[IPV4/IPV6]: inetsw using RCU.

9 files changed, 67 insertions, 47 deletions

diff --git a/include/net/protocol.h b/include/net/protocol.h
index 6f0e4234a442..f67327486a9c 100644
--- a/include/net/protocol.h
+++ b/include/net/protocol.h
@@ -80,11 +80,9 @@ struct inet_protosw {
 
 extern struct inet_protocol *inet_protocol_base;
 extern struct inet_protocol *inet_protos[MAX_INET_PROTOS];
-extern struct list_head inetsw[SOCK_MAX];
 
 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
 extern struct inet6_protocol *inet6_protos[MAX_INET_PROTOS];
-extern struct list_head inetsw6[SOCK_MAX];
 #endif
 
 extern int	inet_add_protocol(struct inet_protocol *prot, unsigned char num);
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 69cfc1691f4a..2ccf51403636 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -94,7 +94,6 @@
 #include <linux/inet.h>
 #include <linux/igmp.h>
 #include <linux/netdevice.h>
-#include <linux/brlock.h>
 #include <net/ip.h>
 #include <net/protocol.h>
 #include <net/arp.h>
@@ -129,7 +128,8 @@ static kmem_cache_t *raw4_sk_cachep;
 /* The inetsw table contains everything that inet_create needs to
  * build a new socket.
  */
-struct list_head inetsw[SOCK_MAX];
+static struct list_head inetsw[SOCK_MAX];
+static spinlock_t inetsw_lock = SPIN_LOCK_UNLOCKED;
 
 /* New destruction routine */
 
@@ -337,8 +337,8 @@ static int inet_create(struct socket *sock, int protocol)
 
 	/* Look for the requested type/protocol pair. */
 	answer = NULL;
-	br_read_lock_bh(BR_NETPROTO_LOCK);
-	list_for_each(p, &inetsw[sock->type]) {
+	rcu_read_lock();
+	list_for_each_rcu(p, &inetsw[sock->type]) {
 		answer = list_entry(p, struct inet_protosw, list);
 
 		/* Check the non-wild match. */
@@ -356,7 +356,6 @@ static int inet_create(struct socket *sock, int protocol)
 		}
 		answer = NULL;
 	}
-	br_read_unlock_bh(BR_NETPROTO_LOCK);
 
 	err = -ESOCKTNOSUPPORT;
 	if (!answer)
@@ -373,6 +372,7 @@ static int inet_create(struct socket *sock, int protocol)
 	sk->no_check = answer->no_check;
 	if (INET_PROTOSW_REUSE & answer->flags)
 		sk->reuse = 1;
+	rcu_read_unlock();
 
 	inet = inet_sk(sk);
 
@@ -427,6 +427,7 @@ static int inet_create(struct socket *sock, int protocol)
 out:
 	return err;
 out_sk_free:
+	rcu_read_unlock();
 	sk_free(sk);
 	goto out;
 }
@@ -979,7 +980,7 @@ void inet_register_protosw(struct inet_protosw *p)
 	int protocol = p->protocol;
 	struct list_head *last_perm;
 
-	br_write_lock_bh(BR_NETPROTO_LOCK);
+	spin_lock_bh(&inetsw_lock);
 
 	if (p->type > SOCK_MAX)
 		goto out_illegal;
@@ -1008,9 +1009,12 @@ void inet_register_protosw(struct inet_protosw *p)
 	 * non-permanent entry.  This means that when we remove this entry, the 
 	 * system automatically returns to the old behavior.
 	 */
-	list_add(&p->list, last_perm);
+	list_add_rcu(&p->list, last_perm);
 out:
-	br_write_unlock_bh(BR_NETPROTO_LOCK);
+	spin_unlock_bh(&inetsw_lock);
+
+	synchronize_kernel();
+
 	return;
 
 out_permanent:
@@ -1032,9 +1036,11 @@ void inet_unregister_protosw(struct inet_protosw *p)
 		       "Attempt to unregister permanent protocol %d.\n",
 		       p->protocol);
 	} else {
-		br_write_lock_bh(BR_NETPROTO_LOCK);
-		list_del(&p->list);
-		br_write_unlock_bh(BR_NETPROTO_LOCK);
+		spin_lock_bh(&inetsw_lock);
+		list_del_rcu(&p->list);
+		spin_unlock_bh(&inetsw_lock);
+
+		synchronize_kernel();
 	}
 }
 
diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c
index 04e3361034d4..4589ca2d16b0 100644
--- a/net/ipv4/icmp.c
+++ b/net/ipv4/icmp.c
@@ -695,15 +695,12 @@ static void icmp_unreach(struct sk_buff *skb)
 	}
 	read_unlock(&raw_v4_lock);
 
-	/*
-	 *	This can't change while we are doing it.
-	 *	Callers have obtained BR_NETPROTO_LOCK so
-	 *	we are OK.
-	 */
-
+	rcu_read_lock();
 	ipprot = inet_protos[hash];
+	smp_read_barrier_depends();
 	if (ipprot && ipprot->err_handler)
 		ipprot->err_handler(skb, info);
+	rcu_read_unlock();
 
 out:
 	return;
diff --git a/net/ipv4/ip_input.c b/net/ipv4/ip_input.c
index 6131e3babfca..fe561aa4a767 100644
--- a/net/ipv4/ip_input.c
+++ b/net/ipv4/ip_input.c
@@ -215,6 +215,7 @@ static inline int ip_local_deliver_finish(struct sk_buff *skb)
         /* Point into the IP datagram, just past the header. */
         skb->h.raw = skb->data;
 
+	rcu_read_lock();
 	{
 		/* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */
 		int protocol = skb->nh.iph->protocol;
@@ -235,10 +236,11 @@ static inline int ip_local_deliver_finish(struct sk_buff *skb)
 		if ((ipprot = inet_protos[hash]) != NULL) {
 			int ret;
 
+			smp_read_barrier_depends();
 			if (!ipprot->no_policy &&
 			    !xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 				kfree_skb(skb);
-				return 0;
+				goto out;
 			}
 			ret = ipprot->handler(skb);
 			if (ret < 0) {
@@ -258,6 +260,8 @@ static inline int ip_local_deliver_finish(struct sk_buff *skb)
 			kfree_skb(skb);
 		}
 	}
+ out:
+	rcu_read_unlock();
 
 	return 0;
 }
diff --git a/net/ipv4/protocol.c b/net/ipv4/protocol.c
index 302e10a3ccfb..9735e45d5b86 100644
--- a/net/ipv4/protocol.c
+++ b/net/ipv4/protocol.c
@@ -37,7 +37,6 @@
 #include <linux/inet.h>
 #include <linux/netdevice.h>
 #include <linux/timer.h>
-#include <linux/brlock.h>
 #include <net/ip.h>
 #include <net/protocol.h>
 #include <net/tcp.h>
@@ -49,6 +48,7 @@
 #include <linux/igmp.h>
 
 struct inet_protocol *inet_protos[MAX_INET_PROTOS];
+static spinlock_t inet_proto_lock = SPIN_LOCK_UNLOCKED;
 
 /*
  *	Add a protocol handler to the hash tables
@@ -60,16 +60,14 @@ int inet_add_protocol(struct inet_protocol *prot, unsigned char protocol)
 
 	hash = protocol & (MAX_INET_PROTOS - 1);
 
-	br_write_lock_bh(BR_NETPROTO_LOCK);
-
+	spin_lock_bh(&inet_proto_lock);
 	if (inet_protos[hash]) {
 		ret = -1;
 	} else {
 		inet_protos[hash] = prot;
 		ret = 0;
 	}
-
-	br_write_unlock_bh(BR_NETPROTO_LOCK);
+	spin_unlock_bh(&inet_proto_lock);
 
 	return ret;
 }
@@ -84,16 +82,15 @@ int inet_del_protocol(struct inet_protocol *prot, unsigned char protocol)
 
 	hash = protocol & (MAX_INET_PROTOS - 1);
 
-	br_write_lock_bh(BR_NETPROTO_LOCK);
-
+	spin_lock_bh(&inet_proto_lock);
 	if (inet_protos[hash] == prot) {
 		inet_protos[hash] = NULL;
 		ret = 0;
 	} else {
 		ret = -1;
 	}
+	spin_unlock_bh(&inet_proto_lock);
 
-	br_write_unlock_bh(BR_NETPROTO_LOCK);
 
 	return ret;
 }
diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c
index 8ae2dfaba7a5..aa799a503823 100644
--- a/net/ipv6/af_inet6.c
+++ b/net/ipv6/af_inet6.c
@@ -45,7 +45,6 @@
 #include <linux/inet.h>
 #include <linux/netdevice.h>
 #include <linux/icmpv6.h>
-#include <linux/brlock.h>
 #include <linux/smp_lock.h>
 
 #include <net/ip.h>
@@ -102,7 +101,8 @@ kmem_cache_t *raw6_sk_cachep;
 /* The inetsw table contains everything that inet_create needs to
  * build a new socket.
  */
-struct list_head inetsw6[SOCK_MAX];
+static struct list_head inetsw6[SOCK_MAX];
+static spinlock_t inetsw6_lock = SPIN_LOCK_UNLOCKED;
 
 static void inet6_sock_destruct(struct sock *sk)
 {
@@ -162,8 +162,8 @@ static int inet6_create(struct socket *sock, int protocol)
 
 	/* Look for the requested type/protocol pair. */
 	answer = NULL;
-	br_read_lock_bh(BR_NETPROTO_LOCK);
-	list_for_each(p, &inetsw6[sock->type]) {
+	rcu_read_lock();
+	list_for_each_rcu(p, &inetsw6[sock->type]) {
 		answer = list_entry(p, struct inet_protosw, list);
 
 		/* Check the non-wild match. */
@@ -181,7 +181,6 @@ static int inet6_create(struct socket *sock, int protocol)
 		}
 		answer = NULL;
 	}
-	br_read_unlock_bh(BR_NETPROTO_LOCK);
 
 	if (!answer)
 		goto free_and_badtype;
@@ -198,6 +197,7 @@ static int inet6_create(struct socket *sock, int protocol)
 	sk->no_check = answer->no_check;
 	if (INET_PROTOSW_REUSE & answer->flags)
 		sk->reuse = 1;
+	rcu_read_unlock();
 
 	inet = inet_sk(sk);
 
@@ -260,12 +260,15 @@ static int inet6_create(struct socket *sock, int protocol)
 	return 0;
 
 free_and_badtype:
+	rcu_read_unlock();
 	sk_free(sk);
 	return -ESOCKTNOSUPPORT;
 free_and_badperm:
+	rcu_read_unlock();
 	sk_free(sk);
 	return -EPERM;
 free_and_noproto:
+	rcu_read_unlock();
 	sk_free(sk);
 	return -EPROTONOSUPPORT;
 do_oom:
@@ -574,7 +577,7 @@ inet6_register_protosw(struct inet_protosw *p)
 	int protocol = p->protocol;
 	struct list_head *last_perm;
 
-	br_write_lock_bh(BR_NETPROTO_LOCK);
+	spin_lock_bh(&inetsw6_lock);
 
 	if (p->type > SOCK_MAX)
 		goto out_illegal;
@@ -603,9 +606,9 @@ inet6_register_protosw(struct inet_protosw *p)
 	 * non-permanent entry.  This means that when we remove this entry, the 
 	 * system automatically returns to the old behavior.
 	 */
-	list_add(&p->list, last_perm);
+	list_add_rcu(&p->list, last_perm);
 out:
-	br_write_unlock_bh(BR_NETPROTO_LOCK);
+	spin_unlock_bh(&inetsw6_lock);
 	return;
 
 out_permanent:
@@ -623,7 +626,17 @@ out_illegal:
 void
 inet6_unregister_protosw(struct inet_protosw *p)
 {
-	inet_unregister_protosw(p);
+	if (INET_PROTOSW_PERMANENT & p->flags) {
+		printk(KERN_ERR
+		       "Attempt to unregister permanent protocol %d.\n",
+		       p->protocol);
+	} else {
+		spin_lock_bh(&inetsw6_lock);
+		list_del_rcu(&p->list);
+		spin_unlock_bh(&inetsw6_lock);
+
+		synchronize_kernel();
+	}
 }
 
 int
diff --git a/net/ipv6/icmp.c b/net/ipv6/icmp.c
index 9f420ea0f94f..8a415c312de0 100644
--- a/net/ipv6/icmp.c
+++ b/net/ipv6/icmp.c
@@ -456,9 +456,12 @@ static void icmpv6_notify(struct sk_buff *skb, int type, int code, u32 info)
 
 	hash = nexthdr & (MAX_INET_PROTOS - 1);
 
+	rcu_read_lock();
 	ipprot = inet6_protos[hash];
+	smp_read_barrier_depends();
 	if (ipprot && ipprot->err_handler)
 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
+	rcu_read_unlock();
 
 	read_lock(&raw_v6_lock);
 	if ((sk = raw_v6_htable[hash]) != NULL) {
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index c737d2f3d486..16feb45246a9 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -152,6 +152,7 @@ static inline int ip6_input_finish(struct sk_buff *skb)
 		skb->h.raw += (skb->h.raw[1]+1)<<3;
 	}
 
+	rcu_read_lock();
 resubmit:
 	if (!pskb_pull(skb, skb->h.raw - skb->data))
 		goto discard;
@@ -165,6 +166,7 @@ resubmit:
 	if ((ipprot = inet6_protos[hash]) != NULL) {
 		int ret;
 		
+		smp_read_barrier_depends();
 		if (ipprot->flags & INET6_PROTO_FINAL) {
 			if (!cksum_sub && skb->ip_summed == CHECKSUM_HW) {
 				skb->csum = csum_sub(skb->csum,
@@ -173,10 +175,8 @@ resubmit:
 			}
 		}
 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
-		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
-			kfree_skb(skb);
-			return 0;
-		}
+		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 
+			goto discard;
 		
 		ret = ipprot->handler(&skb, &nhoff);
 		if (ret > 0)
@@ -194,10 +194,11 @@ resubmit:
 			kfree_skb(skb);
 		}
 	}
-
+	rcu_read_unlock();
 	return 0;
 
 discard:
+	rcu_read_unlock();
 	kfree_skb(skb);
 	return 0;
 }
diff --git a/net/ipv6/protocol.c b/net/ipv6/protocol.c
index d070acfd8e1b..d1dc7ac40ae2 100644
--- a/net/ipv6/protocol.c
+++ b/net/ipv6/protocol.c
@@ -32,7 +32,6 @@
 #include <linux/in6.h>
 #include <linux/netdevice.h>
 #include <linux/if_arp.h>
-#include <linux/brlock.h>
 
 #include <net/sock.h>
 #include <net/snmp.h>
@@ -41,12 +40,14 @@
 #include <net/protocol.h>
 
 struct inet6_protocol *inet6_protos[MAX_INET_PROTOS];
+static spinlock_t inet6_proto_lock = SPIN_LOCK_UNLOCKED;
+
 
 int inet6_add_protocol(struct inet6_protocol *prot, unsigned char protocol)
 {
 	int ret, hash = protocol & (MAX_INET_PROTOS - 1);
 
-	br_write_lock_bh(BR_NETPROTO_LOCK);
+	spin_lock_bh(&inet6_proto_lock);
 
 	if (inet6_protos[hash]) {
 		ret = -1;
@@ -55,7 +56,7 @@ int inet6_add_protocol(struct inet6_protocol *prot, unsigned char protocol)
 		ret = 0;
 	}
 
-	br_write_unlock_bh(BR_NETPROTO_LOCK);
+	spin_unlock_bh(&inet6_proto_lock);
 
 	return ret;
 }
@@ -68,7 +69,7 @@ int inet6_del_protocol(struct inet6_protocol *prot, unsigned char protocol)
 {
 	int ret, hash = protocol & (MAX_INET_PROTOS - 1);
 
-	br_write_lock_bh(BR_NETPROTO_LOCK);
+	spin_lock_bh(&inet6_proto_lock);
 
 	if (inet6_protos[hash] != prot) {
 		ret = -1;
@@ -77,7 +78,7 @@ int inet6_del_protocol(struct inet6_protocol *prot, unsigned char protocol)
 		ret = 0;
 	}
 
-	br_write_unlock_bh(BR_NETPROTO_LOCK);
+	spin_unlock_bh(&inet6_proto_lock);
 
 	return ret;
 }