Merge branch 'net-mitigate-retpoline-overhead'

Paolo Abeni says:

====================
net: mitigate retpoline overhead

The spectre v2 counter-measures, aka retpolines, are a source of measurable
overhead[1]. We can partially address that when the function pointer refers to
a builtin symbol resorting to a list of tests vs well-known builtin function and
direct calls.

Experimental results show that replacing a single indirect call via
retpoline with several branches and a direct call gives performance gains
even when multiple branches are added - 5 or more, as reported in [2].

This may lead to some uglification around the indirect calls. In netconf 2018
Eric Dumazet described a technique to hide the most relevant part of the needed
boilerplate with some macro help.

This series is a [re-]implementation of such idea, exposing the introduced
helpers in a new header file. They are later leveraged to avoid the indirect
call overhead in the GRO path, when possible.

Overall this gives > 10% performance improvement for UDP GRO benchmark and
smaller but measurable for TCP syn flood.

The added infra can be used in follow-up patches to cope with retpoline overhead
in other points of the networking stack (e.g. at the qdisc layer) and possibly
even in other subsystems.

v2  -> v3:
 - fix build error with CONFIG_IPV6=m

v1  -> v2:
 - list explicitly the builtin function names in INDIRECT_CALL_*(),
   as suggested by Ed Cree
 - expand the recipients list

rfc -> v1:
 - use branch prediction hints, as suggested by Eric

[1] http://vger.kernel.org/netconf2018_files/PaoloAbeni_netconf2018.pdf
[2] https://linuxplumbersconf.org/event/2/contributions/99/attachments/98/117/lpc18_paper_af_xdp_perf-v2.pdf
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

9 files changed, 136 insertions, 22 deletions

diff --git a/include/linux/indirect_call_wrapper.h b/include/linux/indirect_call_wrapper.h
new file mode 100644
index 000000000000..7c8b7f4948af
--- /dev/null
+++ b/include/linux/indirect_call_wrapper.h
@@ -0,0 +1,51 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_INDIRECT_CALL_WRAPPER_H
+#define _LINUX_INDIRECT_CALL_WRAPPER_H
+
+#ifdef CONFIG_RETPOLINE
+
+/*
+ * INDIRECT_CALL_$NR - wrapper for indirect calls with $NR known builtin
+ *  @f: function pointer
+ *  @f$NR: builtin functions names, up to $NR of them
+ *  @__VA_ARGS__: arguments for @f
+ *
+ * Avoid retpoline overhead for known builtin, checking @f vs each of them and
+ * eventually invoking directly the builtin function. The functions are check
+ * in the given order. Fallback to the indirect call.
+ */
+#define INDIRECT_CALL_1(f, f1, ...)					\
+	({								\
+		likely(f == f1) ? f1(__VA_ARGS__) : f(__VA_ARGS__);	\
+	})
+#define INDIRECT_CALL_2(f, f2, f1, ...)					\
+	({								\
+		likely(f == f2) ? f2(__VA_ARGS__) :			\
+				  INDIRECT_CALL_1(f, f1, __VA_ARGS__);	\
+	})
+
+#define INDIRECT_CALLABLE_DECLARE(f)	f
+#define INDIRECT_CALLABLE_SCOPE
+
+#else
+#define INDIRECT_CALL_1(f, name, ...) f(__VA_ARGS__)
+#define INDIRECT_CALL_2(f, name, ...) f(__VA_ARGS__)
+#define INDIRECT_CALLABLE_DECLARE(f)
+#define INDIRECT_CALLABLE_SCOPE		static
+#endif
+
+/*
+ * We can use INDIRECT_CALL_$NR for ipv6 related functions only if ipv6 is
+ * builtin, this macro simplify dealing with indirect calls with only ipv4/ipv6
+ * alternatives
+ */
+#if IS_BUILTIN(CONFIG_IPV6)
+#define INDIRECT_CALL_INET(f, f2, f1, ...) \
+	INDIRECT_CALL_2(f, f2, f1, __VA_ARGS__)
+#elif IS_ENABLED(CONFIG_INET)
+#define INDIRECT_CALL_INET(f, f2, f1, ...) INDIRECT_CALL_1(f, f1, __VA_ARGS__)
+#else
+#define INDIRECT_CALL_INET(f, f2, f1, ...) f(__VA_ARGS__)
+#endif
+
+#endif
diff --git a/include/net/inet_common.h b/include/net/inet_common.h
index 3ca969cbd161..975901a95c0f 100644
--- a/include/net/inet_common.h
+++ b/include/net/inet_common.h
@@ -2,6 +2,8 @@
 #ifndef _INET_COMMON_H
 #define _INET_COMMON_H
 
+#include <linux/indirect_call_wrapper.h>
+
 extern const struct proto_ops inet_stream_ops;
 extern const struct proto_ops inet_dgram_ops;
 
@@ -54,4 +56,11 @@ static inline void inet_ctl_sock_destroy(struct sock *sk)
 		sock_release(sk->sk_socket);
 }
 
+#define indirect_call_gro_receive(f2, f1, cb, head, skb)	\
+({								\
+	unlikely(gro_recursion_inc_test(skb)) ?			\
+		NAPI_GRO_CB(skb)->flush |= 1, NULL :		\
+		INDIRECT_CALL_2(cb, f2, f1, head, skb);		\
+})
+
 #endif
diff --git a/net/core/dev.c b/net/core/dev.c
index ed9aa4a91f1f..1b5a4410be0e 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -145,6 +145,7 @@
 #include <linux/sctp.h>
 #include <net/udp_tunnel.h>
 #include <linux/net_namespace.h>
+#include <linux/indirect_call_wrapper.h>
 
 #include "net-sysfs.h"
 
@@ -5338,6 +5339,8 @@ static void flush_all_backlogs(void)
 	put_online_cpus();
 }
 
+INDIRECT_CALLABLE_DECLARE(int inet_gro_complete(struct sk_buff *, int));
+INDIRECT_CALLABLE_DECLARE(int ipv6_gro_complete(struct sk_buff *, int));
 static int napi_gro_complete(struct sk_buff *skb)
 {
 	struct packet_offload *ptype;
@@ -5357,7 +5360,9 @@ static int napi_gro_complete(struct sk_buff *skb)
 		if (ptype->type != type || !ptype->callbacks.gro_complete)
 			continue;
 
-		err = ptype->callbacks.gro_complete(skb, 0);
+		err = INDIRECT_CALL_INET(ptype->callbacks.gro_complete,
+					 ipv6_gro_complete, inet_gro_complete,
+					 skb, 0);
 		break;
 	}
 	rcu_read_unlock();
@@ -5504,6 +5509,10 @@ static void gro_flush_oldest(struct list_head *head)
 	napi_gro_complete(oldest);
 }
 
+INDIRECT_CALLABLE_DECLARE(struct sk_buff *inet_gro_receive(struct list_head *,
+							   struct sk_buff *));
+INDIRECT_CALLABLE_DECLARE(struct sk_buff *ipv6_gro_receive(struct list_head *,
+							   struct sk_buff *));
 static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
 {
 	u32 hash = skb_get_hash_raw(skb) & (GRO_HASH_BUCKETS - 1);
@@ -5553,7 +5562,9 @@ static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff
 			NAPI_GRO_CB(skb)->csum_valid = 0;
 		}
 
-		pp = ptype->callbacks.gro_receive(gro_head, skb);
+		pp = INDIRECT_CALL_INET(ptype->callbacks.gro_receive,
+					ipv6_gro_receive, inet_gro_receive,
+					gro_head, skb);
 		break;
 	}
 	rcu_read_unlock();
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 326c422c22f8..0dfb72c46671 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1385,6 +1385,10 @@ out:
 }
 EXPORT_SYMBOL(inet_gso_segment);
 
+INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp4_gro_receive(struct list_head *,
+							   struct sk_buff *));
+INDIRECT_CALLABLE_DECLARE(struct sk_buff *udp4_gro_receive(struct list_head *,
+							   struct sk_buff *));
 struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb)
 {
 	const struct net_offload *ops;
@@ -1494,7 +1498,8 @@ struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb)
 	skb_gro_pull(skb, sizeof(*iph));
 	skb_set_transport_header(skb, skb_gro_offset(skb));
 
-	pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
+	pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive,
+				       ops->callbacks.gro_receive, head, skb);
 
 out_unlock:
 	rcu_read_unlock();
@@ -1556,6 +1561,8 @@ int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return -EINVAL;
 }
 
+INDIRECT_CALLABLE_DECLARE(int tcp4_gro_complete(struct sk_buff *, int));
+INDIRECT_CALLABLE_DECLARE(int udp4_gro_complete(struct sk_buff *, int));
 int inet_gro_complete(struct sk_buff *skb, int nhoff)
 {
 	__be16 newlen = htons(skb->len - nhoff);
@@ -1581,7 +1588,9 @@ int inet_gro_complete(struct sk_buff *skb, int nhoff)
 	 * because any hdr with option will have been flushed in
 	 * inet_gro_receive().
 	 */
-	err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph));
+	err = INDIRECT_CALL_2(ops->callbacks.gro_complete,
+			      tcp4_gro_complete, udp4_gro_complete,
+			      skb, nhoff + sizeof(*iph));
 
 out_unlock:
 	rcu_read_unlock();
diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
index 870b0a335061..0fbf7d4df9da 100644
--- a/net/ipv4/tcp_offload.c
+++ b/net/ipv4/tcp_offload.c
@@ -10,6 +10,7 @@
  *	TCPv4 GSO/GRO support
  */
 
+#include <linux/indirect_call_wrapper.h>
 #include <linux/skbuff.h>
 #include <net/tcp.h>
 #include <net/protocol.h>
@@ -305,7 +306,8 @@ int tcp_gro_complete(struct sk_buff *skb)
 }
 EXPORT_SYMBOL(tcp_gro_complete);
 
-static struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb)
+INDIRECT_CALLABLE_SCOPE
+struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb)
 {
 	/* Don't bother verifying checksum if we're going to flush anyway. */
 	if (!NAPI_GRO_CB(skb)->flush &&
@@ -318,7 +320,7 @@ static struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *
 	return tcp_gro_receive(head, skb);
 }
 
-static int tcp4_gro_complete(struct sk_buff *skb, int thoff)
+INDIRECT_CALLABLE_SCOPE int tcp4_gro_complete(struct sk_buff *skb, int thoff)
 {
 	const struct iphdr *iph = ip_hdr(skb);
 	struct tcphdr *th = tcp_hdr(skb);
diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
index 0646d61f4fa8..64f9715173ac 100644
--- a/net/ipv4/udp_offload.c
+++ b/net/ipv4/udp_offload.c
@@ -13,6 +13,7 @@
 #include <linux/skbuff.h>
 #include <net/udp.h>
 #include <net/protocol.h>
+#include <net/inet_common.h>
 
 static struct sk_buff *__skb_udp_tunnel_segment(struct sk_buff *skb,
 	netdev_features_t features,
@@ -391,6 +392,8 @@ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
 	return NULL;
 }
 
+INDIRECT_CALLABLE_DECLARE(struct sock *udp6_lib_lookup_skb(struct sk_buff *skb,
+						   __be16 sport, __be16 dport));
 struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
 				struct udphdr *uh, udp_lookup_t lookup)
 {
@@ -402,7 +405,8 @@ struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
 	struct sock *sk;
 
 	rcu_read_lock();
-	sk = (*lookup)(skb, uh->source, uh->dest);
+	sk = INDIRECT_CALL_INET(lookup, udp6_lib_lookup_skb,
+				udp4_lib_lookup_skb, skb, uh->source, uh->dest);
 	if (!sk)
 		goto out_unlock;
 
@@ -451,8 +455,8 @@ out_unlock:
 }
 EXPORT_SYMBOL(udp_gro_receive);
 
-static struct sk_buff *udp4_gro_receive(struct list_head *head,
-					struct sk_buff *skb)
+INDIRECT_CALLABLE_SCOPE
+struct sk_buff *udp4_gro_receive(struct list_head *head, struct sk_buff *skb)
 {
 	struct udphdr *uh = udp_gro_udphdr(skb);
 
@@ -502,7 +506,8 @@ int udp_gro_complete(struct sk_buff *skb, int nhoff,
 	uh->len = newlen;
 
 	rcu_read_lock();
-	sk = (*lookup)(skb, uh->source, uh->dest);
+	sk = INDIRECT_CALL_INET(lookup, udp6_lib_lookup_skb,
+				udp4_lib_lookup_skb, skb, uh->source, uh->dest);
 	if (sk && udp_sk(sk)->gro_enabled) {
 		err = udp_gro_complete_segment(skb);
 	} else if (sk && udp_sk(sk)->gro_complete) {
@@ -525,7 +530,7 @@ int udp_gro_complete(struct sk_buff *skb, int nhoff,
 }
 EXPORT_SYMBOL(udp_gro_complete);
 
-static int udp4_gro_complete(struct sk_buff *skb, int nhoff)
+INDIRECT_CALLABLE_SCOPE int udp4_gro_complete(struct sk_buff *skb, int nhoff)
 {
 	const struct iphdr *iph = ip_hdr(skb);
 	struct udphdr *uh = (struct udphdr *)(skb->data + nhoff);
diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c
index 70f525c33cb6..5c045691c302 100644
--- a/net/ipv6/ip6_offload.c
+++ b/net/ipv6/ip6_offload.c
@@ -20,6 +20,23 @@
 
 #include "ip6_offload.h"
 
+/* All GRO functions are always builtin, except UDP over ipv6, which lays in
+ * ipv6 module, as it depends on UDPv6 lookup function, so we need special care
+ * when ipv6 is built as a module
+ */
+#if IS_BUILTIN(CONFIG_IPV6)
+#define INDIRECT_CALL_L4(f, f2, f1, ...) INDIRECT_CALL_2(f, f2, f1, __VA_ARGS__)
+#else
+#define INDIRECT_CALL_L4(f, f2, f1, ...) INDIRECT_CALL_1(f, f2, __VA_ARGS__)
+#endif
+
+#define indirect_call_gro_receive_l4(f2, f1, cb, head, skb)	\
+({								\
+	unlikely(gro_recursion_inc_test(skb)) ?			\
+		NAPI_GRO_CB(skb)->flush |= 1, NULL :		\
+		INDIRECT_CALL_L4(cb, f2, f1, head, skb);	\
+})
+
 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
 {
 	const struct net_offload *ops = NULL;
@@ -164,8 +181,12 @@ static int ipv6_exthdrs_len(struct ipv6hdr *iph,
 	return len;
 }
 
-static struct sk_buff *ipv6_gro_receive(struct list_head *head,
-					struct sk_buff *skb)
+INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp6_gro_receive(struct list_head *,
+							   struct sk_buff *));
+INDIRECT_CALLABLE_DECLARE(struct sk_buff *udp6_gro_receive(struct list_head *,
+							   struct sk_buff *));
+INDIRECT_CALLABLE_SCOPE struct sk_buff *ipv6_gro_receive(struct list_head *head,
+							 struct sk_buff *skb)
 {
 	const struct net_offload *ops;
 	struct sk_buff *pp = NULL;
@@ -260,7 +281,8 @@ not_same_flow:
 
 	skb_gro_postpull_rcsum(skb, iph, nlen);
 
-	pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
+	pp = indirect_call_gro_receive_l4(tcp6_gro_receive, udp6_gro_receive,
+					 ops->callbacks.gro_receive, head, skb);
 
 out_unlock:
 	rcu_read_unlock();
@@ -301,7 +323,9 @@ static struct sk_buff *ip4ip6_gro_receive(struct list_head *head,
 	return inet_gro_receive(head, skb);
 }
 
-static int ipv6_gro_complete(struct sk_buff *skb, int nhoff)
+INDIRECT_CALLABLE_DECLARE(int tcp6_gro_complete(struct sk_buff *, int));
+INDIRECT_CALLABLE_DECLARE(int udp6_gro_complete(struct sk_buff *, int));
+INDIRECT_CALLABLE_SCOPE int ipv6_gro_complete(struct sk_buff *skb, int nhoff)
 {
 	const struct net_offload *ops;
 	struct ipv6hdr *iph = (struct ipv6hdr *)(skb->data + nhoff);
@@ -320,7 +344,8 @@ static int ipv6_gro_complete(struct sk_buff *skb, int nhoff)
 	if (WARN_ON(!ops || !ops->callbacks.gro_complete))
 		goto out_unlock;
 
-	err = ops->callbacks.gro_complete(skb, nhoff);
+	err = INDIRECT_CALL_L4(ops->callbacks.gro_complete, tcp6_gro_complete,
+			       udp6_gro_complete, skb, nhoff);
 
 out_unlock:
 	rcu_read_unlock();
diff --git a/net/ipv6/tcpv6_offload.c b/net/ipv6/tcpv6_offload.c
index e72947c99454..3179c425d7ff 100644
--- a/net/ipv6/tcpv6_offload.c
+++ b/net/ipv6/tcpv6_offload.c
@@ -9,14 +9,15 @@
  *
  *      TCPv6 GSO/GRO support
  */
+#include <linux/indirect_call_wrapper.h>
 #include <linux/skbuff.h>
 #include <net/protocol.h>
 #include <net/tcp.h>
 #include <net/ip6_checksum.h>
 #include "ip6_offload.h"
 
-static struct sk_buff *tcp6_gro_receive(struct list_head *head,
-					struct sk_buff *skb)
+INDIRECT_CALLABLE_SCOPE
+struct sk_buff *tcp6_gro_receive(struct list_head *head, struct sk_buff *skb)
 {
 	/* Don't bother verifying checksum if we're going to flush anyway. */
 	if (!NAPI_GRO_CB(skb)->flush &&
@@ -29,7 +30,7 @@ static struct sk_buff *tcp6_gro_receive(struct list_head *head,
 	return tcp_gro_receive(head, skb);
 }
 
-static int tcp6_gro_complete(struct sk_buff *skb, int thoff)
+INDIRECT_CALLABLE_SCOPE int tcp6_gro_complete(struct sk_buff *skb, int thoff)
 {
 	const struct ipv6hdr *iph = ipv6_hdr(skb);
 	struct tcphdr *th = tcp_hdr(skb);
diff --git a/net/ipv6/udp_offload.c b/net/ipv6/udp_offload.c
index 828b2457f97b..83b11d0ac091 100644
--- a/net/ipv6/udp_offload.c
+++ b/net/ipv6/udp_offload.c
@@ -11,6 +11,7 @@
  */
 #include <linux/skbuff.h>
 #include <linux/netdevice.h>
+#include <linux/indirect_call_wrapper.h>
 #include <net/protocol.h>
 #include <net/ipv6.h>
 #include <net/udp.h>
@@ -114,8 +115,8 @@ out:
 	return segs;
 }
 
-static struct sk_buff *udp6_gro_receive(struct list_head *head,
-					struct sk_buff *skb)
+INDIRECT_CALLABLE_SCOPE
+struct sk_buff *udp6_gro_receive(struct list_head *head, struct sk_buff *skb)
 {
 	struct udphdr *uh = udp_gro_udphdr(skb);
 
@@ -142,7 +143,7 @@ flush:
 	return NULL;
 }
 
-static int udp6_gro_complete(struct sk_buff *skb, int nhoff)
+INDIRECT_CALLABLE_SCOPE int udp6_gro_complete(struct sk_buff *skb, int nhoff)
 {
 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 	struct udphdr *uh = (struct udphdr *)(skb->data + nhoff);