linux.git/drivers/gpu/drm/amd/amdkfd/kfd_process.c, branch v6.13

drm/amdkfd: wq_release signals dma_fence only when available

2025-01-06T20:17:23+00:00

kfd_process_wq_release() signals eviction fence by
dma_fence_signal() which wanrs if dma_fence
is NULL.

kfd_process->ef is initialized by kfd_process_device_init_vm()
through ioctl. That means the fence is NULL for a new
created kfd_process, and close a kfd_process right
after open it will trigger the warning.

This commit conditionally signals the eviction fence
in kfd_process_wq_release() only when it is available.

[  503.660882] WARNING: CPU: 0 PID: 9 at drivers/dma-buf/dma-fence.c:467 dma_fence_signal+0x74/0xa0
[  503.782940] Workqueue: kfd_process_wq kfd_process_wq_release [amdgpu]
[  503.789640] RIP: 0010:dma_fence_signal+0x74/0xa0
[  503.877620] Call Trace:
[  503.880066]  
[  503.882168]  ? __warn+0xcd/0x260
[  503.885407]  ? dma_fence_signal+0x74/0xa0
[  503.889416]  ? report_bug+0x288/0x2d0
[  503.893089]  ? handle_bug+0x53/0xa0
[  503.896587]  ? exc_invalid_op+0x14/0x50
[  503.900424]  ? asm_exc_invalid_op+0x16/0x20
[  503.904616]  ? dma_fence_signal+0x74/0xa0
[  503.908626]  kfd_process_wq_release+0x6b/0x370 [amdgpu]
[  503.914081]  process_one_work+0x654/0x10a0
[  503.918186]  worker_thread+0x6c3/0xe70
[  503.921943]  ? srso_alias_return_thunk+0x5/0xfbef5
[  503.926735]  ? srso_alias_return_thunk+0x5/0xfbef5
[  503.931527]  ? __kthread_parkme+0x82/0x140
[  503.935631]  ? __pfx_worker_thread+0x10/0x10
[  503.939904]  kthread+0x2a8/0x380
[  503.943132]  ? __pfx_kthread+0x10/0x10
[  503.946882]  ret_from_fork+0x2d/0x70
[  503.950458]  ? __pfx_kthread+0x10/0x10
[  503.954210]  ret_from_fork_asm+0x1a/0x30
[  503.958142]  
[  503.960328] ---[ end trace 0000000000000000 ]---

Fixes: 967d226eaae8 ("dma-buf: add WARN_ON() illegal dma-fence signaling")
Signed-off-by: Zhu Lingshan 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher 
(cherry picked from commit 2774ef7625adb5fb9e9265c26a59dca7b8fd171e)
Cc: stable@vger.kernel.org

drm/amdkfd: pause autosuspend when creating pdd

2024-12-10T15:26:18+00:00

When using MES creating a pdd will require talking to the GPU to
setup the relevant context. The code here forgot to wake up the GPU
in case it was in suspend, this causes KVM to EFAULT for passthrough
GPU for example. This issue can be masked if the GPU was woken up by
other things (e.g. opening the KMS node) first and have not yet gone to sleep.

v4: do the allocation of proc_ctx_bo in a lazy fashion
when the first queue is created in a process (Felix)

Signed-off-by: Jesse Zhang 
Reviewed-by: Yunxiang Li 
Signed-off-by: Alex Deucher 
Cc: stable@vger.kernel.org

drm/amdkfd: change kfd process kref count at creation

2024-11-11T16:56:08+00:00

kfd process kref count(process->ref) is initialized to 1 by kref_init. After
it is created not need to increase its kref. Instad add kfd process kref at kfd
process mmu notifier allocation since we already decrease the kref at
free_notifier of mmu_notifier_ops, so pair them.

When user process opens kfd node multiple times the kfd process kref is
increased each time to balance with kfd node close operation.

Signed-off-by: Xiaogang Chen 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Fix wrong usage of INIT_WORK()

2024-11-08T16:45:29+00:00

In kfd_procfs_show(), the sdma_activity_work_handler is a local variable
and the sdma_activity_work_handler.sdma_activity_work should initialize
with INIT_WORK_ONSTACK() instead of INIT_WORK().

Fixes: 32cb59f31362 ("drm/amdkfd: Track SDMA utilization per process")
Signed-off-by: Yuan Can 
Signed-off-by: Felix Kuehling 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Use dynamic allocation for CU occupancy array in 'kfd_get_cu_occupancy()'

2024-11-04T17:05:29+00:00

The `kfd_get_cu_occupancy` function previously declared a large
`cu_occupancy` array as a local variable, which could lead to stack
overflows due to excessive stack usage. This commit replaces the static
array allocation with dynamic memory allocation using `kcalloc`,
thereby reducing the stack size.

This change avoids the risk of stack overflows in kernel space,  in
scenarios where `AMDGPU_MAX_QUEUES` is large. The  allocated memory is
freed using `kfree` before the function returns  to prevent memory
leaks.

Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c: In function ‘kfd_get_cu_occupancy’:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:322:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=]
  322 | }
      | ^

Fixes: 6ae9e1aba97e ("drm/amdkfd: Update logic for CU occupancy calculations")
Cc: Harish Kasiviswanathan 
Cc: Felix Kuehling 
Cc: Christian König 
Cc: Alex Deucher 
Signed-off-by: Srinivasan Shanmugam 
Suggested-by: Mukul Joshi 
Reviewed-by: Mukul Joshi 
Signed-off-by: Alex Deucher

drm/amdkfd: Accounting pdd vram_usage for svm

2024-10-15T15:22:19+00:00

Process device data pdd->vram_usage is read by rocm-smi via sysfs, this
is currently missing the svm_bo usage accounting, so "rocm-smi
--showpids" per process VRAM usage report is incorrect.

Add pdd->vram_usage accounting when svm_bo allocation and release,
change to atomic64_t type because it is updated outside process mutex
now.

Signed-off-by: Philip Yang 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Fix an eviction fence leak

2024-10-07T18:03:52+00:00

Only creating a new reference for each process instead of each VM.

Fixes: 9a1c1339abf9 ("drm/amdkfd: Run restore_workers on freezable WQs")
Suggested-by: Felix Kuehling 
Signed-off-by: Lang Yu 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Fix CU occupancy for GFX 9.4.3

2024-09-25T16:56:07+00:00

Make CU occupancy calculations work on GFX 9.4.3 by
updating the logic to handle multiple XCCs correctly.

Signed-off-by: Mukul Joshi 
Reviewed-by: Harish Kasiviswanathan 
Signed-off-by: Alex Deucher

drm/amdkfd: Update logic for CU occupancy calculations

2024-09-25T16:56:00+00:00

Currently, the code uses the IH_VMID_X_LUT register to map
a queue's vmid to the corresponding PASID. This logic is racy
since CP can update the VMID-PASID mapping anytime especially
when there are more processes than number of vmids. Update the
logic to calculate CU occupancy by matching doorbell offset of
the queue with valid wave counts against the process's queues.

Signed-off-by: Mukul Joshi 
Reviewed-by: Harish Kasiviswanathan 
Signed-off-by: Alex Deucher

drm/amdkfd: support per-queue reset on gfx9

2024-08-06T14:43:18+00:00

Support per-queue reset for GFX9.  The recommendation is for the driver
to target reset the HW queue via a SPI MMIO register write.

Since this requires pipe and HW queue info and MEC FW is limited to
doorbell reports of hung queues after an unmap failure, scan the HW
queue slots defined by SET_RESOURCES first to identify the user queue
candidates to reset.

Only signal reset events to processes that have had a queue reset.

If queue reset fails, fall back to GPU reset.

Signed-off-by: Jonathan Kim 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher