[INET]: Generalise tcp_tw_bucket, aka TIME_WAIT sockets

This paves the way to generalise the rest of the sock ID lookup
routines and saves some bytes in TCPv4 TIME_WAIT sockets on distro
kernels (where IPv6 is always built as a module):

[root@qemu ~]# grep tw_sock /proc/slabinfo
tw_sock_TCPv6  0  0  128  31  1
tw_sock_TCP    0  0   96  41  1
[root@qemu ~]#

Now if a protocol wants to use the TIME_WAIT generic infrastructure it
only has to set the sk_prot->twsk_obj_size field with the size of its
inet_timewait_sock derived sock and proto_register will create
sk_prot->twsk_slab, for now its only for INET sockets, but we can
introduce timewait_sock later if some non INET transport protocolo
wants to use this stuff.

Next changesets will take advantage of this new infrastructure to
generalise even more TCP code.

[acme@toy net-2.6.14]$ grep built-in /tmp/before.size /tmp/after.size
/tmp/before.size: 188646   11764    5068  205478   322a6 net/ipv4/built-in.o
/tmp/after.size:  188144   11764    5068  204976   320b0 net/ipv4/built-in.o
[acme@toy net-2.6.14]$

Tested with both IPv4 & IPv6 (::1 (localhost) & ::ffff:172.20.0.1
(qemu host)).
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/ipv6.h

@@ -308,6 +308,41 @@ static inline void inet_sk_copy_descendant(struct sock *sk_to,
 
 #define __ipv6_only_sock(sk)	(inet6_sk(sk)->ipv6only)
 #define ipv6_only_sock(sk)	((sk)->sk_family == PF_INET6 && __ipv6_only_sock(sk))
+
+#include <linux/tcp.h>
+
+struct tcp6_timewait_sock {
+	struct tcp_timewait_sock tw_v6_sk;
+	struct in6_addr		 tw_v6_daddr;
+	struct in6_addr		 tw_v6_rcv_saddr;
+};
+
+static inline struct tcp6_timewait_sock *tcp6_twsk(const struct sock *sk)
+{
+	return (struct tcp6_timewait_sock *)sk;
+}
+
+static inline struct in6_addr *__tcp_v6_rcv_saddr(const struct sock *sk)
+{
+	return likely(sk->sk_state != TCP_TIME_WAIT) ?
+		&inet6_sk(sk)->rcv_saddr : &tcp6_twsk(sk)->tw_v6_rcv_saddr;
+}
+
+static inline struct in6_addr *tcp_v6_rcv_saddr(const struct sock *sk)
+{
+	return sk->sk_family == AF_INET6 ? __tcp_v6_rcv_saddr(sk) : NULL;
+}
+
+static inline int tcp_twsk_ipv6only(const struct sock *sk)
+{
+	return inet_twsk(sk)->tw_ipv6only;
+}
+
+static inline int tcp_v6_ipv6only(const struct sock *sk)
+{
+	return likely(sk->sk_state != TCP_TIME_WAIT) ?
+		ipv6_only_sock(sk) : tcp_twsk_ipv6only(sk);
+}
 #else
 #define __ipv6_only_sock(sk)	0
 #define ipv6_only_sock(sk)	0
@@ -322,8 +357,19 @@ static inline struct raw6_sock *raw6_sk(const struct sock *sk)
 	return NULL;
 }
 
-#endif
+#define __tcp_v6_rcv_saddr(__sk)	NULL
+#define tcp_v6_rcv_saddr(__sk)		NULL
+#define tcp_twsk_ipv6only(__sk)		0
+#define tcp_v6_ipv6only(__sk)		0
+#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
 
-#endif
+#define INET6_MATCH(__sk, __saddr, __daddr, __ports, __dif)	   \
+	(((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))  	&& \
+	 ((__sk)->sk_family		== AF_INET6)		&& \
+	 ipv6_addr_equal(&inet6_sk(__sk)->daddr, (__saddr))	&& \
+	 ipv6_addr_equal(&inet6_sk(__sk)->rcv_saddr, (__daddr))	&& \
+	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
 
-#endif
+#endif /* __KERNEL__ */
+
+#endif /* _IPV6_H */

include/linux/tcp.h

@@ -179,6 +179,7 @@ struct tcp_info
 #include <linux/skbuff.h>
 #include <linux/ip.h>
 #include <net/sock.h>
+#include <net/inet_timewait_sock.h>
 
 /* This defines a selective acknowledgement block. */
 struct tcp_sack_block {
@@ -387,6 +388,20 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk)
 	return (struct tcp_sock *)sk;
 }
 
+struct tcp_timewait_sock {
+	struct inet_timewait_sock tw_sk;
+	__u32			  tw_rcv_nxt;
+	__u32			  tw_snd_nxt;
+	__u32			  tw_rcv_wnd;
+	__u32			  tw_ts_recent;
+	long			  tw_ts_recent_stamp;
+};
+
+static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
+{
+	return (struct tcp_timewait_sock *)sk;
+}
+
 static inline void *tcp_ca(const struct tcp_sock *tp)
 {
 	return (void *) tp->ca_priv;

include/net/inet_hashtables.h

@@ -14,6 +14,8 @@
 #ifndef _INET_HASHTABLES_H
 #define _INET_HASHTABLES_H
 
+#include <linux/config.h>
+
 #include <linux/interrupt.h>
 #include <linux/ip.h>
 #include <linux/ipv6.h>
@@ -310,4 +312,43 @@ sherry_cache:
 	read_unlock(&hashinfo->lhash_lock);
 	return sk;
 }
+
+/* Socket demux engine toys. */
+#ifdef __BIG_ENDIAN
+#define INET_COMBINED_PORTS(__sport, __dport) \
+	(((__u32)(__sport) << 16) | (__u32)(__dport))
+#else /* __LITTLE_ENDIAN */
+#define INET_COMBINED_PORTS(__sport, __dport) \
+	(((__u32)(__dport) << 16) | (__u32)(__sport))
+#endif
+
+#if (BITS_PER_LONG == 64)
+#ifdef __BIG_ENDIAN
+#define INET_ADDR_COOKIE(__name, __saddr, __daddr) \
+	const __u64 __name = (((__u64)(__saddr)) << 32) | ((__u64)(__daddr));
+#else /* __LITTLE_ENDIAN */
+#define INET_ADDR_COOKIE(__name, __saddr, __daddr) \
+	const __u64 __name = (((__u64)(__daddr)) << 32) | ((__u64)(__saddr));
+#endif /* __BIG_ENDIAN */
+#define INET_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
+	(((*((__u64 *)&(inet_sk(__sk)->daddr))) == (__cookie))	&&	\
+	 ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))	&&	\
+	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
+#define INET_TW_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
+	(((*((__u64 *)&(inet_twsk(__sk)->tw_daddr))) == (__cookie)) &&	\
+	 ((*((__u32 *)&(inet_twsk(__sk)->tw_dport))) == (__ports)) &&	\
+	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
+#else /* 32-bit arch */
+#define INET_ADDR_COOKIE(__name, __saddr, __daddr)
+#define INET_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)	\
+	((inet_sk(__sk)->daddr		== (__saddr))		&&	\
+	 (inet_sk(__sk)->rcv_saddr	== (__daddr))		&&	\
+	 ((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports))	&&	\
+	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
+#define INET_TW_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)	\
+	((inet_twsk(__sk)->tw_daddr	== (__saddr))		&&	\
+	 (inet_twsk(__sk)->tw_rcv_saddr	== (__daddr))		&&	\
+	 ((*((__u32 *)&(inet_twsk(__sk)->tw_dport))) == (__ports)) &&	\
+	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
+#endif /* 64-bit arch */
 #endif /* _INET_HASHTABLES_H */

include/net/inet_timewait_sock.h

+/*
+ * INET		An implementation of the TCP/IP protocol suite for the LINUX
+ *		operating system.  INET is implemented using the  BSD Socket
+ *		interface as the means of communication with the user level.
+ *
+ *		Definitions for a generic INET TIMEWAIT sock
+ *
+ *		From code originally in net/tcp.h
+ *
+ *		This program is free software; you can redistribute it and/or
+ *		modify it under the terms of the GNU General Public License
+ *		as published by the Free Software Foundation; either version
+ *		2 of the License, or (at your option) any later version.
+ */
+#ifndef _INET_TIMEWAIT_SOCK_
+#define _INET_TIMEWAIT_SOCK_
+
+#include <linux/config.h>
+
+#include <linux/list.h>
+#include <linux/types.h>
+
+#include <net/sock.h>
+#include <net/tcp_states.h>
+
+#include <asm/atomic.h>
+
+#if (BITS_PER_LONG == 64)
+#define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 8
+#else
+#define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 4
+#endif
+
+struct inet_bind_bucket;
+
+/*
+ * This is a TIME_WAIT sock. It works around the memory consumption
+ * problems of sockets in such a state on heavily loaded servers, but
+ * without violating the protocol specification.
+ */
+struct inet_timewait_sock {
+	/*
+	 * Now struct sock also uses sock_common, so please just
+	 * don't add nothing before this first member (__tw_common) --acme
+	 */
+	struct sock_common	__tw_common;
+#define tw_family		__tw_common.skc_family
+#define tw_state		__tw_common.skc_state
+#define tw_reuse		__tw_common.skc_reuse
+#define tw_bound_dev_if		__tw_common.skc_bound_dev_if
+#define tw_node			__tw_common.skc_node
+#define tw_bind_node		__tw_common.skc_bind_node
+#define tw_refcnt		__tw_common.skc_refcnt
+#define tw_prot			__tw_common.skc_prot
+	volatile unsigned char	tw_substate;
+	/* 3 bits hole, try to pack */
+	unsigned char		tw_rcv_wscale;
+	/* Socket demultiplex comparisons on incoming packets. */
+	/* these five are in inet_sock */
+	__u16			tw_sport;
+	__u32			tw_daddr __attribute__((aligned(INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES)));
+	__u32			tw_rcv_saddr;
+	__u16			tw_dport;
+	__u16			tw_num;
+	/* And these are ours. */
+	__u8			tw_ipv6only:1;
+	/* 31 bits hole, try to pack */
+	int			tw_hashent;
+	int			tw_timeout;
+	unsigned long		tw_ttd;
+	struct inet_bind_bucket	*tw_tb;
+	struct hlist_node	tw_death_node;
+};
+
+static inline void inet_twsk_add_node(struct inet_timewait_sock *tw,
+				      struct hlist_head *list)
+{
+	hlist_add_head(&tw->tw_node, list);
+}
+
+static inline void inet_twsk_add_bind_node(struct inet_timewait_sock *tw,
+					   struct hlist_head *list)
+{
+	hlist_add_head(&tw->tw_bind_node, list);
+}
+
+static inline int inet_twsk_dead_hashed(const struct inet_timewait_sock *tw)
+{
+	return tw->tw_death_node.pprev != NULL;
+}
+
+static inline void inet_twsk_dead_node_init(struct inet_timewait_sock *tw)
+{
+	tw->tw_death_node.pprev = NULL;
+}
+
+static inline void __inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
+{
+	__hlist_del(&tw->tw_death_node);
+	inet_twsk_dead_node_init(tw);
+}
+
+static inline int inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
+{
+	if (inet_twsk_dead_hashed(tw)) {
+		__inet_twsk_del_dead_node(tw);
+		return 1;
+	}
+	return 0;
+}
+
+#define inet_twsk_for_each(tw, node, head) \
+	hlist_for_each_entry(tw, node, head, tw_node)
+
+#define inet_twsk_for_each_inmate(tw, node, jail) \
+	hlist_for_each_entry(tw, node, jail, tw_death_node)
+
+#define inet_twsk_for_each_inmate_safe(tw, node, safe, jail) \
+	hlist_for_each_entry_safe(tw, node, safe, jail, tw_death_node)
+
+static inline struct inet_timewait_sock *inet_twsk(const struct sock *sk)
+{
+	return (struct inet_timewait_sock *)sk;
+}
+
+static inline u32 inet_rcv_saddr(const struct sock *sk)
+{
+	return likely(sk->sk_state != TCP_TIME_WAIT) ?
+		inet_sk(sk)->rcv_saddr : inet_twsk(sk)->tw_rcv_saddr;
+}
+
+static inline void inet_twsk_put(struct inet_timewait_sock *tw)
+{
+	if (atomic_dec_and_test(&tw->tw_refcnt)) {
+#ifdef SOCK_REFCNT_DEBUG
+		printk(KERN_DEBUG "%s timewait_sock %p released\n",
+		       tw->tw_prot->name, tw);
+#endif
+		kmem_cache_free(tw->tw_prot->twsk_slab, tw);
+	}
+}
+#endif	/* _INET_TIMEWAIT_SOCK_ */

include/net/sock.h

@@ -88,6 +88,7 @@ do {	spin_lock_init(&((__sk)->sk_lock.slock)); \
 } while(0)
 
 struct sock;
+struct proto;
 
 /**
  *	struct sock_common - minimal network layer representation of sockets
@@ -98,10 +99,11 @@ struct sock;
  *	@skc_node: main hash linkage for various protocol lookup tables
  *	@skc_bind_node: bind hash linkage for various protocol lookup tables
  *	@skc_refcnt: reference count
+ *	@skc_prot: protocol handlers inside a network family
  *
  *	This is the minimal network layer representation of sockets, the header
- *	for struct sock and struct tcp_tw_bucket.
-  */
+ *	for struct sock and struct inet_timewait_sock.
+ */
 struct sock_common {
 	unsigned short		skc_family;
 	volatile unsigned char	skc_state;
@@ -110,11 +112,12 @@ struct sock_common {
 	struct hlist_node	skc_node;
 	struct hlist_node	skc_bind_node;
 	atomic_t		skc_refcnt;
+	struct proto		*skc_prot;
 };
 
 /**
   *	struct sock - network layer representation of sockets
-  *	@__sk_common: shared layout with tcp_tw_bucket
+  *	@__sk_common: shared layout with inet_timewait_sock
   *	@sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
   *	@sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
   *	@sk_lock:	synchronizer
@@ -140,7 +143,6 @@ struct sock_common {
   *	@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
-  *	@sk_prot: protocol handlers inside a network family
   *	@sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, IPV6_ADDRFORM for instance)
   *	@sk_err: last error
   *	@sk_err_soft: errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
@@ -173,7 +175,7 @@ struct sock_common {
  */
 struct sock {
 	/*
-	 * Now struct tcp_tw_bucket also uses sock_common, so please just
+	 * Now struct inet_timewait_sock also uses sock_common, so please just
 	 * don't add nothing before this first member (__sk_common) --acme
 	 */
 	struct sock_common	__sk_common;
@@ -184,6 +186,7 @@ struct sock {
 #define sk_node			__sk_common.skc_node
 #define sk_bind_node		__sk_common.skc_bind_node
 #define sk_refcnt		__sk_common.skc_refcnt
+#define sk_prot			__sk_common.skc_prot
 	unsigned char		sk_shutdown : 2,
 				sk_no_check : 2,
 				sk_userlocks : 4;
@@ -218,7 +221,6 @@ struct sock {
 		struct sk_buff *tail;
 	} sk_backlog;
 	struct sk_buff_head	sk_error_queue;
-	struct proto		*sk_prot;
 	struct proto		*sk_prot_creator;
 	rwlock_t		sk_callback_lock;
 	int			sk_err,
@@ -557,6 +559,9 @@ struct proto {
 	kmem_cache_t		*slab;
 	unsigned int		obj_size;
 
+	kmem_cache_t		*twsk_slab;
+	unsigned int		twsk_obj_size;
+
 	struct request_sock_ops	*rsk_prot;
 
 	struct module		*owner;

include/net/tcp.h

@@ -38,207 +38,14 @@
 #include <net/ip.h>
 #include <net/tcp_states.h>
 
-#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
-#include <linux/ipv6.h>
-#endif
 #include <linux/seq_file.h>
 
 extern struct inet_hashinfo tcp_hashinfo;
 
-#if (BITS_PER_LONG == 64)
-#define TCP_ADDRCMP_ALIGN_BYTES 8
-#else
-#define TCP_ADDRCMP_ALIGN_BYTES 4
-#endif
-
-/* This is a TIME_WAIT bucket.  It works around the memory consumption
- * problems of sockets in such a state on heavily loaded servers, but
- * without violating the protocol specification.
- */
-struct tcp_tw_bucket {
-	/*
-	 * Now struct sock also uses sock_common, so please just
-	 * don't add nothing before this first member (__tw_common) --acme
-	 */
-	struct sock_common	__tw_common;
-#define tw_family		__tw_common.skc_family
-#define tw_state		__tw_common.skc_state
-#define tw_reuse		__tw_common.skc_reuse
-#define tw_bound_dev_if		__tw_common.skc_bound_dev_if
-#define tw_node			__tw_common.skc_node
-#define tw_bind_node		__tw_common.skc_bind_node
-#define tw_refcnt		__tw_common.skc_refcnt
-	volatile unsigned char	tw_substate;
-	unsigned char		tw_rcv_wscale;
-	__u16			tw_sport;
-	/* Socket demultiplex comparisons on incoming packets. */
-	/* these five are in inet_sock */
-	__u32			tw_daddr
-		__attribute__((aligned(TCP_ADDRCMP_ALIGN_BYTES)));
-	__u32			tw_rcv_saddr;
-	__u16			tw_dport;
-	__u16			tw_num;
-	/* And these are ours. */
-	int			tw_hashent;
-	int			tw_timeout;
-	__u32			tw_rcv_nxt;
-	__u32			tw_snd_nxt;
-	__u32			tw_rcv_wnd;
-	__u32			tw_ts_recent;
-	long			tw_ts_recent_stamp;
-	unsigned long		tw_ttd;
-	struct inet_bind_bucket	*tw_tb;
-	struct hlist_node	tw_death_node;
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-	struct in6_addr		tw_v6_daddr;
-	struct in6_addr		tw_v6_rcv_saddr;
-	int			tw_v6_ipv6only;
-#endif
-};
-
-static __inline__ void tw_add_node(struct tcp_tw_bucket *tw,
-				   struct hlist_head *list)
-{
-	hlist_add_head(&tw->tw_node, list);
-}
-
-static __inline__ void tw_add_bind_node(struct tcp_tw_bucket *tw,
-					struct hlist_head *list)
-{
-	hlist_add_head(&tw->tw_bind_node, list);
-}
-
-static inline int tw_dead_hashed(struct tcp_tw_bucket *tw)
-{
-	return tw->tw_death_node.pprev != NULL;
-}
-
-static __inline__ void tw_dead_node_init(struct tcp_tw_bucket *tw)
-{
-	tw->tw_death_node.pprev = NULL;
-}
-
-static __inline__ void __tw_del_dead_node(struct tcp_tw_bucket *tw)
-{
-	__hlist_del(&tw->tw_death_node);
-	tw_dead_node_init(tw);
-}
-
-static __inline__ int tw_del_dead_node(struct tcp_tw_bucket *tw)
-{
-	if (tw_dead_hashed(tw)) {
-		__tw_del_dead_node(tw);
-		return 1;
-	}
-	return 0;
-}
-
-#define tw_for_each(tw, node, head) \
-	hlist_for_each_entry(tw, node, head, tw_node)
-
-#define tw_for_each_inmate(tw, node, jail) \
-	hlist_for_each_entry(tw, node, jail, tw_death_node)
-
-#define tw_for_each_inmate_safe(tw, node, safe, jail) \
-	hlist_for_each_entry_safe(tw, node, safe, jail, tw_death_node)
-
-#define tcptw_sk(__sk)	((struct tcp_tw_bucket *)(__sk))
-
-static inline u32 tcp_v4_rcv_saddr(const struct sock *sk)
-{
-	return likely(sk->sk_state != TCP_TIME_WAIT) ?
-		inet_sk(sk)->rcv_saddr : tcptw_sk(sk)->tw_rcv_saddr;
-}
-
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-static inline struct in6_addr *__tcp_v6_rcv_saddr(const struct sock *sk)
-{
-	return likely(sk->sk_state != TCP_TIME_WAIT) ?
-		&inet6_sk(sk)->rcv_saddr : &tcptw_sk(sk)->tw_v6_rcv_saddr;
-}
-
-static inline struct in6_addr *tcp_v6_rcv_saddr(const struct sock *sk)
-{
-	return sk->sk_family == AF_INET6 ? __tcp_v6_rcv_saddr(sk) : NULL;
-}
-
-#define tcptw_sk_ipv6only(__sk)	(tcptw_sk(__sk)->tw_v6_ipv6only)
-
-static inline int tcp_v6_ipv6only(const struct sock *sk)
-{
-	return likely(sk->sk_state != TCP_TIME_WAIT) ?
-		ipv6_only_sock(sk) : tcptw_sk_ipv6only(sk);
-}
-#else
-# define __tcp_v6_rcv_saddr(__sk)	NULL
-# define tcp_v6_rcv_saddr(__sk)		NULL
-# define tcptw_sk_ipv6only(__sk)	0
-# define tcp_v6_ipv6only(__sk)		0
-#endif
-
-extern kmem_cache_t *tcp_timewait_cachep;
-
-static inline void tcp_tw_put(struct tcp_tw_bucket *tw)
-{
-	if (atomic_dec_and_test(&tw->tw_refcnt)) {
-#ifdef SOCK_REFCNT_DEBUG
-		printk(KERN_DEBUG "tw_bucket %p released\n", tw);
-#endif
-		kmem_cache_free(tcp_timewait_cachep, tw);
-	}
-}
-
 extern atomic_t tcp_orphan_count;
 extern int tcp_tw_count;
 extern void tcp_time_wait(struct sock *sk, int state, int timeo);
-extern void tcp_tw_deschedule(struct tcp_tw_bucket *tw);
-
-
-/* Socket demux engine toys. */
-#ifdef __BIG_ENDIAN
-#define TCP_COMBINED_PORTS(__sport, __dport) \
-	(((__u32)(__sport)<<16) | (__u32)(__dport))
-#else /* __LITTLE_ENDIAN */
-#define TCP_COMBINED_PORTS(__sport, __dport) \
-	(((__u32)(__dport)<<16) | (__u32)(__sport))
-#endif
-
-#if (BITS_PER_LONG == 64)
-#ifdef __BIG_ENDIAN
-#define TCP_V4_ADDR_COOKIE(__name, __saddr, __daddr) \
-	__u64 __name = (((__u64)(__saddr))<<32)|((__u64)(__daddr));
-#else /* __LITTLE_ENDIAN */
-#define TCP_V4_ADDR_COOKIE(__name, __saddr, __daddr) \
-	__u64 __name = (((__u64)(__daddr))<<32)|((__u64)(__saddr));
-#endif /* __BIG_ENDIAN */
-#define TCP_IPV4_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
-	(((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie))	&&	\
-	 ((*((__u32 *)&(inet_sk(__sk)->dport)))== (__ports))	&&	\
-	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
-#define TCP_IPV4_TW_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
-	(((*((__u64 *)&(tcptw_sk(__sk)->tw_daddr))) == (__cookie)) &&	\
-	 ((*((__u32 *)&(tcptw_sk(__sk)->tw_dport))) == (__ports)) &&	\
-	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
-#else /* 32-bit arch */
-#define TCP_V4_ADDR_COOKIE(__name, __saddr, __daddr)
-#define TCP_IPV4_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
-	((inet_sk(__sk)->daddr			== (__saddr))	&&	\
-	 (inet_sk(__sk)->rcv_saddr		== (__daddr))	&&	\
-	 ((*((__u32 *)&(inet_sk(__sk)->dport)))== (__ports))	&&	\
-	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
-#define TCP_IPV4_TW_MATCH(__sk, __cookie, __saddr, __daddr, __ports, __dif)\
-	((tcptw_sk(__sk)->tw_daddr		== (__saddr))	&&	\
-	 (tcptw_sk(__sk)->tw_rcv_saddr		== (__daddr))	&&	\
-	 ((*((__u32 *)&(tcptw_sk(__sk)->tw_dport))) == (__ports)) &&	\
-	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
-#endif /* 64-bit arch */
-
-#define TCP_IPV6_MATCH(__sk, __saddr, __daddr, __ports, __dif)	   \
-	(((*((__u32 *)&(inet_sk(__sk)->dport)))== (__ports))   	&& \
-	 ((__sk)->sk_family		== AF_INET6)		&& \
-	 ipv6_addr_equal(&inet6_sk(__sk)->daddr, (__saddr))	&& \
-	 ipv6_addr_equal(&inet6_sk(__sk)->rcv_saddr, (__daddr))	&& \
-	 (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
+extern void tcp_tw_deschedule(struct inet_timewait_sock *tw);
 
 #define MAX_TCP_HEADER	(128 + MAX_HEADER)
 
@@ -543,7 +350,7 @@ extern int			tcp_v4_rcv(struct sk_buff *skb);
 
 extern int			tcp_v4_remember_stamp(struct sock *sk);
 
-extern int		    	tcp_v4_tw_remember_stamp(struct tcp_tw_bucket *tw);
+extern int		    	tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
 
 extern int			tcp_sendmsg(struct kiocb *iocb, struct sock *sk,
 					    struct msghdr *msg, size_t size);
@@ -616,10 +423,9 @@ enum tcp_tw_status
 };
 
 
-extern enum tcp_tw_status	tcp_timewait_state_process(struct tcp_tw_bucket *tw,
+extern enum tcp_tw_status	tcp_timewait_state_process(struct inet_timewait_sock *tw,
 							   struct sk_buff *skb,
-							   struct tcphdr *th,
-							   unsigned len);
+							   const struct tcphdr *th);
 
 extern struct sock *		tcp_check_req(struct sock *sk,struct sk_buff *skb,
 					      struct request_sock *req,

net/core/sock.c

@@ -1378,7 +1378,8 @@ static LIST_HEAD(proto_list);
 
 int proto_register(struct proto *prot, int alloc_slab)
 {
-	char *request_sock_slab_name;
+	char *request_sock_slab_name = NULL;
+	char *timewait_sock_slab_name;
 	int rc = -ENOBUFS;
 
 	if (alloc_slab) {
@@ -1409,6 +1410,23 @@ int proto_register(struct proto *prot, int alloc_slab)
 				goto out_free_request_sock_slab_name;
 			}
 		}
+
+		if (prot->twsk_obj_size) {
+			static const char mask[] = "tw_sock_%s";
+
+			timewait_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
+
+			if (timewait_sock_slab_name == NULL)
+				goto out_free_request_sock_slab;
+
+			sprintf(timewait_sock_slab_name, mask, prot->name);
+			prot->twsk_slab = kmem_cache_create(timewait_sock_slab_name,
+							    prot->twsk_obj_size,
+							    0, SLAB_HWCACHE_ALIGN,
+							    NULL, NULL);
+			if (prot->twsk_slab == NULL)
+				goto out_free_timewait_sock_slab_name;
+		}
 	}
 
 	write_lock(&proto_list_lock);
@@ -1417,6 +1435,13 @@ int proto_register(struct proto *prot, int alloc_slab)
 	rc = 0;
 out:
 	return rc;
+out_free_timewait_sock_slab_name:
+	kfree(timewait_sock_slab_name);
+out_free_request_sock_slab:
+	if (prot->rsk_prot && prot->rsk_prot->slab) {
+		kmem_cache_destroy(prot->rsk_prot->slab);
+		prot->rsk_prot->slab = NULL;
+	}
 out_free_request_sock_slab_name:
 	kfree(request_sock_slab_name);
 out_free_sock_slab:
@@ -1444,6 +1469,14 @@ void proto_unregister(struct proto *prot)
 		prot->rsk_prot->slab = NULL;
 	}
 
+	if (prot->twsk_slab != NULL) {
+		const char *name = kmem_cache_name(prot->twsk_slab);
+
+		kmem_cache_destroy(prot->twsk_slab);
+		kfree(name);
+		prot->twsk_slab = NULL;
+	}
+
 	list_del(&prot->node);
 	write_unlock(&proto_list_lock);
 }

net/ipv4/tcp.c

@@ -271,8 +271,6 @@ int sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT;
 
 DEFINE_SNMP_STAT(struct tcp_mib, tcp_statistics);
 
-kmem_cache_t *tcp_timewait_cachep;
-
 atomic_t tcp_orphan_count = ATOMIC_INIT(0);
 
 int sysctl_tcp_mem[3];
@@ -2264,13 +2262,6 @@ void __init tcp_init(void)
 	if (!tcp_hashinfo.bind_bucket_cachep)
 		panic("tcp_init: Cannot alloc tcp_bind_bucket cache.");
 
-	tcp_timewait_cachep = kmem_cache_create("tcp_tw_bucket",
-						sizeof(struct tcp_tw_bucket),
-						0, SLAB_HWCACHE_ALIGN,
-						NULL, NULL);
-	if (!tcp_timewait_cachep)
-		panic("tcp_init: Cannot alloc tcp_tw_bucket cache.");
-
 	/* Size and allocate the main established and bind bucket
 	 * hash tables.
 	 *
@@ -2363,4 +2354,3 @@ EXPORT_SYMBOL(tcp_sendpage);
 EXPORT_SYMBOL(tcp_setsockopt);
 EXPORT_SYMBOL(tcp_shutdown);
 EXPORT_SYMBOL(tcp_statistics);
-EXPORT_SYMBOL(tcp_timewait_cachep);

net/ipv4/tcp_diag.c

@@ -81,7 +81,7 @@ static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk,
 	r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
 
 	if (r->tcpdiag_state == TCP_TIME_WAIT) {
-		struct tcp_tw_bucket *tw = (struct tcp_tw_bucket*)sk;
+		const struct inet_timewait_sock *tw = inet_twsk(sk);
 		long tmo = tw->tw_ttd - jiffies;
 		if (tmo < 0)
 			tmo = 0;
@@ -99,10 +99,12 @@ static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk,
 		r->tcpdiag_inode = 0;
 #ifdef CONFIG_IP_TCPDIAG_IPV6
 		if (r->tcpdiag_family == AF_INET6) {
+			const struct tcp6_timewait_sock *tcp6tw = tcp6_twsk(sk);
+
 			ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src,
-				       &tw->tw_v6_rcv_saddr);
+				       &tcp6tw->tw_v6_rcv_saddr);
 			ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst,
-				       &tw->tw_v6_daddr);
+				       &tcp6tw->tw_v6_daddr);
 		}
 #endif
 		nlh->nlmsg_len = skb->tail - b;
@@ -239,7 +241,7 @@ static int tcpdiag_get_exact(struct sk_buff *in_skb, const struct nlmsghdr *nlh)
 out:
 	if (sk) {
 		if (sk->sk_state == TCP_TIME_WAIT)
-			tcp_tw_put((struct tcp_tw_bucket*)sk);
+			inet_twsk_put((struct inet_timewait_sock *)sk);
 		else
 			sock_put(sk);
 	}

net/ipv6/addrconf.c

@@ -1041,7 +1041,7 @@ int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
 	const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
 	const struct in6_addr *sk2_rcv_saddr6 = tcp_v6_rcv_saddr(sk2);
 	u32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
-	u32 sk2_rcv_saddr = tcp_v4_rcv_saddr(sk2);
+	u32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
 	int sk_ipv6only = ipv6_only_sock(sk);
 	int sk2_ipv6only = tcp_v6_ipv6only(sk2);
 	int addr_type = ipv6_addr_type(sk_rcv_saddr6);

net/ipv6/tcp_ipv6.c

@@ -308,33 +308,32 @@ static inline struct sock *__tcp_v6_lookup_established(struct in6_addr *saddr, u
 						       struct in6_addr *daddr, u16 hnum,
 						       int dif)
 {
-	struct inet_ehash_bucket *head;
 	struct sock *sk;
-	struct hlist_node *node;
-	__u32 ports = TCP_COMBINED_PORTS(sport, hnum);
-	int hash;
-
+	const struct hlist_node *node;
+	const __u32 ports = INET_COMBINED_PORTS(sport, hnum);
 	/* Optimize here for direct hit, only listening connections can
 	 * have wildcards anyways.
 	 */
-	hash = tcp_v6_hashfn(daddr, hnum, saddr, sport);
-	head = &tcp_hashinfo.ehash[hash];
+	const int hash = tcp_v6_hashfn(daddr, hnum, saddr, sport);
+	struct inet_ehash_bucket *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. */
-		if(TCP_IPV6_MATCH(sk, saddr, daddr, ports, dif))
+		if (INET6_MATCH(sk, saddr, daddr, ports, dif))
 			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_hashinfo.ehash_size)->chain) {
-		/* FIXME: acme: check this... */
-		struct tcp_tw_bucket *tw = (struct tcp_tw_bucket *)sk;
+		const struct inet_timewait_sock *tw = inet_twsk(sk);
 
 		if(*((__u32 *)&(tw->tw_dport))	== ports	&&
 		   sk->sk_family		== PF_INET6) {
-			if(ipv6_addr_equal(&tw->tw_v6_daddr, saddr)	&&
-			   ipv6_addr_equal(&tw->tw_v6_rcv_saddr, daddr)	&&
-			   (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif))
+			const struct tcp6_timewait_sock *tcp6tw = tcp6_twsk(sk);
+
+			if (ipv6_addr_equal(&tcp6tw->tw_v6_daddr, saddr)	&&
+			    ipv6_addr_equal(&tcp6tw->tw_v6_rcv_saddr, daddr)	&&
+			    (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif))
 				goto hit;
 		}
 	}
@@ -455,43 +454,46 @@ static __u32 tcp_v6_init_sequence(struct sock *sk, struct sk_buff *skb)
 }
 
 static int __tcp_v6_check_established(struct sock *sk, __u16 lport,
-				      struct tcp_tw_bucket **twp)
+				      struct inet_timewait_sock **twp)
 {
 	struct inet_sock *inet = inet_sk(sk);
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct in6_addr *daddr = &np->rcv_saddr;
 	struct in6_addr *saddr = &np->daddr;
 	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);
+	const u32 ports = INET_COMBINED_PORTS(inet->dport, lport);
+	const int hash = tcp_v6_hashfn(daddr, inet->num, saddr, inet->dport);
 	struct inet_ehash_bucket *head = &tcp_hashinfo.ehash[hash];
 	struct sock *sk2;
-	struct hlist_node *node;
-	struct tcp_tw_bucket *tw;
+	const struct hlist_node *node;
+	struct inet_timewait_sock *tw;
 
 	write_lock(&head->lock);
 
 	/* Check TIME-WAIT sockets first. */
 	sk_for_each(sk2, node, &(head + tcp_hashinfo.ehash_size)->chain) {
-		tw = (struct tcp_tw_bucket*)sk2;
+		const struct tcp6_timewait_sock *tcp6tw = tcp6_twsk(sk2);
+
+		tw = inet_twsk(sk2);
 
 		if(*((__u32 *)&(tw->tw_dport))	== ports	&&
 		   sk2->sk_family		== PF_INET6	&&
-		   ipv6_addr_equal(&tw->tw_v6_daddr, saddr)	&&
-		   ipv6_addr_equal(&tw->tw_v6_rcv_saddr, daddr)	&&
+		   ipv6_addr_equal(&tcp6tw->tw_v6_daddr, saddr)	&&
+		   ipv6_addr_equal(&tcp6tw->tw_v6_rcv_saddr, daddr)	&&
 		   sk2->sk_bound_dev_if == sk->sk_bound_dev_if) {
+			const struct tcp_timewait_sock *tcptw = tcp_twsk(sk2);
 			struct tcp_sock *tp = tcp_sk(sk);
 
-			if (tw->tw_ts_recent_stamp &&
-			    (!twp || (sysctl_tcp_tw_reuse &&
-				      xtime.tv_sec - 
-				      tw->tw_ts_recent_stamp > 1))) {
+			if (tcptw->tw_ts_recent_stamp &&
+			    (!twp ||
+			     (sysctl_tcp_tw_reuse &&
+			      xtime.tv_sec - tcptw->tw_ts_recent_stamp > 1))) {
 				/* See comment in tcp_ipv4.c */
-				tp->write_seq = tw->tw_snd_nxt + 65535 + 2;
+				tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
 				if (!tp->write_seq)
 					tp->write_seq = 1;
-				tp->rx_opt.ts_recent = tw->tw_ts_recent;
-				tp->rx_opt.ts_recent_stamp = tw->tw_ts_recent_stamp;
+				tp->rx_opt.ts_recent	   = tcptw->tw_ts_recent;
+				tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp;
 				sock_hold(sk2);
 				goto unique;
 			} else
@@ -502,7 +504,7 @@ static int __tcp_v6_check_established(struct sock *sk, __u16 lport,
 
 	/* And established part... */
 	sk_for_each(sk2, node, &head->chain) {
-		if(TCP_IPV6_MATCH(sk2, saddr, daddr, ports, dif))
+		if (INET6_MATCH(sk2, saddr, daddr, ports, dif))
 			goto not_unique;
 	}
 
@@ -521,7 +523,7 @@ unique:
 		tcp_tw_deschedule(tw);
 		NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
 
-		tcp_tw_put(tw);
+		inet_twsk_put(tw);
 	}
 	return 0;
 
@@ -556,7 +558,7 @@ static int tcp_v6_hash_connect(struct sock *sk)
 		static u32 hint;
 		u32 offset = hint + tcpv6_port_offset(sk);
 		struct hlist_node *node;
- 		struct tcp_tw_bucket *tw = NULL;
+ 		struct inet_timewait_sock *tw = NULL;
 
  		local_bh_disable();
 		for (i = 1; i <= range; i++) {
@@ -609,7 +611,7 @@ ok:
 
  		if (tw) {
  			tcp_tw_deschedule(tw);
- 			tcp_tw_put(tw);
+ 			inet_twsk_put(tw);
  		}
 
 		ret = 0;
@@ -845,7 +847,7 @@ static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 	}
 
 	if (sk->sk_state == TCP_TIME_WAIT) {
-		tcp_tw_put((struct tcp_tw_bucket*)sk);
+		inet_twsk_put((struct inet_timewait_sock *)sk);
 		return;
 	}
 
@@ -1223,12 +1225,14 @@ static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32
 
 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
 {
-	struct tcp_tw_bucket *tw = (struct tcp_tw_bucket *)sk;
+	struct inet_timewait_sock *tw = inet_twsk(sk);
+	const struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
 
-	tcp_v6_send_ack(skb, tw->tw_snd_nxt, tw->tw_rcv_nxt,
-			tw->tw_rcv_wnd >> tw->tw_rcv_wscale, tw->tw_ts_recent);
+	tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
+			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
+			tcptw->tw_ts_recent);
 
-	tcp_tw_put(tw);
+	inet_twsk_put(tw);
 }
 
 static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
@@ -1261,7 +1265,7 @@ static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
 			bh_lock_sock(nsk);
 			return nsk;
 		}
-		tcp_tw_put((struct tcp_tw_bucket*)nsk);
+		inet_twsk_put((struct inet_timewait_sock *)nsk);
 		return NULL;
 	}
 
@@ -1798,26 +1802,26 @@ discard_and_relse:
 
 do_time_wait:
 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
-		tcp_tw_put((struct tcp_tw_bucket *) sk);
+		inet_twsk_put((struct inet_timewait_sock *)sk);
 		goto discard_it;
 	}
 
 	if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
 		TCP_INC_STATS_BH(TCP_MIB_INERRS);
-		tcp_tw_put((struct tcp_tw_bucket *) sk);
+		inet_twsk_put((struct inet_timewait_sock *)sk);
 		goto discard_it;
 	}
 
-	switch(tcp_timewait_state_process((struct tcp_tw_bucket *)sk,
-					  skb, th, skb->len)) {
+	switch (tcp_timewait_state_process((struct inet_timewait_sock *)sk,
+					   skb, th)) {
 	case TCP_TW_SYN:
 	{
 		struct sock *sk2;
 
 		sk2 = tcp_v6_lookup_listener(&skb->nh.ipv6h->daddr, ntohs(th->dest), tcp_v6_iif(skb));
 		if (sk2 != NULL) {
-			tcp_tw_deschedule((struct tcp_tw_bucket *)sk);
-			tcp_tw_put((struct tcp_tw_bucket *)sk);
+			tcp_tw_deschedule((struct inet_timewait_sock *)sk);
+			inet_twsk_put((struct inet_timewait_sock *)sk);
 			sk = sk2;
 			goto process;
 		}
@@ -2137,17 +2141,18 @@ static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
 }
 
 static void get_timewait6_sock(struct seq_file *seq, 
-			       struct tcp_tw_bucket *tw, int i)
+			       struct inet_timewait_sock *tw, int i)
 {
 	struct in6_addr *dest, *src;
 	__u16 destp, srcp;
+	struct tcp6_timewait_sock *tcp6tw = tcp6_twsk((struct sock *)tw);
 	int ttd = tw->tw_ttd - jiffies;
 
 	if (ttd < 0)
 		ttd = 0;
 
-	dest  = &tw->tw_v6_daddr;
-	src   = &tw->tw_v6_rcv_saddr;
+	dest = &tcp6tw->tw_v6_daddr;
+	src  = &tcp6tw->tw_v6_rcv_saddr;
 	destp = ntohs(tw->tw_dport);
 	srcp  = ntohs(tw->tw_sport);
 
@@ -2244,6 +2249,7 @@ struct proto tcpv6_prot = {
 	.sysctl_rmem		= sysctl_tcp_rmem,
 	.max_header		= MAX_TCP_HEADER,
 	.obj_size		= sizeof(struct tcp6_sock),
+	.twsk_obj_size		= sizeof(struct tcp6_timewait_sock),
 	.rsk_prot		= &tcp6_request_sock_ops,
 };