Hard lockup on startx with gf6800GT on latest nvidia driver and xorg.

Synt4x_3rr0r · 01-09-2012, 04:33 PM

Hello.

I'm running Arch Linux on an older computer that is mostly used as a file server, but recently I was given a Geforce 6800GT AGP that I thought I'd use for running MAME games.

There was a Geforce 4 Ti4200 installed before this with the nouveau driver for xorg since the legacy nvidia driver doesn't support newer xorgs. .xinitrc is configured like this in order to only start the wahcade frontend without a window manager.:

Code:

xset -s off
setxkbmap -layout "se"
wahcade

This worked perfectly fine on the Geforce 4 + nouveau driver.

But, now that I had installed the 6800 I followed Arch's wiki page for installing the official NVIDIA driver. It also explains what you need to do to remove nouveau. I then ran nvidia-xconfig and let it do it's magic.

But when I run startx the only thing that happens is the screen going black, the GPU fan slowing down and then nothing. I can't kill xorg with ctrl+alt+backspace and I can't log in with ssh either. If I'm already logged in with ssh on it, that session just freezes.

And there's absolutely nothing in the logs about it either. The only thing I have is this from when I ran startx through ssh:

Code:

X.Org X Server 1.11.3
Release Date: 2011-12-16
X Protocol Version 11, Revision 0
Build Operating System: Linux 3.1.5-1-ARCH i686 
Current Operating System: Linux negerbulle 3.1.7-1-ARCH #1 SMP PREEMPT Wed Jan 4 08:34:05 UTC 2012 i686
Kernel command line: root=/dev/disk/by-uuid/14a16d06-ed52-4eee-bbe9-d1f3b263de0f ro
Build Date: 17 December 2011  09:38:27AM
 
Current version of pixman: 0.24.0
	Before reporting problems, check http://wiki.x.org
	to make sure that you have the latest version.
Markers: (--) probed, (**) from config file, (==) default setting,
	(++) from command line, (!!) notice, (II) informational,
	(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/Xorg.0.log", Time: Mon Jan  9 22:35:08 2012
(==) Using config file: "/etc/X11/xorg.conf"

And these are the last lines of everything.log if that helps:

Code:

Jan  9 22:31:51 localhost kernel: [   29.483344] eth0: no IPv6 routers present
Jan  9 22:32:43 localhost dbus[810]: [system] Activating service name='org.freedesktop.ConsoleKit' (using servicehelper)
Jan  9 22:32:43 localhost dbus[810]: [system] Activating service name='org.freedesktop.PolicyKit1' (using servicehelper)
Jan  9 22:32:43 localhost polkitd[983]: started daemon version 0.103 using authority implementation `local' version `0.103'
Jan  9 22:32:43 localhost dbus[810]: [system] Successfully activated service 'org.freedesktop.PolicyKit1'
Jan  9 22:32:43 localhost dbus[810]: [system] Successfully activated service 'org.freedesktop.ConsoleKit'

Also here's the boot-up part of everything.log:
http://pastebin.com/eF3TmkAA

I've also noticed this error while running some nvidia applications like nvidia-smi on a tty:

Code:

NVIDIA: could not open the device file /dev/nvidiactl (No such file or directory).
Nvidia-smi has failed because it couldn't communicate with NVIDIA driver. Make sure that latest NVIDIA driver is installed and running.

But googling leads me to believe that has something to do with CUDA, which I'm unsure if I need or not. I do not have any CUDA packages installed anyway.

I'm just completely puzzled. I don't know where to begin because the logs won't tell me anything. The nvidiactl error seems to be the light in my darkness (:P) but I tried to remove the packaged nvidia driver that comes from the arch repos and instead manually install the one from nvidias homepage but that didn't make any difference either. What do I do?

EDIT: Oh and the nvidia kernel module is loaded of course.

EDIT2: Sorry... I completely forgot about the xorg.conf:

Code:

# nvidia-xconfig: X configuration file generated by nvidia-xconfig
# nvidia-xconfig:  version 290.10  (buildmeister@swio-display-x86-rhel47-07.nvidia.com)  Wed Nov 16 20:32:22 PST 2011


Section "ServerLayout"
    Identifier     "Layout0"
    Screen      0  "Screen0" 0 0
    InputDevice    "Keyboard0" "CoreKeyboard"
    InputDevice    "Mouse0" "CorePointer"
EndSection

Section "Files"
EndSection

Section "InputDevice"

    # generated from default
    Identifier     "Mouse0"
    Driver         "mouse"
    Option         "Protocol" "auto"
    Option         "Device" "/dev/psaux"
    Option         "Emulate3Buttons" "no"
    Option         "ZAxisMapping" "4 5"
EndSection

Section "InputDevice"

    # generated from default
    Identifier     "Keyboard0"
    Driver         "kbd"
EndSection

Section "Monitor"

#    HorizSync       28.0 - 33.0
#    VertRefresh     43.0 - 72.0
    Identifier     "Monitor0"
    VendorName     "ViewSonic"
    ModelName      "P227f"
    Option         "DPMS" "0"
EndSection

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
EndSection

Section "Screen"
    Identifier     "Screen0"
    Device         "Device0"
    Monitor        "Monitor0"
    DefaultDepth    24
    Option         "ConnectedMonitor" "CRT"
    Option         "DamageEvents" "1"
#    Option         "TripleBuffer"        "1"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

EDIT3: Ok I tried to boot the computer with both SysRescueCD and Backtrack 5 from USB just now, and both of them give this error when starting Xorg:

Code:

[drm] nouveau 0000:01:00.0: PFIFO_DMA_PUSHER - Ch 0 Get 0x04000000 Put 0x00001094 State 0xc0000000 (err: MEM_FAULT) Push 0x00000000

Googling "[drm] PFIFO_DMA_PUSHER (err: MEM_FAULT) Push 0x00000000" I found this page, but it doesn't seem related to the official NVIDIA drivers.
[Bug 63499] nouveau doesn't work with 'nVidia Corporation C51PV [GeForce 6150]' chipset.

RockDoctor · 01-10-2012, 07:26 AM

Just out of curiosity, have you blacklisted the nouveau driver?

Synt4x_3rr0r · 01-10-2012, 09:43 AM

Yes, the Arch linux package for the nvidia driver adds a file under /etc/modprobe.d/ to disable nouveau. Though when I wrote this thread yesterday I had uninstalled that and was running on the driver from NVIDIA's homepage, so nouveau WAS loaded when I wrote the thread. But I noticed this pretty soon after posting and corrected it. Now I'm running on the arch linux package once again and the problem is still there.

How big is the chance that the GPU is broken somehow? It doesn't display artifacts or anything like that in the terminal, but since I can't even get Xorg to start I can't see if that's the case there as well. And I tried to boot other OSes from USB memory sticks but they also refused to start Xorg, giving me that error I pasted in my last edit.

For what it's worth, this particular graphics card HAS been used on Linux in the past so I know it should work.

EDIT: Ok so I tried with the "nv" driver, and Xorg DOES start when using that. But it appears to hang after a while, though I'm not 100% certain if it does actually hang or not because the Xorg.0.log says something about input devices not having drivers. But when I killed X through SSH, nothing happened to the screen output. Since wahcade is only a static screen until you select a game I can't tell if it actually crashed or not. I'm gonna have to try with something else like LXDE or something. Anyway here's the Xorg.0.log from that session:
http://pastebin.com/GBvAfNVM

Edit again...: Ok so I tried with LXDE, and it does seem to work. The CPU monitor and the clock does indeed update. So it's just an input device issue. So now I know the GPU works right? It has to be a software issue, right?!

RockDoctor · 01-10-2012, 04:23 PM

I don't know if it's relevant, but my desktop PC runs both Fedora and Ubuntu; both use the (default) evdev driver for the mouse and keyboard, and the nouveau driver for video. From my Xorg log:

Code:

[12.197] (II) config/udev: Adding input device Power Button (/dev/input/event1)
[    12.197] (**) Power Button: Applying InputClass "evdev keyboard catchall"
[    12.197] (II) LoadModule: "evdev"
[    12.198] (II) Loading /usr/lib/xorg/modules/input/evdev_drv.so
[    12.199] (II) Module evdev: vendor="X.Org Foundation"
[    12.199]    compiled for 1.10.2, module version = 2.6.0
[    12.199]    Module class: X.Org XInput Driver
[    12.199]    ABI class: X.Org XInput driver, version 12.3
[    12.199] (II) Using input driver 'evdev' for 'Power Button'
[    12.199] (II) Loading /usr/lib/xorg/modules/input/evdev_drv.so
[    12.199] (**) Power Button: always reports core events
[    12.199] (**) Power Button: Device: "/dev/input/event1"
[    12.199] (--) Power Button: Found keys
[    12.199] (II) Power Button: Configuring as keyboard
[    12.199] (**) Option "config_info" "udev:/sys/devices/LNXSYSTM:00/LNXPWRBN:00/input/input1/event1"
[    12.199] (II) XINPUT: Adding extended input device "Power Button" (type: KEYBOARD)
[    12.199] (**) Option "xkb_rules" "evdev"
[    12.199] (**) Option "xkb_model" "pc105"
[    12.199] (**) Option "xkb_layout" "us"
[    12.201] (II) config/udev: Adding input device Video Bus (/dev/input/event3)
[    12.201] (**) Video Bus: Applying InputClass "evdev keyboard catchall"
[    12.201] (II) Using input driver 'evdev' for 'Video Bus'
[    12.201] (II) Loading /usr/lib/xorg/modules/input/evdev_drv.so
[    12.201] (**) Video Bus: always reports core events

What happens if you eliminate the "Input Device" sections of your xorg.conf file - will Xorg default to using evdev?

Synt4x_3rr0r · 01-10-2012, 04:43 PM

I'm not at home right now so I'll have to try that tomorrow. But yes that seems to be the logical explanation for why input devices aren't working.

The question still remains though about the official nvidia kernel module and why it makes my system stall when starting xorg. It probably has something to do with old driver files or configs of nouveau still laying around somewhere, but I uninstalled nouveau completely from the system and put a blacklist on the nouveau module as well...

Though I don't like that "MEM_FAULT" thing in the nouveau error message, as well as this line from my Xorg.0.log:

Code:

[    41.780] (II) NV(0): Bad V_BIOS checksum

Could the vram be faulty or something? But then why does Xorg work with the "nv" driver?