]>
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> | |
5a0e3ad6 | 63 | #include <linux/slab.h> |
1da177e4 | 64 | |
457c4cbc | 65 | #include <net/net_namespace.h> |
1da177e4 | 66 | #include <net/icmp.h> |
304a1618 | 67 | #include <net/inet_hashtables.h> |
1da177e4 | 68 | #include <net/tcp.h> |
20380731 | 69 | #include <net/transp_v6.h> |
1da177e4 LT |
70 | #include <net/ipv6.h> |
71 | #include <net/inet_common.h> | |
6d6ee43e | 72 | #include <net/timewait_sock.h> |
1da177e4 | 73 | #include <net/xfrm.h> |
1a2449a8 | 74 | #include <net/netdma.h> |
6e5714ea | 75 | #include <net/secure_seq.h> |
1da177e4 LT |
76 | |
77 | #include <linux/inet.h> | |
78 | #include <linux/ipv6.h> | |
79 | #include <linux/stddef.h> | |
80 | #include <linux/proc_fs.h> | |
81 | #include <linux/seq_file.h> | |
82 | ||
cfb6eeb4 YH |
83 | #include <linux/crypto.h> |
84 | #include <linux/scatterlist.h> | |
85 | ||
ab32ea5d BH |
86 | int sysctl_tcp_tw_reuse __read_mostly; |
87 | int sysctl_tcp_low_latency __read_mostly; | |
4bc2f18b | 88 | EXPORT_SYMBOL(sysctl_tcp_low_latency); |
1da177e4 | 89 | |
1da177e4 | 90 | |
cfb6eeb4 | 91 | #ifdef CONFIG_TCP_MD5SIG |
7174259e ACM |
92 | static struct tcp_md5sig_key *tcp_v4_md5_do_lookup(struct sock *sk, |
93 | __be32 addr); | |
49a72dfb AL |
94 | static int tcp_v4_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, |
95 | __be32 daddr, __be32 saddr, struct tcphdr *th); | |
9501f972 YH |
96 | #else |
97 | static inline | |
98 | struct tcp_md5sig_key *tcp_v4_md5_do_lookup(struct sock *sk, __be32 addr) | |
99 | { | |
100 | return NULL; | |
101 | } | |
cfb6eeb4 YH |
102 | #endif |
103 | ||
5caea4ea | 104 | struct inet_hashinfo tcp_hashinfo; |
4bc2f18b | 105 | EXPORT_SYMBOL(tcp_hashinfo); |
1da177e4 | 106 | |
a94f723d | 107 | static inline __u32 tcp_v4_init_sequence(struct sk_buff *skb) |
1da177e4 | 108 | { |
eddc9ec5 ACM |
109 | return secure_tcp_sequence_number(ip_hdr(skb)->daddr, |
110 | ip_hdr(skb)->saddr, | |
aa8223c7 ACM |
111 | tcp_hdr(skb)->dest, |
112 | tcp_hdr(skb)->source); | |
1da177e4 LT |
113 | } |
114 | ||
6d6ee43e ACM |
115 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp) |
116 | { | |
117 | const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw); | |
118 | struct tcp_sock *tp = tcp_sk(sk); | |
119 | ||
120 | /* With PAWS, it is safe from the viewpoint | |
121 | of data integrity. Even without PAWS it is safe provided sequence | |
122 | spaces do not overlap i.e. at data rates <= 80Mbit/sec. | |
123 | ||
124 | Actually, the idea is close to VJ's one, only timestamp cache is | |
125 | held not per host, but per port pair and TW bucket is used as state | |
126 | holder. | |
127 | ||
128 | If TW bucket has been already destroyed we fall back to VJ's scheme | |
129 | and use initial timestamp retrieved from peer table. | |
130 | */ | |
131 | if (tcptw->tw_ts_recent_stamp && | |
132 | (twp == NULL || (sysctl_tcp_tw_reuse && | |
9d729f72 | 133 | get_seconds() - tcptw->tw_ts_recent_stamp > 1))) { |
6d6ee43e ACM |
134 | tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2; |
135 | if (tp->write_seq == 0) | |
136 | tp->write_seq = 1; | |
137 | tp->rx_opt.ts_recent = tcptw->tw_ts_recent; | |
138 | tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp; | |
139 | sock_hold(sktw); | |
140 | return 1; | |
141 | } | |
142 | ||
143 | return 0; | |
144 | } | |
6d6ee43e ACM |
145 | EXPORT_SYMBOL_GPL(tcp_twsk_unique); |
146 | ||
1da177e4 LT |
147 | /* This will initiate an outgoing connection. */ |
148 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
149 | { | |
2d7192d6 | 150 | struct sockaddr_in *usin = (struct sockaddr_in *)uaddr; |
1da177e4 LT |
151 | struct inet_sock *inet = inet_sk(sk); |
152 | struct tcp_sock *tp = tcp_sk(sk); | |
dca8b089 | 153 | __be16 orig_sport, orig_dport; |
bada8adc | 154 | __be32 daddr, nexthop; |
da905bd1 | 155 | struct flowi4 *fl4; |
2d7192d6 | 156 | struct rtable *rt; |
1da177e4 | 157 | int err; |
f6d8bd05 | 158 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
159 | |
160 | if (addr_len < sizeof(struct sockaddr_in)) | |
161 | return -EINVAL; | |
162 | ||
163 | if (usin->sin_family != AF_INET) | |
164 | return -EAFNOSUPPORT; | |
165 | ||
166 | nexthop = daddr = usin->sin_addr.s_addr; | |
f6d8bd05 ED |
167 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
168 | sock_owned_by_user(sk)); | |
169 | if (inet_opt && inet_opt->opt.srr) { | |
1da177e4 LT |
170 | if (!daddr) |
171 | return -EINVAL; | |
f6d8bd05 | 172 | nexthop = inet_opt->opt.faddr; |
1da177e4 LT |
173 | } |
174 | ||
dca8b089 DM |
175 | orig_sport = inet->inet_sport; |
176 | orig_dport = usin->sin_port; | |
da905bd1 DM |
177 | fl4 = &inet->cork.fl.u.ip4; |
178 | rt = ip_route_connect(fl4, nexthop, inet->inet_saddr, | |
b23dd4fe DM |
179 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if, |
180 | IPPROTO_TCP, | |
181 | orig_sport, orig_dport, sk, true); | |
182 | if (IS_ERR(rt)) { | |
183 | err = PTR_ERR(rt); | |
184 | if (err == -ENETUNREACH) | |
7c73a6fa | 185 | IP_INC_STATS_BH(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); |
b23dd4fe | 186 | return err; |
584bdf8c | 187 | } |
1da177e4 LT |
188 | |
189 | if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) { | |
190 | ip_rt_put(rt); | |
191 | return -ENETUNREACH; | |
192 | } | |
193 | ||
f6d8bd05 | 194 | if (!inet_opt || !inet_opt->opt.srr) |
da905bd1 | 195 | daddr = fl4->daddr; |
1da177e4 | 196 | |
c720c7e8 | 197 | if (!inet->inet_saddr) |
da905bd1 | 198 | inet->inet_saddr = fl4->saddr; |
c720c7e8 | 199 | inet->inet_rcv_saddr = inet->inet_saddr; |
1da177e4 | 200 | |
c720c7e8 | 201 | if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) { |
1da177e4 LT |
202 | /* Reset inherited state */ |
203 | tp->rx_opt.ts_recent = 0; | |
204 | tp->rx_opt.ts_recent_stamp = 0; | |
205 | tp->write_seq = 0; | |
206 | } | |
207 | ||
295ff7ed | 208 | if (tcp_death_row.sysctl_tw_recycle && |
da905bd1 | 209 | !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr) { |
ed2361e6 | 210 | struct inet_peer *peer = rt_get_peer(rt, fl4->daddr); |
7174259e ACM |
211 | /* |
212 | * VJ's idea. We save last timestamp seen from | |
213 | * the destination in peer table, when entering state | |
214 | * TIME-WAIT * and initialize rx_opt.ts_recent from it, | |
215 | * when trying new connection. | |
1da177e4 | 216 | */ |
317fe0e6 ED |
217 | if (peer) { |
218 | inet_peer_refcheck(peer); | |
219 | if ((u32)get_seconds() - peer->tcp_ts_stamp <= TCP_PAWS_MSL) { | |
220 | tp->rx_opt.ts_recent_stamp = peer->tcp_ts_stamp; | |
221 | tp->rx_opt.ts_recent = peer->tcp_ts; | |
222 | } | |
1da177e4 LT |
223 | } |
224 | } | |
225 | ||
c720c7e8 ED |
226 | inet->inet_dport = usin->sin_port; |
227 | inet->inet_daddr = daddr; | |
1da177e4 | 228 | |
d83d8461 | 229 | inet_csk(sk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
230 | if (inet_opt) |
231 | inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
1da177e4 | 232 | |
bee7ca9e | 233 | tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT; |
1da177e4 LT |
234 | |
235 | /* Socket identity is still unknown (sport may be zero). | |
236 | * However we set state to SYN-SENT and not releasing socket | |
237 | * lock select source port, enter ourselves into the hash tables and | |
238 | * complete initialization after this. | |
239 | */ | |
240 | tcp_set_state(sk, TCP_SYN_SENT); | |
a7f5e7f1 | 241 | err = inet_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
242 | if (err) |
243 | goto failure; | |
244 | ||
da905bd1 | 245 | rt = ip_route_newports(fl4, rt, orig_sport, orig_dport, |
b23dd4fe DM |
246 | inet->inet_sport, inet->inet_dport, sk); |
247 | if (IS_ERR(rt)) { | |
248 | err = PTR_ERR(rt); | |
249 | rt = NULL; | |
1da177e4 | 250 | goto failure; |
b23dd4fe | 251 | } |
1da177e4 | 252 | /* OK, now commit destination to socket. */ |
bcd76111 | 253 | sk->sk_gso_type = SKB_GSO_TCPV4; |
d8d1f30b | 254 | sk_setup_caps(sk, &rt->dst); |
1da177e4 LT |
255 | |
256 | if (!tp->write_seq) | |
c720c7e8 ED |
257 | tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr, |
258 | inet->inet_daddr, | |
259 | inet->inet_sport, | |
1da177e4 LT |
260 | usin->sin_port); |
261 | ||
c720c7e8 | 262 | inet->inet_id = tp->write_seq ^ jiffies; |
1da177e4 LT |
263 | |
264 | err = tcp_connect(sk); | |
265 | rt = NULL; | |
266 | if (err) | |
267 | goto failure; | |
268 | ||
269 | return 0; | |
270 | ||
271 | failure: | |
7174259e ACM |
272 | /* |
273 | * This unhashes the socket and releases the local port, | |
274 | * if necessary. | |
275 | */ | |
1da177e4 LT |
276 | tcp_set_state(sk, TCP_CLOSE); |
277 | ip_rt_put(rt); | |
278 | sk->sk_route_caps = 0; | |
c720c7e8 | 279 | inet->inet_dport = 0; |
1da177e4 LT |
280 | return err; |
281 | } | |
4bc2f18b | 282 | EXPORT_SYMBOL(tcp_v4_connect); |
1da177e4 | 283 | |
1da177e4 LT |
284 | /* |
285 | * This routine does path mtu discovery as defined in RFC1191. | |
286 | */ | |
b71d1d42 | 287 | static void do_pmtu_discovery(struct sock *sk, const struct iphdr *iph, u32 mtu) |
1da177e4 LT |
288 | { |
289 | struct dst_entry *dst; | |
290 | struct inet_sock *inet = inet_sk(sk); | |
1da177e4 LT |
291 | |
292 | /* We are not interested in TCP_LISTEN and open_requests (SYN-ACKs | |
293 | * send out by Linux are always <576bytes so they should go through | |
294 | * unfragmented). | |
295 | */ | |
296 | if (sk->sk_state == TCP_LISTEN) | |
297 | return; | |
298 | ||
299 | /* We don't check in the destentry if pmtu discovery is forbidden | |
300 | * on this route. We just assume that no packet_to_big packets | |
301 | * are send back when pmtu discovery is not active. | |
e905a9ed | 302 | * There is a small race when the user changes this flag in the |
1da177e4 LT |
303 | * route, but I think that's acceptable. |
304 | */ | |
305 | if ((dst = __sk_dst_check(sk, 0)) == NULL) | |
306 | return; | |
307 | ||
308 | dst->ops->update_pmtu(dst, mtu); | |
309 | ||
310 | /* Something is about to be wrong... Remember soft error | |
311 | * for the case, if this connection will not able to recover. | |
312 | */ | |
313 | if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst)) | |
314 | sk->sk_err_soft = EMSGSIZE; | |
315 | ||
316 | mtu = dst_mtu(dst); | |
317 | ||
318 | if (inet->pmtudisc != IP_PMTUDISC_DONT && | |
d83d8461 | 319 | inet_csk(sk)->icsk_pmtu_cookie > mtu) { |
1da177e4 LT |
320 | tcp_sync_mss(sk, mtu); |
321 | ||
322 | /* Resend the TCP packet because it's | |
323 | * clear that the old packet has been | |
324 | * dropped. This is the new "fast" path mtu | |
325 | * discovery. | |
326 | */ | |
327 | tcp_simple_retransmit(sk); | |
328 | } /* else let the usual retransmit timer handle it */ | |
329 | } | |
330 | ||
331 | /* | |
332 | * This routine is called by the ICMP module when it gets some | |
333 | * sort of error condition. If err < 0 then the socket should | |
334 | * be closed and the error returned to the user. If err > 0 | |
335 | * it's just the icmp type << 8 | icmp code. After adjustment | |
336 | * header points to the first 8 bytes of the tcp header. We need | |
337 | * to find the appropriate port. | |
338 | * | |
339 | * The locking strategy used here is very "optimistic". When | |
340 | * someone else accesses the socket the ICMP is just dropped | |
341 | * and for some paths there is no check at all. | |
342 | * A more general error queue to queue errors for later handling | |
343 | * is probably better. | |
344 | * | |
345 | */ | |
346 | ||
4d1a2d9e | 347 | void tcp_v4_err(struct sk_buff *icmp_skb, u32 info) |
1da177e4 | 348 | { |
b71d1d42 | 349 | const struct iphdr *iph = (const struct iphdr *)icmp_skb->data; |
4d1a2d9e | 350 | struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2)); |
f1ecd5d9 | 351 | struct inet_connection_sock *icsk; |
1da177e4 LT |
352 | struct tcp_sock *tp; |
353 | struct inet_sock *inet; | |
4d1a2d9e DL |
354 | const int type = icmp_hdr(icmp_skb)->type; |
355 | const int code = icmp_hdr(icmp_skb)->code; | |
1da177e4 | 356 | struct sock *sk; |
f1ecd5d9 | 357 | struct sk_buff *skb; |
1da177e4 | 358 | __u32 seq; |
f1ecd5d9 | 359 | __u32 remaining; |
1da177e4 | 360 | int err; |
4d1a2d9e | 361 | struct net *net = dev_net(icmp_skb->dev); |
1da177e4 | 362 | |
4d1a2d9e | 363 | if (icmp_skb->len < (iph->ihl << 2) + 8) { |
dcfc23ca | 364 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
365 | return; |
366 | } | |
367 | ||
fd54d716 | 368 | sk = inet_lookup(net, &tcp_hashinfo, iph->daddr, th->dest, |
4d1a2d9e | 369 | iph->saddr, th->source, inet_iif(icmp_skb)); |
1da177e4 | 370 | if (!sk) { |
dcfc23ca | 371 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
372 | return; |
373 | } | |
374 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 375 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
376 | return; |
377 | } | |
378 | ||
379 | bh_lock_sock(sk); | |
380 | /* If too many ICMPs get dropped on busy | |
381 | * servers this needs to be solved differently. | |
382 | */ | |
383 | if (sock_owned_by_user(sk)) | |
de0744af | 384 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); |
1da177e4 LT |
385 | |
386 | if (sk->sk_state == TCP_CLOSE) | |
387 | goto out; | |
388 | ||
97e3ecd1 | 389 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
390 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
391 | goto out; | |
392 | } | |
393 | ||
f1ecd5d9 | 394 | icsk = inet_csk(sk); |
1da177e4 LT |
395 | tp = tcp_sk(sk); |
396 | seq = ntohl(th->seq); | |
397 | if (sk->sk_state != TCP_LISTEN && | |
398 | !between(seq, tp->snd_una, tp->snd_nxt)) { | |
de0744af | 399 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
400 | goto out; |
401 | } | |
402 | ||
403 | switch (type) { | |
404 | case ICMP_SOURCE_QUENCH: | |
405 | /* Just silently ignore these. */ | |
406 | goto out; | |
407 | case ICMP_PARAMETERPROB: | |
408 | err = EPROTO; | |
409 | break; | |
410 | case ICMP_DEST_UNREACH: | |
411 | if (code > NR_ICMP_UNREACH) | |
412 | goto out; | |
413 | ||
414 | if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */ | |
415 | if (!sock_owned_by_user(sk)) | |
416 | do_pmtu_discovery(sk, iph, info); | |
417 | goto out; | |
418 | } | |
419 | ||
420 | err = icmp_err_convert[code].errno; | |
f1ecd5d9 DL |
421 | /* check if icmp_skb allows revert of backoff |
422 | * (see draft-zimmermann-tcp-lcd) */ | |
423 | if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH) | |
424 | break; | |
425 | if (seq != tp->snd_una || !icsk->icsk_retransmits || | |
426 | !icsk->icsk_backoff) | |
427 | break; | |
428 | ||
8f49c270 DM |
429 | if (sock_owned_by_user(sk)) |
430 | break; | |
431 | ||
f1ecd5d9 | 432 | icsk->icsk_backoff--; |
9ad7c049 JC |
433 | inet_csk(sk)->icsk_rto = (tp->srtt ? __tcp_set_rto(tp) : |
434 | TCP_TIMEOUT_INIT) << icsk->icsk_backoff; | |
f1ecd5d9 DL |
435 | tcp_bound_rto(sk); |
436 | ||
437 | skb = tcp_write_queue_head(sk); | |
438 | BUG_ON(!skb); | |
439 | ||
440 | remaining = icsk->icsk_rto - min(icsk->icsk_rto, | |
441 | tcp_time_stamp - TCP_SKB_CB(skb)->when); | |
442 | ||
443 | if (remaining) { | |
444 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, | |
445 | remaining, TCP_RTO_MAX); | |
f1ecd5d9 DL |
446 | } else { |
447 | /* RTO revert clocked out retransmission. | |
448 | * Will retransmit now */ | |
449 | tcp_retransmit_timer(sk); | |
450 | } | |
451 | ||
1da177e4 LT |
452 | break; |
453 | case ICMP_TIME_EXCEEDED: | |
454 | err = EHOSTUNREACH; | |
455 | break; | |
456 | default: | |
457 | goto out; | |
458 | } | |
459 | ||
460 | switch (sk->sk_state) { | |
60236fdd | 461 | struct request_sock *req, **prev; |
1da177e4 LT |
462 | case TCP_LISTEN: |
463 | if (sock_owned_by_user(sk)) | |
464 | goto out; | |
465 | ||
463c84b9 ACM |
466 | req = inet_csk_search_req(sk, &prev, th->dest, |
467 | iph->daddr, iph->saddr); | |
1da177e4 LT |
468 | if (!req) |
469 | goto out; | |
470 | ||
471 | /* ICMPs are not backlogged, hence we cannot get | |
472 | an established socket here. | |
473 | */ | |
547b792c | 474 | WARN_ON(req->sk); |
1da177e4 | 475 | |
2e6599cb | 476 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 477 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
478 | goto out; |
479 | } | |
480 | ||
481 | /* | |
482 | * Still in SYN_RECV, just remove it silently. | |
483 | * There is no good way to pass the error to the newly | |
484 | * created socket, and POSIX does not want network | |
485 | * errors returned from accept(). | |
486 | */ | |
463c84b9 | 487 | inet_csk_reqsk_queue_drop(sk, req, prev); |
1da177e4 LT |
488 | goto out; |
489 | ||
490 | case TCP_SYN_SENT: | |
491 | case TCP_SYN_RECV: /* Cannot happen. | |
492 | It can f.e. if SYNs crossed. | |
493 | */ | |
494 | if (!sock_owned_by_user(sk)) { | |
1da177e4 LT |
495 | sk->sk_err = err; |
496 | ||
497 | sk->sk_error_report(sk); | |
498 | ||
499 | tcp_done(sk); | |
500 | } else { | |
501 | sk->sk_err_soft = err; | |
502 | } | |
503 | goto out; | |
504 | } | |
505 | ||
506 | /* If we've already connected we will keep trying | |
507 | * until we time out, or the user gives up. | |
508 | * | |
509 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
510 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
511 | * but it is obsoleted by pmtu discovery). | |
512 | * | |
513 | * Note, that in modern internet, where routing is unreliable | |
514 | * and in each dark corner broken firewalls sit, sending random | |
515 | * errors ordered by their masters even this two messages finally lose | |
516 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
517 | * | |
518 | * Now we are in compliance with RFCs. | |
519 | * --ANK (980905) | |
520 | */ | |
521 | ||
522 | inet = inet_sk(sk); | |
523 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
524 | sk->sk_err = err; | |
525 | sk->sk_error_report(sk); | |
526 | } else { /* Only an error on timeout */ | |
527 | sk->sk_err_soft = err; | |
528 | } | |
529 | ||
530 | out: | |
531 | bh_unlock_sock(sk); | |
532 | sock_put(sk); | |
533 | } | |
534 | ||
419f9f89 HX |
535 | static void __tcp_v4_send_check(struct sk_buff *skb, |
536 | __be32 saddr, __be32 daddr) | |
1da177e4 | 537 | { |
aa8223c7 | 538 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 539 | |
84fa7933 | 540 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
419f9f89 | 541 | th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0); |
663ead3b | 542 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 543 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 544 | } else { |
419f9f89 | 545 | th->check = tcp_v4_check(skb->len, saddr, daddr, |
07f0757a | 546 | csum_partial(th, |
1da177e4 LT |
547 | th->doff << 2, |
548 | skb->csum)); | |
549 | } | |
550 | } | |
551 | ||
419f9f89 | 552 | /* This routine computes an IPv4 TCP checksum. */ |
bb296246 | 553 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb) |
419f9f89 HX |
554 | { |
555 | struct inet_sock *inet = inet_sk(sk); | |
556 | ||
557 | __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr); | |
558 | } | |
4bc2f18b | 559 | EXPORT_SYMBOL(tcp_v4_send_check); |
419f9f89 | 560 | |
a430a43d HX |
561 | int tcp_v4_gso_send_check(struct sk_buff *skb) |
562 | { | |
eddc9ec5 | 563 | const struct iphdr *iph; |
a430a43d HX |
564 | struct tcphdr *th; |
565 | ||
566 | if (!pskb_may_pull(skb, sizeof(*th))) | |
567 | return -EINVAL; | |
568 | ||
eddc9ec5 | 569 | iph = ip_hdr(skb); |
aa8223c7 | 570 | th = tcp_hdr(skb); |
a430a43d HX |
571 | |
572 | th->check = 0; | |
84fa7933 | 573 | skb->ip_summed = CHECKSUM_PARTIAL; |
419f9f89 | 574 | __tcp_v4_send_check(skb, iph->saddr, iph->daddr); |
a430a43d HX |
575 | return 0; |
576 | } | |
577 | ||
1da177e4 LT |
578 | /* |
579 | * This routine will send an RST to the other tcp. | |
580 | * | |
581 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
582 | * for reset. | |
583 | * Answer: if a packet caused RST, it is not for a socket | |
584 | * existing in our system, if it is matched to a socket, | |
585 | * it is just duplicate segment or bug in other side's TCP. | |
586 | * So that we build reply only basing on parameters | |
587 | * arrived with segment. | |
588 | * Exception: precedence violation. We do not implement it in any case. | |
589 | */ | |
590 | ||
cfb6eeb4 | 591 | static void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 592 | { |
aa8223c7 | 593 | struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
594 | struct { |
595 | struct tcphdr th; | |
596 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 597 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
598 | #endif |
599 | } rep; | |
1da177e4 | 600 | struct ip_reply_arg arg; |
cfb6eeb4 YH |
601 | #ifdef CONFIG_TCP_MD5SIG |
602 | struct tcp_md5sig_key *key; | |
603 | #endif | |
a86b1e30 | 604 | struct net *net; |
1da177e4 LT |
605 | |
606 | /* Never send a reset in response to a reset. */ | |
607 | if (th->rst) | |
608 | return; | |
609 | ||
511c3f92 | 610 | if (skb_rtable(skb)->rt_type != RTN_LOCAL) |
1da177e4 LT |
611 | return; |
612 | ||
613 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
614 | memset(&rep, 0, sizeof(rep)); |
615 | rep.th.dest = th->source; | |
616 | rep.th.source = th->dest; | |
617 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
618 | rep.th.rst = 1; | |
1da177e4 LT |
619 | |
620 | if (th->ack) { | |
cfb6eeb4 | 621 | rep.th.seq = th->ack_seq; |
1da177e4 | 622 | } else { |
cfb6eeb4 YH |
623 | rep.th.ack = 1; |
624 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
625 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
626 | } |
627 | ||
7174259e | 628 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
629 | arg.iov[0].iov_base = (unsigned char *)&rep; |
630 | arg.iov[0].iov_len = sizeof(rep.th); | |
631 | ||
632 | #ifdef CONFIG_TCP_MD5SIG | |
eddc9ec5 | 633 | key = sk ? tcp_v4_md5_do_lookup(sk, ip_hdr(skb)->daddr) : NULL; |
cfb6eeb4 YH |
634 | if (key) { |
635 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
636 | (TCPOPT_NOP << 16) | | |
637 | (TCPOPT_MD5SIG << 8) | | |
638 | TCPOLEN_MD5SIG); | |
639 | /* Update length and the length the header thinks exists */ | |
640 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
641 | rep.th.doff = arg.iov[0].iov_len / 4; | |
642 | ||
49a72dfb | 643 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
644 | key, ip_hdr(skb)->saddr, |
645 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
646 | } |
647 | #endif | |
eddc9ec5 ACM |
648 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
649 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 650 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 651 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
88ef4a5a | 652 | arg.flags = (sk && inet_sk(sk)->transparent) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
1da177e4 | 653 | |
adf30907 | 654 | net = dev_net(skb_dst(skb)->dev); |
0a5ebb80 | 655 | ip_send_reply(net->ipv4.tcp_sock, skb, ip_hdr(skb)->saddr, |
7feb49c8 | 656 | &arg, arg.iov[0].iov_len); |
1da177e4 | 657 | |
63231bdd PE |
658 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
659 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
1da177e4 LT |
660 | } |
661 | ||
662 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
663 | outside socket context is ugly, certainly. What can I do? | |
664 | */ | |
665 | ||
9501f972 YH |
666 | static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack, |
667 | u32 win, u32 ts, int oif, | |
88ef4a5a KK |
668 | struct tcp_md5sig_key *key, |
669 | int reply_flags) | |
1da177e4 | 670 | { |
aa8223c7 | 671 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
672 | struct { |
673 | struct tcphdr th; | |
714e85be | 674 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 675 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 676 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
677 | #endif |
678 | ]; | |
1da177e4 LT |
679 | } rep; |
680 | struct ip_reply_arg arg; | |
adf30907 | 681 | struct net *net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
682 | |
683 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 684 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
685 | |
686 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
687 | arg.iov[0].iov_len = sizeof(rep.th); | |
688 | if (ts) { | |
cfb6eeb4 YH |
689 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
690 | (TCPOPT_TIMESTAMP << 8) | | |
691 | TCPOLEN_TIMESTAMP); | |
692 | rep.opt[1] = htonl(tcp_time_stamp); | |
693 | rep.opt[2] = htonl(ts); | |
cb48cfe8 | 694 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
695 | } |
696 | ||
697 | /* Swap the send and the receive. */ | |
698 | rep.th.dest = th->source; | |
699 | rep.th.source = th->dest; | |
700 | rep.th.doff = arg.iov[0].iov_len / 4; | |
701 | rep.th.seq = htonl(seq); | |
702 | rep.th.ack_seq = htonl(ack); | |
703 | rep.th.ack = 1; | |
704 | rep.th.window = htons(win); | |
705 | ||
cfb6eeb4 | 706 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
707 | if (key) { |
708 | int offset = (ts) ? 3 : 0; | |
709 | ||
710 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
711 | (TCPOPT_NOP << 16) | | |
712 | (TCPOPT_MD5SIG << 8) | | |
713 | TCPOLEN_MD5SIG); | |
714 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
715 | rep.th.doff = arg.iov[0].iov_len/4; | |
716 | ||
49a72dfb | 717 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
718 | key, ip_hdr(skb)->saddr, |
719 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
720 | } |
721 | #endif | |
88ef4a5a | 722 | arg.flags = reply_flags; |
eddc9ec5 ACM |
723 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
724 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
725 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
726 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
727 | if (oif) |
728 | arg.bound_dev_if = oif; | |
1da177e4 | 729 | |
0a5ebb80 | 730 | ip_send_reply(net->ipv4.tcp_sock, skb, ip_hdr(skb)->saddr, |
7feb49c8 | 731 | &arg, arg.iov[0].iov_len); |
1da177e4 | 732 | |
63231bdd | 733 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
1da177e4 LT |
734 | } |
735 | ||
736 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
737 | { | |
8feaf0c0 | 738 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 739 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 740 | |
9501f972 | 741 | tcp_v4_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 742 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
9501f972 YH |
743 | tcptw->tw_ts_recent, |
744 | tw->tw_bound_dev_if, | |
88ef4a5a KK |
745 | tcp_twsk_md5_key(tcptw), |
746 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0 | |
9501f972 | 747 | ); |
1da177e4 | 748 | |
8feaf0c0 | 749 | inet_twsk_put(tw); |
1da177e4 LT |
750 | } |
751 | ||
6edafaaf | 752 | static void tcp_v4_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
7174259e | 753 | struct request_sock *req) |
1da177e4 | 754 | { |
9501f972 | 755 | tcp_v4_send_ack(skb, tcp_rsk(req)->snt_isn + 1, |
cfb6eeb4 | 756 | tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, |
9501f972 YH |
757 | req->ts_recent, |
758 | 0, | |
88ef4a5a KK |
759 | tcp_v4_md5_do_lookup(sk, ip_hdr(skb)->daddr), |
760 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0); | |
1da177e4 LT |
761 | } |
762 | ||
1da177e4 | 763 | /* |
9bf1d83e | 764 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 765 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
766 | * socket. |
767 | */ | |
72659ecc OP |
768 | static int tcp_v4_send_synack(struct sock *sk, struct dst_entry *dst, |
769 | struct request_sock *req, | |
770 | struct request_values *rvp) | |
1da177e4 | 771 | { |
2e6599cb | 772 | const struct inet_request_sock *ireq = inet_rsk(req); |
6bd023f3 | 773 | struct flowi4 fl4; |
1da177e4 LT |
774 | int err = -1; |
775 | struct sk_buff * skb; | |
776 | ||
777 | /* First, grab a route. */ | |
6bd023f3 | 778 | if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL) |
fd80eb94 | 779 | return -1; |
1da177e4 | 780 | |
e6b4d113 | 781 | skb = tcp_make_synack(sk, dst, req, rvp); |
1da177e4 LT |
782 | |
783 | if (skb) { | |
419f9f89 | 784 | __tcp_v4_send_check(skb, ireq->loc_addr, ireq->rmt_addr); |
1da177e4 | 785 | |
2e6599cb ACM |
786 | err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr, |
787 | ireq->rmt_addr, | |
788 | ireq->opt); | |
b9df3cb8 | 789 | err = net_xmit_eval(err); |
1da177e4 LT |
790 | } |
791 | ||
1da177e4 LT |
792 | dst_release(dst); |
793 | return err; | |
794 | } | |
795 | ||
72659ecc | 796 | static int tcp_v4_rtx_synack(struct sock *sk, struct request_sock *req, |
e6b4d113 | 797 | struct request_values *rvp) |
fd80eb94 | 798 | { |
72659ecc OP |
799 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS); |
800 | return tcp_v4_send_synack(sk, NULL, req, rvp); | |
fd80eb94 DL |
801 | } |
802 | ||
1da177e4 | 803 | /* |
60236fdd | 804 | * IPv4 request_sock destructor. |
1da177e4 | 805 | */ |
60236fdd | 806 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 807 | { |
a51482bd | 808 | kfree(inet_rsk(req)->opt); |
1da177e4 LT |
809 | } |
810 | ||
2a1d4bd4 | 811 | static void syn_flood_warning(const struct sk_buff *skb) |
1da177e4 | 812 | { |
2a1d4bd4 | 813 | const char *msg; |
1da177e4 | 814 | |
2a1d4bd4 FW |
815 | #ifdef CONFIG_SYN_COOKIES |
816 | if (sysctl_tcp_syncookies) | |
817 | msg = "Sending cookies"; | |
818 | else | |
80e40daa | 819 | #endif |
2a1d4bd4 FW |
820 | msg = "Dropping request"; |
821 | ||
822 | pr_info("TCP: Possible SYN flooding on port %d. %s.\n", | |
823 | ntohs(tcp_hdr(skb)->dest), msg); | |
824 | } | |
1da177e4 LT |
825 | |
826 | /* | |
60236fdd | 827 | * Save and compile IPv4 options into the request_sock if needed. |
1da177e4 | 828 | */ |
f6d8bd05 ED |
829 | static struct ip_options_rcu *tcp_v4_save_options(struct sock *sk, |
830 | struct sk_buff *skb) | |
1da177e4 | 831 | { |
f6d8bd05 ED |
832 | const struct ip_options *opt = &(IPCB(skb)->opt); |
833 | struct ip_options_rcu *dopt = NULL; | |
1da177e4 LT |
834 | |
835 | if (opt && opt->optlen) { | |
f6d8bd05 ED |
836 | int opt_size = sizeof(*dopt) + opt->optlen; |
837 | ||
1da177e4 LT |
838 | dopt = kmalloc(opt_size, GFP_ATOMIC); |
839 | if (dopt) { | |
f6d8bd05 | 840 | if (ip_options_echo(&dopt->opt, skb)) { |
1da177e4 LT |
841 | kfree(dopt); |
842 | dopt = NULL; | |
843 | } | |
844 | } | |
845 | } | |
846 | return dopt; | |
847 | } | |
848 | ||
cfb6eeb4 YH |
849 | #ifdef CONFIG_TCP_MD5SIG |
850 | /* | |
851 | * RFC2385 MD5 checksumming requires a mapping of | |
852 | * IP address->MD5 Key. | |
853 | * We need to maintain these in the sk structure. | |
854 | */ | |
855 | ||
856 | /* Find the Key structure for an address. */ | |
7174259e ACM |
857 | static struct tcp_md5sig_key * |
858 | tcp_v4_md5_do_lookup(struct sock *sk, __be32 addr) | |
cfb6eeb4 YH |
859 | { |
860 | struct tcp_sock *tp = tcp_sk(sk); | |
861 | int i; | |
862 | ||
863 | if (!tp->md5sig_info || !tp->md5sig_info->entries4) | |
864 | return NULL; | |
865 | for (i = 0; i < tp->md5sig_info->entries4; i++) { | |
866 | if (tp->md5sig_info->keys4[i].addr == addr) | |
f8ab18d2 | 867 | return &tp->md5sig_info->keys4[i].base; |
cfb6eeb4 YH |
868 | } |
869 | return NULL; | |
870 | } | |
871 | ||
872 | struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
873 | struct sock *addr_sk) | |
874 | { | |
c720c7e8 | 875 | return tcp_v4_md5_do_lookup(sk, inet_sk(addr_sk)->inet_daddr); |
cfb6eeb4 | 876 | } |
cfb6eeb4 YH |
877 | EXPORT_SYMBOL(tcp_v4_md5_lookup); |
878 | ||
f5b99bcd AB |
879 | static struct tcp_md5sig_key *tcp_v4_reqsk_md5_lookup(struct sock *sk, |
880 | struct request_sock *req) | |
cfb6eeb4 YH |
881 | { |
882 | return tcp_v4_md5_do_lookup(sk, inet_rsk(req)->rmt_addr); | |
883 | } | |
884 | ||
885 | /* This can be called on a newly created socket, from other files */ | |
886 | int tcp_v4_md5_do_add(struct sock *sk, __be32 addr, | |
887 | u8 *newkey, u8 newkeylen) | |
888 | { | |
889 | /* Add Key to the list */ | |
b0a713e9 | 890 | struct tcp_md5sig_key *key; |
cfb6eeb4 YH |
891 | struct tcp_sock *tp = tcp_sk(sk); |
892 | struct tcp4_md5sig_key *keys; | |
893 | ||
b0a713e9 | 894 | key = tcp_v4_md5_do_lookup(sk, addr); |
cfb6eeb4 YH |
895 | if (key) { |
896 | /* Pre-existing entry - just update that one. */ | |
b0a713e9 MD |
897 | kfree(key->key); |
898 | key->key = newkey; | |
899 | key->keylen = newkeylen; | |
cfb6eeb4 | 900 | } else { |
f6685938 ACM |
901 | struct tcp_md5sig_info *md5sig; |
902 | ||
cfb6eeb4 | 903 | if (!tp->md5sig_info) { |
f6685938 ACM |
904 | tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), |
905 | GFP_ATOMIC); | |
cfb6eeb4 YH |
906 | if (!tp->md5sig_info) { |
907 | kfree(newkey); | |
908 | return -ENOMEM; | |
909 | } | |
a465419b | 910 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
cfb6eeb4 | 911 | } |
aa133076 | 912 | if (tcp_alloc_md5sig_pool(sk) == NULL) { |
cfb6eeb4 YH |
913 | kfree(newkey); |
914 | return -ENOMEM; | |
915 | } | |
f6685938 ACM |
916 | md5sig = tp->md5sig_info; |
917 | ||
918 | if (md5sig->alloced4 == md5sig->entries4) { | |
919 | keys = kmalloc((sizeof(*keys) * | |
e905a9ed | 920 | (md5sig->entries4 + 1)), GFP_ATOMIC); |
cfb6eeb4 YH |
921 | if (!keys) { |
922 | kfree(newkey); | |
923 | tcp_free_md5sig_pool(); | |
924 | return -ENOMEM; | |
925 | } | |
926 | ||
f6685938 ACM |
927 | if (md5sig->entries4) |
928 | memcpy(keys, md5sig->keys4, | |
929 | sizeof(*keys) * md5sig->entries4); | |
cfb6eeb4 YH |
930 | |
931 | /* Free old key list, and reference new one */ | |
a80cc20d | 932 | kfree(md5sig->keys4); |
f6685938 ACM |
933 | md5sig->keys4 = keys; |
934 | md5sig->alloced4++; | |
cfb6eeb4 | 935 | } |
f6685938 | 936 | md5sig->entries4++; |
f8ab18d2 DM |
937 | md5sig->keys4[md5sig->entries4 - 1].addr = addr; |
938 | md5sig->keys4[md5sig->entries4 - 1].base.key = newkey; | |
939 | md5sig->keys4[md5sig->entries4 - 1].base.keylen = newkeylen; | |
cfb6eeb4 YH |
940 | } |
941 | return 0; | |
942 | } | |
cfb6eeb4 YH |
943 | EXPORT_SYMBOL(tcp_v4_md5_do_add); |
944 | ||
945 | static int tcp_v4_md5_add_func(struct sock *sk, struct sock *addr_sk, | |
946 | u8 *newkey, u8 newkeylen) | |
947 | { | |
c720c7e8 | 948 | return tcp_v4_md5_do_add(sk, inet_sk(addr_sk)->inet_daddr, |
cfb6eeb4 YH |
949 | newkey, newkeylen); |
950 | } | |
951 | ||
952 | int tcp_v4_md5_do_del(struct sock *sk, __be32 addr) | |
953 | { | |
954 | struct tcp_sock *tp = tcp_sk(sk); | |
955 | int i; | |
956 | ||
957 | for (i = 0; i < tp->md5sig_info->entries4; i++) { | |
958 | if (tp->md5sig_info->keys4[i].addr == addr) { | |
959 | /* Free the key */ | |
f8ab18d2 | 960 | kfree(tp->md5sig_info->keys4[i].base.key); |
cfb6eeb4 YH |
961 | tp->md5sig_info->entries4--; |
962 | ||
963 | if (tp->md5sig_info->entries4 == 0) { | |
964 | kfree(tp->md5sig_info->keys4); | |
965 | tp->md5sig_info->keys4 = NULL; | |
8228a18d | 966 | tp->md5sig_info->alloced4 = 0; |
7174259e | 967 | } else if (tp->md5sig_info->entries4 != i) { |
cfb6eeb4 | 968 | /* Need to do some manipulation */ |
354faf09 YH |
969 | memmove(&tp->md5sig_info->keys4[i], |
970 | &tp->md5sig_info->keys4[i+1], | |
971 | (tp->md5sig_info->entries4 - i) * | |
972 | sizeof(struct tcp4_md5sig_key)); | |
cfb6eeb4 YH |
973 | } |
974 | tcp_free_md5sig_pool(); | |
975 | return 0; | |
976 | } | |
977 | } | |
978 | return -ENOENT; | |
979 | } | |
cfb6eeb4 YH |
980 | EXPORT_SYMBOL(tcp_v4_md5_do_del); |
981 | ||
7174259e | 982 | static void tcp_v4_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
983 | { |
984 | struct tcp_sock *tp = tcp_sk(sk); | |
985 | ||
986 | /* Free each key, then the set of key keys, | |
987 | * the crypto element, and then decrement our | |
988 | * hold on the last resort crypto. | |
989 | */ | |
990 | if (tp->md5sig_info->entries4) { | |
991 | int i; | |
992 | for (i = 0; i < tp->md5sig_info->entries4; i++) | |
f8ab18d2 | 993 | kfree(tp->md5sig_info->keys4[i].base.key); |
cfb6eeb4 YH |
994 | tp->md5sig_info->entries4 = 0; |
995 | tcp_free_md5sig_pool(); | |
996 | } | |
997 | if (tp->md5sig_info->keys4) { | |
998 | kfree(tp->md5sig_info->keys4); | |
999 | tp->md5sig_info->keys4 = NULL; | |
1000 | tp->md5sig_info->alloced4 = 0; | |
1001 | } | |
1002 | } | |
1003 | ||
7174259e ACM |
1004 | static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval, |
1005 | int optlen) | |
cfb6eeb4 YH |
1006 | { |
1007 | struct tcp_md5sig cmd; | |
1008 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
1009 | u8 *newkey; | |
1010 | ||
1011 | if (optlen < sizeof(cmd)) | |
1012 | return -EINVAL; | |
1013 | ||
7174259e | 1014 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
1015 | return -EFAULT; |
1016 | ||
1017 | if (sin->sin_family != AF_INET) | |
1018 | return -EINVAL; | |
1019 | ||
1020 | if (!cmd.tcpm_key || !cmd.tcpm_keylen) { | |
1021 | if (!tcp_sk(sk)->md5sig_info) | |
1022 | return -ENOENT; | |
1023 | return tcp_v4_md5_do_del(sk, sin->sin_addr.s_addr); | |
1024 | } | |
1025 | ||
1026 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
1027 | return -EINVAL; | |
1028 | ||
1029 | if (!tcp_sk(sk)->md5sig_info) { | |
1030 | struct tcp_sock *tp = tcp_sk(sk); | |
aa133076 | 1031 | struct tcp_md5sig_info *p; |
cfb6eeb4 | 1032 | |
aa133076 | 1033 | p = kzalloc(sizeof(*p), sk->sk_allocation); |
cfb6eeb4 YH |
1034 | if (!p) |
1035 | return -EINVAL; | |
1036 | ||
1037 | tp->md5sig_info = p; | |
a465419b | 1038 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1039 | } |
1040 | ||
aa133076 | 1041 | newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, sk->sk_allocation); |
cfb6eeb4 YH |
1042 | if (!newkey) |
1043 | return -ENOMEM; | |
cfb6eeb4 YH |
1044 | return tcp_v4_md5_do_add(sk, sin->sin_addr.s_addr, |
1045 | newkey, cmd.tcpm_keylen); | |
1046 | } | |
1047 | ||
49a72dfb AL |
1048 | static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
1049 | __be32 daddr, __be32 saddr, int nbytes) | |
cfb6eeb4 | 1050 | { |
cfb6eeb4 | 1051 | struct tcp4_pseudohdr *bp; |
49a72dfb | 1052 | struct scatterlist sg; |
cfb6eeb4 YH |
1053 | |
1054 | bp = &hp->md5_blk.ip4; | |
cfb6eeb4 YH |
1055 | |
1056 | /* | |
49a72dfb | 1057 | * 1. the TCP pseudo-header (in the order: source IP address, |
cfb6eeb4 YH |
1058 | * destination IP address, zero-padded protocol number, and |
1059 | * segment length) | |
1060 | */ | |
1061 | bp->saddr = saddr; | |
1062 | bp->daddr = daddr; | |
1063 | bp->pad = 0; | |
076fb722 | 1064 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1065 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1066 | |
49a72dfb AL |
1067 | sg_init_one(&sg, bp, sizeof(*bp)); |
1068 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
1069 | } | |
1070 | ||
1071 | static int tcp_v4_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, | |
1072 | __be32 daddr, __be32 saddr, struct tcphdr *th) | |
1073 | { | |
1074 | struct tcp_md5sig_pool *hp; | |
1075 | struct hash_desc *desc; | |
1076 | ||
1077 | hp = tcp_get_md5sig_pool(); | |
1078 | if (!hp) | |
1079 | goto clear_hash_noput; | |
1080 | desc = &hp->md5_desc; | |
1081 | ||
1082 | if (crypto_hash_init(desc)) | |
1083 | goto clear_hash; | |
1084 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
1085 | goto clear_hash; | |
1086 | if (tcp_md5_hash_header(hp, th)) | |
1087 | goto clear_hash; | |
1088 | if (tcp_md5_hash_key(hp, key)) | |
1089 | goto clear_hash; | |
1090 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 YH |
1091 | goto clear_hash; |
1092 | ||
cfb6eeb4 | 1093 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1094 | return 0; |
49a72dfb | 1095 | |
cfb6eeb4 YH |
1096 | clear_hash: |
1097 | tcp_put_md5sig_pool(); | |
1098 | clear_hash_noput: | |
1099 | memset(md5_hash, 0, 16); | |
49a72dfb | 1100 | return 1; |
cfb6eeb4 YH |
1101 | } |
1102 | ||
49a72dfb AL |
1103 | int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
1104 | struct sock *sk, struct request_sock *req, | |
1105 | struct sk_buff *skb) | |
cfb6eeb4 | 1106 | { |
49a72dfb AL |
1107 | struct tcp_md5sig_pool *hp; |
1108 | struct hash_desc *desc; | |
1109 | struct tcphdr *th = tcp_hdr(skb); | |
cfb6eeb4 YH |
1110 | __be32 saddr, daddr; |
1111 | ||
1112 | if (sk) { | |
c720c7e8 ED |
1113 | saddr = inet_sk(sk)->inet_saddr; |
1114 | daddr = inet_sk(sk)->inet_daddr; | |
49a72dfb AL |
1115 | } else if (req) { |
1116 | saddr = inet_rsk(req)->loc_addr; | |
1117 | daddr = inet_rsk(req)->rmt_addr; | |
cfb6eeb4 | 1118 | } else { |
49a72dfb AL |
1119 | const struct iphdr *iph = ip_hdr(skb); |
1120 | saddr = iph->saddr; | |
1121 | daddr = iph->daddr; | |
cfb6eeb4 | 1122 | } |
49a72dfb AL |
1123 | |
1124 | hp = tcp_get_md5sig_pool(); | |
1125 | if (!hp) | |
1126 | goto clear_hash_noput; | |
1127 | desc = &hp->md5_desc; | |
1128 | ||
1129 | if (crypto_hash_init(desc)) | |
1130 | goto clear_hash; | |
1131 | ||
1132 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
1133 | goto clear_hash; | |
1134 | if (tcp_md5_hash_header(hp, th)) | |
1135 | goto clear_hash; | |
1136 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1137 | goto clear_hash; | |
1138 | if (tcp_md5_hash_key(hp, key)) | |
1139 | goto clear_hash; | |
1140 | if (crypto_hash_final(desc, md5_hash)) | |
1141 | goto clear_hash; | |
1142 | ||
1143 | tcp_put_md5sig_pool(); | |
1144 | return 0; | |
1145 | ||
1146 | clear_hash: | |
1147 | tcp_put_md5sig_pool(); | |
1148 | clear_hash_noput: | |
1149 | memset(md5_hash, 0, 16); | |
1150 | return 1; | |
cfb6eeb4 | 1151 | } |
49a72dfb | 1152 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1153 | |
7174259e | 1154 | static int tcp_v4_inbound_md5_hash(struct sock *sk, struct sk_buff *skb) |
cfb6eeb4 YH |
1155 | { |
1156 | /* | |
1157 | * This gets called for each TCP segment that arrives | |
1158 | * so we want to be efficient. | |
1159 | * We have 3 drop cases: | |
1160 | * o No MD5 hash and one expected. | |
1161 | * o MD5 hash and we're not expecting one. | |
1162 | * o MD5 hash and its wrong. | |
1163 | */ | |
1164 | __u8 *hash_location = NULL; | |
1165 | struct tcp_md5sig_key *hash_expected; | |
eddc9ec5 | 1166 | const struct iphdr *iph = ip_hdr(skb); |
aa8223c7 | 1167 | struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1168 | int genhash; |
cfb6eeb4 YH |
1169 | unsigned char newhash[16]; |
1170 | ||
1171 | hash_expected = tcp_v4_md5_do_lookup(sk, iph->saddr); | |
7d5d5525 | 1172 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1173 | |
cfb6eeb4 YH |
1174 | /* We've parsed the options - do we have a hash? */ |
1175 | if (!hash_expected && !hash_location) | |
1176 | return 0; | |
1177 | ||
1178 | if (hash_expected && !hash_location) { | |
785957d3 | 1179 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
cfb6eeb4 YH |
1180 | return 1; |
1181 | } | |
1182 | ||
1183 | if (!hash_expected && hash_location) { | |
785957d3 | 1184 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
cfb6eeb4 YH |
1185 | return 1; |
1186 | } | |
1187 | ||
1188 | /* Okay, so this is hash_expected and hash_location - | |
1189 | * so we need to calculate the checksum. | |
1190 | */ | |
49a72dfb AL |
1191 | genhash = tcp_v4_md5_hash_skb(newhash, |
1192 | hash_expected, | |
1193 | NULL, NULL, skb); | |
cfb6eeb4 YH |
1194 | |
1195 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
1196 | if (net_ratelimit()) { | |
673d57e7 HH |
1197 | printk(KERN_INFO "MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1198 | &iph->saddr, ntohs(th->source), | |
1199 | &iph->daddr, ntohs(th->dest), | |
cfb6eeb4 | 1200 | genhash ? " tcp_v4_calc_md5_hash failed" : ""); |
cfb6eeb4 YH |
1201 | } |
1202 | return 1; | |
1203 | } | |
1204 | return 0; | |
1205 | } | |
1206 | ||
1207 | #endif | |
1208 | ||
72a3effa | 1209 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1210 | .family = PF_INET, |
2e6599cb | 1211 | .obj_size = sizeof(struct tcp_request_sock), |
72659ecc | 1212 | .rtx_syn_ack = tcp_v4_rtx_synack, |
60236fdd ACM |
1213 | .send_ack = tcp_v4_reqsk_send_ack, |
1214 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 | 1215 | .send_reset = tcp_v4_send_reset, |
72659ecc | 1216 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
1217 | }; |
1218 | ||
cfb6eeb4 | 1219 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1220 | static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
cfb6eeb4 | 1221 | .md5_lookup = tcp_v4_reqsk_md5_lookup, |
e3afe7b7 | 1222 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1223 | }; |
b6332e6c | 1224 | #endif |
cfb6eeb4 | 1225 | |
1da177e4 LT |
1226 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1227 | { | |
4957faad | 1228 | struct tcp_extend_values tmp_ext; |
1da177e4 | 1229 | struct tcp_options_received tmp_opt; |
4957faad | 1230 | u8 *hash_location; |
60236fdd | 1231 | struct request_sock *req; |
e6b4d113 | 1232 | struct inet_request_sock *ireq; |
4957faad | 1233 | struct tcp_sock *tp = tcp_sk(sk); |
e6b4d113 | 1234 | struct dst_entry *dst = NULL; |
eddc9ec5 ACM |
1235 | __be32 saddr = ip_hdr(skb)->saddr; |
1236 | __be32 daddr = ip_hdr(skb)->daddr; | |
1da177e4 | 1237 | __u32 isn = TCP_SKB_CB(skb)->when; |
1da177e4 LT |
1238 | #ifdef CONFIG_SYN_COOKIES |
1239 | int want_cookie = 0; | |
1240 | #else | |
1241 | #define want_cookie 0 /* Argh, why doesn't gcc optimize this :( */ | |
1242 | #endif | |
1243 | ||
1244 | /* Never answer to SYNs send to broadcast or multicast */ | |
511c3f92 | 1245 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1246 | goto drop; |
1247 | ||
1248 | /* TW buckets are converted to open requests without | |
1249 | * limitations, they conserve resources and peer is | |
1250 | * evidently real one. | |
1251 | */ | |
463c84b9 | 1252 | if (inet_csk_reqsk_queue_is_full(sk) && !isn) { |
2a1d4bd4 FW |
1253 | if (net_ratelimit()) |
1254 | syn_flood_warning(skb); | |
1da177e4 LT |
1255 | #ifdef CONFIG_SYN_COOKIES |
1256 | if (sysctl_tcp_syncookies) { | |
1257 | want_cookie = 1; | |
1258 | } else | |
1259 | #endif | |
1260 | goto drop; | |
1261 | } | |
1262 | ||
1263 | /* Accept backlog is full. If we have already queued enough | |
1264 | * of warm entries in syn queue, drop request. It is better than | |
1265 | * clogging syn queue with openreqs with exponentially increasing | |
1266 | * timeout. | |
1267 | */ | |
463c84b9 | 1268 | if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) |
1da177e4 LT |
1269 | goto drop; |
1270 | ||
ce4a7d0d | 1271 | req = inet_reqsk_alloc(&tcp_request_sock_ops); |
1da177e4 LT |
1272 | if (!req) |
1273 | goto drop; | |
1274 | ||
cfb6eeb4 YH |
1275 | #ifdef CONFIG_TCP_MD5SIG |
1276 | tcp_rsk(req)->af_specific = &tcp_request_sock_ipv4_ops; | |
1277 | #endif | |
1278 | ||
1da177e4 | 1279 | tcp_clear_options(&tmp_opt); |
bee7ca9e | 1280 | tmp_opt.mss_clamp = TCP_MSS_DEFAULT; |
4957faad | 1281 | tmp_opt.user_mss = tp->rx_opt.user_mss; |
bb5b7c11 | 1282 | tcp_parse_options(skb, &tmp_opt, &hash_location, 0); |
4957faad WAS |
1283 | |
1284 | if (tmp_opt.cookie_plus > 0 && | |
1285 | tmp_opt.saw_tstamp && | |
1286 | !tp->rx_opt.cookie_out_never && | |
1287 | (sysctl_tcp_cookie_size > 0 || | |
1288 | (tp->cookie_values != NULL && | |
1289 | tp->cookie_values->cookie_desired > 0))) { | |
1290 | u8 *c; | |
1291 | u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS]; | |
1292 | int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE; | |
1293 | ||
1294 | if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0) | |
1295 | goto drop_and_release; | |
1296 | ||
1297 | /* Secret recipe starts with IP addresses */ | |
0eae88f3 ED |
1298 | *mess++ ^= (__force u32)daddr; |
1299 | *mess++ ^= (__force u32)saddr; | |
1da177e4 | 1300 | |
4957faad WAS |
1301 | /* plus variable length Initiator Cookie */ |
1302 | c = (u8 *)mess; | |
1303 | while (l-- > 0) | |
1304 | *c++ ^= *hash_location++; | |
1305 | ||
1306 | #ifdef CONFIG_SYN_COOKIES | |
1307 | want_cookie = 0; /* not our kind of cookie */ | |
1308 | #endif | |
1309 | tmp_ext.cookie_out_never = 0; /* false */ | |
1310 | tmp_ext.cookie_plus = tmp_opt.cookie_plus; | |
1311 | } else if (!tp->rx_opt.cookie_in_always) { | |
1312 | /* redundant indications, but ensure initialization. */ | |
1313 | tmp_ext.cookie_out_never = 1; /* true */ | |
1314 | tmp_ext.cookie_plus = 0; | |
1315 | } else { | |
1316 | goto drop_and_release; | |
1317 | } | |
1318 | tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always; | |
1da177e4 | 1319 | |
4dfc2817 | 1320 | if (want_cookie && !tmp_opt.saw_tstamp) |
1da177e4 | 1321 | tcp_clear_options(&tmp_opt); |
1da177e4 | 1322 | |
1da177e4 | 1323 | tmp_opt.tstamp_ok = tmp_opt.saw_tstamp; |
1da177e4 LT |
1324 | tcp_openreq_init(req, &tmp_opt, skb); |
1325 | ||
bb5b7c11 DM |
1326 | ireq = inet_rsk(req); |
1327 | ireq->loc_addr = daddr; | |
1328 | ireq->rmt_addr = saddr; | |
1329 | ireq->no_srccheck = inet_sk(sk)->transparent; | |
1330 | ireq->opt = tcp_v4_save_options(sk, skb); | |
1331 | ||
284904aa | 1332 | if (security_inet_conn_request(sk, skb, req)) |
bb5b7c11 | 1333 | goto drop_and_free; |
284904aa | 1334 | |
172d69e6 | 1335 | if (!want_cookie || tmp_opt.tstamp_ok) |
aa8223c7 | 1336 | TCP_ECN_create_request(req, tcp_hdr(skb)); |
1da177e4 LT |
1337 | |
1338 | if (want_cookie) { | |
1da177e4 | 1339 | isn = cookie_v4_init_sequence(sk, skb, &req->mss); |
172d69e6 | 1340 | req->cookie_ts = tmp_opt.tstamp_ok; |
1da177e4 LT |
1341 | } else if (!isn) { |
1342 | struct inet_peer *peer = NULL; | |
6bd023f3 | 1343 | struct flowi4 fl4; |
1da177e4 LT |
1344 | |
1345 | /* VJ's idea. We save last timestamp seen | |
1346 | * from the destination in peer table, when entering | |
1347 | * state TIME-WAIT, and check against it before | |
1348 | * accepting new connection request. | |
1349 | * | |
1350 | * If "isn" is not zero, this request hit alive | |
1351 | * timewait bucket, so that all the necessary checks | |
1352 | * are made in the function processing timewait state. | |
1353 | */ | |
1354 | if (tmp_opt.saw_tstamp && | |
295ff7ed | 1355 | tcp_death_row.sysctl_tw_recycle && |
6bd023f3 | 1356 | (dst = inet_csk_route_req(sk, &fl4, req)) != NULL && |
ed2361e6 DM |
1357 | fl4.daddr == saddr && |
1358 | (peer = rt_get_peer((struct rtable *)dst, fl4.daddr)) != NULL) { | |
317fe0e6 | 1359 | inet_peer_refcheck(peer); |
2c1409a0 | 1360 | if ((u32)get_seconds() - peer->tcp_ts_stamp < TCP_PAWS_MSL && |
1da177e4 LT |
1361 | (s32)(peer->tcp_ts - req->ts_recent) > |
1362 | TCP_PAWS_WINDOW) { | |
de0744af | 1363 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED); |
7cd04fa7 | 1364 | goto drop_and_release; |
1da177e4 LT |
1365 | } |
1366 | } | |
1367 | /* Kill the following clause, if you dislike this way. */ | |
1368 | else if (!sysctl_tcp_syncookies && | |
463c84b9 | 1369 | (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < |
1da177e4 LT |
1370 | (sysctl_max_syn_backlog >> 2)) && |
1371 | (!peer || !peer->tcp_ts_stamp) && | |
1372 | (!dst || !dst_metric(dst, RTAX_RTT))) { | |
1373 | /* Without syncookies last quarter of | |
1374 | * backlog is filled with destinations, | |
1375 | * proven to be alive. | |
1376 | * It means that we continue to communicate | |
1377 | * to destinations, already remembered | |
1378 | * to the moment of synflood. | |
1379 | */ | |
673d57e7 HH |
1380 | LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI4/%u\n", |
1381 | &saddr, ntohs(tcp_hdr(skb)->source)); | |
7cd04fa7 | 1382 | goto drop_and_release; |
1da177e4 LT |
1383 | } |
1384 | ||
a94f723d | 1385 | isn = tcp_v4_init_sequence(skb); |
1da177e4 | 1386 | } |
2e6599cb | 1387 | tcp_rsk(req)->snt_isn = isn; |
9ad7c049 | 1388 | tcp_rsk(req)->snt_synack = tcp_time_stamp; |
1da177e4 | 1389 | |
72659ecc OP |
1390 | if (tcp_v4_send_synack(sk, dst, req, |
1391 | (struct request_values *)&tmp_ext) || | |
4957faad | 1392 | want_cookie) |
1da177e4 LT |
1393 | goto drop_and_free; |
1394 | ||
7cd04fa7 | 1395 | inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT); |
1da177e4 LT |
1396 | return 0; |
1397 | ||
7cd04fa7 DL |
1398 | drop_and_release: |
1399 | dst_release(dst); | |
1da177e4 | 1400 | drop_and_free: |
60236fdd | 1401 | reqsk_free(req); |
1da177e4 | 1402 | drop: |
1da177e4 LT |
1403 | return 0; |
1404 | } | |
4bc2f18b | 1405 | EXPORT_SYMBOL(tcp_v4_conn_request); |
1da177e4 LT |
1406 | |
1407 | ||
1408 | /* | |
1409 | * The three way handshake has completed - we got a valid synack - | |
1410 | * now create the new socket. | |
1411 | */ | |
1412 | struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1413 | struct request_sock *req, |
1da177e4 LT |
1414 | struct dst_entry *dst) |
1415 | { | |
2e6599cb | 1416 | struct inet_request_sock *ireq; |
1da177e4 LT |
1417 | struct inet_sock *newinet; |
1418 | struct tcp_sock *newtp; | |
1419 | struct sock *newsk; | |
cfb6eeb4 YH |
1420 | #ifdef CONFIG_TCP_MD5SIG |
1421 | struct tcp_md5sig_key *key; | |
1422 | #endif | |
f6d8bd05 | 1423 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
1424 | |
1425 | if (sk_acceptq_is_full(sk)) | |
1426 | goto exit_overflow; | |
1427 | ||
1da177e4 LT |
1428 | newsk = tcp_create_openreq_child(sk, req, skb); |
1429 | if (!newsk) | |
093d2823 | 1430 | goto exit_nonewsk; |
1da177e4 | 1431 | |
bcd76111 | 1432 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
1da177e4 LT |
1433 | |
1434 | newtp = tcp_sk(newsk); | |
1435 | newinet = inet_sk(newsk); | |
2e6599cb | 1436 | ireq = inet_rsk(req); |
c720c7e8 ED |
1437 | newinet->inet_daddr = ireq->rmt_addr; |
1438 | newinet->inet_rcv_saddr = ireq->loc_addr; | |
1439 | newinet->inet_saddr = ireq->loc_addr; | |
f6d8bd05 ED |
1440 | inet_opt = ireq->opt; |
1441 | rcu_assign_pointer(newinet->inet_opt, inet_opt); | |
2e6599cb | 1442 | ireq->opt = NULL; |
463c84b9 | 1443 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1444 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
d83d8461 | 1445 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
1446 | if (inet_opt) |
1447 | inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
c720c7e8 | 1448 | newinet->inet_id = newtp->write_seq ^ jiffies; |
1da177e4 | 1449 | |
0e734419 DM |
1450 | if (!dst && (dst = inet_csk_route_child_sock(sk, newsk, req)) == NULL) |
1451 | goto put_and_exit; | |
1452 | ||
1453 | sk_setup_caps(newsk, dst); | |
1454 | ||
5d424d5a | 1455 | tcp_mtup_init(newsk); |
1da177e4 | 1456 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1457 | newtp->advmss = dst_metric_advmss(dst); |
f5fff5dc TQ |
1458 | if (tcp_sk(sk)->rx_opt.user_mss && |
1459 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1460 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1461 | ||
1da177e4 | 1462 | tcp_initialize_rcv_mss(newsk); |
9ad7c049 JC |
1463 | if (tcp_rsk(req)->snt_synack) |
1464 | tcp_valid_rtt_meas(newsk, | |
1465 | tcp_time_stamp - tcp_rsk(req)->snt_synack); | |
1466 | newtp->total_retrans = req->retrans; | |
1da177e4 | 1467 | |
cfb6eeb4 YH |
1468 | #ifdef CONFIG_TCP_MD5SIG |
1469 | /* Copy over the MD5 key from the original socket */ | |
c720c7e8 ED |
1470 | key = tcp_v4_md5_do_lookup(sk, newinet->inet_daddr); |
1471 | if (key != NULL) { | |
cfb6eeb4 YH |
1472 | /* |
1473 | * We're using one, so create a matching key | |
1474 | * on the newsk structure. If we fail to get | |
1475 | * memory, then we end up not copying the key | |
1476 | * across. Shucks. | |
1477 | */ | |
f6685938 ACM |
1478 | char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC); |
1479 | if (newkey != NULL) | |
c720c7e8 | 1480 | tcp_v4_md5_do_add(newsk, newinet->inet_daddr, |
cfb6eeb4 | 1481 | newkey, key->keylen); |
a465419b | 1482 | sk_nocaps_add(newsk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1483 | } |
1484 | #endif | |
1485 | ||
0e734419 DM |
1486 | if (__inet_inherit_port(sk, newsk) < 0) |
1487 | goto put_and_exit; | |
9327f705 | 1488 | __inet_hash_nolisten(newsk, NULL); |
1da177e4 LT |
1489 | |
1490 | return newsk; | |
1491 | ||
1492 | exit_overflow: | |
de0744af | 1493 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 BS |
1494 | exit_nonewsk: |
1495 | dst_release(dst); | |
1da177e4 | 1496 | exit: |
de0744af | 1497 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 | 1498 | return NULL; |
0e734419 DM |
1499 | put_and_exit: |
1500 | sock_put(newsk); | |
1501 | goto exit; | |
1da177e4 | 1502 | } |
4bc2f18b | 1503 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); |
1da177e4 LT |
1504 | |
1505 | static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb) | |
1506 | { | |
aa8223c7 | 1507 | struct tcphdr *th = tcp_hdr(skb); |
eddc9ec5 | 1508 | const struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 1509 | struct sock *nsk; |
60236fdd | 1510 | struct request_sock **prev; |
1da177e4 | 1511 | /* Find possible connection requests. */ |
463c84b9 ACM |
1512 | struct request_sock *req = inet_csk_search_req(sk, &prev, th->source, |
1513 | iph->saddr, iph->daddr); | |
1da177e4 LT |
1514 | if (req) |
1515 | return tcp_check_req(sk, skb, req, prev); | |
1516 | ||
3b1e0a65 | 1517 | nsk = inet_lookup_established(sock_net(sk), &tcp_hashinfo, iph->saddr, |
c67499c0 | 1518 | th->source, iph->daddr, th->dest, inet_iif(skb)); |
1da177e4 LT |
1519 | |
1520 | if (nsk) { | |
1521 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
1522 | bh_lock_sock(nsk); | |
1523 | return nsk; | |
1524 | } | |
9469c7b4 | 1525 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
1526 | return NULL; |
1527 | } | |
1528 | ||
1529 | #ifdef CONFIG_SYN_COOKIES | |
af9b4738 | 1530 | if (!th->syn) |
1da177e4 LT |
1531 | sk = cookie_v4_check(sk, skb, &(IPCB(skb)->opt)); |
1532 | #endif | |
1533 | return sk; | |
1534 | } | |
1535 | ||
b51655b9 | 1536 | static __sum16 tcp_v4_checksum_init(struct sk_buff *skb) |
1da177e4 | 1537 | { |
eddc9ec5 ACM |
1538 | const struct iphdr *iph = ip_hdr(skb); |
1539 | ||
84fa7933 | 1540 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
eddc9ec5 ACM |
1541 | if (!tcp_v4_check(skb->len, iph->saddr, |
1542 | iph->daddr, skb->csum)) { | |
fb286bb2 | 1543 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
1da177e4 | 1544 | return 0; |
fb286bb2 | 1545 | } |
1da177e4 | 1546 | } |
fb286bb2 | 1547 | |
eddc9ec5 | 1548 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, |
fb286bb2 HX |
1549 | skb->len, IPPROTO_TCP, 0); |
1550 | ||
1da177e4 | 1551 | if (skb->len <= 76) { |
fb286bb2 | 1552 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1553 | } |
1554 | return 0; | |
1555 | } | |
1556 | ||
1557 | ||
1558 | /* The socket must have it's spinlock held when we get | |
1559 | * here. | |
1560 | * | |
1561 | * We have a potential double-lock case here, so even when | |
1562 | * doing backlog processing we use the BH locking scheme. | |
1563 | * This is because we cannot sleep with the original spinlock | |
1564 | * held. | |
1565 | */ | |
1566 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1567 | { | |
cfb6eeb4 YH |
1568 | struct sock *rsk; |
1569 | #ifdef CONFIG_TCP_MD5SIG | |
1570 | /* | |
1571 | * We really want to reject the packet as early as possible | |
1572 | * if: | |
1573 | * o We're expecting an MD5'd packet and this is no MD5 tcp option | |
1574 | * o There is an MD5 option and we're not expecting one | |
1575 | */ | |
7174259e | 1576 | if (tcp_v4_inbound_md5_hash(sk, skb)) |
cfb6eeb4 YH |
1577 | goto discard; |
1578 | #endif | |
1579 | ||
1da177e4 | 1580 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
ca55158c | 1581 | sock_rps_save_rxhash(sk, skb->rxhash); |
aa8223c7 | 1582 | if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) { |
cfb6eeb4 | 1583 | rsk = sk; |
1da177e4 | 1584 | goto reset; |
cfb6eeb4 | 1585 | } |
1da177e4 LT |
1586 | return 0; |
1587 | } | |
1588 | ||
ab6a5bb6 | 1589 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1590 | goto csum_err; |
1591 | ||
1592 | if (sk->sk_state == TCP_LISTEN) { | |
1593 | struct sock *nsk = tcp_v4_hnd_req(sk, skb); | |
1594 | if (!nsk) | |
1595 | goto discard; | |
1596 | ||
1597 | if (nsk != sk) { | |
1eddcead | 1598 | sock_rps_save_rxhash(nsk, skb->rxhash); |
cfb6eeb4 YH |
1599 | if (tcp_child_process(sk, nsk, skb)) { |
1600 | rsk = nsk; | |
1da177e4 | 1601 | goto reset; |
cfb6eeb4 | 1602 | } |
1da177e4 LT |
1603 | return 0; |
1604 | } | |
ca55158c ED |
1605 | } else |
1606 | sock_rps_save_rxhash(sk, skb->rxhash); | |
1607 | ||
aa8223c7 | 1608 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) { |
cfb6eeb4 | 1609 | rsk = sk; |
1da177e4 | 1610 | goto reset; |
cfb6eeb4 | 1611 | } |
1da177e4 LT |
1612 | return 0; |
1613 | ||
1614 | reset: | |
cfb6eeb4 | 1615 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1616 | discard: |
1617 | kfree_skb(skb); | |
1618 | /* Be careful here. If this function gets more complicated and | |
1619 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1620 | * might be destroyed here. This current version compiles correctly, | |
1621 | * but you have been warned. | |
1622 | */ | |
1623 | return 0; | |
1624 | ||
1625 | csum_err: | |
63231bdd | 1626 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1627 | goto discard; |
1628 | } | |
4bc2f18b | 1629 | EXPORT_SYMBOL(tcp_v4_do_rcv); |
1da177e4 LT |
1630 | |
1631 | /* | |
1632 | * From tcp_input.c | |
1633 | */ | |
1634 | ||
1635 | int tcp_v4_rcv(struct sk_buff *skb) | |
1636 | { | |
eddc9ec5 | 1637 | const struct iphdr *iph; |
1da177e4 LT |
1638 | struct tcphdr *th; |
1639 | struct sock *sk; | |
1640 | int ret; | |
a86b1e30 | 1641 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1642 | |
1643 | if (skb->pkt_type != PACKET_HOST) | |
1644 | goto discard_it; | |
1645 | ||
1646 | /* Count it even if it's bad */ | |
63231bdd | 1647 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1648 | |
1649 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1650 | goto discard_it; | |
1651 | ||
aa8223c7 | 1652 | th = tcp_hdr(skb); |
1da177e4 LT |
1653 | |
1654 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1655 | goto bad_packet; | |
1656 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1657 | goto discard_it; | |
1658 | ||
1659 | /* An explanation is required here, I think. | |
1660 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1661 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1662 | * So, we defer the checks. */ |
60476372 | 1663 | if (!skb_csum_unnecessary(skb) && tcp_v4_checksum_init(skb)) |
1da177e4 LT |
1664 | goto bad_packet; |
1665 | ||
aa8223c7 | 1666 | th = tcp_hdr(skb); |
eddc9ec5 | 1667 | iph = ip_hdr(skb); |
1da177e4 LT |
1668 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1669 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1670 | skb->len - th->doff * 4); | |
1671 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1672 | TCP_SKB_CB(skb)->when = 0; | |
eddc9ec5 | 1673 | TCP_SKB_CB(skb)->flags = iph->tos; |
1da177e4 LT |
1674 | TCP_SKB_CB(skb)->sacked = 0; |
1675 | ||
9a1f27c4 | 1676 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1677 | if (!sk) |
1678 | goto no_tcp_socket; | |
1679 | ||
bb134d5d ED |
1680 | process: |
1681 | if (sk->sk_state == TCP_TIME_WAIT) | |
1682 | goto do_time_wait; | |
1683 | ||
6cce09f8 ED |
1684 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
1685 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
d218d111 | 1686 | goto discard_and_relse; |
6cce09f8 | 1687 | } |
d218d111 | 1688 | |
1da177e4 LT |
1689 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1690 | goto discard_and_relse; | |
b59c2701 | 1691 | nf_reset(skb); |
1da177e4 | 1692 | |
fda9ef5d | 1693 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1694 | goto discard_and_relse; |
1695 | ||
1696 | skb->dev = NULL; | |
1697 | ||
c6366184 | 1698 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1699 | ret = 0; |
1700 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 CL |
1701 | #ifdef CONFIG_NET_DMA |
1702 | struct tcp_sock *tp = tcp_sk(sk); | |
1703 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) | |
f67b4599 | 1704 | tp->ucopy.dma_chan = dma_find_channel(DMA_MEMCPY); |
1a2449a8 | 1705 | if (tp->ucopy.dma_chan) |
1da177e4 | 1706 | ret = tcp_v4_do_rcv(sk, skb); |
1a2449a8 CL |
1707 | else |
1708 | #endif | |
1709 | { | |
1710 | if (!tcp_prequeue(sk, skb)) | |
ae8d7f88 | 1711 | ret = tcp_v4_do_rcv(sk, skb); |
1a2449a8 | 1712 | } |
6cce09f8 | 1713 | } else if (unlikely(sk_add_backlog(sk, skb))) { |
6b03a53a | 1714 | bh_unlock_sock(sk); |
6cce09f8 | 1715 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1716 | goto discard_and_relse; |
1717 | } | |
1da177e4 LT |
1718 | bh_unlock_sock(sk); |
1719 | ||
1720 | sock_put(sk); | |
1721 | ||
1722 | return ret; | |
1723 | ||
1724 | no_tcp_socket: | |
1725 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1726 | goto discard_it; | |
1727 | ||
1728 | if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) { | |
1729 | bad_packet: | |
63231bdd | 1730 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1731 | } else { |
cfb6eeb4 | 1732 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1733 | } |
1734 | ||
1735 | discard_it: | |
1736 | /* Discard frame. */ | |
1737 | kfree_skb(skb); | |
e905a9ed | 1738 | return 0; |
1da177e4 LT |
1739 | |
1740 | discard_and_relse: | |
1741 | sock_put(sk); | |
1742 | goto discard_it; | |
1743 | ||
1744 | do_time_wait: | |
1745 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1746 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1747 | goto discard_it; |
1748 | } | |
1749 | ||
1750 | if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) { | |
63231bdd | 1751 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
9469c7b4 | 1752 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1753 | goto discard_it; |
1754 | } | |
9469c7b4 | 1755 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1756 | case TCP_TW_SYN: { |
c346dca1 | 1757 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
c67499c0 | 1758 | &tcp_hashinfo, |
eddc9ec5 | 1759 | iph->daddr, th->dest, |
463c84b9 | 1760 | inet_iif(skb)); |
1da177e4 | 1761 | if (sk2) { |
9469c7b4 YH |
1762 | inet_twsk_deschedule(inet_twsk(sk), &tcp_death_row); |
1763 | inet_twsk_put(inet_twsk(sk)); | |
1da177e4 LT |
1764 | sk = sk2; |
1765 | goto process; | |
1766 | } | |
1767 | /* Fall through to ACK */ | |
1768 | } | |
1769 | case TCP_TW_ACK: | |
1770 | tcp_v4_timewait_ack(sk, skb); | |
1771 | break; | |
1772 | case TCP_TW_RST: | |
1773 | goto no_tcp_socket; | |
1774 | case TCP_TW_SUCCESS:; | |
1775 | } | |
1776 | goto discard_it; | |
1777 | } | |
1778 | ||
3f419d2d | 1779 | struct inet_peer *tcp_v4_get_peer(struct sock *sk, bool *release_it) |
1da177e4 | 1780 | { |
3f419d2d | 1781 | struct rtable *rt = (struct rtable *) __sk_dst_get(sk); |
1da177e4 | 1782 | struct inet_sock *inet = inet_sk(sk); |
3f419d2d | 1783 | struct inet_peer *peer; |
1da177e4 | 1784 | |
c5216cc7 DM |
1785 | if (!rt || |
1786 | inet->cork.fl.u.ip4.daddr != inet->inet_daddr) { | |
b534ecf1 | 1787 | peer = inet_getpeer_v4(inet->inet_daddr, 1); |
3f419d2d | 1788 | *release_it = true; |
1da177e4 LT |
1789 | } else { |
1790 | if (!rt->peer) | |
a48eff12 | 1791 | rt_bind_peer(rt, inet->inet_daddr, 1); |
1da177e4 | 1792 | peer = rt->peer; |
3f419d2d | 1793 | *release_it = false; |
1da177e4 LT |
1794 | } |
1795 | ||
3f419d2d | 1796 | return peer; |
1da177e4 | 1797 | } |
3f419d2d | 1798 | EXPORT_SYMBOL(tcp_v4_get_peer); |
1da177e4 | 1799 | |
ccb7c410 | 1800 | void *tcp_v4_tw_get_peer(struct sock *sk) |
1da177e4 | 1801 | { |
ccb7c410 | 1802 | struct inet_timewait_sock *tw = inet_twsk(sk); |
1da177e4 | 1803 | |
ccb7c410 | 1804 | return inet_getpeer_v4(tw->tw_daddr, 1); |
1da177e4 | 1805 | } |
ccb7c410 DM |
1806 | EXPORT_SYMBOL(tcp_v4_tw_get_peer); |
1807 | ||
1808 | static struct timewait_sock_ops tcp_timewait_sock_ops = { | |
1809 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1810 | .twsk_unique = tcp_twsk_unique, | |
1811 | .twsk_destructor= tcp_twsk_destructor, | |
1812 | .twsk_getpeer = tcp_v4_tw_get_peer, | |
1813 | }; | |
1da177e4 | 1814 | |
3b401a81 | 1815 | const struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1816 | .queue_xmit = ip_queue_xmit, |
1817 | .send_check = tcp_v4_send_check, | |
1818 | .rebuild_header = inet_sk_rebuild_header, | |
1819 | .conn_request = tcp_v4_conn_request, | |
1820 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
3f419d2d | 1821 | .get_peer = tcp_v4_get_peer, |
543d9cfe ACM |
1822 | .net_header_len = sizeof(struct iphdr), |
1823 | .setsockopt = ip_setsockopt, | |
1824 | .getsockopt = ip_getsockopt, | |
1825 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1826 | .sockaddr_len = sizeof(struct sockaddr_in), | |
ab1e0a13 | 1827 | .bind_conflict = inet_csk_bind_conflict, |
3fdadf7d | 1828 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1829 | .compat_setsockopt = compat_ip_setsockopt, |
1830 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1831 | #endif |
1da177e4 | 1832 | }; |
4bc2f18b | 1833 | EXPORT_SYMBOL(ipv4_specific); |
1da177e4 | 1834 | |
cfb6eeb4 | 1835 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1836 | static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1837 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1838 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 YH |
1839 | .md5_add = tcp_v4_md5_add_func, |
1840 | .md5_parse = tcp_v4_parse_md5_keys, | |
cfb6eeb4 | 1841 | }; |
b6332e6c | 1842 | #endif |
cfb6eeb4 | 1843 | |
1da177e4 LT |
1844 | /* NOTE: A lot of things set to zero explicitly by call to |
1845 | * sk_alloc() so need not be done here. | |
1846 | */ | |
1847 | static int tcp_v4_init_sock(struct sock *sk) | |
1848 | { | |
6687e988 | 1849 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 LT |
1850 | struct tcp_sock *tp = tcp_sk(sk); |
1851 | ||
1852 | skb_queue_head_init(&tp->out_of_order_queue); | |
1853 | tcp_init_xmit_timers(sk); | |
1854 | tcp_prequeue_init(tp); | |
1855 | ||
6687e988 | 1856 | icsk->icsk_rto = TCP_TIMEOUT_INIT; |
1da177e4 LT |
1857 | tp->mdev = TCP_TIMEOUT_INIT; |
1858 | ||
1859 | /* So many TCP implementations out there (incorrectly) count the | |
1860 | * initial SYN frame in their delayed-ACK and congestion control | |
1861 | * algorithms that we must have the following bandaid to talk | |
1862 | * efficiently to them. -DaveM | |
1863 | */ | |
9ad7c049 | 1864 | tp->snd_cwnd = TCP_INIT_CWND; |
1da177e4 LT |
1865 | |
1866 | /* See draft-stevens-tcpca-spec-01 for discussion of the | |
1867 | * initialization of these values. | |
1868 | */ | |
0b6a05c1 | 1869 | tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; |
1da177e4 | 1870 | tp->snd_cwnd_clamp = ~0; |
bee7ca9e | 1871 | tp->mss_cache = TCP_MSS_DEFAULT; |
1da177e4 LT |
1872 | |
1873 | tp->reordering = sysctl_tcp_reordering; | |
6687e988 | 1874 | icsk->icsk_ca_ops = &tcp_init_congestion_ops; |
1da177e4 LT |
1875 | |
1876 | sk->sk_state = TCP_CLOSE; | |
1877 | ||
1878 | sk->sk_write_space = sk_stream_write_space; | |
1879 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
1880 | ||
8292a17a | 1881 | icsk->icsk_af_ops = &ipv4_specific; |
d83d8461 | 1882 | icsk->icsk_sync_mss = tcp_sync_mss; |
cfb6eeb4 YH |
1883 | #ifdef CONFIG_TCP_MD5SIG |
1884 | tp->af_specific = &tcp_sock_ipv4_specific; | |
1885 | #endif | |
1da177e4 | 1886 | |
435cf559 WAS |
1887 | /* TCP Cookie Transactions */ |
1888 | if (sysctl_tcp_cookie_size > 0) { | |
1889 | /* Default, cookies without s_data_payload. */ | |
1890 | tp->cookie_values = | |
1891 | kzalloc(sizeof(*tp->cookie_values), | |
1892 | sk->sk_allocation); | |
1893 | if (tp->cookie_values != NULL) | |
1894 | kref_init(&tp->cookie_values->kref); | |
1895 | } | |
1896 | /* Presumed zeroed, in order of appearance: | |
1897 | * cookie_in_always, cookie_out_never, | |
1898 | * s_data_constant, s_data_in, s_data_out | |
1899 | */ | |
1da177e4 LT |
1900 | sk->sk_sndbuf = sysctl_tcp_wmem[1]; |
1901 | sk->sk_rcvbuf = sysctl_tcp_rmem[1]; | |
1902 | ||
eb4dea58 | 1903 | local_bh_disable(); |
1748376b | 1904 | percpu_counter_inc(&tcp_sockets_allocated); |
eb4dea58 | 1905 | local_bh_enable(); |
1da177e4 LT |
1906 | |
1907 | return 0; | |
1908 | } | |
1909 | ||
7d06b2e0 | 1910 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1911 | { |
1912 | struct tcp_sock *tp = tcp_sk(sk); | |
1913 | ||
1914 | tcp_clear_xmit_timers(sk); | |
1915 | ||
6687e988 | 1916 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1917 | |
1da177e4 | 1918 | /* Cleanup up the write buffer. */ |
fe067e8a | 1919 | tcp_write_queue_purge(sk); |
1da177e4 LT |
1920 | |
1921 | /* Cleans up our, hopefully empty, out_of_order_queue. */ | |
e905a9ed | 1922 | __skb_queue_purge(&tp->out_of_order_queue); |
1da177e4 | 1923 | |
cfb6eeb4 YH |
1924 | #ifdef CONFIG_TCP_MD5SIG |
1925 | /* Clean up the MD5 key list, if any */ | |
1926 | if (tp->md5sig_info) { | |
1927 | tcp_v4_clear_md5_list(sk); | |
1928 | kfree(tp->md5sig_info); | |
1929 | tp->md5sig_info = NULL; | |
1930 | } | |
1931 | #endif | |
1932 | ||
1a2449a8 CL |
1933 | #ifdef CONFIG_NET_DMA |
1934 | /* Cleans up our sk_async_wait_queue */ | |
e905a9ed | 1935 | __skb_queue_purge(&sk->sk_async_wait_queue); |
1a2449a8 CL |
1936 | #endif |
1937 | ||
1da177e4 LT |
1938 | /* Clean prequeue, it must be empty really */ |
1939 | __skb_queue_purge(&tp->ucopy.prequeue); | |
1940 | ||
1941 | /* Clean up a referenced TCP bind bucket. */ | |
463c84b9 | 1942 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1943 | inet_put_port(sk); |
1da177e4 LT |
1944 | |
1945 | /* | |
1946 | * If sendmsg cached page exists, toss it. | |
1947 | */ | |
1948 | if (sk->sk_sndmsg_page) { | |
1949 | __free_page(sk->sk_sndmsg_page); | |
1950 | sk->sk_sndmsg_page = NULL; | |
1951 | } | |
1952 | ||
435cf559 WAS |
1953 | /* TCP Cookie Transactions */ |
1954 | if (tp->cookie_values != NULL) { | |
1955 | kref_put(&tp->cookie_values->kref, | |
1956 | tcp_cookie_values_release); | |
1957 | tp->cookie_values = NULL; | |
1958 | } | |
1959 | ||
1748376b | 1960 | percpu_counter_dec(&tcp_sockets_allocated); |
1da177e4 | 1961 | } |
1da177e4 LT |
1962 | EXPORT_SYMBOL(tcp_v4_destroy_sock); |
1963 | ||
1964 | #ifdef CONFIG_PROC_FS | |
1965 | /* Proc filesystem TCP sock list dumping. */ | |
1966 | ||
3ab5aee7 | 1967 | static inline struct inet_timewait_sock *tw_head(struct hlist_nulls_head *head) |
1da177e4 | 1968 | { |
3ab5aee7 | 1969 | return hlist_nulls_empty(head) ? NULL : |
8feaf0c0 | 1970 | list_entry(head->first, struct inet_timewait_sock, tw_node); |
1da177e4 LT |
1971 | } |
1972 | ||
8feaf0c0 | 1973 | static inline struct inet_timewait_sock *tw_next(struct inet_timewait_sock *tw) |
1da177e4 | 1974 | { |
3ab5aee7 ED |
1975 | return !is_a_nulls(tw->tw_node.next) ? |
1976 | hlist_nulls_entry(tw->tw_node.next, typeof(*tw), tw_node) : NULL; | |
1da177e4 LT |
1977 | } |
1978 | ||
a8b690f9 TH |
1979 | /* |
1980 | * Get next listener socket follow cur. If cur is NULL, get first socket | |
1981 | * starting from bucket given in st->bucket; when st->bucket is zero the | |
1982 | * very first socket in the hash table is returned. | |
1983 | */ | |
1da177e4 LT |
1984 | static void *listening_get_next(struct seq_file *seq, void *cur) |
1985 | { | |
463c84b9 | 1986 | struct inet_connection_sock *icsk; |
c25eb3bf | 1987 | struct hlist_nulls_node *node; |
1da177e4 | 1988 | struct sock *sk = cur; |
5caea4ea | 1989 | struct inet_listen_hashbucket *ilb; |
5799de0b | 1990 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1991 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1992 | |
1993 | if (!sk) { | |
a8b690f9 | 1994 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
5caea4ea | 1995 | spin_lock_bh(&ilb->lock); |
c25eb3bf | 1996 | sk = sk_nulls_head(&ilb->head); |
a8b690f9 | 1997 | st->offset = 0; |
1da177e4 LT |
1998 | goto get_sk; |
1999 | } | |
5caea4ea | 2000 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 | 2001 | ++st->num; |
a8b690f9 | 2002 | ++st->offset; |
1da177e4 LT |
2003 | |
2004 | if (st->state == TCP_SEQ_STATE_OPENREQ) { | |
60236fdd | 2005 | struct request_sock *req = cur; |
1da177e4 | 2006 | |
72a3effa | 2007 | icsk = inet_csk(st->syn_wait_sk); |
1da177e4 LT |
2008 | req = req->dl_next; |
2009 | while (1) { | |
2010 | while (req) { | |
bdccc4ca | 2011 | if (req->rsk_ops->family == st->family) { |
1da177e4 LT |
2012 | cur = req; |
2013 | goto out; | |
2014 | } | |
2015 | req = req->dl_next; | |
2016 | } | |
72a3effa | 2017 | if (++st->sbucket >= icsk->icsk_accept_queue.listen_opt->nr_table_entries) |
1da177e4 LT |
2018 | break; |
2019 | get_req: | |
463c84b9 | 2020 | req = icsk->icsk_accept_queue.listen_opt->syn_table[st->sbucket]; |
1da177e4 | 2021 | } |
1bde5ac4 | 2022 | sk = sk_nulls_next(st->syn_wait_sk); |
1da177e4 | 2023 | st->state = TCP_SEQ_STATE_LISTENING; |
463c84b9 | 2024 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 2025 | } else { |
e905a9ed | 2026 | icsk = inet_csk(sk); |
463c84b9 ACM |
2027 | read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
2028 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) | |
1da177e4 | 2029 | goto start_req; |
463c84b9 | 2030 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1bde5ac4 | 2031 | sk = sk_nulls_next(sk); |
1da177e4 LT |
2032 | } |
2033 | get_sk: | |
c25eb3bf | 2034 | sk_nulls_for_each_from(sk, node) { |
8475ef9f PE |
2035 | if (!net_eq(sock_net(sk), net)) |
2036 | continue; | |
2037 | if (sk->sk_family == st->family) { | |
1da177e4 LT |
2038 | cur = sk; |
2039 | goto out; | |
2040 | } | |
e905a9ed | 2041 | icsk = inet_csk(sk); |
463c84b9 ACM |
2042 | read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
2043 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) { | |
1da177e4 LT |
2044 | start_req: |
2045 | st->uid = sock_i_uid(sk); | |
2046 | st->syn_wait_sk = sk; | |
2047 | st->state = TCP_SEQ_STATE_OPENREQ; | |
2048 | st->sbucket = 0; | |
2049 | goto get_req; | |
2050 | } | |
463c84b9 | 2051 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 2052 | } |
5caea4ea | 2053 | spin_unlock_bh(&ilb->lock); |
a8b690f9 | 2054 | st->offset = 0; |
0f7ff927 | 2055 | if (++st->bucket < INET_LHTABLE_SIZE) { |
5caea4ea ED |
2056 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
2057 | spin_lock_bh(&ilb->lock); | |
c25eb3bf | 2058 | sk = sk_nulls_head(&ilb->head); |
1da177e4 LT |
2059 | goto get_sk; |
2060 | } | |
2061 | cur = NULL; | |
2062 | out: | |
2063 | return cur; | |
2064 | } | |
2065 | ||
2066 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
2067 | { | |
a8b690f9 TH |
2068 | struct tcp_iter_state *st = seq->private; |
2069 | void *rc; | |
2070 | ||
2071 | st->bucket = 0; | |
2072 | st->offset = 0; | |
2073 | rc = listening_get_next(seq, NULL); | |
1da177e4 LT |
2074 | |
2075 | while (rc && *pos) { | |
2076 | rc = listening_get_next(seq, rc); | |
2077 | --*pos; | |
2078 | } | |
2079 | return rc; | |
2080 | } | |
2081 | ||
6eac5604 AK |
2082 | static inline int empty_bucket(struct tcp_iter_state *st) |
2083 | { | |
3ab5aee7 ED |
2084 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain) && |
2085 | hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].twchain); | |
6eac5604 AK |
2086 | } |
2087 | ||
a8b690f9 TH |
2088 | /* |
2089 | * Get first established socket starting from bucket given in st->bucket. | |
2090 | * If st->bucket is zero, the very first socket in the hash is returned. | |
2091 | */ | |
1da177e4 LT |
2092 | static void *established_get_first(struct seq_file *seq) |
2093 | { | |
5799de0b | 2094 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2095 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2096 | void *rc = NULL; |
2097 | ||
a8b690f9 TH |
2098 | st->offset = 0; |
2099 | for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) { | |
1da177e4 | 2100 | struct sock *sk; |
3ab5aee7 | 2101 | struct hlist_nulls_node *node; |
8feaf0c0 | 2102 | struct inet_timewait_sock *tw; |
9db66bdc | 2103 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 2104 | |
6eac5604 AK |
2105 | /* Lockless fast path for the common case of empty buckets */ |
2106 | if (empty_bucket(st)) | |
2107 | continue; | |
2108 | ||
9db66bdc | 2109 | spin_lock_bh(lock); |
3ab5aee7 | 2110 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 2111 | if (sk->sk_family != st->family || |
878628fb | 2112 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
2113 | continue; |
2114 | } | |
2115 | rc = sk; | |
2116 | goto out; | |
2117 | } | |
2118 | st->state = TCP_SEQ_STATE_TIME_WAIT; | |
8feaf0c0 | 2119 | inet_twsk_for_each(tw, node, |
dbca9b27 | 2120 | &tcp_hashinfo.ehash[st->bucket].twchain) { |
28518fc1 | 2121 | if (tw->tw_family != st->family || |
878628fb | 2122 | !net_eq(twsk_net(tw), net)) { |
1da177e4 LT |
2123 | continue; |
2124 | } | |
2125 | rc = tw; | |
2126 | goto out; | |
2127 | } | |
9db66bdc | 2128 | spin_unlock_bh(lock); |
1da177e4 LT |
2129 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2130 | } | |
2131 | out: | |
2132 | return rc; | |
2133 | } | |
2134 | ||
2135 | static void *established_get_next(struct seq_file *seq, void *cur) | |
2136 | { | |
2137 | struct sock *sk = cur; | |
8feaf0c0 | 2138 | struct inet_timewait_sock *tw; |
3ab5aee7 | 2139 | struct hlist_nulls_node *node; |
5799de0b | 2140 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2141 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2142 | |
2143 | ++st->num; | |
a8b690f9 | 2144 | ++st->offset; |
1da177e4 LT |
2145 | |
2146 | if (st->state == TCP_SEQ_STATE_TIME_WAIT) { | |
2147 | tw = cur; | |
2148 | tw = tw_next(tw); | |
2149 | get_tw: | |
878628fb | 2150 | while (tw && (tw->tw_family != st->family || !net_eq(twsk_net(tw), net))) { |
1da177e4 LT |
2151 | tw = tw_next(tw); |
2152 | } | |
2153 | if (tw) { | |
2154 | cur = tw; | |
2155 | goto out; | |
2156 | } | |
9db66bdc | 2157 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2158 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2159 | ||
6eac5604 | 2160 | /* Look for next non empty bucket */ |
a8b690f9 | 2161 | st->offset = 0; |
f373b53b | 2162 | while (++st->bucket <= tcp_hashinfo.ehash_mask && |
6eac5604 AK |
2163 | empty_bucket(st)) |
2164 | ; | |
f373b53b | 2165 | if (st->bucket > tcp_hashinfo.ehash_mask) |
6eac5604 AK |
2166 | return NULL; |
2167 | ||
9db66bdc | 2168 | spin_lock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
3ab5aee7 | 2169 | sk = sk_nulls_head(&tcp_hashinfo.ehash[st->bucket].chain); |
1da177e4 | 2170 | } else |
3ab5aee7 | 2171 | sk = sk_nulls_next(sk); |
1da177e4 | 2172 | |
3ab5aee7 | 2173 | sk_nulls_for_each_from(sk, node) { |
878628fb | 2174 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
1da177e4 LT |
2175 | goto found; |
2176 | } | |
2177 | ||
2178 | st->state = TCP_SEQ_STATE_TIME_WAIT; | |
dbca9b27 | 2179 | tw = tw_head(&tcp_hashinfo.ehash[st->bucket].twchain); |
1da177e4 LT |
2180 | goto get_tw; |
2181 | found: | |
2182 | cur = sk; | |
2183 | out: | |
2184 | return cur; | |
2185 | } | |
2186 | ||
2187 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2188 | { | |
a8b690f9 TH |
2189 | struct tcp_iter_state *st = seq->private; |
2190 | void *rc; | |
2191 | ||
2192 | st->bucket = 0; | |
2193 | rc = established_get_first(seq); | |
1da177e4 LT |
2194 | |
2195 | while (rc && pos) { | |
2196 | rc = established_get_next(seq, rc); | |
2197 | --pos; | |
7174259e | 2198 | } |
1da177e4 LT |
2199 | return rc; |
2200 | } | |
2201 | ||
2202 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2203 | { | |
2204 | void *rc; | |
5799de0b | 2205 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2206 | |
1da177e4 LT |
2207 | st->state = TCP_SEQ_STATE_LISTENING; |
2208 | rc = listening_get_idx(seq, &pos); | |
2209 | ||
2210 | if (!rc) { | |
1da177e4 LT |
2211 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2212 | rc = established_get_idx(seq, pos); | |
2213 | } | |
2214 | ||
2215 | return rc; | |
2216 | } | |
2217 | ||
a8b690f9 TH |
2218 | static void *tcp_seek_last_pos(struct seq_file *seq) |
2219 | { | |
2220 | struct tcp_iter_state *st = seq->private; | |
2221 | int offset = st->offset; | |
2222 | int orig_num = st->num; | |
2223 | void *rc = NULL; | |
2224 | ||
2225 | switch (st->state) { | |
2226 | case TCP_SEQ_STATE_OPENREQ: | |
2227 | case TCP_SEQ_STATE_LISTENING: | |
2228 | if (st->bucket >= INET_LHTABLE_SIZE) | |
2229 | break; | |
2230 | st->state = TCP_SEQ_STATE_LISTENING; | |
2231 | rc = listening_get_next(seq, NULL); | |
2232 | while (offset-- && rc) | |
2233 | rc = listening_get_next(seq, rc); | |
2234 | if (rc) | |
2235 | break; | |
2236 | st->bucket = 0; | |
2237 | /* Fallthrough */ | |
2238 | case TCP_SEQ_STATE_ESTABLISHED: | |
2239 | case TCP_SEQ_STATE_TIME_WAIT: | |
2240 | st->state = TCP_SEQ_STATE_ESTABLISHED; | |
2241 | if (st->bucket > tcp_hashinfo.ehash_mask) | |
2242 | break; | |
2243 | rc = established_get_first(seq); | |
2244 | while (offset-- && rc) | |
2245 | rc = established_get_next(seq, rc); | |
2246 | } | |
2247 | ||
2248 | st->num = orig_num; | |
2249 | ||
2250 | return rc; | |
2251 | } | |
2252 | ||
1da177e4 LT |
2253 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) |
2254 | { | |
5799de0b | 2255 | struct tcp_iter_state *st = seq->private; |
a8b690f9 TH |
2256 | void *rc; |
2257 | ||
2258 | if (*pos && *pos == st->last_pos) { | |
2259 | rc = tcp_seek_last_pos(seq); | |
2260 | if (rc) | |
2261 | goto out; | |
2262 | } | |
2263 | ||
1da177e4 LT |
2264 | st->state = TCP_SEQ_STATE_LISTENING; |
2265 | st->num = 0; | |
a8b690f9 TH |
2266 | st->bucket = 0; |
2267 | st->offset = 0; | |
2268 | rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2269 | ||
2270 | out: | |
2271 | st->last_pos = *pos; | |
2272 | return rc; | |
1da177e4 LT |
2273 | } |
2274 | ||
2275 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2276 | { | |
a8b690f9 | 2277 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2278 | void *rc = NULL; |
1da177e4 LT |
2279 | |
2280 | if (v == SEQ_START_TOKEN) { | |
2281 | rc = tcp_get_idx(seq, 0); | |
2282 | goto out; | |
2283 | } | |
1da177e4 LT |
2284 | |
2285 | switch (st->state) { | |
2286 | case TCP_SEQ_STATE_OPENREQ: | |
2287 | case TCP_SEQ_STATE_LISTENING: | |
2288 | rc = listening_get_next(seq, v); | |
2289 | if (!rc) { | |
1da177e4 | 2290 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2291 | st->bucket = 0; |
2292 | st->offset = 0; | |
1da177e4 LT |
2293 | rc = established_get_first(seq); |
2294 | } | |
2295 | break; | |
2296 | case TCP_SEQ_STATE_ESTABLISHED: | |
2297 | case TCP_SEQ_STATE_TIME_WAIT: | |
2298 | rc = established_get_next(seq, v); | |
2299 | break; | |
2300 | } | |
2301 | out: | |
2302 | ++*pos; | |
a8b690f9 | 2303 | st->last_pos = *pos; |
1da177e4 LT |
2304 | return rc; |
2305 | } | |
2306 | ||
2307 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2308 | { | |
5799de0b | 2309 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2310 | |
2311 | switch (st->state) { | |
2312 | case TCP_SEQ_STATE_OPENREQ: | |
2313 | if (v) { | |
463c84b9 ACM |
2314 | struct inet_connection_sock *icsk = inet_csk(st->syn_wait_sk); |
2315 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); | |
1da177e4 LT |
2316 | } |
2317 | case TCP_SEQ_STATE_LISTENING: | |
2318 | if (v != SEQ_START_TOKEN) | |
5caea4ea | 2319 | spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 LT |
2320 | break; |
2321 | case TCP_SEQ_STATE_TIME_WAIT: | |
2322 | case TCP_SEQ_STATE_ESTABLISHED: | |
2323 | if (v) | |
9db66bdc | 2324 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2325 | break; |
2326 | } | |
2327 | } | |
2328 | ||
2329 | static int tcp_seq_open(struct inode *inode, struct file *file) | |
2330 | { | |
2331 | struct tcp_seq_afinfo *afinfo = PDE(inode)->data; | |
1da177e4 | 2332 | struct tcp_iter_state *s; |
52d6f3f1 | 2333 | int err; |
1da177e4 | 2334 | |
52d6f3f1 DL |
2335 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2336 | sizeof(struct tcp_iter_state)); | |
2337 | if (err < 0) | |
2338 | return err; | |
f40c8174 | 2339 | |
52d6f3f1 | 2340 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2341 | s->family = afinfo->family; |
a8b690f9 | 2342 | s->last_pos = 0; |
f40c8174 DL |
2343 | return 0; |
2344 | } | |
2345 | ||
6f8b13bc | 2346 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2347 | { |
2348 | int rc = 0; | |
2349 | struct proc_dir_entry *p; | |
2350 | ||
68fcadd1 DL |
2351 | afinfo->seq_fops.open = tcp_seq_open; |
2352 | afinfo->seq_fops.read = seq_read; | |
2353 | afinfo->seq_fops.llseek = seq_lseek; | |
2354 | afinfo->seq_fops.release = seq_release_net; | |
7174259e | 2355 | |
9427c4b3 DL |
2356 | afinfo->seq_ops.start = tcp_seq_start; |
2357 | afinfo->seq_ops.next = tcp_seq_next; | |
2358 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2359 | ||
84841c3c DL |
2360 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
2361 | &afinfo->seq_fops, afinfo); | |
2362 | if (!p) | |
1da177e4 LT |
2363 | rc = -ENOMEM; |
2364 | return rc; | |
2365 | } | |
4bc2f18b | 2366 | EXPORT_SYMBOL(tcp_proc_register); |
1da177e4 | 2367 | |
6f8b13bc | 2368 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2369 | { |
6f8b13bc | 2370 | proc_net_remove(net, afinfo->name); |
1da177e4 | 2371 | } |
4bc2f18b | 2372 | EXPORT_SYMBOL(tcp_proc_unregister); |
1da177e4 | 2373 | |
60236fdd | 2374 | static void get_openreq4(struct sock *sk, struct request_sock *req, |
5e659e4c | 2375 | struct seq_file *f, int i, int uid, int *len) |
1da177e4 | 2376 | { |
2e6599cb | 2377 | const struct inet_request_sock *ireq = inet_rsk(req); |
1da177e4 LT |
2378 | int ttd = req->expires - jiffies; |
2379 | ||
5e659e4c | 2380 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
71338aa7 | 2381 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %u %d %pK%n", |
1da177e4 | 2382 | i, |
2e6599cb | 2383 | ireq->loc_addr, |
c720c7e8 | 2384 | ntohs(inet_sk(sk)->inet_sport), |
2e6599cb ACM |
2385 | ireq->rmt_addr, |
2386 | ntohs(ireq->rmt_port), | |
1da177e4 LT |
2387 | TCP_SYN_RECV, |
2388 | 0, 0, /* could print option size, but that is af dependent. */ | |
2389 | 1, /* timers active (only the expire timer) */ | |
2390 | jiffies_to_clock_t(ttd), | |
2391 | req->retrans, | |
2392 | uid, | |
2393 | 0, /* non standard timer */ | |
2394 | 0, /* open_requests have no inode */ | |
2395 | atomic_read(&sk->sk_refcnt), | |
5e659e4c PE |
2396 | req, |
2397 | len); | |
1da177e4 LT |
2398 | } |
2399 | ||
5e659e4c | 2400 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i, int *len) |
1da177e4 LT |
2401 | { |
2402 | int timer_active; | |
2403 | unsigned long timer_expires; | |
cf4c6bf8 IJ |
2404 | struct tcp_sock *tp = tcp_sk(sk); |
2405 | const struct inet_connection_sock *icsk = inet_csk(sk); | |
2406 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 ED |
2407 | __be32 dest = inet->inet_daddr; |
2408 | __be32 src = inet->inet_rcv_saddr; | |
2409 | __u16 destp = ntohs(inet->inet_dport); | |
2410 | __u16 srcp = ntohs(inet->inet_sport); | |
49d09007 | 2411 | int rx_queue; |
1da177e4 | 2412 | |
463c84b9 | 2413 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { |
1da177e4 | 2414 | timer_active = 1; |
463c84b9 ACM |
2415 | timer_expires = icsk->icsk_timeout; |
2416 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2417 | timer_active = 4; |
463c84b9 | 2418 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2419 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2420 | timer_active = 2; |
cf4c6bf8 | 2421 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2422 | } else { |
2423 | timer_active = 0; | |
2424 | timer_expires = jiffies; | |
2425 | } | |
2426 | ||
49d09007 ED |
2427 | if (sk->sk_state == TCP_LISTEN) |
2428 | rx_queue = sk->sk_ack_backlog; | |
2429 | else | |
2430 | /* | |
2431 | * because we dont lock socket, we might find a transient negative value | |
2432 | */ | |
2433 | rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0); | |
2434 | ||
5e659e4c | 2435 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
71338aa7 | 2436 | "%08X %5d %8d %lu %d %pK %lu %lu %u %u %d%n", |
cf4c6bf8 | 2437 | i, src, srcp, dest, destp, sk->sk_state, |
47da8ee6 | 2438 | tp->write_seq - tp->snd_una, |
49d09007 | 2439 | rx_queue, |
1da177e4 LT |
2440 | timer_active, |
2441 | jiffies_to_clock_t(timer_expires - jiffies), | |
463c84b9 | 2442 | icsk->icsk_retransmits, |
cf4c6bf8 | 2443 | sock_i_uid(sk), |
6687e988 | 2444 | icsk->icsk_probes_out, |
cf4c6bf8 IJ |
2445 | sock_i_ino(sk), |
2446 | atomic_read(&sk->sk_refcnt), sk, | |
7be87351 SH |
2447 | jiffies_to_clock_t(icsk->icsk_rto), |
2448 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2449 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2450 | tp->snd_cwnd, |
0b6a05c1 | 2451 | tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh, |
5e659e4c | 2452 | len); |
1da177e4 LT |
2453 | } |
2454 | ||
7174259e | 2455 | static void get_timewait4_sock(struct inet_timewait_sock *tw, |
5e659e4c | 2456 | struct seq_file *f, int i, int *len) |
1da177e4 | 2457 | { |
23f33c2d | 2458 | __be32 dest, src; |
1da177e4 LT |
2459 | __u16 destp, srcp; |
2460 | int ttd = tw->tw_ttd - jiffies; | |
2461 | ||
2462 | if (ttd < 0) | |
2463 | ttd = 0; | |
2464 | ||
2465 | dest = tw->tw_daddr; | |
2466 | src = tw->tw_rcv_saddr; | |
2467 | destp = ntohs(tw->tw_dport); | |
2468 | srcp = ntohs(tw->tw_sport); | |
2469 | ||
5e659e4c | 2470 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
71338aa7 | 2471 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK%n", |
1da177e4 LT |
2472 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
2473 | 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0, | |
5e659e4c | 2474 | atomic_read(&tw->tw_refcnt), tw, len); |
1da177e4 LT |
2475 | } |
2476 | ||
2477 | #define TMPSZ 150 | |
2478 | ||
2479 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2480 | { | |
5799de0b | 2481 | struct tcp_iter_state *st; |
5e659e4c | 2482 | int len; |
1da177e4 LT |
2483 | |
2484 | if (v == SEQ_START_TOKEN) { | |
2485 | seq_printf(seq, "%-*s\n", TMPSZ - 1, | |
2486 | " sl local_address rem_address st tx_queue " | |
2487 | "rx_queue tr tm->when retrnsmt uid timeout " | |
2488 | "inode"); | |
2489 | goto out; | |
2490 | } | |
2491 | st = seq->private; | |
2492 | ||
2493 | switch (st->state) { | |
2494 | case TCP_SEQ_STATE_LISTENING: | |
2495 | case TCP_SEQ_STATE_ESTABLISHED: | |
5e659e4c | 2496 | get_tcp4_sock(v, seq, st->num, &len); |
1da177e4 LT |
2497 | break; |
2498 | case TCP_SEQ_STATE_OPENREQ: | |
5e659e4c | 2499 | get_openreq4(st->syn_wait_sk, v, seq, st->num, st->uid, &len); |
1da177e4 LT |
2500 | break; |
2501 | case TCP_SEQ_STATE_TIME_WAIT: | |
5e659e4c | 2502 | get_timewait4_sock(v, seq, st->num, &len); |
1da177e4 LT |
2503 | break; |
2504 | } | |
5e659e4c | 2505 | seq_printf(seq, "%*s\n", TMPSZ - 1 - len, ""); |
1da177e4 LT |
2506 | out: |
2507 | return 0; | |
2508 | } | |
2509 | ||
1da177e4 | 2510 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2511 | .name = "tcp", |
2512 | .family = AF_INET, | |
5f4472c5 DL |
2513 | .seq_fops = { |
2514 | .owner = THIS_MODULE, | |
2515 | }, | |
9427c4b3 DL |
2516 | .seq_ops = { |
2517 | .show = tcp4_seq_show, | |
2518 | }, | |
1da177e4 LT |
2519 | }; |
2520 | ||
2c8c1e72 | 2521 | static int __net_init tcp4_proc_init_net(struct net *net) |
757764f6 PE |
2522 | { |
2523 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2524 | } | |
2525 | ||
2c8c1e72 | 2526 | static void __net_exit tcp4_proc_exit_net(struct net *net) |
757764f6 PE |
2527 | { |
2528 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2529 | } | |
2530 | ||
2531 | static struct pernet_operations tcp4_net_ops = { | |
2532 | .init = tcp4_proc_init_net, | |
2533 | .exit = tcp4_proc_exit_net, | |
2534 | }; | |
2535 | ||
1da177e4 LT |
2536 | int __init tcp4_proc_init(void) |
2537 | { | |
757764f6 | 2538 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2539 | } |
2540 | ||
2541 | void tcp4_proc_exit(void) | |
2542 | { | |
757764f6 | 2543 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2544 | } |
2545 | #endif /* CONFIG_PROC_FS */ | |
2546 | ||
bf296b12 HX |
2547 | struct sk_buff **tcp4_gro_receive(struct sk_buff **head, struct sk_buff *skb) |
2548 | { | |
b71d1d42 | 2549 | const struct iphdr *iph = skb_gro_network_header(skb); |
bf296b12 HX |
2550 | |
2551 | switch (skb->ip_summed) { | |
2552 | case CHECKSUM_COMPLETE: | |
86911732 | 2553 | if (!tcp_v4_check(skb_gro_len(skb), iph->saddr, iph->daddr, |
bf296b12 HX |
2554 | skb->csum)) { |
2555 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
2556 | break; | |
2557 | } | |
2558 | ||
2559 | /* fall through */ | |
2560 | case CHECKSUM_NONE: | |
2561 | NAPI_GRO_CB(skb)->flush = 1; | |
2562 | return NULL; | |
2563 | } | |
2564 | ||
2565 | return tcp_gro_receive(head, skb); | |
2566 | } | |
bf296b12 HX |
2567 | |
2568 | int tcp4_gro_complete(struct sk_buff *skb) | |
2569 | { | |
b71d1d42 | 2570 | const struct iphdr *iph = ip_hdr(skb); |
bf296b12 HX |
2571 | struct tcphdr *th = tcp_hdr(skb); |
2572 | ||
2573 | th->check = ~tcp_v4_check(skb->len - skb_transport_offset(skb), | |
2574 | iph->saddr, iph->daddr, 0); | |
2575 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; | |
2576 | ||
2577 | return tcp_gro_complete(skb); | |
2578 | } | |
bf296b12 | 2579 | |
1da177e4 LT |
2580 | struct proto tcp_prot = { |
2581 | .name = "TCP", | |
2582 | .owner = THIS_MODULE, | |
2583 | .close = tcp_close, | |
2584 | .connect = tcp_v4_connect, | |
2585 | .disconnect = tcp_disconnect, | |
463c84b9 | 2586 | .accept = inet_csk_accept, |
1da177e4 LT |
2587 | .ioctl = tcp_ioctl, |
2588 | .init = tcp_v4_init_sock, | |
2589 | .destroy = tcp_v4_destroy_sock, | |
2590 | .shutdown = tcp_shutdown, | |
2591 | .setsockopt = tcp_setsockopt, | |
2592 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 2593 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2594 | .sendmsg = tcp_sendmsg, |
2595 | .sendpage = tcp_sendpage, | |
1da177e4 | 2596 | .backlog_rcv = tcp_v4_do_rcv, |
ab1e0a13 ACM |
2597 | .hash = inet_hash, |
2598 | .unhash = inet_unhash, | |
2599 | .get_port = inet_csk_get_port, | |
1da177e4 LT |
2600 | .enter_memory_pressure = tcp_enter_memory_pressure, |
2601 | .sockets_allocated = &tcp_sockets_allocated, | |
0a5578cf | 2602 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2603 | .memory_allocated = &tcp_memory_allocated, |
2604 | .memory_pressure = &tcp_memory_pressure, | |
2605 | .sysctl_mem = sysctl_tcp_mem, | |
2606 | .sysctl_wmem = sysctl_tcp_wmem, | |
2607 | .sysctl_rmem = sysctl_tcp_rmem, | |
2608 | .max_header = MAX_TCP_HEADER, | |
2609 | .obj_size = sizeof(struct tcp_sock), | |
3ab5aee7 | 2610 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2611 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2612 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2613 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2614 | .no_autobind = true, |
543d9cfe ACM |
2615 | #ifdef CONFIG_COMPAT |
2616 | .compat_setsockopt = compat_tcp_setsockopt, | |
2617 | .compat_getsockopt = compat_tcp_getsockopt, | |
2618 | #endif | |
1da177e4 | 2619 | }; |
4bc2f18b | 2620 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 | 2621 | |
046ee902 DL |
2622 | |
2623 | static int __net_init tcp_sk_init(struct net *net) | |
2624 | { | |
2625 | return inet_ctl_sock_create(&net->ipv4.tcp_sock, | |
2626 | PF_INET, SOCK_RAW, IPPROTO_TCP, net); | |
2627 | } | |
2628 | ||
2629 | static void __net_exit tcp_sk_exit(struct net *net) | |
2630 | { | |
2631 | inet_ctl_sock_destroy(net->ipv4.tcp_sock); | |
b099ce26 EB |
2632 | } |
2633 | ||
2634 | static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list) | |
2635 | { | |
2636 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET); | |
046ee902 DL |
2637 | } |
2638 | ||
2639 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
b099ce26 EB |
2640 | .init = tcp_sk_init, |
2641 | .exit = tcp_sk_exit, | |
2642 | .exit_batch = tcp_sk_exit_batch, | |
046ee902 DL |
2643 | }; |
2644 | ||
9b0f976f | 2645 | void __init tcp_v4_init(void) |
1da177e4 | 2646 | { |
5caea4ea | 2647 | inet_hashinfo_init(&tcp_hashinfo); |
6a1b3054 | 2648 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2649 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 | 2650 | } |