drm/i915: Don't clobber plane state on internal disables
Subsystems (2)
PATCH 1/1 drm/i915: Don't clobber plane state on internal disables
We need to disable all sprite planes when disabling the CRTC. We had been using the top-level atomic 'disable' entrypoint to accomplish this, which was wrong. Not only can this lead to various locking issues, it also modifies the actual plane state, making it impossible to restore the plane properly later. For example, a DPMS off followed by a DPMS on will result in any sprite planes in use not being restored properly.
The proper solution here is to call directly into our 'commit plane' hook with a copy of the plane's current state that has 'visible' set to false. Committing this dummy state will turn off the plane, but will not touch the actual plane->state pointer, allowing us to properly restore the plane state later.
---
This was a pretty big problem so I suspect this is probably tied to some of our existing bugzilla's but I'm not sure exactly which ones. It seems that we lack an i-g-t test for plane status across DPMS, so I'll send along a new i-g-t test for that shortly.
drivers/gpu/drm/i915/intel_display.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_display.c +20 -2
@@ -4201,6 +4201,24 @@ static void intel_enable_sprite_planes(struct drm_crtc *crtc) } } +/* +* Disable a plane internally without actually modifying the plane's state. +* This will allow us to easily restore the plane later by just reprogramming +* its state. +*/ +static void disable_plane_internal (struct drm_plane *plane) +{ +struct intel_plane *intel_plane =to_intel_plane (plane); +struct drm_plane_state *state = + plane->funcs->atomic_duplicate_state (plane); +struct intel_plane_state *intel_state =to_intel_plane_state (state); + + intel_state->visible = false; + intel_plane->commit_plane (plane, intel_state); + +intel_plane_destroy_state (plane, state); +} +static void intel_disable_sprite_planes (struct drm_crtc *crtc) {struct drm_device *dev = crtc->dev;
@@ -4210,8 +4228,8 @@ static void intel_disable_sprite_planes(struct drm_crtc *crtc)drm_for_each_legacy_plane (plane, &dev->mode_config.plane_list) { intel_plane =to_intel_plane (plane); -if (intel_plane->pipe == pipe) - plane->funcs->disable_plane (plane); +if (plane->fb && intel_plane->pipe == pipe) +disable_plane_internal (plane); } }
Signature · 6 lines
1.8.5.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 1/1 tests/kms_plane: Ensure planes recover from DPMS
i915 was using the main atomic 'disable plane' to turn off sprite planes during a CRTC disable. This was problematic because it modified the plane state, preventing us from recovering the original state later. One such case was that during a DPMS OFF followed by a DPMS ON, any sprite planes would not be restored properly.
Let's add a test that toggles DPMS off and on and ensures that the CRC remains the same (i.e., planes are successfully restored unchanged).
--- tests/kms_plane.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)
tests/kms_plane.c +20 -1
@@ -145,6 +145,7 @@ create_fb_for_mode__position(data_t *data, drmModeModeInfo *mode,enum {TEST_POSITION_FULLY_COVERED =1 <<0 , +TEST_DPMS =1 <<1 , }; static void
@@ -158,7 +159,7 @@ test_plane_position_with_output(data_t *data,igt_plane_t *primary, *sprite;struct igt_fb primary_fb, sprite_fb;drmModeModeInfo *mode; -igt_crc_t crc; +igt_crc_t crc, crc2;igt_info ("Testing connector %s using pipe %s plane %d\n" ,igt_output_name (output),kmstest_pipe_name (pipe), plane);
@@ -194,11 +195,24 @@ test_plane_position_with_output(data_t *data,igt_pipe_crc_collect_crc (data->pipe_crc, &crc); +if (flags &TEST_DPMS ) { +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_OFF ); +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_ON ); + } + +igt_pipe_crc_collect_crc (data->pipe_crc, &crc2); +if (flags &TEST_POSITION_FULLY_COVERED )igt_assert (igt_crc_equal (&test.reference_crc, &crc));else igt_assert (!igt_crc_equal (&test.reference_crc, &crc)); +igt_assert (igt_crc_equal (&crc, &crc2)); +igt_plane_set_fb (primary,NULL );igt_plane_set_fb (sprite,NULL );
@@ -358,6 +372,11 @@ run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) kmstest_pipe_name(pipe), plane) test_plane_position(data, pipe, plane, 0); +igt_subtest_f ("plane-position-hole-dpms-pipe-%s-plane-%d" , + kmstest_pipe_name(pipe), plane) +test_plane_position (data, pipe, plane, +TEST_DPMS ); +igt_subtest_f ("plane-panning-top-left-pipe-%s-plane-%d" , kmstest_pipe_name(pipe), plane)test_plane_panning (data, pipe, plane,TEST_PANNING_TOP_LEFT );
Signature · 6 lines
1.8.5.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH i-g-t] tests/kms_plane: Ensure planes recover from DPMS
Quoted 37 lines · Matt Roper
i915 was using the main atomic 'disable plane' to turn off sprite planes during a CRTC disable. This was problematic because it modified the plane state, preventing us from recovering the original state later. One such case was that during a DPMS OFF followed by a DPMS ON, any sprite planes would not be restored properly.
Let's add a test that toggles DPMS off and on and ensures that the CRC remains the same (i.e., planes are successfully restored unchanged).
Signed-off-by Matt Roper <[email protected]>--- tests/kms_plane.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)tests/kms_plane.c +2
-1@@ -145,6 +145,7 @@ create_fb_for_mode__position(data_t *data, drmModeModeInfo *mode,enum {TEST_POSITION_FULLY_COVERED =1 <<0 , +TEST_DPMS =1 <<1 , }; static void@@ -158,7 +159,7 @@ test_plane_position_with_output(data_t *data,igt_plane_t *primary, *sprite;struct igt_fb primary_fb, sprite_fb;drmModeModeInfo *mode; -igt_crc_t crc; +igt_crc_t crc, crc2;igt_info ("Testing connector %s using pipe %s plane %d\n" ,igt_output_name (output),kmstest_pipe_name (pipe), plane);@@ -194,11 +195,24 @@ test_plane_position_with_output(data_t *data,
tests/kms_plane.c test_plane_position_with_output(data_t *data, · line 205
Quoted 10 lines
tests/kms_plane.c +8
-0igt_pipe_crc_collect_crc (data->pipe_crc, &crc); +if (flags &TEST_DPMS ) { +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_OFF ); +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_ON ); + }
I think yet one more subtest to test against system suspend/resume in this spot would be good.
Quoted 14 lines
tests/kms_plane.c +5
-0+ +igt_pipe_crc_collect_crc (data->pipe_crc, &crc2); +if (flags &TEST_POSITION_FULLY_COVERED )igt_assert (igt_crc_equal (&test.reference_crc, &crc));else igt_assert (!igt_crc_equal (&test.reference_crc, &crc)); +igt_assert (igt_crc_equal (&crc, &crc2)); +igt_plane_set_fb (primary,NULL );igt_plane_set_fb (sprite,NULL );
tests/kms_plane.c run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) · line 375
Quoted 4 lines
tests/kms_plane.c +1
-0@@ -358,6 +372,11 @@ run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) kmstest_pipe_name(pipe), plane)test_plane_position (data, pipe, plane,0 ); + igt_subtest_f("plane-position-hole-dpms-pipe-%s-plane-%d",
Usually we call these -vs-dpms (and -vs-suspend for the suspend/resume one).
-Daniel
Quoted 14 lines
tests/kms_plane.c +4
-0+ kmstest_pipe_name(pipe), plane) +test_plane_position (data ,pipe ,plane , +TEST_DPMS ); +igt_subtest_f ("plane-panning-top-left-pipe-%s-plane-%d" , kmstest_pipe_name(pipe), plane)test_plane_panning (data, pipe, plane,TEST_PANNING_TOP_LEFT );Signature · 6 lines
1.8.5.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH i-g-t] tests/kms_plane: Ensure planes recover from DPMS
Quoted 40 lines · Daniel Vetter
Quoted 39 lines · Matt Roper
i915 was using the main atomic 'disable plane' to turn off sprite planes during a CRTC disable. This was problematic because it modified the plane state, preventing us from recovering the original state later. One such case was that during a DPMS OFF followed by a DPMS ON, any sprite planes would not be restored properly.
Let's add a test that toggles DPMS off and on and ensures that the CRC remains the same (i.e., planes are successfully restored unchanged).
Signed-off-by Matt Roper <[email protected]>--- tests/kms_plane.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)tests/kms_plane.c +2
-1@@ -145,6 +145,7 @@ create_fb_for_mode__position(data_t *data, drmModeModeInfo *mode,enum {TEST_POSITION_FULLY_COVERED =1 <<0 , +TEST_DPMS =1 <<1 , }; static void@@ -158,7 +159,7 @@ test_plane_position_with_output(data_t *data,igt_plane_t *primary, *sprite;struct igt_fb primary_fb, sprite_fb;drmModeModeInfo *mode; -igt_crc_t crc; +igt_crc_t crc, crc2;igt_info ("Testing connector %s using pipe %s plane %d\n" ,igt_output_name (output),kmstest_pipe_name (pipe), plane);@@ -194,11 +195,24 @@ test_plane_position_with_output(data_t *data,igt_pipe_crc_collect_crc (data->pipe_crc, &crc);
Quoted 11 lines
tests/kms_plane.c test_plane_position_with_output(data_t *data, · line 205
Quoted 8 lines
tests/kms_plane.c +8
-0+if (flags &TEST_DPMS ) { +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_OFF ); +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_ON ); + }I think yet one more subtest to test against system suspend/resume in this spot would be good.
We actually already do have a subtest for suspend/resume ("plane-panning-bottom-right-suspend-pipe-%s-plane-%d") which was probably failing before my kernel patch (although when I tried running it my system had bigger problems and simply wasn't ever coming back from suspend). I just tried it again and it does seem to be running properly now.
Quoted 14 lines
Quoted 14 lines
tests/kms_plane.c +5
-0+ +igt_pipe_crc_collect_crc (data->pipe_crc, &crc2); +if (flags &TEST_POSITION_FULLY_COVERED )igt_assert (igt_crc_equal (&test.reference_crc, &crc));else igt_assert (!igt_crc_equal (&test.reference_crc, &crc)); +igt_assert (igt_crc_equal (&crc, &crc2)); +igt_plane_set_fb (primary,NULL );igt_plane_set_fb (sprite,NULL );
Quoted 8 lines
tests/kms_plane.c run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) · line 375
Quoted 5 lines
tests/kms_plane.c +1
-0@@ -358,6 +372,11 @@ run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) kmstest_pipe_name(pipe), plane)test_plane_position (data, pipe, plane,0 ); + igt_subtest_f("plane-position-hole-dpms-pipe-%s-plane-%d",Usually we call these -vs-dpms (and -vs-suspend for the suspend/resume one).
Yeah, that's what I usually see, but the existing suspend test for kms_plane lacked the "vs" so I left it off dpms as well for consistency. Should I update the name of the existing test as well or leave it as is? I'm not sure if renaming an existing test would cause problems for PRTS or QA tracking.
Matt
Quoted 21 lines
-Daniel
Quoted 14 lines
tests/kms_plane.c +4
-0+ kmstest_pipe_name(pipe), plane) +test_plane_position (data ,pipe ,plane , +TEST_DPMS ); +igt_subtest_f ("plane-panning-top-left-pipe-%s-plane-%d" , kmstest_pipe_name(pipe), plane)test_plane_panning (data, pipe, plane,TEST_PANNING_TOP_LEFT );Signature · 6 lines
1.8.5.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfxSignature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Signature · 9 lines
Matt Roper Graphics Software Engineer IoTG Platform Enabling & Development Intel Corporation (916) 356-2795 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH i-g-t] tests/kms_plane: Ensure planes recover from DPMS
Quoted 53 lines · Matt Roper
Quoted 40 lines · Daniel Vetter
Quoted 39 lines · Matt Roper
i915 was using the main atomic 'disable plane' to turn off sprite planes during a CRTC disable. This was problematic because it modified the plane state, preventing us from recovering the original state later. One such case was that during a DPMS OFF followed by a DPMS ON, any sprite planes would not be restored properly.
Let's add a test that toggles DPMS off and on and ensures that the CRC remains the same (i.e., planes are successfully restored unchanged).
Signed-off-by Matt Roper <[email protected]>--- tests/kms_plane.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)tests/kms_plane.c +2
-1@@ -145,6 +145,7 @@ create_fb_for_mode__position(data_t *data, drmModeModeInfo *mode,enum {TEST_POSITION_FULLY_COVERED =1 <<0 , +TEST_DPMS =1 <<1 , }; static void@@ -158,7 +159,7 @@ test_plane_position_with_output(data_t *data,igt_plane_t *primary, *sprite;struct igt_fb primary_fb, sprite_fb;drmModeModeInfo *mode; -igt_crc_t crc; +igt_crc_t crc, crc2;igt_info ("Testing connector %s using pipe %s plane %d\n" ,igt_output_name (output),kmstest_pipe_name (pipe), plane);@@ -194,11 +195,24 @@ test_plane_position_with_output(data_t *data,igt_pipe_crc_collect_crc (data->pipe_crc, &crc);Quoted 11 lines
tests/kms_plane.c test_plane_position_with_output(data_t *data, · line 205
Quoted 8 lines
tests/kms_plane.c +8
-0+if (flags &TEST_DPMS ) { +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_OFF ); +kmstest_set_connector_dpms (data->drm_fd, + output->config.connector, +DRM_MODE_DPMS_ON ); + }I think yet one more subtest to test against system suspend/resume in this spot would be good.
Quoted 6 lines
We actually already do have a subtest for suspend/resume ("plane-panning-bottom-right-suspend-pipe-%s-plane-%d") which was probably failing before my kernel patch (although when I tried running it my system had bigger problems and simply wasn't ever coming back from suspend). I just tried it again and it does seem to be running properly now.
Hm I guess just yet another regression that failed to get through our QA/bug team maze :( A quick check in bugzilla at least didn't turn up anything. Oh well.
Quoted 22 lines
Quoted 13 lines
Quoted 13 lines
tests/kms_plane.c +5
-0+ +igt_pipe_crc_collect_crc (data->pipe_crc, &crc2); +if (flags &TEST_POSITION_FULLY_COVERED )igt_assert (igt_crc_equal (&test.reference_crc, &crc));else igt_assert (!igt_crc_equal (&test.reference_crc, &crc)); +igt_assert (igt_crc_equal (&crc, &crc2)); +igt_plane_set_fb (primary,NULL );igt_plane_set_fb (sprite,NULL );Quoted 8 lines
tests/kms_plane.c run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) · line 375
Quoted 5 lines
tests/kms_plane.c +1
-0@@ -358,6 +372,11 @@ run_tests_for_pipe_plane(data_t *data, enum pipe pipe, enum igt_plane plane) kmstest_pipe_name(pipe), plane)test_plane_position (data, pipe, plane,0 ); + igt_subtest_f("plane-position-hole-dpms-pipe-%s-plane-%d",Usually we call these -vs-dpms (and -vs-suspend for the suspend/resume one).
Quoted 5 lines
Yeah, that's what I usually see, but the existing suspend test for
kms_planelacked the "vs" so I left it off dpms as well for consistency. Should I update the name of the existing test as well or leave it as is? I'm not sure if renaming an existing test would cause problems for PRTS or QA tracking.
Naming convention just says that it should have suspend/dpms in the name somewhere. So if this is more consistent then we're good I think. I'll apply your patch, thanks.
-Daniel
Quoted 32 lines
Matt
Quoted 21 lines
-Daniel
Quoted 14 lines
tests/kms_plane.c +4
-0+ kmstest_pipe_name(pipe), plane) +test_plane_position (data ,pipe ,plane , +TEST_DPMS ); +igt_subtest_f ("plane-panning-top-left-pipe-%s-plane-%d" , kmstest_pipe_name(pipe), plane)test_plane_panning (data, pipe, plane,TEST_PANNING_TOP_LEFT );Signature · 6 lines
1.8.5.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfxSignature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.chSignature · 5 lines
Matt Roper Graphics Software Engineer IoTG Platform Enabling & Development Intel Corporation (916) 356-2795
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Task id: 5887 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 280/280 280/280 ILK 308/308 308/308 SNB -1 328/328 327/328 IVB -1 379/379 378/379 BYT 294/294 294/294 HSW 387/387 387/387 BDW -1 316/316 315/316 -------------------------------------Detailed------------------------------------- Platform Test drm-intel-nightly Series Applied SNB igt_kms_rotation_crc_sprite-rotation DMESG_WARN(1)PASS(4) DMESG_WARN(2) *IVB igt_kms_rotation_crc_sprite-rotation PASS(2) DMESG_WARN(2) *BDW igt_gem_gtt_hog PASS(3) DMESG_WARN(1)PASS(1)
Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list [email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 37 lines · Matt Roper
We need to disable all sprite planes when disabling the CRTC. We had been using the top-level atomic 'disable' entrypoint to accomplish this, which was wrong. Not only can this lead to various locking issues, it also modifies the actual plane state, making it impossible to restore the plane properly later. For example, a DPMS off followed by a DPMS on will result in any sprite planes in use not being restored properly.
The proper solution here is to call directly into our 'commit plane' hook with a copy of the plane's current state that has 'visible' set to false. Committing this dummy state will turn off the plane, but will not touch the actual
plane->statepointer, allowing us to properly restore the plane state later.Signed-off-by Matt Roper <[email protected]>---
This was a pretty big problem so I suspect this is probably tied to some of our existing bugzilla's but I'm not sure exactly which ones. It seems that we lack an i-g-t test for plane status across DPMS, so I'll send along a new i-g-t test for that shortly.drivers/gpu/drm/i915/intel_display.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/intel_display.c +6
-0@@ -4201,6 +4201,24 @@ static void intel_enable_sprite_planes(struct drm_crtc *crtc) } } +/* +* Disable a plane internally without actually modifying the plane's state. +* This will allow us to easily restore the plane later by just reprogramming +* its state. +*/ +static void disable_plane_internal (struct drm_plane *plane)
drivers/gpu/drm/i915/intel_display.c static void intel_enable_sprite_planes(struct drm_crtc *crtc) · line 4219
Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10
-0+{ +struct intel_plane *intel_plane =to_intel_plane (plane); +struct drm_plane_state *state = + plane->funcs->atomic_duplicate_state (plane); +struct intel_plane_state *intel_state =to_intel_plane_state (state); + + intel_state->visible = false; + intel_plane->commit_plane (plane, intel_state); + +intel_plane_destroy_state (plane, state);
I wonder whether long-term we should do a full atomic commit including the begin/end dance here. But this is more than good enough for now, so merged to dinq.
Thanks, Daniel
Quoted 23 lines
drivers/gpu/drm/i915/intel_display.c +4
-2+} +static void intel_disable_sprite_planes (struct drm_crtc *crtc) {struct drm_device *dev = crtc->dev;@@ -4210,8 +4228,8 @@ static void intel_disable_sprite_planes(struct drm_crtc *crtc)drm_for_each_legacy_plane (plane, &dev->mode_config.plane_list) { intel_plane =to_intel_plane (plane); -if (intel_plane->pipe == pipe) - plane->funcs->disable_plane (plane); +if (plane->fb && intel_plane->pipe == pipe) +disable_plane_internal (plane); } }Signature · 6 lines
1.8.5.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx