]>
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> |
076bb0c8 | 76 | #include <net/busy_poll.h> |
1da177e4 LT |
77 | |
78 | #include <linux/inet.h> | |
79 | #include <linux/ipv6.h> | |
80 | #include <linux/stddef.h> | |
81 | #include <linux/proc_fs.h> | |
82 | #include <linux/seq_file.h> | |
6797318e | 83 | #include <linux/inetdevice.h> |
1da177e4 | 84 | |
cf80e0e4 | 85 | #include <crypto/hash.h> |
cfb6eeb4 YH |
86 | #include <linux/scatterlist.h> |
87 | ||
c24b14c4 SL |
88 | #include <trace/events/tcp.h> |
89 | ||
cfb6eeb4 | 90 | #ifdef CONFIG_TCP_MD5SIG |
a915da9b | 91 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 92 | __be32 daddr, __be32 saddr, const struct tcphdr *th); |
cfb6eeb4 YH |
93 | #endif |
94 | ||
5caea4ea | 95 | struct inet_hashinfo tcp_hashinfo; |
4bc2f18b | 96 | EXPORT_SYMBOL(tcp_hashinfo); |
1da177e4 | 97 | |
84b114b9 | 98 | static u32 tcp_v4_init_seq(const struct sk_buff *skb) |
1da177e4 | 99 | { |
84b114b9 ED |
100 | return secure_tcp_seq(ip_hdr(skb)->daddr, |
101 | ip_hdr(skb)->saddr, | |
102 | tcp_hdr(skb)->dest, | |
103 | tcp_hdr(skb)->source); | |
104 | } | |
105 | ||
5d2ed052 | 106 | static u32 tcp_v4_init_ts_off(const struct net *net, const struct sk_buff *skb) |
84b114b9 | 107 | { |
5d2ed052 | 108 | return secure_tcp_ts_off(net, ip_hdr(skb)->daddr, ip_hdr(skb)->saddr); |
1da177e4 LT |
109 | } |
110 | ||
6d6ee43e ACM |
111 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp) |
112 | { | |
113 | const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw); | |
114 | struct tcp_sock *tp = tcp_sk(sk); | |
115 | ||
116 | /* With PAWS, it is safe from the viewpoint | |
117 | of data integrity. Even without PAWS it is safe provided sequence | |
118 | spaces do not overlap i.e. at data rates <= 80Mbit/sec. | |
119 | ||
120 | Actually, the idea is close to VJ's one, only timestamp cache is | |
121 | held not per host, but per port pair and TW bucket is used as state | |
122 | holder. | |
123 | ||
124 | If TW bucket has been already destroyed we fall back to VJ's scheme | |
125 | and use initial timestamp retrieved from peer table. | |
126 | */ | |
127 | if (tcptw->tw_ts_recent_stamp && | |
56ab6b93 | 128 | (!twp || (sock_net(sk)->ipv4.sysctl_tcp_tw_reuse && |
9d729f72 | 129 | get_seconds() - tcptw->tw_ts_recent_stamp > 1))) { |
6d6ee43e ACM |
130 | tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2; |
131 | if (tp->write_seq == 0) | |
132 | tp->write_seq = 1; | |
133 | tp->rx_opt.ts_recent = tcptw->tw_ts_recent; | |
134 | tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp; | |
135 | sock_hold(sktw); | |
136 | return 1; | |
137 | } | |
138 | ||
139 | return 0; | |
140 | } | |
6d6ee43e ACM |
141 | EXPORT_SYMBOL_GPL(tcp_twsk_unique); |
142 | ||
1da177e4 LT |
143 | /* This will initiate an outgoing connection. */ |
144 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
145 | { | |
2d7192d6 | 146 | struct sockaddr_in *usin = (struct sockaddr_in *)uaddr; |
1da177e4 LT |
147 | struct inet_sock *inet = inet_sk(sk); |
148 | struct tcp_sock *tp = tcp_sk(sk); | |
dca8b089 | 149 | __be16 orig_sport, orig_dport; |
bada8adc | 150 | __be32 daddr, nexthop; |
da905bd1 | 151 | struct flowi4 *fl4; |
2d7192d6 | 152 | struct rtable *rt; |
1da177e4 | 153 | int err; |
f6d8bd05 | 154 | struct ip_options_rcu *inet_opt; |
1946e672 | 155 | struct inet_timewait_death_row *tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row; |
1da177e4 LT |
156 | |
157 | if (addr_len < sizeof(struct sockaddr_in)) | |
158 | return -EINVAL; | |
159 | ||
160 | if (usin->sin_family != AF_INET) | |
161 | return -EAFNOSUPPORT; | |
162 | ||
163 | nexthop = daddr = usin->sin_addr.s_addr; | |
f6d8bd05 | 164 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1e1d04e6 | 165 | lockdep_sock_is_held(sk)); |
f6d8bd05 | 166 | if (inet_opt && inet_opt->opt.srr) { |
1da177e4 LT |
167 | if (!daddr) |
168 | return -EINVAL; | |
f6d8bd05 | 169 | nexthop = inet_opt->opt.faddr; |
1da177e4 LT |
170 | } |
171 | ||
dca8b089 DM |
172 | orig_sport = inet->inet_sport; |
173 | orig_dport = usin->sin_port; | |
da905bd1 DM |
174 | fl4 = &inet->cork.fl.u.ip4; |
175 | rt = ip_route_connect(fl4, nexthop, inet->inet_saddr, | |
b23dd4fe DM |
176 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if, |
177 | IPPROTO_TCP, | |
0e0d44ab | 178 | orig_sport, orig_dport, sk); |
b23dd4fe DM |
179 | if (IS_ERR(rt)) { |
180 | err = PTR_ERR(rt); | |
181 | if (err == -ENETUNREACH) | |
f1d8cba6 | 182 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); |
b23dd4fe | 183 | return err; |
584bdf8c | 184 | } |
1da177e4 LT |
185 | |
186 | if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) { | |
187 | ip_rt_put(rt); | |
188 | return -ENETUNREACH; | |
189 | } | |
190 | ||
f6d8bd05 | 191 | if (!inet_opt || !inet_opt->opt.srr) |
da905bd1 | 192 | daddr = fl4->daddr; |
1da177e4 | 193 | |
c720c7e8 | 194 | if (!inet->inet_saddr) |
da905bd1 | 195 | inet->inet_saddr = fl4->saddr; |
d1e559d0 | 196 | sk_rcv_saddr_set(sk, inet->inet_saddr); |
1da177e4 | 197 | |
c720c7e8 | 198 | if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) { |
1da177e4 LT |
199 | /* Reset inherited state */ |
200 | tp->rx_opt.ts_recent = 0; | |
201 | tp->rx_opt.ts_recent_stamp = 0; | |
ee995283 PE |
202 | if (likely(!tp->repair)) |
203 | tp->write_seq = 0; | |
1da177e4 LT |
204 | } |
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); | |
1946e672 | 221 | err = inet_hash_connect(tcp_death_row, sk); |
1da177e4 LT |
222 | if (err) |
223 | goto failure; | |
224 | ||
877d1f62 | 225 | sk_set_txhash(sk); |
9e7ceb06 | 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); |
19f6d3f3 | 237 | rt = NULL; |
1da177e4 | 238 | |
00355fa5 | 239 | if (likely(!tp->repair)) { |
00355fa5 | 240 | if (!tp->write_seq) |
84b114b9 ED |
241 | tp->write_seq = secure_tcp_seq(inet->inet_saddr, |
242 | inet->inet_daddr, | |
243 | inet->inet_sport, | |
244 | usin->sin_port); | |
5d2ed052 ED |
245 | tp->tsoffset = secure_tcp_ts_off(sock_net(sk), |
246 | inet->inet_saddr, | |
84b114b9 | 247 | inet->inet_daddr); |
00355fa5 | 248 | } |
1da177e4 | 249 | |
02f20d1e | 250 | inet->inet_id = prandom_u32(); |
1da177e4 | 251 | |
19f6d3f3 WW |
252 | if (tcp_fastopen_defer_connect(sk, &err)) |
253 | return err; | |
254 | if (err) | |
255 | goto failure; | |
256 | ||
2b916477 | 257 | err = tcp_connect(sk); |
ee995283 | 258 | |
1da177e4 LT |
259 | if (err) |
260 | goto failure; | |
261 | ||
262 | return 0; | |
263 | ||
264 | failure: | |
7174259e ACM |
265 | /* |
266 | * This unhashes the socket and releases the local port, | |
267 | * if necessary. | |
268 | */ | |
1da177e4 LT |
269 | tcp_set_state(sk, TCP_CLOSE); |
270 | ip_rt_put(rt); | |
271 | sk->sk_route_caps = 0; | |
c720c7e8 | 272 | inet->inet_dport = 0; |
1da177e4 LT |
273 | return err; |
274 | } | |
4bc2f18b | 275 | EXPORT_SYMBOL(tcp_v4_connect); |
1da177e4 | 276 | |
1da177e4 | 277 | /* |
563d34d0 ED |
278 | * This routine reacts to ICMP_FRAG_NEEDED mtu indications as defined in RFC1191. |
279 | * It can be called through tcp_release_cb() if socket was owned by user | |
280 | * at the time tcp_v4_err() was called to handle ICMP message. | |
1da177e4 | 281 | */ |
4fab9071 | 282 | void tcp_v4_mtu_reduced(struct sock *sk) |
1da177e4 | 283 | { |
1da177e4 | 284 | struct inet_sock *inet = inet_sk(sk); |
02b2faaf ED |
285 | struct dst_entry *dst; |
286 | u32 mtu; | |
1da177e4 | 287 | |
02b2faaf ED |
288 | if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) |
289 | return; | |
290 | mtu = tcp_sk(sk)->mtu_info; | |
80d0a69f DM |
291 | dst = inet_csk_update_pmtu(sk, mtu); |
292 | if (!dst) | |
1da177e4 LT |
293 | return; |
294 | ||
1da177e4 LT |
295 | /* Something is about to be wrong... Remember soft error |
296 | * for the case, if this connection will not able to recover. | |
297 | */ | |
298 | if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst)) | |
299 | sk->sk_err_soft = EMSGSIZE; | |
300 | ||
301 | mtu = dst_mtu(dst); | |
302 | ||
303 | if (inet->pmtudisc != IP_PMTUDISC_DONT && | |
482fc609 | 304 | ip_sk_accept_pmtu(sk) && |
d83d8461 | 305 | inet_csk(sk)->icsk_pmtu_cookie > mtu) { |
1da177e4 LT |
306 | tcp_sync_mss(sk, mtu); |
307 | ||
308 | /* Resend the TCP packet because it's | |
309 | * clear that the old packet has been | |
310 | * dropped. This is the new "fast" path mtu | |
311 | * discovery. | |
312 | */ | |
313 | tcp_simple_retransmit(sk); | |
314 | } /* else let the usual retransmit timer handle it */ | |
315 | } | |
4fab9071 | 316 | EXPORT_SYMBOL(tcp_v4_mtu_reduced); |
1da177e4 | 317 | |
55be7a9c DM |
318 | static void do_redirect(struct sk_buff *skb, struct sock *sk) |
319 | { | |
320 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
321 | ||
1ed5c48f | 322 | if (dst) |
6700c270 | 323 | dst->ops->redirect(dst, sk, skb); |
55be7a9c DM |
324 | } |
325 | ||
26e37360 ED |
326 | |
327 | /* handle ICMP messages on TCP_NEW_SYN_RECV request sockets */ | |
9cf74903 | 328 | void tcp_req_err(struct sock *sk, u32 seq, bool abort) |
26e37360 ED |
329 | { |
330 | struct request_sock *req = inet_reqsk(sk); | |
331 | struct net *net = sock_net(sk); | |
332 | ||
333 | /* ICMPs are not backlogged, hence we cannot get | |
334 | * an established socket here. | |
335 | */ | |
26e37360 | 336 | if (seq != tcp_rsk(req)->snt_isn) { |
02a1d6e7 | 337 | __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS); |
9cf74903 | 338 | } else if (abort) { |
26e37360 ED |
339 | /* |
340 | * Still in SYN_RECV, just remove it silently. | |
341 | * There is no good way to pass the error to the newly | |
342 | * created socket, and POSIX does not want network | |
343 | * errors returned from accept(). | |
344 | */ | |
c6973669 | 345 | inet_csk_reqsk_queue_drop(req->rsk_listener, req); |
9caad864 | 346 | tcp_listendrop(req->rsk_listener); |
26e37360 | 347 | } |
ef84d8ce | 348 | reqsk_put(req); |
26e37360 ED |
349 | } |
350 | EXPORT_SYMBOL(tcp_req_err); | |
351 | ||
1da177e4 LT |
352 | /* |
353 | * This routine is called by the ICMP module when it gets some | |
354 | * sort of error condition. If err < 0 then the socket should | |
355 | * be closed and the error returned to the user. If err > 0 | |
356 | * it's just the icmp type << 8 | icmp code. After adjustment | |
357 | * header points to the first 8 bytes of the tcp header. We need | |
358 | * to find the appropriate port. | |
359 | * | |
360 | * The locking strategy used here is very "optimistic". When | |
361 | * someone else accesses the socket the ICMP is just dropped | |
362 | * and for some paths there is no check at all. | |
363 | * A more general error queue to queue errors for later handling | |
364 | * is probably better. | |
365 | * | |
366 | */ | |
367 | ||
4d1a2d9e | 368 | void tcp_v4_err(struct sk_buff *icmp_skb, u32 info) |
1da177e4 | 369 | { |
b71d1d42 | 370 | const struct iphdr *iph = (const struct iphdr *)icmp_skb->data; |
4d1a2d9e | 371 | struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2)); |
f1ecd5d9 | 372 | struct inet_connection_sock *icsk; |
1da177e4 LT |
373 | struct tcp_sock *tp; |
374 | struct inet_sock *inet; | |
4d1a2d9e DL |
375 | const int type = icmp_hdr(icmp_skb)->type; |
376 | const int code = icmp_hdr(icmp_skb)->code; | |
1da177e4 | 377 | struct sock *sk; |
f1ecd5d9 | 378 | struct sk_buff *skb; |
0a672f74 | 379 | struct request_sock *fastopen; |
9a568de4 ED |
380 | u32 seq, snd_una; |
381 | s32 remaining; | |
382 | u32 delta_us; | |
1da177e4 | 383 | int err; |
4d1a2d9e | 384 | struct net *net = dev_net(icmp_skb->dev); |
1da177e4 | 385 | |
26e37360 ED |
386 | sk = __inet_lookup_established(net, &tcp_hashinfo, iph->daddr, |
387 | th->dest, iph->saddr, ntohs(th->source), | |
3fa6f616 | 388 | inet_iif(icmp_skb), 0); |
1da177e4 | 389 | if (!sk) { |
5d3848bc | 390 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
391 | return; |
392 | } | |
393 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 394 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
395 | return; |
396 | } | |
26e37360 ED |
397 | seq = ntohl(th->seq); |
398 | if (sk->sk_state == TCP_NEW_SYN_RECV) | |
9cf74903 ED |
399 | return tcp_req_err(sk, seq, |
400 | type == ICMP_PARAMETERPROB || | |
401 | type == ICMP_TIME_EXCEEDED || | |
402 | (type == ICMP_DEST_UNREACH && | |
403 | (code == ICMP_NET_UNREACH || | |
404 | code == ICMP_HOST_UNREACH))); | |
1da177e4 LT |
405 | |
406 | bh_lock_sock(sk); | |
407 | /* If too many ICMPs get dropped on busy | |
408 | * servers this needs to be solved differently. | |
563d34d0 ED |
409 | * We do take care of PMTU discovery (RFC1191) special case : |
410 | * we can receive locally generated ICMP messages while socket is held. | |
1da177e4 | 411 | */ |
b74aa930 ED |
412 | if (sock_owned_by_user(sk)) { |
413 | if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)) | |
02a1d6e7 | 414 | __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS); |
b74aa930 | 415 | } |
1da177e4 LT |
416 | if (sk->sk_state == TCP_CLOSE) |
417 | goto out; | |
418 | ||
97e3ecd1 | 419 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
02a1d6e7 | 420 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); |
97e3ecd1 | 421 | goto out; |
422 | } | |
423 | ||
f1ecd5d9 | 424 | icsk = inet_csk(sk); |
1da177e4 | 425 | tp = tcp_sk(sk); |
0a672f74 YC |
426 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ |
427 | fastopen = tp->fastopen_rsk; | |
428 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; | |
1da177e4 | 429 | if (sk->sk_state != TCP_LISTEN && |
0a672f74 | 430 | !between(seq, snd_una, tp->snd_nxt)) { |
02a1d6e7 | 431 | __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
432 | goto out; |
433 | } | |
434 | ||
435 | switch (type) { | |
55be7a9c | 436 | case ICMP_REDIRECT: |
45caeaa5 JM |
437 | if (!sock_owned_by_user(sk)) |
438 | do_redirect(icmp_skb, sk); | |
55be7a9c | 439 | goto out; |
1da177e4 LT |
440 | case ICMP_SOURCE_QUENCH: |
441 | /* Just silently ignore these. */ | |
442 | goto out; | |
443 | case ICMP_PARAMETERPROB: | |
444 | err = EPROTO; | |
445 | break; | |
446 | case ICMP_DEST_UNREACH: | |
447 | if (code > NR_ICMP_UNREACH) | |
448 | goto out; | |
449 | ||
450 | if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */ | |
0d4f0608 ED |
451 | /* We are not interested in TCP_LISTEN and open_requests |
452 | * (SYN-ACKs send out by Linux are always <576bytes so | |
453 | * they should go through unfragmented). | |
454 | */ | |
455 | if (sk->sk_state == TCP_LISTEN) | |
456 | goto out; | |
457 | ||
563d34d0 | 458 | tp->mtu_info = info; |
144d56e9 | 459 | if (!sock_owned_by_user(sk)) { |
563d34d0 | 460 | tcp_v4_mtu_reduced(sk); |
144d56e9 | 461 | } else { |
7aa5470c | 462 | if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &sk->sk_tsq_flags)) |
144d56e9 ED |
463 | sock_hold(sk); |
464 | } | |
1da177e4 LT |
465 | goto out; |
466 | } | |
467 | ||
468 | err = icmp_err_convert[code].errno; | |
f1ecd5d9 DL |
469 | /* check if icmp_skb allows revert of backoff |
470 | * (see draft-zimmermann-tcp-lcd) */ | |
471 | if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH) | |
472 | break; | |
473 | if (seq != tp->snd_una || !icsk->icsk_retransmits || | |
0a672f74 | 474 | !icsk->icsk_backoff || fastopen) |
f1ecd5d9 DL |
475 | break; |
476 | ||
8f49c270 DM |
477 | if (sock_owned_by_user(sk)) |
478 | break; | |
479 | ||
f26101c7 ED |
480 | skb = tcp_rtx_queue_head(sk); |
481 | if (WARN_ON_ONCE(!skb)) | |
482 | break; | |
483 | ||
f1ecd5d9 | 484 | icsk->icsk_backoff--; |
fcdd1cf4 ED |
485 | icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) : |
486 | TCP_TIMEOUT_INIT; | |
487 | icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX); | |
f1ecd5d9 | 488 | |
9a568de4 ED |
489 | tcp_mstamp_refresh(tp); |
490 | delta_us = (u32)(tp->tcp_mstamp - skb->skb_mstamp); | |
7faee5c0 | 491 | remaining = icsk->icsk_rto - |
9a568de4 | 492 | usecs_to_jiffies(delta_us); |
f1ecd5d9 | 493 | |
9a568de4 | 494 | if (remaining > 0) { |
f1ecd5d9 DL |
495 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, |
496 | remaining, TCP_RTO_MAX); | |
f1ecd5d9 DL |
497 | } else { |
498 | /* RTO revert clocked out retransmission. | |
499 | * Will retransmit now */ | |
500 | tcp_retransmit_timer(sk); | |
501 | } | |
502 | ||
1da177e4 LT |
503 | break; |
504 | case ICMP_TIME_EXCEEDED: | |
505 | err = EHOSTUNREACH; | |
506 | break; | |
507 | default: | |
508 | goto out; | |
509 | } | |
510 | ||
511 | switch (sk->sk_state) { | |
1da177e4 | 512 | case TCP_SYN_SENT: |
0a672f74 YC |
513 | case TCP_SYN_RECV: |
514 | /* Only in fast or simultaneous open. If a fast open socket is | |
515 | * is already accepted it is treated as a connected one below. | |
516 | */ | |
51456b29 | 517 | if (fastopen && !fastopen->sk) |
0a672f74 YC |
518 | break; |
519 | ||
1da177e4 | 520 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
521 | sk->sk_err = err; |
522 | ||
523 | sk->sk_error_report(sk); | |
524 | ||
525 | tcp_done(sk); | |
526 | } else { | |
527 | sk->sk_err_soft = err; | |
528 | } | |
529 | goto out; | |
530 | } | |
531 | ||
532 | /* If we've already connected we will keep trying | |
533 | * until we time out, or the user gives up. | |
534 | * | |
535 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
536 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
537 | * but it is obsoleted by pmtu discovery). | |
538 | * | |
539 | * Note, that in modern internet, where routing is unreliable | |
540 | * and in each dark corner broken firewalls sit, sending random | |
541 | * errors ordered by their masters even this two messages finally lose | |
542 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
543 | * | |
544 | * Now we are in compliance with RFCs. | |
545 | * --ANK (980905) | |
546 | */ | |
547 | ||
548 | inet = inet_sk(sk); | |
549 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
550 | sk->sk_err = err; | |
551 | sk->sk_error_report(sk); | |
552 | } else { /* Only an error on timeout */ | |
553 | sk->sk_err_soft = err; | |
554 | } | |
555 | ||
556 | out: | |
557 | bh_unlock_sock(sk); | |
558 | sock_put(sk); | |
559 | } | |
560 | ||
28850dc7 | 561 | void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 562 | { |
aa8223c7 | 563 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 564 | |
84fa7933 | 565 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
419f9f89 | 566 | th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0); |
663ead3b | 567 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 568 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 569 | } else { |
419f9f89 | 570 | th->check = tcp_v4_check(skb->len, saddr, daddr, |
07f0757a | 571 | csum_partial(th, |
1da177e4 LT |
572 | th->doff << 2, |
573 | skb->csum)); | |
574 | } | |
575 | } | |
576 | ||
419f9f89 | 577 | /* This routine computes an IPv4 TCP checksum. */ |
bb296246 | 578 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb) |
419f9f89 | 579 | { |
cf533ea5 | 580 | const struct inet_sock *inet = inet_sk(sk); |
419f9f89 HX |
581 | |
582 | __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr); | |
583 | } | |
4bc2f18b | 584 | EXPORT_SYMBOL(tcp_v4_send_check); |
419f9f89 | 585 | |
1da177e4 LT |
586 | /* |
587 | * This routine will send an RST to the other tcp. | |
588 | * | |
589 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
590 | * for reset. | |
591 | * Answer: if a packet caused RST, it is not for a socket | |
592 | * existing in our system, if it is matched to a socket, | |
593 | * it is just duplicate segment or bug in other side's TCP. | |
594 | * So that we build reply only basing on parameters | |
595 | * arrived with segment. | |
596 | * Exception: precedence violation. We do not implement it in any case. | |
597 | */ | |
598 | ||
a00e7444 | 599 | static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) |
1da177e4 | 600 | { |
cf533ea5 | 601 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
602 | struct { |
603 | struct tcphdr th; | |
604 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 605 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
606 | #endif |
607 | } rep; | |
1da177e4 | 608 | struct ip_reply_arg arg; |
cfb6eeb4 | 609 | #ifdef CONFIG_TCP_MD5SIG |
e46787f0 | 610 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
611 | const __u8 *hash_location = NULL; |
612 | unsigned char newhash[16]; | |
613 | int genhash; | |
614 | struct sock *sk1 = NULL; | |
cfb6eeb4 | 615 | #endif |
a86b1e30 | 616 | struct net *net; |
1da177e4 LT |
617 | |
618 | /* Never send a reset in response to a reset. */ | |
619 | if (th->rst) | |
620 | return; | |
621 | ||
c3658e8d ED |
622 | /* If sk not NULL, it means we did a successful lookup and incoming |
623 | * route had to be correct. prequeue might have dropped our dst. | |
624 | */ | |
625 | if (!sk && skb_rtable(skb)->rt_type != RTN_LOCAL) | |
1da177e4 LT |
626 | return; |
627 | ||
628 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
629 | memset(&rep, 0, sizeof(rep)); |
630 | rep.th.dest = th->source; | |
631 | rep.th.source = th->dest; | |
632 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
633 | rep.th.rst = 1; | |
1da177e4 LT |
634 | |
635 | if (th->ack) { | |
cfb6eeb4 | 636 | rep.th.seq = th->ack_seq; |
1da177e4 | 637 | } else { |
cfb6eeb4 YH |
638 | rep.th.ack = 1; |
639 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
640 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
641 | } |
642 | ||
7174259e | 643 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
644 | arg.iov[0].iov_base = (unsigned char *)&rep; |
645 | arg.iov[0].iov_len = sizeof(rep.th); | |
646 | ||
0f85feae | 647 | net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); |
cfb6eeb4 | 648 | #ifdef CONFIG_TCP_MD5SIG |
3b24d854 | 649 | rcu_read_lock(); |
658ddaaf | 650 | hash_location = tcp_parse_md5sig_option(th); |
271c3b9b | 651 | if (sk && sk_fullsock(sk)) { |
e46787f0 FW |
652 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *) |
653 | &ip_hdr(skb)->saddr, AF_INET); | |
654 | } else if (hash_location) { | |
658ddaaf SL |
655 | /* |
656 | * active side is lost. Try to find listening socket through | |
657 | * source port, and then find md5 key through listening socket. | |
658 | * we are not loose security here: | |
659 | * Incoming packet is checked with md5 hash with finding key, | |
660 | * no RST generated if md5 hash doesn't match. | |
661 | */ | |
a583636a CG |
662 | sk1 = __inet_lookup_listener(net, &tcp_hashinfo, NULL, 0, |
663 | ip_hdr(skb)->saddr, | |
da5e3630 | 664 | th->source, ip_hdr(skb)->daddr, |
3fa6f616 DA |
665 | ntohs(th->source), inet_iif(skb), |
666 | tcp_v4_sdif(skb)); | |
658ddaaf SL |
667 | /* don't send rst if it can't find key */ |
668 | if (!sk1) | |
3b24d854 ED |
669 | goto out; |
670 | ||
658ddaaf SL |
671 | key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *) |
672 | &ip_hdr(skb)->saddr, AF_INET); | |
673 | if (!key) | |
3b24d854 ED |
674 | goto out; |
675 | ||
658ddaaf | 676 | |
39f8e58e | 677 | genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb); |
658ddaaf | 678 | if (genhash || memcmp(hash_location, newhash, 16) != 0) |
3b24d854 ED |
679 | goto out; |
680 | ||
658ddaaf SL |
681 | } |
682 | ||
cfb6eeb4 YH |
683 | if (key) { |
684 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
685 | (TCPOPT_NOP << 16) | | |
686 | (TCPOPT_MD5SIG << 8) | | |
687 | TCPOLEN_MD5SIG); | |
688 | /* Update length and the length the header thinks exists */ | |
689 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
690 | rep.th.doff = arg.iov[0].iov_len / 4; | |
691 | ||
49a72dfb | 692 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
693 | key, ip_hdr(skb)->saddr, |
694 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
695 | } |
696 | #endif | |
eddc9ec5 ACM |
697 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
698 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 699 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 700 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
271c3b9b FW |
701 | arg.flags = (sk && inet_sk_transparent(sk)) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
702 | ||
e2446eaa | 703 | /* When socket is gone, all binding information is lost. |
4c675258 AK |
704 | * routing might fail in this case. No choice here, if we choose to force |
705 | * input interface, we will misroute in case of asymmetric route. | |
e2446eaa | 706 | */ |
c24b14c4 | 707 | if (sk) { |
4c675258 | 708 | arg.bound_dev_if = sk->sk_bound_dev_if; |
c7798abf SL |
709 | if (sk_fullsock(sk)) |
710 | trace_tcp_send_reset(sk, skb); | |
c24b14c4 | 711 | } |
1da177e4 | 712 | |
271c3b9b FW |
713 | BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) != |
714 | offsetof(struct inet_timewait_sock, tw_bound_dev_if)); | |
715 | ||
66b13d99 | 716 | arg.tos = ip_hdr(skb)->tos; |
e2d118a1 | 717 | arg.uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL); |
47dcc20a | 718 | local_bh_disable(); |
bdbbb852 ED |
719 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
720 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
721 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
722 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 723 | |
90bbcc60 ED |
724 | __TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
725 | __TCP_INC_STATS(net, TCP_MIB_OUTRSTS); | |
47dcc20a | 726 | local_bh_enable(); |
658ddaaf SL |
727 | |
728 | #ifdef CONFIG_TCP_MD5SIG | |
3b24d854 ED |
729 | out: |
730 | rcu_read_unlock(); | |
658ddaaf | 731 | #endif |
1da177e4 LT |
732 | } |
733 | ||
734 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
735 | outside socket context is ugly, certainly. What can I do? | |
736 | */ | |
737 | ||
e2d118a1 | 738 | static void tcp_v4_send_ack(const struct sock *sk, |
e62a123b | 739 | struct sk_buff *skb, u32 seq, u32 ack, |
ee684b6f | 740 | u32 win, u32 tsval, u32 tsecr, int oif, |
88ef4a5a | 741 | struct tcp_md5sig_key *key, |
66b13d99 | 742 | int reply_flags, u8 tos) |
1da177e4 | 743 | { |
cf533ea5 | 744 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
745 | struct { |
746 | struct tcphdr th; | |
714e85be | 747 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 748 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 749 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
750 | #endif |
751 | ]; | |
1da177e4 | 752 | } rep; |
e2d118a1 | 753 | struct net *net = sock_net(sk); |
1da177e4 LT |
754 | struct ip_reply_arg arg; |
755 | ||
756 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 757 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
758 | |
759 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
760 | arg.iov[0].iov_len = sizeof(rep.th); | |
ee684b6f | 761 | if (tsecr) { |
cfb6eeb4 YH |
762 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
763 | (TCPOPT_TIMESTAMP << 8) | | |
764 | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
765 | rep.opt[1] = htonl(tsval); |
766 | rep.opt[2] = htonl(tsecr); | |
cb48cfe8 | 767 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
768 | } |
769 | ||
770 | /* Swap the send and the receive. */ | |
771 | rep.th.dest = th->source; | |
772 | rep.th.source = th->dest; | |
773 | rep.th.doff = arg.iov[0].iov_len / 4; | |
774 | rep.th.seq = htonl(seq); | |
775 | rep.th.ack_seq = htonl(ack); | |
776 | rep.th.ack = 1; | |
777 | rep.th.window = htons(win); | |
778 | ||
cfb6eeb4 | 779 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 780 | if (key) { |
ee684b6f | 781 | int offset = (tsecr) ? 3 : 0; |
cfb6eeb4 YH |
782 | |
783 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
784 | (TCPOPT_NOP << 16) | | |
785 | (TCPOPT_MD5SIG << 8) | | |
786 | TCPOLEN_MD5SIG); | |
787 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
788 | rep.th.doff = arg.iov[0].iov_len/4; | |
789 | ||
49a72dfb | 790 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
791 | key, ip_hdr(skb)->saddr, |
792 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
793 | } |
794 | #endif | |
88ef4a5a | 795 | arg.flags = reply_flags; |
eddc9ec5 ACM |
796 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
797 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
798 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
799 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
800 | if (oif) |
801 | arg.bound_dev_if = oif; | |
66b13d99 | 802 | arg.tos = tos; |
e2d118a1 | 803 | arg.uid = sock_net_uid(net, sk_fullsock(sk) ? sk : NULL); |
47dcc20a | 804 | local_bh_disable(); |
bdbbb852 ED |
805 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
806 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
807 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
808 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 809 | |
90bbcc60 | 810 | __TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
47dcc20a | 811 | local_bh_enable(); |
1da177e4 LT |
812 | } |
813 | ||
814 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
815 | { | |
8feaf0c0 | 816 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 817 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 818 | |
e2d118a1 | 819 | tcp_v4_send_ack(sk, skb, |
e62a123b | 820 | tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 821 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
9a568de4 | 822 | tcp_time_stamp_raw() + tcptw->tw_ts_offset, |
9501f972 YH |
823 | tcptw->tw_ts_recent, |
824 | tw->tw_bound_dev_if, | |
88ef4a5a | 825 | tcp_twsk_md5_key(tcptw), |
66b13d99 ED |
826 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0, |
827 | tw->tw_tos | |
9501f972 | 828 | ); |
1da177e4 | 829 | |
8feaf0c0 | 830 | inet_twsk_put(tw); |
1da177e4 LT |
831 | } |
832 | ||
a00e7444 | 833 | static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb, |
7174259e | 834 | struct request_sock *req) |
1da177e4 | 835 | { |
168a8f58 JC |
836 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
837 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
838 | */ | |
e62a123b ED |
839 | u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 : |
840 | tcp_sk(sk)->snd_nxt; | |
841 | ||
20a2b49f ED |
842 | /* RFC 7323 2.3 |
843 | * The window field (SEG.WND) of every outgoing segment, with the | |
844 | * exception of <SYN> segments, MUST be right-shifted by | |
845 | * Rcv.Wind.Shift bits: | |
846 | */ | |
e2d118a1 | 847 | tcp_v4_send_ack(sk, skb, seq, |
20a2b49f ED |
848 | tcp_rsk(req)->rcv_nxt, |
849 | req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale, | |
9a568de4 | 850 | tcp_time_stamp_raw() + tcp_rsk(req)->ts_off, |
9501f972 YH |
851 | req->ts_recent, |
852 | 0, | |
30791ac4 | 853 | tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->saddr, |
a915da9b | 854 | AF_INET), |
66b13d99 ED |
855 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0, |
856 | ip_hdr(skb)->tos); | |
1da177e4 LT |
857 | } |
858 | ||
1da177e4 | 859 | /* |
9bf1d83e | 860 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 861 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
862 | * socket. |
863 | */ | |
0f935dbe | 864 | static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst, |
d6274bd8 | 865 | struct flowi *fl, |
72659ecc | 866 | struct request_sock *req, |
ca6fb065 | 867 | struct tcp_fastopen_cookie *foc, |
b3d05147 | 868 | enum tcp_synack_type synack_type) |
1da177e4 | 869 | { |
2e6599cb | 870 | const struct inet_request_sock *ireq = inet_rsk(req); |
6bd023f3 | 871 | struct flowi4 fl4; |
1da177e4 | 872 | int err = -1; |
d41db5af | 873 | struct sk_buff *skb; |
1da177e4 LT |
874 | |
875 | /* First, grab a route. */ | |
ba3f7f04 | 876 | if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL) |
fd80eb94 | 877 | return -1; |
1da177e4 | 878 | |
b3d05147 | 879 | skb = tcp_make_synack(sk, dst, req, foc, synack_type); |
1da177e4 LT |
880 | |
881 | if (skb) { | |
634fb979 | 882 | __tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr); |
1da177e4 | 883 | |
e93c388d | 884 | rcu_read_lock(); |
634fb979 ED |
885 | err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr, |
886 | ireq->ir_rmt_addr, | |
e93c388d ED |
887 | rcu_dereference(ireq->ireq_opt)); |
888 | rcu_read_unlock(); | |
b9df3cb8 | 889 | err = net_xmit_eval(err); |
1da177e4 LT |
890 | } |
891 | ||
1da177e4 LT |
892 | return err; |
893 | } | |
894 | ||
895 | /* | |
60236fdd | 896 | * IPv4 request_sock destructor. |
1da177e4 | 897 | */ |
60236fdd | 898 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 899 | { |
c92e8c02 | 900 | kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1)); |
1da177e4 LT |
901 | } |
902 | ||
cfb6eeb4 YH |
903 | #ifdef CONFIG_TCP_MD5SIG |
904 | /* | |
905 | * RFC2385 MD5 checksumming requires a mapping of | |
906 | * IP address->MD5 Key. | |
907 | * We need to maintain these in the sk structure. | |
908 | */ | |
909 | ||
910 | /* Find the Key structure for an address. */ | |
b83e3deb | 911 | struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk, |
a915da9b ED |
912 | const union tcp_md5_addr *addr, |
913 | int family) | |
cfb6eeb4 | 914 | { |
fd3a154a | 915 | const struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 916 | struct tcp_md5sig_key *key; |
fd3a154a | 917 | const struct tcp_md5sig_info *md5sig; |
6797318e ID |
918 | __be32 mask; |
919 | struct tcp_md5sig_key *best_match = NULL; | |
920 | bool match; | |
cfb6eeb4 | 921 | |
a8afca03 ED |
922 | /* caller either holds rcu_read_lock() or socket lock */ |
923 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
1e1d04e6 | 924 | lockdep_sock_is_held(sk)); |
a8afca03 | 925 | if (!md5sig) |
cfb6eeb4 | 926 | return NULL; |
083a0326 | 927 | |
b67bfe0d | 928 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { |
a915da9b ED |
929 | if (key->family != family) |
930 | continue; | |
6797318e ID |
931 | |
932 | if (family == AF_INET) { | |
933 | mask = inet_make_mask(key->prefixlen); | |
934 | match = (key->addr.a4.s_addr & mask) == | |
935 | (addr->a4.s_addr & mask); | |
936 | #if IS_ENABLED(CONFIG_IPV6) | |
937 | } else if (family == AF_INET6) { | |
938 | match = ipv6_prefix_equal(&key->addr.a6, &addr->a6, | |
939 | key->prefixlen); | |
940 | #endif | |
941 | } else { | |
942 | match = false; | |
943 | } | |
944 | ||
945 | if (match && (!best_match || | |
946 | key->prefixlen > best_match->prefixlen)) | |
947 | best_match = key; | |
948 | } | |
949 | return best_match; | |
950 | } | |
951 | EXPORT_SYMBOL(tcp_md5_do_lookup); | |
952 | ||
e8f37d57 WF |
953 | static struct tcp_md5sig_key *tcp_md5_do_lookup_exact(const struct sock *sk, |
954 | const union tcp_md5_addr *addr, | |
955 | int family, u8 prefixlen) | |
6797318e ID |
956 | { |
957 | const struct tcp_sock *tp = tcp_sk(sk); | |
958 | struct tcp_md5sig_key *key; | |
959 | unsigned int size = sizeof(struct in_addr); | |
960 | const struct tcp_md5sig_info *md5sig; | |
961 | ||
962 | /* caller either holds rcu_read_lock() or socket lock */ | |
963 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
964 | lockdep_sock_is_held(sk)); | |
965 | if (!md5sig) | |
966 | return NULL; | |
967 | #if IS_ENABLED(CONFIG_IPV6) | |
968 | if (family == AF_INET6) | |
969 | size = sizeof(struct in6_addr); | |
970 | #endif | |
971 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { | |
972 | if (key->family != family) | |
973 | continue; | |
974 | if (!memcmp(&key->addr, addr, size) && | |
975 | key->prefixlen == prefixlen) | |
a915da9b | 976 | return key; |
cfb6eeb4 YH |
977 | } |
978 | return NULL; | |
979 | } | |
980 | ||
b83e3deb | 981 | struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk, |
fd3a154a | 982 | const struct sock *addr_sk) |
cfb6eeb4 | 983 | { |
b52e6921 | 984 | const union tcp_md5_addr *addr; |
a915da9b | 985 | |
b52e6921 | 986 | addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr; |
a915da9b | 987 | return tcp_md5_do_lookup(sk, addr, AF_INET); |
cfb6eeb4 | 988 | } |
cfb6eeb4 YH |
989 | EXPORT_SYMBOL(tcp_v4_md5_lookup); |
990 | ||
cfb6eeb4 | 991 | /* This can be called on a newly created socket, from other files */ |
a915da9b | 992 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
6797318e ID |
993 | int family, u8 prefixlen, const u8 *newkey, u8 newkeylen, |
994 | gfp_t gfp) | |
cfb6eeb4 YH |
995 | { |
996 | /* Add Key to the list */ | |
b0a713e9 | 997 | struct tcp_md5sig_key *key; |
cfb6eeb4 | 998 | struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 999 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 1000 | |
6797318e | 1001 | key = tcp_md5_do_lookup_exact(sk, addr, family, prefixlen); |
cfb6eeb4 YH |
1002 | if (key) { |
1003 | /* Pre-existing entry - just update that one. */ | |
a915da9b | 1004 | memcpy(key->key, newkey, newkeylen); |
b0a713e9 | 1005 | key->keylen = newkeylen; |
a915da9b ED |
1006 | return 0; |
1007 | } | |
260fcbeb | 1008 | |
a8afca03 | 1009 | md5sig = rcu_dereference_protected(tp->md5sig_info, |
1e1d04e6 | 1010 | lockdep_sock_is_held(sk)); |
a915da9b ED |
1011 | if (!md5sig) { |
1012 | md5sig = kmalloc(sizeof(*md5sig), gfp); | |
1013 | if (!md5sig) | |
cfb6eeb4 | 1014 | return -ENOMEM; |
cfb6eeb4 | 1015 | |
a915da9b ED |
1016 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
1017 | INIT_HLIST_HEAD(&md5sig->head); | |
a8afca03 | 1018 | rcu_assign_pointer(tp->md5sig_info, md5sig); |
a915da9b | 1019 | } |
cfb6eeb4 | 1020 | |
5f3d9cb2 | 1021 | key = sock_kmalloc(sk, sizeof(*key), gfp); |
a915da9b ED |
1022 | if (!key) |
1023 | return -ENOMEM; | |
71cea17e | 1024 | if (!tcp_alloc_md5sig_pool()) { |
5f3d9cb2 | 1025 | sock_kfree_s(sk, key, sizeof(*key)); |
a915da9b | 1026 | return -ENOMEM; |
cfb6eeb4 | 1027 | } |
a915da9b ED |
1028 | |
1029 | memcpy(key->key, newkey, newkeylen); | |
1030 | key->keylen = newkeylen; | |
1031 | key->family = family; | |
6797318e | 1032 | key->prefixlen = prefixlen; |
a915da9b ED |
1033 | memcpy(&key->addr, addr, |
1034 | (family == AF_INET6) ? sizeof(struct in6_addr) : | |
1035 | sizeof(struct in_addr)); | |
1036 | hlist_add_head_rcu(&key->node, &md5sig->head); | |
cfb6eeb4 YH |
1037 | return 0; |
1038 | } | |
a915da9b | 1039 | EXPORT_SYMBOL(tcp_md5_do_add); |
cfb6eeb4 | 1040 | |
6797318e ID |
1041 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family, |
1042 | u8 prefixlen) | |
cfb6eeb4 | 1043 | { |
a915da9b ED |
1044 | struct tcp_md5sig_key *key; |
1045 | ||
6797318e | 1046 | key = tcp_md5_do_lookup_exact(sk, addr, family, prefixlen); |
a915da9b ED |
1047 | if (!key) |
1048 | return -ENOENT; | |
1049 | hlist_del_rcu(&key->node); | |
5f3d9cb2 | 1050 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1051 | kfree_rcu(key, rcu); |
a915da9b | 1052 | return 0; |
cfb6eeb4 | 1053 | } |
a915da9b | 1054 | EXPORT_SYMBOL(tcp_md5_do_del); |
cfb6eeb4 | 1055 | |
e0683e70 | 1056 | static void tcp_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
1057 | { |
1058 | struct tcp_sock *tp = tcp_sk(sk); | |
a915da9b | 1059 | struct tcp_md5sig_key *key; |
b67bfe0d | 1060 | struct hlist_node *n; |
a8afca03 | 1061 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 1062 | |
a8afca03 ED |
1063 | md5sig = rcu_dereference_protected(tp->md5sig_info, 1); |
1064 | ||
b67bfe0d | 1065 | hlist_for_each_entry_safe(key, n, &md5sig->head, node) { |
a915da9b | 1066 | hlist_del_rcu(&key->node); |
5f3d9cb2 | 1067 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1068 | kfree_rcu(key, rcu); |
cfb6eeb4 YH |
1069 | } |
1070 | } | |
1071 | ||
8917a777 ID |
1072 | static int tcp_v4_parse_md5_keys(struct sock *sk, int optname, |
1073 | char __user *optval, int optlen) | |
cfb6eeb4 YH |
1074 | { |
1075 | struct tcp_md5sig cmd; | |
1076 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
8917a777 | 1077 | u8 prefixlen = 32; |
cfb6eeb4 YH |
1078 | |
1079 | if (optlen < sizeof(cmd)) | |
1080 | return -EINVAL; | |
1081 | ||
7174259e | 1082 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
1083 | return -EFAULT; |
1084 | ||
1085 | if (sin->sin_family != AF_INET) | |
1086 | return -EINVAL; | |
1087 | ||
8917a777 ID |
1088 | if (optname == TCP_MD5SIG_EXT && |
1089 | cmd.tcpm_flags & TCP_MD5SIG_FLAG_PREFIX) { | |
1090 | prefixlen = cmd.tcpm_prefixlen; | |
1091 | if (prefixlen > 32) | |
1092 | return -EINVAL; | |
1093 | } | |
1094 | ||
64a124ed | 1095 | if (!cmd.tcpm_keylen) |
a915da9b | 1096 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
8917a777 | 1097 | AF_INET, prefixlen); |
cfb6eeb4 YH |
1098 | |
1099 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
1100 | return -EINVAL; | |
1101 | ||
a915da9b | 1102 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
8917a777 | 1103 | AF_INET, prefixlen, cmd.tcpm_key, cmd.tcpm_keylen, |
a915da9b | 1104 | GFP_KERNEL); |
cfb6eeb4 YH |
1105 | } |
1106 | ||
19689e38 ED |
1107 | static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp, |
1108 | __be32 daddr, __be32 saddr, | |
1109 | const struct tcphdr *th, int nbytes) | |
cfb6eeb4 | 1110 | { |
cfb6eeb4 | 1111 | struct tcp4_pseudohdr *bp; |
49a72dfb | 1112 | struct scatterlist sg; |
19689e38 | 1113 | struct tcphdr *_th; |
cfb6eeb4 | 1114 | |
19689e38 | 1115 | bp = hp->scratch; |
cfb6eeb4 YH |
1116 | bp->saddr = saddr; |
1117 | bp->daddr = daddr; | |
1118 | bp->pad = 0; | |
076fb722 | 1119 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1120 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1121 | |
19689e38 ED |
1122 | _th = (struct tcphdr *)(bp + 1); |
1123 | memcpy(_th, th, sizeof(*th)); | |
1124 | _th->check = 0; | |
1125 | ||
1126 | sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th)); | |
1127 | ahash_request_set_crypt(hp->md5_req, &sg, NULL, | |
1128 | sizeof(*bp) + sizeof(*th)); | |
cf80e0e4 | 1129 | return crypto_ahash_update(hp->md5_req); |
49a72dfb AL |
1130 | } |
1131 | ||
a915da9b | 1132 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 1133 | __be32 daddr, __be32 saddr, const struct tcphdr *th) |
49a72dfb AL |
1134 | { |
1135 | struct tcp_md5sig_pool *hp; | |
cf80e0e4 | 1136 | struct ahash_request *req; |
49a72dfb AL |
1137 | |
1138 | hp = tcp_get_md5sig_pool(); | |
1139 | if (!hp) | |
1140 | goto clear_hash_noput; | |
cf80e0e4 | 1141 | req = hp->md5_req; |
49a72dfb | 1142 | |
cf80e0e4 | 1143 | if (crypto_ahash_init(req)) |
49a72dfb | 1144 | goto clear_hash; |
19689e38 | 1145 | if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2)) |
49a72dfb AL |
1146 | goto clear_hash; |
1147 | if (tcp_md5_hash_key(hp, key)) | |
1148 | goto clear_hash; | |
cf80e0e4 HX |
1149 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
1150 | if (crypto_ahash_final(req)) | |
cfb6eeb4 YH |
1151 | goto clear_hash; |
1152 | ||
cfb6eeb4 | 1153 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1154 | return 0; |
49a72dfb | 1155 | |
cfb6eeb4 YH |
1156 | clear_hash: |
1157 | tcp_put_md5sig_pool(); | |
1158 | clear_hash_noput: | |
1159 | memset(md5_hash, 0, 16); | |
49a72dfb | 1160 | return 1; |
cfb6eeb4 YH |
1161 | } |
1162 | ||
39f8e58e ED |
1163 | int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, |
1164 | const struct sock *sk, | |
318cf7aa | 1165 | const struct sk_buff *skb) |
cfb6eeb4 | 1166 | { |
49a72dfb | 1167 | struct tcp_md5sig_pool *hp; |
cf80e0e4 | 1168 | struct ahash_request *req; |
318cf7aa | 1169 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
1170 | __be32 saddr, daddr; |
1171 | ||
39f8e58e ED |
1172 | if (sk) { /* valid for establish/request sockets */ |
1173 | saddr = sk->sk_rcv_saddr; | |
1174 | daddr = sk->sk_daddr; | |
cfb6eeb4 | 1175 | } else { |
49a72dfb AL |
1176 | const struct iphdr *iph = ip_hdr(skb); |
1177 | saddr = iph->saddr; | |
1178 | daddr = iph->daddr; | |
cfb6eeb4 | 1179 | } |
49a72dfb AL |
1180 | |
1181 | hp = tcp_get_md5sig_pool(); | |
1182 | if (!hp) | |
1183 | goto clear_hash_noput; | |
cf80e0e4 | 1184 | req = hp->md5_req; |
49a72dfb | 1185 | |
cf80e0e4 | 1186 | if (crypto_ahash_init(req)) |
49a72dfb AL |
1187 | goto clear_hash; |
1188 | ||
19689e38 | 1189 | if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, skb->len)) |
49a72dfb AL |
1190 | goto clear_hash; |
1191 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1192 | goto clear_hash; | |
1193 | if (tcp_md5_hash_key(hp, key)) | |
1194 | goto clear_hash; | |
cf80e0e4 HX |
1195 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
1196 | if (crypto_ahash_final(req)) | |
49a72dfb AL |
1197 | goto clear_hash; |
1198 | ||
1199 | tcp_put_md5sig_pool(); | |
1200 | return 0; | |
1201 | ||
1202 | clear_hash: | |
1203 | tcp_put_md5sig_pool(); | |
1204 | clear_hash_noput: | |
1205 | memset(md5_hash, 0, 16); | |
1206 | return 1; | |
cfb6eeb4 | 1207 | } |
49a72dfb | 1208 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1209 | |
ba8e275a ED |
1210 | #endif |
1211 | ||
ff74e23f | 1212 | /* Called with rcu_read_lock() */ |
ba8e275a | 1213 | static bool tcp_v4_inbound_md5_hash(const struct sock *sk, |
ff74e23f | 1214 | const struct sk_buff *skb) |
cfb6eeb4 | 1215 | { |
ba8e275a | 1216 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
1217 | /* |
1218 | * This gets called for each TCP segment that arrives | |
1219 | * so we want to be efficient. | |
1220 | * We have 3 drop cases: | |
1221 | * o No MD5 hash and one expected. | |
1222 | * o MD5 hash and we're not expecting one. | |
1223 | * o MD5 hash and its wrong. | |
1224 | */ | |
cf533ea5 | 1225 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 1226 | struct tcp_md5sig_key *hash_expected; |
eddc9ec5 | 1227 | const struct iphdr *iph = ip_hdr(skb); |
cf533ea5 | 1228 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1229 | int genhash; |
cfb6eeb4 YH |
1230 | unsigned char newhash[16]; |
1231 | ||
a915da9b ED |
1232 | hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr, |
1233 | AF_INET); | |
7d5d5525 | 1234 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1235 | |
cfb6eeb4 YH |
1236 | /* We've parsed the options - do we have a hash? */ |
1237 | if (!hash_expected && !hash_location) | |
a2a385d6 | 1238 | return false; |
cfb6eeb4 YH |
1239 | |
1240 | if (hash_expected && !hash_location) { | |
c10d9310 | 1241 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
a2a385d6 | 1242 | return true; |
cfb6eeb4 YH |
1243 | } |
1244 | ||
1245 | if (!hash_expected && hash_location) { | |
c10d9310 | 1246 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
a2a385d6 | 1247 | return true; |
cfb6eeb4 YH |
1248 | } |
1249 | ||
1250 | /* Okay, so this is hash_expected and hash_location - | |
1251 | * so we need to calculate the checksum. | |
1252 | */ | |
49a72dfb AL |
1253 | genhash = tcp_v4_md5_hash_skb(newhash, |
1254 | hash_expected, | |
39f8e58e | 1255 | NULL, skb); |
cfb6eeb4 YH |
1256 | |
1257 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
72145a68 | 1258 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE); |
e87cc472 JP |
1259 | net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1260 | &iph->saddr, ntohs(th->source), | |
1261 | &iph->daddr, ntohs(th->dest), | |
1262 | genhash ? " tcp_v4_calc_md5_hash failed" | |
1263 | : ""); | |
a2a385d6 | 1264 | return true; |
cfb6eeb4 | 1265 | } |
a2a385d6 | 1266 | return false; |
cfb6eeb4 | 1267 | #endif |
ba8e275a ED |
1268 | return false; |
1269 | } | |
cfb6eeb4 | 1270 | |
b40cf18e ED |
1271 | static void tcp_v4_init_req(struct request_sock *req, |
1272 | const struct sock *sk_listener, | |
16bea70a OP |
1273 | struct sk_buff *skb) |
1274 | { | |
1275 | struct inet_request_sock *ireq = inet_rsk(req); | |
c92e8c02 | 1276 | struct net *net = sock_net(sk_listener); |
16bea70a | 1277 | |
08d2cc3b ED |
1278 | sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr); |
1279 | sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr); | |
c92e8c02 | 1280 | RCU_INIT_POINTER(ireq->ireq_opt, tcp_v4_save_options(net, skb)); |
16bea70a OP |
1281 | } |
1282 | ||
f964629e ED |
1283 | static struct dst_entry *tcp_v4_route_req(const struct sock *sk, |
1284 | struct flowi *fl, | |
4396e461 | 1285 | const struct request_sock *req) |
d94e0417 | 1286 | { |
4396e461 | 1287 | return inet_csk_route_req(sk, &fl->u.ip4, req); |
d94e0417 OP |
1288 | } |
1289 | ||
72a3effa | 1290 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1291 | .family = PF_INET, |
2e6599cb | 1292 | .obj_size = sizeof(struct tcp_request_sock), |
5db92c99 | 1293 | .rtx_syn_ack = tcp_rtx_synack, |
60236fdd ACM |
1294 | .send_ack = tcp_v4_reqsk_send_ack, |
1295 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 | 1296 | .send_reset = tcp_v4_send_reset, |
688d1945 | 1297 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
1298 | }; |
1299 | ||
b2e4b3de | 1300 | static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
2aec4a29 | 1301 | .mss_clamp = TCP_MSS_DEFAULT, |
16bea70a | 1302 | #ifdef CONFIG_TCP_MD5SIG |
fd3a154a | 1303 | .req_md5_lookup = tcp_v4_md5_lookup, |
e3afe7b7 | 1304 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
b6332e6c | 1305 | #endif |
16bea70a | 1306 | .init_req = tcp_v4_init_req, |
fb7b37a7 OP |
1307 | #ifdef CONFIG_SYN_COOKIES |
1308 | .cookie_init_seq = cookie_v4_init_sequence, | |
1309 | #endif | |
d94e0417 | 1310 | .route_req = tcp_v4_route_req, |
84b114b9 ED |
1311 | .init_seq = tcp_v4_init_seq, |
1312 | .init_ts_off = tcp_v4_init_ts_off, | |
d6274bd8 | 1313 | .send_synack = tcp_v4_send_synack, |
16bea70a | 1314 | }; |
cfb6eeb4 | 1315 | |
1da177e4 LT |
1316 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1317 | { | |
1da177e4 | 1318 | /* Never answer to SYNs send to broadcast or multicast */ |
511c3f92 | 1319 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1320 | goto drop; |
1321 | ||
1fb6f159 OP |
1322 | return tcp_conn_request(&tcp_request_sock_ops, |
1323 | &tcp_request_sock_ipv4_ops, sk, skb); | |
1da177e4 | 1324 | |
1da177e4 | 1325 | drop: |
9caad864 | 1326 | tcp_listendrop(sk); |
1da177e4 LT |
1327 | return 0; |
1328 | } | |
4bc2f18b | 1329 | EXPORT_SYMBOL(tcp_v4_conn_request); |
1da177e4 LT |
1330 | |
1331 | ||
1332 | /* | |
1333 | * The three way handshake has completed - we got a valid synack - | |
1334 | * now create the new socket. | |
1335 | */ | |
0c27171e | 1336 | struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb, |
60236fdd | 1337 | struct request_sock *req, |
5e0724d0 ED |
1338 | struct dst_entry *dst, |
1339 | struct request_sock *req_unhash, | |
1340 | bool *own_req) | |
1da177e4 | 1341 | { |
2e6599cb | 1342 | struct inet_request_sock *ireq; |
1da177e4 LT |
1343 | struct inet_sock *newinet; |
1344 | struct tcp_sock *newtp; | |
1345 | struct sock *newsk; | |
cfb6eeb4 YH |
1346 | #ifdef CONFIG_TCP_MD5SIG |
1347 | struct tcp_md5sig_key *key; | |
1348 | #endif | |
f6d8bd05 | 1349 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
1350 | |
1351 | if (sk_acceptq_is_full(sk)) | |
1352 | goto exit_overflow; | |
1353 | ||
1da177e4 LT |
1354 | newsk = tcp_create_openreq_child(sk, req, skb); |
1355 | if (!newsk) | |
093d2823 | 1356 | goto exit_nonewsk; |
1da177e4 | 1357 | |
bcd76111 | 1358 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
fae6ef87 | 1359 | inet_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1360 | |
1361 | newtp = tcp_sk(newsk); | |
1362 | newinet = inet_sk(newsk); | |
2e6599cb | 1363 | ireq = inet_rsk(req); |
d1e559d0 ED |
1364 | sk_daddr_set(newsk, ireq->ir_rmt_addr); |
1365 | sk_rcv_saddr_set(newsk, ireq->ir_loc_addr); | |
6dd9a14e | 1366 | newsk->sk_bound_dev_if = ireq->ir_iif; |
c92e8c02 ED |
1367 | newinet->inet_saddr = ireq->ir_loc_addr; |
1368 | inet_opt = rcu_dereference(ireq->ireq_opt); | |
1369 | RCU_INIT_POINTER(newinet->inet_opt, inet_opt); | |
463c84b9 | 1370 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1371 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
4c507d28 | 1372 | newinet->rcv_tos = ip_hdr(skb)->tos; |
d83d8461 | 1373 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
1374 | if (inet_opt) |
1375 | inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
02f20d1e | 1376 | newinet->inet_id = prandom_u32(); |
1da177e4 | 1377 | |
dfd25fff ED |
1378 | if (!dst) { |
1379 | dst = inet_csk_route_child_sock(sk, newsk, req); | |
1380 | if (!dst) | |
1381 | goto put_and_exit; | |
1382 | } else { | |
1383 | /* syncookie case : see end of cookie_v4_check() */ | |
1384 | } | |
0e734419 DM |
1385 | sk_setup_caps(newsk, dst); |
1386 | ||
81164413 DB |
1387 | tcp_ca_openreq_child(newsk, dst); |
1388 | ||
1da177e4 | 1389 | tcp_sync_mss(newsk, dst_mtu(dst)); |
3541f9e8 | 1390 | newtp->advmss = tcp_mss_clamp(tcp_sk(sk), dst_metric_advmss(dst)); |
f5fff5dc | 1391 | |
1da177e4 LT |
1392 | tcp_initialize_rcv_mss(newsk); |
1393 | ||
cfb6eeb4 YH |
1394 | #ifdef CONFIG_TCP_MD5SIG |
1395 | /* Copy over the MD5 key from the original socket */ | |
a915da9b ED |
1396 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1397 | AF_INET); | |
00db4124 | 1398 | if (key) { |
cfb6eeb4 YH |
1399 | /* |
1400 | * We're using one, so create a matching key | |
1401 | * on the newsk structure. If we fail to get | |
1402 | * memory, then we end up not copying the key | |
1403 | * across. Shucks. | |
1404 | */ | |
a915da9b | 1405 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newinet->inet_daddr, |
6797318e | 1406 | AF_INET, 32, key->key, key->keylen, GFP_ATOMIC); |
a465419b | 1407 | sk_nocaps_add(newsk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1408 | } |
1409 | #endif | |
1410 | ||
0e734419 DM |
1411 | if (__inet_inherit_port(sk, newsk) < 0) |
1412 | goto put_and_exit; | |
5e0724d0 | 1413 | *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash)); |
c92e8c02 | 1414 | if (likely(*own_req)) { |
49a496c9 | 1415 | tcp_move_syn(newtp, req); |
c92e8c02 ED |
1416 | ireq->ireq_opt = NULL; |
1417 | } else { | |
1418 | newinet->inet_opt = NULL; | |
1419 | } | |
1da177e4 LT |
1420 | return newsk; |
1421 | ||
1422 | exit_overflow: | |
c10d9310 | 1423 | NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 BS |
1424 | exit_nonewsk: |
1425 | dst_release(dst); | |
1da177e4 | 1426 | exit: |
9caad864 | 1427 | tcp_listendrop(sk); |
1da177e4 | 1428 | return NULL; |
0e734419 | 1429 | put_and_exit: |
c92e8c02 | 1430 | newinet->inet_opt = NULL; |
e337e24d CP |
1431 | inet_csk_prepare_forced_close(newsk); |
1432 | tcp_done(newsk); | |
0e734419 | 1433 | goto exit; |
1da177e4 | 1434 | } |
4bc2f18b | 1435 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); |
1da177e4 | 1436 | |
079096f1 | 1437 | static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1438 | { |
079096f1 | 1439 | #ifdef CONFIG_SYN_COOKIES |
52452c54 | 1440 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 1441 | |
af9b4738 | 1442 | if (!th->syn) |
461b74c3 | 1443 | sk = cookie_v4_check(sk, skb); |
1da177e4 LT |
1444 | #endif |
1445 | return sk; | |
1446 | } | |
1447 | ||
1da177e4 | 1448 | /* The socket must have it's spinlock held when we get |
e994b2f0 | 1449 | * here, unless it is a TCP_LISTEN socket. |
1da177e4 LT |
1450 | * |
1451 | * We have a potential double-lock case here, so even when | |
1452 | * doing backlog processing we use the BH locking scheme. | |
1453 | * This is because we cannot sleep with the original spinlock | |
1454 | * held. | |
1455 | */ | |
1456 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1457 | { | |
cfb6eeb4 | 1458 | struct sock *rsk; |
cfb6eeb4 | 1459 | |
1da177e4 | 1460 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
404e0a8b ED |
1461 | struct dst_entry *dst = sk->sk_rx_dst; |
1462 | ||
bdeab991 | 1463 | sock_rps_save_rxhash(sk, skb); |
3d97379a | 1464 | sk_mark_napi_id(sk, skb); |
404e0a8b | 1465 | if (dst) { |
505fbcf0 | 1466 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || |
51456b29 | 1467 | !dst->ops->check(dst, 0)) { |
92101b3b DM |
1468 | dst_release(dst); |
1469 | sk->sk_rx_dst = NULL; | |
1470 | } | |
1471 | } | |
e42e24c3 | 1472 | tcp_rcv_established(sk, skb, tcp_hdr(skb)); |
1da177e4 LT |
1473 | return 0; |
1474 | } | |
1475 | ||
12e25e10 | 1476 | if (tcp_checksum_complete(skb)) |
1da177e4 LT |
1477 | goto csum_err; |
1478 | ||
1479 | if (sk->sk_state == TCP_LISTEN) { | |
079096f1 ED |
1480 | struct sock *nsk = tcp_v4_cookie_check(sk, skb); |
1481 | ||
1da177e4 LT |
1482 | if (!nsk) |
1483 | goto discard; | |
1da177e4 | 1484 | if (nsk != sk) { |
cfb6eeb4 YH |
1485 | if (tcp_child_process(sk, nsk, skb)) { |
1486 | rsk = nsk; | |
1da177e4 | 1487 | goto reset; |
cfb6eeb4 | 1488 | } |
1da177e4 LT |
1489 | return 0; |
1490 | } | |
ca55158c | 1491 | } else |
bdeab991 | 1492 | sock_rps_save_rxhash(sk, skb); |
ca55158c | 1493 | |
72ab4a86 | 1494 | if (tcp_rcv_state_process(sk, skb)) { |
cfb6eeb4 | 1495 | rsk = sk; |
1da177e4 | 1496 | goto reset; |
cfb6eeb4 | 1497 | } |
1da177e4 LT |
1498 | return 0; |
1499 | ||
1500 | reset: | |
cfb6eeb4 | 1501 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1502 | discard: |
1503 | kfree_skb(skb); | |
1504 | /* Be careful here. If this function gets more complicated and | |
1505 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1506 | * might be destroyed here. This current version compiles correctly, | |
1507 | * but you have been warned. | |
1508 | */ | |
1509 | return 0; | |
1510 | ||
1511 | csum_err: | |
c10d9310 ED |
1512 | TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS); |
1513 | TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS); | |
1da177e4 LT |
1514 | goto discard; |
1515 | } | |
4bc2f18b | 1516 | EXPORT_SYMBOL(tcp_v4_do_rcv); |
1da177e4 | 1517 | |
7487449c | 1518 | int tcp_v4_early_demux(struct sk_buff *skb) |
41063e9d | 1519 | { |
41063e9d DM |
1520 | const struct iphdr *iph; |
1521 | const struct tcphdr *th; | |
1522 | struct sock *sk; | |
41063e9d | 1523 | |
41063e9d | 1524 | if (skb->pkt_type != PACKET_HOST) |
7487449c | 1525 | return 0; |
41063e9d | 1526 | |
45f00f99 | 1527 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) |
7487449c | 1528 | return 0; |
41063e9d DM |
1529 | |
1530 | iph = ip_hdr(skb); | |
45f00f99 | 1531 | th = tcp_hdr(skb); |
41063e9d DM |
1532 | |
1533 | if (th->doff < sizeof(struct tcphdr) / 4) | |
7487449c | 1534 | return 0; |
41063e9d | 1535 | |
45f00f99 | 1536 | sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo, |
41063e9d | 1537 | iph->saddr, th->source, |
7011d085 | 1538 | iph->daddr, ntohs(th->dest), |
3fa6f616 | 1539 | skb->skb_iif, inet_sdif(skb)); |
41063e9d DM |
1540 | if (sk) { |
1541 | skb->sk = sk; | |
1542 | skb->destructor = sock_edemux; | |
f7e4eb03 | 1543 | if (sk_fullsock(sk)) { |
d0c294c5 | 1544 | struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst); |
505fbcf0 | 1545 | |
41063e9d DM |
1546 | if (dst) |
1547 | dst = dst_check(dst, 0); | |
92101b3b | 1548 | if (dst && |
505fbcf0 | 1549 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
92101b3b | 1550 | skb_dst_set_noref(skb, dst); |
41063e9d DM |
1551 | } |
1552 | } | |
7487449c | 1553 | return 0; |
41063e9d DM |
1554 | } |
1555 | ||
c9c33212 ED |
1556 | bool tcp_add_backlog(struct sock *sk, struct sk_buff *skb) |
1557 | { | |
1558 | u32 limit = sk->sk_rcvbuf + sk->sk_sndbuf; | |
1559 | ||
1560 | /* Only socket owner can try to collapse/prune rx queues | |
1561 | * to reduce memory overhead, so add a little headroom here. | |
1562 | * Few sockets backlog are possibly concurrently non empty. | |
1563 | */ | |
1564 | limit += 64*1024; | |
1565 | ||
1566 | /* In case all data was pulled from skb frags (in __pskb_pull_tail()), | |
1567 | * we can fix skb->truesize to its real value to avoid future drops. | |
1568 | * This is valid because skb is not yet charged to the socket. | |
1569 | * It has been noticed pure SACK packets were sometimes dropped | |
1570 | * (if cooked by drivers without copybreak feature). | |
1571 | */ | |
60b1af33 | 1572 | skb_condense(skb); |
c9c33212 ED |
1573 | |
1574 | if (unlikely(sk_add_backlog(sk, skb, limit))) { | |
1575 | bh_unlock_sock(sk); | |
1576 | __NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPBACKLOGDROP); | |
1577 | return true; | |
1578 | } | |
1579 | return false; | |
1580 | } | |
1581 | EXPORT_SYMBOL(tcp_add_backlog); | |
1582 | ||
ac6e7800 ED |
1583 | int tcp_filter(struct sock *sk, struct sk_buff *skb) |
1584 | { | |
1585 | struct tcphdr *th = (struct tcphdr *)skb->data; | |
ac6e7800 | 1586 | |
9f7dd87a | 1587 | return sk_filter_trim_cap(sk, skb, th->doff * 4); |
ac6e7800 ED |
1588 | } |
1589 | EXPORT_SYMBOL(tcp_filter); | |
1590 | ||
eeea10b8 ED |
1591 | static void tcp_v4_restore_cb(struct sk_buff *skb) |
1592 | { | |
1593 | memmove(IPCB(skb), &TCP_SKB_CB(skb)->header.h4, | |
1594 | sizeof(struct inet_skb_parm)); | |
1595 | } | |
1596 | ||
1597 | static void tcp_v4_fill_cb(struct sk_buff *skb, const struct iphdr *iph, | |
1598 | const struct tcphdr *th) | |
1599 | { | |
1600 | /* This is tricky : We move IPCB at its correct location into TCP_SKB_CB() | |
1601 | * barrier() makes sure compiler wont play fool^Waliasing games. | |
1602 | */ | |
1603 | memmove(&TCP_SKB_CB(skb)->header.h4, IPCB(skb), | |
1604 | sizeof(struct inet_skb_parm)); | |
1605 | barrier(); | |
1606 | ||
1607 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); | |
1608 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1609 | skb->len - th->doff * 4); | |
1610 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1611 | TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th); | |
1612 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; | |
1613 | TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph); | |
1614 | TCP_SKB_CB(skb)->sacked = 0; | |
1615 | TCP_SKB_CB(skb)->has_rxtstamp = | |
1616 | skb->tstamp || skb_hwtstamps(skb)->hwtstamp; | |
1617 | } | |
1618 | ||
1da177e4 LT |
1619 | /* |
1620 | * From tcp_input.c | |
1621 | */ | |
1622 | ||
1623 | int tcp_v4_rcv(struct sk_buff *skb) | |
1624 | { | |
3b24d854 | 1625 | struct net *net = dev_net(skb->dev); |
3fa6f616 | 1626 | int sdif = inet_sdif(skb); |
eddc9ec5 | 1627 | const struct iphdr *iph; |
cf533ea5 | 1628 | const struct tcphdr *th; |
3b24d854 | 1629 | bool refcounted; |
1da177e4 LT |
1630 | struct sock *sk; |
1631 | int ret; | |
1632 | ||
1633 | if (skb->pkt_type != PACKET_HOST) | |
1634 | goto discard_it; | |
1635 | ||
1636 | /* Count it even if it's bad */ | |
90bbcc60 | 1637 | __TCP_INC_STATS(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1638 | |
1639 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1640 | goto discard_it; | |
1641 | ||
ea1627c2 | 1642 | th = (const struct tcphdr *)skb->data; |
1da177e4 | 1643 | |
ea1627c2 | 1644 | if (unlikely(th->doff < sizeof(struct tcphdr) / 4)) |
1da177e4 LT |
1645 | goto bad_packet; |
1646 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1647 | goto discard_it; | |
1648 | ||
1649 | /* An explanation is required here, I think. | |
1650 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1651 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1652 | * So, we defer the checks. */ |
ed70fcfc TH |
1653 | |
1654 | if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo)) | |
6a5dc9e5 | 1655 | goto csum_error; |
1da177e4 | 1656 | |
ea1627c2 | 1657 | th = (const struct tcphdr *)skb->data; |
eddc9ec5 | 1658 | iph = ip_hdr(skb); |
4bdc3d66 | 1659 | lookup: |
a583636a | 1660 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source, |
3fa6f616 | 1661 | th->dest, sdif, &refcounted); |
1da177e4 LT |
1662 | if (!sk) |
1663 | goto no_tcp_socket; | |
1664 | ||
bb134d5d ED |
1665 | process: |
1666 | if (sk->sk_state == TCP_TIME_WAIT) | |
1667 | goto do_time_wait; | |
1668 | ||
079096f1 ED |
1669 | if (sk->sk_state == TCP_NEW_SYN_RECV) { |
1670 | struct request_sock *req = inet_reqsk(sk); | |
7716682c | 1671 | struct sock *nsk; |
079096f1 ED |
1672 | |
1673 | sk = req->rsk_listener; | |
72923555 | 1674 | if (unlikely(tcp_v4_inbound_md5_hash(sk, skb))) { |
e65c332d | 1675 | sk_drops_add(sk, skb); |
72923555 ED |
1676 | reqsk_put(req); |
1677 | goto discard_it; | |
1678 | } | |
dd319c07 FL |
1679 | if (tcp_checksum_complete(skb)) { |
1680 | reqsk_put(req); | |
1681 | goto csum_error; | |
1682 | } | |
7716682c | 1683 | if (unlikely(sk->sk_state != TCP_LISTEN)) { |
f03f2e15 | 1684 | inet_csk_reqsk_queue_drop_and_put(sk, req); |
4bdc3d66 ED |
1685 | goto lookup; |
1686 | } | |
3b24d854 ED |
1687 | /* We own a reference on the listener, increase it again |
1688 | * as we might lose it too soon. | |
1689 | */ | |
7716682c | 1690 | sock_hold(sk); |
3b24d854 | 1691 | refcounted = true; |
1f3b359f | 1692 | nsk = NULL; |
eeea10b8 ED |
1693 | if (!tcp_filter(sk, skb)) { |
1694 | th = (const struct tcphdr *)skb->data; | |
1695 | iph = ip_hdr(skb); | |
1696 | tcp_v4_fill_cb(skb, iph, th); | |
1f3b359f | 1697 | nsk = tcp_check_req(sk, skb, req, false); |
eeea10b8 | 1698 | } |
079096f1 ED |
1699 | if (!nsk) { |
1700 | reqsk_put(req); | |
7716682c | 1701 | goto discard_and_relse; |
079096f1 ED |
1702 | } |
1703 | if (nsk == sk) { | |
079096f1 | 1704 | reqsk_put(req); |
eeea10b8 | 1705 | tcp_v4_restore_cb(skb); |
079096f1 ED |
1706 | } else if (tcp_child_process(sk, nsk, skb)) { |
1707 | tcp_v4_send_reset(nsk, skb); | |
7716682c | 1708 | goto discard_and_relse; |
079096f1 | 1709 | } else { |
7716682c | 1710 | sock_put(sk); |
079096f1 ED |
1711 | return 0; |
1712 | } | |
1713 | } | |
6cce09f8 | 1714 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
02a1d6e7 | 1715 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); |
d218d111 | 1716 | goto discard_and_relse; |
6cce09f8 | 1717 | } |
d218d111 | 1718 | |
1da177e4 LT |
1719 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1720 | goto discard_and_relse; | |
9ea88a15 | 1721 | |
9ea88a15 DP |
1722 | if (tcp_v4_inbound_md5_hash(sk, skb)) |
1723 | goto discard_and_relse; | |
9ea88a15 | 1724 | |
b59c2701 | 1725 | nf_reset(skb); |
1da177e4 | 1726 | |
ac6e7800 | 1727 | if (tcp_filter(sk, skb)) |
1da177e4 | 1728 | goto discard_and_relse; |
ac6e7800 ED |
1729 | th = (const struct tcphdr *)skb->data; |
1730 | iph = ip_hdr(skb); | |
eeea10b8 | 1731 | tcp_v4_fill_cb(skb, iph, th); |
1da177e4 LT |
1732 | |
1733 | skb->dev = NULL; | |
1734 | ||
e994b2f0 ED |
1735 | if (sk->sk_state == TCP_LISTEN) { |
1736 | ret = tcp_v4_do_rcv(sk, skb); | |
1737 | goto put_and_return; | |
1738 | } | |
1739 | ||
1740 | sk_incoming_cpu_update(sk); | |
1741 | ||
c6366184 | 1742 | bh_lock_sock_nested(sk); |
a44d6eac | 1743 | tcp_segs_in(tcp_sk(sk), skb); |
1da177e4 LT |
1744 | ret = 0; |
1745 | if (!sock_owned_by_user(sk)) { | |
e7942d06 | 1746 | ret = tcp_v4_do_rcv(sk, skb); |
c9c33212 | 1747 | } else if (tcp_add_backlog(sk, skb)) { |
6b03a53a ZY |
1748 | goto discard_and_relse; |
1749 | } | |
1da177e4 LT |
1750 | bh_unlock_sock(sk); |
1751 | ||
e994b2f0 | 1752 | put_and_return: |
3b24d854 ED |
1753 | if (refcounted) |
1754 | sock_put(sk); | |
1da177e4 LT |
1755 | |
1756 | return ret; | |
1757 | ||
1758 | no_tcp_socket: | |
1759 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1760 | goto discard_it; | |
1761 | ||
eeea10b8 ED |
1762 | tcp_v4_fill_cb(skb, iph, th); |
1763 | ||
12e25e10 | 1764 | if (tcp_checksum_complete(skb)) { |
6a5dc9e5 | 1765 | csum_error: |
90bbcc60 | 1766 | __TCP_INC_STATS(net, TCP_MIB_CSUMERRORS); |
1da177e4 | 1767 | bad_packet: |
90bbcc60 | 1768 | __TCP_INC_STATS(net, TCP_MIB_INERRS); |
1da177e4 | 1769 | } else { |
cfb6eeb4 | 1770 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1771 | } |
1772 | ||
1773 | discard_it: | |
1774 | /* Discard frame. */ | |
1775 | kfree_skb(skb); | |
e905a9ed | 1776 | return 0; |
1da177e4 LT |
1777 | |
1778 | discard_and_relse: | |
532182cd | 1779 | sk_drops_add(sk, skb); |
3b24d854 ED |
1780 | if (refcounted) |
1781 | sock_put(sk); | |
1da177e4 LT |
1782 | goto discard_it; |
1783 | ||
1784 | do_time_wait: | |
1785 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1786 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1787 | goto discard_it; |
1788 | } | |
1789 | ||
eeea10b8 ED |
1790 | tcp_v4_fill_cb(skb, iph, th); |
1791 | ||
6a5dc9e5 ED |
1792 | if (tcp_checksum_complete(skb)) { |
1793 | inet_twsk_put(inet_twsk(sk)); | |
1794 | goto csum_error; | |
1da177e4 | 1795 | } |
9469c7b4 | 1796 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1797 | case TCP_TW_SYN: { |
c346dca1 | 1798 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
a583636a CG |
1799 | &tcp_hashinfo, skb, |
1800 | __tcp_hdrlen(th), | |
da5e3630 | 1801 | iph->saddr, th->source, |
eddc9ec5 | 1802 | iph->daddr, th->dest, |
3fa6f616 DA |
1803 | inet_iif(skb), |
1804 | sdif); | |
1da177e4 | 1805 | if (sk2) { |
dbe7faa4 | 1806 | inet_twsk_deschedule_put(inet_twsk(sk)); |
1da177e4 | 1807 | sk = sk2; |
eeea10b8 | 1808 | tcp_v4_restore_cb(skb); |
3b24d854 | 1809 | refcounted = false; |
1da177e4 LT |
1810 | goto process; |
1811 | } | |
1da177e4 | 1812 | } |
fcfd6dfa GS |
1813 | /* to ACK */ |
1814 | /* fall through */ | |
1da177e4 LT |
1815 | case TCP_TW_ACK: |
1816 | tcp_v4_timewait_ack(sk, skb); | |
1817 | break; | |
1818 | case TCP_TW_RST: | |
271c3b9b FW |
1819 | tcp_v4_send_reset(sk, skb); |
1820 | inet_twsk_deschedule_put(inet_twsk(sk)); | |
1821 | goto discard_it; | |
1da177e4 LT |
1822 | case TCP_TW_SUCCESS:; |
1823 | } | |
1824 | goto discard_it; | |
1825 | } | |
1826 | ||
ccb7c410 DM |
1827 | static struct timewait_sock_ops tcp_timewait_sock_ops = { |
1828 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1829 | .twsk_unique = tcp_twsk_unique, | |
1830 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 | 1831 | }; |
1da177e4 | 1832 | |
63d02d15 | 1833 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
5d299f3d ED |
1834 | { |
1835 | struct dst_entry *dst = skb_dst(skb); | |
1836 | ||
5037e9ef | 1837 | if (dst && dst_hold_safe(dst)) { |
ca777eff ED |
1838 | sk->sk_rx_dst = dst; |
1839 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
1840 | } | |
5d299f3d | 1841 | } |
63d02d15 | 1842 | EXPORT_SYMBOL(inet_sk_rx_dst_set); |
5d299f3d | 1843 | |
3b401a81 | 1844 | const struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1845 | .queue_xmit = ip_queue_xmit, |
1846 | .send_check = tcp_v4_send_check, | |
1847 | .rebuild_header = inet_sk_rebuild_header, | |
5d299f3d | 1848 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1849 | .conn_request = tcp_v4_conn_request, |
1850 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
543d9cfe ACM |
1851 | .net_header_len = sizeof(struct iphdr), |
1852 | .setsockopt = ip_setsockopt, | |
1853 | .getsockopt = ip_getsockopt, | |
1854 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1855 | .sockaddr_len = sizeof(struct sockaddr_in), | |
3fdadf7d | 1856 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1857 | .compat_setsockopt = compat_ip_setsockopt, |
1858 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1859 | #endif |
4fab9071 | 1860 | .mtu_reduced = tcp_v4_mtu_reduced, |
1da177e4 | 1861 | }; |
4bc2f18b | 1862 | EXPORT_SYMBOL(ipv4_specific); |
1da177e4 | 1863 | |
cfb6eeb4 | 1864 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1865 | static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1866 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1867 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1868 | .md5_parse = tcp_v4_parse_md5_keys, |
cfb6eeb4 | 1869 | }; |
b6332e6c | 1870 | #endif |
cfb6eeb4 | 1871 | |
1da177e4 LT |
1872 | /* NOTE: A lot of things set to zero explicitly by call to |
1873 | * sk_alloc() so need not be done here. | |
1874 | */ | |
1875 | static int tcp_v4_init_sock(struct sock *sk) | |
1876 | { | |
6687e988 | 1877 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1878 | |
900f65d3 | 1879 | tcp_init_sock(sk); |
1da177e4 | 1880 | |
8292a17a | 1881 | icsk->icsk_af_ops = &ipv4_specific; |
900f65d3 | 1882 | |
cfb6eeb4 | 1883 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1884 | tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific; |
cfb6eeb4 | 1885 | #endif |
1da177e4 | 1886 | |
1da177e4 LT |
1887 | return 0; |
1888 | } | |
1889 | ||
7d06b2e0 | 1890 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1891 | { |
1892 | struct tcp_sock *tp = tcp_sk(sk); | |
1893 | ||
e1a4aa50 SL |
1894 | trace_tcp_destroy_sock(sk); |
1895 | ||
1da177e4 LT |
1896 | tcp_clear_xmit_timers(sk); |
1897 | ||
6687e988 | 1898 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1899 | |
734942cc DW |
1900 | tcp_cleanup_ulp(sk); |
1901 | ||
1da177e4 | 1902 | /* Cleanup up the write buffer. */ |
fe067e8a | 1903 | tcp_write_queue_purge(sk); |
1da177e4 | 1904 | |
cf1ef3f0 WW |
1905 | /* Check if we want to disable active TFO */ |
1906 | tcp_fastopen_active_disable_ofo_check(sk); | |
1907 | ||
1da177e4 | 1908 | /* Cleans up our, hopefully empty, out_of_order_queue. */ |
9f5afeae | 1909 | skb_rbtree_purge(&tp->out_of_order_queue); |
1da177e4 | 1910 | |
cfb6eeb4 YH |
1911 | #ifdef CONFIG_TCP_MD5SIG |
1912 | /* Clean up the MD5 key list, if any */ | |
1913 | if (tp->md5sig_info) { | |
a915da9b | 1914 | tcp_clear_md5_list(sk); |
a8afca03 | 1915 | kfree_rcu(tp->md5sig_info, rcu); |
cfb6eeb4 YH |
1916 | tp->md5sig_info = NULL; |
1917 | } | |
1918 | #endif | |
1a2449a8 | 1919 | |
1da177e4 | 1920 | /* Clean up a referenced TCP bind bucket. */ |
463c84b9 | 1921 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1922 | inet_put_port(sk); |
1da177e4 | 1923 | |
00db4124 | 1924 | BUG_ON(tp->fastopen_rsk); |
435cf559 | 1925 | |
cf60af03 YC |
1926 | /* If socket is aborted during connect operation */ |
1927 | tcp_free_fastopen_req(tp); | |
1fba70e5 | 1928 | tcp_fastopen_destroy_cipher(sk); |
cd8ae852 | 1929 | tcp_saved_syn_free(tp); |
cf60af03 | 1930 | |
180d8cd9 | 1931 | sk_sockets_allocated_dec(sk); |
1da177e4 | 1932 | } |
1da177e4 LT |
1933 | EXPORT_SYMBOL(tcp_v4_destroy_sock); |
1934 | ||
1935 | #ifdef CONFIG_PROC_FS | |
1936 | /* Proc filesystem TCP sock list dumping. */ | |
1937 | ||
a8b690f9 TH |
1938 | /* |
1939 | * Get next listener socket follow cur. If cur is NULL, get first socket | |
1940 | * starting from bucket given in st->bucket; when st->bucket is zero the | |
1941 | * very first socket in the hash table is returned. | |
1942 | */ | |
1da177e4 LT |
1943 | static void *listening_get_next(struct seq_file *seq, void *cur) |
1944 | { | |
5799de0b | 1945 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1946 | struct net *net = seq_file_net(seq); |
3b24d854 | 1947 | struct inet_listen_hashbucket *ilb; |
6cb8e266 | 1948 | struct hlist_nulls_node *node; |
3b24d854 | 1949 | struct sock *sk = cur; |
1da177e4 LT |
1950 | |
1951 | if (!sk) { | |
3b24d854 | 1952 | get_head: |
a8b690f9 | 1953 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
9652dc2e | 1954 | spin_lock(&ilb->lock); |
6cb8e266 | 1955 | sk = sk_nulls_head(&ilb->nulls_head); |
a8b690f9 | 1956 | st->offset = 0; |
1da177e4 LT |
1957 | goto get_sk; |
1958 | } | |
5caea4ea | 1959 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 | 1960 | ++st->num; |
a8b690f9 | 1961 | ++st->offset; |
1da177e4 | 1962 | |
6cb8e266 | 1963 | sk = sk_nulls_next(sk); |
1da177e4 | 1964 | get_sk: |
6cb8e266 | 1965 | sk_nulls_for_each_from(sk, node) { |
8475ef9f PE |
1966 | if (!net_eq(sock_net(sk), net)) |
1967 | continue; | |
3b24d854 ED |
1968 | if (sk->sk_family == st->family) |
1969 | return sk; | |
1da177e4 | 1970 | } |
9652dc2e | 1971 | spin_unlock(&ilb->lock); |
a8b690f9 | 1972 | st->offset = 0; |
3b24d854 ED |
1973 | if (++st->bucket < INET_LHTABLE_SIZE) |
1974 | goto get_head; | |
1975 | return NULL; | |
1da177e4 LT |
1976 | } |
1977 | ||
1978 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
1979 | { | |
a8b690f9 TH |
1980 | struct tcp_iter_state *st = seq->private; |
1981 | void *rc; | |
1982 | ||
1983 | st->bucket = 0; | |
1984 | st->offset = 0; | |
1985 | rc = listening_get_next(seq, NULL); | |
1da177e4 LT |
1986 | |
1987 | while (rc && *pos) { | |
1988 | rc = listening_get_next(seq, rc); | |
1989 | --*pos; | |
1990 | } | |
1991 | return rc; | |
1992 | } | |
1993 | ||
05dbc7b5 | 1994 | static inline bool empty_bucket(const struct tcp_iter_state *st) |
6eac5604 | 1995 | { |
05dbc7b5 | 1996 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain); |
6eac5604 AK |
1997 | } |
1998 | ||
a8b690f9 TH |
1999 | /* |
2000 | * Get first established socket starting from bucket given in st->bucket. | |
2001 | * If st->bucket is zero, the very first socket in the hash is returned. | |
2002 | */ | |
1da177e4 LT |
2003 | static void *established_get_first(struct seq_file *seq) |
2004 | { | |
5799de0b | 2005 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2006 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2007 | void *rc = NULL; |
2008 | ||
a8b690f9 TH |
2009 | st->offset = 0; |
2010 | for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) { | |
1da177e4 | 2011 | struct sock *sk; |
3ab5aee7 | 2012 | struct hlist_nulls_node *node; |
9db66bdc | 2013 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 2014 | |
6eac5604 AK |
2015 | /* Lockless fast path for the common case of empty buckets */ |
2016 | if (empty_bucket(st)) | |
2017 | continue; | |
2018 | ||
9db66bdc | 2019 | spin_lock_bh(lock); |
3ab5aee7 | 2020 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 2021 | if (sk->sk_family != st->family || |
878628fb | 2022 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
2023 | continue; |
2024 | } | |
2025 | rc = sk; | |
2026 | goto out; | |
2027 | } | |
9db66bdc | 2028 | spin_unlock_bh(lock); |
1da177e4 LT |
2029 | } |
2030 | out: | |
2031 | return rc; | |
2032 | } | |
2033 | ||
2034 | static void *established_get_next(struct seq_file *seq, void *cur) | |
2035 | { | |
2036 | struct sock *sk = cur; | |
3ab5aee7 | 2037 | struct hlist_nulls_node *node; |
5799de0b | 2038 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2039 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2040 | |
2041 | ++st->num; | |
a8b690f9 | 2042 | ++st->offset; |
1da177e4 | 2043 | |
05dbc7b5 | 2044 | sk = sk_nulls_next(sk); |
1da177e4 | 2045 | |
3ab5aee7 | 2046 | sk_nulls_for_each_from(sk, node) { |
878628fb | 2047 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
05dbc7b5 | 2048 | return sk; |
1da177e4 LT |
2049 | } |
2050 | ||
05dbc7b5 ED |
2051 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
2052 | ++st->bucket; | |
2053 | return established_get_first(seq); | |
1da177e4 LT |
2054 | } |
2055 | ||
2056 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2057 | { | |
a8b690f9 TH |
2058 | struct tcp_iter_state *st = seq->private; |
2059 | void *rc; | |
2060 | ||
2061 | st->bucket = 0; | |
2062 | rc = established_get_first(seq); | |
1da177e4 LT |
2063 | |
2064 | while (rc && pos) { | |
2065 | rc = established_get_next(seq, rc); | |
2066 | --pos; | |
7174259e | 2067 | } |
1da177e4 LT |
2068 | return rc; |
2069 | } | |
2070 | ||
2071 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2072 | { | |
2073 | void *rc; | |
5799de0b | 2074 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2075 | |
1da177e4 LT |
2076 | st->state = TCP_SEQ_STATE_LISTENING; |
2077 | rc = listening_get_idx(seq, &pos); | |
2078 | ||
2079 | if (!rc) { | |
1da177e4 LT |
2080 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2081 | rc = established_get_idx(seq, pos); | |
2082 | } | |
2083 | ||
2084 | return rc; | |
2085 | } | |
2086 | ||
a8b690f9 TH |
2087 | static void *tcp_seek_last_pos(struct seq_file *seq) |
2088 | { | |
2089 | struct tcp_iter_state *st = seq->private; | |
2090 | int offset = st->offset; | |
2091 | int orig_num = st->num; | |
2092 | void *rc = NULL; | |
2093 | ||
2094 | switch (st->state) { | |
a8b690f9 TH |
2095 | case TCP_SEQ_STATE_LISTENING: |
2096 | if (st->bucket >= INET_LHTABLE_SIZE) | |
2097 | break; | |
2098 | st->state = TCP_SEQ_STATE_LISTENING; | |
2099 | rc = listening_get_next(seq, NULL); | |
2100 | while (offset-- && rc) | |
2101 | rc = listening_get_next(seq, rc); | |
2102 | if (rc) | |
2103 | break; | |
2104 | st->bucket = 0; | |
05dbc7b5 | 2105 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2106 | /* Fallthrough */ |
2107 | case TCP_SEQ_STATE_ESTABLISHED: | |
a8b690f9 TH |
2108 | if (st->bucket > tcp_hashinfo.ehash_mask) |
2109 | break; | |
2110 | rc = established_get_first(seq); | |
2111 | while (offset-- && rc) | |
2112 | rc = established_get_next(seq, rc); | |
2113 | } | |
2114 | ||
2115 | st->num = orig_num; | |
2116 | ||
2117 | return rc; | |
2118 | } | |
2119 | ||
1da177e4 LT |
2120 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) |
2121 | { | |
5799de0b | 2122 | struct tcp_iter_state *st = seq->private; |
a8b690f9 TH |
2123 | void *rc; |
2124 | ||
2125 | if (*pos && *pos == st->last_pos) { | |
2126 | rc = tcp_seek_last_pos(seq); | |
2127 | if (rc) | |
2128 | goto out; | |
2129 | } | |
2130 | ||
1da177e4 LT |
2131 | st->state = TCP_SEQ_STATE_LISTENING; |
2132 | st->num = 0; | |
a8b690f9 TH |
2133 | st->bucket = 0; |
2134 | st->offset = 0; | |
2135 | rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2136 | ||
2137 | out: | |
2138 | st->last_pos = *pos; | |
2139 | return rc; | |
1da177e4 LT |
2140 | } |
2141 | ||
2142 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2143 | { | |
a8b690f9 | 2144 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2145 | void *rc = NULL; |
1da177e4 LT |
2146 | |
2147 | if (v == SEQ_START_TOKEN) { | |
2148 | rc = tcp_get_idx(seq, 0); | |
2149 | goto out; | |
2150 | } | |
1da177e4 LT |
2151 | |
2152 | switch (st->state) { | |
1da177e4 LT |
2153 | case TCP_SEQ_STATE_LISTENING: |
2154 | rc = listening_get_next(seq, v); | |
2155 | if (!rc) { | |
1da177e4 | 2156 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2157 | st->bucket = 0; |
2158 | st->offset = 0; | |
1da177e4 LT |
2159 | rc = established_get_first(seq); |
2160 | } | |
2161 | break; | |
2162 | case TCP_SEQ_STATE_ESTABLISHED: | |
1da177e4 LT |
2163 | rc = established_get_next(seq, v); |
2164 | break; | |
2165 | } | |
2166 | out: | |
2167 | ++*pos; | |
a8b690f9 | 2168 | st->last_pos = *pos; |
1da177e4 LT |
2169 | return rc; |
2170 | } | |
2171 | ||
2172 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2173 | { | |
5799de0b | 2174 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2175 | |
2176 | switch (st->state) { | |
1da177e4 LT |
2177 | case TCP_SEQ_STATE_LISTENING: |
2178 | if (v != SEQ_START_TOKEN) | |
9652dc2e | 2179 | spin_unlock(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 | 2180 | break; |
1da177e4 LT |
2181 | case TCP_SEQ_STATE_ESTABLISHED: |
2182 | if (v) | |
9db66bdc | 2183 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2184 | break; |
2185 | } | |
2186 | } | |
2187 | ||
73cb88ec | 2188 | int tcp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2189 | { |
d9dda78b | 2190 | struct tcp_seq_afinfo *afinfo = PDE_DATA(inode); |
1da177e4 | 2191 | struct tcp_iter_state *s; |
52d6f3f1 | 2192 | int err; |
1da177e4 | 2193 | |
52d6f3f1 DL |
2194 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2195 | sizeof(struct tcp_iter_state)); | |
2196 | if (err < 0) | |
2197 | return err; | |
f40c8174 | 2198 | |
52d6f3f1 | 2199 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2200 | s->family = afinfo->family; |
688d1945 | 2201 | s->last_pos = 0; |
f40c8174 DL |
2202 | return 0; |
2203 | } | |
73cb88ec | 2204 | EXPORT_SYMBOL(tcp_seq_open); |
f40c8174 | 2205 | |
6f8b13bc | 2206 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2207 | { |
2208 | int rc = 0; | |
2209 | struct proc_dir_entry *p; | |
2210 | ||
9427c4b3 DL |
2211 | afinfo->seq_ops.start = tcp_seq_start; |
2212 | afinfo->seq_ops.next = tcp_seq_next; | |
2213 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2214 | ||
84841c3c | 2215 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2216 | afinfo->seq_fops, afinfo); |
84841c3c | 2217 | if (!p) |
1da177e4 LT |
2218 | rc = -ENOMEM; |
2219 | return rc; | |
2220 | } | |
4bc2f18b | 2221 | EXPORT_SYMBOL(tcp_proc_register); |
1da177e4 | 2222 | |
6f8b13bc | 2223 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2224 | { |
ece31ffd | 2225 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2226 | } |
4bc2f18b | 2227 | EXPORT_SYMBOL(tcp_proc_unregister); |
1da177e4 | 2228 | |
d4f06873 | 2229 | static void get_openreq4(const struct request_sock *req, |
aa3a0c8c | 2230 | struct seq_file *f, int i) |
1da177e4 | 2231 | { |
2e6599cb | 2232 | const struct inet_request_sock *ireq = inet_rsk(req); |
fa76ce73 | 2233 | long delta = req->rsk_timer.expires - jiffies; |
1da177e4 | 2234 | |
5e659e4c | 2235 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2236 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK", |
1da177e4 | 2237 | i, |
634fb979 | 2238 | ireq->ir_loc_addr, |
d4f06873 | 2239 | ireq->ir_num, |
634fb979 ED |
2240 | ireq->ir_rmt_addr, |
2241 | ntohs(ireq->ir_rmt_port), | |
1da177e4 LT |
2242 | TCP_SYN_RECV, |
2243 | 0, 0, /* could print option size, but that is af dependent. */ | |
2244 | 1, /* timers active (only the expire timer) */ | |
a399a805 | 2245 | jiffies_delta_to_clock_t(delta), |
e6c022a4 | 2246 | req->num_timeout, |
aa3a0c8c ED |
2247 | from_kuid_munged(seq_user_ns(f), |
2248 | sock_i_uid(req->rsk_listener)), | |
1da177e4 LT |
2249 | 0, /* non standard timer */ |
2250 | 0, /* open_requests have no inode */ | |
d4f06873 | 2251 | 0, |
652586df | 2252 | req); |
1da177e4 LT |
2253 | } |
2254 | ||
652586df | 2255 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i) |
1da177e4 LT |
2256 | { |
2257 | int timer_active; | |
2258 | unsigned long timer_expires; | |
cf533ea5 | 2259 | const struct tcp_sock *tp = tcp_sk(sk); |
cf4c6bf8 | 2260 | const struct inet_connection_sock *icsk = inet_csk(sk); |
cf533ea5 | 2261 | const struct inet_sock *inet = inet_sk(sk); |
0536fcc0 | 2262 | const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq; |
c720c7e8 ED |
2263 | __be32 dest = inet->inet_daddr; |
2264 | __be32 src = inet->inet_rcv_saddr; | |
2265 | __u16 destp = ntohs(inet->inet_dport); | |
2266 | __u16 srcp = ntohs(inet->inet_sport); | |
49d09007 | 2267 | int rx_queue; |
00fd38d9 | 2268 | int state; |
1da177e4 | 2269 | |
6ba8a3b1 | 2270 | if (icsk->icsk_pending == ICSK_TIME_RETRANS || |
57dde7f7 | 2271 | icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || |
6ba8a3b1 | 2272 | icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { |
1da177e4 | 2273 | timer_active = 1; |
463c84b9 ACM |
2274 | timer_expires = icsk->icsk_timeout; |
2275 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2276 | timer_active = 4; |
463c84b9 | 2277 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2278 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2279 | timer_active = 2; |
cf4c6bf8 | 2280 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2281 | } else { |
2282 | timer_active = 0; | |
2283 | timer_expires = jiffies; | |
2284 | } | |
2285 | ||
00fd38d9 ED |
2286 | state = sk_state_load(sk); |
2287 | if (state == TCP_LISTEN) | |
49d09007 ED |
2288 | rx_queue = sk->sk_ack_backlog; |
2289 | else | |
00fd38d9 ED |
2290 | /* Because we don't lock the socket, |
2291 | * we might find a transient negative value. | |
49d09007 ED |
2292 | */ |
2293 | rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0); | |
2294 | ||
5e659e4c | 2295 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
652586df | 2296 | "%08X %5u %8d %lu %d %pK %lu %lu %u %u %d", |
00fd38d9 | 2297 | i, src, srcp, dest, destp, state, |
47da8ee6 | 2298 | tp->write_seq - tp->snd_una, |
49d09007 | 2299 | rx_queue, |
1da177e4 | 2300 | timer_active, |
a399a805 | 2301 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 2302 | icsk->icsk_retransmits, |
a7cb5a49 | 2303 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)), |
6687e988 | 2304 | icsk->icsk_probes_out, |
cf4c6bf8 | 2305 | sock_i_ino(sk), |
41c6d650 | 2306 | refcount_read(&sk->sk_refcnt), sk, |
7be87351 SH |
2307 | jiffies_to_clock_t(icsk->icsk_rto), |
2308 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2309 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2310 | tp->snd_cwnd, |
00fd38d9 ED |
2311 | state == TCP_LISTEN ? |
2312 | fastopenq->max_qlen : | |
652586df | 2313 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)); |
1da177e4 LT |
2314 | } |
2315 | ||
cf533ea5 | 2316 | static void get_timewait4_sock(const struct inet_timewait_sock *tw, |
652586df | 2317 | struct seq_file *f, int i) |
1da177e4 | 2318 | { |
789f558c | 2319 | long delta = tw->tw_timer.expires - jiffies; |
23f33c2d | 2320 | __be32 dest, src; |
1da177e4 | 2321 | __u16 destp, srcp; |
1da177e4 LT |
2322 | |
2323 | dest = tw->tw_daddr; | |
2324 | src = tw->tw_rcv_saddr; | |
2325 | destp = ntohs(tw->tw_dport); | |
2326 | srcp = ntohs(tw->tw_sport); | |
2327 | ||
5e659e4c | 2328 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2329 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK", |
1da177e4 | 2330 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
a399a805 | 2331 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
41c6d650 | 2332 | refcount_read(&tw->tw_refcnt), tw); |
1da177e4 LT |
2333 | } |
2334 | ||
2335 | #define TMPSZ 150 | |
2336 | ||
2337 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2338 | { | |
5799de0b | 2339 | struct tcp_iter_state *st; |
05dbc7b5 | 2340 | struct sock *sk = v; |
1da177e4 | 2341 | |
652586df | 2342 | seq_setwidth(seq, TMPSZ - 1); |
1da177e4 | 2343 | if (v == SEQ_START_TOKEN) { |
652586df | 2344 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
1da177e4 LT |
2345 | "rx_queue tr tm->when retrnsmt uid timeout " |
2346 | "inode"); | |
2347 | goto out; | |
2348 | } | |
2349 | st = seq->private; | |
2350 | ||
079096f1 ED |
2351 | if (sk->sk_state == TCP_TIME_WAIT) |
2352 | get_timewait4_sock(v, seq, st->num); | |
2353 | else if (sk->sk_state == TCP_NEW_SYN_RECV) | |
aa3a0c8c | 2354 | get_openreq4(v, seq, st->num); |
079096f1 ED |
2355 | else |
2356 | get_tcp4_sock(v, seq, st->num); | |
1da177e4 | 2357 | out: |
652586df | 2358 | seq_pad(seq, '\n'); |
1da177e4 LT |
2359 | return 0; |
2360 | } | |
2361 | ||
73cb88ec AV |
2362 | static const struct file_operations tcp_afinfo_seq_fops = { |
2363 | .owner = THIS_MODULE, | |
2364 | .open = tcp_seq_open, | |
2365 | .read = seq_read, | |
2366 | .llseek = seq_lseek, | |
2367 | .release = seq_release_net | |
2368 | }; | |
2369 | ||
1da177e4 | 2370 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2371 | .name = "tcp", |
2372 | .family = AF_INET, | |
73cb88ec | 2373 | .seq_fops = &tcp_afinfo_seq_fops, |
9427c4b3 DL |
2374 | .seq_ops = { |
2375 | .show = tcp4_seq_show, | |
2376 | }, | |
1da177e4 LT |
2377 | }; |
2378 | ||
2c8c1e72 | 2379 | static int __net_init tcp4_proc_init_net(struct net *net) |
757764f6 PE |
2380 | { |
2381 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2382 | } | |
2383 | ||
2c8c1e72 | 2384 | static void __net_exit tcp4_proc_exit_net(struct net *net) |
757764f6 PE |
2385 | { |
2386 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2387 | } | |
2388 | ||
2389 | static struct pernet_operations tcp4_net_ops = { | |
2390 | .init = tcp4_proc_init_net, | |
2391 | .exit = tcp4_proc_exit_net, | |
2392 | }; | |
2393 | ||
1da177e4 LT |
2394 | int __init tcp4_proc_init(void) |
2395 | { | |
757764f6 | 2396 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2397 | } |
2398 | ||
2399 | void tcp4_proc_exit(void) | |
2400 | { | |
757764f6 | 2401 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2402 | } |
2403 | #endif /* CONFIG_PROC_FS */ | |
2404 | ||
2405 | struct proto tcp_prot = { | |
2406 | .name = "TCP", | |
2407 | .owner = THIS_MODULE, | |
2408 | .close = tcp_close, | |
2409 | .connect = tcp_v4_connect, | |
2410 | .disconnect = tcp_disconnect, | |
463c84b9 | 2411 | .accept = inet_csk_accept, |
1da177e4 LT |
2412 | .ioctl = tcp_ioctl, |
2413 | .init = tcp_v4_init_sock, | |
2414 | .destroy = tcp_v4_destroy_sock, | |
2415 | .shutdown = tcp_shutdown, | |
2416 | .setsockopt = tcp_setsockopt, | |
2417 | .getsockopt = tcp_getsockopt, | |
4b9d07a4 | 2418 | .keepalive = tcp_set_keepalive, |
1da177e4 | 2419 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2420 | .sendmsg = tcp_sendmsg, |
2421 | .sendpage = tcp_sendpage, | |
1da177e4 | 2422 | .backlog_rcv = tcp_v4_do_rcv, |
46d3ceab | 2423 | .release_cb = tcp_release_cb, |
ab1e0a13 ACM |
2424 | .hash = inet_hash, |
2425 | .unhash = inet_unhash, | |
2426 | .get_port = inet_csk_get_port, | |
1da177e4 | 2427 | .enter_memory_pressure = tcp_enter_memory_pressure, |
06044751 | 2428 | .leave_memory_pressure = tcp_leave_memory_pressure, |
c9bee3b7 | 2429 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 | 2430 | .sockets_allocated = &tcp_sockets_allocated, |
0a5578cf | 2431 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2432 | .memory_allocated = &tcp_memory_allocated, |
2433 | .memory_pressure = &tcp_memory_pressure, | |
a4fe34bf | 2434 | .sysctl_mem = sysctl_tcp_mem, |
356d1833 ED |
2435 | .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_tcp_wmem), |
2436 | .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_tcp_rmem), | |
1da177e4 LT |
2437 | .max_header = MAX_TCP_HEADER, |
2438 | .obj_size = sizeof(struct tcp_sock), | |
5f0d5a3a | 2439 | .slab_flags = SLAB_TYPESAFE_BY_RCU, |
6d6ee43e | 2440 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2441 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2442 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2443 | .no_autobind = true, |
543d9cfe ACM |
2444 | #ifdef CONFIG_COMPAT |
2445 | .compat_setsockopt = compat_tcp_setsockopt, | |
2446 | .compat_getsockopt = compat_tcp_getsockopt, | |
d1a4c0b3 | 2447 | #endif |
c1e64e29 | 2448 | .diag_destroy = tcp_abort, |
1da177e4 | 2449 | }; |
4bc2f18b | 2450 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 | 2451 | |
bdbbb852 ED |
2452 | static void __net_exit tcp_sk_exit(struct net *net) |
2453 | { | |
2454 | int cpu; | |
2455 | ||
e6972770 DL |
2456 | if (net->ipv4.tcp_congestion_control) |
2457 | module_put(net->ipv4.tcp_congestion_control->owner); | |
6670e152 | 2458 | |
bdbbb852 ED |
2459 | for_each_possible_cpu(cpu) |
2460 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.tcp_sk, cpu)); | |
2461 | free_percpu(net->ipv4.tcp_sk); | |
2462 | } | |
2463 | ||
046ee902 DL |
2464 | static int __net_init tcp_sk_init(struct net *net) |
2465 | { | |
fee83d09 | 2466 | int res, cpu, cnt; |
bdbbb852 ED |
2467 | |
2468 | net->ipv4.tcp_sk = alloc_percpu(struct sock *); | |
2469 | if (!net->ipv4.tcp_sk) | |
2470 | return -ENOMEM; | |
2471 | ||
2472 | for_each_possible_cpu(cpu) { | |
2473 | struct sock *sk; | |
2474 | ||
2475 | res = inet_ctl_sock_create(&sk, PF_INET, SOCK_RAW, | |
2476 | IPPROTO_TCP, net); | |
2477 | if (res) | |
2478 | goto fail; | |
a9d6532b | 2479 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); |
587d8a17 ED |
2480 | |
2481 | /* Please enforce IP_DF and IPID==0 for RST and | |
2482 | * ACK sent in SYN-RECV and TIME-WAIT state. | |
2483 | */ | |
2484 | inet_sk(sk)->pmtudisc = IP_PMTUDISC_DO; | |
2485 | ||
bdbbb852 ED |
2486 | *per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk; |
2487 | } | |
49213555 | 2488 | |
5d134f1c | 2489 | net->ipv4.sysctl_tcp_ecn = 2; |
49213555 DB |
2490 | net->ipv4.sysctl_tcp_ecn_fallback = 1; |
2491 | ||
b0f9ca53 | 2492 | net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS; |
be208663 | 2493 | net->ipv4.sysctl_tcp_min_snd_mss = TCP_MIN_SND_MSS; |
6b58e0a5 | 2494 | net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD; |
05cbc0db | 2495 | net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL; |
046ee902 | 2496 | |
13b287e8 | 2497 | net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME; |
9bd6861b | 2498 | net->ipv4.sysctl_tcp_keepalive_probes = TCP_KEEPALIVE_PROBES; |
b840d15d | 2499 | net->ipv4.sysctl_tcp_keepalive_intvl = TCP_KEEPALIVE_INTVL; |
13b287e8 | 2500 | |
6fa25166 | 2501 | net->ipv4.sysctl_tcp_syn_retries = TCP_SYN_RETRIES; |
7c083ecb | 2502 | net->ipv4.sysctl_tcp_synack_retries = TCP_SYNACK_RETRIES; |
0aca737d | 2503 | net->ipv4.sysctl_tcp_syncookies = 1; |
1043e25f | 2504 | net->ipv4.sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH; |
ae5c3f40 | 2505 | net->ipv4.sysctl_tcp_retries1 = TCP_RETR1; |
c6214a97 | 2506 | net->ipv4.sysctl_tcp_retries2 = TCP_RETR2; |
c402d9be | 2507 | net->ipv4.sysctl_tcp_orphan_retries = 0; |
1e579caa | 2508 | net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT; |
4979f2d9 | 2509 | net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX; |
56ab6b93 | 2510 | net->ipv4.sysctl_tcp_tw_reuse = 0; |
12ed8244 | 2511 | |
fee83d09 | 2512 | cnt = tcp_hashinfo.ehash_mask + 1; |
fee83d09 | 2513 | net->ipv4.tcp_death_row.sysctl_max_tw_buckets = (cnt + 1) / 2; |
1946e672 HY |
2514 | net->ipv4.tcp_death_row.hashinfo = &tcp_hashinfo; |
2515 | ||
fee83d09 | 2516 | net->ipv4.sysctl_max_syn_backlog = max(128, cnt / 256); |
f9301034 | 2517 | net->ipv4.sysctl_tcp_sack = 1; |
9bb37ef0 | 2518 | net->ipv4.sysctl_tcp_window_scaling = 1; |
5d2ed052 | 2519 | net->ipv4.sysctl_tcp_timestamps = 1; |
2ae21cf5 | 2520 | net->ipv4.sysctl_tcp_early_retrans = 3; |
e20223f1 | 2521 | net->ipv4.sysctl_tcp_recovery = TCP_RACK_LOSS_DETECTION; |
b510f0d2 | 2522 | net->ipv4.sysctl_tcp_slow_start_after_idle = 1; /* By default, RFC2861 behavior. */ |
e0a1e5b5 | 2523 | net->ipv4.sysctl_tcp_retrans_collapse = 1; |
c6e21803 | 2524 | net->ipv4.sysctl_tcp_max_reordering = 300; |
6496f6bd | 2525 | net->ipv4.sysctl_tcp_dsack = 1; |
0c12654a | 2526 | net->ipv4.sysctl_tcp_app_win = 31; |
94f0893e | 2527 | net->ipv4.sysctl_tcp_adv_win_scale = 1; |
af9b69a7 | 2528 | net->ipv4.sysctl_tcp_frto = 2; |
4540c0cf | 2529 | net->ipv4.sysctl_tcp_moderate_rcvbuf = 1; |
d06a9904 ED |
2530 | /* This limits the percentage of the congestion window which we |
2531 | * will allow a single TSO frame to consume. Building TSO frames | |
2532 | * which are too large can cause TCP streams to be bursty. | |
2533 | */ | |
2534 | net->ipv4.sysctl_tcp_tso_win_divisor = 3; | |
9184d8bb ED |
2535 | /* Default TSQ limit of four TSO segments */ |
2536 | net->ipv4.sysctl_tcp_limit_output_bytes = 262144; | |
b530b681 ED |
2537 | /* rfc5961 challenge ack rate limiting */ |
2538 | net->ipv4.sysctl_tcp_challenge_ack_limit = 1000; | |
26e9596e | 2539 | net->ipv4.sysctl_tcp_min_tso_segs = 2; |
bd239704 | 2540 | net->ipv4.sysctl_tcp_min_rtt_wlen = 300; |
790f00e1 | 2541 | net->ipv4.sysctl_tcp_autocorking = 1; |
4170ba6b | 2542 | net->ipv4.sysctl_tcp_invalid_ratelimit = HZ/2; |
23a7102a | 2543 | net->ipv4.sysctl_tcp_pacing_ss_ratio = 200; |
c26e91f8 | 2544 | net->ipv4.sysctl_tcp_pacing_ca_ratio = 120; |
356d1833 ED |
2545 | if (net != &init_net) { |
2546 | memcpy(net->ipv4.sysctl_tcp_rmem, | |
2547 | init_net.ipv4.sysctl_tcp_rmem, | |
2548 | sizeof(init_net.ipv4.sysctl_tcp_rmem)); | |
2549 | memcpy(net->ipv4.sysctl_tcp_wmem, | |
2550 | init_net.ipv4.sysctl_tcp_wmem, | |
2551 | sizeof(init_net.ipv4.sysctl_tcp_wmem)); | |
2552 | } | |
e1cfcbe8 | 2553 | net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE; |
43713848 | 2554 | spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock); |
3733be14 HY |
2555 | net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 60 * 60; |
2556 | atomic_set(&net->ipv4.tfo_active_disable_times, 0); | |
e1cfcbe8 | 2557 | |
6670e152 SH |
2558 | /* Reno is always built in */ |
2559 | if (!net_eq(net, &init_net) && | |
2560 | try_module_get(init_net.ipv4.tcp_congestion_control->owner)) | |
2561 | net->ipv4.tcp_congestion_control = init_net.ipv4.tcp_congestion_control; | |
2562 | else | |
2563 | net->ipv4.tcp_congestion_control = &tcp_reno; | |
2564 | ||
49213555 | 2565 | return 0; |
bdbbb852 ED |
2566 | fail: |
2567 | tcp_sk_exit(net); | |
2568 | ||
2569 | return res; | |
b099ce26 EB |
2570 | } |
2571 | ||
2572 | static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list) | |
2573 | { | |
43713848 HY |
2574 | struct net *net; |
2575 | ||
1946e672 | 2576 | inet_twsk_purge(&tcp_hashinfo, AF_INET); |
43713848 HY |
2577 | |
2578 | list_for_each_entry(net, net_exit_list, exit_list) | |
2579 | tcp_fastopen_ctx_destroy(net); | |
046ee902 DL |
2580 | } |
2581 | ||
2582 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
b099ce26 EB |
2583 | .init = tcp_sk_init, |
2584 | .exit = tcp_sk_exit, | |
2585 | .exit_batch = tcp_sk_exit_batch, | |
046ee902 DL |
2586 | }; |
2587 | ||
9b0f976f | 2588 | void __init tcp_v4_init(void) |
1da177e4 | 2589 | { |
6a1b3054 | 2590 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2591 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 | 2592 | } |