Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

Reply
 
Thread Tools
Old 03-18-10, 01:05 PM   #1
Thaorius
Registered User
 
Join Date: Mar 2010
Posts: 5
Default NVIDIA driver on x86_64 freezing X

Hi, new guy here.

I'm setting up a new workstation (triple head) with 2 GeForce 8400 GS cards.

After a lot of pain, I managed to install the NVIDIA driver, edited my xorg.conf, rebooted, and the 3 screen started. So far so good. But then, after moving the mouse a bit, the 2 screens will just freeze, most of the time leaving what was already drawn on screen intact, sometimes a blackscreen on one of the monitors. If a blackscreen is produced, a system crash follows. Otherwise, the system continues to work normally if you access it via SSH, but the screens will continue to be freezed, no matter what I do (killing X, gdm, etc). One thing to note is, that I can't rmmod the nvidia driver after a freeze, "driver in use", even if X and gdm are long gone.

Nothing is ever printed to any log files concerning the error.

Also, if I happen to remove one of the screen (from the second video card), then no freezing will ever occur and everything will work as expected.

In any case, Xinerama doesn't work. When moving the mouse pointer from the center screen to any of the others, the mouse starts to jump from place to place randomly within the secondary screen it is in many times per second. I can drag windows though.

I've tried just about everything I can possibly think of, nothing works.

The driver version is: 190.53.
uname -a: Linux dexter 2.6.32-3-amd64 #1 SMP Wed Feb 24 18:07:42 UTC 2010 x86_64 GNU/Linux

You'll find my xorg.conf and lshw output attached. I can post an X log if needed, but quite frankly, there is nothing odd in them.

Thanks for your time.
Attached Files
File Type: txt xorg.conf.txt (2.5 KB, 126 views)
File Type: txt lshw.txt (24.7 KB, 112 views)
Thaorius is offline   Reply With Quote
Old 03-18-10, 06:56 PM   #2
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: NVIDIA driver on x86_64 freezing X

Thanks for the report.

Could you try the most recent driver 195.36.15? See http://www.nvnews.net/vbulletin/showthread.php?t=122606 for current driver releases.

Also, could you post an nvidia-bug-report.log? It gives us much more complete and detailed information about your system. You can generate it by running nvidia-bug-report.sh as root.
danix is offline   Reply With Quote
Old 03-18-10, 07:16 PM   #3
Thaorius
Registered User
 
Join Date: Mar 2010
Posts: 5
Default Re: NVIDIA driver on x86_64 freezing X

The 195.36.15 driver will crash the system during installation, generating weird hardware IO errors, or a total system crash, or weird characters on screen, or simply a force reboot (cpu triple fault?). I tested it on 4 different kernels, 2 of which where stock debian kernels (2.6.32-3-amd64 and 2.6.32-trunk-amd64), and 2 vanilla kernels (2.6.33.1 and 2.6.32). During 2 of those installs, I was running tail -f /var/log/dmesg on another tty. The crash symtoms mentioned above occur right after the nvidia module prints a hello kind of message.

I have currently downgraded the triple head to a dual head, and moved the extra head to my old workstation. It's damn annoying, that's for sure, but it works for the time being.

Xinerama won't work at all, dual or triple head, producing the error outlined on my first post. one thing to note though, is that the first time I activated Xinerama, it loaded and worked flawlessly, for about 5 seconds until the driver froze for the first time. After that, it glitched as described.

Also, I wasn't able to install the 190.53 driver on a vanilla 2.6.33.1 kernel. The installer simply said that it couldn't find the version information for the currently running kernel within the source tree (yes, it was the right tree). It did find it for the vanilla 2.6.32 kernel test though (kernel version not supported?).

I'll post the bug report later.
Thaorius is offline   Reply With Quote
Old 03-19-10, 03:50 PM   #4
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: NVIDIA driver on x86_64 freezing X

The bug report will be very helpful, especially the output of our installer log to see what is happening during your driver installation, as well as information about your X server and other system configuration items.

I reread your original post, and noticed the part you mentioned about the cursor jumping after crossing screen boundaries in Xinerama. This is an issue we have observed with X.org 1.7 and certain Xinerama configurations, which we are currently investigating. At the moment we believe it to be a bug in X, but we are still in the early stages of investigating the issue.

If you are running a 1.7 X.org, there were some problems with dual GPU setups that are resolved in 195.36.15, so hopefully you'll be able to get that version of the driver installed. Unfortunately, the Xinerama issues are not yet resolved.
danix is offline   Reply With Quote
Old 03-20-10, 06:44 PM   #5
const.crist
Registered User
 
Join Date: Mar 2010
Posts: 3
Default Re: NVIDIA driver on x86_64 freezing X

Quote:
Originally Posted by danix View Post
The bug report will be very helpful, especially the output of our installer log to see what is happening during your driver installation, as well as information about your X server and other system configuration items.

I reread your original post, and noticed the part you mentioned about the cursor jumping after crossing screen boundaries in Xinerama. This is an issue we have observed with X.org 1.7 and certain Xinerama configurations, which we are currently investigating. At the moment we believe it to be a bug in X, but we are still in the early stages of investigating the issue.

If you are running a 1.7 X.org, there were some problems with dual GPU setups that are resolved in 195.36.15, so hopefully you'll be able to get that version of the driver installed. Unfortunately, the Xinerama issues are not yet resolved.
hi!

same thing happens to me. and my configuration is not that fancy.
basically 2 cards (one geforce 6, the other one geforce 7) and an amd64 processor.

right away after the kernel module gets compiled, the installation crashes, the console
gets garbled with junk. the whole machine freezes and it cannot be accessed with
ssh either. I have tried the --ui=no to no avail; same thing again. at the beggining I have suspected the fb support in the kernel. it is disabled right now. after the machine
reboots, the logs are full of junk. the _same_ problem appears when I try to install 17x drivers.

the cards are o.k. since they have been functioning perfectly with 2.6.28 / 18x driver.

I saw that every driver release has some min. software req. explicitly stated. can you please state the "max." software req. (like which is the latest kernel you have tested against).

details about the machine:
- running debian squeeze/sid
- I attach the kernel configuration and the lspci dump.

bye now!
cristian
Attached Files
File Type: txt kernel_config.txt (84.4 KB, 87 views)
File Type: txt pci_config.txt (26.5 KB, 98 views)
const.crist is offline   Reply With Quote
Old 03-20-10, 06:53 PM   #6
Malvineous
Registered User
 
Join Date: Sep 2006
Posts: 17
Default Re: NVIDIA driver on x86_64 freezing X

This sounds very similar to the problem I am experiencing (in this thread, nvidia-bug-report included there.)

The unusual part of Xorg.log is:

Code:
(WW) Mar 20 19:22:33 NVIDIA(0): WAIT (2, 6, 0x8000, 0x00005ca8, 0x000065a0)
(WW) Mar 20 19:22:40 NVIDIA(0): WAIT (1, 6, 0x8000, 0x00005ca8, 0x000065a0)

Backtrace:
Segmentation fault at address 0x7f29ecd38000

Fatal server error:
Caught signal 11 (Segmentation fault). Server aborting
and dmesg:
Code:
NVRM: Xid (0002:00): 6, PE007f 
NVRM: Xid (0002:00): 7, Ch 0000007f M 00001ffc D ffffffff intr ffffffff
NVRM: Xid (0002:00): 26, Ch 0000007f M 00001ffc D ffffffff intr ffffffff
NVRM: Xid (0002:00): 4, Ch 0000007e acquireValue 00000000 dmaPut 00000000 dmaGet 00000000 
NVRM: Xid (0001:00): 6, PE0001
But see the other thread for the full nvidia-bug-report output. The problem seems to occur as soon as you try running two nVidia cards at the same time.
Malvineous is offline   Reply With Quote
Old 03-20-10, 07:41 PM   #7
const.crist
Registered User
 
Join Date: Mar 2010
Posts: 3
Default Re: NVIDIA driver on x86_64 freezing X

Quote:
Originally Posted by const.crist View Post
hi!

same thing happens to me. and my configuration is not that fancy.
basically 2 cards (one geforce 6, the other one geforce 7) and an amd64 processor.

right away after the kernel module gets compiled, the installation crashes, the console
gets garbled with junk. the whole machine freezes and it cannot be accessed with
ssh either. I have tried the --ui=no to no avail; same thing again. at the beggining I have suspected the fb support in the kernel. it is disabled right now. after the machine
reboots, the logs are full of junk. the _same_ problem appears when I try to install 17x drivers.

the cards are o.k. since they have been functioning perfectly with 2.6.28 / 18x driver.

I saw that every driver release has some min. software req. explicitly stated. can you please state the "max." software req. (like which is the latest kernel you have tested against).

details about the machine:
- running debian squeeze/sid
- I attach the kernel configuration and the lspci dump.

bye now!
cristian
hi again!

it is definitely a kernel problem. at least 2.6.3[2-3] have problems with 17x, 19x drivers
(machine _crashes_ even _during_ the installation of nvidia software through the
provided installing scripts). I am not sure though if there is some strange setting in the kernel config that affects so badly the nvidia driver. my configurations are similar
up to the new features introduced by the .32, .33 kernels.

I went back to 2.6.28.2 and both 173.14.25 and 195.36.15 compile and work with xorg 1:7.5+3 from squeeze/sid. 185.18.36/xorg reports some missing symbol and
X won't start.

several more questions:

- what is the semantics of the numbers of the releases? right now I see 17x, 18x, 19x being actively maintained.

- is anyone maintaining a matrix with the versions of the linux kernel + xorg version + nvidia driver version that are supposed to work?

thanks a lot!
bye now!
cristian
const.crist is offline   Reply With Quote
Old 03-22-10, 11:58 AM   #8
Thaorius
Registered User
 
Join Date: Mar 2010
Posts: 5
Default Re: NVIDIA driver on x86_64 freezing X

Sorry for the delay, I've been busy.

You'll find all the relevant system logs, the xorg config and the nvidia-bug-report output attached.

Some new developments:
I tried installing the stock version of the driver 173.X from the debian repositories (I figured, if it's on the debian stable repositories, it must work), which required me to install a kernel image only available on lenny. I downloaded and installed the .deb package, made sure grub.cfg was properly set up and rebooted. I couldn't get the kernel to boot my system (/sbin/init not found in root fs).

Also, while dumping the logs and the bug report, I noticed that an X stacktrace gets printed in the logs apparently after 2 minutes or so.

reading const.crist's post, I'm also interested in a known to work configuration. At this point I could sacrifice my need for the latest stable release of every package and install lenny with a bunch of manually upgraded packages (this is, after all, a developer's computer). Quite frankly, it's incredibly annoying to develop software without the third screen.

Oh, and I got a random crash on the dual head last night; allow me to elaborate:
I was installing debian in 2 VirtualBox OSE virtual machines (at the same time) for an automated deployment test, and all the sudden the system froze while I was typing something in the guest's command line. I tried to switch to tty1 with Ctrl+Alt+F1, and the screen got filled with junk, very similar to the junk I get when installing the 195.X driver (on this case, grey background, every line of the tty filled with flashing E letters that change from white to grey) (VESA text buffer corruption?). A couple seconds later, I started getting a bunch of weird IO errors on screen, right below the junk. At this point, the system was non responsive to keyboard input. SSH access was not possible either. I wasn't able to find anything relevant about this on the system logs, but I don't remember what the IO errors where, so they might be on the attached logs.

Thanks for your time.
Attached Files
File Type: gz nvidia.tar.gz (140.1 KB, 90 views)
Thaorius is offline   Reply With Quote

Old 03-24-10, 01:50 PM   #9
Thaorius
Registered User
 
Join Date: Mar 2010
Posts: 5
Default Re: NVIDIA driver on x86_64 freezing X

Anyone, please? I really need this up and running.
Thaorius is offline   Reply With Quote
Old 03-25-10, 05:37 PM   #10
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: NVIDIA driver on x86_64 freezing X

I noticed this in your dmesg output:

[ 0.410815] PCI: Using MMCONFIG at e0000000 - efffffff

MMCONFIG can be problematic with multi-cpu, multi-gpu, 2.6.x kernel configurations. Try adding "pci=nommconf" to your boot parameters and see if that improves things.

Also, I'd still like to see an nvidia-installer-log demonstrating the crash during installation of the 195.36.15 driver. The one you provided was for the 190.53 version.
danix is offline   Reply With Quote
Old 03-25-10, 08:01 PM   #11
Thaorius
Registered User
 
Join Date: Mar 2010
Posts: 5
Default Re: NVIDIA driver on x86_64 freezing X

About MMCONFIG, I forgot to mention, I had already read somewhere that it may cause problems, and I tested all kernels with nommconfig after failing with the stock config. It makes no difference.

As for the crashing install log, I'll generate one later tonight, once I'm done coding something urgent.
Thaorius is offline   Reply With Quote
Old 03-28-10, 02:22 PM   #12
const.crist
Registered User
 
Join Date: Mar 2010
Posts: 3
Default Re: NVIDIA driver on x86_64 freezing X

Quote:
Originally Posted by danix View Post
I noticed this in your dmesg output:

[ 0.410815] PCI: Using MMCONFIG at e0000000 - efffffff

MMCONFIG can be problematic with multi-cpu, multi-gpu, 2.6.x kernel configurations. Try adding "pci=nommconf" to your boot parameters and see if that improves things.

Also, I'd still like to see an nvidia-installer-log demonstrating the crash during installation of the 195.36.15 driver. The one you provided was for the 190.53 version.
I could do it but, since the _machine_ crashes while this log is being dumped to the fs, after the reboot the file contains rubbish. (lots of '\0' for example). I do not think it is very useful.

is there any other way to debug the installation process?

thanks!
bye now
const.crist is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 08:20 AM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.