Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * TCP over IPv6 | |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 7 | * |
1ab1457c | 8 | * Based on: |
1da177e4 LT |
9 | * linux/net/ipv4/tcp.c |
10 | * linux/net/ipv4/tcp_input.c | |
11 | * linux/net/ipv4/tcp_output.c | |
12 | * | |
13 | * Fixes: | |
14 | * Hideaki YOSHIFUJI : sin6_scope_id support | |
15 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
16 | * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind | |
17 | * a single port at the same time. | |
18 | * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file. | |
19 | * | |
20 | * This program is free software; you can redistribute it and/or | |
21 | * modify it under the terms of the GNU General Public License | |
22 | * as published by the Free Software Foundation; either version | |
23 | * 2 of the License, or (at your option) any later version. | |
24 | */ | |
25 | ||
eb4dea58 | 26 | #include <linux/bottom_half.h> |
1da177e4 | 27 | #include <linux/module.h> |
1da177e4 LT |
28 | #include <linux/errno.h> |
29 | #include <linux/types.h> | |
30 | #include <linux/socket.h> | |
31 | #include <linux/sockios.h> | |
32 | #include <linux/net.h> | |
33 | #include <linux/jiffies.h> | |
34 | #include <linux/in.h> | |
35 | #include <linux/in6.h> | |
36 | #include <linux/netdevice.h> | |
37 | #include <linux/init.h> | |
38 | #include <linux/jhash.h> | |
39 | #include <linux/ipsec.h> | |
40 | #include <linux/times.h> | |
5a0e3ad6 | 41 | #include <linux/slab.h> |
1da177e4 LT |
42 | |
43 | #include <linux/ipv6.h> | |
44 | #include <linux/icmpv6.h> | |
45 | #include <linux/random.h> | |
46 | ||
47 | #include <net/tcp.h> | |
48 | #include <net/ndisc.h> | |
5324a040 | 49 | #include <net/inet6_hashtables.h> |
8129765a | 50 | #include <net/inet6_connection_sock.h> |
1da177e4 LT |
51 | #include <net/ipv6.h> |
52 | #include <net/transp_v6.h> | |
53 | #include <net/addrconf.h> | |
54 | #include <net/ip6_route.h> | |
55 | #include <net/ip6_checksum.h> | |
56 | #include <net/inet_ecn.h> | |
57 | #include <net/protocol.h> | |
58 | #include <net/xfrm.h> | |
1da177e4 LT |
59 | #include <net/snmp.h> |
60 | #include <net/dsfield.h> | |
6d6ee43e | 61 | #include <net/timewait_sock.h> |
18134bed | 62 | #include <net/netdma.h> |
3d58b5fa | 63 | #include <net/inet_common.h> |
6e5714ea | 64 | #include <net/secure_seq.h> |
d1a4c0b3 | 65 | #include <net/tcp_memcontrol.h> |
1da177e4 LT |
66 | |
67 | #include <asm/uaccess.h> | |
68 | ||
69 | #include <linux/proc_fs.h> | |
70 | #include <linux/seq_file.h> | |
71 | ||
cfb6eeb4 YH |
72 | #include <linux/crypto.h> |
73 | #include <linux/scatterlist.h> | |
74 | ||
cfb6eeb4 | 75 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb); |
6edafaaf GJ |
76 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
77 | struct request_sock *req); | |
1da177e4 LT |
78 | |
79 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb); | |
1da177e4 | 80 | |
3b401a81 SH |
81 | static const struct inet_connection_sock_af_ops ipv6_mapped; |
82 | static const struct inet_connection_sock_af_ops ipv6_specific; | |
a928630a | 83 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de SH |
84 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific; |
85 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific; | |
9501f972 YH |
86 | #else |
87 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
b71d1d42 | 88 | const struct in6_addr *addr) |
9501f972 YH |
89 | { |
90 | return NULL; | |
91 | } | |
a928630a | 92 | #endif |
1da177e4 | 93 | |
fae6ef87 NC |
94 | static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
95 | { | |
96 | struct dst_entry *dst = skb_dst(skb); | |
97 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
98 | ||
99 | dst_hold(dst); | |
100 | sk->sk_rx_dst = dst; | |
101 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
102 | if (rt->rt6i_node) | |
103 | inet6_sk(sk)->rx_dst_cookie = rt->rt6i_node->fn_sernum; | |
104 | } | |
105 | ||
1da177e4 LT |
106 | static void tcp_v6_hash(struct sock *sk) |
107 | { | |
108 | if (sk->sk_state != TCP_CLOSE) { | |
8292a17a | 109 | if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) { |
1da177e4 LT |
110 | tcp_prot.hash(sk); |
111 | return; | |
112 | } | |
113 | local_bh_disable(); | |
9327f705 | 114 | __inet6_hash(sk, NULL); |
1da177e4 LT |
115 | local_bh_enable(); |
116 | } | |
117 | } | |
118 | ||
cf533ea5 | 119 | static __u32 tcp_v6_init_sequence(const struct sk_buff *skb) |
1da177e4 | 120 | { |
0660e03f ACM |
121 | return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, |
122 | ipv6_hdr(skb)->saddr.s6_addr32, | |
aa8223c7 ACM |
123 | tcp_hdr(skb)->dest, |
124 | tcp_hdr(skb)->source); | |
1da177e4 LT |
125 | } |
126 | ||
1ab1457c | 127 | static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
1da177e4 LT |
128 | int addr_len) |
129 | { | |
130 | struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr; | |
1ab1457c | 131 | struct inet_sock *inet = inet_sk(sk); |
d83d8461 | 132 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 LT |
133 | struct ipv6_pinfo *np = inet6_sk(sk); |
134 | struct tcp_sock *tp = tcp_sk(sk); | |
20c59de2 | 135 | struct in6_addr *saddr = NULL, *final_p, final; |
493f377d | 136 | struct rt6_info *rt; |
4c9483b2 | 137 | struct flowi6 fl6; |
1da177e4 LT |
138 | struct dst_entry *dst; |
139 | int addr_type; | |
140 | int err; | |
141 | ||
1ab1457c | 142 | if (addr_len < SIN6_LEN_RFC2133) |
1da177e4 LT |
143 | return -EINVAL; |
144 | ||
1ab1457c | 145 | if (usin->sin6_family != AF_INET6) |
a02cec21 | 146 | return -EAFNOSUPPORT; |
1da177e4 | 147 | |
4c9483b2 | 148 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
149 | |
150 | if (np->sndflow) { | |
4c9483b2 DM |
151 | fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
152 | IP6_ECN_flow_init(fl6.flowlabel); | |
153 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1da177e4 | 154 | struct ip6_flowlabel *flowlabel; |
4c9483b2 | 155 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); |
1da177e4 LT |
156 | if (flowlabel == NULL) |
157 | return -EINVAL; | |
4e3fd7a0 | 158 | usin->sin6_addr = flowlabel->dst; |
1da177e4 LT |
159 | fl6_sock_release(flowlabel); |
160 | } | |
161 | } | |
162 | ||
163 | /* | |
1ab1457c YH |
164 | * connect() to INADDR_ANY means loopback (BSD'ism). |
165 | */ | |
166 | ||
167 | if(ipv6_addr_any(&usin->sin6_addr)) | |
168 | usin->sin6_addr.s6_addr[15] = 0x1; | |
1da177e4 LT |
169 | |
170 | addr_type = ipv6_addr_type(&usin->sin6_addr); | |
171 | ||
172 | if(addr_type & IPV6_ADDR_MULTICAST) | |
173 | return -ENETUNREACH; | |
174 | ||
175 | if (addr_type&IPV6_ADDR_LINKLOCAL) { | |
176 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
177 | usin->sin6_scope_id) { | |
178 | /* If interface is set while binding, indices | |
179 | * must coincide. | |
180 | */ | |
181 | if (sk->sk_bound_dev_if && | |
182 | sk->sk_bound_dev_if != usin->sin6_scope_id) | |
183 | return -EINVAL; | |
184 | ||
185 | sk->sk_bound_dev_if = usin->sin6_scope_id; | |
186 | } | |
187 | ||
188 | /* Connect to link-local address requires an interface */ | |
189 | if (!sk->sk_bound_dev_if) | |
190 | return -EINVAL; | |
191 | } | |
192 | ||
193 | if (tp->rx_opt.ts_recent_stamp && | |
194 | !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) { | |
195 | tp->rx_opt.ts_recent = 0; | |
196 | tp->rx_opt.ts_recent_stamp = 0; | |
197 | tp->write_seq = 0; | |
198 | } | |
199 | ||
4e3fd7a0 | 200 | np->daddr = usin->sin6_addr; |
4c9483b2 | 201 | np->flow_label = fl6.flowlabel; |
1da177e4 LT |
202 | |
203 | /* | |
204 | * TCP over IPv4 | |
205 | */ | |
206 | ||
207 | if (addr_type == IPV6_ADDR_MAPPED) { | |
d83d8461 | 208 | u32 exthdrlen = icsk->icsk_ext_hdr_len; |
1da177e4 LT |
209 | struct sockaddr_in sin; |
210 | ||
211 | SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); | |
212 | ||
213 | if (__ipv6_only_sock(sk)) | |
214 | return -ENETUNREACH; | |
215 | ||
216 | sin.sin_family = AF_INET; | |
217 | sin.sin_port = usin->sin6_port; | |
218 | sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; | |
219 | ||
d83d8461 | 220 | icsk->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 221 | sk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
222 | #ifdef CONFIG_TCP_MD5SIG |
223 | tp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
224 | #endif | |
1da177e4 LT |
225 | |
226 | err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); | |
227 | ||
228 | if (err) { | |
d83d8461 ACM |
229 | icsk->icsk_ext_hdr_len = exthdrlen; |
230 | icsk->icsk_af_ops = &ipv6_specific; | |
1da177e4 | 231 | sk->sk_backlog_rcv = tcp_v6_do_rcv; |
cfb6eeb4 YH |
232 | #ifdef CONFIG_TCP_MD5SIG |
233 | tp->af_specific = &tcp_sock_ipv6_specific; | |
234 | #endif | |
1da177e4 LT |
235 | goto failure; |
236 | } else { | |
c720c7e8 ED |
237 | ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); |
238 | ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, | |
239 | &np->rcv_saddr); | |
1da177e4 LT |
240 | } |
241 | ||
242 | return err; | |
243 | } | |
244 | ||
245 | if (!ipv6_addr_any(&np->rcv_saddr)) | |
246 | saddr = &np->rcv_saddr; | |
247 | ||
4c9483b2 | 248 | fl6.flowi6_proto = IPPROTO_TCP; |
4e3fd7a0 AD |
249 | fl6.daddr = np->daddr; |
250 | fl6.saddr = saddr ? *saddr : np->saddr; | |
4c9483b2 DM |
251 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
252 | fl6.flowi6_mark = sk->sk_mark; | |
1958b856 DM |
253 | fl6.fl6_dport = usin->sin6_port; |
254 | fl6.fl6_sport = inet->inet_sport; | |
1da177e4 | 255 | |
4c9483b2 | 256 | final_p = fl6_update_dst(&fl6, np->opt, &final); |
1da177e4 | 257 | |
4c9483b2 | 258 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
beb8d13b | 259 | |
4c9483b2 | 260 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true); |
68d0c6d3 DM |
261 | if (IS_ERR(dst)) { |
262 | err = PTR_ERR(dst); | |
1da177e4 | 263 | goto failure; |
14e50e57 | 264 | } |
1da177e4 LT |
265 | |
266 | if (saddr == NULL) { | |
4c9483b2 | 267 | saddr = &fl6.saddr; |
4e3fd7a0 | 268 | np->rcv_saddr = *saddr; |
1da177e4 LT |
269 | } |
270 | ||
271 | /* set the source address */ | |
4e3fd7a0 | 272 | np->saddr = *saddr; |
c720c7e8 | 273 | inet->inet_rcv_saddr = LOOPBACK4_IPV6; |
1da177e4 | 274 | |
f83ef8c0 | 275 | sk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 276 | __ip6_dst_store(sk, dst, NULL, NULL); |
1da177e4 | 277 | |
493f377d DM |
278 | rt = (struct rt6_info *) dst; |
279 | if (tcp_death_row.sysctl_tw_recycle && | |
280 | !tp->rx_opt.ts_recent_stamp && | |
81166dd6 DM |
281 | ipv6_addr_equal(&rt->rt6i_dst.addr, &np->daddr)) |
282 | tcp_fetch_timewait_stamp(sk, dst); | |
493f377d | 283 | |
d83d8461 | 284 | icsk->icsk_ext_hdr_len = 0; |
1da177e4 | 285 | if (np->opt) |
d83d8461 ACM |
286 | icsk->icsk_ext_hdr_len = (np->opt->opt_flen + |
287 | np->opt->opt_nflen); | |
1da177e4 LT |
288 | |
289 | tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
290 | ||
c720c7e8 | 291 | inet->inet_dport = usin->sin6_port; |
1da177e4 LT |
292 | |
293 | tcp_set_state(sk, TCP_SYN_SENT); | |
d8313f5c | 294 | err = inet6_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
295 | if (err) |
296 | goto late_failure; | |
297 | ||
2b916477 | 298 | if (!tp->write_seq && likely(!tp->repair)) |
1da177e4 LT |
299 | tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32, |
300 | np->daddr.s6_addr32, | |
c720c7e8 ED |
301 | inet->inet_sport, |
302 | inet->inet_dport); | |
1da177e4 LT |
303 | |
304 | err = tcp_connect(sk); | |
305 | if (err) | |
306 | goto late_failure; | |
307 | ||
308 | return 0; | |
309 | ||
310 | late_failure: | |
311 | tcp_set_state(sk, TCP_CLOSE); | |
312 | __sk_dst_reset(sk); | |
313 | failure: | |
c720c7e8 | 314 | inet->inet_dport = 0; |
1da177e4 LT |
315 | sk->sk_route_caps = 0; |
316 | return err; | |
317 | } | |
318 | ||
563d34d0 ED |
319 | static void tcp_v6_mtu_reduced(struct sock *sk) |
320 | { | |
321 | struct dst_entry *dst; | |
322 | ||
323 | if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) | |
324 | return; | |
325 | ||
326 | dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info); | |
327 | if (!dst) | |
328 | return; | |
329 | ||
330 | if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { | |
331 | tcp_sync_mss(sk, dst_mtu(dst)); | |
332 | tcp_simple_retransmit(sk); | |
333 | } | |
334 | } | |
335 | ||
1da177e4 | 336 | static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
d5fdd6ba | 337 | u8 type, u8 code, int offset, __be32 info) |
1da177e4 | 338 | { |
b71d1d42 | 339 | const struct ipv6hdr *hdr = (const struct ipv6hdr*)skb->data; |
505cbfc5 | 340 | const struct tcphdr *th = (struct tcphdr *)(skb->data+offset); |
1da177e4 LT |
341 | struct ipv6_pinfo *np; |
342 | struct sock *sk; | |
343 | int err; | |
1ab1457c | 344 | struct tcp_sock *tp; |
1da177e4 | 345 | __u32 seq; |
ca12a1a4 | 346 | struct net *net = dev_net(skb->dev); |
1da177e4 | 347 | |
ca12a1a4 | 348 | sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr, |
d86e0dac | 349 | th->dest, &hdr->saddr, th->source, skb->dev->ifindex); |
1da177e4 LT |
350 | |
351 | if (sk == NULL) { | |
e41b5368 DL |
352 | ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), |
353 | ICMP6_MIB_INERRORS); | |
1da177e4 LT |
354 | return; |
355 | } | |
356 | ||
357 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 358 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
359 | return; |
360 | } | |
361 | ||
362 | bh_lock_sock(sk); | |
563d34d0 | 363 | if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG) |
de0744af | 364 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); |
1da177e4 LT |
365 | |
366 | if (sk->sk_state == TCP_CLOSE) | |
367 | goto out; | |
368 | ||
e802af9c SH |
369 | if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) { |
370 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
371 | goto out; | |
372 | } | |
373 | ||
1da177e4 | 374 | tp = tcp_sk(sk); |
1ab1457c | 375 | seq = ntohl(th->seq); |
1da177e4 LT |
376 | if (sk->sk_state != TCP_LISTEN && |
377 | !between(seq, tp->snd_una, tp->snd_nxt)) { | |
de0744af | 378 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
379 | goto out; |
380 | } | |
381 | ||
382 | np = inet6_sk(sk); | |
383 | ||
ec18d9a2 DM |
384 | if (type == NDISC_REDIRECT) { |
385 | struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie); | |
386 | ||
1ed5c48f | 387 | if (dst) |
6700c270 | 388 | dst->ops->redirect(dst, sk, skb); |
50a75a89 | 389 | goto out; |
ec18d9a2 DM |
390 | } |
391 | ||
1da177e4 | 392 | if (type == ICMPV6_PKT_TOOBIG) { |
0d4f0608 ED |
393 | /* We are not interested in TCP_LISTEN and open_requests |
394 | * (SYN-ACKs send out by Linux are always <576bytes so | |
395 | * they should go through unfragmented). | |
396 | */ | |
397 | if (sk->sk_state == TCP_LISTEN) | |
398 | goto out; | |
399 | ||
563d34d0 ED |
400 | tp->mtu_info = ntohl(info); |
401 | if (!sock_owned_by_user(sk)) | |
402 | tcp_v6_mtu_reduced(sk); | |
d013ef2a JA |
403 | else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, |
404 | &tp->tsq_flags)) | |
405 | sock_hold(sk); | |
1da177e4 LT |
406 | goto out; |
407 | } | |
408 | ||
409 | icmpv6_err_convert(type, code, &err); | |
410 | ||
60236fdd | 411 | /* Might be for an request_sock */ |
1da177e4 | 412 | switch (sk->sk_state) { |
60236fdd | 413 | struct request_sock *req, **prev; |
1da177e4 LT |
414 | case TCP_LISTEN: |
415 | if (sock_owned_by_user(sk)) | |
416 | goto out; | |
417 | ||
8129765a ACM |
418 | req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr, |
419 | &hdr->saddr, inet6_iif(skb)); | |
1da177e4 LT |
420 | if (!req) |
421 | goto out; | |
422 | ||
423 | /* ICMPs are not backlogged, hence we cannot get | |
424 | * an established socket here. | |
425 | */ | |
547b792c | 426 | WARN_ON(req->sk != NULL); |
1da177e4 | 427 | |
2e6599cb | 428 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 429 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
430 | goto out; |
431 | } | |
432 | ||
463c84b9 | 433 | inet_csk_reqsk_queue_drop(sk, req, prev); |
5f1e942c | 434 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
435 | goto out; |
436 | ||
437 | case TCP_SYN_SENT: | |
438 | case TCP_SYN_RECV: /* Cannot happen. | |
1ab1457c | 439 | It can, it SYNs are crossed. --ANK */ |
1da177e4 | 440 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
441 | sk->sk_err = err; |
442 | sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */ | |
443 | ||
444 | tcp_done(sk); | |
445 | } else | |
446 | sk->sk_err_soft = err; | |
447 | goto out; | |
448 | } | |
449 | ||
450 | if (!sock_owned_by_user(sk) && np->recverr) { | |
451 | sk->sk_err = err; | |
452 | sk->sk_error_report(sk); | |
453 | } else | |
454 | sk->sk_err_soft = err; | |
455 | ||
456 | out: | |
457 | bh_unlock_sock(sk); | |
458 | sock_put(sk); | |
459 | } | |
460 | ||
461 | ||
9f10d3f6 NC |
462 | static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst, |
463 | struct flowi6 *fl6, | |
3840a06e | 464 | struct request_sock *req, |
fff32699 | 465 | u16 queue_mapping) |
1da177e4 | 466 | { |
ca304b61 | 467 | struct inet6_request_sock *treq = inet6_rsk(req); |
1da177e4 LT |
468 | struct ipv6_pinfo *np = inet6_sk(sk); |
469 | struct sk_buff * skb; | |
9494218f | 470 | int err = -ENOMEM; |
1da177e4 | 471 | |
9f10d3f6 NC |
472 | /* First, grab a route. */ |
473 | if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL) | |
fd80eb94 | 474 | goto done; |
9494218f | 475 | |
1a2c6181 | 476 | skb = tcp_make_synack(sk, dst, req, NULL); |
9494218f | 477 | |
1da177e4 | 478 | if (skb) { |
8ad50d96 | 479 | __tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr); |
1da177e4 | 480 | |
9f10d3f6 | 481 | fl6->daddr = treq->rmt_addr; |
fff32699 | 482 | skb_set_queue_mapping(skb, queue_mapping); |
43264e0b | 483 | err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass); |
b9df3cb8 | 484 | err = net_xmit_eval(err); |
1da177e4 LT |
485 | } |
486 | ||
487 | done: | |
1da177e4 LT |
488 | return err; |
489 | } | |
490 | ||
1a2c6181 | 491 | static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req) |
72659ecc | 492 | { |
9f10d3f6 | 493 | struct flowi6 fl6; |
e6c022a4 | 494 | int res; |
9f10d3f6 | 495 | |
1a2c6181 | 496 | res = tcp_v6_send_synack(sk, NULL, &fl6, req, 0); |
e6c022a4 ED |
497 | if (!res) |
498 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS); | |
499 | return res; | |
72659ecc OP |
500 | } |
501 | ||
60236fdd | 502 | static void tcp_v6_reqsk_destructor(struct request_sock *req) |
1da177e4 | 503 | { |
800d55f1 | 504 | kfree_skb(inet6_rsk(req)->pktopts); |
1da177e4 LT |
505 | } |
506 | ||
cfb6eeb4 YH |
507 | #ifdef CONFIG_TCP_MD5SIG |
508 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
b71d1d42 | 509 | const struct in6_addr *addr) |
cfb6eeb4 | 510 | { |
a915da9b | 511 | return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6); |
cfb6eeb4 YH |
512 | } |
513 | ||
514 | static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk, | |
515 | struct sock *addr_sk) | |
516 | { | |
517 | return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr); | |
518 | } | |
519 | ||
520 | static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk, | |
521 | struct request_sock *req) | |
522 | { | |
523 | return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr); | |
524 | } | |
525 | ||
cfb6eeb4 YH |
526 | static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval, |
527 | int optlen) | |
528 | { | |
529 | struct tcp_md5sig cmd; | |
530 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr; | |
cfb6eeb4 YH |
531 | |
532 | if (optlen < sizeof(cmd)) | |
533 | return -EINVAL; | |
534 | ||
535 | if (copy_from_user(&cmd, optval, sizeof(cmd))) | |
536 | return -EFAULT; | |
537 | ||
538 | if (sin6->sin6_family != AF_INET6) | |
539 | return -EINVAL; | |
540 | ||
541 | if (!cmd.tcpm_keylen) { | |
e773e4fa | 542 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
a915da9b ED |
543 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], |
544 | AF_INET); | |
545 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr, | |
546 | AF_INET6); | |
cfb6eeb4 YH |
547 | } |
548 | ||
549 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
550 | return -EINVAL; | |
551 | ||
a915da9b ED |
552 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
553 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], | |
554 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
cfb6eeb4 | 555 | |
a915da9b ED |
556 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr, |
557 | AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
cfb6eeb4 YH |
558 | } |
559 | ||
49a72dfb | 560 | static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
b71d1d42 ED |
561 | const struct in6_addr *daddr, |
562 | const struct in6_addr *saddr, int nbytes) | |
cfb6eeb4 | 563 | { |
cfb6eeb4 | 564 | struct tcp6_pseudohdr *bp; |
49a72dfb | 565 | struct scatterlist sg; |
8d26d76d | 566 | |
cfb6eeb4 | 567 | bp = &hp->md5_blk.ip6; |
cfb6eeb4 | 568 | /* 1. TCP pseudo-header (RFC2460) */ |
4e3fd7a0 AD |
569 | bp->saddr = *saddr; |
570 | bp->daddr = *daddr; | |
49a72dfb | 571 | bp->protocol = cpu_to_be32(IPPROTO_TCP); |
00b1304c | 572 | bp->len = cpu_to_be32(nbytes); |
cfb6eeb4 | 573 | |
49a72dfb AL |
574 | sg_init_one(&sg, bp, sizeof(*bp)); |
575 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
576 | } | |
c7da57a1 | 577 | |
49a72dfb | 578 | static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, |
b71d1d42 | 579 | const struct in6_addr *daddr, struct in6_addr *saddr, |
318cf7aa | 580 | const struct tcphdr *th) |
49a72dfb AL |
581 | { |
582 | struct tcp_md5sig_pool *hp; | |
583 | struct hash_desc *desc; | |
584 | ||
585 | hp = tcp_get_md5sig_pool(); | |
586 | if (!hp) | |
587 | goto clear_hash_noput; | |
588 | desc = &hp->md5_desc; | |
589 | ||
590 | if (crypto_hash_init(desc)) | |
591 | goto clear_hash; | |
592 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
593 | goto clear_hash; | |
594 | if (tcp_md5_hash_header(hp, th)) | |
595 | goto clear_hash; | |
596 | if (tcp_md5_hash_key(hp, key)) | |
597 | goto clear_hash; | |
598 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 | 599 | goto clear_hash; |
cfb6eeb4 | 600 | |
cfb6eeb4 | 601 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 602 | return 0; |
49a72dfb | 603 | |
cfb6eeb4 YH |
604 | clear_hash: |
605 | tcp_put_md5sig_pool(); | |
606 | clear_hash_noput: | |
607 | memset(md5_hash, 0, 16); | |
49a72dfb | 608 | return 1; |
cfb6eeb4 YH |
609 | } |
610 | ||
49a72dfb | 611 | static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
318cf7aa ED |
612 | const struct sock *sk, |
613 | const struct request_sock *req, | |
614 | const struct sk_buff *skb) | |
cfb6eeb4 | 615 | { |
b71d1d42 | 616 | const struct in6_addr *saddr, *daddr; |
49a72dfb AL |
617 | struct tcp_md5sig_pool *hp; |
618 | struct hash_desc *desc; | |
318cf7aa | 619 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
620 | |
621 | if (sk) { | |
622 | saddr = &inet6_sk(sk)->saddr; | |
623 | daddr = &inet6_sk(sk)->daddr; | |
49a72dfb | 624 | } else if (req) { |
cfb6eeb4 YH |
625 | saddr = &inet6_rsk(req)->loc_addr; |
626 | daddr = &inet6_rsk(req)->rmt_addr; | |
49a72dfb | 627 | } else { |
b71d1d42 | 628 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
49a72dfb AL |
629 | saddr = &ip6h->saddr; |
630 | daddr = &ip6h->daddr; | |
cfb6eeb4 | 631 | } |
49a72dfb AL |
632 | |
633 | hp = tcp_get_md5sig_pool(); | |
634 | if (!hp) | |
635 | goto clear_hash_noput; | |
636 | desc = &hp->md5_desc; | |
637 | ||
638 | if (crypto_hash_init(desc)) | |
639 | goto clear_hash; | |
640 | ||
641 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
642 | goto clear_hash; | |
643 | if (tcp_md5_hash_header(hp, th)) | |
644 | goto clear_hash; | |
645 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
646 | goto clear_hash; | |
647 | if (tcp_md5_hash_key(hp, key)) | |
648 | goto clear_hash; | |
649 | if (crypto_hash_final(desc, md5_hash)) | |
650 | goto clear_hash; | |
651 | ||
652 | tcp_put_md5sig_pool(); | |
653 | return 0; | |
654 | ||
655 | clear_hash: | |
656 | tcp_put_md5sig_pool(); | |
657 | clear_hash_noput: | |
658 | memset(md5_hash, 0, 16); | |
659 | return 1; | |
cfb6eeb4 YH |
660 | } |
661 | ||
318cf7aa | 662 | static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb) |
cfb6eeb4 | 663 | { |
cf533ea5 | 664 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 665 | struct tcp_md5sig_key *hash_expected; |
b71d1d42 | 666 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
318cf7aa | 667 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 668 | int genhash; |
cfb6eeb4 YH |
669 | u8 newhash[16]; |
670 | ||
671 | hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr); | |
7d5d5525 | 672 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 673 | |
785957d3 DM |
674 | /* We've parsed the options - do we have a hash? */ |
675 | if (!hash_expected && !hash_location) | |
676 | return 0; | |
677 | ||
678 | if (hash_expected && !hash_location) { | |
679 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); | |
cfb6eeb4 YH |
680 | return 1; |
681 | } | |
682 | ||
785957d3 DM |
683 | if (!hash_expected && hash_location) { |
684 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); | |
cfb6eeb4 YH |
685 | return 1; |
686 | } | |
687 | ||
688 | /* check the signature */ | |
49a72dfb AL |
689 | genhash = tcp_v6_md5_hash_skb(newhash, |
690 | hash_expected, | |
691 | NULL, NULL, skb); | |
692 | ||
cfb6eeb4 | 693 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { |
e87cc472 JP |
694 | net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n", |
695 | genhash ? "failed" : "mismatch", | |
696 | &ip6h->saddr, ntohs(th->source), | |
697 | &ip6h->daddr, ntohs(th->dest)); | |
cfb6eeb4 YH |
698 | return 1; |
699 | } | |
700 | return 0; | |
701 | } | |
702 | #endif | |
703 | ||
c6aefafb | 704 | struct request_sock_ops tcp6_request_sock_ops __read_mostly = { |
1da177e4 | 705 | .family = AF_INET6, |
2e6599cb | 706 | .obj_size = sizeof(struct tcp6_request_sock), |
72659ecc | 707 | .rtx_syn_ack = tcp_v6_rtx_synack, |
60236fdd ACM |
708 | .send_ack = tcp_v6_reqsk_send_ack, |
709 | .destructor = tcp_v6_reqsk_destructor, | |
72659ecc OP |
710 | .send_reset = tcp_v6_send_reset, |
711 | .syn_ack_timeout = tcp_syn_ack_timeout, | |
1da177e4 LT |
712 | }; |
713 | ||
cfb6eeb4 | 714 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 715 | static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = { |
cfb6eeb4 | 716 | .md5_lookup = tcp_v6_reqsk_md5_lookup, |
e3afe7b7 | 717 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 718 | }; |
b6332e6c | 719 | #endif |
cfb6eeb4 | 720 | |
626e264d | 721 | static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win, |
ee684b6f AV |
722 | u32 tsval, u32 tsecr, |
723 | struct tcp_md5sig_key *key, int rst, u8 tclass) | |
1da177e4 | 724 | { |
cf533ea5 ED |
725 | const struct tcphdr *th = tcp_hdr(skb); |
726 | struct tcphdr *t1; | |
1da177e4 | 727 | struct sk_buff *buff; |
4c9483b2 | 728 | struct flowi6 fl6; |
adf30907 | 729 | struct net *net = dev_net(skb_dst(skb)->dev); |
e5047992 | 730 | struct sock *ctl_sk = net->ipv6.tcp_sk; |
77c676da | 731 | unsigned int tot_len = sizeof(struct tcphdr); |
adf30907 | 732 | struct dst_entry *dst; |
81ada62d | 733 | __be32 *topt; |
1da177e4 | 734 | |
ee684b6f | 735 | if (tsecr) |
626e264d | 736 | tot_len += TCPOLEN_TSTAMP_ALIGNED; |
cfb6eeb4 | 737 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
738 | if (key) |
739 | tot_len += TCPOLEN_MD5SIG_ALIGNED; | |
740 | #endif | |
741 | ||
cfb6eeb4 | 742 | buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len, |
1da177e4 | 743 | GFP_ATOMIC); |
1ab1457c YH |
744 | if (buff == NULL) |
745 | return; | |
1da177e4 | 746 | |
cfb6eeb4 | 747 | skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len); |
1da177e4 | 748 | |
cfb6eeb4 | 749 | t1 = (struct tcphdr *) skb_push(buff, tot_len); |
6651ffc8 | 750 | skb_reset_transport_header(buff); |
1da177e4 LT |
751 | |
752 | /* Swap the send and the receive. */ | |
753 | memset(t1, 0, sizeof(*t1)); | |
754 | t1->dest = th->source; | |
755 | t1->source = th->dest; | |
cfb6eeb4 | 756 | t1->doff = tot_len / 4; |
626e264d IJ |
757 | t1->seq = htonl(seq); |
758 | t1->ack_seq = htonl(ack); | |
759 | t1->ack = !rst || !th->ack; | |
760 | t1->rst = rst; | |
761 | t1->window = htons(win); | |
1da177e4 | 762 | |
81ada62d IJ |
763 | topt = (__be32 *)(t1 + 1); |
764 | ||
ee684b6f | 765 | if (tsecr) { |
626e264d IJ |
766 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
767 | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
768 | *topt++ = htonl(tsval); |
769 | *topt++ = htonl(tsecr); | |
626e264d IJ |
770 | } |
771 | ||
cfb6eeb4 YH |
772 | #ifdef CONFIG_TCP_MD5SIG |
773 | if (key) { | |
81ada62d IJ |
774 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
775 | (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG); | |
776 | tcp_v6_md5_hash_hdr((__u8 *)topt, key, | |
78e645cb IJ |
777 | &ipv6_hdr(skb)->saddr, |
778 | &ipv6_hdr(skb)->daddr, t1); | |
cfb6eeb4 YH |
779 | } |
780 | #endif | |
781 | ||
4c9483b2 | 782 | memset(&fl6, 0, sizeof(fl6)); |
4e3fd7a0 AD |
783 | fl6.daddr = ipv6_hdr(skb)->saddr; |
784 | fl6.saddr = ipv6_hdr(skb)->daddr; | |
1da177e4 | 785 | |
e5700aff DM |
786 | buff->ip_summed = CHECKSUM_PARTIAL; |
787 | buff->csum = 0; | |
788 | ||
4c9483b2 | 789 | __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr); |
1da177e4 | 790 | |
4c9483b2 | 791 | fl6.flowi6_proto = IPPROTO_TCP; |
4c675258 AK |
792 | if (ipv6_addr_type(&fl6.daddr) & IPV6_ADDR_LINKLOCAL) |
793 | fl6.flowi6_oif = inet6_iif(skb); | |
1958b856 DM |
794 | fl6.fl6_dport = t1->dest; |
795 | fl6.fl6_sport = t1->source; | |
4c9483b2 | 796 | security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); |
1da177e4 | 797 | |
c20121ae DL |
798 | /* Pass a socket to ip6_dst_lookup either it is for RST |
799 | * Underlying function will use this to retrieve the network | |
800 | * namespace | |
801 | */ | |
4c9483b2 | 802 | dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false); |
68d0c6d3 DM |
803 | if (!IS_ERR(dst)) { |
804 | skb_dst_set(buff, dst); | |
b903d324 | 805 | ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass); |
68d0c6d3 DM |
806 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
807 | if (rst) | |
808 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
809 | return; | |
1da177e4 LT |
810 | } |
811 | ||
812 | kfree_skb(buff); | |
813 | } | |
814 | ||
626e264d | 815 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 816 | { |
cf533ea5 | 817 | const struct tcphdr *th = tcp_hdr(skb); |
626e264d | 818 | u32 seq = 0, ack_seq = 0; |
fa3e5b4e | 819 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
820 | #ifdef CONFIG_TCP_MD5SIG |
821 | const __u8 *hash_location = NULL; | |
822 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
823 | unsigned char newhash[16]; | |
824 | int genhash; | |
825 | struct sock *sk1 = NULL; | |
826 | #endif | |
1da177e4 | 827 | |
626e264d | 828 | if (th->rst) |
1da177e4 LT |
829 | return; |
830 | ||
626e264d IJ |
831 | if (!ipv6_unicast_destination(skb)) |
832 | return; | |
1da177e4 | 833 | |
cfb6eeb4 | 834 | #ifdef CONFIG_TCP_MD5SIG |
658ddaaf SL |
835 | hash_location = tcp_parse_md5sig_option(th); |
836 | if (!sk && hash_location) { | |
837 | /* | |
838 | * active side is lost. Try to find listening socket through | |
839 | * source port, and then find md5 key through listening socket. | |
840 | * we are not loose security here: | |
841 | * Incoming packet is checked with md5 hash with finding key, | |
842 | * no RST generated if md5 hash doesn't match. | |
843 | */ | |
844 | sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev), | |
5ba24953 TH |
845 | &tcp_hashinfo, &ipv6h->saddr, |
846 | th->source, &ipv6h->daddr, | |
658ddaaf SL |
847 | ntohs(th->source), inet6_iif(skb)); |
848 | if (!sk1) | |
849 | return; | |
850 | ||
851 | rcu_read_lock(); | |
852 | key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr); | |
853 | if (!key) | |
854 | goto release_sk1; | |
855 | ||
856 | genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb); | |
857 | if (genhash || memcmp(hash_location, newhash, 16) != 0) | |
858 | goto release_sk1; | |
859 | } else { | |
860 | key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL; | |
861 | } | |
cfb6eeb4 YH |
862 | #endif |
863 | ||
626e264d IJ |
864 | if (th->ack) |
865 | seq = ntohl(th->ack_seq); | |
866 | else | |
867 | ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len - | |
868 | (th->doff << 2); | |
1da177e4 | 869 | |
ee684b6f | 870 | tcp_v6_send_response(skb, seq, ack_seq, 0, 0, 0, key, 1, 0); |
658ddaaf SL |
871 | |
872 | #ifdef CONFIG_TCP_MD5SIG | |
873 | release_sk1: | |
874 | if (sk1) { | |
875 | rcu_read_unlock(); | |
876 | sock_put(sk1); | |
877 | } | |
878 | #endif | |
626e264d | 879 | } |
1da177e4 | 880 | |
ee684b6f AV |
881 | static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, |
882 | u32 win, u32 tsval, u32 tsecr, | |
b903d324 | 883 | struct tcp_md5sig_key *key, u8 tclass) |
626e264d | 884 | { |
ee684b6f | 885 | tcp_v6_send_response(skb, seq, ack, win, tsval, tsecr, key, 0, tclass); |
1da177e4 LT |
886 | } |
887 | ||
888 | static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
889 | { | |
8feaf0c0 | 890 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 891 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 892 | |
9501f972 | 893 | tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
8feaf0c0 | 894 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
ee684b6f | 895 | tcp_time_stamp + tcptw->tw_ts_offset, |
b903d324 ED |
896 | tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw), |
897 | tw->tw_tclass); | |
1da177e4 | 898 | |
8feaf0c0 | 899 | inet_twsk_put(tw); |
1da177e4 LT |
900 | } |
901 | ||
6edafaaf GJ |
902 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
903 | struct request_sock *req) | |
1da177e4 | 904 | { |
ee684b6f AV |
905 | tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, |
906 | req->rcv_wnd, tcp_time_stamp, req->ts_recent, | |
b903d324 | 907 | tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0); |
1da177e4 LT |
908 | } |
909 | ||
910 | ||
911 | static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb) | |
912 | { | |
60236fdd | 913 | struct request_sock *req, **prev; |
aa8223c7 | 914 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
915 | struct sock *nsk; |
916 | ||
917 | /* Find possible connection requests. */ | |
8129765a | 918 | req = inet6_csk_search_req(sk, &prev, th->source, |
0660e03f ACM |
919 | &ipv6_hdr(skb)->saddr, |
920 | &ipv6_hdr(skb)->daddr, inet6_iif(skb)); | |
1da177e4 | 921 | if (req) |
8336886f | 922 | return tcp_check_req(sk, skb, req, prev, false); |
1da177e4 | 923 | |
3b1e0a65 | 924 | nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo, |
d86e0dac PE |
925 | &ipv6_hdr(skb)->saddr, th->source, |
926 | &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb)); | |
1da177e4 LT |
927 | |
928 | if (nsk) { | |
929 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
930 | bh_lock_sock(nsk); | |
931 | return nsk; | |
932 | } | |
9469c7b4 | 933 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
934 | return NULL; |
935 | } | |
936 | ||
c6aefafb | 937 | #ifdef CONFIG_SYN_COOKIES |
af9b4738 | 938 | if (!th->syn) |
c6aefafb | 939 | sk = cookie_v6_check(sk, skb); |
1da177e4 LT |
940 | #endif |
941 | return sk; | |
942 | } | |
943 | ||
1da177e4 LT |
944 | /* FIXME: this is substantially similar to the ipv4 code. |
945 | * Can some kind of merge be done? -- erics | |
946 | */ | |
947 | static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb) | |
948 | { | |
e6b4d113 WAS |
949 | struct tcp_options_received tmp_opt; |
950 | struct request_sock *req; | |
ca304b61 | 951 | struct inet6_request_sock *treq; |
1da177e4 | 952 | struct ipv6_pinfo *np = inet6_sk(sk); |
1da177e4 | 953 | struct tcp_sock *tp = tcp_sk(sk); |
e6b4d113 | 954 | __u32 isn = TCP_SKB_CB(skb)->when; |
493f377d | 955 | struct dst_entry *dst = NULL; |
3840a06e | 956 | struct flowi6 fl6; |
a2a385d6 | 957 | bool want_cookie = false; |
1da177e4 LT |
958 | |
959 | if (skb->protocol == htons(ETH_P_IP)) | |
960 | return tcp_v4_conn_request(sk, skb); | |
961 | ||
962 | if (!ipv6_unicast_destination(skb)) | |
1ab1457c | 963 | goto drop; |
1da177e4 | 964 | |
463c84b9 | 965 | if (inet_csk_reqsk_queue_is_full(sk) && !isn) { |
946cedcc ED |
966 | want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6"); |
967 | if (!want_cookie) | |
968 | goto drop; | |
1da177e4 LT |
969 | } |
970 | ||
5f1e942c VS |
971 | if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) { |
972 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); | |
1da177e4 | 973 | goto drop; |
5f1e942c | 974 | } |
1da177e4 | 975 | |
ca304b61 | 976 | req = inet6_reqsk_alloc(&tcp6_request_sock_ops); |
1da177e4 LT |
977 | if (req == NULL) |
978 | goto drop; | |
979 | ||
cfb6eeb4 YH |
980 | #ifdef CONFIG_TCP_MD5SIG |
981 | tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops; | |
982 | #endif | |
983 | ||
1da177e4 LT |
984 | tcp_clear_options(&tmp_opt); |
985 | tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
986 | tmp_opt.user_mss = tp->rx_opt.user_mss; | |
1a2c6181 | 987 | tcp_parse_options(skb, &tmp_opt, 0, NULL); |
1da177e4 | 988 | |
4dfc2817 | 989 | if (want_cookie && !tmp_opt.saw_tstamp) |
c6aefafb | 990 | tcp_clear_options(&tmp_opt); |
c6aefafb | 991 | |
1da177e4 LT |
992 | tmp_opt.tstamp_ok = tmp_opt.saw_tstamp; |
993 | tcp_openreq_init(req, &tmp_opt, skb); | |
994 | ||
ca304b61 | 995 | treq = inet6_rsk(req); |
4e3fd7a0 AD |
996 | treq->rmt_addr = ipv6_hdr(skb)->saddr; |
997 | treq->loc_addr = ipv6_hdr(skb)->daddr; | |
172d69e6 | 998 | if (!want_cookie || tmp_opt.tstamp_ok) |
5d134f1c | 999 | TCP_ECN_create_request(req, skb, sock_net(sk)); |
c6aefafb | 1000 | |
4d0fe50c ED |
1001 | treq->iif = sk->sk_bound_dev_if; |
1002 | ||
1003 | /* So that link locals have meaning */ | |
1004 | if (!sk->sk_bound_dev_if && | |
1005 | ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL) | |
1006 | treq->iif = inet6_iif(skb); | |
1007 | ||
2bbdf389 | 1008 | if (!isn) { |
c6aefafb GG |
1009 | if (ipv6_opt_accepted(sk, skb) || |
1010 | np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || | |
1011 | np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { | |
1012 | atomic_inc(&skb->users); | |
1013 | treq->pktopts = skb; | |
1014 | } | |
493f377d DM |
1015 | |
1016 | if (want_cookie) { | |
2bbdf389 FW |
1017 | isn = cookie_v6_init_sequence(sk, skb, &req->mss); |
1018 | req->cookie_ts = tmp_opt.tstamp_ok; | |
493f377d DM |
1019 | goto have_isn; |
1020 | } | |
1021 | ||
1022 | /* VJ's idea. We save last timestamp seen | |
1023 | * from the destination in peer table, when entering | |
1024 | * state TIME-WAIT, and check against it before | |
1025 | * accepting new connection request. | |
1026 | * | |
1027 | * If "isn" is not zero, this request hit alive | |
1028 | * timewait bucket, so that all the necessary checks | |
1029 | * are made in the function processing timewait state. | |
1030 | */ | |
1031 | if (tmp_opt.saw_tstamp && | |
1032 | tcp_death_row.sysctl_tw_recycle && | |
81166dd6 DM |
1033 | (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL) { |
1034 | if (!tcp_peer_is_proven(req, dst, true)) { | |
493f377d DM |
1035 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED); |
1036 | goto drop_and_release; | |
1037 | } | |
1038 | } | |
1039 | /* Kill the following clause, if you dislike this way. */ | |
1040 | else if (!sysctl_tcp_syncookies && | |
1041 | (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < | |
1042 | (sysctl_max_syn_backlog >> 2)) && | |
81166dd6 | 1043 | !tcp_peer_is_proven(req, dst, false)) { |
493f377d DM |
1044 | /* Without syncookies last quarter of |
1045 | * backlog is filled with destinations, | |
1046 | * proven to be alive. | |
1047 | * It means that we continue to communicate | |
1048 | * to destinations, already remembered | |
1049 | * to the moment of synflood. | |
1050 | */ | |
1051 | LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n", | |
1052 | &treq->rmt_addr, ntohs(tcp_hdr(skb)->source)); | |
1053 | goto drop_and_release; | |
2bbdf389 | 1054 | } |
493f377d DM |
1055 | |
1056 | isn = tcp_v6_init_sequence(skb); | |
c6aefafb | 1057 | } |
493f377d | 1058 | have_isn: |
2e6599cb | 1059 | tcp_rsk(req)->snt_isn = isn; |
1da177e4 | 1060 | |
437c5b53 NC |
1061 | if (security_inet_conn_request(sk, skb, req)) |
1062 | goto drop_and_release; | |
4237c75c | 1063 | |
9f10d3f6 | 1064 | if (tcp_v6_send_synack(sk, dst, &fl6, req, |
fff32699 | 1065 | skb_get_queue_mapping(skb)) || |
4957faad | 1066 | want_cookie) |
e6b4d113 | 1067 | goto drop_and_free; |
1da177e4 | 1068 | |
016818d0 | 1069 | tcp_rsk(req)->snt_synack = tcp_time_stamp; |
8336886f | 1070 | tcp_rsk(req)->listener = NULL; |
e6b4d113 WAS |
1071 | inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT); |
1072 | return 0; | |
1da177e4 | 1073 | |
493f377d DM |
1074 | drop_and_release: |
1075 | dst_release(dst); | |
e6b4d113 WAS |
1076 | drop_and_free: |
1077 | reqsk_free(req); | |
1da177e4 | 1078 | drop: |
5f1e942c | 1079 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
1080 | return 0; /* don't send reset */ |
1081 | } | |
1082 | ||
1083 | static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1084 | struct request_sock *req, |
1da177e4 LT |
1085 | struct dst_entry *dst) |
1086 | { | |
78d15e82 | 1087 | struct inet6_request_sock *treq; |
1da177e4 LT |
1088 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); |
1089 | struct tcp6_sock *newtcp6sk; | |
1090 | struct inet_sock *newinet; | |
1091 | struct tcp_sock *newtp; | |
1092 | struct sock *newsk; | |
cfb6eeb4 YH |
1093 | #ifdef CONFIG_TCP_MD5SIG |
1094 | struct tcp_md5sig_key *key; | |
1095 | #endif | |
3840a06e | 1096 | struct flowi6 fl6; |
1da177e4 LT |
1097 | |
1098 | if (skb->protocol == htons(ETH_P_IP)) { | |
1099 | /* | |
1100 | * v6 mapped | |
1101 | */ | |
1102 | ||
1103 | newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst); | |
1104 | ||
1ab1457c | 1105 | if (newsk == NULL) |
1da177e4 LT |
1106 | return NULL; |
1107 | ||
1108 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1109 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1110 | ||
1111 | newinet = inet_sk(newsk); | |
1112 | newnp = inet6_sk(newsk); | |
1113 | newtp = tcp_sk(newsk); | |
1114 | ||
1115 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1116 | ||
c720c7e8 | 1117 | ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr); |
1da177e4 | 1118 | |
c720c7e8 | 1119 | ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); |
1da177e4 | 1120 | |
4e3fd7a0 | 1121 | newnp->rcv_saddr = newnp->saddr; |
1da177e4 | 1122 | |
8292a17a | 1123 | inet_csk(newsk)->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 1124 | newsk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
1125 | #ifdef CONFIG_TCP_MD5SIG |
1126 | newtp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
1127 | #endif | |
1128 | ||
676a1184 YZ |
1129 | newnp->ipv6_ac_list = NULL; |
1130 | newnp->ipv6_fl_list = NULL; | |
1da177e4 LT |
1131 | newnp->pktoptions = NULL; |
1132 | newnp->opt = NULL; | |
505cbfc5 | 1133 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1134 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
e7219858 | 1135 | newnp->rcv_tclass = ipv6_get_dsfield(ipv6_hdr(skb)); |
1da177e4 | 1136 | |
e6848976 ACM |
1137 | /* |
1138 | * No need to charge this sock to the relevant IPv6 refcnt debug socks count | |
1139 | * here, tcp_create_openreq_child now does this for us, see the comment in | |
1140 | * that function for the gory details. -acme | |
1da177e4 | 1141 | */ |
1da177e4 LT |
1142 | |
1143 | /* It is tricky place. Until this moment IPv4 tcp | |
8292a17a | 1144 | worked with IPv6 icsk.icsk_af_ops. |
1da177e4 LT |
1145 | Sync it now. |
1146 | */ | |
d83d8461 | 1147 | tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); |
1da177e4 LT |
1148 | |
1149 | return newsk; | |
1150 | } | |
1151 | ||
78d15e82 | 1152 | treq = inet6_rsk(req); |
1da177e4 LT |
1153 | |
1154 | if (sk_acceptq_is_full(sk)) | |
1155 | goto out_overflow; | |
1156 | ||
493f377d | 1157 | if (!dst) { |
3840a06e | 1158 | dst = inet6_csk_route_req(sk, &fl6, req); |
493f377d | 1159 | if (!dst) |
1da177e4 | 1160 | goto out; |
1ab1457c | 1161 | } |
1da177e4 LT |
1162 | |
1163 | newsk = tcp_create_openreq_child(sk, req, skb); | |
1164 | if (newsk == NULL) | |
093d2823 | 1165 | goto out_nonewsk; |
1da177e4 | 1166 | |
e6848976 ACM |
1167 | /* |
1168 | * No need to charge this sock to the relevant IPv6 refcnt debug socks | |
1169 | * count here, tcp_create_openreq_child now does this for us, see the | |
1170 | * comment in that function for the gory details. -acme | |
1171 | */ | |
1da177e4 | 1172 | |
59eed279 | 1173 | newsk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 1174 | __ip6_dst_store(newsk, dst, NULL, NULL); |
fae6ef87 | 1175 | inet6_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1176 | |
1177 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1178 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1179 | ||
1180 | newtp = tcp_sk(newsk); | |
1181 | newinet = inet_sk(newsk); | |
1182 | newnp = inet6_sk(newsk); | |
1183 | ||
1184 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1185 | ||
4e3fd7a0 AD |
1186 | newnp->daddr = treq->rmt_addr; |
1187 | newnp->saddr = treq->loc_addr; | |
1188 | newnp->rcv_saddr = treq->loc_addr; | |
2e6599cb | 1189 | newsk->sk_bound_dev_if = treq->iif; |
1da177e4 | 1190 | |
1ab1457c | 1191 | /* Now IPv6 options... |
1da177e4 LT |
1192 | |
1193 | First: no IPv4 options. | |
1194 | */ | |
f6d8bd05 | 1195 | newinet->inet_opt = NULL; |
676a1184 | 1196 | newnp->ipv6_ac_list = NULL; |
d35690be | 1197 | newnp->ipv6_fl_list = NULL; |
1da177e4 LT |
1198 | |
1199 | /* Clone RX bits */ | |
1200 | newnp->rxopt.all = np->rxopt.all; | |
1201 | ||
1202 | /* Clone pktoptions received with SYN */ | |
1203 | newnp->pktoptions = NULL; | |
2e6599cb | 1204 | if (treq->pktopts != NULL) { |
99a1dec7 MG |
1205 | newnp->pktoptions = skb_clone(treq->pktopts, |
1206 | sk_gfp_atomic(sk, GFP_ATOMIC)); | |
ab185d7b | 1207 | consume_skb(treq->pktopts); |
2e6599cb | 1208 | treq->pktopts = NULL; |
1da177e4 LT |
1209 | if (newnp->pktoptions) |
1210 | skb_set_owner_r(newnp->pktoptions, newsk); | |
1211 | } | |
1212 | newnp->opt = NULL; | |
505cbfc5 | 1213 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1214 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
e7219858 | 1215 | newnp->rcv_tclass = ipv6_get_dsfield(ipv6_hdr(skb)); |
1da177e4 LT |
1216 | |
1217 | /* Clone native IPv6 options from listening socket (if any) | |
1218 | ||
1219 | Yes, keeping reference count would be much more clever, | |
1220 | but we make one more one thing there: reattach optmem | |
1221 | to newsk. | |
1222 | */ | |
43264e0b RL |
1223 | if (np->opt) |
1224 | newnp->opt = ipv6_dup_options(newsk, np->opt); | |
1da177e4 | 1225 | |
d83d8461 | 1226 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
1da177e4 | 1227 | if (newnp->opt) |
d83d8461 ACM |
1228 | inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + |
1229 | newnp->opt->opt_flen); | |
1da177e4 | 1230 | |
5d424d5a | 1231 | tcp_mtup_init(newsk); |
1da177e4 | 1232 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1233 | newtp->advmss = dst_metric_advmss(dst); |
d135c522 NC |
1234 | if (tcp_sk(sk)->rx_opt.user_mss && |
1235 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1236 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1237 | ||
1da177e4 | 1238 | tcp_initialize_rcv_mss(newsk); |
623df484 | 1239 | tcp_synack_rtt_meas(newsk, req); |
e6c022a4 | 1240 | newtp->total_retrans = req->num_retrans; |
1da177e4 | 1241 | |
c720c7e8 ED |
1242 | newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; |
1243 | newinet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
1da177e4 | 1244 | |
cfb6eeb4 YH |
1245 | #ifdef CONFIG_TCP_MD5SIG |
1246 | /* Copy over the MD5 key from the original socket */ | |
1247 | if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) { | |
1248 | /* We're using one, so create a matching key | |
1249 | * on the newsk structure. If we fail to get | |
1250 | * memory, then we end up not copying the key | |
1251 | * across. Shucks. | |
1252 | */ | |
a915da9b | 1253 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr, |
99a1dec7 MG |
1254 | AF_INET6, key->key, key->keylen, |
1255 | sk_gfp_atomic(sk, GFP_ATOMIC)); | |
cfb6eeb4 YH |
1256 | } |
1257 | #endif | |
1258 | ||
093d2823 | 1259 | if (__inet_inherit_port(sk, newsk) < 0) { |
e337e24d CP |
1260 | inet_csk_prepare_forced_close(newsk); |
1261 | tcp_done(newsk); | |
093d2823 BS |
1262 | goto out; |
1263 | } | |
9327f705 | 1264 | __inet6_hash(newsk, NULL); |
1da177e4 LT |
1265 | |
1266 | return newsk; | |
1267 | ||
1268 | out_overflow: | |
de0744af | 1269 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 | 1270 | out_nonewsk: |
1da177e4 | 1271 | dst_release(dst); |
093d2823 BS |
1272 | out: |
1273 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
1da177e4 LT |
1274 | return NULL; |
1275 | } | |
1276 | ||
b51655b9 | 1277 | static __sum16 tcp_v6_checksum_init(struct sk_buff *skb) |
1da177e4 | 1278 | { |
84fa7933 | 1279 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
684f2176 | 1280 | if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr, |
0660e03f | 1281 | &ipv6_hdr(skb)->daddr, skb->csum)) { |
fb286bb2 | 1282 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
1da177e4 | 1283 | return 0; |
fb286bb2 | 1284 | } |
1da177e4 | 1285 | } |
fb286bb2 | 1286 | |
684f2176 | 1287 | skb->csum = ~csum_unfold(tcp_v6_check(skb->len, |
0660e03f ACM |
1288 | &ipv6_hdr(skb)->saddr, |
1289 | &ipv6_hdr(skb)->daddr, 0)); | |
fb286bb2 | 1290 | |
1da177e4 | 1291 | if (skb->len <= 76) { |
fb286bb2 | 1292 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1293 | } |
1294 | return 0; | |
1295 | } | |
1296 | ||
1297 | /* The socket must have it's spinlock held when we get | |
1298 | * here. | |
1299 | * | |
1300 | * We have a potential double-lock case here, so even when | |
1301 | * doing backlog processing we use the BH locking scheme. | |
1302 | * This is because we cannot sleep with the original spinlock | |
1303 | * held. | |
1304 | */ | |
1305 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1306 | { | |
1307 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1308 | struct tcp_sock *tp; | |
1309 | struct sk_buff *opt_skb = NULL; | |
1310 | ||
1311 | /* Imagine: socket is IPv6. IPv4 packet arrives, | |
1312 | goes to IPv4 receive handler and backlogged. | |
1313 | From backlog it always goes here. Kerboom... | |
1314 | Fortunately, tcp_rcv_established and rcv_established | |
1315 | handle them correctly, but it is not case with | |
1316 | tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK | |
1317 | */ | |
1318 | ||
1319 | if (skb->protocol == htons(ETH_P_IP)) | |
1320 | return tcp_v4_do_rcv(sk, skb); | |
1321 | ||
cfb6eeb4 YH |
1322 | #ifdef CONFIG_TCP_MD5SIG |
1323 | if (tcp_v6_inbound_md5_hash (sk, skb)) | |
1324 | goto discard; | |
1325 | #endif | |
1326 | ||
fda9ef5d | 1327 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1328 | goto discard; |
1329 | ||
1330 | /* | |
1331 | * socket locking is here for SMP purposes as backlog rcv | |
1332 | * is currently called with bh processing disabled. | |
1333 | */ | |
1334 | ||
1335 | /* Do Stevens' IPV6_PKTOPTIONS. | |
1336 | ||
1337 | Yes, guys, it is the only place in our code, where we | |
1338 | may make it not affecting IPv4. | |
1339 | The rest of code is protocol independent, | |
1340 | and I do not like idea to uglify IPv4. | |
1341 | ||
1342 | Actually, all the idea behind IPV6_PKTOPTIONS | |
1343 | looks not very well thought. For now we latch | |
1344 | options, received in the last packet, enqueued | |
1345 | by tcp. Feel free to propose better solution. | |
1ab1457c | 1346 | --ANK (980728) |
1da177e4 LT |
1347 | */ |
1348 | if (np->rxopt.all) | |
99a1dec7 | 1349 | opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC)); |
1da177e4 LT |
1350 | |
1351 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ | |
5d299f3d ED |
1352 | struct dst_entry *dst = sk->sk_rx_dst; |
1353 | ||
bdeab991 | 1354 | sock_rps_save_rxhash(sk, skb); |
5d299f3d ED |
1355 | if (dst) { |
1356 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || | |
1357 | dst->ops->check(dst, np->rx_dst_cookie) == NULL) { | |
1358 | dst_release(dst); | |
1359 | sk->sk_rx_dst = NULL; | |
1360 | } | |
1361 | } | |
1362 | ||
aa8223c7 | 1363 | if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1364 | goto reset; |
1da177e4 LT |
1365 | if (opt_skb) |
1366 | goto ipv6_pktoptions; | |
1367 | return 0; | |
1368 | } | |
1369 | ||
ab6a5bb6 | 1370 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1371 | goto csum_err; |
1372 | ||
1ab1457c | 1373 | if (sk->sk_state == TCP_LISTEN) { |
1da177e4 LT |
1374 | struct sock *nsk = tcp_v6_hnd_req(sk, skb); |
1375 | if (!nsk) | |
1376 | goto discard; | |
1377 | ||
1378 | /* | |
1379 | * Queue it on the new socket if the new socket is active, | |
1380 | * otherwise we just shortcircuit this and continue with | |
1381 | * the new socket.. | |
1382 | */ | |
1ab1457c | 1383 | if(nsk != sk) { |
bdeab991 | 1384 | sock_rps_save_rxhash(nsk, skb); |
1da177e4 LT |
1385 | if (tcp_child_process(sk, nsk, skb)) |
1386 | goto reset; | |
1387 | if (opt_skb) | |
1388 | __kfree_skb(opt_skb); | |
1389 | return 0; | |
1390 | } | |
47482f13 | 1391 | } else |
bdeab991 | 1392 | sock_rps_save_rxhash(sk, skb); |
1da177e4 | 1393 | |
aa8223c7 | 1394 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1395 | goto reset; |
1da177e4 LT |
1396 | if (opt_skb) |
1397 | goto ipv6_pktoptions; | |
1398 | return 0; | |
1399 | ||
1400 | reset: | |
cfb6eeb4 | 1401 | tcp_v6_send_reset(sk, skb); |
1da177e4 LT |
1402 | discard: |
1403 | if (opt_skb) | |
1404 | __kfree_skb(opt_skb); | |
1405 | kfree_skb(skb); | |
1406 | return 0; | |
1407 | csum_err: | |
6a5dc9e5 | 1408 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS); |
63231bdd | 1409 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1410 | goto discard; |
1411 | ||
1412 | ||
1413 | ipv6_pktoptions: | |
1414 | /* Do you ask, what is it? | |
1415 | ||
1416 | 1. skb was enqueued by tcp. | |
1417 | 2. skb is added to tail of read queue, rather than out of order. | |
1418 | 3. socket is not in passive state. | |
1419 | 4. Finally, it really contains options, which user wants to receive. | |
1420 | */ | |
1421 | tp = tcp_sk(sk); | |
1422 | if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt && | |
1423 | !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { | |
333fad53 | 1424 | if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) |
505cbfc5 | 1425 | np->mcast_oif = inet6_iif(opt_skb); |
333fad53 | 1426 | if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) |
0660e03f | 1427 | np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; |
4c507d28 | 1428 | if (np->rxopt.bits.rxtclass) |
e7219858 | 1429 | np->rcv_tclass = ipv6_get_dsfield(ipv6_hdr(skb)); |
1da177e4 LT |
1430 | if (ipv6_opt_accepted(sk, opt_skb)) { |
1431 | skb_set_owner_r(opt_skb, sk); | |
1432 | opt_skb = xchg(&np->pktoptions, opt_skb); | |
1433 | } else { | |
1434 | __kfree_skb(opt_skb); | |
1435 | opt_skb = xchg(&np->pktoptions, NULL); | |
1436 | } | |
1437 | } | |
1438 | ||
800d55f1 | 1439 | kfree_skb(opt_skb); |
1da177e4 LT |
1440 | return 0; |
1441 | } | |
1442 | ||
e5bbef20 | 1443 | static int tcp_v6_rcv(struct sk_buff *skb) |
1da177e4 | 1444 | { |
cf533ea5 | 1445 | const struct tcphdr *th; |
b71d1d42 | 1446 | const struct ipv6hdr *hdr; |
1da177e4 LT |
1447 | struct sock *sk; |
1448 | int ret; | |
a86b1e30 | 1449 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1450 | |
1451 | if (skb->pkt_type != PACKET_HOST) | |
1452 | goto discard_it; | |
1453 | ||
1454 | /* | |
1455 | * Count it even if it's bad. | |
1456 | */ | |
63231bdd | 1457 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1458 | |
1459 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1460 | goto discard_it; | |
1461 | ||
aa8223c7 | 1462 | th = tcp_hdr(skb); |
1da177e4 LT |
1463 | |
1464 | if (th->doff < sizeof(struct tcphdr)/4) | |
1465 | goto bad_packet; | |
1466 | if (!pskb_may_pull(skb, th->doff*4)) | |
1467 | goto discard_it; | |
1468 | ||
60476372 | 1469 | if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb)) |
6a5dc9e5 | 1470 | goto csum_error; |
1da177e4 | 1471 | |
aa8223c7 | 1472 | th = tcp_hdr(skb); |
e802af9c | 1473 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
1474 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1475 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1476 | skb->len - th->doff*4); | |
1477 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1478 | TCP_SKB_CB(skb)->when = 0; | |
b82d1bb4 | 1479 | TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr); |
1da177e4 LT |
1480 | TCP_SKB_CB(skb)->sacked = 0; |
1481 | ||
9a1f27c4 | 1482 | sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1483 | if (!sk) |
1484 | goto no_tcp_socket; | |
1485 | ||
1486 | process: | |
1487 | if (sk->sk_state == TCP_TIME_WAIT) | |
1488 | goto do_time_wait; | |
1489 | ||
e802af9c SH |
1490 | if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) { |
1491 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
1492 | goto discard_and_relse; | |
1493 | } | |
1494 | ||
1da177e4 LT |
1495 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
1496 | goto discard_and_relse; | |
1497 | ||
fda9ef5d | 1498 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1499 | goto discard_and_relse; |
1500 | ||
1501 | skb->dev = NULL; | |
1502 | ||
293b9c42 | 1503 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1504 | ret = 0; |
1505 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 | 1506 | #ifdef CONFIG_NET_DMA |
1ab1457c | 1507 | struct tcp_sock *tp = tcp_sk(sk); |
b4caea8a | 1508 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) |
a2bd1140 | 1509 | tp->ucopy.dma_chan = net_dma_find_channel(); |
1ab1457c YH |
1510 | if (tp->ucopy.dma_chan) |
1511 | ret = tcp_v6_do_rcv(sk, skb); | |
1512 | else | |
1a2449a8 CL |
1513 | #endif |
1514 | { | |
1515 | if (!tcp_prequeue(sk, skb)) | |
1516 | ret = tcp_v6_do_rcv(sk, skb); | |
1517 | } | |
da882c1f ED |
1518 | } else if (unlikely(sk_add_backlog(sk, skb, |
1519 | sk->sk_rcvbuf + sk->sk_sndbuf))) { | |
6b03a53a | 1520 | bh_unlock_sock(sk); |
6cce09f8 | 1521 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1522 | goto discard_and_relse; |
1523 | } | |
1da177e4 LT |
1524 | bh_unlock_sock(sk); |
1525 | ||
1526 | sock_put(sk); | |
1527 | return ret ? -1 : 0; | |
1528 | ||
1529 | no_tcp_socket: | |
1530 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1531 | goto discard_it; | |
1532 | ||
1533 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1534 | csum_error: |
1535 | TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS); | |
1da177e4 | 1536 | bad_packet: |
63231bdd | 1537 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1538 | } else { |
cfb6eeb4 | 1539 | tcp_v6_send_reset(NULL, skb); |
1da177e4 LT |
1540 | } |
1541 | ||
1542 | discard_it: | |
1da177e4 LT |
1543 | kfree_skb(skb); |
1544 | return 0; | |
1545 | ||
1546 | discard_and_relse: | |
1547 | sock_put(sk); | |
1548 | goto discard_it; | |
1549 | ||
1550 | do_time_wait: | |
1551 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1552 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1553 | goto discard_it; |
1554 | } | |
1555 | ||
6a5dc9e5 | 1556 | if (skb->len < (th->doff<<2)) { |
9469c7b4 | 1557 | inet_twsk_put(inet_twsk(sk)); |
6a5dc9e5 ED |
1558 | goto bad_packet; |
1559 | } | |
1560 | if (tcp_checksum_complete(skb)) { | |
1561 | inet_twsk_put(inet_twsk(sk)); | |
1562 | goto csum_error; | |
1da177e4 LT |
1563 | } |
1564 | ||
9469c7b4 | 1565 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 LT |
1566 | case TCP_TW_SYN: |
1567 | { | |
1568 | struct sock *sk2; | |
1569 | ||
c346dca1 | 1570 | sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo, |
5ba24953 | 1571 | &ipv6_hdr(skb)->saddr, th->source, |
0660e03f | 1572 | &ipv6_hdr(skb)->daddr, |
505cbfc5 | 1573 | ntohs(th->dest), inet6_iif(skb)); |
1da177e4 | 1574 | if (sk2 != NULL) { |
295ff7ed ACM |
1575 | struct inet_timewait_sock *tw = inet_twsk(sk); |
1576 | inet_twsk_deschedule(tw, &tcp_death_row); | |
1577 | inet_twsk_put(tw); | |
1da177e4 LT |
1578 | sk = sk2; |
1579 | goto process; | |
1580 | } | |
1581 | /* Fall through to ACK */ | |
1582 | } | |
1583 | case TCP_TW_ACK: | |
1584 | tcp_v6_timewait_ack(sk, skb); | |
1585 | break; | |
1586 | case TCP_TW_RST: | |
1587 | goto no_tcp_socket; | |
1588 | case TCP_TW_SUCCESS:; | |
1589 | } | |
1590 | goto discard_it; | |
1591 | } | |
1592 | ||
c7109986 ED |
1593 | static void tcp_v6_early_demux(struct sk_buff *skb) |
1594 | { | |
1595 | const struct ipv6hdr *hdr; | |
1596 | const struct tcphdr *th; | |
1597 | struct sock *sk; | |
1598 | ||
1599 | if (skb->pkt_type != PACKET_HOST) | |
1600 | return; | |
1601 | ||
1602 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) | |
1603 | return; | |
1604 | ||
1605 | hdr = ipv6_hdr(skb); | |
1606 | th = tcp_hdr(skb); | |
1607 | ||
1608 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1609 | return; | |
1610 | ||
1611 | sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo, | |
1612 | &hdr->saddr, th->source, | |
1613 | &hdr->daddr, ntohs(th->dest), | |
1614 | inet6_iif(skb)); | |
1615 | if (sk) { | |
1616 | skb->sk = sk; | |
1617 | skb->destructor = sock_edemux; | |
1618 | if (sk->sk_state != TCP_TIME_WAIT) { | |
1619 | struct dst_entry *dst = sk->sk_rx_dst; | |
f3f12135 | 1620 | |
c7109986 | 1621 | if (dst) |
5d299f3d | 1622 | dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); |
c7109986 | 1623 | if (dst && |
f3f12135 | 1624 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
c7109986 ED |
1625 | skb_dst_set_noref(skb, dst); |
1626 | } | |
1627 | } | |
1628 | } | |
1629 | ||
ccb7c410 DM |
1630 | static struct timewait_sock_ops tcp6_timewait_sock_ops = { |
1631 | .twsk_obj_size = sizeof(struct tcp6_timewait_sock), | |
1632 | .twsk_unique = tcp_twsk_unique, | |
1633 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 DM |
1634 | }; |
1635 | ||
3b401a81 | 1636 | static const struct inet_connection_sock_af_ops ipv6_specific = { |
543d9cfe ACM |
1637 | .queue_xmit = inet6_csk_xmit, |
1638 | .send_check = tcp_v6_send_check, | |
1639 | .rebuild_header = inet6_sk_rebuild_header, | |
5d299f3d | 1640 | .sk_rx_dst_set = inet6_sk_rx_dst_set, |
543d9cfe ACM |
1641 | .conn_request = tcp_v6_conn_request, |
1642 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
543d9cfe | 1643 | .net_header_len = sizeof(struct ipv6hdr), |
67469601 | 1644 | .net_frag_header_len = sizeof(struct frag_hdr), |
543d9cfe ACM |
1645 | .setsockopt = ipv6_setsockopt, |
1646 | .getsockopt = ipv6_getsockopt, | |
1647 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1648 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1649 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1650 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1651 | .compat_setsockopt = compat_ipv6_setsockopt, |
1652 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1653 | #endif |
1da177e4 LT |
1654 | }; |
1655 | ||
cfb6eeb4 | 1656 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1657 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = { |
cfb6eeb4 | 1658 | .md5_lookup = tcp_v6_md5_lookup, |
49a72dfb | 1659 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 1660 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1661 | }; |
a928630a | 1662 | #endif |
cfb6eeb4 | 1663 | |
1da177e4 LT |
1664 | /* |
1665 | * TCP over IPv4 via INET6 API | |
1666 | */ | |
1667 | ||
3b401a81 | 1668 | static const struct inet_connection_sock_af_ops ipv6_mapped = { |
543d9cfe ACM |
1669 | .queue_xmit = ip_queue_xmit, |
1670 | .send_check = tcp_v4_send_check, | |
1671 | .rebuild_header = inet_sk_rebuild_header, | |
63d02d15 | 1672 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1673 | .conn_request = tcp_v6_conn_request, |
1674 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
543d9cfe ACM |
1675 | .net_header_len = sizeof(struct iphdr), |
1676 | .setsockopt = ipv6_setsockopt, | |
1677 | .getsockopt = ipv6_getsockopt, | |
1678 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1679 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1680 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1681 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1682 | .compat_setsockopt = compat_ipv6_setsockopt, |
1683 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1684 | #endif |
1da177e4 LT |
1685 | }; |
1686 | ||
cfb6eeb4 | 1687 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1688 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = { |
cfb6eeb4 | 1689 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1690 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1691 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1692 | }; |
a928630a | 1693 | #endif |
cfb6eeb4 | 1694 | |
1da177e4 LT |
1695 | /* NOTE: A lot of things set to zero explicitly by call to |
1696 | * sk_alloc() so need not be done here. | |
1697 | */ | |
1698 | static int tcp_v6_init_sock(struct sock *sk) | |
1699 | { | |
6687e988 | 1700 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1701 | |
900f65d3 | 1702 | tcp_init_sock(sk); |
1da177e4 | 1703 | |
8292a17a | 1704 | icsk->icsk_af_ops = &ipv6_specific; |
1da177e4 | 1705 | |
cfb6eeb4 | 1706 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1707 | tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific; |
cfb6eeb4 YH |
1708 | #endif |
1709 | ||
1da177e4 LT |
1710 | return 0; |
1711 | } | |
1712 | ||
7d06b2e0 | 1713 | static void tcp_v6_destroy_sock(struct sock *sk) |
1da177e4 | 1714 | { |
1da177e4 | 1715 | tcp_v4_destroy_sock(sk); |
7d06b2e0 | 1716 | inet6_destroy_sock(sk); |
1da177e4 LT |
1717 | } |
1718 | ||
952a10be | 1719 | #ifdef CONFIG_PROC_FS |
1da177e4 | 1720 | /* Proc filesystem TCPv6 sock list dumping. */ |
1ab1457c | 1721 | static void get_openreq6(struct seq_file *seq, |
a7cb5a49 | 1722 | const struct sock *sk, struct request_sock *req, int i, kuid_t uid) |
1da177e4 | 1723 | { |
1da177e4 | 1724 | int ttd = req->expires - jiffies; |
b71d1d42 ED |
1725 | const struct in6_addr *src = &inet6_rsk(req)->loc_addr; |
1726 | const struct in6_addr *dest = &inet6_rsk(req)->rmt_addr; | |
1da177e4 LT |
1727 | |
1728 | if (ttd < 0) | |
1729 | ttd = 0; | |
1730 | ||
1da177e4 LT |
1731 | seq_printf(seq, |
1732 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1733 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1734 | i, |
1735 | src->s6_addr32[0], src->s6_addr32[1], | |
1736 | src->s6_addr32[2], src->s6_addr32[3], | |
fd507037 | 1737 | ntohs(inet_rsk(req)->loc_port), |
1da177e4 LT |
1738 | dest->s6_addr32[0], dest->s6_addr32[1], |
1739 | dest->s6_addr32[2], dest->s6_addr32[3], | |
2e6599cb | 1740 | ntohs(inet_rsk(req)->rmt_port), |
1da177e4 LT |
1741 | TCP_SYN_RECV, |
1742 | 0,0, /* could print option size, but that is af dependent. */ | |
1ab1457c YH |
1743 | 1, /* timers active (only the expire timer) */ |
1744 | jiffies_to_clock_t(ttd), | |
e6c022a4 | 1745 | req->num_timeout, |
a7cb5a49 | 1746 | from_kuid_munged(seq_user_ns(seq), uid), |
1ab1457c | 1747 | 0, /* non standard timer */ |
1da177e4 LT |
1748 | 0, /* open_requests have no inode */ |
1749 | 0, req); | |
1750 | } | |
1751 | ||
1752 | static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i) | |
1753 | { | |
b71d1d42 | 1754 | const struct in6_addr *dest, *src; |
1da177e4 LT |
1755 | __u16 destp, srcp; |
1756 | int timer_active; | |
1757 | unsigned long timer_expires; | |
cf533ea5 ED |
1758 | const struct inet_sock *inet = inet_sk(sp); |
1759 | const struct tcp_sock *tp = tcp_sk(sp); | |
463c84b9 | 1760 | const struct inet_connection_sock *icsk = inet_csk(sp); |
cf533ea5 | 1761 | const struct ipv6_pinfo *np = inet6_sk(sp); |
1da177e4 LT |
1762 | |
1763 | dest = &np->daddr; | |
1764 | src = &np->rcv_saddr; | |
c720c7e8 ED |
1765 | destp = ntohs(inet->inet_dport); |
1766 | srcp = ntohs(inet->inet_sport); | |
463c84b9 ACM |
1767 | |
1768 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { | |
1da177e4 | 1769 | timer_active = 1; |
463c84b9 ACM |
1770 | timer_expires = icsk->icsk_timeout; |
1771 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 1772 | timer_active = 4; |
463c84b9 | 1773 | timer_expires = icsk->icsk_timeout; |
1da177e4 LT |
1774 | } else if (timer_pending(&sp->sk_timer)) { |
1775 | timer_active = 2; | |
1776 | timer_expires = sp->sk_timer.expires; | |
1777 | } else { | |
1778 | timer_active = 0; | |
1779 | timer_expires = jiffies; | |
1780 | } | |
1781 | ||
1782 | seq_printf(seq, | |
1783 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1784 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n", |
1da177e4 LT |
1785 | i, |
1786 | src->s6_addr32[0], src->s6_addr32[1], | |
1787 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1788 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1789 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1ab1457c | 1790 | sp->sk_state, |
47da8ee6 SS |
1791 | tp->write_seq-tp->snd_una, |
1792 | (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq), | |
1da177e4 | 1793 | timer_active, |
a399a805 | 1794 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 1795 | icsk->icsk_retransmits, |
a7cb5a49 | 1796 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), |
6687e988 | 1797 | icsk->icsk_probes_out, |
1da177e4 LT |
1798 | sock_i_ino(sp), |
1799 | atomic_read(&sp->sk_refcnt), sp, | |
7be87351 SH |
1800 | jiffies_to_clock_t(icsk->icsk_rto), |
1801 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 1802 | (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong, |
0b6a05c1 IJ |
1803 | tp->snd_cwnd, |
1804 | tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh | |
1da177e4 LT |
1805 | ); |
1806 | } | |
1807 | ||
1ab1457c | 1808 | static void get_timewait6_sock(struct seq_file *seq, |
8feaf0c0 | 1809 | struct inet_timewait_sock *tw, int i) |
1da177e4 | 1810 | { |
b71d1d42 | 1811 | const struct in6_addr *dest, *src; |
1da177e4 | 1812 | __u16 destp, srcp; |
cf533ea5 | 1813 | const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw); |
a399a805 | 1814 | long delta = tw->tw_ttd - jiffies; |
1da177e4 | 1815 | |
0fa1a53e ACM |
1816 | dest = &tw6->tw_v6_daddr; |
1817 | src = &tw6->tw_v6_rcv_saddr; | |
1da177e4 LT |
1818 | destp = ntohs(tw->tw_dport); |
1819 | srcp = ntohs(tw->tw_sport); | |
1820 | ||
1821 | seq_printf(seq, | |
1822 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1823 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1824 | i, |
1825 | src->s6_addr32[0], src->s6_addr32[1], | |
1826 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1827 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1828 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1829 | tw->tw_substate, 0, 0, | |
a399a805 | 1830 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
1da177e4 LT |
1831 | atomic_read(&tw->tw_refcnt), tw); |
1832 | } | |
1833 | ||
1da177e4 LT |
1834 | static int tcp6_seq_show(struct seq_file *seq, void *v) |
1835 | { | |
1836 | struct tcp_iter_state *st; | |
1837 | ||
1838 | if (v == SEQ_START_TOKEN) { | |
1839 | seq_puts(seq, | |
1840 | " sl " | |
1841 | "local_address " | |
1842 | "remote_address " | |
1843 | "st tx_queue rx_queue tr tm->when retrnsmt" | |
1844 | " uid timeout inode\n"); | |
1845 | goto out; | |
1846 | } | |
1847 | st = seq->private; | |
1848 | ||
1849 | switch (st->state) { | |
1850 | case TCP_SEQ_STATE_LISTENING: | |
1851 | case TCP_SEQ_STATE_ESTABLISHED: | |
1852 | get_tcp6_sock(seq, v, st->num); | |
1853 | break; | |
1854 | case TCP_SEQ_STATE_OPENREQ: | |
1855 | get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid); | |
1856 | break; | |
1857 | case TCP_SEQ_STATE_TIME_WAIT: | |
1858 | get_timewait6_sock(seq, v, st->num); | |
1859 | break; | |
1860 | } | |
1861 | out: | |
1862 | return 0; | |
1863 | } | |
1864 | ||
73cb88ec AV |
1865 | static const struct file_operations tcp6_afinfo_seq_fops = { |
1866 | .owner = THIS_MODULE, | |
1867 | .open = tcp_seq_open, | |
1868 | .read = seq_read, | |
1869 | .llseek = seq_lseek, | |
1870 | .release = seq_release_net | |
1871 | }; | |
1872 | ||
1da177e4 | 1873 | static struct tcp_seq_afinfo tcp6_seq_afinfo = { |
1da177e4 LT |
1874 | .name = "tcp6", |
1875 | .family = AF_INET6, | |
73cb88ec | 1876 | .seq_fops = &tcp6_afinfo_seq_fops, |
9427c4b3 DL |
1877 | .seq_ops = { |
1878 | .show = tcp6_seq_show, | |
1879 | }, | |
1da177e4 LT |
1880 | }; |
1881 | ||
2c8c1e72 | 1882 | int __net_init tcp6_proc_init(struct net *net) |
1da177e4 | 1883 | { |
6f8b13bc | 1884 | return tcp_proc_register(net, &tcp6_seq_afinfo); |
1da177e4 LT |
1885 | } |
1886 | ||
6f8b13bc | 1887 | void tcp6_proc_exit(struct net *net) |
1da177e4 | 1888 | { |
6f8b13bc | 1889 | tcp_proc_unregister(net, &tcp6_seq_afinfo); |
1da177e4 LT |
1890 | } |
1891 | #endif | |
1892 | ||
f77d6021 ED |
1893 | static void tcp_v6_clear_sk(struct sock *sk, int size) |
1894 | { | |
1895 | struct inet_sock *inet = inet_sk(sk); | |
1896 | ||
1897 | /* we do not want to clear pinet6 field, because of RCU lookups */ | |
1898 | sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6)); | |
1899 | ||
1900 | size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6); | |
1901 | memset(&inet->pinet6 + 1, 0, size); | |
1902 | } | |
1903 | ||
1da177e4 LT |
1904 | struct proto tcpv6_prot = { |
1905 | .name = "TCPv6", | |
1906 | .owner = THIS_MODULE, | |
1907 | .close = tcp_close, | |
1908 | .connect = tcp_v6_connect, | |
1909 | .disconnect = tcp_disconnect, | |
463c84b9 | 1910 | .accept = inet_csk_accept, |
1da177e4 LT |
1911 | .ioctl = tcp_ioctl, |
1912 | .init = tcp_v6_init_sock, | |
1913 | .destroy = tcp_v6_destroy_sock, | |
1914 | .shutdown = tcp_shutdown, | |
1915 | .setsockopt = tcp_setsockopt, | |
1916 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 1917 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
1918 | .sendmsg = tcp_sendmsg, |
1919 | .sendpage = tcp_sendpage, | |
1da177e4 | 1920 | .backlog_rcv = tcp_v6_do_rcv, |
46d3ceab | 1921 | .release_cb = tcp_release_cb, |
563d34d0 | 1922 | .mtu_reduced = tcp_v6_mtu_reduced, |
1da177e4 | 1923 | .hash = tcp_v6_hash, |
ab1e0a13 ACM |
1924 | .unhash = inet_unhash, |
1925 | .get_port = inet_csk_get_port, | |
1da177e4 LT |
1926 | .enter_memory_pressure = tcp_enter_memory_pressure, |
1927 | .sockets_allocated = &tcp_sockets_allocated, | |
1928 | .memory_allocated = &tcp_memory_allocated, | |
1929 | .memory_pressure = &tcp_memory_pressure, | |
0a5578cf | 1930 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
1931 | .sysctl_wmem = sysctl_tcp_wmem, |
1932 | .sysctl_rmem = sysctl_tcp_rmem, | |
1933 | .max_header = MAX_TCP_HEADER, | |
1934 | .obj_size = sizeof(struct tcp6_sock), | |
3ab5aee7 | 1935 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 1936 | .twsk_prot = &tcp6_timewait_sock_ops, |
60236fdd | 1937 | .rsk_prot = &tcp6_request_sock_ops, |
39d8cda7 | 1938 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 1939 | .no_autobind = true, |
543d9cfe ACM |
1940 | #ifdef CONFIG_COMPAT |
1941 | .compat_setsockopt = compat_tcp_setsockopt, | |
1942 | .compat_getsockopt = compat_tcp_getsockopt, | |
1943 | #endif | |
c255a458 | 1944 | #ifdef CONFIG_MEMCG_KMEM |
d1a4c0b3 GC |
1945 | .proto_cgroup = tcp_proto_cgroup, |
1946 | #endif | |
f77d6021 | 1947 | .clear_sk = tcp_v6_clear_sk, |
1da177e4 LT |
1948 | }; |
1949 | ||
41135cc8 | 1950 | static const struct inet6_protocol tcpv6_protocol = { |
c7109986 | 1951 | .early_demux = tcp_v6_early_demux, |
1da177e4 LT |
1952 | .handler = tcp_v6_rcv, |
1953 | .err_handler = tcp_v6_err, | |
1954 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
1955 | }; | |
1956 | ||
1da177e4 LT |
1957 | static struct inet_protosw tcpv6_protosw = { |
1958 | .type = SOCK_STREAM, | |
1959 | .protocol = IPPROTO_TCP, | |
1960 | .prot = &tcpv6_prot, | |
1961 | .ops = &inet6_stream_ops, | |
1da177e4 | 1962 | .no_check = 0, |
d83d8461 ACM |
1963 | .flags = INET_PROTOSW_PERMANENT | |
1964 | INET_PROTOSW_ICSK, | |
1da177e4 LT |
1965 | }; |
1966 | ||
2c8c1e72 | 1967 | static int __net_init tcpv6_net_init(struct net *net) |
93ec926b | 1968 | { |
5677242f DL |
1969 | return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6, |
1970 | SOCK_RAW, IPPROTO_TCP, net); | |
93ec926b DL |
1971 | } |
1972 | ||
2c8c1e72 | 1973 | static void __net_exit tcpv6_net_exit(struct net *net) |
93ec926b | 1974 | { |
5677242f | 1975 | inet_ctl_sock_destroy(net->ipv6.tcp_sk); |
b099ce26 EB |
1976 | } |
1977 | ||
2c8c1e72 | 1978 | static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list) |
b099ce26 EB |
1979 | { |
1980 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6); | |
93ec926b DL |
1981 | } |
1982 | ||
1983 | static struct pernet_operations tcpv6_net_ops = { | |
b099ce26 EB |
1984 | .init = tcpv6_net_init, |
1985 | .exit = tcpv6_net_exit, | |
1986 | .exit_batch = tcpv6_net_exit_batch, | |
93ec926b DL |
1987 | }; |
1988 | ||
7f4e4868 | 1989 | int __init tcpv6_init(void) |
1da177e4 | 1990 | { |
7f4e4868 DL |
1991 | int ret; |
1992 | ||
3336288a VY |
1993 | ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP); |
1994 | if (ret) | |
c6b641a4 | 1995 | goto out; |
3336288a | 1996 | |
1da177e4 | 1997 | /* register inet6 protocol */ |
7f4e4868 DL |
1998 | ret = inet6_register_protosw(&tcpv6_protosw); |
1999 | if (ret) | |
2000 | goto out_tcpv6_protocol; | |
2001 | ||
93ec926b | 2002 | ret = register_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2003 | if (ret) |
2004 | goto out_tcpv6_protosw; | |
2005 | out: | |
2006 | return ret; | |
ae0f7d5f | 2007 | |
7f4e4868 DL |
2008 | out_tcpv6_protosw: |
2009 | inet6_unregister_protosw(&tcpv6_protosw); | |
3336288a VY |
2010 | out_tcpv6_protocol: |
2011 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
7f4e4868 DL |
2012 | goto out; |
2013 | } | |
2014 | ||
09f7709f | 2015 | void tcpv6_exit(void) |
7f4e4868 | 2016 | { |
93ec926b | 2017 | unregister_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2018 | inet6_unregister_protosw(&tcpv6_protosw); |
2019 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1da177e4 | 2020 | } |