Add ftrace events around user queue creation and destruction to profile queue setup and teardown latency.
Signed-off-by: Prike Liang <[email protected]> --- drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 58 +++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 9 ++++ 2 files changed, 67 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index d13e64a69e25..df98be22f1f5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -578,6 +578,64 @@ TRACE_EVENT(amdgpu_reset_reg_dumps, __entry->value) ); +DECLARE_EVENT_CLASS(amdgpu_userq_queue, + TP_PROTO(struct amdgpu_usermode_queue *queue), + TP_ARGS(queue), + TP_STRUCT__entry( + __field(void *, queue) + __field(u64, doorbell_index) + __field(int, queue_type) + __field(int, state) + __field(u32, xcp_id) + ), + TP_fast_assign( + __entry->queue = queue; + __entry->doorbell_index = queue->doorbell_index; + __entry->queue_type = queue->queue_type; + __entry->state = queue->state; + __entry->xcp_id = queue->xcp_id; + ), + TP_printk("queue=%p, doorbell=%llu, type=%d, state=%d, xcp_id=%u", + __entry->queue, __entry->doorbell_index, + __entry->queue_type, __entry->state, __entry->xcp_id) +); +DEFINE_EVENT(amdgpu_userq_queue, amdgpu_userq_create_start, + TP_PROTO(struct amdgpu_usermode_queue *queue), + TP_ARGS(queue)); +DEFINE_EVENT(amdgpu_userq_queue, amdgpu_userq_destroy_start, + TP_PROTO(struct amdgpu_usermode_queue *queue), + TP_ARGS(queue)); +DECLARE_EVENT_CLASS(amdgpu_userq_queue_result, + TP_PROTO(struct amdgpu_usermode_queue *queue, int result), + TP_ARGS(queue, result), + TP_STRUCT__entry( + __field(void *, queue) + __field(u64, doorbell_index) + __field(int, queue_type) + __field(int, state) + __field(u32, xcp_id) + __field(int, result) + ), + TP_fast_assign( + __entry->queue = queue; + __entry->doorbell_index = queue->doorbell_index; + __entry->queue_type = queue->queue_type; + __entry->state = queue->state; + __entry->xcp_id = queue->xcp_id; + __entry->result = result; + ), + TP_printk("queue=%p, doorbell=%llu, type=%d, state=%d, xcp_id=%u, result=%d", + __entry->queue, __entry->doorbell_index, + __entry->queue_type, __entry->state, + __entry->xcp_id, __entry->result) +); +DEFINE_EVENT(amdgpu_userq_queue_result, amdgpu_userq_create_end, + TP_PROTO(struct amdgpu_usermode_queue *queue, int result), + TP_ARGS(queue, result)); +DEFINE_EVENT(amdgpu_userq_queue_result, amdgpu_userq_destroy_end, + TP_PROTO(struct amdgpu_usermode_queue *queue, int result), + TP_ARGS(queue, result)); + #undef AMDGPU_JOB_GET_TIMELINE_NAME #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 38e310a8694d..9dc6cb579ac7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -33,6 +33,7 @@ #include "amdgpu_userq.h" #include "amdgpu_hmm.h" #include "amdgpu_userq_fence.h" +#include "amdgpu_trace.h" u32 amdgpu_userq_get_supported_ip_mask(struct amdgpu_device *adev) { @@ -509,6 +510,8 @@ amdgpu_userq_destroy(struct amdgpu_userq_mgr *uq_mgr, struct amdgpu_usermode_que const struct amdgpu_userq_funcs *uq_funcs = adev->userq_funcs[queue->queue_type]; int r = 0; + trace_amdgpu_userq_destroy_start(queue); + cancel_delayed_work_sync(&uq_mgr->resume_work); /* Cancel any pending hang detection work and cleanup */ @@ -538,6 +541,7 @@ amdgpu_userq_destroy(struct amdgpu_userq_mgr *uq_mgr, struct amdgpu_usermode_que amdgpu_bo_unpin(queue->wptr_obj.obj); amdgpu_bo_unreserve(queue->wptr_obj.obj); amdgpu_bo_unref(&queue->wptr_obj.obj); + trace_amdgpu_userq_destroy_end(queue, r); kfree(queue); pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); @@ -639,6 +643,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) INIT_DELAYED_WORK(&queue->hang_detect_work, amdgpu_userq_hang_detect_work); + trace_amdgpu_userq_create_start(queue); r = amdgpu_userq_fence_driver_alloc(adev, &queue->fence_drv); if (r) goto free_queue; @@ -700,6 +705,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) r = amdgpu_userq_map_helper(queue); if (r) { drm_file_err(uq_mgr->file, "Failed to map Queue\n"); + trace_amdgpu_userq_create_end(queue, r); mutex_unlock(&uq_mgr->userq_mutex); goto erase_doorbell; } @@ -716,11 +722,13 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) * This drops the last reference which should take care of * all cleanup. */ + trace_amdgpu_userq_create_end(queue, r); amdgpu_userq_put(queue); return r; } amdgpu_debugfs_userq_init(filp, queue, qid); + trace_amdgpu_userq_create_end(queue, 0); args->out.queue_id = qid; return 0; @@ -736,6 +744,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) free_fence_drv: amdgpu_userq_fence_driver_free(queue); free_queue: + trace_amdgpu_userq_create_end(queue, r); kfree(queue); err_pm_runtime: pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); -- 2.34.1
