On Thu, Sep 28, 2023 at 06:24:49PM +0300, Ville Syrjala wrote:
> From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
>
> Assuming crtc->state is pointing at the correct thing for the
> async flip commit is nonsense. If we had already queued up multiple
> commits this would point at the very lates crtc state even if the
> older commits hadn't even happened yet.
>
> Instead properly stage/arm the event like we do for async flips.
> Since we don't need to arm multiple of these at the same time we
> don't need a list like the normal vblank even processing uses.
>
> Signed-off-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> ---
> drivers/gpu/drm/i915/display/intel_crtc.c | 9 ++++++++-
> drivers/gpu/drm/i915/display/intel_display_irq.c | 9 ++++-----
> drivers/gpu/drm/i915/display/intel_display_types.h | 3 +++
> 3 files changed, 15 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_crtc.c b/drivers/gpu/drm/i915/display/intel_crtc.c
> index 1fd068e6e26c..8a84a31c7b48 100644
> --- a/drivers/gpu/drm/i915/display/intel_crtc.c
> +++ b/drivers/gpu/drm/i915/display/intel_crtc.c
> @@ -553,8 +553,15 @@ void intel_pipe_update_start(struct intel_atomic_state *state,
>
> intel_psr_lock(new_crtc_state);
>
> - if (new_crtc_state->do_async_flip)
> + if (new_crtc_state->do_async_flip) {
> + spin_lock_irq(&crtc->base.dev->event_lock);
> + /* arm the event for the flip done irq handler */
> + crtc->flip_done_event = new_crtc_state->uapi.event;
> + spin_unlock_irq(&crtc->base.dev->event_lock);
> +
> + new_crtc_state->uapi.event = NULL;
> return;
> + }
>
> if (intel_crtc_needs_vblank_work(new_crtc_state))
> intel_crtc_vblank_work_init(new_crtc_state);
> diff --git a/drivers/gpu/drm/i915/display/intel_display_irq.c b/drivers/gpu/drm/i915/display/intel_display_irq.c
> index bff4a76310c0..d3df615f0e48 100644
> --- a/drivers/gpu/drm/i915/display/intel_display_irq.c
> +++ b/drivers/gpu/drm/i915/display/intel_display_irq.c
> @@ -340,16 +340,15 @@ static void flip_done_handler(struct drm_i915_private *i915,
> enum pipe pipe)
> {
> struct intel_crtc *crtc = intel_crtc_for_pipe(i915, pipe);
> - struct drm_crtc_state *crtc_state = crtc->base.state;
> - struct drm_pending_vblank_event *e = crtc_state->event;
> struct drm_device *dev = &i915->drm;
> unsigned long irqflags;
>
> spin_lock_irqsave(&dev->event_lock, irqflags);
>
> - crtc_state->event = NULL;
> -
> - drm_crtc_send_vblank_event(&crtc->base, e);
> + if (crtc->flip_done_event) {
> + drm_crtc_send_vblank_event(&crtc->base, crtc->flip_done_event);
> + crtc->flip_done_event = NULL;
> + }
I just observed an oops here due to e==NULL with the current code.
I *think* I've seen it once before as well. Pstore also caught what
seemed to some kind of spurious DE interrupt, which might explain
the oops. But not really sure what happened as the machine died before
I could poke at it more.
>
> spin_unlock_irqrestore(&dev->event_lock, irqflags);
> }
> diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h b/drivers/gpu/drm/i915/display/intel_display_types.h
> index 8d8b2f8d37a9..a8ae1a25a550 100644
> --- a/drivers/gpu/drm/i915/display/intel_display_types.h
> +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
> @@ -1461,6 +1461,9 @@ struct intel_crtc {
>
> struct intel_crtc_state *config;
>
> + /* armed event for async flip */
> + struct drm_pending_vblank_event *flip_done_event;
> +
> /* Access to these should be protected by dev_priv->irq_lock. */
> bool cpu_fifo_underrun_disabled;
> bool pch_fifo_underrun_disabled;
> --
> 2.41.0
--
Ville Syrjälä
Intel
