aboutsummaryrefslogtreecommitdiffstats
path: root/net/ipv4/tcp.c
diff options
context:
space:
mode:
Diffstat (limited to 'net/ipv4/tcp.c')
-rw-r--r--net/ipv4/tcp.c102
1 files changed, 78 insertions, 24 deletions
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 1d723de18686..0b491bf03db4 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -5,8 +5,6 @@
5 * 5 *
6 * Implementation of the Transmission Control Protocol(TCP). 6 * Implementation of the Transmission Control Protocol(TCP).
7 * 7 *
8 * Version: $Id: tcp.c,v 1.216 2002/02/01 22:01:04 davem Exp $
9 *
10 * Authors: Ross Biro 8 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Mark Evans, <evansmp@uhura.aston.ac.uk> 10 * Mark Evans, <evansmp@uhura.aston.ac.uk>
@@ -279,8 +277,6 @@
279 277
280int sysctl_tcp_fin_timeout __read_mostly = TCP_FIN_TIMEOUT; 278int sysctl_tcp_fin_timeout __read_mostly = TCP_FIN_TIMEOUT;
281 279
282DEFINE_SNMP_STAT(struct tcp_mib, tcp_statistics) __read_mostly;
283
284atomic_t tcp_orphan_count = ATOMIC_INIT(0); 280atomic_t tcp_orphan_count = ATOMIC_INIT(0);
285 281
286EXPORT_SYMBOL_GPL(tcp_orphan_count); 282EXPORT_SYMBOL_GPL(tcp_orphan_count);
@@ -318,10 +314,10 @@ int tcp_memory_pressure __read_mostly;
318 314
319EXPORT_SYMBOL(tcp_memory_pressure); 315EXPORT_SYMBOL(tcp_memory_pressure);
320 316
321void tcp_enter_memory_pressure(void) 317void tcp_enter_memory_pressure(struct sock *sk)
322{ 318{
323 if (!tcp_memory_pressure) { 319 if (!tcp_memory_pressure) {
324 NET_INC_STATS(LINUX_MIB_TCPMEMORYPRESSURES); 320 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMEMORYPRESSURES);
325 tcp_memory_pressure = 1; 321 tcp_memory_pressure = 1;
326 } 322 }
327} 323}
@@ -346,8 +342,8 @@ unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
346 return inet_csk_listen_poll(sk); 342 return inet_csk_listen_poll(sk);
347 343
348 /* Socket is not locked. We are protected from async events 344 /* Socket is not locked. We are protected from async events
349 by poll logic and correct handling of state changes 345 * by poll logic and correct handling of state changes
350 made by another threads is impossible in any case. 346 * made by other threads is impossible in any case.
351 */ 347 */
352 348
353 mask = 0; 349 mask = 0;
@@ -373,10 +369,10 @@ unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
373 * in state CLOSE_WAIT. One solution is evident --- to set POLLHUP 369 * in state CLOSE_WAIT. One solution is evident --- to set POLLHUP
374 * if and only if shutdown has been made in both directions. 370 * if and only if shutdown has been made in both directions.
375 * Actually, it is interesting to look how Solaris and DUX 371 * Actually, it is interesting to look how Solaris and DUX
376 * solve this dilemma. I would prefer, if PULLHUP were maskable, 372 * solve this dilemma. I would prefer, if POLLHUP were maskable,
377 * then we could set it on SND_SHUTDOWN. BTW examples given 373 * then we could set it on SND_SHUTDOWN. BTW examples given
378 * in Stevens' books assume exactly this behaviour, it explains 374 * in Stevens' books assume exactly this behaviour, it explains
379 * why PULLHUP is incompatible with POLLOUT. --ANK 375 * why POLLHUP is incompatible with POLLOUT. --ANK
380 * 376 *
381 * NOTE. Check for TCP_CLOSE is added. The goal is to prevent 377 * NOTE. Check for TCP_CLOSE is added. The goal is to prevent
382 * blocking on fresh not-connected or disconnected socket. --ANK 378 * blocking on fresh not-connected or disconnected socket. --ANK
@@ -651,7 +647,7 @@ struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp)
651 } 647 }
652 __kfree_skb(skb); 648 __kfree_skb(skb);
653 } else { 649 } else {
654 sk->sk_prot->enter_memory_pressure(); 650 sk->sk_prot->enter_memory_pressure(sk);
655 sk_stream_moderate_sndbuf(sk); 651 sk_stream_moderate_sndbuf(sk);
656 } 652 }
657 return NULL; 653 return NULL;
@@ -1155,7 +1151,7 @@ static void tcp_prequeue_process(struct sock *sk)
1155 struct sk_buff *skb; 1151 struct sk_buff *skb;
1156 struct tcp_sock *tp = tcp_sk(sk); 1152 struct tcp_sock *tp = tcp_sk(sk);
1157 1153
1158 NET_INC_STATS_USER(LINUX_MIB_TCPPREQUEUED); 1154 NET_INC_STATS_USER(sock_net(sk), LINUX_MIB_TCPPREQUEUED);
1159 1155
1160 /* RX process wants to run with disabled BHs, though it is not 1156 /* RX process wants to run with disabled BHs, though it is not
1161 * necessary */ 1157 * necessary */
@@ -1477,7 +1473,7 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1477 /* __ Restore normal policy in scheduler __ */ 1473 /* __ Restore normal policy in scheduler __ */
1478 1474
1479 if ((chunk = len - tp->ucopy.len) != 0) { 1475 if ((chunk = len - tp->ucopy.len) != 0) {
1480 NET_ADD_STATS_USER(LINUX_MIB_TCPDIRECTCOPYFROMBACKLOG, chunk); 1476 NET_ADD_STATS_USER(sock_net(sk), LINUX_MIB_TCPDIRECTCOPYFROMBACKLOG, chunk);
1481 len -= chunk; 1477 len -= chunk;
1482 copied += chunk; 1478 copied += chunk;
1483 } 1479 }
@@ -1488,7 +1484,7 @@ do_prequeue:
1488 tcp_prequeue_process(sk); 1484 tcp_prequeue_process(sk);
1489 1485
1490 if ((chunk = len - tp->ucopy.len) != 0) { 1486 if ((chunk = len - tp->ucopy.len) != 0) {
1491 NET_ADD_STATS_USER(LINUX_MIB_TCPDIRECTCOPYFROMPREQUEUE, chunk); 1487 NET_ADD_STATS_USER(sock_net(sk), LINUX_MIB_TCPDIRECTCOPYFROMPREQUEUE, chunk);
1492 len -= chunk; 1488 len -= chunk;
1493 copied += chunk; 1489 copied += chunk;
1494 } 1490 }
@@ -1603,7 +1599,7 @@ skip_copy:
1603 tcp_prequeue_process(sk); 1599 tcp_prequeue_process(sk);
1604 1600
1605 if (copied > 0 && (chunk = len - tp->ucopy.len) != 0) { 1601 if (copied > 0 && (chunk = len - tp->ucopy.len) != 0) {
1606 NET_ADD_STATS_USER(LINUX_MIB_TCPDIRECTCOPYFROMPREQUEUE, chunk); 1602 NET_ADD_STATS_USER(sock_net(sk), LINUX_MIB_TCPDIRECTCOPYFROMPREQUEUE, chunk);
1607 len -= chunk; 1603 len -= chunk;
1608 copied += chunk; 1604 copied += chunk;
1609 } 1605 }
@@ -1670,12 +1666,12 @@ void tcp_set_state(struct sock *sk, int state)
1670 switch (state) { 1666 switch (state) {
1671 case TCP_ESTABLISHED: 1667 case TCP_ESTABLISHED:
1672 if (oldstate != TCP_ESTABLISHED) 1668 if (oldstate != TCP_ESTABLISHED)
1673 TCP_INC_STATS(TCP_MIB_CURRESTAB); 1669 TCP_INC_STATS(sock_net(sk), TCP_MIB_CURRESTAB);
1674 break; 1670 break;
1675 1671
1676 case TCP_CLOSE: 1672 case TCP_CLOSE:
1677 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED) 1673 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
1678 TCP_INC_STATS(TCP_MIB_ESTABRESETS); 1674 TCP_INC_STATS(sock_net(sk), TCP_MIB_ESTABRESETS);
1679 1675
1680 sk->sk_prot->unhash(sk); 1676 sk->sk_prot->unhash(sk);
1681 if (inet_csk(sk)->icsk_bind_hash && 1677 if (inet_csk(sk)->icsk_bind_hash &&
@@ -1684,7 +1680,7 @@ void tcp_set_state(struct sock *sk, int state)
1684 /* fall through */ 1680 /* fall through */
1685 default: 1681 default:
1686 if (oldstate==TCP_ESTABLISHED) 1682 if (oldstate==TCP_ESTABLISHED)
1687 TCP_DEC_STATS(TCP_MIB_CURRESTAB); 1683 TCP_DEC_STATS(sock_net(sk), TCP_MIB_CURRESTAB);
1688 } 1684 }
1689 1685
1690 /* Change state AFTER socket is unhashed to avoid closed 1686 /* Change state AFTER socket is unhashed to avoid closed
@@ -1795,13 +1791,13 @@ void tcp_close(struct sock *sk, long timeout)
1795 */ 1791 */
1796 if (data_was_unread) { 1792 if (data_was_unread) {
1797 /* Unread data was tossed, zap the connection. */ 1793 /* Unread data was tossed, zap the connection. */
1798 NET_INC_STATS_USER(LINUX_MIB_TCPABORTONCLOSE); 1794 NET_INC_STATS_USER(sock_net(sk), LINUX_MIB_TCPABORTONCLOSE);
1799 tcp_set_state(sk, TCP_CLOSE); 1795 tcp_set_state(sk, TCP_CLOSE);
1800 tcp_send_active_reset(sk, GFP_KERNEL); 1796 tcp_send_active_reset(sk, GFP_KERNEL);
1801 } else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) { 1797 } else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1802 /* Check zero linger _after_ checking for unread data. */ 1798 /* Check zero linger _after_ checking for unread data. */
1803 sk->sk_prot->disconnect(sk, 0); 1799 sk->sk_prot->disconnect(sk, 0);
1804 NET_INC_STATS_USER(LINUX_MIB_TCPABORTONDATA); 1800 NET_INC_STATS_USER(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
1805 } else if (tcp_close_state(sk)) { 1801 } else if (tcp_close_state(sk)) {
1806 /* We FIN if the application ate all the data before 1802 /* We FIN if the application ate all the data before
1807 * zapping the connection. 1803 * zapping the connection.
@@ -1873,7 +1869,8 @@ adjudge_to_death:
1873 if (tp->linger2 < 0) { 1869 if (tp->linger2 < 0) {
1874 tcp_set_state(sk, TCP_CLOSE); 1870 tcp_set_state(sk, TCP_CLOSE);
1875 tcp_send_active_reset(sk, GFP_ATOMIC); 1871 tcp_send_active_reset(sk, GFP_ATOMIC);
1876 NET_INC_STATS_BH(LINUX_MIB_TCPABORTONLINGER); 1872 NET_INC_STATS_BH(sock_net(sk),
1873 LINUX_MIB_TCPABORTONLINGER);
1877 } else { 1874 } else {
1878 const int tmo = tcp_fin_time(sk); 1875 const int tmo = tcp_fin_time(sk);
1879 1876
@@ -1895,7 +1892,8 @@ adjudge_to_death:
1895 "sockets\n"); 1892 "sockets\n");
1896 tcp_set_state(sk, TCP_CLOSE); 1893 tcp_set_state(sk, TCP_CLOSE);
1897 tcp_send_active_reset(sk, GFP_ATOMIC); 1894 tcp_send_active_reset(sk, GFP_ATOMIC);
1898 NET_INC_STATS_BH(LINUX_MIB_TCPABORTONMEMORY); 1895 NET_INC_STATS_BH(sock_net(sk),
1896 LINUX_MIB_TCPABORTONMEMORY);
1899 } 1897 }
1900 } 1898 }
1901 1899
@@ -2590,12 +2588,69 @@ void __tcp_put_md5sig_pool(void)
2590} 2588}
2591 2589
2592EXPORT_SYMBOL(__tcp_put_md5sig_pool); 2590EXPORT_SYMBOL(__tcp_put_md5sig_pool);
2591
2592int tcp_md5_hash_header(struct tcp_md5sig_pool *hp,
2593 struct tcphdr *th)
2594{
2595 struct scatterlist sg;
2596 int err;
2597
2598 __sum16 old_checksum = th->check;
2599 th->check = 0;
2600 /* options aren't included in the hash */
2601 sg_init_one(&sg, th, sizeof(struct tcphdr));
2602 err = crypto_hash_update(&hp->md5_desc, &sg, sizeof(struct tcphdr));
2603 th->check = old_checksum;
2604 return err;
2605}
2606
2607EXPORT_SYMBOL(tcp_md5_hash_header);
2608
2609int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *hp,
2610 struct sk_buff *skb, unsigned header_len)
2611{
2612 struct scatterlist sg;
2613 const struct tcphdr *tp = tcp_hdr(skb);
2614 struct hash_desc *desc = &hp->md5_desc;
2615 unsigned i;
2616 const unsigned head_data_len = skb_headlen(skb) > header_len ?
2617 skb_headlen(skb) - header_len : 0;
2618 const struct skb_shared_info *shi = skb_shinfo(skb);
2619
2620 sg_init_table(&sg, 1);
2621
2622 sg_set_buf(&sg, ((u8 *) tp) + header_len, head_data_len);
2623 if (crypto_hash_update(desc, &sg, head_data_len))
2624 return 1;
2625
2626 for (i = 0; i < shi->nr_frags; ++i) {
2627 const struct skb_frag_struct *f = &shi->frags[i];
2628 sg_set_page(&sg, f->page, f->size, f->page_offset);
2629 if (crypto_hash_update(desc, &sg, f->size))
2630 return 1;
2631 }
2632
2633 return 0;
2634}
2635
2636EXPORT_SYMBOL(tcp_md5_hash_skb_data);
2637
2638int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, struct tcp_md5sig_key *key)
2639{
2640 struct scatterlist sg;
2641
2642 sg_init_one(&sg, key->key, key->keylen);
2643 return crypto_hash_update(&hp->md5_desc, &sg, key->keylen);
2644}
2645
2646EXPORT_SYMBOL(tcp_md5_hash_key);
2647
2593#endif 2648#endif
2594 2649
2595void tcp_done(struct sock *sk) 2650void tcp_done(struct sock *sk)
2596{ 2651{
2597 if(sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV) 2652 if(sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV)
2598 TCP_INC_STATS_BH(TCP_MIB_ATTEMPTFAILS); 2653 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_ATTEMPTFAILS);
2599 2654
2600 tcp_set_state(sk, TCP_CLOSE); 2655 tcp_set_state(sk, TCP_CLOSE);
2601 tcp_clear_xmit_timers(sk); 2656 tcp_clear_xmit_timers(sk);
@@ -2732,4 +2787,3 @@ EXPORT_SYMBOL(tcp_splice_read);
2732EXPORT_SYMBOL(tcp_sendpage); 2787EXPORT_SYMBOL(tcp_sendpage);
2733EXPORT_SYMBOL(tcp_setsockopt); 2788EXPORT_SYMBOL(tcp_setsockopt);
2734EXPORT_SYMBOL(tcp_shutdown); 2789EXPORT_SYMBOL(tcp_shutdown);
2735EXPORT_SYMBOL(tcp_statistics);