]>
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> | |
83 | ||
cf80e0e4 | 84 | #include <crypto/hash.h> |
cfb6eeb4 YH |
85 | #include <linux/scatterlist.h> |
86 | ||
ab32ea5d | 87 | int sysctl_tcp_low_latency __read_mostly; |
1da177e4 | 88 | |
cfb6eeb4 | 89 | #ifdef CONFIG_TCP_MD5SIG |
a915da9b | 90 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 91 | __be32 daddr, __be32 saddr, const struct tcphdr *th); |
cfb6eeb4 YH |
92 | #endif |
93 | ||
5caea4ea | 94 | struct inet_hashinfo tcp_hashinfo; |
4bc2f18b | 95 | EXPORT_SYMBOL(tcp_hashinfo); |
1da177e4 | 96 | |
95a22cae | 97 | static u32 tcp_v4_init_sequence(const struct sk_buff *skb, u32 *tsoff) |
1da177e4 | 98 | { |
eddc9ec5 ACM |
99 | return secure_tcp_sequence_number(ip_hdr(skb)->daddr, |
100 | ip_hdr(skb)->saddr, | |
aa8223c7 | 101 | tcp_hdr(skb)->dest, |
95a22cae | 102 | tcp_hdr(skb)->source, tsoff); |
1da177e4 LT |
103 | } |
104 | ||
6d6ee43e ACM |
105 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp) |
106 | { | |
107 | const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw); | |
108 | struct tcp_sock *tp = tcp_sk(sk); | |
109 | ||
110 | /* With PAWS, it is safe from the viewpoint | |
111 | of data integrity. Even without PAWS it is safe provided sequence | |
112 | spaces do not overlap i.e. at data rates <= 80Mbit/sec. | |
113 | ||
114 | Actually, the idea is close to VJ's one, only timestamp cache is | |
115 | held not per host, but per port pair and TW bucket is used as state | |
116 | holder. | |
117 | ||
118 | If TW bucket has been already destroyed we fall back to VJ's scheme | |
119 | and use initial timestamp retrieved from peer table. | |
120 | */ | |
121 | if (tcptw->tw_ts_recent_stamp && | |
56ab6b93 | 122 | (!twp || (sock_net(sk)->ipv4.sysctl_tcp_tw_reuse && |
9d729f72 | 123 | get_seconds() - tcptw->tw_ts_recent_stamp > 1))) { |
6d6ee43e ACM |
124 | tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2; |
125 | if (tp->write_seq == 0) | |
126 | tp->write_seq = 1; | |
127 | tp->rx_opt.ts_recent = tcptw->tw_ts_recent; | |
128 | tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp; | |
129 | sock_hold(sktw); | |
130 | return 1; | |
131 | } | |
132 | ||
133 | return 0; | |
134 | } | |
6d6ee43e ACM |
135 | EXPORT_SYMBOL_GPL(tcp_twsk_unique); |
136 | ||
1da177e4 LT |
137 | /* This will initiate an outgoing connection. */ |
138 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
139 | { | |
2d7192d6 | 140 | struct sockaddr_in *usin = (struct sockaddr_in *)uaddr; |
1da177e4 LT |
141 | struct inet_sock *inet = inet_sk(sk); |
142 | struct tcp_sock *tp = tcp_sk(sk); | |
dca8b089 | 143 | __be16 orig_sport, orig_dport; |
bada8adc | 144 | __be32 daddr, nexthop; |
da905bd1 | 145 | struct flowi4 *fl4; |
2d7192d6 | 146 | struct rtable *rt; |
1da177e4 | 147 | int err; |
f6d8bd05 | 148 | struct ip_options_rcu *inet_opt; |
1946e672 | 149 | struct inet_timewait_death_row *tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row; |
1da177e4 LT |
150 | |
151 | if (addr_len < sizeof(struct sockaddr_in)) | |
152 | return -EINVAL; | |
153 | ||
154 | if (usin->sin_family != AF_INET) | |
155 | return -EAFNOSUPPORT; | |
156 | ||
157 | nexthop = daddr = usin->sin_addr.s_addr; | |
f6d8bd05 | 158 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1e1d04e6 | 159 | lockdep_sock_is_held(sk)); |
f6d8bd05 | 160 | if (inet_opt && inet_opt->opt.srr) { |
1da177e4 LT |
161 | if (!daddr) |
162 | return -EINVAL; | |
f6d8bd05 | 163 | nexthop = inet_opt->opt.faddr; |
1da177e4 LT |
164 | } |
165 | ||
dca8b089 DM |
166 | orig_sport = inet->inet_sport; |
167 | orig_dport = usin->sin_port; | |
da905bd1 DM |
168 | fl4 = &inet->cork.fl.u.ip4; |
169 | rt = ip_route_connect(fl4, nexthop, inet->inet_saddr, | |
b23dd4fe DM |
170 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if, |
171 | IPPROTO_TCP, | |
0e0d44ab | 172 | orig_sport, orig_dport, sk); |
b23dd4fe DM |
173 | if (IS_ERR(rt)) { |
174 | err = PTR_ERR(rt); | |
175 | if (err == -ENETUNREACH) | |
f1d8cba6 | 176 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); |
b23dd4fe | 177 | return err; |
584bdf8c | 178 | } |
1da177e4 LT |
179 | |
180 | if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) { | |
181 | ip_rt_put(rt); | |
182 | return -ENETUNREACH; | |
183 | } | |
184 | ||
f6d8bd05 | 185 | if (!inet_opt || !inet_opt->opt.srr) |
da905bd1 | 186 | daddr = fl4->daddr; |
1da177e4 | 187 | |
c720c7e8 | 188 | if (!inet->inet_saddr) |
da905bd1 | 189 | inet->inet_saddr = fl4->saddr; |
d1e559d0 | 190 | sk_rcv_saddr_set(sk, inet->inet_saddr); |
1da177e4 | 191 | |
c720c7e8 | 192 | if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) { |
1da177e4 LT |
193 | /* Reset inherited state */ |
194 | tp->rx_opt.ts_recent = 0; | |
195 | tp->rx_opt.ts_recent_stamp = 0; | |
ee995283 PE |
196 | if (likely(!tp->repair)) |
197 | tp->write_seq = 0; | |
1da177e4 LT |
198 | } |
199 | ||
1946e672 | 200 | if (tcp_death_row->sysctl_tw_recycle && |
81166dd6 DM |
201 | !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr) |
202 | tcp_fetch_timewait_stamp(sk, &rt->dst); | |
1da177e4 | 203 | |
c720c7e8 | 204 | inet->inet_dport = usin->sin_port; |
d1e559d0 | 205 | sk_daddr_set(sk, daddr); |
1da177e4 | 206 | |
d83d8461 | 207 | inet_csk(sk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
208 | if (inet_opt) |
209 | inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
1da177e4 | 210 | |
bee7ca9e | 211 | tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT; |
1da177e4 LT |
212 | |
213 | /* Socket identity is still unknown (sport may be zero). | |
214 | * However we set state to SYN-SENT and not releasing socket | |
215 | * lock select source port, enter ourselves into the hash tables and | |
216 | * complete initialization after this. | |
217 | */ | |
218 | tcp_set_state(sk, TCP_SYN_SENT); | |
1946e672 | 219 | err = inet_hash_connect(tcp_death_row, sk); |
1da177e4 LT |
220 | if (err) |
221 | goto failure; | |
222 | ||
877d1f62 | 223 | sk_set_txhash(sk); |
9e7ceb06 | 224 | |
da905bd1 | 225 | rt = ip_route_newports(fl4, rt, orig_sport, orig_dport, |
b23dd4fe DM |
226 | inet->inet_sport, inet->inet_dport, sk); |
227 | if (IS_ERR(rt)) { | |
228 | err = PTR_ERR(rt); | |
229 | rt = NULL; | |
1da177e4 | 230 | goto failure; |
b23dd4fe | 231 | } |
1da177e4 | 232 | /* OK, now commit destination to socket. */ |
bcd76111 | 233 | sk->sk_gso_type = SKB_GSO_TCPV4; |
d8d1f30b | 234 | sk_setup_caps(sk, &rt->dst); |
1da177e4 | 235 | |
ee995283 | 236 | if (!tp->write_seq && likely(!tp->repair)) |
c720c7e8 ED |
237 | tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr, |
238 | inet->inet_daddr, | |
239 | inet->inet_sport, | |
95a22cae FW |
240 | usin->sin_port, |
241 | &tp->tsoffset); | |
1da177e4 | 242 | |
c720c7e8 | 243 | inet->inet_id = tp->write_seq ^ jiffies; |
1da177e4 | 244 | |
2b916477 | 245 | err = tcp_connect(sk); |
ee995283 | 246 | |
1da177e4 LT |
247 | rt = NULL; |
248 | if (err) | |
249 | goto failure; | |
250 | ||
251 | return 0; | |
252 | ||
253 | failure: | |
7174259e ACM |
254 | /* |
255 | * This unhashes the socket and releases the local port, | |
256 | * if necessary. | |
257 | */ | |
1da177e4 LT |
258 | tcp_set_state(sk, TCP_CLOSE); |
259 | ip_rt_put(rt); | |
260 | sk->sk_route_caps = 0; | |
c720c7e8 | 261 | inet->inet_dport = 0; |
1da177e4 LT |
262 | return err; |
263 | } | |
4bc2f18b | 264 | EXPORT_SYMBOL(tcp_v4_connect); |
1da177e4 | 265 | |
1da177e4 | 266 | /* |
563d34d0 ED |
267 | * This routine reacts to ICMP_FRAG_NEEDED mtu indications as defined in RFC1191. |
268 | * It can be called through tcp_release_cb() if socket was owned by user | |
269 | * at the time tcp_v4_err() was called to handle ICMP message. | |
1da177e4 | 270 | */ |
4fab9071 | 271 | void tcp_v4_mtu_reduced(struct sock *sk) |
1da177e4 LT |
272 | { |
273 | struct dst_entry *dst; | |
274 | struct inet_sock *inet = inet_sk(sk); | |
563d34d0 | 275 | u32 mtu = tcp_sk(sk)->mtu_info; |
1da177e4 | 276 | |
80d0a69f DM |
277 | dst = inet_csk_update_pmtu(sk, mtu); |
278 | if (!dst) | |
1da177e4 LT |
279 | return; |
280 | ||
1da177e4 LT |
281 | /* Something is about to be wrong... Remember soft error |
282 | * for the case, if this connection will not able to recover. | |
283 | */ | |
284 | if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst)) | |
285 | sk->sk_err_soft = EMSGSIZE; | |
286 | ||
287 | mtu = dst_mtu(dst); | |
288 | ||
289 | if (inet->pmtudisc != IP_PMTUDISC_DONT && | |
482fc609 | 290 | ip_sk_accept_pmtu(sk) && |
d83d8461 | 291 | inet_csk(sk)->icsk_pmtu_cookie > mtu) { |
1da177e4 LT |
292 | tcp_sync_mss(sk, mtu); |
293 | ||
294 | /* Resend the TCP packet because it's | |
295 | * clear that the old packet has been | |
296 | * dropped. This is the new "fast" path mtu | |
297 | * discovery. | |
298 | */ | |
299 | tcp_simple_retransmit(sk); | |
300 | } /* else let the usual retransmit timer handle it */ | |
301 | } | |
4fab9071 | 302 | EXPORT_SYMBOL(tcp_v4_mtu_reduced); |
1da177e4 | 303 | |
55be7a9c DM |
304 | static void do_redirect(struct sk_buff *skb, struct sock *sk) |
305 | { | |
306 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
307 | ||
1ed5c48f | 308 | if (dst) |
6700c270 | 309 | dst->ops->redirect(dst, sk, skb); |
55be7a9c DM |
310 | } |
311 | ||
26e37360 ED |
312 | |
313 | /* handle ICMP messages on TCP_NEW_SYN_RECV request sockets */ | |
9cf74903 | 314 | void tcp_req_err(struct sock *sk, u32 seq, bool abort) |
26e37360 ED |
315 | { |
316 | struct request_sock *req = inet_reqsk(sk); | |
317 | struct net *net = sock_net(sk); | |
318 | ||
319 | /* ICMPs are not backlogged, hence we cannot get | |
320 | * an established socket here. | |
321 | */ | |
26e37360 | 322 | if (seq != tcp_rsk(req)->snt_isn) { |
02a1d6e7 | 323 | __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS); |
9cf74903 | 324 | } else if (abort) { |
26e37360 ED |
325 | /* |
326 | * Still in SYN_RECV, just remove it silently. | |
327 | * There is no good way to pass the error to the newly | |
328 | * created socket, and POSIX does not want network | |
329 | * errors returned from accept(). | |
330 | */ | |
c6973669 | 331 | inet_csk_reqsk_queue_drop(req->rsk_listener, req); |
9caad864 | 332 | tcp_listendrop(req->rsk_listener); |
26e37360 | 333 | } |
ef84d8ce | 334 | reqsk_put(req); |
26e37360 ED |
335 | } |
336 | EXPORT_SYMBOL(tcp_req_err); | |
337 | ||
1da177e4 LT |
338 | /* |
339 | * This routine is called by the ICMP module when it gets some | |
340 | * sort of error condition. If err < 0 then the socket should | |
341 | * be closed and the error returned to the user. If err > 0 | |
342 | * it's just the icmp type << 8 | icmp code. After adjustment | |
343 | * header points to the first 8 bytes of the tcp header. We need | |
344 | * to find the appropriate port. | |
345 | * | |
346 | * The locking strategy used here is very "optimistic". When | |
347 | * someone else accesses the socket the ICMP is just dropped | |
348 | * and for some paths there is no check at all. | |
349 | * A more general error queue to queue errors for later handling | |
350 | * is probably better. | |
351 | * | |
352 | */ | |
353 | ||
4d1a2d9e | 354 | void tcp_v4_err(struct sk_buff *icmp_skb, u32 info) |
1da177e4 | 355 | { |
b71d1d42 | 356 | const struct iphdr *iph = (const struct iphdr *)icmp_skb->data; |
4d1a2d9e | 357 | struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2)); |
f1ecd5d9 | 358 | struct inet_connection_sock *icsk; |
1da177e4 LT |
359 | struct tcp_sock *tp; |
360 | struct inet_sock *inet; | |
4d1a2d9e DL |
361 | const int type = icmp_hdr(icmp_skb)->type; |
362 | const int code = icmp_hdr(icmp_skb)->code; | |
1da177e4 | 363 | struct sock *sk; |
f1ecd5d9 | 364 | struct sk_buff *skb; |
0a672f74 YC |
365 | struct request_sock *fastopen; |
366 | __u32 seq, snd_una; | |
f1ecd5d9 | 367 | __u32 remaining; |
1da177e4 | 368 | int err; |
4d1a2d9e | 369 | struct net *net = dev_net(icmp_skb->dev); |
1da177e4 | 370 | |
26e37360 ED |
371 | sk = __inet_lookup_established(net, &tcp_hashinfo, iph->daddr, |
372 | th->dest, iph->saddr, ntohs(th->source), | |
373 | inet_iif(icmp_skb)); | |
1da177e4 | 374 | if (!sk) { |
5d3848bc | 375 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
376 | return; |
377 | } | |
378 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 379 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
380 | return; |
381 | } | |
26e37360 ED |
382 | seq = ntohl(th->seq); |
383 | if (sk->sk_state == TCP_NEW_SYN_RECV) | |
9cf74903 ED |
384 | return tcp_req_err(sk, seq, |
385 | type == ICMP_PARAMETERPROB || | |
386 | type == ICMP_TIME_EXCEEDED || | |
387 | (type == ICMP_DEST_UNREACH && | |
388 | (code == ICMP_NET_UNREACH || | |
389 | code == ICMP_HOST_UNREACH))); | |
1da177e4 LT |
390 | |
391 | bh_lock_sock(sk); | |
392 | /* If too many ICMPs get dropped on busy | |
393 | * servers this needs to be solved differently. | |
563d34d0 ED |
394 | * We do take care of PMTU discovery (RFC1191) special case : |
395 | * we can receive locally generated ICMP messages while socket is held. | |
1da177e4 | 396 | */ |
b74aa930 ED |
397 | if (sock_owned_by_user(sk)) { |
398 | if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)) | |
02a1d6e7 | 399 | __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS); |
b74aa930 | 400 | } |
1da177e4 LT |
401 | if (sk->sk_state == TCP_CLOSE) |
402 | goto out; | |
403 | ||
97e3ecd1 | 404 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
02a1d6e7 | 405 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); |
97e3ecd1 | 406 | goto out; |
407 | } | |
408 | ||
f1ecd5d9 | 409 | icsk = inet_csk(sk); |
1da177e4 | 410 | tp = tcp_sk(sk); |
0a672f74 YC |
411 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ |
412 | fastopen = tp->fastopen_rsk; | |
413 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; | |
1da177e4 | 414 | if (sk->sk_state != TCP_LISTEN && |
0a672f74 | 415 | !between(seq, snd_una, tp->snd_nxt)) { |
02a1d6e7 | 416 | __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
417 | goto out; |
418 | } | |
419 | ||
420 | switch (type) { | |
55be7a9c DM |
421 | case ICMP_REDIRECT: |
422 | do_redirect(icmp_skb, sk); | |
423 | goto out; | |
1da177e4 LT |
424 | case ICMP_SOURCE_QUENCH: |
425 | /* Just silently ignore these. */ | |
426 | goto out; | |
427 | case ICMP_PARAMETERPROB: | |
428 | err = EPROTO; | |
429 | break; | |
430 | case ICMP_DEST_UNREACH: | |
431 | if (code > NR_ICMP_UNREACH) | |
432 | goto out; | |
433 | ||
434 | if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */ | |
0d4f0608 ED |
435 | /* We are not interested in TCP_LISTEN and open_requests |
436 | * (SYN-ACKs send out by Linux are always <576bytes so | |
437 | * they should go through unfragmented). | |
438 | */ | |
439 | if (sk->sk_state == TCP_LISTEN) | |
440 | goto out; | |
441 | ||
563d34d0 | 442 | tp->mtu_info = info; |
144d56e9 | 443 | if (!sock_owned_by_user(sk)) { |
563d34d0 | 444 | tcp_v4_mtu_reduced(sk); |
144d56e9 | 445 | } else { |
7aa5470c | 446 | if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &sk->sk_tsq_flags)) |
144d56e9 ED |
447 | sock_hold(sk); |
448 | } | |
1da177e4 LT |
449 | goto out; |
450 | } | |
451 | ||
452 | err = icmp_err_convert[code].errno; | |
f1ecd5d9 DL |
453 | /* check if icmp_skb allows revert of backoff |
454 | * (see draft-zimmermann-tcp-lcd) */ | |
455 | if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH) | |
456 | break; | |
457 | if (seq != tp->snd_una || !icsk->icsk_retransmits || | |
0a672f74 | 458 | !icsk->icsk_backoff || fastopen) |
f1ecd5d9 DL |
459 | break; |
460 | ||
8f49c270 DM |
461 | if (sock_owned_by_user(sk)) |
462 | break; | |
463 | ||
f1ecd5d9 | 464 | icsk->icsk_backoff--; |
fcdd1cf4 ED |
465 | icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) : |
466 | TCP_TIMEOUT_INIT; | |
467 | icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX); | |
f1ecd5d9 DL |
468 | |
469 | skb = tcp_write_queue_head(sk); | |
470 | BUG_ON(!skb); | |
471 | ||
7faee5c0 ED |
472 | remaining = icsk->icsk_rto - |
473 | min(icsk->icsk_rto, | |
474 | tcp_time_stamp - tcp_skb_timestamp(skb)); | |
f1ecd5d9 DL |
475 | |
476 | if (remaining) { | |
477 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, | |
478 | remaining, TCP_RTO_MAX); | |
f1ecd5d9 DL |
479 | } else { |
480 | /* RTO revert clocked out retransmission. | |
481 | * Will retransmit now */ | |
482 | tcp_retransmit_timer(sk); | |
483 | } | |
484 | ||
1da177e4 LT |
485 | break; |
486 | case ICMP_TIME_EXCEEDED: | |
487 | err = EHOSTUNREACH; | |
488 | break; | |
489 | default: | |
490 | goto out; | |
491 | } | |
492 | ||
493 | switch (sk->sk_state) { | |
1da177e4 | 494 | case TCP_SYN_SENT: |
0a672f74 YC |
495 | case TCP_SYN_RECV: |
496 | /* Only in fast or simultaneous open. If a fast open socket is | |
497 | * is already accepted it is treated as a connected one below. | |
498 | */ | |
51456b29 | 499 | if (fastopen && !fastopen->sk) |
0a672f74 YC |
500 | break; |
501 | ||
1da177e4 | 502 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
503 | sk->sk_err = err; |
504 | ||
505 | sk->sk_error_report(sk); | |
506 | ||
507 | tcp_done(sk); | |
508 | } else { | |
509 | sk->sk_err_soft = err; | |
510 | } | |
511 | goto out; | |
512 | } | |
513 | ||
514 | /* If we've already connected we will keep trying | |
515 | * until we time out, or the user gives up. | |
516 | * | |
517 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
518 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
519 | * but it is obsoleted by pmtu discovery). | |
520 | * | |
521 | * Note, that in modern internet, where routing is unreliable | |
522 | * and in each dark corner broken firewalls sit, sending random | |
523 | * errors ordered by their masters even this two messages finally lose | |
524 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
525 | * | |
526 | * Now we are in compliance with RFCs. | |
527 | * --ANK (980905) | |
528 | */ | |
529 | ||
530 | inet = inet_sk(sk); | |
531 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
532 | sk->sk_err = err; | |
533 | sk->sk_error_report(sk); | |
534 | } else { /* Only an error on timeout */ | |
535 | sk->sk_err_soft = err; | |
536 | } | |
537 | ||
538 | out: | |
539 | bh_unlock_sock(sk); | |
540 | sock_put(sk); | |
541 | } | |
542 | ||
28850dc7 | 543 | void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 544 | { |
aa8223c7 | 545 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 546 | |
84fa7933 | 547 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
419f9f89 | 548 | th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0); |
663ead3b | 549 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 550 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 551 | } else { |
419f9f89 | 552 | th->check = tcp_v4_check(skb->len, saddr, daddr, |
07f0757a | 553 | csum_partial(th, |
1da177e4 LT |
554 | th->doff << 2, |
555 | skb->csum)); | |
556 | } | |
557 | } | |
558 | ||
419f9f89 | 559 | /* This routine computes an IPv4 TCP checksum. */ |
bb296246 | 560 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb) |
419f9f89 | 561 | { |
cf533ea5 | 562 | const struct inet_sock *inet = inet_sk(sk); |
419f9f89 HX |
563 | |
564 | __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr); | |
565 | } | |
4bc2f18b | 566 | EXPORT_SYMBOL(tcp_v4_send_check); |
419f9f89 | 567 | |
1da177e4 LT |
568 | /* |
569 | * This routine will send an RST to the other tcp. | |
570 | * | |
571 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
572 | * for reset. | |
573 | * Answer: if a packet caused RST, it is not for a socket | |
574 | * existing in our system, if it is matched to a socket, | |
575 | * it is just duplicate segment or bug in other side's TCP. | |
576 | * So that we build reply only basing on parameters | |
577 | * arrived with segment. | |
578 | * Exception: precedence violation. We do not implement it in any case. | |
579 | */ | |
580 | ||
a00e7444 | 581 | static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) |
1da177e4 | 582 | { |
cf533ea5 | 583 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
584 | struct { |
585 | struct tcphdr th; | |
586 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 587 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
588 | #endif |
589 | } rep; | |
1da177e4 | 590 | struct ip_reply_arg arg; |
cfb6eeb4 | 591 | #ifdef CONFIG_TCP_MD5SIG |
e46787f0 | 592 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
593 | const __u8 *hash_location = NULL; |
594 | unsigned char newhash[16]; | |
595 | int genhash; | |
596 | struct sock *sk1 = NULL; | |
cfb6eeb4 | 597 | #endif |
a86b1e30 | 598 | struct net *net; |
1da177e4 LT |
599 | |
600 | /* Never send a reset in response to a reset. */ | |
601 | if (th->rst) | |
602 | return; | |
603 | ||
c3658e8d ED |
604 | /* If sk not NULL, it means we did a successful lookup and incoming |
605 | * route had to be correct. prequeue might have dropped our dst. | |
606 | */ | |
607 | if (!sk && skb_rtable(skb)->rt_type != RTN_LOCAL) | |
1da177e4 LT |
608 | return; |
609 | ||
610 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
611 | memset(&rep, 0, sizeof(rep)); |
612 | rep.th.dest = th->source; | |
613 | rep.th.source = th->dest; | |
614 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
615 | rep.th.rst = 1; | |
1da177e4 LT |
616 | |
617 | if (th->ack) { | |
cfb6eeb4 | 618 | rep.th.seq = th->ack_seq; |
1da177e4 | 619 | } else { |
cfb6eeb4 YH |
620 | rep.th.ack = 1; |
621 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
622 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
623 | } |
624 | ||
7174259e | 625 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
626 | arg.iov[0].iov_base = (unsigned char *)&rep; |
627 | arg.iov[0].iov_len = sizeof(rep.th); | |
628 | ||
0f85feae | 629 | net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); |
cfb6eeb4 | 630 | #ifdef CONFIG_TCP_MD5SIG |
3b24d854 | 631 | rcu_read_lock(); |
658ddaaf | 632 | hash_location = tcp_parse_md5sig_option(th); |
271c3b9b | 633 | if (sk && sk_fullsock(sk)) { |
e46787f0 FW |
634 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *) |
635 | &ip_hdr(skb)->saddr, AF_INET); | |
636 | } else if (hash_location) { | |
658ddaaf SL |
637 | /* |
638 | * active side is lost. Try to find listening socket through | |
639 | * source port, and then find md5 key through listening socket. | |
640 | * we are not loose security here: | |
641 | * Incoming packet is checked with md5 hash with finding key, | |
642 | * no RST generated if md5 hash doesn't match. | |
643 | */ | |
a583636a CG |
644 | sk1 = __inet_lookup_listener(net, &tcp_hashinfo, NULL, 0, |
645 | ip_hdr(skb)->saddr, | |
da5e3630 | 646 | th->source, ip_hdr(skb)->daddr, |
658ddaaf SL |
647 | ntohs(th->source), inet_iif(skb)); |
648 | /* don't send rst if it can't find key */ | |
649 | if (!sk1) | |
3b24d854 ED |
650 | goto out; |
651 | ||
658ddaaf SL |
652 | key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *) |
653 | &ip_hdr(skb)->saddr, AF_INET); | |
654 | if (!key) | |
3b24d854 ED |
655 | goto out; |
656 | ||
658ddaaf | 657 | |
39f8e58e | 658 | genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb); |
658ddaaf | 659 | if (genhash || memcmp(hash_location, newhash, 16) != 0) |
3b24d854 ED |
660 | goto out; |
661 | ||
658ddaaf SL |
662 | } |
663 | ||
cfb6eeb4 YH |
664 | if (key) { |
665 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
666 | (TCPOPT_NOP << 16) | | |
667 | (TCPOPT_MD5SIG << 8) | | |
668 | TCPOLEN_MD5SIG); | |
669 | /* Update length and the length the header thinks exists */ | |
670 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
671 | rep.th.doff = arg.iov[0].iov_len / 4; | |
672 | ||
49a72dfb | 673 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
674 | key, ip_hdr(skb)->saddr, |
675 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
676 | } |
677 | #endif | |
eddc9ec5 ACM |
678 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
679 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 680 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 681 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
271c3b9b FW |
682 | arg.flags = (sk && inet_sk_transparent(sk)) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
683 | ||
e2446eaa | 684 | /* When socket is gone, all binding information is lost. |
4c675258 AK |
685 | * routing might fail in this case. No choice here, if we choose to force |
686 | * input interface, we will misroute in case of asymmetric route. | |
e2446eaa | 687 | */ |
4c675258 AK |
688 | if (sk) |
689 | arg.bound_dev_if = sk->sk_bound_dev_if; | |
1da177e4 | 690 | |
271c3b9b FW |
691 | BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) != |
692 | offsetof(struct inet_timewait_sock, tw_bound_dev_if)); | |
693 | ||
66b13d99 | 694 | arg.tos = ip_hdr(skb)->tos; |
e2d118a1 | 695 | arg.uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL); |
47dcc20a | 696 | local_bh_disable(); |
bdbbb852 ED |
697 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
698 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
699 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
700 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 701 | |
90bbcc60 ED |
702 | __TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
703 | __TCP_INC_STATS(net, TCP_MIB_OUTRSTS); | |
47dcc20a | 704 | local_bh_enable(); |
658ddaaf SL |
705 | |
706 | #ifdef CONFIG_TCP_MD5SIG | |
3b24d854 ED |
707 | out: |
708 | rcu_read_unlock(); | |
658ddaaf | 709 | #endif |
1da177e4 LT |
710 | } |
711 | ||
712 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
713 | outside socket context is ugly, certainly. What can I do? | |
714 | */ | |
715 | ||
e2d118a1 | 716 | static void tcp_v4_send_ack(const struct sock *sk, |
e62a123b | 717 | struct sk_buff *skb, u32 seq, u32 ack, |
ee684b6f | 718 | u32 win, u32 tsval, u32 tsecr, int oif, |
88ef4a5a | 719 | struct tcp_md5sig_key *key, |
66b13d99 | 720 | int reply_flags, u8 tos) |
1da177e4 | 721 | { |
cf533ea5 | 722 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
723 | struct { |
724 | struct tcphdr th; | |
714e85be | 725 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 726 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 727 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
728 | #endif |
729 | ]; | |
1da177e4 | 730 | } rep; |
e2d118a1 | 731 | struct net *net = sock_net(sk); |
1da177e4 LT |
732 | struct ip_reply_arg arg; |
733 | ||
734 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 735 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
736 | |
737 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
738 | arg.iov[0].iov_len = sizeof(rep.th); | |
ee684b6f | 739 | if (tsecr) { |
cfb6eeb4 YH |
740 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
741 | (TCPOPT_TIMESTAMP << 8) | | |
742 | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
743 | rep.opt[1] = htonl(tsval); |
744 | rep.opt[2] = htonl(tsecr); | |
cb48cfe8 | 745 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
746 | } |
747 | ||
748 | /* Swap the send and the receive. */ | |
749 | rep.th.dest = th->source; | |
750 | rep.th.source = th->dest; | |
751 | rep.th.doff = arg.iov[0].iov_len / 4; | |
752 | rep.th.seq = htonl(seq); | |
753 | rep.th.ack_seq = htonl(ack); | |
754 | rep.th.ack = 1; | |
755 | rep.th.window = htons(win); | |
756 | ||
cfb6eeb4 | 757 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 758 | if (key) { |
ee684b6f | 759 | int offset = (tsecr) ? 3 : 0; |
cfb6eeb4 YH |
760 | |
761 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
762 | (TCPOPT_NOP << 16) | | |
763 | (TCPOPT_MD5SIG << 8) | | |
764 | TCPOLEN_MD5SIG); | |
765 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
766 | rep.th.doff = arg.iov[0].iov_len/4; | |
767 | ||
49a72dfb | 768 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
769 | key, ip_hdr(skb)->saddr, |
770 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
771 | } |
772 | #endif | |
88ef4a5a | 773 | arg.flags = reply_flags; |
eddc9ec5 ACM |
774 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
775 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
776 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
777 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
778 | if (oif) |
779 | arg.bound_dev_if = oif; | |
66b13d99 | 780 | arg.tos = tos; |
e2d118a1 | 781 | arg.uid = sock_net_uid(net, sk_fullsock(sk) ? sk : NULL); |
47dcc20a | 782 | local_bh_disable(); |
bdbbb852 ED |
783 | ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk), |
784 | skb, &TCP_SKB_CB(skb)->header.h4.opt, | |
24a2d43d ED |
785 | ip_hdr(skb)->saddr, ip_hdr(skb)->daddr, |
786 | &arg, arg.iov[0].iov_len); | |
1da177e4 | 787 | |
90bbcc60 | 788 | __TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
47dcc20a | 789 | local_bh_enable(); |
1da177e4 LT |
790 | } |
791 | ||
792 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
793 | { | |
8feaf0c0 | 794 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 795 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 796 | |
e2d118a1 | 797 | tcp_v4_send_ack(sk, skb, |
e62a123b | 798 | tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 799 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
ee684b6f | 800 | tcp_time_stamp + tcptw->tw_ts_offset, |
9501f972 YH |
801 | tcptw->tw_ts_recent, |
802 | tw->tw_bound_dev_if, | |
88ef4a5a | 803 | tcp_twsk_md5_key(tcptw), |
66b13d99 ED |
804 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0, |
805 | tw->tw_tos | |
9501f972 | 806 | ); |
1da177e4 | 807 | |
8feaf0c0 | 808 | inet_twsk_put(tw); |
1da177e4 LT |
809 | } |
810 | ||
a00e7444 | 811 | static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb, |
7174259e | 812 | struct request_sock *req) |
1da177e4 | 813 | { |
168a8f58 JC |
814 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
815 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
816 | */ | |
e62a123b ED |
817 | u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 : |
818 | tcp_sk(sk)->snd_nxt; | |
819 | ||
20a2b49f ED |
820 | /* RFC 7323 2.3 |
821 | * The window field (SEG.WND) of every outgoing segment, with the | |
822 | * exception of <SYN> segments, MUST be right-shifted by | |
823 | * Rcv.Wind.Shift bits: | |
824 | */ | |
e2d118a1 | 825 | tcp_v4_send_ack(sk, skb, seq, |
20a2b49f ED |
826 | tcp_rsk(req)->rcv_nxt, |
827 | req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale, | |
95a22cae | 828 | tcp_time_stamp + tcp_rsk(req)->ts_off, |
9501f972 YH |
829 | req->ts_recent, |
830 | 0, | |
a915da9b ED |
831 | tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr, |
832 | AF_INET), | |
66b13d99 ED |
833 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0, |
834 | ip_hdr(skb)->tos); | |
1da177e4 LT |
835 | } |
836 | ||
1da177e4 | 837 | /* |
9bf1d83e | 838 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 839 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
840 | * socket. |
841 | */ | |
0f935dbe | 842 | static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst, |
d6274bd8 | 843 | struct flowi *fl, |
72659ecc | 844 | struct request_sock *req, |
ca6fb065 | 845 | struct tcp_fastopen_cookie *foc, |
b3d05147 | 846 | enum tcp_synack_type synack_type) |
1da177e4 | 847 | { |
2e6599cb | 848 | const struct inet_request_sock *ireq = inet_rsk(req); |
6bd023f3 | 849 | struct flowi4 fl4; |
1da177e4 | 850 | int err = -1; |
d41db5af | 851 | struct sk_buff *skb; |
1da177e4 LT |
852 | |
853 | /* First, grab a route. */ | |
ba3f7f04 | 854 | if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL) |
fd80eb94 | 855 | return -1; |
1da177e4 | 856 | |
b3d05147 | 857 | skb = tcp_make_synack(sk, dst, req, foc, synack_type); |
1da177e4 LT |
858 | |
859 | if (skb) { | |
634fb979 | 860 | __tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr); |
1da177e4 | 861 | |
634fb979 ED |
862 | err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr, |
863 | ireq->ir_rmt_addr, | |
2e6599cb | 864 | ireq->opt); |
b9df3cb8 | 865 | err = net_xmit_eval(err); |
1da177e4 LT |
866 | } |
867 | ||
1da177e4 LT |
868 | return err; |
869 | } | |
870 | ||
871 | /* | |
60236fdd | 872 | * IPv4 request_sock destructor. |
1da177e4 | 873 | */ |
60236fdd | 874 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 875 | { |
a51482bd | 876 | kfree(inet_rsk(req)->opt); |
1da177e4 LT |
877 | } |
878 | ||
cfb6eeb4 YH |
879 | #ifdef CONFIG_TCP_MD5SIG |
880 | /* | |
881 | * RFC2385 MD5 checksumming requires a mapping of | |
882 | * IP address->MD5 Key. | |
883 | * We need to maintain these in the sk structure. | |
884 | */ | |
885 | ||
886 | /* Find the Key structure for an address. */ | |
b83e3deb | 887 | struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk, |
a915da9b ED |
888 | const union tcp_md5_addr *addr, |
889 | int family) | |
cfb6eeb4 | 890 | { |
fd3a154a | 891 | const struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 892 | struct tcp_md5sig_key *key; |
a915da9b | 893 | unsigned int size = sizeof(struct in_addr); |
fd3a154a | 894 | const struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 895 | |
a8afca03 ED |
896 | /* caller either holds rcu_read_lock() or socket lock */ |
897 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
1e1d04e6 | 898 | lockdep_sock_is_held(sk)); |
a8afca03 | 899 | if (!md5sig) |
cfb6eeb4 | 900 | return NULL; |
a915da9b ED |
901 | #if IS_ENABLED(CONFIG_IPV6) |
902 | if (family == AF_INET6) | |
903 | size = sizeof(struct in6_addr); | |
904 | #endif | |
b67bfe0d | 905 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { |
a915da9b ED |
906 | if (key->family != family) |
907 | continue; | |
908 | if (!memcmp(&key->addr, addr, size)) | |
909 | return key; | |
cfb6eeb4 YH |
910 | } |
911 | return NULL; | |
912 | } | |
a915da9b | 913 | EXPORT_SYMBOL(tcp_md5_do_lookup); |
cfb6eeb4 | 914 | |
b83e3deb | 915 | struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk, |
fd3a154a | 916 | const struct sock *addr_sk) |
cfb6eeb4 | 917 | { |
b52e6921 | 918 | const union tcp_md5_addr *addr; |
a915da9b | 919 | |
b52e6921 | 920 | addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr; |
a915da9b | 921 | return tcp_md5_do_lookup(sk, addr, AF_INET); |
cfb6eeb4 | 922 | } |
cfb6eeb4 YH |
923 | EXPORT_SYMBOL(tcp_v4_md5_lookup); |
924 | ||
cfb6eeb4 | 925 | /* This can be called on a newly created socket, from other files */ |
a915da9b ED |
926 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
927 | int family, const u8 *newkey, u8 newkeylen, gfp_t gfp) | |
cfb6eeb4 YH |
928 | { |
929 | /* Add Key to the list */ | |
b0a713e9 | 930 | struct tcp_md5sig_key *key; |
cfb6eeb4 | 931 | struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 932 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 933 | |
c0353c7b | 934 | key = tcp_md5_do_lookup(sk, addr, family); |
cfb6eeb4 YH |
935 | if (key) { |
936 | /* Pre-existing entry - just update that one. */ | |
a915da9b | 937 | memcpy(key->key, newkey, newkeylen); |
b0a713e9 | 938 | key->keylen = newkeylen; |
a915da9b ED |
939 | return 0; |
940 | } | |
260fcbeb | 941 | |
a8afca03 | 942 | md5sig = rcu_dereference_protected(tp->md5sig_info, |
1e1d04e6 | 943 | lockdep_sock_is_held(sk)); |
a915da9b ED |
944 | if (!md5sig) { |
945 | md5sig = kmalloc(sizeof(*md5sig), gfp); | |
946 | if (!md5sig) | |
cfb6eeb4 | 947 | return -ENOMEM; |
cfb6eeb4 | 948 | |
a915da9b ED |
949 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
950 | INIT_HLIST_HEAD(&md5sig->head); | |
a8afca03 | 951 | rcu_assign_pointer(tp->md5sig_info, md5sig); |
a915da9b | 952 | } |
cfb6eeb4 | 953 | |
5f3d9cb2 | 954 | key = sock_kmalloc(sk, sizeof(*key), gfp); |
a915da9b ED |
955 | if (!key) |
956 | return -ENOMEM; | |
71cea17e | 957 | if (!tcp_alloc_md5sig_pool()) { |
5f3d9cb2 | 958 | sock_kfree_s(sk, key, sizeof(*key)); |
a915da9b | 959 | return -ENOMEM; |
cfb6eeb4 | 960 | } |
a915da9b ED |
961 | |
962 | memcpy(key->key, newkey, newkeylen); | |
963 | key->keylen = newkeylen; | |
964 | key->family = family; | |
965 | memcpy(&key->addr, addr, | |
966 | (family == AF_INET6) ? sizeof(struct in6_addr) : | |
967 | sizeof(struct in_addr)); | |
968 | hlist_add_head_rcu(&key->node, &md5sig->head); | |
cfb6eeb4 YH |
969 | return 0; |
970 | } | |
a915da9b | 971 | EXPORT_SYMBOL(tcp_md5_do_add); |
cfb6eeb4 | 972 | |
a915da9b | 973 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family) |
cfb6eeb4 | 974 | { |
a915da9b ED |
975 | struct tcp_md5sig_key *key; |
976 | ||
c0353c7b | 977 | key = tcp_md5_do_lookup(sk, addr, family); |
a915da9b ED |
978 | if (!key) |
979 | return -ENOENT; | |
980 | hlist_del_rcu(&key->node); | |
5f3d9cb2 | 981 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 982 | kfree_rcu(key, rcu); |
a915da9b | 983 | return 0; |
cfb6eeb4 | 984 | } |
a915da9b | 985 | EXPORT_SYMBOL(tcp_md5_do_del); |
cfb6eeb4 | 986 | |
e0683e70 | 987 | static void tcp_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
988 | { |
989 | struct tcp_sock *tp = tcp_sk(sk); | |
a915da9b | 990 | struct tcp_md5sig_key *key; |
b67bfe0d | 991 | struct hlist_node *n; |
a8afca03 | 992 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 993 | |
a8afca03 ED |
994 | md5sig = rcu_dereference_protected(tp->md5sig_info, 1); |
995 | ||
b67bfe0d | 996 | hlist_for_each_entry_safe(key, n, &md5sig->head, node) { |
a915da9b | 997 | hlist_del_rcu(&key->node); |
5f3d9cb2 | 998 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 999 | kfree_rcu(key, rcu); |
cfb6eeb4 YH |
1000 | } |
1001 | } | |
1002 | ||
7174259e ACM |
1003 | static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval, |
1004 | int optlen) | |
cfb6eeb4 YH |
1005 | { |
1006 | struct tcp_md5sig cmd; | |
1007 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
cfb6eeb4 YH |
1008 | |
1009 | if (optlen < sizeof(cmd)) | |
1010 | return -EINVAL; | |
1011 | ||
7174259e | 1012 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
1013 | return -EFAULT; |
1014 | ||
1015 | if (sin->sin_family != AF_INET) | |
1016 | return -EINVAL; | |
1017 | ||
64a124ed | 1018 | if (!cmd.tcpm_keylen) |
a915da9b ED |
1019 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
1020 | AF_INET); | |
cfb6eeb4 YH |
1021 | |
1022 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
1023 | return -EINVAL; | |
1024 | ||
a915da9b ED |
1025 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
1026 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, | |
1027 | GFP_KERNEL); | |
cfb6eeb4 YH |
1028 | } |
1029 | ||
19689e38 ED |
1030 | static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp, |
1031 | __be32 daddr, __be32 saddr, | |
1032 | const struct tcphdr *th, int nbytes) | |
cfb6eeb4 | 1033 | { |
cfb6eeb4 | 1034 | struct tcp4_pseudohdr *bp; |
49a72dfb | 1035 | struct scatterlist sg; |
19689e38 | 1036 | struct tcphdr *_th; |
cfb6eeb4 | 1037 | |
19689e38 | 1038 | bp = hp->scratch; |
cfb6eeb4 YH |
1039 | bp->saddr = saddr; |
1040 | bp->daddr = daddr; | |
1041 | bp->pad = 0; | |
076fb722 | 1042 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1043 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1044 | |
19689e38 ED |
1045 | _th = (struct tcphdr *)(bp + 1); |
1046 | memcpy(_th, th, sizeof(*th)); | |
1047 | _th->check = 0; | |
1048 | ||
1049 | sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th)); | |
1050 | ahash_request_set_crypt(hp->md5_req, &sg, NULL, | |
1051 | sizeof(*bp) + sizeof(*th)); | |
cf80e0e4 | 1052 | return crypto_ahash_update(hp->md5_req); |
49a72dfb AL |
1053 | } |
1054 | ||
a915da9b | 1055 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 1056 | __be32 daddr, __be32 saddr, const struct tcphdr *th) |
49a72dfb AL |
1057 | { |
1058 | struct tcp_md5sig_pool *hp; | |
cf80e0e4 | 1059 | struct ahash_request *req; |
49a72dfb AL |
1060 | |
1061 | hp = tcp_get_md5sig_pool(); | |
1062 | if (!hp) | |
1063 | goto clear_hash_noput; | |
cf80e0e4 | 1064 | req = hp->md5_req; |
49a72dfb | 1065 | |
cf80e0e4 | 1066 | if (crypto_ahash_init(req)) |
49a72dfb | 1067 | goto clear_hash; |
19689e38 | 1068 | if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2)) |
49a72dfb AL |
1069 | goto clear_hash; |
1070 | if (tcp_md5_hash_key(hp, key)) | |
1071 | goto clear_hash; | |
cf80e0e4 HX |
1072 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
1073 | if (crypto_ahash_final(req)) | |
cfb6eeb4 YH |
1074 | goto clear_hash; |
1075 | ||
cfb6eeb4 | 1076 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1077 | return 0; |
49a72dfb | 1078 | |
cfb6eeb4 YH |
1079 | clear_hash: |
1080 | tcp_put_md5sig_pool(); | |
1081 | clear_hash_noput: | |
1082 | memset(md5_hash, 0, 16); | |
49a72dfb | 1083 | return 1; |
cfb6eeb4 YH |
1084 | } |
1085 | ||
39f8e58e ED |
1086 | int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, |
1087 | const struct sock *sk, | |
318cf7aa | 1088 | const struct sk_buff *skb) |
cfb6eeb4 | 1089 | { |
49a72dfb | 1090 | struct tcp_md5sig_pool *hp; |
cf80e0e4 | 1091 | struct ahash_request *req; |
318cf7aa | 1092 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
1093 | __be32 saddr, daddr; |
1094 | ||
39f8e58e ED |
1095 | if (sk) { /* valid for establish/request sockets */ |
1096 | saddr = sk->sk_rcv_saddr; | |
1097 | daddr = sk->sk_daddr; | |
cfb6eeb4 | 1098 | } else { |
49a72dfb AL |
1099 | const struct iphdr *iph = ip_hdr(skb); |
1100 | saddr = iph->saddr; | |
1101 | daddr = iph->daddr; | |
cfb6eeb4 | 1102 | } |
49a72dfb AL |
1103 | |
1104 | hp = tcp_get_md5sig_pool(); | |
1105 | if (!hp) | |
1106 | goto clear_hash_noput; | |
cf80e0e4 | 1107 | req = hp->md5_req; |
49a72dfb | 1108 | |
cf80e0e4 | 1109 | if (crypto_ahash_init(req)) |
49a72dfb AL |
1110 | goto clear_hash; |
1111 | ||
19689e38 | 1112 | if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, skb->len)) |
49a72dfb AL |
1113 | goto clear_hash; |
1114 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1115 | goto clear_hash; | |
1116 | if (tcp_md5_hash_key(hp, key)) | |
1117 | goto clear_hash; | |
cf80e0e4 HX |
1118 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
1119 | if (crypto_ahash_final(req)) | |
49a72dfb AL |
1120 | goto clear_hash; |
1121 | ||
1122 | tcp_put_md5sig_pool(); | |
1123 | return 0; | |
1124 | ||
1125 | clear_hash: | |
1126 | tcp_put_md5sig_pool(); | |
1127 | clear_hash_noput: | |
1128 | memset(md5_hash, 0, 16); | |
1129 | return 1; | |
cfb6eeb4 | 1130 | } |
49a72dfb | 1131 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1132 | |
ba8e275a ED |
1133 | #endif |
1134 | ||
ff74e23f | 1135 | /* Called with rcu_read_lock() */ |
ba8e275a | 1136 | static bool tcp_v4_inbound_md5_hash(const struct sock *sk, |
ff74e23f | 1137 | const struct sk_buff *skb) |
cfb6eeb4 | 1138 | { |
ba8e275a | 1139 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
1140 | /* |
1141 | * This gets called for each TCP segment that arrives | |
1142 | * so we want to be efficient. | |
1143 | * We have 3 drop cases: | |
1144 | * o No MD5 hash and one expected. | |
1145 | * o MD5 hash and we're not expecting one. | |
1146 | * o MD5 hash and its wrong. | |
1147 | */ | |
cf533ea5 | 1148 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 1149 | struct tcp_md5sig_key *hash_expected; |
eddc9ec5 | 1150 | const struct iphdr *iph = ip_hdr(skb); |
cf533ea5 | 1151 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1152 | int genhash; |
cfb6eeb4 YH |
1153 | unsigned char newhash[16]; |
1154 | ||
a915da9b ED |
1155 | hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr, |
1156 | AF_INET); | |
7d5d5525 | 1157 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1158 | |
cfb6eeb4 YH |
1159 | /* We've parsed the options - do we have a hash? */ |
1160 | if (!hash_expected && !hash_location) | |
a2a385d6 | 1161 | return false; |
cfb6eeb4 YH |
1162 | |
1163 | if (hash_expected && !hash_location) { | |
c10d9310 | 1164 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
a2a385d6 | 1165 | return true; |
cfb6eeb4 YH |
1166 | } |
1167 | ||
1168 | if (!hash_expected && hash_location) { | |
c10d9310 | 1169 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
a2a385d6 | 1170 | return true; |
cfb6eeb4 YH |
1171 | } |
1172 | ||
1173 | /* Okay, so this is hash_expected and hash_location - | |
1174 | * so we need to calculate the checksum. | |
1175 | */ | |
49a72dfb AL |
1176 | genhash = tcp_v4_md5_hash_skb(newhash, |
1177 | hash_expected, | |
39f8e58e | 1178 | NULL, skb); |
cfb6eeb4 YH |
1179 | |
1180 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
72145a68 | 1181 | NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE); |
e87cc472 JP |
1182 | net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1183 | &iph->saddr, ntohs(th->source), | |
1184 | &iph->daddr, ntohs(th->dest), | |
1185 | genhash ? " tcp_v4_calc_md5_hash failed" | |
1186 | : ""); | |
a2a385d6 | 1187 | return true; |
cfb6eeb4 | 1188 | } |
a2a385d6 | 1189 | return false; |
cfb6eeb4 | 1190 | #endif |
ba8e275a ED |
1191 | return false; |
1192 | } | |
cfb6eeb4 | 1193 | |
b40cf18e ED |
1194 | static void tcp_v4_init_req(struct request_sock *req, |
1195 | const struct sock *sk_listener, | |
16bea70a OP |
1196 | struct sk_buff *skb) |
1197 | { | |
1198 | struct inet_request_sock *ireq = inet_rsk(req); | |
1199 | ||
08d2cc3b ED |
1200 | sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr); |
1201 | sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr); | |
16bea70a OP |
1202 | ireq->opt = tcp_v4_save_options(skb); |
1203 | } | |
1204 | ||
f964629e ED |
1205 | static struct dst_entry *tcp_v4_route_req(const struct sock *sk, |
1206 | struct flowi *fl, | |
d94e0417 OP |
1207 | const struct request_sock *req, |
1208 | bool *strict) | |
1209 | { | |
1210 | struct dst_entry *dst = inet_csk_route_req(sk, &fl->u.ip4, req); | |
1211 | ||
1212 | if (strict) { | |
1213 | if (fl->u.ip4.daddr == inet_rsk(req)->ir_rmt_addr) | |
1214 | *strict = true; | |
1215 | else | |
1216 | *strict = false; | |
1217 | } | |
1218 | ||
1219 | return dst; | |
1220 | } | |
1221 | ||
72a3effa | 1222 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1223 | .family = PF_INET, |
2e6599cb | 1224 | .obj_size = sizeof(struct tcp_request_sock), |
5db92c99 | 1225 | .rtx_syn_ack = tcp_rtx_synack, |
60236fdd ACM |
1226 | .send_ack = tcp_v4_reqsk_send_ack, |
1227 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 | 1228 | .send_reset = tcp_v4_send_reset, |
688d1945 | 1229 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
1230 | }; |
1231 | ||
b2e4b3de | 1232 | static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
2aec4a29 | 1233 | .mss_clamp = TCP_MSS_DEFAULT, |
16bea70a | 1234 | #ifdef CONFIG_TCP_MD5SIG |
fd3a154a | 1235 | .req_md5_lookup = tcp_v4_md5_lookup, |
e3afe7b7 | 1236 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
b6332e6c | 1237 | #endif |
16bea70a | 1238 | .init_req = tcp_v4_init_req, |
fb7b37a7 OP |
1239 | #ifdef CONFIG_SYN_COOKIES |
1240 | .cookie_init_seq = cookie_v4_init_sequence, | |
1241 | #endif | |
d94e0417 | 1242 | .route_req = tcp_v4_route_req, |
936b8bdb | 1243 | .init_seq = tcp_v4_init_sequence, |
d6274bd8 | 1244 | .send_synack = tcp_v4_send_synack, |
16bea70a | 1245 | }; |
cfb6eeb4 | 1246 | |
1da177e4 LT |
1247 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1248 | { | |
1da177e4 | 1249 | /* Never answer to SYNs send to broadcast or multicast */ |
511c3f92 | 1250 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1251 | goto drop; |
1252 | ||
1fb6f159 OP |
1253 | return tcp_conn_request(&tcp_request_sock_ops, |
1254 | &tcp_request_sock_ipv4_ops, sk, skb); | |
1da177e4 | 1255 | |
1da177e4 | 1256 | drop: |
9caad864 | 1257 | tcp_listendrop(sk); |
1da177e4 LT |
1258 | return 0; |
1259 | } | |
4bc2f18b | 1260 | EXPORT_SYMBOL(tcp_v4_conn_request); |
1da177e4 LT |
1261 | |
1262 | ||
1263 | /* | |
1264 | * The three way handshake has completed - we got a valid synack - | |
1265 | * now create the new socket. | |
1266 | */ | |
0c27171e | 1267 | struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb, |
60236fdd | 1268 | struct request_sock *req, |
5e0724d0 ED |
1269 | struct dst_entry *dst, |
1270 | struct request_sock *req_unhash, | |
1271 | bool *own_req) | |
1da177e4 | 1272 | { |
2e6599cb | 1273 | struct inet_request_sock *ireq; |
1da177e4 LT |
1274 | struct inet_sock *newinet; |
1275 | struct tcp_sock *newtp; | |
1276 | struct sock *newsk; | |
cfb6eeb4 YH |
1277 | #ifdef CONFIG_TCP_MD5SIG |
1278 | struct tcp_md5sig_key *key; | |
1279 | #endif | |
f6d8bd05 | 1280 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
1281 | |
1282 | if (sk_acceptq_is_full(sk)) | |
1283 | goto exit_overflow; | |
1284 | ||
1da177e4 LT |
1285 | newsk = tcp_create_openreq_child(sk, req, skb); |
1286 | if (!newsk) | |
093d2823 | 1287 | goto exit_nonewsk; |
1da177e4 | 1288 | |
bcd76111 | 1289 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
fae6ef87 | 1290 | inet_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1291 | |
1292 | newtp = tcp_sk(newsk); | |
1293 | newinet = inet_sk(newsk); | |
2e6599cb | 1294 | ireq = inet_rsk(req); |
d1e559d0 ED |
1295 | sk_daddr_set(newsk, ireq->ir_rmt_addr); |
1296 | sk_rcv_saddr_set(newsk, ireq->ir_loc_addr); | |
6dd9a14e | 1297 | newsk->sk_bound_dev_if = ireq->ir_iif; |
634fb979 | 1298 | newinet->inet_saddr = ireq->ir_loc_addr; |
f6d8bd05 ED |
1299 | inet_opt = ireq->opt; |
1300 | rcu_assign_pointer(newinet->inet_opt, inet_opt); | |
2e6599cb | 1301 | ireq->opt = NULL; |
463c84b9 | 1302 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1303 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
4c507d28 | 1304 | newinet->rcv_tos = ip_hdr(skb)->tos; |
d83d8461 | 1305 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
1306 | if (inet_opt) |
1307 | inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
c720c7e8 | 1308 | newinet->inet_id = newtp->write_seq ^ jiffies; |
1da177e4 | 1309 | |
dfd25fff ED |
1310 | if (!dst) { |
1311 | dst = inet_csk_route_child_sock(sk, newsk, req); | |
1312 | if (!dst) | |
1313 | goto put_and_exit; | |
1314 | } else { | |
1315 | /* syncookie case : see end of cookie_v4_check() */ | |
1316 | } | |
0e734419 DM |
1317 | sk_setup_caps(newsk, dst); |
1318 | ||
81164413 DB |
1319 | tcp_ca_openreq_child(newsk, dst); |
1320 | ||
1da177e4 | 1321 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1322 | newtp->advmss = dst_metric_advmss(dst); |
f5fff5dc TQ |
1323 | if (tcp_sk(sk)->rx_opt.user_mss && |
1324 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1325 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1326 | ||
1da177e4 LT |
1327 | tcp_initialize_rcv_mss(newsk); |
1328 | ||
cfb6eeb4 YH |
1329 | #ifdef CONFIG_TCP_MD5SIG |
1330 | /* Copy over the MD5 key from the original socket */ | |
a915da9b ED |
1331 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1332 | AF_INET); | |
00db4124 | 1333 | if (key) { |
cfb6eeb4 YH |
1334 | /* |
1335 | * We're using one, so create a matching key | |
1336 | * on the newsk structure. If we fail to get | |
1337 | * memory, then we end up not copying the key | |
1338 | * across. Shucks. | |
1339 | */ | |
a915da9b ED |
1340 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1341 | AF_INET, key->key, key->keylen, GFP_ATOMIC); | |
a465419b | 1342 | sk_nocaps_add(newsk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1343 | } |
1344 | #endif | |
1345 | ||
0e734419 DM |
1346 | if (__inet_inherit_port(sk, newsk) < 0) |
1347 | goto put_and_exit; | |
5e0724d0 | 1348 | *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash)); |
805c4bc0 | 1349 | if (*own_req) |
49a496c9 | 1350 | tcp_move_syn(newtp, req); |
1da177e4 LT |
1351 | |
1352 | return newsk; | |
1353 | ||
1354 | exit_overflow: | |
c10d9310 | 1355 | NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 BS |
1356 | exit_nonewsk: |
1357 | dst_release(dst); | |
1da177e4 | 1358 | exit: |
9caad864 | 1359 | tcp_listendrop(sk); |
1da177e4 | 1360 | return NULL; |
0e734419 | 1361 | put_and_exit: |
e337e24d CP |
1362 | inet_csk_prepare_forced_close(newsk); |
1363 | tcp_done(newsk); | |
0e734419 | 1364 | goto exit; |
1da177e4 | 1365 | } |
4bc2f18b | 1366 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); |
1da177e4 | 1367 | |
079096f1 | 1368 | static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1369 | { |
079096f1 | 1370 | #ifdef CONFIG_SYN_COOKIES |
52452c54 | 1371 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 1372 | |
af9b4738 | 1373 | if (!th->syn) |
461b74c3 | 1374 | sk = cookie_v4_check(sk, skb); |
1da177e4 LT |
1375 | #endif |
1376 | return sk; | |
1377 | } | |
1378 | ||
1da177e4 | 1379 | /* The socket must have it's spinlock held when we get |
e994b2f0 | 1380 | * here, unless it is a TCP_LISTEN socket. |
1da177e4 LT |
1381 | * |
1382 | * We have a potential double-lock case here, so even when | |
1383 | * doing backlog processing we use the BH locking scheme. | |
1384 | * This is because we cannot sleep with the original spinlock | |
1385 | * held. | |
1386 | */ | |
1387 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1388 | { | |
cfb6eeb4 | 1389 | struct sock *rsk; |
cfb6eeb4 | 1390 | |
1da177e4 | 1391 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
404e0a8b ED |
1392 | struct dst_entry *dst = sk->sk_rx_dst; |
1393 | ||
bdeab991 | 1394 | sock_rps_save_rxhash(sk, skb); |
3d97379a | 1395 | sk_mark_napi_id(sk, skb); |
404e0a8b | 1396 | if (dst) { |
505fbcf0 | 1397 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || |
51456b29 | 1398 | !dst->ops->check(dst, 0)) { |
92101b3b DM |
1399 | dst_release(dst); |
1400 | sk->sk_rx_dst = NULL; | |
1401 | } | |
1402 | } | |
c995ae22 | 1403 | tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len); |
1da177e4 LT |
1404 | return 0; |
1405 | } | |
1406 | ||
12e25e10 | 1407 | if (tcp_checksum_complete(skb)) |
1da177e4 LT |
1408 | goto csum_err; |
1409 | ||
1410 | if (sk->sk_state == TCP_LISTEN) { | |
079096f1 ED |
1411 | struct sock *nsk = tcp_v4_cookie_check(sk, skb); |
1412 | ||
1da177e4 LT |
1413 | if (!nsk) |
1414 | goto discard; | |
1da177e4 | 1415 | if (nsk != sk) { |
bdeab991 | 1416 | sock_rps_save_rxhash(nsk, skb); |
38cb5245 | 1417 | sk_mark_napi_id(nsk, skb); |
cfb6eeb4 YH |
1418 | if (tcp_child_process(sk, nsk, skb)) { |
1419 | rsk = nsk; | |
1da177e4 | 1420 | goto reset; |
cfb6eeb4 | 1421 | } |
1da177e4 LT |
1422 | return 0; |
1423 | } | |
ca55158c | 1424 | } else |
bdeab991 | 1425 | sock_rps_save_rxhash(sk, skb); |
ca55158c | 1426 | |
72ab4a86 | 1427 | if (tcp_rcv_state_process(sk, skb)) { |
cfb6eeb4 | 1428 | rsk = sk; |
1da177e4 | 1429 | goto reset; |
cfb6eeb4 | 1430 | } |
1da177e4 LT |
1431 | return 0; |
1432 | ||
1433 | reset: | |
cfb6eeb4 | 1434 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1435 | discard: |
1436 | kfree_skb(skb); | |
1437 | /* Be careful here. If this function gets more complicated and | |
1438 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1439 | * might be destroyed here. This current version compiles correctly, | |
1440 | * but you have been warned. | |
1441 | */ | |
1442 | return 0; | |
1443 | ||
1444 | csum_err: | |
c10d9310 ED |
1445 | TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS); |
1446 | TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS); | |
1da177e4 LT |
1447 | goto discard; |
1448 | } | |
4bc2f18b | 1449 | EXPORT_SYMBOL(tcp_v4_do_rcv); |
1da177e4 | 1450 | |
160eb5a6 | 1451 | void tcp_v4_early_demux(struct sk_buff *skb) |
41063e9d | 1452 | { |
41063e9d DM |
1453 | const struct iphdr *iph; |
1454 | const struct tcphdr *th; | |
1455 | struct sock *sk; | |
41063e9d | 1456 | |
41063e9d | 1457 | if (skb->pkt_type != PACKET_HOST) |
160eb5a6 | 1458 | return; |
41063e9d | 1459 | |
45f00f99 | 1460 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) |
160eb5a6 | 1461 | return; |
41063e9d DM |
1462 | |
1463 | iph = ip_hdr(skb); | |
45f00f99 | 1464 | th = tcp_hdr(skb); |
41063e9d DM |
1465 | |
1466 | if (th->doff < sizeof(struct tcphdr) / 4) | |
160eb5a6 | 1467 | return; |
41063e9d | 1468 | |
45f00f99 | 1469 | sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo, |
41063e9d | 1470 | iph->saddr, th->source, |
7011d085 | 1471 | iph->daddr, ntohs(th->dest), |
9cb429d6 | 1472 | skb->skb_iif); |
41063e9d DM |
1473 | if (sk) { |
1474 | skb->sk = sk; | |
1475 | skb->destructor = sock_edemux; | |
f7e4eb03 | 1476 | if (sk_fullsock(sk)) { |
d0c294c5 | 1477 | struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst); |
505fbcf0 | 1478 | |
41063e9d DM |
1479 | if (dst) |
1480 | dst = dst_check(dst, 0); | |
92101b3b | 1481 | if (dst && |
505fbcf0 | 1482 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
92101b3b | 1483 | skb_dst_set_noref(skb, dst); |
41063e9d DM |
1484 | } |
1485 | } | |
41063e9d DM |
1486 | } |
1487 | ||
b2fb4f54 ED |
1488 | /* Packet is added to VJ-style prequeue for processing in process |
1489 | * context, if a reader task is waiting. Apparently, this exciting | |
1490 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
1491 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
1492 | * see, why it failed. 8)8) --ANK | |
1493 | * | |
1494 | */ | |
1495 | bool tcp_prequeue(struct sock *sk, struct sk_buff *skb) | |
1496 | { | |
1497 | struct tcp_sock *tp = tcp_sk(sk); | |
1498 | ||
1499 | if (sysctl_tcp_low_latency || !tp->ucopy.task) | |
1500 | return false; | |
1501 | ||
1502 | if (skb->len <= tcp_hdrlen(skb) && | |
1503 | skb_queue_len(&tp->ucopy.prequeue) == 0) | |
1504 | return false; | |
1505 | ||
ca777eff ED |
1506 | /* Before escaping RCU protected region, we need to take care of skb |
1507 | * dst. Prequeue is only enabled for established sockets. | |
1508 | * For such sockets, we might need the skb dst only to set sk->sk_rx_dst | |
1509 | * Instead of doing full sk_rx_dst validity here, let's perform | |
1510 | * an optimistic check. | |
1511 | */ | |
1512 | if (likely(sk->sk_rx_dst)) | |
1513 | skb_dst_drop(skb); | |
1514 | else | |
5037e9ef | 1515 | skb_dst_force_safe(skb); |
ca777eff | 1516 | |
b2fb4f54 ED |
1517 | __skb_queue_tail(&tp->ucopy.prequeue, skb); |
1518 | tp->ucopy.memory += skb->truesize; | |
0cef6a4c ED |
1519 | if (skb_queue_len(&tp->ucopy.prequeue) >= 32 || |
1520 | tp->ucopy.memory + atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) { | |
b2fb4f54 ED |
1521 | struct sk_buff *skb1; |
1522 | ||
1523 | BUG_ON(sock_owned_by_user(sk)); | |
0cef6a4c ED |
1524 | __NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPPREQUEUEDROPPED, |
1525 | skb_queue_len(&tp->ucopy.prequeue)); | |
b2fb4f54 | 1526 | |
0cef6a4c | 1527 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) |
b2fb4f54 | 1528 | sk_backlog_rcv(sk, skb1); |
b2fb4f54 ED |
1529 | |
1530 | tp->ucopy.memory = 0; | |
1531 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
1532 | wake_up_interruptible_sync_poll(sk_sleep(sk), | |
1533 | POLLIN | POLLRDNORM | POLLRDBAND); | |
1534 | if (!inet_csk_ack_scheduled(sk)) | |
1535 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
1536 | (3 * tcp_rto_min(sk)) / 4, | |
1537 | TCP_RTO_MAX); | |
1538 | } | |
1539 | return true; | |
1540 | } | |
1541 | EXPORT_SYMBOL(tcp_prequeue); | |
1542 | ||
c9c33212 ED |
1543 | bool tcp_add_backlog(struct sock *sk, struct sk_buff *skb) |
1544 | { | |
1545 | u32 limit = sk->sk_rcvbuf + sk->sk_sndbuf; | |
1546 | ||
1547 | /* Only socket owner can try to collapse/prune rx queues | |
1548 | * to reduce memory overhead, so add a little headroom here. | |
1549 | * Few sockets backlog are possibly concurrently non empty. | |
1550 | */ | |
1551 | limit += 64*1024; | |
1552 | ||
1553 | /* In case all data was pulled from skb frags (in __pskb_pull_tail()), | |
1554 | * we can fix skb->truesize to its real value to avoid future drops. | |
1555 | * This is valid because skb is not yet charged to the socket. | |
1556 | * It has been noticed pure SACK packets were sometimes dropped | |
1557 | * (if cooked by drivers without copybreak feature). | |
1558 | */ | |
60b1af33 | 1559 | skb_condense(skb); |
c9c33212 ED |
1560 | |
1561 | if (unlikely(sk_add_backlog(sk, skb, limit))) { | |
1562 | bh_unlock_sock(sk); | |
1563 | __NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPBACKLOGDROP); | |
1564 | return true; | |
1565 | } | |
1566 | return false; | |
1567 | } | |
1568 | EXPORT_SYMBOL(tcp_add_backlog); | |
1569 | ||
ac6e7800 ED |
1570 | int tcp_filter(struct sock *sk, struct sk_buff *skb) |
1571 | { | |
1572 | struct tcphdr *th = (struct tcphdr *)skb->data; | |
1573 | unsigned int eaten = skb->len; | |
1574 | int err; | |
1575 | ||
1576 | err = sk_filter_trim_cap(sk, skb, th->doff * 4); | |
1577 | if (!err) { | |
1578 | eaten -= skb->len; | |
1579 | TCP_SKB_CB(skb)->end_seq -= eaten; | |
1580 | } | |
1581 | return err; | |
1582 | } | |
1583 | EXPORT_SYMBOL(tcp_filter); | |
1584 | ||
1da177e4 LT |
1585 | /* |
1586 | * From tcp_input.c | |
1587 | */ | |
1588 | ||
1589 | int tcp_v4_rcv(struct sk_buff *skb) | |
1590 | { | |
3b24d854 | 1591 | struct net *net = dev_net(skb->dev); |
eddc9ec5 | 1592 | const struct iphdr *iph; |
cf533ea5 | 1593 | const struct tcphdr *th; |
3b24d854 | 1594 | bool refcounted; |
1da177e4 LT |
1595 | struct sock *sk; |
1596 | int ret; | |
1597 | ||
1598 | if (skb->pkt_type != PACKET_HOST) | |
1599 | goto discard_it; | |
1600 | ||
1601 | /* Count it even if it's bad */ | |
90bbcc60 | 1602 | __TCP_INC_STATS(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1603 | |
1604 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1605 | goto discard_it; | |
1606 | ||
ea1627c2 | 1607 | th = (const struct tcphdr *)skb->data; |
1da177e4 | 1608 | |
ea1627c2 | 1609 | if (unlikely(th->doff < sizeof(struct tcphdr) / 4)) |
1da177e4 LT |
1610 | goto bad_packet; |
1611 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1612 | goto discard_it; | |
1613 | ||
1614 | /* An explanation is required here, I think. | |
1615 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1616 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1617 | * So, we defer the checks. */ |
ed70fcfc TH |
1618 | |
1619 | if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo)) | |
6a5dc9e5 | 1620 | goto csum_error; |
1da177e4 | 1621 | |
ea1627c2 | 1622 | th = (const struct tcphdr *)skb->data; |
eddc9ec5 | 1623 | iph = ip_hdr(skb); |
971f10ec ED |
1624 | /* This is tricky : We move IPCB at its correct location into TCP_SKB_CB() |
1625 | * barrier() makes sure compiler wont play fool^Waliasing games. | |
1626 | */ | |
1627 | memmove(&TCP_SKB_CB(skb)->header.h4, IPCB(skb), | |
1628 | sizeof(struct inet_skb_parm)); | |
1629 | barrier(); | |
1630 | ||
1da177e4 LT |
1631 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1632 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1633 | skb->len - th->doff * 4); | |
1634 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
e11ecddf | 1635 | TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th); |
04317daf | 1636 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; |
b82d1bb4 | 1637 | TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph); |
1da177e4 LT |
1638 | TCP_SKB_CB(skb)->sacked = 0; |
1639 | ||
4bdc3d66 | 1640 | lookup: |
a583636a | 1641 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source, |
3b24d854 | 1642 | th->dest, &refcounted); |
1da177e4 LT |
1643 | if (!sk) |
1644 | goto no_tcp_socket; | |
1645 | ||
bb134d5d ED |
1646 | process: |
1647 | if (sk->sk_state == TCP_TIME_WAIT) | |
1648 | goto do_time_wait; | |
1649 | ||
079096f1 ED |
1650 | if (sk->sk_state == TCP_NEW_SYN_RECV) { |
1651 | struct request_sock *req = inet_reqsk(sk); | |
7716682c | 1652 | struct sock *nsk; |
079096f1 ED |
1653 | |
1654 | sk = req->rsk_listener; | |
72923555 | 1655 | if (unlikely(tcp_v4_inbound_md5_hash(sk, skb))) { |
e65c332d | 1656 | sk_drops_add(sk, skb); |
72923555 ED |
1657 | reqsk_put(req); |
1658 | goto discard_it; | |
1659 | } | |
7716682c | 1660 | if (unlikely(sk->sk_state != TCP_LISTEN)) { |
f03f2e15 | 1661 | inet_csk_reqsk_queue_drop_and_put(sk, req); |
4bdc3d66 ED |
1662 | goto lookup; |
1663 | } | |
3b24d854 ED |
1664 | /* We own a reference on the listener, increase it again |
1665 | * as we might lose it too soon. | |
1666 | */ | |
7716682c | 1667 | sock_hold(sk); |
3b24d854 | 1668 | refcounted = true; |
7716682c | 1669 | nsk = tcp_check_req(sk, skb, req, false); |
079096f1 ED |
1670 | if (!nsk) { |
1671 | reqsk_put(req); | |
7716682c | 1672 | goto discard_and_relse; |
079096f1 ED |
1673 | } |
1674 | if (nsk == sk) { | |
079096f1 ED |
1675 | reqsk_put(req); |
1676 | } else if (tcp_child_process(sk, nsk, skb)) { | |
1677 | tcp_v4_send_reset(nsk, skb); | |
7716682c | 1678 | goto discard_and_relse; |
079096f1 | 1679 | } else { |
7716682c | 1680 | sock_put(sk); |
079096f1 ED |
1681 | return 0; |
1682 | } | |
1683 | } | |
6cce09f8 | 1684 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
02a1d6e7 | 1685 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); |
d218d111 | 1686 | goto discard_and_relse; |
6cce09f8 | 1687 | } |
d218d111 | 1688 | |
1da177e4 LT |
1689 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1690 | goto discard_and_relse; | |
9ea88a15 | 1691 | |
9ea88a15 DP |
1692 | if (tcp_v4_inbound_md5_hash(sk, skb)) |
1693 | goto discard_and_relse; | |
9ea88a15 | 1694 | |
b59c2701 | 1695 | nf_reset(skb); |
1da177e4 | 1696 | |
ac6e7800 | 1697 | if (tcp_filter(sk, skb)) |
1da177e4 | 1698 | goto discard_and_relse; |
ac6e7800 ED |
1699 | th = (const struct tcphdr *)skb->data; |
1700 | iph = ip_hdr(skb); | |
1da177e4 LT |
1701 | |
1702 | skb->dev = NULL; | |
1703 | ||
e994b2f0 ED |
1704 | if (sk->sk_state == TCP_LISTEN) { |
1705 | ret = tcp_v4_do_rcv(sk, skb); | |
1706 | goto put_and_return; | |
1707 | } | |
1708 | ||
1709 | sk_incoming_cpu_update(sk); | |
1710 | ||
c6366184 | 1711 | bh_lock_sock_nested(sk); |
a44d6eac | 1712 | tcp_segs_in(tcp_sk(sk), skb); |
1da177e4 LT |
1713 | ret = 0; |
1714 | if (!sock_owned_by_user(sk)) { | |
7bced397 | 1715 | if (!tcp_prequeue(sk, skb)) |
1da177e4 | 1716 | ret = tcp_v4_do_rcv(sk, skb); |
c9c33212 | 1717 | } else if (tcp_add_backlog(sk, skb)) { |
6b03a53a ZY |
1718 | goto discard_and_relse; |
1719 | } | |
1da177e4 LT |
1720 | bh_unlock_sock(sk); |
1721 | ||
e994b2f0 | 1722 | put_and_return: |
3b24d854 ED |
1723 | if (refcounted) |
1724 | sock_put(sk); | |
1da177e4 LT |
1725 | |
1726 | return ret; | |
1727 | ||
1728 | no_tcp_socket: | |
1729 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1730 | goto discard_it; | |
1731 | ||
12e25e10 | 1732 | if (tcp_checksum_complete(skb)) { |
6a5dc9e5 | 1733 | csum_error: |
90bbcc60 | 1734 | __TCP_INC_STATS(net, TCP_MIB_CSUMERRORS); |
1da177e4 | 1735 | bad_packet: |
90bbcc60 | 1736 | __TCP_INC_STATS(net, TCP_MIB_INERRS); |
1da177e4 | 1737 | } else { |
cfb6eeb4 | 1738 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1739 | } |
1740 | ||
1741 | discard_it: | |
1742 | /* Discard frame. */ | |
1743 | kfree_skb(skb); | |
e905a9ed | 1744 | return 0; |
1da177e4 LT |
1745 | |
1746 | discard_and_relse: | |
532182cd | 1747 | sk_drops_add(sk, skb); |
3b24d854 ED |
1748 | if (refcounted) |
1749 | sock_put(sk); | |
1da177e4 LT |
1750 | goto discard_it; |
1751 | ||
1752 | do_time_wait: | |
1753 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1754 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1755 | goto discard_it; |
1756 | } | |
1757 | ||
6a5dc9e5 ED |
1758 | if (tcp_checksum_complete(skb)) { |
1759 | inet_twsk_put(inet_twsk(sk)); | |
1760 | goto csum_error; | |
1da177e4 | 1761 | } |
9469c7b4 | 1762 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1763 | case TCP_TW_SYN: { |
c346dca1 | 1764 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
a583636a CG |
1765 | &tcp_hashinfo, skb, |
1766 | __tcp_hdrlen(th), | |
da5e3630 | 1767 | iph->saddr, th->source, |
eddc9ec5 | 1768 | iph->daddr, th->dest, |
463c84b9 | 1769 | inet_iif(skb)); |
1da177e4 | 1770 | if (sk2) { |
dbe7faa4 | 1771 | inet_twsk_deschedule_put(inet_twsk(sk)); |
1da177e4 | 1772 | sk = sk2; |
3b24d854 | 1773 | refcounted = false; |
1da177e4 LT |
1774 | goto process; |
1775 | } | |
1776 | /* Fall through to ACK */ | |
1777 | } | |
1778 | case TCP_TW_ACK: | |
1779 | tcp_v4_timewait_ack(sk, skb); | |
1780 | break; | |
1781 | case TCP_TW_RST: | |
271c3b9b FW |
1782 | tcp_v4_send_reset(sk, skb); |
1783 | inet_twsk_deschedule_put(inet_twsk(sk)); | |
1784 | goto discard_it; | |
1da177e4 LT |
1785 | case TCP_TW_SUCCESS:; |
1786 | } | |
1787 | goto discard_it; | |
1788 | } | |
1789 | ||
ccb7c410 DM |
1790 | static struct timewait_sock_ops tcp_timewait_sock_ops = { |
1791 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1792 | .twsk_unique = tcp_twsk_unique, | |
1793 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 | 1794 | }; |
1da177e4 | 1795 | |
63d02d15 | 1796 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
5d299f3d ED |
1797 | { |
1798 | struct dst_entry *dst = skb_dst(skb); | |
1799 | ||
5037e9ef | 1800 | if (dst && dst_hold_safe(dst)) { |
ca777eff ED |
1801 | sk->sk_rx_dst = dst; |
1802 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
1803 | } | |
5d299f3d | 1804 | } |
63d02d15 | 1805 | EXPORT_SYMBOL(inet_sk_rx_dst_set); |
5d299f3d | 1806 | |
3b401a81 | 1807 | const struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1808 | .queue_xmit = ip_queue_xmit, |
1809 | .send_check = tcp_v4_send_check, | |
1810 | .rebuild_header = inet_sk_rebuild_header, | |
5d299f3d | 1811 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1812 | .conn_request = tcp_v4_conn_request, |
1813 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
543d9cfe ACM |
1814 | .net_header_len = sizeof(struct iphdr), |
1815 | .setsockopt = ip_setsockopt, | |
1816 | .getsockopt = ip_getsockopt, | |
1817 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1818 | .sockaddr_len = sizeof(struct sockaddr_in), | |
3fdadf7d | 1819 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1820 | .compat_setsockopt = compat_ip_setsockopt, |
1821 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1822 | #endif |
4fab9071 | 1823 | .mtu_reduced = tcp_v4_mtu_reduced, |
1da177e4 | 1824 | }; |
4bc2f18b | 1825 | EXPORT_SYMBOL(ipv4_specific); |
1da177e4 | 1826 | |
cfb6eeb4 | 1827 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1828 | static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1829 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1830 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1831 | .md5_parse = tcp_v4_parse_md5_keys, |
cfb6eeb4 | 1832 | }; |
b6332e6c | 1833 | #endif |
cfb6eeb4 | 1834 | |
1da177e4 LT |
1835 | /* NOTE: A lot of things set to zero explicitly by call to |
1836 | * sk_alloc() so need not be done here. | |
1837 | */ | |
1838 | static int tcp_v4_init_sock(struct sock *sk) | |
1839 | { | |
6687e988 | 1840 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1841 | |
900f65d3 | 1842 | tcp_init_sock(sk); |
1da177e4 | 1843 | |
8292a17a | 1844 | icsk->icsk_af_ops = &ipv4_specific; |
900f65d3 | 1845 | |
cfb6eeb4 | 1846 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1847 | tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific; |
cfb6eeb4 | 1848 | #endif |
1da177e4 | 1849 | |
1da177e4 LT |
1850 | return 0; |
1851 | } | |
1852 | ||
7d06b2e0 | 1853 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1854 | { |
1855 | struct tcp_sock *tp = tcp_sk(sk); | |
1856 | ||
1857 | tcp_clear_xmit_timers(sk); | |
1858 | ||
6687e988 | 1859 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1860 | |
1da177e4 | 1861 | /* Cleanup up the write buffer. */ |
fe067e8a | 1862 | tcp_write_queue_purge(sk); |
1da177e4 LT |
1863 | |
1864 | /* Cleans up our, hopefully empty, out_of_order_queue. */ | |
9f5afeae | 1865 | skb_rbtree_purge(&tp->out_of_order_queue); |
1da177e4 | 1866 | |
cfb6eeb4 YH |
1867 | #ifdef CONFIG_TCP_MD5SIG |
1868 | /* Clean up the MD5 key list, if any */ | |
1869 | if (tp->md5sig_info) { | |
a915da9b | 1870 | tcp_clear_md5_list(sk); |
a8afca03 | 1871 | kfree_rcu(tp->md5sig_info, rcu); |
cfb6eeb4 YH |
1872 | tp->md5sig_info = NULL; |
1873 | } | |
1874 | #endif | |
1a2449a8 | 1875 | |
1da177e4 LT |
1876 | /* Clean prequeue, it must be empty really */ |
1877 | __skb_queue_purge(&tp->ucopy.prequeue); | |
1878 | ||
1879 | /* Clean up a referenced TCP bind bucket. */ | |
463c84b9 | 1880 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1881 | inet_put_port(sk); |
1da177e4 | 1882 | |
00db4124 | 1883 | BUG_ON(tp->fastopen_rsk); |
435cf559 | 1884 | |
cf60af03 YC |
1885 | /* If socket is aborted during connect operation */ |
1886 | tcp_free_fastopen_req(tp); | |
cd8ae852 | 1887 | tcp_saved_syn_free(tp); |
cf60af03 | 1888 | |
180d8cd9 | 1889 | sk_sockets_allocated_dec(sk); |
1da177e4 | 1890 | } |
1da177e4 LT |
1891 | EXPORT_SYMBOL(tcp_v4_destroy_sock); |
1892 | ||
1893 | #ifdef CONFIG_PROC_FS | |
1894 | /* Proc filesystem TCP sock list dumping. */ | |
1895 | ||
a8b690f9 TH |
1896 | /* |
1897 | * Get next listener socket follow cur. If cur is NULL, get first socket | |
1898 | * starting from bucket given in st->bucket; when st->bucket is zero the | |
1899 | * very first socket in the hash table is returned. | |
1900 | */ | |
1da177e4 LT |
1901 | static void *listening_get_next(struct seq_file *seq, void *cur) |
1902 | { | |
5799de0b | 1903 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1904 | struct net *net = seq_file_net(seq); |
3b24d854 | 1905 | struct inet_listen_hashbucket *ilb; |
3b24d854 | 1906 | struct sock *sk = cur; |
1da177e4 LT |
1907 | |
1908 | if (!sk) { | |
3b24d854 | 1909 | get_head: |
a8b690f9 | 1910 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
9652dc2e | 1911 | spin_lock(&ilb->lock); |
3b24d854 | 1912 | sk = sk_head(&ilb->head); |
a8b690f9 | 1913 | st->offset = 0; |
1da177e4 LT |
1914 | goto get_sk; |
1915 | } | |
5caea4ea | 1916 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 | 1917 | ++st->num; |
a8b690f9 | 1918 | ++st->offset; |
1da177e4 | 1919 | |
3b24d854 | 1920 | sk = sk_next(sk); |
1da177e4 | 1921 | get_sk: |
3b24d854 | 1922 | sk_for_each_from(sk) { |
8475ef9f PE |
1923 | if (!net_eq(sock_net(sk), net)) |
1924 | continue; | |
3b24d854 ED |
1925 | if (sk->sk_family == st->family) |
1926 | return sk; | |
1da177e4 | 1927 | } |
9652dc2e | 1928 | spin_unlock(&ilb->lock); |
a8b690f9 | 1929 | st->offset = 0; |
3b24d854 ED |
1930 | if (++st->bucket < INET_LHTABLE_SIZE) |
1931 | goto get_head; | |
1932 | return NULL; | |
1da177e4 LT |
1933 | } |
1934 | ||
1935 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
1936 | { | |
a8b690f9 TH |
1937 | struct tcp_iter_state *st = seq->private; |
1938 | void *rc; | |
1939 | ||
1940 | st->bucket = 0; | |
1941 | st->offset = 0; | |
1942 | rc = listening_get_next(seq, NULL); | |
1da177e4 LT |
1943 | |
1944 | while (rc && *pos) { | |
1945 | rc = listening_get_next(seq, rc); | |
1946 | --*pos; | |
1947 | } | |
1948 | return rc; | |
1949 | } | |
1950 | ||
05dbc7b5 | 1951 | static inline bool empty_bucket(const struct tcp_iter_state *st) |
6eac5604 | 1952 | { |
05dbc7b5 | 1953 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain); |
6eac5604 AK |
1954 | } |
1955 | ||
a8b690f9 TH |
1956 | /* |
1957 | * Get first established socket starting from bucket given in st->bucket. | |
1958 | * If st->bucket is zero, the very first socket in the hash is returned. | |
1959 | */ | |
1da177e4 LT |
1960 | static void *established_get_first(struct seq_file *seq) |
1961 | { | |
5799de0b | 1962 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1963 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1964 | void *rc = NULL; |
1965 | ||
a8b690f9 TH |
1966 | st->offset = 0; |
1967 | for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) { | |
1da177e4 | 1968 | struct sock *sk; |
3ab5aee7 | 1969 | struct hlist_nulls_node *node; |
9db66bdc | 1970 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 1971 | |
6eac5604 AK |
1972 | /* Lockless fast path for the common case of empty buckets */ |
1973 | if (empty_bucket(st)) | |
1974 | continue; | |
1975 | ||
9db66bdc | 1976 | spin_lock_bh(lock); |
3ab5aee7 | 1977 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 1978 | if (sk->sk_family != st->family || |
878628fb | 1979 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
1980 | continue; |
1981 | } | |
1982 | rc = sk; | |
1983 | goto out; | |
1984 | } | |
9db66bdc | 1985 | spin_unlock_bh(lock); |
1da177e4 LT |
1986 | } |
1987 | out: | |
1988 | return rc; | |
1989 | } | |
1990 | ||
1991 | static void *established_get_next(struct seq_file *seq, void *cur) | |
1992 | { | |
1993 | struct sock *sk = cur; | |
3ab5aee7 | 1994 | struct hlist_nulls_node *node; |
5799de0b | 1995 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1996 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1997 | |
1998 | ++st->num; | |
a8b690f9 | 1999 | ++st->offset; |
1da177e4 | 2000 | |
05dbc7b5 | 2001 | sk = sk_nulls_next(sk); |
1da177e4 | 2002 | |
3ab5aee7 | 2003 | sk_nulls_for_each_from(sk, node) { |
878628fb | 2004 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
05dbc7b5 | 2005 | return sk; |
1da177e4 LT |
2006 | } |
2007 | ||
05dbc7b5 ED |
2008 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
2009 | ++st->bucket; | |
2010 | return established_get_first(seq); | |
1da177e4 LT |
2011 | } |
2012 | ||
2013 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2014 | { | |
a8b690f9 TH |
2015 | struct tcp_iter_state *st = seq->private; |
2016 | void *rc; | |
2017 | ||
2018 | st->bucket = 0; | |
2019 | rc = established_get_first(seq); | |
1da177e4 LT |
2020 | |
2021 | while (rc && pos) { | |
2022 | rc = established_get_next(seq, rc); | |
2023 | --pos; | |
7174259e | 2024 | } |
1da177e4 LT |
2025 | return rc; |
2026 | } | |
2027 | ||
2028 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2029 | { | |
2030 | void *rc; | |
5799de0b | 2031 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2032 | |
1da177e4 LT |
2033 | st->state = TCP_SEQ_STATE_LISTENING; |
2034 | rc = listening_get_idx(seq, &pos); | |
2035 | ||
2036 | if (!rc) { | |
1da177e4 LT |
2037 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2038 | rc = established_get_idx(seq, pos); | |
2039 | } | |
2040 | ||
2041 | return rc; | |
2042 | } | |
2043 | ||
a8b690f9 TH |
2044 | static void *tcp_seek_last_pos(struct seq_file *seq) |
2045 | { | |
2046 | struct tcp_iter_state *st = seq->private; | |
2047 | int offset = st->offset; | |
2048 | int orig_num = st->num; | |
2049 | void *rc = NULL; | |
2050 | ||
2051 | switch (st->state) { | |
a8b690f9 TH |
2052 | case TCP_SEQ_STATE_LISTENING: |
2053 | if (st->bucket >= INET_LHTABLE_SIZE) | |
2054 | break; | |
2055 | st->state = TCP_SEQ_STATE_LISTENING; | |
2056 | rc = listening_get_next(seq, NULL); | |
2057 | while (offset-- && rc) | |
2058 | rc = listening_get_next(seq, rc); | |
2059 | if (rc) | |
2060 | break; | |
2061 | st->bucket = 0; | |
05dbc7b5 | 2062 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2063 | /* Fallthrough */ |
2064 | case TCP_SEQ_STATE_ESTABLISHED: | |
a8b690f9 TH |
2065 | if (st->bucket > tcp_hashinfo.ehash_mask) |
2066 | break; | |
2067 | rc = established_get_first(seq); | |
2068 | while (offset-- && rc) | |
2069 | rc = established_get_next(seq, rc); | |
2070 | } | |
2071 | ||
2072 | st->num = orig_num; | |
2073 | ||
2074 | return rc; | |
2075 | } | |
2076 | ||
1da177e4 LT |
2077 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) |
2078 | { | |
5799de0b | 2079 | struct tcp_iter_state *st = seq->private; |
a8b690f9 TH |
2080 | void *rc; |
2081 | ||
2082 | if (*pos && *pos == st->last_pos) { | |
2083 | rc = tcp_seek_last_pos(seq); | |
2084 | if (rc) | |
2085 | goto out; | |
2086 | } | |
2087 | ||
1da177e4 LT |
2088 | st->state = TCP_SEQ_STATE_LISTENING; |
2089 | st->num = 0; | |
a8b690f9 TH |
2090 | st->bucket = 0; |
2091 | st->offset = 0; | |
2092 | rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2093 | ||
2094 | out: | |
2095 | st->last_pos = *pos; | |
2096 | return rc; | |
1da177e4 LT |
2097 | } |
2098 | ||
2099 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2100 | { | |
a8b690f9 | 2101 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2102 | void *rc = NULL; |
1da177e4 LT |
2103 | |
2104 | if (v == SEQ_START_TOKEN) { | |
2105 | rc = tcp_get_idx(seq, 0); | |
2106 | goto out; | |
2107 | } | |
1da177e4 LT |
2108 | |
2109 | switch (st->state) { | |
1da177e4 LT |
2110 | case TCP_SEQ_STATE_LISTENING: |
2111 | rc = listening_get_next(seq, v); | |
2112 | if (!rc) { | |
1da177e4 | 2113 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2114 | st->bucket = 0; |
2115 | st->offset = 0; | |
1da177e4 LT |
2116 | rc = established_get_first(seq); |
2117 | } | |
2118 | break; | |
2119 | case TCP_SEQ_STATE_ESTABLISHED: | |
1da177e4 LT |
2120 | rc = established_get_next(seq, v); |
2121 | break; | |
2122 | } | |
2123 | out: | |
2124 | ++*pos; | |
a8b690f9 | 2125 | st->last_pos = *pos; |
1da177e4 LT |
2126 | return rc; |
2127 | } | |
2128 | ||
2129 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2130 | { | |
5799de0b | 2131 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2132 | |
2133 | switch (st->state) { | |
1da177e4 LT |
2134 | case TCP_SEQ_STATE_LISTENING: |
2135 | if (v != SEQ_START_TOKEN) | |
9652dc2e | 2136 | spin_unlock(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 | 2137 | break; |
1da177e4 LT |
2138 | case TCP_SEQ_STATE_ESTABLISHED: |
2139 | if (v) | |
9db66bdc | 2140 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2141 | break; |
2142 | } | |
2143 | } | |
2144 | ||
73cb88ec | 2145 | int tcp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2146 | { |
d9dda78b | 2147 | struct tcp_seq_afinfo *afinfo = PDE_DATA(inode); |
1da177e4 | 2148 | struct tcp_iter_state *s; |
52d6f3f1 | 2149 | int err; |
1da177e4 | 2150 | |
52d6f3f1 DL |
2151 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2152 | sizeof(struct tcp_iter_state)); | |
2153 | if (err < 0) | |
2154 | return err; | |
f40c8174 | 2155 | |
52d6f3f1 | 2156 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2157 | s->family = afinfo->family; |
688d1945 | 2158 | s->last_pos = 0; |
f40c8174 DL |
2159 | return 0; |
2160 | } | |
73cb88ec | 2161 | EXPORT_SYMBOL(tcp_seq_open); |
f40c8174 | 2162 | |
6f8b13bc | 2163 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2164 | { |
2165 | int rc = 0; | |
2166 | struct proc_dir_entry *p; | |
2167 | ||
9427c4b3 DL |
2168 | afinfo->seq_ops.start = tcp_seq_start; |
2169 | afinfo->seq_ops.next = tcp_seq_next; | |
2170 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2171 | ||
84841c3c | 2172 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2173 | afinfo->seq_fops, afinfo); |
84841c3c | 2174 | if (!p) |
1da177e4 LT |
2175 | rc = -ENOMEM; |
2176 | return rc; | |
2177 | } | |
4bc2f18b | 2178 | EXPORT_SYMBOL(tcp_proc_register); |
1da177e4 | 2179 | |
6f8b13bc | 2180 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2181 | { |
ece31ffd | 2182 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2183 | } |
4bc2f18b | 2184 | EXPORT_SYMBOL(tcp_proc_unregister); |
1da177e4 | 2185 | |
d4f06873 | 2186 | static void get_openreq4(const struct request_sock *req, |
aa3a0c8c | 2187 | struct seq_file *f, int i) |
1da177e4 | 2188 | { |
2e6599cb | 2189 | const struct inet_request_sock *ireq = inet_rsk(req); |
fa76ce73 | 2190 | long delta = req->rsk_timer.expires - jiffies; |
1da177e4 | 2191 | |
5e659e4c | 2192 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2193 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK", |
1da177e4 | 2194 | i, |
634fb979 | 2195 | ireq->ir_loc_addr, |
d4f06873 | 2196 | ireq->ir_num, |
634fb979 ED |
2197 | ireq->ir_rmt_addr, |
2198 | ntohs(ireq->ir_rmt_port), | |
1da177e4 LT |
2199 | TCP_SYN_RECV, |
2200 | 0, 0, /* could print option size, but that is af dependent. */ | |
2201 | 1, /* timers active (only the expire timer) */ | |
a399a805 | 2202 | jiffies_delta_to_clock_t(delta), |
e6c022a4 | 2203 | req->num_timeout, |
aa3a0c8c ED |
2204 | from_kuid_munged(seq_user_ns(f), |
2205 | sock_i_uid(req->rsk_listener)), | |
1da177e4 LT |
2206 | 0, /* non standard timer */ |
2207 | 0, /* open_requests have no inode */ | |
d4f06873 | 2208 | 0, |
652586df | 2209 | req); |
1da177e4 LT |
2210 | } |
2211 | ||
652586df | 2212 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i) |
1da177e4 LT |
2213 | { |
2214 | int timer_active; | |
2215 | unsigned long timer_expires; | |
cf533ea5 | 2216 | const struct tcp_sock *tp = tcp_sk(sk); |
cf4c6bf8 | 2217 | const struct inet_connection_sock *icsk = inet_csk(sk); |
cf533ea5 | 2218 | const struct inet_sock *inet = inet_sk(sk); |
0536fcc0 | 2219 | const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq; |
c720c7e8 ED |
2220 | __be32 dest = inet->inet_daddr; |
2221 | __be32 src = inet->inet_rcv_saddr; | |
2222 | __u16 destp = ntohs(inet->inet_dport); | |
2223 | __u16 srcp = ntohs(inet->inet_sport); | |
49d09007 | 2224 | int rx_queue; |
00fd38d9 | 2225 | int state; |
1da177e4 | 2226 | |
6ba8a3b1 | 2227 | if (icsk->icsk_pending == ICSK_TIME_RETRANS || |
57dde7f7 | 2228 | icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || |
6ba8a3b1 | 2229 | icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { |
1da177e4 | 2230 | timer_active = 1; |
463c84b9 ACM |
2231 | timer_expires = icsk->icsk_timeout; |
2232 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2233 | timer_active = 4; |
463c84b9 | 2234 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2235 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2236 | timer_active = 2; |
cf4c6bf8 | 2237 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2238 | } else { |
2239 | timer_active = 0; | |
2240 | timer_expires = jiffies; | |
2241 | } | |
2242 | ||
00fd38d9 ED |
2243 | state = sk_state_load(sk); |
2244 | if (state == TCP_LISTEN) | |
49d09007 ED |
2245 | rx_queue = sk->sk_ack_backlog; |
2246 | else | |
00fd38d9 ED |
2247 | /* Because we don't lock the socket, |
2248 | * we might find a transient negative value. | |
49d09007 ED |
2249 | */ |
2250 | rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0); | |
2251 | ||
5e659e4c | 2252 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
652586df | 2253 | "%08X %5u %8d %lu %d %pK %lu %lu %u %u %d", |
00fd38d9 | 2254 | i, src, srcp, dest, destp, state, |
47da8ee6 | 2255 | tp->write_seq - tp->snd_una, |
49d09007 | 2256 | rx_queue, |
1da177e4 | 2257 | timer_active, |
a399a805 | 2258 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 2259 | icsk->icsk_retransmits, |
a7cb5a49 | 2260 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)), |
6687e988 | 2261 | icsk->icsk_probes_out, |
cf4c6bf8 IJ |
2262 | sock_i_ino(sk), |
2263 | atomic_read(&sk->sk_refcnt), sk, | |
7be87351 SH |
2264 | jiffies_to_clock_t(icsk->icsk_rto), |
2265 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2266 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2267 | tp->snd_cwnd, |
00fd38d9 ED |
2268 | state == TCP_LISTEN ? |
2269 | fastopenq->max_qlen : | |
652586df | 2270 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)); |
1da177e4 LT |
2271 | } |
2272 | ||
cf533ea5 | 2273 | static void get_timewait4_sock(const struct inet_timewait_sock *tw, |
652586df | 2274 | struct seq_file *f, int i) |
1da177e4 | 2275 | { |
789f558c | 2276 | long delta = tw->tw_timer.expires - jiffies; |
23f33c2d | 2277 | __be32 dest, src; |
1da177e4 | 2278 | __u16 destp, srcp; |
1da177e4 LT |
2279 | |
2280 | dest = tw->tw_daddr; | |
2281 | src = tw->tw_rcv_saddr; | |
2282 | destp = ntohs(tw->tw_dport); | |
2283 | srcp = ntohs(tw->tw_sport); | |
2284 | ||
5e659e4c | 2285 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2286 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK", |
1da177e4 | 2287 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
a399a805 | 2288 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
652586df | 2289 | atomic_read(&tw->tw_refcnt), tw); |
1da177e4 LT |
2290 | } |
2291 | ||
2292 | #define TMPSZ 150 | |
2293 | ||
2294 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2295 | { | |
5799de0b | 2296 | struct tcp_iter_state *st; |
05dbc7b5 | 2297 | struct sock *sk = v; |
1da177e4 | 2298 | |
652586df | 2299 | seq_setwidth(seq, TMPSZ - 1); |
1da177e4 | 2300 | if (v == SEQ_START_TOKEN) { |
652586df | 2301 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
1da177e4 LT |
2302 | "rx_queue tr tm->when retrnsmt uid timeout " |
2303 | "inode"); | |
2304 | goto out; | |
2305 | } | |
2306 | st = seq->private; | |
2307 | ||
079096f1 ED |
2308 | if (sk->sk_state == TCP_TIME_WAIT) |
2309 | get_timewait4_sock(v, seq, st->num); | |
2310 | else if (sk->sk_state == TCP_NEW_SYN_RECV) | |
aa3a0c8c | 2311 | get_openreq4(v, seq, st->num); |
079096f1 ED |
2312 | else |
2313 | get_tcp4_sock(v, seq, st->num); | |
1da177e4 | 2314 | out: |
652586df | 2315 | seq_pad(seq, '\n'); |
1da177e4 LT |
2316 | return 0; |
2317 | } | |
2318 | ||
73cb88ec AV |
2319 | static const struct file_operations tcp_afinfo_seq_fops = { |
2320 | .owner = THIS_MODULE, | |
2321 | .open = tcp_seq_open, | |
2322 | .read = seq_read, | |
2323 | .llseek = seq_lseek, | |
2324 | .release = seq_release_net | |
2325 | }; | |
2326 | ||
1da177e4 | 2327 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2328 | .name = "tcp", |
2329 | .family = AF_INET, | |
73cb88ec | 2330 | .seq_fops = &tcp_afinfo_seq_fops, |
9427c4b3 DL |
2331 | .seq_ops = { |
2332 | .show = tcp4_seq_show, | |
2333 | }, | |
1da177e4 LT |
2334 | }; |
2335 | ||
2c8c1e72 | 2336 | static int __net_init tcp4_proc_init_net(struct net *net) |
757764f6 PE |
2337 | { |
2338 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2339 | } | |
2340 | ||
2c8c1e72 | 2341 | static void __net_exit tcp4_proc_exit_net(struct net *net) |
757764f6 PE |
2342 | { |
2343 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2344 | } | |
2345 | ||
2346 | static struct pernet_operations tcp4_net_ops = { | |
2347 | .init = tcp4_proc_init_net, | |
2348 | .exit = tcp4_proc_exit_net, | |
2349 | }; | |
2350 | ||
1da177e4 LT |
2351 | int __init tcp4_proc_init(void) |
2352 | { | |
757764f6 | 2353 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2354 | } |
2355 | ||
2356 | void tcp4_proc_exit(void) | |
2357 | { | |
757764f6 | 2358 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2359 | } |
2360 | #endif /* CONFIG_PROC_FS */ | |
2361 | ||
2362 | struct proto tcp_prot = { | |
2363 | .name = "TCP", | |
2364 | .owner = THIS_MODULE, | |
2365 | .close = tcp_close, | |
2366 | .connect = tcp_v4_connect, | |
2367 | .disconnect = tcp_disconnect, | |
463c84b9 | 2368 | .accept = inet_csk_accept, |
1da177e4 LT |
2369 | .ioctl = tcp_ioctl, |
2370 | .init = tcp_v4_init_sock, | |
2371 | .destroy = tcp_v4_destroy_sock, | |
2372 | .shutdown = tcp_shutdown, | |
2373 | .setsockopt = tcp_setsockopt, | |
2374 | .getsockopt = tcp_getsockopt, | |
4b9d07a4 | 2375 | .keepalive = tcp_set_keepalive, |
1da177e4 | 2376 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2377 | .sendmsg = tcp_sendmsg, |
2378 | .sendpage = tcp_sendpage, | |
1da177e4 | 2379 | .backlog_rcv = tcp_v4_do_rcv, |
46d3ceab | 2380 | .release_cb = tcp_release_cb, |
ab1e0a13 ACM |
2381 | .hash = inet_hash, |
2382 | .unhash = inet_unhash, | |
2383 | .get_port = inet_csk_get_port, | |
1da177e4 | 2384 | .enter_memory_pressure = tcp_enter_memory_pressure, |
c9bee3b7 | 2385 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 | 2386 | .sockets_allocated = &tcp_sockets_allocated, |
0a5578cf | 2387 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2388 | .memory_allocated = &tcp_memory_allocated, |
2389 | .memory_pressure = &tcp_memory_pressure, | |
a4fe34bf | 2390 | .sysctl_mem = sysctl_tcp_mem, |
1da177e4 LT |
2391 | .sysctl_wmem = sysctl_tcp_wmem, |
2392 | .sysctl_rmem = sysctl_tcp_rmem, | |
2393 | .max_header = MAX_TCP_HEADER, | |
2394 | .obj_size = sizeof(struct tcp_sock), | |
3ab5aee7 | 2395 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2396 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2397 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2398 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2399 | .no_autobind = true, |
543d9cfe ACM |
2400 | #ifdef CONFIG_COMPAT |
2401 | .compat_setsockopt = compat_tcp_setsockopt, | |
2402 | .compat_getsockopt = compat_tcp_getsockopt, | |
d1a4c0b3 | 2403 | #endif |
c1e64e29 | 2404 | .diag_destroy = tcp_abort, |
1da177e4 | 2405 | }; |
4bc2f18b | 2406 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 | 2407 | |
bdbbb852 ED |
2408 | static void __net_exit tcp_sk_exit(struct net *net) |
2409 | { | |
2410 | int cpu; | |
2411 | ||
2412 | for_each_possible_cpu(cpu) | |
2413 | inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.tcp_sk, cpu)); | |
2414 | free_percpu(net->ipv4.tcp_sk); | |
2415 | } | |
2416 | ||
046ee902 DL |
2417 | static int __net_init tcp_sk_init(struct net *net) |
2418 | { | |
fee83d09 | 2419 | int res, cpu, cnt; |
bdbbb852 ED |
2420 | |
2421 | net->ipv4.tcp_sk = alloc_percpu(struct sock *); | |
2422 | if (!net->ipv4.tcp_sk) | |
2423 | return -ENOMEM; | |
2424 | ||
2425 | for_each_possible_cpu(cpu) { | |
2426 | struct sock *sk; | |
2427 | ||
2428 | res = inet_ctl_sock_create(&sk, PF_INET, SOCK_RAW, | |
2429 | IPPROTO_TCP, net); | |
2430 | if (res) | |
2431 | goto fail; | |
a9d6532b | 2432 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); |
bdbbb852 ED |
2433 | *per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk; |
2434 | } | |
49213555 | 2435 | |
5d134f1c | 2436 | net->ipv4.sysctl_tcp_ecn = 2; |
49213555 DB |
2437 | net->ipv4.sysctl_tcp_ecn_fallback = 1; |
2438 | ||
b0f9ca53 | 2439 | net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS; |
6b58e0a5 | 2440 | net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD; |
05cbc0db | 2441 | net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL; |
046ee902 | 2442 | |
13b287e8 | 2443 | net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME; |
9bd6861b | 2444 | net->ipv4.sysctl_tcp_keepalive_probes = TCP_KEEPALIVE_PROBES; |
b840d15d | 2445 | net->ipv4.sysctl_tcp_keepalive_intvl = TCP_KEEPALIVE_INTVL; |
13b287e8 | 2446 | |
6fa25166 | 2447 | net->ipv4.sysctl_tcp_syn_retries = TCP_SYN_RETRIES; |
7c083ecb | 2448 | net->ipv4.sysctl_tcp_synack_retries = TCP_SYNACK_RETRIES; |
0aca737d | 2449 | net->ipv4.sysctl_tcp_syncookies = 1; |
1043e25f | 2450 | net->ipv4.sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH; |
ae5c3f40 | 2451 | net->ipv4.sysctl_tcp_retries1 = TCP_RETR1; |
c6214a97 | 2452 | net->ipv4.sysctl_tcp_retries2 = TCP_RETR2; |
c402d9be | 2453 | net->ipv4.sysctl_tcp_orphan_retries = 0; |
1e579caa | 2454 | net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT; |
4979f2d9 | 2455 | net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX; |
56ab6b93 | 2456 | net->ipv4.sysctl_tcp_tw_reuse = 0; |
12ed8244 | 2457 | |
fee83d09 | 2458 | cnt = tcp_hashinfo.ehash_mask + 1; |
1946e672 | 2459 | net->ipv4.tcp_death_row.sysctl_tw_recycle = 0; |
fee83d09 | 2460 | net->ipv4.tcp_death_row.sysctl_max_tw_buckets = (cnt + 1) / 2; |
1946e672 HY |
2461 | net->ipv4.tcp_death_row.hashinfo = &tcp_hashinfo; |
2462 | ||
fee83d09 HY |
2463 | net->ipv4.sysctl_max_syn_backlog = max(128, cnt / 256); |
2464 | ||
49213555 | 2465 | return 0; |
bdbbb852 ED |
2466 | fail: |
2467 | tcp_sk_exit(net); | |
2468 | ||
2469 | return res; | |
b099ce26 EB |
2470 | } |
2471 | ||
2472 | static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list) | |
2473 | { | |
1946e672 | 2474 | inet_twsk_purge(&tcp_hashinfo, AF_INET); |
046ee902 DL |
2475 | } |
2476 | ||
2477 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
b099ce26 EB |
2478 | .init = tcp_sk_init, |
2479 | .exit = tcp_sk_exit, | |
2480 | .exit_batch = tcp_sk_exit_batch, | |
046ee902 DL |
2481 | }; |
2482 | ||
9b0f976f | 2483 | void __init tcp_v4_init(void) |
1da177e4 | 2484 | { |
6a1b3054 | 2485 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2486 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 | 2487 | } |