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 | |
c720c7e8 | 250 | inet->inet_id = tp->write_seq ^ jiffies; |
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 | ||
f1ecd5d9 | 480 | icsk->icsk_backoff--; |
fcdd1cf4 ED |
481 | icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) : |
482 | TCP_TIMEOUT_INIT; | |
483 | icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX); | |
f1ecd5d9 | 484 | |
75c119af | 485 | skb = tcp_rtx_queue_head(sk); |
f1ecd5d9 DL |
486 | BUG_ON(!skb); |
487 | ||
9a568de4 ED |
488 | tcp_mstamp_refresh(tp); |
489 | delta_us = (u32)(tp->tcp_mstamp - skb->skb_mstamp); | |
7faee5c0 | 490 | remaining = icsk->icsk_rto - |
9a568de4 | 491 | usecs_to_jiffies(delta_us); |
f1ecd5d9 | 492 | |
9a568de4 | 493 | if (remaining > 0) { |
f1ecd5d9 DL |
494 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, |
495 | remaining, TCP_RTO_MAX); | |
f1ecd5d9 DL |
496 | } else { |
497 | /* RTO revert clocked out retransmission. | |
498 | * Will retransmit now */ | |
499 | tcp_retransmit_timer(sk); | |
500 | } | |
501 | ||
1da177e4 LT |
502 | break; |
503 | case ICMP_TIME_EXCEEDED: | |
504 | err = EHOSTUNREACH; | |
505 | break; | |
506 | default: | |
507 | goto out; | |
508 | } | |
509 | ||
510 | switch (sk->sk_state) { | |
1da177e4 | 511 | case TCP_SYN_SENT: |
0a672f74 YC |
512 | case TCP_SYN_RECV: |
513 | /* Only in fast or simultaneous open. If a fast open socket is | |
514 | * is already accepted it is treated as a connected one below. | |
515 | */ | |
51456b29 | 516 | if (fastopen && !fastopen->sk) |
0a672f74 YC |
517 | break; |
518 | ||
1da177e4 | 519 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
520 | sk->sk_err = err; |
521 | ||
522 | sk->sk_error_report(sk); | |
523 | ||
524 | tcp_done(sk); | |
525 | } else { | |
526 | sk->sk_err_soft = err; | |
527 | } | |
528 | goto out; | |
529 | } | |
530 | ||
531 | /* If we've already connected we will keep trying | |
532 | * until we time out, or the user gives up. | |
533 | * | |
534 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
535 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
536 | * but it is obsoleted by pmtu discovery). | |
537 | * | |
538 | * Note, that in modern internet, where routing is unreliable | |
539 | * and in each dark corner broken firewalls sit, sending random | |
540 | * errors ordered by their masters even this two messages finally lose | |
541 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
542 | * | |
543 | * Now we are in compliance with RFCs. | |
544 | * --ANK (980905) | |
545 | */ | |
546 | ||
547 | inet = inet_sk(sk); | |
548 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
549 | sk->sk_err = err; | |
550 | sk->sk_error_report(sk); | |
551 | } else { /* Only an error on timeout */ | |
552 | sk->sk_err_soft = err; | |
553 | } | |
554 | ||
555 | out: | |
556 | bh_unlock_sock(sk); | |
557 | sock_put(sk); | |
558 | } | |
559 | ||
28850dc7 | 560 | void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 561 | { |
aa8223c7 | 562 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 563 | |
84fa7933 | 564 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
419f9f89 | 565 | th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0); |
663ead3b | 566 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 567 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 568 | } else { |
419f9f89 | 569 | th->check = tcp_v4_check(skb->len, saddr, daddr, |
07f0757a | 570 | csum_partial(th, |
1da177e4 LT |
571 | th->doff << 2, |
572 | skb->csum)); | |
573 | } | |
574 | } | |
575 | ||
419f9f89 | 576 | /* This routine computes an IPv4 TCP checksum. */ |
bb296246 | 577 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb) |
419f9f89 | 578 | { |
cf533ea5 | 579 | const struct inet_sock *inet = inet_sk(sk); |
419f9f89 HX |
580 | |
581 | __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr); | |
582 | } | |
4bc2f18b | 583 | EXPORT_SYMBOL(tcp_v4_send_check); |
419f9f89 | 584 | |
1da177e4 LT |
585 | /* |
586 | * This routine will send an RST to the other tcp. | |
587 | * | |
588 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
589 | * for reset. | |
590 | * Answer: if a packet caused RST, it is not for a socket | |
591 | * existing in our system, if it is matched to a socket, | |
592 | * it is just duplicate segment or bug in other side's TCP. | |
593 | * So that we build reply only basing on parameters | |
594 | * arrived with segment. | |
595 | * Exception: precedence violation. We do not implement it in any case. | |
596 | */ | |
597 | ||
a00e7444 | 598 | static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) |
1da177e4 | 599 | { |
cf533ea5 | 600 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
601 | struct { |
602 | struct tcphdr th; | |
603 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 604 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
605 | #endif |
606 | } rep; | |
1da177e4 | 607 | struct ip_reply_arg arg; |
cfb6eeb4 | 608 | #ifdef CONFIG_TCP_MD5SIG |
e46787f0 | 609 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
610 | const __u8 *hash_location = NULL; |
611 | unsigned char newhash[16]; | |
612 | int genhash; | |
613 | struct sock *sk1 = NULL; | |
cfb6eeb4 | 614 | #endif |
a86b1e30 | 615 | struct net *net; |
1da177e4 LT |
616 | |
617 | /* Never send a reset in response to a reset. */ | |
618 | if (th->rst) | |
619 | return; | |
620 | ||
c3658e8d ED |
621 | /* If sk not NULL, it means we did a successful lookup and incoming |
622 | * route had to be correct. prequeue might have dropped our dst. | |
623 | */ | |
624 | if (!sk && skb_rtable(skb)->rt_type != RTN_LOCAL) | |
1da177e4 LT |
625 | return; |
626 | ||
627 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
628 | memset(&rep, 0, sizeof(rep)); |
629 | rep.th.dest = th->source; | |
630 | rep.th.source = th->dest; | |
631 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
632 | rep.th.rst = 1; | |
1da177e4 LT |
633 | |
634 | if (th->ack) { | |
cfb6eeb4 | 635 | rep.th.seq = th->ack_seq; |
1da177e4 | 636 | } else { |
cfb6eeb4 YH |
637 | rep.th.ack = 1; |
638 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
639 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
640 | } |
641 | ||
7174259e | 642 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
643 | arg.iov[0].iov_base = (unsigned char *)&rep; |
644 | arg.iov[0].iov_len = sizeof(rep.th); | |
645 | ||
0f85feae | 646 | net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); |
cfb6eeb4 | 647 | #ifdef CONFIG_TCP_MD5SIG |
3b24d854 | 648 | rcu_read_lock(); |
658ddaaf | 649 | hash_location = tcp_parse_md5sig_option(th); |
271c3b9b | 650 | if (sk && sk_fullsock(sk)) { |
e46787f0 FW |
651 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *) |
652 | &ip_hdr(skb)->saddr, AF_INET); | |
653 | } else if (hash_location) { | |
658ddaaf SL |
654 | /* |
655 | * active side is lost. Try to find listening socket through | |
656 | * source port, and then find md5 key through listening socket. | |
657 | * we are not loose security here: | |
658 | * Incoming packet is checked with md5 hash with finding key, | |
659 | * no RST generated if md5 hash doesn't match. | |
660 | */ | |
a583636a CG |
661 | sk1 = __inet_lookup_listener(net, &tcp_hashinfo, NULL, 0, |
662 | ip_hdr(skb)->saddr, | |
da5e3630 | 663 | th->source, ip_hdr(skb)->daddr, |
3fa6f616 DA |
664 | ntohs(th->source), inet_iif(skb), |
665 | tcp_v4_sdif(skb)); | |
658ddaaf SL |
666 | /* don't send rst if it can't find key */ |
667 | if (!sk1) | |
3b24d854 ED |
668 | goto out; |
669 | ||
658ddaaf SL |
670 | key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *) |
671 | &ip_hdr(skb)->saddr, AF_INET); | |
672 | if (!key) | |
3b24d854 ED |
673 | goto out; |
674 | ||
658ddaaf | 675 | |
39f8e58e | 676 | genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb); |
658ddaaf | 677 | if (genhash || memcmp(hash_location, newhash, 16) != 0) |
3b24d854 ED |
678 | goto out; |
679 | ||
658ddaaf SL |
680 | } |
681 | ||
cfb6eeb4 YH |
682 | if (key) { |
683 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
684 | (TCPOPT_NOP << 16) | | |
685 | (TCPOPT_MD5SIG << 8) | | |
686 | TCPOLEN_MD5SIG); | |
687 | /* Update length and the length the header thinks exists */ | |
688 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
689 | rep.th.doff = arg.iov[0].iov_len / 4; | |
690 | ||
49a72dfb | 691 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
692 | key, ip_hdr(skb)->saddr, |
693 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
694 | } |
695 | #endif | |
eddc9ec5 ACM |
696 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
697 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 698 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 699 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
271c3b9b FW |
700 | arg.flags = (sk && inet_sk_transparent(sk)) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
701 | ||
e2446eaa | 702 | /* When socket is gone, all binding information is lost. |
4c675258 AK |
703 | * routing might fail in this case. No choice here, if we choose to force |
704 | * input interface, we will misroute in case of asymmetric route. | |
e2446eaa | 705 | */ |
c24b14c4 | 706 | if (sk) { |
4c675258 | 707 | arg.bound_dev_if = sk->sk_bound_dev_if; |
c24b14c4 SL |
708 | trace_tcp_send_reset(sk, skb); |
709 | } | |
1da177e4 | 710 | |
271c3b9b FW |
711 | BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) != |
712 | offsetof(struct inet_timewait_sock, tw_bound_dev_if)); | |
713 | ||
66b13d99 | 714 | arg.tos = ip_hdr(skb)->tos; |
e2d118a1 | 715 | arg.uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL); |
47dcc20a | 716 | local_bh_disable(); |
bdbbb852 ED |
717 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
718 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
719 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
720 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 721 | |
90bbcc60 ED |
722 | __TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
723 | __TCP_INC_STATS(net, TCP_MIB_OUTRSTS); | |
47dcc20a | 724 | local_bh_enable(); |
658ddaaf SL |
725 | |
726 | #ifdef CONFIG_TCP_MD5SIG | |
3b24d854 ED |
727 | out: |
728 | rcu_read_unlock(); | |
658ddaaf | 729 | #endif |
1da177e4 LT |
730 | } |
731 | ||
732 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
733 | outside socket context is ugly, certainly. What can I do? | |
734 | */ | |
735 | ||
e2d118a1 | 736 | static void tcp_v4_send_ack(const struct sock *sk, |
e62a123b | 737 | struct sk_buff *skb, u32 seq, u32 ack, |
ee684b6f | 738 | u32 win, u32 tsval, u32 tsecr, int oif, |
88ef4a5a | 739 | struct tcp_md5sig_key *key, |
66b13d99 | 740 | int reply_flags, u8 tos) |
1da177e4 | 741 | { |
cf533ea5 | 742 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
743 | struct { |
744 | struct tcphdr th; | |
714e85be | 745 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 746 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 747 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
748 | #endif |
749 | ]; | |
1da177e4 | 750 | } rep; |
e2d118a1 | 751 | struct net *net = sock_net(sk); |
1da177e4 LT |
752 | struct ip_reply_arg arg; |
753 | ||
754 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 755 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
756 | |
757 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
758 | arg.iov[0].iov_len = sizeof(rep.th); | |
ee684b6f | 759 | if (tsecr) { |
cfb6eeb4 YH |
760 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
761 | (TCPOPT_TIMESTAMP << 8) | | |
762 | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
763 | rep.opt[1] = htonl(tsval); |
764 | rep.opt[2] = htonl(tsecr); | |
cb48cfe8 | 765 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
766 | } |
767 | ||
768 | /* Swap the send and the receive. */ | |
769 | rep.th.dest = th->source; | |
770 | rep.th.source = th->dest; | |
771 | rep.th.doff = arg.iov[0].iov_len / 4; | |
772 | rep.th.seq = htonl(seq); | |
773 | rep.th.ack_seq = htonl(ack); | |
774 | rep.th.ack = 1; | |
775 | rep.th.window = htons(win); | |
776 | ||
cfb6eeb4 | 777 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 778 | if (key) { |
ee684b6f | 779 | int offset = (tsecr) ? 3 : 0; |
cfb6eeb4 YH |
780 | |
781 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
782 | (TCPOPT_NOP << 16) | | |
783 | (TCPOPT_MD5SIG << 8) | | |
784 | TCPOLEN_MD5SIG); | |
785 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
786 | rep.th.doff = arg.iov[0].iov_len/4; | |
787 | ||
49a72dfb | 788 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
789 | key, ip_hdr(skb)->saddr, |
790 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
791 | } |
792 | #endif | |
88ef4a5a | 793 | arg.flags = reply_flags; |
eddc9ec5 ACM |
794 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
795 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
796 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
797 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
798 | if (oif) |
799 | arg.bound_dev_if = oif; | |
66b13d99 | 800 | arg.tos = tos; |
e2d118a1 | 801 | arg.uid = sock_net_uid(net, sk_fullsock(sk) ? sk : NULL); |
47dcc20a | 802 | local_bh_disable(); |
bdbbb852 ED |
803 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
804 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
805 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
806 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 807 | |
90bbcc60 | 808 | __TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
47dcc20a | 809 | local_bh_enable(); |
1da177e4 LT |
810 | } |
811 | ||
812 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
813 | { | |
8feaf0c0 | 814 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 815 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 816 | |
e2d118a1 | 817 | tcp_v4_send_ack(sk, skb, |
e62a123b | 818 | tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 819 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
9a568de4 | 820 | tcp_time_stamp_raw() + tcptw->tw_ts_offset, |
9501f972 YH |
821 | tcptw->tw_ts_recent, |
822 | tw->tw_bound_dev_if, | |
88ef4a5a | 823 | tcp_twsk_md5_key(tcptw), |
66b13d99 ED |
824 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0, |
825 | tw->tw_tos | |
9501f972 | 826 | ); |
1da177e4 | 827 | |
8feaf0c0 | 828 | inet_twsk_put(tw); |
1da177e4 LT |
829 | } |
830 | ||
a00e7444 | 831 | static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb, |
7174259e | 832 | struct request_sock *req) |
1da177e4 | 833 | { |
168a8f58 JC |
834 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
835 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
836 | */ | |
e62a123b ED |
837 | u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 : |
838 | tcp_sk(sk)->snd_nxt; | |
839 | ||
20a2b49f ED |
840 | /* RFC 7323 2.3 |
841 | * The window field (SEG.WND) of every outgoing segment, with the | |
842 | * exception of <SYN> segments, MUST be right-shifted by | |
843 | * Rcv.Wind.Shift bits: | |
844 | */ | |
e2d118a1 | 845 | tcp_v4_send_ack(sk, skb, seq, |
20a2b49f ED |
846 | tcp_rsk(req)->rcv_nxt, |
847 | req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale, | |
9a568de4 | 848 | tcp_time_stamp_raw() + tcp_rsk(req)->ts_off, |
9501f972 YH |
849 | req->ts_recent, |
850 | 0, | |
a915da9b ED |
851 | tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr, |
852 | AF_INET), | |
66b13d99 ED |
853 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0, |
854 | ip_hdr(skb)->tos); | |
1da177e4 LT |
855 | } |
856 | ||
1da177e4 | 857 | /* |
9bf1d83e | 858 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 859 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
860 | * socket. |
861 | */ | |
0f935dbe | 862 | static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst, |
d6274bd8 | 863 | struct flowi *fl, |
72659ecc | 864 | struct request_sock *req, |
ca6fb065 | 865 | struct tcp_fastopen_cookie *foc, |
b3d05147 | 866 | enum tcp_synack_type synack_type) |
1da177e4 | 867 | { |
2e6599cb | 868 | const struct inet_request_sock *ireq = inet_rsk(req); |
6bd023f3 | 869 | struct flowi4 fl4; |
1da177e4 | 870 | int err = -1; |
d41db5af | 871 | struct sk_buff *skb; |
1da177e4 LT |
872 | |
873 | /* First, grab a route. */ | |
ba3f7f04 | 874 | if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL) |
fd80eb94 | 875 | return -1; |
1da177e4 | 876 | |
b3d05147 | 877 | skb = tcp_make_synack(sk, dst, req, foc, synack_type); |
1da177e4 LT |
878 | |
879 | if (skb) { | |
634fb979 | 880 | __tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr); |
1da177e4 | 881 | |
634fb979 ED |
882 | err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr, |
883 | ireq->ir_rmt_addr, | |
c92e8c02 | 884 | rcu_dereference(ireq->ireq_opt)); |
b9df3cb8 | 885 | err = net_xmit_eval(err); |
1da177e4 LT |
886 | } |
887 | ||
1da177e4 LT |
888 | return err; |
889 | } | |
890 | ||
891 | /* | |
60236fdd | 892 | * IPv4 request_sock destructor. |
1da177e4 | 893 | */ |
60236fdd | 894 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 895 | { |
c92e8c02 | 896 | kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1)); |
1da177e4 LT |
897 | } |
898 | ||
cfb6eeb4 YH |
899 | #ifdef CONFIG_TCP_MD5SIG |
900 | /* | |
901 | * RFC2385 MD5 checksumming requires a mapping of | |
902 | * IP address->MD5 Key. | |
903 | * We need to maintain these in the sk structure. | |
904 | */ | |
905 | ||
906 | /* Find the Key structure for an address. */ | |
b83e3deb | 907 | struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk, |
a915da9b ED |
908 | const union tcp_md5_addr *addr, |
909 | int family) | |
cfb6eeb4 | 910 | { |
fd3a154a | 911 | const struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 912 | struct tcp_md5sig_key *key; |
fd3a154a | 913 | const struct tcp_md5sig_info *md5sig; |
6797318e ID |
914 | __be32 mask; |
915 | struct tcp_md5sig_key *best_match = NULL; | |
916 | bool match; | |
cfb6eeb4 | 917 | |
a8afca03 ED |
918 | /* caller either holds rcu_read_lock() or socket lock */ |
919 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
1e1d04e6 | 920 | lockdep_sock_is_held(sk)); |
a8afca03 | 921 | if (!md5sig) |
cfb6eeb4 | 922 | return NULL; |
083a0326 | 923 | |
b67bfe0d | 924 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { |
a915da9b ED |
925 | if (key->family != family) |
926 | continue; | |
6797318e ID |
927 | |
928 | if (family == AF_INET) { | |
929 | mask = inet_make_mask(key->prefixlen); | |
930 | match = (key->addr.a4.s_addr & mask) == | |
931 | (addr->a4.s_addr & mask); | |
932 | #if IS_ENABLED(CONFIG_IPV6) | |
933 | } else if (family == AF_INET6) { | |
934 | match = ipv6_prefix_equal(&key->addr.a6, &addr->a6, | |
935 | key->prefixlen); | |
936 | #endif | |
937 | } else { | |
938 | match = false; | |
939 | } | |
940 | ||
941 | if (match && (!best_match || | |
942 | key->prefixlen > best_match->prefixlen)) | |
943 | best_match = key; | |
944 | } | |
945 | return best_match; | |
946 | } | |
947 | EXPORT_SYMBOL(tcp_md5_do_lookup); | |
948 | ||
e8f37d57 WF |
949 | static struct tcp_md5sig_key *tcp_md5_do_lookup_exact(const struct sock *sk, |
950 | const union tcp_md5_addr *addr, | |
951 | int family, u8 prefixlen) | |
6797318e ID |
952 | { |
953 | const struct tcp_sock *tp = tcp_sk(sk); | |
954 | struct tcp_md5sig_key *key; | |
955 | unsigned int size = sizeof(struct in_addr); | |
956 | const struct tcp_md5sig_info *md5sig; | |
957 | ||
958 | /* caller either holds rcu_read_lock() or socket lock */ | |
959 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
960 | lockdep_sock_is_held(sk)); | |
961 | if (!md5sig) | |
962 | return NULL; | |
963 | #if IS_ENABLED(CONFIG_IPV6) | |
964 | if (family == AF_INET6) | |
965 | size = sizeof(struct in6_addr); | |
966 | #endif | |
967 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { | |
968 | if (key->family != family) | |
969 | continue; | |
970 | if (!memcmp(&key->addr, addr, size) && | |
971 | key->prefixlen == prefixlen) | |
a915da9b | 972 | return key; |
cfb6eeb4 YH |
973 | } |
974 | return NULL; | |
975 | } | |
976 | ||
b83e3deb | 977 | struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk, |
fd3a154a | 978 | const struct sock *addr_sk) |
cfb6eeb4 | 979 | { |
b52e6921 | 980 | const union tcp_md5_addr *addr; |
a915da9b | 981 | |
b52e6921 | 982 | addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr; |
a915da9b | 983 | return tcp_md5_do_lookup(sk, addr, AF_INET); |
cfb6eeb4 | 984 | } |
cfb6eeb4 YH |
985 | EXPORT_SYMBOL(tcp_v4_md5_lookup); |
986 | ||
cfb6eeb4 | 987 | /* This can be called on a newly created socket, from other files */ |
a915da9b | 988 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
6797318e ID |
989 | int family, u8 prefixlen, const u8 *newkey, u8 newkeylen, |
990 | gfp_t gfp) | |
cfb6eeb4 YH |
991 | { |
992 | /* Add Key to the list */ | |
b0a713e9 | 993 | struct tcp_md5sig_key *key; |
cfb6eeb4 | 994 | struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 995 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 996 | |
6797318e | 997 | key = tcp_md5_do_lookup_exact(sk, addr, family, prefixlen); |
cfb6eeb4 YH |
998 | if (key) { |
999 | /* Pre-existing entry - just update that one. */ | |
a915da9b | 1000 | memcpy(key->key, newkey, newkeylen); |
b0a713e9 | 1001 | key->keylen = newkeylen; |
a915da9b ED |
1002 | return 0; |
1003 | } | |
260fcbeb | 1004 | |
a8afca03 | 1005 | md5sig = rcu_dereference_protected(tp->md5sig_info, |
1e1d04e6 | 1006 | lockdep_sock_is_held(sk)); |
a915da9b ED |
1007 | if (!md5sig) { |
1008 | md5sig = kmalloc(sizeof(*md5sig), gfp); | |
1009 | if (!md5sig) | |
cfb6eeb4 | 1010 | return -ENOMEM; |
cfb6eeb4 | 1011 | |
a915da9b ED |
1012 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
1013 | INIT_HLIST_HEAD(&md5sig->head); | |
a8afca03 | 1014 | rcu_assign_pointer(tp->md5sig_info, md5sig); |
a915da9b | 1015 | } |
cfb6eeb4 | 1016 | |
5f3d9cb2 | 1017 | key = sock_kmalloc(sk, sizeof(*key), gfp); |
a915da9b ED |
1018 | if (!key) |
1019 | return -ENOMEM; | |
71cea17e | 1020 | if (!tcp_alloc_md5sig_pool()) { |
5f3d9cb2 | 1021 | sock_kfree_s(sk, key, sizeof(*key)); |
a915da9b | 1022 | return -ENOMEM; |
cfb6eeb4 | 1023 | } |
a915da9b ED |
1024 | |
1025 | memcpy(key->key, newkey, newkeylen); | |
1026 | key->keylen = newkeylen; | |
1027 | key->family = family; | |
6797318e | 1028 | key->prefixlen = prefixlen; |
a915da9b ED |
1029 | memcpy(&key->addr, addr, |
1030 | (family == AF_INET6) ? sizeof(struct in6_addr) : | |
1031 | sizeof(struct in_addr)); | |
1032 | hlist_add_head_rcu(&key->node, &md5sig->head); | |
cfb6eeb4 YH |
1033 | return 0; |
1034 | } | |
a915da9b | 1035 | EXPORT_SYMBOL(tcp_md5_do_add); |
cfb6eeb4 | 1036 | |
6797318e ID |
1037 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family, |
1038 | u8 prefixlen) | |
cfb6eeb4 | 1039 | { |
a915da9b ED |
1040 | struct tcp_md5sig_key *key; |
1041 | ||
6797318e | 1042 | key = tcp_md5_do_lookup_exact(sk, addr, family, prefixlen); |
a915da9b ED |
1043 | if (!key) |
1044 | return -ENOENT; | |
1045 | hlist_del_rcu(&key->node); | |
5f3d9cb2 | 1046 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1047 | kfree_rcu(key, rcu); |
a915da9b | 1048 | return 0; |
cfb6eeb4 | 1049 | } |
a915da9b | 1050 | EXPORT_SYMBOL(tcp_md5_do_del); |
cfb6eeb4 | 1051 | |
e0683e70 | 1052 | static void tcp_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
1053 | { |
1054 | struct tcp_sock *tp = tcp_sk(sk); | |
a915da9b | 1055 | struct tcp_md5sig_key *key; |
b67bfe0d | 1056 | struct hlist_node *n; |
a8afca03 | 1057 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 1058 | |
a8afca03 ED |
1059 | md5sig = rcu_dereference_protected(tp->md5sig_info, 1); |
1060 | ||
b67bfe0d | 1061 | hlist_for_each_entry_safe(key, n, &md5sig->head, node) { |
a915da9b | 1062 | hlist_del_rcu(&key->node); |
5f3d9cb2 | 1063 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1064 | kfree_rcu(key, rcu); |
cfb6eeb4 YH |
1065 | } |
1066 | } | |
1067 | ||
8917a777 ID |
1068 | static int tcp_v4_parse_md5_keys(struct sock *sk, int optname, |
1069 | char __user *optval, int optlen) | |
cfb6eeb4 YH |
1070 | { |
1071 | struct tcp_md5sig cmd; | |
1072 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
8917a777 | 1073 | u8 prefixlen = 32; |
cfb6eeb4 YH |
1074 | |
1075 | if (optlen < sizeof(cmd)) | |
1076 | return -EINVAL; | |
1077 | ||
7174259e | 1078 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
1079 | return -EFAULT; |
1080 | ||
1081 | if (sin->sin_family != AF_INET) | |
1082 | return -EINVAL; | |
1083 | ||
8917a777 ID |
1084 | if (optname == TCP_MD5SIG_EXT && |
1085 | cmd.tcpm_flags & TCP_MD5SIG_FLAG_PREFIX) { | |
1086 | prefixlen = cmd.tcpm_prefixlen; | |
1087 | if (prefixlen > 32) | |
1088 | return -EINVAL; | |
1089 | } | |
1090 | ||
64a124ed | 1091 | if (!cmd.tcpm_keylen) |
a915da9b | 1092 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
8917a777 | 1093 | AF_INET, prefixlen); |
cfb6eeb4 YH |
1094 | |
1095 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
1096 | return -EINVAL; | |
1097 | ||
a915da9b | 1098 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
8917a777 | 1099 | AF_INET, prefixlen, cmd.tcpm_key, cmd.tcpm_keylen, |
a915da9b | 1100 | GFP_KERNEL); |
cfb6eeb4 YH |
1101 | } |
1102 | ||
19689e38 ED |
1103 | static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp, |
1104 | __be32 daddr, __be32 saddr, | |
1105 | const struct tcphdr *th, int nbytes) | |
cfb6eeb4 | 1106 | { |
cfb6eeb4 | 1107 | struct tcp4_pseudohdr *bp; |
49a72dfb | 1108 | struct scatterlist sg; |
19689e38 | 1109 | struct tcphdr *_th; |
cfb6eeb4 | 1110 | |
19689e38 | 1111 | bp = hp->scratch; |
cfb6eeb4 YH |
1112 | bp->saddr = saddr; |
1113 | bp->daddr = daddr; | |
1114 | bp->pad = 0; | |
076fb722 | 1115 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1116 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1117 | |
19689e38 ED |
1118 | _th = (struct tcphdr *)(bp + 1); |
1119 | memcpy(_th, th, sizeof(*th)); | |
1120 | _th->check = 0; | |
1121 | ||
1122 | sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th)); | |
1123 | ahash_request_set_crypt(hp->md5_req, &sg, NULL, | |
1124 | sizeof(*bp) + sizeof(*th)); | |
cf80e0e4 | 1125 | return crypto_ahash_update(hp->md5_req); |
49a72dfb AL |
1126 | } |
1127 | ||
a915da9b | 1128 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 1129 | __be32 daddr, __be32 saddr, const struct tcphdr *th) |
49a72dfb AL |
1130 | { |
1131 | struct tcp_md5sig_pool *hp; | |
cf80e0e4 | 1132 | struct ahash_request *req; |
49a72dfb AL |
1133 | |
1134 | hp = tcp_get_md5sig_pool(); | |
1135 | if (!hp) | |
1136 | goto clear_hash_noput; | |
cf80e0e4 | 1137 | req = hp->md5_req; |
49a72dfb | 1138 | |
cf80e0e4 | 1139 | if (crypto_ahash_init(req)) |
49a72dfb | 1140 | goto clear_hash; |
19689e38 | 1141 | if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2)) |
49a72dfb AL |
1142 | goto clear_hash; |
1143 | if (tcp_md5_hash_key(hp, key)) | |
1144 | goto clear_hash; | |
cf80e0e4 HX |
1145 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
1146 | if (crypto_ahash_final(req)) | |
cfb6eeb4 YH |
1147 | goto clear_hash; |
1148 | ||
cfb6eeb4 | 1149 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1150 | return 0; |
49a72dfb | 1151 | |
cfb6eeb4 YH |
1152 | clear_hash: |
1153 | tcp_put_md5sig_pool(); | |
1154 | clear_hash_noput: | |
1155 | memset(md5_hash, 0, 16); | |
49a72dfb | 1156 | return 1; |
cfb6eeb4 YH |
1157 | } |
1158 | ||
39f8e58e ED |
1159 | int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, |
1160 | const struct sock *sk, | |
318cf7aa | 1161 | const struct sk_buff *skb) |
cfb6eeb4 | 1162 | { |
49a72dfb | 1163 | struct tcp_md5sig_pool *hp; |
cf80e0e4 | 1164 | struct ahash_request *req; |
318cf7aa | 1165 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
1166 | __be32 saddr, daddr; |
1167 | ||
39f8e58e ED |
1168 | if (sk) { /* valid for establish/request sockets */ |
1169 | saddr = sk->sk_rcv_saddr; | |
1170 | daddr = sk->sk_daddr; | |
cfb6eeb4 | 1171 | } else { |
49a72dfb AL |
1172 | const struct iphdr *iph = ip_hdr(skb); |
1173 | saddr = iph->saddr; | |
1174 | daddr = iph->daddr; | |
cfb6eeb4 | 1175 | } |
49a72dfb AL |
1176 | |
1177 | hp = tcp_get_md5sig_pool(); | |
1178 | if (!hp) | |
1179 | goto clear_hash_noput; | |
cf80e0e4 | 1180 | req = hp->md5_req; |
49a72dfb | 1181 | |
cf80e0e4 | 1182 | if (crypto_ahash_init(req)) |
49a72dfb AL |
1183 | goto clear_hash; |
1184 | ||
19689e38 | 1185 | if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, skb->len)) |
49a72dfb AL |
1186 | goto clear_hash; |
1187 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1188 | goto clear_hash; | |
1189 | if (tcp_md5_hash_key(hp, key)) | |
1190 | goto clear_hash; | |
cf80e0e4 HX |
1191 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
1192 | if (crypto_ahash_final(req)) | |
49a72dfb AL |
1193 | goto clear_hash; |
1194 | ||
1195 | tcp_put_md5sig_pool(); | |
1196 | return 0; | |
1197 | ||
1198 | clear_hash: | |
1199 | tcp_put_md5sig_pool(); | |
1200 | clear_hash_noput: | |
1201 | memset(md5_hash, 0, 16); | |
1202 | return 1; | |
cfb6eeb4 | 1203 | } |
49a72dfb | 1204 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1205 | |
ba8e275a ED |
1206 | #endif |
1207 | ||
ff74e23f | 1208 | /* Called with rcu_read_lock() */ |
ba8e275a | 1209 | static bool tcp_v4_inbound_md5_hash(const struct sock *sk, |
ff74e23f | 1210 | const struct sk_buff *skb) |
cfb6eeb4 | 1211 | { |
ba8e275a | 1212 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
1213 | /* |
1214 | * This gets called for each TCP segment that arrives | |
1215 | * so we want to be efficient. | |
1216 | * We have 3 drop cases: | |
1217 | * o No MD5 hash and one expected. | |
1218 | * o MD5 hash and we're not expecting one. | |
1219 | * o MD5 hash and its wrong. | |
1220 | */ | |
cf533ea5 | 1221 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 1222 | struct tcp_md5sig_key *hash_expected; |
eddc9ec5 | 1223 | const struct iphdr *iph = ip_hdr(skb); |
cf533ea5 | 1224 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1225 | int genhash; |
cfb6eeb4 YH |
1226 | unsigned char newhash[16]; |
1227 | ||
a915da9b ED |
1228 | hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr, |
1229 | AF_INET); | |
7d5d5525 | 1230 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1231 | |
cfb6eeb4 YH |
1232 | /* We've parsed the options - do we have a hash? */ |
1233 | if (!hash_expected && !hash_location) | |
a2a385d6 | 1234 | return false; |
cfb6eeb4 YH |
1235 | |
1236 | if (hash_expected && !hash_location) { | |
c10d9310 | 1237 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
a2a385d6 | 1238 | return true; |
cfb6eeb4 YH |
1239 | } |
1240 | ||
1241 | if (!hash_expected && hash_location) { | |
c10d9310 | 1242 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
a2a385d6 | 1243 | return true; |
cfb6eeb4 YH |
1244 | } |
1245 | ||
1246 | /* Okay, so this is hash_expected and hash_location - | |
1247 | * so we need to calculate the checksum. | |
1248 | */ | |
49a72dfb AL |
1249 | genhash = tcp_v4_md5_hash_skb(newhash, |
1250 | hash_expected, | |
39f8e58e | 1251 | NULL, skb); |
cfb6eeb4 YH |
1252 | |
1253 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
72145a68 | 1254 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE); |
e87cc472 JP |
1255 | net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1256 | &iph->saddr, ntohs(th->source), | |
1257 | &iph->daddr, ntohs(th->dest), | |
1258 | genhash ? " tcp_v4_calc_md5_hash failed" | |
1259 | : ""); | |
a2a385d6 | 1260 | return true; |
cfb6eeb4 | 1261 | } |
a2a385d6 | 1262 | return false; |
cfb6eeb4 | 1263 | #endif |
ba8e275a ED |
1264 | return false; |
1265 | } | |
cfb6eeb4 | 1266 | |
b40cf18e ED |
1267 | static void tcp_v4_init_req(struct request_sock *req, |
1268 | const struct sock *sk_listener, | |
16bea70a OP |
1269 | struct sk_buff *skb) |
1270 | { | |
1271 | struct inet_request_sock *ireq = inet_rsk(req); | |
c92e8c02 | 1272 | struct net *net = sock_net(sk_listener); |
16bea70a | 1273 | |
08d2cc3b ED |
1274 | sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr); |
1275 | sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr); | |
c92e8c02 | 1276 | RCU_INIT_POINTER(ireq->ireq_opt, tcp_v4_save_options(net, skb)); |
16bea70a OP |
1277 | } |
1278 | ||
f964629e ED |
1279 | static struct dst_entry *tcp_v4_route_req(const struct sock *sk, |
1280 | struct flowi *fl, | |
4396e461 | 1281 | const struct request_sock *req) |
d94e0417 | 1282 | { |
4396e461 | 1283 | return inet_csk_route_req(sk, &fl->u.ip4, req); |
d94e0417 OP |
1284 | } |
1285 | ||
72a3effa | 1286 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1287 | .family = PF_INET, |
2e6599cb | 1288 | .obj_size = sizeof(struct tcp_request_sock), |
5db92c99 | 1289 | .rtx_syn_ack = tcp_rtx_synack, |
60236fdd ACM |
1290 | .send_ack = tcp_v4_reqsk_send_ack, |
1291 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 | 1292 | .send_reset = tcp_v4_send_reset, |
688d1945 | 1293 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
1294 | }; |
1295 | ||
b2e4b3de | 1296 | static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
2aec4a29 | 1297 | .mss_clamp = TCP_MSS_DEFAULT, |
16bea70a | 1298 | #ifdef CONFIG_TCP_MD5SIG |
fd3a154a | 1299 | .req_md5_lookup = tcp_v4_md5_lookup, |
e3afe7b7 | 1300 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
b6332e6c | 1301 | #endif |
16bea70a | 1302 | .init_req = tcp_v4_init_req, |
fb7b37a7 OP |
1303 | #ifdef CONFIG_SYN_COOKIES |
1304 | .cookie_init_seq = cookie_v4_init_sequence, | |
1305 | #endif | |
d94e0417 | 1306 | .route_req = tcp_v4_route_req, |
84b114b9 ED |
1307 | .init_seq = tcp_v4_init_seq, |
1308 | .init_ts_off = tcp_v4_init_ts_off, | |
d6274bd8 | 1309 | .send_synack = tcp_v4_send_synack, |
16bea70a | 1310 | }; |
cfb6eeb4 | 1311 | |
1da177e4 LT |
1312 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1313 | { | |
1da177e4 | 1314 | /* Never answer to SYNs send to broadcast or multicast */ |
511c3f92 | 1315 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1316 | goto drop; |
1317 | ||
1fb6f159 OP |
1318 | return tcp_conn_request(&tcp_request_sock_ops, |
1319 | &tcp_request_sock_ipv4_ops, sk, skb); | |
1da177e4 | 1320 | |
1da177e4 | 1321 | drop: |
9caad864 | 1322 | tcp_listendrop(sk); |
1da177e4 LT |
1323 | return 0; |
1324 | } | |
4bc2f18b | 1325 | EXPORT_SYMBOL(tcp_v4_conn_request); |
1da177e4 LT |
1326 | |
1327 | ||
1328 | /* | |
1329 | * The three way handshake has completed - we got a valid synack - | |
1330 | * now create the new socket. | |
1331 | */ | |
0c27171e | 1332 | struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb, |
60236fdd | 1333 | struct request_sock *req, |
5e0724d0 ED |
1334 | struct dst_entry *dst, |
1335 | struct request_sock *req_unhash, | |
1336 | bool *own_req) | |
1da177e4 | 1337 | { |
2e6599cb | 1338 | struct inet_request_sock *ireq; |
1da177e4 LT |
1339 | struct inet_sock *newinet; |
1340 | struct tcp_sock *newtp; | |
1341 | struct sock *newsk; | |
cfb6eeb4 YH |
1342 | #ifdef CONFIG_TCP_MD5SIG |
1343 | struct tcp_md5sig_key *key; | |
1344 | #endif | |
f6d8bd05 | 1345 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
1346 | |
1347 | if (sk_acceptq_is_full(sk)) | |
1348 | goto exit_overflow; | |
1349 | ||
1da177e4 LT |
1350 | newsk = tcp_create_openreq_child(sk, req, skb); |
1351 | if (!newsk) | |
093d2823 | 1352 | goto exit_nonewsk; |
1da177e4 | 1353 | |
bcd76111 | 1354 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
fae6ef87 | 1355 | inet_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1356 | |
1357 | newtp = tcp_sk(newsk); | |
1358 | newinet = inet_sk(newsk); | |
2e6599cb | 1359 | ireq = inet_rsk(req); |
d1e559d0 ED |
1360 | sk_daddr_set(newsk, ireq->ir_rmt_addr); |
1361 | sk_rcv_saddr_set(newsk, ireq->ir_loc_addr); | |
6dd9a14e | 1362 | newsk->sk_bound_dev_if = ireq->ir_iif; |
c92e8c02 ED |
1363 | newinet->inet_saddr = ireq->ir_loc_addr; |
1364 | inet_opt = rcu_dereference(ireq->ireq_opt); | |
1365 | RCU_INIT_POINTER(newinet->inet_opt, inet_opt); | |
463c84b9 | 1366 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1367 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
4c507d28 | 1368 | newinet->rcv_tos = ip_hdr(skb)->tos; |
d83d8461 | 1369 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
1370 | if (inet_opt) |
1371 | inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
c720c7e8 | 1372 | newinet->inet_id = newtp->write_seq ^ jiffies; |
1da177e4 | 1373 | |
dfd25fff ED |
1374 | if (!dst) { |
1375 | dst = inet_csk_route_child_sock(sk, newsk, req); | |
1376 | if (!dst) | |
1377 | goto put_and_exit; | |
1378 | } else { | |
1379 | /* syncookie case : see end of cookie_v4_check() */ | |
1380 | } | |
0e734419 DM |
1381 | sk_setup_caps(newsk, dst); |
1382 | ||
81164413 DB |
1383 | tcp_ca_openreq_child(newsk, dst); |
1384 | ||
1da177e4 | 1385 | tcp_sync_mss(newsk, dst_mtu(dst)); |
3541f9e8 | 1386 | newtp->advmss = tcp_mss_clamp(tcp_sk(sk), dst_metric_advmss(dst)); |
f5fff5dc | 1387 | |
1da177e4 LT |
1388 | tcp_initialize_rcv_mss(newsk); |
1389 | ||
cfb6eeb4 YH |
1390 | #ifdef CONFIG_TCP_MD5SIG |
1391 | /* Copy over the MD5 key from the original socket */ | |
a915da9b ED |
1392 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1393 | AF_INET); | |
00db4124 | 1394 | if (key) { |
cfb6eeb4 YH |
1395 | /* |
1396 | * We're using one, so create a matching key | |
1397 | * on the newsk structure. If we fail to get | |
1398 | * memory, then we end up not copying the key | |
1399 | * across. Shucks. | |
1400 | */ | |
a915da9b | 1401 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newinet->inet_daddr, |
6797318e | 1402 | AF_INET, 32, key->key, key->keylen, GFP_ATOMIC); |
a465419b | 1403 | sk_nocaps_add(newsk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1404 | } |
1405 | #endif | |
1406 | ||
0e734419 DM |
1407 | if (__inet_inherit_port(sk, newsk) < 0) |
1408 | goto put_and_exit; | |
5e0724d0 | 1409 | *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash)); |
c92e8c02 | 1410 | if (likely(*own_req)) { |
49a496c9 | 1411 | tcp_move_syn(newtp, req); |
c92e8c02 ED |
1412 | ireq->ireq_opt = NULL; |
1413 | } else { | |
1414 | newinet->inet_opt = NULL; | |
1415 | } | |
1da177e4 LT |
1416 | return newsk; |
1417 | ||
1418 | exit_overflow: | |
c10d9310 | 1419 | NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 BS |
1420 | exit_nonewsk: |
1421 | dst_release(dst); | |
1da177e4 | 1422 | exit: |
9caad864 | 1423 | tcp_listendrop(sk); |
1da177e4 | 1424 | return NULL; |
0e734419 | 1425 | put_and_exit: |
c92e8c02 | 1426 | newinet->inet_opt = NULL; |
e337e24d CP |
1427 | inet_csk_prepare_forced_close(newsk); |
1428 | tcp_done(newsk); | |
0e734419 | 1429 | goto exit; |
1da177e4 | 1430 | } |
4bc2f18b | 1431 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); |
1da177e4 | 1432 | |
079096f1 | 1433 | static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1434 | { |
079096f1 | 1435 | #ifdef CONFIG_SYN_COOKIES |
52452c54 | 1436 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 1437 | |
af9b4738 | 1438 | if (!th->syn) |
461b74c3 | 1439 | sk = cookie_v4_check(sk, skb); |
1da177e4 LT |
1440 | #endif |
1441 | return sk; | |
1442 | } | |
1443 | ||
1da177e4 | 1444 | /* The socket must have it's spinlock held when we get |
e994b2f0 | 1445 | * here, unless it is a TCP_LISTEN socket. |
1da177e4 LT |
1446 | * |
1447 | * We have a potential double-lock case here, so even when | |
1448 | * doing backlog processing we use the BH locking scheme. | |
1449 | * This is because we cannot sleep with the original spinlock | |
1450 | * held. | |
1451 | */ | |
1452 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1453 | { | |
cfb6eeb4 | 1454 | struct sock *rsk; |
cfb6eeb4 | 1455 | |
1da177e4 | 1456 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
404e0a8b ED |
1457 | struct dst_entry *dst = sk->sk_rx_dst; |
1458 | ||
bdeab991 | 1459 | sock_rps_save_rxhash(sk, skb); |
3d97379a | 1460 | sk_mark_napi_id(sk, skb); |
404e0a8b | 1461 | if (dst) { |
505fbcf0 | 1462 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || |
51456b29 | 1463 | !dst->ops->check(dst, 0)) { |
92101b3b DM |
1464 | dst_release(dst); |
1465 | sk->sk_rx_dst = NULL; | |
1466 | } | |
1467 | } | |
e42e24c3 | 1468 | tcp_rcv_established(sk, skb, tcp_hdr(skb)); |
1da177e4 LT |
1469 | return 0; |
1470 | } | |
1471 | ||
12e25e10 | 1472 | if (tcp_checksum_complete(skb)) |
1da177e4 LT |
1473 | goto csum_err; |
1474 | ||
1475 | if (sk->sk_state == TCP_LISTEN) { | |
079096f1 ED |
1476 | struct sock *nsk = tcp_v4_cookie_check(sk, skb); |
1477 | ||
1da177e4 LT |
1478 | if (!nsk) |
1479 | goto discard; | |
1da177e4 | 1480 | if (nsk != sk) { |
cfb6eeb4 YH |
1481 | if (tcp_child_process(sk, nsk, skb)) { |
1482 | rsk = nsk; | |
1da177e4 | 1483 | goto reset; |
cfb6eeb4 | 1484 | } |
1da177e4 LT |
1485 | return 0; |
1486 | } | |
ca55158c | 1487 | } else |
bdeab991 | 1488 | sock_rps_save_rxhash(sk, skb); |
ca55158c | 1489 | |
72ab4a86 | 1490 | if (tcp_rcv_state_process(sk, skb)) { |
cfb6eeb4 | 1491 | rsk = sk; |
1da177e4 | 1492 | goto reset; |
cfb6eeb4 | 1493 | } |
1da177e4 LT |
1494 | return 0; |
1495 | ||
1496 | reset: | |
cfb6eeb4 | 1497 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1498 | discard: |
1499 | kfree_skb(skb); | |
1500 | /* Be careful here. If this function gets more complicated and | |
1501 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1502 | * might be destroyed here. This current version compiles correctly, | |
1503 | * but you have been warned. | |
1504 | */ | |
1505 | return 0; | |
1506 | ||
1507 | csum_err: | |
c10d9310 ED |
1508 | TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS); |
1509 | TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS); | |
1da177e4 LT |
1510 | goto discard; |
1511 | } | |
4bc2f18b | 1512 | EXPORT_SYMBOL(tcp_v4_do_rcv); |
1da177e4 | 1513 | |
7487449c | 1514 | int tcp_v4_early_demux(struct sk_buff *skb) |
41063e9d | 1515 | { |
41063e9d DM |
1516 | const struct iphdr *iph; |
1517 | const struct tcphdr *th; | |
1518 | struct sock *sk; | |
41063e9d | 1519 | |
41063e9d | 1520 | if (skb->pkt_type != PACKET_HOST) |
7487449c | 1521 | return 0; |
41063e9d | 1522 | |
45f00f99 | 1523 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) |
7487449c | 1524 | return 0; |
41063e9d DM |
1525 | |
1526 | iph = ip_hdr(skb); | |
45f00f99 | 1527 | th = tcp_hdr(skb); |
41063e9d DM |
1528 | |
1529 | if (th->doff < sizeof(struct tcphdr) / 4) | |
7487449c | 1530 | return 0; |
41063e9d | 1531 | |
45f00f99 | 1532 | sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo, |
41063e9d | 1533 | iph->saddr, th->source, |
7011d085 | 1534 | iph->daddr, ntohs(th->dest), |
3fa6f616 | 1535 | skb->skb_iif, inet_sdif(skb)); |
41063e9d DM |
1536 | if (sk) { |
1537 | skb->sk = sk; | |
1538 | skb->destructor = sock_edemux; | |
f7e4eb03 | 1539 | if (sk_fullsock(sk)) { |
d0c294c5 | 1540 | struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst); |
505fbcf0 | 1541 | |
41063e9d DM |
1542 | if (dst) |
1543 | dst = dst_check(dst, 0); | |
92101b3b | 1544 | if (dst && |
505fbcf0 | 1545 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
92101b3b | 1546 | skb_dst_set_noref(skb, dst); |
41063e9d DM |
1547 | } |
1548 | } | |
7487449c | 1549 | return 0; |
41063e9d DM |
1550 | } |
1551 | ||
c9c33212 ED |
1552 | bool tcp_add_backlog(struct sock *sk, struct sk_buff *skb) |
1553 | { | |
1554 | u32 limit = sk->sk_rcvbuf + sk->sk_sndbuf; | |
1555 | ||
1556 | /* Only socket owner can try to collapse/prune rx queues | |
1557 | * to reduce memory overhead, so add a little headroom here. | |
1558 | * Few sockets backlog are possibly concurrently non empty. | |
1559 | */ | |
1560 | limit += 64*1024; | |
1561 | ||
1562 | /* In case all data was pulled from skb frags (in __pskb_pull_tail()), | |
1563 | * we can fix skb->truesize to its real value to avoid future drops. | |
1564 | * This is valid because skb is not yet charged to the socket. | |
1565 | * It has been noticed pure SACK packets were sometimes dropped | |
1566 | * (if cooked by drivers without copybreak feature). | |
1567 | */ | |
60b1af33 | 1568 | skb_condense(skb); |
c9c33212 ED |
1569 | |
1570 | if (unlikely(sk_add_backlog(sk, skb, limit))) { | |
1571 | bh_unlock_sock(sk); | |
1572 | __NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPBACKLOGDROP); | |
1573 | return true; | |
1574 | } | |
1575 | return false; | |
1576 | } | |
1577 | EXPORT_SYMBOL(tcp_add_backlog); | |
1578 | ||
ac6e7800 ED |
1579 | int tcp_filter(struct sock *sk, struct sk_buff *skb) |
1580 | { | |
1581 | struct tcphdr *th = (struct tcphdr *)skb->data; | |
1582 | unsigned int eaten = skb->len; | |
1583 | int err; | |
1584 | ||
1585 | err = sk_filter_trim_cap(sk, skb, th->doff * 4); | |
1586 | if (!err) { | |
1587 | eaten -= skb->len; | |
1588 | TCP_SKB_CB(skb)->end_seq -= eaten; | |
1589 | } | |
1590 | return err; | |
1591 | } | |
1592 | EXPORT_SYMBOL(tcp_filter); | |
1593 | ||
1da177e4 LT |
1594 | /* |
1595 | * From tcp_input.c | |
1596 | */ | |
1597 | ||
1598 | int tcp_v4_rcv(struct sk_buff *skb) | |
1599 | { | |
3b24d854 | 1600 | struct net *net = dev_net(skb->dev); |
3fa6f616 | 1601 | int sdif = inet_sdif(skb); |
eddc9ec5 | 1602 | const struct iphdr *iph; |
cf533ea5 | 1603 | const struct tcphdr *th; |
3b24d854 | 1604 | bool refcounted; |
1da177e4 LT |
1605 | struct sock *sk; |
1606 | int ret; | |
1607 | ||
1608 | if (skb->pkt_type != PACKET_HOST) | |
1609 | goto discard_it; | |
1610 | ||
1611 | /* Count it even if it's bad */ | |
90bbcc60 | 1612 | __TCP_INC_STATS(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1613 | |
1614 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1615 | goto discard_it; | |
1616 | ||
ea1627c2 | 1617 | th = (const struct tcphdr *)skb->data; |
1da177e4 | 1618 | |
ea1627c2 | 1619 | if (unlikely(th->doff < sizeof(struct tcphdr) / 4)) |
1da177e4 LT |
1620 | goto bad_packet; |
1621 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1622 | goto discard_it; | |
1623 | ||
1624 | /* An explanation is required here, I think. | |
1625 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1626 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1627 | * So, we defer the checks. */ |
ed70fcfc TH |
1628 | |
1629 | if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo)) | |
6a5dc9e5 | 1630 | goto csum_error; |
1da177e4 | 1631 | |
ea1627c2 | 1632 | th = (const struct tcphdr *)skb->data; |
eddc9ec5 | 1633 | iph = ip_hdr(skb); |
971f10ec ED |
1634 | /* This is tricky : We move IPCB at its correct location into TCP_SKB_CB() |
1635 | * barrier() makes sure compiler wont play fool^Waliasing games. | |
1636 | */ | |
1637 | memmove(&TCP_SKB_CB(skb)->header.h4, IPCB(skb), | |
1638 | sizeof(struct inet_skb_parm)); | |
1639 | barrier(); | |
1640 | ||
1da177e4 LT |
1641 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1642 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1643 | skb->len - th->doff * 4); | |
1644 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
e11ecddf | 1645 | TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th); |
04317daf | 1646 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; |
b82d1bb4 | 1647 | TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph); |
1da177e4 | 1648 | TCP_SKB_CB(skb)->sacked = 0; |
98aaa913 MM |
1649 | TCP_SKB_CB(skb)->has_rxtstamp = |
1650 | skb->tstamp || skb_hwtstamps(skb)->hwtstamp; | |
1da177e4 | 1651 | |
4bdc3d66 | 1652 | lookup: |
a583636a | 1653 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source, |
3fa6f616 | 1654 | th->dest, sdif, &refcounted); |
1da177e4 LT |
1655 | if (!sk) |
1656 | goto no_tcp_socket; | |
1657 | ||
bb134d5d ED |
1658 | process: |
1659 | if (sk->sk_state == TCP_TIME_WAIT) | |
1660 | goto do_time_wait; | |
1661 | ||
079096f1 ED |
1662 | if (sk->sk_state == TCP_NEW_SYN_RECV) { |
1663 | struct request_sock *req = inet_reqsk(sk); | |
7716682c | 1664 | struct sock *nsk; |
079096f1 ED |
1665 | |
1666 | sk = req->rsk_listener; | |
72923555 | 1667 | if (unlikely(tcp_v4_inbound_md5_hash(sk, skb))) { |
e65c332d | 1668 | sk_drops_add(sk, skb); |
72923555 ED |
1669 | reqsk_put(req); |
1670 | goto discard_it; | |
1671 | } | |
7716682c | 1672 | if (unlikely(sk->sk_state != TCP_LISTEN)) { |
f03f2e15 | 1673 | inet_csk_reqsk_queue_drop_and_put(sk, req); |
4bdc3d66 ED |
1674 | goto lookup; |
1675 | } | |
3b24d854 ED |
1676 | /* We own a reference on the listener, increase it again |
1677 | * as we might lose it too soon. | |
1678 | */ | |
7716682c | 1679 | sock_hold(sk); |
3b24d854 | 1680 | refcounted = true; |
1f3b359f ED |
1681 | nsk = NULL; |
1682 | if (!tcp_filter(sk, skb)) | |
1683 | nsk = tcp_check_req(sk, skb, req, false); | |
079096f1 ED |
1684 | if (!nsk) { |
1685 | reqsk_put(req); | |
7716682c | 1686 | goto discard_and_relse; |
079096f1 ED |
1687 | } |
1688 | if (nsk == sk) { | |
079096f1 ED |
1689 | reqsk_put(req); |
1690 | } else if (tcp_child_process(sk, nsk, skb)) { | |
1691 | tcp_v4_send_reset(nsk, skb); | |
7716682c | 1692 | goto discard_and_relse; |
079096f1 | 1693 | } else { |
7716682c | 1694 | sock_put(sk); |
079096f1 ED |
1695 | return 0; |
1696 | } | |
1697 | } | |
6cce09f8 | 1698 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
02a1d6e7 | 1699 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); |
d218d111 | 1700 | goto discard_and_relse; |
6cce09f8 | 1701 | } |
d218d111 | 1702 | |
1da177e4 LT |
1703 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1704 | goto discard_and_relse; | |
9ea88a15 | 1705 | |
9ea88a15 DP |
1706 | if (tcp_v4_inbound_md5_hash(sk, skb)) |
1707 | goto discard_and_relse; | |
9ea88a15 | 1708 | |
b59c2701 | 1709 | nf_reset(skb); |
1da177e4 | 1710 | |
ac6e7800 | 1711 | if (tcp_filter(sk, skb)) |
1da177e4 | 1712 | goto discard_and_relse; |
ac6e7800 ED |
1713 | th = (const struct tcphdr *)skb->data; |
1714 | iph = ip_hdr(skb); | |
1da177e4 LT |
1715 | |
1716 | skb->dev = NULL; | |
1717 | ||
e994b2f0 ED |
1718 | if (sk->sk_state == TCP_LISTEN) { |
1719 | ret = tcp_v4_do_rcv(sk, skb); | |
1720 | goto put_and_return; | |
1721 | } | |
1722 | ||
1723 | sk_incoming_cpu_update(sk); | |
1724 | ||
c6366184 | 1725 | bh_lock_sock_nested(sk); |
a44d6eac | 1726 | tcp_segs_in(tcp_sk(sk), skb); |
1da177e4 LT |
1727 | ret = 0; |
1728 | if (!sock_owned_by_user(sk)) { | |
e7942d06 | 1729 | ret = tcp_v4_do_rcv(sk, skb); |
c9c33212 | 1730 | } else if (tcp_add_backlog(sk, skb)) { |
6b03a53a ZY |
1731 | goto discard_and_relse; |
1732 | } | |
1da177e4 LT |
1733 | bh_unlock_sock(sk); |
1734 | ||
e994b2f0 | 1735 | put_and_return: |
3b24d854 ED |
1736 | if (refcounted) |
1737 | sock_put(sk); | |
1da177e4 LT |
1738 | |
1739 | return ret; | |
1740 | ||
1741 | no_tcp_socket: | |
1742 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1743 | goto discard_it; | |
1744 | ||
12e25e10 | 1745 | if (tcp_checksum_complete(skb)) { |
6a5dc9e5 | 1746 | csum_error: |
90bbcc60 | 1747 | __TCP_INC_STATS(net, TCP_MIB_CSUMERRORS); |
1da177e4 | 1748 | bad_packet: |
90bbcc60 | 1749 | __TCP_INC_STATS(net, TCP_MIB_INERRS); |
1da177e4 | 1750 | } else { |
cfb6eeb4 | 1751 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1752 | } |
1753 | ||
1754 | discard_it: | |
1755 | /* Discard frame. */ | |
1756 | kfree_skb(skb); | |
e905a9ed | 1757 | return 0; |
1da177e4 LT |
1758 | |
1759 | discard_and_relse: | |
532182cd | 1760 | sk_drops_add(sk, skb); |
3b24d854 ED |
1761 | if (refcounted) |
1762 | sock_put(sk); | |
1da177e4 LT |
1763 | goto discard_it; |
1764 | ||
1765 | do_time_wait: | |
1766 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1767 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1768 | goto discard_it; |
1769 | } | |
1770 | ||
6a5dc9e5 ED |
1771 | if (tcp_checksum_complete(skb)) { |
1772 | inet_twsk_put(inet_twsk(sk)); | |
1773 | goto csum_error; | |
1da177e4 | 1774 | } |
9469c7b4 | 1775 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1776 | case TCP_TW_SYN: { |
c346dca1 | 1777 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
a583636a CG |
1778 | &tcp_hashinfo, skb, |
1779 | __tcp_hdrlen(th), | |
da5e3630 | 1780 | iph->saddr, th->source, |
eddc9ec5 | 1781 | iph->daddr, th->dest, |
3fa6f616 DA |
1782 | inet_iif(skb), |
1783 | sdif); | |
1da177e4 | 1784 | if (sk2) { |
dbe7faa4 | 1785 | inet_twsk_deschedule_put(inet_twsk(sk)); |
1da177e4 | 1786 | sk = sk2; |
3b24d854 | 1787 | refcounted = false; |
1da177e4 LT |
1788 | goto process; |
1789 | } | |
1da177e4 | 1790 | } |
fcfd6dfa GS |
1791 | /* to ACK */ |
1792 | /* fall through */ | |
1da177e4 LT |
1793 | case TCP_TW_ACK: |
1794 | tcp_v4_timewait_ack(sk, skb); | |
1795 | break; | |
1796 | case TCP_TW_RST: | |
271c3b9b FW |
1797 | tcp_v4_send_reset(sk, skb); |
1798 | inet_twsk_deschedule_put(inet_twsk(sk)); | |
1799 | goto discard_it; | |
1da177e4 LT |
1800 | case TCP_TW_SUCCESS:; |
1801 | } | |
1802 | goto discard_it; | |
1803 | } | |
1804 | ||
ccb7c410 DM |
1805 | static struct timewait_sock_ops tcp_timewait_sock_ops = { |
1806 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1807 | .twsk_unique = tcp_twsk_unique, | |
1808 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 | 1809 | }; |
1da177e4 | 1810 | |
63d02d15 | 1811 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
5d299f3d ED |
1812 | { |
1813 | struct dst_entry *dst = skb_dst(skb); | |
1814 | ||
5037e9ef | 1815 | if (dst && dst_hold_safe(dst)) { |
ca777eff ED |
1816 | sk->sk_rx_dst = dst; |
1817 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
1818 | } | |
5d299f3d | 1819 | } |
63d02d15 | 1820 | EXPORT_SYMBOL(inet_sk_rx_dst_set); |
5d299f3d | 1821 | |
3b401a81 | 1822 | const struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1823 | .queue_xmit = ip_queue_xmit, |
1824 | .send_check = tcp_v4_send_check, | |
1825 | .rebuild_header = inet_sk_rebuild_header, | |
5d299f3d | 1826 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1827 | .conn_request = tcp_v4_conn_request, |
1828 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
543d9cfe ACM |
1829 | .net_header_len = sizeof(struct iphdr), |
1830 | .setsockopt = ip_setsockopt, | |
1831 | .getsockopt = ip_getsockopt, | |
1832 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1833 | .sockaddr_len = sizeof(struct sockaddr_in), | |
3fdadf7d | 1834 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1835 | .compat_setsockopt = compat_ip_setsockopt, |
1836 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1837 | #endif |
4fab9071 | 1838 | .mtu_reduced = tcp_v4_mtu_reduced, |
1da177e4 | 1839 | }; |
4bc2f18b | 1840 | EXPORT_SYMBOL(ipv4_specific); |
1da177e4 | 1841 | |
cfb6eeb4 | 1842 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1843 | static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1844 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1845 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1846 | .md5_parse = tcp_v4_parse_md5_keys, |
cfb6eeb4 | 1847 | }; |
b6332e6c | 1848 | #endif |
cfb6eeb4 | 1849 | |
1da177e4 LT |
1850 | /* NOTE: A lot of things set to zero explicitly by call to |
1851 | * sk_alloc() so need not be done here. | |
1852 | */ | |
1853 | static int tcp_v4_init_sock(struct sock *sk) | |
1854 | { | |
6687e988 | 1855 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1856 | |
900f65d3 | 1857 | tcp_init_sock(sk); |
1da177e4 | 1858 | |
8292a17a | 1859 | icsk->icsk_af_ops = &ipv4_specific; |
900f65d3 | 1860 | |
cfb6eeb4 | 1861 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1862 | tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific; |
cfb6eeb4 | 1863 | #endif |
1da177e4 | 1864 | |
1da177e4 LT |
1865 | return 0; |
1866 | } | |
1867 | ||
7d06b2e0 | 1868 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1869 | { |
1870 | struct tcp_sock *tp = tcp_sk(sk); | |
1871 | ||
e1a4aa50 SL |
1872 | trace_tcp_destroy_sock(sk); |
1873 | ||
1da177e4 LT |
1874 | tcp_clear_xmit_timers(sk); |
1875 | ||
6687e988 | 1876 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1877 | |
734942cc DW |
1878 | tcp_cleanup_ulp(sk); |
1879 | ||
1da177e4 | 1880 | /* Cleanup up the write buffer. */ |
fe067e8a | 1881 | tcp_write_queue_purge(sk); |
1da177e4 | 1882 | |
cf1ef3f0 WW |
1883 | /* Check if we want to disable active TFO */ |
1884 | tcp_fastopen_active_disable_ofo_check(sk); | |
1885 | ||
1da177e4 | 1886 | /* Cleans up our, hopefully empty, out_of_order_queue. */ |
9f5afeae | 1887 | skb_rbtree_purge(&tp->out_of_order_queue); |
1da177e4 | 1888 | |
cfb6eeb4 YH |
1889 | #ifdef CONFIG_TCP_MD5SIG |
1890 | /* Clean up the MD5 key list, if any */ | |
1891 | if (tp->md5sig_info) { | |
a915da9b | 1892 | tcp_clear_md5_list(sk); |
a8afca03 | 1893 | kfree_rcu(tp->md5sig_info, rcu); |
cfb6eeb4 YH |
1894 | tp->md5sig_info = NULL; |
1895 | } | |
1896 | #endif | |
1a2449a8 | 1897 | |
1da177e4 | 1898 | /* Clean up a referenced TCP bind bucket. */ |
463c84b9 | 1899 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1900 | inet_put_port(sk); |
1da177e4 | 1901 | |
00db4124 | 1902 | BUG_ON(tp->fastopen_rsk); |
435cf559 | 1903 | |
cf60af03 YC |
1904 | /* If socket is aborted during connect operation */ |
1905 | tcp_free_fastopen_req(tp); | |
1fba70e5 | 1906 | tcp_fastopen_destroy_cipher(sk); |
cd8ae852 | 1907 | tcp_saved_syn_free(tp); |
cf60af03 | 1908 | |
180d8cd9 | 1909 | sk_sockets_allocated_dec(sk); |
1da177e4 | 1910 | } |
1da177e4 LT |
1911 | EXPORT_SYMBOL(tcp_v4_destroy_sock); |
1912 | ||
1913 | #ifdef CONFIG_PROC_FS | |
1914 | /* Proc filesystem TCP sock list dumping. */ | |
1915 | ||
a8b690f9 TH |
1916 | /* |
1917 | * Get next listener socket follow cur. If cur is NULL, get first socket | |
1918 | * starting from bucket given in st->bucket; when st->bucket is zero the | |
1919 | * very first socket in the hash table is returned. | |
1920 | */ | |
1da177e4 LT |
1921 | static void *listening_get_next(struct seq_file *seq, void *cur) |
1922 | { | |
5799de0b | 1923 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1924 | struct net *net = seq_file_net(seq); |
3b24d854 | 1925 | struct inet_listen_hashbucket *ilb; |
3b24d854 | 1926 | struct sock *sk = cur; |
1da177e4 LT |
1927 | |
1928 | if (!sk) { | |
3b24d854 | 1929 | get_head: |
a8b690f9 | 1930 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
9652dc2e | 1931 | spin_lock(&ilb->lock); |
3b24d854 | 1932 | sk = sk_head(&ilb->head); |
a8b690f9 | 1933 | st->offset = 0; |
1da177e4 LT |
1934 | goto get_sk; |
1935 | } | |
5caea4ea | 1936 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 | 1937 | ++st->num; |
a8b690f9 | 1938 | ++st->offset; |
1da177e4 | 1939 | |
3b24d854 | 1940 | sk = sk_next(sk); |
1da177e4 | 1941 | get_sk: |
3b24d854 | 1942 | sk_for_each_from(sk) { |
8475ef9f PE |
1943 | if (!net_eq(sock_net(sk), net)) |
1944 | continue; | |
3b24d854 ED |
1945 | if (sk->sk_family == st->family) |
1946 | return sk; | |
1da177e4 | 1947 | } |
9652dc2e | 1948 | spin_unlock(&ilb->lock); |
a8b690f9 | 1949 | st->offset = 0; |
3b24d854 ED |
1950 | if (++st->bucket < INET_LHTABLE_SIZE) |
1951 | goto get_head; | |
1952 | return NULL; | |
1da177e4 LT |
1953 | } |
1954 | ||
1955 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
1956 | { | |
a8b690f9 TH |
1957 | struct tcp_iter_state *st = seq->private; |
1958 | void *rc; | |
1959 | ||
1960 | st->bucket = 0; | |
1961 | st->offset = 0; | |
1962 | rc = listening_get_next(seq, NULL); | |
1da177e4 LT |
1963 | |
1964 | while (rc && *pos) { | |
1965 | rc = listening_get_next(seq, rc); | |
1966 | --*pos; | |
1967 | } | |
1968 | return rc; | |
1969 | } | |
1970 | ||
05dbc7b5 | 1971 | static inline bool empty_bucket(const struct tcp_iter_state *st) |
6eac5604 | 1972 | { |
05dbc7b5 | 1973 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain); |
6eac5604 AK |
1974 | } |
1975 | ||
a8b690f9 TH |
1976 | /* |
1977 | * Get first established socket starting from bucket given in st->bucket. | |
1978 | * If st->bucket is zero, the very first socket in the hash is returned. | |
1979 | */ | |
1da177e4 LT |
1980 | static void *established_get_first(struct seq_file *seq) |
1981 | { | |
5799de0b | 1982 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1983 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1984 | void *rc = NULL; |
1985 | ||
a8b690f9 TH |
1986 | st->offset = 0; |
1987 | for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) { | |
1da177e4 | 1988 | struct sock *sk; |
3ab5aee7 | 1989 | struct hlist_nulls_node *node; |
9db66bdc | 1990 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 1991 | |
6eac5604 AK |
1992 | /* Lockless fast path for the common case of empty buckets */ |
1993 | if (empty_bucket(st)) | |
1994 | continue; | |
1995 | ||
9db66bdc | 1996 | spin_lock_bh(lock); |
3ab5aee7 | 1997 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 1998 | if (sk->sk_family != st->family || |
878628fb | 1999 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
2000 | continue; |
2001 | } | |
2002 | rc = sk; | |
2003 | goto out; | |
2004 | } | |
9db66bdc | 2005 | spin_unlock_bh(lock); |
1da177e4 LT |
2006 | } |
2007 | out: | |
2008 | return rc; | |
2009 | } | |
2010 | ||
2011 | static void *established_get_next(struct seq_file *seq, void *cur) | |
2012 | { | |
2013 | struct sock *sk = cur; | |
3ab5aee7 | 2014 | struct hlist_nulls_node *node; |
5799de0b | 2015 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2016 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2017 | |
2018 | ++st->num; | |
a8b690f9 | 2019 | ++st->offset; |
1da177e4 | 2020 | |
05dbc7b5 | 2021 | sk = sk_nulls_next(sk); |
1da177e4 | 2022 | |
3ab5aee7 | 2023 | sk_nulls_for_each_from(sk, node) { |
878628fb | 2024 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
05dbc7b5 | 2025 | return sk; |
1da177e4 LT |
2026 | } |
2027 | ||
05dbc7b5 ED |
2028 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
2029 | ++st->bucket; | |
2030 | return established_get_first(seq); | |
1da177e4 LT |
2031 | } |
2032 | ||
2033 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2034 | { | |
a8b690f9 TH |
2035 | struct tcp_iter_state *st = seq->private; |
2036 | void *rc; | |
2037 | ||
2038 | st->bucket = 0; | |
2039 | rc = established_get_first(seq); | |
1da177e4 LT |
2040 | |
2041 | while (rc && pos) { | |
2042 | rc = established_get_next(seq, rc); | |
2043 | --pos; | |
7174259e | 2044 | } |
1da177e4 LT |
2045 | return rc; |
2046 | } | |
2047 | ||
2048 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2049 | { | |
2050 | void *rc; | |
5799de0b | 2051 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2052 | |
1da177e4 LT |
2053 | st->state = TCP_SEQ_STATE_LISTENING; |
2054 | rc = listening_get_idx(seq, &pos); | |
2055 | ||
2056 | if (!rc) { | |
1da177e4 LT |
2057 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2058 | rc = established_get_idx(seq, pos); | |
2059 | } | |
2060 | ||
2061 | return rc; | |
2062 | } | |
2063 | ||
a8b690f9 TH |
2064 | static void *tcp_seek_last_pos(struct seq_file *seq) |
2065 | { | |
2066 | struct tcp_iter_state *st = seq->private; | |
2067 | int offset = st->offset; | |
2068 | int orig_num = st->num; | |
2069 | void *rc = NULL; | |
2070 | ||
2071 | switch (st->state) { | |
a8b690f9 TH |
2072 | case TCP_SEQ_STATE_LISTENING: |
2073 | if (st->bucket >= INET_LHTABLE_SIZE) | |
2074 | break; | |
2075 | st->state = TCP_SEQ_STATE_LISTENING; | |
2076 | rc = listening_get_next(seq, NULL); | |
2077 | while (offset-- && rc) | |
2078 | rc = listening_get_next(seq, rc); | |
2079 | if (rc) | |
2080 | break; | |
2081 | st->bucket = 0; | |
05dbc7b5 | 2082 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2083 | /* Fallthrough */ |
2084 | case TCP_SEQ_STATE_ESTABLISHED: | |
a8b690f9 TH |
2085 | if (st->bucket > tcp_hashinfo.ehash_mask) |
2086 | break; | |
2087 | rc = established_get_first(seq); | |
2088 | while (offset-- && rc) | |
2089 | rc = established_get_next(seq, rc); | |
2090 | } | |
2091 | ||
2092 | st->num = orig_num; | |
2093 | ||
2094 | return rc; | |
2095 | } | |
2096 | ||
1da177e4 LT |
2097 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) |
2098 | { | |
5799de0b | 2099 | struct tcp_iter_state *st = seq->private; |
a8b690f9 TH |
2100 | void *rc; |
2101 | ||
2102 | if (*pos && *pos == st->last_pos) { | |
2103 | rc = tcp_seek_last_pos(seq); | |
2104 | if (rc) | |
2105 | goto out; | |
2106 | } | |
2107 | ||
1da177e4 LT |
2108 | st->state = TCP_SEQ_STATE_LISTENING; |
2109 | st->num = 0; | |
a8b690f9 TH |
2110 | st->bucket = 0; |
2111 | st->offset = 0; | |
2112 | rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2113 | ||
2114 | out: | |
2115 | st->last_pos = *pos; | |
2116 | return rc; | |
1da177e4 LT |
2117 | } |
2118 | ||
2119 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2120 | { | |
a8b690f9 | 2121 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2122 | void *rc = NULL; |
1da177e4 LT |
2123 | |
2124 | if (v == SEQ_START_TOKEN) { | |
2125 | rc = tcp_get_idx(seq, 0); | |
2126 | goto out; | |
2127 | } | |
1da177e4 LT |
2128 | |
2129 | switch (st->state) { | |
1da177e4 LT |
2130 | case TCP_SEQ_STATE_LISTENING: |
2131 | rc = listening_get_next(seq, v); | |
2132 | if (!rc) { | |
1da177e4 | 2133 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2134 | st->bucket = 0; |
2135 | st->offset = 0; | |
1da177e4 LT |
2136 | rc = established_get_first(seq); |
2137 | } | |
2138 | break; | |
2139 | case TCP_SEQ_STATE_ESTABLISHED: | |
1da177e4 LT |
2140 | rc = established_get_next(seq, v); |
2141 | break; | |
2142 | } | |
2143 | out: | |
2144 | ++*pos; | |
a8b690f9 | 2145 | st->last_pos = *pos; |
1da177e4 LT |
2146 | return rc; |
2147 | } | |
2148 | ||
2149 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2150 | { | |
5799de0b | 2151 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2152 | |
2153 | switch (st->state) { | |
1da177e4 LT |
2154 | case TCP_SEQ_STATE_LISTENING: |
2155 | if (v != SEQ_START_TOKEN) | |
9652dc2e | 2156 | spin_unlock(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 | 2157 | break; |
1da177e4 LT |
2158 | case TCP_SEQ_STATE_ESTABLISHED: |
2159 | if (v) | |
9db66bdc | 2160 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2161 | break; |
2162 | } | |
2163 | } | |
2164 | ||
73cb88ec | 2165 | int tcp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2166 | { |
d9dda78b | 2167 | struct tcp_seq_afinfo *afinfo = PDE_DATA(inode); |
1da177e4 | 2168 | struct tcp_iter_state *s; |
52d6f3f1 | 2169 | int err; |
1da177e4 | 2170 | |
52d6f3f1 DL |
2171 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2172 | sizeof(struct tcp_iter_state)); | |
2173 | if (err < 0) | |
2174 | return err; | |
f40c8174 | 2175 | |
52d6f3f1 | 2176 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2177 | s->family = afinfo->family; |
688d1945 | 2178 | s->last_pos = 0; |
f40c8174 DL |
2179 | return 0; |
2180 | } | |
73cb88ec | 2181 | EXPORT_SYMBOL(tcp_seq_open); |
f40c8174 | 2182 | |
6f8b13bc | 2183 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2184 | { |
2185 | int rc = 0; | |
2186 | struct proc_dir_entry *p; | |
2187 | ||
9427c4b3 DL |
2188 | afinfo->seq_ops.start = tcp_seq_start; |
2189 | afinfo->seq_ops.next = tcp_seq_next; | |
2190 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2191 | ||
84841c3c | 2192 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2193 | afinfo->seq_fops, afinfo); |
84841c3c | 2194 | if (!p) |
1da177e4 LT |
2195 | rc = -ENOMEM; |
2196 | return rc; | |
2197 | } | |
4bc2f18b | 2198 | EXPORT_SYMBOL(tcp_proc_register); |
1da177e4 | 2199 | |
6f8b13bc | 2200 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2201 | { |
ece31ffd | 2202 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2203 | } |
4bc2f18b | 2204 | EXPORT_SYMBOL(tcp_proc_unregister); |
1da177e4 | 2205 | |
d4f06873 | 2206 | static void get_openreq4(const struct request_sock *req, |
aa3a0c8c | 2207 | struct seq_file *f, int i) |
1da177e4 | 2208 | { |
2e6599cb | 2209 | const struct inet_request_sock *ireq = inet_rsk(req); |
fa76ce73 | 2210 | long delta = req->rsk_timer.expires - jiffies; |
1da177e4 | 2211 | |
5e659e4c | 2212 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2213 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK", |
1da177e4 | 2214 | i, |
634fb979 | 2215 | ireq->ir_loc_addr, |
d4f06873 | 2216 | ireq->ir_num, |
634fb979 ED |
2217 | ireq->ir_rmt_addr, |
2218 | ntohs(ireq->ir_rmt_port), | |
1da177e4 LT |
2219 | TCP_SYN_RECV, |
2220 | 0, 0, /* could print option size, but that is af dependent. */ | |
2221 | 1, /* timers active (only the expire timer) */ | |
a399a805 | 2222 | jiffies_delta_to_clock_t(delta), |
e6c022a4 | 2223 | req->num_timeout, |
aa3a0c8c ED |
2224 | from_kuid_munged(seq_user_ns(f), |
2225 | sock_i_uid(req->rsk_listener)), | |
1da177e4 LT |
2226 | 0, /* non standard timer */ |
2227 | 0, /* open_requests have no inode */ | |
d4f06873 | 2228 | 0, |
652586df | 2229 | req); |
1da177e4 LT |
2230 | } |
2231 | ||
652586df | 2232 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i) |
1da177e4 LT |
2233 | { |
2234 | int timer_active; | |
2235 | unsigned long timer_expires; | |
cf533ea5 | 2236 | const struct tcp_sock *tp = tcp_sk(sk); |
cf4c6bf8 | 2237 | const struct inet_connection_sock *icsk = inet_csk(sk); |
cf533ea5 | 2238 | const struct inet_sock *inet = inet_sk(sk); |
0536fcc0 | 2239 | const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq; |
c720c7e8 ED |
2240 | __be32 dest = inet->inet_daddr; |
2241 | __be32 src = inet->inet_rcv_saddr; | |
2242 | __u16 destp = ntohs(inet->inet_dport); | |
2243 | __u16 srcp = ntohs(inet->inet_sport); | |
49d09007 | 2244 | int rx_queue; |
00fd38d9 | 2245 | int state; |
1da177e4 | 2246 | |
6ba8a3b1 | 2247 | if (icsk->icsk_pending == ICSK_TIME_RETRANS || |
57dde7f7 | 2248 | icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || |
6ba8a3b1 | 2249 | icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { |
1da177e4 | 2250 | timer_active = 1; |
463c84b9 ACM |
2251 | timer_expires = icsk->icsk_timeout; |
2252 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2253 | timer_active = 4; |
463c84b9 | 2254 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2255 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2256 | timer_active = 2; |
cf4c6bf8 | 2257 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2258 | } else { |
2259 | timer_active = 0; | |
2260 | timer_expires = jiffies; | |
2261 | } | |
2262 | ||
00fd38d9 ED |
2263 | state = sk_state_load(sk); |
2264 | if (state == TCP_LISTEN) | |
49d09007 ED |
2265 | rx_queue = sk->sk_ack_backlog; |
2266 | else | |
00fd38d9 ED |
2267 | /* Because we don't lock the socket, |
2268 | * we might find a transient negative value. | |
49d09007 ED |
2269 | */ |
2270 | rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0); | |
2271 | ||
5e659e4c | 2272 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
652586df | 2273 | "%08X %5u %8d %lu %d %pK %lu %lu %u %u %d", |
00fd38d9 | 2274 | i, src, srcp, dest, destp, state, |
47da8ee6 | 2275 | tp->write_seq - tp->snd_una, |
49d09007 | 2276 | rx_queue, |
1da177e4 | 2277 | timer_active, |
a399a805 | 2278 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 2279 | icsk->icsk_retransmits, |
a7cb5a49 | 2280 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)), |
6687e988 | 2281 | icsk->icsk_probes_out, |
cf4c6bf8 | 2282 | sock_i_ino(sk), |
41c6d650 | 2283 | refcount_read(&sk->sk_refcnt), sk, |
7be87351 SH |
2284 | jiffies_to_clock_t(icsk->icsk_rto), |
2285 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2286 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2287 | tp->snd_cwnd, |
00fd38d9 ED |
2288 | state == TCP_LISTEN ? |
2289 | fastopenq->max_qlen : | |
652586df | 2290 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)); |
1da177e4 LT |
2291 | } |
2292 | ||
cf533ea5 | 2293 | static void get_timewait4_sock(const struct inet_timewait_sock *tw, |
652586df | 2294 | struct seq_file *f, int i) |
1da177e4 | 2295 | { |
789f558c | 2296 | long delta = tw->tw_timer.expires - jiffies; |
23f33c2d | 2297 | __be32 dest, src; |
1da177e4 | 2298 | __u16 destp, srcp; |
1da177e4 LT |
2299 | |
2300 | dest = tw->tw_daddr; | |
2301 | src = tw->tw_rcv_saddr; | |
2302 | destp = ntohs(tw->tw_dport); | |
2303 | srcp = ntohs(tw->tw_sport); | |
2304 | ||
5e659e4c | 2305 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2306 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK", |
1da177e4 | 2307 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
a399a805 | 2308 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
41c6d650 | 2309 | refcount_read(&tw->tw_refcnt), tw); |
1da177e4 LT |
2310 | } |
2311 | ||
2312 | #define TMPSZ 150 | |
2313 | ||
2314 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2315 | { | |
5799de0b | 2316 | struct tcp_iter_state *st; |
05dbc7b5 | 2317 | struct sock *sk = v; |
1da177e4 | 2318 | |
652586df | 2319 | seq_setwidth(seq, TMPSZ - 1); |
1da177e4 | 2320 | if (v == SEQ_START_TOKEN) { |
652586df | 2321 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
1da177e4 LT |
2322 | "rx_queue tr tm->when retrnsmt uid timeout " |
2323 | "inode"); | |
2324 | goto out; | |
2325 | } | |
2326 | st = seq->private; | |
2327 | ||
079096f1 ED |
2328 | if (sk->sk_state == TCP_TIME_WAIT) |
2329 | get_timewait4_sock(v, seq, st->num); | |
2330 | else if (sk->sk_state == TCP_NEW_SYN_RECV) | |
aa3a0c8c | 2331 | get_openreq4(v, seq, st->num); |
079096f1 ED |
2332 | else |
2333 | get_tcp4_sock(v, seq, st->num); | |
1da177e4 | 2334 | out: |
652586df | 2335 | seq_pad(seq, '\n'); |
1da177e4 LT |
2336 | return 0; |
2337 | } | |
2338 | ||
73cb88ec AV |
2339 | static const struct file_operations tcp_afinfo_seq_fops = { |
2340 | .owner = THIS_MODULE, | |
2341 | .open = tcp_seq_open, | |
2342 | .read = seq_read, | |
2343 | .llseek = seq_lseek, | |
2344 | .release = seq_release_net | |
2345 | }; | |
2346 | ||
1da177e4 | 2347 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2348 | .name = "tcp", |
2349 | .family = AF_INET, | |
73cb88ec | 2350 | .seq_fops = &tcp_afinfo_seq_fops, |
9427c4b3 DL |
2351 | .seq_ops = { |
2352 | .show = tcp4_seq_show, | |
2353 | }, | |
1da177e4 LT |
2354 | }; |
2355 | ||
2c8c1e72 | 2356 | static int __net_init tcp4_proc_init_net(struct net *net) |
757764f6 PE |
2357 | { |
2358 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2359 | } | |
2360 | ||
2c8c1e72 | 2361 | static void __net_exit tcp4_proc_exit_net(struct net *net) |
757764f6 PE |
2362 | { |
2363 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2364 | } | |
2365 | ||
2366 | static struct pernet_operations tcp4_net_ops = { | |
2367 | .init = tcp4_proc_init_net, | |
2368 | .exit = tcp4_proc_exit_net, | |
2369 | }; | |
2370 | ||
1da177e4 LT |
2371 | int __init tcp4_proc_init(void) |
2372 | { | |
757764f6 | 2373 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2374 | } |
2375 | ||
2376 | void tcp4_proc_exit(void) | |
2377 | { | |
757764f6 | 2378 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2379 | } |
2380 | #endif /* CONFIG_PROC_FS */ | |
2381 | ||
2382 | struct proto tcp_prot = { | |
2383 | .name = "TCP", | |
2384 | .owner = THIS_MODULE, | |
2385 | .close = tcp_close, | |
2386 | .connect = tcp_v4_connect, | |
2387 | .disconnect = tcp_disconnect, | |
463c84b9 | 2388 | .accept = inet_csk_accept, |
1da177e4 LT |
2389 | .ioctl = tcp_ioctl, |
2390 | .init = tcp_v4_init_sock, | |
2391 | .destroy = tcp_v4_destroy_sock, | |
2392 | .shutdown = tcp_shutdown, | |
2393 | .setsockopt = tcp_setsockopt, | |
2394 | .getsockopt = tcp_getsockopt, | |
4b9d07a4 | 2395 | .keepalive = tcp_set_keepalive, |
1da177e4 | 2396 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2397 | .sendmsg = tcp_sendmsg, |
2398 | .sendpage = tcp_sendpage, | |
1da177e4 | 2399 | .backlog_rcv = tcp_v4_do_rcv, |
46d3ceab | 2400 | .release_cb = tcp_release_cb, |
ab1e0a13 ACM |
2401 | .hash = inet_hash, |
2402 | .unhash = inet_unhash, | |
2403 | .get_port = inet_csk_get_port, | |
1da177e4 | 2404 | .enter_memory_pressure = tcp_enter_memory_pressure, |
06044751 | 2405 | .leave_memory_pressure = tcp_leave_memory_pressure, |
c9bee3b7 | 2406 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 | 2407 | .sockets_allocated = &tcp_sockets_allocated, |
0a5578cf | 2408 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2409 | .memory_allocated = &tcp_memory_allocated, |
2410 | .memory_pressure = &tcp_memory_pressure, | |
a4fe34bf | 2411 | .sysctl_mem = sysctl_tcp_mem, |
1da177e4 LT |
2412 | .sysctl_wmem = sysctl_tcp_wmem, |
2413 | .sysctl_rmem = sysctl_tcp_rmem, | |
2414 | .max_header = MAX_TCP_HEADER, | |
2415 | .obj_size = sizeof(struct tcp_sock), | |
5f0d5a3a | 2416 | .slab_flags = SLAB_TYPESAFE_BY_RCU, |
6d6ee43e | 2417 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2418 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2419 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2420 | .no_autobind = true, |
543d9cfe ACM |
2421 | #ifdef CONFIG_COMPAT |
2422 | .compat_setsockopt = compat_tcp_setsockopt, | |
2423 | .compat_getsockopt = compat_tcp_getsockopt, | |
d1a4c0b3 | 2424 | #endif |
c1e64e29 | 2425 | .diag_destroy = tcp_abort, |
1da177e4 | 2426 | }; |
4bc2f18b | 2427 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 | 2428 | |
bdbbb852 ED |
2429 | static void __net_exit tcp_sk_exit(struct net *net) |
2430 | { | |
2431 | int cpu; | |
2432 | ||
2433 | for_each_possible_cpu(cpu) | |
2434 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.tcp_sk, cpu)); | |
2435 | free_percpu(net->ipv4.tcp_sk); | |
2436 | } | |
2437 | ||
046ee902 DL |
2438 | static int __net_init tcp_sk_init(struct net *net) |
2439 | { | |
fee83d09 | 2440 | int res, cpu, cnt; |
bdbbb852 ED |
2441 | |
2442 | net->ipv4.tcp_sk = alloc_percpu(struct sock *); | |
2443 | if (!net->ipv4.tcp_sk) | |
2444 | return -ENOMEM; | |
2445 | ||
2446 | for_each_possible_cpu(cpu) { | |
2447 | struct sock *sk; | |
2448 | ||
2449 | res = inet_ctl_sock_create(&sk, PF_INET, SOCK_RAW, | |
2450 | IPPROTO_TCP, net); | |
2451 | if (res) | |
2452 | goto fail; | |
a9d6532b | 2453 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); |
bdbbb852 ED |
2454 | *per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk; |
2455 | } | |
49213555 | 2456 | |
5d134f1c | 2457 | net->ipv4.sysctl_tcp_ecn = 2; |
49213555 DB |
2458 | net->ipv4.sysctl_tcp_ecn_fallback = 1; |
2459 | ||
b0f9ca53 | 2460 | net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS; |
6b58e0a5 | 2461 | net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD; |
05cbc0db | 2462 | net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL; |
046ee902 | 2463 | |
13b287e8 | 2464 | net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME; |
9bd6861b | 2465 | net->ipv4.sysctl_tcp_keepalive_probes = TCP_KEEPALIVE_PROBES; |
b840d15d | 2466 | net->ipv4.sysctl_tcp_keepalive_intvl = TCP_KEEPALIVE_INTVL; |
13b287e8 | 2467 | |
6fa25166 | 2468 | net->ipv4.sysctl_tcp_syn_retries = TCP_SYN_RETRIES; |
7c083ecb | 2469 | net->ipv4.sysctl_tcp_synack_retries = TCP_SYNACK_RETRIES; |
0aca737d | 2470 | net->ipv4.sysctl_tcp_syncookies = 1; |
1043e25f | 2471 | net->ipv4.sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH; |
ae5c3f40 | 2472 | net->ipv4.sysctl_tcp_retries1 = TCP_RETR1; |
c6214a97 | 2473 | net->ipv4.sysctl_tcp_retries2 = TCP_RETR2; |
c402d9be | 2474 | net->ipv4.sysctl_tcp_orphan_retries = 0; |
1e579caa | 2475 | net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT; |
4979f2d9 | 2476 | net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX; |
56ab6b93 | 2477 | net->ipv4.sysctl_tcp_tw_reuse = 0; |
12ed8244 | 2478 | |
fee83d09 | 2479 | cnt = tcp_hashinfo.ehash_mask + 1; |
fee83d09 | 2480 | net->ipv4.tcp_death_row.sysctl_max_tw_buckets = (cnt + 1) / 2; |
1946e672 HY |
2481 | net->ipv4.tcp_death_row.hashinfo = &tcp_hashinfo; |
2482 | ||
fee83d09 | 2483 | net->ipv4.sysctl_max_syn_backlog = max(128, cnt / 256); |
f9301034 | 2484 | net->ipv4.sysctl_tcp_sack = 1; |
9bb37ef0 | 2485 | net->ipv4.sysctl_tcp_window_scaling = 1; |
5d2ed052 | 2486 | net->ipv4.sysctl_tcp_timestamps = 1; |
2ae21cf5 | 2487 | net->ipv4.sysctl_tcp_early_retrans = 3; |
e20223f1 | 2488 | net->ipv4.sysctl_tcp_recovery = TCP_RACK_LOSS_DETECTION; |
fee83d09 | 2489 | |
e1cfcbe8 | 2490 | net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE; |
43713848 | 2491 | spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock); |
3733be14 HY |
2492 | net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 60 * 60; |
2493 | atomic_set(&net->ipv4.tfo_active_disable_times, 0); | |
e1cfcbe8 | 2494 | |
49213555 | 2495 | return 0; |
bdbbb852 ED |
2496 | fail: |
2497 | tcp_sk_exit(net); | |
2498 | ||
2499 | return res; | |
b099ce26 EB |
2500 | } |
2501 | ||
2502 | static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list) | |
2503 | { | |
43713848 HY |
2504 | struct net *net; |
2505 | ||
1946e672 | 2506 | inet_twsk_purge(&tcp_hashinfo, AF_INET); |
43713848 HY |
2507 | |
2508 | list_for_each_entry(net, net_exit_list, exit_list) | |
2509 | tcp_fastopen_ctx_destroy(net); | |
046ee902 DL |
2510 | } |
2511 | ||
2512 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
b099ce26 EB |
2513 | .init = tcp_sk_init, |
2514 | .exit = tcp_sk_exit, | |
2515 | .exit_batch = tcp_sk_exit_batch, | |
046ee902 DL |
2516 | }; |
2517 | ||
9b0f976f | 2518 | void __init tcp_v4_init(void) |
1da177e4 | 2519 | { |
6a1b3054 | 2520 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2521 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 | 2522 | } |