm clueless...

[korngerd]

I'm just wondering if people are getting any errors using verbose logging in X. From the Nvidia docs:

The NVIDIA X driver will output more messages when the verbosity
level is at or above 5 (X defaults to verbosity level 1 for stderr
and level 3 for the log file). So, to enable verbose messaging from
the NVIDIA X driver to both the log file and stderr, you could start
X by doing the following: 'startx -- -verbose 5 -logverbose 5'.

I've just started a verbose output of X (and hence Nvidia's drivers as well), so maybe this would shed some light on the mysteries of the X lockups...

Another thing I noticed from the Nvidia docs is:

You should use the AGP module that works best with your AGP chip set.
If you are experiencing problems with stability, you may want to start
by disabling AGP and observing if that solves the problems. Then you
can experiment with either of the other AGP modules.

So, has anyone played around with:

Option "NvAGP" "integer"
Configure AGP support. Integer argument can be one of:
0 : disable agp
1 : use NVIDIA's internal AGP support, if possible
2 : use AGPGART, if possible
3 : use any agp support (try AGPGART, then NVIDIA's AGP)
Please note that NVIDIA's internal AGP support cannot
work if AGPGART is either statically compiled into your
kernel or is built as a module, but loaded into your
kernel (some distributions load AGPGART into the kernel
at boot up). Default: 3 (the default was 1 until after
1.0-1251).

Also, some guy on the forums has compiled a list of Nvidia/Xorg related crashes (sorry, I'm not an ATI user, although I do understand that ATI people are having similar crashes):
http://forums.gentoo.org/viewtopic.php?t=237282
Sorry for the long post

[bunder]

So, has anyone played around with:

Option "NvAGP" "integer"
Configure AGP support. Integer argument can be one of:
0 : disable agp
1 : use NVIDIA's internal AGP support, if possible
2 : use AGPGART, if possible
3 : use any agp support (try AGPGART, then NVIDIA's AGP)
Please note that NVIDIA's internal AGP support cannot
work if AGPGART is either statically compiled into your
kernel or is built as a module, but loaded into your
kernel (some distributions load AGPGART into the kernel
at boot up). Default: 3 (the default was 1 until after
1.0-1251).

Sorry for the long post

A lot. I went from AGPGART and back several times thinking that was the problem.

[korngerd]

A lot. I went from AGPGART and back several times thinking that was the problem.

Hm... Did you compile AGPGART into your kernel, or as a module? I might try to see if that's the problem, but I'm still testing the verbose logging stuff (amazingly, it's been running xscreensaver for about 9 hours straight now). It'll eventually crash... Then, maybe I could scrape some information from the logs

Anyway, Happy New Year all

[Dolio]

I've had agpgart as both a module and compiled in, and had freezes both times.

You need to have agpgart as a module to use nvagp at all, though, I think.

[korngerd]

OK, I guess Nvidia's docs aren't too accurate afterall. Xorg finally crashed after about 13 hours. I did try the verbose logging, but wasn't able to obtain any information from it (I don't think the Nvidia driver logged a thing to my Xorg.log). I'm clueless now, since people seemed to have tried Xorg with and without AGPGART. Anyway, I'm going off to celebrate New Years Hope this bug gets irradicated sooner or later...

[jannis]

So I'm another user having this problem.
No, not on my own PC but on the one of my friend.

My (own) PC:
X.org 6.8.1.901 (6.8.2 RC 1)
ATI Radeon 9000 Pro
Kernel-Driver (not ati-driver)
2.6.9-nitro4
AMD Athlon XP 2000+ running as 2100+
SiS-Chipset on K7S5A
This box runs perfectly fine.

My friend's PC:
X.org 6.8.1.901 (6.8.2 RC 1)
GeForce4 MX 440
Nvidia-Kernel and Nvidia-glx 1.0.6629-r1 using the AGPART-Interface
2.6.9-nitro4
Intel Pentium 4 2 GHz
SiS-Chipset (don't know the MB)

This box doesn't work that perfect:

It's the described X-Server-Freeze with it having 100 %CPU, mouse moving, SSH possible, no keyboard....... I'm sure you know what I mean.

It doesn't matter what I do with that box. It freezes when I'm working with it, meaning some apps (no glx) or when I do nothing but login (doing all the emerge-stuff via SSH).
I just played Quake3 to verify it works and had no freeze there yet (I played max. 3 minutes).
The WM is not the prob, because when using the "default" WM (I don't know how it's called but you get 3 xterms and 1 clock) it freezes

I don't think I'll have the time to change the kernel-config or using the nv-driver as we know it doesn't fix it.

Actually I just wanted to paste the config of the 2 boxes. They are not really different (expect ATI <-> nVidia) and one works, the other doesn't.

P.S.: some say it's the driver which clocks the card faster than the card can handle. can someone tell me how to tell the driver how fast to clock the card? is there any tool for doing that in linux?

[yaneurabeya]

Heh. I'll join in the fun a bit too now. The thing is that I did get my Xorg to work properly with the nvidia driver and 2.6.9-r4, but considering that it crashes now due to a different kernel setup, I've been trying everything within my power to get it to run.

-I am running 2.6.9-r13 currently.
-I get a hard crash when running APM (so I'm going to try uncompiling APM and APIC and see what happens) and using the nvidia driver.
-When I run startx the splash loads and then does a hard freeze on the console.
-Supposedly this is power related. I am a part of a Linux listserv on my school's network and an admin did disable something in the kernel and it did work properly. That fix might only be Dell related.

My specs:
P4 2.4 Ghz, 512 Mb RD Ram, 64 Mb Geforce 4 Ti 4200, latest version of Xorg, gentoo-dev-sources-2.6.9-r13 compiled using the 'virgin method' (ie no genkernel ).

I will see if disabling the APM works for me. I'd rather not lose APM/APIC but seeing as how the APM might conflict with my X and the APIC conflicts with my realtek 8139 card, I might have no other choice but to do this.

//EDIT:
Not compiling APM support and Power Management support into the kernel source did the trick it seems. ACPI works perfectly fine with my 8139 card now (woot for better low power level support!). I'll try to reproduce my error exactly, but it has to deal with APM I think. My _verbose_ bug submit can be found at http://bugs.gentoo.org/show_bug.cgi?id=76503.

[jannis]

Ok, changing the grapiccard's clocks didn't fix it ://

But I've been off the box for about 8 hours and it didn't yet freeze.

Can someone tell me how to disable Font-Antialiasing in X? Just wanna play a bit.

[ikaro]

dont load Freetype in xorg.

[jannis]

Ok, done this but doesn't fix it.
OK, now some hints :
Do you have this line in you /var/log/Xorg.0.log:
"(WW) Open APM failed (/dev/apm_bios) (No such file or directory)" ?

I had one X-Server crashing and I had a look at the logs. When a server is running and has not yet crashed, the last line of it's log is:
"SetClientVersion: 0 8"
When I look at a Xorg.0.log.old of a crashed server, the last lines are:

Code: Select all

SetClientVersion: 0 8
(WW) NVIDIA(0): WAIT (2, 6, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (1, 6, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (2, 6, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (1, 6, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (2, 7, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (1, 7, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (2, 7, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (1, 7, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (2, 7, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (1, 7, 0x800004x, 0x08x, 0x08x)
(WW) NVIDIA(0): WAIT (2, 6, 0x800004x, 0x08x, 0x08x)

might have something do do with graphics? it's certain!

[jannis]

Good News.
I diasbled ACPI in my kernel an enabled all (sane) options for APM.
Now the /var/log/Xorg.0.log ends with:

Code: Select all

SetClientVersion: 0 8
(II) Open APM successful
(II) NVIDIA(0): Setting mode "1600x1200"

Well and I didn't have a freeze yet. Please try this too and report if it helps. *hope*
(make sure you have /dev/apm_bios and your Xorg.0.log tells you it worked.)

[iarwain]

Good News.
I diasbled ACPI in my kernel an enabled all (sane) options for APM.
Now the /var/log/Xorg.0.log ends with:

Code: Select all

SetClientVersion: 0 8
(II) Open APM successful
(II) NVIDIA(0): Setting mode "1600x1200"

Well and I didn't have a freeze yet. Please try this too and report if it helps. *hope*
(make sure you have /dev/apm_bios and your Xorg.0.log tells you it worked.)

Hi jannis. I have ACPI disabled and APM enabled (CONFIG_APM and CONFIG_APM_RTC_IS_GMT in 2.4.26-r9). I have a correct /dev/apm_bios and Xorg.0.log tells me "(II) Open APM successful" too. I've had all this for months and unfortunately I sometimes have crashes. I've never tried disabling completely APM. I'll try it as soon as I can ( Oh no, I'll have to push the power button when shutting down!).

NVIDIA's readme says: "Power management support is still under development and a beta feature. As a result, some functionality is still missing or unreliable". Maybe it's more beta than I thought.

However, I guess that many people with this freeze has already disabled both ACPI and APM.[/url]

[jannis]

What do you exactly mean with "sometimes I have crahes"? I had a crash every 30 mins with ACPI and I didn't have one crash in 2 hours without ACPI. I'll continue testing tomorrow.

Has someone tried to fix this problem by updating his/her BIOS so that ACPI might work properly?
On my own PC I did so, but I don't know if it would work with my old BIOS.

The problem at my friend's box is, that it's a fujitsu-siemens and they always change the bios so I have to download an damn-old "update"

[iarwain]

What do you exactly mean with "sometimes I have crahes"? I had a crash every 30 mins with ACPI and I didn't have one crash in 2 hours without ACPI. I'll continue testing tomorrow.

I can't predict the crashes. With "sometimes" I mean that I can have 2 crashes in 1 day or also no crashes in 2 months.

[korngerd]

I just upgraded to the latest NVidia drivers (1.0.6629-r1), and got the following in dmesg:

Code: Select all

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

Not sure if that helps, but I'll look into it.

Fortunately (or unfortunately), the nvnews forum has a topic on this too:
http://www.nvnews.net/vbulletin/showthr ... 875&page=1

[korngerd]

Crashed again, but another address it seems:

Code: Select all

NVRM: Xid: 6, PE0000 030c 0004002a 00000000 00010001 0004002a

I think 1.0.6629 is worse than the current stable for x86. I'm getting crashes every 2-3 hours now...
I don't know if it applies, but there seems to be a patch that fixes this memory leak:
http://www.linux-gamers.net/modules/new ... toryid=561
Haven't tried the patch, but will post back later once I get the patch applied.

[yaneurabeya]

Crashed again, but another address it seems:

Code: Select all

NVRM: Xid: 6, PE0000 030c 0004002a 00000000 00010001 0004002a

I think 1.0.6629 is worse than the current stable for x86. I'm getting crashes every 2-3 hours now...
I don't know if it applies, but there seems to be a patch that fixes this memory leak:
http://www.linux-gamers.net/modules/new ... toryid=561
Haven't tried the patch, but will post back later once I get the patch applied.

I'm seriously curious what the Nvidia chipset users have for their power settings. Please list them and let's compare this.

Another thing, maybe the issue can't be totally related to either ACPI or APM, but rather one or the other depending on the motherboard. Someone who wrote into my linux forum here at school said he was able to stop the problem with Dell PCs by disabling something power management related. I am going to try the nv driver, not autoload the nvidia module on startup for Xorg and see what happens when I reenable APM support. The point is that APM console blanking at least for me appears to be a point of issue, as well as some other things possibly with APM internally. Here are the options I selected previously for my kernel:

Code: Select all

                  <M> APM (Advanced Power Management) BIOS support
                  [ ]   Ignore USER SUSPEND  
                  [*]   Enable PM at boot time
                  [ ]   Make CPU Idle calls when idle (NEW) 
                  [*]   Enable console blanking using APM
                  [ ]   RTC stores time in GMT (NEW) 
                  [*]   Allow interrupts during APM BIOS calls
                  [*]   Use real mode APM BIOS call to power off

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?

My kernel is...

Code: Select all

bash-2.05b# uname -a
Linux sprsd 2.6.9-gentoo-r13 #1 SMP Mon Jan 3 16:40:32 PST 2005 i686 Intel(R) Pentium(R) 4 CPU 2.40GHz GenuineIntel GNU/Linux

[korngerd]

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

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

[ikaro]

udev and acpi