X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M)

NovHak · 03-25-12, 03:34 AM

Dear forum readers and contributors,

Help !

It seems to me I've tried whatever could be tried but I still can't use the proprietary driver on my system. I attached a bug report.

Symptoms : A short time after the login screen has been displayed, X becomes unresponsive, except for the mouse which can still be moved. Switching to another VT via Ctrl-Alt-Fn doesn't work any more. I can switch to another VT before the freeze. If I try to reboot from a text VT the system gets stuck, the fans become mad and I get a message "stall on CPU1" or something like that.

dmesg displays a "fallen off the bus" error, as well as some ACPI errors related to MXMI/MXMS methods being passed too much arguments. I tried booting with acpi=off with no success (except that the ACPI messages were gone...), same freeze at startup.

Flashed the system BIOS and vBIOS to more recent versions with no change, except for slightly less ACPI MXM-related error messages. Anyway, I doubt this is ACPI-related because of my acpi=off try. I passed NVreg_RegisterForACPIEvents=0 anyway, with no success either.

I tried X server options too, such as disabling the Composite extension, setting ConnectToAcpid to "False" => no success. Both even seem to be ignored, since I still see
"(II) Initializing built-in extension COMPOSITE", and
"(II) Open ACPI successful (/var/run/acpid.socket)"
despite being acknowledged earlier in the log file (the log file even explicitly says the Composite extension will be disabled).

Any advice is welcome !

EDIT : FWIW, my machine is a Clevo M571RU-U

NovHak · 03-26-12, 12:10 AM

Here's another bug report with more information. I've set DISPLAY and XAUTHORITY environment variables accordingly and more information has been gathered.

I'm still using driver version 295.33, the most up to date "stable" version as of now.

It may be worth mentioning that this X hangup problem is not new for me. It started after a driver update sometime around last summer I think. I was using a Knoppix live CD at that time. I don't remember the last functional driver version , neither do I remember the precise symptoms, but it looked the same, i.e. systematic hangups at startup.

Could this bug be related ?

apriori · 04-03-12, 07:51 AM

Hey man, let me add some stuff I tried on my Clevo 570RU (with an 8800 M GTX) which don't seem to help either.

Deactivate acpi all together - in that scenario the card will "fall of the bus" 28 secs after the last call to acpi, which failed
Code:
```
ACPI Warning: For \_SB_.PCI0.PEGP.VGA_.MXMS: Excess arguments - needs 1, found 2 (20110623/nspredef-326)
```
in normal mode (full apci with WMI): Same as in 1. but it needs 58 - 60 secs for it. Why on earth that difference? No idea.
Completely blacklisting or even uninstalling nouveau - doesn't help at all (the module nouveau wasn't loaded when the nvidia driver were used anyway) + "nouveau.modeset=0" (which is mandatory when using nvidia binary drivers)
Fix the DSDT - I had 2 errors and 3 warnings in it. The errors were minor (invalid parameter window lengths) so I could fix those easily. However, I can't do anything about the earlier mentioned excess arguments problem, because I don't know enough about that stuff. I doubt that these two calls are the cause of this misbehavior, though, because completely deactiving acpi didn't help either.
Upgrade the BIOS to an unlocked variant for more options (warning: of course out of vendor warranty) - Version 10.13aSL, KBC/EC 1.00.10

5.1 Deactivate ASPM for the PCIe ports - no change whatsoever
5.2 Deactivate processor power management - (I had to do that with kernel parameter processor.max_cstate=0 anyway, otherwise the machine would always stall after some time independant of nvidia drivers)
Upgrade the video cards BIOS to version 69.92.40.18 - no change
Other tried versions: 69.23.00.12 - no change either

Several other "solutions" (none of them works for me) floating around which didn't help:

Using persistance mode as fast as possible on bootup using nvidia-smi
Changing the PCIe port - funny stuff, how shall I do that on a nootebook?

So far the only workaround which is working is - guess what: Reverting to ancient versions of the nvidia driver. Latest working driver is 270.41.19, which of course comes with the implication, that you're stuck with a 2.6 kernel and Xorg 1.10. It's a (excuse me) pain in the ass to actually revert most distros to such old versions of the respective packages. So far every following nvidia driver did not work.

Since this is not my first post about this issue, and there are several others who have the same problem, it would be nice if NVIDIA would actually say a word about this or how we could assist in the solution. This "bug" (and I think it is a bug, because the card works flawlessly under windows) now is VERY old and renders my card useless on linux (I need the performance for graphics dev).

One other thing though, if you went to all the trouble to revert to these ancient versions, the funny thing is, you can even get OpenCL working with it, using the respective lib of the newest drivers. Definetly not an intended feature, I guess, but it works.

EDIT: I will attach the bugreport tar.gz/more logs later this day

Regards

apriori

momann · 04-04-12, 04:08 AM

I have the same laptop and the same problem. The X server hangs right after logging in. The time between logging and freezing depends on the drivers version as it happens with the symptoms. With some drivers it is just a black screen from the very beginning, but sometimes I have the time to open an application before the system freezes. With the most recent drivers versions (>285) the pointer is still active, but the computer is otherwise unresponsive to any mouse and keyboard instructions and I need to power it down.

This happens with all recent drivers (at least from 275.x) but not with the old drivers. I am running Debian 6 (64-bit) and the system works fine with version 195.31.36.

The problems happens regardless of the installation method (apt, compiling to the kernel in the Debian way, or installing the Nvidia bundles). Removing and blacklisting nouveau, framebuffers, etc, does not help.

This problem occurs in all Linux distros I have tried (Debian-based, RHEL-based, Slackware-based, etc), but not in FreeBSD. In FreeBSD even the most recent driver versions run fine.

The only workaround I have found this far is sticking to old driver versions (which in Debian Stable is not a problem for the time being). This also forces me to use CUDA version 3 (CUDA 4 requires at least the version 285.x of the drivers).

My question is: Is this problem specific for this particular laptop/motherboard or is the card deprecated for Linux systems?

ADDENDUM: A similar problem also happened for a while in my workstation, which has a Tyan S7025 motherboard, a Quadro FX 4800 and three Tesla C1060. However, the problem appears to be corrected from version 285.x. With the laptop, though, the problem persists with all recent versions of the drivers...

momann · 04-04-12, 09:11 AM

UPDATE: Version 270.41.19 (Nvidia Bundle) seems to run fine. I have been using this driver for about one hour without experiencing any problems. This has allowed me to install Cuda 4.0 and to profit from GPU acceleration in VMD, which was actually my goal.

momann · 05-02-12, 01:56 PM

In Linux, the Nouveau driver works just fine.

I have also tried all kind of things with my BIOS (flashing it to different versions of the BIOS and EC firmware, unlocking the BIOS, trying different settings combinations) to no avail. In desperation, I have even flashed the VBIOS with a 9800M GT ROM. It makes no difference whatsoever.

As I mentioned somewhere else, FreeBSD 9 runs flawlessly with the 290.10 driver. OpenIndiana (Solaris) also works well with a relatively new version of the drivers (I do not remember which one).

Provided this seems to be a laptop model-specific issue and a Linux-specific issue, I do not think that NVIDIA is going to fix the problem.

I wonder why this problem is only present in Linux, but in my case, the sound card also gives problems with Linux kernels newer than 2.6.32 and this problem is also Linux-specific.

I have been playing with FreeBSD recently (PC-BSD, in fact) and I think I am going to migrate this laptop to BSD as soon as I have some time to deal with the learning curve...

Licaon · 05-02-12, 02:33 PM

302.07 too?

momann · 05-03-12, 03:15 AM

Quote:

Originally Posted by Licaon

302.07 too?

Yes. With the 302.07 driver, the delay between login and freezing is maybe a bit longer. I was able to run glxgears, which reported an extremely poor performance (even if that is irrelevant at this stage).

All drivers newer than 270.41.19 freeze this laptop in all Linux distros and with all kernel and xorg-server versions.

The problem with 270.41.19 is two-fold:

1. It cannot be installed with kernels of the 3.x series.

2. It is not compatible with xorg-server 1.11 (and above, I guess).

momann · 05-06-12, 02:06 PM

Somebody seems to have found a very long workaround:

http://koroshiyaitchy.wordpress.com/...70ru-u-laptop/

It is an interesting approach, tough, which illustrate the power of free software, which allows one to create an operating system specifically for his/her laptop.

sandipt · 05-08-12, 01:48 PM

Looks like most of you facing issue with Clevo notebooks . Please attach nvidia bug report as soon as you observed issue by running nvidia-bug-report.sh script as root user.

momann · 05-09-12, 03:07 AM

Quote:

Originally Posted by sandipt

Looks like most of you facing issue with Clevo notebooks . Please attach nvidia bug report as soon as you observed issue by running nvidia-bug-report.sh script as root user.

The problem is the the freezing makes the monitor and the keyboard completely unresponsive. The mouse pointer is still active with some driver versions, though, which make me guess that the OS is still running in the background. In other words, it looks more like an X server freezing than as a system freezing.

However, the only way I have found to use the system after the freezing is powering it down by pressing the power button for a few seconds. No other options seem to work. Therefore, I would like to ask you whether or not it would make sense to run the Nvidia bug report script after rebooting or before the freezing happens (it happens one minute after the xserver had started).

momann · 05-09-12, 03:41 AM

Quote:

Originally Posted by sandipt

Looks like most of you facing issue with Clevo notebooks . Please attach nvidia bug report as soon as you observed issue by running nvidia-bug-report.sh script as root user.

I attach two log files from the nvidia-bug-report.sh script:

- One for the latest working drivers (270.41.19).

- One for the latest drivers (295.49) right before the freezing (as explained before, I have not found a way to do it after the freezing).

I hope this helps. As mentioned before, this bug is Linux-specific (FreeBSD and OpenSolaris are not affected). In Linux it affects all distributions and kernels I have tried so far. It also appears to be model-specific. In my case I have a Clevo M570RU-U.

04-03-12, 07:51 AM	#3
apriori Registered User Join Date: Nov 2008 Posts: 22	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) Hey man, let me add some stuff I tried on my Clevo 570RU (with an 8800 M GTX) which don't seem to help either. Deactivate acpi all together - in that scenario the card will "fall of the bus" 28 secs after the last call to acpi, which failed Code: ACPI Warning: For \_SB_.PCI0.PEGP.VGA_.MXMS: Excess arguments - needs 1, found 2 (20110623/nspredef-326) in normal mode (full apci with WMI): Same as in 1. but it needs 58 - 60 secs for it. Why on earth that difference? No idea. Completely blacklisting or even uninstalling nouveau - doesn't help at all (the module nouveau wasn't loaded when the nvidia driver were used anyway) + "nouveau.modeset=0" (which is mandatory when using nvidia binary drivers) Fix the DSDT - I had 2 errors and 3 warnings in it. The errors were minor (invalid parameter window lengths) so I could fix those easily. However, I can't do anything about the earlier mentioned excess arguments problem, because I don't know enough about that stuff. I doubt that these two calls are the cause of this misbehavior, though, because completely deactiving acpi didn't help either. Upgrade the BIOS to an unlocked variant for more options (warning: of course out of vendor warranty) - Version 10.13aSL, KBC/EC 1.00.10 5.1 Deactivate ASPM for the PCIe ports - no change whatsoever 5.2 Deactivate processor power management - (I had to do that with kernel parameter processor.max_cstate=0 anyway, otherwise the machine would always stall after some time independant of nvidia drivers) Upgrade the video cards BIOS to version 69.92.40.18 - no change Other tried versions: 69.23.00.12 - no change either Several other "solutions" (none of them works for me) floating around which didn't help: Using persistance mode as fast as possible on bootup using nvidia-smi Changing the PCIe port - funny stuff, how shall I do that on a nootebook? So far the only workaround which is working is - guess what: Reverting to ancient versions of the nvidia driver. Latest working driver is 270.41.19, which of course comes with the implication, that you're stuck with a 2.6 kernel and Xorg 1.10. It's a (excuse me) pain in the ass to actually revert most distros to such old versions of the respective packages. So far every following nvidia driver did not work. Since this is not my first post about this issue, and there are several others who have the same problem, it would be nice if NVIDIA would actually say a word about this or how we could assist in the solution. This "bug" (and I think it is a bug, because the card works flawlessly under windows) now is VERY old and renders my card useless on linux (I need the performance for graphics dev). One other thing though, if you went to all the trouble to revert to these ancient versions, the funny thing is, you can even get OpenCL working with it, using the respective lib of the newest drivers. Definetly not an intended feature, I guess, but it works. EDIT: I will attach the bugreport tar.gz/more logs later this day Regards apriori

04-04-12, 04:08 AM	#4
momann Registered User Join Date: Apr 2012 Posts: 22	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) I have the same laptop and the same problem. The X server hangs right after logging in. The time between logging and freezing depends on the drivers version as it happens with the symptoms. With some drivers it is just a black screen from the very beginning, but sometimes I have the time to open an application before the system freezes. With the most recent drivers versions (>285) the pointer is still active, but the computer is otherwise unresponsive to any mouse and keyboard instructions and I need to power it down. This happens with all recent drivers (at least from 275.x) but not with the old drivers. I am running Debian 6 (64-bit) and the system works fine with version 195.31.36. The problems happens regardless of the installation method (apt, compiling to the kernel in the Debian way, or installing the Nvidia bundles). Removing and blacklisting nouveau, framebuffers, etc, does not help. This problem occurs in all Linux distros I have tried (Debian-based, RHEL-based, Slackware-based, etc), but not in FreeBSD. In FreeBSD even the most recent driver versions run fine. The only workaround I have found this far is sticking to old driver versions (which in Debian Stable is not a problem for the time being). This also forces me to use CUDA version 3 (CUDA 4 requires at least the version 285.x of the drivers). My question is: Is this problem specific for this particular laptop/motherboard or is the card deprecated for Linux systems? ADDENDUM: A similar problem also happened for a while in my workstation, which has a Tyan S7025 motherboard, a Quadro FX 4800 and three Tesla C1060. However, the problem appears to be corrected from version 285.x. With the laptop, though, the problem persists with all recent versions of the drivers...

04-04-12, 09:11 AM	#5
momann Registered User Join Date: Apr 2012 Posts: 22	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) UPDATE: Version 270.41.19 (Nvidia Bundle) seems to run fine. I have been using this driver for about one hour without experiencing any problems. This has allowed me to install Cuda 4.0 and to profit from GPU acceleration in VMD, which was actually my goal.

05-02-12, 01:56 PM	#6
momann Registered User Join Date: Apr 2012 Posts: 22	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) In Linux, the Nouveau driver works just fine. I have also tried all kind of things with my BIOS (flashing it to different versions of the BIOS and EC firmware, unlocking the BIOS, trying different settings combinations) to no avail. In desperation, I have even flashed the VBIOS with a 9800M GT ROM. It makes no difference whatsoever. As I mentioned somewhere else, FreeBSD 9 runs flawlessly with the 290.10 driver. OpenIndiana (Solaris) also works well with a relatively new version of the drivers (I do not remember which one). Provided this seems to be a laptop model-specific issue and a Linux-specific issue, I do not think that NVIDIA is going to fix the problem. I wonder why this problem is only present in Linux, but in my case, the sound card also gives problems with Linux kernels newer than 2.6.32 and this problem is also Linux-specific. I have been playing with FreeBSD recently (PC-BSD, in fact) and I think I am going to migrate this laptop to BSD as soon as I have some time to deal with the learning curve...

05-02-12, 02:33 PM	#7
Licaon Registered User Join Date: Nov 2004 Location: Between the keyboard and the chair. Posts: 490	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) 302.07 too?

05-06-12, 02:06 PM	#9
momann Registered User Join Date: Apr 2012 Posts: 22	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) Somebody seems to have found a very long workaround: http://koroshiyaitchy.wordpress.com/...70ru-u-laptop/ It is an interesting approach, tough, which illustrate the power of free software, which allows one to create an operating system specifically for his/her laptop.

05-08-12, 01:48 PM	#10
sandipt NVIDIA Corporation Join Date: Dec 2010 Posts: 259	Re: X crashes systematically at startup (Ubuntu 11.10, 8800M GTX/G92M) Looks like most of you facing issue with Clevo notebooks . Please attach nvidia bug report as soon as you observed issue by running nvidia-bug-report.sh script as root user.