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.
Fixes: cc009e613de6 ("drm/amdkfd: Add KFD support for soc21 v3")
Signed-off-by: Yunxiang Li <Yunxiang.Li@xxxxxxx>
---
v3: remove the cleanup in kfd_bind_process_to_device and document why
this issue doesn't always happen
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 555a892fcf963..c81c020af75d1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1635,12 +1635,19 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev,
atomic64_set(&pdd->evict_duration_counter, 0);
if (dev->kfd->shared_resources.enable_mes) {
+ retval = pm_runtime_resume_and_get(bdev);
+ if (retval < 0) {
+ pr_err("failed to stop autosuspend\n");
+ goto err_free_pdd;
+ }
retval = amdgpu_amdkfd_alloc_gtt_mem(adev,
AMDGPU_MES_PROC_CTX_SIZE,
&pdd->proc_ctx_bo,
&pdd->proc_ctx_gpu_addr,
&pdd->proc_ctx_cpu_ptr,
false);
+ pm_runtime_mark_last_busy(bdev);
+ pm_runtime_put_autosuspend(bdev);
if (retval) {
dev_err(bdev,
"failed to allocate process context bo\n");
--
2.34.1
