blob: 0ea10eefa60f005911b0849dd14218a1eb9ccf1d [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Implementation of the Transmission Control Protocol(TCP).
7 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * IPv4 specific functions
9 *
10 *
11 * code split from:
12 * linux/ipv4/tcp.c
13 * linux/ipv4/tcp_input.c
14 * linux/ipv4/tcp_output.c
15 *
16 * See tcp.c for author information
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 */
23
24/*
25 * Changes:
26 * David S. Miller : New socket lookup architecture.
27 * This code is dedicated to John Dyson.
28 * David S. Miller : Change semantics of established hash,
29 * half is devoted to TIME_WAIT sockets
30 * and the rest go in the other half.
31 * Andi Kleen : Add support for syncookies and fixed
32 * some bugs: ip options weren't passed to
33 * the TCP layer, missed a check for an
34 * ACK bit.
35 * Andi Kleen : Implemented fast path mtu discovery.
36 * Fixed many serious bugs in the
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070037 * request_sock handling and moved
Linus Torvalds1da177e2005-04-16 15:20:36 -070038 * most of it into the af independent code.
39 * Added tail drop and some other bugfixes.
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -080040 * Added new listen semantics.
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 * Mike McLagan : Routing by source
42 * Juan Jose Ciarlante: ip_dynaddr bits
43 * Andi Kleen: various fixes.
44 * Vitaly E. Lavrov : Transparent proxy revived after year
45 * coma.
46 * Andi Kleen : Fix new listen.
47 * Andi Kleen : Fix accept error reporting.
48 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
49 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
50 * a single port at the same time.
51 */
52
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
Herbert Xueb4dea52008-12-29 23:04:08 -080054#include <linux/bottom_half.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#include <linux/types.h>
56#include <linux/fcntl.h>
57#include <linux/module.h>
58#include <linux/random.h>
59#include <linux/cache.h>
60#include <linux/jhash.h>
61#include <linux/init.h>
62#include <linux/times.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090063#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020065#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066#include <net/icmp.h>
Arnaldo Carvalho de Melo304a1612005-08-09 19:59:20 -070067#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <net/tcp.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030069#include <net/transp_v6.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#include <net/ipv6.h>
71#include <net/inet_common.h>
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -080072#include <net/timewait_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#include <net/xfrm.h>
Chris Leech1a2449a2006-05-23 18:05:53 -070074#include <net/netdma.h>
David S. Miller6e5714e2011-08-03 20:50:44 -070075#include <net/secure_seq.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076
77#include <linux/inet.h>
78#include <linux/ipv6.h>
79#include <linux/stddef.h>
80#include <linux/proc_fs.h>
81#include <linux/seq_file.h>
82
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -080083#include <linux/crypto.h>
84#include <linux/scatterlist.h>
85
Brian Haleyab32ea52006-09-22 14:15:41 -070086int sysctl_tcp_tw_reuse __read_mostly;
87int sysctl_tcp_low_latency __read_mostly;
Eric Dumazet4bc2f182010-07-09 21:22:10 +000088EXPORT_SYMBOL(sysctl_tcp_low_latency);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -080091#ifdef CONFIG_TCP_MD5SIG
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -020092static struct tcp_md5sig_key *tcp_v4_md5_do_lookup(struct sock *sk,
93 __be32 addr);
Adam Langley49a72df2008-07-19 00:01:42 -070094static int tcp_v4_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
Eric Dumazet318cf7a2011-10-24 02:46:04 -040095 __be32 daddr, __be32 saddr, const struct tcphdr *th);
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +090096#else
97static inline
98struct tcp_md5sig_key *tcp_v4_md5_do_lookup(struct sock *sk, __be32 addr)
99{
100 return NULL;
101}
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800102#endif
103
Eric Dumazet5caea4e2008-11-20 00:40:07 -0800104struct inet_hashinfo tcp_hashinfo;
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000105EXPORT_SYMBOL(tcp_hashinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400107static inline __u32 tcp_v4_init_sequence(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700109 return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
110 ip_hdr(skb)->saddr,
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -0700111 tcp_hdr(skb)->dest,
112 tcp_hdr(skb)->source);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113}
114
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800115int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp)
116{
117 const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw);
118 struct tcp_sock *tp = tcp_sk(sk);
119
120 /* With PAWS, it is safe from the viewpoint
121 of data integrity. Even without PAWS it is safe provided sequence
122 spaces do not overlap i.e. at data rates <= 80Mbit/sec.
123
124 Actually, the idea is close to VJ's one, only timestamp cache is
125 held not per host, but per port pair and TW bucket is used as state
126 holder.
127
128 If TW bucket has been already destroyed we fall back to VJ's scheme
129 and use initial timestamp retrieved from peer table.
130 */
131 if (tcptw->tw_ts_recent_stamp &&
132 (twp == NULL || (sysctl_tcp_tw_reuse &&
James Morris9d729f72007-03-04 16:12:44 -0800133 get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800134 tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
135 if (tp->write_seq == 0)
136 tp->write_seq = 1;
137 tp->rx_opt.ts_recent = tcptw->tw_ts_recent;
138 tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp;
139 sock_hold(sktw);
140 return 1;
141 }
142
143 return 0;
144}
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800145EXPORT_SYMBOL_GPL(tcp_twsk_unique);
146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147/* This will initiate an outgoing connection. */
148int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
149{
David S. Miller2d7192d2011-04-26 13:28:44 -0700150 struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 struct inet_sock *inet = inet_sk(sk);
152 struct tcp_sock *tp = tcp_sk(sk);
David S. Millerdca8b082011-02-24 13:38:12 -0800153 __be16 orig_sport, orig_dport;
Al Virobada8ad2006-09-26 21:27:15 -0700154 __be32 daddr, nexthop;
David S. Millerda905bd2011-05-06 16:11:19 -0700155 struct flowi4 *fl4;
David S. Miller2d7192d2011-04-26 13:28:44 -0700156 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 int err;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000158 struct ip_options_rcu *inet_opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
160 if (addr_len < sizeof(struct sockaddr_in))
161 return -EINVAL;
162
163 if (usin->sin_family != AF_INET)
164 return -EAFNOSUPPORT;
165
166 nexthop = daddr = usin->sin_addr.s_addr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000167 inet_opt = rcu_dereference_protected(inet->inet_opt,
168 sock_owned_by_user(sk));
169 if (inet_opt && inet_opt->opt.srr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 if (!daddr)
171 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000172 nexthop = inet_opt->opt.faddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 }
174
David S. Millerdca8b082011-02-24 13:38:12 -0800175 orig_sport = inet->inet_sport;
176 orig_dport = usin->sin_port;
David S. Millerda905bd2011-05-06 16:11:19 -0700177 fl4 = &inet->cork.fl.u.ip4;
178 rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
David S. Millerb23dd4f2011-03-02 14:31:35 -0800179 RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
180 IPPROTO_TCP,
181 orig_sport, orig_dport, sk, true);
182 if (IS_ERR(rt)) {
183 err = PTR_ERR(rt);
184 if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700185 IP_INC_STATS_BH(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800186 return err;
Wei Dong584bdf82007-05-31 22:49:28 -0700187 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
189 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
190 ip_rt_put(rt);
191 return -ENETUNREACH;
192 }
193
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000194 if (!inet_opt || !inet_opt->opt.srr)
David S. Millerda905bd2011-05-06 16:11:19 -0700195 daddr = fl4->daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000197 if (!inet->inet_saddr)
David S. Millerda905bd2011-05-06 16:11:19 -0700198 inet->inet_saddr = fl4->saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000199 inet->inet_rcv_saddr = inet->inet_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000201 if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 /* Reset inherited state */
203 tp->rx_opt.ts_recent = 0;
204 tp->rx_opt.ts_recent_stamp = 0;
205 tp->write_seq = 0;
206 }
207
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700208 if (tcp_death_row.sysctl_tw_recycle &&
David S. Millerda905bd2011-05-06 16:11:19 -0700209 !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr) {
David S. Millered2361e2011-05-18 18:38:54 -0400210 struct inet_peer *peer = rt_get_peer(rt, fl4->daddr);
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200211 /*
212 * VJ's idea. We save last timestamp seen from
213 * the destination in peer table, when entering state
214 * TIME-WAIT * and initialize rx_opt.ts_recent from it,
215 * when trying new connection.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 */
Eric Dumazet317fe0e2010-06-16 04:52:13 +0000217 if (peer) {
218 inet_peer_refcheck(peer);
219 if ((u32)get_seconds() - peer->tcp_ts_stamp <= TCP_PAWS_MSL) {
220 tp->rx_opt.ts_recent_stamp = peer->tcp_ts_stamp;
221 tp->rx_opt.ts_recent = peer->tcp_ts;
222 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 }
224 }
225
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000226 inet->inet_dport = usin->sin_port;
227 inet->inet_daddr = daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800229 inet_csk(sk)->icsk_ext_hdr_len = 0;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000230 if (inet_opt)
231 inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232
William Allen Simpsonbee7ca92009-11-10 09:51:18 +0000233 tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234
235 /* Socket identity is still unknown (sport may be zero).
236 * However we set state to SYN-SENT and not releasing socket
237 * lock select source port, enter ourselves into the hash tables and
238 * complete initialization after this.
239 */
240 tcp_set_state(sk, TCP_SYN_SENT);
Arnaldo Carvalho de Meloa7f5e7f2005-12-13 23:25:31 -0800241 err = inet_hash_connect(&tcp_death_row, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 if (err)
243 goto failure;
244
David S. Millerda905bd2011-05-06 16:11:19 -0700245 rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
David S. Millerb23dd4f2011-03-02 14:31:35 -0800246 inet->inet_sport, inet->inet_dport, sk);
247 if (IS_ERR(rt)) {
248 err = PTR_ERR(rt);
249 rt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 goto failure;
David S. Millerb23dd4f2011-03-02 14:31:35 -0800251 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 /* OK, now commit destination to socket. */
Herbert Xubcd76112006-06-30 13:36:35 -0700253 sk->sk_gso_type = SKB_GSO_TCPV4;
Changli Gaod8d1f302010-06-10 23:31:35 -0700254 sk_setup_caps(sk, &rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256 if (!tp->write_seq)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000257 tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr,
258 inet->inet_daddr,
259 inet->inet_sport,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 usin->sin_port);
261
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000262 inet->inet_id = tp->write_seq ^ jiffies;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
264 err = tcp_connect(sk);
265 rt = NULL;
266 if (err)
267 goto failure;
268
269 return 0;
270
271failure:
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200272 /*
273 * This unhashes the socket and releases the local port,
274 * if necessary.
275 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 tcp_set_state(sk, TCP_CLOSE);
277 ip_rt_put(rt);
278 sk->sk_route_caps = 0;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000279 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 return err;
281}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000282EXPORT_SYMBOL(tcp_v4_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284/*
285 * This routine does path mtu discovery as defined in RFC1191.
286 */
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000287static void do_pmtu_discovery(struct sock *sk, const struct iphdr *iph, u32 mtu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
289 struct dst_entry *dst;
290 struct inet_sock *inet = inet_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291
292 /* We are not interested in TCP_LISTEN and open_requests (SYN-ACKs
293 * send out by Linux are always <576bytes so they should go through
294 * unfragmented).
295 */
296 if (sk->sk_state == TCP_LISTEN)
297 return;
298
299 /* We don't check in the destentry if pmtu discovery is forbidden
300 * on this route. We just assume that no packet_to_big packets
301 * are send back when pmtu discovery is not active.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900302 * There is a small race when the user changes this flag in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 * route, but I think that's acceptable.
304 */
305 if ((dst = __sk_dst_check(sk, 0)) == NULL)
306 return;
307
308 dst->ops->update_pmtu(dst, mtu);
309
310 /* Something is about to be wrong... Remember soft error
311 * for the case, if this connection will not able to recover.
312 */
313 if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
314 sk->sk_err_soft = EMSGSIZE;
315
316 mtu = dst_mtu(dst);
317
318 if (inet->pmtudisc != IP_PMTUDISC_DONT &&
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800319 inet_csk(sk)->icsk_pmtu_cookie > mtu) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 tcp_sync_mss(sk, mtu);
321
322 /* Resend the TCP packet because it's
323 * clear that the old packet has been
324 * dropped. This is the new "fast" path mtu
325 * discovery.
326 */
327 tcp_simple_retransmit(sk);
328 } /* else let the usual retransmit timer handle it */
329}
330
331/*
332 * This routine is called by the ICMP module when it gets some
333 * sort of error condition. If err < 0 then the socket should
334 * be closed and the error returned to the user. If err > 0
335 * it's just the icmp type << 8 | icmp code. After adjustment
336 * header points to the first 8 bytes of the tcp header. We need
337 * to find the appropriate port.
338 *
339 * The locking strategy used here is very "optimistic". When
340 * someone else accesses the socket the ICMP is just dropped
341 * and for some paths there is no check at all.
342 * A more general error queue to queue errors for later handling
343 * is probably better.
344 *
345 */
346
Damian Lukowski4d1a2d92009-08-26 00:16:27 +0000347void tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000349 const struct iphdr *iph = (const struct iphdr *)icmp_skb->data;
Damian Lukowski4d1a2d92009-08-26 00:16:27 +0000350 struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2));
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000351 struct inet_connection_sock *icsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 struct tcp_sock *tp;
353 struct inet_sock *inet;
Damian Lukowski4d1a2d92009-08-26 00:16:27 +0000354 const int type = icmp_hdr(icmp_skb)->type;
355 const int code = icmp_hdr(icmp_skb)->code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 struct sock *sk;
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000357 struct sk_buff *skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 __u32 seq;
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000359 __u32 remaining;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 int err;
Damian Lukowski4d1a2d92009-08-26 00:16:27 +0000361 struct net *net = dev_net(icmp_skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
Damian Lukowski4d1a2d92009-08-26 00:16:27 +0000363 if (icmp_skb->len < (iph->ihl << 2) + 8) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700364 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 return;
366 }
367
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700368 sk = inet_lookup(net, &tcp_hashinfo, iph->daddr, th->dest,
Damian Lukowski4d1a2d92009-08-26 00:16:27 +0000369 iph->saddr, th->source, inet_iif(icmp_skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 if (!sk) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700371 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 return;
373 }
374 if (sk->sk_state == TCP_TIME_WAIT) {
YOSHIFUJI Hideaki9469c7b2006-10-10 19:41:46 -0700375 inet_twsk_put(inet_twsk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 return;
377 }
378
379 bh_lock_sock(sk);
380 /* If too many ICMPs get dropped on busy
381 * servers this needs to be solved differently.
382 */
383 if (sock_owned_by_user(sk))
Pavel Emelyanovde0744a2008-07-16 20:31:16 -0700384 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
386 if (sk->sk_state == TCP_CLOSE)
387 goto out;
388
stephen hemminger97e3ecd12010-03-18 11:27:32 +0000389 if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
390 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
391 goto out;
392 }
393
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000394 icsk = inet_csk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 tp = tcp_sk(sk);
396 seq = ntohl(th->seq);
397 if (sk->sk_state != TCP_LISTEN &&
398 !between(seq, tp->snd_una, tp->snd_nxt)) {
Pavel Emelyanovde0744a2008-07-16 20:31:16 -0700399 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 goto out;
401 }
402
403 switch (type) {
404 case ICMP_SOURCE_QUENCH:
405 /* Just silently ignore these. */
406 goto out;
407 case ICMP_PARAMETERPROB:
408 err = EPROTO;
409 break;
410 case ICMP_DEST_UNREACH:
411 if (code > NR_ICMP_UNREACH)
412 goto out;
413
414 if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
415 if (!sock_owned_by_user(sk))
416 do_pmtu_discovery(sk, iph, info);
417 goto out;
418 }
419
420 err = icmp_err_convert[code].errno;
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000421 /* check if icmp_skb allows revert of backoff
422 * (see draft-zimmermann-tcp-lcd) */
423 if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH)
424 break;
425 if (seq != tp->snd_una || !icsk->icsk_retransmits ||
426 !icsk->icsk_backoff)
427 break;
428
David S. Miller8f49c272010-11-12 13:35:00 -0800429 if (sock_owned_by_user(sk))
430 break;
431
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000432 icsk->icsk_backoff--;
Jerry Chu9ad7c042011-06-08 11:08:38 +0000433 inet_csk(sk)->icsk_rto = (tp->srtt ? __tcp_set_rto(tp) :
434 TCP_TIMEOUT_INIT) << icsk->icsk_backoff;
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000435 tcp_bound_rto(sk);
436
437 skb = tcp_write_queue_head(sk);
438 BUG_ON(!skb);
439
440 remaining = icsk->icsk_rto - min(icsk->icsk_rto,
441 tcp_time_stamp - TCP_SKB_CB(skb)->when);
442
443 if (remaining) {
444 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
445 remaining, TCP_RTO_MAX);
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000446 } else {
447 /* RTO revert clocked out retransmission.
448 * Will retransmit now */
449 tcp_retransmit_timer(sk);
450 }
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 break;
453 case ICMP_TIME_EXCEEDED:
454 err = EHOSTUNREACH;
455 break;
456 default:
457 goto out;
458 }
459
460 switch (sk->sk_state) {
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700461 struct request_sock *req, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 case TCP_LISTEN:
463 if (sock_owned_by_user(sk))
464 goto out;
465
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700466 req = inet_csk_search_req(sk, &prev, th->dest,
467 iph->daddr, iph->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 if (!req)
469 goto out;
470
471 /* ICMPs are not backlogged, hence we cannot get
472 an established socket here.
473 */
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700474 WARN_ON(req->sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700476 if (seq != tcp_rsk(req)->snt_isn) {
Pavel Emelyanovde0744a2008-07-16 20:31:16 -0700477 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 goto out;
479 }
480
481 /*
482 * Still in SYN_RECV, just remove it silently.
483 * There is no good way to pass the error to the newly
484 * created socket, and POSIX does not want network
485 * errors returned from accept().
486 */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700487 inet_csk_reqsk_queue_drop(sk, req, prev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 goto out;
489
490 case TCP_SYN_SENT:
491 case TCP_SYN_RECV: /* Cannot happen.
492 It can f.e. if SYNs crossed.
493 */
494 if (!sock_owned_by_user(sk)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 sk->sk_err = err;
496
497 sk->sk_error_report(sk);
498
499 tcp_done(sk);
500 } else {
501 sk->sk_err_soft = err;
502 }
503 goto out;
504 }
505
506 /* If we've already connected we will keep trying
507 * until we time out, or the user gives up.
508 *
509 * rfc1122 4.2.3.9 allows to consider as hard errors
510 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
511 * but it is obsoleted by pmtu discovery).
512 *
513 * Note, that in modern internet, where routing is unreliable
514 * and in each dark corner broken firewalls sit, sending random
515 * errors ordered by their masters even this two messages finally lose
516 * their original sense (even Linux sends invalid PORT_UNREACHs)
517 *
518 * Now we are in compliance with RFCs.
519 * --ANK (980905)
520 */
521
522 inet = inet_sk(sk);
523 if (!sock_owned_by_user(sk) && inet->recverr) {
524 sk->sk_err = err;
525 sk->sk_error_report(sk);
526 } else { /* Only an error on timeout */
527 sk->sk_err_soft = err;
528 }
529
530out:
531 bh_unlock_sock(sk);
532 sock_put(sk);
533}
534
Herbert Xu419f9f82010-04-11 02:15:53 +0000535static void __tcp_v4_send_check(struct sk_buff *skb,
536 __be32 saddr, __be32 daddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -0700538 struct tcphdr *th = tcp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Patrick McHardy84fa7932006-08-29 16:44:56 -0700540 if (skb->ip_summed == CHECKSUM_PARTIAL) {
Herbert Xu419f9f82010-04-11 02:15:53 +0000541 th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0);
Herbert Xu663ead32007-04-09 11:59:07 -0700542 skb->csum_start = skb_transport_header(skb) - skb->head;
Al Viroff1dcad2006-11-20 18:07:29 -0800543 skb->csum_offset = offsetof(struct tcphdr, check);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 } else {
Herbert Xu419f9f82010-04-11 02:15:53 +0000545 th->check = tcp_v4_check(skb->len, saddr, daddr,
Joe Perches07f07572008-11-19 15:44:53 -0800546 csum_partial(th,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 th->doff << 2,
548 skb->csum));
549 }
550}
551
Herbert Xu419f9f82010-04-11 02:15:53 +0000552/* This routine computes an IPv4 TCP checksum. */
Herbert Xubb296242010-04-11 02:15:55 +0000553void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb)
Herbert Xu419f9f82010-04-11 02:15:53 +0000554{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400555 const struct inet_sock *inet = inet_sk(sk);
Herbert Xu419f9f82010-04-11 02:15:53 +0000556
557 __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr);
558}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000559EXPORT_SYMBOL(tcp_v4_send_check);
Herbert Xu419f9f82010-04-11 02:15:53 +0000560
Herbert Xua430a432006-07-08 13:34:56 -0700561int tcp_v4_gso_send_check(struct sk_buff *skb)
562{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700563 const struct iphdr *iph;
Herbert Xua430a432006-07-08 13:34:56 -0700564 struct tcphdr *th;
565
566 if (!pskb_may_pull(skb, sizeof(*th)))
567 return -EINVAL;
568
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700569 iph = ip_hdr(skb);
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -0700570 th = tcp_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -0700571
572 th->check = 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -0700573 skb->ip_summed = CHECKSUM_PARTIAL;
Herbert Xu419f9f82010-04-11 02:15:53 +0000574 __tcp_v4_send_check(skb, iph->saddr, iph->daddr);
Herbert Xua430a432006-07-08 13:34:56 -0700575 return 0;
576}
577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578/*
579 * This routine will send an RST to the other tcp.
580 *
581 * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.)
582 * for reset.
583 * Answer: if a packet caused RST, it is not for a socket
584 * existing in our system, if it is matched to a socket,
585 * it is just duplicate segment or bug in other side's TCP.
586 * So that we build reply only basing on parameters
587 * arrived with segment.
588 * Exception: precedence violation. We do not implement it in any case.
589 */
590
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800591static void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400593 const struct tcphdr *th = tcp_hdr(skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800594 struct {
595 struct tcphdr th;
596#ifdef CONFIG_TCP_MD5SIG
Al Viro714e85b2006-11-14 20:51:49 -0800597 __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)];
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800598#endif
599 } rep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 struct ip_reply_arg arg;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800601#ifdef CONFIG_TCP_MD5SIG
602 struct tcp_md5sig_key *key;
603#endif
Pavel Emelyanova86b1e32008-07-16 20:20:58 -0700604 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
606 /* Never send a reset in response to a reset. */
607 if (th->rst)
608 return;
609
Eric Dumazet511c3f92009-06-02 05:14:27 +0000610 if (skb_rtable(skb)->rt_type != RTN_LOCAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 return;
612
613 /* Swap the send and the receive. */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800614 memset(&rep, 0, sizeof(rep));
615 rep.th.dest = th->source;
616 rep.th.source = th->dest;
617 rep.th.doff = sizeof(struct tcphdr) / 4;
618 rep.th.rst = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619
620 if (th->ack) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800621 rep.th.seq = th->ack_seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 } else {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800623 rep.th.ack = 1;
624 rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin +
625 skb->len - (th->doff << 2));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
627
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200628 memset(&arg, 0, sizeof(arg));
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800629 arg.iov[0].iov_base = (unsigned char *)&rep;
630 arg.iov[0].iov_len = sizeof(rep.th);
631
632#ifdef CONFIG_TCP_MD5SIG
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700633 key = sk ? tcp_v4_md5_do_lookup(sk, ip_hdr(skb)->daddr) : NULL;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800634 if (key) {
635 rep.opt[0] = htonl((TCPOPT_NOP << 24) |
636 (TCPOPT_NOP << 16) |
637 (TCPOPT_MD5SIG << 8) |
638 TCPOLEN_MD5SIG);
639 /* Update length and the length the header thinks exists */
640 arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED;
641 rep.th.doff = arg.iov[0].iov_len / 4;
642
Adam Langley49a72df2008-07-19 00:01:42 -0700643 tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1],
Ilpo Järvinen78e645cb2008-10-09 14:37:47 -0700644 key, ip_hdr(skb)->saddr,
645 ip_hdr(skb)->daddr, &rep.th);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800646 }
647#endif
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700648 arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
649 ip_hdr(skb)->saddr, /* XXX */
Ilpo Järvinen52cd5752008-10-08 11:34:06 -0700650 arg.iov[0].iov_len, IPPROTO_TCP, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 arg.csumoffset = offsetof(struct tcphdr, check) / 2;
KOVACS Krisztian88ef4a52008-10-01 07:41:00 -0700652 arg.flags = (sk && inet_sk(sk)->transparent) ? IP_REPLY_ARG_NOSRCCHECK : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Eric Dumazetadf30902009-06-02 05:19:30 +0000654 net = dev_net(skb_dst(skb)->dev);
Eric Dumazet66b13d92011-10-24 03:06:21 -0400655 arg.tos = ip_hdr(skb)->tos;
David S. Miller0a5ebb82011-05-09 13:22:43 -0700656 ip_send_reply(net->ipv4.tcp_sock, skb, ip_hdr(skb)->saddr,
Denis V. Lunev7feb49c2008-04-03 14:32:00 -0700657 &arg, arg.iov[0].iov_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Pavel Emelyanov63231bd2008-07-16 20:22:25 -0700659 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
660 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661}
662
663/* The code following below sending ACKs in SYN-RECV and TIME-WAIT states
664 outside socket context is ugly, certainly. What can I do?
665 */
666
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900667static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
668 u32 win, u32 ts, int oif,
KOVACS Krisztian88ef4a52008-10-01 07:41:00 -0700669 struct tcp_md5sig_key *key,
Eric Dumazet66b13d92011-10-24 03:06:21 -0400670 int reply_flags, u8 tos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400672 const struct tcphdr *th = tcp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 struct {
674 struct tcphdr th;
Al Viro714e85b2006-11-14 20:51:49 -0800675 __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800676#ifdef CONFIG_TCP_MD5SIG
Al Viro714e85b2006-11-14 20:51:49 -0800677 + (TCPOLEN_MD5SIG_ALIGNED >> 2)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800678#endif
679 ];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 } rep;
681 struct ip_reply_arg arg;
Eric Dumazetadf30902009-06-02 05:19:30 +0000682 struct net *net = dev_net(skb_dst(skb)->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
684 memset(&rep.th, 0, sizeof(struct tcphdr));
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200685 memset(&arg, 0, sizeof(arg));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
687 arg.iov[0].iov_base = (unsigned char *)&rep;
688 arg.iov[0].iov_len = sizeof(rep.th);
689 if (ts) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800690 rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
691 (TCPOPT_TIMESTAMP << 8) |
692 TCPOLEN_TIMESTAMP);
693 rep.opt[1] = htonl(tcp_time_stamp);
694 rep.opt[2] = htonl(ts);
Craig Schlentercb48cfe2007-01-09 00:11:15 -0800695 arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 }
697
698 /* Swap the send and the receive. */
699 rep.th.dest = th->source;
700 rep.th.source = th->dest;
701 rep.th.doff = arg.iov[0].iov_len / 4;
702 rep.th.seq = htonl(seq);
703 rep.th.ack_seq = htonl(ack);
704 rep.th.ack = 1;
705 rep.th.window = htons(win);
706
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800707#ifdef CONFIG_TCP_MD5SIG
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800708 if (key) {
709 int offset = (ts) ? 3 : 0;
710
711 rep.opt[offset++] = htonl((TCPOPT_NOP << 24) |
712 (TCPOPT_NOP << 16) |
713 (TCPOPT_MD5SIG << 8) |
714 TCPOLEN_MD5SIG);
715 arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED;
716 rep.th.doff = arg.iov[0].iov_len/4;
717
Adam Langley49a72df2008-07-19 00:01:42 -0700718 tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset],
Adam Langley90b7e112008-07-31 20:49:48 -0700719 key, ip_hdr(skb)->saddr,
720 ip_hdr(skb)->daddr, &rep.th);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800721 }
722#endif
KOVACS Krisztian88ef4a52008-10-01 07:41:00 -0700723 arg.flags = reply_flags;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700724 arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
725 ip_hdr(skb)->saddr, /* XXX */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 arg.iov[0].iov_len, IPPROTO_TCP, 0);
727 arg.csumoffset = offsetof(struct tcphdr, check) / 2;
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900728 if (oif)
729 arg.bound_dev_if = oif;
Eric Dumazet66b13d92011-10-24 03:06:21 -0400730 arg.tos = tos;
David S. Miller0a5ebb82011-05-09 13:22:43 -0700731 ip_send_reply(net->ipv4.tcp_sock, skb, ip_hdr(skb)->saddr,
Denis V. Lunev7feb49c2008-04-03 14:32:00 -0700732 &arg, arg.iov[0].iov_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733
Pavel Emelyanov63231bd2008-07-16 20:22:25 -0700734 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735}
736
737static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
738{
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700739 struct inet_timewait_sock *tw = inet_twsk(sk);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800740 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900742 tcp_v4_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200743 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900744 tcptw->tw_ts_recent,
745 tw->tw_bound_dev_if,
KOVACS Krisztian88ef4a52008-10-01 07:41:00 -0700746 tcp_twsk_md5_key(tcptw),
Eric Dumazet66b13d92011-10-24 03:06:21 -0400747 tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0,
748 tw->tw_tos
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900749 );
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700751 inet_twsk_put(tw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752}
753
Gui Jianfeng6edafaa2008-08-06 23:50:04 -0700754static void tcp_v4_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200755 struct request_sock *req)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756{
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900757 tcp_v4_send_ack(skb, tcp_rsk(req)->snt_isn + 1,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800758 tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd,
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +0900759 req->ts_recent,
760 0,
KOVACS Krisztian88ef4a52008-10-01 07:41:00 -0700761 tcp_v4_md5_do_lookup(sk, ip_hdr(skb)->daddr),
Eric Dumazet66b13d92011-10-24 03:06:21 -0400762 inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0,
763 ip_hdr(skb)->tos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764}
765
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766/*
Kris Katterjohn9bf1d832008-02-17 22:29:19 -0800767 * Send a SYN-ACK after having received a SYN.
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700768 * This still operates on a request_sock only, not on a big
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 * socket.
770 */
Octavian Purdila72659ec2010-01-17 19:09:39 -0800771static int tcp_v4_send_synack(struct sock *sk, struct dst_entry *dst,
772 struct request_sock *req,
773 struct request_values *rvp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700775 const struct inet_request_sock *ireq = inet_rsk(req);
David S. Miller6bd023f2011-05-18 18:32:03 -0400776 struct flowi4 fl4;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 int err = -1;
778 struct sk_buff * skb;
779
780 /* First, grab a route. */
David S. Miller6bd023f2011-05-18 18:32:03 -0400781 if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL)
Denis V. Lunevfd80eb92008-02-29 11:43:03 -0800782 return -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
William Allen Simpsone6b4d112009-12-02 18:07:39 +0000784 skb = tcp_make_synack(sk, dst, req, rvp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785
786 if (skb) {
Herbert Xu419f9f82010-04-11 02:15:53 +0000787 __tcp_v4_send_check(skb, ireq->loc_addr, ireq->rmt_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700789 err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr,
790 ireq->rmt_addr,
791 ireq->opt);
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -0200792 err = net_xmit_eval(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 }
794
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 dst_release(dst);
796 return err;
797}
798
Octavian Purdila72659ec2010-01-17 19:09:39 -0800799static int tcp_v4_rtx_synack(struct sock *sk, struct request_sock *req,
William Allen Simpsone6b4d112009-12-02 18:07:39 +0000800 struct request_values *rvp)
Denis V. Lunevfd80eb92008-02-29 11:43:03 -0800801{
Octavian Purdila72659ec2010-01-17 19:09:39 -0800802 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
803 return tcp_v4_send_synack(sk, NULL, req, rvp);
Denis V. Lunevfd80eb92008-02-29 11:43:03 -0800804}
805
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806/*
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700807 * IPv4 request_sock destructor.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 */
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700809static void tcp_v4_reqsk_destructor(struct request_sock *req)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Jesper Juhla51482b2005-11-08 09:41:34 -0800811 kfree(inet_rsk(req)->opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812}
813
Eric Dumazet946cedc2011-08-30 03:21:44 +0000814/*
815 * Return 1 if a syncookie should be sent
816 */
817int tcp_syn_flood_action(struct sock *sk,
818 const struct sk_buff *skb,
819 const char *proto)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820{
Eric Dumazet946cedc2011-08-30 03:21:44 +0000821 const char *msg = "Dropping request";
822 int want_cookie = 0;
823 struct listen_sock *lopt;
824
825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Florian Westphal2a1d4bd2010-06-03 00:43:12 +0000827#ifdef CONFIG_SYN_COOKIES
Eric Dumazet946cedc2011-08-30 03:21:44 +0000828 if (sysctl_tcp_syncookies) {
Florian Westphal2a1d4bd2010-06-03 00:43:12 +0000829 msg = "Sending cookies";
Eric Dumazet946cedc2011-08-30 03:21:44 +0000830 want_cookie = 1;
831 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES);
832 } else
Arnaldo Carvalho de Melo80e40da2006-01-04 01:58:06 -0200833#endif
Eric Dumazet946cedc2011-08-30 03:21:44 +0000834 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP);
Florian Westphal2a1d4bd2010-06-03 00:43:12 +0000835
Eric Dumazet946cedc2011-08-30 03:21:44 +0000836 lopt = inet_csk(sk)->icsk_accept_queue.listen_opt;
837 if (!lopt->synflood_warned) {
838 lopt->synflood_warned = 1;
839 pr_info("%s: Possible SYN flooding on port %d. %s. "
840 " Check SNMP counters.\n",
841 proto, ntohs(tcp_hdr(skb)->dest), msg);
842 }
843 return want_cookie;
Florian Westphal2a1d4bd2010-06-03 00:43:12 +0000844}
Eric Dumazet946cedc2011-08-30 03:21:44 +0000845EXPORT_SYMBOL(tcp_syn_flood_action);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
847/*
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700848 * Save and compile IPv4 options into the request_sock if needed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000850static struct ip_options_rcu *tcp_v4_save_options(struct sock *sk,
851 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852{
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000853 const struct ip_options *opt = &(IPCB(skb)->opt);
854 struct ip_options_rcu *dopt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
856 if (opt && opt->optlen) {
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000857 int opt_size = sizeof(*dopt) + opt->optlen;
858
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 dopt = kmalloc(opt_size, GFP_ATOMIC);
860 if (dopt) {
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000861 if (ip_options_echo(&dopt->opt, skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 kfree(dopt);
863 dopt = NULL;
864 }
865 }
866 }
867 return dopt;
868}
869
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800870#ifdef CONFIG_TCP_MD5SIG
871/*
872 * RFC2385 MD5 checksumming requires a mapping of
873 * IP address->MD5 Key.
874 * We need to maintain these in the sk structure.
875 */
876
877/* Find the Key structure for an address. */
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200878static struct tcp_md5sig_key *
879 tcp_v4_md5_do_lookup(struct sock *sk, __be32 addr)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800880{
881 struct tcp_sock *tp = tcp_sk(sk);
882 int i;
883
884 if (!tp->md5sig_info || !tp->md5sig_info->entries4)
885 return NULL;
886 for (i = 0; i < tp->md5sig_info->entries4; i++) {
887 if (tp->md5sig_info->keys4[i].addr == addr)
David S. Millerf8ab18d2007-09-28 15:18:35 -0700888 return &tp->md5sig_info->keys4[i].base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800889 }
890 return NULL;
891}
892
893struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
894 struct sock *addr_sk)
895{
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000896 return tcp_v4_md5_do_lookup(sk, inet_sk(addr_sk)->inet_daddr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800897}
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800898EXPORT_SYMBOL(tcp_v4_md5_lookup);
899
Adrian Bunkf5b99bc2006-11-30 17:22:29 -0800900static struct tcp_md5sig_key *tcp_v4_reqsk_md5_lookup(struct sock *sk,
901 struct request_sock *req)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800902{
903 return tcp_v4_md5_do_lookup(sk, inet_rsk(req)->rmt_addr);
904}
905
906/* This can be called on a newly created socket, from other files */
907int tcp_v4_md5_do_add(struct sock *sk, __be32 addr,
908 u8 *newkey, u8 newkeylen)
909{
910 /* Add Key to the list */
Matthias M. Dellwegb0a713e2007-10-29 20:55:27 -0700911 struct tcp_md5sig_key *key;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800912 struct tcp_sock *tp = tcp_sk(sk);
913 struct tcp4_md5sig_key *keys;
914
Matthias M. Dellwegb0a713e2007-10-29 20:55:27 -0700915 key = tcp_v4_md5_do_lookup(sk, addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800916 if (key) {
917 /* Pre-existing entry - just update that one. */
Matthias M. Dellwegb0a713e2007-10-29 20:55:27 -0700918 kfree(key->key);
919 key->key = newkey;
920 key->keylen = newkeylen;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800921 } else {
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -0200922 struct tcp_md5sig_info *md5sig;
923
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800924 if (!tp->md5sig_info) {
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -0200925 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info),
926 GFP_ATOMIC);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800927 if (!tp->md5sig_info) {
928 kfree(newkey);
929 return -ENOMEM;
930 }
Eric Dumazeta4654192010-05-16 00:36:33 -0700931 sk_nocaps_add(sk, NETIF_F_GSO_MASK);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800932 }
Yan, Zheng260fcbe2011-09-29 17:10:10 +0000933
934 md5sig = tp->md5sig_info;
935 if (md5sig->entries4 == 0 &&
936 tcp_alloc_md5sig_pool(sk) == NULL) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800937 kfree(newkey);
938 return -ENOMEM;
939 }
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -0200940
941 if (md5sig->alloced4 == md5sig->entries4) {
942 keys = kmalloc((sizeof(*keys) *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900943 (md5sig->entries4 + 1)), GFP_ATOMIC);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800944 if (!keys) {
945 kfree(newkey);
Yan, Zheng260fcbe2011-09-29 17:10:10 +0000946 if (md5sig->entries4 == 0)
947 tcp_free_md5sig_pool();
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800948 return -ENOMEM;
949 }
950
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -0200951 if (md5sig->entries4)
952 memcpy(keys, md5sig->keys4,
953 sizeof(*keys) * md5sig->entries4);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800954
955 /* Free old key list, and reference new one */
YOSHIFUJI Hideakia80cc202007-11-20 17:30:06 -0800956 kfree(md5sig->keys4);
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -0200957 md5sig->keys4 = keys;
958 md5sig->alloced4++;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800959 }
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -0200960 md5sig->entries4++;
David S. Millerf8ab18d2007-09-28 15:18:35 -0700961 md5sig->keys4[md5sig->entries4 - 1].addr = addr;
962 md5sig->keys4[md5sig->entries4 - 1].base.key = newkey;
963 md5sig->keys4[md5sig->entries4 - 1].base.keylen = newkeylen;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800964 }
965 return 0;
966}
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800967EXPORT_SYMBOL(tcp_v4_md5_do_add);
968
969static int tcp_v4_md5_add_func(struct sock *sk, struct sock *addr_sk,
970 u8 *newkey, u8 newkeylen)
971{
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000972 return tcp_v4_md5_do_add(sk, inet_sk(addr_sk)->inet_daddr,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800973 newkey, newkeylen);
974}
975
976int tcp_v4_md5_do_del(struct sock *sk, __be32 addr)
977{
978 struct tcp_sock *tp = tcp_sk(sk);
979 int i;
980
981 for (i = 0; i < tp->md5sig_info->entries4; i++) {
982 if (tp->md5sig_info->keys4[i].addr == addr) {
983 /* Free the key */
David S. Millerf8ab18d2007-09-28 15:18:35 -0700984 kfree(tp->md5sig_info->keys4[i].base.key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800985 tp->md5sig_info->entries4--;
986
987 if (tp->md5sig_info->entries4 == 0) {
988 kfree(tp->md5sig_info->keys4);
989 tp->md5sig_info->keys4 = NULL;
Leigh Brown8228a18d2006-12-17 17:12:30 -0800990 tp->md5sig_info->alloced4 = 0;
Yan, Zheng260fcbe2011-09-29 17:10:10 +0000991 tcp_free_md5sig_pool();
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -0200992 } else if (tp->md5sig_info->entries4 != i) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800993 /* Need to do some manipulation */
YOSHIFUJI Hideaki354faf02007-11-20 17:30:31 -0800994 memmove(&tp->md5sig_info->keys4[i],
995 &tp->md5sig_info->keys4[i+1],
996 (tp->md5sig_info->entries4 - i) *
997 sizeof(struct tcp4_md5sig_key));
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800998 }
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800999 return 0;
1000 }
1001 }
1002 return -ENOENT;
1003}
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001004EXPORT_SYMBOL(tcp_v4_md5_do_del);
1005
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -02001006static void tcp_v4_clear_md5_list(struct sock *sk)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001007{
1008 struct tcp_sock *tp = tcp_sk(sk);
1009
1010 /* Free each key, then the set of key keys,
1011 * the crypto element, and then decrement our
1012 * hold on the last resort crypto.
1013 */
1014 if (tp->md5sig_info->entries4) {
1015 int i;
1016 for (i = 0; i < tp->md5sig_info->entries4; i++)
David S. Millerf8ab18d2007-09-28 15:18:35 -07001017 kfree(tp->md5sig_info->keys4[i].base.key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001018 tp->md5sig_info->entries4 = 0;
1019 tcp_free_md5sig_pool();
1020 }
1021 if (tp->md5sig_info->keys4) {
1022 kfree(tp->md5sig_info->keys4);
1023 tp->md5sig_info->keys4 = NULL;
1024 tp->md5sig_info->alloced4 = 0;
1025 }
1026}
1027
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -02001028static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval,
1029 int optlen)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001030{
1031 struct tcp_md5sig cmd;
1032 struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr;
1033 u8 *newkey;
1034
1035 if (optlen < sizeof(cmd))
1036 return -EINVAL;
1037
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -02001038 if (copy_from_user(&cmd, optval, sizeof(cmd)))
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001039 return -EFAULT;
1040
1041 if (sin->sin_family != AF_INET)
1042 return -EINVAL;
1043
1044 if (!cmd.tcpm_key || !cmd.tcpm_keylen) {
1045 if (!tcp_sk(sk)->md5sig_info)
1046 return -ENOENT;
1047 return tcp_v4_md5_do_del(sk, sin->sin_addr.s_addr);
1048 }
1049
1050 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
1051 return -EINVAL;
1052
1053 if (!tcp_sk(sk)->md5sig_info) {
1054 struct tcp_sock *tp = tcp_sk(sk);
Wu Fengguangaa133072009-09-02 23:45:45 -07001055 struct tcp_md5sig_info *p;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001056
Wu Fengguangaa133072009-09-02 23:45:45 -07001057 p = kzalloc(sizeof(*p), sk->sk_allocation);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001058 if (!p)
1059 return -EINVAL;
1060
1061 tp->md5sig_info = p;
Eric Dumazeta4654192010-05-16 00:36:33 -07001062 sk_nocaps_add(sk, NETIF_F_GSO_MASK);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001063 }
1064
Wu Fengguangaa133072009-09-02 23:45:45 -07001065 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, sk->sk_allocation);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001066 if (!newkey)
1067 return -ENOMEM;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001068 return tcp_v4_md5_do_add(sk, sin->sin_addr.s_addr,
1069 newkey, cmd.tcpm_keylen);
1070}
1071
Adam Langley49a72df2008-07-19 00:01:42 -07001072static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
1073 __be32 daddr, __be32 saddr, int nbytes)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001074{
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001075 struct tcp4_pseudohdr *bp;
Adam Langley49a72df2008-07-19 00:01:42 -07001076 struct scatterlist sg;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001077
1078 bp = &hp->md5_blk.ip4;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001079
1080 /*
Adam Langley49a72df2008-07-19 00:01:42 -07001081 * 1. the TCP pseudo-header (in the order: source IP address,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001082 * destination IP address, zero-padded protocol number, and
1083 * segment length)
1084 */
1085 bp->saddr = saddr;
1086 bp->daddr = daddr;
1087 bp->pad = 0;
YOSHIFUJI Hideaki076fb722008-04-17 12:48:12 +09001088 bp->protocol = IPPROTO_TCP;
Adam Langley49a72df2008-07-19 00:01:42 -07001089 bp->len = cpu_to_be16(nbytes);
David S. Millerc7da57a2007-10-26 00:41:21 -07001090
Adam Langley49a72df2008-07-19 00:01:42 -07001091 sg_init_one(&sg, bp, sizeof(*bp));
1092 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
1093}
1094
1095static int tcp_v4_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001096 __be32 daddr, __be32 saddr, const struct tcphdr *th)
Adam Langley49a72df2008-07-19 00:01:42 -07001097{
1098 struct tcp_md5sig_pool *hp;
1099 struct hash_desc *desc;
1100
1101 hp = tcp_get_md5sig_pool();
1102 if (!hp)
1103 goto clear_hash_noput;
1104 desc = &hp->md5_desc;
1105
1106 if (crypto_hash_init(desc))
1107 goto clear_hash;
1108 if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
1109 goto clear_hash;
1110 if (tcp_md5_hash_header(hp, th))
1111 goto clear_hash;
1112 if (tcp_md5_hash_key(hp, key))
1113 goto clear_hash;
1114 if (crypto_hash_final(desc, md5_hash))
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001115 goto clear_hash;
1116
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001117 tcp_put_md5sig_pool();
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001118 return 0;
Adam Langley49a72df2008-07-19 00:01:42 -07001119
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001120clear_hash:
1121 tcp_put_md5sig_pool();
1122clear_hash_noput:
1123 memset(md5_hash, 0, 16);
Adam Langley49a72df2008-07-19 00:01:42 -07001124 return 1;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001125}
1126
Adam Langley49a72df2008-07-19 00:01:42 -07001127int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001128 const struct sock *sk, const struct request_sock *req,
1129 const struct sk_buff *skb)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001130{
Adam Langley49a72df2008-07-19 00:01:42 -07001131 struct tcp_md5sig_pool *hp;
1132 struct hash_desc *desc;
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001133 const struct tcphdr *th = tcp_hdr(skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001134 __be32 saddr, daddr;
1135
1136 if (sk) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001137 saddr = inet_sk(sk)->inet_saddr;
1138 daddr = inet_sk(sk)->inet_daddr;
Adam Langley49a72df2008-07-19 00:01:42 -07001139 } else if (req) {
1140 saddr = inet_rsk(req)->loc_addr;
1141 daddr = inet_rsk(req)->rmt_addr;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001142 } else {
Adam Langley49a72df2008-07-19 00:01:42 -07001143 const struct iphdr *iph = ip_hdr(skb);
1144 saddr = iph->saddr;
1145 daddr = iph->daddr;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001146 }
Adam Langley49a72df2008-07-19 00:01:42 -07001147
1148 hp = tcp_get_md5sig_pool();
1149 if (!hp)
1150 goto clear_hash_noput;
1151 desc = &hp->md5_desc;
1152
1153 if (crypto_hash_init(desc))
1154 goto clear_hash;
1155
1156 if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
1157 goto clear_hash;
1158 if (tcp_md5_hash_header(hp, th))
1159 goto clear_hash;
1160 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
1161 goto clear_hash;
1162 if (tcp_md5_hash_key(hp, key))
1163 goto clear_hash;
1164 if (crypto_hash_final(desc, md5_hash))
1165 goto clear_hash;
1166
1167 tcp_put_md5sig_pool();
1168 return 0;
1169
1170clear_hash:
1171 tcp_put_md5sig_pool();
1172clear_hash_noput:
1173 memset(md5_hash, 0, 16);
1174 return 1;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001175}
Adam Langley49a72df2008-07-19 00:01:42 -07001176EXPORT_SYMBOL(tcp_v4_md5_hash_skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001177
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001178static int tcp_v4_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001179{
1180 /*
1181 * This gets called for each TCP segment that arrives
1182 * so we want to be efficient.
1183 * We have 3 drop cases:
1184 * o No MD5 hash and one expected.
1185 * o MD5 hash and we're not expecting one.
1186 * o MD5 hash and its wrong.
1187 */
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001188 const __u8 *hash_location = NULL;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001189 struct tcp_md5sig_key *hash_expected;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001190 const struct iphdr *iph = ip_hdr(skb);
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001191 const struct tcphdr *th = tcp_hdr(skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001192 int genhash;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001193 unsigned char newhash[16];
1194
1195 hash_expected = tcp_v4_md5_do_lookup(sk, iph->saddr);
YOSHIFUJI Hideaki7d5d5522008-04-17 12:29:53 +09001196 hash_location = tcp_parse_md5sig_option(th);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001197
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001198 /* We've parsed the options - do we have a hash? */
1199 if (!hash_expected && !hash_location)
1200 return 0;
1201
1202 if (hash_expected && !hash_location) {
David S. Miller785957d2008-07-30 03:03:15 -07001203 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001204 return 1;
1205 }
1206
1207 if (!hash_expected && hash_location) {
David S. Miller785957d2008-07-30 03:03:15 -07001208 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001209 return 1;
1210 }
1211
1212 /* Okay, so this is hash_expected and hash_location -
1213 * so we need to calculate the checksum.
1214 */
Adam Langley49a72df2008-07-19 00:01:42 -07001215 genhash = tcp_v4_md5_hash_skb(newhash,
1216 hash_expected,
1217 NULL, NULL, skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001218
1219 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
1220 if (net_ratelimit()) {
Harvey Harrison673d57e2008-10-31 00:53:57 -07001221 printk(KERN_INFO "MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n",
1222 &iph->saddr, ntohs(th->source),
1223 &iph->daddr, ntohs(th->dest),
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001224 genhash ? " tcp_v4_calc_md5_hash failed" : "");
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001225 }
1226 return 1;
1227 }
1228 return 0;
1229}
1230
1231#endif
1232
Eric Dumazet72a3eff2006-11-16 02:30:37 -08001233struct request_sock_ops tcp_request_sock_ops __read_mostly = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 .family = PF_INET,
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001235 .obj_size = sizeof(struct tcp_request_sock),
Octavian Purdila72659ec2010-01-17 19:09:39 -08001236 .rtx_syn_ack = tcp_v4_rtx_synack,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001237 .send_ack = tcp_v4_reqsk_send_ack,
1238 .destructor = tcp_v4_reqsk_destructor,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 .send_reset = tcp_v4_send_reset,
Octavian Purdila72659ec2010-01-17 19:09:39 -08001240 .syn_ack_timeout = tcp_syn_ack_timeout,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241};
1242
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001243#ifdef CONFIG_TCP_MD5SIG
Stephen Hemmingerb2e4b3d2009-09-01 19:25:03 +00001244static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001245 .md5_lookup = tcp_v4_reqsk_md5_lookup,
John Dykstrae3afe7b2009-07-16 05:04:51 +00001246 .calc_md5_hash = tcp_v4_md5_hash_skb,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001247};
Andrew Mortonb6332e62006-11-30 19:16:28 -08001248#endif
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001249
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
1251{
William Allen Simpson4957faade2009-12-02 18:25:27 +00001252 struct tcp_extend_values tmp_ext;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 struct tcp_options_received tmp_opt;
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001254 const u8 *hash_location;
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001255 struct request_sock *req;
William Allen Simpsone6b4d112009-12-02 18:07:39 +00001256 struct inet_request_sock *ireq;
William Allen Simpson4957faade2009-12-02 18:25:27 +00001257 struct tcp_sock *tp = tcp_sk(sk);
William Allen Simpsone6b4d112009-12-02 18:07:39 +00001258 struct dst_entry *dst = NULL;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001259 __be32 saddr = ip_hdr(skb)->saddr;
1260 __be32 daddr = ip_hdr(skb)->daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 __u32 isn = TCP_SKB_CB(skb)->when;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 int want_cookie = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263
1264 /* Never answer to SYNs send to broadcast or multicast */
Eric Dumazet511c3f92009-06-02 05:14:27 +00001265 if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 goto drop;
1267
1268 /* TW buckets are converted to open requests without
1269 * limitations, they conserve resources and peer is
1270 * evidently real one.
1271 */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001272 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
Eric Dumazet946cedc2011-08-30 03:21:44 +00001273 want_cookie = tcp_syn_flood_action(sk, skb, "TCP");
1274 if (!want_cookie)
1275 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 }
1277
1278 /* Accept backlog is full. If we have already queued enough
1279 * of warm entries in syn queue, drop request. It is better than
1280 * clogging syn queue with openreqs with exponentially increasing
1281 * timeout.
1282 */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001283 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 goto drop;
1285
Arnaldo Carvalho de Meloce4a7d02008-06-10 12:39:35 -07001286 req = inet_reqsk_alloc(&tcp_request_sock_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 if (!req)
1288 goto drop;
1289
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001290#ifdef CONFIG_TCP_MD5SIG
1291 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv4_ops;
1292#endif
1293
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 tcp_clear_options(&tmp_opt);
William Allen Simpsonbee7ca92009-11-10 09:51:18 +00001295 tmp_opt.mss_clamp = TCP_MSS_DEFAULT;
William Allen Simpson4957faade2009-12-02 18:25:27 +00001296 tmp_opt.user_mss = tp->rx_opt.user_mss;
David S. Millerbb5b7c12009-12-15 20:56:42 -08001297 tcp_parse_options(skb, &tmp_opt, &hash_location, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298
William Allen Simpson4957faade2009-12-02 18:25:27 +00001299 if (tmp_opt.cookie_plus > 0 &&
1300 tmp_opt.saw_tstamp &&
1301 !tp->rx_opt.cookie_out_never &&
1302 (sysctl_tcp_cookie_size > 0 ||
1303 (tp->cookie_values != NULL &&
1304 tp->cookie_values->cookie_desired > 0))) {
1305 u8 *c;
1306 u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS];
1307 int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE;
1308
1309 if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0)
1310 goto drop_and_release;
1311
1312 /* Secret recipe starts with IP addresses */
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001313 *mess++ ^= (__force u32)daddr;
1314 *mess++ ^= (__force u32)saddr;
William Allen Simpson4957faade2009-12-02 18:25:27 +00001315
1316 /* plus variable length Initiator Cookie */
1317 c = (u8 *)mess;
1318 while (l-- > 0)
1319 *c++ ^= *hash_location++;
1320
William Allen Simpson4957faade2009-12-02 18:25:27 +00001321 want_cookie = 0; /* not our kind of cookie */
William Allen Simpson4957faade2009-12-02 18:25:27 +00001322 tmp_ext.cookie_out_never = 0; /* false */
1323 tmp_ext.cookie_plus = tmp_opt.cookie_plus;
1324 } else if (!tp->rx_opt.cookie_in_always) {
1325 /* redundant indications, but ensure initialization. */
1326 tmp_ext.cookie_out_never = 1; /* true */
1327 tmp_ext.cookie_plus = 0;
1328 } else {
1329 goto drop_and_release;
1330 }
1331 tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332
Florian Westphal4dfc2812008-04-10 03:12:40 -07001333 if (want_cookie && !tmp_opt.saw_tstamp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 tcp_clear_options(&tmp_opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 tcp_openreq_init(req, &tmp_opt, skb);
1338
David S. Millerbb5b7c12009-12-15 20:56:42 -08001339 ireq = inet_rsk(req);
1340 ireq->loc_addr = daddr;
1341 ireq->rmt_addr = saddr;
1342 ireq->no_srccheck = inet_sk(sk)->transparent;
1343 ireq->opt = tcp_v4_save_options(sk, skb);
1344
Paul Moore284904a2009-03-27 17:10:28 -04001345 if (security_inet_conn_request(sk, skb, req))
David S. Millerbb5b7c12009-12-15 20:56:42 -08001346 goto drop_and_free;
Paul Moore284904a2009-03-27 17:10:28 -04001347
Florian Westphal172d69e2010-06-21 11:48:45 +00001348 if (!want_cookie || tmp_opt.tstamp_ok)
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001349 TCP_ECN_create_request(req, tcp_hdr(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
1351 if (want_cookie) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 isn = cookie_v4_init_sequence(sk, skb, &req->mss);
Florian Westphal172d69e2010-06-21 11:48:45 +00001353 req->cookie_ts = tmp_opt.tstamp_ok;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 } else if (!isn) {
1355 struct inet_peer *peer = NULL;
David S. Miller6bd023f2011-05-18 18:32:03 -04001356 struct flowi4 fl4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357
1358 /* VJ's idea. We save last timestamp seen
1359 * from the destination in peer table, when entering
1360 * state TIME-WAIT, and check against it before
1361 * accepting new connection request.
1362 *
1363 * If "isn" is not zero, this request hit alive
1364 * timewait bucket, so that all the necessary checks
1365 * are made in the function processing timewait state.
1366 */
1367 if (tmp_opt.saw_tstamp &&
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -07001368 tcp_death_row.sysctl_tw_recycle &&
David S. Miller6bd023f2011-05-18 18:32:03 -04001369 (dst = inet_csk_route_req(sk, &fl4, req)) != NULL &&
David S. Millered2361e2011-05-18 18:38:54 -04001370 fl4.daddr == saddr &&
1371 (peer = rt_get_peer((struct rtable *)dst, fl4.daddr)) != NULL) {
Eric Dumazet317fe0e2010-06-16 04:52:13 +00001372 inet_peer_refcheck(peer);
Eric Dumazet2c1409a2009-11-12 09:33:09 +00001373 if ((u32)get_seconds() - peer->tcp_ts_stamp < TCP_PAWS_MSL &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 (s32)(peer->tcp_ts - req->ts_recent) >
1375 TCP_PAWS_WINDOW) {
Pavel Emelyanovde0744a2008-07-16 20:31:16 -07001376 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED);
Denis V. Lunev7cd04fa2008-03-03 11:59:32 -08001377 goto drop_and_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 }
1379 }
1380 /* Kill the following clause, if you dislike this way. */
1381 else if (!sysctl_tcp_syncookies &&
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001382 (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 (sysctl_max_syn_backlog >> 2)) &&
1384 (!peer || !peer->tcp_ts_stamp) &&
1385 (!dst || !dst_metric(dst, RTAX_RTT))) {
1386 /* Without syncookies last quarter of
1387 * backlog is filled with destinations,
1388 * proven to be alive.
1389 * It means that we continue to communicate
1390 * to destinations, already remembered
1391 * to the moment of synflood.
1392 */
Harvey Harrison673d57e2008-10-31 00:53:57 -07001393 LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI4/%u\n",
1394 &saddr, ntohs(tcp_hdr(skb)->source));
Denis V. Lunev7cd04fa2008-03-03 11:59:32 -08001395 goto drop_and_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 }
1397
Gerrit Renkera94f7232006-11-10 14:06:49 -08001398 isn = tcp_v4_init_sequence(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 }
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001400 tcp_rsk(req)->snt_isn = isn;
Jerry Chu9ad7c042011-06-08 11:08:38 +00001401 tcp_rsk(req)->snt_synack = tcp_time_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
Octavian Purdila72659ec2010-01-17 19:09:39 -08001403 if (tcp_v4_send_synack(sk, dst, req,
1404 (struct request_values *)&tmp_ext) ||
William Allen Simpson4957faade2009-12-02 18:25:27 +00001405 want_cookie)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 goto drop_and_free;
1407
Denis V. Lunev7cd04fa2008-03-03 11:59:32 -08001408 inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 return 0;
1410
Denis V. Lunev7cd04fa2008-03-03 11:59:32 -08001411drop_and_release:
1412 dst_release(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413drop_and_free:
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001414 reqsk_free(req);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415drop:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 return 0;
1417}
Eric Dumazet4bc2f182010-07-09 21:22:10 +00001418EXPORT_SYMBOL(tcp_v4_conn_request);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419
1420
1421/*
1422 * The three way handshake has completed - we got a valid synack -
1423 * now create the new socket.
1424 */
1425struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001426 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 struct dst_entry *dst)
1428{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001429 struct inet_request_sock *ireq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 struct inet_sock *newinet;
1431 struct tcp_sock *newtp;
1432 struct sock *newsk;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001433#ifdef CONFIG_TCP_MD5SIG
1434 struct tcp_md5sig_key *key;
1435#endif
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001436 struct ip_options_rcu *inet_opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
1438 if (sk_acceptq_is_full(sk))
1439 goto exit_overflow;
1440
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 newsk = tcp_create_openreq_child(sk, req, skb);
1442 if (!newsk)
Balazs Scheidler093d2822010-10-21 13:06:43 +02001443 goto exit_nonewsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
Herbert Xubcd76112006-06-30 13:36:35 -07001445 newsk->sk_gso_type = SKB_GSO_TCPV4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446
1447 newtp = tcp_sk(newsk);
1448 newinet = inet_sk(newsk);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001449 ireq = inet_rsk(req);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001450 newinet->inet_daddr = ireq->rmt_addr;
1451 newinet->inet_rcv_saddr = ireq->loc_addr;
1452 newinet->inet_saddr = ireq->loc_addr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001453 inet_opt = ireq->opt;
1454 rcu_assign_pointer(newinet->inet_opt, inet_opt);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001455 ireq->opt = NULL;
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001456 newinet->mc_index = inet_iif(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001457 newinet->mc_ttl = ip_hdr(skb)->ttl;
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001458 inet_csk(newsk)->icsk_ext_hdr_len = 0;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001459 if (inet_opt)
1460 inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001461 newinet->inet_id = newtp->write_seq ^ jiffies;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462
David S. Miller0e734412011-05-08 15:28:03 -07001463 if (!dst && (dst = inet_csk_route_child_sock(sk, newsk, req)) == NULL)
1464 goto put_and_exit;
1465
1466 sk_setup_caps(newsk, dst);
1467
John Heffner5d424d52006-03-20 17:53:41 -08001468 tcp_mtup_init(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 tcp_sync_mss(newsk, dst_mtu(dst));
David S. Miller0dbaee32010-12-13 12:52:14 -08001470 newtp->advmss = dst_metric_advmss(dst);
Tom Quetchenbachf5fff5d2008-09-21 00:21:51 -07001471 if (tcp_sk(sk)->rx_opt.user_mss &&
1472 tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
1473 newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;
1474
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 tcp_initialize_rcv_mss(newsk);
Jerry Chu9ad7c042011-06-08 11:08:38 +00001476 if (tcp_rsk(req)->snt_synack)
1477 tcp_valid_rtt_meas(newsk,
1478 tcp_time_stamp - tcp_rsk(req)->snt_synack);
1479 newtp->total_retrans = req->retrans;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001481#ifdef CONFIG_TCP_MD5SIG
1482 /* Copy over the MD5 key from the original socket */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001483 key = tcp_v4_md5_do_lookup(sk, newinet->inet_daddr);
1484 if (key != NULL) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001485 /*
1486 * We're using one, so create a matching key
1487 * on the newsk structure. If we fail to get
1488 * memory, then we end up not copying the key
1489 * across. Shucks.
1490 */
Arnaldo Carvalho de Melof6685932006-11-17 11:06:01 -02001491 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1492 if (newkey != NULL)
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001493 tcp_v4_md5_do_add(newsk, newinet->inet_daddr,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001494 newkey, key->keylen);
Eric Dumazeta4654192010-05-16 00:36:33 -07001495 sk_nocaps_add(newsk, NETIF_F_GSO_MASK);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001496 }
1497#endif
1498
David S. Miller0e734412011-05-08 15:28:03 -07001499 if (__inet_inherit_port(sk, newsk) < 0)
1500 goto put_and_exit;
Eric Dumazet9327f702009-12-04 03:46:54 +00001501 __inet_hash_nolisten(newsk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
1503 return newsk;
1504
1505exit_overflow:
Pavel Emelyanovde0744a2008-07-16 20:31:16 -07001506 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
Balazs Scheidler093d2822010-10-21 13:06:43 +02001507exit_nonewsk:
1508 dst_release(dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509exit:
Pavel Emelyanovde0744a2008-07-16 20:31:16 -07001510 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 return NULL;
David S. Miller0e734412011-05-08 15:28:03 -07001512put_and_exit:
1513 sock_put(newsk);
1514 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515}
Eric Dumazet4bc2f182010-07-09 21:22:10 +00001516EXPORT_SYMBOL(tcp_v4_syn_recv_sock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517
1518static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
1519{
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001520 struct tcphdr *th = tcp_hdr(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001521 const struct iphdr *iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 struct sock *nsk;
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001523 struct request_sock **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 /* Find possible connection requests. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001525 struct request_sock *req = inet_csk_search_req(sk, &prev, th->source,
1526 iph->saddr, iph->daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 if (req)
1528 return tcp_check_req(sk, skb, req, prev);
1529
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001530 nsk = inet_lookup_established(sock_net(sk), &tcp_hashinfo, iph->saddr,
Pavel Emelyanovc67499c2008-01-31 05:06:40 -08001531 th->source, iph->daddr, th->dest, inet_iif(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
1533 if (nsk) {
1534 if (nsk->sk_state != TCP_TIME_WAIT) {
1535 bh_lock_sock(nsk);
1536 return nsk;
1537 }
YOSHIFUJI Hideaki9469c7b2006-10-10 19:41:46 -07001538 inet_twsk_put(inet_twsk(nsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 return NULL;
1540 }
1541
1542#ifdef CONFIG_SYN_COOKIES
Florian Westphalaf9b4732010-06-03 00:43:44 +00001543 if (!th->syn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 sk = cookie_v4_check(sk, skb, &(IPCB(skb)->opt));
1545#endif
1546 return sk;
1547}
1548
Al Virob51655b2006-11-14 21:40:42 -08001549static __sum16 tcp_v4_checksum_init(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001551 const struct iphdr *iph = ip_hdr(skb);
1552
Patrick McHardy84fa7932006-08-29 16:44:56 -07001553 if (skb->ip_summed == CHECKSUM_COMPLETE) {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001554 if (!tcp_v4_check(skb->len, iph->saddr,
1555 iph->daddr, skb->csum)) {
Herbert Xufb286bb2005-11-10 13:01:24 -08001556 skb->ip_summed = CHECKSUM_UNNECESSARY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 return 0;
Herbert Xufb286bb2005-11-10 13:01:24 -08001558 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 }
Herbert Xufb286bb2005-11-10 13:01:24 -08001560
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001561 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
Herbert Xufb286bb2005-11-10 13:01:24 -08001562 skb->len, IPPROTO_TCP, 0);
1563
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564 if (skb->len <= 76) {
Herbert Xufb286bb2005-11-10 13:01:24 -08001565 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 }
1567 return 0;
1568}
1569
1570
1571/* The socket must have it's spinlock held when we get
1572 * here.
1573 *
1574 * We have a potential double-lock case here, so even when
1575 * doing backlog processing we use the BH locking scheme.
1576 * This is because we cannot sleep with the original spinlock
1577 * held.
1578 */
1579int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
1580{
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001581 struct sock *rsk;
1582#ifdef CONFIG_TCP_MD5SIG
1583 /*
1584 * We really want to reject the packet as early as possible
1585 * if:
1586 * o We're expecting an MD5'd packet and this is no MD5 tcp option
1587 * o There is an MD5 option and we're not expecting one
1588 */
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -02001589 if (tcp_v4_inbound_md5_hash(sk, skb))
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001590 goto discard;
1591#endif
1592
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
Tom Herbertbdeab992011-08-14 19:45:55 +00001594 sock_rps_save_rxhash(sk, skb);
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001595 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001596 rsk = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 goto reset;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001598 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 return 0;
1600 }
1601
Arnaldo Carvalho de Meloab6a5bb2007-03-18 17:43:48 -07001602 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 goto csum_err;
1604
1605 if (sk->sk_state == TCP_LISTEN) {
1606 struct sock *nsk = tcp_v4_hnd_req(sk, skb);
1607 if (!nsk)
1608 goto discard;
1609
1610 if (nsk != sk) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001611 sock_rps_save_rxhash(nsk, skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001612 if (tcp_child_process(sk, nsk, skb)) {
1613 rsk = nsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 goto reset;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001615 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 return 0;
1617 }
Eric Dumazetca551582010-06-03 09:03:58 +00001618 } else
Tom Herbertbdeab992011-08-14 19:45:55 +00001619 sock_rps_save_rxhash(sk, skb);
Eric Dumazetca551582010-06-03 09:03:58 +00001620
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001621 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001622 rsk = sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 goto reset;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001624 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 return 0;
1626
1627reset:
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001628 tcp_v4_send_reset(rsk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629discard:
1630 kfree_skb(skb);
1631 /* Be careful here. If this function gets more complicated and
1632 * gcc suffers from register pressure on the x86, sk (in %ebx)
1633 * might be destroyed here. This current version compiles correctly,
1634 * but you have been warned.
1635 */
1636 return 0;
1637
1638csum_err:
Pavel Emelyanov63231bd2008-07-16 20:22:25 -07001639 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 goto discard;
1641}
Eric Dumazet4bc2f182010-07-09 21:22:10 +00001642EXPORT_SYMBOL(tcp_v4_do_rcv);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643
1644/*
1645 * From tcp_input.c
1646 */
1647
1648int tcp_v4_rcv(struct sk_buff *skb)
1649{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001650 const struct iphdr *iph;
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001651 const struct tcphdr *th;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 struct sock *sk;
1653 int ret;
Pavel Emelyanova86b1e32008-07-16 20:20:58 -07001654 struct net *net = dev_net(skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655
1656 if (skb->pkt_type != PACKET_HOST)
1657 goto discard_it;
1658
1659 /* Count it even if it's bad */
Pavel Emelyanov63231bd2008-07-16 20:22:25 -07001660 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661
1662 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1663 goto discard_it;
1664
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001665 th = tcp_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
1667 if (th->doff < sizeof(struct tcphdr) / 4)
1668 goto bad_packet;
1669 if (!pskb_may_pull(skb, th->doff * 4))
1670 goto discard_it;
1671
1672 /* An explanation is required here, I think.
1673 * Packet length and doff are validated by header prediction,
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -08001674 * provided case of th->doff==0 is eliminated.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 * So, we defer the checks. */
Herbert Xu60476372007-04-09 11:59:39 -07001676 if (!skb_csum_unnecessary(skb) && tcp_v4_checksum_init(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 goto bad_packet;
1678
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001679 th = tcp_hdr(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001680 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1682 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1683 skb->len - th->doff * 4);
1684 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1685 TCP_SKB_CB(skb)->when = 0;
Eric Dumazetb82d1bb2011-09-27 02:20:08 -04001686 TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 TCP_SKB_CB(skb)->sacked = 0;
1688
Arnaldo Carvalho de Melo9a1f27c2008-10-07 11:41:57 -07001689 sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 if (!sk)
1691 goto no_tcp_socket;
1692
Eric Dumazetbb134d52010-03-09 05:55:56 +00001693process:
1694 if (sk->sk_state == TCP_TIME_WAIT)
1695 goto do_time_wait;
1696
Eric Dumazet6cce09f2010-03-07 23:21:57 +00001697 if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
1698 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
Stephen Hemmingerd218d112010-01-11 16:28:01 -08001699 goto discard_and_relse;
Eric Dumazet6cce09f2010-03-07 23:21:57 +00001700 }
Stephen Hemmingerd218d112010-01-11 16:28:01 -08001701
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1703 goto discard_and_relse;
Patrick McHardyb59c2702006-01-06 23:06:10 -08001704 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Dmitry Mishinfda9ef52006-08-31 15:28:39 -07001706 if (sk_filter(sk, skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 goto discard_and_relse;
1708
1709 skb->dev = NULL;
1710
Ingo Molnarc6366182006-07-03 00:25:13 -07001711 bh_lock_sock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 ret = 0;
1713 if (!sock_owned_by_user(sk)) {
Chris Leech1a2449a2006-05-23 18:05:53 -07001714#ifdef CONFIG_NET_DMA
1715 struct tcp_sock *tp = tcp_sk(sk);
1716 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
Dan Williamsf67b4592009-01-06 11:38:15 -07001717 tp->ucopy.dma_chan = dma_find_channel(DMA_MEMCPY);
Chris Leech1a2449a2006-05-23 18:05:53 -07001718 if (tp->ucopy.dma_chan)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 ret = tcp_v4_do_rcv(sk, skb);
Chris Leech1a2449a2006-05-23 18:05:53 -07001720 else
1721#endif
1722 {
1723 if (!tcp_prequeue(sk, skb))
Shan Weiae8d7f82009-05-05 01:01:29 +00001724 ret = tcp_v4_do_rcv(sk, skb);
Chris Leech1a2449a2006-05-23 18:05:53 -07001725 }
Eric Dumazet6cce09f2010-03-07 23:21:57 +00001726 } else if (unlikely(sk_add_backlog(sk, skb))) {
Zhu Yi6b03a532010-03-04 18:01:41 +00001727 bh_unlock_sock(sk);
Eric Dumazet6cce09f2010-03-07 23:21:57 +00001728 NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Zhu Yi6b03a532010-03-04 18:01:41 +00001729 goto discard_and_relse;
1730 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 bh_unlock_sock(sk);
1732
1733 sock_put(sk);
1734
1735 return ret;
1736
1737no_tcp_socket:
1738 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1739 goto discard_it;
1740
1741 if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) {
1742bad_packet:
Pavel Emelyanov63231bd2008-07-16 20:22:25 -07001743 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 } else {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001745 tcp_v4_send_reset(NULL, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 }
1747
1748discard_it:
1749 /* Discard frame. */
1750 kfree_skb(skb);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001751 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752
1753discard_and_relse:
1754 sock_put(sk);
1755 goto discard_it;
1756
1757do_time_wait:
1758 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
YOSHIFUJI Hideaki9469c7b2006-10-10 19:41:46 -07001759 inet_twsk_put(inet_twsk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 goto discard_it;
1761 }
1762
1763 if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) {
Pavel Emelyanov63231bd2008-07-16 20:22:25 -07001764 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
YOSHIFUJI Hideaki9469c7b2006-10-10 19:41:46 -07001765 inet_twsk_put(inet_twsk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 goto discard_it;
1767 }
YOSHIFUJI Hideaki9469c7b2006-10-10 19:41:46 -07001768 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 case TCP_TW_SYN: {
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001770 struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
Pavel Emelyanovc67499c2008-01-31 05:06:40 -08001771 &tcp_hashinfo,
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001772 iph->daddr, th->dest,
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001773 inet_iif(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 if (sk2) {
YOSHIFUJI Hideaki9469c7b2006-10-10 19:41:46 -07001775 inet_twsk_deschedule(inet_twsk(sk), &tcp_death_row);
1776 inet_twsk_put(inet_twsk(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 sk = sk2;
1778 goto process;
1779 }
1780 /* Fall through to ACK */
1781 }
1782 case TCP_TW_ACK:
1783 tcp_v4_timewait_ack(sk, skb);
1784 break;
1785 case TCP_TW_RST:
1786 goto no_tcp_socket;
1787 case TCP_TW_SUCCESS:;
1788 }
1789 goto discard_it;
1790}
1791
David S. Miller3f419d22010-11-29 13:37:14 -08001792struct inet_peer *tcp_v4_get_peer(struct sock *sk, bool *release_it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793{
David S. Miller3f419d22010-11-29 13:37:14 -08001794 struct rtable *rt = (struct rtable *) __sk_dst_get(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 struct inet_sock *inet = inet_sk(sk);
David S. Miller3f419d22010-11-29 13:37:14 -08001796 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797
David S. Millerc5216cc2011-05-06 22:36:30 -07001798 if (!rt ||
1799 inet->cork.fl.u.ip4.daddr != inet->inet_daddr) {
David S. Millerb534ecf2010-11-30 11:54:19 -08001800 peer = inet_getpeer_v4(inet->inet_daddr, 1);
David S. Miller3f419d22010-11-29 13:37:14 -08001801 *release_it = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 } else {
1803 if (!rt->peer)
David S. Millera48eff12011-05-18 18:42:43 -04001804 rt_bind_peer(rt, inet->inet_daddr, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 peer = rt->peer;
David S. Miller3f419d22010-11-29 13:37:14 -08001806 *release_it = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 }
1808
David S. Miller3f419d22010-11-29 13:37:14 -08001809 return peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810}
David S. Miller3f419d22010-11-29 13:37:14 -08001811EXPORT_SYMBOL(tcp_v4_get_peer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812
David S. Millerccb7c412010-12-01 18:09:13 -08001813void *tcp_v4_tw_get_peer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814{
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001815 const struct inet_timewait_sock *tw = inet_twsk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816
David S. Millerccb7c412010-12-01 18:09:13 -08001817 return inet_getpeer_v4(tw->tw_daddr, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818}
David S. Millerccb7c412010-12-01 18:09:13 -08001819EXPORT_SYMBOL(tcp_v4_tw_get_peer);
1820
1821static struct timewait_sock_ops tcp_timewait_sock_ops = {
1822 .twsk_obj_size = sizeof(struct tcp_timewait_sock),
1823 .twsk_unique = tcp_twsk_unique,
1824 .twsk_destructor= tcp_twsk_destructor,
1825 .twsk_getpeer = tcp_v4_tw_get_peer,
1826};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827
Stephen Hemminger3b401a82009-09-01 19:25:04 +00001828const struct inet_connection_sock_af_ops ipv4_specific = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001829 .queue_xmit = ip_queue_xmit,
1830 .send_check = tcp_v4_send_check,
1831 .rebuild_header = inet_sk_rebuild_header,
1832 .conn_request = tcp_v4_conn_request,
1833 .syn_recv_sock = tcp_v4_syn_recv_sock,
David S. Miller3f419d22010-11-29 13:37:14 -08001834 .get_peer = tcp_v4_get_peer,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001835 .net_header_len = sizeof(struct iphdr),
1836 .setsockopt = ip_setsockopt,
1837 .getsockopt = ip_getsockopt,
1838 .addr2sockaddr = inet_csk_addr2sockaddr,
1839 .sockaddr_len = sizeof(struct sockaddr_in),
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001840 .bind_conflict = inet_csk_bind_conflict,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001841#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001842 .compat_setsockopt = compat_ip_setsockopt,
1843 .compat_getsockopt = compat_ip_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001844#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845};
Eric Dumazet4bc2f182010-07-09 21:22:10 +00001846EXPORT_SYMBOL(ipv4_specific);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001848#ifdef CONFIG_TCP_MD5SIG
Stephen Hemmingerb2e4b3d2009-09-01 19:25:03 +00001849static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001850 .md5_lookup = tcp_v4_md5_lookup,
Adam Langley49a72df2008-07-19 00:01:42 -07001851 .calc_md5_hash = tcp_v4_md5_hash_skb,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001852 .md5_add = tcp_v4_md5_add_func,
1853 .md5_parse = tcp_v4_parse_md5_keys,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001854};
Andrew Mortonb6332e62006-11-30 19:16:28 -08001855#endif
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001856
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857/* NOTE: A lot of things set to zero explicitly by call to
1858 * sk_alloc() so need not be done here.
1859 */
1860static int tcp_v4_init_sock(struct sock *sk)
1861{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -03001862 struct inet_connection_sock *icsk = inet_csk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 struct tcp_sock *tp = tcp_sk(sk);
1864
1865 skb_queue_head_init(&tp->out_of_order_queue);
1866 tcp_init_xmit_timers(sk);
1867 tcp_prequeue_init(tp);
1868
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -03001869 icsk->icsk_rto = TCP_TIMEOUT_INIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 tp->mdev = TCP_TIMEOUT_INIT;
1871
1872 /* So many TCP implementations out there (incorrectly) count the
1873 * initial SYN frame in their delayed-ACK and congestion control
1874 * algorithms that we must have the following bandaid to talk
1875 * efficiently to them. -DaveM
1876 */
Jerry Chu9ad7c042011-06-08 11:08:38 +00001877 tp->snd_cwnd = TCP_INIT_CWND;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878
1879 /* See draft-stevens-tcpca-spec-01 for discussion of the
1880 * initialization of these values.
1881 */
Ilpo Järvinen0b6a05c2009-09-15 01:30:10 -07001882 tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 tp->snd_cwnd_clamp = ~0;
William Allen Simpsonbee7ca92009-11-10 09:51:18 +00001884 tp->mss_cache = TCP_MSS_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885
1886 tp->reordering = sysctl_tcp_reordering;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -03001887 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888
1889 sk->sk_state = TCP_CLOSE;
1890
1891 sk->sk_write_space = sk_stream_write_space;
1892 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1893
Arnaldo Carvalho de Melo8292a172005-12-13 23:15:52 -08001894 icsk->icsk_af_ops = &ipv4_specific;
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001895 icsk->icsk_sync_mss = tcp_sync_mss;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001896#ifdef CONFIG_TCP_MD5SIG
1897 tp->af_specific = &tcp_sock_ipv4_specific;
1898#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899
William Allen Simpson435cf552009-12-02 18:17:05 +00001900 /* TCP Cookie Transactions */
1901 if (sysctl_tcp_cookie_size > 0) {
1902 /* Default, cookies without s_data_payload. */
1903 tp->cookie_values =
1904 kzalloc(sizeof(*tp->cookie_values),
1905 sk->sk_allocation);
1906 if (tp->cookie_values != NULL)
1907 kref_init(&tp->cookie_values->kref);
1908 }
1909 /* Presumed zeroed, in order of appearance:
1910 * cookie_in_always, cookie_out_never,
1911 * s_data_constant, s_data_in, s_data_out
1912 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1914 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1915
Herbert Xueb4dea52008-12-29 23:04:08 -08001916 local_bh_disable();
Eric Dumazet17483762008-11-25 21:16:35 -08001917 percpu_counter_inc(&tcp_sockets_allocated);
Herbert Xueb4dea52008-12-29 23:04:08 -08001918 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
1920 return 0;
1921}
1922
Brian Haley7d06b2e2008-06-14 17:04:49 -07001923void tcp_v4_destroy_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924{
1925 struct tcp_sock *tp = tcp_sk(sk);
1926
1927 tcp_clear_xmit_timers(sk);
1928
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -03001929 tcp_cleanup_congestion_control(sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -07001930
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 /* Cleanup up the write buffer. */
David S. Millerfe067e82007-03-07 12:12:44 -08001932 tcp_write_queue_purge(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933
1934 /* Cleans up our, hopefully empty, out_of_order_queue. */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001935 __skb_queue_purge(&tp->out_of_order_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001937#ifdef CONFIG_TCP_MD5SIG
1938 /* Clean up the MD5 key list, if any */
1939 if (tp->md5sig_info) {
1940 tcp_v4_clear_md5_list(sk);
1941 kfree(tp->md5sig_info);
1942 tp->md5sig_info = NULL;
1943 }
1944#endif
1945
Chris Leech1a2449a2006-05-23 18:05:53 -07001946#ifdef CONFIG_NET_DMA
1947 /* Cleans up our sk_async_wait_queue */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001948 __skb_queue_purge(&sk->sk_async_wait_queue);
Chris Leech1a2449a2006-05-23 18:05:53 -07001949#endif
1950
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 /* Clean prequeue, it must be empty really */
1952 __skb_queue_purge(&tp->ucopy.prequeue);
1953
1954 /* Clean up a referenced TCP bind bucket. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001955 if (inet_csk(sk)->icsk_bind_hash)
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001956 inet_put_port(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
1958 /*
1959 * If sendmsg cached page exists, toss it.
1960 */
1961 if (sk->sk_sndmsg_page) {
1962 __free_page(sk->sk_sndmsg_page);
1963 sk->sk_sndmsg_page = NULL;
1964 }
1965
William Allen Simpson435cf552009-12-02 18:17:05 +00001966 /* TCP Cookie Transactions */
1967 if (tp->cookie_values != NULL) {
1968 kref_put(&tp->cookie_values->kref,
1969 tcp_cookie_values_release);
1970 tp->cookie_values = NULL;
1971 }
1972
Eric Dumazet17483762008-11-25 21:16:35 -08001973 percpu_counter_dec(&tcp_sockets_allocated);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975EXPORT_SYMBOL(tcp_v4_destroy_sock);
1976
1977#ifdef CONFIG_PROC_FS
1978/* Proc filesystem TCP sock list dumping. */
1979
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08001980static inline struct inet_timewait_sock *tw_head(struct hlist_nulls_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981{
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08001982 return hlist_nulls_empty(head) ? NULL :
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001983 list_entry(head->first, struct inet_timewait_sock, tw_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984}
1985
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001986static inline struct inet_timewait_sock *tw_next(struct inet_timewait_sock *tw)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987{
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08001988 return !is_a_nulls(tw->tw_node.next) ?
1989 hlist_nulls_entry(tw->tw_node.next, typeof(*tw), tw_node) : NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990}
1991
Tom Herberta8b690f2010-06-07 00:43:42 -07001992/*
1993 * Get next listener socket follow cur. If cur is NULL, get first socket
1994 * starting from bucket given in st->bucket; when st->bucket is zero the
1995 * very first socket in the hash table is returned.
1996 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997static void *listening_get_next(struct seq_file *seq, void *cur)
1998{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001999 struct inet_connection_sock *icsk;
Eric Dumazetc25eb3b2008-11-23 17:22:55 -08002000 struct hlist_nulls_node *node;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 struct sock *sk = cur;
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002002 struct inet_listen_hashbucket *ilb;
Jianjun Kong5799de02008-11-03 02:49:10 -08002003 struct tcp_iter_state *st = seq->private;
Denis V. Luneva4146b12008-04-13 22:11:14 -07002004 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005
2006 if (!sk) {
Tom Herberta8b690f2010-06-07 00:43:42 -07002007 ilb = &tcp_hashinfo.listening_hash[st->bucket];
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002008 spin_lock_bh(&ilb->lock);
Eric Dumazetc25eb3b2008-11-23 17:22:55 -08002009 sk = sk_nulls_head(&ilb->head);
Tom Herberta8b690f2010-06-07 00:43:42 -07002010 st->offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 goto get_sk;
2012 }
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002013 ilb = &tcp_hashinfo.listening_hash[st->bucket];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 ++st->num;
Tom Herberta8b690f2010-06-07 00:43:42 -07002015 ++st->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
2017 if (st->state == TCP_SEQ_STATE_OPENREQ) {
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07002018 struct request_sock *req = cur;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019
Eric Dumazet72a3eff2006-11-16 02:30:37 -08002020 icsk = inet_csk(st->syn_wait_sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 req = req->dl_next;
2022 while (1) {
2023 while (req) {
Daniel Lezcanobdccc4c2008-07-19 00:15:13 -07002024 if (req->rsk_ops->family == st->family) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 cur = req;
2026 goto out;
2027 }
2028 req = req->dl_next;
2029 }
Eric Dumazet72a3eff2006-11-16 02:30:37 -08002030 if (++st->sbucket >= icsk->icsk_accept_queue.listen_opt->nr_table_entries)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 break;
2032get_req:
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002033 req = icsk->icsk_accept_queue.listen_opt->syn_table[st->sbucket];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 }
Eric Dumazet1bde5ac2010-12-23 09:32:46 -08002035 sk = sk_nulls_next(st->syn_wait_sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 st->state = TCP_SEQ_STATE_LISTENING;
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002037 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 } else {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002039 icsk = inet_csk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002040 read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
2041 if (reqsk_queue_len(&icsk->icsk_accept_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 goto start_req;
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002043 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
Eric Dumazet1bde5ac2010-12-23 09:32:46 -08002044 sk = sk_nulls_next(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 }
2046get_sk:
Eric Dumazetc25eb3b2008-11-23 17:22:55 -08002047 sk_nulls_for_each_from(sk, node) {
Pavel Emelyanov8475ef92010-11-22 03:26:12 +00002048 if (!net_eq(sock_net(sk), net))
2049 continue;
2050 if (sk->sk_family == st->family) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 cur = sk;
2052 goto out;
2053 }
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002054 icsk = inet_csk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002055 read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
2056 if (reqsk_queue_len(&icsk->icsk_accept_queue)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057start_req:
2058 st->uid = sock_i_uid(sk);
2059 st->syn_wait_sk = sk;
2060 st->state = TCP_SEQ_STATE_OPENREQ;
2061 st->sbucket = 0;
2062 goto get_req;
2063 }
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002064 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 }
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002066 spin_unlock_bh(&ilb->lock);
Tom Herberta8b690f2010-06-07 00:43:42 -07002067 st->offset = 0;
Arnaldo Carvalho de Melo0f7ff922005-08-09 19:59:44 -07002068 if (++st->bucket < INET_LHTABLE_SIZE) {
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002069 ilb = &tcp_hashinfo.listening_hash[st->bucket];
2070 spin_lock_bh(&ilb->lock);
Eric Dumazetc25eb3b2008-11-23 17:22:55 -08002071 sk = sk_nulls_head(&ilb->head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 goto get_sk;
2073 }
2074 cur = NULL;
2075out:
2076 return cur;
2077}
2078
2079static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
2080{
Tom Herberta8b690f2010-06-07 00:43:42 -07002081 struct tcp_iter_state *st = seq->private;
2082 void *rc;
2083
2084 st->bucket = 0;
2085 st->offset = 0;
2086 rc = listening_get_next(seq, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087
2088 while (rc && *pos) {
2089 rc = listening_get_next(seq, rc);
2090 --*pos;
2091 }
2092 return rc;
2093}
2094
Andi Kleen6eac5602008-08-28 01:08:02 -07002095static inline int empty_bucket(struct tcp_iter_state *st)
2096{
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002097 return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain) &&
2098 hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].twchain);
Andi Kleen6eac5602008-08-28 01:08:02 -07002099}
2100
Tom Herberta8b690f2010-06-07 00:43:42 -07002101/*
2102 * Get first established socket starting from bucket given in st->bucket.
2103 * If st->bucket is zero, the very first socket in the hash is returned.
2104 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105static void *established_get_first(struct seq_file *seq)
2106{
Jianjun Kong5799de02008-11-03 02:49:10 -08002107 struct tcp_iter_state *st = seq->private;
Denis V. Luneva4146b12008-04-13 22:11:14 -07002108 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 void *rc = NULL;
2110
Tom Herberta8b690f2010-06-07 00:43:42 -07002111 st->offset = 0;
2112 for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 struct sock *sk;
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002114 struct hlist_nulls_node *node;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002115 struct inet_timewait_sock *tw;
Eric Dumazet9db66bd2008-11-20 20:39:09 -08002116 spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117
Andi Kleen6eac5602008-08-28 01:08:02 -07002118 /* Lockless fast path for the common case of empty buckets */
2119 if (empty_bucket(st))
2120 continue;
2121
Eric Dumazet9db66bd2008-11-20 20:39:09 -08002122 spin_lock_bh(lock);
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002123 sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
Daniel Lezcanof40c8172008-03-21 04:13:54 -07002124 if (sk->sk_family != st->family ||
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002125 !net_eq(sock_net(sk), net)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 continue;
2127 }
2128 rc = sk;
2129 goto out;
2130 }
2131 st->state = TCP_SEQ_STATE_TIME_WAIT;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002132 inet_twsk_for_each(tw, node,
Eric Dumazetdbca9b2752007-02-08 14:16:46 -08002133 &tcp_hashinfo.ehash[st->bucket].twchain) {
Pavel Emelyanov28518fc2008-03-21 15:52:00 -07002134 if (tw->tw_family != st->family ||
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002135 !net_eq(twsk_net(tw), net)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136 continue;
2137 }
2138 rc = tw;
2139 goto out;
2140 }
Eric Dumazet9db66bd2008-11-20 20:39:09 -08002141 spin_unlock_bh(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 st->state = TCP_SEQ_STATE_ESTABLISHED;
2143 }
2144out:
2145 return rc;
2146}
2147
2148static void *established_get_next(struct seq_file *seq, void *cur)
2149{
2150 struct sock *sk = cur;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002151 struct inet_timewait_sock *tw;
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002152 struct hlist_nulls_node *node;
Jianjun Kong5799de02008-11-03 02:49:10 -08002153 struct tcp_iter_state *st = seq->private;
Denis V. Luneva4146b12008-04-13 22:11:14 -07002154 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155
2156 ++st->num;
Tom Herberta8b690f2010-06-07 00:43:42 -07002157 ++st->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158
2159 if (st->state == TCP_SEQ_STATE_TIME_WAIT) {
2160 tw = cur;
2161 tw = tw_next(tw);
2162get_tw:
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002163 while (tw && (tw->tw_family != st->family || !net_eq(twsk_net(tw), net))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164 tw = tw_next(tw);
2165 }
2166 if (tw) {
2167 cur = tw;
2168 goto out;
2169 }
Eric Dumazet9db66bd2008-11-20 20:39:09 -08002170 spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171 st->state = TCP_SEQ_STATE_ESTABLISHED;
2172
Andi Kleen6eac5602008-08-28 01:08:02 -07002173 /* Look for next non empty bucket */
Tom Herberta8b690f2010-06-07 00:43:42 -07002174 st->offset = 0;
Eric Dumazetf373b532009-10-09 00:16:19 +00002175 while (++st->bucket <= tcp_hashinfo.ehash_mask &&
Andi Kleen6eac5602008-08-28 01:08:02 -07002176 empty_bucket(st))
2177 ;
Eric Dumazetf373b532009-10-09 00:16:19 +00002178 if (st->bucket > tcp_hashinfo.ehash_mask)
Andi Kleen6eac5602008-08-28 01:08:02 -07002179 return NULL;
2180
Eric Dumazet9db66bd2008-11-20 20:39:09 -08002181 spin_lock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002182 sk = sk_nulls_head(&tcp_hashinfo.ehash[st->bucket].chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 } else
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002184 sk = sk_nulls_next(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002186 sk_nulls_for_each_from(sk, node) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002187 if (sk->sk_family == st->family && net_eq(sock_net(sk), net))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 goto found;
2189 }
2190
2191 st->state = TCP_SEQ_STATE_TIME_WAIT;
Eric Dumazetdbca9b2752007-02-08 14:16:46 -08002192 tw = tw_head(&tcp_hashinfo.ehash[st->bucket].twchain);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 goto get_tw;
2194found:
2195 cur = sk;
2196out:
2197 return cur;
2198}
2199
2200static void *established_get_idx(struct seq_file *seq, loff_t pos)
2201{
Tom Herberta8b690f2010-06-07 00:43:42 -07002202 struct tcp_iter_state *st = seq->private;
2203 void *rc;
2204
2205 st->bucket = 0;
2206 rc = established_get_first(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207
2208 while (rc && pos) {
2209 rc = established_get_next(seq, rc);
2210 --pos;
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -02002211 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 return rc;
2213}
2214
2215static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
2216{
2217 void *rc;
Jianjun Kong5799de02008-11-03 02:49:10 -08002218 struct tcp_iter_state *st = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 st->state = TCP_SEQ_STATE_LISTENING;
2221 rc = listening_get_idx(seq, &pos);
2222
2223 if (!rc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 st->state = TCP_SEQ_STATE_ESTABLISHED;
2225 rc = established_get_idx(seq, pos);
2226 }
2227
2228 return rc;
2229}
2230
Tom Herberta8b690f2010-06-07 00:43:42 -07002231static void *tcp_seek_last_pos(struct seq_file *seq)
2232{
2233 struct tcp_iter_state *st = seq->private;
2234 int offset = st->offset;
2235 int orig_num = st->num;
2236 void *rc = NULL;
2237
2238 switch (st->state) {
2239 case TCP_SEQ_STATE_OPENREQ:
2240 case TCP_SEQ_STATE_LISTENING:
2241 if (st->bucket >= INET_LHTABLE_SIZE)
2242 break;
2243 st->state = TCP_SEQ_STATE_LISTENING;
2244 rc = listening_get_next(seq, NULL);
2245 while (offset-- && rc)
2246 rc = listening_get_next(seq, rc);
2247 if (rc)
2248 break;
2249 st->bucket = 0;
2250 /* Fallthrough */
2251 case TCP_SEQ_STATE_ESTABLISHED:
2252 case TCP_SEQ_STATE_TIME_WAIT:
2253 st->state = TCP_SEQ_STATE_ESTABLISHED;
2254 if (st->bucket > tcp_hashinfo.ehash_mask)
2255 break;
2256 rc = established_get_first(seq);
2257 while (offset-- && rc)
2258 rc = established_get_next(seq, rc);
2259 }
2260
2261 st->num = orig_num;
2262
2263 return rc;
2264}
2265
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
2267{
Jianjun Kong5799de02008-11-03 02:49:10 -08002268 struct tcp_iter_state *st = seq->private;
Tom Herberta8b690f2010-06-07 00:43:42 -07002269 void *rc;
2270
2271 if (*pos && *pos == st->last_pos) {
2272 rc = tcp_seek_last_pos(seq);
2273 if (rc)
2274 goto out;
2275 }
2276
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277 st->state = TCP_SEQ_STATE_LISTENING;
2278 st->num = 0;
Tom Herberta8b690f2010-06-07 00:43:42 -07002279 st->bucket = 0;
2280 st->offset = 0;
2281 rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2282
2283out:
2284 st->last_pos = *pos;
2285 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286}
2287
2288static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2289{
Tom Herberta8b690f2010-06-07 00:43:42 -07002290 struct tcp_iter_state *st = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 void *rc = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292
2293 if (v == SEQ_START_TOKEN) {
2294 rc = tcp_get_idx(seq, 0);
2295 goto out;
2296 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297
2298 switch (st->state) {
2299 case TCP_SEQ_STATE_OPENREQ:
2300 case TCP_SEQ_STATE_LISTENING:
2301 rc = listening_get_next(seq, v);
2302 if (!rc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303 st->state = TCP_SEQ_STATE_ESTABLISHED;
Tom Herberta8b690f2010-06-07 00:43:42 -07002304 st->bucket = 0;
2305 st->offset = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 rc = established_get_first(seq);
2307 }
2308 break;
2309 case TCP_SEQ_STATE_ESTABLISHED:
2310 case TCP_SEQ_STATE_TIME_WAIT:
2311 rc = established_get_next(seq, v);
2312 break;
2313 }
2314out:
2315 ++*pos;
Tom Herberta8b690f2010-06-07 00:43:42 -07002316 st->last_pos = *pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 return rc;
2318}
2319
2320static void tcp_seq_stop(struct seq_file *seq, void *v)
2321{
Jianjun Kong5799de02008-11-03 02:49:10 -08002322 struct tcp_iter_state *st = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323
2324 switch (st->state) {
2325 case TCP_SEQ_STATE_OPENREQ:
2326 if (v) {
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002327 struct inet_connection_sock *icsk = inet_csk(st->syn_wait_sk);
2328 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 }
2330 case TCP_SEQ_STATE_LISTENING:
2331 if (v != SEQ_START_TOKEN)
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002332 spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333 break;
2334 case TCP_SEQ_STATE_TIME_WAIT:
2335 case TCP_SEQ_STATE_ESTABLISHED:
2336 if (v)
Eric Dumazet9db66bd2008-11-20 20:39:09 -08002337 spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 break;
2339 }
2340}
2341
2342static int tcp_seq_open(struct inode *inode, struct file *file)
2343{
2344 struct tcp_seq_afinfo *afinfo = PDE(inode)->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345 struct tcp_iter_state *s;
Denis V. Lunev52d6f3f2008-04-13 22:12:41 -07002346 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347
Denis V. Lunev52d6f3f2008-04-13 22:12:41 -07002348 err = seq_open_net(inode, file, &afinfo->seq_ops,
2349 sizeof(struct tcp_iter_state));
2350 if (err < 0)
2351 return err;
Daniel Lezcanof40c8172008-03-21 04:13:54 -07002352
Denis V. Lunev52d6f3f2008-04-13 22:12:41 -07002353 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354 s->family = afinfo->family;
Tom Herberta8b690f2010-06-07 00:43:42 -07002355 s->last_pos = 0;
Daniel Lezcanof40c8172008-03-21 04:13:54 -07002356 return 0;
2357}
2358
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07002359int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360{
2361 int rc = 0;
2362 struct proc_dir_entry *p;
2363
Denis V. Lunev68fcadd2008-04-13 22:13:30 -07002364 afinfo->seq_fops.open = tcp_seq_open;
2365 afinfo->seq_fops.read = seq_read;
2366 afinfo->seq_fops.llseek = seq_lseek;
2367 afinfo->seq_fops.release = seq_release_net;
Arnaldo Carvalho de Melo71742592006-11-17 10:57:30 -02002368
Denis V. Lunev9427c4b2008-04-13 22:12:13 -07002369 afinfo->seq_ops.start = tcp_seq_start;
2370 afinfo->seq_ops.next = tcp_seq_next;
2371 afinfo->seq_ops.stop = tcp_seq_stop;
2372
Denis V. Lunev84841c32008-05-02 04:10:08 -07002373 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2374 &afinfo->seq_fops, afinfo);
2375 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 rc = -ENOMEM;
2377 return rc;
2378}
Eric Dumazet4bc2f182010-07-09 21:22:10 +00002379EXPORT_SYMBOL(tcp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07002381void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382{
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07002383 proc_net_remove(net, afinfo->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002384}
Eric Dumazet4bc2f182010-07-09 21:22:10 +00002385EXPORT_SYMBOL(tcp_proc_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386
Eric Dumazetcf533ea2011-10-21 05:22:42 -04002387static void get_openreq4(const struct sock *sk, const struct request_sock *req,
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002388 struct seq_file *f, int i, int uid, int *len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002390 const struct inet_request_sock *ireq = inet_rsk(req);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391 int ttd = req->expires - jiffies;
2392
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002393 seq_printf(f, "%4d: %08X:%04X %08X:%04X"
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002394 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %u %d %pK%n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395 i,
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002396 ireq->loc_addr,
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002397 ntohs(inet_sk(sk)->inet_sport),
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002398 ireq->rmt_addr,
2399 ntohs(ireq->rmt_port),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 TCP_SYN_RECV,
2401 0, 0, /* could print option size, but that is af dependent. */
2402 1, /* timers active (only the expire timer) */
2403 jiffies_to_clock_t(ttd),
2404 req->retrans,
2405 uid,
2406 0, /* non standard timer */
2407 0, /* open_requests have no inode */
2408 atomic_read(&sk->sk_refcnt),
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002409 req,
2410 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411}
2412
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002413static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i, int *len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414{
2415 int timer_active;
2416 unsigned long timer_expires;
Eric Dumazetcf533ea2011-10-21 05:22:42 -04002417 const struct tcp_sock *tp = tcp_sk(sk);
Ilpo Järvinencf4c6bf2007-02-22 01:13:58 -08002418 const struct inet_connection_sock *icsk = inet_csk(sk);
Eric Dumazetcf533ea2011-10-21 05:22:42 -04002419 const struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002420 __be32 dest = inet->inet_daddr;
2421 __be32 src = inet->inet_rcv_saddr;
2422 __u16 destp = ntohs(inet->inet_dport);
2423 __u16 srcp = ntohs(inet->inet_sport);
Eric Dumazet49d09002009-12-03 16:06:13 -08002424 int rx_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002426 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427 timer_active = 1;
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002428 timer_expires = icsk->icsk_timeout;
2429 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430 timer_active = 4;
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002431 timer_expires = icsk->icsk_timeout;
Ilpo Järvinencf4c6bf2007-02-22 01:13:58 -08002432 } else if (timer_pending(&sk->sk_timer)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433 timer_active = 2;
Ilpo Järvinencf4c6bf2007-02-22 01:13:58 -08002434 timer_expires = sk->sk_timer.expires;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 } else {
2436 timer_active = 0;
2437 timer_expires = jiffies;
2438 }
2439
Eric Dumazet49d09002009-12-03 16:06:13 -08002440 if (sk->sk_state == TCP_LISTEN)
2441 rx_queue = sk->sk_ack_backlog;
2442 else
2443 /*
2444 * because we dont lock socket, we might find a transient negative value
2445 */
2446 rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);
2447
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002448 seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002449 "%08X %5d %8d %lu %d %pK %lu %lu %u %u %d%n",
Ilpo Järvinencf4c6bf2007-02-22 01:13:58 -08002450 i, src, srcp, dest, destp, sk->sk_state,
Sridhar Samudrala47da8ee2006-06-27 13:29:00 -07002451 tp->write_seq - tp->snd_una,
Eric Dumazet49d09002009-12-03 16:06:13 -08002452 rx_queue,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 timer_active,
2454 jiffies_to_clock_t(timer_expires - jiffies),
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002455 icsk->icsk_retransmits,
Ilpo Järvinencf4c6bf2007-02-22 01:13:58 -08002456 sock_i_uid(sk),
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -03002457 icsk->icsk_probes_out,
Ilpo Järvinencf4c6bf2007-02-22 01:13:58 -08002458 sock_i_ino(sk),
2459 atomic_read(&sk->sk_refcnt), sk,
Stephen Hemminger7be87352008-06-27 20:00:19 -07002460 jiffies_to_clock_t(icsk->icsk_rto),
2461 jiffies_to_clock_t(icsk->icsk_ack.ato),
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002462 (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 tp->snd_cwnd,
Ilpo Järvinen0b6a05c2009-09-15 01:30:10 -07002464 tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh,
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002465 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466}
2467
Eric Dumazetcf533ea2011-10-21 05:22:42 -04002468static void get_timewait4_sock(const struct inet_timewait_sock *tw,
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002469 struct seq_file *f, int i, int *len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470{
Al Viro23f33c22006-09-27 18:43:50 -07002471 __be32 dest, src;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472 __u16 destp, srcp;
2473 int ttd = tw->tw_ttd - jiffies;
2474
2475 if (ttd < 0)
2476 ttd = 0;
2477
2478 dest = tw->tw_daddr;
2479 src = tw->tw_rcv_saddr;
2480 destp = ntohs(tw->tw_dport);
2481 srcp = ntohs(tw->tw_sport);
2482
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002483 seq_printf(f, "%4d: %08X:%04X %08X:%04X"
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002484 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK%n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002485 i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
2486 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002487 atomic_read(&tw->tw_refcnt), tw, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002488}
2489
2490#define TMPSZ 150
2491
2492static int tcp4_seq_show(struct seq_file *seq, void *v)
2493{
Jianjun Kong5799de02008-11-03 02:49:10 -08002494 struct tcp_iter_state *st;
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002495 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496
2497 if (v == SEQ_START_TOKEN) {
2498 seq_printf(seq, "%-*s\n", TMPSZ - 1,
2499 " sl local_address rem_address st tx_queue "
2500 "rx_queue tr tm->when retrnsmt uid timeout "
2501 "inode");
2502 goto out;
2503 }
2504 st = seq->private;
2505
2506 switch (st->state) {
2507 case TCP_SEQ_STATE_LISTENING:
2508 case TCP_SEQ_STATE_ESTABLISHED:
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002509 get_tcp4_sock(v, seq, st->num, &len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510 break;
2511 case TCP_SEQ_STATE_OPENREQ:
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002512 get_openreq4(st->syn_wait_sk, v, seq, st->num, st->uid, &len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513 break;
2514 case TCP_SEQ_STATE_TIME_WAIT:
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002515 get_timewait4_sock(v, seq, st->num, &len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 break;
2517 }
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002518 seq_printf(seq, "%*s\n", TMPSZ - 1 - len, "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002519out:
2520 return 0;
2521}
2522
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523static struct tcp_seq_afinfo tcp4_seq_afinfo = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524 .name = "tcp",
2525 .family = AF_INET,
Denis V. Lunev5f4472c2008-04-13 22:13:53 -07002526 .seq_fops = {
2527 .owner = THIS_MODULE,
2528 },
Denis V. Lunev9427c4b2008-04-13 22:12:13 -07002529 .seq_ops = {
2530 .show = tcp4_seq_show,
2531 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07002532};
2533
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002534static int __net_init tcp4_proc_init_net(struct net *net)
Pavel Emelyanov757764f2008-03-24 14:56:02 -07002535{
2536 return tcp_proc_register(net, &tcp4_seq_afinfo);
2537}
2538
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002539static void __net_exit tcp4_proc_exit_net(struct net *net)
Pavel Emelyanov757764f2008-03-24 14:56:02 -07002540{
2541 tcp_proc_unregister(net, &tcp4_seq_afinfo);
2542}
2543
2544static struct pernet_operations tcp4_net_ops = {
2545 .init = tcp4_proc_init_net,
2546 .exit = tcp4_proc_exit_net,
2547};
2548
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549int __init tcp4_proc_init(void)
2550{
Pavel Emelyanov757764f2008-03-24 14:56:02 -07002551 return register_pernet_subsys(&tcp4_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002552}
2553
2554void tcp4_proc_exit(void)
2555{
Pavel Emelyanov757764f2008-03-24 14:56:02 -07002556 unregister_pernet_subsys(&tcp4_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557}
2558#endif /* CONFIG_PROC_FS */
2559
Herbert Xubf296b12008-12-15 23:43:36 -08002560struct sk_buff **tcp4_gro_receive(struct sk_buff **head, struct sk_buff *skb)
2561{
Eric Dumazetb71d1d42011-04-22 04:53:02 +00002562 const struct iphdr *iph = skb_gro_network_header(skb);
Herbert Xubf296b12008-12-15 23:43:36 -08002563
2564 switch (skb->ip_summed) {
2565 case CHECKSUM_COMPLETE:
Herbert Xu86911732009-01-29 14:19:50 +00002566 if (!tcp_v4_check(skb_gro_len(skb), iph->saddr, iph->daddr,
Herbert Xubf296b12008-12-15 23:43:36 -08002567 skb->csum)) {
2568 skb->ip_summed = CHECKSUM_UNNECESSARY;
2569 break;
2570 }
2571
2572 /* fall through */
2573 case CHECKSUM_NONE:
2574 NAPI_GRO_CB(skb)->flush = 1;
2575 return NULL;
2576 }
2577
2578 return tcp_gro_receive(head, skb);
2579}
Herbert Xubf296b12008-12-15 23:43:36 -08002580
2581int tcp4_gro_complete(struct sk_buff *skb)
2582{
Eric Dumazetb71d1d42011-04-22 04:53:02 +00002583 const struct iphdr *iph = ip_hdr(skb);
Herbert Xubf296b12008-12-15 23:43:36 -08002584 struct tcphdr *th = tcp_hdr(skb);
2585
2586 th->check = ~tcp_v4_check(skb->len - skb_transport_offset(skb),
2587 iph->saddr, iph->daddr, 0);
2588 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
2589
2590 return tcp_gro_complete(skb);
2591}
Herbert Xubf296b12008-12-15 23:43:36 -08002592
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593struct proto tcp_prot = {
2594 .name = "TCP",
2595 .owner = THIS_MODULE,
2596 .close = tcp_close,
2597 .connect = tcp_v4_connect,
2598 .disconnect = tcp_disconnect,
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07002599 .accept = inet_csk_accept,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002600 .ioctl = tcp_ioctl,
2601 .init = tcp_v4_init_sock,
2602 .destroy = tcp_v4_destroy_sock,
2603 .shutdown = tcp_shutdown,
2604 .setsockopt = tcp_setsockopt,
2605 .getsockopt = tcp_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002606 .recvmsg = tcp_recvmsg,
Changli Gao7ba42912010-07-10 20:41:55 +00002607 .sendmsg = tcp_sendmsg,
2608 .sendpage = tcp_sendpage,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609 .backlog_rcv = tcp_v4_do_rcv,
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08002610 .hash = inet_hash,
2611 .unhash = inet_unhash,
2612 .get_port = inet_csk_get_port,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002613 .enter_memory_pressure = tcp_enter_memory_pressure,
2614 .sockets_allocated = &tcp_sockets_allocated,
Arnaldo Carvalho de Melo0a5578c2005-08-09 20:11:41 -07002615 .orphan_count = &tcp_orphan_count,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002616 .memory_allocated = &tcp_memory_allocated,
2617 .memory_pressure = &tcp_memory_pressure,
2618 .sysctl_mem = sysctl_tcp_mem,
2619 .sysctl_wmem = sysctl_tcp_wmem,
2620 .sysctl_rmem = sysctl_tcp_rmem,
2621 .max_header = MAX_TCP_HEADER,
2622 .obj_size = sizeof(struct tcp_sock),
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002623 .slab_flags = SLAB_DESTROY_BY_RCU,
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002624 .twsk_prot = &tcp_timewait_sock_ops,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07002625 .rsk_prot = &tcp_request_sock_ops,
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07002626 .h.hashinfo = &tcp_hashinfo,
Changli Gao7ba42912010-07-10 20:41:55 +00002627 .no_autobind = true,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002628#ifdef CONFIG_COMPAT
2629 .compat_setsockopt = compat_tcp_setsockopt,
2630 .compat_getsockopt = compat_tcp_getsockopt,
2631#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002632};
Eric Dumazet4bc2f182010-07-09 21:22:10 +00002633EXPORT_SYMBOL(tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634
Denis V. Lunev046ee902008-04-03 14:31:33 -07002635
2636static int __net_init tcp_sk_init(struct net *net)
2637{
2638 return inet_ctl_sock_create(&net->ipv4.tcp_sock,
2639 PF_INET, SOCK_RAW, IPPROTO_TCP, net);
2640}
2641
2642static void __net_exit tcp_sk_exit(struct net *net)
2643{
2644 inet_ctl_sock_destroy(net->ipv4.tcp_sock);
Eric W. Biedermanb099ce22009-12-03 02:29:09 +00002645}
2646
2647static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
2648{
2649 inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET);
Denis V. Lunev046ee902008-04-03 14:31:33 -07002650}
2651
2652static struct pernet_operations __net_initdata tcp_sk_ops = {
Eric W. Biedermanb099ce22009-12-03 02:29:09 +00002653 .init = tcp_sk_init,
2654 .exit = tcp_sk_exit,
2655 .exit_batch = tcp_sk_exit_batch,
Denis V. Lunev046ee902008-04-03 14:31:33 -07002656};
2657
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08002658void __init tcp_v4_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659{
Eric Dumazet5caea4e2008-11-20 00:40:07 -08002660 inet_hashinfo_init(&tcp_hashinfo);
Eric W. Biederman6a1b3052009-02-22 00:10:18 -08002661 if (register_pernet_subsys(&tcp_sk_ops))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662 panic("Failed to create the TCP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663}