Before the offending commit in msm_atomic_commit_tail wait_flush was called once per frame, after the commit was submitted. After it wait_flush is also called at the beginning to ensure previous potentially async commits are done.
For cmd panels the source of wait_flush is a ping-pong irq notifying a completion. The completion needs to be notified with complete_all so multiple waiting parties (new async committers) can proceed.
Signed-off-by: Iskren Chernev iskren.chernev@gmail.com Suggested-by: Rob Clark robdclark@gmail.com Fixes: 2d99ced787e3d ("drm/msm: async commit support") --- drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c index 0c8f9f88301fa..f5d71b2740793 100644 --- a/drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c +++ b/drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c @@ -1180,7 +1180,7 @@ static void mdp5_crtc_pp_done_irq(struct mdp_irq *irq, uint32_t irqstatus) struct mdp5_crtc *mdp5_crtc = container_of(irq, struct mdp5_crtc, pp_done);
- complete(&mdp5_crtc->pp_completion); + complete_all(&mdp5_crtc->pp_completion); }
static void mdp5_crtc_wait_for_pp_done(struct drm_crtc *crtc)
base-commit: 59fa6a163ffabc1bf25c5e0e33899e268a96d3cc
On Wed, Jan 27, 2021 at 05:24:40PM +0200, Iskren Chernev wrote:
Before the offending commit in msm_atomic_commit_tail wait_flush was called once per frame, after the commit was submitted. After it wait_flush is also called at the beginning to ensure previous potentially async commits are done.
For cmd panels the source of wait_flush is a ping-pong irq notifying a completion. The completion needs to be notified with complete_all so multiple waiting parties (new async committers) can proceed.
Signed-off-by: Iskren Chernev iskren.chernev@gmail.com Suggested-by: Rob Clark robdclark@gmail.com Fixes: 2d99ced787e3d ("drm/msm: async commit support")
Nice job tracking down this fix!
Reviewed-by: Brian Masney masneyb@onstation.org Tested-by: Brian Masney masneyb@onstation.org
Hi Iskren,
On Mittwoch, 27. Jänner 2021 16:24:40 CET Iskren Chernev wrote:
Before the offending commit in msm_atomic_commit_tail wait_flush was called once per frame, after the commit was submitted. After it wait_flush is also called at the beginning to ensure previous potentially async commits are done.
For cmd panels the source of wait_flush is a ping-pong irq notifying a completion. The completion needs to be notified with complete_all so multiple waiting parties (new async committers) can proceed.
Signed-off-by: Iskren Chernev iskren.chernev@gmail.com Suggested-by: Rob Clark robdclark@gmail.com Fixes: 2d99ced787e3d ("drm/msm: async commit support")
I've tested this now on fairphone-fp2 and lge-nexus5-hammerhead, works great!
Tested-by: Luca Weiss luca@z3ntu.xyz
Regards Luca
On 1/27/21 6:24 PM, Iskren Chernev wrote:
Before the offending commit in msm_atomic_commit_tail wait_flush was called once per frame, after the commit was submitted. After it wait_flush is also called at the beginning to ensure previous potentially async commits are done.
For cmd panels the source of wait_flush is a ping-pong irq notifying a completion. The completion needs to be notified with complete_all so multiple waiting parties (new async committers) can proceed.
Signed-off-by: Iskren Chernev iskren.chernev@gmail.com Suggested-by: Rob Clark robdclark@gmail.com Fixes: 2d99ced787e3d ("drm/msm: async commit support")
Tested on msm8974pro samsung-klte, finally got 60 fps instead of 13 in kmscube.
Tested-by: Alexey Minnekhanov alexeymin@postmarketos.org
dri-devel@lists.freedesktop.org
-
Alexey Minnekhanov -
Brian Masney -
Iskren Chernev -
Luca Weiss