libva strikes again.

I don't see libva related message in your description. Is your APP using hardware video acceleration ?

I have no idea which app caused this GPU hang.

The logs I pasted in the bug description are kernel logs from /var/log/kern.log. Here are all of the relevant lines with timestamps:

Dec 25 00:27:02 hostname kernel: [302308.803013] i915 0000:00:02.0: BAR 6: [??? 0x00000000 flags 0x2] has bogus alignment
Dec 25 00:37:31 hostname kernel: [302937.612941] [drm] stuck on render ring
Dec 25 00:37:31 hostname kernel: [302937.613504] [drm] GPU HANG: ecode 6:0:0x87e8fffd, in Xorg [4626], reason: Ring hung, action: reset
Dec 25 00:37:31 hostname kernel: [302937.613505] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Dec 25 00:37:31 hostname kernel: [302937.613506] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Dec 25 00:37:31 hostname kernel: [302937.613507] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Dec 25 00:37:31 hostname kernel: [302937.613507] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Dec 25 00:37:31 hostname kernel: [302937.613508] [drm] GPU crash dump saved to /sys/class/drm/card0/error
Dec 25 00:37:31 hostname kernel: [302937.615579] drm/i915: Resetting chip after gpu hang

Hi, 
    Will you please try to add the boot option of "i915.enable_ppgtt = 0" and see whether it is helpful?

Thanks

(In reply to Chris Wilson from comment #1)
> libva strikes again.

Hi, Chris

    I checked the GPU commands in GPU_error message. It seems that it is not related with libva.

    Anyway, we can wait whether the boot option of "i915.enable_ppgtt=0" is helpful and decide how to check it.

thanks

(In reply to ykzhao from comment #4)
> Will you please try to add the boot option of "i915.enable_ppgtt = 0"
> and see whether it is helpful?

I've changed /etc/modprobe.d/i915.conf like this:

  --- a/modprobe.d/i915.conf
  +++ b/modprobe.d/i915.conf
  @@ -1,2 +1,2 @@
   # https://wiki.archlinux.org/index.php/Intel#Module-based_Powersaving_Options
  -options i915 i915_enable_rc6=1 i915_enable_fbc=1 lvds_downclock=1
  +options i915 i915_enable_rc6=1 i915_enable_fbc=1 lvds_downclock=1 enable_ppgtt=0

and rebooted my laptop. Is there an easy way to check that the enable_ppgtt option is correctly set to 0?

modinfo only shows me the default, not the current value:

  $ sudo modinfo i915 | grep ppgtt
  parm:           enable_ppgtt:Override PPGTT usage. (-1=auto [default], 0=disabled, 1=aliasing, 2=full) (int)

Speaking of boot options, it looks like "i915_enable_rc6", "i915_enable_fbc" and "lvds_downclock" aren't part of the "modinfo -p i915" output. I guess I should remove them?

The first two however do exist without the "i915_" prefix but maybe I should with the defaults?

Hi, Francois

    You can use the following command to check whether the module parameter is configured?
    > cat /sys/module/i915/parameters/enable_ppgtt
    > cat /sys/module/i915/parameters/enable_rc6 

    All the parameters for I915 driver can be found under the directory of "/sys/module/i915/parameters/".

    Yes. For the parameter of "enable_rc6/enable_fbc" the "i915_" prefix should be removed. You can get it from the output of "modinfo i915".

Thanks
    Yakui

(In reply to ykzhao from comment #5)
> Anyway, we can wait whether the boot option of "i915.enable_ppgtt=0" is
> helpful and decide how to check it.

It doesn't seem to be helpful. I'm running with ppgtt disabled and got two similar GPU HANG in the last few days: bug 97247 and bug 97271.

I suppose we can close this bug and move the discussion over to bug 97271.

Closed it as 'wont fix' as it is not a video specific issue.