View Single Post
Old 03-25-12, 05:57 PM   #8
Registered User
Join Date: Oct 2004
Posts: 11
Default Re: Random video hangs after upgrading from 290.10 to 295.9 and 295.17

Originally Posted by alan242 View Post
Well, in your original post, your bug report shows there's problems with your X setup. Specifically,

[3235440.663] (WW) Warning, couldn't open module dri
[3235440.663] (II) UnloadModule: "dri"
[3235440.663] (II) Unloading dri
[3235440.663] (EE) Failed to load module "dri" (module does not exist, 0)
[3235440.663] (II) LoadModule: "dri2"
[3235440.663] (WW) Warning, couldn't open module dri2
[3235440.663] (II) UnloadModule: "dri2"
[3235440.663] (II) Unloading dri2
[3235440.663] (EE) Failed to load module "dri2" (module does not exist, 0)
The DRI modules are not used by the nvidia blob driver (except recently to present some VDPAU API naming interface), and are not necessary. I have never loaded the DRI modules and did not have the ISR problems until upgrading to the 295.xx driver series.

The README for the driver installation also indicates that the DRI module should be removed from the X configuration.

Remove the following lines:

    Load "dri"
    Load "GLCore"
[3235442.928] (WW) NVIDIA(0): The EDID for Mitsubishi MEUSPTV10 (DFP-0) contradicts itself:
[3235442.928] (WW) NVIDIA(0): mode "1920x1080" is specified in the EDID; however, the
[3235442.928] (WW) NVIDIA(0): EDID's valid VertRefresh range (59.000-61.000 Hz) would
[3235442.928] (WW) NVIDIA(0): exclude this mode's VertRefresh (30.0 Hz); ignoring
[3235442.928] (WW) NVIDIA(0): VertRefresh check for mode "1920x1080".
I have no problems with the resolution or refresh rate being used. The message here is just an indication that the EDID information that the display is providing is inconsistent with itself. As indicated in the message, the driver is behaving sanely and just ignoring the VertRefresh range which is incorrect.

[3235443.093] (WW) NVIDIA(0): Option "TwinViewXineramaInfoOrder" requested "CRT", but no
[3235443.093] (WW) NVIDIA(0): such display device could be found, or all display devices
[3235443.093] (WW) NVIDIA(0): by that name are currently unavailable.
[3235443.093] (WW) NVIDIA(0): Option "TwinViewXineramaInfoOrder" requested "TV", but no such
[3235443.093] (WW) NVIDIA(0): display device could be found, or all display devices by
[3235443.093] (WW) NVIDIA(0): that name are currently unavailable.
I am not using Xinerama and these messages have no relevance to my setup.

[3235443.094] (EE) NVIDIA: Failed to load module "dri2" (module does not exist, 0)
[3235443.094] (II) NVIDIA(0): The X server will not be able to send the VDPAU driver name to
[3235443.094] (II) NVIDIA(0): libvdpau.
Again, this is the message regarding the naming interface for VDPAU, but the DRI modules are not necessary and the driver functions correctly without loading the DRI extensions.

Once your X setup is cleaned up and there are no more warnings or errors in the log, if you still have the problem, you may need to disable MSI interrupts. Either the motherboard or the video card might be having problems with using that interrupt mechanism. To disable MSI, you'll need to pass the parm NVreg_EnableMSI=0 to the nvidia kernel driver.
I have been using MSI interrupts ever since I began using this video card without problems (until upgrading to 295.xx). I am reporting a regression in the 295.xx series driver. MSI interrupts worked perfectly fine with all driver series prior to 295.xx. I need to upgrade to 295.xx to avoid segmentation faults related to OpenGL multithreading in the 290.xx driver series, that were fixed in the 295.xx series.
PhoenixUT is offline   Reply With Quote