[Intel-gfx] [PATCH] drm/i915/guc: In the submission cleanup, do not bail out if there is no execbuf_client

Joonas Lahtinen joonas.lahtinen at linux.intel.com
Tue Feb 14 14:20:04 UTC 2017


On ma, 2017-02-13 at 07:55 -0800, Oscar Mateo wrote:
> 
> On 02/10/2017 04:04 AM, Chris Wilson wrote:
> > 
> > On Thu, Feb 09, 2017 at 02:24:25AM -0800, Oscar Mateo wrote:
> > > 
> > > There is other stuff that potentially needs cleaning, even if we didn't get to the point of
> > > creating an execbuf_client.
> > Just because the allocator doesn't employ onion unwinding?
> > Or is there more to come?
> No, nothing more to come. I just saw this was wrong in passing and 
> decided to send a patch. But I see Joonas is spearheading a bigger 
> cleanup, so maybe this can be fixed there.

I was actually inspired by this change (I had it in my TODO list) which
I completed while going through the code. Would be great if you can do
a similar fixup of the code to include proper onion teardown to
submission_init and make submission_fini without condition checks.

In the guc_client case it brought up quite good fixes and
irregularities in phasing the init and teardown.

Regards, Joonas
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation


More information about the Intel-gfx mailing list