drm/amdkfd: Add PTL control IOCTL Option and unify refcount logic

Introduce a new IOCTL option to allow userspace explicit control over
the Peak Tops Limiter (PTL) state for profiling

Link: https://github.com/ROCm/rocm-systems/tree/develop/projects/rocprofiler-sdk
Signed-off-by: Perry Yuan <perry.yuan@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

5 files changed, 118 insertions, 0 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 467a3dbe1bfa..aab6a4de54fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -2400,6 +2400,8 @@ static int gfx_v9_4_3_perf_monitor_ptl_init(struct amdgpu_device *adev, bool ena
 
 	ptl->hw_supported = true;
 
+	atomic_set(&ptl->disable_ref, 0);
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index fc00d0418684..883de31df04d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1774,6 +1774,104 @@ static int kfd_ioctl_svm(struct file *filep, struct kfd_process *p, void *data)
 }
 #endif
 
+static int kfd_ptl_control(struct kfd_process_device *pdd, bool enable)
+{
+	struct amdgpu_device *adev = pdd->dev->adev;
+	struct amdgpu_ptl *ptl = &adev->psp.ptl;
+	enum amdgpu_ptl_fmt pref_format1 = ptl->fmt1;
+	enum amdgpu_ptl_fmt pref_format2 = ptl->fmt2;
+	uint32_t ptl_state = enable ? 1 : 0;
+	int ret;
+
+	if (!ptl->hw_supported)
+		return -EOPNOTSUPP;
+
+	if (!pdd->dev->kfd2kgd || !pdd->dev->kfd2kgd->ptl_ctrl)
+		return -EOPNOTSUPP;
+
+	ret = pdd->dev->kfd2kgd->ptl_ctrl(adev, PSP_PTL_PERF_MON_SET,
+					  &ptl_state,
+					  &pref_format1,
+					  &pref_format2);
+	return ret;
+}
+
+int kfd_ptl_disable_request(struct kfd_process_device *pdd,
+		struct kfd_process *p)
+{
+	struct amdgpu_device *adev = pdd->dev->adev;
+	struct amdgpu_ptl *ptl = &adev->psp.ptl;
+	int ret = 0;
+
+	mutex_lock(&ptl->mutex);
+
+	if (pdd->ptl_disable_req)
+		goto out;
+
+	if (atomic_inc_return(&ptl->disable_ref) == 1) {
+		ret = kfd_ptl_control(pdd, false);
+		if (ret) {
+			atomic_dec(&ptl->disable_ref);
+			dev_warn(pdd->dev->adev->dev,
+					"failed to disable PTL\n");
+			goto out;
+		}
+	}
+	pdd->ptl_disable_req = true;
+
+out:
+	mutex_unlock(&ptl->mutex);
+	return ret;
+}
+
+int kfd_ptl_disable_release(struct kfd_process_device *pdd,
+		struct kfd_process *p)
+{
+	struct amdgpu_device *adev = pdd->dev->adev;
+	struct amdgpu_ptl *ptl = &adev->psp.ptl;
+	int ret = 0;
+
+	mutex_lock(&ptl->mutex);
+
+	if (!pdd->ptl_disable_req)
+		goto out;
+
+	if (atomic_dec_return(&ptl->disable_ref) == 0) {
+		ret = kfd_ptl_control(pdd, true);
+		if (ret) {
+			atomic_inc(&ptl->disable_ref);
+			dev_warn(adev->dev, "Failed to enable PTL on release: %d\n", ret);
+			goto out;
+		}
+	}
+	pdd->ptl_disable_req = false;
+
+out:
+	mutex_unlock(&ptl->mutex);
+	return ret;
+}
+
+static int kfd_profiler_ptl_control(struct kfd_process *p,
+		struct kfd_ioctl_ptl_control *args)
+{
+	struct kfd_process_device *pdd;
+	int ret;
+
+	mutex_lock(&p->mutex);
+	pdd = kfd_process_device_data_by_id(p, args->gpu_id);
+	mutex_unlock(&p->mutex);
+
+	if (!pdd || !pdd->dev || !pdd->dev->kfd)
+		return -EINVAL;
+
+	if (args->enable == 0)
+		ret = kfd_ptl_disable_request(pdd, p);
+	else
+		ret = kfd_ptl_disable_release(pdd, p);
+
+	return ret;
+}
+
 static int criu_checkpoint_process(struct kfd_process *p,
 			     uint8_t __user *user_priv_data,
 			     uint64_t *priv_offset)
@@ -3242,6 +3340,7 @@ static inline uint32_t profile_lock_device(struct kfd_process *p,
 		if (!kfd->profiler_process) {
 			kfd->profiler_process = p;
 			status = 0;
+			kfd_ptl_disable_request(pdd, p);
 		} else if (kfd->profiler_process == p) {
 			status = -EALREADY;
 		} else {
@@ -3250,6 +3349,7 @@ static inline uint32_t profile_lock_device(struct kfd_process *p,
 	} else if (op == 0 && kfd->profiler_process == p) {
 		kfd->profiler_process = NULL;
 		status = 0;
+		kfd_ptl_disable_release(pdd, p);
 	}
 	mutex_unlock(&kfd->profiler_lock);
 
@@ -3292,6 +3392,8 @@ static int kfd_ioctl_profiler(struct file *filep, struct kfd_process *p, void *d
 		return 0;
 	case KFD_IOC_PROFILER_PMC:
 		return kfd_profiler_pmc(p, &args->pmc);
+	case KFD_IOC_PROFILER_PTL_CONTROL:
+		return kfd_profiler_ptl_control(p, &args->ptl);
 	}
 	return -EINVAL;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 903386e0740b..482bcfa10f82 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -872,6 +872,8 @@ struct kfd_process_device {
 	bool has_reset_queue;
 
 	u32 pasid;
+	/* Indicates this process has requested PTL stay disabled */
+	bool ptl_disable_req;
 };
 
 #define qpd_to_pdd(x) container_of(x, struct kfd_process_device, qpd)
@@ -1603,6 +1605,12 @@ static inline bool kfd_is_first_node(struct kfd_node *node)
 	return (node == node->kfd->nodes[0]);
 }
 
+/* PTL support */
+int kfd_ptl_disable_request(struct kfd_process_device *pdd,
+		struct kfd_process *p);
+int kfd_ptl_disable_release(struct kfd_process_device *pdd,
+		struct kfd_process *p);
+
 /* Debugfs */
 #if defined(CONFIG_DEBUG_FS)
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 1a8cb512dfe3..368283d53077 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1128,6 +1128,10 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
 		pr_debug("Releasing pdd (topology id %d, for pid %d)\n",
 			pdd->dev->id, p->lead_thread->pid);
 		kfd_process_profiler_release(p, pdd);
+
+		if (pdd->ptl_disable_req)
+			kfd_ptl_disable_release(pdd, p);
+
 		kfd_process_device_destroy_cwsr_dgpu(pdd);
 		kfd_process_device_destroy_ib_mem(pdd);
 
diff --git a/drivers/gpu/drm/amd/include/amdgpu_ptl.h b/drivers/gpu/drm/amd/include/amdgpu_ptl.h
index ffed443a14ae..9e63a9a9680a 100644
--- a/drivers/gpu/drm/amd/include/amdgpu_ptl.h
+++ b/drivers/gpu/drm/amd/include/amdgpu_ptl.h
@@ -39,6 +39,8 @@ struct amdgpu_ptl {
 	enum amdgpu_ptl_fmt		fmt2;
 	bool				enabled;
 	bool				hw_supported;
+	/* PTL disable reference counting */
+	atomic_t			disable_ref;
 	struct mutex			mutex;
 };