| Message ID | 20230929204120.120795-1-hamza.mahfooz@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Revert "drm/amd/display: Check all enabled planes in dm_check_crtc_cursor" | expand |
On 9/29/23 22:41, Hamza Mahfooz wrote: > From: Ivan Lipski <ivlipski@amd.com> > > This reverts commit 45e1ade04b4d60fe5df859076005779f27c4c9be. > > Since, it causes the following IGT tests to fail: > kms_cursor_legacy@cursor-vs-flip.* > kms_cursor_legacy@flip-vs-cursor.* Any information about how those tests fail? Maybe they accidentally rely on the broken behaviour? FWIW, something like the reverted commit is definitely needed, see https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3177#note_1829068 . That MR is blocked by the reverted fix.
On 10/2/23 12:05, Michel Dänzer wrote: > On 9/29/23 22:41, Hamza Mahfooz wrote: >> From: Ivan Lipski <ivlipski@amd.com> >> >> This reverts commit 45e1ade04b4d60fe5df859076005779f27c4c9be. >> >> Since, it causes the following IGT tests to fail: >> kms_cursor_legacy@cursor-vs-flip.* >> kms_cursor_legacy@flip-vs-cursor.* > > Any information about how those tests fail? Maybe they accidentally rely on the broken behaviour? I was able to reproduce, that doesn't seem to be the case. They just rely on multiple legacy cursor ioctl calls being able to complete between consecutive flips, which I suppose is broken by always pulling in non-cursor plane state with any cursor plane state changes. I'll see if I can find a better solution. Meanwhile, Acked-by: Michel Dänzer <mdaenzer@redhat.com> for the revert.
On 10/2/23 12:48, Michel Dänzer wrote: > On 10/2/23 12:05, Michel Dänzer wrote: >> On 9/29/23 22:41, Hamza Mahfooz wrote: >>> From: Ivan Lipski <ivlipski@amd.com> >>> >>> This reverts commit 45e1ade04b4d60fe5df859076005779f27c4c9be. >>> >>> Since, it causes the following IGT tests to fail: >>> kms_cursor_legacy@cursor-vs-flip.* >>> kms_cursor_legacy@flip-vs-cursor.* >> >> Any information about how those tests fail? Maybe they accidentally rely on the broken behaviour? > > I was able to reproduce, that doesn't seem to be the case. They just rely on multiple legacy cursor ioctl calls being able to complete between consecutive flips, which I suppose is broken by always pulling in non-cursor plane state with any cursor plane state changes. > > I'll see if I can find a better solution. Meanwhile, > > Acked-by: Michel Dänzer <mdaenzer@redhat.com> > > for the revert. Alternatively, here's an incremental fix: https://patchwork.freedesktop.org/series/124527/
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 32156609fbcf..49ffb4d6e9cc 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -10290,24 +10290,14 @@ static int dm_check_crtc_cursor(struct drm_atomic_state *state, * blending properties match the underlying planes'. */ - new_cursor_state = drm_atomic_get_plane_state(state, cursor); - if (IS_ERR(new_cursor_state)) - return PTR_ERR(new_cursor_state); - - if (!new_cursor_state->fb) + new_cursor_state = drm_atomic_get_new_plane_state(state, cursor); + if (!new_cursor_state || !new_cursor_state->fb) return 0; dm_get_oriented_plane_size(new_cursor_state, &cursor_src_w, &cursor_src_h); cursor_scale_w = new_cursor_state->crtc_w * 1000 / cursor_src_w; cursor_scale_h = new_cursor_state->crtc_h * 1000 / cursor_src_h; - /* Need to check all enabled planes, even if this commit doesn't change - * their state - */ - i = drm_atomic_add_affected_planes(state, crtc); - if (i) - return i; - for_each_new_plane_in_state_reverse(state, underlying, new_underlying_state, i) { /* Narrow down to non-cursor planes on the same CRTC as the cursor */ if (new_underlying_state->crtc != crtc || underlying == crtc->cursor)