The uvd/vce/vcn code accesses the IB at predefined offsets without
checking that the IB is large enough. Check the bounds here. The caller
is responsible for making sure it can handle arbitrary return values.
Also make the idx a uint32_t to prevent overflows causing the condition
to fail.
Signed-off-by: Benjamin Cheng <benjamin.cheng@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
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>
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>
If when we backup the ring contents for reemit before a
ring reset, we skip jobs associated with the bad
context, however, we need to make sure the fences
are reemited as unprocessed submissions may depend on
them.
v2: clean up fence handling, make helpers static
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If we cancel a bad job and reemit the ring contents, and
we get another timeout, cancel everything rather than reemitting.
The wptr markers are only relevant for the original emit. If
we reemit, the wptr markers are no longer correct.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds robust reset handling for user queues (userq) to improve
recovery from queue failures. The key components include:
1. Queue detection and reset logic:
- amdgpu_userq_detect_and_reset_queues() identifies failed queues
- Per-IP detect_and_reset callbacks for targeted recovery
- Falls back to full GPU reset when needed
2. Reset infrastructure:
- Adds userq_reset_work workqueue for async reset handling
- Implements pre/post reset handlers for queue state management
- Integrates with existing GPU reset framework
3. Error handling improvements:
- Enhanced state tracking with HUNG state
- Automatic reset triggering on critical failures
- VRAM loss handling during recovery
4. Integration points:
- Added to device init/reset paths
- Called during queue destroy, suspend, and isolation events
- Handles both individual queue and full GPU resets
The reset functionality works with both gfx/compute and sdma queues,
providing better resilience against queue failures while minimizing
disruption to unaffected queues.
v2: add detection and reset calls when preemption/unmaped fails.
add a per device userq counter for each user queue type.(Alex)
v3: make sure we hold the adev->userq_mutex when we call amdgpu_userq_detect_and_reset_queues. (Alex)
warn if the adev->userq_mutex is not held.
v4: make sure we have all of the uqm->userq_mutex held.
warn if the uqm->userq_mutex is not held.
v5: Use array for user queue type counters.(Alex)
all of the uqm->userq_mutex need to be held when calling detect and reset. (Alex)
v6: fix lock dep warning in amdgpu_userq_fence_dence_driver_process
v7: add the queue types in an array and use a loop in amdgpu_userq_detect_and_reset_queues (Lijo)
v8: remove atomic_set(&userq_mgr->userq_count[i], 0).
it should already be 0 since we kzalloc the structure (Alex)
v9: For consistency with kernel queues, We may want something like:
amdgpu_userq_is_reset_type_supported (Alex)
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>
Decouple the amdgpu fence from the amdgpu_job structure.
This lets us clean up the separate fence ops for the embedded
fence and other fences. This also allows us to allocate the
vm fence up front when we allocate the job.
v2: Additional cleanup suggested by Christian
v3: Additional cleanups suggested by Christian
v4: Additional cleanups suggested by David and
vm fence fix
v5: cast seqno (David)
Cc: David.Wu3@amd.com
Cc: christian.koenig@amd.com
Tested-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When we backup ring contents to reemit after a queue reset,
we don't backup ring contents from the bad context. When
we signal the fences, we should set an error on those
fences as well.
v2: misc cleanups
v3: add locking for fence error, fix comment (Christian)
v4: fix wrap around, locking (Christian)
Fixes: 77cc0da39c ("drm/amdgpu: track ring state associated with a fence")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rename extra_dw to extra_bytes and document what it's for.
The value is already used as if it were bytes in vcn_v4_0.c
and in amdgpu_ring_init. Just adjust the dword count in
jpeg_v1_0.c so that it becomes a byte count.
v2:
Rename extra_dw to extra_bytes as discussed during review.
Fixes: c8c1a1d2ef ("drm/amdgpu: define and add extra dword for jpeg ring")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
None of the pointer operations handled by the ring file requires
volatile, for this reason, this commit removes all occurrences of
volatile associated with rings.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rather than checking in the callbacks, check if the reset
type is supported in the caller.
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need to know the wptr and sequence number associated
with a fence so that we can re-emit the unprocessed state
after a ring reset. Pre-allocate storage space for
the ring buffer contents and add helpers to save off
and re-emit the unprocessed state so that it can be
re-emitted after the queue is reset.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move guilty logic into the ring reset callbacks. This
allows each ring reset callback to better handle fence
errors and force completions in line with the reset
behavior for each IP. It also allows us to remove
the ring guilty callback since that logic now lives
in the reset callback.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
What we actually care about is the amdgpu_fence object
so pass that in explicitly to avoid possible mistakes
in the future.
The job_run_counter handling can be safely removed at this
point as we no longer support job resubmission.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the amdgpu fence container so we can store additional
data in the fence. This also fixes the start_time handling
for MCBP since we were casting the fence to an amdgpu_fence
and it wasn't.
Fixes: 3f4c175d62 ("drm/amdgpu: MCBP based on DRM scheduler (v9)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add some random documentation associated with the ring buffer
manipulations and writeback.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This would be set by IPs which only accept submissions
from the kernel, not userspace, such as when kernel
queues are disabled. Don't expose the rings to userspace
and reject any submissions in the CS IOCTL.
v2: fix error code (Alex)
Reviewed-by: Sunil Khatri<sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This was leftover from MES bring up when we had MES
user queues in the kernel. It's no longer used so
remove it.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Increase the maximum number of rings supported by the AMDGPU driver from 133 to 149.
This change is necessary to enable support for the SDMA page ring.
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Benjamin ChengandAlex Deucher
Srinivasan ShanmugamandAlex Deucher
Rodrigo SiqueiraandAlex Deucher
Jesse.ZhangandAlex Deucher
Tvrtko UrsulinandAlex Deucher
Timur KristófandAlex Deucher
Tao ZhouandAlex Deucher