ipv6: Move ipv6_fl_list from ipv6_pinfo to inet_sock.

In {tcp6,udp6,raw6}_sock, struct ipv6_pinfo is always placed at
the beginning of a new cache line because

  1. __alignof__(struct tcp_sock) is 64 due to ____cacheline_aligned
     of __cacheline_group_begin(tcp_sock_write_tx)

  2. __alignof__(struct udp_sock) is 64 due to ____cacheline_aligned
     of struct numa_drop_counters

  3. in raw6_sock, struct numa_drop_counters is placed before
     struct ipv6_pinfo

.  struct ipv6_pinfo is 136 bytes, but the last cache line is
only used by ipv6_fl_list:

  $ pahole -C ipv6_pinfo vmlinux
  struct ipv6_pinfo {
  ...
  	/* --- cacheline 2 boundary (128 bytes) --- */
  	struct ipv6_fl_socklist *  ipv6_fl_list;         /*   128     8 */

  	/* size: 136, cachelines: 3, members: 23 */

Let's move ipv6_fl_list from struct ipv6_pinfo to struct inet_sock
to save a full cache line for {tcp6,udp6,raw6}_sock.

Now, struct ipv6_pinfo is 128 bytes, and {tcp6,udp6,raw6}_sock have
64 bytes less, while {tcp,udp,raw}_sock retain the same size.

Before:

  # grep -E "^(RAW|UDP[^L\-]|TCP)" /proc/slabinfo | awk '{print $1, "\t", $4}'
  RAWv6 	 1408
  UDPv6 	 1472
  TCPv6 	 2560
  RAW 		 1152
  UDP	 	 1280
  TCP 		 2368

After:

  # grep -E "^(RAW|UDP[^L\-]|TCP)" /proc/slabinfo | awk '{print $1, "\t", $4}'
  RAWv6 	 1344
  UDPv6 	 1408
  TCPv6 	 2496
  RAW 		 1152
  UDP	 	 1280
  TCP 		 2368

Also, ipv6_fl_list and inet_flags (SNDFLOW bit) are placed in the
same cache line.

  $ pahole -C inet_sock vmlinux
  ...
  	/* --- cacheline 11 boundary (704 bytes) was 56 bytes ago --- */
  	struct ipv6_pinfo *        pinet6;               /*   760     8 */
  	/* --- cacheline 12 boundary (768 bytes) --- */
  	struct ipv6_fl_socklist *  ipv6_fl_list;         /*   768     8 */
  	unsigned long              inet_flags;           /*   776     8 */

Doc churn is due to the insufficient Type column (only 1 space short).

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20251014224210.2964778-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

2 files changed, 28 insertions, 29 deletions

diff --git a/net/ipv6/ip6_flowlabel.c b/net/ipv6/ip6_flowlabel.c
index a3ff575798dd..60d0be47a9f3 100644
--- a/net/ipv6/ip6_flowlabel.c
+++ b/net/ipv6/ip6_flowlabel.c
@@ -66,8 +66,8 @@ EXPORT_SYMBOL(ipv6_flowlabel_exclusive);
 	     fl != NULL;					\
 	     fl = rcu_dereference(fl->next))
 
-#define for_each_sk_fl_rcu(np, sfl)				\
-	for (sfl = rcu_dereference(np->ipv6_fl_list);	\
+#define for_each_sk_fl_rcu(sk, sfl)				\
+	for (sfl = rcu_dereference(inet_sk(sk)->ipv6_fl_list);	\
 	     sfl != NULL;					\
 	     sfl = rcu_dereference(sfl->next))
 
@@ -262,12 +262,11 @@ static struct ip6_flowlabel *fl_intern(struct net *net,
 struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label)
 {
 	struct ipv6_fl_socklist *sfl;
-	struct ipv6_pinfo *np = inet6_sk(sk);
 
 	label &= IPV6_FLOWLABEL_MASK;
 
 	rcu_read_lock();
-	for_each_sk_fl_rcu(np, sfl) {
+	for_each_sk_fl_rcu(sk, sfl) {
 		struct ip6_flowlabel *fl = sfl->fl;
 
 		if (fl->label == label && atomic_inc_not_zero(&fl->users)) {
@@ -283,16 +282,16 @@ EXPORT_SYMBOL_GPL(__fl6_sock_lookup);
 
 void fl6_free_socklist(struct sock *sk)
 {
-	struct ipv6_pinfo *np = inet6_sk(sk);
+	struct inet_sock *inet = inet_sk(sk);
 	struct ipv6_fl_socklist *sfl;
 
-	if (!rcu_access_pointer(np->ipv6_fl_list))
+	if (!rcu_access_pointer(inet->ipv6_fl_list))
 		return;
 
 	spin_lock_bh(&ip6_sk_fl_lock);
-	while ((sfl = rcu_dereference_protected(np->ipv6_fl_list,
+	while ((sfl = rcu_dereference_protected(inet->ipv6_fl_list,
 						lockdep_is_held(&ip6_sk_fl_lock))) != NULL) {
-		np->ipv6_fl_list = sfl->next;
+		inet->ipv6_fl_list = sfl->next;
 		spin_unlock_bh(&ip6_sk_fl_lock);
 
 		fl_release(sfl->fl);
@@ -470,16 +469,15 @@ done:
 
 static int mem_check(struct sock *sk)
 {
-	struct ipv6_pinfo *np = inet6_sk(sk);
-	struct ipv6_fl_socklist *sfl;
 	int room = FL_MAX_SIZE - atomic_read(&fl_size);
+	struct ipv6_fl_socklist *sfl;
 	int count = 0;
 
 	if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
 		return 0;
 
 	rcu_read_lock();
-	for_each_sk_fl_rcu(np, sfl)
+	for_each_sk_fl_rcu(sk, sfl)
 		count++;
 	rcu_read_unlock();
 
@@ -492,13 +490,15 @@ static int mem_check(struct sock *sk)
 	return 0;
 }
 
-static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
-		struct ip6_flowlabel *fl)
+static inline void fl_link(struct sock *sk, struct ipv6_fl_socklist *sfl,
+			   struct ip6_flowlabel *fl)
 {
+	struct inet_sock *inet = inet_sk(sk);
+
 	spin_lock_bh(&ip6_sk_fl_lock);
 	sfl->fl = fl;
-	sfl->next = np->ipv6_fl_list;
-	rcu_assign_pointer(np->ipv6_fl_list, sfl);
+	sfl->next = inet->ipv6_fl_list;
+	rcu_assign_pointer(inet->ipv6_fl_list, sfl);
 	spin_unlock_bh(&ip6_sk_fl_lock);
 }
 
@@ -520,7 +520,7 @@ int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
 
 	rcu_read_lock();
 
-	for_each_sk_fl_rcu(np, sfl) {
+	for_each_sk_fl_rcu(sk, sfl) {
 		if (sfl->fl->label == (np->flow_label & IPV6_FLOWLABEL_MASK)) {
 			spin_lock_bh(&ip6_fl_lock);
 			freq->flr_label = sfl->fl->label;
@@ -559,7 +559,7 @@ static int ipv6_flowlabel_put(struct sock *sk, struct in6_flowlabel_req *freq)
 	}
 
 	spin_lock_bh(&ip6_sk_fl_lock);
-	for (sflp = &np->ipv6_fl_list;
+	for (sflp = &inet_sk(sk)->ipv6_fl_list;
 	     (sfl = socklist_dereference(*sflp)) != NULL;
 	     sflp = &sfl->next) {
 		if (sfl->fl->label == freq->flr_label)
@@ -579,13 +579,12 @@ found:
 
 static int ipv6_flowlabel_renew(struct sock *sk, struct in6_flowlabel_req *freq)
 {
-	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct net *net = sock_net(sk);
 	struct ipv6_fl_socklist *sfl;
 	int err;
 
 	rcu_read_lock();
-	for_each_sk_fl_rcu(np, sfl) {
+	for_each_sk_fl_rcu(sk, sfl) {
 		if (sfl->fl->label == freq->flr_label) {
 			err = fl6_renew(sfl->fl, freq->flr_linger,
 					freq->flr_expires);
@@ -614,7 +613,6 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq,
 {
 	struct ipv6_fl_socklist *sfl, *sfl1 = NULL;
 	struct ip6_flowlabel *fl, *fl1 = NULL;
-	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct net *net = sock_net(sk);
 	int err;
 
@@ -645,7 +643,7 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq,
 	if (freq->flr_label) {
 		err = -EEXIST;
 		rcu_read_lock();
-		for_each_sk_fl_rcu(np, sfl) {
+		for_each_sk_fl_rcu(sk, sfl) {
 			if (sfl->fl->label == freq->flr_label) {
 				if (freq->flr_flags & IPV6_FL_F_EXCL) {
 					rcu_read_unlock();
@@ -682,7 +680,7 @@ recheck:
 				fl1->linger = fl->linger;
 			if ((long)(fl->expires - fl1->expires) > 0)
 				fl1->expires = fl->expires;
-			fl_link(np, sfl1, fl1);
+			fl_link(sk, sfl1, fl1);
 			fl_free(fl);
 			return 0;
 
@@ -716,7 +714,7 @@ release:
 		}
 	}
 
-	fl_link(np, sfl1, fl);
+	fl_link(sk, sfl1, fl);
 	return 0;
 done:
 	fl_free(fl);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 59c4977a811a..6197dd4e6261 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1386,7 +1386,9 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 		if (!newsk)
 			return NULL;
 
-		inet_sk(newsk)->pinet6 = tcp_inet6_sk(newsk);
+		newinet = inet_sk(newsk);
+		newinet->pinet6 = tcp_inet6_sk(newsk);
+		newinet->ipv6_fl_list = NULL;
 
 		newnp = tcp_inet6_sk(newsk);
 		newtp = tcp_sk(newsk);
@@ -1405,7 +1407,6 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 
 		newnp->ipv6_mc_list = NULL;
 		newnp->ipv6_ac_list = NULL;
-		newnp->ipv6_fl_list = NULL;
 		newnp->pktoptions  = NULL;
 		newnp->opt	   = NULL;
 		newnp->mcast_oif   = inet_iif(skb);
@@ -1453,10 +1454,12 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 	newsk->sk_gso_type = SKB_GSO_TCPV6;
 	inet6_sk_rx_dst_set(newsk, skb);
 
-	inet_sk(newsk)->pinet6 = tcp_inet6_sk(newsk);
+	newinet = inet_sk(newsk);
+	newinet->pinet6 = tcp_inet6_sk(newsk);
+	newinet->ipv6_fl_list = NULL;
+	newinet->inet_opt = NULL;
 
 	newtp = tcp_sk(newsk);
-	newinet = inet_sk(newsk);
 	newnp = tcp_inet6_sk(newsk);
 
 	memcpy(newnp, np, sizeof(struct ipv6_pinfo));
@@ -1469,10 +1472,8 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 
 	   First: no IPv4 options.
 	 */
-	newinet->inet_opt = NULL;
 	newnp->ipv6_mc_list = NULL;
 	newnp->ipv6_ac_list = NULL;
-	newnp->ipv6_fl_list = NULL;
 
 	/* Clone RX bits */
 	newnp->rxopt.all = np->rxopt.all;