]>
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 | ||
afd46503 | 53 | #define pr_fmt(fmt) "TCP: " fmt |
1da177e4 | 54 | |
eb4dea58 | 55 | #include <linux/bottom_half.h> |
1da177e4 LT |
56 | #include <linux/types.h> |
57 | #include <linux/fcntl.h> | |
58 | #include <linux/module.h> | |
59 | #include <linux/random.h> | |
60 | #include <linux/cache.h> | |
61 | #include <linux/jhash.h> | |
62 | #include <linux/init.h> | |
63 | #include <linux/times.h> | |
5a0e3ad6 | 64 | #include <linux/slab.h> |
1da177e4 | 65 | |
457c4cbc | 66 | #include <net/net_namespace.h> |
1da177e4 | 67 | #include <net/icmp.h> |
304a1618 | 68 | #include <net/inet_hashtables.h> |
1da177e4 | 69 | #include <net/tcp.h> |
20380731 | 70 | #include <net/transp_v6.h> |
1da177e4 LT |
71 | #include <net/ipv6.h> |
72 | #include <net/inet_common.h> | |
6d6ee43e | 73 | #include <net/timewait_sock.h> |
1da177e4 | 74 | #include <net/xfrm.h> |
6e5714ea | 75 | #include <net/secure_seq.h> |
d1a4c0b3 | 76 | #include <net/tcp_memcontrol.h> |
076bb0c8 | 77 | #include <net/busy_poll.h> |
1da177e4 LT |
78 | |
79 | #include <linux/inet.h> | |
80 | #include <linux/ipv6.h> | |
81 | #include <linux/stddef.h> | |
82 | #include <linux/proc_fs.h> | |
83 | #include <linux/seq_file.h> | |
84 | ||
cfb6eeb4 YH |
85 | #include <linux/crypto.h> |
86 | #include <linux/scatterlist.h> | |
87 | ||
ab32ea5d BH |
88 | int sysctl_tcp_tw_reuse __read_mostly; |
89 | int sysctl_tcp_low_latency __read_mostly; | |
4bc2f18b | 90 | EXPORT_SYMBOL(sysctl_tcp_low_latency); |
1da177e4 | 91 | |
cfb6eeb4 | 92 | #ifdef CONFIG_TCP_MD5SIG |
a915da9b | 93 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 94 | __be32 daddr, __be32 saddr, const struct tcphdr *th); |
cfb6eeb4 YH |
95 | #endif |
96 | ||
5caea4ea | 97 | struct inet_hashinfo tcp_hashinfo; |
4bc2f18b | 98 | EXPORT_SYMBOL(tcp_hashinfo); |
1da177e4 | 99 | |
936b8bdb | 100 | static __u32 tcp_v4_init_sequence(const struct sk_buff *skb) |
1da177e4 | 101 | { |
eddc9ec5 ACM |
102 | return secure_tcp_sequence_number(ip_hdr(skb)->daddr, |
103 | ip_hdr(skb)->saddr, | |
aa8223c7 ACM |
104 | tcp_hdr(skb)->dest, |
105 | tcp_hdr(skb)->source); | |
1da177e4 LT |
106 | } |
107 | ||
6d6ee43e ACM |
108 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp) |
109 | { | |
110 | const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw); | |
111 | struct tcp_sock *tp = tcp_sk(sk); | |
112 | ||
113 | /* With PAWS, it is safe from the viewpoint | |
114 | of data integrity. Even without PAWS it is safe provided sequence | |
115 | spaces do not overlap i.e. at data rates <= 80Mbit/sec. | |
116 | ||
117 | Actually, the idea is close to VJ's one, only timestamp cache is | |
118 | held not per host, but per port pair and TW bucket is used as state | |
119 | holder. | |
120 | ||
121 | If TW bucket has been already destroyed we fall back to VJ's scheme | |
122 | and use initial timestamp retrieved from peer table. | |
123 | */ | |
124 | if (tcptw->tw_ts_recent_stamp && | |
125 | (twp == NULL || (sysctl_tcp_tw_reuse && | |
9d729f72 | 126 | get_seconds() - tcptw->tw_ts_recent_stamp > 1))) { |
6d6ee43e ACM |
127 | tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2; |
128 | if (tp->write_seq == 0) | |
129 | tp->write_seq = 1; | |
130 | tp->rx_opt.ts_recent = tcptw->tw_ts_recent; | |
131 | tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp; | |
132 | sock_hold(sktw); | |
133 | return 1; | |
134 | } | |
135 | ||
136 | return 0; | |
137 | } | |
6d6ee43e ACM |
138 | EXPORT_SYMBOL_GPL(tcp_twsk_unique); |
139 | ||
1da177e4 LT |
140 | /* This will initiate an outgoing connection. */ |
141 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
142 | { | |
2d7192d6 | 143 | struct sockaddr_in *usin = (struct sockaddr_in *)uaddr; |
1da177e4 LT |
144 | struct inet_sock *inet = inet_sk(sk); |
145 | struct tcp_sock *tp = tcp_sk(sk); | |
dca8b089 | 146 | __be16 orig_sport, orig_dport; |
bada8adc | 147 | __be32 daddr, nexthop; |
da905bd1 | 148 | struct flowi4 *fl4; |
2d7192d6 | 149 | struct rtable *rt; |
1da177e4 | 150 | int err; |
f6d8bd05 | 151 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
152 | |
153 | if (addr_len < sizeof(struct sockaddr_in)) | |
154 | return -EINVAL; | |
155 | ||
156 | if (usin->sin_family != AF_INET) | |
157 | return -EAFNOSUPPORT; | |
158 | ||
159 | nexthop = daddr = usin->sin_addr.s_addr; | |
f6d8bd05 ED |
160 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
161 | sock_owned_by_user(sk)); | |
162 | if (inet_opt && inet_opt->opt.srr) { | |
1da177e4 LT |
163 | if (!daddr) |
164 | return -EINVAL; | |
f6d8bd05 | 165 | nexthop = inet_opt->opt.faddr; |
1da177e4 LT |
166 | } |
167 | ||
dca8b089 DM |
168 | orig_sport = inet->inet_sport; |
169 | orig_dport = usin->sin_port; | |
da905bd1 DM |
170 | fl4 = &inet->cork.fl.u.ip4; |
171 | rt = ip_route_connect(fl4, nexthop, inet->inet_saddr, | |
b23dd4fe DM |
172 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if, |
173 | IPPROTO_TCP, | |
0e0d44ab | 174 | orig_sport, orig_dport, sk); |
b23dd4fe DM |
175 | if (IS_ERR(rt)) { |
176 | err = PTR_ERR(rt); | |
177 | if (err == -ENETUNREACH) | |
f1d8cba6 | 178 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); |
b23dd4fe | 179 | return err; |
584bdf8c | 180 | } |
1da177e4 LT |
181 | |
182 | if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) { | |
183 | ip_rt_put(rt); | |
184 | return -ENETUNREACH; | |
185 | } | |
186 | ||
f6d8bd05 | 187 | if (!inet_opt || !inet_opt->opt.srr) |
da905bd1 | 188 | daddr = fl4->daddr; |
1da177e4 | 189 | |
c720c7e8 | 190 | if (!inet->inet_saddr) |
da905bd1 | 191 | inet->inet_saddr = fl4->saddr; |
d1e559d0 | 192 | sk_rcv_saddr_set(sk, inet->inet_saddr); |
1da177e4 | 193 | |
c720c7e8 | 194 | if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) { |
1da177e4 LT |
195 | /* Reset inherited state */ |
196 | tp->rx_opt.ts_recent = 0; | |
197 | tp->rx_opt.ts_recent_stamp = 0; | |
ee995283 PE |
198 | if (likely(!tp->repair)) |
199 | tp->write_seq = 0; | |
1da177e4 LT |
200 | } |
201 | ||
295ff7ed | 202 | if (tcp_death_row.sysctl_tw_recycle && |
81166dd6 DM |
203 | !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr) |
204 | tcp_fetch_timewait_stamp(sk, &rt->dst); | |
1da177e4 | 205 | |
c720c7e8 | 206 | inet->inet_dport = usin->sin_port; |
d1e559d0 | 207 | sk_daddr_set(sk, daddr); |
1da177e4 | 208 | |
d83d8461 | 209 | inet_csk(sk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
210 | if (inet_opt) |
211 | inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
1da177e4 | 212 | |
bee7ca9e | 213 | tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT; |
1da177e4 LT |
214 | |
215 | /* Socket identity is still unknown (sport may be zero). | |
216 | * However we set state to SYN-SENT and not releasing socket | |
217 | * lock select source port, enter ourselves into the hash tables and | |
218 | * complete initialization after this. | |
219 | */ | |
220 | tcp_set_state(sk, TCP_SYN_SENT); | |
a7f5e7f1 | 221 | err = inet_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
222 | if (err) |
223 | goto failure; | |
224 | ||
9e7ceb06 SP |
225 | inet_set_txhash(sk); |
226 | ||
da905bd1 | 227 | rt = ip_route_newports(fl4, rt, orig_sport, orig_dport, |
b23dd4fe DM |
228 | inet->inet_sport, inet->inet_dport, sk); |
229 | if (IS_ERR(rt)) { | |
230 | err = PTR_ERR(rt); | |
231 | rt = NULL; | |
1da177e4 | 232 | goto failure; |
b23dd4fe | 233 | } |
1da177e4 | 234 | /* OK, now commit destination to socket. */ |
bcd76111 | 235 | sk->sk_gso_type = SKB_GSO_TCPV4; |
d8d1f30b | 236 | sk_setup_caps(sk, &rt->dst); |
1da177e4 | 237 | |
ee995283 | 238 | if (!tp->write_seq && likely(!tp->repair)) |
c720c7e8 ED |
239 | tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr, |
240 | inet->inet_daddr, | |
241 | inet->inet_sport, | |
1da177e4 LT |
242 | usin->sin_port); |
243 | ||
c720c7e8 | 244 | inet->inet_id = tp->write_seq ^ jiffies; |
1da177e4 | 245 | |
2b916477 | 246 | err = tcp_connect(sk); |
ee995283 | 247 | |
1da177e4 LT |
248 | rt = NULL; |
249 | if (err) | |
250 | goto failure; | |
251 | ||
252 | return 0; | |
253 | ||
254 | failure: | |
7174259e ACM |
255 | /* |
256 | * This unhashes the socket and releases the local port, | |
257 | * if necessary. | |
258 | */ | |
1da177e4 LT |
259 | tcp_set_state(sk, TCP_CLOSE); |
260 | ip_rt_put(rt); | |
261 | sk->sk_route_caps = 0; | |
c720c7e8 | 262 | inet->inet_dport = 0; |
1da177e4 LT |
263 | return err; |
264 | } | |
4bc2f18b | 265 | EXPORT_SYMBOL(tcp_v4_connect); |
1da177e4 | 266 | |
1da177e4 | 267 | /* |
563d34d0 ED |
268 | * This routine reacts to ICMP_FRAG_NEEDED mtu indications as defined in RFC1191. |
269 | * It can be called through tcp_release_cb() if socket was owned by user | |
270 | * at the time tcp_v4_err() was called to handle ICMP message. | |
1da177e4 | 271 | */ |
4fab9071 | 272 | void tcp_v4_mtu_reduced(struct sock *sk) |
1da177e4 LT |
273 | { |
274 | struct dst_entry *dst; | |
275 | struct inet_sock *inet = inet_sk(sk); | |
563d34d0 | 276 | u32 mtu = tcp_sk(sk)->mtu_info; |
1da177e4 | 277 | |
80d0a69f DM |
278 | dst = inet_csk_update_pmtu(sk, mtu); |
279 | if (!dst) | |
1da177e4 LT |
280 | return; |
281 | ||
1da177e4 LT |
282 | /* Something is about to be wrong... Remember soft error |
283 | * for the case, if this connection will not able to recover. | |
284 | */ | |
285 | if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst)) | |
286 | sk->sk_err_soft = EMSGSIZE; | |
287 | ||
288 | mtu = dst_mtu(dst); | |
289 | ||
290 | if (inet->pmtudisc != IP_PMTUDISC_DONT && | |
482fc609 | 291 | ip_sk_accept_pmtu(sk) && |
d83d8461 | 292 | inet_csk(sk)->icsk_pmtu_cookie > mtu) { |
1da177e4 LT |
293 | tcp_sync_mss(sk, mtu); |
294 | ||
295 | /* Resend the TCP packet because it's | |
296 | * clear that the old packet has been | |
297 | * dropped. This is the new "fast" path mtu | |
298 | * discovery. | |
299 | */ | |
300 | tcp_simple_retransmit(sk); | |
301 | } /* else let the usual retransmit timer handle it */ | |
302 | } | |
4fab9071 | 303 | EXPORT_SYMBOL(tcp_v4_mtu_reduced); |
1da177e4 | 304 | |
55be7a9c DM |
305 | static void do_redirect(struct sk_buff *skb, struct sock *sk) |
306 | { | |
307 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
308 | ||
1ed5c48f | 309 | if (dst) |
6700c270 | 310 | dst->ops->redirect(dst, sk, skb); |
55be7a9c DM |
311 | } |
312 | ||
26e37360 ED |
313 | |
314 | /* handle ICMP messages on TCP_NEW_SYN_RECV request sockets */ | |
315 | void tcp_req_err(struct sock *sk, u32 seq) | |
316 | { | |
317 | struct request_sock *req = inet_reqsk(sk); | |
318 | struct net *net = sock_net(sk); | |
319 | ||
320 | /* ICMPs are not backlogged, hence we cannot get | |
321 | * an established socket here. | |
322 | */ | |
323 | WARN_ON(req->sk); | |
324 | ||
325 | if (seq != tcp_rsk(req)->snt_isn) { | |
326 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); | |
c6973669 | 327 | reqsk_put(req); |
26e37360 ED |
328 | } else { |
329 | /* | |
330 | * Still in SYN_RECV, just remove it silently. | |
331 | * There is no good way to pass the error to the newly | |
332 | * created socket, and POSIX does not want network | |
333 | * errors returned from accept(). | |
334 | */ | |
26e37360 | 335 | NET_INC_STATS_BH(net, LINUX_MIB_LISTENDROPS); |
c6973669 | 336 | inet_csk_reqsk_queue_drop(req->rsk_listener, req); |
26e37360 | 337 | } |
26e37360 ED |
338 | } |
339 | EXPORT_SYMBOL(tcp_req_err); | |
340 | ||
1da177e4 LT |
341 | /* |
342 | * This routine is called by the ICMP module when it gets some | |
343 | * sort of error condition. If err < 0 then the socket should | |
344 | * be closed and the error returned to the user. If err > 0 | |
345 | * it's just the icmp type << 8 | icmp code. After adjustment | |
346 | * header points to the first 8 bytes of the tcp header. We need | |
347 | * to find the appropriate port. | |
348 | * | |
349 | * The locking strategy used here is very "optimistic". When | |
350 | * someone else accesses the socket the ICMP is just dropped | |
351 | * and for some paths there is no check at all. | |
352 | * A more general error queue to queue errors for later handling | |
353 | * is probably better. | |
354 | * | |
355 | */ | |
356 | ||
4d1a2d9e | 357 | void tcp_v4_err(struct sk_buff *icmp_skb, u32 info) |
1da177e4 | 358 | { |
b71d1d42 | 359 | const struct iphdr *iph = (const struct iphdr *)icmp_skb->data; |
4d1a2d9e | 360 | struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2)); |
f1ecd5d9 | 361 | struct inet_connection_sock *icsk; |
1da177e4 LT |
362 | struct tcp_sock *tp; |
363 | struct inet_sock *inet; | |
4d1a2d9e DL |
364 | const int type = icmp_hdr(icmp_skb)->type; |
365 | const int code = icmp_hdr(icmp_skb)->code; | |
1da177e4 | 366 | struct sock *sk; |
f1ecd5d9 | 367 | struct sk_buff *skb; |
0a672f74 YC |
368 | struct request_sock *fastopen; |
369 | __u32 seq, snd_una; | |
f1ecd5d9 | 370 | __u32 remaining; |
1da177e4 | 371 | int err; |
4d1a2d9e | 372 | struct net *net = dev_net(icmp_skb->dev); |
1da177e4 | 373 | |
26e37360 ED |
374 | sk = __inet_lookup_established(net, &tcp_hashinfo, iph->daddr, |
375 | th->dest, iph->saddr, ntohs(th->source), | |
376 | inet_iif(icmp_skb)); | |
1da177e4 | 377 | if (!sk) { |
dcfc23ca | 378 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
379 | return; |
380 | } | |
381 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 382 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
383 | return; |
384 | } | |
26e37360 ED |
385 | seq = ntohl(th->seq); |
386 | if (sk->sk_state == TCP_NEW_SYN_RECV) | |
387 | return tcp_req_err(sk, seq); | |
1da177e4 LT |
388 | |
389 | bh_lock_sock(sk); | |
390 | /* If too many ICMPs get dropped on busy | |
391 | * servers this needs to be solved differently. | |
563d34d0 ED |
392 | * We do take care of PMTU discovery (RFC1191) special case : |
393 | * we can receive locally generated ICMP messages while socket is held. | |
1da177e4 | 394 | */ |
b74aa930 ED |
395 | if (sock_owned_by_user(sk)) { |
396 | if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)) | |
397 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); | |
398 | } | |
1da177e4 LT |
399 | if (sk->sk_state == TCP_CLOSE) |
400 | goto out; | |
401 | ||
97e3ecd1 | 402 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
403 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
404 | goto out; | |
405 | } | |
406 | ||
f1ecd5d9 | 407 | icsk = inet_csk(sk); |
1da177e4 | 408 | tp = tcp_sk(sk); |
0a672f74 YC |
409 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ |
410 | fastopen = tp->fastopen_rsk; | |
411 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; | |
1da177e4 | 412 | if (sk->sk_state != TCP_LISTEN && |
0a672f74 | 413 | !between(seq, snd_una, tp->snd_nxt)) { |
de0744af | 414 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
415 | goto out; |
416 | } | |
417 | ||
418 | switch (type) { | |
55be7a9c DM |
419 | case ICMP_REDIRECT: |
420 | do_redirect(icmp_skb, sk); | |
421 | goto out; | |
1da177e4 LT |
422 | case ICMP_SOURCE_QUENCH: |
423 | /* Just silently ignore these. */ | |
424 | goto out; | |
425 | case ICMP_PARAMETERPROB: | |
426 | err = EPROTO; | |
427 | break; | |
428 | case ICMP_DEST_UNREACH: | |
429 | if (code > NR_ICMP_UNREACH) | |
430 | goto out; | |
431 | ||
432 | if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */ | |
0d4f0608 ED |
433 | /* We are not interested in TCP_LISTEN and open_requests |
434 | * (SYN-ACKs send out by Linux are always <576bytes so | |
435 | * they should go through unfragmented). | |
436 | */ | |
437 | if (sk->sk_state == TCP_LISTEN) | |
438 | goto out; | |
439 | ||
563d34d0 | 440 | tp->mtu_info = info; |
144d56e9 | 441 | if (!sock_owned_by_user(sk)) { |
563d34d0 | 442 | tcp_v4_mtu_reduced(sk); |
144d56e9 ED |
443 | } else { |
444 | if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &tp->tsq_flags)) | |
445 | sock_hold(sk); | |
446 | } | |
1da177e4 LT |
447 | goto out; |
448 | } | |
449 | ||
450 | err = icmp_err_convert[code].errno; | |
f1ecd5d9 DL |
451 | /* check if icmp_skb allows revert of backoff |
452 | * (see draft-zimmermann-tcp-lcd) */ | |
453 | if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH) | |
454 | break; | |
455 | if (seq != tp->snd_una || !icsk->icsk_retransmits || | |
0a672f74 | 456 | !icsk->icsk_backoff || fastopen) |
f1ecd5d9 DL |
457 | break; |
458 | ||
8f49c270 DM |
459 | if (sock_owned_by_user(sk)) |
460 | break; | |
461 | ||
f1ecd5d9 | 462 | icsk->icsk_backoff--; |
fcdd1cf4 ED |
463 | icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) : |
464 | TCP_TIMEOUT_INIT; | |
465 | icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX); | |
f1ecd5d9 DL |
466 | |
467 | skb = tcp_write_queue_head(sk); | |
468 | BUG_ON(!skb); | |
469 | ||
7faee5c0 ED |
470 | remaining = icsk->icsk_rto - |
471 | min(icsk->icsk_rto, | |
472 | tcp_time_stamp - tcp_skb_timestamp(skb)); | |
f1ecd5d9 DL |
473 | |
474 | if (remaining) { | |
475 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, | |
476 | remaining, TCP_RTO_MAX); | |
f1ecd5d9 DL |
477 | } else { |
478 | /* RTO revert clocked out retransmission. | |
479 | * Will retransmit now */ | |
480 | tcp_retransmit_timer(sk); | |
481 | } | |
482 | ||
1da177e4 LT |
483 | break; |
484 | case ICMP_TIME_EXCEEDED: | |
485 | err = EHOSTUNREACH; | |
486 | break; | |
487 | default: | |
488 | goto out; | |
489 | } | |
490 | ||
491 | switch (sk->sk_state) { | |
1da177e4 | 492 | case TCP_SYN_SENT: |
0a672f74 YC |
493 | case TCP_SYN_RECV: |
494 | /* Only in fast or simultaneous open. If a fast open socket is | |
495 | * is already accepted it is treated as a connected one below. | |
496 | */ | |
497 | if (fastopen && fastopen->sk == NULL) | |
498 | break; | |
499 | ||
1da177e4 | 500 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
501 | sk->sk_err = err; |
502 | ||
503 | sk->sk_error_report(sk); | |
504 | ||
505 | tcp_done(sk); | |
506 | } else { | |
507 | sk->sk_err_soft = err; | |
508 | } | |
509 | goto out; | |
510 | } | |
511 | ||
512 | /* If we've already connected we will keep trying | |
513 | * until we time out, or the user gives up. | |
514 | * | |
515 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
516 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
517 | * but it is obsoleted by pmtu discovery). | |
518 | * | |
519 | * Note, that in modern internet, where routing is unreliable | |
520 | * and in each dark corner broken firewalls sit, sending random | |
521 | * errors ordered by their masters even this two messages finally lose | |
522 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
523 | * | |
524 | * Now we are in compliance with RFCs. | |
525 | * --ANK (980905) | |
526 | */ | |
527 | ||
528 | inet = inet_sk(sk); | |
529 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
530 | sk->sk_err = err; | |
531 | sk->sk_error_report(sk); | |
532 | } else { /* Only an error on timeout */ | |
533 | sk->sk_err_soft = err; | |
534 | } | |
535 | ||
536 | out: | |
537 | bh_unlock_sock(sk); | |
538 | sock_put(sk); | |
539 | } | |
540 | ||
28850dc7 | 541 | void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 542 | { |
aa8223c7 | 543 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 544 | |
84fa7933 | 545 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
419f9f89 | 546 | th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0); |
663ead3b | 547 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 548 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 549 | } else { |
419f9f89 | 550 | th->check = tcp_v4_check(skb->len, saddr, daddr, |
07f0757a | 551 | csum_partial(th, |
1da177e4 LT |
552 | th->doff << 2, |
553 | skb->csum)); | |
554 | } | |
555 | } | |
556 | ||
419f9f89 | 557 | /* This routine computes an IPv4 TCP checksum. */ |
bb296246 | 558 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb) |
419f9f89 | 559 | { |
cf533ea5 | 560 | const struct inet_sock *inet = inet_sk(sk); |
419f9f89 HX |
561 | |
562 | __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr); | |
563 | } | |
4bc2f18b | 564 | EXPORT_SYMBOL(tcp_v4_send_check); |
419f9f89 | 565 | |
1da177e4 LT |
566 | /* |
567 | * This routine will send an RST to the other tcp. | |
568 | * | |
569 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
570 | * for reset. | |
571 | * Answer: if a packet caused RST, it is not for a socket | |
572 | * existing in our system, if it is matched to a socket, | |
573 | * it is just duplicate segment or bug in other side's TCP. | |
574 | * So that we build reply only basing on parameters | |
575 | * arrived with segment. | |
576 | * Exception: precedence violation. We do not implement it in any case. | |
577 | */ | |
578 | ||
cfb6eeb4 | 579 | static void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 580 | { |
cf533ea5 | 581 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
582 | struct { |
583 | struct tcphdr th; | |
584 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 585 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
586 | #endif |
587 | } rep; | |
1da177e4 | 588 | struct ip_reply_arg arg; |
cfb6eeb4 YH |
589 | #ifdef CONFIG_TCP_MD5SIG |
590 | struct tcp_md5sig_key *key; | |
658ddaaf SL |
591 | const __u8 *hash_location = NULL; |
592 | unsigned char newhash[16]; | |
593 | int genhash; | |
594 | struct sock *sk1 = NULL; | |
cfb6eeb4 | 595 | #endif |
a86b1e30 | 596 | struct net *net; |
1da177e4 LT |
597 | |
598 | /* Never send a reset in response to a reset. */ | |
599 | if (th->rst) | |
600 | return; | |
601 | ||
c3658e8d ED |
602 | /* If sk not NULL, it means we did a successful lookup and incoming |
603 | * route had to be correct. prequeue might have dropped our dst. | |
604 | */ | |
605 | if (!sk && skb_rtable(skb)->rt_type != RTN_LOCAL) | |
1da177e4 LT |
606 | return; |
607 | ||
608 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
609 | memset(&rep, 0, sizeof(rep)); |
610 | rep.th.dest = th->source; | |
611 | rep.th.source = th->dest; | |
612 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
613 | rep.th.rst = 1; | |
1da177e4 LT |
614 | |
615 | if (th->ack) { | |
cfb6eeb4 | 616 | rep.th.seq = th->ack_seq; |
1da177e4 | 617 | } else { |
cfb6eeb4 YH |
618 | rep.th.ack = 1; |
619 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
620 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
621 | } |
622 | ||
7174259e | 623 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
624 | arg.iov[0].iov_base = (unsigned char *)&rep; |
625 | arg.iov[0].iov_len = sizeof(rep.th); | |
626 | ||
0f85feae | 627 | net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); |
cfb6eeb4 | 628 | #ifdef CONFIG_TCP_MD5SIG |
658ddaaf SL |
629 | hash_location = tcp_parse_md5sig_option(th); |
630 | if (!sk && hash_location) { | |
631 | /* | |
632 | * active side is lost. Try to find listening socket through | |
633 | * source port, and then find md5 key through listening socket. | |
634 | * we are not loose security here: | |
635 | * Incoming packet is checked with md5 hash with finding key, | |
636 | * no RST generated if md5 hash doesn't match. | |
637 | */ | |
0f85feae | 638 | sk1 = __inet_lookup_listener(net, |
da5e3630 TH |
639 | &tcp_hashinfo, ip_hdr(skb)->saddr, |
640 | th->source, ip_hdr(skb)->daddr, | |
658ddaaf SL |
641 | ntohs(th->source), inet_iif(skb)); |
642 | /* don't send rst if it can't find key */ | |
643 | if (!sk1) | |
644 | return; | |
645 | rcu_read_lock(); | |
646 | key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *) | |
647 | &ip_hdr(skb)->saddr, AF_INET); | |
648 | if (!key) | |
649 | goto release_sk1; | |
650 | ||
39f8e58e | 651 | genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb); |
658ddaaf SL |
652 | if (genhash || memcmp(hash_location, newhash, 16) != 0) |
653 | goto release_sk1; | |
654 | } else { | |
655 | key = sk ? tcp_md5_do_lookup(sk, (union tcp_md5_addr *) | |
656 | &ip_hdr(skb)->saddr, | |
657 | AF_INET) : NULL; | |
658 | } | |
659 | ||
cfb6eeb4 YH |
660 | if (key) { |
661 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
662 | (TCPOPT_NOP << 16) | | |
663 | (TCPOPT_MD5SIG << 8) | | |
664 | TCPOLEN_MD5SIG); | |
665 | /* Update length and the length the header thinks exists */ | |
666 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
667 | rep.th.doff = arg.iov[0].iov_len / 4; | |
668 | ||
49a72dfb | 669 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
670 | key, ip_hdr(skb)->saddr, |
671 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
672 | } |
673 | #endif | |
eddc9ec5 ACM |
674 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
675 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 676 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 677 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
88ef4a5a | 678 | arg.flags = (sk && inet_sk(sk)->transparent) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
e2446eaa | 679 | /* When socket is gone, all binding information is lost. |
4c675258 AK |
680 | * routing might fail in this case. No choice here, if we choose to force |
681 | * input interface, we will misroute in case of asymmetric route. | |
e2446eaa | 682 | */ |
4c675258 AK |
683 | if (sk) |
684 | arg.bound_dev_if = sk->sk_bound_dev_if; | |
1da177e4 | 685 | |
66b13d99 | 686 | arg.tos = ip_hdr(skb)->tos; |
bdbbb852 ED |
687 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
688 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
689 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
690 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 691 | |
63231bdd PE |
692 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
693 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
658ddaaf SL |
694 | |
695 | #ifdef CONFIG_TCP_MD5SIG | |
696 | release_sk1: | |
697 | if (sk1) { | |
698 | rcu_read_unlock(); | |
699 | sock_put(sk1); | |
700 | } | |
701 | #endif | |
1da177e4 LT |
702 | } |
703 | ||
704 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
705 | outside socket context is ugly, certainly. What can I do? | |
706 | */ | |
707 | ||
9501f972 | 708 | static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack, |
ee684b6f | 709 | u32 win, u32 tsval, u32 tsecr, int oif, |
88ef4a5a | 710 | struct tcp_md5sig_key *key, |
66b13d99 | 711 | int reply_flags, u8 tos) |
1da177e4 | 712 | { |
cf533ea5 | 713 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
714 | struct { |
715 | struct tcphdr th; | |
714e85be | 716 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 717 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 718 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
719 | #endif |
720 | ]; | |
1da177e4 LT |
721 | } rep; |
722 | struct ip_reply_arg arg; | |
adf30907 | 723 | struct net *net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
724 | |
725 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 726 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
727 | |
728 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
729 | arg.iov[0].iov_len = sizeof(rep.th); | |
ee684b6f | 730 | if (tsecr) { |
cfb6eeb4 YH |
731 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
732 | (TCPOPT_TIMESTAMP << 8) | | |
733 | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
734 | rep.opt[1] = htonl(tsval); |
735 | rep.opt[2] = htonl(tsecr); | |
cb48cfe8 | 736 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
737 | } |
738 | ||
739 | /* Swap the send and the receive. */ | |
740 | rep.th.dest = th->source; | |
741 | rep.th.source = th->dest; | |
742 | rep.th.doff = arg.iov[0].iov_len / 4; | |
743 | rep.th.seq = htonl(seq); | |
744 | rep.th.ack_seq = htonl(ack); | |
745 | rep.th.ack = 1; | |
746 | rep.th.window = htons(win); | |
747 | ||
cfb6eeb4 | 748 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 749 | if (key) { |
ee684b6f | 750 | int offset = (tsecr) ? 3 : 0; |
cfb6eeb4 YH |
751 | |
752 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
753 | (TCPOPT_NOP << 16) | | |
754 | (TCPOPT_MD5SIG << 8) | | |
755 | TCPOLEN_MD5SIG); | |
756 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
757 | rep.th.doff = arg.iov[0].iov_len/4; | |
758 | ||
49a72dfb | 759 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
760 | key, ip_hdr(skb)->saddr, |
761 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
762 | } |
763 | #endif | |
88ef4a5a | 764 | arg.flags = reply_flags; |
eddc9ec5 ACM |
765 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
766 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
767 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
768 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
769 | if (oif) |
770 | arg.bound_dev_if = oif; | |
66b13d99 | 771 | arg.tos = tos; |
bdbbb852 ED |
772 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
773 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
774 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
775 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 776 | |
63231bdd | 777 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
1da177e4 LT |
778 | } |
779 | ||
780 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
781 | { | |
8feaf0c0 | 782 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 783 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 784 | |
9501f972 | 785 | tcp_v4_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 786 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
ee684b6f | 787 | tcp_time_stamp + tcptw->tw_ts_offset, |
9501f972 YH |
788 | tcptw->tw_ts_recent, |
789 | tw->tw_bound_dev_if, | |
88ef4a5a | 790 | tcp_twsk_md5_key(tcptw), |
66b13d99 ED |
791 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0, |
792 | tw->tw_tos | |
9501f972 | 793 | ); |
1da177e4 | 794 | |
8feaf0c0 | 795 | inet_twsk_put(tw); |
1da177e4 LT |
796 | } |
797 | ||
6edafaaf | 798 | static void tcp_v4_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
7174259e | 799 | struct request_sock *req) |
1da177e4 | 800 | { |
168a8f58 JC |
801 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
802 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
803 | */ | |
804 | tcp_v4_send_ack(skb, (sk->sk_state == TCP_LISTEN) ? | |
805 | tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt, | |
806 | tcp_rsk(req)->rcv_nxt, req->rcv_wnd, | |
ee684b6f | 807 | tcp_time_stamp, |
9501f972 YH |
808 | req->ts_recent, |
809 | 0, | |
a915da9b ED |
810 | tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr, |
811 | AF_INET), | |
66b13d99 ED |
812 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0, |
813 | ip_hdr(skb)->tos); | |
1da177e4 LT |
814 | } |
815 | ||
1da177e4 | 816 | /* |
9bf1d83e | 817 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 818 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
819 | * socket. |
820 | */ | |
72659ecc | 821 | static int tcp_v4_send_synack(struct sock *sk, struct dst_entry *dst, |
d6274bd8 | 822 | struct flowi *fl, |
72659ecc | 823 | struct request_sock *req, |
843f4a55 YC |
824 | u16 queue_mapping, |
825 | struct tcp_fastopen_cookie *foc) | |
1da177e4 | 826 | { |
2e6599cb | 827 | const struct inet_request_sock *ireq = inet_rsk(req); |
6bd023f3 | 828 | struct flowi4 fl4; |
1da177e4 | 829 | int err = -1; |
d41db5af | 830 | struct sk_buff *skb; |
1da177e4 LT |
831 | |
832 | /* First, grab a route. */ | |
ba3f7f04 | 833 | if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL) |
fd80eb94 | 834 | return -1; |
1da177e4 | 835 | |
843f4a55 | 836 | skb = tcp_make_synack(sk, dst, req, foc); |
1da177e4 LT |
837 | |
838 | if (skb) { | |
634fb979 | 839 | __tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr); |
1da177e4 | 840 | |
fff32699 | 841 | skb_set_queue_mapping(skb, queue_mapping); |
634fb979 ED |
842 | err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr, |
843 | ireq->ir_rmt_addr, | |
2e6599cb | 844 | ireq->opt); |
b9df3cb8 | 845 | err = net_xmit_eval(err); |
1da177e4 LT |
846 | } |
847 | ||
1da177e4 LT |
848 | return err; |
849 | } | |
850 | ||
851 | /* | |
60236fdd | 852 | * IPv4 request_sock destructor. |
1da177e4 | 853 | */ |
60236fdd | 854 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 855 | { |
a51482bd | 856 | kfree(inet_rsk(req)->opt); |
1da177e4 LT |
857 | } |
858 | ||
946cedcc | 859 | /* |
a2a385d6 | 860 | * Return true if a syncookie should be sent |
946cedcc | 861 | */ |
a2a385d6 | 862 | bool tcp_syn_flood_action(struct sock *sk, |
946cedcc ED |
863 | const struct sk_buff *skb, |
864 | const char *proto) | |
1da177e4 | 865 | { |
946cedcc | 866 | const char *msg = "Dropping request"; |
a2a385d6 | 867 | bool want_cookie = false; |
946cedcc ED |
868 | struct listen_sock *lopt; |
869 | ||
2a1d4bd4 | 870 | #ifdef CONFIG_SYN_COOKIES |
946cedcc | 871 | if (sysctl_tcp_syncookies) { |
2a1d4bd4 | 872 | msg = "Sending cookies"; |
a2a385d6 | 873 | want_cookie = true; |
946cedcc ED |
874 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES); |
875 | } else | |
80e40daa | 876 | #endif |
946cedcc ED |
877 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP); |
878 | ||
879 | lopt = inet_csk(sk)->icsk_accept_queue.listen_opt; | |
5ad37d5d | 880 | if (!lopt->synflood_warned && sysctl_tcp_syncookies != 2) { |
946cedcc | 881 | lopt->synflood_warned = 1; |
afd46503 | 882 | pr_info("%s: Possible SYN flooding on port %d. %s. Check SNMP counters.\n", |
946cedcc ED |
883 | proto, ntohs(tcp_hdr(skb)->dest), msg); |
884 | } | |
885 | return want_cookie; | |
2a1d4bd4 | 886 | } |
946cedcc | 887 | EXPORT_SYMBOL(tcp_syn_flood_action); |
1da177e4 | 888 | |
cfb6eeb4 YH |
889 | #ifdef CONFIG_TCP_MD5SIG |
890 | /* | |
891 | * RFC2385 MD5 checksumming requires a mapping of | |
892 | * IP address->MD5 Key. | |
893 | * We need to maintain these in the sk structure. | |
894 | */ | |
895 | ||
896 | /* Find the Key structure for an address. */ | |
a915da9b ED |
897 | struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, |
898 | const union tcp_md5_addr *addr, | |
899 | int family) | |
cfb6eeb4 | 900 | { |
fd3a154a | 901 | const struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 902 | struct tcp_md5sig_key *key; |
a915da9b | 903 | unsigned int size = sizeof(struct in_addr); |
fd3a154a | 904 | const struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 905 | |
a8afca03 ED |
906 | /* caller either holds rcu_read_lock() or socket lock */ |
907 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
b4fb05ea ED |
908 | sock_owned_by_user(sk) || |
909 | lockdep_is_held(&sk->sk_lock.slock)); | |
a8afca03 | 910 | if (!md5sig) |
cfb6eeb4 | 911 | return NULL; |
a915da9b ED |
912 | #if IS_ENABLED(CONFIG_IPV6) |
913 | if (family == AF_INET6) | |
914 | size = sizeof(struct in6_addr); | |
915 | #endif | |
b67bfe0d | 916 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { |
a915da9b ED |
917 | if (key->family != family) |
918 | continue; | |
919 | if (!memcmp(&key->addr, addr, size)) | |
920 | return key; | |
cfb6eeb4 YH |
921 | } |
922 | return NULL; | |
923 | } | |
a915da9b | 924 | EXPORT_SYMBOL(tcp_md5_do_lookup); |
cfb6eeb4 YH |
925 | |
926 | struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
fd3a154a | 927 | const struct sock *addr_sk) |
cfb6eeb4 | 928 | { |
a915da9b ED |
929 | union tcp_md5_addr *addr; |
930 | ||
fd3a154a | 931 | addr = (union tcp_md5_addr *)&sk->sk_daddr; |
a915da9b | 932 | return tcp_md5_do_lookup(sk, addr, AF_INET); |
cfb6eeb4 | 933 | } |
cfb6eeb4 YH |
934 | EXPORT_SYMBOL(tcp_v4_md5_lookup); |
935 | ||
cfb6eeb4 | 936 | /* This can be called on a newly created socket, from other files */ |
a915da9b ED |
937 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
938 | int family, const u8 *newkey, u8 newkeylen, gfp_t gfp) | |
cfb6eeb4 YH |
939 | { |
940 | /* Add Key to the list */ | |
b0a713e9 | 941 | struct tcp_md5sig_key *key; |
cfb6eeb4 | 942 | struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 943 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 944 | |
c0353c7b | 945 | key = tcp_md5_do_lookup(sk, addr, family); |
cfb6eeb4 YH |
946 | if (key) { |
947 | /* Pre-existing entry - just update that one. */ | |
a915da9b | 948 | memcpy(key->key, newkey, newkeylen); |
b0a713e9 | 949 | key->keylen = newkeylen; |
a915da9b ED |
950 | return 0; |
951 | } | |
260fcbeb | 952 | |
a8afca03 ED |
953 | md5sig = rcu_dereference_protected(tp->md5sig_info, |
954 | sock_owned_by_user(sk)); | |
a915da9b ED |
955 | if (!md5sig) { |
956 | md5sig = kmalloc(sizeof(*md5sig), gfp); | |
957 | if (!md5sig) | |
cfb6eeb4 | 958 | return -ENOMEM; |
cfb6eeb4 | 959 | |
a915da9b ED |
960 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
961 | INIT_HLIST_HEAD(&md5sig->head); | |
a8afca03 | 962 | rcu_assign_pointer(tp->md5sig_info, md5sig); |
a915da9b | 963 | } |
cfb6eeb4 | 964 | |
5f3d9cb2 | 965 | key = sock_kmalloc(sk, sizeof(*key), gfp); |
a915da9b ED |
966 | if (!key) |
967 | return -ENOMEM; | |
71cea17e | 968 | if (!tcp_alloc_md5sig_pool()) { |
5f3d9cb2 | 969 | sock_kfree_s(sk, key, sizeof(*key)); |
a915da9b | 970 | return -ENOMEM; |
cfb6eeb4 | 971 | } |
a915da9b ED |
972 | |
973 | memcpy(key->key, newkey, newkeylen); | |
974 | key->keylen = newkeylen; | |
975 | key->family = family; | |
976 | memcpy(&key->addr, addr, | |
977 | (family == AF_INET6) ? sizeof(struct in6_addr) : | |
978 | sizeof(struct in_addr)); | |
979 | hlist_add_head_rcu(&key->node, &md5sig->head); | |
cfb6eeb4 YH |
980 | return 0; |
981 | } | |
a915da9b | 982 | EXPORT_SYMBOL(tcp_md5_do_add); |
cfb6eeb4 | 983 | |
a915da9b | 984 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family) |
cfb6eeb4 | 985 | { |
a915da9b ED |
986 | struct tcp_md5sig_key *key; |
987 | ||
c0353c7b | 988 | key = tcp_md5_do_lookup(sk, addr, family); |
a915da9b ED |
989 | if (!key) |
990 | return -ENOENT; | |
991 | hlist_del_rcu(&key->node); | |
5f3d9cb2 | 992 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 993 | kfree_rcu(key, rcu); |
a915da9b | 994 | return 0; |
cfb6eeb4 | 995 | } |
a915da9b | 996 | EXPORT_SYMBOL(tcp_md5_do_del); |
cfb6eeb4 | 997 | |
e0683e70 | 998 | static void tcp_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
999 | { |
1000 | struct tcp_sock *tp = tcp_sk(sk); | |
a915da9b | 1001 | struct tcp_md5sig_key *key; |
b67bfe0d | 1002 | struct hlist_node *n; |
a8afca03 | 1003 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 1004 | |
a8afca03 ED |
1005 | md5sig = rcu_dereference_protected(tp->md5sig_info, 1); |
1006 | ||
b67bfe0d | 1007 | hlist_for_each_entry_safe(key, n, &md5sig->head, node) { |
a915da9b | 1008 | hlist_del_rcu(&key->node); |
5f3d9cb2 | 1009 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1010 | kfree_rcu(key, rcu); |
cfb6eeb4 YH |
1011 | } |
1012 | } | |
1013 | ||
7174259e ACM |
1014 | static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval, |
1015 | int optlen) | |
cfb6eeb4 YH |
1016 | { |
1017 | struct tcp_md5sig cmd; | |
1018 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
cfb6eeb4 YH |
1019 | |
1020 | if (optlen < sizeof(cmd)) | |
1021 | return -EINVAL; | |
1022 | ||
7174259e | 1023 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
1024 | return -EFAULT; |
1025 | ||
1026 | if (sin->sin_family != AF_INET) | |
1027 | return -EINVAL; | |
1028 | ||
64a124ed | 1029 | if (!cmd.tcpm_keylen) |
a915da9b ED |
1030 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
1031 | AF_INET); | |
cfb6eeb4 YH |
1032 | |
1033 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
1034 | return -EINVAL; | |
1035 | ||
a915da9b ED |
1036 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
1037 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, | |
1038 | GFP_KERNEL); | |
cfb6eeb4 YH |
1039 | } |
1040 | ||
49a72dfb AL |
1041 | static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
1042 | __be32 daddr, __be32 saddr, int nbytes) | |
cfb6eeb4 | 1043 | { |
cfb6eeb4 | 1044 | struct tcp4_pseudohdr *bp; |
49a72dfb | 1045 | struct scatterlist sg; |
cfb6eeb4 YH |
1046 | |
1047 | bp = &hp->md5_blk.ip4; | |
cfb6eeb4 YH |
1048 | |
1049 | /* | |
49a72dfb | 1050 | * 1. the TCP pseudo-header (in the order: source IP address, |
cfb6eeb4 YH |
1051 | * destination IP address, zero-padded protocol number, and |
1052 | * segment length) | |
1053 | */ | |
1054 | bp->saddr = saddr; | |
1055 | bp->daddr = daddr; | |
1056 | bp->pad = 0; | |
076fb722 | 1057 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1058 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1059 | |
49a72dfb AL |
1060 | sg_init_one(&sg, bp, sizeof(*bp)); |
1061 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
1062 | } | |
1063 | ||
a915da9b | 1064 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 1065 | __be32 daddr, __be32 saddr, const struct tcphdr *th) |
49a72dfb AL |
1066 | { |
1067 | struct tcp_md5sig_pool *hp; | |
1068 | struct hash_desc *desc; | |
1069 | ||
1070 | hp = tcp_get_md5sig_pool(); | |
1071 | if (!hp) | |
1072 | goto clear_hash_noput; | |
1073 | desc = &hp->md5_desc; | |
1074 | ||
1075 | if (crypto_hash_init(desc)) | |
1076 | goto clear_hash; | |
1077 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
1078 | goto clear_hash; | |
1079 | if (tcp_md5_hash_header(hp, th)) | |
1080 | goto clear_hash; | |
1081 | if (tcp_md5_hash_key(hp, key)) | |
1082 | goto clear_hash; | |
1083 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 YH |
1084 | goto clear_hash; |
1085 | ||
cfb6eeb4 | 1086 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1087 | return 0; |
49a72dfb | 1088 | |
cfb6eeb4 YH |
1089 | clear_hash: |
1090 | tcp_put_md5sig_pool(); | |
1091 | clear_hash_noput: | |
1092 | memset(md5_hash, 0, 16); | |
49a72dfb | 1093 | return 1; |
cfb6eeb4 YH |
1094 | } |
1095 | ||
39f8e58e ED |
1096 | int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, |
1097 | const struct sock *sk, | |
318cf7aa | 1098 | const struct sk_buff *skb) |
cfb6eeb4 | 1099 | { |
49a72dfb AL |
1100 | struct tcp_md5sig_pool *hp; |
1101 | struct hash_desc *desc; | |
318cf7aa | 1102 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
1103 | __be32 saddr, daddr; |
1104 | ||
39f8e58e ED |
1105 | if (sk) { /* valid for establish/request sockets */ |
1106 | saddr = sk->sk_rcv_saddr; | |
1107 | daddr = sk->sk_daddr; | |
cfb6eeb4 | 1108 | } else { |
49a72dfb AL |
1109 | const struct iphdr *iph = ip_hdr(skb); |
1110 | saddr = iph->saddr; | |
1111 | daddr = iph->daddr; | |
cfb6eeb4 | 1112 | } |
49a72dfb AL |
1113 | |
1114 | hp = tcp_get_md5sig_pool(); | |
1115 | if (!hp) | |
1116 | goto clear_hash_noput; | |
1117 | desc = &hp->md5_desc; | |
1118 | ||
1119 | if (crypto_hash_init(desc)) | |
1120 | goto clear_hash; | |
1121 | ||
1122 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
1123 | goto clear_hash; | |
1124 | if (tcp_md5_hash_header(hp, th)) | |
1125 | goto clear_hash; | |
1126 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1127 | goto clear_hash; | |
1128 | if (tcp_md5_hash_key(hp, key)) | |
1129 | goto clear_hash; | |
1130 | if (crypto_hash_final(desc, md5_hash)) | |
1131 | goto clear_hash; | |
1132 | ||
1133 | tcp_put_md5sig_pool(); | |
1134 | return 0; | |
1135 | ||
1136 | clear_hash: | |
1137 | tcp_put_md5sig_pool(); | |
1138 | clear_hash_noput: | |
1139 | memset(md5_hash, 0, 16); | |
1140 | return 1; | |
cfb6eeb4 | 1141 | } |
49a72dfb | 1142 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1143 | |
ff74e23f ED |
1144 | /* Called with rcu_read_lock() */ |
1145 | static bool tcp_v4_inbound_md5_hash(struct sock *sk, | |
1146 | const struct sk_buff *skb) | |
cfb6eeb4 YH |
1147 | { |
1148 | /* | |
1149 | * This gets called for each TCP segment that arrives | |
1150 | * so we want to be efficient. | |
1151 | * We have 3 drop cases: | |
1152 | * o No MD5 hash and one expected. | |
1153 | * o MD5 hash and we're not expecting one. | |
1154 | * o MD5 hash and its wrong. | |
1155 | */ | |
cf533ea5 | 1156 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 1157 | struct tcp_md5sig_key *hash_expected; |
eddc9ec5 | 1158 | const struct iphdr *iph = ip_hdr(skb); |
cf533ea5 | 1159 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1160 | int genhash; |
cfb6eeb4 YH |
1161 | unsigned char newhash[16]; |
1162 | ||
a915da9b ED |
1163 | hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr, |
1164 | AF_INET); | |
7d5d5525 | 1165 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1166 | |
cfb6eeb4 YH |
1167 | /* We've parsed the options - do we have a hash? */ |
1168 | if (!hash_expected && !hash_location) | |
a2a385d6 | 1169 | return false; |
cfb6eeb4 YH |
1170 | |
1171 | if (hash_expected && !hash_location) { | |
785957d3 | 1172 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
a2a385d6 | 1173 | return true; |
cfb6eeb4 YH |
1174 | } |
1175 | ||
1176 | if (!hash_expected && hash_location) { | |
785957d3 | 1177 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
a2a385d6 | 1178 | return true; |
cfb6eeb4 YH |
1179 | } |
1180 | ||
1181 | /* Okay, so this is hash_expected and hash_location - | |
1182 | * so we need to calculate the checksum. | |
1183 | */ | |
49a72dfb AL |
1184 | genhash = tcp_v4_md5_hash_skb(newhash, |
1185 | hash_expected, | |
39f8e58e | 1186 | NULL, skb); |
cfb6eeb4 YH |
1187 | |
1188 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
e87cc472 JP |
1189 | net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1190 | &iph->saddr, ntohs(th->source), | |
1191 | &iph->daddr, ntohs(th->dest), | |
1192 | genhash ? " tcp_v4_calc_md5_hash failed" | |
1193 | : ""); | |
a2a385d6 | 1194 | return true; |
cfb6eeb4 | 1195 | } |
a2a385d6 | 1196 | return false; |
cfb6eeb4 | 1197 | } |
cfb6eeb4 YH |
1198 | #endif |
1199 | ||
08d2cc3b | 1200 | static void tcp_v4_init_req(struct request_sock *req, struct sock *sk_listener, |
16bea70a OP |
1201 | struct sk_buff *skb) |
1202 | { | |
1203 | struct inet_request_sock *ireq = inet_rsk(req); | |
1204 | ||
08d2cc3b ED |
1205 | sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr); |
1206 | sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr); | |
1207 | ireq->no_srccheck = inet_sk(sk_listener)->transparent; | |
16bea70a | 1208 | ireq->opt = tcp_v4_save_options(skb); |
3f66b083 | 1209 | ireq->ireq_family = AF_INET; |
16bea70a OP |
1210 | } |
1211 | ||
d94e0417 OP |
1212 | static struct dst_entry *tcp_v4_route_req(struct sock *sk, struct flowi *fl, |
1213 | const struct request_sock *req, | |
1214 | bool *strict) | |
1215 | { | |
1216 | struct dst_entry *dst = inet_csk_route_req(sk, &fl->u.ip4, req); | |
1217 | ||
1218 | if (strict) { | |
1219 | if (fl->u.ip4.daddr == inet_rsk(req)->ir_rmt_addr) | |
1220 | *strict = true; | |
1221 | else | |
1222 | *strict = false; | |
1223 | } | |
1224 | ||
1225 | return dst; | |
1226 | } | |
1227 | ||
72a3effa | 1228 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1229 | .family = PF_INET, |
2e6599cb | 1230 | .obj_size = sizeof(struct tcp_request_sock), |
5db92c99 | 1231 | .rtx_syn_ack = tcp_rtx_synack, |
60236fdd ACM |
1232 | .send_ack = tcp_v4_reqsk_send_ack, |
1233 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 | 1234 | .send_reset = tcp_v4_send_reset, |
688d1945 | 1235 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
1236 | }; |
1237 | ||
b2e4b3de | 1238 | static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
2aec4a29 | 1239 | .mss_clamp = TCP_MSS_DEFAULT, |
16bea70a | 1240 | #ifdef CONFIG_TCP_MD5SIG |
fd3a154a | 1241 | .req_md5_lookup = tcp_v4_md5_lookup, |
e3afe7b7 | 1242 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
b6332e6c | 1243 | #endif |
16bea70a | 1244 | .init_req = tcp_v4_init_req, |
fb7b37a7 OP |
1245 | #ifdef CONFIG_SYN_COOKIES |
1246 | .cookie_init_seq = cookie_v4_init_sequence, | |
1247 | #endif | |
d94e0417 | 1248 | .route_req = tcp_v4_route_req, |
936b8bdb | 1249 | .init_seq = tcp_v4_init_sequence, |
d6274bd8 | 1250 | .send_synack = tcp_v4_send_synack, |
695da14e | 1251 | .queue_hash_add = inet_csk_reqsk_queue_hash_add, |
16bea70a | 1252 | }; |
cfb6eeb4 | 1253 | |
1da177e4 LT |
1254 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1255 | { | |
1da177e4 | 1256 | /* Never answer to SYNs send to broadcast or multicast */ |
511c3f92 | 1257 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1258 | goto drop; |
1259 | ||
1fb6f159 OP |
1260 | return tcp_conn_request(&tcp_request_sock_ops, |
1261 | &tcp_request_sock_ipv4_ops, sk, skb); | |
1da177e4 | 1262 | |
1da177e4 | 1263 | drop: |
848bf15f | 1264 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
1265 | return 0; |
1266 | } | |
4bc2f18b | 1267 | EXPORT_SYMBOL(tcp_v4_conn_request); |
1da177e4 LT |
1268 | |
1269 | ||
1270 | /* | |
1271 | * The three way handshake has completed - we got a valid synack - | |
1272 | * now create the new socket. | |
1273 | */ | |
1274 | struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1275 | struct request_sock *req, |
1da177e4 LT |
1276 | struct dst_entry *dst) |
1277 | { | |
2e6599cb | 1278 | struct inet_request_sock *ireq; |
1da177e4 LT |
1279 | struct inet_sock *newinet; |
1280 | struct tcp_sock *newtp; | |
1281 | struct sock *newsk; | |
cfb6eeb4 YH |
1282 | #ifdef CONFIG_TCP_MD5SIG |
1283 | struct tcp_md5sig_key *key; | |
1284 | #endif | |
f6d8bd05 | 1285 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
1286 | |
1287 | if (sk_acceptq_is_full(sk)) | |
1288 | goto exit_overflow; | |
1289 | ||
1da177e4 LT |
1290 | newsk = tcp_create_openreq_child(sk, req, skb); |
1291 | if (!newsk) | |
093d2823 | 1292 | goto exit_nonewsk; |
1da177e4 | 1293 | |
bcd76111 | 1294 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
fae6ef87 | 1295 | inet_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1296 | |
1297 | newtp = tcp_sk(newsk); | |
1298 | newinet = inet_sk(newsk); | |
2e6599cb | 1299 | ireq = inet_rsk(req); |
d1e559d0 ED |
1300 | sk_daddr_set(newsk, ireq->ir_rmt_addr); |
1301 | sk_rcv_saddr_set(newsk, ireq->ir_loc_addr); | |
634fb979 | 1302 | newinet->inet_saddr = ireq->ir_loc_addr; |
f6d8bd05 ED |
1303 | inet_opt = ireq->opt; |
1304 | rcu_assign_pointer(newinet->inet_opt, inet_opt); | |
2e6599cb | 1305 | ireq->opt = NULL; |
463c84b9 | 1306 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1307 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
4c507d28 | 1308 | newinet->rcv_tos = ip_hdr(skb)->tos; |
d83d8461 | 1309 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
b73c3d0e | 1310 | inet_set_txhash(newsk); |
f6d8bd05 ED |
1311 | if (inet_opt) |
1312 | inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
c720c7e8 | 1313 | newinet->inet_id = newtp->write_seq ^ jiffies; |
1da177e4 | 1314 | |
dfd25fff ED |
1315 | if (!dst) { |
1316 | dst = inet_csk_route_child_sock(sk, newsk, req); | |
1317 | if (!dst) | |
1318 | goto put_and_exit; | |
1319 | } else { | |
1320 | /* syncookie case : see end of cookie_v4_check() */ | |
1321 | } | |
0e734419 DM |
1322 | sk_setup_caps(newsk, dst); |
1323 | ||
81164413 DB |
1324 | tcp_ca_openreq_child(newsk, dst); |
1325 | ||
1da177e4 | 1326 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1327 | newtp->advmss = dst_metric_advmss(dst); |
f5fff5dc TQ |
1328 | if (tcp_sk(sk)->rx_opt.user_mss && |
1329 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1330 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1331 | ||
1da177e4 LT |
1332 | tcp_initialize_rcv_mss(newsk); |
1333 | ||
cfb6eeb4 YH |
1334 | #ifdef CONFIG_TCP_MD5SIG |
1335 | /* Copy over the MD5 key from the original socket */ | |
a915da9b ED |
1336 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1337 | AF_INET); | |
c720c7e8 | 1338 | if (key != NULL) { |
cfb6eeb4 YH |
1339 | /* |
1340 | * We're using one, so create a matching key | |
1341 | * on the newsk structure. If we fail to get | |
1342 | * memory, then we end up not copying the key | |
1343 | * across. Shucks. | |
1344 | */ | |
a915da9b ED |
1345 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1346 | AF_INET, key->key, key->keylen, GFP_ATOMIC); | |
a465419b | 1347 | sk_nocaps_add(newsk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1348 | } |
1349 | #endif | |
1350 | ||
0e734419 DM |
1351 | if (__inet_inherit_port(sk, newsk) < 0) |
1352 | goto put_and_exit; | |
9327f705 | 1353 | __inet_hash_nolisten(newsk, NULL); |
1da177e4 LT |
1354 | |
1355 | return newsk; | |
1356 | ||
1357 | exit_overflow: | |
de0744af | 1358 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 BS |
1359 | exit_nonewsk: |
1360 | dst_release(dst); | |
1da177e4 | 1361 | exit: |
de0744af | 1362 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 | 1363 | return NULL; |
0e734419 | 1364 | put_and_exit: |
e337e24d CP |
1365 | inet_csk_prepare_forced_close(newsk); |
1366 | tcp_done(newsk); | |
0e734419 | 1367 | goto exit; |
1da177e4 | 1368 | } |
4bc2f18b | 1369 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); |
1da177e4 LT |
1370 | |
1371 | static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb) | |
1372 | { | |
52452c54 | 1373 | const struct tcphdr *th = tcp_hdr(skb); |
eddc9ec5 | 1374 | const struct iphdr *iph = ip_hdr(skb); |
52452c54 | 1375 | struct request_sock *req; |
1da177e4 | 1376 | struct sock *nsk; |
52452c54 ED |
1377 | |
1378 | req = inet_csk_search_req(sk, th->source, iph->saddr, iph->daddr); | |
fa76ce73 ED |
1379 | if (req) { |
1380 | nsk = tcp_check_req(sk, skb, req, false); | |
1381 | reqsk_put(req); | |
1382 | return nsk; | |
1383 | } | |
1da177e4 | 1384 | |
3b1e0a65 | 1385 | nsk = inet_lookup_established(sock_net(sk), &tcp_hashinfo, iph->saddr, |
c67499c0 | 1386 | th->source, iph->daddr, th->dest, inet_iif(skb)); |
1da177e4 LT |
1387 | |
1388 | if (nsk) { | |
1389 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
1390 | bh_lock_sock(nsk); | |
1391 | return nsk; | |
1392 | } | |
9469c7b4 | 1393 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
1394 | return NULL; |
1395 | } | |
1396 | ||
1397 | #ifdef CONFIG_SYN_COOKIES | |
af9b4738 | 1398 | if (!th->syn) |
461b74c3 | 1399 | sk = cookie_v4_check(sk, skb); |
1da177e4 LT |
1400 | #endif |
1401 | return sk; | |
1402 | } | |
1403 | ||
1da177e4 LT |
1404 | /* The socket must have it's spinlock held when we get |
1405 | * here. | |
1406 | * | |
1407 | * We have a potential double-lock case here, so even when | |
1408 | * doing backlog processing we use the BH locking scheme. | |
1409 | * This is because we cannot sleep with the original spinlock | |
1410 | * held. | |
1411 | */ | |
1412 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1413 | { | |
cfb6eeb4 | 1414 | struct sock *rsk; |
cfb6eeb4 | 1415 | |
1da177e4 | 1416 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
404e0a8b ED |
1417 | struct dst_entry *dst = sk->sk_rx_dst; |
1418 | ||
bdeab991 | 1419 | sock_rps_save_rxhash(sk, skb); |
3d97379a | 1420 | sk_mark_napi_id(sk, skb); |
404e0a8b | 1421 | if (dst) { |
505fbcf0 ED |
1422 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || |
1423 | dst->ops->check(dst, 0) == NULL) { | |
92101b3b DM |
1424 | dst_release(dst); |
1425 | sk->sk_rx_dst = NULL; | |
1426 | } | |
1427 | } | |
c995ae22 | 1428 | tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len); |
1da177e4 LT |
1429 | return 0; |
1430 | } | |
1431 | ||
ab6a5bb6 | 1432 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1433 | goto csum_err; |
1434 | ||
1435 | if (sk->sk_state == TCP_LISTEN) { | |
1436 | struct sock *nsk = tcp_v4_hnd_req(sk, skb); | |
1437 | if (!nsk) | |
1438 | goto discard; | |
1439 | ||
1440 | if (nsk != sk) { | |
bdeab991 | 1441 | sock_rps_save_rxhash(nsk, skb); |
3d97379a | 1442 | sk_mark_napi_id(sk, skb); |
cfb6eeb4 YH |
1443 | if (tcp_child_process(sk, nsk, skb)) { |
1444 | rsk = nsk; | |
1da177e4 | 1445 | goto reset; |
cfb6eeb4 | 1446 | } |
1da177e4 LT |
1447 | return 0; |
1448 | } | |
ca55158c | 1449 | } else |
bdeab991 | 1450 | sock_rps_save_rxhash(sk, skb); |
ca55158c | 1451 | |
aa8223c7 | 1452 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) { |
cfb6eeb4 | 1453 | rsk = sk; |
1da177e4 | 1454 | goto reset; |
cfb6eeb4 | 1455 | } |
1da177e4 LT |
1456 | return 0; |
1457 | ||
1458 | reset: | |
cfb6eeb4 | 1459 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1460 | discard: |
1461 | kfree_skb(skb); | |
1462 | /* Be careful here. If this function gets more complicated and | |
1463 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1464 | * might be destroyed here. This current version compiles correctly, | |
1465 | * but you have been warned. | |
1466 | */ | |
1467 | return 0; | |
1468 | ||
1469 | csum_err: | |
6a5dc9e5 | 1470 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS); |
63231bdd | 1471 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1472 | goto discard; |
1473 | } | |
4bc2f18b | 1474 | EXPORT_SYMBOL(tcp_v4_do_rcv); |
1da177e4 | 1475 | |
160eb5a6 | 1476 | void tcp_v4_early_demux(struct sk_buff *skb) |
41063e9d | 1477 | { |
41063e9d DM |
1478 | const struct iphdr *iph; |
1479 | const struct tcphdr *th; | |
1480 | struct sock *sk; | |
41063e9d | 1481 | |
41063e9d | 1482 | if (skb->pkt_type != PACKET_HOST) |
160eb5a6 | 1483 | return; |
41063e9d | 1484 | |
45f00f99 | 1485 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) |
160eb5a6 | 1486 | return; |
41063e9d DM |
1487 | |
1488 | iph = ip_hdr(skb); | |
45f00f99 | 1489 | th = tcp_hdr(skb); |
41063e9d DM |
1490 | |
1491 | if (th->doff < sizeof(struct tcphdr) / 4) | |
160eb5a6 | 1492 | return; |
41063e9d | 1493 | |
45f00f99 | 1494 | sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo, |
41063e9d | 1495 | iph->saddr, th->source, |
7011d085 | 1496 | iph->daddr, ntohs(th->dest), |
9cb429d6 | 1497 | skb->skb_iif); |
41063e9d DM |
1498 | if (sk) { |
1499 | skb->sk = sk; | |
1500 | skb->destructor = sock_edemux; | |
f7e4eb03 | 1501 | if (sk_fullsock(sk)) { |
41063e9d | 1502 | struct dst_entry *dst = sk->sk_rx_dst; |
505fbcf0 | 1503 | |
41063e9d DM |
1504 | if (dst) |
1505 | dst = dst_check(dst, 0); | |
92101b3b | 1506 | if (dst && |
505fbcf0 | 1507 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
92101b3b | 1508 | skb_dst_set_noref(skb, dst); |
41063e9d DM |
1509 | } |
1510 | } | |
41063e9d DM |
1511 | } |
1512 | ||
b2fb4f54 ED |
1513 | /* Packet is added to VJ-style prequeue for processing in process |
1514 | * context, if a reader task is waiting. Apparently, this exciting | |
1515 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
1516 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
1517 | * see, why it failed. 8)8) --ANK | |
1518 | * | |
1519 | */ | |
1520 | bool tcp_prequeue(struct sock *sk, struct sk_buff *skb) | |
1521 | { | |
1522 | struct tcp_sock *tp = tcp_sk(sk); | |
1523 | ||
1524 | if (sysctl_tcp_low_latency || !tp->ucopy.task) | |
1525 | return false; | |
1526 | ||
1527 | if (skb->len <= tcp_hdrlen(skb) && | |
1528 | skb_queue_len(&tp->ucopy.prequeue) == 0) | |
1529 | return false; | |
1530 | ||
ca777eff ED |
1531 | /* Before escaping RCU protected region, we need to take care of skb |
1532 | * dst. Prequeue is only enabled for established sockets. | |
1533 | * For such sockets, we might need the skb dst only to set sk->sk_rx_dst | |
1534 | * Instead of doing full sk_rx_dst validity here, let's perform | |
1535 | * an optimistic check. | |
1536 | */ | |
1537 | if (likely(sk->sk_rx_dst)) | |
1538 | skb_dst_drop(skb); | |
1539 | else | |
1540 | skb_dst_force(skb); | |
1541 | ||
b2fb4f54 ED |
1542 | __skb_queue_tail(&tp->ucopy.prequeue, skb); |
1543 | tp->ucopy.memory += skb->truesize; | |
1544 | if (tp->ucopy.memory > sk->sk_rcvbuf) { | |
1545 | struct sk_buff *skb1; | |
1546 | ||
1547 | BUG_ON(sock_owned_by_user(sk)); | |
1548 | ||
1549 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { | |
1550 | sk_backlog_rcv(sk, skb1); | |
1551 | NET_INC_STATS_BH(sock_net(sk), | |
1552 | LINUX_MIB_TCPPREQUEUEDROPPED); | |
1553 | } | |
1554 | ||
1555 | tp->ucopy.memory = 0; | |
1556 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
1557 | wake_up_interruptible_sync_poll(sk_sleep(sk), | |
1558 | POLLIN | POLLRDNORM | POLLRDBAND); | |
1559 | if (!inet_csk_ack_scheduled(sk)) | |
1560 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
1561 | (3 * tcp_rto_min(sk)) / 4, | |
1562 | TCP_RTO_MAX); | |
1563 | } | |
1564 | return true; | |
1565 | } | |
1566 | EXPORT_SYMBOL(tcp_prequeue); | |
1567 | ||
1da177e4 LT |
1568 | /* |
1569 | * From tcp_input.c | |
1570 | */ | |
1571 | ||
1572 | int tcp_v4_rcv(struct sk_buff *skb) | |
1573 | { | |
eddc9ec5 | 1574 | const struct iphdr *iph; |
cf533ea5 | 1575 | const struct tcphdr *th; |
1da177e4 LT |
1576 | struct sock *sk; |
1577 | int ret; | |
a86b1e30 | 1578 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1579 | |
1580 | if (skb->pkt_type != PACKET_HOST) | |
1581 | goto discard_it; | |
1582 | ||
1583 | /* Count it even if it's bad */ | |
63231bdd | 1584 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1585 | |
1586 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1587 | goto discard_it; | |
1588 | ||
aa8223c7 | 1589 | th = tcp_hdr(skb); |
1da177e4 LT |
1590 | |
1591 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1592 | goto bad_packet; | |
1593 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1594 | goto discard_it; | |
1595 | ||
1596 | /* An explanation is required here, I think. | |
1597 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1598 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1599 | * So, we defer the checks. */ |
ed70fcfc TH |
1600 | |
1601 | if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo)) | |
6a5dc9e5 | 1602 | goto csum_error; |
1da177e4 | 1603 | |
aa8223c7 | 1604 | th = tcp_hdr(skb); |
eddc9ec5 | 1605 | iph = ip_hdr(skb); |
971f10ec ED |
1606 | /* This is tricky : We move IPCB at its correct location into TCP_SKB_CB() |
1607 | * barrier() makes sure compiler wont play fool^Waliasing games. | |
1608 | */ | |
1609 | memmove(&TCP_SKB_CB(skb)->header.h4, IPCB(skb), | |
1610 | sizeof(struct inet_skb_parm)); | |
1611 | barrier(); | |
1612 | ||
1da177e4 LT |
1613 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1614 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1615 | skb->len - th->doff * 4); | |
1616 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
e11ecddf | 1617 | TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th); |
04317daf | 1618 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; |
b82d1bb4 | 1619 | TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph); |
1da177e4 LT |
1620 | TCP_SKB_CB(skb)->sacked = 0; |
1621 | ||
9a1f27c4 | 1622 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1623 | if (!sk) |
1624 | goto no_tcp_socket; | |
1625 | ||
bb134d5d ED |
1626 | process: |
1627 | if (sk->sk_state == TCP_TIME_WAIT) | |
1628 | goto do_time_wait; | |
1629 | ||
6cce09f8 ED |
1630 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
1631 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
d218d111 | 1632 | goto discard_and_relse; |
6cce09f8 | 1633 | } |
d218d111 | 1634 | |
1da177e4 LT |
1635 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1636 | goto discard_and_relse; | |
9ea88a15 DP |
1637 | |
1638 | #ifdef CONFIG_TCP_MD5SIG | |
1639 | /* | |
1640 | * We really want to reject the packet as early as possible | |
1641 | * if: | |
1642 | * o We're expecting an MD5'd packet and this is no MD5 tcp option | |
1643 | * o There is an MD5 option and we're not expecting one | |
1644 | */ | |
1645 | if (tcp_v4_inbound_md5_hash(sk, skb)) | |
1646 | goto discard_and_relse; | |
1647 | #endif | |
1648 | ||
b59c2701 | 1649 | nf_reset(skb); |
1da177e4 | 1650 | |
fda9ef5d | 1651 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1652 | goto discard_and_relse; |
1653 | ||
2c8c56e1 | 1654 | sk_incoming_cpu_update(sk); |
1da177e4 LT |
1655 | skb->dev = NULL; |
1656 | ||
c6366184 | 1657 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1658 | ret = 0; |
1659 | if (!sock_owned_by_user(sk)) { | |
7bced397 | 1660 | if (!tcp_prequeue(sk, skb)) |
1da177e4 | 1661 | ret = tcp_v4_do_rcv(sk, skb); |
da882c1f ED |
1662 | } else if (unlikely(sk_add_backlog(sk, skb, |
1663 | sk->sk_rcvbuf + sk->sk_sndbuf))) { | |
6b03a53a | 1664 | bh_unlock_sock(sk); |
6cce09f8 | 1665 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1666 | goto discard_and_relse; |
1667 | } | |
1da177e4 LT |
1668 | bh_unlock_sock(sk); |
1669 | ||
1670 | sock_put(sk); | |
1671 | ||
1672 | return ret; | |
1673 | ||
1674 | no_tcp_socket: | |
1675 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1676 | goto discard_it; | |
1677 | ||
1678 | if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1679 | csum_error: |
1680 | TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS); | |
1da177e4 | 1681 | bad_packet: |
63231bdd | 1682 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1683 | } else { |
cfb6eeb4 | 1684 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1685 | } |
1686 | ||
1687 | discard_it: | |
1688 | /* Discard frame. */ | |
1689 | kfree_skb(skb); | |
e905a9ed | 1690 | return 0; |
1da177e4 LT |
1691 | |
1692 | discard_and_relse: | |
1693 | sock_put(sk); | |
1694 | goto discard_it; | |
1695 | ||
1696 | do_time_wait: | |
1697 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1698 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1699 | goto discard_it; |
1700 | } | |
1701 | ||
6a5dc9e5 | 1702 | if (skb->len < (th->doff << 2)) { |
9469c7b4 | 1703 | inet_twsk_put(inet_twsk(sk)); |
6a5dc9e5 ED |
1704 | goto bad_packet; |
1705 | } | |
1706 | if (tcp_checksum_complete(skb)) { | |
1707 | inet_twsk_put(inet_twsk(sk)); | |
1708 | goto csum_error; | |
1da177e4 | 1709 | } |
9469c7b4 | 1710 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1711 | case TCP_TW_SYN: { |
c346dca1 | 1712 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
c67499c0 | 1713 | &tcp_hashinfo, |
da5e3630 | 1714 | iph->saddr, th->source, |
eddc9ec5 | 1715 | iph->daddr, th->dest, |
463c84b9 | 1716 | inet_iif(skb)); |
1da177e4 | 1717 | if (sk2) { |
9469c7b4 YH |
1718 | inet_twsk_deschedule(inet_twsk(sk), &tcp_death_row); |
1719 | inet_twsk_put(inet_twsk(sk)); | |
1da177e4 LT |
1720 | sk = sk2; |
1721 | goto process; | |
1722 | } | |
1723 | /* Fall through to ACK */ | |
1724 | } | |
1725 | case TCP_TW_ACK: | |
1726 | tcp_v4_timewait_ack(sk, skb); | |
1727 | break; | |
1728 | case TCP_TW_RST: | |
1729 | goto no_tcp_socket; | |
1730 | case TCP_TW_SUCCESS:; | |
1731 | } | |
1732 | goto discard_it; | |
1733 | } | |
1734 | ||
ccb7c410 DM |
1735 | static struct timewait_sock_ops tcp_timewait_sock_ops = { |
1736 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1737 | .twsk_unique = tcp_twsk_unique, | |
1738 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 | 1739 | }; |
1da177e4 | 1740 | |
63d02d15 | 1741 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
5d299f3d ED |
1742 | { |
1743 | struct dst_entry *dst = skb_dst(skb); | |
1744 | ||
ca777eff ED |
1745 | if (dst) { |
1746 | dst_hold(dst); | |
1747 | sk->sk_rx_dst = dst; | |
1748 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
1749 | } | |
5d299f3d | 1750 | } |
63d02d15 | 1751 | EXPORT_SYMBOL(inet_sk_rx_dst_set); |
5d299f3d | 1752 | |
3b401a81 | 1753 | const struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1754 | .queue_xmit = ip_queue_xmit, |
1755 | .send_check = tcp_v4_send_check, | |
1756 | .rebuild_header = inet_sk_rebuild_header, | |
5d299f3d | 1757 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1758 | .conn_request = tcp_v4_conn_request, |
1759 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
543d9cfe ACM |
1760 | .net_header_len = sizeof(struct iphdr), |
1761 | .setsockopt = ip_setsockopt, | |
1762 | .getsockopt = ip_getsockopt, | |
1763 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1764 | .sockaddr_len = sizeof(struct sockaddr_in), | |
ab1e0a13 | 1765 | .bind_conflict = inet_csk_bind_conflict, |
3fdadf7d | 1766 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1767 | .compat_setsockopt = compat_ip_setsockopt, |
1768 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1769 | #endif |
4fab9071 | 1770 | .mtu_reduced = tcp_v4_mtu_reduced, |
1da177e4 | 1771 | }; |
4bc2f18b | 1772 | EXPORT_SYMBOL(ipv4_specific); |
1da177e4 | 1773 | |
cfb6eeb4 | 1774 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1775 | static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1776 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1777 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1778 | .md5_parse = tcp_v4_parse_md5_keys, |
cfb6eeb4 | 1779 | }; |
b6332e6c | 1780 | #endif |
cfb6eeb4 | 1781 | |
1da177e4 LT |
1782 | /* NOTE: A lot of things set to zero explicitly by call to |
1783 | * sk_alloc() so need not be done here. | |
1784 | */ | |
1785 | static int tcp_v4_init_sock(struct sock *sk) | |
1786 | { | |
6687e988 | 1787 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1788 | |
900f65d3 | 1789 | tcp_init_sock(sk); |
1da177e4 | 1790 | |
8292a17a | 1791 | icsk->icsk_af_ops = &ipv4_specific; |
900f65d3 | 1792 | |
cfb6eeb4 | 1793 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1794 | tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific; |
cfb6eeb4 | 1795 | #endif |
1da177e4 | 1796 | |
1da177e4 LT |
1797 | return 0; |
1798 | } | |
1799 | ||
7d06b2e0 | 1800 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1801 | { |
1802 | struct tcp_sock *tp = tcp_sk(sk); | |
1803 | ||
1804 | tcp_clear_xmit_timers(sk); | |
1805 | ||
6687e988 | 1806 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1807 | |
1da177e4 | 1808 | /* Cleanup up the write buffer. */ |
fe067e8a | 1809 | tcp_write_queue_purge(sk); |
1da177e4 LT |
1810 | |
1811 | /* Cleans up our, hopefully empty, out_of_order_queue. */ | |
e905a9ed | 1812 | __skb_queue_purge(&tp->out_of_order_queue); |
1da177e4 | 1813 | |
cfb6eeb4 YH |
1814 | #ifdef CONFIG_TCP_MD5SIG |
1815 | /* Clean up the MD5 key list, if any */ | |
1816 | if (tp->md5sig_info) { | |
a915da9b | 1817 | tcp_clear_md5_list(sk); |
a8afca03 | 1818 | kfree_rcu(tp->md5sig_info, rcu); |
cfb6eeb4 YH |
1819 | tp->md5sig_info = NULL; |
1820 | } | |
1821 | #endif | |
1a2449a8 | 1822 | |
1da177e4 LT |
1823 | /* Clean prequeue, it must be empty really */ |
1824 | __skb_queue_purge(&tp->ucopy.prequeue); | |
1825 | ||
1826 | /* Clean up a referenced TCP bind bucket. */ | |
463c84b9 | 1827 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1828 | inet_put_port(sk); |
1da177e4 | 1829 | |
168a8f58 | 1830 | BUG_ON(tp->fastopen_rsk != NULL); |
435cf559 | 1831 | |
cf60af03 YC |
1832 | /* If socket is aborted during connect operation */ |
1833 | tcp_free_fastopen_req(tp); | |
1834 | ||
180d8cd9 | 1835 | sk_sockets_allocated_dec(sk); |
d1a4c0b3 | 1836 | sock_release_memcg(sk); |
1da177e4 | 1837 | } |
1da177e4 LT |
1838 | EXPORT_SYMBOL(tcp_v4_destroy_sock); |
1839 | ||
1840 | #ifdef CONFIG_PROC_FS | |
1841 | /* Proc filesystem TCP sock list dumping. */ | |
1842 | ||
a8b690f9 TH |
1843 | /* |
1844 | * Get next listener socket follow cur. If cur is NULL, get first socket | |
1845 | * starting from bucket given in st->bucket; when st->bucket is zero the | |
1846 | * very first socket in the hash table is returned. | |
1847 | */ | |
1da177e4 LT |
1848 | static void *listening_get_next(struct seq_file *seq, void *cur) |
1849 | { | |
463c84b9 | 1850 | struct inet_connection_sock *icsk; |
c25eb3bf | 1851 | struct hlist_nulls_node *node; |
1da177e4 | 1852 | struct sock *sk = cur; |
5caea4ea | 1853 | struct inet_listen_hashbucket *ilb; |
5799de0b | 1854 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1855 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1856 | |
1857 | if (!sk) { | |
a8b690f9 | 1858 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
5caea4ea | 1859 | spin_lock_bh(&ilb->lock); |
c25eb3bf | 1860 | sk = sk_nulls_head(&ilb->head); |
a8b690f9 | 1861 | st->offset = 0; |
1da177e4 LT |
1862 | goto get_sk; |
1863 | } | |
5caea4ea | 1864 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 | 1865 | ++st->num; |
a8b690f9 | 1866 | ++st->offset; |
1da177e4 LT |
1867 | |
1868 | if (st->state == TCP_SEQ_STATE_OPENREQ) { | |
60236fdd | 1869 | struct request_sock *req = cur; |
1da177e4 | 1870 | |
72a3effa | 1871 | icsk = inet_csk(st->syn_wait_sk); |
1da177e4 LT |
1872 | req = req->dl_next; |
1873 | while (1) { | |
1874 | while (req) { | |
bdccc4ca | 1875 | if (req->rsk_ops->family == st->family) { |
1da177e4 LT |
1876 | cur = req; |
1877 | goto out; | |
1878 | } | |
1879 | req = req->dl_next; | |
1880 | } | |
72a3effa | 1881 | if (++st->sbucket >= icsk->icsk_accept_queue.listen_opt->nr_table_entries) |
1da177e4 LT |
1882 | break; |
1883 | get_req: | |
463c84b9 | 1884 | req = icsk->icsk_accept_queue.listen_opt->syn_table[st->sbucket]; |
1da177e4 | 1885 | } |
1bde5ac4 | 1886 | sk = sk_nulls_next(st->syn_wait_sk); |
1da177e4 | 1887 | st->state = TCP_SEQ_STATE_LISTENING; |
b2827053 | 1888 | spin_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 1889 | } else { |
e905a9ed | 1890 | icsk = inet_csk(sk); |
b2827053 | 1891 | spin_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
463c84b9 | 1892 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) |
1da177e4 | 1893 | goto start_req; |
b2827053 | 1894 | spin_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1bde5ac4 | 1895 | sk = sk_nulls_next(sk); |
1da177e4 LT |
1896 | } |
1897 | get_sk: | |
c25eb3bf | 1898 | sk_nulls_for_each_from(sk, node) { |
8475ef9f PE |
1899 | if (!net_eq(sock_net(sk), net)) |
1900 | continue; | |
1901 | if (sk->sk_family == st->family) { | |
1da177e4 LT |
1902 | cur = sk; |
1903 | goto out; | |
1904 | } | |
e905a9ed | 1905 | icsk = inet_csk(sk); |
b2827053 | 1906 | spin_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
463c84b9 | 1907 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) { |
1da177e4 LT |
1908 | start_req: |
1909 | st->uid = sock_i_uid(sk); | |
1910 | st->syn_wait_sk = sk; | |
1911 | st->state = TCP_SEQ_STATE_OPENREQ; | |
1912 | st->sbucket = 0; | |
1913 | goto get_req; | |
1914 | } | |
b2827053 | 1915 | spin_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 1916 | } |
5caea4ea | 1917 | spin_unlock_bh(&ilb->lock); |
a8b690f9 | 1918 | st->offset = 0; |
0f7ff927 | 1919 | if (++st->bucket < INET_LHTABLE_SIZE) { |
5caea4ea ED |
1920 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1921 | spin_lock_bh(&ilb->lock); | |
c25eb3bf | 1922 | sk = sk_nulls_head(&ilb->head); |
1da177e4 LT |
1923 | goto get_sk; |
1924 | } | |
1925 | cur = NULL; | |
1926 | out: | |
1927 | return cur; | |
1928 | } | |
1929 | ||
1930 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
1931 | { | |
a8b690f9 TH |
1932 | struct tcp_iter_state *st = seq->private; |
1933 | void *rc; | |
1934 | ||
1935 | st->bucket = 0; | |
1936 | st->offset = 0; | |
1937 | rc = listening_get_next(seq, NULL); | |
1da177e4 LT |
1938 | |
1939 | while (rc && *pos) { | |
1940 | rc = listening_get_next(seq, rc); | |
1941 | --*pos; | |
1942 | } | |
1943 | return rc; | |
1944 | } | |
1945 | ||
05dbc7b5 | 1946 | static inline bool empty_bucket(const struct tcp_iter_state *st) |
6eac5604 | 1947 | { |
05dbc7b5 | 1948 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain); |
6eac5604 AK |
1949 | } |
1950 | ||
a8b690f9 TH |
1951 | /* |
1952 | * Get first established socket starting from bucket given in st->bucket. | |
1953 | * If st->bucket is zero, the very first socket in the hash is returned. | |
1954 | */ | |
1da177e4 LT |
1955 | static void *established_get_first(struct seq_file *seq) |
1956 | { | |
5799de0b | 1957 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1958 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1959 | void *rc = NULL; |
1960 | ||
a8b690f9 TH |
1961 | st->offset = 0; |
1962 | for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) { | |
1da177e4 | 1963 | struct sock *sk; |
3ab5aee7 | 1964 | struct hlist_nulls_node *node; |
9db66bdc | 1965 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 1966 | |
6eac5604 AK |
1967 | /* Lockless fast path for the common case of empty buckets */ |
1968 | if (empty_bucket(st)) | |
1969 | continue; | |
1970 | ||
9db66bdc | 1971 | spin_lock_bh(lock); |
3ab5aee7 | 1972 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 1973 | if (sk->sk_family != st->family || |
878628fb | 1974 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
1975 | continue; |
1976 | } | |
1977 | rc = sk; | |
1978 | goto out; | |
1979 | } | |
9db66bdc | 1980 | spin_unlock_bh(lock); |
1da177e4 LT |
1981 | } |
1982 | out: | |
1983 | return rc; | |
1984 | } | |
1985 | ||
1986 | static void *established_get_next(struct seq_file *seq, void *cur) | |
1987 | { | |
1988 | struct sock *sk = cur; | |
3ab5aee7 | 1989 | struct hlist_nulls_node *node; |
5799de0b | 1990 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1991 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1992 | |
1993 | ++st->num; | |
a8b690f9 | 1994 | ++st->offset; |
1da177e4 | 1995 | |
05dbc7b5 | 1996 | sk = sk_nulls_next(sk); |
1da177e4 | 1997 | |
3ab5aee7 | 1998 | sk_nulls_for_each_from(sk, node) { |
878628fb | 1999 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
05dbc7b5 | 2000 | return sk; |
1da177e4 LT |
2001 | } |
2002 | ||
05dbc7b5 ED |
2003 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
2004 | ++st->bucket; | |
2005 | return established_get_first(seq); | |
1da177e4 LT |
2006 | } |
2007 | ||
2008 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2009 | { | |
a8b690f9 TH |
2010 | struct tcp_iter_state *st = seq->private; |
2011 | void *rc; | |
2012 | ||
2013 | st->bucket = 0; | |
2014 | rc = established_get_first(seq); | |
1da177e4 LT |
2015 | |
2016 | while (rc && pos) { | |
2017 | rc = established_get_next(seq, rc); | |
2018 | --pos; | |
7174259e | 2019 | } |
1da177e4 LT |
2020 | return rc; |
2021 | } | |
2022 | ||
2023 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2024 | { | |
2025 | void *rc; | |
5799de0b | 2026 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2027 | |
1da177e4 LT |
2028 | st->state = TCP_SEQ_STATE_LISTENING; |
2029 | rc = listening_get_idx(seq, &pos); | |
2030 | ||
2031 | if (!rc) { | |
1da177e4 LT |
2032 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2033 | rc = established_get_idx(seq, pos); | |
2034 | } | |
2035 | ||
2036 | return rc; | |
2037 | } | |
2038 | ||
a8b690f9 TH |
2039 | static void *tcp_seek_last_pos(struct seq_file *seq) |
2040 | { | |
2041 | struct tcp_iter_state *st = seq->private; | |
2042 | int offset = st->offset; | |
2043 | int orig_num = st->num; | |
2044 | void *rc = NULL; | |
2045 | ||
2046 | switch (st->state) { | |
2047 | case TCP_SEQ_STATE_OPENREQ: | |
2048 | case TCP_SEQ_STATE_LISTENING: | |
2049 | if (st->bucket >= INET_LHTABLE_SIZE) | |
2050 | break; | |
2051 | st->state = TCP_SEQ_STATE_LISTENING; | |
2052 | rc = listening_get_next(seq, NULL); | |
2053 | while (offset-- && rc) | |
2054 | rc = listening_get_next(seq, rc); | |
2055 | if (rc) | |
2056 | break; | |
2057 | st->bucket = 0; | |
05dbc7b5 | 2058 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2059 | /* Fallthrough */ |
2060 | case TCP_SEQ_STATE_ESTABLISHED: | |
a8b690f9 TH |
2061 | if (st->bucket > tcp_hashinfo.ehash_mask) |
2062 | break; | |
2063 | rc = established_get_first(seq); | |
2064 | while (offset-- && rc) | |
2065 | rc = established_get_next(seq, rc); | |
2066 | } | |
2067 | ||
2068 | st->num = orig_num; | |
2069 | ||
2070 | return rc; | |
2071 | } | |
2072 | ||
1da177e4 LT |
2073 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) |
2074 | { | |
5799de0b | 2075 | struct tcp_iter_state *st = seq->private; |
a8b690f9 TH |
2076 | void *rc; |
2077 | ||
2078 | if (*pos && *pos == st->last_pos) { | |
2079 | rc = tcp_seek_last_pos(seq); | |
2080 | if (rc) | |
2081 | goto out; | |
2082 | } | |
2083 | ||
1da177e4 LT |
2084 | st->state = TCP_SEQ_STATE_LISTENING; |
2085 | st->num = 0; | |
a8b690f9 TH |
2086 | st->bucket = 0; |
2087 | st->offset = 0; | |
2088 | rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2089 | ||
2090 | out: | |
2091 | st->last_pos = *pos; | |
2092 | return rc; | |
1da177e4 LT |
2093 | } |
2094 | ||
2095 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2096 | { | |
a8b690f9 | 2097 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2098 | void *rc = NULL; |
1da177e4 LT |
2099 | |
2100 | if (v == SEQ_START_TOKEN) { | |
2101 | rc = tcp_get_idx(seq, 0); | |
2102 | goto out; | |
2103 | } | |
1da177e4 LT |
2104 | |
2105 | switch (st->state) { | |
2106 | case TCP_SEQ_STATE_OPENREQ: | |
2107 | case TCP_SEQ_STATE_LISTENING: | |
2108 | rc = listening_get_next(seq, v); | |
2109 | if (!rc) { | |
1da177e4 | 2110 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2111 | st->bucket = 0; |
2112 | st->offset = 0; | |
1da177e4 LT |
2113 | rc = established_get_first(seq); |
2114 | } | |
2115 | break; | |
2116 | case TCP_SEQ_STATE_ESTABLISHED: | |
1da177e4 LT |
2117 | rc = established_get_next(seq, v); |
2118 | break; | |
2119 | } | |
2120 | out: | |
2121 | ++*pos; | |
a8b690f9 | 2122 | st->last_pos = *pos; |
1da177e4 LT |
2123 | return rc; |
2124 | } | |
2125 | ||
2126 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2127 | { | |
5799de0b | 2128 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2129 | |
2130 | switch (st->state) { | |
2131 | case TCP_SEQ_STATE_OPENREQ: | |
2132 | if (v) { | |
463c84b9 | 2133 | struct inet_connection_sock *icsk = inet_csk(st->syn_wait_sk); |
b2827053 | 2134 | spin_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 LT |
2135 | } |
2136 | case TCP_SEQ_STATE_LISTENING: | |
2137 | if (v != SEQ_START_TOKEN) | |
5caea4ea | 2138 | spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 | 2139 | break; |
1da177e4 LT |
2140 | case TCP_SEQ_STATE_ESTABLISHED: |
2141 | if (v) | |
9db66bdc | 2142 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2143 | break; |
2144 | } | |
2145 | } | |
2146 | ||
73cb88ec | 2147 | int tcp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2148 | { |
d9dda78b | 2149 | struct tcp_seq_afinfo *afinfo = PDE_DATA(inode); |
1da177e4 | 2150 | struct tcp_iter_state *s; |
52d6f3f1 | 2151 | int err; |
1da177e4 | 2152 | |
52d6f3f1 DL |
2153 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2154 | sizeof(struct tcp_iter_state)); | |
2155 | if (err < 0) | |
2156 | return err; | |
f40c8174 | 2157 | |
52d6f3f1 | 2158 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2159 | s->family = afinfo->family; |
688d1945 | 2160 | s->last_pos = 0; |
f40c8174 DL |
2161 | return 0; |
2162 | } | |
73cb88ec | 2163 | EXPORT_SYMBOL(tcp_seq_open); |
f40c8174 | 2164 | |
6f8b13bc | 2165 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2166 | { |
2167 | int rc = 0; | |
2168 | struct proc_dir_entry *p; | |
2169 | ||
9427c4b3 DL |
2170 | afinfo->seq_ops.start = tcp_seq_start; |
2171 | afinfo->seq_ops.next = tcp_seq_next; | |
2172 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2173 | ||
84841c3c | 2174 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2175 | afinfo->seq_fops, afinfo); |
84841c3c | 2176 | if (!p) |
1da177e4 LT |
2177 | rc = -ENOMEM; |
2178 | return rc; | |
2179 | } | |
4bc2f18b | 2180 | EXPORT_SYMBOL(tcp_proc_register); |
1da177e4 | 2181 | |
6f8b13bc | 2182 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2183 | { |
ece31ffd | 2184 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2185 | } |
4bc2f18b | 2186 | EXPORT_SYMBOL(tcp_proc_unregister); |
1da177e4 | 2187 | |
d4f06873 | 2188 | static void get_openreq4(const struct request_sock *req, |
652586df | 2189 | struct seq_file *f, int i, kuid_t uid) |
1da177e4 | 2190 | { |
2e6599cb | 2191 | const struct inet_request_sock *ireq = inet_rsk(req); |
fa76ce73 | 2192 | long delta = req->rsk_timer.expires - jiffies; |
1da177e4 | 2193 | |
5e659e4c | 2194 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2195 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK", |
1da177e4 | 2196 | i, |
634fb979 | 2197 | ireq->ir_loc_addr, |
d4f06873 | 2198 | ireq->ir_num, |
634fb979 ED |
2199 | ireq->ir_rmt_addr, |
2200 | ntohs(ireq->ir_rmt_port), | |
1da177e4 LT |
2201 | TCP_SYN_RECV, |
2202 | 0, 0, /* could print option size, but that is af dependent. */ | |
2203 | 1, /* timers active (only the expire timer) */ | |
a399a805 | 2204 | jiffies_delta_to_clock_t(delta), |
e6c022a4 | 2205 | req->num_timeout, |
a7cb5a49 | 2206 | from_kuid_munged(seq_user_ns(f), uid), |
1da177e4 LT |
2207 | 0, /* non standard timer */ |
2208 | 0, /* open_requests have no inode */ | |
d4f06873 | 2209 | 0, |
652586df | 2210 | req); |
1da177e4 LT |
2211 | } |
2212 | ||
652586df | 2213 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i) |
1da177e4 LT |
2214 | { |
2215 | int timer_active; | |
2216 | unsigned long timer_expires; | |
cf533ea5 | 2217 | const struct tcp_sock *tp = tcp_sk(sk); |
cf4c6bf8 | 2218 | const struct inet_connection_sock *icsk = inet_csk(sk); |
cf533ea5 | 2219 | const struct inet_sock *inet = inet_sk(sk); |
168a8f58 | 2220 | struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq; |
c720c7e8 ED |
2221 | __be32 dest = inet->inet_daddr; |
2222 | __be32 src = inet->inet_rcv_saddr; | |
2223 | __u16 destp = ntohs(inet->inet_dport); | |
2224 | __u16 srcp = ntohs(inet->inet_sport); | |
49d09007 | 2225 | int rx_queue; |
1da177e4 | 2226 | |
6ba8a3b1 ND |
2227 | if (icsk->icsk_pending == ICSK_TIME_RETRANS || |
2228 | icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS || | |
2229 | icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { | |
1da177e4 | 2230 | timer_active = 1; |
463c84b9 ACM |
2231 | timer_expires = icsk->icsk_timeout; |
2232 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2233 | timer_active = 4; |
463c84b9 | 2234 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2235 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2236 | timer_active = 2; |
cf4c6bf8 | 2237 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2238 | } else { |
2239 | timer_active = 0; | |
2240 | timer_expires = jiffies; | |
2241 | } | |
2242 | ||
49d09007 ED |
2243 | if (sk->sk_state == TCP_LISTEN) |
2244 | rx_queue = sk->sk_ack_backlog; | |
2245 | else | |
2246 | /* | |
2247 | * because we dont lock socket, we might find a transient negative value | |
2248 | */ | |
2249 | rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0); | |
2250 | ||
5e659e4c | 2251 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
652586df | 2252 | "%08X %5u %8d %lu %d %pK %lu %lu %u %u %d", |
cf4c6bf8 | 2253 | i, src, srcp, dest, destp, sk->sk_state, |
47da8ee6 | 2254 | tp->write_seq - tp->snd_una, |
49d09007 | 2255 | rx_queue, |
1da177e4 | 2256 | timer_active, |
a399a805 | 2257 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 2258 | icsk->icsk_retransmits, |
a7cb5a49 | 2259 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)), |
6687e988 | 2260 | icsk->icsk_probes_out, |
cf4c6bf8 IJ |
2261 | sock_i_ino(sk), |
2262 | atomic_read(&sk->sk_refcnt), sk, | |
7be87351 SH |
2263 | jiffies_to_clock_t(icsk->icsk_rto), |
2264 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2265 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2266 | tp->snd_cwnd, |
168a8f58 JC |
2267 | sk->sk_state == TCP_LISTEN ? |
2268 | (fastopenq ? fastopenq->max_qlen : 0) : | |
652586df | 2269 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)); |
1da177e4 LT |
2270 | } |
2271 | ||
cf533ea5 | 2272 | static void get_timewait4_sock(const struct inet_timewait_sock *tw, |
652586df | 2273 | struct seq_file *f, int i) |
1da177e4 | 2274 | { |
23f33c2d | 2275 | __be32 dest, src; |
1da177e4 | 2276 | __u16 destp, srcp; |
e2a1d3e4 | 2277 | s32 delta = tw->tw_ttd - inet_tw_time_stamp(); |
1da177e4 LT |
2278 | |
2279 | dest = tw->tw_daddr; | |
2280 | src = tw->tw_rcv_saddr; | |
2281 | destp = ntohs(tw->tw_dport); | |
2282 | srcp = ntohs(tw->tw_sport); | |
2283 | ||
5e659e4c | 2284 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2285 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK", |
1da177e4 | 2286 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
a399a805 | 2287 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
652586df | 2288 | atomic_read(&tw->tw_refcnt), tw); |
1da177e4 LT |
2289 | } |
2290 | ||
2291 | #define TMPSZ 150 | |
2292 | ||
2293 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2294 | { | |
5799de0b | 2295 | struct tcp_iter_state *st; |
05dbc7b5 | 2296 | struct sock *sk = v; |
1da177e4 | 2297 | |
652586df | 2298 | seq_setwidth(seq, TMPSZ - 1); |
1da177e4 | 2299 | if (v == SEQ_START_TOKEN) { |
652586df | 2300 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
1da177e4 LT |
2301 | "rx_queue tr tm->when retrnsmt uid timeout " |
2302 | "inode"); | |
2303 | goto out; | |
2304 | } | |
2305 | st = seq->private; | |
2306 | ||
2307 | switch (st->state) { | |
2308 | case TCP_SEQ_STATE_LISTENING: | |
2309 | case TCP_SEQ_STATE_ESTABLISHED: | |
05dbc7b5 | 2310 | if (sk->sk_state == TCP_TIME_WAIT) |
652586df | 2311 | get_timewait4_sock(v, seq, st->num); |
05dbc7b5 | 2312 | else |
652586df | 2313 | get_tcp4_sock(v, seq, st->num); |
1da177e4 LT |
2314 | break; |
2315 | case TCP_SEQ_STATE_OPENREQ: | |
d4f06873 | 2316 | get_openreq4(v, seq, st->num, st->uid); |
1da177e4 LT |
2317 | break; |
2318 | } | |
1da177e4 | 2319 | out: |
652586df | 2320 | seq_pad(seq, '\n'); |
1da177e4 LT |
2321 | return 0; |
2322 | } | |
2323 | ||
73cb88ec AV |
2324 | static const struct file_operations tcp_afinfo_seq_fops = { |
2325 | .owner = THIS_MODULE, | |
2326 | .open = tcp_seq_open, | |
2327 | .read = seq_read, | |
2328 | .llseek = seq_lseek, | |
2329 | .release = seq_release_net | |
2330 | }; | |
2331 | ||
1da177e4 | 2332 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2333 | .name = "tcp", |
2334 | .family = AF_INET, | |
73cb88ec | 2335 | .seq_fops = &tcp_afinfo_seq_fops, |
9427c4b3 DL |
2336 | .seq_ops = { |
2337 | .show = tcp4_seq_show, | |
2338 | }, | |
1da177e4 LT |
2339 | }; |
2340 | ||
2c8c1e72 | 2341 | static int __net_init tcp4_proc_init_net(struct net *net) |
757764f6 PE |
2342 | { |
2343 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2344 | } | |
2345 | ||
2c8c1e72 | 2346 | static void __net_exit tcp4_proc_exit_net(struct net *net) |
757764f6 PE |
2347 | { |
2348 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2349 | } | |
2350 | ||
2351 | static struct pernet_operations tcp4_net_ops = { | |
2352 | .init = tcp4_proc_init_net, | |
2353 | .exit = tcp4_proc_exit_net, | |
2354 | }; | |
2355 | ||
1da177e4 LT |
2356 | int __init tcp4_proc_init(void) |
2357 | { | |
757764f6 | 2358 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2359 | } |
2360 | ||
2361 | void tcp4_proc_exit(void) | |
2362 | { | |
757764f6 | 2363 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2364 | } |
2365 | #endif /* CONFIG_PROC_FS */ | |
2366 | ||
2367 | struct proto tcp_prot = { | |
2368 | .name = "TCP", | |
2369 | .owner = THIS_MODULE, | |
2370 | .close = tcp_close, | |
2371 | .connect = tcp_v4_connect, | |
2372 | .disconnect = tcp_disconnect, | |
463c84b9 | 2373 | .accept = inet_csk_accept, |
1da177e4 LT |
2374 | .ioctl = tcp_ioctl, |
2375 | .init = tcp_v4_init_sock, | |
2376 | .destroy = tcp_v4_destroy_sock, | |
2377 | .shutdown = tcp_shutdown, | |
2378 | .setsockopt = tcp_setsockopt, | |
2379 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 2380 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2381 | .sendmsg = tcp_sendmsg, |
2382 | .sendpage = tcp_sendpage, | |
1da177e4 | 2383 | .backlog_rcv = tcp_v4_do_rcv, |
46d3ceab | 2384 | .release_cb = tcp_release_cb, |
ab1e0a13 ACM |
2385 | .hash = inet_hash, |
2386 | .unhash = inet_unhash, | |
2387 | .get_port = inet_csk_get_port, | |
1da177e4 | 2388 | .enter_memory_pressure = tcp_enter_memory_pressure, |
c9bee3b7 | 2389 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 | 2390 | .sockets_allocated = &tcp_sockets_allocated, |
0a5578cf | 2391 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2392 | .memory_allocated = &tcp_memory_allocated, |
2393 | .memory_pressure = &tcp_memory_pressure, | |
a4fe34bf | 2394 | .sysctl_mem = sysctl_tcp_mem, |
1da177e4 LT |
2395 | .sysctl_wmem = sysctl_tcp_wmem, |
2396 | .sysctl_rmem = sysctl_tcp_rmem, | |
2397 | .max_header = MAX_TCP_HEADER, | |
2398 | .obj_size = sizeof(struct tcp_sock), | |
3ab5aee7 | 2399 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2400 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2401 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2402 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2403 | .no_autobind = true, |
543d9cfe ACM |
2404 | #ifdef CONFIG_COMPAT |
2405 | .compat_setsockopt = compat_tcp_setsockopt, | |
2406 | .compat_getsockopt = compat_tcp_getsockopt, | |
2407 | #endif | |
c255a458 | 2408 | #ifdef CONFIG_MEMCG_KMEM |
d1a4c0b3 GC |
2409 | .init_cgroup = tcp_init_cgroup, |
2410 | .destroy_cgroup = tcp_destroy_cgroup, | |
2411 | .proto_cgroup = tcp_proto_cgroup, | |
2412 | #endif | |
1da177e4 | 2413 | }; |
4bc2f18b | 2414 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 | 2415 | |
bdbbb852 ED |
2416 | static void __net_exit tcp_sk_exit(struct net *net) |
2417 | { | |
2418 | int cpu; | |
2419 | ||
2420 | for_each_possible_cpu(cpu) | |
2421 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.tcp_sk, cpu)); | |
2422 | free_percpu(net->ipv4.tcp_sk); | |
2423 | } | |
2424 | ||
046ee902 DL |
2425 | static int __net_init tcp_sk_init(struct net *net) |
2426 | { | |
bdbbb852 ED |
2427 | int res, cpu; |
2428 | ||
2429 | net->ipv4.tcp_sk = alloc_percpu(struct sock *); | |
2430 | if (!net->ipv4.tcp_sk) | |
2431 | return -ENOMEM; | |
2432 | ||
2433 | for_each_possible_cpu(cpu) { | |
2434 | struct sock *sk; | |
2435 | ||
2436 | res = inet_ctl_sock_create(&sk, PF_INET, SOCK_RAW, | |
2437 | IPPROTO_TCP, net); | |
2438 | if (res) | |
2439 | goto fail; | |
2440 | *per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk; | |
2441 | } | |
5d134f1c | 2442 | net->ipv4.sysctl_tcp_ecn = 2; |
b0f9ca53 | 2443 | net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS; |
6b58e0a5 | 2444 | net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD; |
05cbc0db | 2445 | net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL; |
be9f4a44 | 2446 | return 0; |
046ee902 | 2447 | |
bdbbb852 ED |
2448 | fail: |
2449 | tcp_sk_exit(net); | |
2450 | ||
2451 | return res; | |
b099ce26 EB |
2452 | } |
2453 | ||
2454 | static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list) | |
2455 | { | |
2456 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET); | |
046ee902 DL |
2457 | } |
2458 | ||
2459 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
b099ce26 EB |
2460 | .init = tcp_sk_init, |
2461 | .exit = tcp_sk_exit, | |
2462 | .exit_batch = tcp_sk_exit_batch, | |
046ee902 DL |
2463 | }; |
2464 | ||
9b0f976f | 2465 | void __init tcp_v4_init(void) |
1da177e4 | 2466 | { |
5caea4ea | 2467 | inet_hashinfo_init(&tcp_hashinfo); |
6a1b3054 | 2468 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2469 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 | 2470 | } |