]>
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> |
1a2449a8 | 75 | #include <net/netdma.h> |
6e5714ea | 76 | #include <net/secure_seq.h> |
d1a4c0b3 | 77 | #include <net/tcp_memcontrol.h> |
076bb0c8 | 78 | #include <net/busy_poll.h> |
1da177e4 LT |
79 | |
80 | #include <linux/inet.h> | |
81 | #include <linux/ipv6.h> | |
82 | #include <linux/stddef.h> | |
83 | #include <linux/proc_fs.h> | |
84 | #include <linux/seq_file.h> | |
85 | ||
cfb6eeb4 YH |
86 | #include <linux/crypto.h> |
87 | #include <linux/scatterlist.h> | |
88 | ||
ab32ea5d BH |
89 | int sysctl_tcp_tw_reuse __read_mostly; |
90 | int sysctl_tcp_low_latency __read_mostly; | |
4bc2f18b | 91 | EXPORT_SYMBOL(sysctl_tcp_low_latency); |
1da177e4 | 92 | |
cfb6eeb4 | 93 | #ifdef CONFIG_TCP_MD5SIG |
a915da9b | 94 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 95 | __be32 daddr, __be32 saddr, const struct tcphdr *th); |
cfb6eeb4 YH |
96 | #endif |
97 | ||
5caea4ea | 98 | struct inet_hashinfo tcp_hashinfo; |
4bc2f18b | 99 | EXPORT_SYMBOL(tcp_hashinfo); |
1da177e4 | 100 | |
936b8bdb | 101 | static __u32 tcp_v4_init_sequence(const struct sk_buff *skb) |
1da177e4 | 102 | { |
eddc9ec5 ACM |
103 | return secure_tcp_sequence_number(ip_hdr(skb)->daddr, |
104 | ip_hdr(skb)->saddr, | |
aa8223c7 ACM |
105 | tcp_hdr(skb)->dest, |
106 | tcp_hdr(skb)->source); | |
1da177e4 LT |
107 | } |
108 | ||
6d6ee43e ACM |
109 | int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp) |
110 | { | |
111 | const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw); | |
112 | struct tcp_sock *tp = tcp_sk(sk); | |
113 | ||
114 | /* With PAWS, it is safe from the viewpoint | |
115 | of data integrity. Even without PAWS it is safe provided sequence | |
116 | spaces do not overlap i.e. at data rates <= 80Mbit/sec. | |
117 | ||
118 | Actually, the idea is close to VJ's one, only timestamp cache is | |
119 | held not per host, but per port pair and TW bucket is used as state | |
120 | holder. | |
121 | ||
122 | If TW bucket has been already destroyed we fall back to VJ's scheme | |
123 | and use initial timestamp retrieved from peer table. | |
124 | */ | |
125 | if (tcptw->tw_ts_recent_stamp && | |
126 | (twp == NULL || (sysctl_tcp_tw_reuse && | |
9d729f72 | 127 | get_seconds() - tcptw->tw_ts_recent_stamp > 1))) { |
6d6ee43e ACM |
128 | tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2; |
129 | if (tp->write_seq == 0) | |
130 | tp->write_seq = 1; | |
131 | tp->rx_opt.ts_recent = tcptw->tw_ts_recent; | |
132 | tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp; | |
133 | sock_hold(sktw); | |
134 | return 1; | |
135 | } | |
136 | ||
137 | return 0; | |
138 | } | |
6d6ee43e ACM |
139 | EXPORT_SYMBOL_GPL(tcp_twsk_unique); |
140 | ||
1da177e4 LT |
141 | /* This will initiate an outgoing connection. */ |
142 | int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
143 | { | |
2d7192d6 | 144 | struct sockaddr_in *usin = (struct sockaddr_in *)uaddr; |
1da177e4 LT |
145 | struct inet_sock *inet = inet_sk(sk); |
146 | struct tcp_sock *tp = tcp_sk(sk); | |
dca8b089 | 147 | __be16 orig_sport, orig_dport; |
bada8adc | 148 | __be32 daddr, nexthop; |
da905bd1 | 149 | struct flowi4 *fl4; |
2d7192d6 | 150 | struct rtable *rt; |
1da177e4 | 151 | int err; |
f6d8bd05 | 152 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
153 | |
154 | if (addr_len < sizeof(struct sockaddr_in)) | |
155 | return -EINVAL; | |
156 | ||
157 | if (usin->sin_family != AF_INET) | |
158 | return -EAFNOSUPPORT; | |
159 | ||
160 | nexthop = daddr = usin->sin_addr.s_addr; | |
f6d8bd05 ED |
161 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
162 | sock_owned_by_user(sk)); | |
163 | if (inet_opt && inet_opt->opt.srr) { | |
1da177e4 LT |
164 | if (!daddr) |
165 | return -EINVAL; | |
f6d8bd05 | 166 | nexthop = inet_opt->opt.faddr; |
1da177e4 LT |
167 | } |
168 | ||
dca8b089 DM |
169 | orig_sport = inet->inet_sport; |
170 | orig_dport = usin->sin_port; | |
da905bd1 DM |
171 | fl4 = &inet->cork.fl.u.ip4; |
172 | rt = ip_route_connect(fl4, nexthop, inet->inet_saddr, | |
b23dd4fe DM |
173 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if, |
174 | IPPROTO_TCP, | |
0e0d44ab | 175 | orig_sport, orig_dport, sk); |
b23dd4fe DM |
176 | if (IS_ERR(rt)) { |
177 | err = PTR_ERR(rt); | |
178 | if (err == -ENETUNREACH) | |
f1d8cba6 | 179 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); |
b23dd4fe | 180 | return err; |
584bdf8c | 181 | } |
1da177e4 LT |
182 | |
183 | if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) { | |
184 | ip_rt_put(rt); | |
185 | return -ENETUNREACH; | |
186 | } | |
187 | ||
f6d8bd05 | 188 | if (!inet_opt || !inet_opt->opt.srr) |
da905bd1 | 189 | daddr = fl4->daddr; |
1da177e4 | 190 | |
c720c7e8 | 191 | if (!inet->inet_saddr) |
da905bd1 | 192 | inet->inet_saddr = fl4->saddr; |
c720c7e8 | 193 | inet->inet_rcv_saddr = inet->inet_saddr; |
1da177e4 | 194 | |
c720c7e8 | 195 | if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) { |
1da177e4 LT |
196 | /* Reset inherited state */ |
197 | tp->rx_opt.ts_recent = 0; | |
198 | tp->rx_opt.ts_recent_stamp = 0; | |
ee995283 PE |
199 | if (likely(!tp->repair)) |
200 | tp->write_seq = 0; | |
1da177e4 LT |
201 | } |
202 | ||
295ff7ed | 203 | if (tcp_death_row.sysctl_tw_recycle && |
81166dd6 DM |
204 | !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr) |
205 | tcp_fetch_timewait_stamp(sk, &rt->dst); | |
1da177e4 | 206 | |
c720c7e8 ED |
207 | inet->inet_dport = usin->sin_port; |
208 | inet->inet_daddr = daddr; | |
1da177e4 | 209 | |
b73c3d0e TH |
210 | inet_set_txhash(sk); |
211 | ||
d83d8461 | 212 | inet_csk(sk)->icsk_ext_hdr_len = 0; |
f6d8bd05 ED |
213 | if (inet_opt) |
214 | inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
1da177e4 | 215 | |
bee7ca9e | 216 | tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT; |
1da177e4 LT |
217 | |
218 | /* Socket identity is still unknown (sport may be zero). | |
219 | * However we set state to SYN-SENT and not releasing socket | |
220 | * lock select source port, enter ourselves into the hash tables and | |
221 | * complete initialization after this. | |
222 | */ | |
223 | tcp_set_state(sk, TCP_SYN_SENT); | |
a7f5e7f1 | 224 | err = inet_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
225 | if (err) |
226 | goto failure; | |
227 | ||
da905bd1 | 228 | rt = ip_route_newports(fl4, rt, orig_sport, orig_dport, |
b23dd4fe DM |
229 | inet->inet_sport, inet->inet_dport, sk); |
230 | if (IS_ERR(rt)) { | |
231 | err = PTR_ERR(rt); | |
232 | rt = NULL; | |
1da177e4 | 233 | goto failure; |
b23dd4fe | 234 | } |
1da177e4 | 235 | /* OK, now commit destination to socket. */ |
bcd76111 | 236 | sk->sk_gso_type = SKB_GSO_TCPV4; |
d8d1f30b | 237 | sk_setup_caps(sk, &rt->dst); |
1da177e4 | 238 | |
ee995283 | 239 | if (!tp->write_seq && likely(!tp->repair)) |
c720c7e8 ED |
240 | tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr, |
241 | inet->inet_daddr, | |
242 | inet->inet_sport, | |
1da177e4 LT |
243 | usin->sin_port); |
244 | ||
c720c7e8 | 245 | inet->inet_id = tp->write_seq ^ jiffies; |
1da177e4 | 246 | |
2b916477 | 247 | err = tcp_connect(sk); |
ee995283 | 248 | |
1da177e4 LT |
249 | rt = NULL; |
250 | if (err) | |
251 | goto failure; | |
252 | ||
253 | return 0; | |
254 | ||
255 | failure: | |
7174259e ACM |
256 | /* |
257 | * This unhashes the socket and releases the local port, | |
258 | * if necessary. | |
259 | */ | |
1da177e4 LT |
260 | tcp_set_state(sk, TCP_CLOSE); |
261 | ip_rt_put(rt); | |
262 | sk->sk_route_caps = 0; | |
c720c7e8 | 263 | inet->inet_dport = 0; |
1da177e4 LT |
264 | return err; |
265 | } | |
4bc2f18b | 266 | EXPORT_SYMBOL(tcp_v4_connect); |
1da177e4 | 267 | |
1da177e4 | 268 | /* |
563d34d0 ED |
269 | * This routine reacts to ICMP_FRAG_NEEDED mtu indications as defined in RFC1191. |
270 | * It can be called through tcp_release_cb() if socket was owned by user | |
271 | * at the time tcp_v4_err() was called to handle ICMP message. | |
1da177e4 | 272 | */ |
4fab9071 | 273 | void tcp_v4_mtu_reduced(struct sock *sk) |
1da177e4 LT |
274 | { |
275 | struct dst_entry *dst; | |
276 | struct inet_sock *inet = inet_sk(sk); | |
563d34d0 | 277 | u32 mtu = tcp_sk(sk)->mtu_info; |
1da177e4 | 278 | |
80d0a69f DM |
279 | dst = inet_csk_update_pmtu(sk, mtu); |
280 | if (!dst) | |
1da177e4 LT |
281 | return; |
282 | ||
1da177e4 LT |
283 | /* Something is about to be wrong... Remember soft error |
284 | * for the case, if this connection will not able to recover. | |
285 | */ | |
286 | if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst)) | |
287 | sk->sk_err_soft = EMSGSIZE; | |
288 | ||
289 | mtu = dst_mtu(dst); | |
290 | ||
291 | if (inet->pmtudisc != IP_PMTUDISC_DONT && | |
482fc609 | 292 | ip_sk_accept_pmtu(sk) && |
d83d8461 | 293 | inet_csk(sk)->icsk_pmtu_cookie > mtu) { |
1da177e4 LT |
294 | tcp_sync_mss(sk, mtu); |
295 | ||
296 | /* Resend the TCP packet because it's | |
297 | * clear that the old packet has been | |
298 | * dropped. This is the new "fast" path mtu | |
299 | * discovery. | |
300 | */ | |
301 | tcp_simple_retransmit(sk); | |
302 | } /* else let the usual retransmit timer handle it */ | |
303 | } | |
4fab9071 | 304 | EXPORT_SYMBOL(tcp_v4_mtu_reduced); |
1da177e4 | 305 | |
55be7a9c DM |
306 | static void do_redirect(struct sk_buff *skb, struct sock *sk) |
307 | { | |
308 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
309 | ||
1ed5c48f | 310 | if (dst) |
6700c270 | 311 | dst->ops->redirect(dst, sk, skb); |
55be7a9c DM |
312 | } |
313 | ||
1da177e4 LT |
314 | /* |
315 | * This routine is called by the ICMP module when it gets some | |
316 | * sort of error condition. If err < 0 then the socket should | |
317 | * be closed and the error returned to the user. If err > 0 | |
318 | * it's just the icmp type << 8 | icmp code. After adjustment | |
319 | * header points to the first 8 bytes of the tcp header. We need | |
320 | * to find the appropriate port. | |
321 | * | |
322 | * The locking strategy used here is very "optimistic". When | |
323 | * someone else accesses the socket the ICMP is just dropped | |
324 | * and for some paths there is no check at all. | |
325 | * A more general error queue to queue errors for later handling | |
326 | * is probably better. | |
327 | * | |
328 | */ | |
329 | ||
4d1a2d9e | 330 | void tcp_v4_err(struct sk_buff *icmp_skb, u32 info) |
1da177e4 | 331 | { |
b71d1d42 | 332 | const struct iphdr *iph = (const struct iphdr *)icmp_skb->data; |
4d1a2d9e | 333 | struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2)); |
f1ecd5d9 | 334 | struct inet_connection_sock *icsk; |
1da177e4 LT |
335 | struct tcp_sock *tp; |
336 | struct inet_sock *inet; | |
4d1a2d9e DL |
337 | const int type = icmp_hdr(icmp_skb)->type; |
338 | const int code = icmp_hdr(icmp_skb)->code; | |
1da177e4 | 339 | struct sock *sk; |
f1ecd5d9 | 340 | struct sk_buff *skb; |
0a672f74 YC |
341 | struct request_sock *fastopen; |
342 | __u32 seq, snd_una; | |
f1ecd5d9 | 343 | __u32 remaining; |
1da177e4 | 344 | int err; |
4d1a2d9e | 345 | struct net *net = dev_net(icmp_skb->dev); |
1da177e4 | 346 | |
fd54d716 | 347 | sk = inet_lookup(net, &tcp_hashinfo, iph->daddr, th->dest, |
4d1a2d9e | 348 | iph->saddr, th->source, inet_iif(icmp_skb)); |
1da177e4 | 349 | if (!sk) { |
dcfc23ca | 350 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
1da177e4 LT |
351 | return; |
352 | } | |
353 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 354 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
355 | return; |
356 | } | |
357 | ||
358 | bh_lock_sock(sk); | |
359 | /* If too many ICMPs get dropped on busy | |
360 | * servers this needs to be solved differently. | |
563d34d0 ED |
361 | * We do take care of PMTU discovery (RFC1191) special case : |
362 | * we can receive locally generated ICMP messages while socket is held. | |
1da177e4 | 363 | */ |
b74aa930 ED |
364 | if (sock_owned_by_user(sk)) { |
365 | if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)) | |
366 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); | |
367 | } | |
1da177e4 LT |
368 | if (sk->sk_state == TCP_CLOSE) |
369 | goto out; | |
370 | ||
97e3ecd1 | 371 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
372 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
373 | goto out; | |
374 | } | |
375 | ||
f1ecd5d9 | 376 | icsk = inet_csk(sk); |
1da177e4 LT |
377 | tp = tcp_sk(sk); |
378 | seq = ntohl(th->seq); | |
0a672f74 YC |
379 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ |
380 | fastopen = tp->fastopen_rsk; | |
381 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; | |
1da177e4 | 382 | if (sk->sk_state != TCP_LISTEN && |
0a672f74 | 383 | !between(seq, snd_una, tp->snd_nxt)) { |
de0744af | 384 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
385 | goto out; |
386 | } | |
387 | ||
388 | switch (type) { | |
55be7a9c DM |
389 | case ICMP_REDIRECT: |
390 | do_redirect(icmp_skb, sk); | |
391 | goto out; | |
1da177e4 LT |
392 | case ICMP_SOURCE_QUENCH: |
393 | /* Just silently ignore these. */ | |
394 | goto out; | |
395 | case ICMP_PARAMETERPROB: | |
396 | err = EPROTO; | |
397 | break; | |
398 | case ICMP_DEST_UNREACH: | |
399 | if (code > NR_ICMP_UNREACH) | |
400 | goto out; | |
401 | ||
402 | if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */ | |
0d4f0608 ED |
403 | /* We are not interested in TCP_LISTEN and open_requests |
404 | * (SYN-ACKs send out by Linux are always <576bytes so | |
405 | * they should go through unfragmented). | |
406 | */ | |
407 | if (sk->sk_state == TCP_LISTEN) | |
408 | goto out; | |
409 | ||
563d34d0 | 410 | tp->mtu_info = info; |
144d56e9 | 411 | if (!sock_owned_by_user(sk)) { |
563d34d0 | 412 | tcp_v4_mtu_reduced(sk); |
144d56e9 ED |
413 | } else { |
414 | if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &tp->tsq_flags)) | |
415 | sock_hold(sk); | |
416 | } | |
1da177e4 LT |
417 | goto out; |
418 | } | |
419 | ||
420 | err = icmp_err_convert[code].errno; | |
f1ecd5d9 DL |
421 | /* check if icmp_skb allows revert of backoff |
422 | * (see draft-zimmermann-tcp-lcd) */ | |
423 | if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH) | |
424 | break; | |
425 | if (seq != tp->snd_una || !icsk->icsk_retransmits || | |
0a672f74 | 426 | !icsk->icsk_backoff || fastopen) |
f1ecd5d9 DL |
427 | break; |
428 | ||
8f49c270 DM |
429 | if (sock_owned_by_user(sk)) |
430 | break; | |
431 | ||
f1ecd5d9 | 432 | icsk->icsk_backoff--; |
740b0f18 | 433 | inet_csk(sk)->icsk_rto = (tp->srtt_us ? __tcp_set_rto(tp) : |
9ad7c049 | 434 | TCP_TIMEOUT_INIT) << icsk->icsk_backoff; |
f1ecd5d9 DL |
435 | tcp_bound_rto(sk); |
436 | ||
437 | skb = tcp_write_queue_head(sk); | |
438 | BUG_ON(!skb); | |
439 | ||
7faee5c0 ED |
440 | remaining = icsk->icsk_rto - |
441 | min(icsk->icsk_rto, | |
442 | tcp_time_stamp - tcp_skb_timestamp(skb)); | |
f1ecd5d9 DL |
443 | |
444 | if (remaining) { | |
445 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, | |
446 | remaining, TCP_RTO_MAX); | |
f1ecd5d9 DL |
447 | } else { |
448 | /* RTO revert clocked out retransmission. | |
449 | * Will retransmit now */ | |
450 | tcp_retransmit_timer(sk); | |
451 | } | |
452 | ||
1da177e4 LT |
453 | break; |
454 | case ICMP_TIME_EXCEEDED: | |
455 | err = EHOSTUNREACH; | |
456 | break; | |
457 | default: | |
458 | goto out; | |
459 | } | |
460 | ||
461 | switch (sk->sk_state) { | |
60236fdd | 462 | struct request_sock *req, **prev; |
1da177e4 LT |
463 | case TCP_LISTEN: |
464 | if (sock_owned_by_user(sk)) | |
465 | goto out; | |
466 | ||
463c84b9 ACM |
467 | req = inet_csk_search_req(sk, &prev, th->dest, |
468 | iph->daddr, iph->saddr); | |
1da177e4 LT |
469 | if (!req) |
470 | goto out; | |
471 | ||
472 | /* ICMPs are not backlogged, hence we cannot get | |
473 | an established socket here. | |
474 | */ | |
547b792c | 475 | WARN_ON(req->sk); |
1da177e4 | 476 | |
2e6599cb | 477 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 478 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
479 | goto out; |
480 | } | |
481 | ||
482 | /* | |
483 | * Still in SYN_RECV, just remove it silently. | |
484 | * There is no good way to pass the error to the newly | |
485 | * created socket, and POSIX does not want network | |
486 | * errors returned from accept(). | |
487 | */ | |
463c84b9 | 488 | inet_csk_reqsk_queue_drop(sk, req, prev); |
848bf15f | 489 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
490 | goto out; |
491 | ||
492 | case TCP_SYN_SENT: | |
0a672f74 YC |
493 | case TCP_SYN_RECV: |
494 | /* Only in fast or simultaneous open. If a fast open socket is | |
495 | * is already accepted it is treated as a connected one below. | |
496 | */ | |
497 | if (fastopen && fastopen->sk == NULL) | |
498 | break; | |
499 | ||
1da177e4 | 500 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
501 | sk->sk_err = err; |
502 | ||
503 | sk->sk_error_report(sk); | |
504 | ||
505 | tcp_done(sk); | |
506 | } else { | |
507 | sk->sk_err_soft = err; | |
508 | } | |
509 | goto out; | |
510 | } | |
511 | ||
512 | /* If we've already connected we will keep trying | |
513 | * until we time out, or the user gives up. | |
514 | * | |
515 | * rfc1122 4.2.3.9 allows to consider as hard errors | |
516 | * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too, | |
517 | * but it is obsoleted by pmtu discovery). | |
518 | * | |
519 | * Note, that in modern internet, where routing is unreliable | |
520 | * and in each dark corner broken firewalls sit, sending random | |
521 | * errors ordered by their masters even this two messages finally lose | |
522 | * their original sense (even Linux sends invalid PORT_UNREACHs) | |
523 | * | |
524 | * Now we are in compliance with RFCs. | |
525 | * --ANK (980905) | |
526 | */ | |
527 | ||
528 | inet = inet_sk(sk); | |
529 | if (!sock_owned_by_user(sk) && inet->recverr) { | |
530 | sk->sk_err = err; | |
531 | sk->sk_error_report(sk); | |
532 | } else { /* Only an error on timeout */ | |
533 | sk->sk_err_soft = err; | |
534 | } | |
535 | ||
536 | out: | |
537 | bh_unlock_sock(sk); | |
538 | sock_put(sk); | |
539 | } | |
540 | ||
28850dc7 | 541 | void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr) |
1da177e4 | 542 | { |
aa8223c7 | 543 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 544 | |
84fa7933 | 545 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
419f9f89 | 546 | th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0); |
663ead3b | 547 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 548 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 549 | } else { |
419f9f89 | 550 | th->check = tcp_v4_check(skb->len, saddr, daddr, |
07f0757a | 551 | csum_partial(th, |
1da177e4 LT |
552 | th->doff << 2, |
553 | skb->csum)); | |
554 | } | |
555 | } | |
556 | ||
419f9f89 | 557 | /* This routine computes an IPv4 TCP checksum. */ |
bb296246 | 558 | void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb) |
419f9f89 | 559 | { |
cf533ea5 | 560 | const struct inet_sock *inet = inet_sk(sk); |
419f9f89 HX |
561 | |
562 | __tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr); | |
563 | } | |
4bc2f18b | 564 | EXPORT_SYMBOL(tcp_v4_send_check); |
419f9f89 | 565 | |
1da177e4 LT |
566 | /* |
567 | * This routine will send an RST to the other tcp. | |
568 | * | |
569 | * Someone asks: why I NEVER use socket parameters (TOS, TTL etc.) | |
570 | * for reset. | |
571 | * Answer: if a packet caused RST, it is not for a socket | |
572 | * existing in our system, if it is matched to a socket, | |
573 | * it is just duplicate segment or bug in other side's TCP. | |
574 | * So that we build reply only basing on parameters | |
575 | * arrived with segment. | |
576 | * Exception: precedence violation. We do not implement it in any case. | |
577 | */ | |
578 | ||
cfb6eeb4 | 579 | static void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 580 | { |
cf533ea5 | 581 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
582 | struct { |
583 | struct tcphdr th; | |
584 | #ifdef CONFIG_TCP_MD5SIG | |
714e85be | 585 | __be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)]; |
cfb6eeb4 YH |
586 | #endif |
587 | } rep; | |
1da177e4 | 588 | struct ip_reply_arg arg; |
cfb6eeb4 YH |
589 | #ifdef CONFIG_TCP_MD5SIG |
590 | struct tcp_md5sig_key *key; | |
658ddaaf SL |
591 | const __u8 *hash_location = NULL; |
592 | unsigned char newhash[16]; | |
593 | int genhash; | |
594 | struct sock *sk1 = NULL; | |
cfb6eeb4 | 595 | #endif |
a86b1e30 | 596 | struct net *net; |
1da177e4 LT |
597 | |
598 | /* Never send a reset in response to a reset. */ | |
599 | if (th->rst) | |
600 | return; | |
601 | ||
511c3f92 | 602 | if (skb_rtable(skb)->rt_type != RTN_LOCAL) |
1da177e4 LT |
603 | return; |
604 | ||
605 | /* Swap the send and the receive. */ | |
cfb6eeb4 YH |
606 | memset(&rep, 0, sizeof(rep)); |
607 | rep.th.dest = th->source; | |
608 | rep.th.source = th->dest; | |
609 | rep.th.doff = sizeof(struct tcphdr) / 4; | |
610 | rep.th.rst = 1; | |
1da177e4 LT |
611 | |
612 | if (th->ack) { | |
cfb6eeb4 | 613 | rep.th.seq = th->ack_seq; |
1da177e4 | 614 | } else { |
cfb6eeb4 YH |
615 | rep.th.ack = 1; |
616 | rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin + | |
617 | skb->len - (th->doff << 2)); | |
1da177e4 LT |
618 | } |
619 | ||
7174259e | 620 | memset(&arg, 0, sizeof(arg)); |
cfb6eeb4 YH |
621 | arg.iov[0].iov_base = (unsigned char *)&rep; |
622 | arg.iov[0].iov_len = sizeof(rep.th); | |
623 | ||
624 | #ifdef CONFIG_TCP_MD5SIG | |
658ddaaf SL |
625 | hash_location = tcp_parse_md5sig_option(th); |
626 | if (!sk && hash_location) { | |
627 | /* | |
628 | * active side is lost. Try to find listening socket through | |
629 | * source port, and then find md5 key through listening socket. | |
630 | * we are not loose security here: | |
631 | * Incoming packet is checked with md5 hash with finding key, | |
632 | * no RST generated if md5 hash doesn't match. | |
633 | */ | |
634 | sk1 = __inet_lookup_listener(dev_net(skb_dst(skb)->dev), | |
da5e3630 TH |
635 | &tcp_hashinfo, ip_hdr(skb)->saddr, |
636 | th->source, ip_hdr(skb)->daddr, | |
658ddaaf SL |
637 | ntohs(th->source), inet_iif(skb)); |
638 | /* don't send rst if it can't find key */ | |
639 | if (!sk1) | |
640 | return; | |
641 | rcu_read_lock(); | |
642 | key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *) | |
643 | &ip_hdr(skb)->saddr, AF_INET); | |
644 | if (!key) | |
645 | goto release_sk1; | |
646 | ||
647 | genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, NULL, skb); | |
648 | if (genhash || memcmp(hash_location, newhash, 16) != 0) | |
649 | goto release_sk1; | |
650 | } else { | |
651 | key = sk ? tcp_md5_do_lookup(sk, (union tcp_md5_addr *) | |
652 | &ip_hdr(skb)->saddr, | |
653 | AF_INET) : NULL; | |
654 | } | |
655 | ||
cfb6eeb4 YH |
656 | if (key) { |
657 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | | |
658 | (TCPOPT_NOP << 16) | | |
659 | (TCPOPT_MD5SIG << 8) | | |
660 | TCPOLEN_MD5SIG); | |
661 | /* Update length and the length the header thinks exists */ | |
662 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
663 | rep.th.doff = arg.iov[0].iov_len / 4; | |
664 | ||
49a72dfb | 665 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1], |
78e645cb IJ |
666 | key, ip_hdr(skb)->saddr, |
667 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
668 | } |
669 | #endif | |
eddc9ec5 ACM |
670 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
671 | ip_hdr(skb)->saddr, /* XXX */ | |
52cd5750 | 672 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
1da177e4 | 673 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; |
88ef4a5a | 674 | arg.flags = (sk && inet_sk(sk)->transparent) ? IP_REPLY_ARG_NOSRCCHECK : 0; |
e2446eaa | 675 | /* When socket is gone, all binding information is lost. |
4c675258 AK |
676 | * routing might fail in this case. No choice here, if we choose to force |
677 | * input interface, we will misroute in case of asymmetric route. | |
e2446eaa | 678 | */ |
4c675258 AK |
679 | if (sk) |
680 | arg.bound_dev_if = sk->sk_bound_dev_if; | |
1da177e4 | 681 | |
adf30907 | 682 | net = dev_net(skb_dst(skb)->dev); |
66b13d99 | 683 | arg.tos = ip_hdr(skb)->tos; |
be9f4a44 | 684 | ip_send_unicast_reply(net, skb, ip_hdr(skb)->saddr, |
70e73416 | 685 | ip_hdr(skb)->daddr, &arg, arg.iov[0].iov_len); |
1da177e4 | 686 | |
63231bdd PE |
687 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
688 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
658ddaaf SL |
689 | |
690 | #ifdef CONFIG_TCP_MD5SIG | |
691 | release_sk1: | |
692 | if (sk1) { | |
693 | rcu_read_unlock(); | |
694 | sock_put(sk1); | |
695 | } | |
696 | #endif | |
1da177e4 LT |
697 | } |
698 | ||
699 | /* The code following below sending ACKs in SYN-RECV and TIME-WAIT states | |
700 | outside socket context is ugly, certainly. What can I do? | |
701 | */ | |
702 | ||
9501f972 | 703 | static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack, |
ee684b6f | 704 | u32 win, u32 tsval, u32 tsecr, int oif, |
88ef4a5a | 705 | struct tcp_md5sig_key *key, |
66b13d99 | 706 | int reply_flags, u8 tos) |
1da177e4 | 707 | { |
cf533ea5 | 708 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
709 | struct { |
710 | struct tcphdr th; | |
714e85be | 711 | __be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2) |
cfb6eeb4 | 712 | #ifdef CONFIG_TCP_MD5SIG |
714e85be | 713 | + (TCPOLEN_MD5SIG_ALIGNED >> 2) |
cfb6eeb4 YH |
714 | #endif |
715 | ]; | |
1da177e4 LT |
716 | } rep; |
717 | struct ip_reply_arg arg; | |
adf30907 | 718 | struct net *net = dev_net(skb_dst(skb)->dev); |
1da177e4 LT |
719 | |
720 | memset(&rep.th, 0, sizeof(struct tcphdr)); | |
7174259e | 721 | memset(&arg, 0, sizeof(arg)); |
1da177e4 LT |
722 | |
723 | arg.iov[0].iov_base = (unsigned char *)&rep; | |
724 | arg.iov[0].iov_len = sizeof(rep.th); | |
ee684b6f | 725 | if (tsecr) { |
cfb6eeb4 YH |
726 | rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
727 | (TCPOPT_TIMESTAMP << 8) | | |
728 | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
729 | rep.opt[1] = htonl(tsval); |
730 | rep.opt[2] = htonl(tsecr); | |
cb48cfe8 | 731 | arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED; |
1da177e4 LT |
732 | } |
733 | ||
734 | /* Swap the send and the receive. */ | |
735 | rep.th.dest = th->source; | |
736 | rep.th.source = th->dest; | |
737 | rep.th.doff = arg.iov[0].iov_len / 4; | |
738 | rep.th.seq = htonl(seq); | |
739 | rep.th.ack_seq = htonl(ack); | |
740 | rep.th.ack = 1; | |
741 | rep.th.window = htons(win); | |
742 | ||
cfb6eeb4 | 743 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 744 | if (key) { |
ee684b6f | 745 | int offset = (tsecr) ? 3 : 0; |
cfb6eeb4 YH |
746 | |
747 | rep.opt[offset++] = htonl((TCPOPT_NOP << 24) | | |
748 | (TCPOPT_NOP << 16) | | |
749 | (TCPOPT_MD5SIG << 8) | | |
750 | TCPOLEN_MD5SIG); | |
751 | arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED; | |
752 | rep.th.doff = arg.iov[0].iov_len/4; | |
753 | ||
49a72dfb | 754 | tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset], |
90b7e112 AL |
755 | key, ip_hdr(skb)->saddr, |
756 | ip_hdr(skb)->daddr, &rep.th); | |
cfb6eeb4 YH |
757 | } |
758 | #endif | |
88ef4a5a | 759 | arg.flags = reply_flags; |
eddc9ec5 ACM |
760 | arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr, |
761 | ip_hdr(skb)->saddr, /* XXX */ | |
1da177e4 LT |
762 | arg.iov[0].iov_len, IPPROTO_TCP, 0); |
763 | arg.csumoffset = offsetof(struct tcphdr, check) / 2; | |
9501f972 YH |
764 | if (oif) |
765 | arg.bound_dev_if = oif; | |
66b13d99 | 766 | arg.tos = tos; |
be9f4a44 | 767 | ip_send_unicast_reply(net, skb, ip_hdr(skb)->saddr, |
70e73416 | 768 | ip_hdr(skb)->daddr, &arg, arg.iov[0].iov_len); |
1da177e4 | 769 | |
63231bdd | 770 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
1da177e4 LT |
771 | } |
772 | ||
773 | static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
774 | { | |
8feaf0c0 | 775 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 776 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 777 | |
9501f972 | 778 | tcp_v4_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
7174259e | 779 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
ee684b6f | 780 | tcp_time_stamp + tcptw->tw_ts_offset, |
9501f972 YH |
781 | tcptw->tw_ts_recent, |
782 | tw->tw_bound_dev_if, | |
88ef4a5a | 783 | tcp_twsk_md5_key(tcptw), |
66b13d99 ED |
784 | tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0, |
785 | tw->tw_tos | |
9501f972 | 786 | ); |
1da177e4 | 787 | |
8feaf0c0 | 788 | inet_twsk_put(tw); |
1da177e4 LT |
789 | } |
790 | ||
6edafaaf | 791 | static void tcp_v4_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
7174259e | 792 | struct request_sock *req) |
1da177e4 | 793 | { |
168a8f58 JC |
794 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
795 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
796 | */ | |
797 | tcp_v4_send_ack(skb, (sk->sk_state == TCP_LISTEN) ? | |
798 | tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt, | |
799 | tcp_rsk(req)->rcv_nxt, req->rcv_wnd, | |
ee684b6f | 800 | tcp_time_stamp, |
9501f972 YH |
801 | req->ts_recent, |
802 | 0, | |
a915da9b ED |
803 | tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr, |
804 | AF_INET), | |
66b13d99 ED |
805 | inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0, |
806 | ip_hdr(skb)->tos); | |
1da177e4 LT |
807 | } |
808 | ||
1da177e4 | 809 | /* |
9bf1d83e | 810 | * Send a SYN-ACK after having received a SYN. |
60236fdd | 811 | * This still operates on a request_sock only, not on a big |
1da177e4 LT |
812 | * socket. |
813 | */ | |
72659ecc | 814 | static int tcp_v4_send_synack(struct sock *sk, struct dst_entry *dst, |
d6274bd8 | 815 | struct flowi *fl, |
72659ecc | 816 | struct request_sock *req, |
843f4a55 YC |
817 | u16 queue_mapping, |
818 | struct tcp_fastopen_cookie *foc) | |
1da177e4 | 819 | { |
2e6599cb | 820 | const struct inet_request_sock *ireq = inet_rsk(req); |
6bd023f3 | 821 | struct flowi4 fl4; |
1da177e4 | 822 | int err = -1; |
d41db5af | 823 | struct sk_buff *skb; |
1da177e4 LT |
824 | |
825 | /* First, grab a route. */ | |
ba3f7f04 | 826 | if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL) |
fd80eb94 | 827 | return -1; |
1da177e4 | 828 | |
843f4a55 | 829 | skb = tcp_make_synack(sk, dst, req, foc); |
1da177e4 LT |
830 | |
831 | if (skb) { | |
634fb979 | 832 | __tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr); |
1da177e4 | 833 | |
fff32699 | 834 | skb_set_queue_mapping(skb, queue_mapping); |
634fb979 ED |
835 | err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr, |
836 | ireq->ir_rmt_addr, | |
2e6599cb | 837 | ireq->opt); |
b9df3cb8 | 838 | err = net_xmit_eval(err); |
1da177e4 LT |
839 | } |
840 | ||
1da177e4 LT |
841 | return err; |
842 | } | |
843 | ||
844 | /* | |
60236fdd | 845 | * IPv4 request_sock destructor. |
1da177e4 | 846 | */ |
60236fdd | 847 | static void tcp_v4_reqsk_destructor(struct request_sock *req) |
1da177e4 | 848 | { |
a51482bd | 849 | kfree(inet_rsk(req)->opt); |
1da177e4 LT |
850 | } |
851 | ||
946cedcc | 852 | /* |
a2a385d6 | 853 | * Return true if a syncookie should be sent |
946cedcc | 854 | */ |
a2a385d6 | 855 | bool tcp_syn_flood_action(struct sock *sk, |
946cedcc ED |
856 | const struct sk_buff *skb, |
857 | const char *proto) | |
1da177e4 | 858 | { |
946cedcc | 859 | const char *msg = "Dropping request"; |
a2a385d6 | 860 | bool want_cookie = false; |
946cedcc ED |
861 | struct listen_sock *lopt; |
862 | ||
2a1d4bd4 | 863 | #ifdef CONFIG_SYN_COOKIES |
946cedcc | 864 | if (sysctl_tcp_syncookies) { |
2a1d4bd4 | 865 | msg = "Sending cookies"; |
a2a385d6 | 866 | want_cookie = true; |
946cedcc ED |
867 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES); |
868 | } else | |
80e40daa | 869 | #endif |
946cedcc ED |
870 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP); |
871 | ||
872 | lopt = inet_csk(sk)->icsk_accept_queue.listen_opt; | |
5ad37d5d | 873 | if (!lopt->synflood_warned && sysctl_tcp_syncookies != 2) { |
946cedcc | 874 | lopt->synflood_warned = 1; |
afd46503 | 875 | pr_info("%s: Possible SYN flooding on port %d. %s. Check SNMP counters.\n", |
946cedcc ED |
876 | proto, ntohs(tcp_hdr(skb)->dest), msg); |
877 | } | |
878 | return want_cookie; | |
2a1d4bd4 | 879 | } |
946cedcc | 880 | EXPORT_SYMBOL(tcp_syn_flood_action); |
1da177e4 LT |
881 | |
882 | /* | |
60236fdd | 883 | * Save and compile IPv4 options into the request_sock if needed. |
1da177e4 | 884 | */ |
5dff747b | 885 | static struct ip_options_rcu *tcp_v4_save_options(struct sk_buff *skb) |
1da177e4 | 886 | { |
f6d8bd05 ED |
887 | const struct ip_options *opt = &(IPCB(skb)->opt); |
888 | struct ip_options_rcu *dopt = NULL; | |
1da177e4 LT |
889 | |
890 | if (opt && opt->optlen) { | |
f6d8bd05 ED |
891 | int opt_size = sizeof(*dopt) + opt->optlen; |
892 | ||
1da177e4 LT |
893 | dopt = kmalloc(opt_size, GFP_ATOMIC); |
894 | if (dopt) { | |
f6d8bd05 | 895 | if (ip_options_echo(&dopt->opt, skb)) { |
1da177e4 LT |
896 | kfree(dopt); |
897 | dopt = NULL; | |
898 | } | |
899 | } | |
900 | } | |
901 | return dopt; | |
902 | } | |
903 | ||
cfb6eeb4 YH |
904 | #ifdef CONFIG_TCP_MD5SIG |
905 | /* | |
906 | * RFC2385 MD5 checksumming requires a mapping of | |
907 | * IP address->MD5 Key. | |
908 | * We need to maintain these in the sk structure. | |
909 | */ | |
910 | ||
911 | /* Find the Key structure for an address. */ | |
a915da9b ED |
912 | struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, |
913 | const union tcp_md5_addr *addr, | |
914 | int family) | |
cfb6eeb4 YH |
915 | { |
916 | struct tcp_sock *tp = tcp_sk(sk); | |
a915da9b | 917 | struct tcp_md5sig_key *key; |
a915da9b | 918 | unsigned int size = sizeof(struct in_addr); |
a8afca03 | 919 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 920 | |
a8afca03 ED |
921 | /* caller either holds rcu_read_lock() or socket lock */ |
922 | md5sig = rcu_dereference_check(tp->md5sig_info, | |
b4fb05ea ED |
923 | sock_owned_by_user(sk) || |
924 | lockdep_is_held(&sk->sk_lock.slock)); | |
a8afca03 | 925 | if (!md5sig) |
cfb6eeb4 | 926 | return NULL; |
a915da9b ED |
927 | #if IS_ENABLED(CONFIG_IPV6) |
928 | if (family == AF_INET6) | |
929 | size = sizeof(struct in6_addr); | |
930 | #endif | |
b67bfe0d | 931 | hlist_for_each_entry_rcu(key, &md5sig->head, node) { |
a915da9b ED |
932 | if (key->family != family) |
933 | continue; | |
934 | if (!memcmp(&key->addr, addr, size)) | |
935 | return key; | |
cfb6eeb4 YH |
936 | } |
937 | return NULL; | |
938 | } | |
a915da9b | 939 | EXPORT_SYMBOL(tcp_md5_do_lookup); |
cfb6eeb4 YH |
940 | |
941 | struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, | |
942 | struct sock *addr_sk) | |
943 | { | |
a915da9b ED |
944 | union tcp_md5_addr *addr; |
945 | ||
946 | addr = (union tcp_md5_addr *)&inet_sk(addr_sk)->inet_daddr; | |
947 | return tcp_md5_do_lookup(sk, addr, AF_INET); | |
cfb6eeb4 | 948 | } |
cfb6eeb4 YH |
949 | EXPORT_SYMBOL(tcp_v4_md5_lookup); |
950 | ||
f5b99bcd AB |
951 | static struct tcp_md5sig_key *tcp_v4_reqsk_md5_lookup(struct sock *sk, |
952 | struct request_sock *req) | |
cfb6eeb4 | 953 | { |
a915da9b ED |
954 | union tcp_md5_addr *addr; |
955 | ||
634fb979 | 956 | addr = (union tcp_md5_addr *)&inet_rsk(req)->ir_rmt_addr; |
a915da9b | 957 | return tcp_md5_do_lookup(sk, addr, AF_INET); |
cfb6eeb4 YH |
958 | } |
959 | ||
960 | /* This can be called on a newly created socket, from other files */ | |
a915da9b ED |
961 | int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, |
962 | int family, const u8 *newkey, u8 newkeylen, gfp_t gfp) | |
cfb6eeb4 YH |
963 | { |
964 | /* Add Key to the list */ | |
b0a713e9 | 965 | struct tcp_md5sig_key *key; |
cfb6eeb4 | 966 | struct tcp_sock *tp = tcp_sk(sk); |
a915da9b | 967 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 968 | |
c0353c7b | 969 | key = tcp_md5_do_lookup(sk, addr, family); |
cfb6eeb4 YH |
970 | if (key) { |
971 | /* Pre-existing entry - just update that one. */ | |
a915da9b | 972 | memcpy(key->key, newkey, newkeylen); |
b0a713e9 | 973 | key->keylen = newkeylen; |
a915da9b ED |
974 | return 0; |
975 | } | |
260fcbeb | 976 | |
a8afca03 ED |
977 | md5sig = rcu_dereference_protected(tp->md5sig_info, |
978 | sock_owned_by_user(sk)); | |
a915da9b ED |
979 | if (!md5sig) { |
980 | md5sig = kmalloc(sizeof(*md5sig), gfp); | |
981 | if (!md5sig) | |
cfb6eeb4 | 982 | return -ENOMEM; |
cfb6eeb4 | 983 | |
a915da9b ED |
984 | sk_nocaps_add(sk, NETIF_F_GSO_MASK); |
985 | INIT_HLIST_HEAD(&md5sig->head); | |
a8afca03 | 986 | rcu_assign_pointer(tp->md5sig_info, md5sig); |
a915da9b | 987 | } |
cfb6eeb4 | 988 | |
5f3d9cb2 | 989 | key = sock_kmalloc(sk, sizeof(*key), gfp); |
a915da9b ED |
990 | if (!key) |
991 | return -ENOMEM; | |
71cea17e | 992 | if (!tcp_alloc_md5sig_pool()) { |
5f3d9cb2 | 993 | sock_kfree_s(sk, key, sizeof(*key)); |
a915da9b | 994 | return -ENOMEM; |
cfb6eeb4 | 995 | } |
a915da9b ED |
996 | |
997 | memcpy(key->key, newkey, newkeylen); | |
998 | key->keylen = newkeylen; | |
999 | key->family = family; | |
1000 | memcpy(&key->addr, addr, | |
1001 | (family == AF_INET6) ? sizeof(struct in6_addr) : | |
1002 | sizeof(struct in_addr)); | |
1003 | hlist_add_head_rcu(&key->node, &md5sig->head); | |
cfb6eeb4 YH |
1004 | return 0; |
1005 | } | |
a915da9b | 1006 | EXPORT_SYMBOL(tcp_md5_do_add); |
cfb6eeb4 | 1007 | |
a915da9b | 1008 | int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family) |
cfb6eeb4 | 1009 | { |
a915da9b ED |
1010 | struct tcp_md5sig_key *key; |
1011 | ||
c0353c7b | 1012 | key = tcp_md5_do_lookup(sk, addr, family); |
a915da9b ED |
1013 | if (!key) |
1014 | return -ENOENT; | |
1015 | hlist_del_rcu(&key->node); | |
5f3d9cb2 | 1016 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1017 | kfree_rcu(key, rcu); |
a915da9b | 1018 | return 0; |
cfb6eeb4 | 1019 | } |
a915da9b | 1020 | EXPORT_SYMBOL(tcp_md5_do_del); |
cfb6eeb4 | 1021 | |
e0683e70 | 1022 | static void tcp_clear_md5_list(struct sock *sk) |
cfb6eeb4 YH |
1023 | { |
1024 | struct tcp_sock *tp = tcp_sk(sk); | |
a915da9b | 1025 | struct tcp_md5sig_key *key; |
b67bfe0d | 1026 | struct hlist_node *n; |
a8afca03 | 1027 | struct tcp_md5sig_info *md5sig; |
cfb6eeb4 | 1028 | |
a8afca03 ED |
1029 | md5sig = rcu_dereference_protected(tp->md5sig_info, 1); |
1030 | ||
b67bfe0d | 1031 | hlist_for_each_entry_safe(key, n, &md5sig->head, node) { |
a915da9b | 1032 | hlist_del_rcu(&key->node); |
5f3d9cb2 | 1033 | atomic_sub(sizeof(*key), &sk->sk_omem_alloc); |
a915da9b | 1034 | kfree_rcu(key, rcu); |
cfb6eeb4 YH |
1035 | } |
1036 | } | |
1037 | ||
7174259e ACM |
1038 | static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval, |
1039 | int optlen) | |
cfb6eeb4 YH |
1040 | { |
1041 | struct tcp_md5sig cmd; | |
1042 | struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr; | |
cfb6eeb4 YH |
1043 | |
1044 | if (optlen < sizeof(cmd)) | |
1045 | return -EINVAL; | |
1046 | ||
7174259e | 1047 | if (copy_from_user(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
1048 | return -EFAULT; |
1049 | ||
1050 | if (sin->sin_family != AF_INET) | |
1051 | return -EINVAL; | |
1052 | ||
64a124ed | 1053 | if (!cmd.tcpm_keylen) |
a915da9b ED |
1054 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
1055 | AF_INET); | |
cfb6eeb4 YH |
1056 | |
1057 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
1058 | return -EINVAL; | |
1059 | ||
a915da9b ED |
1060 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr, |
1061 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, | |
1062 | GFP_KERNEL); | |
cfb6eeb4 YH |
1063 | } |
1064 | ||
49a72dfb AL |
1065 | static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
1066 | __be32 daddr, __be32 saddr, int nbytes) | |
cfb6eeb4 | 1067 | { |
cfb6eeb4 | 1068 | struct tcp4_pseudohdr *bp; |
49a72dfb | 1069 | struct scatterlist sg; |
cfb6eeb4 YH |
1070 | |
1071 | bp = &hp->md5_blk.ip4; | |
cfb6eeb4 YH |
1072 | |
1073 | /* | |
49a72dfb | 1074 | * 1. the TCP pseudo-header (in the order: source IP address, |
cfb6eeb4 YH |
1075 | * destination IP address, zero-padded protocol number, and |
1076 | * segment length) | |
1077 | */ | |
1078 | bp->saddr = saddr; | |
1079 | bp->daddr = daddr; | |
1080 | bp->pad = 0; | |
076fb722 | 1081 | bp->protocol = IPPROTO_TCP; |
49a72dfb | 1082 | bp->len = cpu_to_be16(nbytes); |
c7da57a1 | 1083 | |
49a72dfb AL |
1084 | sg_init_one(&sg, bp, sizeof(*bp)); |
1085 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
1086 | } | |
1087 | ||
a915da9b | 1088 | static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
318cf7aa | 1089 | __be32 daddr, __be32 saddr, const struct tcphdr *th) |
49a72dfb AL |
1090 | { |
1091 | struct tcp_md5sig_pool *hp; | |
1092 | struct hash_desc *desc; | |
1093 | ||
1094 | hp = tcp_get_md5sig_pool(); | |
1095 | if (!hp) | |
1096 | goto clear_hash_noput; | |
1097 | desc = &hp->md5_desc; | |
1098 | ||
1099 | if (crypto_hash_init(desc)) | |
1100 | goto clear_hash; | |
1101 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
1102 | goto clear_hash; | |
1103 | if (tcp_md5_hash_header(hp, th)) | |
1104 | goto clear_hash; | |
1105 | if (tcp_md5_hash_key(hp, key)) | |
1106 | goto clear_hash; | |
1107 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 YH |
1108 | goto clear_hash; |
1109 | ||
cfb6eeb4 | 1110 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 1111 | return 0; |
49a72dfb | 1112 | |
cfb6eeb4 YH |
1113 | clear_hash: |
1114 | tcp_put_md5sig_pool(); | |
1115 | clear_hash_noput: | |
1116 | memset(md5_hash, 0, 16); | |
49a72dfb | 1117 | return 1; |
cfb6eeb4 YH |
1118 | } |
1119 | ||
49a72dfb | 1120 | int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
318cf7aa ED |
1121 | const struct sock *sk, const struct request_sock *req, |
1122 | const struct sk_buff *skb) | |
cfb6eeb4 | 1123 | { |
49a72dfb AL |
1124 | struct tcp_md5sig_pool *hp; |
1125 | struct hash_desc *desc; | |
318cf7aa | 1126 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
1127 | __be32 saddr, daddr; |
1128 | ||
1129 | if (sk) { | |
c720c7e8 ED |
1130 | saddr = inet_sk(sk)->inet_saddr; |
1131 | daddr = inet_sk(sk)->inet_daddr; | |
49a72dfb | 1132 | } else if (req) { |
634fb979 ED |
1133 | saddr = inet_rsk(req)->ir_loc_addr; |
1134 | daddr = inet_rsk(req)->ir_rmt_addr; | |
cfb6eeb4 | 1135 | } else { |
49a72dfb AL |
1136 | const struct iphdr *iph = ip_hdr(skb); |
1137 | saddr = iph->saddr; | |
1138 | daddr = iph->daddr; | |
cfb6eeb4 | 1139 | } |
49a72dfb AL |
1140 | |
1141 | hp = tcp_get_md5sig_pool(); | |
1142 | if (!hp) | |
1143 | goto clear_hash_noput; | |
1144 | desc = &hp->md5_desc; | |
1145 | ||
1146 | if (crypto_hash_init(desc)) | |
1147 | goto clear_hash; | |
1148 | ||
1149 | if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
1150 | goto clear_hash; | |
1151 | if (tcp_md5_hash_header(hp, th)) | |
1152 | goto clear_hash; | |
1153 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
1154 | goto clear_hash; | |
1155 | if (tcp_md5_hash_key(hp, key)) | |
1156 | goto clear_hash; | |
1157 | if (crypto_hash_final(desc, md5_hash)) | |
1158 | goto clear_hash; | |
1159 | ||
1160 | tcp_put_md5sig_pool(); | |
1161 | return 0; | |
1162 | ||
1163 | clear_hash: | |
1164 | tcp_put_md5sig_pool(); | |
1165 | clear_hash_noput: | |
1166 | memset(md5_hash, 0, 16); | |
1167 | return 1; | |
cfb6eeb4 | 1168 | } |
49a72dfb | 1169 | EXPORT_SYMBOL(tcp_v4_md5_hash_skb); |
cfb6eeb4 | 1170 | |
9ea88a15 DP |
1171 | static bool __tcp_v4_inbound_md5_hash(struct sock *sk, |
1172 | const struct sk_buff *skb) | |
cfb6eeb4 YH |
1173 | { |
1174 | /* | |
1175 | * This gets called for each TCP segment that arrives | |
1176 | * so we want to be efficient. | |
1177 | * We have 3 drop cases: | |
1178 | * o No MD5 hash and one expected. | |
1179 | * o MD5 hash and we're not expecting one. | |
1180 | * o MD5 hash and its wrong. | |
1181 | */ | |
cf533ea5 | 1182 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 1183 | struct tcp_md5sig_key *hash_expected; |
eddc9ec5 | 1184 | const struct iphdr *iph = ip_hdr(skb); |
cf533ea5 | 1185 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 1186 | int genhash; |
cfb6eeb4 YH |
1187 | unsigned char newhash[16]; |
1188 | ||
a915da9b ED |
1189 | hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr, |
1190 | AF_INET); | |
7d5d5525 | 1191 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 1192 | |
cfb6eeb4 YH |
1193 | /* We've parsed the options - do we have a hash? */ |
1194 | if (!hash_expected && !hash_location) | |
a2a385d6 | 1195 | return false; |
cfb6eeb4 YH |
1196 | |
1197 | if (hash_expected && !hash_location) { | |
785957d3 | 1198 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); |
a2a385d6 | 1199 | return true; |
cfb6eeb4 YH |
1200 | } |
1201 | ||
1202 | if (!hash_expected && hash_location) { | |
785957d3 | 1203 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); |
a2a385d6 | 1204 | return true; |
cfb6eeb4 YH |
1205 | } |
1206 | ||
1207 | /* Okay, so this is hash_expected and hash_location - | |
1208 | * so we need to calculate the checksum. | |
1209 | */ | |
49a72dfb AL |
1210 | genhash = tcp_v4_md5_hash_skb(newhash, |
1211 | hash_expected, | |
1212 | NULL, NULL, skb); | |
cfb6eeb4 YH |
1213 | |
1214 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
e87cc472 JP |
1215 | net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n", |
1216 | &iph->saddr, ntohs(th->source), | |
1217 | &iph->daddr, ntohs(th->dest), | |
1218 | genhash ? " tcp_v4_calc_md5_hash failed" | |
1219 | : ""); | |
a2a385d6 | 1220 | return true; |
cfb6eeb4 | 1221 | } |
a2a385d6 | 1222 | return false; |
cfb6eeb4 YH |
1223 | } |
1224 | ||
9ea88a15 DP |
1225 | static bool tcp_v4_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb) |
1226 | { | |
1227 | bool ret; | |
1228 | ||
1229 | rcu_read_lock(); | |
1230 | ret = __tcp_v4_inbound_md5_hash(sk, skb); | |
1231 | rcu_read_unlock(); | |
1232 | ||
1233 | return ret; | |
1234 | } | |
1235 | ||
cfb6eeb4 YH |
1236 | #endif |
1237 | ||
16bea70a OP |
1238 | static void tcp_v4_init_req(struct request_sock *req, struct sock *sk, |
1239 | struct sk_buff *skb) | |
1240 | { | |
1241 | struct inet_request_sock *ireq = inet_rsk(req); | |
1242 | ||
1243 | ireq->ir_loc_addr = ip_hdr(skb)->daddr; | |
1244 | ireq->ir_rmt_addr = ip_hdr(skb)->saddr; | |
1245 | ireq->no_srccheck = inet_sk(sk)->transparent; | |
1246 | ireq->opt = tcp_v4_save_options(skb); | |
1247 | } | |
1248 | ||
d94e0417 OP |
1249 | static struct dst_entry *tcp_v4_route_req(struct sock *sk, struct flowi *fl, |
1250 | const struct request_sock *req, | |
1251 | bool *strict) | |
1252 | { | |
1253 | struct dst_entry *dst = inet_csk_route_req(sk, &fl->u.ip4, req); | |
1254 | ||
1255 | if (strict) { | |
1256 | if (fl->u.ip4.daddr == inet_rsk(req)->ir_rmt_addr) | |
1257 | *strict = true; | |
1258 | else | |
1259 | *strict = false; | |
1260 | } | |
1261 | ||
1262 | return dst; | |
1263 | } | |
1264 | ||
72a3effa | 1265 | struct request_sock_ops tcp_request_sock_ops __read_mostly = { |
1da177e4 | 1266 | .family = PF_INET, |
2e6599cb | 1267 | .obj_size = sizeof(struct tcp_request_sock), |
5db92c99 | 1268 | .rtx_syn_ack = tcp_rtx_synack, |
60236fdd ACM |
1269 | .send_ack = tcp_v4_reqsk_send_ack, |
1270 | .destructor = tcp_v4_reqsk_destructor, | |
1da177e4 | 1271 | .send_reset = tcp_v4_send_reset, |
688d1945 | 1272 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
1273 | }; |
1274 | ||
b2e4b3de | 1275 | static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = { |
2aec4a29 | 1276 | .mss_clamp = TCP_MSS_DEFAULT, |
16bea70a | 1277 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 1278 | .md5_lookup = tcp_v4_reqsk_md5_lookup, |
e3afe7b7 | 1279 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
b6332e6c | 1280 | #endif |
16bea70a | 1281 | .init_req = tcp_v4_init_req, |
fb7b37a7 OP |
1282 | #ifdef CONFIG_SYN_COOKIES |
1283 | .cookie_init_seq = cookie_v4_init_sequence, | |
1284 | #endif | |
d94e0417 | 1285 | .route_req = tcp_v4_route_req, |
936b8bdb | 1286 | .init_seq = tcp_v4_init_sequence, |
d6274bd8 | 1287 | .send_synack = tcp_v4_send_synack, |
695da14e | 1288 | .queue_hash_add = inet_csk_reqsk_queue_hash_add, |
16bea70a | 1289 | }; |
cfb6eeb4 | 1290 | |
1da177e4 LT |
1291 | int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb) |
1292 | { | |
1da177e4 | 1293 | /* Never answer to SYNs send to broadcast or multicast */ |
511c3f92 | 1294 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) |
1da177e4 LT |
1295 | goto drop; |
1296 | ||
1fb6f159 OP |
1297 | return tcp_conn_request(&tcp_request_sock_ops, |
1298 | &tcp_request_sock_ipv4_ops, sk, skb); | |
1da177e4 | 1299 | |
1da177e4 | 1300 | drop: |
848bf15f | 1301 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
1302 | return 0; |
1303 | } | |
4bc2f18b | 1304 | EXPORT_SYMBOL(tcp_v4_conn_request); |
1da177e4 LT |
1305 | |
1306 | ||
1307 | /* | |
1308 | * The three way handshake has completed - we got a valid synack - | |
1309 | * now create the new socket. | |
1310 | */ | |
1311 | struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1312 | struct request_sock *req, |
1da177e4 LT |
1313 | struct dst_entry *dst) |
1314 | { | |
2e6599cb | 1315 | struct inet_request_sock *ireq; |
1da177e4 LT |
1316 | struct inet_sock *newinet; |
1317 | struct tcp_sock *newtp; | |
1318 | struct sock *newsk; | |
cfb6eeb4 YH |
1319 | #ifdef CONFIG_TCP_MD5SIG |
1320 | struct tcp_md5sig_key *key; | |
1321 | #endif | |
f6d8bd05 | 1322 | struct ip_options_rcu *inet_opt; |
1da177e4 LT |
1323 | |
1324 | if (sk_acceptq_is_full(sk)) | |
1325 | goto exit_overflow; | |
1326 | ||
1da177e4 LT |
1327 | newsk = tcp_create_openreq_child(sk, req, skb); |
1328 | if (!newsk) | |
093d2823 | 1329 | goto exit_nonewsk; |
1da177e4 | 1330 | |
bcd76111 | 1331 | newsk->sk_gso_type = SKB_GSO_TCPV4; |
fae6ef87 | 1332 | inet_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1333 | |
1334 | newtp = tcp_sk(newsk); | |
1335 | newinet = inet_sk(newsk); | |
2e6599cb | 1336 | ireq = inet_rsk(req); |
634fb979 ED |
1337 | newinet->inet_daddr = ireq->ir_rmt_addr; |
1338 | newinet->inet_rcv_saddr = ireq->ir_loc_addr; | |
1339 | newinet->inet_saddr = ireq->ir_loc_addr; | |
f6d8bd05 ED |
1340 | inet_opt = ireq->opt; |
1341 | rcu_assign_pointer(newinet->inet_opt, inet_opt); | |
2e6599cb | 1342 | ireq->opt = NULL; |
463c84b9 | 1343 | newinet->mc_index = inet_iif(skb); |
eddc9ec5 | 1344 | newinet->mc_ttl = ip_hdr(skb)->ttl; |
4c507d28 | 1345 | newinet->rcv_tos = ip_hdr(skb)->tos; |
d83d8461 | 1346 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
b73c3d0e | 1347 | inet_set_txhash(newsk); |
f6d8bd05 ED |
1348 | if (inet_opt) |
1349 | inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen; | |
c720c7e8 | 1350 | newinet->inet_id = newtp->write_seq ^ jiffies; |
1da177e4 | 1351 | |
dfd25fff ED |
1352 | if (!dst) { |
1353 | dst = inet_csk_route_child_sock(sk, newsk, req); | |
1354 | if (!dst) | |
1355 | goto put_and_exit; | |
1356 | } else { | |
1357 | /* syncookie case : see end of cookie_v4_check() */ | |
1358 | } | |
0e734419 DM |
1359 | sk_setup_caps(newsk, dst); |
1360 | ||
1da177e4 | 1361 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1362 | newtp->advmss = dst_metric_advmss(dst); |
f5fff5dc TQ |
1363 | if (tcp_sk(sk)->rx_opt.user_mss && |
1364 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1365 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1366 | ||
1da177e4 LT |
1367 | tcp_initialize_rcv_mss(newsk); |
1368 | ||
cfb6eeb4 YH |
1369 | #ifdef CONFIG_TCP_MD5SIG |
1370 | /* Copy over the MD5 key from the original socket */ | |
a915da9b ED |
1371 | key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1372 | AF_INET); | |
c720c7e8 | 1373 | if (key != NULL) { |
cfb6eeb4 YH |
1374 | /* |
1375 | * We're using one, so create a matching key | |
1376 | * on the newsk structure. If we fail to get | |
1377 | * memory, then we end up not copying the key | |
1378 | * across. Shucks. | |
1379 | */ | |
a915da9b ED |
1380 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newinet->inet_daddr, |
1381 | AF_INET, key->key, key->keylen, GFP_ATOMIC); | |
a465419b | 1382 | sk_nocaps_add(newsk, NETIF_F_GSO_MASK); |
cfb6eeb4 YH |
1383 | } |
1384 | #endif | |
1385 | ||
0e734419 DM |
1386 | if (__inet_inherit_port(sk, newsk) < 0) |
1387 | goto put_and_exit; | |
9327f705 | 1388 | __inet_hash_nolisten(newsk, NULL); |
1da177e4 LT |
1389 | |
1390 | return newsk; | |
1391 | ||
1392 | exit_overflow: | |
de0744af | 1393 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 BS |
1394 | exit_nonewsk: |
1395 | dst_release(dst); | |
1da177e4 | 1396 | exit: |
de0744af | 1397 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 | 1398 | return NULL; |
0e734419 | 1399 | put_and_exit: |
e337e24d CP |
1400 | inet_csk_prepare_forced_close(newsk); |
1401 | tcp_done(newsk); | |
0e734419 | 1402 | goto exit; |
1da177e4 | 1403 | } |
4bc2f18b | 1404 | EXPORT_SYMBOL(tcp_v4_syn_recv_sock); |
1da177e4 LT |
1405 | |
1406 | static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb) | |
1407 | { | |
aa8223c7 | 1408 | struct tcphdr *th = tcp_hdr(skb); |
eddc9ec5 | 1409 | const struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 1410 | struct sock *nsk; |
60236fdd | 1411 | struct request_sock **prev; |
1da177e4 | 1412 | /* Find possible connection requests. */ |
463c84b9 ACM |
1413 | struct request_sock *req = inet_csk_search_req(sk, &prev, th->source, |
1414 | iph->saddr, iph->daddr); | |
1da177e4 | 1415 | if (req) |
8336886f | 1416 | return tcp_check_req(sk, skb, req, prev, false); |
1da177e4 | 1417 | |
3b1e0a65 | 1418 | nsk = inet_lookup_established(sock_net(sk), &tcp_hashinfo, iph->saddr, |
c67499c0 | 1419 | th->source, iph->daddr, th->dest, inet_iif(skb)); |
1da177e4 LT |
1420 | |
1421 | if (nsk) { | |
1422 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
1423 | bh_lock_sock(nsk); | |
1424 | return nsk; | |
1425 | } | |
9469c7b4 | 1426 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
1427 | return NULL; |
1428 | } | |
1429 | ||
1430 | #ifdef CONFIG_SYN_COOKIES | |
af9b4738 | 1431 | if (!th->syn) |
1da177e4 LT |
1432 | sk = cookie_v4_check(sk, skb, &(IPCB(skb)->opt)); |
1433 | #endif | |
1434 | return sk; | |
1435 | } | |
1436 | ||
1da177e4 LT |
1437 | /* The socket must have it's spinlock held when we get |
1438 | * here. | |
1439 | * | |
1440 | * We have a potential double-lock case here, so even when | |
1441 | * doing backlog processing we use the BH locking scheme. | |
1442 | * This is because we cannot sleep with the original spinlock | |
1443 | * held. | |
1444 | */ | |
1445 | int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1446 | { | |
cfb6eeb4 | 1447 | struct sock *rsk; |
cfb6eeb4 | 1448 | |
1da177e4 | 1449 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
404e0a8b ED |
1450 | struct dst_entry *dst = sk->sk_rx_dst; |
1451 | ||
bdeab991 | 1452 | sock_rps_save_rxhash(sk, skb); |
404e0a8b | 1453 | if (dst) { |
505fbcf0 ED |
1454 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || |
1455 | dst->ops->check(dst, 0) == NULL) { | |
92101b3b DM |
1456 | dst_release(dst); |
1457 | sk->sk_rx_dst = NULL; | |
1458 | } | |
1459 | } | |
c995ae22 | 1460 | tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len); |
1da177e4 LT |
1461 | return 0; |
1462 | } | |
1463 | ||
ab6a5bb6 | 1464 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1465 | goto csum_err; |
1466 | ||
1467 | if (sk->sk_state == TCP_LISTEN) { | |
1468 | struct sock *nsk = tcp_v4_hnd_req(sk, skb); | |
1469 | if (!nsk) | |
1470 | goto discard; | |
1471 | ||
1472 | if (nsk != sk) { | |
bdeab991 | 1473 | sock_rps_save_rxhash(nsk, skb); |
cfb6eeb4 YH |
1474 | if (tcp_child_process(sk, nsk, skb)) { |
1475 | rsk = nsk; | |
1da177e4 | 1476 | goto reset; |
cfb6eeb4 | 1477 | } |
1da177e4 LT |
1478 | return 0; |
1479 | } | |
ca55158c | 1480 | } else |
bdeab991 | 1481 | sock_rps_save_rxhash(sk, skb); |
ca55158c | 1482 | |
aa8223c7 | 1483 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) { |
cfb6eeb4 | 1484 | rsk = sk; |
1da177e4 | 1485 | goto reset; |
cfb6eeb4 | 1486 | } |
1da177e4 LT |
1487 | return 0; |
1488 | ||
1489 | reset: | |
cfb6eeb4 | 1490 | tcp_v4_send_reset(rsk, skb); |
1da177e4 LT |
1491 | discard: |
1492 | kfree_skb(skb); | |
1493 | /* Be careful here. If this function gets more complicated and | |
1494 | * gcc suffers from register pressure on the x86, sk (in %ebx) | |
1495 | * might be destroyed here. This current version compiles correctly, | |
1496 | * but you have been warned. | |
1497 | */ | |
1498 | return 0; | |
1499 | ||
1500 | csum_err: | |
6a5dc9e5 | 1501 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS); |
63231bdd | 1502 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1503 | goto discard; |
1504 | } | |
4bc2f18b | 1505 | EXPORT_SYMBOL(tcp_v4_do_rcv); |
1da177e4 | 1506 | |
160eb5a6 | 1507 | void tcp_v4_early_demux(struct sk_buff *skb) |
41063e9d | 1508 | { |
41063e9d DM |
1509 | const struct iphdr *iph; |
1510 | const struct tcphdr *th; | |
1511 | struct sock *sk; | |
41063e9d | 1512 | |
41063e9d | 1513 | if (skb->pkt_type != PACKET_HOST) |
160eb5a6 | 1514 | return; |
41063e9d | 1515 | |
45f00f99 | 1516 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) |
160eb5a6 | 1517 | return; |
41063e9d DM |
1518 | |
1519 | iph = ip_hdr(skb); | |
45f00f99 | 1520 | th = tcp_hdr(skb); |
41063e9d DM |
1521 | |
1522 | if (th->doff < sizeof(struct tcphdr) / 4) | |
160eb5a6 | 1523 | return; |
41063e9d | 1524 | |
45f00f99 | 1525 | sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo, |
41063e9d | 1526 | iph->saddr, th->source, |
7011d085 | 1527 | iph->daddr, ntohs(th->dest), |
9cb429d6 | 1528 | skb->skb_iif); |
41063e9d DM |
1529 | if (sk) { |
1530 | skb->sk = sk; | |
1531 | skb->destructor = sock_edemux; | |
1532 | if (sk->sk_state != TCP_TIME_WAIT) { | |
1533 | struct dst_entry *dst = sk->sk_rx_dst; | |
505fbcf0 | 1534 | |
41063e9d DM |
1535 | if (dst) |
1536 | dst = dst_check(dst, 0); | |
92101b3b | 1537 | if (dst && |
505fbcf0 | 1538 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
92101b3b | 1539 | skb_dst_set_noref(skb, dst); |
41063e9d DM |
1540 | } |
1541 | } | |
41063e9d DM |
1542 | } |
1543 | ||
b2fb4f54 ED |
1544 | /* Packet is added to VJ-style prequeue for processing in process |
1545 | * context, if a reader task is waiting. Apparently, this exciting | |
1546 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) | |
1547 | * failed somewhere. Latency? Burstiness? Well, at least now we will | |
1548 | * see, why it failed. 8)8) --ANK | |
1549 | * | |
1550 | */ | |
1551 | bool tcp_prequeue(struct sock *sk, struct sk_buff *skb) | |
1552 | { | |
1553 | struct tcp_sock *tp = tcp_sk(sk); | |
1554 | ||
1555 | if (sysctl_tcp_low_latency || !tp->ucopy.task) | |
1556 | return false; | |
1557 | ||
1558 | if (skb->len <= tcp_hdrlen(skb) && | |
1559 | skb_queue_len(&tp->ucopy.prequeue) == 0) | |
1560 | return false; | |
1561 | ||
ca777eff ED |
1562 | /* Before escaping RCU protected region, we need to take care of skb |
1563 | * dst. Prequeue is only enabled for established sockets. | |
1564 | * For such sockets, we might need the skb dst only to set sk->sk_rx_dst | |
1565 | * Instead of doing full sk_rx_dst validity here, let's perform | |
1566 | * an optimistic check. | |
1567 | */ | |
1568 | if (likely(sk->sk_rx_dst)) | |
1569 | skb_dst_drop(skb); | |
1570 | else | |
1571 | skb_dst_force(skb); | |
1572 | ||
b2fb4f54 ED |
1573 | __skb_queue_tail(&tp->ucopy.prequeue, skb); |
1574 | tp->ucopy.memory += skb->truesize; | |
1575 | if (tp->ucopy.memory > sk->sk_rcvbuf) { | |
1576 | struct sk_buff *skb1; | |
1577 | ||
1578 | BUG_ON(sock_owned_by_user(sk)); | |
1579 | ||
1580 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { | |
1581 | sk_backlog_rcv(sk, skb1); | |
1582 | NET_INC_STATS_BH(sock_net(sk), | |
1583 | LINUX_MIB_TCPPREQUEUEDROPPED); | |
1584 | } | |
1585 | ||
1586 | tp->ucopy.memory = 0; | |
1587 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { | |
1588 | wake_up_interruptible_sync_poll(sk_sleep(sk), | |
1589 | POLLIN | POLLRDNORM | POLLRDBAND); | |
1590 | if (!inet_csk_ack_scheduled(sk)) | |
1591 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
1592 | (3 * tcp_rto_min(sk)) / 4, | |
1593 | TCP_RTO_MAX); | |
1594 | } | |
1595 | return true; | |
1596 | } | |
1597 | EXPORT_SYMBOL(tcp_prequeue); | |
1598 | ||
1da177e4 LT |
1599 | /* |
1600 | * From tcp_input.c | |
1601 | */ | |
1602 | ||
1603 | int tcp_v4_rcv(struct sk_buff *skb) | |
1604 | { | |
eddc9ec5 | 1605 | const struct iphdr *iph; |
cf533ea5 | 1606 | const struct tcphdr *th; |
1da177e4 LT |
1607 | struct sock *sk; |
1608 | int ret; | |
a86b1e30 | 1609 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1610 | |
1611 | if (skb->pkt_type != PACKET_HOST) | |
1612 | goto discard_it; | |
1613 | ||
1614 | /* Count it even if it's bad */ | |
63231bdd | 1615 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1616 | |
1617 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1618 | goto discard_it; | |
1619 | ||
aa8223c7 | 1620 | th = tcp_hdr(skb); |
1da177e4 LT |
1621 | |
1622 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1623 | goto bad_packet; | |
1624 | if (!pskb_may_pull(skb, th->doff * 4)) | |
1625 | goto discard_it; | |
1626 | ||
1627 | /* An explanation is required here, I think. | |
1628 | * Packet length and doff are validated by header prediction, | |
caa20d9a | 1629 | * provided case of th->doff==0 is eliminated. |
1da177e4 | 1630 | * So, we defer the checks. */ |
ed70fcfc TH |
1631 | |
1632 | if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo)) | |
6a5dc9e5 | 1633 | goto csum_error; |
1da177e4 | 1634 | |
aa8223c7 | 1635 | th = tcp_hdr(skb); |
eddc9ec5 | 1636 | iph = ip_hdr(skb); |
1da177e4 LT |
1637 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1638 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1639 | skb->len - th->doff * 4); | |
1640 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
04317daf | 1641 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; |
b82d1bb4 | 1642 | TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph); |
1da177e4 LT |
1643 | TCP_SKB_CB(skb)->sacked = 0; |
1644 | ||
9a1f27c4 | 1645 | sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1646 | if (!sk) |
1647 | goto no_tcp_socket; | |
1648 | ||
bb134d5d ED |
1649 | process: |
1650 | if (sk->sk_state == TCP_TIME_WAIT) | |
1651 | goto do_time_wait; | |
1652 | ||
6cce09f8 ED |
1653 | if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { |
1654 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
d218d111 | 1655 | goto discard_and_relse; |
6cce09f8 | 1656 | } |
d218d111 | 1657 | |
1da177e4 LT |
1658 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1659 | goto discard_and_relse; | |
9ea88a15 DP |
1660 | |
1661 | #ifdef CONFIG_TCP_MD5SIG | |
1662 | /* | |
1663 | * We really want to reject the packet as early as possible | |
1664 | * if: | |
1665 | * o We're expecting an MD5'd packet and this is no MD5 tcp option | |
1666 | * o There is an MD5 option and we're not expecting one | |
1667 | */ | |
1668 | if (tcp_v4_inbound_md5_hash(sk, skb)) | |
1669 | goto discard_and_relse; | |
1670 | #endif | |
1671 | ||
b59c2701 | 1672 | nf_reset(skb); |
1da177e4 | 1673 | |
fda9ef5d | 1674 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1675 | goto discard_and_relse; |
1676 | ||
8b80cda5 | 1677 | sk_mark_napi_id(sk, skb); |
1da177e4 LT |
1678 | skb->dev = NULL; |
1679 | ||
c6366184 | 1680 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1681 | ret = 0; |
1682 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 CL |
1683 | #ifdef CONFIG_NET_DMA |
1684 | struct tcp_sock *tp = tcp_sk(sk); | |
1685 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) | |
a2bd1140 | 1686 | tp->ucopy.dma_chan = net_dma_find_channel(); |
1a2449a8 | 1687 | if (tp->ucopy.dma_chan) |
1da177e4 | 1688 | ret = tcp_v4_do_rcv(sk, skb); |
1a2449a8 CL |
1689 | else |
1690 | #endif | |
1691 | { | |
1692 | if (!tcp_prequeue(sk, skb)) | |
ae8d7f88 | 1693 | ret = tcp_v4_do_rcv(sk, skb); |
1a2449a8 | 1694 | } |
da882c1f ED |
1695 | } else if (unlikely(sk_add_backlog(sk, skb, |
1696 | sk->sk_rcvbuf + sk->sk_sndbuf))) { | |
6b03a53a | 1697 | bh_unlock_sock(sk); |
6cce09f8 | 1698 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1699 | goto discard_and_relse; |
1700 | } | |
1da177e4 LT |
1701 | bh_unlock_sock(sk); |
1702 | ||
1703 | sock_put(sk); | |
1704 | ||
1705 | return ret; | |
1706 | ||
1707 | no_tcp_socket: | |
1708 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1709 | goto discard_it; | |
1710 | ||
1711 | if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1712 | csum_error: |
1713 | TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS); | |
1da177e4 | 1714 | bad_packet: |
63231bdd | 1715 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1716 | } else { |
cfb6eeb4 | 1717 | tcp_v4_send_reset(NULL, skb); |
1da177e4 LT |
1718 | } |
1719 | ||
1720 | discard_it: | |
1721 | /* Discard frame. */ | |
1722 | kfree_skb(skb); | |
e905a9ed | 1723 | return 0; |
1da177e4 LT |
1724 | |
1725 | discard_and_relse: | |
1726 | sock_put(sk); | |
1727 | goto discard_it; | |
1728 | ||
1729 | do_time_wait: | |
1730 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1731 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1732 | goto discard_it; |
1733 | } | |
1734 | ||
6a5dc9e5 | 1735 | if (skb->len < (th->doff << 2)) { |
9469c7b4 | 1736 | inet_twsk_put(inet_twsk(sk)); |
6a5dc9e5 ED |
1737 | goto bad_packet; |
1738 | } | |
1739 | if (tcp_checksum_complete(skb)) { | |
1740 | inet_twsk_put(inet_twsk(sk)); | |
1741 | goto csum_error; | |
1da177e4 | 1742 | } |
9469c7b4 | 1743 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 | 1744 | case TCP_TW_SYN: { |
c346dca1 | 1745 | struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev), |
c67499c0 | 1746 | &tcp_hashinfo, |
da5e3630 | 1747 | iph->saddr, th->source, |
eddc9ec5 | 1748 | iph->daddr, th->dest, |
463c84b9 | 1749 | inet_iif(skb)); |
1da177e4 | 1750 | if (sk2) { |
9469c7b4 YH |
1751 | inet_twsk_deschedule(inet_twsk(sk), &tcp_death_row); |
1752 | inet_twsk_put(inet_twsk(sk)); | |
1da177e4 LT |
1753 | sk = sk2; |
1754 | goto process; | |
1755 | } | |
1756 | /* Fall through to ACK */ | |
1757 | } | |
1758 | case TCP_TW_ACK: | |
1759 | tcp_v4_timewait_ack(sk, skb); | |
1760 | break; | |
1761 | case TCP_TW_RST: | |
1762 | goto no_tcp_socket; | |
1763 | case TCP_TW_SUCCESS:; | |
1764 | } | |
1765 | goto discard_it; | |
1766 | } | |
1767 | ||
ccb7c410 DM |
1768 | static struct timewait_sock_ops tcp_timewait_sock_ops = { |
1769 | .twsk_obj_size = sizeof(struct tcp_timewait_sock), | |
1770 | .twsk_unique = tcp_twsk_unique, | |
1771 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 | 1772 | }; |
1da177e4 | 1773 | |
63d02d15 | 1774 | void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
5d299f3d ED |
1775 | { |
1776 | struct dst_entry *dst = skb_dst(skb); | |
1777 | ||
ca777eff ED |
1778 | if (dst) { |
1779 | dst_hold(dst); | |
1780 | sk->sk_rx_dst = dst; | |
1781 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
1782 | } | |
5d299f3d | 1783 | } |
63d02d15 | 1784 | EXPORT_SYMBOL(inet_sk_rx_dst_set); |
5d299f3d | 1785 | |
3b401a81 | 1786 | const struct inet_connection_sock_af_ops ipv4_specific = { |
543d9cfe ACM |
1787 | .queue_xmit = ip_queue_xmit, |
1788 | .send_check = tcp_v4_send_check, | |
1789 | .rebuild_header = inet_sk_rebuild_header, | |
5d299f3d | 1790 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1791 | .conn_request = tcp_v4_conn_request, |
1792 | .syn_recv_sock = tcp_v4_syn_recv_sock, | |
543d9cfe ACM |
1793 | .net_header_len = sizeof(struct iphdr), |
1794 | .setsockopt = ip_setsockopt, | |
1795 | .getsockopt = ip_getsockopt, | |
1796 | .addr2sockaddr = inet_csk_addr2sockaddr, | |
1797 | .sockaddr_len = sizeof(struct sockaddr_in), | |
ab1e0a13 | 1798 | .bind_conflict = inet_csk_bind_conflict, |
3fdadf7d | 1799 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1800 | .compat_setsockopt = compat_ip_setsockopt, |
1801 | .compat_getsockopt = compat_ip_getsockopt, | |
3fdadf7d | 1802 | #endif |
4fab9071 | 1803 | .mtu_reduced = tcp_v4_mtu_reduced, |
1da177e4 | 1804 | }; |
4bc2f18b | 1805 | EXPORT_SYMBOL(ipv4_specific); |
1da177e4 | 1806 | |
cfb6eeb4 | 1807 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1808 | static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = { |
cfb6eeb4 | 1809 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1810 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1811 | .md5_parse = tcp_v4_parse_md5_keys, |
cfb6eeb4 | 1812 | }; |
b6332e6c | 1813 | #endif |
cfb6eeb4 | 1814 | |
1da177e4 LT |
1815 | /* NOTE: A lot of things set to zero explicitly by call to |
1816 | * sk_alloc() so need not be done here. | |
1817 | */ | |
1818 | static int tcp_v4_init_sock(struct sock *sk) | |
1819 | { | |
6687e988 | 1820 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1821 | |
900f65d3 | 1822 | tcp_init_sock(sk); |
1da177e4 | 1823 | |
8292a17a | 1824 | icsk->icsk_af_ops = &ipv4_specific; |
900f65d3 | 1825 | |
cfb6eeb4 | 1826 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1827 | tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific; |
cfb6eeb4 | 1828 | #endif |
1da177e4 | 1829 | |
1da177e4 LT |
1830 | return 0; |
1831 | } | |
1832 | ||
7d06b2e0 | 1833 | void tcp_v4_destroy_sock(struct sock *sk) |
1da177e4 LT |
1834 | { |
1835 | struct tcp_sock *tp = tcp_sk(sk); | |
1836 | ||
1837 | tcp_clear_xmit_timers(sk); | |
1838 | ||
6687e988 | 1839 | tcp_cleanup_congestion_control(sk); |
317a76f9 | 1840 | |
1da177e4 | 1841 | /* Cleanup up the write buffer. */ |
fe067e8a | 1842 | tcp_write_queue_purge(sk); |
1da177e4 LT |
1843 | |
1844 | /* Cleans up our, hopefully empty, out_of_order_queue. */ | |
e905a9ed | 1845 | __skb_queue_purge(&tp->out_of_order_queue); |
1da177e4 | 1846 | |
cfb6eeb4 YH |
1847 | #ifdef CONFIG_TCP_MD5SIG |
1848 | /* Clean up the MD5 key list, if any */ | |
1849 | if (tp->md5sig_info) { | |
a915da9b | 1850 | tcp_clear_md5_list(sk); |
a8afca03 | 1851 | kfree_rcu(tp->md5sig_info, rcu); |
cfb6eeb4 YH |
1852 | tp->md5sig_info = NULL; |
1853 | } | |
1854 | #endif | |
1855 | ||
1a2449a8 CL |
1856 | #ifdef CONFIG_NET_DMA |
1857 | /* Cleans up our sk_async_wait_queue */ | |
e905a9ed | 1858 | __skb_queue_purge(&sk->sk_async_wait_queue); |
1a2449a8 CL |
1859 | #endif |
1860 | ||
1da177e4 LT |
1861 | /* Clean prequeue, it must be empty really */ |
1862 | __skb_queue_purge(&tp->ucopy.prequeue); | |
1863 | ||
1864 | /* Clean up a referenced TCP bind bucket. */ | |
463c84b9 | 1865 | if (inet_csk(sk)->icsk_bind_hash) |
ab1e0a13 | 1866 | inet_put_port(sk); |
1da177e4 | 1867 | |
168a8f58 | 1868 | BUG_ON(tp->fastopen_rsk != NULL); |
435cf559 | 1869 | |
cf60af03 YC |
1870 | /* If socket is aborted during connect operation */ |
1871 | tcp_free_fastopen_req(tp); | |
1872 | ||
180d8cd9 | 1873 | sk_sockets_allocated_dec(sk); |
d1a4c0b3 | 1874 | sock_release_memcg(sk); |
1da177e4 | 1875 | } |
1da177e4 LT |
1876 | EXPORT_SYMBOL(tcp_v4_destroy_sock); |
1877 | ||
1878 | #ifdef CONFIG_PROC_FS | |
1879 | /* Proc filesystem TCP sock list dumping. */ | |
1880 | ||
a8b690f9 TH |
1881 | /* |
1882 | * Get next listener socket follow cur. If cur is NULL, get first socket | |
1883 | * starting from bucket given in st->bucket; when st->bucket is zero the | |
1884 | * very first socket in the hash table is returned. | |
1885 | */ | |
1da177e4 LT |
1886 | static void *listening_get_next(struct seq_file *seq, void *cur) |
1887 | { | |
463c84b9 | 1888 | struct inet_connection_sock *icsk; |
c25eb3bf | 1889 | struct hlist_nulls_node *node; |
1da177e4 | 1890 | struct sock *sk = cur; |
5caea4ea | 1891 | struct inet_listen_hashbucket *ilb; |
5799de0b | 1892 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1893 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1894 | |
1895 | if (!sk) { | |
a8b690f9 | 1896 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
5caea4ea | 1897 | spin_lock_bh(&ilb->lock); |
c25eb3bf | 1898 | sk = sk_nulls_head(&ilb->head); |
a8b690f9 | 1899 | st->offset = 0; |
1da177e4 LT |
1900 | goto get_sk; |
1901 | } | |
5caea4ea | 1902 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1da177e4 | 1903 | ++st->num; |
a8b690f9 | 1904 | ++st->offset; |
1da177e4 LT |
1905 | |
1906 | if (st->state == TCP_SEQ_STATE_OPENREQ) { | |
60236fdd | 1907 | struct request_sock *req = cur; |
1da177e4 | 1908 | |
72a3effa | 1909 | icsk = inet_csk(st->syn_wait_sk); |
1da177e4 LT |
1910 | req = req->dl_next; |
1911 | while (1) { | |
1912 | while (req) { | |
bdccc4ca | 1913 | if (req->rsk_ops->family == st->family) { |
1da177e4 LT |
1914 | cur = req; |
1915 | goto out; | |
1916 | } | |
1917 | req = req->dl_next; | |
1918 | } | |
72a3effa | 1919 | if (++st->sbucket >= icsk->icsk_accept_queue.listen_opt->nr_table_entries) |
1da177e4 LT |
1920 | break; |
1921 | get_req: | |
463c84b9 | 1922 | req = icsk->icsk_accept_queue.listen_opt->syn_table[st->sbucket]; |
1da177e4 | 1923 | } |
1bde5ac4 | 1924 | sk = sk_nulls_next(st->syn_wait_sk); |
1da177e4 | 1925 | st->state = TCP_SEQ_STATE_LISTENING; |
463c84b9 | 1926 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 1927 | } else { |
e905a9ed | 1928 | icsk = inet_csk(sk); |
463c84b9 ACM |
1929 | read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1930 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) | |
1da177e4 | 1931 | goto start_req; |
463c84b9 | 1932 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1bde5ac4 | 1933 | sk = sk_nulls_next(sk); |
1da177e4 LT |
1934 | } |
1935 | get_sk: | |
c25eb3bf | 1936 | sk_nulls_for_each_from(sk, node) { |
8475ef9f PE |
1937 | if (!net_eq(sock_net(sk), net)) |
1938 | continue; | |
1939 | if (sk->sk_family == st->family) { | |
1da177e4 LT |
1940 | cur = sk; |
1941 | goto out; | |
1942 | } | |
e905a9ed | 1943 | icsk = inet_csk(sk); |
463c84b9 ACM |
1944 | read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1945 | if (reqsk_queue_len(&icsk->icsk_accept_queue)) { | |
1da177e4 LT |
1946 | start_req: |
1947 | st->uid = sock_i_uid(sk); | |
1948 | st->syn_wait_sk = sk; | |
1949 | st->state = TCP_SEQ_STATE_OPENREQ; | |
1950 | st->sbucket = 0; | |
1951 | goto get_req; | |
1952 | } | |
463c84b9 | 1953 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); |
1da177e4 | 1954 | } |
5caea4ea | 1955 | spin_unlock_bh(&ilb->lock); |
a8b690f9 | 1956 | st->offset = 0; |
0f7ff927 | 1957 | if (++st->bucket < INET_LHTABLE_SIZE) { |
5caea4ea ED |
1958 | ilb = &tcp_hashinfo.listening_hash[st->bucket]; |
1959 | spin_lock_bh(&ilb->lock); | |
c25eb3bf | 1960 | sk = sk_nulls_head(&ilb->head); |
1da177e4 LT |
1961 | goto get_sk; |
1962 | } | |
1963 | cur = NULL; | |
1964 | out: | |
1965 | return cur; | |
1966 | } | |
1967 | ||
1968 | static void *listening_get_idx(struct seq_file *seq, loff_t *pos) | |
1969 | { | |
a8b690f9 TH |
1970 | struct tcp_iter_state *st = seq->private; |
1971 | void *rc; | |
1972 | ||
1973 | st->bucket = 0; | |
1974 | st->offset = 0; | |
1975 | rc = listening_get_next(seq, NULL); | |
1da177e4 LT |
1976 | |
1977 | while (rc && *pos) { | |
1978 | rc = listening_get_next(seq, rc); | |
1979 | --*pos; | |
1980 | } | |
1981 | return rc; | |
1982 | } | |
1983 | ||
05dbc7b5 | 1984 | static inline bool empty_bucket(const struct tcp_iter_state *st) |
6eac5604 | 1985 | { |
05dbc7b5 | 1986 | return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain); |
6eac5604 AK |
1987 | } |
1988 | ||
a8b690f9 TH |
1989 | /* |
1990 | * Get first established socket starting from bucket given in st->bucket. | |
1991 | * If st->bucket is zero, the very first socket in the hash is returned. | |
1992 | */ | |
1da177e4 LT |
1993 | static void *established_get_first(struct seq_file *seq) |
1994 | { | |
5799de0b | 1995 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 1996 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1997 | void *rc = NULL; |
1998 | ||
a8b690f9 TH |
1999 | st->offset = 0; |
2000 | for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) { | |
1da177e4 | 2001 | struct sock *sk; |
3ab5aee7 | 2002 | struct hlist_nulls_node *node; |
9db66bdc | 2003 | spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket); |
1da177e4 | 2004 | |
6eac5604 AK |
2005 | /* Lockless fast path for the common case of empty buckets */ |
2006 | if (empty_bucket(st)) | |
2007 | continue; | |
2008 | ||
9db66bdc | 2009 | spin_lock_bh(lock); |
3ab5aee7 | 2010 | sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) { |
f40c8174 | 2011 | if (sk->sk_family != st->family || |
878628fb | 2012 | !net_eq(sock_net(sk), net)) { |
1da177e4 LT |
2013 | continue; |
2014 | } | |
2015 | rc = sk; | |
2016 | goto out; | |
2017 | } | |
9db66bdc | 2018 | spin_unlock_bh(lock); |
1da177e4 LT |
2019 | } |
2020 | out: | |
2021 | return rc; | |
2022 | } | |
2023 | ||
2024 | static void *established_get_next(struct seq_file *seq, void *cur) | |
2025 | { | |
2026 | struct sock *sk = cur; | |
3ab5aee7 | 2027 | struct hlist_nulls_node *node; |
5799de0b | 2028 | struct tcp_iter_state *st = seq->private; |
a4146b1b | 2029 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2030 | |
2031 | ++st->num; | |
a8b690f9 | 2032 | ++st->offset; |
1da177e4 | 2033 | |
05dbc7b5 | 2034 | sk = sk_nulls_next(sk); |
1da177e4 | 2035 | |
3ab5aee7 | 2036 | sk_nulls_for_each_from(sk, node) { |
878628fb | 2037 | if (sk->sk_family == st->family && net_eq(sock_net(sk), net)) |
05dbc7b5 | 2038 | return sk; |
1da177e4 LT |
2039 | } |
2040 | ||
05dbc7b5 ED |
2041 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
2042 | ++st->bucket; | |
2043 | return established_get_first(seq); | |
1da177e4 LT |
2044 | } |
2045 | ||
2046 | static void *established_get_idx(struct seq_file *seq, loff_t pos) | |
2047 | { | |
a8b690f9 TH |
2048 | struct tcp_iter_state *st = seq->private; |
2049 | void *rc; | |
2050 | ||
2051 | st->bucket = 0; | |
2052 | rc = established_get_first(seq); | |
1da177e4 LT |
2053 | |
2054 | while (rc && pos) { | |
2055 | rc = established_get_next(seq, rc); | |
2056 | --pos; | |
7174259e | 2057 | } |
1da177e4 LT |
2058 | return rc; |
2059 | } | |
2060 | ||
2061 | static void *tcp_get_idx(struct seq_file *seq, loff_t pos) | |
2062 | { | |
2063 | void *rc; | |
5799de0b | 2064 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2065 | |
1da177e4 LT |
2066 | st->state = TCP_SEQ_STATE_LISTENING; |
2067 | rc = listening_get_idx(seq, &pos); | |
2068 | ||
2069 | if (!rc) { | |
1da177e4 LT |
2070 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
2071 | rc = established_get_idx(seq, pos); | |
2072 | } | |
2073 | ||
2074 | return rc; | |
2075 | } | |
2076 | ||
a8b690f9 TH |
2077 | static void *tcp_seek_last_pos(struct seq_file *seq) |
2078 | { | |
2079 | struct tcp_iter_state *st = seq->private; | |
2080 | int offset = st->offset; | |
2081 | int orig_num = st->num; | |
2082 | void *rc = NULL; | |
2083 | ||
2084 | switch (st->state) { | |
2085 | case TCP_SEQ_STATE_OPENREQ: | |
2086 | case TCP_SEQ_STATE_LISTENING: | |
2087 | if (st->bucket >= INET_LHTABLE_SIZE) | |
2088 | break; | |
2089 | st->state = TCP_SEQ_STATE_LISTENING; | |
2090 | rc = listening_get_next(seq, NULL); | |
2091 | while (offset-- && rc) | |
2092 | rc = listening_get_next(seq, rc); | |
2093 | if (rc) | |
2094 | break; | |
2095 | st->bucket = 0; | |
05dbc7b5 | 2096 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2097 | /* Fallthrough */ |
2098 | case TCP_SEQ_STATE_ESTABLISHED: | |
a8b690f9 TH |
2099 | if (st->bucket > tcp_hashinfo.ehash_mask) |
2100 | break; | |
2101 | rc = established_get_first(seq); | |
2102 | while (offset-- && rc) | |
2103 | rc = established_get_next(seq, rc); | |
2104 | } | |
2105 | ||
2106 | st->num = orig_num; | |
2107 | ||
2108 | return rc; | |
2109 | } | |
2110 | ||
1da177e4 LT |
2111 | static void *tcp_seq_start(struct seq_file *seq, loff_t *pos) |
2112 | { | |
5799de0b | 2113 | struct tcp_iter_state *st = seq->private; |
a8b690f9 TH |
2114 | void *rc; |
2115 | ||
2116 | if (*pos && *pos == st->last_pos) { | |
2117 | rc = tcp_seek_last_pos(seq); | |
2118 | if (rc) | |
2119 | goto out; | |
2120 | } | |
2121 | ||
1da177e4 LT |
2122 | st->state = TCP_SEQ_STATE_LISTENING; |
2123 | st->num = 0; | |
a8b690f9 TH |
2124 | st->bucket = 0; |
2125 | st->offset = 0; | |
2126 | rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
2127 | ||
2128 | out: | |
2129 | st->last_pos = *pos; | |
2130 | return rc; | |
1da177e4 LT |
2131 | } |
2132 | ||
2133 | static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2134 | { | |
a8b690f9 | 2135 | struct tcp_iter_state *st = seq->private; |
1da177e4 | 2136 | void *rc = NULL; |
1da177e4 LT |
2137 | |
2138 | if (v == SEQ_START_TOKEN) { | |
2139 | rc = tcp_get_idx(seq, 0); | |
2140 | goto out; | |
2141 | } | |
1da177e4 LT |
2142 | |
2143 | switch (st->state) { | |
2144 | case TCP_SEQ_STATE_OPENREQ: | |
2145 | case TCP_SEQ_STATE_LISTENING: | |
2146 | rc = listening_get_next(seq, v); | |
2147 | if (!rc) { | |
1da177e4 | 2148 | st->state = TCP_SEQ_STATE_ESTABLISHED; |
a8b690f9 TH |
2149 | st->bucket = 0; |
2150 | st->offset = 0; | |
1da177e4 LT |
2151 | rc = established_get_first(seq); |
2152 | } | |
2153 | break; | |
2154 | case TCP_SEQ_STATE_ESTABLISHED: | |
1da177e4 LT |
2155 | rc = established_get_next(seq, v); |
2156 | break; | |
2157 | } | |
2158 | out: | |
2159 | ++*pos; | |
a8b690f9 | 2160 | st->last_pos = *pos; |
1da177e4 LT |
2161 | return rc; |
2162 | } | |
2163 | ||
2164 | static void tcp_seq_stop(struct seq_file *seq, void *v) | |
2165 | { | |
5799de0b | 2166 | struct tcp_iter_state *st = seq->private; |
1da177e4 LT |
2167 | |
2168 | switch (st->state) { | |
2169 | case TCP_SEQ_STATE_OPENREQ: | |
2170 | if (v) { | |
463c84b9 ACM |
2171 | struct inet_connection_sock *icsk = inet_csk(st->syn_wait_sk); |
2172 | read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); | |
1da177e4 LT |
2173 | } |
2174 | case TCP_SEQ_STATE_LISTENING: | |
2175 | if (v != SEQ_START_TOKEN) | |
5caea4ea | 2176 | spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock); |
1da177e4 | 2177 | break; |
1da177e4 LT |
2178 | case TCP_SEQ_STATE_ESTABLISHED: |
2179 | if (v) | |
9db66bdc | 2180 | spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket)); |
1da177e4 LT |
2181 | break; |
2182 | } | |
2183 | } | |
2184 | ||
73cb88ec | 2185 | int tcp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2186 | { |
d9dda78b | 2187 | struct tcp_seq_afinfo *afinfo = PDE_DATA(inode); |
1da177e4 | 2188 | struct tcp_iter_state *s; |
52d6f3f1 | 2189 | int err; |
1da177e4 | 2190 | |
52d6f3f1 DL |
2191 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2192 | sizeof(struct tcp_iter_state)); | |
2193 | if (err < 0) | |
2194 | return err; | |
f40c8174 | 2195 | |
52d6f3f1 | 2196 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2197 | s->family = afinfo->family; |
688d1945 | 2198 | s->last_pos = 0; |
f40c8174 DL |
2199 | return 0; |
2200 | } | |
73cb88ec | 2201 | EXPORT_SYMBOL(tcp_seq_open); |
f40c8174 | 2202 | |
6f8b13bc | 2203 | int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 LT |
2204 | { |
2205 | int rc = 0; | |
2206 | struct proc_dir_entry *p; | |
2207 | ||
9427c4b3 DL |
2208 | afinfo->seq_ops.start = tcp_seq_start; |
2209 | afinfo->seq_ops.next = tcp_seq_next; | |
2210 | afinfo->seq_ops.stop = tcp_seq_stop; | |
2211 | ||
84841c3c | 2212 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2213 | afinfo->seq_fops, afinfo); |
84841c3c | 2214 | if (!p) |
1da177e4 LT |
2215 | rc = -ENOMEM; |
2216 | return rc; | |
2217 | } | |
4bc2f18b | 2218 | EXPORT_SYMBOL(tcp_proc_register); |
1da177e4 | 2219 | |
6f8b13bc | 2220 | void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo) |
1da177e4 | 2221 | { |
ece31ffd | 2222 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2223 | } |
4bc2f18b | 2224 | EXPORT_SYMBOL(tcp_proc_unregister); |
1da177e4 | 2225 | |
cf533ea5 | 2226 | static void get_openreq4(const struct sock *sk, const struct request_sock *req, |
652586df | 2227 | struct seq_file *f, int i, kuid_t uid) |
1da177e4 | 2228 | { |
2e6599cb | 2229 | const struct inet_request_sock *ireq = inet_rsk(req); |
a399a805 | 2230 | long delta = req->expires - jiffies; |
1da177e4 | 2231 | |
5e659e4c | 2232 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2233 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK", |
1da177e4 | 2234 | i, |
634fb979 | 2235 | ireq->ir_loc_addr, |
c720c7e8 | 2236 | ntohs(inet_sk(sk)->inet_sport), |
634fb979 ED |
2237 | ireq->ir_rmt_addr, |
2238 | ntohs(ireq->ir_rmt_port), | |
1da177e4 LT |
2239 | TCP_SYN_RECV, |
2240 | 0, 0, /* could print option size, but that is af dependent. */ | |
2241 | 1, /* timers active (only the expire timer) */ | |
a399a805 | 2242 | jiffies_delta_to_clock_t(delta), |
e6c022a4 | 2243 | req->num_timeout, |
a7cb5a49 | 2244 | from_kuid_munged(seq_user_ns(f), uid), |
1da177e4 LT |
2245 | 0, /* non standard timer */ |
2246 | 0, /* open_requests have no inode */ | |
2247 | atomic_read(&sk->sk_refcnt), | |
652586df | 2248 | req); |
1da177e4 LT |
2249 | } |
2250 | ||
652586df | 2251 | static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i) |
1da177e4 LT |
2252 | { |
2253 | int timer_active; | |
2254 | unsigned long timer_expires; | |
cf533ea5 | 2255 | const struct tcp_sock *tp = tcp_sk(sk); |
cf4c6bf8 | 2256 | const struct inet_connection_sock *icsk = inet_csk(sk); |
cf533ea5 | 2257 | const struct inet_sock *inet = inet_sk(sk); |
168a8f58 | 2258 | struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq; |
c720c7e8 ED |
2259 | __be32 dest = inet->inet_daddr; |
2260 | __be32 src = inet->inet_rcv_saddr; | |
2261 | __u16 destp = ntohs(inet->inet_dport); | |
2262 | __u16 srcp = ntohs(inet->inet_sport); | |
49d09007 | 2263 | int rx_queue; |
1da177e4 | 2264 | |
6ba8a3b1 ND |
2265 | if (icsk->icsk_pending == ICSK_TIME_RETRANS || |
2266 | icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS || | |
2267 | icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { | |
1da177e4 | 2268 | timer_active = 1; |
463c84b9 ACM |
2269 | timer_expires = icsk->icsk_timeout; |
2270 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2271 | timer_active = 4; |
463c84b9 | 2272 | timer_expires = icsk->icsk_timeout; |
cf4c6bf8 | 2273 | } else if (timer_pending(&sk->sk_timer)) { |
1da177e4 | 2274 | timer_active = 2; |
cf4c6bf8 | 2275 | timer_expires = sk->sk_timer.expires; |
1da177e4 LT |
2276 | } else { |
2277 | timer_active = 0; | |
2278 | timer_expires = jiffies; | |
2279 | } | |
2280 | ||
49d09007 ED |
2281 | if (sk->sk_state == TCP_LISTEN) |
2282 | rx_queue = sk->sk_ack_backlog; | |
2283 | else | |
2284 | /* | |
2285 | * because we dont lock socket, we might find a transient negative value | |
2286 | */ | |
2287 | rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0); | |
2288 | ||
5e659e4c | 2289 | seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX " |
652586df | 2290 | "%08X %5u %8d %lu %d %pK %lu %lu %u %u %d", |
cf4c6bf8 | 2291 | i, src, srcp, dest, destp, sk->sk_state, |
47da8ee6 | 2292 | tp->write_seq - tp->snd_una, |
49d09007 | 2293 | rx_queue, |
1da177e4 | 2294 | timer_active, |
a399a805 | 2295 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 2296 | icsk->icsk_retransmits, |
a7cb5a49 | 2297 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)), |
6687e988 | 2298 | icsk->icsk_probes_out, |
cf4c6bf8 IJ |
2299 | sock_i_ino(sk), |
2300 | atomic_read(&sk->sk_refcnt), sk, | |
7be87351 SH |
2301 | jiffies_to_clock_t(icsk->icsk_rto), |
2302 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 2303 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
1da177e4 | 2304 | tp->snd_cwnd, |
168a8f58 JC |
2305 | sk->sk_state == TCP_LISTEN ? |
2306 | (fastopenq ? fastopenq->max_qlen : 0) : | |
652586df | 2307 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)); |
1da177e4 LT |
2308 | } |
2309 | ||
cf533ea5 | 2310 | static void get_timewait4_sock(const struct inet_timewait_sock *tw, |
652586df | 2311 | struct seq_file *f, int i) |
1da177e4 | 2312 | { |
23f33c2d | 2313 | __be32 dest, src; |
1da177e4 | 2314 | __u16 destp, srcp; |
e2a1d3e4 | 2315 | s32 delta = tw->tw_ttd - inet_tw_time_stamp(); |
1da177e4 LT |
2316 | |
2317 | dest = tw->tw_daddr; | |
2318 | src = tw->tw_rcv_saddr; | |
2319 | destp = ntohs(tw->tw_dport); | |
2320 | srcp = ntohs(tw->tw_sport); | |
2321 | ||
5e659e4c | 2322 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
652586df | 2323 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK", |
1da177e4 | 2324 | i, src, srcp, dest, destp, tw->tw_substate, 0, 0, |
a399a805 | 2325 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
652586df | 2326 | atomic_read(&tw->tw_refcnt), tw); |
1da177e4 LT |
2327 | } |
2328 | ||
2329 | #define TMPSZ 150 | |
2330 | ||
2331 | static int tcp4_seq_show(struct seq_file *seq, void *v) | |
2332 | { | |
5799de0b | 2333 | struct tcp_iter_state *st; |
05dbc7b5 | 2334 | struct sock *sk = v; |
1da177e4 | 2335 | |
652586df | 2336 | seq_setwidth(seq, TMPSZ - 1); |
1da177e4 | 2337 | if (v == SEQ_START_TOKEN) { |
652586df | 2338 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
1da177e4 LT |
2339 | "rx_queue tr tm->when retrnsmt uid timeout " |
2340 | "inode"); | |
2341 | goto out; | |
2342 | } | |
2343 | st = seq->private; | |
2344 | ||
2345 | switch (st->state) { | |
2346 | case TCP_SEQ_STATE_LISTENING: | |
2347 | case TCP_SEQ_STATE_ESTABLISHED: | |
05dbc7b5 | 2348 | if (sk->sk_state == TCP_TIME_WAIT) |
652586df | 2349 | get_timewait4_sock(v, seq, st->num); |
05dbc7b5 | 2350 | else |
652586df | 2351 | get_tcp4_sock(v, seq, st->num); |
1da177e4 LT |
2352 | break; |
2353 | case TCP_SEQ_STATE_OPENREQ: | |
652586df | 2354 | get_openreq4(st->syn_wait_sk, v, seq, st->num, st->uid); |
1da177e4 LT |
2355 | break; |
2356 | } | |
1da177e4 | 2357 | out: |
652586df | 2358 | seq_pad(seq, '\n'); |
1da177e4 LT |
2359 | return 0; |
2360 | } | |
2361 | ||
73cb88ec AV |
2362 | static const struct file_operations tcp_afinfo_seq_fops = { |
2363 | .owner = THIS_MODULE, | |
2364 | .open = tcp_seq_open, | |
2365 | .read = seq_read, | |
2366 | .llseek = seq_lseek, | |
2367 | .release = seq_release_net | |
2368 | }; | |
2369 | ||
1da177e4 | 2370 | static struct tcp_seq_afinfo tcp4_seq_afinfo = { |
1da177e4 LT |
2371 | .name = "tcp", |
2372 | .family = AF_INET, | |
73cb88ec | 2373 | .seq_fops = &tcp_afinfo_seq_fops, |
9427c4b3 DL |
2374 | .seq_ops = { |
2375 | .show = tcp4_seq_show, | |
2376 | }, | |
1da177e4 LT |
2377 | }; |
2378 | ||
2c8c1e72 | 2379 | static int __net_init tcp4_proc_init_net(struct net *net) |
757764f6 PE |
2380 | { |
2381 | return tcp_proc_register(net, &tcp4_seq_afinfo); | |
2382 | } | |
2383 | ||
2c8c1e72 | 2384 | static void __net_exit tcp4_proc_exit_net(struct net *net) |
757764f6 PE |
2385 | { |
2386 | tcp_proc_unregister(net, &tcp4_seq_afinfo); | |
2387 | } | |
2388 | ||
2389 | static struct pernet_operations tcp4_net_ops = { | |
2390 | .init = tcp4_proc_init_net, | |
2391 | .exit = tcp4_proc_exit_net, | |
2392 | }; | |
2393 | ||
1da177e4 LT |
2394 | int __init tcp4_proc_init(void) |
2395 | { | |
757764f6 | 2396 | return register_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2397 | } |
2398 | ||
2399 | void tcp4_proc_exit(void) | |
2400 | { | |
757764f6 | 2401 | unregister_pernet_subsys(&tcp4_net_ops); |
1da177e4 LT |
2402 | } |
2403 | #endif /* CONFIG_PROC_FS */ | |
2404 | ||
2405 | struct proto tcp_prot = { | |
2406 | .name = "TCP", | |
2407 | .owner = THIS_MODULE, | |
2408 | .close = tcp_close, | |
2409 | .connect = tcp_v4_connect, | |
2410 | .disconnect = tcp_disconnect, | |
463c84b9 | 2411 | .accept = inet_csk_accept, |
1da177e4 LT |
2412 | .ioctl = tcp_ioctl, |
2413 | .init = tcp_v4_init_sock, | |
2414 | .destroy = tcp_v4_destroy_sock, | |
2415 | .shutdown = tcp_shutdown, | |
2416 | .setsockopt = tcp_setsockopt, | |
2417 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 2418 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2419 | .sendmsg = tcp_sendmsg, |
2420 | .sendpage = tcp_sendpage, | |
1da177e4 | 2421 | .backlog_rcv = tcp_v4_do_rcv, |
46d3ceab | 2422 | .release_cb = tcp_release_cb, |
ab1e0a13 ACM |
2423 | .hash = inet_hash, |
2424 | .unhash = inet_unhash, | |
2425 | .get_port = inet_csk_get_port, | |
1da177e4 | 2426 | .enter_memory_pressure = tcp_enter_memory_pressure, |
c9bee3b7 | 2427 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 | 2428 | .sockets_allocated = &tcp_sockets_allocated, |
0a5578cf | 2429 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2430 | .memory_allocated = &tcp_memory_allocated, |
2431 | .memory_pressure = &tcp_memory_pressure, | |
a4fe34bf | 2432 | .sysctl_mem = sysctl_tcp_mem, |
1da177e4 LT |
2433 | .sysctl_wmem = sysctl_tcp_wmem, |
2434 | .sysctl_rmem = sysctl_tcp_rmem, | |
2435 | .max_header = MAX_TCP_HEADER, | |
2436 | .obj_size = sizeof(struct tcp_sock), | |
3ab5aee7 | 2437 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2438 | .twsk_prot = &tcp_timewait_sock_ops, |
60236fdd | 2439 | .rsk_prot = &tcp_request_sock_ops, |
39d8cda7 | 2440 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2441 | .no_autobind = true, |
543d9cfe ACM |
2442 | #ifdef CONFIG_COMPAT |
2443 | .compat_setsockopt = compat_tcp_setsockopt, | |
2444 | .compat_getsockopt = compat_tcp_getsockopt, | |
2445 | #endif | |
c255a458 | 2446 | #ifdef CONFIG_MEMCG_KMEM |
d1a4c0b3 GC |
2447 | .init_cgroup = tcp_init_cgroup, |
2448 | .destroy_cgroup = tcp_destroy_cgroup, | |
2449 | .proto_cgroup = tcp_proto_cgroup, | |
2450 | #endif | |
1da177e4 | 2451 | }; |
4bc2f18b | 2452 | EXPORT_SYMBOL(tcp_prot); |
1da177e4 | 2453 | |
046ee902 DL |
2454 | static int __net_init tcp_sk_init(struct net *net) |
2455 | { | |
5d134f1c | 2456 | net->ipv4.sysctl_tcp_ecn = 2; |
be9f4a44 | 2457 | return 0; |
046ee902 DL |
2458 | } |
2459 | ||
2460 | static void __net_exit tcp_sk_exit(struct net *net) | |
2461 | { | |
b099ce26 EB |
2462 | } |
2463 | ||
2464 | static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list) | |
2465 | { | |
2466 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET); | |
046ee902 DL |
2467 | } |
2468 | ||
2469 | static struct pernet_operations __net_initdata tcp_sk_ops = { | |
b099ce26 EB |
2470 | .init = tcp_sk_init, |
2471 | .exit = tcp_sk_exit, | |
2472 | .exit_batch = tcp_sk_exit_batch, | |
046ee902 DL |
2473 | }; |
2474 | ||
9b0f976f | 2475 | void __init tcp_v4_init(void) |
1da177e4 | 2476 | { |
5caea4ea | 2477 | inet_hashinfo_init(&tcp_hashinfo); |
6a1b3054 | 2478 | if (register_pernet_subsys(&tcp_sk_ops)) |
1da177e4 | 2479 | panic("Failed to create the TCP control socket.\n"); |
1da177e4 | 2480 | } |