[TIMEWAIT]: Introduce inet_timewait_death_row

That groups all of the tables and variables associated to the TCP timewait
schedulling/recycling/killing code, that now can be isolated from the TCP
specific code and used by other transport protocols, such as DCCP.

Next changeset will move this code to net/ipv4/inet_timewait_sock.c
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/inet_timewait_sock.h

@@ -19,13 +19,69 @@
 
 #include <linux/ip.h>
 #include <linux/list.h>
+#include <linux/timer.h>
 #include <linux/types.h>
+#include <linux/workqueue.h>
 
 #include <net/sock.h>
 #include <net/tcp_states.h>
 
 #include <asm/atomic.h>
 
+struct inet_hashinfo;
+
+#define INET_TWDR_RECYCLE_SLOTS_LOG	5
+#define INET_TWDR_RECYCLE_SLOTS		(1 << INET_TWDR_RECYCLE_SLOTS_LOG)
+
+/*
+ * If time > 4sec, it is "slow" path, no recycling is required,
+ * so that we select tick to get range about 4 seconds.
+ */
+#if HZ <= 16 || HZ > 4096
+# error Unsupported: HZ <= 16 or HZ > 4096
+#elif HZ <= 32
+# define INET_TWDR_RECYCLE_TICK (5 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#elif HZ <= 64
+# define INET_TWDR_RECYCLE_TICK (6 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#elif HZ <= 128
+# define INET_TWDR_RECYCLE_TICK (7 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#elif HZ <= 256
+# define INET_TWDR_RECYCLE_TICK (8 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#elif HZ <= 512
+# define INET_TWDR_RECYCLE_TICK (9 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#elif HZ <= 1024
+# define INET_TWDR_RECYCLE_TICK (10 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#elif HZ <= 2048
+# define INET_TWDR_RECYCLE_TICK (11 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#else
+# define INET_TWDR_RECYCLE_TICK (12 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
+#endif
+
+/* TIME_WAIT reaping mechanism. */
+#define INET_TWDR_TWKILL_SLOTS	8 /* Please keep this a power of 2. */
+
+#define INET_TWDR_TWKILL_QUOTA 100
+
+struct inet_timewait_death_row {
+	/* Short-time timewait calendar */
+	int			twcal_hand;
+	int			twcal_jiffie;
+	struct timer_list	twcal_timer;
+	struct hlist_head	twcal_row[INET_TWDR_RECYCLE_SLOTS];
+
+	spinlock_t		death_lock;
+	int			tw_count;
+	int			period;
+	u32			thread_slots;
+	struct work_struct	twkill_work;
+	struct timer_list	tw_timer;
+	int			slot;
+	struct hlist_head	cells[INET_TWDR_TWKILL_SLOTS];
+	struct inet_hashinfo 	*hashinfo;
+	int			sysctl_tw_recycle;
+	int			sysctl_max_tw_buckets;
+};
+
 #if (BITS_PER_LONG == 64)
 #define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 8
 #else
@@ -33,7 +89,6 @@
 #endif
 
 struct inet_bind_bucket;
-struct inet_hashinfo;
 
 /*
  * This is a TIME_WAIT sock. It works around the memory consumption

include/net/tcp.h

@@ -29,6 +29,7 @@
 #include <linux/percpu.h>
 
 #include <net/inet_connection_sock.h>
+#include <net/inet_timewait_sock.h>
 #include <net/inet_hashtables.h>
 #include <net/checksum.h>
 #include <net/request_sock.h>
@@ -42,9 +43,9 @@
 extern struct inet_hashinfo tcp_hashinfo;
 
 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 inet_timewait_sock *tw);
+extern void inet_twsk_deschedule(struct inet_timewait_sock *tw,
+				 struct inet_timewait_death_row *twdr);
 
 #define MAX_TCP_HEADER	(128 + MAX_HEADER)
 
@@ -148,33 +149,6 @@ extern void tcp_tw_deschedule(struct inet_timewait_sock *tw);
 					 * timestamps. It must be less than
 					 * minimal timewait lifetime.
 					 */
-
-#define TCP_TW_RECYCLE_SLOTS_LOG	5
-#define TCP_TW_RECYCLE_SLOTS		(1<<TCP_TW_RECYCLE_SLOTS_LOG)
-
-/* If time > 4sec, it is "slow" path, no recycling is required,
-   so that we select tick to get range about 4 seconds.
- */
-
-#if HZ <= 16 || HZ > 4096
-# error Unsupported: HZ <= 16 or HZ > 4096
-#elif HZ <= 32
-# define TCP_TW_RECYCLE_TICK (5+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#elif HZ <= 64
-# define TCP_TW_RECYCLE_TICK (6+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#elif HZ <= 128
-# define TCP_TW_RECYCLE_TICK (7+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#elif HZ <= 256
-# define TCP_TW_RECYCLE_TICK (8+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#elif HZ <= 512
-# define TCP_TW_RECYCLE_TICK (9+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#elif HZ <= 1024
-# define TCP_TW_RECYCLE_TICK (10+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#elif HZ <= 2048
-# define TCP_TW_RECYCLE_TICK (11+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#else
-# define TCP_TW_RECYCLE_TICK (12+2-TCP_TW_RECYCLE_SLOTS_LOG)
-#endif
 /*
  *	TCP option
  */
@@ -209,12 +183,13 @@ extern void tcp_tw_deschedule(struct inet_timewait_sock *tw);
 #define TCP_NAGLE_CORK		2	/* Socket is corked	    */
 #define TCP_NAGLE_PUSH		4	/* Cork is overriden for already queued data */
 
+extern struct inet_timewait_death_row tcp_death_row;
+
 /* sysctl variables for tcp */
 extern int sysctl_tcp_timestamps;
 extern int sysctl_tcp_window_scaling;
 extern int sysctl_tcp_sack;
 extern int sysctl_tcp_fin_timeout;
-extern int sysctl_tcp_tw_recycle;
 extern int sysctl_tcp_keepalive_time;
 extern int sysctl_tcp_keepalive_probes;
 extern int sysctl_tcp_keepalive_intvl;
@@ -229,7 +204,6 @@ extern int sysctl_tcp_stdurg;
 extern int sysctl_tcp_rfc1337;
 extern int sysctl_tcp_abort_on_overflow;
 extern int sysctl_tcp_max_orphans;
-extern int sysctl_tcp_max_tw_buckets;
 extern int sysctl_tcp_fack;
 extern int sysctl_tcp_reordering;
 extern int sysctl_tcp_ecn;

net/ipv4/proc.c

@@ -65,7 +65,7 @@ static int sockstat_seq_show(struct seq_file *seq, void *v)
 	socket_seq_show(seq);
 	seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %d\n",
 		   fold_prot_inuse(&tcp_prot), atomic_read(&tcp_orphan_count),
-		   tcp_tw_count, atomic_read(&tcp_sockets_allocated),
+		   tcp_death_row.tw_count, atomic_read(&tcp_sockets_allocated),
 		   atomic_read(&tcp_memory_allocated));
 	seq_printf(seq, "UDP: inuse %d\n", fold_prot_inuse(&udp_prot));
 	seq_printf(seq, "RAW: inuse %d\n", fold_prot_inuse(&raw_prot));

net/ipv4/sysctl_net_ipv4.c

@@ -259,7 +259,7 @@ ctl_table ipv4_table[] = {
 	{
 		.ctl_name	= NET_TCP_MAX_TW_BUCKETS,
 		.procname	= "tcp_max_tw_buckets",
-		.data		= &sysctl_tcp_max_tw_buckets,
+		.data		= &tcp_death_row.sysctl_max_tw_buckets,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec
@@ -363,7 +363,7 @@ ctl_table ipv4_table[] = {
 	{
 		.ctl_name	= NET_TCP_TW_RECYCLE,
 		.procname	= "tcp_tw_recycle",
-		.data		= &sysctl_tcp_tw_recycle,
+		.data		= &tcp_death_row.sysctl_tw_recycle,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec

net/ipv4/tcp.c

@@ -2109,12 +2109,12 @@ void __init tcp_init(void)
 	if (order >= 4) {
 		sysctl_local_port_range[0] = 32768;
 		sysctl_local_port_range[1] = 61000;
-		sysctl_tcp_max_tw_buckets = 180000;
+		tcp_death_row.sysctl_max_tw_buckets = 180000;
 		sysctl_tcp_max_orphans = 4096 << (order - 4);
 		sysctl_max_syn_backlog = 1024;
 	} else if (order < 3) {
 		sysctl_local_port_range[0] = 1024 * (3 - order);
-		sysctl_tcp_max_tw_buckets >>= (3 - order);
+		tcp_death_row.sysctl_max_tw_buckets >>= (3 - order);
 		sysctl_tcp_max_orphans >>= (3 - order);
 		sysctl_max_syn_backlog = 128;
 	}

net/ipv4/tcp_ipv4.c

@@ -199,7 +199,7 @@ unique:
 		NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
 	} else if (tw) {
 		/* Silly. Should hash-dance instead... */
-		tcp_tw_deschedule(tw);
+		inet_twsk_deschedule(tw, &tcp_death_row);
 		NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
 
 		inet_twsk_put(tw);
@@ -291,7 +291,7 @@ ok:
  		spin_unlock(&head->lock);
 
  		if (tw) {
- 			tcp_tw_deschedule(tw);
+ 			inet_twsk_deschedule(tw, &tcp_death_row);;
  			inet_twsk_put(tw);
  		}
 
@@ -366,7 +366,7 @@ int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 		tp->write_seq		   = 0;
 	}
 
-	if (sysctl_tcp_tw_recycle &&
+	if (tcp_death_row.sysctl_tw_recycle &&
 	    !tp->rx_opt.ts_recent_stamp && rt->rt_dst == daddr) {
 		struct inet_peer *peer = rt_get_peer(rt);
 
@@ -965,7 +965,7 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
 		 * are made in the function processing timewait state.
 		 */
 		if (tmp_opt.saw_tstamp &&
-		    sysctl_tcp_tw_recycle &&
+		    tcp_death_row.sysctl_tw_recycle &&
 		    (dst = inet_csk_route_req(sk, req)) != NULL &&
 		    (peer = rt_get_peer((struct rtable *)dst)) != NULL &&
 		    peer->v4daddr == saddr) {
@@ -1305,7 +1305,8 @@ do_time_wait:
 							ntohs(th->dest),
 							inet_iif(skb));
 		if (sk2) {
-			tcp_tw_deschedule((struct inet_timewait_sock *)sk);
+			inet_twsk_deschedule((struct inet_timewait_sock *)sk,
+					     &tcp_death_row);
 			inet_twsk_put((struct inet_timewait_sock *)sk);
 			sk = sk2;
 			goto process;

net/ipv6/tcp_ipv6.c

@@ -521,7 +521,7 @@ unique:
 		NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
 	} else if (tw) {
 		/* Silly. Should hash-dance instead... */
-		tcp_tw_deschedule(tw);
+		inet_twsk_deschedule(tw, &tcp_death_row);
 		NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
 
 		inet_twsk_put(tw);
@@ -611,7 +611,7 @@ ok:
  		spin_unlock(&head->lock);
 
  		if (tw) {
- 			tcp_tw_deschedule(tw);
+ 			inet_twsk_deschedule(tw, &tcp_death_row);
  			inet_twsk_put(tw);
  		}
 
@@ -1820,8 +1820,9 @@ do_time_wait:
 
 		sk2 = tcp_v6_lookup_listener(&skb->nh.ipv6h->daddr, ntohs(th->dest), tcp_v6_iif(skb));
 		if (sk2 != NULL) {
-			tcp_tw_deschedule((struct inet_timewait_sock *)sk);
-			inet_twsk_put((struct inet_timewait_sock *)sk);
+			struct inet_timewait_sock *tw = inet_twsk(sk);
+			inet_twsk_deschedule(tw, &tcp_death_row);
+			inet_twsk_put(tw);
 			sk = sk2;
 			goto process;
 		}