[Intel-gfx] [PATCH 02/16] kselftests: Exercise hw-independent mock tests for i915.ko

[Chris Wilson]

On Thu, Dec 08, 2016 at 08:50:17AM -0700, Shuah Khan wrote:
> On 12/07/2016 07:09 AM, Chris Wilson wrote:
> > On Wed, Dec 07, 2016 at 01:58:19PM +0000, Chris Wilson wrote:
> >> Although being a GPU driver most functionality of i915.ko depends upon
> >> real hardware, many of its internal interfaces can be "mocked" and so
> >> tested independently of any hardware. Expanding the test coverage is not
> >> only useful for i915.ko, but should provide some integration tests for
> >> core infrastructure as well.
> >>
> >> Loading i915.ko with mock_selftests=-1 will cause it to execute its mock
> >> tests then fail with -ENOTTY. If the driver is already loaded and bound
> >> to hardware, it requires a few more steps to unbind, and so the simple
> >> preliminary modprobe -r will fail.
> > 
> > I changed the exit condition to return 0 after successfully completing
> > the mock tests (when passed mock_selftests=-1) so modprobe reports
> > success/fail clearly.
> > -Chris
> > 
> 
> Hi Chris,
> 
> Is this patch slated to go in through drm tree with the other patches
> in the series. Just make sure you create .gitignore for any generated
> files that should be ignored by git

If you agree to having it turn up via the drm tree, then yes :)

Is there any guideline on test behaviour I can read? Is there a
particular directory structure you like? I have a few selftests planned
for i915.ko and core drm.ko, and maybe a few other gpus will join in
over time (certainly should write a few for drivers/gpu/drm/vgem/vgem.ko as
well), is selftests/drivers/gpu/(drm.sh, i915.sh, vgem.sh, etc)
suitable?
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre