[Intel-gfx] [PATCH 2/5] drm/i915/error: capture ringbuffer pointed to by START

[Arun Siluvery]

On 29/01/2016 11:47, Chris Wilson wrote:
> On Thu, Jan 28, 2016 at 07:01:21PM +0000, Arun Siluvery wrote:
>> From: Dave Gordon <david.s.gordon at intel.com>
>>
>> For: VIZ-2021
>> Signed-off-by: Dave Gordon <david.s.gordon at intel.com>
>
> What information does this actually provide over and above the hw is not
> executing the ring we expect? How have you used this, how do you plan
> to?
Most of the times this matches with the ringbuffer we dump but when 
there is inconsistency it helps to know what hw is actually executing as 
opposed to what we expect, otherwise the head, tail values we capture 
and the instructions at those offsets don't make sense. Without this we 
don't have any idea what the HW was doing and what caused hang.

regards
Arun

>
> As it stands adding more fragile loops is just increasing the potential
> for an OOPS in that code, even more so as we can eliminate the current
> dangerous list iteration for extracting the current ctx object.
> -Chris
>