[Intel-gfx] [RFC PATCH 00/12] AubCrash

Lionel Landwerlin lionel.g.landwerlin at intel.com
Fri Oct 27 20:17:00 UTC 2017


On 27/10/17 19:31, Oscar Mateo wrote:
>
>
> On 10/27/2017 11:20 AM, Chris Wilson wrote:
>> Quoting Oscar Mateo (2017-10-27 19:01:03)
>>> AubCrash is a companion to i915_gpu_error. It gives us the 
>>> possibility to
>>> dump an AUB file that describes the state of the system at the point of
>>> the crash (GTTs, contexts, BBs, BOs, etc...). Being an AUB file, it 
>>> can be
>>> used by a number of already existing tools (graphical AUB file 
>>> browsers,
>>> simulators, emulators, etc...) that facilitate debugging (an 
>>> improvement
>>> over the current text-based crash dump).
>> Serious question: why both? If the aub is useful for pm debugging, it
>> should be either a replacement for sysfs/error or an addendum.
>>
>> I expect tools to use the new format as well.
>> -Chris
>
> That's why I sent this as an RFC, to get suggestions :)
>
> There are things an AUB file cannot replace (fences, seqnos, error 
> registers, etc...). It could still be used as an addendum, where the 
> things that go into the AUB do not go into sysfs/error anymore (maybe 
> with a message to look for the rest in the AUB file?). Or even better: 
> I could try to provide the remaining text inside the AUB as comments, 
> so that only one file is required...
>
> As for tools, I still need to test how compatible Mesas's Aubinator is 
> at the moment. There a number of non-opensource tools as well, but I 
> have to discover which ones are publicly available outside of Intel 
> before I provide a list here. 

I don't think there is any problem having yet another tool for a new 
format as long as it brings some interesting bits.
When aubinator_error_decode was introduced Chris actually brought up the 
idea of a new format to make parsing of the error state a bit easier.


More information about the Intel-gfx mailing list