drm/amdgpu: use correct gfp_t for job allocation

For job allocation in GPU reset and page fault handling we must use
GFP_ATOMIC to guarantee that we don't cycle back and depend on a
dma_fence submission for the memory allocation.

Add gfp_flags argument to amdgpu_job_alloc() and expose the
gfp_flags of IB pools with amdgpu_ib_pool_gfp_flags() so that
we can use different flags when allocating jobs.

Assisted-by: Claude:Sonnet 4
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

6 files changed, 22 insertions, 7 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index a496c383cf6e..121282dd30c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -711,7 +711,8 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 		goto err;
 	}
 
-	ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, 0, &job);
+	ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, 0, GFP_KERNEL,
+			       &job);
 	if (ret)
 		goto err;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 2a6859b6ef05..e8eaa3908ca0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -268,7 +268,7 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 	for (i = 0; i < p->gang_size; ++i) {
 		ret = amdgpu_job_alloc(p->adev, vm, p->entities[i], vm,
 				       num_ibs[i], p->filp->client_id,
-				       &p->jobs[i]);
+				       GFP_KERNEL, &p->jobs[i]);
 		if (ret)
 			goto free_all_kdata;
 		switch (p->adev->enforce_isolation[fpriv->xcp_id]) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index dca96f71efd1..634b3f7a5fff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -417,6 +417,17 @@ void amdgpu_ib_pool_fini(struct amdgpu_device *adev)
 }
 
 /**
+ * amdgpu_ib_pool_gfp_flags - Returns the gfp flags to use for each pool
+ * @adev: amdgpu device pointer
+ * @type: the IB pool type
+ */
+gfp_t amdgpu_ib_pool_gfp_flags(struct amdgpu_device *adev,
+			       enum amdgpu_ib_pool_type type)
+{
+	return adev->ib_pools[type].gfp_flags;
+}
+
+/**
  * amdgpu_ib_ring_tests - test IBs on the rings
  *
  * @adev: amdgpu_device pointer
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index fa695bdcf6eb..d97cf1212e0f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -199,7 +199,7 @@ exit:
 int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		     struct drm_sched_entity *entity, void *owner,
 		     unsigned int num_ibs, u64 drm_client_id,
-		     struct amdgpu_job **job)
+		     gfp_t gfp_flags, struct amdgpu_job **job)
 {
 	struct amdgpu_fence *af;
 	int r;
@@ -207,18 +207,18 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (num_ibs == 0)
 		return -EINVAL;
 
-	*job = kzalloc_flex(**job, ibs, num_ibs);
+	*job = kzalloc_flex(**job, ibs, num_ibs, gfp_flags);
 	if (!*job)
 		return -ENOMEM;
 
-	af = kzalloc_obj(struct amdgpu_fence);
+	af = kzalloc_obj(struct amdgpu_fence, gfp_flags);
 	if (!af) {
 		r = -ENOMEM;
 		goto err_job;
 	}
 	(*job)->hw_fence = af;
 
-	af = kzalloc_obj(struct amdgpu_fence);
+	af = kzalloc_obj(struct amdgpu_fence, gfp_flags);
 	if (!af) {
 		r = -ENOMEM;
 		goto err_fence;
@@ -257,6 +257,7 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 	int r;
 
 	r = amdgpu_job_alloc(adev, NULL, entity, owner, 1, k_job_id,
+			     amdgpu_ib_pool_gfp_flags(adev, pool_type),
 			     job);
 	if (r)
 		return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
index 822849693b3a..7e03bd2ff639 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
@@ -116,7 +116,7 @@ static inline struct amdgpu_ring *amdgpu_job_ring(struct amdgpu_job *job)
 int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		     struct drm_sched_entity *entity, void *owner,
 		     unsigned int num_ibs, u64 drm_client_id,
-		     struct amdgpu_job **job);
+		     gfp_t gfp_flags, struct amdgpu_job **job);
 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 			     struct drm_sched_entity *entity, void *owner,
 			     size_t size, enum amdgpu_ib_pool_type pool_type,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index b11e9f244b53..015623b8fd05 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -583,6 +583,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		       struct dma_fence **f);
 int amdgpu_ib_pool_init(struct amdgpu_device *adev);
 void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
+gfp_t amdgpu_ib_pool_gfp_flags(struct amdgpu_device *adev,
+			       enum amdgpu_ib_pool_type type);
 int amdgpu_ib_ring_tests(struct amdgpu_device *adev);
 bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring);
 void amdgpu_ring_backup_unprocessed_commands(struct amdgpu_ring *ring,