drm/amdgpu: Provide page_flip_target hook

Now we can program a flip during a vertical blank period, if it's the
one targeted by the flip (or a later one). This allows simplifying
amdgpu_flip_work_func considerably.

agd: update dce_virtual.c as well.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -730,10 +730,11 @@ void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
  */
 
 struct amdgpu_flip_work {
-	struct work_struct		flip_work;
+	struct delayed_work		flip_work;
 	struct work_struct		unpin_work;
 	struct amdgpu_device		*adev;
 	int				crtc_id;
+	u32				target_vblank;
 	uint64_t			base;
 	struct drm_pending_vblank_event *event;
 	struct amdgpu_bo		*old_rbo;

drivers/gpu/drm/amd/amdgpu/amdgpu_display.c

@@ -41,7 +41,7 @@ static void amdgpu_flip_callback(struct fence *f, struct fence_cb *cb)
 		container_of(cb, struct amdgpu_flip_work, cb);
 
 	fence_put(f);
-	schedule_work(&work->flip_work);
+	schedule_work(&work->flip_work.work);
 }
 
 static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work,
@@ -63,16 +63,17 @@ static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work,
 
 static void amdgpu_flip_work_func(struct work_struct *__work)
 {
+	struct delayed_work *delayed_work =
+		container_of(__work, struct delayed_work, work);
 	struct amdgpu_flip_work *work =
-		container_of(__work, struct amdgpu_flip_work, flip_work);
+		container_of(delayed_work, struct amdgpu_flip_work, flip_work);
 	struct amdgpu_device *adev = work->adev;
 	struct amdgpu_crtc *amdgpuCrtc = adev->mode_info.crtcs[work->crtc_id];
 
 	struct drm_crtc *crtc = &amdgpuCrtc->base;
 	unsigned long flags;
-	unsigned i, repcnt = 4;
-	int vpos, hpos, stat, min_udelay = 0;
-	struct drm_vblank_crtc *vblank = &crtc->dev->vblank[work->crtc_id];
+	unsigned i;
+	int vpos, hpos;
 
 	if (amdgpu_flip_handle_fence(work, &work->excl))
 		return;
@@ -81,55 +82,23 @@ static void amdgpu_flip_work_func(struct work_struct *__work)
 		if (amdgpu_flip_handle_fence(work, &work->shared[i]))
 			return;
 
-	/* We borrow the event spin lock for protecting flip_status */
-	spin_lock_irqsave(&crtc->dev->event_lock, flags);
-
-	/* If this happens to execute within the "virtually extended" vblank
-	 * interval before the start of the real vblank interval then it needs
-	 * to delay programming the mmio flip until the real vblank is entered.
-	 * This prevents completing a flip too early due to the way we fudge
-	 * our vblank counter and vblank timestamps in order to work around the
-	 * problem that the hw fires vblank interrupts before actual start of
-	 * vblank (when line buffer refilling is done for a frame). It
-	 * complements the fudging logic in amdgpu_get_crtc_scanoutpos() for
-	 * timestamping and amdgpu_get_vblank_counter_kms() for vblank counts.
-	 *
-	 * In practice this won't execute very often unless on very fast
-	 * machines because the time window for this to happen is very small.
+	/* Wait until we're out of the vertical blank period before the one
+	 * targeted by the flip
 	 */
-	while (amdgpuCrtc->enabled && --repcnt) {
-		/* GET_DISTANCE_TO_VBLANKSTART returns distance to real vblank
-		 * start in hpos, and to the "fudged earlier" vblank start in
-		 * vpos.
-		 */
-		stat = amdgpu_get_crtc_scanoutpos(adev->ddev, work->crtc_id,
-						  GET_DISTANCE_TO_VBLANKSTART,
-						  &vpos, &hpos, NULL, NULL,
-						  &crtc->hwmode);
-
-		if ((stat & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE)) !=
-		    (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE) ||
-		    !(vpos >= 0 && hpos <= 0))
-			break;
-
-		/* Sleep at least until estimated real start of hw vblank */
-		min_udelay = (-hpos + 1) * max(vblank->linedur_ns / 1000, 5);
-		if (min_udelay > vblank->framedur_ns / 2000) {
-			/* Don't wait ridiculously long - something is wrong */
-			repcnt = 0;
-			break;
-		}
-		spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
-		usleep_range(min_udelay, 2 * min_udelay);
-		spin_lock_irqsave(&crtc->dev->event_lock, flags);
+	if (amdgpuCrtc->enabled &&
+	    (amdgpu_get_crtc_scanoutpos(adev->ddev, work->crtc_id, 0,
+					&vpos, &hpos, NULL, NULL,
+					&crtc->hwmode)
+	     & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) ==
+	    (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) &&
+	    (int)(work->target_vblank -
+		  amdgpu_get_vblank_counter_kms(adev->ddev, amdgpuCrtc->crtc_id)) > 0) {
+		schedule_delayed_work(&work->flip_work, usecs_to_jiffies(1000));
+		return;
 	}
 
-	if (!repcnt)
-		DRM_DEBUG_DRIVER("Delay problem on crtc %d: min_udelay %d, "
-				 "framedur %d, linedur %d, stat %d, vpos %d, "
-				 "hpos %d\n", work->crtc_id, min_udelay,
-				 vblank->framedur_ns / 1000,
-				 vblank->linedur_ns / 1000, stat, vpos, hpos);
+	/* We borrow the event spin lock for protecting flip_status */
+	spin_lock_irqsave(&crtc->dev->event_lock, flags);
 
 	/* Do the flip (mmio) */
 	adev->mode_info.funcs->page_flip(adev, work->crtc_id, work->base, work->async);
@@ -169,10 +138,10 @@ static void amdgpu_unpin_work_func(struct work_struct *__work)
 	kfree(work);
 }
 
-int amdgpu_crtc_page_flip(struct drm_crtc *crtc,
-			  struct drm_framebuffer *fb,
-			  struct drm_pending_vblank_event *event,
-			  uint32_t page_flip_flags)
+int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc,
+				 struct drm_framebuffer *fb,
+				 struct drm_pending_vblank_event *event,
+				 uint32_t page_flip_flags, uint32_t target)
 {
 	struct drm_device *dev = crtc->dev;
 	struct amdgpu_device *adev = dev->dev_private;
@@ -191,7 +160,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc,
 	if (work == NULL)
 		return -ENOMEM;
 
-	INIT_WORK(&work->flip_work, amdgpu_flip_work_func);
+	INIT_DELAYED_WORK(&work->flip_work, amdgpu_flip_work_func);
 	INIT_WORK(&work->unpin_work, amdgpu_unpin_work_func);
 
 	work->event = event;
@@ -237,12 +206,8 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc,
 	amdgpu_bo_unreserve(new_rbo);
 
 	work->base = base;
-
-	r = drm_crtc_vblank_get(crtc);
-	if (r) {
-		DRM_ERROR("failed to get vblank before flip\n");
-		goto pflip_cleanup;
-	}
+	work->target_vblank = target - drm_crtc_vblank_count(crtc) +
+		amdgpu_get_vblank_counter_kms(dev, work->crtc_id);
 
 	/* we borrow the event spin lock for protecting flip_wrok */
 	spin_lock_irqsave(&crtc->dev->event_lock, flags);
@@ -250,7 +215,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc,
 		DRM_DEBUG_DRIVER("flip queue: crtc already busy\n");
 		spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
 		r = -EBUSY;
-		goto vblank_cleanup;
+		goto pflip_cleanup;
 	}
 
 	amdgpu_crtc->pflip_status = AMDGPU_FLIP_PENDING;
@@ -262,12 +227,9 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc,
 	/* update crtc fb */
 	crtc->primary->fb = fb;
 	spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
-	amdgpu_flip_work_func(&work->flip_work);
+	amdgpu_flip_work_func(&work->flip_work.work);
 	return 0;
 
-vblank_cleanup:
-	drm_crtc_vblank_put(crtc);
-
 pflip_cleanup:
 	if (unlikely(amdgpu_bo_reserve(new_rbo, false) != 0)) {
 		DRM_ERROR("failed to reserve new rbo in error path\n");

drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h

@@ -591,10 +591,10 @@ int amdgpu_align_pitch(struct amdgpu_device *adev, int width, int bpp, bool tile
 void amdgpu_print_display_setup(struct drm_device *dev);
 int amdgpu_modeset_create_props(struct amdgpu_device *adev);
 int amdgpu_crtc_set_config(struct drm_mode_set *set);
-int amdgpu_crtc_page_flip(struct drm_crtc *crtc,
-			  struct drm_framebuffer *fb,
-			  struct drm_pending_vblank_event *event,
-			  uint32_t page_flip_flags);
+int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc,
+				 struct drm_framebuffer *fb,
+				 struct drm_pending_vblank_event *event,
+				 uint32_t page_flip_flags, uint32_t target);
 extern const struct drm_mode_config_funcs amdgpu_mode_funcs;
 
 #endif

drivers/gpu/drm/amd/amdgpu/dce_v10_0.c

@@ -2737,7 +2737,7 @@ static const struct drm_crtc_funcs dce_v10_0_crtc_funcs = {
 	.gamma_set = dce_v10_0_crtc_gamma_set,
 	.set_config = amdgpu_crtc_set_config,
 	.destroy = dce_v10_0_crtc_destroy,
-	.page_flip = amdgpu_crtc_page_flip,
+	.page_flip_target = amdgpu_crtc_page_flip_target,
 };
 
 static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode)

drivers/gpu/drm/amd/amdgpu/dce_v11_0.c

@@ -2755,7 +2755,7 @@ static const struct drm_crtc_funcs dce_v11_0_crtc_funcs = {
 	.gamma_set = dce_v11_0_crtc_gamma_set,
 	.set_config = amdgpu_crtc_set_config,
 	.destroy = dce_v11_0_crtc_destroy,
-	.page_flip = amdgpu_crtc_page_flip,
+	.page_flip_target = amdgpu_crtc_page_flip_target,
 };
 
 static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode)

drivers/gpu/drm/amd/amdgpu/dce_v8_0.c

@@ -2598,7 +2598,7 @@ static const struct drm_crtc_funcs dce_v8_0_crtc_funcs = {
 	.gamma_set = dce_v8_0_crtc_gamma_set,
 	.set_config = amdgpu_crtc_set_config,
 	.destroy = dce_v8_0_crtc_destroy,
-	.page_flip = amdgpu_crtc_page_flip,
+	.page_flip_target = amdgpu_crtc_page_flip_target,
 };
 
 static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode)

drivers/gpu/drm/amd/amdgpu/dce_virtual.c

@@ -188,7 +188,7 @@ static const struct drm_crtc_funcs dce_virtual_crtc_funcs = {
 	.gamma_set = dce_virtual_crtc_gamma_set,
 	.set_config = amdgpu_crtc_set_config,
 	.destroy = dce_virtual_crtc_destroy,
-	.page_flip = amdgpu_crtc_page_flip,
+	.page_flip_target = amdgpu_crtc_page_flip_target,
 };
 
 static void dce_virtual_crtc_dpms(struct drm_crtc *crtc, int mode)