On 31 March 2016 at 03:25, Mark yao mark.yao@rock-chips.com wrote:

On 2016年03月30日 21:48, Tomeu Vizoso wrote:

...

As per the docs, atomic_commit should return -EBUSY "if an asycnhronous updated is requested and there is an earlier updated pending".

Also wait for the pending event to complete when a sync update is requested.

Signed-off-by: Tomeu Vizoso tomeu.vizoso@collabora.com

drivers/gpu/drm/rockchip/rockchip_drm_drv.h | 1 + drivers/gpu/drm/rockchip/rockchip_drm_fb.c | 14 +++++++++++++- drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 6 ++++++ 3 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h index 3529f692edb8..37952eefd33d 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h @@ -69,6 +69,7 @@ int rockchip_register_crtc_funcs(struct drm_crtc *crtc, void rockchip_unregister_crtc_funcs(struct drm_crtc *crtc); int rockchip_drm_crtc_mode_config(struct drm_crtc *crtc, int connector_type, int out_mode); +bool rockchip_drm_crtc_has_pending_event(struct drm_crtc *crtc); int rockchip_drm_dma_attach_device(struct drm_device *drm_dev, struct device *dev); void rockchip_drm_dma_detach_device(struct drm_device *drm_dev, diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_fb.c b/drivers/gpu/drm/rockchip/rockchip_drm_fb.c index 3b8f652698f8..028fd2f8f47b 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_fb.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_fb.c @@ -280,7 +280,19 @@ int rockchip_drm_atomic_commit(struct drm_device *dev, { struct rockchip_drm_private *private = dev->dev_private; struct rockchip_atomic_commit *commit = &private->commit;

•   int ret;
  

•   struct drm_crtc_state *crtc_state;
  

•   struct drm_crtc *crtc;
  

•   int i, ret;
  

•   for_each_crtc_in_state(state, crtc, crtc_state, i) {
  

•           if (crtc->state->event ||
  

•               rockchip_drm_crtc_has_pending_event(crtc)) {
  

•                   if (async)
  

•                           return -EBUSY;
  

•                   else
  

•                           rockchip_crtc_wait_for_update(crtc);
  

...

•           }
  

•   }
    ret = drm_atomic_helper_prepare_planes(dev, state);
    if (ret)
  

diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c index e2118e62345b..5240bc8fe088 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c @@ -848,6 +848,12 @@ int rockchip_drm_crtc_mode_config(struct drm_crtc *crtc, } EXPORT_SYMBOL_GPL(rockchip_drm_crtc_mode_config); +bool rockchip_drm_crtc_has_pending_event(struct drm_crtc *crtc) +{

•   return to_vop(crtc)->event;
  

+} +EXPORT_SYMBOL_GPL(rockchip_drm_crtc_has_pending_event);

No need to do EXPORT_SYMBOL_GPL, rockchip_drm_vop.c and rockchip_drm_fb.c will build into same object.

...

static int vop_crtc_enable_vblank(struct drm_crtc *crtc) { struct vop *vop = to_vop(crtc);

Hi Tomeu

Since rockchip atomic async use "serialize outstanding asynchronous commits", async commit will wait on "flush_work(&commit->work);" to ensure last pending event is done, so vop->event should be NULL after flush. if not, it should be a BUG.

I missed that, thanks.

Anyway, serialize async commits is not a good way, but I don't think your patch is better way.

Per-CRTC workqueues should improve that, but I'm right now more interested in correctness than performance.

I think maybe we can check the "commit->work" status(finish or running) to judge return BUSY or not, instead of flush it.

That sounds better indeed.

Thanks,

Tomeu

Thanks.

-- Ｍark Yao

---

dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel