On Fri, Apr 28, 2023 at 4:05 AM Tvrtko Ursulin
wrote:
>
>
> On 27/04/2023 18:53, Rob Clark wrote:
> > From: Rob Clark
> >
> > These are useful in particular for VM scenarios where the process which
> > has opened to drm device file is just a proxy for the real user in a VM
> > guest.
> >
> >
From: Rob Clark
Use the new helper to export stats about memory usage.
v2: Drop unintended hunk
v3: Rebase
Signed-off-by: Rob Clark
Reviewed-by: Emil Velikov
---
drivers/gpu/drm/msm/msm_drv.c | 2 ++
drivers/gpu/drm/msm/msm_gem.c | 15 +++
2 files changed, 17 insertions(+)
From: Rob Clark
These are useful in particular for VM scenarios where the process which
has opened to drm device file is just a proxy for the real user in a VM
guest.
v2: doc cleanups
Signed-off-by: Rob Clark
---
Documentation/gpu/drm-usage-stats.rst | 10 ++
From: Rob Clark
The restriction about no whitespace, etc, really only applies to the
usage of strings in keys. Values can contain anything (other than
newline).
Signed-off-by: Rob Clark
Acked-by: Tvrtko Ursulin
---
Documentation/gpu/drm-usage-stats.rst | 27 ++-
1
From: Rob Clark
Also store the override strings in drm_file so that fdinfo can display
them. We still need to keep our original copy as we could need these
override strings after the device file has been closed and drm_file
freed.
Signed-off-by: Rob Clark
---
From: Rob Clark
Now that we have a common helper, use it.
Signed-off-by: Rob Clark
Reviewed-by: Dmitry Baryshkov
---
drivers/gpu/drm/msm/msm_drv.c | 11 +--
drivers/gpu/drm/msm/msm_gpu.c | 2 --
2 files changed, 5 insertions(+), 8 deletions(-)
diff --git
From: Rob Clark
Add support to dump GEM stats to fdinfo.
v2: Fix typos, change size units to match docs, use div_u64
v3: Do it in core
v4: more kerneldoc
v5: doc fixes
Signed-off-by: Rob Clark
Reviewed-by: Emil Velikov
Reviewed-by: Daniel Vetter
---
Documentation/gpu/drm-usage-stats.rst |
From: Rob Clark
Handle a bit of the boiler-plate in a single case, and make it easier to
add some core tracked stats. This also ensures consistent behavior
across drivers for standardised fields.
v2: Update drm-usage-stats.rst, 64b client-id, rename drm_show_fdinfo
Reviewed-by: Daniel Vetter
From: Rob Clark
Signed-off-by: Rob Clark
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c| 3 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c | 16 ++--
drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h | 2 +-
3 files changed, 9 insertions(+), 12
From: Rob Clark
Similar motivation to other similar recent attempt[1]. But with an
attempt to have some shared code for this. As well as documentation.
It is probably a bit UMA-centric, I guess devices with VRAM might want
some placement stats as well. But this seems like a reasonable start.
From: Rob Clark
Fix a couple missing ':'s.
Signed-off-by: Rob Clark
Reviewed-by: Rodrigo Vivi
---
Documentation/gpu/drm-usage-stats.rst | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/Documentation/gpu/drm-usage-stats.rst
b/Documentation/gpu/drm-usage-stats.rst
index
On 4/29/2023 2:25 PM, Dmitry Baryshkov wrote:
The driver doesn't support hsic/memcolor and pcc SSPP subblocks.
Drop corresponding definitions.
Signed-off-by: Dmitry Baryshkov
---
Changes since v2:
- Fixed commit message to remove igc block mention.
Changes since v1:
- Rebased on top of
On 01/05/2023 21:49, Abhinav Kumar wrote:
On 4/29/2023 2:25 PM, Dmitry Baryshkov wrote:
The driver doesn't support hsic/memcolor and pcc SSPP subblocks.
Drop corresponding definitions.
Signed-off-by: Dmitry Baryshkov
---
Changes since v2:
- Fixed commit message to remove igc block
From: Rob Clark
For errors after msm_submitqueue_get(), we need to drop the submitqueue
reference. Additionally after get_unused_fd() we need to drop the fd.
The ordering for dropping the queue lock and put_unused_fd() is not
important, so just move this all into out_post_unlock.
v2: Only drop
On 5/1/2023 11:56 AM, Dmitry Baryshkov wrote:
On 01/05/2023 21:49, Abhinav Kumar wrote:
On 4/29/2023 2:25 PM, Dmitry Baryshkov wrote:
The driver doesn't support hsic/memcolor and pcc SSPP subblocks.
Drop corresponding definitions.
Signed-off-by: Dmitry Baryshkov
---
Changes since v2:
From: Rob Clark
For errors after msm_submitqueue_get(), we need to drop the submitqueue
reference. Additionally after get_unused_fd() we need to drop the fd.
The ordering for dropping the queue lock and put_unused_fd() is not
important, so just move this all into out_post_unlock.
Reported-by:
On 1.05.2023 14:39, Dmitry Baryshkov wrote:
> On 01/05/2023 15:37, Konrad Dybcio wrote:
>>
>>
>> On 1.05.2023 03:12, Dmitry Baryshkov wrote:
>>> If the dispcc uses CLK_OPS_PARENT_ENABLE (e.g. on QCM2290), CCF can try
>>> enabling VCO before the rate has been programmed. This can cause clock
>>>
On 1.05.2023 03:12, Dmitry Baryshkov wrote:
> If the dispcc uses CLK_OPS_PARENT_ENABLE (e.g. on QCM2290), CCF can try
> enabling VCO before the rate has been programmed. This can cause clock
> lockups and/or other boot issues. Program the VCO to the minimal PLL
> rate if the read rate is 0 Hz.
On 01/05/2023 15:37, Konrad Dybcio wrote:
On 1.05.2023 03:12, Dmitry Baryshkov wrote:
If the dispcc uses CLK_OPS_PARENT_ENABLE (e.g. on QCM2290), CCF can try
enabling VCO before the rate has been programmed. This can cause clock
lockups and/or other boot issues. Program the VCO to the minimal
On 5/1/2023 1:45 PM, Dmitry Baryshkov wrote:
On 01/05/2023 22:58, Abhinav Kumar wrote:
On 4/30/2023 4:57 PM, Dmitry Baryshkov wrote:
There is no reason to split the dpu_encoder interface into separate
_init() and _setup() phases. Merge them into a single function.
I think the reason
On Mon, May 1, 2023 at 1:56 PM Dmitry Baryshkov
wrote:
>
> On 01/05/2023 23:44, Rob Clark wrote:
> > From: Rob Clark
> >
> > Fixes the following splat on a6xx gen2+ (a640, a650, a660 families),
> > a6xx gen1 has smaller GMU allocations so they fit under the default
> > 64K max segment size.
> >
On 2023-04-30 23:35:53, Dmitry Baryshkov wrote:
> Using IS_ERR_OR_NULL() together with PTR_ERR() is a typical mistake. If
> the value is NULL, then the function will return 0 instead of a proper
> return code. Moreover dpu_hw_dsc_init() can not return NULL.
More specifically, it allows the init
On 02/05/2023 01:17, Marijn Suijten wrote:
On 2023-04-30 23:35:55, Dmitry Baryshkov wrote:
On msm8998/sdm845 some LM blocks do not have corresponding PINGPONG
block. Currently the driver uses PINGPONG_MAX for such cases. Switch
that to use PINGPONG_NONE instead, which is more logical.
On 2023-05-02 01:15:24, Dmitry Baryshkov wrote:
> On 02/05/2023 01:13, Marijn Suijten wrote:
> > On 2023-04-30 23:35:54, Dmitry Baryshkov wrote:
> >> dpu_rm_init() contains checks for block->id values. These were logical
> >> in the vendor driver, when one can not be sure which values were passed
On 4/28/2023 5:30 PM, Dmitry Baryshkov wrote:
On 29/04/2023 02:45, Kuogee Hsieh wrote:
Add support for DSC 1.2 by providing the necessary hooks to program
the DPU DSC 1.2 encoder.
Reported-by: kernel test robot
What exactly was reported?
Signed-off-by: Kuogee Hsieh
---
From: Rob Clark
Fixes the following splat on a6xx gen2+ (a640, a650, a660 families),
a6xx gen1 has smaller GMU allocations so they fit under the default
64K max segment size.
[ cut here ]
DMA-API: msm_dpu ae01000.display-controller: mapping sg segment longer than
On 5/1/2023 1:50 PM, Dmitry Baryshkov wrote:
On 01/05/2023 23:40, Kuogee Hsieh wrote:
On 4/28/2023 5:30 PM, Dmitry Baryshkov wrote:
On 29/04/2023 02:45, Kuogee Hsieh wrote:
Add support for DSC 1.2 by providing the necessary hooks to program
the DPU DSC 1.2 encoder.
Reported-by: kernel
On 4/30/2023 1:57 PM, Dmitry Baryshkov wrote:
The function dpu_plane_sspp_update_pipe() contains code to skip enabling
the QoS and OT limitis for CURSOR pipes. However all DPU since sdm845
repurpose DMA SSPP for the cursor planes because they lack the real
CURSOR SSPP. Fix the condition to
On 01/05/2023 23:40, Kuogee Hsieh wrote:
On 4/28/2023 5:30 PM, Dmitry Baryshkov wrote:
On 29/04/2023 02:45, Kuogee Hsieh wrote:
Add support for DSC 1.2 by providing the necessary hooks to program
the DPU DSC 1.2 encoder.
Reported-by: kernel test robot
What exactly was reported?
On 2023-04-30 23:35:55, Dmitry Baryshkov wrote:
> On msm8998/sdm845 some LM blocks do not have corresponding PINGPONG
> block. Currently the driver uses PINGPONG_MAX for such cases. Switch
> that to use PINGPONG_NONE instead, which is more logical.
dpu_lm_cfg::pingpong still has a doc-comment
On 01/05/2023 22:58, Abhinav Kumar wrote:
On 4/30/2023 4:57 PM, Dmitry Baryshkov wrote:
There is no reason to split the dpu_encoder interface into separate
_init() and _setup() phases. Merge them into a single function.
I think the reason for having this split was to pass a valid encoder
On 01/05/2023 23:44, Rob Clark wrote:
From: Rob Clark
Fixes the following splat on a6xx gen2+ (a640, a650, a660 families),
a6xx gen1 has smaller GMU allocations so they fit under the default
64K max segment size.
[ cut here ]
DMA-API: msm_dpu
On 1.05.2023 21:40, Akhil P Oommen wrote:
> On Fri, Mar 31, 2023 at 01:25:20AM +0200, Konrad Dybcio wrote:
>> Some (particularly SMD_RPM, a.k.a non-RPMh) SoCs implement A6XX GPUs
>> but don't implement the associated GMUs. This is due to the fact that
>> the GMU directly pokes at RPMh. Sadly,
On 2023-04-30 23:35:54, Dmitry Baryshkov wrote:
> dpu_rm_init() contains checks for block->id values. These were logical
> in the vendor driver, when one can not be sure which values were passed
> from DT. In the upstream driver this is not necessary: the catalog is a
> part of the driver, we
On 2023-04-30 23:35:56, Dmitry Baryshkov wrote:
> Move the check for lm->pingpong being not NONE from dpu_rm_init() to
> dpu_lm_init(), following the change to dpu_hw_intf_init().
>
> Signed-off-by: Dmitry Baryshkov
In some way I also suggested this change in [1], but did not receive a
(N)ACK.
On 4/30/2023 1:57 PM, Dmitry Baryshkov wrote:
Get rid of intermediatory configuration structure and defines. Pass the
format and the enablement bit directly to the new helper. The
WB_CDP_CNTL register ignores BIT(2), so we can write it for both SSPP
and WB CDP settings.
Signed-off-by: Dmitry
On Fri, Mar 31, 2023 at 01:25:20AM +0200, Konrad Dybcio wrote:
> Some (particularly SMD_RPM, a.k.a non-RPMh) SoCs implement A6XX GPUs
> but don't implement the associated GMUs. This is due to the fact that
> the GMU directly pokes at RPMh. Sadly, this means we have to take care
> of enabling &
On 4/30/2023 4:57 PM, Dmitry Baryshkov wrote:
There is no reason to split the dpu_encoder interface into separate
_init() and _setup() phases. Merge them into a single function.
I think the reason for having this split was to pass a valid encoder to
the interface_modeset_init() and then
On 02/05/2023 00:22, Abhinav Kumar wrote:
On 5/1/2023 1:45 PM, Dmitry Baryshkov wrote:
On 01/05/2023 22:58, Abhinav Kumar wrote:
On 4/30/2023 4:57 PM, Dmitry Baryshkov wrote:
There is no reason to split the dpu_encoder interface into separate
_init() and _setup() phases. Merge them into a
On 02/05/2023 01:13, Marijn Suijten wrote:
On 2023-04-30 23:35:54, Dmitry Baryshkov wrote:
dpu_rm_init() contains checks for block->id values. These were logical
in the vendor driver, when one can not be sure which values were passed
from DT. In the upstream driver this is not necessary: the
On 02/05/2023 01:23, Marijn Suijten wrote:
On 2023-04-30 23:35:56, Dmitry Baryshkov wrote:
Move the check for lm->pingpong being not NONE from dpu_rm_init() to
dpu_lm_init(), following the change to dpu_hw_intf_init().
Signed-off-by: Dmitry Baryshkov
In some way I also suggested this change
41 matches
Mail list logo