]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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 | * | |
1da177e4 LT |
8 | * 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 | |
60236fdd | 37 | * request_sock handling and moved |
1da177e4 LT |
38 | * most of it into the af independent code. |
39 | * Added tail drop and some other bugfixes. | |
caa20d9a | 40 | * Added new listen semantics. |
1da177e4 LT |
41 | * 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 | ||
1da177e4 | 53 | |
eb4dea58 | 54 | #include <linux/bottom_half.h> |
1da177e4 LT |
55 | #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> | |
63 | ||
457c4cbc | 64 | #include <net/net_namespace.h> |
1da177e4 | 65 | #include <net/icmp.h> |
304a1618 | 66 | #include <net/inet_hashtables.h> |
1da177e4 | 67 | #include <net/tcp.h> |
20380731 | 68 | #include <net/transp_v6.h> |
1da177e4 LT |
69 | #include <net/ipv6.h> |
70 | #include <net/inet_common.h> | |
6d6ee43e | 71 | #include <net/timewait_sock.h> |
1da177e4 | 72 | #include <net/xfrm.h> |
1a2449a8 | 73 | #include <net/netdma.h> |
1da177e4 LT |
74 | |
75 | #include <linux/inet.h> | |
76 | #include <linux/ipv6.h> | |
77 | #include <linux/stddef.h> | |
78 | #include <linux/proc_fs.h> | |
79 | #include <linux/seq_file.h> | |
80 | ||
cfb6eeb4 YH |
81 | #include <linux/crypto.h> |
82 | #include <linux/scatterlist.h> | |
83 | ||
ab32ea5d BH |
84 | int sysctl_tcp_tw_reuse __read_mostly; |
85 | int sysctl_tcp_low_latency __read_mostly; | |
1da177e4 | 86 | |
1da177e4 | 87 | |
cfb6eeb4 | 88 | #ifdef CONFIG_TCP_MD5SIG |
7174259e ACM |
89 | static struct tcp_md5sig_key *tcp_v4_md5_do_lookup(struct sock *sk, |
90 | __be32 addr); | |
49a72dfb AL |
91 | static int tcp_v4_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, |
92 | __be32 daddr, __be32 saddr, struct tcphdr *th); | |
9501f972 YH |
93 | #else |
94 | static inline | |
95 | struct tcp_md5sig_key *tcp_v4_md5_do_lookup(struct sock *sk, __be32 addr) | |
96 | { | |
97 | return NULL; | |
98 | } | |
cfb6eeb4 YH |
99 | #endif |
100 | ||
5caea4ea | 101 | struct inet_hashinfo tcp_hashinfo; |
1da177e4 | 102 | |
a94f723d | 103 | static inline __u32 tcp_v4_init_sequence(struct sk_buff *skb) |
1da177e4 | 104 | { |
eddc9ec5 ACM |
105 | return secure_tcp_sequence_number(ip_hdr(skb)->daddr, |
106 | ip_hdr(skb)->saddr, | |
aa8223c7 ACM |
107 | tcp_hdr(skb)->dest, |
108 | tcp_hdr(skb)->source); | |
1da177e4 LT |
109 | } |
110 | ||
6d6ee43e ACM |
111 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp) |
112 | { | |
113 | const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw); | |
114 | struct tcp_sock *tp = tcp_sk(sk); | |
115 | ||
116 | /* With PAWS, it is safe from the viewpoint | |
117 | of data integrity. Even without PAWS it is safe provided sequence | |
118 | spaces do not overlap i.e. at data rates <= 80Mbit/sec. | |
119 | ||
120 | Actually, the idea is close to VJ's one, only timestamp cache is | |
121 | held not per host, but per port pair and TW bucket is used as state | |
122 | holder. | |
123 | ||
124 | If TW bucket has been already destroyed we fall back to VJ's scheme | |
125 | and use initial timestamp retrieved from peer table. | |
126 | */ | |
127 | if (tcptw->tw_ts_recent_stamp && | |
128 | (twp == NULL || (sysctl_tcp_tw_reuse && | |
9d729f72 | 129 | get_seconds() - tcptw->tw_ts_recent_stamp > 1))) { |
6d6ee43e ACM |
130 | tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2; |
131 | if (tp->write_seq == 0) | |
132 | tp->write_seq = 1; | |
133 | tp->rx_opt.ts_recent = tcptw->tw_ts_recent; | |
134 | tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp; | |
135 | sock_hold(sktw); | |
136 | return 1; | |
137 | } | |
138 | ||
139 | return 0; | |
140 | } | |
141 | ||
142 | EXPORT_SYMBOL_GPL(tcp_twsk_unique); | |
143 | ||
1da177e4 LT |
144 | /* This will initiate an outgoing connection. */ |
145 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
146 | { | |
147 | struct inet_sock *inet = inet_sk(sk); | |
148 | struct tcp_sock *tp = tcp_sk(sk); | |
149 | struct sockaddr_in *usin = (struct sockaddr_in *)uaddr; | |
150 | struct rtable *rt; | |
bada8adc | 151 | __be32 daddr, nexthop; |
1da177e4 LT |
152 | int tmp; |
153 | int err; | |
154 | ||
155 | if (addr_len < sizeof(struct sockaddr_in)) | |
156 | return -EINVAL; | |
157 | ||
158 | if (usin->sin_family != AF_INET) | |
159 | return -EAFNOSUPPORT; | |
160 | ||
161 | nexthop = daddr = usin->sin_addr.s_addr; | |
162 | if (inet->opt && inet->opt->srr) { | |
163 | if (!daddr) | |
164 | return -EINVAL; | |
165 | nexthop = inet->opt->faddr; | |
166 | } | |
167 | ||
168 | tmp = ip_route_connect(&rt, nexthop, inet->saddr, | |
169 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if, | |
170 | IPPROTO_TCP, | |
8eb9086f | 171 | inet->sport, usin->sin_port, sk, 1); |
584bdf8c WD |
172 | if (tmp < 0) { |
173 | if (tmp == -ENETUNREACH) | |
7c73a6fa | 174 | IP_INC_STATS_BH(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); |
1da177e4 | 175 | return tmp; |
584bdf8c | 176 | } |
1da177e4 LT |
177 | |
178 | if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) { | |
179 | ip_rt_put(rt); | |
180 | return -ENETUNREACH; | |
181 | } | |
182 | ||
183 | if (!inet->opt || !inet->opt->srr) | |
184 | daddr = rt->rt_dst; | |
185 | ||
186 | if (!inet->saddr) | |
187 | inet->saddr = rt->rt_src; | |
188 | inet->rcv_saddr = inet->saddr; | |
189 | ||
190 | if (tp->rx_opt.ts_recent_stamp && inet->daddr != daddr) { | |
191 | /* Reset inherited state */ | |
192 | tp->rx_opt.ts_recent = 0; | |
193 | tp->rx_opt.ts_recent_stamp = 0; | |
194 | tp->write_seq = 0; | |
195 | } | |
196 | ||
295ff7ed | 197 | if (tcp_death_row.sysctl_tw_recycle && |
1da177e4 LT |
198 | !tp->rx_opt.ts_recent_stamp && rt->rt_dst == daddr) { |
199 | struct inet_peer *peer = rt_get_peer(rt); | |
7174259e ACM |
200 | /* |
201 | * VJ's idea. We save last timestamp seen from | |
202 | * the destination in peer table, when entering state | |
203 | * TIME-WAIT * and initialize rx_opt.ts_recent from it, | |
204 | * when trying new connection. | |
1da177e4 | 205 | */ |
7174259e | 206 | if (peer != NULL && |
9d729f72 | 207 | peer->tcp_ts_stamp + TCP_PAWS_MSL >= get_seconds()) { |
1da177e4 LT |
208 | tp->rx_opt.ts_recent_stamp = peer->tcp_ts_stamp; |
209 | tp->rx_opt.ts_recent = peer->tcp_ts; | |
210 | } | |
211 | } | |
212 | ||
213 | inet->dport = usin->sin_port; | |
214 | inet->daddr = daddr; | |
215 | ||
d83d8461 | 216 | inet_csk(sk)->icsk_ext_hdr_len = 0; |
1da177e4 | 217 | if (inet->opt) |
d83d8461 | 218 | inet_csk(sk)->icsk_ext_hdr_len = inet->opt->optlen; |
1da177e4 LT |
219 | |
220 | tp->rx_opt.mss_clamp = 536; | |
221 | ||
222 | /* Socket identity is still unknown (sport may be zero). | |
223 | * However we set state to SYN-SENT and not releasing socket | |
224 | * lock select source port, enter ourselves into the hash tables and | |
225 | * complete initialization after this. | |
226 | */ | |
227 | tcp_set_state(sk, TCP_SYN_SENT); | |
a7f5e7f1 | 228 | err = inet_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
229 | if (err) |
230 | goto failure; | |
231 | ||
7174259e ACM |
232 | err = ip_route_newports(&rt, IPPROTO_TCP, |
233 | inet->sport, inet->dport, sk); | |
1da177e4 LT |
234 | if (err) |
235 | goto failure; | |
236 | ||
237 | /* OK, now commit destination to socket. */ | |
bcd76111 | 238 | sk->sk_gso_type = SKB_GSO_TCPV4; |
6cbb0df7 | 239 | sk_setup_caps(sk, &rt->u.dst); |
1da177e4 LT |
240 | |
241 | if (!tp->write_seq) | |
242 | tp->write_seq = secure_tcp_sequence_number(inet->saddr, | |
243 | inet->daddr, | |
244 | inet->sport, | |
245 | usin->sin_port); | |
246 | ||
247 | inet->id = tp->write_seq ^ jiffies; | |
248 | ||
249 | err = tcp_connect(sk); | |
250 | rt = NULL; | |
251 | if (err) | |
252 | goto failure; | |
253 | ||
254 | return 0; | |
255 | ||
256 | failure: | |
7174259e ACM |
257 | /* |
258 | * This unhashes the socket and releases the local port, | |
259 | * if necessary. | |
260 | */ | |
1da177e4 LT |
261 | tcp_set_state(sk, TCP_CLOSE); |
262 | ip_rt_put(rt); | |
263 | sk->sk_route_caps = 0; | |
264 | inet->dport = 0; | |
265 | return err; | |
266 | } | |
267 | ||
1da177e4 LT |
268 | /* |
269 | * This routine does path mtu discovery as defined in RFC1191. | |
270 | */ | |
40efc6fa | 271 | static void do_pmtu_discovery(struct sock *sk, struct iphdr *iph, u32 mtu) |
1da177e4 LT |
272 | { |
273 | struct dst_entry *dst; | |
274 | struct inet_sock *inet = inet_sk(sk); | |
1da177e4 LT |
275 | |
276 | /* We are not interested in TCP_LISTEN and open_requests (SYN-ACKs | |
277 | * send out by Linux are always <576bytes so they should go through | |
278 | * unfragmented). | |
279 | */ | |
280 | if (sk->sk_state == TCP_LISTEN) | |
281 | return; | |
282 | ||
283 | /* We don't check in the destentry if pmtu discovery is forbidden | |
284 | * on this route. We just assume that no packet_to_big packets | |
285 | * are send back when pmtu discovery is not active. | |
e905a9ed | 286 | * There is a small race when the user changes this flag in the |
1da177e4 LT |
287 | * route, but I think that's acceptable. |
288 | */ | |
289 | if ((dst = __sk_dst_check(sk, 0)) == NULL) | |
290 | return; | |
291 | ||
292 | dst->ops->update_pmtu(dst, mtu); | |
293 | ||
294 | /* Something is about to be wrong... Remember soft error | |
295 | * for the case, if this connection will not able to recover. | |
296 | */ | |
297 | if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst)) | |
298 | sk->sk_err_soft = EMSGSIZE; | |
299 | ||
300 | mtu = dst_mtu(dst); | |
301 | ||
302 | if (inet->pmtudisc != IP_PMTUDISC_DONT && | |
d83d8461 | 303 | inet_csk(sk)->icsk_pmtu_cookie > mtu) { |
1da177e4 LT |
304 | tcp_sync_mss(sk, mtu); |
305 | ||
306 | /* Resend the TCP packet because it's | |
307 | * clear that the old packet has been | |
308 | * dropped. This is the new "fast" path mtu | |
309 | * discovery. | |
310 | */ | |
311 | tcp_simple_retransmit(sk); | |
312 | } /* else let the usual retransmit timer handle it */ | |
313 | } | |
314 | ||
315 | /* | |
316 | * This routine is called by the ICMP module when it gets some | |
317 | * sort of error condition. If err < 0 then the socket should | |
318 | * be closed and the error returned to the user. If err > 0 | |
319 | * it's just the icmp type << 8 | icmp code. After adjustment | |
320 | * header points to the first 8 bytes of the tcp header. We need | |
321 | * to find the appropriate port. | |
322 | * | |
323 | * The locking strategy used here is very "optimistic". When | |
324 | * someone else accesses the socket the ICMP is just dropped | |
325 | * and for some paths there is no check at all. | |
326 | * A more general error queue to queue errors for later handling | |
327 | * is probably better. | |
328 | * | |
329 | */ | |
330 | ||
331 | void tcp_v4_err(struct sk_buff *skb, u32 info) | |
332 | { | |
333 | struct iphdr *iph = (struct iphdr *)skb->data; | |
334 | struct tcphdr *th = (struct tcphdr *)(skb->data + (iph->ihl << 2)); | |
335 | struct tcp_sock *tp; | |
336 | struct inet_sock *inet; | |
88c7664f ACM |
337 | const int type = icmp_hdr(skb)->type; |
338 | const int code = icmp_hdr(skb)->code; | |
1da177e4 LT |
339 | struct sock *sk; |
340 | __u32 seq; | |
341 | int err; | |
fd54d716 | 342 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
343 | |
344 | if (skb->len < (iph->ihl << 2) + 8) { | |
dcfc23ca | 345 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
346 | return; |
347 | } | |
348 | ||
fd54d716 | 349 | sk = inet_lookup(net, &tcp_hashinfo, iph->daddr, th->dest, |
c67499c0 | 350 | iph->saddr, th->source, inet_iif(skb)); |
1da177e4 | 351 | if (!sk) { |
dcfc23ca | 352 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
353 | return; |
354 | } | |
355 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 356 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
357 | return; |
358 | } | |
359 | ||
360 | bh_lock_sock(sk); | |
361 | /* If too many ICMPs get dropped on busy | |
362 | * servers this needs to be solved differently. | |
363 | */ | |
364 | if (sock_owned_by_user(sk)) | |
de0744af | 365 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); |
1da177e4 LT |
366 | |
367 | if (sk->sk_state == TCP_CLOSE) | |
368 | goto out; | |
369 | ||
370 | tp = tcp_sk(sk); | |
371 | seq = ntohl(th->seq); | |
372 | if (sk->sk_state != TCP_LISTEN && | |
373 | !between(seq, tp->snd_una, tp->snd_nxt)) { | |
de0744af | 374 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
375 | goto out; |
376 | } | |
377 | ||
378 | switch (type) { | |
379 | case ICMP_SOURCE_QUENCH: | |
380 | /* Just silently ignore these. */ | |
381 | goto out; | |
382 | case ICMP_PARAMETERPROB: | |
383 | err = EPROTO; | |
384 | break; | |
385 | case ICMP_DEST_UNREACH: | |
386 | if (code > NR_ICMP_UNREACH) | |
387 | goto out; | |
388 | ||
389 | if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */ | |
390 | if (!sock_owned_by_user(sk)) | |
391 | do_pmtu_discovery(sk, iph, info); | |
392 | goto out; | |
393 | } | |
394 | ||
395 | err = icmp_err_convert[code].errno; | |
396 | break; | |
397 | case ICMP_TIME_EXCEEDED: | |
398 | err = EHOSTUNREACH; | |
399 | break; | |
400 | default: | |
401 | goto out; | |
402 | } | |
403 | ||
404 | switch (sk->sk_state) { | |
60236fdd | 405 | struct request_sock *req, **prev; |
1da177e4 LT |
406 | case TCP_LISTEN: |
407 | if (sock_owned_by_user(sk)) | |
408 | goto out; | |
409 | ||
463c84b9 ACM |
410 | req = inet_csk_search_req(sk, &prev, th->dest, |
411 | iph->daddr, iph->saddr); | |
1da177e4 LT |
412 | if (!req) |
413 | goto out; | |
414 | ||
415 | /* ICMPs are not backlogged, hence we cannot get | |
416 | an established socket here. | |
417 | */ | |
547b792c | 418 | WARN_ON(req->sk); |
1da177e4 | 419 | |
2e6599cb | 420 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 421 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
422 | goto out; |
423 | } | |
424 | ||
425 | /* | |
426 | * Still in SYN_RECV, just remove it silently. | |
427 | * There is no good way to pass the error to the newly | |
428 | * created socket, and POSIX does not want network | |
429 | * errors returned from accept(). | |
430 | */ | |
463c84b9 | 431 | inet_csk_reqsk_queue_drop(sk, req, prev); |
1da177e4 LT |
432 | goto out; |
433 | ||
434 | case TCP_SYN_SENT: | |
435 | case TCP_SYN_RECV: /* Cannot happen. | |
436 | It can f.e. if SYNs crossed. | |
437 | */ | |
438 | if (!sock_owned_by_user(sk)) { | |
1da177e4 LT |
439 | sk->sk_err = err; |
440 | ||
441 | sk->sk_error_report(sk); | |
442 | ||
443 | tcp_done(sk); | |
444 | } else { | |
445 | sk->sk_err_soft = err; | |
446 | } | |
447 | goto out; | |
448 | } | |
449 | ||
450 | /* If we've already connected we will keep trying | |
451 | * until we time out, or the user gives up. | |
452 | * | |
453 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
454 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
455 | * but it is obsoleted by pmtu discovery). | |
456 | * | |
457 | * Note, that in modern internet, where routing is unreliable | |
458 | * and in each dark corner broken firewalls sit, sending random | |
459 | * errors ordered by their masters even this two messages finally lose | |
460 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
461 | * | |
462 | * Now we are in compliance with RFCs. | |
463 | * --ANK (980905) | |
464 | */ | |
465 | ||
466 | inet = inet_sk(sk); | |
467 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
468 | sk->sk_err = err; | |
469 | sk->sk_error_report(sk); | |
470 | } else { /* Only an error on timeout */ | |
471 | sk->sk_err_soft = err; | |
472 | } | |
473 | ||
474 | out: | |
475 | bh_unlock_sock(sk); | |
476 | sock_put(sk); | |
477 | } | |
478 | ||
479 | /* This routine computes an IPv4 TCP checksum. */ | |
8292a17a | 480 | void tcp_v4_send_check(struct sock *sk, int len, struct sk_buff *skb) |
1da177e4 LT |
481 | { |
482 | struct inet_sock *inet = inet_sk(sk); | |
aa8223c7 | 483 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 484 | |
84fa7933 | 485 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
ba7808ea FD |
486 | th->check = ~tcp_v4_check(len, inet->saddr, |
487 | inet->daddr, 0); | |
663ead3b | 488 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 489 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 490 | } else { |
ba7808ea | 491 | th->check = tcp_v4_check(len, inet->saddr, inet->daddr, |
07f0757a | 492 | csum_partial(th, |
1da177e4 LT |
493 | th->doff << 2, |
494 | skb->csum)); | |
495 | } | |
496 | } | |
497 | ||
a430a43d HX |
498 | int tcp_v4_gso_send_check(struct sk_buff *skb) |
499 | { | |
eddc9ec5 | 500 | const struct iphdr *iph; |
a430a43d HX |
501 | struct tcphdr *th; |
502 | ||
503 | if (!pskb_may_pull(skb, sizeof(*th))) | |
504 | return -EINVAL; | |
505 | ||
eddc9ec5 | 506 | iph = ip_hdr(skb); |
aa8223c7 | 507 | th = tcp_hdr(skb); |
a430a43d HX |
508 | |
509 | th->check = 0; | |
ba7808ea | 510 | th->check = ~tcp_v4_check(skb->len, iph->saddr, iph->daddr, 0); |
663ead3b | 511 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 512 | skb->csum_offset = offsetof(struct tcphdr, check); |
84fa7933 | 513 | skb->ip_summed = CHECKSUM_PARTIAL; |
a430a43d HX |
514 | return 0; |
515 | } | |
516 | ||
1da177e4 LT |
517 | /* |
518 | * This routine will send an RST to the other tcp. | |
519 | * | |
520 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
521 | * for reset. | |
522 | * Answer: if a packet caused RST, it is not for a socket | |
523 | * existing in our system, if it is matched to a socket, | |
524 | * it is just duplicate segment or bug in other side's TCP. | |
525 | * So that we build reply only basing on parameters | |
526 | * arrived with segment. | |
527 | * Exception: precedence violation. We do not implement it in any case. | |
528 | */ | |
529 | ||
cfb6eeb4 | 530 | static void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 531 | { |
aa8223c7 | 532 | struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
533 | struct { |
534 | struct tcphdr th; | |
535 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 536 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
537 | #endif |
538 | } rep; | |
1da177e4 | 539 | struct ip_reply_arg arg; |
cfb6eeb4 YH |
540 | #ifdef CONFIG_TCP_MD5SIG |
541 | struct tcp_md5sig_key *key; | |
542 | #endif | |
a86b1e30 | 543 | struct net *net; |
1da177e4 LT |
544 | |
545 | /* Never send a reset in response to a reset. */ | |
546 | if (th->rst) | |
547 | return; | |
548 | ||
ee6b9673 | 549 | if (skb->rtable->rt_type != RTN_LOCAL) |
1da177e4 LT |
550 | return; |
551 | ||
552 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
553 | memset(&rep, 0, sizeof(rep)); |
554 | rep.th.dest = th->source; | |
555 | rep.th.source = th->dest; | |
556 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
557 | rep.th.rst = 1; | |
1da177e4 LT |
558 | |
559 | if (th->ack) { | |
cfb6eeb4 | 560 | rep.th.seq = th->ack_seq; |
1da177e4 | 561 | } else { |
cfb6eeb4 YH |
562 | rep.th.ack = 1; |
563 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
564 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
565 | } |
566 | ||
7174259e | 567 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
568 | arg.iov[0].iov_base = (unsigned char *)&rep; |
569 | arg.iov[0].iov_len = sizeof(rep.th); | |
570 | ||
571 | #ifdef CONFIG_TCP_MD5SIG | |
eddc9ec5 | 572 | key = sk ? tcp_v4_md5_do_lookup(sk, ip_hdr(skb)->daddr) : NULL; |
cfb6eeb4 YH |
573 | if (key) { |
574 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
575 | (TCPOPT_NOP << 16) | | |
576 | (TCPOPT_MD5SIG << 8) | | |
577 | TCPOLEN_MD5SIG); | |
578 | /* Update length and the length the header thinks exists */ | |
579 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
580 | rep.th.doff = arg.iov[0].iov_len / 4; | |
581 | ||
49a72dfb | 582 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
583 | key, ip_hdr(skb)->saddr, |
584 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
585 | } |
586 | #endif | |
eddc9ec5 ACM |
587 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
588 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 589 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 590 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
88ef4a5a | 591 | arg.flags = (sk && inet_sk(sk)->transparent) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
1da177e4 | 592 | |
a86b1e30 PE |
593 | net = dev_net(skb->dst->dev); |
594 | ip_send_reply(net->ipv4.tcp_sock, skb, | |
7feb49c8 | 595 | &arg, arg.iov[0].iov_len); |
1da177e4 | 596 | |
63231bdd PE |
597 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
598 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
1da177e4 LT |
599 | } |
600 | ||
601 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
602 | outside socket context is ugly, certainly. What can I do? | |
603 | */ | |
604 | ||
9501f972 YH |
605 | static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack, |
606 | u32 win, u32 ts, int oif, | |
88ef4a5a KK |
607 | struct tcp_md5sig_key *key, |
608 | int reply_flags) | |
1da177e4 | 609 | { |
aa8223c7 | 610 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
611 | struct { |
612 | struct tcphdr th; | |
714e85be | 613 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 614 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 615 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
616 | #endif |
617 | ]; | |
1da177e4 LT |
618 | } rep; |
619 | struct ip_reply_arg arg; | |
4dd7972d | 620 | struct net *net = dev_net(skb->dst->dev); |
1da177e4 LT |
621 | |
622 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 623 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
624 | |
625 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
626 | arg.iov[0].iov_len = sizeof(rep.th); | |
627 | if (ts) { | |
cfb6eeb4 YH |
628 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
629 | (TCPOPT_TIMESTAMP << 8) | | |
630 | TCPOLEN_TIMESTAMP); | |
631 | rep.opt[1] = htonl(tcp_time_stamp); | |
632 | rep.opt[2] = htonl(ts); | |
cb48cfe8 | 633 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
634 | } |
635 | ||
636 | /* Swap the send and the receive. */ | |
637 | rep.th.dest = th->source; | |
638 | rep.th.source = th->dest; | |
639 | rep.th.doff = arg.iov[0].iov_len / 4; | |
640 | rep.th.seq = htonl(seq); | |
641 | rep.th.ack_seq = htonl(ack); | |
642 | rep.th.ack = 1; | |
643 | rep.th.window = htons(win); | |
644 | ||
cfb6eeb4 | 645 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
646 | if (key) { |
647 | int offset = (ts) ? 3 : 0; | |
648 | ||
649 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
650 | (TCPOPT_NOP << 16) | | |
651 | (TCPOPT_MD5SIG << 8) | | |
652 | TCPOLEN_MD5SIG); | |
653 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
654 | rep.th.doff = arg.iov[0].iov_len/4; | |
655 | ||
49a72dfb | 656 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
657 | key, ip_hdr(skb)->saddr, |
658 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
659 | } |
660 | #endif | |
88ef4a5a | 661 | arg.flags = reply_flags; |
eddc9ec5 ACM |
662 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
663 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
664 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
665 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
666 | if (oif) |
667 | arg.bound_dev_if = oif; | |
1da177e4 | 668 | |
a86b1e30 | 669 | ip_send_reply(net->ipv4.tcp_sock, skb, |
7feb49c8 | 670 | &arg, arg.iov[0].iov_len); |
1da177e4 | 671 | |
63231bdd | 672 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
1da177e4 LT |
673 | } |
674 | ||
675 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
676 | { | |
8feaf0c0 | 677 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 678 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 679 | |
9501f972 | 680 | tcp_v4_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 681 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
9501f972 YH |
682 | tcptw->tw_ts_recent, |
683 | tw->tw_bound_dev_if, | |
88ef4a5a KK |
684 | tcp_twsk_md5_key(tcptw), |
685 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0 | |
9501f972 | 686 | ); |
1da177e4 | 687 | |
8feaf0c0 | 688 | inet_twsk_put(tw); |
1da177e4 LT |
689 | } |
690 | ||
6edafaaf | 691 | static void tcp_v4_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
7174259e | 692 | struct request_sock *req) |
1da177e4 | 693 | { |
9501f972 | 694 | tcp_v4_send_ack(skb, tcp_rsk(req)->snt_isn + 1, |
cfb6eeb4 | 695 | tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, |
9501f972 YH |
696 | req->ts_recent, |
697 | 0, | |
88ef4a5a KK |
698 | tcp_v4_md5_do_lookup(sk, ip_hdr(skb)->daddr), |
699 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0); | |
1da177e4 LT |
700 | } |
701 | ||
1da177e4 | 702 | /* |
9bf1d83e | 703 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 704 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
705 | * socket. |
706 | */ | |
fd80eb94 DL |
707 | static int __tcp_v4_send_synack(struct sock *sk, struct request_sock *req, |
708 | struct dst_entry *dst) | |
1da177e4 | 709 | { |
2e6599cb | 710 | const struct inet_request_sock *ireq = inet_rsk(req); |
1da177e4 LT |
711 | int err = -1; |
712 | struct sk_buff * skb; | |
713 | ||
714 | /* First, grab a route. */ | |
463c84b9 | 715 | if (!dst && (dst = inet_csk_route_req(sk, req)) == NULL) |
fd80eb94 | 716 | return -1; |
1da177e4 LT |
717 | |
718 | skb = tcp_make_synack(sk, dst, req); | |
719 | ||
720 | if (skb) { | |
aa8223c7 | 721 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 722 | |
ba7808ea | 723 | th->check = tcp_v4_check(skb->len, |
2e6599cb ACM |
724 | ireq->loc_addr, |
725 | ireq->rmt_addr, | |
07f0757a | 726 | csum_partial(th, skb->len, |
1da177e4 LT |
727 | skb->csum)); |
728 | ||
2e6599cb ACM |
729 | err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr, |
730 | ireq->rmt_addr, | |
731 | ireq->opt); | |
b9df3cb8 | 732 | err = net_xmit_eval(err); |
1da177e4 LT |
733 | } |
734 | ||
1da177e4 LT |
735 | dst_release(dst); |
736 | return err; | |
737 | } | |
738 | ||
fd80eb94 DL |
739 | static int tcp_v4_send_synack(struct sock *sk, struct request_sock *req) |
740 | { | |
741 | return __tcp_v4_send_synack(sk, req, NULL); | |
742 | } | |
743 | ||
1da177e4 | 744 | /* |
60236fdd | 745 | * IPv4 request_sock destructor. |
1da177e4 | 746 | */ |
60236fdd | 747 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 748 | { |
a51482bd | 749 | kfree(inet_rsk(req)->opt); |
1da177e4 LT |
750 | } |
751 | ||
80e40daa | 752 | #ifdef CONFIG_SYN_COOKIES |
40efc6fa | 753 | static void syn_flood_warning(struct sk_buff *skb) |
1da177e4 LT |
754 | { |
755 | static unsigned long warntime; | |
756 | ||
757 | if (time_after(jiffies, (warntime + HZ * 60))) { | |
758 | warntime = jiffies; | |
759 | printk(KERN_INFO | |
760 | "possible SYN flooding on port %d. Sending cookies.\n", | |
aa8223c7 | 761 | ntohs(tcp_hdr(skb)->dest)); |
1da177e4 LT |
762 | } |
763 | } | |
80e40daa | 764 | #endif |
1da177e4 LT |
765 | |
766 | /* | |
60236fdd | 767 | * Save and compile IPv4 options into the request_sock if needed. |
1da177e4 | 768 | */ |
40efc6fa SH |
769 | static struct ip_options *tcp_v4_save_options(struct sock *sk, |
770 | struct sk_buff *skb) | |
1da177e4 LT |
771 | { |
772 | struct ip_options *opt = &(IPCB(skb)->opt); | |
773 | struct ip_options *dopt = NULL; | |
774 | ||
775 | if (opt && opt->optlen) { | |
776 | int opt_size = optlength(opt); | |
777 | dopt = kmalloc(opt_size, GFP_ATOMIC); | |
778 | if (dopt) { | |
779 | if (ip_options_echo(dopt, skb)) { | |
780 | kfree(dopt); | |
781 | dopt = NULL; | |
782 | } | |
783 | } | |
784 | } | |
785 | return dopt; | |
786 | } | |
787 | ||
cfb6eeb4 YH |
788 | #ifdef CONFIG_TCP_MD5SIG |
789 | /* | |
790 | * RFC2385 MD5 checksumming requires a mapping of | |
791 | * IP address->MD5 Key. | |
792 | * We need to maintain these in the sk structure. | |
793 | */ | |
794 | ||
795 | /* Find the Key structure for an address. */ | |
7174259e ACM |
796 | static struct tcp_md5sig_key * |
797 | tcp_v4_md5_do_lookup(struct sock *sk, __be32 addr) | |
cfb6eeb4 YH |
798 | { |
799 | struct tcp_sock *tp = tcp_sk(sk); | |
800 | int i; | |
801 | ||
802 | if (!tp->md5sig_info || !tp->md5sig_info->entries4) | |
803 | return NULL; | |
804 | for (i = 0; i < tp->md5sig_info->entries4; i++) { | |
805 | if (tp->md5sig_info->keys4[i].addr == addr) | |
f8ab18d2 | 806 | return &tp->md5sig_info->keys4[i].base; |
cfb6eeb4 YH |
807 | } |
808 | return NULL; | |
809 | } | |
810 | ||
811 | struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
812 | struct sock *addr_sk) | |
813 | { | |
814 | return tcp_v4_md5_do_lookup(sk, inet_sk(addr_sk)->daddr); | |
815 | } | |
816 | ||
817 | EXPORT_SYMBOL(tcp_v4_md5_lookup); | |
818 | ||
f5b99bcd AB |
819 | static struct tcp_md5sig_key *tcp_v4_reqsk_md5_lookup(struct sock *sk, |
820 | struct request_sock *req) | |
cfb6eeb4 YH |
821 | { |
822 | return tcp_v4_md5_do_lookup(sk, inet_rsk(req)->rmt_addr); | |
823 | } | |
824 | ||
825 | /* This can be called on a newly created socket, from other files */ | |
826 | int tcp_v4_md5_do_add(struct sock *sk, __be32 addr, | |
827 | u8 *newkey, u8 newkeylen) | |
828 | { | |
829 | /* Add Key to the list */ | |
b0a713e9 | 830 | struct tcp_md5sig_key *key; |
cfb6eeb4 YH |
831 | struct tcp_sock *tp = tcp_sk(sk); |
832 | struct tcp4_md5sig_key *keys; | |
833 | ||
b0a713e9 | 834 | key = tcp_v4_md5_do_lookup(sk, addr); |
cfb6eeb4 YH |
835 | if (key) { |
836 | /* Pre-existing entry - just update that one. */ | |
b0a713e9 MD |
837 | kfree(key->key); |
838 | key->key = newkey; | |
839 | key->keylen = newkeylen; | |
cfb6eeb4 | 840 | } else { |
f6685938 ACM |
841 | struct tcp_md5sig_info *md5sig; |
842 | ||
cfb6eeb4 | 843 | if (!tp->md5sig_info) { |
f6685938 ACM |
844 | tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), |
845 | GFP_ATOMIC); | |
cfb6eeb4 YH |
846 | if (!tp->md5sig_info) { |
847 | kfree(newkey); | |
848 | return -ENOMEM; | |
849 | } | |
3d7dbeac | 850 | sk->sk_route_caps &= ~NETIF_F_GSO_MASK; |
cfb6eeb4 YH |
851 | } |
852 | if (tcp_alloc_md5sig_pool() == NULL) { | |
853 | kfree(newkey); | |
854 | return -ENOMEM; | |
855 | } | |
f6685938 ACM |
856 | md5sig = tp->md5sig_info; |
857 | ||
858 | if (md5sig->alloced4 == md5sig->entries4) { | |
859 | keys = kmalloc((sizeof(*keys) * | |
e905a9ed | 860 | (md5sig->entries4 + 1)), GFP_ATOMIC); |
cfb6eeb4 YH |
861 | if (!keys) { |
862 | kfree(newkey); | |
863 | tcp_free_md5sig_pool(); | |
864 | return -ENOMEM; | |
865 | } | |
866 | ||
f6685938 ACM |
867 | if (md5sig->entries4) |
868 | memcpy(keys, md5sig->keys4, | |
869 | sizeof(*keys) * md5sig->entries4); | |
cfb6eeb4 YH |
870 | |
871 | /* Free old key list, and reference new one */ | |
a80cc20d | 872 | kfree(md5sig->keys4); |
f6685938 ACM |
873 | md5sig->keys4 = keys; |
874 | md5sig->alloced4++; | |
cfb6eeb4 | 875 | } |
f6685938 | 876 | md5sig->entries4++; |
f8ab18d2 DM |
877 | md5sig->keys4[md5sig->entries4 - 1].addr = addr; |
878 | md5sig->keys4[md5sig->entries4 - 1].base.key = newkey; | |
879 | md5sig->keys4[md5sig->entries4 - 1].base.keylen = newkeylen; | |
cfb6eeb4 YH |
880 | } |
881 | return 0; | |
882 | } | |
883 | ||
884 | EXPORT_SYMBOL(tcp_v4_md5_do_add); | |
885 | ||
886 | static int tcp_v4_md5_add_func(struct sock *sk, struct sock *addr_sk, | |
887 | u8 *newkey, u8 newkeylen) | |
888 | { | |
889 | return tcp_v4_md5_do_add(sk, inet_sk(addr_sk)->daddr, | |
890 | newkey, newkeylen); | |
891 | } | |
892 | ||
893 | int tcp_v4_md5_do_del(struct sock *sk, __be32 addr) | |
894 | { | |
895 | struct tcp_sock *tp = tcp_sk(sk); | |
896 | int i; | |
897 | ||
898 | for (i = 0; i < tp->md5sig_info->entries4; i++) { | |
899 | if (tp->md5sig_info->keys4[i].addr == addr) { | |
900 | /* Free the key */ | |
f8ab18d2 | 901 | kfree(tp->md5sig_info->keys4[i].base.key); |
cfb6eeb4 YH |
902 | tp->md5sig_info->entries4--; |
903 | ||
904 | if (tp->md5sig_info->entries4 == 0) { | |
905 | kfree(tp->md5sig_info->keys4); | |
906 | tp->md5sig_info->keys4 = NULL; | |
8228a18d | 907 | tp->md5sig_info->alloced4 = 0; |
7174259e | 908 | } else if (tp->md5sig_info->entries4 != i) { |
cfb6eeb4 | 909 | /* Need to do some manipulation */ |
354faf09 YH |
910 | memmove(&tp->md5sig_info->keys4[i], |
911 | &tp->md5sig_info->keys4[i+1], | |
912 | (tp->md5sig_info->entries4 - i) * | |
913 | sizeof(struct tcp4_md5sig_key)); | |
cfb6eeb4 YH |
914 | } |
915 | tcp_free_md5sig_pool(); | |
916 | return 0; | |
917 | } | |
918 | } | |
919 | return -ENOENT; | |
920 | } | |
921 | ||
922 | EXPORT_SYMBOL(tcp_v4_md5_do_del); | |
923 | ||
7174259e | 924 | static void tcp_v4_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
925 | { |
926 | struct tcp_sock *tp = tcp_sk(sk); | |
927 | ||
928 | /* Free each key, then the set of key keys, | |
929 | * the crypto element, and then decrement our | |
930 | * hold on the last resort crypto. | |
931 | */ | |
932 | if (tp->md5sig_info->entries4) { | |
933 | int i; | |
934 | for (i = 0; i < tp->md5sig_info->entries4; i++) | |
f8ab18d2 | 935 | kfree(tp->md5sig_info->keys4[i].base.key); |
cfb6eeb4 YH |
936 | tp->md5sig_info->entries4 = 0; |
937 | tcp_free_md5sig_pool(); | |
938 | } | |
939 | if (tp->md5sig_info->keys4) { | |
940 | kfree(tp->md5sig_info->keys4); | |
941 | tp->md5sig_info->keys4 = NULL; | |
942 | tp->md5sig_info->alloced4 = 0; | |
943 | } | |
944 | } | |
945 | ||
7174259e ACM |
946 | static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval, |
947 | int optlen) | |
cfb6eeb4 YH |
948 | { |
949 | struct tcp_md5sig cmd; | |
950 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
951 | u8 *newkey; | |
952 | ||
953 | if (optlen < sizeof(cmd)) | |
954 | return -EINVAL; | |
955 | ||
7174259e | 956 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
957 | return -EFAULT; |
958 | ||
959 | if (sin->sin_family != AF_INET) | |
960 | return -EINVAL; | |
961 | ||
962 | if (!cmd.tcpm_key || !cmd.tcpm_keylen) { | |
963 | if (!tcp_sk(sk)->md5sig_info) | |
964 | return -ENOENT; | |
965 | return tcp_v4_md5_do_del(sk, sin->sin_addr.s_addr); | |
966 | } | |
967 | ||
968 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
969 | return -EINVAL; | |
970 | ||
971 | if (!tcp_sk(sk)->md5sig_info) { | |
972 | struct tcp_sock *tp = tcp_sk(sk); | |
7174259e | 973 | struct tcp_md5sig_info *p = kzalloc(sizeof(*p), GFP_KERNEL); |
cfb6eeb4 | 974 | |
cfb6eeb4 YH |
975 | if (!p) |
976 | return -EINVAL; | |
977 | ||
978 | tp->md5sig_info = p; | |
3d7dbeac | 979 | sk->sk_route_caps &= ~NETIF_F_GSO_MASK; |
cfb6eeb4 YH |
980 | } |
981 | ||
f6685938 | 982 | newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); |
cfb6eeb4 YH |
983 | if (!newkey) |
984 | return -ENOMEM; | |
cfb6eeb4 YH |
985 | return tcp_v4_md5_do_add(sk, sin->sin_addr.s_addr, |
986 | newkey, cmd.tcpm_keylen); | |
987 | } | |
988 | ||
49a72dfb AL |
989 | static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
990 | __be32 daddr, __be32 saddr, int nbytes) | |
cfb6eeb4 | 991 | { |
cfb6eeb4 | 992 | struct tcp4_pseudohdr *bp; |
49a72dfb | 993 | struct scatterlist sg; |
cfb6eeb4 YH |
994 | |
995 | bp = &hp->md5_blk.ip4; | |
cfb6eeb4 YH |
996 | |
997 | /* | |
49a72dfb | 998 | * 1. the TCP pseudo-header (in the order: source IP address, |
cfb6eeb4 YH |
999 | * destination IP address, zero-padded protocol number, and |
1000 | * segment length) | |
1001 | */ | |
1002 | bp->saddr = saddr; | |
1003 | bp->daddr = daddr; | |
1004 | bp->pad = 0; | |
076fb722 | 1005 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1006 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1007 | |
49a72dfb AL |
1008 | sg_init_one(&sg, bp, sizeof(*bp)); |
1009 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
1010 | } | |
1011 | ||
1012 | static int tcp_v4_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, | |
1013 | __be32 daddr, __be32 saddr, struct tcphdr *th) | |
1014 | { | |
1015 | struct tcp_md5sig_pool *hp; | |
1016 | struct hash_desc *desc; | |
1017 | ||
1018 | hp = tcp_get_md5sig_pool(); | |
1019 | if (!hp) | |
1020 | goto clear_hash_noput; | |
1021 | desc = &hp->md5_desc; | |
1022 | ||
1023 | if (crypto_hash_init(desc)) | |
1024 | goto clear_hash; | |
1025 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
1026 | goto clear_hash; | |
1027 | if (tcp_md5_hash_header(hp, th)) | |
1028 | goto clear_hash; | |
1029 | if (tcp_md5_hash_key(hp, key)) | |
1030 | goto clear_hash; | |
1031 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 YH |
1032 | goto clear_hash; |
1033 | ||
cfb6eeb4 | 1034 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1035 | return 0; |
49a72dfb | 1036 | |
cfb6eeb4 YH |
1037 | clear_hash: |
1038 | tcp_put_md5sig_pool(); | |
1039 | clear_hash_noput: | |
1040 | memset(md5_hash, 0, 16); | |
49a72dfb | 1041 | return 1; |
cfb6eeb4 YH |
1042 | } |
1043 | ||
49a72dfb AL |
1044 | int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
1045 | struct sock *sk, struct request_sock *req, | |
1046 | struct sk_buff *skb) | |
cfb6eeb4 | 1047 | { |
49a72dfb AL |
1048 | struct tcp_md5sig_pool *hp; |
1049 | struct hash_desc *desc; | |
1050 | struct tcphdr *th = tcp_hdr(skb); | |
cfb6eeb4 YH |
1051 | __be32 saddr, daddr; |
1052 | ||
1053 | if (sk) { | |
1054 | saddr = inet_sk(sk)->saddr; | |
1055 | daddr = inet_sk(sk)->daddr; | |
49a72dfb AL |
1056 | } else if (req) { |
1057 | saddr = inet_rsk(req)->loc_addr; | |
1058 | daddr = inet_rsk(req)->rmt_addr; | |
cfb6eeb4 | 1059 | } else { |
49a72dfb AL |
1060 | const struct iphdr *iph = ip_hdr(skb); |
1061 | saddr = iph->saddr; | |
1062 | daddr = iph->daddr; | |
cfb6eeb4 | 1063 | } |
49a72dfb AL |
1064 | |
1065 | hp = tcp_get_md5sig_pool(); | |
1066 | if (!hp) | |
1067 | goto clear_hash_noput; | |
1068 | desc = &hp->md5_desc; | |
1069 | ||
1070 | if (crypto_hash_init(desc)) | |
1071 | goto clear_hash; | |
1072 | ||
1073 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
1074 | goto clear_hash; | |
1075 | if (tcp_md5_hash_header(hp, th)) | |
1076 | goto clear_hash; | |
1077 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1078 | goto clear_hash; | |
1079 | if (tcp_md5_hash_key(hp, key)) | |
1080 | goto clear_hash; | |
1081 | if (crypto_hash_final(desc, md5_hash)) | |
1082 | goto clear_hash; | |
1083 | ||
1084 | tcp_put_md5sig_pool(); | |
1085 | return 0; | |
1086 | ||
1087 | clear_hash: | |
1088 | tcp_put_md5sig_pool(); | |
1089 | clear_hash_noput: | |
1090 | memset(md5_hash, 0, 16); | |
1091 | return 1; | |
cfb6eeb4 YH |
1092 | } |
1093 | ||
49a72dfb | 1094 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1095 | |
7174259e | 1096 | static int tcp_v4_inbound_md5_hash(struct sock *sk, struct sk_buff *skb) |
cfb6eeb4 YH |
1097 | { |
1098 | /* | |
1099 | * This gets called for each TCP segment that arrives | |
1100 | * so we want to be efficient. | |
1101 | * We have 3 drop cases: | |
1102 | * o No MD5 hash and one expected. | |
1103 | * o MD5 hash and we're not expecting one. | |
1104 | * o MD5 hash and its wrong. | |
1105 | */ | |
1106 | __u8 *hash_location = NULL; | |
1107 | struct tcp_md5sig_key *hash_expected; | |
eddc9ec5 | 1108 | const struct iphdr *iph = ip_hdr(skb); |
aa8223c7 | 1109 | struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1110 | int genhash; |
cfb6eeb4 YH |
1111 | unsigned char newhash[16]; |
1112 | ||
1113 | hash_expected = tcp_v4_md5_do_lookup(sk, iph->saddr); | |
7d5d5525 | 1114 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1115 | |
cfb6eeb4 YH |
1116 | /* We've parsed the options - do we have a hash? */ |
1117 | if (!hash_expected && !hash_location) | |
1118 | return 0; | |
1119 | ||
1120 | if (hash_expected && !hash_location) { | |
785957d3 | 1121 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
cfb6eeb4 YH |
1122 | return 1; |
1123 | } | |
1124 | ||
1125 | if (!hash_expected && hash_location) { | |
785957d3 | 1126 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
cfb6eeb4 YH |
1127 | return 1; |
1128 | } | |
1129 | ||
1130 | /* Okay, so this is hash_expected and hash_location - | |
1131 | * so we need to calculate the checksum. | |
1132 | */ | |
49a72dfb AL |
1133 | genhash = tcp_v4_md5_hash_skb(newhash, |
1134 | hash_expected, | |
1135 | NULL, NULL, skb); | |
cfb6eeb4 YH |
1136 | |
1137 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
1138 | if (net_ratelimit()) { | |
673d57e7 HH |
1139 | printk(KERN_INFO "MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1140 | &iph->saddr, ntohs(th->source), | |
1141 | &iph->daddr, ntohs(th->dest), | |
cfb6eeb4 | 1142 | genhash ? " tcp_v4_calc_md5_hash failed" : ""); |
cfb6eeb4 YH |
1143 | } |
1144 | return 1; | |
1145 | } | |
1146 | return 0; | |
1147 | } | |
1148 | ||
1149 | #endif | |
1150 | ||
72a3effa | 1151 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1152 | .family = PF_INET, |
2e6599cb | 1153 | .obj_size = sizeof(struct tcp_request_sock), |
1da177e4 | 1154 | .rtx_syn_ack = tcp_v4_send_synack, |
60236fdd ACM |
1155 | .send_ack = tcp_v4_reqsk_send_ack, |
1156 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 LT |
1157 | .send_reset = tcp_v4_send_reset, |
1158 | }; | |
1159 | ||
cfb6eeb4 | 1160 | #ifdef CONFIG_TCP_MD5SIG |
b6332e6c | 1161 | static struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
cfb6eeb4 | 1162 | .md5_lookup = tcp_v4_reqsk_md5_lookup, |
cfb6eeb4 | 1163 | }; |
b6332e6c | 1164 | #endif |
cfb6eeb4 | 1165 | |
6d6ee43e ACM |
1166 | static struct timewait_sock_ops tcp_timewait_sock_ops = { |
1167 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1168 | .twsk_unique = tcp_twsk_unique, | |
cfb6eeb4 | 1169 | .twsk_destructor= tcp_twsk_destructor, |
6d6ee43e ACM |
1170 | }; |
1171 | ||
1da177e4 LT |
1172 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1173 | { | |
2e6599cb | 1174 | struct inet_request_sock *ireq; |
1da177e4 | 1175 | struct tcp_options_received tmp_opt; |
60236fdd | 1176 | struct request_sock *req; |
eddc9ec5 ACM |
1177 | __be32 saddr = ip_hdr(skb)->saddr; |
1178 | __be32 daddr = ip_hdr(skb)->daddr; | |
1da177e4 LT |
1179 | __u32 isn = TCP_SKB_CB(skb)->when; |
1180 | struct dst_entry *dst = NULL; | |
1181 | #ifdef CONFIG_SYN_COOKIES | |
1182 | int want_cookie = 0; | |
1183 | #else | |
1184 | #define want_cookie 0 /* Argh, why doesn't gcc optimize this :( */ | |
1185 | #endif | |
1186 | ||
1187 | /* Never answer to SYNs send to broadcast or multicast */ | |
ee6b9673 | 1188 | if (skb->rtable->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1189 | goto drop; |
1190 | ||
1191 | /* TW buckets are converted to open requests without | |
1192 | * limitations, they conserve resources and peer is | |
1193 | * evidently real one. | |
1194 | */ | |
463c84b9 | 1195 | if (inet_csk_reqsk_queue_is_full(sk) && !isn) { |
1da177e4 LT |
1196 | #ifdef CONFIG_SYN_COOKIES |
1197 | if (sysctl_tcp_syncookies) { | |
1198 | want_cookie = 1; | |
1199 | } else | |
1200 | #endif | |
1201 | goto drop; | |
1202 | } | |
1203 | ||
1204 | /* Accept backlog is full. If we have already queued enough | |
1205 | * of warm entries in syn queue, drop request. It is better than | |
1206 | * clogging syn queue with openreqs with exponentially increasing | |
1207 | * timeout. | |
1208 | */ | |
463c84b9 | 1209 | if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) |
1da177e4 LT |
1210 | goto drop; |
1211 | ||
ce4a7d0d | 1212 | req = inet_reqsk_alloc(&tcp_request_sock_ops); |
1da177e4 LT |
1213 | if (!req) |
1214 | goto drop; | |
1215 | ||
cfb6eeb4 YH |
1216 | #ifdef CONFIG_TCP_MD5SIG |
1217 | tcp_rsk(req)->af_specific = &tcp_request_sock_ipv4_ops; | |
1218 | #endif | |
1219 | ||
1da177e4 LT |
1220 | tcp_clear_options(&tmp_opt); |
1221 | tmp_opt.mss_clamp = 536; | |
1222 | tmp_opt.user_mss = tcp_sk(sk)->rx_opt.user_mss; | |
1223 | ||
1224 | tcp_parse_options(skb, &tmp_opt, 0); | |
1225 | ||
4dfc2817 | 1226 | if (want_cookie && !tmp_opt.saw_tstamp) |
1da177e4 | 1227 | tcp_clear_options(&tmp_opt); |
1da177e4 LT |
1228 | |
1229 | if (tmp_opt.saw_tstamp && !tmp_opt.rcv_tsval) { | |
1230 | /* Some OSes (unknown ones, but I see them on web server, which | |
1231 | * contains information interesting only for windows' | |
1232 | * users) do not send their stamp in SYN. It is easy case. | |
1233 | * We simply do not advertise TS support. | |
1234 | */ | |
1235 | tmp_opt.saw_tstamp = 0; | |
1236 | tmp_opt.tstamp_ok = 0; | |
1237 | } | |
1238 | tmp_opt.tstamp_ok = tmp_opt.saw_tstamp; | |
1239 | ||
1240 | tcp_openreq_init(req, &tmp_opt, skb); | |
1241 | ||
4237c75c VY |
1242 | if (security_inet_conn_request(sk, skb, req)) |
1243 | goto drop_and_free; | |
1244 | ||
2e6599cb ACM |
1245 | ireq = inet_rsk(req); |
1246 | ireq->loc_addr = daddr; | |
1247 | ireq->rmt_addr = saddr; | |
88ef4a5a | 1248 | ireq->no_srccheck = inet_sk(sk)->transparent; |
2e6599cb | 1249 | ireq->opt = tcp_v4_save_options(sk, skb); |
1da177e4 | 1250 | if (!want_cookie) |
aa8223c7 | 1251 | TCP_ECN_create_request(req, tcp_hdr(skb)); |
1da177e4 LT |
1252 | |
1253 | if (want_cookie) { | |
1254 | #ifdef CONFIG_SYN_COOKIES | |
1255 | syn_flood_warning(skb); | |
4dfc2817 | 1256 | req->cookie_ts = tmp_opt.tstamp_ok; |
1da177e4 LT |
1257 | #endif |
1258 | isn = cookie_v4_init_sequence(sk, skb, &req->mss); | |
1259 | } else if (!isn) { | |
1260 | struct inet_peer *peer = NULL; | |
1261 | ||
1262 | /* VJ's idea. We save last timestamp seen | |
1263 | * from the destination in peer table, when entering | |
1264 | * state TIME-WAIT, and check against it before | |
1265 | * accepting new connection request. | |
1266 | * | |
1267 | * If "isn" is not zero, this request hit alive | |
1268 | * timewait bucket, so that all the necessary checks | |
1269 | * are made in the function processing timewait state. | |
1270 | */ | |
1271 | if (tmp_opt.saw_tstamp && | |
295ff7ed | 1272 | tcp_death_row.sysctl_tw_recycle && |
463c84b9 | 1273 | (dst = inet_csk_route_req(sk, req)) != NULL && |
1da177e4 LT |
1274 | (peer = rt_get_peer((struct rtable *)dst)) != NULL && |
1275 | peer->v4daddr == saddr) { | |
9d729f72 | 1276 | if (get_seconds() < peer->tcp_ts_stamp + TCP_PAWS_MSL && |
1da177e4 LT |
1277 | (s32)(peer->tcp_ts - req->ts_recent) > |
1278 | TCP_PAWS_WINDOW) { | |
de0744af | 1279 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED); |
7cd04fa7 | 1280 | goto drop_and_release; |
1da177e4 LT |
1281 | } |
1282 | } | |
1283 | /* Kill the following clause, if you dislike this way. */ | |
1284 | else if (!sysctl_tcp_syncookies && | |
463c84b9 | 1285 | (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < |
1da177e4 LT |
1286 | (sysctl_max_syn_backlog >> 2)) && |
1287 | (!peer || !peer->tcp_ts_stamp) && | |
1288 | (!dst || !dst_metric(dst, RTAX_RTT))) { | |
1289 | /* Without syncookies last quarter of | |
1290 | * backlog is filled with destinations, | |
1291 | * proven to be alive. | |
1292 | * It means that we continue to communicate | |
1293 | * to destinations, already remembered | |
1294 | * to the moment of synflood. | |
1295 | */ | |
673d57e7 HH |
1296 | LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI4/%u\n", |
1297 | &saddr, ntohs(tcp_hdr(skb)->source)); | |
7cd04fa7 | 1298 | goto drop_and_release; |
1da177e4 LT |
1299 | } |
1300 | ||
a94f723d | 1301 | isn = tcp_v4_init_sequence(skb); |
1da177e4 | 1302 | } |
2e6599cb | 1303 | tcp_rsk(req)->snt_isn = isn; |
1da177e4 | 1304 | |
7cd04fa7 | 1305 | if (__tcp_v4_send_synack(sk, req, dst) || want_cookie) |
1da177e4 LT |
1306 | goto drop_and_free; |
1307 | ||
7cd04fa7 | 1308 | inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT); |
1da177e4 LT |
1309 | return 0; |
1310 | ||
7cd04fa7 DL |
1311 | drop_and_release: |
1312 | dst_release(dst); | |
1da177e4 | 1313 | drop_and_free: |
60236fdd | 1314 | reqsk_free(req); |
1da177e4 | 1315 | drop: |
1da177e4 LT |
1316 | return 0; |
1317 | } | |
1318 | ||
1319 | ||
1320 | /* | |
1321 | * The three way handshake has completed - we got a valid synack - | |
1322 | * now create the new socket. | |
1323 | */ | |
1324 | struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1325 | struct request_sock *req, |
1da177e4 LT |
1326 | struct dst_entry *dst) |
1327 | { | |
2e6599cb | 1328 | struct inet_request_sock *ireq; |
1da177e4 LT |
1329 | struct inet_sock *newinet; |
1330 | struct tcp_sock *newtp; | |
1331 | struct sock *newsk; | |
cfb6eeb4 YH |
1332 | #ifdef CONFIG_TCP_MD5SIG |
1333 | struct tcp_md5sig_key *key; | |
1334 | #endif | |
1da177e4 LT |
1335 | |
1336 | if (sk_acceptq_is_full(sk)) | |
1337 | goto exit_overflow; | |
1338 | ||
463c84b9 | 1339 | if (!dst && (dst = inet_csk_route_req(sk, req)) == NULL) |
1da177e4 LT |
1340 | goto exit; |
1341 | ||
1342 | newsk = tcp_create_openreq_child(sk, req, skb); | |
1343 | if (!newsk) | |
1344 | goto exit; | |
1345 | ||
bcd76111 | 1346 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
6cbb0df7 | 1347 | sk_setup_caps(newsk, dst); |
1da177e4 LT |
1348 | |
1349 | newtp = tcp_sk(newsk); | |
1350 | newinet = inet_sk(newsk); | |
2e6599cb ACM |
1351 | ireq = inet_rsk(req); |
1352 | newinet->daddr = ireq->rmt_addr; | |
1353 | newinet->rcv_saddr = ireq->loc_addr; | |
1354 | newinet->saddr = ireq->loc_addr; | |
1355 | newinet->opt = ireq->opt; | |
1356 | ireq->opt = NULL; | |
463c84b9 | 1357 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1358 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
d83d8461 | 1359 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
1da177e4 | 1360 | if (newinet->opt) |
d83d8461 | 1361 | inet_csk(newsk)->icsk_ext_hdr_len = newinet->opt->optlen; |
1da177e4 LT |
1362 | newinet->id = newtp->write_seq ^ jiffies; |
1363 | ||
5d424d5a | 1364 | tcp_mtup_init(newsk); |
1da177e4 LT |
1365 | tcp_sync_mss(newsk, dst_mtu(dst)); |
1366 | newtp->advmss = dst_metric(dst, RTAX_ADVMSS); | |
f5fff5dc TQ |
1367 | if (tcp_sk(sk)->rx_opt.user_mss && |
1368 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1369 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1370 | ||
1da177e4 LT |
1371 | tcp_initialize_rcv_mss(newsk); |
1372 | ||
cfb6eeb4 YH |
1373 | #ifdef CONFIG_TCP_MD5SIG |
1374 | /* Copy over the MD5 key from the original socket */ | |
1375 | if ((key = tcp_v4_md5_do_lookup(sk, newinet->daddr)) != NULL) { | |
1376 | /* | |
1377 | * We're using one, so create a matching key | |
1378 | * on the newsk structure. If we fail to get | |
1379 | * memory, then we end up not copying the key | |
1380 | * across. Shucks. | |
1381 | */ | |
f6685938 ACM |
1382 | char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC); |
1383 | if (newkey != NULL) | |
cfb6eeb4 YH |
1384 | tcp_v4_md5_do_add(newsk, inet_sk(sk)->daddr, |
1385 | newkey, key->keylen); | |
49a72dfb | 1386 | newsk->sk_route_caps &= ~NETIF_F_GSO_MASK; |
cfb6eeb4 YH |
1387 | } |
1388 | #endif | |
1389 | ||
ab1e0a13 ACM |
1390 | __inet_hash_nolisten(newsk); |
1391 | __inet_inherit_port(sk, newsk); | |
1da177e4 LT |
1392 | |
1393 | return newsk; | |
1394 | ||
1395 | exit_overflow: | |
de0744af | 1396 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
1da177e4 | 1397 | exit: |
de0744af | 1398 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
1399 | dst_release(dst); |
1400 | return NULL; | |
1401 | } | |
1402 | ||
1403 | static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb) | |
1404 | { | |
aa8223c7 | 1405 | struct tcphdr *th = tcp_hdr(skb); |
eddc9ec5 | 1406 | const struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 1407 | struct sock *nsk; |
60236fdd | 1408 | struct request_sock **prev; |
1da177e4 | 1409 | /* Find possible connection requests. */ |
463c84b9 ACM |
1410 | struct request_sock *req = inet_csk_search_req(sk, &prev, th->source, |
1411 | iph->saddr, iph->daddr); | |
1da177e4 LT |
1412 | if (req) |
1413 | return tcp_check_req(sk, skb, req, prev); | |
1414 | ||
3b1e0a65 | 1415 | nsk = inet_lookup_established(sock_net(sk), &tcp_hashinfo, iph->saddr, |
c67499c0 | 1416 | th->source, iph->daddr, th->dest, inet_iif(skb)); |
1da177e4 LT |
1417 | |
1418 | if (nsk) { | |
1419 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
1420 | bh_lock_sock(nsk); | |
1421 | return nsk; | |
1422 | } | |
9469c7b4 | 1423 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
1424 | return NULL; |
1425 | } | |
1426 | ||
1427 | #ifdef CONFIG_SYN_COOKIES | |
1428 | if (!th->rst && !th->syn && th->ack) | |
1429 | sk = cookie_v4_check(sk, skb, &(IPCB(skb)->opt)); | |
1430 | #endif | |
1431 | return sk; | |
1432 | } | |
1433 | ||
b51655b9 | 1434 | static __sum16 tcp_v4_checksum_init(struct sk_buff *skb) |
1da177e4 | 1435 | { |
eddc9ec5 ACM |
1436 | const struct iphdr *iph = ip_hdr(skb); |
1437 | ||
84fa7933 | 1438 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
eddc9ec5 ACM |
1439 | if (!tcp_v4_check(skb->len, iph->saddr, |
1440 | iph->daddr, skb->csum)) { | |
fb286bb2 | 1441 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
1da177e4 | 1442 | return 0; |
fb286bb2 | 1443 | } |
1da177e4 | 1444 | } |
fb286bb2 | 1445 | |
eddc9ec5 | 1446 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, |
fb286bb2 HX |
1447 | skb->len, IPPROTO_TCP, 0); |
1448 | ||
1da177e4 | 1449 | if (skb->len <= 76) { |
fb286bb2 | 1450 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1451 | } |
1452 | return 0; | |
1453 | } | |
1454 | ||
1455 | ||
1456 | /* The socket must have it's spinlock held when we get | |
1457 | * here. | |
1458 | * | |
1459 | * We have a potential double-lock case here, so even when | |
1460 | * doing backlog processing we use the BH locking scheme. | |
1461 | * This is because we cannot sleep with the original spinlock | |
1462 | * held. | |
1463 | */ | |
1464 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1465 | { | |
cfb6eeb4 YH |
1466 | struct sock *rsk; |
1467 | #ifdef CONFIG_TCP_MD5SIG | |
1468 | /* | |
1469 | * We really want to reject the packet as early as possible | |
1470 | * if: | |
1471 | * o We're expecting an MD5'd packet and this is no MD5 tcp option | |
1472 | * o There is an MD5 option and we're not expecting one | |
1473 | */ | |
7174259e | 1474 | if (tcp_v4_inbound_md5_hash(sk, skb)) |
cfb6eeb4 YH |
1475 | goto discard; |
1476 | #endif | |
1477 | ||
1da177e4 LT |
1478 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
1479 | TCP_CHECK_TIMER(sk); | |
aa8223c7 | 1480 | if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) { |
cfb6eeb4 | 1481 | rsk = sk; |
1da177e4 | 1482 | goto reset; |
cfb6eeb4 | 1483 | } |
1da177e4 LT |
1484 | TCP_CHECK_TIMER(sk); |
1485 | return 0; | |
1486 | } | |
1487 | ||
ab6a5bb6 | 1488 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1489 | goto csum_err; |
1490 | ||
1491 | if (sk->sk_state == TCP_LISTEN) { | |
1492 | struct sock *nsk = tcp_v4_hnd_req(sk, skb); | |
1493 | if (!nsk) | |
1494 | goto discard; | |
1495 | ||
1496 | if (nsk != sk) { | |
cfb6eeb4 YH |
1497 | if (tcp_child_process(sk, nsk, skb)) { |
1498 | rsk = nsk; | |
1da177e4 | 1499 | goto reset; |
cfb6eeb4 | 1500 | } |
1da177e4 LT |
1501 | return 0; |
1502 | } | |
1503 | } | |
1504 | ||
1505 | TCP_CHECK_TIMER(sk); | |
aa8223c7 | 1506 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) { |
cfb6eeb4 | 1507 | rsk = sk; |
1da177e4 | 1508 | goto reset; |
cfb6eeb4 | 1509 | } |
1da177e4 LT |
1510 | TCP_CHECK_TIMER(sk); |
1511 | return 0; | |
1512 | ||
1513 | reset: | |
cfb6eeb4 | 1514 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1515 | discard: |
1516 | kfree_skb(skb); | |
1517 | /* Be careful here. If this function gets more complicated and | |
1518 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1519 | * might be destroyed here. This current version compiles correctly, | |
1520 | * but you have been warned. | |
1521 | */ | |
1522 | return 0; | |
1523 | ||
1524 | csum_err: | |
63231bdd | 1525 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1526 | goto discard; |
1527 | } | |
1528 | ||
1529 | /* | |
1530 | * From tcp_input.c | |
1531 | */ | |
1532 | ||
1533 | int tcp_v4_rcv(struct sk_buff *skb) | |
1534 | { | |
eddc9ec5 | 1535 | const struct iphdr *iph; |
1da177e4 LT |
1536 | struct tcphdr *th; |
1537 | struct sock *sk; | |
1538 | int ret; | |
a86b1e30 | 1539 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1540 | |
1541 | if (skb->pkt_type != PACKET_HOST) | |
1542 | goto discard_it; | |
1543 | ||
1544 | /* Count it even if it's bad */ | |
63231bdd | 1545 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1546 | |
1547 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1548 | goto discard_it; | |
1549 | ||
aa8223c7 | 1550 | th = tcp_hdr(skb); |
1da177e4 LT |
1551 | |
1552 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1553 | goto bad_packet; | |
1554 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1555 | goto discard_it; | |
1556 | ||
1557 | /* An explanation is required here, I think. | |
1558 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1559 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1560 | * So, we defer the checks. */ |
60476372 | 1561 | if (!skb_csum_unnecessary(skb) && tcp_v4_checksum_init(skb)) |
1da177e4 LT |
1562 | goto bad_packet; |
1563 | ||
aa8223c7 | 1564 | th = tcp_hdr(skb); |
eddc9ec5 | 1565 | iph = ip_hdr(skb); |
1da177e4 LT |
1566 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1567 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1568 | skb->len - th->doff * 4); | |
1569 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1570 | TCP_SKB_CB(skb)->when = 0; | |
eddc9ec5 | 1571 | TCP_SKB_CB(skb)->flags = iph->tos; |
1da177e4 LT |
1572 | TCP_SKB_CB(skb)->sacked = 0; |
1573 | ||
9a1f27c4 | 1574 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1575 | if (!sk) |
1576 | goto no_tcp_socket; | |
1577 | ||
1578 | process: | |
1579 | if (sk->sk_state == TCP_TIME_WAIT) | |
1580 | goto do_time_wait; | |
1581 | ||
1582 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1583 | goto discard_and_relse; | |
b59c2701 | 1584 | nf_reset(skb); |
1da177e4 | 1585 | |
fda9ef5d | 1586 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1587 | goto discard_and_relse; |
1588 | ||
1589 | skb->dev = NULL; | |
1590 | ||
c6366184 | 1591 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1592 | ret = 0; |
1593 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 CL |
1594 | #ifdef CONFIG_NET_DMA |
1595 | struct tcp_sock *tp = tcp_sk(sk); | |
1596 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) | |
f67b4599 | 1597 | tp->ucopy.dma_chan = dma_find_channel(DMA_MEMCPY); |
1a2449a8 | 1598 | if (tp->ucopy.dma_chan) |
1da177e4 | 1599 | ret = tcp_v4_do_rcv(sk, skb); |
1a2449a8 CL |
1600 | else |
1601 | #endif | |
1602 | { | |
1603 | if (!tcp_prequeue(sk, skb)) | |
1604 | ret = tcp_v4_do_rcv(sk, skb); | |
1605 | } | |
1da177e4 LT |
1606 | } else |
1607 | sk_add_backlog(sk, skb); | |
1608 | bh_unlock_sock(sk); | |
1609 | ||
1610 | sock_put(sk); | |
1611 | ||
1612 | return ret; | |
1613 | ||
1614 | no_tcp_socket: | |
1615 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1616 | goto discard_it; | |
1617 | ||
1618 | if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) { | |
1619 | bad_packet: | |
63231bdd | 1620 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1621 | } else { |
cfb6eeb4 | 1622 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1623 | } |
1624 | ||
1625 | discard_it: | |
1626 | /* Discard frame. */ | |
1627 | kfree_skb(skb); | |
e905a9ed | 1628 | return 0; |
1da177e4 LT |
1629 | |
1630 | discard_and_relse: | |
1631 | sock_put(sk); | |
1632 | goto discard_it; | |
1633 | ||
1634 | do_time_wait: | |
1635 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1636 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1637 | goto discard_it; |
1638 | } | |
1639 | ||
1640 | if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) { | |
63231bdd | 1641 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
9469c7b4 | 1642 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1643 | goto discard_it; |
1644 | } | |
9469c7b4 | 1645 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1646 | case TCP_TW_SYN: { |
c346dca1 | 1647 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
c67499c0 | 1648 | &tcp_hashinfo, |
eddc9ec5 | 1649 | iph->daddr, th->dest, |
463c84b9 | 1650 | inet_iif(skb)); |
1da177e4 | 1651 | if (sk2) { |
9469c7b4 YH |
1652 | inet_twsk_deschedule(inet_twsk(sk), &tcp_death_row); |
1653 | inet_twsk_put(inet_twsk(sk)); | |
1da177e4 LT |
1654 | sk = sk2; |
1655 | goto process; | |
1656 | } | |
1657 | /* Fall through to ACK */ | |
1658 | } | |
1659 | case TCP_TW_ACK: | |
1660 | tcp_v4_timewait_ack(sk, skb); | |
1661 | break; | |
1662 | case TCP_TW_RST: | |
1663 | goto no_tcp_socket; | |
1664 | case TCP_TW_SUCCESS:; | |
1665 | } | |
1666 | goto discard_it; | |
1667 | } | |
1668 | ||
1da177e4 LT |
1669 | /* VJ's idea. Save last timestamp seen from this destination |
1670 | * and hold it at least for normal timewait interval to use for duplicate | |
1671 | * segment detection in subsequent connections, before they enter synchronized | |
1672 | * state. | |
1673 | */ | |
1674 | ||
1675 | int tcp_v4_remember_stamp(struct sock *sk) | |
1676 | { | |
1677 | struct inet_sock *inet = inet_sk(sk); | |
1678 | struct tcp_sock *tp = tcp_sk(sk); | |
1679 | struct rtable *rt = (struct rtable *)__sk_dst_get(sk); | |
1680 | struct inet_peer *peer = NULL; | |
1681 | int release_it = 0; | |
1682 | ||
1683 | if (!rt || rt->rt_dst != inet->daddr) { | |
1684 | peer = inet_getpeer(inet->daddr, 1); | |
1685 | release_it = 1; | |
1686 | } else { | |
1687 | if (!rt->peer) | |
1688 | rt_bind_peer(rt, 1); | |
1689 | peer = rt->peer; | |
1690 | } | |
1691 | ||
1692 | if (peer) { | |
1693 | if ((s32)(peer->tcp_ts - tp->rx_opt.ts_recent) <= 0 || | |
9d729f72 | 1694 | (peer->tcp_ts_stamp + TCP_PAWS_MSL < get_seconds() && |
1da177e4 LT |
1695 | peer->tcp_ts_stamp <= tp->rx_opt.ts_recent_stamp)) { |
1696 | peer->tcp_ts_stamp = tp->rx_opt.ts_recent_stamp; | |
1697 | peer->tcp_ts = tp->rx_opt.ts_recent; | |
1698 | } | |
1699 | if (release_it) | |
1700 | inet_putpeer(peer); | |
1701 | return 1; | |
1702 | } | |
1703 | ||
1704 | return 0; | |
1705 | } | |
1706 | ||
8feaf0c0 | 1707 | int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw) |
1da177e4 | 1708 | { |
8feaf0c0 | 1709 | struct inet_peer *peer = inet_getpeer(tw->tw_daddr, 1); |
1da177e4 LT |
1710 | |
1711 | if (peer) { | |
8feaf0c0 ACM |
1712 | const struct tcp_timewait_sock *tcptw = tcp_twsk((struct sock *)tw); |
1713 | ||
1714 | if ((s32)(peer->tcp_ts - tcptw->tw_ts_recent) <= 0 || | |
9d729f72 | 1715 | (peer->tcp_ts_stamp + TCP_PAWS_MSL < get_seconds() && |
8feaf0c0 ACM |
1716 | peer->tcp_ts_stamp <= tcptw->tw_ts_recent_stamp)) { |
1717 | peer->tcp_ts_stamp = tcptw->tw_ts_recent_stamp; | |
1718 | peer->tcp_ts = tcptw->tw_ts_recent; | |
1da177e4 LT |
1719 | } |
1720 | inet_putpeer(peer); | |
1721 | return 1; | |
1722 | } | |
1723 | ||
1724 | return 0; | |
1725 | } | |
1726 | ||
8292a17a | 1727 | struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1728 | .queue_xmit = ip_queue_xmit, |
1729 | .send_check = tcp_v4_send_check, | |
1730 | .rebuild_header = inet_sk_rebuild_header, | |
1731 | .conn_request = tcp_v4_conn_request, | |
1732 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
1733 | .remember_stamp = tcp_v4_remember_stamp, | |
1734 | .net_header_len = sizeof(struct iphdr), | |
1735 | .setsockopt = ip_setsockopt, | |
1736 | .getsockopt = ip_getsockopt, | |
1737 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1738 | .sockaddr_len = sizeof(struct sockaddr_in), | |
ab1e0a13 | 1739 | .bind_conflict = inet_csk_bind_conflict, |
3fdadf7d | 1740 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1741 | .compat_setsockopt = compat_ip_setsockopt, |
1742 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1743 | #endif |
1da177e4 LT |
1744 | }; |
1745 | ||
cfb6eeb4 | 1746 | #ifdef CONFIG_TCP_MD5SIG |
b6332e6c | 1747 | static struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1748 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1749 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 YH |
1750 | .md5_add = tcp_v4_md5_add_func, |
1751 | .md5_parse = tcp_v4_parse_md5_keys, | |
cfb6eeb4 | 1752 | }; |
b6332e6c | 1753 | #endif |
cfb6eeb4 | 1754 | |
1da177e4 LT |
1755 | /* NOTE: A lot of things set to zero explicitly by call to |
1756 | * sk_alloc() so need not be done here. | |
1757 | */ | |
1758 | static int tcp_v4_init_sock(struct sock *sk) | |
1759 | { | |
6687e988 | 1760 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 LT |
1761 | struct tcp_sock *tp = tcp_sk(sk); |
1762 | ||
1763 | skb_queue_head_init(&tp->out_of_order_queue); | |
1764 | tcp_init_xmit_timers(sk); | |
1765 | tcp_prequeue_init(tp); | |
1766 | ||
6687e988 | 1767 | icsk->icsk_rto = TCP_TIMEOUT_INIT; |
1da177e4 LT |
1768 | tp->mdev = TCP_TIMEOUT_INIT; |
1769 | ||
1770 | /* So many TCP implementations out there (incorrectly) count the | |
1771 | * initial SYN frame in their delayed-ACK and congestion control | |
1772 | * algorithms that we must have the following bandaid to talk | |
1773 | * efficiently to them. -DaveM | |
1774 | */ | |
1775 | tp->snd_cwnd = 2; | |
1776 | ||
1777 | /* See draft-stevens-tcpca-spec-01 for discussion of the | |
1778 | * initialization of these values. | |
1779 | */ | |
1780 | tp->snd_ssthresh = 0x7fffffff; /* Infinity */ | |
1781 | tp->snd_cwnd_clamp = ~0; | |
c1b4a7e6 | 1782 | tp->mss_cache = 536; |
1da177e4 LT |
1783 | |
1784 | tp->reordering = sysctl_tcp_reordering; | |
6687e988 | 1785 | icsk->icsk_ca_ops = &tcp_init_congestion_ops; |
1da177e4 LT |
1786 | |
1787 | sk->sk_state = TCP_CLOSE; | |
1788 | ||
1789 | sk->sk_write_space = sk_stream_write_space; | |
1790 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
1791 | ||
8292a17a | 1792 | icsk->icsk_af_ops = &ipv4_specific; |
d83d8461 | 1793 | icsk->icsk_sync_mss = tcp_sync_mss; |
cfb6eeb4 YH |
1794 | #ifdef CONFIG_TCP_MD5SIG |
1795 | tp->af_specific = &tcp_sock_ipv4_specific; | |
1796 | #endif | |
1da177e4 LT |
1797 | |
1798 | sk->sk_sndbuf = sysctl_tcp_wmem[1]; | |
1799 | sk->sk_rcvbuf = sysctl_tcp_rmem[1]; | |
1800 | ||
eb4dea58 | 1801 | local_bh_disable(); |
1748376b | 1802 | percpu_counter_inc(&tcp_sockets_allocated); |
eb4dea58 | 1803 | local_bh_enable(); |
1da177e4 LT |
1804 | |
1805 | return 0; | |
1806 | } | |
1807 | ||
7d06b2e0 | 1808 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1809 | { |
1810 | struct tcp_sock *tp = tcp_sk(sk); | |
1811 | ||
1812 | tcp_clear_xmit_timers(sk); | |
1813 | ||
6687e988 | 1814 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1815 | |
1da177e4 | 1816 | /* Cleanup up the write buffer. */ |
fe067e8a | 1817 | tcp_write_queue_purge(sk); |
1da177e4 LT |
1818 | |
1819 | /* Cleans up our, hopefully empty, out_of_order_queue. */ | |
e905a9ed | 1820 | __skb_queue_purge(&tp->out_of_order_queue); |
1da177e4 | 1821 | |
cfb6eeb4 YH |
1822 | #ifdef CONFIG_TCP_MD5SIG |
1823 | /* Clean up the MD5 key list, if any */ | |
1824 | if (tp->md5sig_info) { | |
1825 | tcp_v4_clear_md5_list(sk); | |
1826 | kfree(tp->md5sig_info); | |
1827 | tp->md5sig_info = NULL; | |
1828 | } | |
1829 | #endif | |
1830 | ||
1a2449a8 CL |
1831 | #ifdef CONFIG_NET_DMA |
1832 | /* Cleans up our sk_async_wait_queue */ | |
e905a9ed | 1833 | __skb_queue_purge(&sk->sk_async_wait_queue); |
1a2449a8 CL |
1834 | #endif |
1835 | ||
1da177e4 LT |
1836 | /* Clean prequeue, it must be empty really */ |
1837 | __skb_queue_purge(&tp->ucopy.prequeue); | |
1838 | ||
1839 | /* Clean up a referenced TCP bind bucket. */ | |
463c84b9 | 1840 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1841 | inet_put_port(sk); |
1da177e4 LT |
1842 | |
1843 | /* | |
1844 | * If sendmsg cached page exists, toss it. | |
1845 | */ | |
1846 | if (sk->sk_sndmsg_page) { | |
1847 | __free_page(sk->sk_sndmsg_page); | |
1848 | sk->sk_sndmsg_page = NULL; | |
1849 | } | |
1850 | ||
1748376b | 1851 | percpu_counter_dec(&tcp_sockets_allocated); |
1da177e4 LT |
1852 | } |
1853 | ||
1854 | EXPORT_SYMBOL(tcp_v4_destroy_sock); | |
1855 | ||
1856 | #ifdef CONFIG_PROC_FS | |
1857 | /* Proc filesystem TCP sock list dumping. */ | |
1858 | ||
3ab5aee7 | 1859 | static inline struct inet_timewait_sock *tw_head(struct hlist_nulls_head *head) |
1da177e4 | 1860 | { |
3ab5aee7 | 1861 | return hlist_nulls_empty(head) ? NULL : |
8feaf0c0 | 1862 | list_entry(head->first, struct inet_timewait_sock, tw_node); |
1da177e4 LT |
1863 | } |
1864 | ||
8feaf0c0 | 1865 | static inline struct inet_timewait_sock *tw_next(struct inet_timewait_sock *tw) |
1da177e4 | 1866 | { |
3ab5aee7 ED |
1867 | return !is_a_nulls(tw->tw_node.next) ? |
1868 | hlist_nulls_entry(tw->tw_node.next, typeof(*tw), tw_node) : NULL; | |
1da177e4 LT |
1869 | } |
1870 | ||
1871 | static void *listening_get_next(struct seq_file *seq, void *cur) | |
1872 | { | |
463c84b9 | 1873 | struct inet_connection_sock *icsk; |
c25eb3bf | 1874 | struct hlist_nulls_node *node; |
1da177e4 | 1875 | struct sock *sk = cur; |
5caea4ea | 1876 | struct inet_listen_hashbucket *ilb; |
5799de0b | 1877 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1878 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1879 | |
1880 | if (!sk) { | |
1881 | st->bucket = 0; | |
5caea4ea ED |
1882 | ilb = &tcp_hashinfo.listening_hash[0]; |
1883 | spin_lock_bh(&ilb->lock); | |
c25eb3bf | 1884 | sk = sk_nulls_head(&ilb->head); |
1da177e4 LT |
1885 | goto get_sk; |
1886 | } | |
5caea4ea | 1887 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 LT |
1888 | ++st->num; |
1889 | ||
1890 | if (st->state == TCP_SEQ_STATE_OPENREQ) { | |
60236fdd | 1891 | struct request_sock *req = cur; |
1da177e4 | 1892 | |
72a3effa | 1893 | icsk = inet_csk(st->syn_wait_sk); |
1da177e4 LT |
1894 | req = req->dl_next; |
1895 | while (1) { | |
1896 | while (req) { | |
bdccc4ca | 1897 | if (req->rsk_ops->family == st->family) { |
1da177e4 LT |
1898 | cur = req; |
1899 | goto out; | |
1900 | } | |
1901 | req = req->dl_next; | |
1902 | } | |
72a3effa | 1903 | if (++st->sbucket >= icsk->icsk_accept_queue.listen_opt->nr_table_entries) |
1da177e4 LT |
1904 | break; |
1905 | get_req: | |
463c84b9 | 1906 | req = icsk->icsk_accept_queue.listen_opt->syn_table[st->sbucket]; |
1da177e4 LT |
1907 | } |
1908 | sk = sk_next(st->syn_wait_sk); | |
1909 | st->state = TCP_SEQ_STATE_LISTENING; | |
463c84b9 | 1910 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 1911 | } else { |
e905a9ed | 1912 | icsk = inet_csk(sk); |
463c84b9 ACM |
1913 | read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1914 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) | |
1da177e4 | 1915 | goto start_req; |
463c84b9 | 1916 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 LT |
1917 | sk = sk_next(sk); |
1918 | } | |
1919 | get_sk: | |
c25eb3bf | 1920 | sk_nulls_for_each_from(sk, node) { |
878628fb | 1921 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) { |
1da177e4 LT |
1922 | cur = sk; |
1923 | goto out; | |
1924 | } | |
e905a9ed | 1925 | icsk = inet_csk(sk); |
463c84b9 ACM |
1926 | read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1927 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) { | |
1da177e4 LT |
1928 | start_req: |
1929 | st->uid = sock_i_uid(sk); | |
1930 | st->syn_wait_sk = sk; | |
1931 | st->state = TCP_SEQ_STATE_OPENREQ; | |
1932 | st->sbucket = 0; | |
1933 | goto get_req; | |
1934 | } | |
463c84b9 | 1935 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 1936 | } |
5caea4ea | 1937 | spin_unlock_bh(&ilb->lock); |
0f7ff927 | 1938 | if (++st->bucket < INET_LHTABLE_SIZE) { |
5caea4ea ED |
1939 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1940 | spin_lock_bh(&ilb->lock); | |
c25eb3bf | 1941 | sk = sk_nulls_head(&ilb->head); |
1da177e4 LT |
1942 | goto get_sk; |
1943 | } | |
1944 | cur = NULL; | |
1945 | out: | |
1946 | return cur; | |
1947 | } | |
1948 | ||
1949 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
1950 | { | |
1951 | void *rc = listening_get_next(seq, NULL); | |
1952 | ||
1953 | while (rc && *pos) { | |
1954 | rc = listening_get_next(seq, rc); | |
1955 | --*pos; | |
1956 | } | |
1957 | return rc; | |
1958 | } | |
1959 | ||
6eac5604 AK |
1960 | static inline int empty_bucket(struct tcp_iter_state *st) |
1961 | { | |
3ab5aee7 ED |
1962 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain) && |
1963 | hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].twchain); | |
6eac5604 AK |
1964 | } |
1965 | ||
1da177e4 LT |
1966 | static void *established_get_first(struct seq_file *seq) |
1967 | { | |
5799de0b | 1968 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1969 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1970 | void *rc = NULL; |
1971 | ||
6e04e021 | 1972 | for (st->bucket = 0; st->bucket < tcp_hashinfo.ehash_size; ++st->bucket) { |
1da177e4 | 1973 | struct sock *sk; |
3ab5aee7 | 1974 | struct hlist_nulls_node *node; |
8feaf0c0 | 1975 | struct inet_timewait_sock *tw; |
9db66bdc | 1976 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 1977 | |
6eac5604 AK |
1978 | /* Lockless fast path for the common case of empty buckets */ |
1979 | if (empty_bucket(st)) | |
1980 | continue; | |
1981 | ||
9db66bdc | 1982 | spin_lock_bh(lock); |
3ab5aee7 | 1983 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 1984 | if (sk->sk_family != st->family || |
878628fb | 1985 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
1986 | continue; |
1987 | } | |
1988 | rc = sk; | |
1989 | goto out; | |
1990 | } | |
1991 | st->state = TCP_SEQ_STATE_TIME_WAIT; | |
8feaf0c0 | 1992 | inet_twsk_for_each(tw, node, |
dbca9b27 | 1993 | &tcp_hashinfo.ehash[st->bucket].twchain) { |
28518fc1 | 1994 | if (tw->tw_family != st->family || |
878628fb | 1995 | !net_eq(twsk_net(tw), net)) { |
1da177e4 LT |
1996 | continue; |
1997 | } | |
1998 | rc = tw; | |
1999 | goto out; | |
2000 | } | |
9db66bdc | 2001 | spin_unlock_bh(lock); |
1da177e4 LT |
2002 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2003 | } | |
2004 | out: | |
2005 | return rc; | |
2006 | } | |
2007 | ||
2008 | static void *established_get_next(struct seq_file *seq, void *cur) | |
2009 | { | |
2010 | struct sock *sk = cur; | |
8feaf0c0 | 2011 | struct inet_timewait_sock *tw; |
3ab5aee7 | 2012 | struct hlist_nulls_node *node; |
5799de0b | 2013 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2014 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2015 | |
2016 | ++st->num; | |
2017 | ||
2018 | if (st->state == TCP_SEQ_STATE_TIME_WAIT) { | |
2019 | tw = cur; | |
2020 | tw = tw_next(tw); | |
2021 | get_tw: | |
878628fb | 2022 | while (tw && (tw->tw_family != st->family || !net_eq(twsk_net(tw), net))) { |
1da177e4 LT |
2023 | tw = tw_next(tw); |
2024 | } | |
2025 | if (tw) { | |
2026 | cur = tw; | |
2027 | goto out; | |
2028 | } | |
9db66bdc | 2029 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2030 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2031 | ||
6eac5604 AK |
2032 | /* Look for next non empty bucket */ |
2033 | while (++st->bucket < tcp_hashinfo.ehash_size && | |
2034 | empty_bucket(st)) | |
2035 | ; | |
2036 | if (st->bucket >= tcp_hashinfo.ehash_size) | |
2037 | return NULL; | |
2038 | ||
9db66bdc | 2039 | spin_lock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
3ab5aee7 | 2040 | sk = sk_nulls_head(&tcp_hashinfo.ehash[st->bucket].chain); |
1da177e4 | 2041 | } else |
3ab5aee7 | 2042 | sk = sk_nulls_next(sk); |
1da177e4 | 2043 | |
3ab5aee7 | 2044 | sk_nulls_for_each_from(sk, node) { |
878628fb | 2045 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
1da177e4 LT |
2046 | goto found; |
2047 | } | |
2048 | ||
2049 | st->state = TCP_SEQ_STATE_TIME_WAIT; | |
dbca9b27 | 2050 | tw = tw_head(&tcp_hashinfo.ehash[st->bucket].twchain); |
1da177e4 LT |
2051 | goto get_tw; |
2052 | found: | |
2053 | cur = sk; | |
2054 | out: | |
2055 | return cur; | |
2056 | } | |
2057 | ||
2058 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2059 | { | |
2060 | void *rc = established_get_first(seq); | |
2061 | ||
2062 | while (rc && pos) { | |
2063 | rc = established_get_next(seq, rc); | |
2064 | --pos; | |
7174259e | 2065 | } |
1da177e4 LT |
2066 | return rc; |
2067 | } | |
2068 | ||
2069 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2070 | { | |
2071 | void *rc; | |
5799de0b | 2072 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2073 | |
1da177e4 LT |
2074 | st->state = TCP_SEQ_STATE_LISTENING; |
2075 | rc = listening_get_idx(seq, &pos); | |
2076 | ||
2077 | if (!rc) { | |
1da177e4 LT |
2078 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2079 | rc = established_get_idx(seq, pos); | |
2080 | } | |
2081 | ||
2082 | return rc; | |
2083 | } | |
2084 | ||
2085 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) | |
2086 | { | |
5799de0b | 2087 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2088 | st->state = TCP_SEQ_STATE_LISTENING; |
2089 | st->num = 0; | |
2090 | return *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2091 | } | |
2092 | ||
2093 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2094 | { | |
2095 | void *rc = NULL; | |
5799de0b | 2096 | struct tcp_iter_state *st; |
1da177e4 LT |
2097 | |
2098 | if (v == SEQ_START_TOKEN) { | |
2099 | rc = tcp_get_idx(seq, 0); | |
2100 | goto out; | |
2101 | } | |
2102 | st = seq->private; | |
2103 | ||
2104 | switch (st->state) { | |
2105 | case TCP_SEQ_STATE_OPENREQ: | |
2106 | case TCP_SEQ_STATE_LISTENING: | |
2107 | rc = listening_get_next(seq, v); | |
2108 | if (!rc) { | |
1da177e4 LT |
2109 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2110 | rc = established_get_first(seq); | |
2111 | } | |
2112 | break; | |
2113 | case TCP_SEQ_STATE_ESTABLISHED: | |
2114 | case TCP_SEQ_STATE_TIME_WAIT: | |
2115 | rc = established_get_next(seq, v); | |
2116 | break; | |
2117 | } | |
2118 | out: | |
2119 | ++*pos; | |
2120 | return rc; | |
2121 | } | |
2122 | ||
2123 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2124 | { | |
5799de0b | 2125 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2126 | |
2127 | switch (st->state) { | |
2128 | case TCP_SEQ_STATE_OPENREQ: | |
2129 | if (v) { | |
463c84b9 ACM |
2130 | struct inet_connection_sock *icsk = inet_csk(st->syn_wait_sk); |
2131 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); | |
1da177e4 LT |
2132 | } |
2133 | case TCP_SEQ_STATE_LISTENING: | |
2134 | if (v != SEQ_START_TOKEN) | |
5caea4ea | 2135 | spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 LT |
2136 | break; |
2137 | case TCP_SEQ_STATE_TIME_WAIT: | |
2138 | case TCP_SEQ_STATE_ESTABLISHED: | |
2139 | if (v) | |
9db66bdc | 2140 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2141 | break; |
2142 | } | |
2143 | } | |
2144 | ||
2145 | static int tcp_seq_open(struct inode *inode, struct file *file) | |
2146 | { | |
2147 | struct tcp_seq_afinfo *afinfo = PDE(inode)->data; | |
1da177e4 | 2148 | struct tcp_iter_state *s; |
52d6f3f1 | 2149 | int err; |
1da177e4 | 2150 | |
52d6f3f1 DL |
2151 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2152 | sizeof(struct tcp_iter_state)); | |
2153 | if (err < 0) | |
2154 | return err; | |
f40c8174 | 2155 | |
52d6f3f1 | 2156 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2157 | s->family = afinfo->family; |
f40c8174 DL |
2158 | return 0; |
2159 | } | |
2160 | ||
6f8b13bc | 2161 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2162 | { |
2163 | int rc = 0; | |
2164 | struct proc_dir_entry *p; | |
2165 | ||
68fcadd1 DL |
2166 | afinfo->seq_fops.open = tcp_seq_open; |
2167 | afinfo->seq_fops.read = seq_read; | |
2168 | afinfo->seq_fops.llseek = seq_lseek; | |
2169 | afinfo->seq_fops.release = seq_release_net; | |
7174259e | 2170 | |
9427c4b3 DL |
2171 | afinfo->seq_ops.start = tcp_seq_start; |
2172 | afinfo->seq_ops.next = tcp_seq_next; | |
2173 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2174 | ||
84841c3c DL |
2175 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
2176 | &afinfo->seq_fops, afinfo); | |
2177 | if (!p) | |
1da177e4 LT |
2178 | rc = -ENOMEM; |
2179 | return rc; | |
2180 | } | |
2181 | ||
6f8b13bc | 2182 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2183 | { |
6f8b13bc | 2184 | proc_net_remove(net, afinfo->name); |
1da177e4 LT |
2185 | } |
2186 | ||
60236fdd | 2187 | static void get_openreq4(struct sock *sk, struct request_sock *req, |
5e659e4c | 2188 | struct seq_file *f, int i, int uid, int *len) |
1da177e4 | 2189 | { |
2e6599cb | 2190 | const struct inet_request_sock *ireq = inet_rsk(req); |
1da177e4 LT |
2191 | int ttd = req->expires - jiffies; |
2192 | ||
5e659e4c PE |
2193 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
2194 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %u %d %p%n", | |
1da177e4 | 2195 | i, |
2e6599cb | 2196 | ireq->loc_addr, |
1da177e4 | 2197 | ntohs(inet_sk(sk)->sport), |
2e6599cb ACM |
2198 | ireq->rmt_addr, |
2199 | ntohs(ireq->rmt_port), | |
1da177e4 LT |
2200 | TCP_SYN_RECV, |
2201 | 0, 0, /* could print option size, but that is af dependent. */ | |
2202 | 1, /* timers active (only the expire timer) */ | |
2203 | jiffies_to_clock_t(ttd), | |
2204 | req->retrans, | |
2205 | uid, | |
2206 | 0, /* non standard timer */ | |
2207 | 0, /* open_requests have no inode */ | |
2208 | atomic_read(&sk->sk_refcnt), | |
5e659e4c PE |
2209 | req, |
2210 | len); | |
1da177e4 LT |
2211 | } |
2212 | ||
5e659e4c | 2213 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i, int *len) |
1da177e4 LT |
2214 | { |
2215 | int timer_active; | |
2216 | unsigned long timer_expires; | |
cf4c6bf8 IJ |
2217 | struct tcp_sock *tp = tcp_sk(sk); |
2218 | const struct inet_connection_sock *icsk = inet_csk(sk); | |
2219 | struct inet_sock *inet = inet_sk(sk); | |
714e85be AV |
2220 | __be32 dest = inet->daddr; |
2221 | __be32 src = inet->rcv_saddr; | |
1da177e4 LT |
2222 | __u16 destp = ntohs(inet->dport); |
2223 | __u16 srcp = ntohs(inet->sport); | |
2224 | ||
463c84b9 | 2225 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { |
1da177e4 | 2226 | timer_active = 1; |
463c84b9 ACM |
2227 | timer_expires = icsk->icsk_timeout; |
2228 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2229 | timer_active = 4; |
463c84b9 | 2230 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2231 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2232 | timer_active = 2; |
cf4c6bf8 | 2233 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2234 | } else { |
2235 | timer_active = 0; | |
2236 | timer_expires = jiffies; | |
2237 | } | |
2238 | ||
5e659e4c | 2239 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
7be87351 | 2240 | "%08X %5d %8d %lu %d %p %lu %lu %u %u %d%n", |
cf4c6bf8 | 2241 | i, src, srcp, dest, destp, sk->sk_state, |
47da8ee6 | 2242 | tp->write_seq - tp->snd_una, |
cf4c6bf8 | 2243 | sk->sk_state == TCP_LISTEN ? sk->sk_ack_backlog : |
7174259e | 2244 | (tp->rcv_nxt - tp->copied_seq), |
1da177e4 LT |
2245 | timer_active, |
2246 | jiffies_to_clock_t(timer_expires - jiffies), | |
463c84b9 | 2247 | icsk->icsk_retransmits, |
cf4c6bf8 | 2248 | sock_i_uid(sk), |
6687e988 | 2249 | icsk->icsk_probes_out, |
cf4c6bf8 IJ |
2250 | sock_i_ino(sk), |
2251 | atomic_read(&sk->sk_refcnt), sk, | |
7be87351 SH |
2252 | jiffies_to_clock_t(icsk->icsk_rto), |
2253 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2254 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2255 | tp->snd_cwnd, |
5e659e4c PE |
2256 | tp->snd_ssthresh >= 0xFFFF ? -1 : tp->snd_ssthresh, |
2257 | len); | |
1da177e4 LT |
2258 | } |
2259 | ||
7174259e | 2260 | static void get_timewait4_sock(struct inet_timewait_sock *tw, |
5e659e4c | 2261 | struct seq_file *f, int i, int *len) |
1da177e4 | 2262 | { |
23f33c2d | 2263 | __be32 dest, src; |
1da177e4 LT |
2264 | __u16 destp, srcp; |
2265 | int ttd = tw->tw_ttd - jiffies; | |
2266 | ||
2267 | if (ttd < 0) | |
2268 | ttd = 0; | |
2269 | ||
2270 | dest = tw->tw_daddr; | |
2271 | src = tw->tw_rcv_saddr; | |
2272 | destp = ntohs(tw->tw_dport); | |
2273 | srcp = ntohs(tw->tw_sport); | |
2274 | ||
5e659e4c PE |
2275 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
2276 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p%n", | |
1da177e4 LT |
2277 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
2278 | 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0, | |
5e659e4c | 2279 | atomic_read(&tw->tw_refcnt), tw, len); |
1da177e4 LT |
2280 | } |
2281 | ||
2282 | #define TMPSZ 150 | |
2283 | ||
2284 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2285 | { | |
5799de0b | 2286 | struct tcp_iter_state *st; |
5e659e4c | 2287 | int len; |
1da177e4 LT |
2288 | |
2289 | if (v == SEQ_START_TOKEN) { | |
2290 | seq_printf(seq, "%-*s\n", TMPSZ - 1, | |
2291 | " sl local_address rem_address st tx_queue " | |
2292 | "rx_queue tr tm->when retrnsmt uid timeout " | |
2293 | "inode"); | |
2294 | goto out; | |
2295 | } | |
2296 | st = seq->private; | |
2297 | ||
2298 | switch (st->state) { | |
2299 | case TCP_SEQ_STATE_LISTENING: | |
2300 | case TCP_SEQ_STATE_ESTABLISHED: | |
5e659e4c | 2301 | get_tcp4_sock(v, seq, st->num, &len); |
1da177e4 LT |
2302 | break; |
2303 | case TCP_SEQ_STATE_OPENREQ: | |
5e659e4c | 2304 | get_openreq4(st->syn_wait_sk, v, seq, st->num, st->uid, &len); |
1da177e4 LT |
2305 | break; |
2306 | case TCP_SEQ_STATE_TIME_WAIT: | |
5e659e4c | 2307 | get_timewait4_sock(v, seq, st->num, &len); |
1da177e4 LT |
2308 | break; |
2309 | } | |
5e659e4c | 2310 | seq_printf(seq, "%*s\n", TMPSZ - 1 - len, ""); |
1da177e4 LT |
2311 | out: |
2312 | return 0; | |
2313 | } | |
2314 | ||
1da177e4 | 2315 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2316 | .name = "tcp", |
2317 | .family = AF_INET, | |
5f4472c5 DL |
2318 | .seq_fops = { |
2319 | .owner = THIS_MODULE, | |
2320 | }, | |
9427c4b3 DL |
2321 | .seq_ops = { |
2322 | .show = tcp4_seq_show, | |
2323 | }, | |
1da177e4 LT |
2324 | }; |
2325 | ||
757764f6 PE |
2326 | static int tcp4_proc_init_net(struct net *net) |
2327 | { | |
2328 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2329 | } | |
2330 | ||
2331 | static void tcp4_proc_exit_net(struct net *net) | |
2332 | { | |
2333 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2334 | } | |
2335 | ||
2336 | static struct pernet_operations tcp4_net_ops = { | |
2337 | .init = tcp4_proc_init_net, | |
2338 | .exit = tcp4_proc_exit_net, | |
2339 | }; | |
2340 | ||
1da177e4 LT |
2341 | int __init tcp4_proc_init(void) |
2342 | { | |
757764f6 | 2343 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2344 | } |
2345 | ||
2346 | void tcp4_proc_exit(void) | |
2347 | { | |
757764f6 | 2348 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2349 | } |
2350 | #endif /* CONFIG_PROC_FS */ | |
2351 | ||
bf296b12 HX |
2352 | struct sk_buff **tcp4_gro_receive(struct sk_buff **head, struct sk_buff *skb) |
2353 | { | |
2354 | struct iphdr *iph = ip_hdr(skb); | |
2355 | ||
2356 | switch (skb->ip_summed) { | |
2357 | case CHECKSUM_COMPLETE: | |
2358 | if (!tcp_v4_check(skb->len, iph->saddr, iph->daddr, | |
2359 | skb->csum)) { | |
2360 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
2361 | break; | |
2362 | } | |
2363 | ||
2364 | /* fall through */ | |
2365 | case CHECKSUM_NONE: | |
2366 | NAPI_GRO_CB(skb)->flush = 1; | |
2367 | return NULL; | |
2368 | } | |
2369 | ||
2370 | return tcp_gro_receive(head, skb); | |
2371 | } | |
2372 | EXPORT_SYMBOL(tcp4_gro_receive); | |
2373 | ||
2374 | int tcp4_gro_complete(struct sk_buff *skb) | |
2375 | { | |
2376 | struct iphdr *iph = ip_hdr(skb); | |
2377 | struct tcphdr *th = tcp_hdr(skb); | |
2378 | ||
2379 | th->check = ~tcp_v4_check(skb->len - skb_transport_offset(skb), | |
2380 | iph->saddr, iph->daddr, 0); | |
2381 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; | |
2382 | ||
2383 | return tcp_gro_complete(skb); | |
2384 | } | |
2385 | EXPORT_SYMBOL(tcp4_gro_complete); | |
2386 | ||
1da177e4 LT |
2387 | struct proto tcp_prot = { |
2388 | .name = "TCP", | |
2389 | .owner = THIS_MODULE, | |
2390 | .close = tcp_close, | |
2391 | .connect = tcp_v4_connect, | |
2392 | .disconnect = tcp_disconnect, | |
463c84b9 | 2393 | .accept = inet_csk_accept, |
1da177e4 LT |
2394 | .ioctl = tcp_ioctl, |
2395 | .init = tcp_v4_init_sock, | |
2396 | .destroy = tcp_v4_destroy_sock, | |
2397 | .shutdown = tcp_shutdown, | |
2398 | .setsockopt = tcp_setsockopt, | |
2399 | .getsockopt = tcp_getsockopt, | |
1da177e4 LT |
2400 | .recvmsg = tcp_recvmsg, |
2401 | .backlog_rcv = tcp_v4_do_rcv, | |
ab1e0a13 ACM |
2402 | .hash = inet_hash, |
2403 | .unhash = inet_unhash, | |
2404 | .get_port = inet_csk_get_port, | |
1da177e4 LT |
2405 | .enter_memory_pressure = tcp_enter_memory_pressure, |
2406 | .sockets_allocated = &tcp_sockets_allocated, | |
0a5578cf | 2407 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2408 | .memory_allocated = &tcp_memory_allocated, |
2409 | .memory_pressure = &tcp_memory_pressure, | |
2410 | .sysctl_mem = sysctl_tcp_mem, | |
2411 | .sysctl_wmem = sysctl_tcp_wmem, | |
2412 | .sysctl_rmem = sysctl_tcp_rmem, | |
2413 | .max_header = MAX_TCP_HEADER, | |
2414 | .obj_size = sizeof(struct tcp_sock), | |
3ab5aee7 | 2415 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2416 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2417 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2418 | .h.hashinfo = &tcp_hashinfo, |
543d9cfe ACM |
2419 | #ifdef CONFIG_COMPAT |
2420 | .compat_setsockopt = compat_tcp_setsockopt, | |
2421 | .compat_getsockopt = compat_tcp_getsockopt, | |
2422 | #endif | |
1da177e4 LT |
2423 | }; |
2424 | ||
046ee902 DL |
2425 | |
2426 | static int __net_init tcp_sk_init(struct net *net) | |
2427 | { | |
2428 | return inet_ctl_sock_create(&net->ipv4.tcp_sock, | |
2429 | PF_INET, SOCK_RAW, IPPROTO_TCP, net); | |
2430 | } | |
2431 | ||
2432 | static void __net_exit tcp_sk_exit(struct net *net) | |
2433 | { | |
2434 | inet_ctl_sock_destroy(net->ipv4.tcp_sock); | |
d315492b | 2435 | inet_twsk_purge(net, &tcp_hashinfo, &tcp_death_row, AF_INET); |
046ee902 DL |
2436 | } |
2437 | ||
2438 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
2439 | .init = tcp_sk_init, | |
2440 | .exit = tcp_sk_exit, | |
2441 | }; | |
2442 | ||
9b0f976f | 2443 | void __init tcp_v4_init(void) |
1da177e4 | 2444 | { |
5caea4ea | 2445 | inet_hashinfo_init(&tcp_hashinfo); |
2f20d2e6 | 2446 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2447 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 LT |
2448 | } |
2449 | ||
2450 | EXPORT_SYMBOL(ipv4_specific); | |
1da177e4 | 2451 | EXPORT_SYMBOL(tcp_hashinfo); |
1da177e4 | 2452 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 LT |
2453 | EXPORT_SYMBOL(tcp_v4_conn_request); |
2454 | EXPORT_SYMBOL(tcp_v4_connect); | |
2455 | EXPORT_SYMBOL(tcp_v4_do_rcv); | |
1da177e4 LT |
2456 | EXPORT_SYMBOL(tcp_v4_remember_stamp); |
2457 | EXPORT_SYMBOL(tcp_v4_send_check); | |
2458 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); | |
2459 | ||
2460 | #ifdef CONFIG_PROC_FS | |
2461 | EXPORT_SYMBOL(tcp_proc_register); | |
2462 | EXPORT_SYMBOL(tcp_proc_unregister); | |
2463 | #endif | |
1da177e4 | 2464 | EXPORT_SYMBOL(sysctl_tcp_low_latency); |
1da177e4 | 2465 |