Skip to contentSkip to conversation

Fix hang when entering into dpms-off mode

PATCH In review Aaron Liu 2 participants 6 messages Feb 28, 2019 – Mar 1, 2019 5 messages not in the archive

lore
Share & tools
lore

Aaron Liu

author
#0 lore raw

PATCH 1/1 Fix hang when entering into dpms-off mode

The hang happened in CTS 4.6 glcts. It is dued to inconsistent process in amdgpu_dri2_schedule_swap when system entering into dpms-off mode.

Normally, drmmode_wait_vblank will be excuted. But when entering into dpms-off mode, drmmode_wait_vblank can't be excuted and fallback to amdgpu_dri2_schedule_event to process.
Finally in amdgpu_dri2_deferred_event, vblank_handler will be excuted. It is the root cause of hang. Instead of vblank_handler, it just use amdgpu_dri2_frame_event_handler to process under dpms-off mode.

Change-Id I6bfed9692fc02c8c3c86df2a816884c4a1f46cb1
---
 src/amdgpu_dri2.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
src/amdgpu_dri2.c +7 -2
index a6b76a1..c54ad0c 100644
@@ -884,7 +884,7 @@ CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)
 	if (ret) {
 		xf86DrvMsg(scrn->scrnIndex, X_ERROR,
 			   "%s cannot get current time\n", __func__);
-		if (event_info->drm_queue_seq)
+		if (event_info->drm_queue_seq && amdgpu_crtc_is_enabled(crtc))
 			drmmode_crtc->drmmode->event_context.
 				vblank_handler(pAMDGPUEnt->fd, 0, 0, 0,
 					       (void*)event_info->drm_queue_seq);
@@ -900,7 +900,12 @@ CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)
 	delta_seq = delta_t * drmmode_crtc->dpms_last_fps;
 	delta_seq /= 1000000;
 	frame = (CARD64) drmmode_crtc->dpms_last_seq + delta_seq;
-	if (event_info->drm_queue_seq)
+	/*
+	 * If CRTC is in DPMS off state, it can't use vblank_handler.
+	 * Because drmmode_wait_vblank is not excuted in such as
+	 * amdgpu_dri2_schedule_swap/amdgpu_dri2_schedule_wait_msc.
+	 */
+	if (event_info->drm_queue_seq && amdgpu_crtc_is_enabled(crtc))
 		drmmode_crtc->drmmode->event_context.
 			vblank_handler(pAMDGPUEnt->fd, frame, drm_now / 1000000,
 				       drm_now % 1000000,
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1551322325-19958-1-git-send-email-aaron.liu-5C7GfCeVMHo@public.gmane.org

#2 lore raw

Hi Aaron,

thanks for the patch. xf86-video-amdgpu now uses GitLab merge requests for patch submission and review:

https://gitlab.freedesktop.org/xorg/driver/xf86-video-amdgpu/merge_requests

That said, comments on the patch below:

Quoted 34 lines · Aaron Liu

The hang happened in CTS 4.6 glcts. It is dued to inconsistent process in amdgpu_dri2_schedule_swap when system entering into dpms-off mode.

Normally, drmmode_wait_vblank will be excuted. But when entering into dpms-off mode, drmmode_wait_vblank can't be excuted and fallback to amdgpu_dri2_schedule_event to process.
Finally in amdgpu_dri2_deferred_event, vblank_handler will be excuted. It is the root cause of hang. Instead of vblank_handler, it just use amdgpu_dri2_frame_event_handler to process under dpms-off mode.

Change-Id I6bfed9692fc02c8c3c86df2a816884c4a1f46cb1
---
 src/amdgpu_dri2.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
src/amdgpu_dri2.c +1 -1
index a6b76a1..c54ad0c 100644
@@ -884,7 +884,7 @@ CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)
 	if (ret) {
 		xf86DrvMsg(scrn->scrnIndex, X_ERROR,
 			   "%s cannot get current time\n", __func__);
-		if (event_info->drm_queue_seq)
+		if (event_info->drm_queue_seq && amdgpu_crtc_is_enabled(crtc))
 			drmmode_crtc->drmmode->event_context.
 				vblank_handler(pAMDGPUEnt->fd, 0, 0, 0,
 					       (void*)event_info->drm_queue_seq);
@@ -900,7 +900,12 @@ CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)
 	delta_seq = delta_t * drmmode_crtc->dpms_last_fps;
 	delta_seq /= 1000000;
 	frame = (CARD64) drmmode_crtc->dpms_last_seq + delta_seq;

src/amdgpu_dri2.c CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data) · line 911

Quoted 10 lines
src/amdgpu_dri2.c +6 -1
-	if (event_info->drm_queue_seq)
+	/*
+	 * If CRTC is in DPMS off state, it can't use vblank_handler.
+	 * Because drmmode_wait_vblank is not excuted in such as
+	 * amdgpu_dri2_schedule_swap/amdgpu_dri2_schedule_wait_msc.
+	 */
+	if (event_info->drm_queue_seq && amdgpu_crtc_is_enabled(crtc))
 		drmmode_crtc->drmmode->event_context.
 			vblank_handler(pAMDGPUEnt->fd, frame, drm_now / 1000000,
 				       drm_now % 1000000,

This isn't a good solution I'm afraid, as it'll leave the struct amdgpu_drm_queue_entry memory associated with event_info->drm_queue_seq linked into the amdgpu_drm_queue list, which would gradually slow down processing of that list.

I think I know what the issue is, I'll work on a fix in the afternoon.

Signature · 6 lines
Earthling Michel Dänzer               |              https://www.amd.com
Libre software enthusiast             |             Mesa and X developer
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): dd111c99-c1f9-d1d7-5c40-3fcf7b75b1ea-otUistvHUpPR7s880joybQ@public.gmane.org

#4 lore raw
Quoted 24 lines · Michel Dänzer
Quoted 6 lines · Aaron Liu
src/amdgpu_dri2.c +0 -0
@@ -900,7 +900,12 @@ CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)
 	delta_seq = delta_t * drmmode_crtc->dpms_last_fps;
 	delta_seq /= 1000000;
 	frame = (CARD64) drmmode_crtc->dpms_last_seq + delta_seq;

src/amdgpu_dri2.c CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data) · line 911

Quoted 10 lines
src/amdgpu_dri2.c +6 -1
-	if (event_info->drm_queue_seq)
+	/*
+	 * If CRTC is in DPMS off state, it can't use vblank_handler.
+	 * Because drmmode_wait_vblank is not excuted in such as
+	 * amdgpu_dri2_schedule_swap/amdgpu_dri2_schedule_wait_msc.
+	 */
+	if (event_info->drm_queue_seq && amdgpu_crtc_is_enabled(crtc))
 		drmmode_crtc->drmmode->event_context.
 			vblank_handler(pAMDGPUEnt->fd, frame, drm_now / 1000000,
 				       drm_now % 1000000,

This isn't a good solution I'm afraid, as it'll leave the struct amdgpu_drm_queue_entry memory associated with event_info->drm_queue_seq linked into the amdgpu_drm_queue list, which would gradually slow down processing of that list.

Quoted 1 line

I think I know what the issue is, I'll work on a fix in the afternoon.

Please test
https://gitlab.freedesktop.org/xorg/driver/xf86-video-amdgpu/merge_requests/29
.

Signature · 6 lines
Earthling Michel Dänzer               |              https://www.amd.com
Libre software enthusiast             |             Mesa and X developer
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 814ff51d-cde2-8a3e-6755-fd656a900a87-otUistvHUpPR7s880joybQ@public.gmane.org

author
#6 lore raw

@Michel Dänzer,

I have reviewed your patch and verified it passed.

I couldn't merge this merge request to your master manually. The log is below:

aaliu@lnx-aaliu:~/work/brahma/xf86-video-amdgpu-gitlab$ git push origin master
Username for 'https://gitlab.freedesktop.org': aaliu
Password for 'https://[email protected]': 
remote: HTTP Basic: Access denied
remote: You must use a personal access token with 'api' scope for Git over HTTP.
remote: You can generate one at https://gitlab.freedesktop.org/profile/personal_access_tokens
fatal: Authentication failed for 'https://gitlab.freedesktop.org/xorg/driver/xf86-video-amdgpu.git/'

BR,
Aaron Liu

Quoted 44 lines

-----Original Message-----
From: Michel Dänzer <[email protected]>
Sent: Friday, March 01, 2019 12:36 AM
To: Liu, Aaron <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH xf86-video-amdgpu] Fix hang when entering into dpms- off mode

Quoted 3 lines · Michel Dänzer

Review CARD32

Quoted 2 lines · Aaron Liu
(quoted diff) +0 -0
@@ -900,7 +900,12 @@ CARD32

amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)

Quoted 4 lines
Quoted 4 lines
src/amdgpu_dri2.c +0 -1
 	delta_seq = delta_t * drmmode_crtc->dpms_last_fps;
 	delta_seq /= 1000000;
 	frame = (CARD64) drmmode_crtc->dpms_last_seq + delta_seq;
-	if (event_info->drm_queue_seq)
Quoted 8 lines

src/amdgpu_dri2.c CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data)

Quoted 8 lines
src/amdgpu_dri2.c +6 -0
+	/*
+	 * If CRTC is in DPMS off state, it can't use vblank_handler.
+	 * Because drmmode_wait_vblank is not excuted in such as
+	 * amdgpu_dri2_schedule_swap/amdgpu_dri2_schedule_wait_msc.
+	 */
+	if (event_info->drm_queue_seq && amdgpu_crtc_is_enabled(crtc))
 		drmmode_crtc->drmmode->event_context.
 			vblank_handler(pAMDGPUEnt->fd, frame, drm_now /

1000000,

Quoted 6 lines

src/amdgpu_dri2.c CARD32 amdgpu_dri2_deferred_event(OsTimerPtr timer, CARD32 now, pointer data) · line 911

Quoted 2 lines
src/amdgpu_dri2.c +0 -0
 				       drm_now % 1000000,

This isn't a good solution I'm afraid, as it'll leave the struct amdgpu_drm_queue_entry memory associated with event_info->drm_queue_seq linked into the amdgpu_drm_queue list,

which

Quoted 4 lines

would gradually slow down processing of that list.

I think I know what the issue is, I'll work on a fix in the afternoon.

Please test
https://gitlab.freedesktop.org/xorg/driver/xf86-video-
amdgpu/merge_requests/29
.

Quoted 3 lines

--
Earthling Michel Dänzer | https://www.amd.com
Libre software enthusiast | Mesa and X developer

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): DM6PR12MB361173B0EF321F8862519D85F0760-lmeGfMZKVrGUZgGQSl7qawdYzm3356FpvxpqHgZTriW3zl9H0oFU5g@public.gmane.org

Reviewed-by Aaron Liu Tested-by Aaron Liu
#8 lore raw
Quoted 3 lines · Aaron

@Michel Dänzer,

I have reviewed your patch and verified it passed.

Thanks Aaron, so I assume I can add

?

Quoted 1 line

I couldn't merge this merge request to your master manually.

Don't worry, I'll merge it via the GitLab merge request UI.

Signature · 6 lines
Earthling Michel Dänzer               |              https://www.amd.com
Libre software enthusiast             |             Mesa and X developer
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): a274ff35-60ba-d735-2656-0508556ad5b1-otUistvHUpPR7s880joybQ@public.gmane.org

author
#10 lore raw
Quoted 11 lines

-----Original Message-----
From: Michel Dänzer <[email protected]>
Sent: Friday, March 01, 2019 4:43 PM
To: Liu, Aaron <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH xf86-video-amdgpu] Fix hang when entering into dpms- off mode

Quoted 2 lines · Aaron

@Michel Dänzer,

Quoted 8 lines
Quoted 1 line

I have reviewed your patch and verified it passed.

Thanks Aaron, so I assume I can add

?

Sure😊

Quoted 10 lines
Quoted 3 lines

I couldn't merge this merge request to your master manually.

Don't worry, I'll merge it via the GitLab merge request UI.

--
Earthling Michel Dänzer | https://www.amd.com
Libre software enthusiast | Mesa and X developer

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx