Age | Commit message (Collapse) | Author | Files | Lines |
|
The last TODO item here that was not marked as done was
the display portion, which came along with the pull-request.
So, now that Xe is part of drm-next and it includes the
display portion, let's entirely kill this RFC here.
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240110190427.63095-1-rodrigo.vivi@intel.com
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
|
|
Nothing else to be done on this front from Xe perspective.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231201042158.80009-6-rodrigo.vivi@intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
The must-have part of the documentation was already added to the existing
/gpu/drm-vm-bind-async. The other extra discussion around GPUVM helpers
are currently active in the community. None of those discussion should
block Xe since documentation, specially around locking was completed in
a community consensus.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231201042158.80009-5-rodrigo.vivi@intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
As already indicated in this block, the consensus was already
reached out and documented as:
The ASYNC VM_BIND document </gpu/drm-vm-bind-async>
However this was item was not moved to the completed section.
Let's move and clean up the WIP block.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231201042158.80009-4-rodrigo.vivi@intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Current drm-xe-next doesn't have any drm/scheduler patch that is not
already accepted in drm-misc-next. This completed this goal with
the consensus of how the drm/scheduler fits to the fw scheduling and
the relationship between drm_gpu_scheduler and drm_sched_entity.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231201042158.80009-3-rodrigo.vivi@intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
No DRM scheduler changes required, drivers just return NULL in run_job
vfunc.
The rough consensus is that no helper or extra scaffolding is needed
around long-running jobs and no further changes to drm-scheduler.
At least for now. Other drivers that currently do long-running workloads
have no plat to use drm-scheduler. Besides, the current consensus is
that this solution of simply returning NULL to the run_job function should
work without extra code duplication or complication.
On top of that, this item was already a non-blocking one for upstreaming Xe,
so let's move that to the 'Completed' section and revisit the long-running
solution as a community after Xe is integrated in DRM.
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231201042158.80009-2-rodrigo.vivi@intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Add the first version of the VM_BIND locking document which is
intended to be part of the xe driver upstreaming agreement.
The document describes and discuss the locking used during exec-
functions, evicton and for userptr gpu-vmas. Intention is to be using the
same nomenclature as the drm-vm-bind-async.rst.
v2:
- s/gvm/gpu_vm/g (Rodrigo Vivi)
- Clarify the userptr seqlock with a pointer to mm/mmu_notifier.c
(Rodrigo Vivi)
- Adjust commit message accordingly.
- Add SPDX license header.
v3:
- Large update to align with the drm_gpuvm manager locking
- Add "Efficient userptr gpu_vma exec function iteration" section
- Add "Locking at bind- and unbind time" section.
v4:
- Fix tabs vs space errors by untabifying (Rodrigo Vivi)
- Minor style fixes and typos (Rodrigo Vivi)
- Clarify situations where stale GPU mappings are occurring and how
access through these mappings are blocked. (Rodrigo Vivi)
- Insert into the toctree in implementation_guidelines.rst
v5:
- Add a section about recoverable page-faults.
- Use local references to other documentation where possible
(Bagas Sanjaya)
- General documentation fixes and typos (Danilo Krummrich and
Boris Brezillon)
- Improve the documentation around locks that need to be grabbed from the
dm-fence critical section (Boris Brezillon)
- Add more references to the DRM GPUVM helpers (Danilo Krummrich and
Boriz Brezillon)
- Update the rfc/xe.rst document.
v6:
- Rework wording to improve readability (Boris Brezillon, Rodrigo Vivi,
Bagas Sanjaya)
- Various minor fixes across the document (Boris Brezillon)
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Danilo Krummrich <dakr@redhat.com>
Acked-by: John Hubbard <jhubbard@nvidia.com> # Documentation/core-api/pin_user_pages.rst changes
Link: https://patchwork.freedesktop.org/patch/msgid/20231129090637.2629-1-thomas.hellstrom@linux.intel.com
|
|
Add a motivation for and description of asynchronous VM_BIND operation
v2:
- Fix typos (Nirmoy Das)
- Improve the description of a memory fence (Oak Zeng)
- Add a reference to the document in the Xe RFC.
- Add pointers to sample uAPI suggestions
v3:
- Address review comments (Danilo Krummrich)
- Formatting fixes
v4:
- Address typos (Francois Dugast)
- Explain why in-fences are not allowed for VM_BIND operations for long-
running workloads (Matthew Brost)
v5:
- More typo- and style fixing
- Further clarify the implications of disallowing in-fences for VM_BIND
operations for long-running workloads (Matthew Brost)
v6:
- Point out that a gpu_vm is a virtual GPU Address space.
(Danilo Krummrich)
- For an explanation of dma-fences point to the dma-fence documentation.
(Paulo Zanoni)
- Clarify that VM_BIND errors are reported synchronously. (Paulo Zanoni)
- Use an rst doc reference when pointing to the async vm_bind document
from the xe merge plan.
- Add the VM_BIND documentation to the drm documentation table-of-content,
using an intermediate "Misc DRM driver uAPI- and feature implementation
guidelines"
v7:
- Update the error handling documentation to remove the VM error state.
v8:
- Clarify error handling and difference in operation support between
async VM_BIND and sync VM_BIND. (Paulo Zanoni)
- Update the sample uAPI with a self-contained example. (Paulo Zanoni)
Cc: Paulo R Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Acked-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Danilo Krummrich <dakr@redhat.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231012132552.20196-1-thomas.hellstrom@linux.intel.com
|
|
Nouveau has landed the GPU VA helpers, support and documentation
already and Xe is already using the upstream GPU VA.
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/commit/ea4ae69e66b2940107e74f240ecb9dae87bf1ff1
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/commits/drm-xe-next?ref_type=heads
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230829163005.54067-4-rodrigo.vivi@intel.com
|
|
The consensus is for individual drivers VM_BIND uapis with
the GPUVA helpers that are already implemented and merged
upstream.
The merged GPUVA documentation also establish some overall
rules for the locking to be followed by the drivers.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230829163005.54067-3-rodrigo.vivi@intel.com
|
|
Xe is already using devcoredump infrastructure as the primary
error state and all the changes needed for user space error
replay and other useful logs are getting added into xe_devcoredump.
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/blob/drm-xe-next/drivers/gpu/drm/xe/xe_devcoredump.c
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20230829163005.54067-2-rodrigo.vivi@intel.com
|
|
Also the uapi should be reviewed and scrutinized before xe
is accepted upstream and we shouldn't cause regression.
Link: https://lore.kernel.org/all/20230630100059.122881-1-thomas.hellstrom@linux.intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230829163005.54067-1-rodrigo.vivi@intel.com
|
|
Let’s establish a merge plan for Xe, by writing down clear pre-merge goals, in
order to avoid unnecessary delays.
This initial document starts with a TODO list containing items with clear and
measurable key results. Xe’s initial pull request should only be sent to
dri-devel after all the items are clearly resolved.
Since many of them involve some level of a community consensus, in many cases,
the consensus will be reached in follow-up patches to this document with more
details of the API or helpers that will be developed or modified.
Besides of the items that are highlighted in this document, it is important
to highlight that Oded, has been volunteered to give the overall ack on Xe
driver as the way to confirm that it looks good for upstream.
v2: Incorporated Daniel's feedback:
- Do not make long-running compute a blocker.
- Add a mention to drm-exec that that ties to vm_bind and long-running
compute jobs. Then I also added GPUVA since I recently noticed that this
ties also to the work Matt is doing on that front.
- Added the devcoredump section.
- Add the mention to Oded being volunteered for the overall ack.
v3: Reword a bit the Async VM_BIND to incorporate Daniel's feedback on ensuring
the async vmbind consensus explicitly include Mesa, besides other kernel
drivers.
Cc: Dave Airlie <airlied@redhat.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Francois Dugast <francois.dugast@intel.com>
Signed-off-by: Luis Strano <luis.strano@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230419191913.158807-1-rodrigo.vivi@intel.com
|