diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index a560844103727dce62d2b8ef88290cd321a28936..27ea00489ecc41b08adc8b780bd18f62fa36a791 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -28,6 +28,7 @@
 #include <drm/radeon_drm.h>
 #include "radeon_reg.h"
 #include "radeon.h"
+#include "radeon_trace.h"
 
 static int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 {
@@ -559,6 +560,8 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		return r;
 	}
 
+	trace_radeon_cs(&parser);
+
 	r = radeon_cs_ib_chunk(rdev, &parser);
 	if (r) {
 		goto out;
diff --git a/drivers/gpu/drm/radeon/radeon_trace.h b/drivers/gpu/drm/radeon/radeon_trace.h
index a7d7c6df68e295efd56f17eb1f5fca5819c0e717..f7e367815964f34756a13691935018ae95978ca1 100644
--- a/drivers/gpu/drm/radeon/radeon_trace.h
+++ b/drivers/gpu/drm/radeon/radeon_trace.h
@@ -27,6 +27,26 @@ TRACE_EVENT(radeon_bo_create,
 	    TP_printk("bo=%p, pages=%u", __entry->bo, __entry->pages)
 );
 
+TRACE_EVENT(radeon_cs,
+	    TP_PROTO(struct radeon_cs_parser *p),
+	    TP_ARGS(p),
+	    TP_STRUCT__entry(
+			     __field(u32, ring)
+			     __field(u32, dw)
+			     __field(u32, fences)
+			     ),
+
+	    TP_fast_assign(
+			   __entry->ring = p->ring;
+			   __entry->dw = p->chunks[p->chunk_ib_idx].length_dw;
+			   __entry->fences = radeon_fence_count_emitted(
+				p->rdev, p->ring);
+			   ),
+	    TP_printk("ring=%u, dw=%u, fences=%u",
+		      __entry->ring, __entry->dw,
+		      __entry->fences)
+);
+
 DECLARE_EVENT_CLASS(radeon_fence_request,
 
 	    TP_PROTO(struct drm_device *dev, u32 seqno),