[PATCH v2 0/3] Selective Fetch and async flip

Intel-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed

* [PATCH v2 0/3] Selective Fetch and async flip
@ 2025-12-01 11:32 Jouni Högander
  2025-12-01 11:32 ` [PATCH v2 1/3] drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for PSR Jouni Högander
                   ` (3 more replies)
  0 siblings, 4 replies; 5+ messages in thread
From: Jouni Högander @ 2025-12-01 11:32 UTC (permalink / raw)
  To: intel-gfx, intel-xe; +Cc: Jouni Högander

This patch set contains fixes for Selective Fetch async flip
sequences. On async flip selective fetch is choosing full ftrame
update. Also subsequent flip/update is still using full frame update
to ensure plane with pending async flip is not taken in to selective
fetch/update.

Jouni Högander (3):
  drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for
    PSR
  drm/i915/psr: Perform full frame update on async flip
  drm/i915/psr: Allow async flip when Selective Fetch enabled

 drivers/gpu/drm/i915/display/intel_display.c |  8 ---
 drivers/gpu/drm/i915/display/intel_plane.c   | 10 ++-
 drivers/gpu/drm/i915/display/intel_psr.c     | 72 +++++++++++---------
 3 files changed, 49 insertions(+), 41 deletions(-)

-- 
2.43.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH v2 1/3] drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for PSR
  2025-12-01 11:32 [PATCH v2 0/3] Selective Fetch and async flip Jouni Högander
@ 2025-12-01 11:32 ` Jouni Högander
  2025-12-01 11:32 ` [PATCH v2 2/3] drm/i915/psr: Perform full frame update on async flip Jouni Högander
                   ` (2 subsequent siblings)
  3 siblings, 0 replies; 5+ messages in thread
From: Jouni Högander @ 2025-12-01 11:32 UTC (permalink / raw)
  To: intel-gfx, intel-xe; +Cc: Jouni Högander

Currently plane id bit is set in crtc_state->async_flip_planes only when
async flip toggle workaround is needed. We want to utilize
crtc_state->async_flip_planes further in Selective Fetch calculation.

Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
---
 drivers/gpu/drm/i915/display/intel_plane.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/display/intel_plane.c b/drivers/gpu/drm/i915/display/intel_plane.c
index a7fec5ba6ac0..c7df72239e64 100644
--- a/drivers/gpu/drm/i915/display/intel_plane.c
+++ b/drivers/gpu/drm/i915/display/intel_plane.c
@@ -599,8 +599,7 @@ static int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_cr
 	if (intel_plane_do_async_flip(plane, old_crtc_state, new_crtc_state)) {
 		new_crtc_state->do_async_flip = true;
 		new_crtc_state->async_flip_planes |= BIT(plane->id);
-	} else if (plane->need_async_flip_toggle_wa &&
-		   new_crtc_state->uapi.async_flip) {
+	} else if (new_crtc_state->uapi.async_flip) {
 		/*
 		 * On platforms with double buffered async flip bit we
 		 * set the bit already one frame early during the sync
@@ -608,6 +607,13 @@ static int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_cr
 		 * hardware will therefore be ready to perform a real
 		 * async flip during the next commit, without having
 		 * to wait yet another frame for the bit to latch.
+		 *
+		 * async_flip_planes bitmask is also used by selective
+		 * fetch calculation to continue full frame updates as
+		 * long as there may be pending async flip on any
+		 * plane which is part of selective
+		 * update. I.e. old_crtc_state->async_flip_planes &
+		 * BIT(<plane in su area>->id).
 		 */
 		new_crtc_state->async_flip_planes |= BIT(plane->id);
 	}
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH v2 2/3] drm/i915/psr: Perform full frame update on async flip
  2025-12-01 11:32 [PATCH v2 0/3] Selective Fetch and async flip Jouni Högander
  2025-12-01 11:32 ` [PATCH v2 1/3] drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for PSR Jouni Högander
@ 2025-12-01 11:32 ` Jouni Högander
  2025-12-01 11:32 ` [PATCH v2 3/3] drm/i915/psr: Allow async flip when Selective Fetch enabled Jouni Högander
  2025-12-01 11:36 ` ✗ Fi.CI.BUILD: failure for Selective Fetch and async flip (rev2) Patchwork
  3 siblings, 0 replies; 5+ messages in thread
From: Jouni Högander @ 2025-12-01 11:32 UTC (permalink / raw)
  To: intel-gfx, intel-xe; +Cc: Jouni Högander

According to bspec selective fetch is not supported with async flips and
instructing full frame update on async flip.

v2:
  - check also crtc_state->async_flip_planes in
    psr2_sel_fetch_plane_state_supported

Bspec: 55229
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
---
 drivers/gpu/drm/i915/display/intel_psr.c | 72 ++++++++++++++----------
 1 file changed, 41 insertions(+), 31 deletions(-)

diff --git a/drivers/gpu/drm/i915/display/intel_psr.c b/drivers/gpu/drm/i915/display/intel_psr.c
index 08bca4573974..c99d875bbf16 100644
--- a/drivers/gpu/drm/i915/display/intel_psr.c
+++ b/drivers/gpu/drm/i915/display/intel_psr.c
@@ -2728,13 +2728,20 @@ intel_psr2_sel_fetch_et_alignment(struct intel_atomic_state *state,
  * Plane scaling and rotation is not supported by selective fetch and both
  * properties can change without a modeset, so need to be check at every
  * atomic commit.
+ *
+ * If plane was having async flip previously we can't use selective
+ * fetch as we don't know if the flip is completed.
  */
-static bool psr2_sel_fetch_plane_state_supported(const struct intel_plane_state *plane_state)
+static bool psr2_sel_fetch_plane_state_supported(const struct intel_crtc_state *old_crtc_state,
+						 const struct intel_plane_state *plane_state)
 {
+	struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
+
 	if (plane_state->uapi.dst.y1 < 0 ||
 	    plane_state->uapi.dst.x1 < 0 ||
 	    plane_state->scaler_id >= 0 ||
-	    plane_state->uapi.rotation != DRM_MODE_ROTATE_0)
+	    plane_state->uapi.rotation != DRM_MODE_ROTATE_0 ||
+	    old_crtc_state->async_flip_planes & plane->id)
 		return false;
 
 	return true;
@@ -2749,7 +2756,8 @@ static bool psr2_sel_fetch_plane_state_supported(const struct intel_plane_state
  */
 static bool psr2_sel_fetch_pipe_state_supported(const struct intel_crtc_state *crtc_state)
 {
-	if (crtc_state->scaler_state.scaler_id >= 0)
+	if (crtc_state->scaler_state.scaler_id >= 0 ||
+	    crtc_state->uapi.async_flip)
 		return false;
 
 	return true;
@@ -2808,24 +2816,25 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 				struct intel_crtc *crtc)
 {
 	struct intel_display *display = to_intel_display(state);
-	struct intel_crtc_state *crtc_state = intel_atomic_get_new_crtc_state(state, crtc);
+	struct intel_crtc_state *new_crtc_state = intel_atomic_get_new_crtc_state(state, crtc);
+	struct intel_crtc_state *old_crtc_state = intel_atomic_get_new_crtc_state(state, crtc);
 	struct intel_plane_state *new_plane_state, *old_plane_state;
 	struct intel_plane *plane;
 	bool full_update = false, cursor_in_su_area = false;
 	int i, ret;
 
-	if (!crtc_state->enable_psr2_sel_fetch)
+	if (!new_crtc_state->enable_psr2_sel_fetch)
 		return 0;
 
-	if (!psr2_sel_fetch_pipe_state_supported(crtc_state)) {
+	if (!psr2_sel_fetch_pipe_state_supported(new_crtc_state)) {
 		full_update = true;
 		goto skip_sel_fetch_set_loop;
 	}
 
-	crtc_state->psr2_su_area.x1 = 0;
-	crtc_state->psr2_su_area.y1 = -1;
-	crtc_state->psr2_su_area.x2 = drm_rect_width(&crtc_state->pipe_src);
-	crtc_state->psr2_su_area.y2 = -1;
+	new_crtc_state->psr2_su_area.x1 = 0;
+	new_crtc_state->psr2_su_area.y1 = -1;
+	new_crtc_state->psr2_su_area.x2 = drm_rect_width(&new_crtc_state->pipe_src);
+	new_crtc_state->psr2_su_area.y2 = -1;
 
 	/*
 	 * Calculate minimal selective fetch area of each plane and calculate
@@ -2838,14 +2847,14 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 		struct drm_rect src, damaged_area = { .x1 = 0, .y1 = -1,
 						      .x2 = INT_MAX };
 
-		if (new_plane_state->uapi.crtc != crtc_state->uapi.crtc)
+		if (new_plane_state->uapi.crtc != new_crtc_state->uapi.crtc)
 			continue;
 
 		if (!new_plane_state->uapi.visible &&
 		    !old_plane_state->uapi.visible)
 			continue;
 
-		if (!psr2_sel_fetch_plane_state_supported(new_plane_state)) {
+		if (!psr2_sel_fetch_plane_state_supported(old_crtc_state, new_plane_state)) {
 			full_update = true;
 			break;
 		}
@@ -2861,23 +2870,23 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 			if (old_plane_state->uapi.visible) {
 				damaged_area.y1 = old_plane_state->uapi.dst.y1;
 				damaged_area.y2 = old_plane_state->uapi.dst.y2;
-				clip_area_update(&crtc_state->psr2_su_area, &damaged_area,
-						 &crtc_state->pipe_src);
+				clip_area_update(&new_crtc_state->psr2_su_area, &damaged_area,
+						 &new_crtc_state->pipe_src);
 			}
 
 			if (new_plane_state->uapi.visible) {
 				damaged_area.y1 = new_plane_state->uapi.dst.y1;
 				damaged_area.y2 = new_plane_state->uapi.dst.y2;
-				clip_area_update(&crtc_state->psr2_su_area, &damaged_area,
-						 &crtc_state->pipe_src);
+				clip_area_update(&new_crtc_state->psr2_su_area, &damaged_area,
+						 &new_crtc_state->pipe_src);
 			}
 			continue;
 		} else if (new_plane_state->uapi.alpha != old_plane_state->uapi.alpha) {
 			/* If alpha changed mark the whole plane area as damaged */
 			damaged_area.y1 = new_plane_state->uapi.dst.y1;
 			damaged_area.y2 = new_plane_state->uapi.dst.y2;
-			clip_area_update(&crtc_state->psr2_su_area, &damaged_area,
-					 &crtc_state->pipe_src);
+			clip_area_update(&new_crtc_state->psr2_su_area, &damaged_area,
+					 &new_crtc_state->pipe_src);
 			continue;
 		}
 
@@ -2893,7 +2902,8 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 		damaged_area.x1 += new_plane_state->uapi.dst.x1 - src.x1;
 		damaged_area.x2 += new_plane_state->uapi.dst.x1 - src.x1;
 
-		clip_area_update(&crtc_state->psr2_su_area, &damaged_area, &crtc_state->pipe_src);
+		clip_area_update(&new_crtc_state->psr2_su_area, &damaged_area,
+				 &new_crtc_state->pipe_src);
 	}
 
 	/*
@@ -2902,7 +2912,7 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 	 * should identify cases where this happens and fix the area
 	 * calculation for those.
 	 */
-	if (crtc_state->psr2_su_area.y1 == -1) {
+	if (new_crtc_state->psr2_su_area.y1 == -1) {
 		drm_info_once(display->drm,
 			      "Selective fetch area calculation failed in pipe %c\n",
 			      pipe_name(crtc->pipe));
@@ -2912,7 +2922,7 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 	if (full_update)
 		goto skip_sel_fetch_set_loop;
 
-	intel_psr_apply_su_area_workarounds(crtc_state);
+	intel_psr_apply_su_area_workarounds(new_crtc_state);
 
 	ret = drm_atomic_add_affected_planes(&state->base, &crtc->base);
 	if (ret)
@@ -2926,7 +2936,7 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 	 */
 	intel_psr2_sel_fetch_et_alignment(state, crtc, &cursor_in_su_area);
 
-	intel_psr2_sel_fetch_pipe_alignment(crtc_state);
+	intel_psr2_sel_fetch_pipe_alignment(new_crtc_state);
 
 	/*
 	 * Now that we have the pipe damaged area check if it intersect with
@@ -2937,11 +2947,11 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 		struct drm_rect *sel_fetch_area, inter;
 		struct intel_plane *linked = new_plane_state->planar_linked_plane;
 
-		if (new_plane_state->uapi.crtc != crtc_state->uapi.crtc ||
+		if (new_plane_state->uapi.crtc != new_crtc_state->uapi.crtc ||
 		    !new_plane_state->uapi.visible)
 			continue;
 
-		inter = crtc_state->psr2_su_area;
+		inter = new_crtc_state->psr2_su_area;
 		sel_fetch_area = &new_plane_state->psr2_sel_fetch_area;
 		if (!drm_rect_intersect(&inter, &new_plane_state->uapi.dst)) {
 			sel_fetch_area->y1 = -1;
@@ -2951,12 +2961,12 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 			 * disable it
 			 */
 			if (drm_rect_height(&old_plane_state->psr2_sel_fetch_area) > 0)
-				crtc_state->update_planes |= BIT(plane->id);
+				new_crtc_state->update_planes |= BIT(plane->id);
 
 			continue;
 		}
 
-		if (!psr2_sel_fetch_plane_state_supported(new_plane_state)) {
+		if (!psr2_sel_fetch_plane_state_supported(old_crtc_state, new_plane_state)) {
 			full_update = true;
 			break;
 		}
@@ -2964,7 +2974,7 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 		sel_fetch_area = &new_plane_state->psr2_sel_fetch_area;
 		sel_fetch_area->y1 = inter.y1 - new_plane_state->uapi.dst.y1;
 		sel_fetch_area->y2 = inter.y2 - new_plane_state->uapi.dst.y1;
-		crtc_state->update_planes |= BIT(plane->id);
+		new_crtc_state->update_planes |= BIT(plane->id);
 
 		/*
 		 * Sel_fetch_area is calculated for UV plane. Use
@@ -2981,14 +2991,14 @@ int intel_psr2_sel_fetch_update(struct intel_atomic_state *state,
 			linked_sel_fetch_area = &linked_new_plane_state->psr2_sel_fetch_area;
 			linked_sel_fetch_area->y1 = sel_fetch_area->y1;
 			linked_sel_fetch_area->y2 = sel_fetch_area->y2;
-			crtc_state->update_planes |= BIT(linked->id);
+			new_crtc_state->update_planes |= BIT(linked->id);
 		}
 	}
 
 skip_sel_fetch_set_loop:
-	psr2_man_trk_ctl_calc(crtc_state, full_update);
-	crtc_state->pipe_srcsz_early_tpt =
-		psr2_pipe_srcsz_early_tpt_calc(crtc_state, full_update);
+	psr2_man_trk_ctl_calc(new_crtc_state, full_update);
+	old_crtc_state->pipe_srcsz_early_tpt =
+		psr2_pipe_srcsz_early_tpt_calc(new_crtc_state, full_update);
 	return 0;
 }
 
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH v2 3/3] drm/i915/psr: Allow async flip when Selective Fetch enabled
  2025-12-01 11:32 [PATCH v2 0/3] Selective Fetch and async flip Jouni Högander
  2025-12-01 11:32 ` [PATCH v2 1/3] drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for PSR Jouni Högander
  2025-12-01 11:32 ` [PATCH v2 2/3] drm/i915/psr: Perform full frame update on async flip Jouni Högander
@ 2025-12-01 11:32 ` Jouni Högander
  2025-12-01 11:36 ` ✗ Fi.CI.BUILD: failure for Selective Fetch and async flip (rev2) Patchwork
  3 siblings, 0 replies; 5+ messages in thread
From: Jouni Högander @ 2025-12-01 11:32 UTC (permalink / raw)
  To: intel-gfx, intel-xe; +Cc: Jouni Högander

Now as Selective Fetch is performing full frame update on async flip and
vblank evasion is done as needed we can allow async flip even when
Selective Fetch is enabled.

Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
---
 drivers/gpu/drm/i915/display/intel_display.c | 8 --------
 1 file changed, 8 deletions(-)

diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
index 04f5c488f399..a8a3e8000187 100644
--- a/drivers/gpu/drm/i915/display/intel_display.c
+++ b/drivers/gpu/drm/i915/display/intel_display.c
@@ -6017,14 +6017,6 @@ static int intel_async_flip_check_uapi(struct intel_atomic_state *state,
 		return -EINVAL;
 	}
 
-	/* FIXME: selective fetch should be disabled for async flips */
-	if (new_crtc_state->enable_psr2_sel_fetch) {
-		drm_dbg_kms(display->drm,
-			    "[CRTC:%d:%s] async flip disallowed with PSR2 selective fetch\n",
-			    crtc->base.base.id, crtc->base.name);
-		return -EINVAL;
-	}
-
 	for_each_oldnew_intel_plane_in_state(state, plane, old_plane_state,
 					     new_plane_state, i) {
 		if (plane->pipe != crtc->pipe)
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* ✗ Fi.CI.BUILD: failure for Selective Fetch and async flip (rev2)
  2025-12-01 11:32 [PATCH v2 0/3] Selective Fetch and async flip Jouni Högander
                   ` (2 preceding siblings ...)
  2025-12-01 11:32 ` [PATCH v2 3/3] drm/i915/psr: Allow async flip when Selective Fetch enabled Jouni Högander
@ 2025-12-01 11:36 ` Patchwork
  3 siblings, 0 replies; 5+ messages in thread
From: Patchwork @ 2025-12-01 11:36 UTC (permalink / raw)
  To: Jouni Högander; +Cc: intel-gfx

== Series Details ==

Series: Selective Fetch and async flip (rev2)
URL   : https://patchwork.freedesktop.org/series/158003/
State : failure

== Summary ==

Error: patch https://patchwork.freedesktop.org/api/1.0/series/158003/revisions/2/mbox/ not applied
Applying: drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for PSR
Applying: drm/i915/psr: Perform full frame update on async flip
Using index info to reconstruct a base tree...
M	drivers/gpu/drm/i915/display/intel_psr.c
Falling back to patching base and 3-way merge...
Auto-merging drivers/gpu/drm/i915/display/intel_psr.c
CONFLICT (content): Merge conflict in drivers/gpu/drm/i915/display/intel_psr.c
error: Failed to merge in the changes.
hint: Use 'git am --show-current-patch=diff' to see the failed patch
Patch failed at 0002 drm/i915/psr: Perform full frame update on async flip
When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".
Build failed, no error log produced

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2025-12-01 11:36 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-12-01 11:32 [PATCH v2 0/3] Selective Fetch and async flip Jouni Högander
2025-12-01 11:32 ` [PATCH v2 1/3] drm/i915/psr: Set plane id bit in crtc_state->async_flip_planes for PSR Jouni Högander
2025-12-01 11:32 ` [PATCH v2 2/3] drm/i915/psr: Perform full frame update on async flip Jouni Högander
2025-12-01 11:32 ` [PATCH v2 3/3] drm/i915/psr: Allow async flip when Selective Fetch enabled Jouni Högander
2025-12-01 11:36 ` ✗ Fi.CI.BUILD: failure for Selective Fetch and async flip (rev2) Patchwork

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox