Add build number, version and date to the existing part number print.
Example:
[drm] ATOM BIOS: 113-PN000108-103, build: 00159017, ver: 022.040.003.043.000001, 2025/07/27
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Store the start wptr and ib size in the IB fence. On queue
reset, save the ring contents of all IBs.
For reemit, reemit the entire IB state for non-guilty contexts.
For guilty contexts, replace the IB submission with nops, but reemit
the rest. Split the reemit per fence and when we reemit, update the
wptr with the new values from reemit. This allows us to reemit jobs
repeatedly as the wptrs get properly updated each time.
v2: further simplify the logic
v3: reemit vm state, not just vm fence
v4: just nop the IB and possibly the VM portion of the submission
v5: simplify the vm fence check
v6: split the vm and ib fences
v7: fix commit message
v8: use wptr rather than count_dw to calculate offsets
v9: fix missing documenation update spotted by the kernel test robot
Reviewed-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Now that DC supports external DP bridge encoders,
it has reached feature parity with the legacy non-DC display
driver on CIK APUs: Kaveri, Kabini, Mullins.
Use the DC display driver by default on SI APUs, unless it is
explicitly disabled using the amdgpu.dc=0 module parameter.
DC brings proper support for DP/HDMI audio, DP MST, VRR,
10-bit colors, some HDR features, atomic modesetting, etc.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the pre-existing implementation in the BIOS parser, but call
the ExternalEncoderControl function for external encoders instead
of the built-in DAC load detection function.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Implement link creation for external DP bridge encoders
such as NUTMEG and TRAVIS used with CIK APUs such as
Kaveri for supporting analog and LVDS connections.
Typically found in CIK APU based laptops or on FM2
motherboards that have analog connectors.
When we query connector information from the VBIOS and
discover a connector using such an encoder, let's find the
real DisplayPort encoder and use that. Set the connector
signal type to DP, so the pre-existing DP code paths can
work with it without refactoring every signal type check
in the DC code base.
In the DM, make sure to report correct connector type and
also to initialize DP specifics such as the AUX channel.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Deal with some minor idiosyncracies of TRAVIS and NUTMEG chips.
- Always use DP signal type with these chips so that the
normal DP code paths can work with them without a major
refactor of the code base. Properly set this.
- NUTMEG seems to only work with HBR, not RBR, so set a
preferred link rate for this chip.
See amdgpu_atombios_dp_get_dp_link_config() for reference.
- NUTMEG is recognized as a DP branch device but reports 0 sinks,
which is wrong and confuses DC (it hits an early return).
Fix that by hardcoding the sink count to 1.
- Fixup old DC code selecting a special panel mode necessary
for NUTMEG and TRAVIS.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Some GPUs use external DP bridge encoders NUTMEG and TRAVIS
to implement analog and/or LVDS connections. Typically found in
CIK APU based laptops or on FM2 motherboards that have analog
connectors. These were necessary at the time because Kaveri
didn't have a built-in DAC nor LVDS support.
These devices sadly don't work transparently and need to be
controlled by the driver. Implement the necessary control for
the NUTMEG and TRAVIS encoders in the DCE HWSS.
For reference, see the legacy non-DC amdgpu display code:
amdgpu_atombios_encoder_setup_external_encoder()
amdgpu_atombios_encoder_setup_dig()
amdgpu_atombios_encoder_setup_ext_encoder_ddc()
- Prepare DDC before using it:
Call the EXTERNAL_ENCODER_CONTROL_DDC_SETUP action so that
the encoder knows to set up DDC over the AUX channel.
- When a stream is enabled or disabled:
Call the EXTERNAL_ENCODER_CONTROL_ENABLE/DISABLE actions.
- Before enabling the DP link:
Call the EXTERNAL_ENCODER_CONTROL_SETUP action.
This commit just hooks up the HWSS support.
Detecting the external DP bridge encoders will be done in
a subsequent commit.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The DisplayPort code already has the concept of preferred link
settings, but it only allows setting a preferred lane count and
link width at the same time. It does not consider the possiblity
that some devices may not work on lower link rates but may
support various lane counts.
Allow specifying a preferred link rate which will be used as
the initial link rate when deciding the DP link settings.
This is necessary to support NUTMEG which only works with HBR
but not with RBR.
For reference, see the legacy non-DC amdgpu display code:
amdgpu_atombios_dp_get_dp_link_config()
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Make it possible to add a HWSS function to prepare the DDC
before trying to use it. This is going to be necessary for
external DP bridge encoders.
This commit just adds the function to common DC code.
The actual implementation of this function for DCE is
done in a subsequent commit.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
DDC probe means that we use the DDC (Display Data Channel) to
verify whether a display is connected or not.
This was written with mainly analog connectors in mind, so it
didn't work over the AUX channel. However, in order to support
external DP bridge encoders found in CIK APUs, it needs to work
over the AUX channel too.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The VBIOS has a function called ExternalEncoderControl which
controls the DP bridge encoders that some GPUs use for analog
and LVDS output. Fixup this old functionality.
For reference, see the legacy non-DC amdgpu display code:
amdgpu_atombios_encoder_setup_external_encoder()
- Set same parameters for the ENABLE action as the SETUP action
- Add missing enum values for DDC setup and DAC load detection
- Fix the bits per color field
- Clarify the code that sets the link rate
- Expose the function so that it can be called by rest of DC
A subsequent commit will call this function from DCE HWSS.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Slightly refactor and simplify DAC load detection.
This prepares the code to be used for also executing DAC
load detection on external DP bridge encoders.
DAC load detection belongs better in the hardware sequencer
code because the implementation is HW dependent and not all
chips support the functionality. The code is cleaner when
link detection probably doesn't call the VBIOS directly.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To improve consistency and avoid duplicating the same code.
Also, properly handle all enum values where they weren't
handled correctly before.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The dce60_register_irq_handlers function was basically identical
to dce110_register_irq_handlers. They can use the same function,
reducing duplicated code and easing the maintenance burden for
old DCE versions.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
dce60_clk_mgr was basically identical to dce_clk_mgr other than
a few minor details. They can be all handled in the same file,
reducing duplicated code and easing the maintenance burden for
old DCE versions.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rodrigo Siqueira <siqueira@igalia.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This file was not compiled or included anywhere and not
modified for years. And it has duplicate function
definitions of many functions that are defined in
other files.
That leads to a lot of confusion (for both developers
and for code editors), eg. when searching for a definition
of a function it can end up on one of the functions in
this file instead of the real implementations.
I suspect that it was forgotten that this should be
deleted after the various clock manager functions were
split into hardware generation specific files.
Delete it now.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rodrigo Siqueira <siqueira@igalia.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The only difference between DCE 6 and other DCE versions is
that DCE 6 doesn't support DCCG_AUDIO_DTO2_USE_512FBR_DTO.
Recently a check was added to dce_aud_wall_dto_setup() to
check that. So now DCE 6 can just use dce_aud_wall_dto_setup()
just like other DCE versions.
Clean up DCE 6 specific audio functions which were otherwise
identical to the rest.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rodrigo Siqueira <siqueira@igalia.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
amdgpu_vm_flush() is used during job submission and is not expected to
fail. Convert it to return void and simplify the caller.
Initialize the COND_EXEC patch location to 0 so it is safe to call
amdgpu_ring_patch_cond_exec() when init_cond_exec is not supported.
Suggested-by: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
dma_fence_wait(old, false) is not interruptible and cannot return an
error. Drop the unreachable error handling in amdgpu_fence_emit().
Since the function can no longer fail, convert amdgpu_fence_emit() to
return void and remove return value handling from all callers.
v2:
- Add comment explaining why dma_fence_wait(..., false)
return value is ignored (Alex)
Suggested-by: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reorders the IB schedule sequence to cleanly
separate the vm operation from the IB submission.
This makes the two independent so we can cleanly
associate each one with its respective fence.
v2: fixes for VCN
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The mes and mes_kiq parameters we originally added for
mes bring up. However, mes is required for operation
on gfx11 and newer so these parameters aren't actually
used by the driver anymore. Remove them.
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We only want to stop the work queues, not mess with the
fences, etc.
v2: add the job back to the pending list.
v3: return the proper job status so scheduler adds the
job back to the pending list
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Re-order the struct members a bit to avoid some holes:
/* size: 408, cachelines: 7, members: 15 */
/* sum members: 393, holes: 4, sum holes: 15 */
/* last cacheline: 24 bytes */
/* size: 400, cachelines: 7, members: 15 */
/* sum members: 393, holes: 1, sum holes: 7 */
/* last cacheline: 16 bytes */
While doing so we notice a duplicate but will address than in the
following patch.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Extend the GFX12 compute MQD initialization to support
Compute Unit (CU) masking for fine-grained resource allocation.
This allows compute queues to be limited to specific CUs for
performance isolation and debugging purposes.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Extend the GFX11 compute MQD initialization to support
Compute Unit (CU) masking for fine-grained resource allocation.
This allows compute queues to be limited to specific CUs for
performance isolation and debugging purposes.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add new fields to the amdgpu_mqd_prop structure to track CU (Compute Unit)
mask information, including the mask itself, count, flags, and a flag to
indicate if user-specified CU masking is active.
v2: Create a generic function amdgpu_gfx_mqd_symmetrically_map_cu_mask()
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
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>
Extend the AMDGPU user queue function interface to support MQD
updates by adding an mqd_update callback.
v2: add the input paramter struct drm_amdgpu_userq_in in mqd_update
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
All sdma versions used the same logic, so add a helper and move the
common code to a single place.
---
v2: pass amdgpu_vm_pte_funcs as well
v3: drop all the *_set_vm_pte_funcs one liners
v5: rebased
---
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm_sched_job_arm and drm_sched_entity_push_job must be called
under the same lock to guarantee the order of execution.
This commit adds a check in amdgpu_ttm_job_submit and fix the
places where the lock was missing.
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of reserving a number of GTT pages for VCE 1.0 this
commit now uses amdgpu_gtt_mgr_alloc_entries to allocate
the pages when initializing vce 1.0.
While at it remove the "does the VCPU BO already have a
32-bit address" check as suggested by Timur.
This decouples vce init from gtt init.
---
v7: renamed variables (Christian)
---
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This allows to have init/fini functions to hold all the init and
teardown code for amdgpu_ttm_buffer_entity.
For now only drm_sched_entity init/destroy function calls are moved
here, but as entities gain new members it will make code simpler.
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If multiple entities share the same window we must make sure
that jobs using them are executed sequentially.
This commit gives separate windows to each entity, so jobs
from multiple entities could execute in parallel if needed.
(for now they all use the first sdma engine, so it makes no
difference yet).
The entity stores the gart window offsets to centralize the
"window id" to "window offset" in a single place.
default_entity doesn't get any windows reserved since there is
no use for them.
---
v3:
- renamed gart_window_lock -> lock (Christian)
- added amdgpu_ttm_buffer_entity_init (Christian)
- fixed gart_addr in svm_migrate_gart_map (Felix)
- renamed gart_window_idX -> gart_window_offs[]
- added amdgpu_compute_gart_address
v4:
- u32 -> u64
- added kerneldoc
v5:
- removed gtt_window_lock
- simplified gart window creation and use: entities using a
single window now uses window #0 instead of #1
- fix dst_addr calculation in kfd_migrate.c
---
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
