Hi Dave, Daniel,
A bit bigger than normal, but this is several weeks of fixes.
The following changes since commit d7ca2d19c751b6715e9cb899a6b94f47b3499d02:
Merge tag 'drm-msm-next-2020-01-14' of https://gitlab.freedesktop.org/drm/msm
into drm-next (2020-01-20 14:09:43 +1000)
are available
Acked-by: Evan Quan
-Original Message-
From: amd-gfx On Behalf Of Jack Zhang
Sent: Wednesday, February 5, 2020 5:18 PM
To: amd-gfx@lists.freedesktop.org
Cc: Zhang, Jack (Jian)
Subject: [PATCH] drm/amdgpu/sriov Don't send msg when smu suspend
For sriov and pp_onevf_mode, do not send
On Wed, Feb 5, 2020 at 4:18 AM Jack Zhang wrote:
>
> For sriov and pp_onevf_mode, do not send message to set smu
> status, becasue smu doesn't support these messages under VF.
Typo: becasue -> because
With that fixed:
Acked-by: Alex Deucher
>
> Besides, it should skip smu_suspend when
On 2020-02-05 10:22 a.m., Christian König wrote:
Make use of the better performance here as well.
This patch is only compile tested!
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 38 +++--
1 file changed, 23 insertions(+), 15 deletions(-)
If we're using the BAR, we should probably flush HDP cache/buffers
before reading or after writing.
Regards,
Felix
On 2020-02-05 10:22 a.m., Christian König wrote:
This should speed up debugging VRAM access a lot.
Signed-off-by: Christian König
---
Please disregard the patch 5 and 6, as I have a new version for them.
Yong
On 2020-02-05 6:39 p.m., Yong Zhao wrote:
One minor fix added.
Yong
On 2020-02-05 6:28 p.m., Yong Zhao wrote:
The sdma_queue_count was only used for inferring whether we should
unmap SDMA queues under HWS mode. In
Those printings are duplicated or useless.
Change-Id: I88fbe8f5748bbd0a20bcf1f6ca67b9dde99733fe
Signed-off-by: Yong Zhao
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 --
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 4 +---
2 files changed, 1 insertion(+), 5
The sdma_queue_count was only used for inferring whether we should
unmap SDMA queues under HWS mode. In contrast, We only mapped active
queues rather than all in map_queues_cpsch(). In order to match the
map and unmap for SDMA queues, we should just count active SDMA
queues.
Moreover, previously
MAX_SDMA_ENGINE_NUM will be used in more than one place.
Change-Id: I99c84086ee62612b373c547a9d29bc4a69e7c72e
Signed-off-by: Yong Zhao
---
drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h| 2 +-
drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 1 +
2 files changed, 2 insertions(+), 1
One minor fix added.
Yong
On 2020-02-05 6:28 p.m., Yong Zhao wrote:
The sdma_queue_count was only used for inferring whether we should
unmap SDMA queues under HWS mode. In contrast, We mapped active queues
rather than all in map_queues_cpsch(). In order to match the map and unmap
for SDMA
Those printings are duplicated or useless.
Change-Id: I88fbe8f5748bbd0a20bcf1f6ca67b9dde99733fe
Signed-off-by: Yong Zhao
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 --
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 4 +---
2 files changed, 1 insertion(+), 5
When the queue creation is failed, some resources were not freed. Fix it.
Change-Id: Ia24b6ad31528dceddfd4d1c58bb1d22c35d3eabf
Signed-off-by: Yong Zhao
---
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 3 +++
1 file changed, 3 insertions(+)
diff --git
The queues represented in queue_bitmap are only CP queues.
Change-Id: I7e6a75de39718d7c4da608166b85b9377d06d1b3
Signed-off-by: Yong Zhao
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 4 ++--
.../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c| 12 ++--
The sdma_queue_count was only used for inferring whether we should
unmap SDMA queues under HWS mode. In contrast, We mapped active queues
rather than all in map_queues_cpsch(). In order to match the map and unmap
for SDMA queues, we should just count active SDMA queues. Meanwhile,
rename
The previous code of calculating active CP queues is problematic if
some SDMA queues are inactive. Fix that by counting CP queues directly.
Change-Id: I5ffaa75a95cbebc984558199ba2f3db6909c52a9
Signed-off-by: Yong Zhao
---
.../drm/amd/amdkfd/kfd_device_queue_manager.c | 45 +--
The name is easier to understand the code.
Change-Id: I9064dab1d022e02780023131f940fff578a06b72
Signed-off-by: Yong Zhao
---
.../drm/amd/amdkfd/kfd_device_queue_manager.c | 38 +--
.../drm/amd/amdkfd/kfd_device_queue_manager.h | 2 +-
.../gpu/drm/amd/amdkfd/kfd_packet_manager.c
Series is:
Reviewed-by: Alex Deucher
On Wed, Feb 5, 2020 at 10:22 AM Christian König
wrote:
>
> Make use of the better performance here as well.
>
> This patch is only compile tested!
>
> Signed-off-by: Christian König
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 38
just a couple of typos in the patch title:
drm\amdgpu: [DALI] Dali Varient Detection
It should be:
drm/amdgpu: [DALI] Dali Variant Detection
With that fixed, patch is:
Reviewed-by: Alex Deucher
On Wed, Feb 5, 2020 at 9:59 AM Tawfik, Aly wrote:
>
> Hi,
>
>
>
> Dali is a raven2 based asic that
Applied. Thanks!
Alex
On Wed, Feb 5, 2020 at 1:44 PM Bhawanpreet Lakha
wrote:
>
> Reviewed-by: Bhawanpreet Lakha
>
> On 2020-02-05 1:38 p.m., Dan Carpenter wrote:
> > These frees need to be re-ordered so that we don't dereference "hdcp_work"
> > right after it's freed. Also in
On Wed, Feb 5, 2020 at 11:55 AM Takashi Iwai wrote:
>
> On Thu, 28 Nov 2019 15:35:23 +0100,
> Harry Wentland wrote:
> >
> > On 2019-11-28 3:27 a.m., Takashi Iwai wrote:
> > > Currently, gcc spews a warning as:
> > > drivers/gpu/drm/amd/amdgpu/../display/dc/dcn10/dcn10_hubbub.c: In
> > >
Reviewed-by: Bhawanpreet Lakha
On 2020-02-05 1:38 p.m., Dan Carpenter wrote:
These frees need to be re-ordered so that we don't dereference "hdcp_work"
right after it's freed. Also in hdcp_create_workqueue() there is a
problem that "hdcp_work" can be NULL if the allocation fails so it would
Hi Andrey,
This commit breaks all drivers, which may bail out of the timeout
processing as they wish to extend the timeout (etnaviv, v3d).
Those drivers currently just return from the timeout handler before
calling drm_sched_stop(), which means with this commit applied we are
removing the first
These frees need to be re-ordered so that we don't dereference "hdcp_work"
right after it's freed. Also in hdcp_create_workqueue() there is a
problem that "hdcp_work" can be NULL if the allocation fails so it would
lead to a NULL dereference in the cleanup code.
Fixes: 9aeb8a134a0a
[AMD Public Use]
Tested on Vega20 via proc mem op reads.
Old MMIO ~2.7MB/s, Improved MMIO ~3.2MB/s, BAR ~44MB/s
Acked-by: Jonathan Kim
-Original Message-
From: Christian König
Sent: Wednesday, February 5, 2020 10:23 AM
To: amd-gfx@lists.freedesktop.org; Kuehling, Felix ;
Kim,
On Thu, 28 Nov 2019 15:35:23 +0100,
Harry Wentland wrote:
>
> On 2019-11-28 3:27 a.m., Takashi Iwai wrote:
> > Currently, gcc spews a warning as:
> > drivers/gpu/drm/amd/amdgpu/../display/dc/dcn10/dcn10_hubbub.c: In
> > function ‘hubbub1_verify_allow_pstate_change_high’:
> >
On 2020-02-05 10:22 a.m., Jerry (Fangzhi) Zuo wrote:
> Unlike DP 1.2 edid corruption test, DP 1.4 requires to calculate
> real CRC value of the last edid data block, and write it back.
> Current edid CRC calculates routine adds the last CRC byte,
> and check if non-zero.
>
> This behavior is
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for gem.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 4
To handle debugfs setup on non DP MST connectors.
Reviewed-by: Harry Wentland
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
.../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 18 ++
1 file changed, 14 insertions(+), 4 deletions(-)
diff --git
The core does this for us now.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c| 1 -
drivers/gpu/drm/amd/amdgpu/dce_virtual.c | 1 -
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 1 -
3 files changed, 3 deletions(-)
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for rings.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 23 -
Split into init and register functions to avoid a segfault
in some configs when the load/unload callbacks are removed.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c| 10 ++
drivers/gpu/drm/amd/amdgpu/atombios_dp.c | 8 +---
We've moved the debugfs handling into a centralized place
so we can remove the legacy load an unload callbacks.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 -
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c| 13 +++--
2
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for register access files.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 4
Into the function that creates the debugfs files rather
than setting them explicitly in the callers.
Reviewed-by: Harry Wentland
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 2 --
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for fence handling.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 3 +++
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for pm.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 7 +++
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for display.
Reviewed-by: Harry Wentland
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c |
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for ttm.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 10 ++
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for SA (sub allocator).
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 4
In order to remove the load and unload drm callbacks,
we need to reorder the init sequence to move all the drm
debugfs file handling. Do this for firmware.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 4
to amdgpu_debugfs_fini. It will be used for other things in
the future.
Acked-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 4 ++--
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
3
These are deprecated and the drm will soon start warning when drivers still
use them. It was a long and twisty road, but seems to work.
v2: Add additional patch (13/15) which should fix the crash reported by
Thomas Zimmermann.
Alex Deucher (15):
drm/amdgpu: rename
On 2020-02-05 9:45 a.m., James Zhu wrote:
Support pause_state for multiple instance, and it will fix vcn2.5 DPG mode
power off issue on instance 1.
Signed-off-by: James Zhu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h | 3 +--
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 14 --
Unlike DP 1.2 edid corruption test, DP 1.4 requires to calculate
real CRC value of the last edid data block, and write it back.
Current edid CRC calculates routine adds the last CRC byte,
and check if non-zero.
This behavior is not accurate; actually, we need to return
the actual CRC value when
Make use of the better performance here as well.
This patch is only compile tested!
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 38 +++--
1 file changed, 23 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
This should speed up debugging VRAM access a lot.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 21 +
1 file changed, 21 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
Only write the _HI register when necessary.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 17 +++--
1 file changed, 11 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
This speeds up the access quite a bit from 2.2 MB/s to
2.9 MB/s on 32bit and 12,8 MB/s on 64bit.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 27 ++---
1 file changed, 11 insertions(+), 16 deletions(-)
diff --git
Hi,
Dali is a raven2 based asic that drives at a lower (6W) TDP than other raven 2
chips. Currently the fused internal id is the same on all raven 2 boards, this
means that the detection process must be done through PCIE REV ID.
Unfortunately PCIE REV ID is not defined inside the scope of
Support pause_state for multiple instance, and it will fix vcn2.5 DPG mode
power off issue on instance 1.
Signed-off-by: James Zhu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h | 3 +--
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 14 --
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 6
Patches 10-12 are
Reviewed-by: Harry Wentland
Harry
On 2020-02-04 10:48 p.m., Alex Deucher wrote:
> These are deprecated and the drm will soon start warning when drivers still
> use them. It was a long and twisty road, but seems to work.
>
> Alex Deucher (14):
> drm/amdgpu: rename
Am 05.02.20 um 13:56 schrieb Pan, Xinhui:
No need to trigger eviction as the memory mapping will not be used anymore.
All pt/pd bos share same resv, hence the same shared eviction fence. Everytime
page table is freed, the fence will be signled and that cuases kfd unexcepted
evictions.
kfd bo
No need to trigger eviction as the memory mapping will not be used anymore.
All pt/pd bos share same resv, hence the same shared eviction fence. Everytime
page table is freed, the fence will be signled and that cuases kfd unexcepted
evictions.
kfd bo uses its own resv, so it is not affetced.
Hi, Team,
Would you please help to take a look at this patch?
Regards,
Jack
-Original Message-
From: amd-gfx On Behalf Of Jack Zhang
Sent: Wednesday, February 5, 2020 5:18 PM
To: amd-gfx@lists.freedesktop.org
Cc: Zhang, Jack (Jian)
Subject: [PATCH] drm/amdgpu/sriov Don't send msg when
For sriov and pp_onevf_mode, do not send message to set smu
status, becasue smu doesn't support these messages under VF.
Besides, it should skip smu_suspend when pp_onevf_mode is disabled.
Signed-off-by: Jack Zhang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 ---
Hi Alex
Am 05.02.20 um 04:48 schrieb Alex Deucher:
> These are deprecated and the drm will soon start warning when drivers still
> use them. It was a long and twisty road, but seems to work.
>
> Alex Deucher (14):
> drm/amdgpu: rename amdgpu_debugfs_preempt_cleanup
> drm/amdgpu/ttm: move
Am 05.02.20 um 04:48 schrieb Alex Deucher:
These are deprecated and the drm will soon start warning when drivers still
use them. It was a long and twisty road, but seems to work.
Acked-by: Christian König for the whole series.
Alex Deucher (14):
drm/amdgpu: rename
57 matches
Mail list logo
