[Intel-gfx] [PATCH] drm/i915: Record batch buffer following GPU error

[2points.mailinglist at gmx.org]

Hi Chris,

tried to gather some debugging output with this patch, but came upon the following problem: The moment an error is to be recorded, the kernel panics. From what I gather from the backtrace, it seems to crash in clone_obj, somewhere along kmap_atomic, probably.

You can find the dmesg output up to everything dying noisily here:
http://pastebin.ca/1793058

I started from the drm-intel git repository, at commit 0d0e2fee9d6f03a481fa213c3db5171c66e7f7cc, drm/i915: add dynamic performance control support for Ironlake, and applied the following additional patches:

drm/i915: Increase fb alignment to 64k (to avoid that page error at module load)
drm/i915: Selectively enable self-reclaim (for signature changes to  i915_gem_object_get_pages; tried with and without, no changes though)
drm/i915: Record batch buffer following GPU error

Best regards,
 Moritz


-------- Original message --------
> Date: Mon,  8 Feb 2010 11:01:01 +0000
> From: Chris Wilson <chris at chris-wilson.co.uk>
> To: intel-gfx at lists.freedesktop.org
> Subject: [Intel-gfx] [PATCH] drm/i915: Record batch buffer following GPU	error

> In order to improve our diagnostic capabilities following a GPU hang
> and subsequent reset, we need to record the batch buffer that triggered
> the error. We assume that the current batch buffer, plus a few details
> about what else is on the active list, will be sufficient -- at the very
> least an improvement over nothing.
> 
> v2: Avoid excessive work under spinlocks.
> v3: Include ringbuffer for latter analysis.
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>