tmakin/ack-tegra
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

drm/amdkfd: Handle queue destroy buffer access race

Browse Source 
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>
This commit is contained in:
Philip Yang
2024-08-02 11:28:45 -04:00
committed by Alex Deucher
parent 70f83e7706
commit a1fc9f584c
4 changed files with 53 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+5 -3
View File
@@ -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;