Skip to contentSkip to conversation

drm/i915: Hold runtime PM during plane commit

PATCH In review Matt Roper 5 participants 8 messages Dec 15, 2014 – Dec 17, 2014

Subsystems (2)
lore
Share & tools
lore

Matt Roper

author
#0 lore raw

PATCH 1/1 drm/i915: Hold runtime PM during plane commit

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

---
 drivers/gpu/drm/i915/intel_display.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/gpu/drm/i915/intel_display.c +3 -0
index 3044af5..a0ddce5 100644
@@ -11863,6 +11863,7 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc,
 		   uint32_t src_w, uint32_t src_h)
 {
 	struct drm_device *dev = plane->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_framebuffer *old_fb = plane->fb;
 	struct intel_plane_state state;
 	struct intel_plane *intel_plane = to_intel_plane(plane);
@@ -11902,7 +11903,9 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc,
 			return ret;
 	}
 
+	intel_runtime_pm_get(dev_priv);
 	intel_plane->commit_plane(plane, &state);
+	intel_runtime_pm_put(dev_priv);
 
 	if (fb != old_fb && old_fb) {
 		if (intel_crtc->active)
Signature · 6 lines
1.8.5.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 1/1
#1 lore raw

2014-12-15 16:11 GMT-02:00 Matt Roper <[email protected]>:

Quoted 21 lines

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

Quoted 7 lines

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

I see we're in the middle of a very big rework on how all these prepare/commit functions are called, so I don't think it makes sense to spend too much time trying to find the very-best-perfect spot for the get/put calls, since they're likely to be changed later. So I guess that for now it's important to fix the current "regression" reported by QA:

Testcase: igt/pm-rpm/legacy-planes

Quoted 33 lines
---
 drivers/gpu/drm/i915/intel_display.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/gpu/drm/i915/intel_display.c +3 -1
index 3044af5..a0ddce5 100644
@@ -11863,6 +11863,7 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc,
                   uint32_t src_w, uint32_t src_h)
 {
        struct drm_device *dev = plane->dev;
+       struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_framebuffer *old_fb = plane->fb;
        struct intel_plane_state state;
        struct intel_plane *intel_plane = to_intel_plane(plane);
@@ -11902,7 +11903,9 @@ intel_update_plane(struct drm_plane *plane, struct drm_crtc *crtc,
                        return ret;
        }
 
+       intel_runtime_pm_get(dev_priv);
        intel_plane->commit_plane(plane, &state);
+       intel_runtime_pm_put(dev_priv);
 
        if (fb != old_fb && old_fb) {
                if (intel_crtc->active)
--

1.8.5.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 5 lines
Paulo Zanoni
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw
Quoted 21 lines · Paulo Zanoni

2014-12-15 16:11 GMT-02:00 Matt Roper <[email protected]>:

Quoted 20 lines

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

Which kind of registers? If this is just in the system agent then a rpm ref is enough, but if this is also about plane registers then we'd need a reference of the plane power domain. Which would indicate some failure to check for crtc->active somewhere I think.

Quoted 15 lines
Quoted 7 lines

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

I see we're in the middle of a very big rework on how all these prepare/commit functions are called, so I don't think it makes sense to spend too much time trying to find the very-best-perfect spot for the get/put calls, since they're likely to be changed later. So I guess that for now it's important to fix the current "regression" reported by QA:

Quoted 2 lines

Testcase: igt/pm-rpm/legacy-planes

Also Cc: [email protected] I guess? Or is this only for dinq? -Daniel

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
maintainer
#3 lore raw
Quoted 43 lines · Daniel Vetter
Quoted 21 lines · Paulo Zanoni

2014-12-15 16:11 GMT-02:00 Matt Roper <[email protected]>:

Quoted 20 lines

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

Which kind of registers? If this is just in the system agent then a rpm ref is enough, but if this is also about plane registers then we'd need a reference of the plane power domain. Which would indicate some failure to check for crtc->active somewhere I think.

Quoted 15 lines
Quoted 7 lines

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

I see we're in the middle of a very big rework on how all these prepare/commit functions are called, so I don't think it makes sense to spend too much time trying to find the very-best-perfect spot for the get/put calls, since they're likely to be changed later. So I guess that for now it's important to fix the current "regression" reported by QA:

Quoted 4 lines
Quoted 2 lines

Testcase: igt/pm-rpm/legacy-planes

Also Cc: [email protected] I guess? Or is this only for dinq?

At least the partial fix referenced is in 3.17.

Quoted 9 lines

-Daniel

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
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
maintainer
#4 lore raw
Quoted 44 lines · Jani Nikula
Quoted 43 lines · Daniel Vetter
Quoted 21 lines · Paulo Zanoni

2014-12-15 16:11 GMT-02:00 Matt Roper <[email protected]>:

Quoted 20 lines

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

Which kind of registers? If this is just in the system agent then a rpm ref is enough, but if this is also about plane registers then we'd need a reference of the plane power domain. Which would indicate some failure to check for crtc->active somewhere I think.

Quoted 15 lines
Quoted 7 lines

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

I see we're in the middle of a very big rework on how all these prepare/commit functions are called, so I don't think it makes sense to spend too much time trying to find the very-best-perfect spot for the get/put calls, since they're likely to be changed later. So I guess that for now it's important to fix the current "regression" reported by QA:

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

Testcase: igt/pm-rpm/legacy-planes

Also Cc: [email protected] I guess? Or is this only for dinq?

At least the partial fix referenced is in 3.17.

Ugh, but it conflicts badly, there's no ->commit_plane in drm-intel-next-fixes.

Daniel, pick this up for dinq, and if someone wants this to stable, it needs a backported version.

BR,
Jani.

Quoted 13 lines
Quoted 10 lines

-Daniel

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
Signature · 1 line
Jani Nikula, Intel Open Source Technology Center
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw
Quoted 50 lines · Jani Nikula
Quoted 49 lines · Jani Nikula
Quoted 47 lines · Daniel Vetter
Quoted 21 lines · Paulo Zanoni

2014-12-15 16:11 GMT-02:00 Matt Roper <[email protected]>:

Quoted 20 lines

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

Which kind of registers? If this is just in the system agent then a rpm ref is enough, but if this is also about plane registers then we'd need a reference of the plane power domain. Which would indicate some failure to check for crtc->active somewhere I think.

Quoted 15 lines
Quoted 7 lines

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

I see we're in the middle of a very big rework on how all these prepare/commit functions are called, so I don't think it makes sense to spend too much time trying to find the very-best-perfect spot for the get/put calls, since they're likely to be changed later. So I guess that for now it's important to fix the current "regression" reported by QA:

Quoted 2 lines

Testcase: igt/pm-rpm/legacy-planes

Also Cc: [email protected] I guess? Or is this only for dinq?

Quoted 7 lines
Quoted 1 line

At least the partial fix referenced is in 3.17.

Ugh, but it conflicts badly, there's no ->commit_plane in drm-intel-next-fixes.

Daniel, pick this up for dinq, and if someone wants this to stable, it needs a backported version.

Done, and thanks to Paulo for the additional clarification. I've augmented the commit message with that.
-Daniel

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
#6 lore raw
In reply to Daniel Vetter

2014-12-16 7:42 GMT-02:00 Daniel Vetter <[email protected]>:

Quoted 23 lines · Paulo Zanoni
Quoted 21 lines

2014-12-15 16:11 GMT-02:00 Matt Roper <[email protected]>:

Quoted 20 lines

During plane operations, we read/write some registers that only operate properly if we're not runtime suspended. At the moment we're not holding the runtime PM reference across the whole plane operation, so there's a potential for problems.

This issue was already partially addressed by commit

commit d6dd6843ff4a57c662dbc378b9f99a9c034b0956
Author: Paulo Zanoni <[email protected]>
Date: Fri Aug 15 15:59:32 2014 -0300

drm/i915: fix plane/cursor handling when runtime suspended

which took care of holding the runtime PM reference during the pin and fence operations for plane updates. However there are still a few actual plane registers that we also need to hold the runtime PM reference for. Recent refactoring patches in preparation for atomic have rearranged the code and made it increasingly likely that the hardware will have time to suspend between the pin/fence operation and the actual register writes.

Quoted 1 line

Which kind of registers?

Some of the regs touched by ivb_get_colorkey().

Quoted 20 lines

If this is just in the system agent then a rpm
ref is enough, but if this is also about plane registers then we'd need a reference of the plane power domain. Which would indicate some failure to check for crtc->active somewhere I think.

Quoted 15 lines
Quoted 7 lines

The solution here grabs the runtime PM reference around the 'commit' operation for planes, which should cover all the relevant register reads/writes.

Cc: Paulo Zanoni <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87180

I see we're in the middle of a very big rework on how all these prepare/commit functions are called, so I don't think it makes sense to spend too much time trying to find the very-best-perfect spot for the get/put calls, since they're likely to be changed later. So I guess that for now it's important to fix the current "regression" reported by QA:

Quoted 4 lines
Quoted 2 lines

Testcase: igt/pm-rpm/legacy-planes

Also Cc: [email protected] I guess? Or is this only for dinq?

I would suggest only dinq, since it's the only thing that can reproduce the WARN, which got visible after:

commit 6beb8c23ebcc3d3287d8a247d11b73d7d0eaa475 drm/i915: Consolidate plane 'prepare' functions (v2)

Quoted 5 lines

-Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Signature · 5 lines
Paulo Zanoni
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v1 1/1
#7 lore raw
In reply to Matt Roper

-------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 364/364 364/364
ILK +5 360/366 365/366

SNB                                  448/450              448/450
IVB                                  497/498              497/498
BYT                                  289/289              289/289
HSW                                  563/564              563/564
BDW                                  417/417              417/417
-------------------------------------Detailed-------------------------------------
Platform  Test                                drm-intel-nightly          Series Applied
 ILK  igt_drv_suspend_fence-restore-untiled      DMESG_WARN(1, M26)PASS(5, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_bcs-flip-vs-modeset-interruptible      DMESG_WARN(1, M26)PASS(5, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_busy-flip-interruptible      DMESG_WARN(1, M26)PASS(5, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_flip-vs-rmfb-interruptible      DMESG_WARN(1, M26)PASS(5, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_rcs-flip-vs-dpms      DMESG_WARN(1, M26)PASS(4, M37M26)      PASS(1, M37)

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