Am 19.09.23 um 16:17 schrieb Philip Yang:
On GFX v9.4.3 dGPU, applications have random timeout failure when XNACK
on, dmesg log has "amdgpu: IH soft ring buffer overflow 0x900, 0x900",
because dGPU mode has 272 cam entries. After increasing IH soft ring
to 512 entries, no more IH soft ring
Am 19.09.23 um 16:09 schrieb Philip Yang:
If retry cam enabled, we don't use sw retry fault filter and add fault
into sw filter ring, so we shouldn't remove fault from sw filter.
Signed-off-by: Philip Yang
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5
From: Wenjing Liu
[why]
The current ODM change detection only compares first two ODM slices.
If there are 4 ODM slices and the change is within the last two slices, the
logic fails to detect ODM change and cause us to skip ODM programming
unexpectedly.
[how]
Add a is ODM topology changed
From: Aric Cyr
This version brings along the following:
- Fix incorrect odm change detection logic
- Add primary pipe check when building slice table for dcn3x
- Add get primary dpp pipe resource interface
- Add missing function pointer for DCN321 resource
- Rename DisableMinDispClkODM in
From: Wenjing Liu
[why]
There is a corner use case where the same plane pointer can be added into
dc state twice. In addition to plane state pointer check we also need to match
primary pipe before identifying the same plane.
Reviewed-by: Dillon Varone
Acked-by: Wayne Lin
Signed-off-by:
From: Wenjing Liu
[why]
Need to have a helper function to find the primary dp pipe of the plane
associated with a dpp pipe
Reviewed-by: Dillon Varone
Acked-by: Wayne Lin
Signed-off-by: Wenjing Liu
---
drivers/gpu/drm/amd/display/dc/core/dc_resource.c | 14 ++
From: Alvin Lee
[Description]
After repopulating pipes for a split / merge, we want to find the
lowest vlevel that supports the desired mpc combine level.
Reviewed-by: Nevenko Stupar
Reviewed-by: Dillon Varone
Reviewed-by: Aric Cyr
Acked-by: Wayne Lin
Signed-off-by: Alvin Lee
---
From: Wenjing Liu
[why]
acquire_free_pipe_as_secondary_opp_head function pointer is not
assigned. This causes ODM mode to be disabled after enabling
windowed MPO ODM capability on DCN321.
[how]
assign the correct function to the function pointer.
Reviewed-by: Dillon Varone
Acked-by: Wayne Lin
From: Gabe Teeger
[what and why]
The logic for DisableMinDispClkODM is reversed now that odm is
disabled by default. Renaming to EnableMinDispClkODM.
Reviewed-by: Charlene Liu
Acked-by: Wayne Lin
Signed-off-by: Gabe Teeger
---
drivers/gpu/drm/amd/display/dc/dc.h | 2 +-
1 file changed, 1
From: Max Tseng
[Why]
Allow dc report maximum display clock possible at vmin
Reviewed-by: Wayne Lin
Acked-by: Wayne Lin
Signed-off-by: Max Tseng
---
drivers/gpu/drm/amd/display/dc/dc.h | 1 +
drivers/gpu/drm/amd/display/dc/dcn314/dcn314_resource.c | 2 ++
From: Hong-lu Cheng
[Description]
Change to new interface for notifying OS of cursor support changes
Reviewed-by: Gang Chu
Reviewed-by: Jaina Han
Reviewed-by: Jun Lei
Acked-by: Wayne Lin
Signed-off-by: Hong-lu Cheng
---
drivers/gpu/drm/amd/display/dc/core/dc.c | 2 +-
1 file changed, 1
From: Rodrigo Siqueira
bios_parser2 uses space instead of tabs in multiple areas; this commit
converts those spaces into tabs. It also removes some other small
details.
Reviewed-by: Aurabindo Pillai
Acked-by: Wayne Lin
Signed-off-by: Rodrigo Siqueira
---
From: Alex Hung
[WHY & HOW]
IGT's kms_hdmi_inject (subtest inject_audio) triggers a warning message
from dce_aud_az_configure. This can be fixed by checking Virtual signal
that does not need to configure AZ audio since it does not have any.
Reviewed-by: Rodrigo Siqueira
Acked-by: Wayne Lin
From: Dmytro Laktyushkin
This change adds a method to block mpo when it would cause us to
exit pstate support. We should block mpo by default where MPO
causes issues by preventing pstate change.
Reviewed-by: Jun Lei
Reviewed-by: Aric Cyr
Acked-by: Wayne Lin
Signed-off-by: Dmytro Laktyushkin
From: Duncan Ma
[Why]
There is a race condition between x86 and dmcub fw when attempting to exit
IPS2. Scenarios including exiting IPS2 before IPS2 has been entered. This
can cause unexpected hang when DMCUB attempt to exit while PMFW still
tries to enter IPS2.
[How]
A new design has been
From: Ilya Bakoulin
[Why]
Triggering OTG sync before all OTG/HPO clock programming is complete
causes timing sync to fail and a subsequent P-state hang.
[How]
Move DTB clock programming earlier in the sequence to
enable_stream_timing.
Reviewed-by: Ariel Bernstein
Acked-by: Wayne Lin
From: Wenjing Liu
[why]
With the more generic hw minimal state transition sequence,
this limitation has been overcome.
Reviewed-by: Aric Cyr
Acked-by: Wayne Lin
Signed-off-by: Wenjing Liu
---
drivers/gpu/drm/amd/display/dc/core/dc.c | 34 ---
From: Wenjing Liu
[why]
We do not have a programming sequence to support seamless transition
between MPC combine to ODM combine when plane count remains the same.
This can happen with windowed MPO ODM use cases when plane scaling
is changed over ODM Combine's capability.
[How]
We are adding a
From: Wenjing Liu
[why]
ODM combine could prevent us from supporting more planes
we will reset ODM slice count back to 1 when all planes have
been removed to maximize the amount of planes supported when
new planes are added.
[how]
reset ODM slice count when all planes are removed.
Reviewed-by:
From: Wenjing Liu
[why]
commit minimal transition state would update current
state to new state with surface and stream update
applied. If we determine fast update only after we
have committed the minimal transition state based on
new state, we will skip committing the full new state.
[how]
From: Qingqing Zhuo
[Why & How]
mmhub_reg_offsets never used in code. Remove it.
Reviewed-by: Rodrigo Siqueira
Acked-by: Wayne Lin
Signed-off-by: Qingqing Zhuo
---
drivers/gpu/drm/amd/display/dc/dcn35/dcn35_resource.c | 10 --
1 file changed, 10 deletions(-)
diff --git
This DC patchset brings improvements in multiple areas. In summary, we
highlight:
- fix incorrect odm change detection logic
- add primary pipe check when building slice table for dcn3x
- add get primary dpp pipe resource interface
- add missing function pointer for DCN321 resource
- Rename
Hi Alex,
On 9/18/2023 10:05 PM, Alex Deucher wrote:
> On Mon, Sep 11, 2023 at 2:00 AM Ma Jun wrote:
>>
>> Add reset option for fan_curve.
>> User can use command "echo r > fan_cure" to reset the fan_curve
>> to boot value
>>
>> Signed-off-by: Ma Jun
>> ---
>> drivers/gpu/drm/amd/pm/amdgpu_pm.c
Hi,
On 2023/8/25 21:18, Deucher, Alexander wrote:
[Public]
-Original Message-
From: amd-gfx On Behalf Of Sui
Jingfeng
Sent: Friday, August 25, 2023 2:27 AM
To: Bjorn Helgaas
Cc: alsa-de...@alsa-project.org; Sui Jingfeng ;
nouv...@lists.freedesktop.org; linux-ker...@vger.kernel.org;
[AMD Official Use Only - General]
Thanks for catch it, will update the patch.
Tao
> -Original Message-
> From: Wang, Yang(Kevin)
> Sent: Tuesday, September 19, 2023 11:34 PM
> To: Zhou1, Tao ; amd-gfx@lists.freedesktop.org; Zhang,
> Hawking ; Yang, Stanley ;
> Li, Candice ; Chai,
On 2023-09-15 18:13, Xiaogang.Chen wrote:
From: Xiaogang Chen
We do not need free dma address array of svm_range each time we do dma unmap
for pages in svm_range as we can reuse the same array. Only free it when free
svm_range. Seperate these two operations and use them accordinly.
On 2023-09-19 10:21, Philip Yang wrote:
If new range is splited to multiple pranges with max_svm_range_pages
alignment and added to update_list, svm validate and map should keep
going after error to make sure prange->mapped_to_gpu flag is up to date
for the whole range.
svm validate and map
On Tue, Sep 19, 2023 at 8:08 AM Christian König
wrote:
>
> The KIQ code path was ignoring the second flush. Also avoid long lines and
> re-calculating the register offsets over and over again.
>
> Signed-off-by: Christian König
Reviewed-by: Alex Deucher
> ---
>
DRM CI keeps track of which tests are failing, flaking or being skipped
by the ci in the expectations files. Add entries for those files to the
corresponding driver maintainer, so they can be notified when they
change.
Signed-off-by: Helen Koike
---
For reference:
On 2023-09-19 10:17, Philip Yang wrote:
On GFX v9.4.3 dGPU, applications have random timeout failure when XNACK
on, dmesg log has "amdgpu: IH soft ring buffer overflow 0x900, 0x900",
because dGPU mode has 272 cam entries. After increasing IH soft ring
to 512 entries, no more IH soft ring
[AMD Official Use Only - General]
Hi Tao,
Based on your description, I think you should use BITS_PER_TYPE() instead of
sizeof(), right?
Best Regards,
Kevin
-Original Message-
From: Zhou1, Tao
Sent: Tuesday, September 19, 2023 6:10 PM
To: amd-gfx@lists.freedesktop.org; Zhang, Hawking
AGP aperture is deprecated and no longer functional.
v2: fix typo (Alex)
v3: just skip the agp setup call
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
If new range is splited to multiple pranges with max_svm_range_pages
alignment and added to update_list, svm validate and map should keep
going after error to make sure prange->mapped_to_gpu flag is up to date
for the whole range.
svm validate and map update set prange->mapped_to_gpu after
On GFX v9.4.3 dGPU, applications have random timeout failure when XNACK
on, dmesg log has "amdgpu: IH soft ring buffer overflow 0x900, 0x900",
because dGPU mode has 272 cam entries. After increasing IH soft ring
to 512 entries, no more IH soft ring overflow message and application
passed.
Fixes:
If retry cam enabled, we don't use sw retry fault filter and add fault
into sw filter ring, so we shouldn't remove fault from sw filter.
Signed-off-by: Philip Yang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git
On Mon, 18 Sep 2023, Alex Hung wrote:
> [WHY]
> edid_override and drm_edid_override_connector_update, according to drm
> documentation, should not be referred outside drm_edid.
>
> [HOW]
> Remove and replace them accordingly. This can tested by IGT's
> kms_hdmi_inject test.
>
> Signed-off-by:
Fix kernel-doc warnings discovered in AMD gpu display driver.
Fixes these warnings:
./drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h:110: warning:
Function parameter or member 'overlap_only'
not described in 'mpcc_blnd_cfg'.
./drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h:110: warning:
Function
For the PASID flushing we already handled that at a higher layer, apply
those workarounds to the standard flush as well.
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 19 +++
Instead of each implementation doing this more or less correctly
move taking the reset lock at a higher level.
v2: fix typo
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 9 +
The same PASID can be used by more than one VMID, reset each of them.
Use the common KIQ handling.
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 63 --
1 file changed, 19
Testing for reset is pointless since the reset can start right after the
test. Grab the reset semaphore instead.
The same PASID can be used by more than once VMID, build a mask of VMIDs
to invalidate instead of just restting the first one.
Signed-off-by: Christian König
Reviewed-by: Alex
The same PASID can be used by more than one VMID, reset each of them.
Use the common KIQ handling.
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 66 --
1 file changed, 19
Testing for reset is pointless since the reset can start right after the
test.
The same PASID can be used by more than one VMID, invalidate each of them.
Move the KIQ and all the workaround handling into common GMC code.
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
Reviewed-by:
That function never fails, drop the error return.
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
Reviewed-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 7 ---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 6 +++---
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 7
Move the SDMA workaround necessary for Navi 1x into a higher layer.
v2: use dev_err
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 48 +++
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 5 +-
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
Remove leftovers from copying this from the gmc v10 code.
Signed-off-by: Christian König
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 108 ++---
1 file changed, 41 insertions(+), 67 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
Testing for reset is pointless since the reset can start right after the
test. Grab the reset semaphore instead.
The same PASID can be used by more than once VMID, build a mask of VMIDs
to invalidate instead of just restting the first one.
Signed-off-by: Christian König
Reviewed-by: Alex
The KIQ code path was ignoring the second flush. Also avoid long lines and
re-calculating the register offsets over and over again.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 29 +--
1 file changed, 18 insertions(+), 11 deletions(-)
diff
Prepare for bad page retirement.
Signed-off-by: Tao Zhou
---
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 4 +++-
drivers/gpu/drm/amd/amdgpu/umc_v12_0.h | 2 ++
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
The amdgpu_ras_eeprom_control.bad_channel_bitmap is u32 type, but the
channel index could be larger than 32. For the ASICs whose channel
number is more than 32, the amdgpu_dpm_send_hbm_bad_channel_flag
interface is not supported, so we simply bypass channel bitmap update under
this condition.
Print channel index for UMC v12.
Signed-off-by: Tao Zhou
---
drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 10 ++
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
index c6742dd863d4..7714c2ef2cdc
Am 08.09.23 um 20:58 schrieb Felix Kuehling:
On 2023-09-05 02:04, Christian König wrote:
The KIQ code path was ignoring the second flush. Also avoid long
lines and
re-calculating the register offsets over and over again.
Signed-off-by: Christian König
---
Am 14.09.23 um 20:21 schrieb Alex Deucher:
We normally place GART based on the location of VRAM and the
available address space around that, but provide an option
to force a particular location for hardware that needs it.
Ah, somehow that patch arrived delayed in my inbox.
Signed-off-by:
53 matches
Mail list logo