diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_prof.c b/drivers/gpu/nvgpu/os/linux/ioctl_prof.c
index 1b2fdf05e..7c21a4161 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_prof.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_prof.c
@@ -470,6 +470,7 @@ static int nvgpu_prof_ioctl_alloc_pma_stream(struct nvgpu_profiler_object_priv *
 		prof->pma_bytes_available_buffer_va);
 
 	args->pma_buffer_va = pma_buffer_offset;
+	args->pma_channel_id = 0u;
 
 	/* Decrement pma_dmabuf ref count as we already mapped it. */
 	dma_buf_put(pma_dmabuf);
@@ -515,7 +516,8 @@ static void nvgpu_prof_free_pma_stream_priv_data(struct nvgpu_profiler_object_pr
 	prof->pma_bytes_available_buffer_cpuva = NULL;
 }
 
-static int nvgpu_prof_ioctl_free_pma_stream(struct nvgpu_profiler_object_priv *priv)
+static int nvgpu_prof_ioctl_free_pma_stream(struct nvgpu_profiler_object_priv *priv,
+		struct nvgpu_profiler_free_pma_stream_args *args)
 {
 	struct nvgpu_profiler_object *prof = priv->prof;
 	struct gk20a *g = prof->g;
@@ -942,7 +944,8 @@ long nvgpu_prof_fops_ioctl(struct file *filp, unsigned int cmd,
 		break;
 
 	case NVGPU_PROFILER_IOCTL_FREE_PMA_STREAM:
-		err = nvgpu_prof_ioctl_free_pma_stream(prof_priv);
+		err = nvgpu_prof_ioctl_free_pma_stream(prof_priv,
+			(struct nvgpu_profiler_free_pma_stream_args *)buf);
 		break;
 
 	case NVGPU_PROFILER_IOCTL_EXEC_REG_OPS:
diff --git a/include/uapi/linux/nvgpu.h b/include/uapi/linux/nvgpu.h
index e8434f83e..d9d640be2 100644
--- a/include/uapi/linux/nvgpu.h
+++ b/include/uapi/linux/nvgpu.h
@@ -832,7 +832,14 @@ struct nvgpu_profiler_alloc_pma_stream_args {
 #define NVGPU_PROFILER_ALLOC_PMA_STREAM_ARG_FLAG_CTXSW		(1 << 0)
 	__u32 flags;
 
-	__u32 reserved[3];
+	__u32 pma_channel_id;				/* out: PMA hardware stream id */
+
+	__u32 reserved[2];
+};
+
+struct nvgpu_profiler_free_pma_stream_args {
+	__u32 pma_channel_id;				/* in: PMA hardware stream id */
+	__u32 reserved[2];
 };
 
 struct nvgpu_profiler_pma_stream_update_get_put_args {
@@ -851,7 +858,9 @@ struct nvgpu_profiler_pma_stream_update_get_put_args {
 #define NVGPU_PROFILER_PMA_STREAM_UPDATE_GET_PUT_ARG_FLAG_OVERFLOW_TRIGGERED		(1 << 3)
 	__u32 flags;
 
-	__u32 reserved[3];
+	__u32 pma_channel_id;			/* in: PMA channel index */
+
+	__u32 reserved[2];
 };
 
 /*
@@ -955,7 +964,7 @@ struct nvgpu_profiler_vab_flush_state_args {
 #define NVGPU_PROFILER_IOCTL_ALLOC_PMA_STREAM \
 	_IOWR(NVGPU_PROFILER_IOCTL_MAGIC, 4, struct nvgpu_profiler_alloc_pma_stream_args)
 #define NVGPU_PROFILER_IOCTL_FREE_PMA_STREAM \
-	_IO(NVGPU_PROFILER_IOCTL_MAGIC, 5)
+	_IOW(NVGPU_PROFILER_IOCTL_MAGIC, 5, struct nvgpu_profiler_free_pma_stream_args)
 #define NVGPU_PROFILER_IOCTL_BIND_PM_RESOURCES \
 	_IO(NVGPU_PROFILER_IOCTL_MAGIC, 6)
 #define NVGPU_PROFILER_IOCTL_UNBIND_PM_RESOURCES \