diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index c27f2fdcea32..580c5f592408 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1760,6 +1760,7 @@ static int nfsd4_do_async_copy(void *data)
 	struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
 	__be32 nfserr;
 
+	trace_nfsd_copy_do_async(copy);
 	if (nfsd4_ssc_is_inter(copy)) {
 		struct file *filp;
 
@@ -1800,17 +1801,23 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 
 	copy->cp_clp = cstate->clp;
 	if (nfsd4_ssc_is_inter(copy)) {
+		trace_nfsd_copy_inter(copy);
 		if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
 			status = nfserr_notsupp;
 			goto out;
 		}
 		status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
-		if (status)
+		if (status) {
+			trace_nfsd_copy_done(copy, status);
 			return nfserr_offload_denied;
+		}
 	} else {
+		trace_nfsd_copy_intra(copy);
 		status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
-		if (status)
+		if (status) {
+			trace_nfsd_copy_done(copy, status);
 			return status;
+		}
 	}
 
 	memcpy(&copy->fh, &cstate->current_fh.fh_handle,
@@ -1847,6 +1854,7 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 				       copy->nf_dst->nf_file, true);
 	}
 out:
+	trace_nfsd_copy_done(copy, status);
 	release_copy_files(copy);
 	return status;
 out_err:
diff --git a/fs/nfsd/trace.h b/fs/nfsd/trace.h
index 803904348871..fbc0ccb40424 100644
--- a/fs/nfsd/trace.h
+++ b/fs/nfsd/trace.h
@@ -1863,6 +1863,93 @@ TRACE_EVENT(nfsd_end_grace,
 	)
 );
 
+DECLARE_EVENT_CLASS(nfsd_copy_class,
+	TP_PROTO(
+		const struct nfsd4_copy *copy
+	),
+	TP_ARGS(copy),
+	TP_STRUCT__entry(
+		__field(bool, intra)
+		__field(bool, async)
+		__field(u32, src_cl_boot)
+		__field(u32, src_cl_id)
+		__field(u32, src_so_id)
+		__field(u32, src_si_generation)
+		__field(u32, dst_cl_boot)
+		__field(u32, dst_cl_id)
+		__field(u32, dst_so_id)
+		__field(u32, dst_si_generation)
+		__field(u64, src_cp_pos)
+		__field(u64, dst_cp_pos)
+		__field(u64, cp_count)
+		__sockaddr(addr, sizeof(struct sockaddr_in6))
+	),
+	TP_fast_assign(
+		const stateid_t *src_stp = &copy->cp_src_stateid;
+		const stateid_t *dst_stp = &copy->cp_dst_stateid;
+
+		__entry->intra = test_bit(NFSD4_COPY_F_INTRA, &copy->cp_flags);
+		__entry->async = !test_bit(NFSD4_COPY_F_SYNCHRONOUS, &copy->cp_flags);
+		__entry->src_cl_boot = src_stp->si_opaque.so_clid.cl_boot;
+		__entry->src_cl_id = src_stp->si_opaque.so_clid.cl_id;
+		__entry->src_so_id = src_stp->si_opaque.so_id;
+		__entry->src_si_generation = src_stp->si_generation;
+		__entry->dst_cl_boot = dst_stp->si_opaque.so_clid.cl_boot;
+		__entry->dst_cl_id = dst_stp->si_opaque.so_clid.cl_id;
+		__entry->dst_so_id = dst_stp->si_opaque.so_id;
+		__entry->dst_si_generation = dst_stp->si_generation;
+		__entry->src_cp_pos = copy->cp_src_pos;
+		__entry->dst_cp_pos = copy->cp_dst_pos;
+		__entry->cp_count = copy->cp_count;
+		__assign_sockaddr(addr, &copy->cp_clp->cl_addr,
+				sizeof(struct sockaddr_in6));
+	),
+	TP_printk("client=%pISpc intra=%d async=%d "
+		"src_stateid[si_generation:0x%x cl_boot:0x%x cl_id:0x%x so_id:0x%x] "
+		"dst_stateid[si_generation:0x%x cl_boot:0x%x cl_id:0x%x so_id:0x%x] "
+		"cp_src_pos=%llu cp_dst_pos=%llu cp_count=%llu",
+		__get_sockaddr(addr), __entry->intra, __entry->async,
+		__entry->src_si_generation, __entry->src_cl_boot,
+		__entry->src_cl_id, __entry->src_so_id,
+		__entry->dst_si_generation, __entry->dst_cl_boot,
+		__entry->dst_cl_id, __entry->dst_so_id,
+		__entry->src_cp_pos, __entry->dst_cp_pos, __entry->cp_count
+	)
+);
+
+#define DEFINE_COPY_EVENT(name)				\
+DEFINE_EVENT(nfsd_copy_class, nfsd_copy_##name,	\
+	TP_PROTO(const struct nfsd4_copy *copy),	\
+	TP_ARGS(copy))
+
+DEFINE_COPY_EVENT(inter);
+DEFINE_COPY_EVENT(intra);
+DEFINE_COPY_EVENT(do_async);
+
+TRACE_EVENT(nfsd_copy_done,
+	TP_PROTO(
+		const struct nfsd4_copy *copy,
+		__be32 status
+	),
+	TP_ARGS(copy, status),
+	TP_STRUCT__entry(
+		__field(int, status)
+		__field(bool, intra)
+		__field(bool, async)
+		__sockaddr(addr, sizeof(struct sockaddr_in6))
+	),
+	TP_fast_assign(
+		__entry->status = be32_to_cpu(status);
+		__entry->intra = test_bit(NFSD4_COPY_F_INTRA, &copy->cp_flags);
+		__entry->async = !test_bit(NFSD4_COPY_F_SYNCHRONOUS, &copy->cp_flags);
+		__assign_sockaddr(addr, &copy->cp_clp->cl_addr,
+				sizeof(struct sockaddr_in6));
+	),
+	TP_printk("addr=%pISpc status=%d intra=%d async=%d ",
+		__get_sockaddr(addr), __entry->status, __entry->intra, __entry->async
+	)
+);
+
 #endif /* _NFSD_TRACE_H */
 
 #undef TRACE_INCLUDE_PATH