diff --git a/man/man8/tc-actions.8 b/man/man8/tc-actions.8
index f46166e..bee59f7 100644
--- a/man/man8/tc-actions.8
+++ b/man/man8/tc-actions.8
@@ -47,6 +47,8 @@ actions \- independently defined actions in tc
 ] [
 .I COOKIESPEC
 ] [
+.I FLAGS
+] [
 .I CONTROL
 ]
 
@@ -71,6 +73,10 @@ ACTNAME
 :=
 .BI cookie " COOKIE"
 
+.I FLAGS
+:=
+.I no_percpu
+
 .I ACTDETAIL
 :=
 .I ACTNAME ACTPARAMS
@@ -187,6 +193,14 @@ The value to be stored is completely arbitrary and does not require a specific
 format. It is stored inside the action structure itself.
 
 .TP
+.I FLAGS
+Action-specific flags. Currently, the only supported flag is
+.I no_percpu
+which indicates that action is expected to have minimal software data-path
+traffic and doesn't need to allocate stat counters with percpu allocator.
+This option is intended to be used by hardware-offloaded actions.
+
+.TP
 .BI since " MSTIME"
 When dumping large number of actions, a millisecond time-filter can be
 specified
diff --git a/tc/m_action.c b/tc/m_action.c
index bdc6272..c46aeaa 100644
--- a/tc/m_action.c
+++ b/tc/m_action.c
@@ -249,6 +249,16 @@ done0:
 				addattr_l(n, MAX_MSG, TCA_ACT_COOKIE,
 					  &act_ck, act_ck_len);
 
+			if (*argv && strcmp(*argv, "no_percpu") == 0) {
+				struct nla_bitfield32 flags =
+					{ TCA_ACT_FLAGS_NO_PERCPU_STATS,
+					  TCA_ACT_FLAGS_NO_PERCPU_STATS };
+
+				addattr_l(n, MAX_MSG, TCA_ACT_FLAGS, &flags,
+					  sizeof(struct nla_bitfield32));
+				NEXT_ARG_FWD();
+			}
+
 			addattr_nest_end(n, tail);
 			ok++;
 		}
@@ -317,6 +327,15 @@ static int tc_print_one_action(FILE *f, struct rtattr *arg)
 					   strsz, b1, sizeof(b1)));
 		print_string(PRINT_FP, NULL, "%s", _SL_);
 	}
+	if (tb[TCA_ACT_FLAGS]) {
+		struct nla_bitfield32 *flags = RTA_DATA(tb[TCA_ACT_FLAGS]);
+
+		if (flags->selector & TCA_ACT_FLAGS_NO_PERCPU_STATS)
+			print_bool(PRINT_ANY, "no_percpu", "\tno_percpu",
+				   flags->value &
+				   TCA_ACT_FLAGS_NO_PERCPU_STATS);
+		print_string(PRINT_FP, NULL, "%s", _SL_);
+	}
 
 	return 0;
 }