*** some fixings ***
Subsystems (2)
cover *** some fixings ***
*** some bug fixings for SRIOV case ***
Emily Deng (1):
drm/amdgpu: Correct sdma_v4 get_wptr(v2)
Monk Liu (5):
drm/amdgpu: don't use MM idle_work for SRIOV(v2) drm/amdgpu: block till late_init_work done in dev_fini drm/amdgpu: should call drm_helper with dc support drm/amdgpu: adjust timeout for ib_ring_tests(v2) drm/amdgpu: cleanup SA inti and fini(v2)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 +- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 38 ++++++++++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 2 - drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c | 62 ++++++------------------------ drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 12 +++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 9 ++--- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 17 ++++---- 8 files changed, 64 insertions(+), 82 deletions(-)
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): 1519721235-28552-1-git-send-email-Monk.Liu-5C7GfCeVMHo@public.gmane.org
PATCH 1/6 drm/amdgpu: don't use MM idle_work for SRIOV(v2)
SRIOV doesn't give VF cg/pg feature so the MM's idle_work is skipped for SR-IOV
v2:
remove superfluous changes
since idle_work is not scheduled for SR-IOV so the condition check for SR-IOV inside idle_work also can be dropped
--- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 12 ++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 9 ++++----- 2 files changed, 10 insertions(+), 11 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +6 -6
@@ -297,7 +297,8 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)if (adev->uvd.vcpu_bo ==NULL )return 0 ; -cancel_delayed_work_sync (&adev->uvd.idle_work); +if (!amdgpu_sriov_vf (adev)) +cancel_delayed_work_sync (&adev->uvd.idle_work);for (i =0 ; i < adev->uvd.max_handles; ++i)if (atomic_read (&adev->uvd.handles[i]))
@@ -1116,9 +1117,6 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) container_of(work, struct amdgpu_device, uvd.idle_work.work);unsigned fences =amdgpu_fence_count_emitted (&adev->uvd.ring); -if (amdgpu_sriov_vf (adev)) -return ; -if (fences ==0 ) {if (adev->pm.dpm_enabled) {amdgpu_dpm_enable_uvd (adev, false);
@@ -1138,11 +1136,12 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work)void amdgpu_uvd_ring_begin_use (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev; -bool set_clocks = !cancel_delayed_work_sync (&adev->uvd.idle_work); +bool set_clocks;if (amdgpu_sriov_vf (adev))return ; + set_clocks = !cancel_delayed_work_sync (&adev->uvd.idle_work);if (set_clocks) {if (adev->pm.dpm_enabled) {amdgpu_dpm_enable_uvd (adev, true);
@@ -1158,7 +1157,8 @@ void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring)void amdgpu_uvd_ring_end_use (struct amdgpu_ring *ring) { -schedule_delayed_work (&ring->adev->uvd.idle_work,UVD_IDLE_TIMEOUT ); +if (!amdgpu_sriov_vf (ring->adev)) +schedule_delayed_work (&ring->adev->uvd.idle_work,UVD_IDLE_TIMEOUT ); }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -5
@@ -241,7 +241,8 @@ int amdgpu_vce_suspend(struct amdgpu_device *adev)if (i ==AMDGPU_MAX_VCE_HANDLES )return 0 ; -cancel_delayed_work_sync (&adev->vce.idle_work); +if (!amdgpu_sriov_vf (adev)) +cancel_delayed_work_sync (&adev->vce.idle_work);/* TODO: suspending running encoding sessions isn't supported */ return -EINVAL ; }
@@ -300,9 +301,6 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work) container_of(work, struct amdgpu_device, vce.idle_work.work);unsigned i, count =0 ; -if (amdgpu_sriov_vf (adev)) -return ; -for (i =0 ; i < adev->vce.num_rings; i++) count +=amdgpu_fence_count_emitted (&adev->vce.ring[i]);
@@ -362,7 +360,8 @@ void amdgpu_vce_ring_begin_use(struct amdgpu_ring *ring)*/ void amdgpu_vce_ring_end_use (struct amdgpu_ring *ring) { -schedule_delayed_work (&ring->adev->vce.idle_work,VCE_IDLE_TIMEOUT ); +if (!amdgpu_sriov_vf (ring->adev)) +schedule_delayed_work (&ring->adev->vce.idle_work,VCE_IDLE_TIMEOUT ); }/**
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): 1519721235-28552-2-git-send-email-Monk.Liu-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 1/6] drm/amdgpu: don't use MM idle_work for SRIOV(v2)
Am 27.02.2018 um 09:47 schrieb Monk Liu:
Quoted 20 lines
SRIOV doesn't give VF cg/pg feature so the MM's
idle_workis skipped for SR-IOVv2:
remove superfluous changes
sinceidle_workis not scheduled for SR-IOV so the condition check for SR-IOV insideidle_workalso can be droppedChange-Id I6dd7ea48d23b0fee74ecb9e93b53bfe36b0e8164Signed-off-by Monk Liu <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 12 ++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 9 ++++----- 2 files changed, 10 insertions(+), 11 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c int amdgpu_uvd_suspend(struct amdgpu_device *adev) · line 301
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +2
-1@@ -297,7 +297,8 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)if (adev->uvd.vcpu_bo ==NULL )return 0 ; -cancel_delayed_work_sync (&adev->uvd.idle_work); +if (!amdgpu_sriov_vf (adev)) +cancel_delayed_work_sync (&adev->uvd.idle_work);
Drop that, canceling a work item which was never scheduled is harmless and the code isn't performance critical.
Quoted 42 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4
-5for (i =0 ; i < adev->uvd.max_handles; ++i)if (atomic_read (&adev->uvd.handles[i]))@@ -1116,9 +1117,6 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) container_of(work, struct amdgpu_device, uvd.idle_work.work);unsigned fences =amdgpu_fence_count_emitted (&adev->uvd.ring); -if (amdgpu_sriov_vf (adev)) -return ; -if (fences ==0 ) {if (adev->pm.dpm_enabled) {amdgpu_dpm_enable_uvd (adev, false);@@ -1138,11 +1136,12 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work)void amdgpu_uvd_ring_begin_use (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev; -bool set_clocks = !cancel_delayed_work_sync (&adev->uvd.idle_work); +bool set_clocks;if (amdgpu_sriov_vf (adev))return ; + set_clocks = !cancel_delayed_work_sync (&adev->uvd.idle_work);if (set_clocks) {if (adev->pm.dpm_enabled) {amdgpu_dpm_enable_uvd (adev, true);@@ -1158,7 +1157,8 @@ void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring)void amdgpu_uvd_ring_end_use (struct amdgpu_ring *ring) { -schedule_delayed_work (&ring->adev->uvd.idle_work,UVD_IDLE_TIMEOUT ); +if (!amdgpu_sriov_vf (ring->adev)) +schedule_delayed_work (&ring->adev->uvd.idle_work,UVD_IDLE_TIMEOUT ); }/** drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c int amdgpu_vce_suspend(struct amdgpu_device *adev) · line 245
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +2
-1@@ -241,7 +241,8 @@ int amdgpu_vce_suspend(struct amdgpu_device *adev)if (i ==AMDGPU_MAX_VCE_HANDLES )return 0 ; -cancel_delayed_work_sync (&adev->vce.idle_work); +if (!amdgpu_sriov_vf (adev)) +cancel_delayed_work_sync (&adev->vce.idle_work);
Dito.
Apart from that the patch looks good to me,
Christian.
Quoted 14 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +0
-3/* TODO: suspending running encoding sessions isn't supported */ return -EINVAL ; }@@ -300,9 +301,6 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work) container_of(work, struct amdgpu_device, vce.idle_work.work);unsigned i, count =0 ; -if (amdgpu_sriov_vf (adev)) -return ; -for (i =0 ; i < adev->vce.num_rings; i++) count +=amdgpu_fence_count_emitted (&adev->vce.ring[i]);
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c void amdgpu_vce_ring_begin_use(struct amdgpu_ring *ring) · line 367
Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +2
-1@@ -362,7 +360,8 @@ void amdgpu_vce_ring_begin_use(struct amdgpu_ring *ring)*/ void amdgpu_vce_ring_end_use (struct amdgpu_ring *ring) { -schedule_delayed_work (&ring->adev->vce.idle_work,VCE_IDLE_TIMEOUT ); +if (!amdgpu_sriov_vf (ring->adev)) +schedule_delayed_work (&ring->adev->vce.idle_work,VCE_IDLE_TIMEOUT ); }/**
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/6 drm/amdgpu: block till late_init_work done in dev_fini
In reply to a message not in the archiveotherwise page fault hit in the work if driver
finished too quick
--- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +3 -1
@@ -2089,7 +2089,9 @@ void amdgpu_device_fini(struct amdgpu_device *adev) adev->firmware.gpu_info_fw =NULL ; } adev->accel_working = false; -cancel_delayed_work_sync (&adev->late_init_work); + +while (cancel_delayed_work_sync (&adev->late_init_work)) +msleep (1 );/* free i2c buses */ if (!amdgpu_device_has_dc_support (adev))amdgpu_i2c_fini (adev);
Signature · 6 lines
2.7.4 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 3/6 drm/amdgpu: should call drm_helper with dc support
In reply to a message not in the archiveotherwise DRM keep reporting complain on ATOMIC flags during kmd reloading
--- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +1 -1
@@ -345,7 +345,7 @@ int amdgpu_fbdev_init(struct amdgpu_device *adev)drm_fb_helper_single_add_all_connectors (&rfbdev->helper);/* disable all the possible outputs/crtcs before entering KMS mode */ -if (!amdgpu_device_has_dc_support (adev)) +if (amdgpu_device_has_dc_support (adev))drm_helper_disable_unused_functions (adev->ddev);drm_fb_helper_initial_config (&rfbdev->helper, bpp_sel);
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): 1519721235-28552-4-git-send-email-Monk.Liu-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
Quoted 14 lines · Monk Liu
otherwise DRM keep reporting complain on ATOMIC flags during kmd reloading
Change-Id I835b96e6d61c7995bbd5dd5478d056671dde9192Signed-off-by Monk Liu <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c int amdgpu_fbdev_init(struct amdgpu_device *adev) · line 349
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +1
-1@@ -345,7 +345,7 @@ int amdgpu_fbdev_init(struct amdgpu_device *adev)drm_fb_helper_single_add_all_connectors (&rfbdev->helper);/* disable all the possible outputs/crtcs before entering KMS mode */ -if (!amdgpu_device_has_dc_support (adev)) +if (amdgpu_device_has_dc_support (adev))drm_helper_disable_unused_functions (adev->ddev);
This function definitely needs to be called without DC. *If* it also needs to be called with DC (I'll leave that to the DC developers to judge), just remove the conditional.
Signature · 6 lines
Earthling Michel Dänzer | http://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): 9259bfa0-5b0f-10b6-7b26-396adf41d984-otUistvHUpPR7s880joybQ@public.gmane.org
RE: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
I'm not familiar with DC stuffs, I found without this patch DRM always reporting error like this:
[ 89.783235] [drm] Found UVD firmware Version: 1.87 Family ID: 17 [ 89.783285] [drm] PSP loading UVD firmware [ 89.784657] [drm] MM table gpu addr = 0xcfc000, cpu addr = 000000005ec858a9. [ 89.784883] [drm] Found VCE firmware Version: 53.45 Binary ID: 4 [ 89.784936] [drm] PSP loading VCE firmware [ 89.928790] [drm] UVD and UVD ENC initialized successfully. [ 89.928972] [drm] VCE initialized successfully. [ 89.936514] [drm:drm_helper_disable_unused_functions [drm_kms_helper]] *ERROR* Called for atomic driver, this is not what you want. [ 89.958458] [drm] fb mappable at 0xE1200000 [ 89.958460] [drm] vram apper at 0xE0000000 [ 89.958461] [drm] size 9216000 [ 89.958462] [drm] fb depth is 24 [ 89.958463] [drm] pitch is 7680
-----Original Message-----
From: Michel Dänzer [mailto:[email protected]]
Sent: 2018年2月27日 17:17
To: Liu, Monk <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
Quoted 15 lines · Monk Liu
otherwise DRM keep reporting complain on ATOMIC flags during kmd reloading
Change-Id I835b96e6d61c7995bbd5dd5478d056671dde9192Signed-off-by Monk Liu <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c int amdgpu_fbdev_init(struct amdgpu_device *adev) · line 349
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +1
-1@@ -345,7 +345,7 @@ int amdgpu_fbdev_init(struct amdgpu_device *adev)drm_fb_helper_single_add_all_connectors (&rfbdev->helper);/* disable all the possible outputs/crtcs before entering KMS mode */ -if (!amdgpu_device_has_dc_support (adev)) +if (amdgpu_device_has_dc_support (adev))drm_helper_disable_unused_functions (adev->ddev);
This function definitely needs to be called without DC. *If* it also needs to be called with DC (I'll leave that to the DC developers to judge), just remove the conditional.
Signature · 6 lines
Earthling Michel Dänzer | http://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): BLUPR12MB04490304A6CB9A6FD7B28E9D84C00-7LeqcoF/[email protected]
RE: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
Okay, I think maybe the problem is sr-iov related, I'll prepare another one
-----Original Message-----
From: Liu, Monk
Sent: 2018年2月27日 17:27
To: 'Michel Dänzer' <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
I'm not familiar with DC stuffs, I found without this patch DRM always reporting error like this:
[ 89.783235] [drm] Found UVD firmware Version: 1.87 Family ID: 17 [ 89.783285] [drm] PSP loading UVD firmware [ 89.784657] [drm] MM table gpu addr = 0xcfc000, cpu addr = 000000005ec858a9. [ 89.784883] [drm] Found VCE firmware Version: 53.45 Binary ID: 4 [ 89.784936] [drm] PSP loading VCE firmware [ 89.928790] [drm] UVD and UVD ENC initialized successfully. [ 89.928972] [drm] VCE initialized successfully. [ 89.936514] [drm:drm_helper_disable_unused_functions [drm_kms_helper]] *ERROR* Called for atomic driver, this is not what you want. [ 89.958458] [drm] fb mappable at 0xE1200000 [ 89.958460] [drm] vram apper at 0xE0000000 [ 89.958461] [drm] size 9216000 [ 89.958462] [drm] fb depth is 24 [ 89.958463] [drm] pitch is 7680
-----Original Message-----
From: Michel Dänzer [mailto:[email protected]]
Sent: 2018年2月27日 17:17
To: Liu, Monk <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
Quoted 15 lines · Monk Liu
otherwise DRM keep reporting complain on ATOMIC flags during kmd reloading
Change-Id I835b96e6d61c7995bbd5dd5478d056671dde9192Signed-off-by Monk Liu <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c int amdgpu_fbdev_init(struct amdgpu_device *adev) · line 349
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +1
-1@@ -345,7 +345,7 @@ int amdgpu_fbdev_init(struct amdgpu_device *adev)drm_fb_helper_single_add_all_connectors (&rfbdev->helper);/* disable all the possible outputs/crtcs before entering KMS mode */ -if (!amdgpu_device_has_dc_support (adev)) +if (amdgpu_device_has_dc_support (adev))drm_helper_disable_unused_functions (adev->ddev);
This function definitely needs to be called without DC. *If* it also needs to be called with DC (I'll leave that to the DC developers to judge), just remove the conditional.
Signature · 6 lines
Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Re: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
In reply to a message not in the archiveQuoted 1 line · Monk
I'm not familiar with DC stuffs, I found without this patch DRM always reporting error like this:
[...]
Quoted 1 line
[ 89.936514] [drm:drm_helper_disable_unused_functions [drm_kms_helper]] *ERROR* Called for atomic driver, this is not what you want.
"atomic driver" means DC is enabled, in which case drm_helper_disable_unused_functions shouldn't have been called.
Maybe the issue is that DRIVER_ATOMIC is a driver flag, whereas DC can be enabled or not per device. See the discussion on
https://patchwork.freedesktop.org/patch/200304/ .
Meanwhile, NAK for this patch.
Signature · 6 lines
Earthling Michel Dänzer | http://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): 5fab3cbe-a1a6-40f0-7123-bb5310ec5d7f-otUistvHUpPR7s880joybQ@public.gmane.org
RE: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
The cause is found:
In amdgpu_pci_probe(), we set support_atomic to true no matter sriov or BM, because "adev" is not created yet so cannot judge SRIOV or not that time ... But in amdgpu_fbdev_init(), we check if need to call that disable_unused_function() by amdgpu_device_has_dc_support() which would return false For SRIOV case,
That way we have two result checking DC support in amdgpu_pci_probe() and amdgpu_fbdev_init(), and lead to this error reporting.
-----Original Message-----
From: Michel Dänzer [mailto:[email protected]]
Sent: 2018年2月27日 17:38
To: Liu, Monk <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH 3/6] drm/amdgpu: should call drm_helper with dc support
Quoted 1 line · Monk
I'm not familiar with DC stuffs, I found without this patch DRM always reporting error like this:
[...]
Quoted 1 line
[ 89.936514] [drm:drm_helper_disable_unused_functions [drm_kms_helper]] *ERROR* Called for atomic driver, this is not what you want.
"atomic driver" means DC is enabled, in which case drm_helper_disable_unused_functions shouldn't have been called.
Maybe the issue is that DRIVER_ATOMIC is a driver flag, whereas DC can be enabled or not per device. See the discussion on https://patchwork.freedesktop.org/patch/200304/ .
Meanwhile, NAK for this patch.
Signature · 6 lines
Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 4/6 drm/amdgpu: adjust timeout for ib_ring_tests(v2)
In reply to a message not in the archiveissue:
sometime GFX/MM ib test hit timeout under SRIOV env, root cause is that engine doesn't come back soon enough so the current IB test considered as timed out.
fix:
for SRIOV GFX IB test wait time need to be expanded a lot during SRIOV runtimei mode since it couldn't really begin before GFX engine come back.
for SRIOV MM IB test it always need more time since MM scheduling is not go together with GFX engine, it is controled by h/w MM scheduler so no matter runtime or exclusive mode MM IB test always need more time.
v2:
use ring type instead of idx to judge
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 32 +++++++++++++++++++++++++++++++- 1 file changed, 31 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +31 -1
@@ -322,14 +322,44 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) {unsigned i;int r, ret =0 ; +long tmo_gfx, tmo_mm; + + tmo_mm = tmo_gfx =AMDGPU_IB_TEST_TIMEOUT ; +if (amdgpu_sriov_vf (adev)) { +/* for MM engines in hypervisor side they are not scheduled together +* with CP and SDMA engines, so even in exclusive mode MM engine could +* still running on other VF thus the IB TEST TIMEOUT for MM engines +* under SR-IOV should be set to a long time. +*/ + tmo_mm =8 *AMDGPU_IB_TEST_TIMEOUT ;/* 8 sec should be enough for the MM comes back to this VF */ + } + +if (amdgpu_sriov_runtime (adev)) { +/* for CP & SDMA engines since they are scheduled together so +* need to make the timeout width enough to cover the time +* cost waiting for it coming back under RUNTIME only +*/ + tmo_gfx =8 *AMDGPU_IB_TEST_TIMEOUT ; + }for (i =0 ; i <AMDGPU_MAX_RINGS ; ++i) {struct amdgpu_ring *ring = adev->rings[i]; +long tmo;if (!ring || !ring->ready)continue ; - r =amdgpu_ring_test_ib (ring,AMDGPU_IB_TEST_TIMEOUT ); +/* MM engine need more time */ +if (ring->funcs->type ==AMDGPU_RING_TYPE_UVD || + ring->funcs->type ==AMDGPU_RING_TYPE_VCE || + ring->funcs->type ==AMDGPU_RING_TYPE_UVD_ENC || + ring->funcs->type ==AMDGPU_RING_TYPE_VCN_DEC || + ring->funcs->type ==AMDGPU_RING_TYPE_VCN_ENC ) + tmo = tmo_mm; +else + tmo = tmo_gfx; + + r =amdgpu_ring_test_ib (ring, tmo);if (r) { ring->ready = false;
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): 1519721235-28552-5-git-send-email-Monk.Liu-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 4/6] drm/amdgpu: adjust timeout for ib_ring_tests(v2)
Am 27.02.2018 um 09:47 schrieb Monk Liu:
Quoted 32 lines
issue:
sometime GFX/MM ib test hit timeout under SRIOV env, root cause is that engine doesn't come back soon enough so the current IB test considered as timed out.fix:
for SRIOV GFX IB test wait time need to be expanded a lot during SRIOV runtimei mode since it couldn't really begin before GFX engine come back.for SRIOV MM IB test it always need more time since MM scheduling is not go together with GFX engine, it is controled by h/w MM scheduler so no matter runtime or exclusive mode MM IB test always need more time.
v2:
use ring type instead of idx to judgeChange-Id I0342371bc073656476ad850e1f5d9a021846dc8cSigned-off-by Monk Liu <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 32 +++++++++++++++++++++++++++++++- 1 file changed, 31 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +0
-0@@ -322,14 +322,44 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) {unsigned i;int r, ret =0 ;
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c int amdgpu_ib_ring_tests(struct amdgpu_device *adev) · line 334
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +10
-0+long tmo_gfx, tmo_mm; + + tmo_mm = tmo_gfx =AMDGPU_IB_TEST_TIMEOUT ; +if (amdgpu_sriov_vf (adev)) { +/* for MM engines in hypervisor side they are not scheduled together +* with CP and SDMA engines, so even in exclusive mode MM engine could +* still running on other VF thus the IB TEST TIMEOUT for MM engines +* under SR-IOV should be set to a long time. +*/ + tmo_mm =8 *AMDGPU_IB_TEST_TIMEOUT ;/* 8 sec should be enough for the MM comes back to this VF */
Why not adding the "8 sec should be enough for the MM comes back to this VF" to the comment above?
Apart from that the patch is Reviewed-by: Christian König <[email protected]>
Regards,
Christian.
Quoted 22 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +14
-1+ } + +if (amdgpu_sriov_runtime (adev)) { +/* for CP & SDMA engines since they are scheduled together so +* need to make the timeout width enough to cover the time +* cost waiting for it coming back under RUNTIME only +*/ + tmo_gfx =8 *AMDGPU_IB_TEST_TIMEOUT ; + }for (i =0 ; i <AMDGPU_MAX_RINGS ; ++i) {struct amdgpu_ring *ring = adev->rings[i]; +long tmo;if (!ring || !ring->ready)continue ; - r =amdgpu_ring_test_ib (ring,AMDGPU_IB_TEST_TIMEOUT ); +/* MM engine need more time */ + if (ring->funcs->type == AMDGPU_RING_TYPE_UVD || + ring->funcs->type == AMDGPU_RING_TYPE_VCE || + ring->funcs->type == AMDGPU_RING_TYPE_UVD_ENC ||
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c int amdgpu_ib_ring_tests(struct amdgpu_device *adev) · line 365
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +7
-0+ ring->funcs->type == AMDGPU_RING_TYPE_VCN_DEC || + ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) + tmo = tmo_mm; + else + tmo = tmo_gfx; + + r =amdgpu_ring_test_ib (ring, tmo);if (r) { ring->ready = false;
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 5/6 drm/amdgpu: Correct sdma_v4 get_wptr(v2)
In reply to a message not in the archiveFrom: Emily Deng <[email protected]>
the original method will change the wptr value in wb. v2:
furthur cleanup
--- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +8 -9
@@ -238,15 +238,14 @@ static uint64_t sdma_v4_0_ring_get_rptr(struct amdgpu_ring *ring)static uint64_t sdma_v4_0_ring_get_wptr (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev; -u64 *wptr =NULL ; +u64 wptr;uint64_t local_wptr =0 ; +u64 tmp =0 ;if (ring->use_doorbell) {/* XXX check if swapping is necessary on BE */ - wptr = ((u64 *)&adev->wb.wb[ring->wptr_offs]); -DRM_DEBUG ("wptr/doorbell before shift == 0x%016llx\n" , *wptr); - *wptr = (*wptr) >>2 ; -DRM_DEBUG ("wptr/doorbell after shift == 0x%016llx\n" , *wptr); + wptr =READ_ONCE (*((u64 *)&adev->wb.wb[ring->wptr_offs])); +DRM_DEBUG ("wptr/doorbell before shift == 0x%016llx\n" , wptr); }else {u32 lowbit, highbit;int me = (ring == &adev->sdma.instance[0 ].ring) ?0 :1 ;
@@ -257,12 +256,12 @@ static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring)DRM_DEBUG ("wptr [%i]high== 0x%08x low==0x%08x\n" , me, highbit, lowbit); - *wptr = highbit; - *wptr = (*wptr) <<32 ; - *wptr |= lowbit; + wptr = highbit; + wptr = wptr <<32 ; + wptr |= lowbit; } -return *wptr; +return wptr >>2 ; }/**
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): 1519721235-28552-6-git-send-email-Monk.Liu-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 5/6] drm/amdgpu: Correct sdma_v4 get_wptr(v2)
Am 27.02.2018 um 09:47 schrieb Monk Liu:
Quoted 18 lines
From: Emily Deng <[email protected]>
the original method will change the wptr value in wb. v2:
furthur cleanupChange-Id I984fabca35d9dcf1f5fa8ef7779b2afb7f7d7370Signed-off-by Emily Deng <[email protected]>Signed-off-by Monk Liu <[email protected]>--- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-)drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +0
-0
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c static uint64_t sdma_v4_0_ring_get_rptr(struct amdgpu_ring *ring) · line 243
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +2
-1@@ -238,15 +238,14 @@ static uint64_t sdma_v4_0_ring_get_rptr(struct amdgpu_ring *ring)static uint64_t sdma_v4_0_ring_get_wptr (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev; -u64 *wptr =NULL ; +u64 wptr;uint64_t local_wptr =0 ; +u64 tmp =0 ;
As far as I can see local_wptr and tmp are now unused and can be dropped as well.
With that fixed the patch is Reviewed-by: Christian König <[email protected]>
Regards,
Christian.
Quoted 19 lines
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +2
-7if (ring->use_doorbell) {/* XXX check if swapping is necessary on BE */ - wptr = ((u64 *)&adev->wb.wb[ring->wptr_offs]); -DRM_DEBUG ("wptr/doorbell before shift == 0x%016llx\n" , *wptr); - *wptr = (*wptr) >>2 ; -DRM_DEBUG ("wptr/doorbell after shift == 0x%016llx\n" , *wptr); + wptr =READ_ONCE (*((u64 *)&adev->wb.wb[ring->wptr_offs])); +DRM_DEBUG ("wptr/doorbell before shift == 0x%016llx\n" , wptr); }else {u32 lowbit, highbit;int me = (ring == &adev->sdma.instance[0 ].ring) ?0 :1 ;@@ -257,12 +256,12 @@ static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring)DRM_DEBUG ("wptr [%i]high== 0x%08x low==0x%08x\n" , me, highbit, lowbit); - *wptr = highbit; - *wptr = (*wptr) <<32 ; - *wptr |= lowbit;
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring) · line 267
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +4
-1+ wptr = highbit; + wptr = wptr <<32 ; + wptr |= lowbit; } -return *wptr; +return wptr >>2 ; }/**
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 6/6 drm/amdgpu: cleanup SA inti and fini(v2)
In reply to a message not in the archiveshould use bo_create_kernel instead of split to two function that create and pin the SA bo
issue:
before this patch, there are DMAR read error in host side when running SRIOV test, the DMAR address dropped in the range of SA bo.
fix:
after this cleanups of SA init and fini, above DMAR eror gone.
v2:
keep sa_bo's fini instead of suspend, to keep reporting error
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 6 --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 2 - drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c | 62 ++++++------------------------ 3 files changed, 11 insertions(+), 59 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +0 -6
@@ -279,11 +279,6 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev)return r; } - r =amdgpu_sa_bo_manager_start (adev, &adev->ring_tmp_bo); -if (r) { -return r; - } - adev->ib_pool_ready = true;if (amdgpu_debugfs_sa_init (adev)) {dev_err (adev->dev,"failed to register debugfs file for SA\n" );
@@ -302,7 +297,6 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev)void amdgpu_ib_pool_fini (struct amdgpu_device *adev) {if (adev->ib_pool_ready) { -amdgpu_sa_bo_manager_suspend (adev, &adev->ring_tmp_bo);amdgpu_sa_bo_manager_fini (adev, &adev->ring_tmp_bo); adev->ib_pool_ready = false; }
drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +0 -2
@@ -281,8 +281,6 @@ void amdgpu_sa_bo_manager_fini(struct amdgpu_device *adev, struct amdgpu_sa_manager *sa_manager);int amdgpu_sa_bo_manager_start (struct amdgpu_device *adev,struct amdgpu_sa_manager *sa_manager); -int amdgpu_sa_bo_manager_suspend (struct amdgpu_device *adev, -struct amdgpu_sa_manager *sa_manager);int amdgpu_sa_bo_new (struct amdgpu_sa_manager *sa_manager,struct amdgpu_sa_bo **sa_bo,unsigned size,unsigned align);
drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +11 -51
87 lines not on this page · raw message
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): 1519721235-28552-7-git-send-email-Monk.Liu-5C7GfCeVMHo@public.gmane.org
Christian König
Reviewed-byRe: [PATCH 6/6] drm/amdgpu: cleanup SA inti and fini(v2)
Am 27.02.2018 um 09:47 schrieb Monk Liu:
Quoted 12 lines
should use
bo_create_kernelinstead of split to two function that create and pin the SA boissue:
before this patch, there are DMAR read error in host side when running SRIOV test, the DMAR address dropped in the range of SA bo.fix:
after this cleanups of SA init and fini, above DMAR eror gone.
Quoted 6 lines
v2:
keep sa_bo's fini instead of suspend, to keep reporting errorChange-Id I3f299a3342bd7263776bff69e4b31b0d3816749aSigned-off-by Monk Liu <[email protected]>
Quoted 126 lines
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 6 --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 2 - drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c | 62 ++++++------------------------ 3 files changed, 11 insertions(+), 59 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +0
-6@@ -279,11 +279,6 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev)return r; } - r =amdgpu_sa_bo_manager_start (adev, &adev->ring_tmp_bo); -if (r) { -return r; - } - adev->ib_pool_ready = true;if (amdgpu_debugfs_sa_init (adev)) {dev_err (adev->dev,"failed to register debugfs file for SA\n" );@@ -302,7 +297,6 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev)void amdgpu_ib_pool_fini (struct amdgpu_device *adev) {if (adev->ib_pool_ready) { -amdgpu_sa_bo_manager_suspend (adev, &adev->ring_tmp_bo);amdgpu_sa_bo_manager_fini (adev, &adev->ring_tmp_bo); adev->ib_pool_ready = false; }drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +0
-2@@ -281,8 +281,6 @@ void amdgpu_sa_bo_manager_fini(struct amdgpu_device *adev, struct amdgpu_sa_manager *sa_manager);int amdgpu_sa_bo_manager_start (struct amdgpu_device *adev,struct amdgpu_sa_manager *sa_manager); -int amdgpu_sa_bo_manager_suspend (struct amdgpu_device *adev, -struct amdgpu_sa_manager *sa_manager);int amdgpu_sa_bo_new (struct amdgpu_sa_manager *sa_manager,struct amdgpu_sa_bo **sa_bo,unsigned size,unsigned align);drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +9
-46@@ -63,21 +63,27 @@ int amdgpu_sa_bo_manager_init(struct amdgpu_device *adev,for (i =0 ; i <AMDGPU_SA_NUM_FENCE_LISTS ; ++i)INIT_LIST_HEAD (&sa_manager->flist[i]); - r =amdgpu_bo_create (adev, size, align, true, domain, -0 ,NULL ,NULL , &sa_manager->bo); + r =amdgpu_bo_create_kernel (adev, size, align, domain, &sa_manager->bo, + &sa_manager->gpu_addr, &sa_manager->cpu_ptr);if (r) {dev_err (adev->dev,"(%d) failed to allocate bo for manager\n" , r);return r; } +memset (sa_manager->cpu_ptr,0 , sa_manager->size);return r; }void amdgpu_sa_bo_manager_fini (struct amdgpu_device *adev, -struct amdgpu_sa_manager *sa_manager) +struct amdgpu_sa_manager *sa_manager) {struct amdgpu_sa_bo *sa_bo, *tmp; +if (sa_manager->bo ==NULL ) { +dev_err (adev->dev,"no bo for sa manager\n" ); +return ; + } +if (!list_empty (&sa_manager->olist)) { sa_manager->hole = &sa_manager->olist,amdgpu_sa_bo_try_free (sa_manager);@@ -88,55 +94,9 @@ void amdgpu_sa_bo_manager_fini(struct amdgpu_device *adev,list_for_each_entry_safe (sa_bo, tmp, &sa_manager->olist, olist) {amdgpu_sa_bo_remove_locked (sa_bo); } -amdgpu_bo_unref (&sa_manager->bo); - sa_manager->size =0 ; -} - -int amdgpu_sa_bo_manager_start (struct amdgpu_device *adev, -struct amdgpu_sa_manager *sa_manager) -{ -int r; - -if (sa_manager->bo ==NULL ) { -dev_err (adev->dev,"no bo for sa manager\n" ); -return -EINVAL ; - } -/* map the buffer */ - r =amdgpu_bo_reserve (sa_manager->bo, false); -if (r) { -dev_err (adev->dev,"(%d) failed to reserve manager bo\n" , r); -return r; - } - r =amdgpu_bo_pin (sa_manager->bo, sa_manager->domain, &sa_manager->gpu_addr); -if (r) { -amdgpu_bo_unreserve (sa_manager->bo); -dev_err (adev->dev,"(%d) failed to pin manager bo\n" , r); -return r; - } - r =amdgpu_bo_kmap (sa_manager->bo, &sa_manager->cpu_ptr); -memset (sa_manager->cpu_ptr,0 , sa_manager->size); -amdgpu_bo_unreserve (sa_manager->bo); -return r; -} - -int amdgpu_sa_bo_manager_suspend (struct amdgpu_device *adev, -struct amdgpu_sa_manager *sa_manager) -{ -int r; - -if (sa_manager->bo ==NULL ) { -dev_err (adev->dev,"no bo for sa manager\n" ); -return -EINVAL ; - } - - r =amdgpu_bo_reserve (sa_manager->bo, true); -if (!r) {
drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c void amdgpu_sa_bo_manager_fini(struct amdgpu_device *adev, · line 102
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +2
-5-amdgpu_bo_kunmap (sa_manager->bo); -amdgpu_bo_unpin (sa_manager->bo); -amdgpu_bo_unreserve (sa_manager->bo); - } -return r; +amdgpu_bo_free_kernel (&sa_manager->bo, &sa_manager->gpu_addr, &sa_manager->cpu_ptr); + sa_manager->size =0 ; }static void amdgpu_sa_bo_remove_locked (struct amdgpu_sa_bo *sa_bo)
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx