[INET]: Move tcp_port_rover to inet_hashinfo

Also expose all of the tcp_hashinfo members, i.e. killing those
tcp_ehash, etc macros, this will more clearly expose already generic
functions and some that need just a bit of work to become generic, as
we'll see in the upcoming changesets.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/inet_hashtables.h

@@ -117,6 +117,7 @@ struct inet_hashinfo {
 	wait_queue_head_t		lhash_wait;
 	spinlock_t			portalloc_lock;
 	kmem_cache_t			*bind_bucket_cachep;
+	int				port_rover;
 };
 
 static inline int inet_ehashfn(const __u32 laddr, const __u16 lport,

include/net/sock.h

@@ -136,7 +136,7 @@ struct sock_common {
   *	@sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
   *	@sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
   *	@sk_lingertime: %SO_LINGER l_linger setting
-  *	@sk_hashent: hash entry in several tables (e.g. tcp_ehash)
+  *	@sk_hashent: hash entry in several tables (e.g. inet_hashinfo.ehash)
   *	@sk_backlog: always used with the per-socket spinlock held
   *	@sk_callback_lock: used with the callbacks in the end of this struct
   *	@sk_error_queue: rarely used

include/net/tcp.h

@@ -41,19 +41,7 @@
 #endif
 #include <linux/seq_file.h>
 
-extern struct inet_hashinfo	tcp_hashinfo;
-#define tcp_ehash		(tcp_hashinfo.ehash)
-#define tcp_bhash		(tcp_hashinfo.bhash)
-#define tcp_ehash_size		(tcp_hashinfo.ehash_size)
-#define tcp_bhash_size		(tcp_hashinfo.bhash_size)
-#define tcp_listening_hash	(tcp_hashinfo.listening_hash)
-#define tcp_lhash_lock		(tcp_hashinfo.lhash_lock)
-#define tcp_lhash_users		(tcp_hashinfo.lhash_users)
-#define tcp_lhash_wait		(tcp_hashinfo.lhash_wait)
-#define tcp_portalloc_lock	(tcp_hashinfo.portalloc_lock)
-#define tcp_bucket_cachep	(tcp_hashinfo.bind_bucket_cachep)
-
-extern int tcp_port_rover;
+extern struct inet_hashinfo tcp_hashinfo;
 
 #if (BITS_PER_LONG == 64)
 #define TCP_ADDRCMP_ALIGN_BYTES 8
@@ -1463,21 +1451,21 @@ extern void tcp_listen_wlock(void);
 
 /* - We may sleep inside this lock.
  * - If sleeping is not required (or called from BH),
- *   use plain read_(un)lock(&tcp_lhash_lock).
+ *   use plain read_(un)lock(&inet_hashinfo.lhash_lock).
  */
 
 static inline void tcp_listen_lock(void)
 {
 	/* read_lock synchronizes to candidates to writers */
-	read_lock(&tcp_lhash_lock);
-	atomic_inc(&tcp_lhash_users);
-	read_unlock(&tcp_lhash_lock);
+	read_lock(&tcp_hashinfo.lhash_lock);
+	atomic_inc(&tcp_hashinfo.lhash_users);
+	read_unlock(&tcp_hashinfo.lhash_lock);
 }
 
 static inline void tcp_listen_unlock(void)
 {
-	if (atomic_dec_and_test(&tcp_lhash_users))
-		wake_up(&tcp_lhash_wait);
+	if (atomic_dec_and_test(&tcp_hashinfo.lhash_users))
+		wake_up(&tcp_hashinfo.lhash_wait);
 }
 
 static inline int keepalive_intvl_when(const struct tcp_sock *tp)

net/ipv4/tcp.c

@@ -2257,11 +2257,11 @@ void __init tcp_init(void)
 		__skb_cb_too_small_for_tcp(sizeof(struct tcp_skb_cb),
 					   sizeof(skb->cb));
 
-	tcp_bucket_cachep = kmem_cache_create("tcp_bind_bucket",
-					      sizeof(struct inet_bind_bucket),
-					      0, SLAB_HWCACHE_ALIGN,
-					      NULL, NULL);
-	if (!tcp_bucket_cachep)
+	tcp_hashinfo.bind_bucket_cachep =
+		kmem_cache_create("tcp_bind_bucket",
+				  sizeof(struct inet_bind_bucket), 0,
+				  SLAB_HWCACHE_ALIGN, NULL, NULL);
+	if (!tcp_hashinfo.bind_bucket_cachep)
 		panic("tcp_init: Cannot alloc tcp_bind_bucket cache.");
 
 	tcp_timewait_cachep = kmem_cache_create("tcp_tw_bucket",
@@ -2276,7 +2276,7 @@ void __init tcp_init(void)
 	 *
 	 * The methodology is similar to that of the buffer cache.
 	 */
-	tcp_ehash =
+	tcp_hashinfo.ehash =
 		alloc_large_system_hash("TCP established",
 					sizeof(struct inet_ehash_bucket),
 					thash_entries,
@@ -2284,37 +2284,37 @@ void __init tcp_init(void)
 						(25 - PAGE_SHIFT) :
 						(27 - PAGE_SHIFT),
 					HASH_HIGHMEM,
-					&tcp_ehash_size,
+					&tcp_hashinfo.ehash_size,
 					NULL,
 					0);
-	tcp_ehash_size = (1 << tcp_ehash_size) >> 1;
-	for (i = 0; i < (tcp_ehash_size << 1); i++) {
-		rwlock_init(&tcp_ehash[i].lock);
-		INIT_HLIST_HEAD(&tcp_ehash[i].chain);
+	tcp_hashinfo.ehash_size = (1 << tcp_hashinfo.ehash_size) >> 1;
+	for (i = 0; i < (tcp_hashinfo.ehash_size << 1); i++) {
+		rwlock_init(&tcp_hashinfo.ehash[i].lock);
+		INIT_HLIST_HEAD(&tcp_hashinfo.ehash[i].chain);
 	}
 
-	tcp_bhash =
+	tcp_hashinfo.bhash =
 		alloc_large_system_hash("TCP bind",
 					sizeof(struct inet_bind_hashbucket),
-					tcp_ehash_size,
+					tcp_hashinfo.ehash_size,
 					(num_physpages >= 128 * 1024) ?
 						(25 - PAGE_SHIFT) :
 						(27 - PAGE_SHIFT),
 					HASH_HIGHMEM,
-					&tcp_bhash_size,
+					&tcp_hashinfo.bhash_size,
 					NULL,
 					64 * 1024);
-	tcp_bhash_size = 1 << tcp_bhash_size;
-	for (i = 0; i < tcp_bhash_size; i++) {
-		spin_lock_init(&tcp_bhash[i].lock);
-		INIT_HLIST_HEAD(&tcp_bhash[i].chain);
+	tcp_hashinfo.bhash_size = 1 << tcp_hashinfo.bhash_size;
+	for (i = 0; i < tcp_hashinfo.bhash_size; i++) {
+		spin_lock_init(&tcp_hashinfo.bhash[i].lock);
+		INIT_HLIST_HEAD(&tcp_hashinfo.bhash[i].chain);
 	}
 
 	/* Try to be a bit smarter and adjust defaults depending
 	 * on available memory.
 	 */
 	for (order = 0; ((1 << order) << PAGE_SHIFT) <
-			(tcp_bhash_size * sizeof(struct inet_bind_hashbucket));
+			(tcp_hashinfo.bhash_size * sizeof(struct inet_bind_hashbucket));
 			order++)
 		;
 	if (order >= 4) {
@@ -2329,7 +2329,7 @@ void __init tcp_init(void)
 		sysctl_tcp_max_orphans >>= (3 - order);
 		sysctl_max_syn_backlog = 128;
 	}
-	tcp_port_rover = sysctl_local_port_range[0] - 1;
+	tcp_hashinfo.port_rover = sysctl_local_port_range[0] - 1;
 
 	sysctl_tcp_mem[0] =  768 << order;
 	sysctl_tcp_mem[1] = 1024 << order;
@@ -2344,7 +2344,7 @@ void __init tcp_init(void)
 
 	printk(KERN_INFO "TCP: Hash tables configured "
 	       "(established %d bind %d)\n",
-	       tcp_ehash_size << 1, tcp_bhash_size);
+	       tcp_hashinfo.ehash_size << 1, tcp_hashinfo.bhash_size);
 
 	tcp_register_congestion_control(&tcp_reno);
 }

net/ipv4/tcp_diag.c

@@ -595,7 +595,7 @@ static int tcpdiag_dump(struct sk_buff *skb, struct netlink_callback *cb)
 			struct hlist_node *node;
 
 			num = 0;
-			sk_for_each(sk, node, &tcp_listening_hash[i]) {
+			sk_for_each(sk, node, &tcp_hashinfo.listening_hash[i]) {
 				struct inet_sock *inet = inet_sk(sk);
 
 				if (num < s_num) {
@@ -645,8 +645,8 @@ static int tcpdiag_dump(struct sk_buff *skb, struct netlink_callback *cb)
 	if (!(r->tcpdiag_states&~(TCPF_LISTEN|TCPF_SYN_RECV)))
 		return skb->len;
 
-	for (i = s_i; i < tcp_ehash_size; i++) {
-		struct inet_ehash_bucket *head = &tcp_ehash[i];
+	for (i = s_i; i < tcp_hashinfo.ehash_size; i++) {
+		struct inet_ehash_bucket *head = &tcp_hashinfo.ehash[i];
 		struct sock *sk;
 		struct hlist_node *node;
 
@@ -678,7 +678,7 @@ static int tcpdiag_dump(struct sk_buff *skb, struct netlink_callback *cb)
 
 		if (r->tcpdiag_states&TCPF_TIME_WAIT) {
 			sk_for_each(sk, node,
-				    &tcp_ehash[i + tcp_ehash_size].chain) {
+				    &tcp_hashinfo.ehash[i + tcp_hashinfo.ehash_size].chain) {
 				struct inet_sock *inet = inet_sk(sk);
 
 				if (num < s_num)

net/ipv4/tcp_minisocks.c

@@ -60,12 +60,11 @@ int tcp_tw_count;
 /* Must be called with locally disabled BHs. */
 static void tcp_timewait_kill(struct tcp_tw_bucket *tw)
 {
-	struct inet_ehash_bucket *ehead;
 	struct inet_bind_hashbucket *bhead;
 	struct inet_bind_bucket *tb;
-
 	/* Unlink from established hashes. */
-	ehead = &tcp_ehash[tw->tw_hashent];
+	struct inet_ehash_bucket *ehead = &tcp_hashinfo.ehash[tw->tw_hashent];
+
 	write_lock(&ehead->lock);
 	if (hlist_unhashed(&tw->tw_node)) {
 		write_unlock(&ehead->lock);
@@ -76,12 +75,12 @@ static void tcp_timewait_kill(struct tcp_tw_bucket *tw)
 	write_unlock(&ehead->lock);
 
 	/* Disassociate with bind bucket. */
-	bhead = &tcp_bhash[inet_bhashfn(tw->tw_num, tcp_bhash_size)];
+	bhead = &tcp_hashinfo.bhash[inet_bhashfn(tw->tw_num, tcp_hashinfo.bhash_size)];
 	spin_lock(&bhead->lock);
 	tb = tw->tw_tb;
 	__hlist_del(&tw->tw_bind_node);
 	tw->tw_tb = NULL;
-	inet_bind_bucket_destroy(tcp_bucket_cachep, tb);
+	inet_bind_bucket_destroy(tcp_hashinfo.bind_bucket_cachep, tb);
 	spin_unlock(&bhead->lock);
 
 #ifdef SOCK_REFCNT_DEBUG
@@ -297,13 +296,13 @@ tcp_timewait_state_process(struct tcp_tw_bucket *tw, struct sk_buff *skb,
 static void __tcp_tw_hashdance(struct sock *sk, struct tcp_tw_bucket *tw)
 {
 	const struct inet_sock *inet = inet_sk(sk);
-	struct inet_ehash_bucket *ehead = &tcp_ehash[sk->sk_hashent];
+	struct inet_ehash_bucket *ehead = &tcp_hashinfo.ehash[sk->sk_hashent];
 	struct inet_bind_hashbucket *bhead;
 	/* Step 1: Put TW into bind hash. Original socket stays there too.
 	   Note, that any socket with inet->num != 0 MUST be bound in
 	   binding cache, even if it is closed.
 	 */
-	bhead = &tcp_bhash[inet_bhashfn(inet->num, tcp_bhash_size)];
+	bhead = &tcp_hashinfo.bhash[inet_bhashfn(inet->num, tcp_hashinfo.bhash_size)];
 	spin_lock(&bhead->lock);
 	tw->tw_tb = inet->bind_hash;
 	BUG_TRAP(inet->bind_hash);
@@ -317,7 +316,7 @@ static void __tcp_tw_hashdance(struct sock *sk, struct tcp_tw_bucket *tw)
 		sock_prot_dec_use(sk->sk_prot);
 
 	/* Step 3: Hash TW into TIMEWAIT half of established hash table. */
-	tw_add_node(tw, &(ehead + tcp_ehash_size)->chain);
+	tw_add_node(tw, &(ehead + tcp_hashinfo.ehash_size)->chain);
 	atomic_inc(&tw->tw_refcnt);
 
 	write_unlock(&ehead->lock);

net/ipv6/tcp_ipv6.c

@@ -84,7 +84,7 @@ static __inline__ int tcp_v6_hashfn(struct in6_addr *laddr, u16 lport,
 	hashent ^= (laddr->s6_addr32[3] ^ faddr->s6_addr32[3]);
 	hashent ^= hashent>>16;
 	hashent ^= hashent>>8;
-	return (hashent & (tcp_ehash_size - 1));
+	return (hashent & (tcp_hashinfo.ehash_size - 1));
 }
 
 static __inline__ int tcp_v6_sk_hashfn(struct sock *sk)
@@ -138,15 +138,15 @@ static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
 		int remaining = (high - low) + 1;
 		int rover;
 
-		spin_lock(&tcp_portalloc_lock);
-		if (tcp_port_rover < low)
+		spin_lock(&tcp_hashinfo.portalloc_lock);
+		if (tcp_hashinfo.port_rover < low)
 			rover = low;
 		else
-			rover = tcp_port_rover;
+			rover = tcp_hashinfo.port_rover;
 		do {	rover++;
 			if (rover > high)
 				rover = low;
-			head = &tcp_bhash[inet_bhashfn(rover, tcp_bhash_size)];
+			head = &tcp_hashinfo.bhash[inet_bhashfn(rover, tcp_hashinfo.bhash_size)];
 			spin_lock(&head->lock);
 			inet_bind_bucket_for_each(tb, node, &head->chain)
 				if (tb->port == rover)
@@ -155,8 +155,8 @@ static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
 		next:
 			spin_unlock(&head->lock);
 		} while (--remaining > 0);
-		tcp_port_rover = rover;
-		spin_unlock(&tcp_portalloc_lock);
+		tcp_hashinfo.port_rover = rover;
+		spin_unlock(&tcp_hashinfo.portalloc_lock);
 
 		/* Exhausted local port range during search?  It is not
 		 * possible for us to be holding one of the bind hash
@@ -171,7 +171,7 @@ static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
 		/* OK, here is the one we will use. */
 		snum = rover;
 	} else {
-		head = &tcp_bhash[inet_bhashfn(snum, tcp_bhash_size)];
+		head = &tcp_hashinfo.bhash[inet_bhashfn(snum, tcp_hashinfo.bhash_size)];
 		spin_lock(&head->lock);
 		inet_bind_bucket_for_each(tb, node, &head->chain)
 			if (tb->port == snum)
@@ -192,8 +192,11 @@ static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
 	}
 tb_not_found:
 	ret = 1;
-	if (!tb && (tb = inet_bind_bucket_create(tcp_bucket_cachep, head, snum)) == NULL)
-		goto fail_unlock;
+	if (tb == NULL) {
+	       	tb = inet_bind_bucket_create(tcp_hashinfo.bind_bucket_cachep, head, snum);
+		if (tb == NULL)
+			goto fail_unlock;
+	}
 	if (hlist_empty(&tb->owners)) {
 		if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
 			tb->fastreuse = 1;
@@ -224,13 +227,13 @@ static __inline__ void __tcp_v6_hash(struct sock *sk)
 	BUG_TRAP(sk_unhashed(sk));
 
 	if (sk->sk_state == TCP_LISTEN) {
-		list = &tcp_listening_hash[inet_sk_listen_hashfn(sk)];
-		lock = &tcp_lhash_lock;
+		list = &tcp_hashinfo.listening_hash[inet_sk_listen_hashfn(sk)];
+		lock = &tcp_hashinfo.lhash_lock;
 		tcp_listen_wlock();
 	} else {
 		sk->sk_hashent = tcp_v6_sk_hashfn(sk);
-		list = &tcp_ehash[sk->sk_hashent].chain;
-		lock = &tcp_ehash[sk->sk_hashent].lock;
+		list = &tcp_hashinfo.ehash[sk->sk_hashent].chain;
+		lock = &tcp_hashinfo.ehash[sk->sk_hashent].lock;
 		write_lock(lock);
 	}
 
@@ -263,8 +266,8 @@ static struct sock *tcp_v6_lookup_listener(struct in6_addr *daddr, unsigned shor
 	int score, hiscore;
 
 	hiscore=0;
-	read_lock(&tcp_lhash_lock);
-	sk_for_each(sk, node, &tcp_listening_hash[inet_lhashfn(hnum)]) {
+	read_lock(&tcp_hashinfo.lhash_lock);
+	sk_for_each(sk, node, &tcp_hashinfo.listening_hash[inet_lhashfn(hnum)]) {
 		if (inet_sk(sk)->num == hnum && sk->sk_family == PF_INET6) {
 			struct ipv6_pinfo *np = inet6_sk(sk);
 			
@@ -291,7 +294,7 @@ static struct sock *tcp_v6_lookup_listener(struct in6_addr *daddr, unsigned shor
 	}
 	if (result)
 		sock_hold(result);
-	read_unlock(&tcp_lhash_lock);
+	read_unlock(&tcp_hashinfo.lhash_lock);
 	return result;
 }
 
@@ -315,7 +318,7 @@ static inline struct sock *__tcp_v6_lookup_established(struct in6_addr *saddr, u
 	 * have wildcards anyways.
 	 */
 	hash = tcp_v6_hashfn(daddr, hnum, saddr, sport);
-	head = &tcp_ehash[hash];
+	head = &tcp_hashinfo.ehash[hash];
 	read_lock(&head->lock);
 	sk_for_each(sk, node, &head->chain) {
 		/* For IPV6 do the cheaper port and family tests first. */
@@ -323,7 +326,7 @@ static inline struct sock *__tcp_v6_lookup_established(struct in6_addr *saddr, u
 			goto hit; /* You sunk my battleship! */
 	}
 	/* Must check for a TIME_WAIT'er before going to listener hash. */
-	sk_for_each(sk, node, &(head + tcp_ehash_size)->chain) {
+	sk_for_each(sk, node, &(head + tcp_hashinfo.ehash_size)->chain) {
 		/* FIXME: acme: check this... */
 		struct tcp_tw_bucket *tw = (struct tcp_tw_bucket *)sk;
 
@@ -461,7 +464,7 @@ static int __tcp_v6_check_established(struct sock *sk, __u16 lport,
 	int dif = sk->sk_bound_dev_if;
 	u32 ports = TCP_COMBINED_PORTS(inet->dport, lport);
 	int hash = tcp_v6_hashfn(daddr, inet->num, saddr, inet->dport);
-	struct inet_ehash_bucket *head = &tcp_ehash[hash];
+	struct inet_ehash_bucket *head = &tcp_hashinfo.ehash[hash];
 	struct sock *sk2;
 	struct hlist_node *node;
 	struct tcp_tw_bucket *tw;
@@ -469,7 +472,7 @@ static int __tcp_v6_check_established(struct sock *sk, __u16 lport,
 	write_lock(&head->lock);
 
 	/* Check TIME-WAIT sockets first. */
-	sk_for_each(sk2, node, &(head + tcp_ehash_size)->chain) {
+	sk_for_each(sk2, node, &(head + tcp_hashinfo.ehash_size)->chain) {
 		tw = (struct tcp_tw_bucket*)sk2;
 
 		if(*((__u32 *)&(tw->tw_dport))	== ports	&&
@@ -558,7 +561,7 @@ static int tcp_v6_hash_connect(struct sock *sk)
  		local_bh_disable();
 		for (i = 1; i <= range; i++) {
 			port = low + (i + offset) % range;
- 			head = &tcp_bhash[inet_bhashfn(port, tcp_bhash_size)];
+ 			head = &tcp_hashinfo.bhash[inet_bhashfn(port, tcp_hashinfo.bhash_size)];
  			spin_lock(&head->lock);
 
  			/* Does not bother with rcv_saddr checks,
@@ -578,7 +581,7 @@ static int tcp_v6_hash_connect(struct sock *sk)
  				}
  			}
 
- 			tb = inet_bind_bucket_create(tcp_bucket_cachep, head, port);
+ 			tb = inet_bind_bucket_create(tcp_hashinfo.bind_bucket_cachep, head, port);
  			if (!tb) {
  				spin_unlock(&head->lock);
  				break;
@@ -613,7 +616,7 @@ static int tcp_v6_hash_connect(struct sock *sk)
 		goto out;
  	}
 
- 	head = &tcp_bhash[inet_bhashfn(snum, tcp_bhash_size)];
+ 	head = &tcp_hashinfo.bhash[inet_bhashfn(snum, tcp_hashinfo.bhash_size)];
  	tb   = inet_sk(sk)->bind_hash;
 	spin_lock_bh(&head->lock);