[PATCH] drm/amdkfd: Move the process suspend and resume out of full access

[Emily Deng]

For the suspend and resume process, exclusive access is not required.
Therefore, it can be moved out of the full access section to reduce the
duration of exclusive access.

Signed-off-by: Emily Deng <Emily.Deng at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 16 +++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 13 +++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 12 +++++--
 drivers/gpu/drm/amd/amdkfd/kfd_device.c    | 40 ++++++++++++++++++----
 4 files changed, 72 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index c166e059ead3..ccf898a73897 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -265,6 +265,22 @@ int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm)
 	return r;
 }
 
+void amdgpu_amdkfd_suspend_process(struct amdgpu_device *adev, bool run_pm)
+{
+	if (adev->kfd.dev)
+		kgd2kfd_suspend_process(adev->kfd.dev, run_pm, true);
+}
+
+int amdgpu_amdkfd_resume_process(struct amdgpu_device *adev, bool run_pm)
+{
+	int r = 0;
+
+	if (adev->kfd.dev)
+		r = kgd2kfd_resume_process(adev->kfd.dev, run_pm);
+
+	return r;
+}
+
 int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev,
 			    struct amdgpu_reset_context *reset_context)
 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 3a54f90ffcf8..78eaac829a58 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -162,6 +162,8 @@ void amdgpu_amdkfd_fini(void);
 
 void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm);
 int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm);
+void amdgpu_amdkfd_suspend_process(struct amdgpu_device *adev, bool run_pm);
+int amdgpu_amdkfd_resume_process(struct amdgpu_device *adev, bool run_pm);
 void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 			const void *ih_ring_entry);
 void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
@@ -491,6 +493,8 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 void kgd2kfd_device_exit(struct kfd_dev *kfd);
 void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm, bool force);
 int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm);
+void kgd2kfd_suspend_process(struct kfd_dev *kfd, bool run_pm, bool force);
+int kgd2kfd_resume_process(struct kfd_dev *kfd, bool run_pm);
 int kgd2kfd_pre_reset(struct kfd_dev *kfd,
 		      struct amdgpu_reset_context *reset_context);
 int kgd2kfd_post_reset(struct kfd_dev *kfd);
@@ -541,6 +545,15 @@ static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
 	return 0;
 }
 
+static inline void kgd2kfd_suspend_process(struct kfd_dev *kfd, bool run_pm, bool force)
+{
+}
+
+static inline int kgd2kfd_resume_process(struct kfd_dev *kfd, bool run_pm)
+{
+	return 0;
+}
+
 static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd,
 				    struct amdgpu_reset_context *reset_context)
 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 4e343561a17d..34f040820546 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -5032,7 +5032,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool notify_clients)
 	amdgpu_device_ip_suspend_phase1(adev);
 
 	if (!adev->in_s0ix)
-		amdgpu_amdkfd_suspend(adev, adev->in_runpm);
+		amdgpu_amdkfd_suspend(adev, !amdgpu_sriov_runtime(adev) || adev->in_runpm);
 
 	r = amdgpu_device_evict_resources(adev);
 	if (r)
@@ -5049,6 +5049,9 @@ int amdgpu_device_suspend(struct drm_device *dev, bool notify_clients)
 	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_release_full_gpu(adev, false);
 
+	if (!adev->in_s0ix)
+		amdgpu_amdkfd_suspend_process(adev, adev->in_runpm);
+
 	r = amdgpu_dpm_notify_rlc_state(adev, false);
 	if (r)
 		return r;
@@ -5119,7 +5122,7 @@ int amdgpu_device_resume(struct drm_device *dev, bool notify_clients)
 	}
 
 	if (!adev->in_s0ix) {
-		r = amdgpu_amdkfd_resume(adev, adev->in_runpm);
+		r = amdgpu_amdkfd_resume(adev, !amdgpu_sriov_runtime(adev) || adev->in_runpm);
 		if (r)
 			goto exit;
 	}
@@ -5136,6 +5139,11 @@ int amdgpu_device_resume(struct drm_device *dev, bool notify_clients)
 		amdgpu_virt_release_full_gpu(adev, true);
 	}
 
+	if (!adev->in_s0ix) {
+		r = amdgpu_amdkfd_resume_process(adev, adev->in_runpm);
+		if (r)
+			goto exit;
+	}
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index bc8f4e8f9905..23c30320ca3a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -1061,13 +1061,8 @@ void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm, bool force)
 		return;
 
 	/* for runtime suspend, skip locking kfd */
-	if (!run_pm) {
-		mutex_lock(&kfd_processes_mutex);
-		/* For first KFD device suspend all the KFD processes */
-		if (++kfd_locked == 1)
-			kfd_suspend_all_processes(force);
-		mutex_unlock(&kfd_processes_mutex);
-	}
+	kgd2kfd_suspend_process(kfd, run_pm, force);
+
 
 	for (i = 0; i < kfd->num_nodes; i++) {
 		node = kfd->nodes[i];
@@ -1088,6 +1083,37 @@ int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
 			return ret;
 	}
 
+	ret = kgd2kfd_resume_process(kfd, run_pm);
+
+
+	return ret;
+}
+
+void kgd2kfd_suspend_process(struct kfd_dev *kfd, bool run_pm, bool force)
+{
+	struct kfd_node *node;
+	int i;
+
+	if (!kfd->init_complete)
+		return;
+
+	/* for runtime suspend, skip locking kfd */
+	if (!run_pm) {
+		mutex_lock(&kfd_processes_mutex);
+		/* For first KFD device suspend all the KFD processes */
+		if (++kfd_locked == 1)
+			kfd_suspend_all_processes(force);
+		mutex_unlock(&kfd_processes_mutex);
+	}
+}
+
+int kgd2kfd_resume_process(struct kfd_dev *kfd, bool run_pm)
+{
+	int ret, i;
+
+	if (!kfd->init_complete)
+		return 0;
+
 	/* for runtime resume, skip unlocking kfd */
 	if (!run_pm) {
 		mutex_lock(&kfd_processes_mutex);
-- 
2.34.1