[Intel-gfx] [PATCH] drm/i915: Reject non-canonical rotations

Daniel Vetter daniel at ffwll.ch
Wed Mar 23 08:58:31 UTC 2016

On Tue, Mar 22, 2016 at 02:14:38PM +0000, Matthew Auld wrote:
> Hi Daniel,
> >> I thought we do normalize this somewhere.
> I did write an i-g-t test which submits such a rotation value and it
> is not rejected.

Normalize = the drm core makes sure drivers don't see all the
combinations, but only canonical values. Userspace can still submit values
with too many bits set. I'm not sure we want (or can, it's ABI) change

> >> Your patch lacks motivation
> As in I haven't properly conveyed the motivation behind the patch in
> the commit message?
> >> Yes I can usually guess when
> it's due to static analyzer checks, but you need to explain that. And you
> need to explain what exactly the analyzer is complaining about.
> erm, no static analyser, for this patch or any prior, promise, but duly noted ;)
> Joonas actually suggested this patch, and some of the preceding ones
> as beginner tasks for me.

Oh surprising, spotting all these random things all over tends to be stuff
only static analyzers manage ;-) Patch still needs some motivation, since
if your igt passes and the driver behaves correctly it's all fine.
Daniel Vetter
Software Engineer, Intel Corporation

More information about the Intel-gfx mailing list