diff options
| author | Philip Yang <Philip.Yang@amd.com> | 2024-08-02 11:28:45 -0400 |
|---|---|---|
| committer | Alex Deucher <alexander.deucher@amd.com> | 2024-08-13 10:43:16 -0400 |
| commit | a1fc9f584c4aaf8bc1ebfa459fc57a3f26a290d8 (patch) | |
| tree | 92f73c1a86f4892b86270f03f518285bae2d2360 /drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | |
| parent | 70f83e7706e57200edb8ffa36883b2f43d214142 (diff) | |
drm/amdkfd: Handle queue destroy buffer access race
Add helper function kfd_queue_unreference_buffers to reduce queue buffer
refcount, separate it from release queue buffers.
Because it is circular locking to hold dqm_lock to take vm lock,
kfd_ioctl_destroy_queue should take vm lock, unreference queue buffers
first, but not release queue buffers, to handle error in case failed to
hold vm lock. Then hold dqm_lock to remove queue from queue list and
then release queue buffers.
Restore process worker restore queue hold dqm_lock, will always find
the queue with valid queue buffers.
v2 (Felix):
- renamed kfd_queue_unreference_buffer(s) to kfd_queue_unref_bo_va(s)
- added two FIXME comments for follow up
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 8 |
1 files changed, 5 insertions, 3 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index f732ee35b531..20ea745729ee 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -217,6 +217,7 @@ void pqm_uninit(struct process_queue_manager *pqm) list_for_each_entry_safe(pqn, next, &pqm->queues, process_queue_list) { if (pqn->q) { pdd = kfd_get_process_device_data(pqn->q->device, pqm->process); + kfd_queue_unref_bo_vas(pdd, &pqn->q->properties); kfd_queue_release_buffers(pdd, &pqn->q->properties); pqm_clean_queue_resource(pqm, pqn); } @@ -512,7 +513,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) } if (pqn->q) { - retval = kfd_queue_release_buffers(pdd, &pqn->q->properties); + retval = kfd_queue_unref_bo_vas(pdd, &pqn->q->properties); if (retval) goto err_destroy_queue; @@ -526,7 +527,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) if (retval != -ETIME) goto err_destroy_queue; } - + kfd_queue_release_buffers(pdd, &pqn->q->properties); pqm_clean_queue_resource(pqm, pqn); uninit_queue(pqn->q); } @@ -579,7 +580,8 @@ int pqm_update_queue_properties(struct process_queue_manager *pqm, return -EFAULT; } - kfd_queue_buffer_put(vm, &pqn->q->properties.ring_bo); + kfd_queue_unref_bo_va(vm, &pqn->q->properties.ring_bo); + kfd_queue_buffer_put(&pqn->q->properties.ring_bo); amdgpu_bo_unreserve(vm->root.bo); pqn->q->properties.ring_bo = p->ring_bo; |