If the kernel hasn't been compiled with PCIe hotplug support this
can lead to problems with dGPUs that use BOCO because they effectively
drop off the bus.
To prevent issues, disable BOCO support when compiled without PCIe hotplug.
Reported-by: Gabriel Marcano <gabemarcano@yahoo.com> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1707#note_2696862 Acked-by: Alex Deucher <alexander.deucher@amd.com> Link: https://lore.kernel.org/r/20241211155601.3585256-1-superm1@kernel.org Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Alex Deucher [Tue, 12 Nov 2024 21:19:18 +0000 (16:19 -0500)]
drm/amd/display/dc: add helper for panic updates
Add a DC helper for panic updates.
Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: Lu Yao <yaolu@kylinos.cn> Cc: Jocelyn Falempe <jfalempe@redhat.com> Cc: Harry Wentland <harry.wentland@amd.com>
Alex Deucher [Thu, 31 Oct 2024 17:20:19 +0000 (13:20 -0400)]
drm/amd/display: add clear_tiling mi callbacks
This adds clear_tiling callbacks to the mi structure that
will be used for drm panic support to clear the tiling on
a display. Mem input (mi) is used on DCE based display
IPs.
Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: Lu Yao <yaolu@kylinos.cn> Cc: Jocelyn Falempe <jfalempe@redhat.com> Cc: Harry Wentland <harry.wentland@amd.com>
Alex Deucher [Thu, 31 Oct 2024 17:17:06 +0000 (13:17 -0400)]
drm/amd/display: add clear_tiling hubp callbacks
This adds clear_tiling callbacks to the hubp structure that
will be used for drm panic support to clear the tiling on
a display. hubp3 support from Jocelyn's original patch
and the rest from me.
Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: Lu Yao <yaolu@kylinos.cn> Cc: Jocelyn Falempe <jfalempe@redhat.com> Cc: Harry Wentland <harry.wentland@amd.com>
Alex Deucher [Wed, 30 Oct 2024 21:27:23 +0000 (17:27 -0400)]
drm/amdgpu: add generic display panic helper code
Pull this out of Jocelyn's patch and make it generic.
Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: Lu Yao <yaolu@kylinos.cn> Cc: Jocelyn Falempe <jfalempe@redhat.com> Cc: Harry Wentland <harry.wentland@amd.com>
Lijo Lazar [Mon, 9 Dec 2024 03:44:53 +0000 (09:14 +0530)]
drm/amdgpu: Avoid VF for RAS recovery source check
VF device sets the RAS flag when mailbox data can't be read properly.
There is no conclusive way to tell if the real source is RAS error.
Therefore VF schedules a KFD based reset which doesn't set RAS source.
SKip checking RAS source for any VF scheduled recovery.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reported-by: Vojislav Tomasevic <vojislav.tomasevic@amd.com> Reviewed-by: Yiqing Yao <yiqing.yao@amd.com> Tested-by: Yiqing Yao <yiqing.yao@amd.com> Fixes: e1ee2111ca48 ("drm/amdgpu: Prefer RAS recovery for scheduler hang") Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu/sdma7: Implement resume function for each instance
Extracts the resume sequence for per sdma instance from sdma_v7_0_gfx_resume.
This function can be used in start or restart scenarios of specific instances.
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Andrew Martin [Fri, 6 Dec 2024 19:52:34 +0000 (14:52 -0500)]
drm/amdkfd: Uninitialized pointer read
This a pointer that is being passed into other functions, so it is best to
initialize it to NULL prior.
Signed-off-by: Andrew Martin <Andrew.Martin@amd.com> Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Leo Li [Thu, 5 Dec 2024 20:43:42 +0000 (15:43 -0500)]
drm/amdgpu: rename register headers to dcn_2_0_1
They were named with the incorrect dcn version.
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Signed-off-by: Sun peng Li <sunpeng.li@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Aric Cyr [Mon, 2 Dec 2024 04:06:31 +0000 (23:06 -0500)]
drm/amd/display: 3.2.313
* Fix some regressions related to IPS2 and PSR Panel Replay
* Bug fixes in DML
* DMCUB debug improvements
* Other refactors and improvements across multiple components
Reviewed-by: Leo Li <sunpeng.li@amd.com> Signed-off-by: Aric Cyr <aric.cyr@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Taimur Hassan [Sun, 1 Dec 2024 05:46:06 +0000 (00:46 -0500)]
drm/amd/display: [FW Promotion] Release 0.0.246.0
Reviewed-by: Leo Li <sunpeng.li@amd.com> Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Charlene Liu [Fri, 29 Nov 2024 22:18:50 +0000 (17:18 -0500)]
drm/amd/display: update dcn351 used clock offset
[why]
hw register offset delta
Reviewed-by: Martin Leung <martin.leung@amd.com> Signed-off-by: Charlene Liu <Charlene.Liu@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In some test environments causes reporting failures for S0i3/S4.
It shouldn't actually block entry provided there's no race with the
last state being updated, but currently suspecting there's an IPS2
check that's no longer being met.
Fixes: 8488646966fe ("drm/amd/display: Don't allow IPS2 in D0 for RCG Dynamic") Reviewed-by: Ovidiu Bunea <ovidiu.bunea@amd.com> Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Chris Park [Wed, 27 Nov 2024 18:50:04 +0000 (13:50 -0500)]
drm/amd/display: Update color space, bias and scale programming sequence
[Why]
DMColor inaccurately updates color space, bias and scale
destructively in dc_plane_state. This can be resolved by
accurately populating the infos on dc_plane_info where then
translation to plane state can happen as a whole surface update sequence.
[How]
Remove dc_plane_state update in DMColor and update color space,
bias and scale on dc_plane_info.
Reviewed-by: Dillon Varone <dillon.varone@amd.com> Signed-off-by: Chris Park <chris.park@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Ausef Yousof [Tue, 26 Nov 2024 20:26:01 +0000 (15:26 -0500)]
drm/amd/display: Overwriting dualDPP UBF values before usage
[WHY]
Right now in dml2 mode validation we are calculating UBF parameters for
prefetch calculation for single and dual DPP scenarios. Data structure
to store such values are just 1D arrays, the single DPP values are
overwritten by the dualDPP values, and we end up using dualDPP for
prefetch calculations twice (once in place of singleDPP support check
and again for dual).
This naturally leads to many problems, one of which validating a mode in
"singleDPP" (when we used dual DPP parameters) and sending the singleDPP
parameters to mode programming, if we cannot support then we observe the
corruption as described in the ticket.
[HOW]
UBF values need to have 2d arrays to store values specific to single and
dual DPP states to avoid single DPP values being overwritten. Other
parameters are recorded on a per state basis such as prefetch UBF values
but they are in the same loop used for calculation and at that point its
fine to overwrite them, its not the case for plain UBF values.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Ausef Yousof <Ausef.Yousof@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Leo Li [Tue, 26 Nov 2024 17:34:39 +0000 (12:34 -0500)]
drm/amd/display: Make DMCUB tracebuffer debugfs chronological
[Why]
Previously, the debugfs did a simple dump of the tracebuffer region.
Because the tracebuffer is a ring, it meant that the entries printed may
not be in chronological order if the ring rolled over. This makes
parsing the tracelog cumbersome.
[How]
Since dmcub provides the current entry count, use that to determine
the latest tracelog entry and output the log chronologically.
Also, the fb region size is not accurate of the actual tracebuffer size;
it has been padded to alignment requirements. Use the tracebuffer size
reported by the fw meta_info, if available. If not, a fallback to the
hardcoded default is needed. To make this value available to other .c
files, its define was moved to dmub_srv.h.
Also, print a indicator at the start of the log if rollover occurred.
Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Leo Li <sunpeng.li@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why/How]
Mismatch between mode support and mode programming occurs.
Mode support would calculate higher row vblank than mode programming.
As a result, mode programming fails and hardware isn't properly programmed.
Reviewed-by: Dillon Varone <dillon.varone@amd.com> Signed-off-by: Austin Zheng <Austin.Zheng@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Wayne Lin [Thu, 31 Oct 2024 09:32:32 +0000 (17:32 +0800)]
drm/amd/display: Adjust secure_display_context data structure
[Why]
Variables relates to secure display are spreading out within struct
amdgpu_display_manager.
[How]
Encapsulate relevant variables into struct secure_display_context and
adjust relevant affected codes.
Reviewed-by: HaoPing Liu <haoping.liu@amd.com> Signed-off-by: Wayne Lin <Wayne.Lin@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Wayne Lin [Thu, 25 Jul 2024 07:29:44 +0000 (15:29 +0800)]
drm/amd/display: Fix phy id mapping issue for secure display
[Why]
Under mst scenario, mst streams are from the same link_enc_hw_inst.
As the result, can't utilize that as the phy index for distinguising
different stream sinks.
[How]
Sort the connectors by:
link_enc_hw_instance->mst tree depth->mst RAD
After sorting the phy index assignment, store connector's relevant info
into dm mapping array. Once need the index, just look up the static
array.
Reviewed-by: HaoPing Liu <haoping.liu@amd.com> Signed-off-by: Wayne Lin <Wayne.Lin@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Wayne Lin [Tue, 6 Aug 2024 09:21:24 +0000 (17:21 +0800)]
drm/amd/display: Adjust dc_stream_forward_crc_window to accept assignment of phy_id
[Why]
For mst streams under same topology, stream->link->link_enc_hw_inst are the same and
hence can't distinguish the crc window setting.
[How]
Firstly adjust dc_stream_forward_crc_window to accept assignment of phy_id. Follow up
another patch to determine the phy_id at dm layer.
Reviewed-by: HaoPing Liu <haoping.liu@amd.com> Signed-off-by: Wayne Lin <Wayne.Lin@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Joshua Aberback [Sat, 23 Nov 2024 00:46:10 +0000 (19:46 -0500)]
drm/amd/display: Refactor dcn31_panel_construct to avoid assert
[Why]
We want to avoid unnecessary asserts, one of which is hit in
dcn31_panel_construct when booting on a DCN32 asic that has an eDP
connector on a different DIG than A or B. The DIG-based mapping only
applies when edp0_on_dp1 is supported, therefore the check for valid
eng_id can be moved within the appropriate section of the if statement.
Reviewed-by: Alvin Lee <alvin.lee2@amd.com> Signed-off-by: Joshua Aberback <joshua.aberback@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lijo Lazar [Fri, 6 Dec 2024 12:54:06 +0000 (18:24 +0530)]
drm/amd/pm: Revert state if force level fails
Before forcing level, CG/PG is disabled or enabled depending on the new
level. However if the force level operation fails, CG/PG state remains
modified. Revert the state change on failure. Also, move invalid
operation checks to the beginning before any logic that could change SOC
state.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Sathishkumar S [Mon, 11 Nov 2024 22:29:02 +0000 (17:29 -0500)]
drm/amdgpu: update macro for maximum jpeg rings
Update the macro to accomdate more rings.
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com> Acked-by: Leo Liu <leo.liu@amd.com> Reviewed-by: Sonny Jiang <sonjiang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Sathishkumar S [Fri, 18 Oct 2024 12:35:16 +0000 (18:05 +0530)]
drm/amdgpu: add irq source ids for VCN5_0/JPEG5_0
Add interrupt source id macros for VCN5 and JPEG5
V2: Update copyright year (Sonny)
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com> Acked-by: Leo Liu <leo.liu@amd.com> Reviewed-by: Sonny Jiang <sonjiang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd/amdgpu: Add Annotations to Process Isolation functions
This update adds explanations to key functions that manage how the
Kernel Fusion Driver (KFD) and Kernel Graphics Driver (KGD) share the
GPU.
amdgpu_gfx_enforce_isolation_wait_for_kfd: Controls the waiting period
for KFD to ensure it takes turns with KGD in using the GPU. It uses a
mutex to safely manage shared data, like timing and state, and tracks
when KFD starts and stops waiting.
amdgpu_gfx_enforce_isolation_ring_begin_use: Ensures KFD has enough time
to run before new tasks are submitted to the GPU ring. It uses a mutex
to synchronize access and may adjust the KFD scheduler.
amdgpu_gfx_enforce_isolation_ring_end_use: Handles cleanup and state
updates when finishing the use of a GPU ring. It may also adjust the KFD
scheduler, using a mutex to manage shared data access.
Cc: Christian König <christian.koenig@amd.com> Cc: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com> Suggested-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd/amdgpu: Add Descriptions to Process Isolation and Cleaner Shader Sysfs Functions
This update adds explanations to key functions related to process
isolation and cleaner shader execution sysfs interfaces.
- `amdgpu_gfx_set_run_cleaner_shader`: Describes how to manually run a
cleaner shader, which clears the Local Data Store (LDS) and General
Purpose Registers (GPRs) to ensure data isolation between GPU workloads.
- `amdgpu_gfx_get_enforce_isolation`: Describes how to query the current
settings of the 'enforce_isolation' feature for each GPU partition.
- `amdgpu_gfx_set_enforce_isolation`: Describes how to enable or disable
process isolation for GPU partitions through the sysfs interface.
Cc: Christian König <christian.koenig@amd.com> Cc: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com> Suggested-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd: Show an info message about optional firmware missing
With the warning from the core about missing firmware gone,
users still may be notified of missing optional firmware by
a more friendly message to clarify it's optional.
Suggested-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Yang Wang [Fri, 29 Nov 2024 07:41:05 +0000 (15:41 +0800)]
drm/amdgpu: add ACA support for jpeg v4.0.3
Add ACA support for jpeg v4.0.3.
Signed-off-by: Yang Wang <kevinyang.wang@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Tao Zhou <tao.zhou1@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Yang Wang [Thu, 28 Nov 2024 07:19:20 +0000 (15:19 +0800)]
drm/amdgpu: add ACA support for vcn v4.0.3
v1:
Add ACA support for vcn v4.0.3.
v2:
- split VCN ACA(v1) to 2 parts: vcn and jpeg.
- move mmSMNAID_AID0_MCA_SMU to amdgpu_aca.h file.
v3:
- split JPEG ACA to another patch.
Signed-off-by: Yang Wang <kevinyang.wang@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Tao Zhou <tao.zhou1@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Yang Wang [Fri, 29 Nov 2024 07:12:37 +0000 (15:12 +0800)]
drm/amdgpu: move common ACA ipid defines into amdgpu_aca.h
move common ACA ipid defines into amdgpu_aca.h file.
Signed-off-by: Yang Wang <kevinyang.wang@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Tao Zhou <tao.zhou1@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Tao Zhou [Fri, 29 Nov 2024 08:52:41 +0000 (16:52 +0800)]
drm/amdgpu: correct the calculation of RAS bad page
After the introduction of NPS RAS, one bad page record on eeprom may be
related to 1 or 16 bad pages, so the bad page record and bad page are
two different concepts, define a new variable to store bad page number.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd: Add the capability to mark certain firmware as "required"
Some of the firmware that is loaded by amdgpu is not actually required.
For example the ISP firmware on some SoCs is optional, and if it's not
present the ISP IP block just won't be initialized.
The firmware loader core however will show a warning when this happens
like this:
```
Direct firmware load for amdgpu/isp_4_1_0.bin failed with error -2
```
To avoid confusion for non-required firmware, adjust the amd-ucode helper
to take an extra argument indicating if the firmware is required or
optional.
On optional firmware use firmware_request_nowarn() instead of
request_firmware() to avoid the warnings.
Le Ma [Wed, 7 Aug 2024 09:33:00 +0000 (17:33 +0800)]
drm/amdkfd: update the cwsr area size for gfx950
Update cwsr area size for gfx950 to fit the new user queue buffer validation.
The size of LDS calculation is referred from gfx950 thunk implementation.
Signed-off-by: Le Ma <le.ma@amd.com> Acked-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lancelot SIX [Fri, 12 Jul 2024 22:22:29 +0000 (23:22 +0100)]
drm/amdkfd: Handle save/restore of lds allocated in 1280B blocks
The gfx-9 trap handler is reading LDS allocation size in 256 bytes
granularity (from SQ_WAVE_LDS_ALLOC), but it using the assumption that
this value is always even (i.e. the LDS allocation is really done in
multiple of 512 bytes). This was true so far, but gfx-950 allocates LDS
in chunks of 1280 bytes, making this assumption invalid. This can cause
the trap handler to try to save / restore past the end of LDS, and past
the LDS allocated slot in the save are, overriding data from the
following wave.
This patch updates the trap handler to support LDS allocated in 1280
bytes blocks:
- During restore, copy from main memory directly to LDS in batch of 1280
bytes.
- During save, continue to use 512 bytes blocks (we only have 2 VGPRs we
can use to hold data), making sure to mask the upper half of the wave
when handling when the LDS size is not a multiple of 512 bytes.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com> Co-authored-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Jay Cornwall <jay.cornwall@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lancelot SIX [Mon, 17 Jun 2024 11:10:56 +0000 (12:10 +0100)]
drm/amdkfd: Adjust CWSR trap handler for gfx950
In gfx950, the SQ_WAVE_LDS_ALLOC.LDS_SIZE field is extended to bits 12
to 22. The LDS_SIZE granularity remains unchanged (units of 64 dwords,
or 256 bytes). This patch adjusts the CWSR trap handler to read the
full extent of LDS_SIZE.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com> Reviewed-by: Jay Cornwall <jay.cornwall@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lancelot SIX [Fri, 12 Apr 2024 07:41:53 +0000 (08:41 +0100)]
drm/amdkfd: update buffer_{store,load}_* modifiers for gfx940
Instruction modifiers of the untyped vector memory buffer instructions
(MUBUF encoded) changed in gfx940. The slc, scc and glc modifiers have
been replaced with sc0, sc1 and nt.
The current CWSR trap handler is written using pre-gfx940 modifier
names, making the source incompatible with a strict gfx940 assembler.
This patch updates the cwsr_trap_handler_gfx9.s source file to be
compatible with all gfx9 variants of the ISA. The binary assembled code
is unchanged (so the behaviour is unchanged as well), only the source
representation is updated.
Signed-off-by: Lancelot SIX <lancelot.six@amd.com> Reviewed-by: Jay Cornwall <jay.cornwall@amd.com> Acked-by: Felix Kuehling <felix.kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Alex Sierra [Wed, 21 Feb 2024 21:02:15 +0000 (15:02 -0600)]
drm/amdkfd: add gc 9.5.0 support on kfd
Initial support for GC 9.5.0.
v2: squash in pqm_clean_queue_resource() fix from Lijo
Signed-off-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Felix Kuehling <felix.kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Alex Sierra [Wed, 24 Jul 2024 00:29:02 +0000 (19:29 -0500)]
drm/amd: update mtype flags for gfx 9.5.0
Update mtype flags to meet gfx 9.5.0 requirements for remote GPU
memory and system memory.
Signed-off-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Felix Kuehling <felix.kuehling@amd.com> Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Alex Sierra [Thu, 29 Feb 2024 20:55:58 +0000 (14:55 -0600)]
drm/amdgpu: Set proper MTYPE for GC 9.5.0
GC 9.5.0 local memory MTYPE default should be set as RW.
Signed-off-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Felix Kuehling <felix.kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Alex Sierra [Wed, 24 Jul 2024 00:19:42 +0000 (19:19 -0500)]
drm/amd: define gc ip version local variable
For better readability. Also leftover orphaned code.
Signed-off-by: Alex Sierra <alex.sierra@amd.com> Reviewed-by: Felix Kuehling <felix.kuehling@amd.com> Reviewed-by: Harish Kasiviswanathan <harish.kasiviswanathan@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Prike Liang [Mon, 2 Dec 2024 06:13:02 +0000 (14:13 +0800)]
drm/amdgpu: Avoid to release the FW twice in the validated error
There will to release the FW twice when the FW validated error.
Even if the release_firmware() will further validate the FW whether
is empty, but that will be redundant and inefficient.
Randy Dunlap [Thu, 28 Nov 2024 03:20:53 +0000 (19:20 -0800)]
drm/amdgpu: device: fix spellos and punctuation
Make spelling and punctuation changes to ease reading of the comments.
Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Christian König <christian.koenig@amd.com> Cc: Xinhui Pan <Xinhui.Pan@amd.com> Cc: amd-gfx@lists.freedesktop.org Cc: David Airlie <airlied@gmail.com> Cc: Simona Vetter <simona@ffwll.ch> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Ivan Stepchenko [Mon, 2 Dec 2024 08:00:43 +0000 (11:00 +0300)]
drm/amdgpu: Fix potential NULL pointer dereference in atomctrl_get_smc_sclk_range_table
The function atomctrl_get_smc_sclk_range_table() does not check the return
value of smu_atom_get_data_table(). If smu_atom_get_data_table() fails to
retrieve SMU_Info table, it returns NULL which is later dereferenced.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
In practice this should never happen as this code only gets called
on polaris chips and the vbios data table will always be present on
those chips.
Fixes: a23eefa2f461 ("drm/amd/powerplay: enable dpm for baffin.") Signed-off-by: Ivan Stepchenko <sid@itb.spb.ru> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Sathishkumar S [Fri, 29 Nov 2024 04:17:47 +0000 (09:47 +0530)]
drm/amdgpu: Add amdgpu_vcn_sched_mask debugfs
Add debugfs entry to enable or disable job submission to
specific vcn instances. The entry is created only when
there is more than an instance and is unified queue type.
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com> Reviewed-by: Jesse Zhang <jesse.zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd: Add Suspend/Hibernate notification callback support
As part of the suspend sequence VRAM needs to be evicted on dGPUs.
In order to make suspend/resume more reliable we moved this into
the pmops prepare() callback so that the suspend sequence would fail
but the system could remain operational under high memory usage suspend.
Another class of issues exist though where due to memory fragementation
there isn't a large enough contiguous space and swap isn't accessible.
Add support for a suspend/hibernate notification callback that could
evict VRAM before tasks are frozen. This should allow paging out to swap
if necessary.
projects / thirdparty / kernel / linux.git / log
