[Public]
Regards,
Prike
> -Original Message-
> From: Alex Deucher
> Sent: Tuesday, September 9, 2025 5:55 AM
> To: Liang, Prike
> Cc: amd-gfx@lists.freedesktop.org; Deucher, Alexander
> ; Koenig, Christian
> Subject: Re: [PATCH v10 08/14] drm/amdgpu: add userq object va track hel
Hi Peter / Jani,
From the measurements I have done, the difference between RBtree full
iteration and list full iteration
is negligible, even under heavy fragmentation. Based on this, I think it
would be reasonable to include
these macros in rbtree.h as a convenience for cases where a full walk
For SMU v13.0.6 SOCs, move to partition metrics v1.1 schema
Signed-off-by: Lijo Lazar
---
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 6 ++--
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 34 +++
2 files changed, 37 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/d
Fill and publish GPU metrics in v1.9 format for SMUv13.0.6 SOCs
Signed-off-by: Lijo Lazar
---
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 113 +
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 118 ++
2 files changed, 177 insertions(+), 54 deletions(-)
diff -
Fill and publish GPU metrics in v1.9 format for SMUv13.0.12 SOCs
Signed-off-by: Lijo Lazar
---
v2: Keep return type of smu_v13_0_12_get_gpu_metrics as void (Asad)
.../drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 89 ---
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 25 +++---
.
Update SMUv13.0.12 partition metrics to partition metrics v1.1 schema.
Signed-off-by: Lijo Lazar
---
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c
b/drivers/gpu
Add gpu metrics definition which is only a set of gpu metrics
attributes. A field is encoded by its id, type and number of instances.
Signed-off-by: Lijo Lazar
---
v2: Add unit field and change encoding to 64-bit.
v3: Fix unit encoding
.../gpu/drm/amd/include/kgd_pp_interface.h| 117 +++
Use a schema similar to gpu metrics v1.9 for partition metrics also. It
will have field type encoded followed by the field value(s). The
attribute ids used will be shared with gpu metrics. The structure
definition is only to distinguish between gpu metrics and partition
metrics though both gpu metr
[Public]
The current approach in hwmon itself is not uniform.
We have checks like this
if (((adev->family == AMDGPU_FAMILY_SI) ||
((adev->flags & AMD_IS_APU) && (gc_ver < IP_VERSION(9, 3, 0 &&
(attr == &sensor_dev_attr_power1_average.dev_attr.attr))
[Public]
I don’t prefer the original version as that is more rework compared to the
direction of the new solution.
Thanks,
Lijo
From: Wang, Yang(Kevin)
Sent: Tuesday, September 9, 2025 9:05 AM
To: Lazar, Lijo ; Kamal, Asad ;
amd-gfx@lists.freedesktop.org
Cc: Zhang, Hawking ; Ma, Le ; Zhang,
M
[Public]
>> This one is kind of urgent, hence the change is specifically done inside
>> amdgpu_virt.
I prefer to use the origin patch (v1) to fix it if it is an urgent issue.
And the new solution can be submitted later, and all hwmon/nodes will be
managed uniformly, instead of using one method
[AMD Official Use Only - AMD Internal Distribution Only]
Reviewed-by: Asad Kamal
Thanks & Regards
Asad
-Original Message-
From: Wang, Yang(Kevin)
Sent: Tuesday, September 9, 2025 7:51 AM
To: amd-gfx@lists.freedesktop.org
Cc: Kamal, Asad
Subject: [PATCH] drm/amd/pm: make smu_set_temp_f
[AMD Official Use Only - AMD Internal Distribution Only]
Ping on series
-Original Message-
From: Jesse.Zhang
Sent: Friday, September 5, 2025 11:19 AM
To: amd-gfx@lists.freedesktop.org
Cc: Deucher, Alexander ; Koenig, Christian
; Zhang, Jesse(Jie) ; Deucher,
Alexander
Subject: [PATCH
On 9/8/2025 4:50 PM, Bjorn Helgaas wrote:
On Sun, Aug 17, 2025 at 09:00:59PM -0500, Mario Limonciello (AMD) wrote:
For the suspend flow PCIe bridges that have downstream devices are put into
the appropriate low power state (D3hot or D3cold depending upon specific
devices). For the hibernate f
[AMD Official Use Only - AMD Internal Distribution Only]
Yes, it seems there is a new hwmon/sysnode visible attribute check path, but
there is already a check logic in amdgpu_pm.c now, such as checking the return
value - EPONOTUPPORT.
e.g: (amdgpu_dpm_get_fan_speed_pwm(adev, NULL) == -EOPNOTSUP
On 07/18, Melissa Wen wrote:
> On 18-07-2025 00:51, Matthew Schwartz wrote:
> [...]
>
> In short, there is a chance that you are not seeing those glitches
> because there are no changes in the pipe split when transitioning
> between 1-2 overlay planes in your hw, but the split happens on steam
> d
On Mon, Aug 25, 2025 at 4:54 PM Xi Ruoyao wrote:
>
> dml21_map_dc_state_into_dml_display_cfg calls (the call is usually
> inlined by the compiler) populate_dml21_surface_config_from_plane_state
> and populate_dml21_plane_config_from_plane_state which may use FPU. In
> a x86-64 build:
>
> $ ob
In a series that reworked the CSB manipulation, Christian noticed the
use of volatile in some of the function parameters and pointed out that
the use of volatile in amdgpu is probably wrong [1]. He also shared the
following link about this topic:
https://docs.kernel.org/process/volatile-consider
Remove the unnecessary use of volatile in some of the amdgpu.h and
amdgpu_ih.h headers.
Signed-off-by: Rodrigo Siqueira
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h | 6 +++---
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu
Signed-off-by: Rodrigo Siqueira
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 10 +-
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 10 +-
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 10 +-
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c |
I think you can land patches 1, 2, 7, and 12.
Alex
On Mon, Sep 8, 2025 at 4:59 AM Prike Liang wrote:
>
> This will help on validating the userq input args, and
> rejecting for the invalid userq request at the IOCTLs
> first place.
>
> Signed-off-by: Prike Liang
> Reviewed-by: Alex Deucher
> --
On Mon, Sep 8, 2025 at 4:59 AM Prike Liang wrote:
>
> The user queue object destroy requires ensuring its
> VA keeps mapping prior to the queue being destroyed.
> Otherwise, it seems a bug in the user space or VA
> freed wrongly, and the kernel driver should report an
> invalidated state to the us
In subject, s|PCI: PM:|PCI/PM:| to follow previous practice.
On Sun, Aug 17, 2025 at 09:00:55PM -0500, Mario Limonciello (AMD) wrote:
> PCI devices can be programmed as a wakeup source from low power states
> by sysfs. However when using the S4 flow to go into S5 these wakeup
> sources should be
[Public]
> -Original Message-
> From: amd-gfx On Behalf Of Silviu-
> Alexandru Ștefan
> Sent: Monday, September 8, 2025 4:42 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Wentland, Harry ; Li, Sun peng (Leo)
> ; Rodrigo Siqueira ; Silviu-
> Alexandru Ștefan
> Subject: [PATCH] drm/amd/disp
The aux channel is always registered if the connector is DP,
even when the link doesn't contain DP. When using the DisplayPort
connector in DP++ mode, the I2C is done directly over the pins,
not over an aux channel. The practical effect is the creation of
2 I2C devices for DDC, one from the actual
Applied the series.
Thanks!
Alex
On Sun, Aug 24, 2025 at 8:48 PM Rodrigo Siqueira wrote:
>
> The function amdgpu_vcn_sw_fini() returns an integer, but this number is
> always 0. This commit changes the amdgpu_vcn_sw_fini() return to void,
> and eliminates all checks to this return across differ
On Mon, Sep 8, 2025 at 8:54 AM Christian König wrote:
>
> On 05.09.25 20:39, Liu, Shaoyun wrote:
> > [AMD Official Use Only - AMD Internal Distribution Only]
> >
> > I can confirm that during world switch the entire gfx block (including gfx,
> > compute and sdma for gfx10+) been switched together
Update SMUv13.0.12 partition metrics to partition metrics v1.1 schema.
Signed-off-by: Lijo Lazar
---
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c
b/drivers/gpu
On Fri, Sep 05, 2025 at 09:56:25AM -0400, Alex Deucher wrote:
> The size and offset come from userspace, so it's likely a mesa issue.
> I've reported it here:
> https://gitlab.freedesktop.org/mesa/mesa/-/issues/13838
Cool, thanks!
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/note
On 2025-09-03 12:44, Alex Deucher wrote:
> This function can be called from an atomic context so we can't use
> fsleep().
>
> Fixes: 01f60348d8fb ("drm/amd/display: Fix 'failed to blank crtc!'")
> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4549
> Cc: Wen Chen
> Cc: Fangzhi Zuo
>
On Mon, 2025-09-08 at 12:02 -0600, Alex Hung wrote:
>
>
> On 9/8/25 11:45, Timur Kristóf wrote:
> > On Mon, 2025-09-08 at 11:40 -0600, Alex Hung wrote:
> > >
> > >
> > > On 9/8/25 11:36, Alex Deucher wrote:
> > > > @alexh...@amd.com@Wentland, Harry
> > > > Were you planning to pick this up f
On 9/8/25 1:10 PM, Alex Hung wrote:
On 8/24/25 12:23, Kuan-Wei Chiu wrote:
Replace the previous O(N^2) implementation of remove_duplicates() in
with a O(N) version using a fast/slow pointer approach. The new version
keeps only the first occurrence of each element and compacts the array
in p
On 08.09.25 19:05, Alex Hung wrote:
>
>
> On 8/24/25 12:23, Kuan-Wei Chiu wrote:
>> Replace the custom bubble sort used for sorting reserved time
>> candidates in with the kernel's standard sort() helper. The previous
>> code had O(N^2) time complexity, while the generic kernel sort runs in
>> O(
On 08.09.25 15:51, Alex Deucher wrote:
> On Mon, Sep 8, 2025 at 8:54 AM Christian König
> wrote:
>>
>> On 05.09.25 20:39, Liu, Shaoyun wrote:
>>> [AMD Official Use Only - AMD Internal Distribution Only]
>>>
>>> I can confirm that during world switch the entire gfx block (including gfx,
>>> compu
On 8/24/25 12:23, Kuan-Wei Chiu wrote:
Replace the previous O(N^2) implementation of remove_duplicates() in
with a O(N) version using a fast/slow pointer approach. The new version
keeps only the first occurrence of each element and compacts the array
in place, improving efficiency without chan
to get migration pages. dst MIGRATE_PFN_VALID bit and src
MIGRATE_PFN_MIGRATE bit should always be set when migration success.
cpage includes src MIGRATE_PFN_MIGRATE bit set and MIGRATE_PFN_VALID
bit unset pages for both ram and vram when memory is only allocated
without being populated before mig
Add the userq object virtual address get(),mapped() and put()
helpers for tracking the userq obj va address usage.
Signed-off-by: Prike Liang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 172 -
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h | 14 ++
drivers/gpu/drm/amd/amdgp
[AMD Official Use Only - AMD Internal Distribution Only]
ping
-Original Message-
From: amd-gfx On Behalf Of Asad Kamal
Sent: Friday, September 5, 2025 10:13 AM
To: amd-gfx@lists.freedesktop.org; Lazar, Lijo
Cc: Zhang, Hawking ; Ma, Le ; Zhang,
Morris ; Kamal, Asad ; Deucher,
Alexander
On Sat, Sep 06, 2025 at 09:36:31AM -0500, Mario Limonciello (AMD) wrote:
> A variety of issues both in function and in power consumption have been
> raised as a result of devices not being put into a low power state when
> the system is powered off.
>
> There have been some localized changes[1] to
Le 03/08/2025 à 05:57, Jim Cromie a écrit :
Describe the 3 API macros providing dynamic_debug's classmaps
DYNDBG_CLASSMAP_DEFINE - create & export a classmap
DYNDBG_CLASSMAP_USE- refer to exported map
DYNDBG_CLASSMAP_PARAM - bind control param to the classmap
DYNDBG_CLASSMAP_PARAM_REF +
[Public]
Hi all,
This week this patchset was tested on 4 systems, two dGPU and two APU based,
and tested across multiple display and connection types.
APU
* Single Display eDP -> 1080p 60hz, 1920x1200 165hz, 3840x2400 60hz
* Single Display DP (SST DSC) -> 4k144hz, 4k240hz
On Sun, Sep 7, 2025 at 7:06 PM Rodrigo Siqueira wrote:
>
> Loading and unloading the amdgpu module in TTY mode is successful at
> first glance, but attempting to turn off the system after this action
> results in a kernel panic. Sometimes, after trying to run load/unload
> multiple times, the user
Patch looks good to me.
Thanks.
Reviewed-by: Tom Chung
On 9/8/2025 9:09 PM, Ray Wu wrote:
[Why&How]
Remove duplicated code
Signed-off-by: Ray Wu
---
drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c | 3 ---
.../gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c | 3 -
[Why&How]
Remove duplicated code
Signed-off-by: Ray Wu
---
drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c | 3 ---
.../gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c | 3 ---
drivers/gpu/drm/amd/display/dc/resource/dcn36/dcn36_resource.c | 3 ---
3 files changed, 9 de
On 05.09.25 20:39, Liu, Shaoyun wrote:
> [AMD Official Use Only - AMD Internal Distribution Only]
>
> I can confirm that during world switch the entire gfx block (including gfx,
> compute and sdma for gfx10+) been switched together .
Yeah, but that simply doesn't work as expected.
The problem i
On 9/8/25 4:19 AM, Greg Kroah-Hartman wrote:
On Sat, Sep 06, 2025 at 09:36:31AM -0500, Mario Limonciello (AMD) wrote:
A variety of issues both in function and in power consumption have been
raised as a result of devices not being put into a low power state when
the system is powered off.
The
Well good catch, but clear NAK to this workaround.
On 05.09.25 09:19, Jesse.Zhang wrote:
> A general protection fault occurs when signaling DMA fences from
> the user queue fence driver due to an invalid callback function
> pointer. This indicates a use-after-free
> where fence objects are accesse
Currently, gpu_metrics is exported a binary blob with the defintion provided by
means of a structure
and version. Though providing the version works well for backward compatiblity,
it doesn't work the
other way for tools like amd-smi. Installing a driver with a newer definition
of gpu_metrics
so
For SMU v13.0.6 SOCs, move to partition metrics v1.1 schema
Signed-off-by: Lijo Lazar
---
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 6 ++--
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 34 +++
2 files changed, 37 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/d
Use a schema similar to gpu metrics v1.9 for partition metrics also. It
will have field type encoded followed by the field value(s). The
attribute ids used will be shared with gpu metrics. The structure
definition is only to distinguish between gpu metrics and partition
metrics though both gpu metr
Fill and publish GPU metrics in v1.9 format for SMUv13.0.6 SOCs
Signed-off-by: Lijo Lazar
---
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 113 +
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 118 ++
2 files changed, 177 insertions(+), 54 deletions(-)
diff -
Add helper macros to define metrics struct definitions. It will define
structs with field type followed by actual field. A helper macro is also
added to initialize the field encoding for all fields and to initialize
the field members to 0xFFs.
Signed-off-by: Lijo Lazar
---
drivers/gpu/drm/amd/pm
From: Alex Deucher
Query the status of the user queue, currently whether
the queue is hung and whether or not VRAM is lost.
v2: Misc cleanups
Reviewed-by: Sunil Khatri
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 35 ++-
drivers/gpu/drm/amd/
[Public]
Regards,
Prike
> -Original Message-
> From: Alex Deucher
> Sent: Friday, September 5, 2025 11:00 PM
> To: Liang, Prike
> Cc: Koenig, Christian ; amd-
> g...@lists.freedesktop.org; Deucher, Alexander
> Subject: Re: [PATCH v9 08/14] drm/amdgpu: add userq object va track he
It requires validating the userq VA whether is mapped before
trying to resume the queue.
Signed-off-by: Prike Liang
Reviewed-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 8
1 file changed, 8 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
b
Keeping waiting the userq fence infinitely untill
hang detection, and then suspend the hang queue and
set the fence error.
Signed-off-by: Prike Liang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 11 ---
1 file changed, 8 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/
This is no invocation for amdgpu_userq_active().
Signed-off-by: Prike Liang
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 16
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h | 2 --
2 files changed, 18 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu
When a user unmaps a userq VA, the driver must ensure
the queue has no in-flight jobs. If there is pending work,
the kernel should wait for the attached eviction (bookkeeping)
fence to signal before deleting the mapping.
Suggested-by: Christian König
Signed-off-by: Prike Liang
---
drivers/gpu/d
Before destroying the userq buffer object, it requires validating
the userq HW unmap status and ensuring the userq is unmapped from
hardware. If the user HW unmap failed, then it needs to reset the
queue for reusing.
Signed-off-by: Prike Liang
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/a
The user queue object destroy requires ensuring its
VA keeps mapping prior to the queue being destroyed.
Otherwise, it seems a bug in the user space or VA
freed wrongly, and the kernel driver should report an
invalidated state to the user IOCTL request.
Signed-off-by: Prike Liang
---
drivers/gpu
Track the userq obj for its life time, and reference and
dereference the buffer counter at its creating and destroying
period.
Signed-off-by: Prike Liang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq
This will help on validating the userq input args, and
rejecting for the invalid userq request at the IOCTLs
first place.
Signed-off-by: Prike Liang
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 81 +++---
drivers/gpu/drm/amd/amdgpu/mes_userqueue.c |
It needs to validate the userq object virtual address to
determine whether it is residented in a valid vm mapping.
Signed-off-by: Prike Liang
Reviewed-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 40 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h | 2 ++
From: Alex Deucher
Add an API to query queue status such as whether the
queue is hung or whether vram is lost.
Reviewed-by: Christian König
Reviewed-by: Sunil Khatri
Signed-off-by: Alex Deucher
---
include/uapi/drm/amdgpu_drm.h | 14 ++
1 file changed, 14 insertions(+)
diff --gi
On 9/5/25 05:37, Mario Limonciello wrote:
Have a try with this diff on 6.17-rc4.
I confirm that the issue goes away after applying this patch.
65 matches
Mail list logo
