[Intel-gfx] [PATCH] igt/drm_read: Abuse read(drm)

[Chris Wilson]

On Fri, Dec 05, 2014 at 03:46:54PM +0100, Daniel Vetter wrote:
> On Fri, Dec 05, 2014 at 01:14:40PM +0000, Chris Wilson wrote:
> > +static void assert_empty(int fd)
> > +{
> > +	struct pollfd pfd = {fd, POLLIN };
> > +	igt_assert(poll(&pfd, 1, 0) == 0);
> > +}
> > +
> > +static void generate_event(int fd)
> > +{
> > +	union drm_wait_vblank vbl;
> > +
> > +	/* We assume that pipe 0 is running */
> 
> This assumption falls short on gen2/3 with lvds only, and might fall short
> if a previous test wreaked havoc. Also dpms tends to foul up kms tests,
> too. I think we do need the full-blown modeset dance here unfortunately.

I actually hoped that queuing a .relative=0 vblank event the kernel would
have seen that it was a no-op and just put the event into the queue, but
it still does the vblank_get/_put and so requires the active pipe.

Oh well, it's not too complicated to setup a single pipe for my purpose.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre