[Intel-gfx] [RESEND PATCH v14 2/2] drm/i915: Put "cooked" vlank counters in frame CRC lines

[Daniel Vetter]

On Mon, Jan 16, 2017 at 10:12:36AM +0100, Tomeu Vizoso wrote:
> On 10 January 2017 at 17:31, Daniel Vetter <daniel at ffwll.ch> wrote:
> > On Tue, Jan 10, 2017 at 05:54:57PM +0200, Ville Syrjälä wrote:
> >> On Tue, Jan 10, 2017 at 02:43:05PM +0100, Tomeu Vizoso wrote:
> >> > Use drm_accurate_vblank_count so we have the full 32 bit to represent
> >> > the frame counter and userspace has a simpler way of knowing when the
> >> > counter wraps around.
> >> >
> >> > Signed-off-by: Tomeu Vizoso <tomeu.vizoso at collabora.com>
> >> > Reviewed-by: Emil Velikov <emil.velikov at collabora.com>
> >> > Reviewed-by: Robert Foss <robert.foss at collabora.com>
> >> > ---
> >> >
> >> >  drivers/gpu/drm/i915/i915_irq.c | 6 +++---
> >> >  1 file changed, 3 insertions(+), 3 deletions(-)
> >> >
> >> > diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> >> > index b9beb5955dae..75fb1f66cc0c 100644
> >> > --- a/drivers/gpu/drm/i915/i915_irq.c
> >> > +++ b/drivers/gpu/drm/i915/i915_irq.c
> >> > @@ -1557,7 +1557,6 @@ static void display_pipe_crc_irq_handler(struct drm_i915_private *dev_priv,
> >> >     struct drm_driver *driver = dev_priv->drm.driver;
> >> >     uint32_t crcs[5];
> >> >     int head, tail;
> >> > -   u32 frame;
> >> >
> >> >     spin_lock(&pipe_crc->lock);
> >> >     if (pipe_crc->source) {
> >> > @@ -1612,8 +1611,9 @@ static void display_pipe_crc_irq_handler(struct drm_i915_private *dev_priv,
> >> >             crcs[2] = crc2;
> >> >             crcs[3] = crc3;
> >> >             crcs[4] = crc4;
> >> > -           frame = driver->get_vblank_counter(&dev_priv->drm, pipe);
> >> > -           drm_crtc_add_crc_entry(&crtc->base, true, frame, crcs);
> >> > +           drm_crtc_add_crc_entry(&crtc->base, true,
> >> > +                                  drm_accurate_vblank_count(&crtc->base),
> >>
> >> My assumption would be that this gets called after the vblank irq
> >> handler, so using the _accurate version seems a bit overkill.
> >
> > Since we're at like v15 of this I figured I'll pull this in, and we can
> > polish this a bit more later. Tomeu, can you pls do that follow-up patch
> > and get Ville to review+merge it.
> 
> At least on the SKL and SNB I have here, the -sequence subtests in
> kms_pipe_crc_basic fail if I replace the call to
> drm_accurate_vblank_count with drm_crtc_vblank_count.
> 
> Any ideas on why this could be?

No idea at all, on a quick check things seem ordered correctly. Can you
pls paste how the test falls over?
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch