Merge branch 'listener-sock-const'

Eric Dumazet says:

====================
dccp/tcp: constify listener sock

Another patch bomb to prepare lockless TCP/DCCP LISTEN handling.

SYNACK retransmits are built and sent without listener socket
being locked. Soon, initial SYNACK packets will have same property.

This series makes sure we did not something wrong with this model,
by adding a const qualifier in all the paths taken from synack building
and transmit, for IPv4/IPv6 and TCP/dccp.

The only potential problem was the rewrite of ecn bits for connections
with DCTCP as congestion module, but this was a very minor one.
====================

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

4 files changed, 31 insertions, 18 deletions

diff --git a/net/ipv6/datagram.c b/net/ipv6/datagram.c
index 9aadd57..d70b023 100644
--- a/net/ipv6/datagram.c
+++ b/net/ipv6/datagram.c
@@ -263,7 +263,7 @@ void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
 
 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
 {
-	struct ipv6_pinfo *np = inet6_sk(sk);
+	const struct ipv6_pinfo *np = inet6_sk(sk);
 	struct sock_exterr_skb *serr;
 	struct ipv6hdr *iph;
 	struct sk_buff *skb;
diff --git a/net/ipv6/inet6_connection_sock.c b/net/ipv6/inet6_connection_sock.c
index 6927f3f..91b7d33 100644
--- a/net/ipv6/inet6_connection_sock.c
+++ b/net/ipv6/inet6_connection_sock.c
@@ -65,7 +65,7 @@ int inet6_csk_bind_conflict(const struct sock *sk,
 }
 EXPORT_SYMBOL_GPL(inet6_csk_bind_conflict);
 
-struct dst_entry *inet6_csk_route_req(struct sock *sk,
+struct dst_entry *inet6_csk_route_req(const struct sock *sk,
 				      struct flowi6 *fl6,
 				      const struct request_sock *req)
 {
diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index 291a07b..60c5653 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -150,14 +150,16 @@ int ip6_output(struct sock *sk, struct sk_buff *skb)
 }
 
 /*
- *	xmit an sk_buff (used by TCP, SCTP and DCCP)
+ * xmit an sk_buff (used by TCP, SCTP and DCCP)
+ * Note : socket lock is not held for SYNACK packets, but might be modified
+ * by calls to skb_set_owner_w() and ipv6_local_error(),
+ * which are using proper atomic operations or spinlocks.
  */
-
-int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
+int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
 	     struct ipv6_txoptions *opt, int tclass)
 {
 	struct net *net = sock_net(sk);
-	struct ipv6_pinfo *np = inet6_sk(sk);
+	const struct ipv6_pinfo *np = inet6_sk(sk);
 	struct in6_addr *first_hop = &fl6->daddr;
 	struct dst_entry *dst = skb_dst(skb);
 	struct ipv6hdr *hdr;
@@ -186,7 +188,10 @@ int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
 			}
 			consume_skb(skb);
 			skb = skb2;
-			skb_set_owner_w(skb, sk);
+			/* skb_set_owner_w() changes sk->sk_wmem_alloc atomically,
+			 * it is safe to call in our context (socket lock not held)
+			 */
+			skb_set_owner_w(skb, (struct sock *)sk);
 		}
 		if (opt->opt_flen)
 			ipv6_push_frag_opts(skb, opt, &proto);
@@ -224,13 +229,20 @@ int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
 	if ((skb->len <= mtu) || skb->ignore_df || skb_is_gso(skb)) {
 		IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
 			      IPSTATS_MIB_OUT, skb->len);
+		/* hooks should never assume socket lock is held.
+		 * we promote our socket to non const
+		 */
 		return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
-			       net, sk, skb, NULL, dst->dev,
+			       net, (struct sock *)sk, skb, NULL, dst->dev,
 			       dst_output_okfn);
 	}
 
 	skb->dev = dst->dev;
-	ipv6_local_error(sk, EMSGSIZE, fl6, mtu);
+	/* ipv6_local_error() does not require socket lock,
+	 * we promote our socket to non const
+	 */
+	ipv6_local_error((struct sock *)sk, EMSGSIZE, fl6, mtu);
+
 	IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
 	kfree_skb(skb);
 	return -EMSGSIZE;
@@ -883,7 +895,7 @@ out:
 	return dst;
 }
 
-static int ip6_dst_lookup_tail(struct net *net, struct sock *sk,
+static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
 			       struct dst_entry **dst, struct flowi6 *fl6)
 {
 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
@@ -1014,7 +1026,7 @@ EXPORT_SYMBOL_GPL(ip6_dst_lookup);
  *	It returns a valid dst pointer on success, or a pointer encoded
  *	error code.
  */
-struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
+struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
 				      const struct in6_addr *final_dst)
 {
 	struct dst_entry *dst = NULL;
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index a004e0b..16fb299 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -434,7 +434,7 @@ out:
 }
 
 
-static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
+static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst,
 			      struct flowi *fl,
 			      struct request_sock *req,
 			      u16 queue_mapping,
@@ -476,13 +476,13 @@ static void tcp_v6_reqsk_destructor(struct request_sock *req)
 }
 
 #ifdef CONFIG_TCP_MD5SIG
-static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
+static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
 						   const struct in6_addr *addr)
 {
 	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
 }
 
-static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
+static struct tcp_md5sig_key *tcp_v6_md5_lookup(const struct sock *sk,
 						const struct sock *addr_sk)
 {
 	return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
@@ -663,22 +663,23 @@ static bool tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
 }
 #endif
 
-static void tcp_v6_init_req(struct request_sock *req, struct sock *sk,
+static void tcp_v6_init_req(struct request_sock *req,
+			    const struct sock *sk_listener,
 			    struct sk_buff *skb)
 {
 	struct inet_request_sock *ireq = inet_rsk(req);
-	struct ipv6_pinfo *np = inet6_sk(sk);
+	const struct ipv6_pinfo *np = inet6_sk(sk_listener);
 
 	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
 	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
 
 	/* So that link locals have meaning */
-	if (!sk->sk_bound_dev_if &&
+	if (!sk_listener->sk_bound_dev_if &&
 	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
 		ireq->ir_iif = tcp_v6_iif(skb);
 
 	if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
-	    (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) ||
+	    (ipv6_opt_accepted(sk_listener, skb, &TCP_SKB_CB(skb)->header.h6) ||
 	     np->rxopt.bits.rxinfo ||
 	     np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
 	     np->rxopt.bits.rxohlim || np->repflow)) {