[Intel-gfx] GM45: Some strange GPU hang while trying to playback using xv output

[Chris Wilson]

On Wed, 13 Apr 2011 00:22:08 +0300, Dmitry Nezhevenko <dion at inhex.net> wrote:
> Hi,
> 
> This is third time I've catched this.
> [363445.000069] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
> elapsed... GPU hung
> [363445.001196] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request
> returns -11 (awaiting 3712730 at 3712658, next 3712731)
> [363445.504076] [drm:i915_reset] *ERROR* Failed to reset chip.
> 
> It is not easily reproducible. For example my current uptime is ~7 days with a
> lot of s2ram cycles. I've just tried to play some movie using mplayer (with -vo
> xv). And got just black window instead of video with working sound. I'm able to
> quit playing, continue work. However some of UI elements are now painted using
> wrong colors (like yellow test in FF instead of black).
> 
> I'm attaching i915_error_state from such situation. Any ideas?

> 0x0a39f028:      0x61010004: STATE_BASE_ADDRESS
> 0x0a39f02c:      0x00000001:    general state base address 0x00000000
> 0x0a39f030:      0x0a2f9001:    surface state base address 0x0a2f9000
> 0x0a39f034:      0x00000001:    indirect state base address 0x00000000
> 0x0a39f038:      0x10000001:    general state upper bound 0x10000000
> 0x0a39f03c:      0x10000001:    indirect state upper bound 0x10000000

You hit a dormant bug in xf86-video-intel, fixed with

commit 23f9b14df7c102c1036134835dd5d1a508059858
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date:   Sat Feb 12 10:42:34 2011 +0000

    i965: Remove broken maximum base addresses from video

-Chris
-- 
Chris Wilson, Intel Open Source Technology Centre