Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * TCP over IPv6 | |
1ab1457c | 4 | * Linux INET6 implementation |
1da177e4 LT |
5 | * |
6 | * Authors: | |
1ab1457c | 7 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 8 | * |
1ab1457c | 9 | * Based on: |
1da177e4 LT |
10 | * linux/net/ipv4/tcp.c |
11 | * linux/net/ipv4/tcp_input.c | |
12 | * linux/net/ipv4/tcp_output.c | |
13 | * | |
14 | * Fixes: | |
15 | * Hideaki YOSHIFUJI : sin6_scope_id support | |
16 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
17 | * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind | |
18 | * a single port at the same time. | |
19 | * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file. | |
1da177e4 LT |
20 | */ |
21 | ||
eb4dea58 | 22 | #include <linux/bottom_half.h> |
1da177e4 | 23 | #include <linux/module.h> |
1da177e4 LT |
24 | #include <linux/errno.h> |
25 | #include <linux/types.h> | |
26 | #include <linux/socket.h> | |
27 | #include <linux/sockios.h> | |
28 | #include <linux/net.h> | |
29 | #include <linux/jiffies.h> | |
30 | #include <linux/in.h> | |
31 | #include <linux/in6.h> | |
32 | #include <linux/netdevice.h> | |
33 | #include <linux/init.h> | |
34 | #include <linux/jhash.h> | |
35 | #include <linux/ipsec.h> | |
36 | #include <linux/times.h> | |
5a0e3ad6 | 37 | #include <linux/slab.h> |
4aa956d8 | 38 | #include <linux/uaccess.h> |
1da177e4 LT |
39 | #include <linux/ipv6.h> |
40 | #include <linux/icmpv6.h> | |
41 | #include <linux/random.h> | |
0e219ae4 | 42 | #include <linux/indirect_call_wrapper.h> |
1da177e4 LT |
43 | |
44 | #include <net/tcp.h> | |
45 | #include <net/ndisc.h> | |
5324a040 | 46 | #include <net/inet6_hashtables.h> |
8129765a | 47 | #include <net/inet6_connection_sock.h> |
1da177e4 LT |
48 | #include <net/ipv6.h> |
49 | #include <net/transp_v6.h> | |
50 | #include <net/addrconf.h> | |
51 | #include <net/ip6_route.h> | |
52 | #include <net/ip6_checksum.h> | |
53 | #include <net/inet_ecn.h> | |
54 | #include <net/protocol.h> | |
55 | #include <net/xfrm.h> | |
1da177e4 LT |
56 | #include <net/snmp.h> |
57 | #include <net/dsfield.h> | |
6d6ee43e | 58 | #include <net/timewait_sock.h> |
3d58b5fa | 59 | #include <net/inet_common.h> |
6e5714ea | 60 | #include <net/secure_seq.h> |
076bb0c8 | 61 | #include <net/busy_poll.h> |
1da177e4 | 62 | |
1da177e4 LT |
63 | #include <linux/proc_fs.h> |
64 | #include <linux/seq_file.h> | |
65 | ||
cf80e0e4 | 66 | #include <crypto/hash.h> |
cfb6eeb4 YH |
67 | #include <linux/scatterlist.h> |
68 | ||
c24b14c4 SL |
69 | #include <trace/events/tcp.h> |
70 | ||
a00e7444 ED |
71 | static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb); |
72 | static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb, | |
6edafaaf | 73 | struct request_sock *req); |
1da177e4 | 74 | |
d2489c7b | 75 | INDIRECT_CALLABLE_SCOPE int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 76 | |
3b401a81 | 77 | static const struct inet_connection_sock_af_ops ipv6_mapped; |
35b2c321 | 78 | const struct inet_connection_sock_af_ops ipv6_specific; |
a928630a | 79 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de SH |
80 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific; |
81 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific; | |
9501f972 | 82 | #else |
51723935 | 83 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk, |
dea53bb8 DA |
84 | const struct in6_addr *addr, |
85 | int l3index) | |
9501f972 YH |
86 | { |
87 | return NULL; | |
88 | } | |
a928630a | 89 | #endif |
1da177e4 | 90 | |
93a77c11 ED |
91 | /* Helper returning the inet6 address from a given tcp socket. |
92 | * It can be used in TCP stack instead of inet6_sk(sk). | |
93 | * This avoids a dereference and allow compiler optimizations. | |
f5d54767 | 94 | * It is a specialized version of inet6_sk_generic(). |
93a77c11 ED |
95 | */ |
96 | static struct ipv6_pinfo *tcp_inet6_sk(const struct sock *sk) | |
97 | { | |
f5d54767 | 98 | unsigned int offset = sizeof(struct tcp6_sock) - sizeof(struct ipv6_pinfo); |
93a77c11 | 99 | |
f5d54767 | 100 | return (struct ipv6_pinfo *)(((u8 *)sk) + offset); |
93a77c11 ED |
101 | } |
102 | ||
fae6ef87 NC |
103 | static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
104 | { | |
105 | struct dst_entry *dst = skb_dst(skb); | |
fae6ef87 | 106 | |
5037e9ef | 107 | if (dst && dst_hold_safe(dst)) { |
ca777eff ED |
108 | const struct rt6_info *rt = (const struct rt6_info *)dst; |
109 | ||
8f905c0e | 110 | rcu_assign_pointer(sk->sk_rx_dst, dst); |
0c0a5ef8 | 111 | sk->sk_rx_dst_ifindex = skb->skb_iif; |
ef57c161 | 112 | sk->sk_rx_dst_cookie = rt6_get_cookie(rt); |
ca777eff | 113 | } |
fae6ef87 NC |
114 | } |
115 | ||
84b114b9 | 116 | static u32 tcp_v6_init_seq(const struct sk_buff *skb) |
1da177e4 | 117 | { |
84b114b9 ED |
118 | return secure_tcpv6_seq(ipv6_hdr(skb)->daddr.s6_addr32, |
119 | ipv6_hdr(skb)->saddr.s6_addr32, | |
120 | tcp_hdr(skb)->dest, | |
121 | tcp_hdr(skb)->source); | |
122 | } | |
123 | ||
5d2ed052 | 124 | static u32 tcp_v6_init_ts_off(const struct net *net, const struct sk_buff *skb) |
84b114b9 | 125 | { |
5d2ed052 | 126 | return secure_tcpv6_ts_off(net, ipv6_hdr(skb)->daddr.s6_addr32, |
84b114b9 | 127 | ipv6_hdr(skb)->saddr.s6_addr32); |
1da177e4 LT |
128 | } |
129 | ||
d74bad4e AI |
130 | static int tcp_v6_pre_connect(struct sock *sk, struct sockaddr *uaddr, |
131 | int addr_len) | |
132 | { | |
133 | /* This check is replicated from tcp_v6_connect() and intended to | |
134 | * prevent BPF program called below from accessing bytes that are out | |
135 | * of the bound specified by user in addr_len. | |
136 | */ | |
137 | if (addr_len < SIN6_LEN_RFC2133) | |
138 | return -EINVAL; | |
139 | ||
140 | sock_owned_by_me(sk); | |
141 | ||
142 | return BPF_CGROUP_RUN_PROG_INET6_CONNECT(sk, uaddr); | |
143 | } | |
144 | ||
1ab1457c | 145 | static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
1da177e4 LT |
146 | int addr_len) |
147 | { | |
148 | struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr; | |
d83d8461 | 149 | struct inet_connection_sock *icsk = inet_csk(sk); |
08eaef90 | 150 | struct in6_addr *saddr = NULL, *final_p, final; |
fbb82952 | 151 | struct inet_timewait_death_row *tcp_death_row; |
93a77c11 | 152 | struct ipv6_pinfo *np = tcp_inet6_sk(sk); |
08eaef90 | 153 | struct inet_sock *inet = inet_sk(sk); |
1da177e4 | 154 | struct tcp_sock *tp = tcp_sk(sk); |
08eaef90 | 155 | struct net *net = sock_net(sk); |
45f6fad8 | 156 | struct ipv6_txoptions *opt; |
1da177e4 | 157 | struct dst_entry *dst; |
08eaef90 | 158 | struct flowi6 fl6; |
1da177e4 LT |
159 | int addr_type; |
160 | int err; | |
161 | ||
1ab1457c | 162 | if (addr_len < SIN6_LEN_RFC2133) |
1da177e4 LT |
163 | return -EINVAL; |
164 | ||
1ab1457c | 165 | if (usin->sin6_family != AF_INET6) |
a02cec21 | 166 | return -EAFNOSUPPORT; |
1da177e4 | 167 | |
4c9483b2 | 168 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
169 | |
170 | if (np->sndflow) { | |
4c9483b2 DM |
171 | fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
172 | IP6_ECN_flow_init(fl6.flowlabel); | |
173 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1da177e4 | 174 | struct ip6_flowlabel *flowlabel; |
4c9483b2 | 175 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); |
59c820b2 | 176 | if (IS_ERR(flowlabel)) |
1da177e4 | 177 | return -EINVAL; |
1da177e4 LT |
178 | fl6_sock_release(flowlabel); |
179 | } | |
180 | } | |
181 | ||
182 | /* | |
1ab1457c YH |
183 | * connect() to INADDR_ANY means loopback (BSD'ism). |
184 | */ | |
185 | ||
052d2369 JL |
186 | if (ipv6_addr_any(&usin->sin6_addr)) { |
187 | if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) | |
188 | ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), | |
189 | &usin->sin6_addr); | |
190 | else | |
191 | usin->sin6_addr = in6addr_loopback; | |
192 | } | |
1da177e4 LT |
193 | |
194 | addr_type = ipv6_addr_type(&usin->sin6_addr); | |
195 | ||
4c99aa40 | 196 | if (addr_type & IPV6_ADDR_MULTICAST) |
1da177e4 LT |
197 | return -ENETUNREACH; |
198 | ||
199 | if (addr_type&IPV6_ADDR_LINKLOCAL) { | |
200 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
201 | usin->sin6_scope_id) { | |
202 | /* If interface is set while binding, indices | |
203 | * must coincide. | |
204 | */ | |
54dc3e33 | 205 | if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) |
1da177e4 LT |
206 | return -EINVAL; |
207 | ||
208 | sk->sk_bound_dev_if = usin->sin6_scope_id; | |
209 | } | |
210 | ||
211 | /* Connect to link-local address requires an interface */ | |
212 | if (!sk->sk_bound_dev_if) | |
213 | return -EINVAL; | |
214 | } | |
215 | ||
216 | if (tp->rx_opt.ts_recent_stamp && | |
efe4208f | 217 | !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) { |
1da177e4 LT |
218 | tp->rx_opt.ts_recent = 0; |
219 | tp->rx_opt.ts_recent_stamp = 0; | |
0f317464 | 220 | WRITE_ONCE(tp->write_seq, 0); |
1da177e4 LT |
221 | } |
222 | ||
efe4208f | 223 | sk->sk_v6_daddr = usin->sin6_addr; |
4c9483b2 | 224 | np->flow_label = fl6.flowlabel; |
1da177e4 LT |
225 | |
226 | /* | |
227 | * TCP over IPv4 | |
228 | */ | |
229 | ||
052d2369 | 230 | if (addr_type & IPV6_ADDR_MAPPED) { |
d83d8461 | 231 | u32 exthdrlen = icsk->icsk_ext_hdr_len; |
1da177e4 LT |
232 | struct sockaddr_in sin; |
233 | ||
89e9c728 | 234 | if (ipv6_only_sock(sk)) |
1da177e4 LT |
235 | return -ENETUNREACH; |
236 | ||
237 | sin.sin_family = AF_INET; | |
238 | sin.sin_port = usin->sin6_port; | |
239 | sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; | |
240 | ||
f49cd2f4 KI |
241 | /* Paired with READ_ONCE() in tcp_(get|set)sockopt() */ |
242 | WRITE_ONCE(icsk->icsk_af_ops, &ipv6_mapped); | |
cec37a6e | 243 | if (sk_is_mptcp(sk)) |
31484d56 | 244 | mptcpv6_handle_mapped(sk, true); |
1da177e4 | 245 | sk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
246 | #ifdef CONFIG_TCP_MD5SIG |
247 | tp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
248 | #endif | |
1da177e4 LT |
249 | |
250 | err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); | |
251 | ||
252 | if (err) { | |
d83d8461 | 253 | icsk->icsk_ext_hdr_len = exthdrlen; |
f49cd2f4 KI |
254 | /* Paired with READ_ONCE() in tcp_(get|set)sockopt() */ |
255 | WRITE_ONCE(icsk->icsk_af_ops, &ipv6_specific); | |
cec37a6e | 256 | if (sk_is_mptcp(sk)) |
31484d56 | 257 | mptcpv6_handle_mapped(sk, false); |
1da177e4 | 258 | sk->sk_backlog_rcv = tcp_v6_do_rcv; |
cfb6eeb4 YH |
259 | #ifdef CONFIG_TCP_MD5SIG |
260 | tp->af_specific = &tcp_sock_ipv6_specific; | |
261 | #endif | |
1da177e4 | 262 | goto failure; |
1da177e4 | 263 | } |
d1e559d0 | 264 | np->saddr = sk->sk_v6_rcv_saddr; |
1da177e4 LT |
265 | |
266 | return err; | |
267 | } | |
268 | ||
efe4208f ED |
269 | if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) |
270 | saddr = &sk->sk_v6_rcv_saddr; | |
1da177e4 | 271 | |
4c9483b2 | 272 | fl6.flowi6_proto = IPPROTO_TCP; |
efe4208f | 273 | fl6.daddr = sk->sk_v6_daddr; |
4e3fd7a0 | 274 | fl6.saddr = saddr ? *saddr : np->saddr; |
8230680f | 275 | fl6.flowlabel = ip6_make_flowinfo(np->tclass, np->flow_label); |
4c9483b2 DM |
276 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
277 | fl6.flowi6_mark = sk->sk_mark; | |
1958b856 DM |
278 | fl6.fl6_dport = usin->sin6_port; |
279 | fl6.fl6_sport = inet->inet_sport; | |
e2d118a1 | 280 | fl6.flowi6_uid = sk->sk_uid; |
1da177e4 | 281 | |
1e1d04e6 | 282 | opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk)); |
45f6fad8 | 283 | final_p = fl6_update_dst(&fl6, opt, &final); |
1da177e4 | 284 | |
3df98d79 | 285 | security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); |
beb8d13b | 286 | |
08eaef90 | 287 | dst = ip6_dst_lookup_flow(net, sk, &fl6, final_p); |
68d0c6d3 DM |
288 | if (IS_ERR(dst)) { |
289 | err = PTR_ERR(dst); | |
1da177e4 | 290 | goto failure; |
14e50e57 | 291 | } |
1da177e4 | 292 | |
4461568a KI |
293 | tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row; |
294 | ||
63159f29 | 295 | if (!saddr) { |
4c9483b2 | 296 | saddr = &fl6.saddr; |
8c5dae4c KI |
297 | |
298 | err = inet_bhash2_update_saddr(sk, saddr, AF_INET6); | |
299 | if (err) | |
300 | goto failure; | |
1da177e4 LT |
301 | } |
302 | ||
303 | /* set the source address */ | |
4e3fd7a0 | 304 | np->saddr = *saddr; |
c720c7e8 | 305 | inet->inet_rcv_saddr = LOOPBACK4_IPV6; |
1da177e4 | 306 | |
f83ef8c0 | 307 | sk->sk_gso_type = SKB_GSO_TCPV6; |
6bd4f355 | 308 | ip6_dst_store(sk, dst, NULL, NULL); |
1da177e4 | 309 | |
d83d8461 | 310 | icsk->icsk_ext_hdr_len = 0; |
45f6fad8 ED |
311 | if (opt) |
312 | icsk->icsk_ext_hdr_len = opt->opt_flen + | |
313 | opt->opt_nflen; | |
1da177e4 LT |
314 | |
315 | tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
316 | ||
c720c7e8 | 317 | inet->inet_dport = usin->sin6_port; |
1da177e4 LT |
318 | |
319 | tcp_set_state(sk, TCP_SYN_SENT); | |
1946e672 | 320 | err = inet6_hash_connect(tcp_death_row, sk); |
1da177e4 LT |
321 | if (err) |
322 | goto late_failure; | |
323 | ||
877d1f62 | 324 | sk_set_txhash(sk); |
9e7ceb06 | 325 | |
00355fa5 | 326 | if (likely(!tp->repair)) { |
00355fa5 | 327 | if (!tp->write_seq) |
0f317464 ED |
328 | WRITE_ONCE(tp->write_seq, |
329 | secure_tcpv6_seq(np->saddr.s6_addr32, | |
330 | sk->sk_v6_daddr.s6_addr32, | |
331 | inet->inet_sport, | |
332 | inet->inet_dport)); | |
08eaef90 | 333 | tp->tsoffset = secure_tcpv6_ts_off(net, np->saddr.s6_addr32, |
84b114b9 | 334 | sk->sk_v6_daddr.s6_addr32); |
00355fa5 | 335 | } |
1da177e4 | 336 | |
19f6d3f3 WW |
337 | if (tcp_fastopen_defer_connect(sk, &err)) |
338 | return err; | |
339 | if (err) | |
340 | goto late_failure; | |
341 | ||
1da177e4 LT |
342 | err = tcp_connect(sk); |
343 | if (err) | |
344 | goto late_failure; | |
345 | ||
346 | return 0; | |
347 | ||
348 | late_failure: | |
349 | tcp_set_state(sk, TCP_CLOSE); | |
e0833d1f | 350 | inet_bhash2_reset_saddr(sk); |
1da177e4 | 351 | failure: |
c720c7e8 | 352 | inet->inet_dport = 0; |
1da177e4 LT |
353 | sk->sk_route_caps = 0; |
354 | return err; | |
355 | } | |
356 | ||
563d34d0 ED |
357 | static void tcp_v6_mtu_reduced(struct sock *sk) |
358 | { | |
359 | struct dst_entry *dst; | |
c7bb4b89 | 360 | u32 mtu; |
563d34d0 ED |
361 | |
362 | if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) | |
363 | return; | |
364 | ||
c7bb4b89 ED |
365 | mtu = READ_ONCE(tcp_sk(sk)->mtu_info); |
366 | ||
367 | /* Drop requests trying to increase our current mss. | |
368 | * Check done in __ip6_rt_update_pmtu() is too late. | |
369 | */ | |
370 | if (tcp_mtu_to_mss(sk, mtu) >= tcp_sk(sk)->mss_cache) | |
371 | return; | |
372 | ||
373 | dst = inet6_csk_update_pmtu(sk, mtu); | |
563d34d0 ED |
374 | if (!dst) |
375 | return; | |
376 | ||
377 | if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { | |
378 | tcp_sync_mss(sk, dst_mtu(dst)); | |
379 | tcp_simple_retransmit(sk); | |
380 | } | |
381 | } | |
382 | ||
32bbd879 | 383 | static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
d5fdd6ba | 384 | u8 type, u8 code, int offset, __be32 info) |
1da177e4 | 385 | { |
4c99aa40 | 386 | const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; |
505cbfc5 | 387 | const struct tcphdr *th = (struct tcphdr *)(skb->data+offset); |
2215089b ED |
388 | struct net *net = dev_net(skb->dev); |
389 | struct request_sock *fastopen; | |
1da177e4 | 390 | struct ipv6_pinfo *np; |
1ab1457c | 391 | struct tcp_sock *tp; |
0a672f74 | 392 | __u32 seq, snd_una; |
2215089b | 393 | struct sock *sk; |
9cf74903 | 394 | bool fatal; |
2215089b | 395 | int err; |
1da177e4 | 396 | |
4461568a | 397 | sk = __inet6_lookup_established(net, net->ipv4.tcp_death_row.hashinfo, |
2215089b ED |
398 | &hdr->daddr, th->dest, |
399 | &hdr->saddr, ntohs(th->source), | |
4297a0ef | 400 | skb->dev->ifindex, inet6_sdif(skb)); |
1da177e4 | 401 | |
2215089b | 402 | if (!sk) { |
a16292a0 ED |
403 | __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), |
404 | ICMP6_MIB_INERRORS); | |
32bbd879 | 405 | return -ENOENT; |
1da177e4 LT |
406 | } |
407 | ||
408 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 409 | inet_twsk_put(inet_twsk(sk)); |
32bbd879 | 410 | return 0; |
1da177e4 | 411 | } |
2215089b | 412 | seq = ntohl(th->seq); |
9cf74903 | 413 | fatal = icmpv6_err_convert(type, code, &err); |
32bbd879 SB |
414 | if (sk->sk_state == TCP_NEW_SYN_RECV) { |
415 | tcp_req_err(sk, seq, fatal); | |
416 | return 0; | |
417 | } | |
1da177e4 LT |
418 | |
419 | bh_lock_sock(sk); | |
563d34d0 | 420 | if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG) |
02a1d6e7 | 421 | __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS); |
1da177e4 LT |
422 | |
423 | if (sk->sk_state == TCP_CLOSE) | |
424 | goto out; | |
425 | ||
790eb673 ED |
426 | if (static_branch_unlikely(&ip6_min_hopcount)) { |
427 | /* min_hopcount can be changed concurrently from do_ipv6_setsockopt() */ | |
428 | if (ipv6_hdr(skb)->hop_limit < READ_ONCE(tcp_inet6_sk(sk)->min_hopcount)) { | |
429 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); | |
430 | goto out; | |
431 | } | |
e802af9c SH |
432 | } |
433 | ||
1da177e4 | 434 | tp = tcp_sk(sk); |
0a672f74 | 435 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ |
d983ea6f | 436 | fastopen = rcu_dereference(tp->fastopen_rsk); |
0a672f74 | 437 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; |
1da177e4 | 438 | if (sk->sk_state != TCP_LISTEN && |
0a672f74 | 439 | !between(seq, snd_una, tp->snd_nxt)) { |
02a1d6e7 | 440 | __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
441 | goto out; |
442 | } | |
443 | ||
93a77c11 | 444 | np = tcp_inet6_sk(sk); |
1da177e4 | 445 | |
ec18d9a2 | 446 | if (type == NDISC_REDIRECT) { |
45caeaa5 JM |
447 | if (!sock_owned_by_user(sk)) { |
448 | struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie); | |
ec18d9a2 | 449 | |
45caeaa5 JM |
450 | if (dst) |
451 | dst->ops->redirect(dst, sk, skb); | |
452 | } | |
50a75a89 | 453 | goto out; |
ec18d9a2 DM |
454 | } |
455 | ||
1da177e4 | 456 | if (type == ICMPV6_PKT_TOOBIG) { |
c7bb4b89 ED |
457 | u32 mtu = ntohl(info); |
458 | ||
0d4f0608 ED |
459 | /* We are not interested in TCP_LISTEN and open_requests |
460 | * (SYN-ACKs send out by Linux are always <576bytes so | |
461 | * they should go through unfragmented). | |
462 | */ | |
463 | if (sk->sk_state == TCP_LISTEN) | |
464 | goto out; | |
465 | ||
93b36cf3 HFS |
466 | if (!ip6_sk_accept_pmtu(sk)) |
467 | goto out; | |
468 | ||
c7bb4b89 ED |
469 | if (mtu < IPV6_MIN_MTU) |
470 | goto out; | |
471 | ||
472 | WRITE_ONCE(tp->mtu_info, mtu); | |
473 | ||
563d34d0 ED |
474 | if (!sock_owned_by_user(sk)) |
475 | tcp_v6_mtu_reduced(sk); | |
d013ef2a | 476 | else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, |
7aa5470c | 477 | &sk->sk_tsq_flags)) |
d013ef2a | 478 | sock_hold(sk); |
1da177e4 LT |
479 | goto out; |
480 | } | |
481 | ||
1da177e4 | 482 | |
60236fdd | 483 | /* Might be for an request_sock */ |
1da177e4 | 484 | switch (sk->sk_state) { |
1da177e4 | 485 | case TCP_SYN_SENT: |
0a672f74 YC |
486 | case TCP_SYN_RECV: |
487 | /* Only in fast or simultaneous open. If a fast open socket is | |
634a63e7 | 488 | * already accepted it is treated as a connected one below. |
0a672f74 | 489 | */ |
63159f29 | 490 | if (fastopen && !fastopen->sk) |
0a672f74 YC |
491 | break; |
492 | ||
45af29ca ED |
493 | ipv6_icmp_error(sk, skb, err, th->dest, ntohl(info), (u8 *)th); |
494 | ||
1da177e4 | 495 | if (!sock_owned_by_user(sk)) { |
e13ec3da | 496 | WRITE_ONCE(sk->sk_err, err); |
e3ae2365 | 497 | sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */ |
1da177e4 LT |
498 | |
499 | tcp_done(sk); | |
cee1af82 ED |
500 | } else { |
501 | WRITE_ONCE(sk->sk_err_soft, err); | |
502 | } | |
1da177e4 | 503 | goto out; |
d2924569 ED |
504 | case TCP_LISTEN: |
505 | break; | |
506 | default: | |
507 | /* check if this ICMP message allows revert of backoff. | |
508 | * (see RFC 6069) | |
509 | */ | |
510 | if (!fastopen && type == ICMPV6_DEST_UNREACH && | |
511 | code == ICMPV6_NOROUTE) | |
512 | tcp_ld_RTO_revert(sk, seq); | |
1da177e4 LT |
513 | } |
514 | ||
515 | if (!sock_owned_by_user(sk) && np->recverr) { | |
e13ec3da | 516 | WRITE_ONCE(sk->sk_err, err); |
e3ae2365 | 517 | sk_error_report(sk); |
cee1af82 ED |
518 | } else { |
519 | WRITE_ONCE(sk->sk_err_soft, err); | |
520 | } | |
1da177e4 LT |
521 | out: |
522 | bh_unlock_sock(sk); | |
523 | sock_put(sk); | |
32bbd879 | 524 | return 0; |
1da177e4 LT |
525 | } |
526 | ||
527 | ||
0f935dbe | 528 | static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst, |
d6274bd8 | 529 | struct flowi *fl, |
3840a06e | 530 | struct request_sock *req, |
ca6fb065 | 531 | struct tcp_fastopen_cookie *foc, |
331fca43 MKL |
532 | enum tcp_synack_type synack_type, |
533 | struct sk_buff *syn_skb) | |
1da177e4 | 534 | { |
634fb979 | 535 | struct inet_request_sock *ireq = inet_rsk(req); |
93a77c11 | 536 | struct ipv6_pinfo *np = tcp_inet6_sk(sk); |
56ac42bc | 537 | struct ipv6_txoptions *opt; |
d6274bd8 | 538 | struct flowi6 *fl6 = &fl->u.ip6; |
4c99aa40 | 539 | struct sk_buff *skb; |
9494218f | 540 | int err = -ENOMEM; |
ac8f1710 | 541 | u8 tclass; |
1da177e4 | 542 | |
9f10d3f6 | 543 | /* First, grab a route. */ |
f76b33c3 ED |
544 | if (!dst && (dst = inet6_csk_route_req(sk, fl6, req, |
545 | IPPROTO_TCP)) == NULL) | |
fd80eb94 | 546 | goto done; |
9494218f | 547 | |
331fca43 | 548 | skb = tcp_make_synack(sk, dst, req, foc, synack_type, syn_skb); |
9494218f | 549 | |
1da177e4 | 550 | if (skb) { |
634fb979 ED |
551 | __tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr, |
552 | &ireq->ir_v6_rmt_addr); | |
1da177e4 | 553 | |
634fb979 | 554 | fl6->daddr = ireq->ir_v6_rmt_addr; |
53b24b8f | 555 | if (np->repflow && ireq->pktopts) |
df3687ff FF |
556 | fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts)); |
557 | ||
870e3a63 | 558 | tclass = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_reflect_tos) ? |
8ef44b6f WW |
559 | (tcp_rsk(req)->syn_tos & ~INET_ECN_MASK) | |
560 | (np->tclass & INET_ECN_MASK) : | |
861602b5 | 561 | np->tclass; |
407c85c7 AD |
562 | |
563 | if (!INET_ECN_is_capable(tclass) && | |
564 | tcp_bpf_ca_needs_ecn((struct sock *)req)) | |
565 | tclass |= INET_ECN_ECT_0; | |
566 | ||
567 | rcu_read_lock(); | |
568 | opt = ireq->ipv6_opt; | |
56ac42bc HD |
569 | if (!opt) |
570 | opt = rcu_dereference(np->opt); | |
43b90bfa | 571 | err = ip6_xmit(sk, skb, fl6, skb->mark ? : sk->sk_mark, opt, |
861602b5 | 572 | tclass, sk->sk_priority); |
3e4006f0 | 573 | rcu_read_unlock(); |
b9df3cb8 | 574 | err = net_xmit_eval(err); |
1da177e4 LT |
575 | } |
576 | ||
577 | done: | |
1da177e4 LT |
578 | return err; |
579 | } | |
580 | ||
72659ecc | 581 | |
60236fdd | 582 | static void tcp_v6_reqsk_destructor(struct request_sock *req) |
1da177e4 | 583 | { |
56ac42bc | 584 | kfree(inet_rsk(req)->ipv6_opt); |
12c8691d | 585 | consume_skb(inet_rsk(req)->pktopts); |
1da177e4 LT |
586 | } |
587 | ||
cfb6eeb4 | 588 | #ifdef CONFIG_TCP_MD5SIG |
b83e3deb | 589 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk, |
dea53bb8 DA |
590 | const struct in6_addr *addr, |
591 | int l3index) | |
cfb6eeb4 | 592 | { |
dea53bb8 DA |
593 | return tcp_md5_do_lookup(sk, l3index, |
594 | (union tcp_md5_addr *)addr, AF_INET6); | |
cfb6eeb4 YH |
595 | } |
596 | ||
b83e3deb | 597 | static struct tcp_md5sig_key *tcp_v6_md5_lookup(const struct sock *sk, |
fd3a154a | 598 | const struct sock *addr_sk) |
cfb6eeb4 | 599 | { |
dea53bb8 DA |
600 | int l3index; |
601 | ||
602 | l3index = l3mdev_master_ifindex_by_index(sock_net(sk), | |
603 | addr_sk->sk_bound_dev_if); | |
604 | return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr, | |
605 | l3index); | |
cfb6eeb4 YH |
606 | } |
607 | ||
8917a777 | 608 | static int tcp_v6_parse_md5_keys(struct sock *sk, int optname, |
d4c19c49 | 609 | sockptr_t optval, int optlen) |
cfb6eeb4 YH |
610 | { |
611 | struct tcp_md5sig cmd; | |
612 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr; | |
dea53bb8 | 613 | int l3index = 0; |
8917a777 | 614 | u8 prefixlen; |
a76c2315 | 615 | u8 flags; |
cfb6eeb4 YH |
616 | |
617 | if (optlen < sizeof(cmd)) | |
618 | return -EINVAL; | |
619 | ||
d4c19c49 | 620 | if (copy_from_sockptr(&cmd, optval, sizeof(cmd))) |
cfb6eeb4 YH |
621 | return -EFAULT; |
622 | ||
623 | if (sin6->sin6_family != AF_INET6) | |
624 | return -EINVAL; | |
625 | ||
a76c2315 LC |
626 | flags = cmd.tcpm_flags & TCP_MD5SIG_FLAG_IFINDEX; |
627 | ||
8917a777 ID |
628 | if (optname == TCP_MD5SIG_EXT && |
629 | cmd.tcpm_flags & TCP_MD5SIG_FLAG_PREFIX) { | |
630 | prefixlen = cmd.tcpm_prefixlen; | |
631 | if (prefixlen > 128 || (ipv6_addr_v4mapped(&sin6->sin6_addr) && | |
632 | prefixlen > 32)) | |
633 | return -EINVAL; | |
634 | } else { | |
635 | prefixlen = ipv6_addr_v4mapped(&sin6->sin6_addr) ? 32 : 128; | |
636 | } | |
637 | ||
a76c2315 | 638 | if (optname == TCP_MD5SIG_EXT && cmd.tcpm_ifindex && |
6b102db5 DA |
639 | cmd.tcpm_flags & TCP_MD5SIG_FLAG_IFINDEX) { |
640 | struct net_device *dev; | |
641 | ||
642 | rcu_read_lock(); | |
643 | dev = dev_get_by_index_rcu(sock_net(sk), cmd.tcpm_ifindex); | |
644 | if (dev && netif_is_l3_master(dev)) | |
645 | l3index = dev->ifindex; | |
646 | rcu_read_unlock(); | |
647 | ||
648 | /* ok to reference set/not set outside of rcu; | |
649 | * right now device MUST be an L3 master | |
650 | */ | |
651 | if (!dev || !l3index) | |
652 | return -EINVAL; | |
653 | } | |
654 | ||
cfb6eeb4 | 655 | if (!cmd.tcpm_keylen) { |
e773e4fa | 656 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
a915da9b | 657 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], |
6b102db5 | 658 | AF_INET, prefixlen, |
a76c2315 | 659 | l3index, flags); |
a915da9b | 660 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr, |
a76c2315 | 661 | AF_INET6, prefixlen, l3index, flags); |
cfb6eeb4 YH |
662 | } |
663 | ||
664 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
665 | return -EINVAL; | |
666 | ||
a915da9b ED |
667 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
668 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], | |
a76c2315 | 669 | AF_INET, prefixlen, l3index, flags, |
459837b5 | 670 | cmd.tcpm_key, cmd.tcpm_keylen); |
cfb6eeb4 | 671 | |
a915da9b | 672 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr, |
a76c2315 | 673 | AF_INET6, prefixlen, l3index, flags, |
459837b5 | 674 | cmd.tcpm_key, cmd.tcpm_keylen); |
cfb6eeb4 YH |
675 | } |
676 | ||
19689e38 ED |
677 | static int tcp_v6_md5_hash_headers(struct tcp_md5sig_pool *hp, |
678 | const struct in6_addr *daddr, | |
679 | const struct in6_addr *saddr, | |
680 | const struct tcphdr *th, int nbytes) | |
cfb6eeb4 | 681 | { |
cfb6eeb4 | 682 | struct tcp6_pseudohdr *bp; |
49a72dfb | 683 | struct scatterlist sg; |
19689e38 | 684 | struct tcphdr *_th; |
8d26d76d | 685 | |
19689e38 | 686 | bp = hp->scratch; |
cfb6eeb4 | 687 | /* 1. TCP pseudo-header (RFC2460) */ |
4e3fd7a0 AD |
688 | bp->saddr = *saddr; |
689 | bp->daddr = *daddr; | |
49a72dfb | 690 | bp->protocol = cpu_to_be32(IPPROTO_TCP); |
00b1304c | 691 | bp->len = cpu_to_be32(nbytes); |
cfb6eeb4 | 692 | |
19689e38 ED |
693 | _th = (struct tcphdr *)(bp + 1); |
694 | memcpy(_th, th, sizeof(*th)); | |
695 | _th->check = 0; | |
696 | ||
697 | sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th)); | |
698 | ahash_request_set_crypt(hp->md5_req, &sg, NULL, | |
699 | sizeof(*bp) + sizeof(*th)); | |
cf80e0e4 | 700 | return crypto_ahash_update(hp->md5_req); |
49a72dfb | 701 | } |
c7da57a1 | 702 | |
19689e38 | 703 | static int tcp_v6_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key, |
b71d1d42 | 704 | const struct in6_addr *daddr, struct in6_addr *saddr, |
318cf7aa | 705 | const struct tcphdr *th) |
49a72dfb AL |
706 | { |
707 | struct tcp_md5sig_pool *hp; | |
cf80e0e4 | 708 | struct ahash_request *req; |
49a72dfb AL |
709 | |
710 | hp = tcp_get_md5sig_pool(); | |
711 | if (!hp) | |
712 | goto clear_hash_noput; | |
cf80e0e4 | 713 | req = hp->md5_req; |
49a72dfb | 714 | |
cf80e0e4 | 715 | if (crypto_ahash_init(req)) |
49a72dfb | 716 | goto clear_hash; |
19689e38 | 717 | if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2)) |
49a72dfb AL |
718 | goto clear_hash; |
719 | if (tcp_md5_hash_key(hp, key)) | |
720 | goto clear_hash; | |
cf80e0e4 HX |
721 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
722 | if (crypto_ahash_final(req)) | |
cfb6eeb4 | 723 | goto clear_hash; |
cfb6eeb4 | 724 | |
cfb6eeb4 | 725 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 726 | return 0; |
49a72dfb | 727 | |
cfb6eeb4 YH |
728 | clear_hash: |
729 | tcp_put_md5sig_pool(); | |
730 | clear_hash_noput: | |
731 | memset(md5_hash, 0, 16); | |
49a72dfb | 732 | return 1; |
cfb6eeb4 YH |
733 | } |
734 | ||
39f8e58e ED |
735 | static int tcp_v6_md5_hash_skb(char *md5_hash, |
736 | const struct tcp_md5sig_key *key, | |
318cf7aa | 737 | const struct sock *sk, |
318cf7aa | 738 | const struct sk_buff *skb) |
cfb6eeb4 | 739 | { |
b71d1d42 | 740 | const struct in6_addr *saddr, *daddr; |
49a72dfb | 741 | struct tcp_md5sig_pool *hp; |
cf80e0e4 | 742 | struct ahash_request *req; |
318cf7aa | 743 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 744 | |
39f8e58e ED |
745 | if (sk) { /* valid for establish/request sockets */ |
746 | saddr = &sk->sk_v6_rcv_saddr; | |
efe4208f | 747 | daddr = &sk->sk_v6_daddr; |
49a72dfb | 748 | } else { |
b71d1d42 | 749 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
49a72dfb AL |
750 | saddr = &ip6h->saddr; |
751 | daddr = &ip6h->daddr; | |
cfb6eeb4 | 752 | } |
49a72dfb AL |
753 | |
754 | hp = tcp_get_md5sig_pool(); | |
755 | if (!hp) | |
756 | goto clear_hash_noput; | |
cf80e0e4 | 757 | req = hp->md5_req; |
49a72dfb | 758 | |
cf80e0e4 | 759 | if (crypto_ahash_init(req)) |
49a72dfb AL |
760 | goto clear_hash; |
761 | ||
19689e38 | 762 | if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, skb->len)) |
49a72dfb AL |
763 | goto clear_hash; |
764 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
765 | goto clear_hash; | |
766 | if (tcp_md5_hash_key(hp, key)) | |
767 | goto clear_hash; | |
cf80e0e4 HX |
768 | ahash_request_set_crypt(req, NULL, md5_hash, 0); |
769 | if (crypto_ahash_final(req)) | |
49a72dfb AL |
770 | goto clear_hash; |
771 | ||
772 | tcp_put_md5sig_pool(); | |
773 | return 0; | |
774 | ||
775 | clear_hash: | |
776 | tcp_put_md5sig_pool(); | |
777 | clear_hash_noput: | |
778 | memset(md5_hash, 0, 16); | |
779 | return 1; | |
cfb6eeb4 YH |
780 | } |
781 | ||
ba8e275a ED |
782 | #endif |
783 | ||
b40cf18e ED |
784 | static void tcp_v6_init_req(struct request_sock *req, |
785 | const struct sock *sk_listener, | |
16bea70a OP |
786 | struct sk_buff *skb) |
787 | { | |
c2027d1e | 788 | bool l3_slave = ipv6_l3mdev_skb(TCP_SKB_CB(skb)->header.h6.flags); |
16bea70a | 789 | struct inet_request_sock *ireq = inet_rsk(req); |
93a77c11 | 790 | const struct ipv6_pinfo *np = tcp_inet6_sk(sk_listener); |
16bea70a OP |
791 | |
792 | ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr; | |
793 | ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr; | |
794 | ||
16bea70a | 795 | /* So that link locals have meaning */ |
c2027d1e | 796 | if ((!sk_listener->sk_bound_dev_if || l3_slave) && |
16bea70a | 797 | ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL) |
870c3151 | 798 | ireq->ir_iif = tcp_v6_iif(skb); |
16bea70a | 799 | |
04317daf | 800 | if (!TCP_SKB_CB(skb)->tcp_tw_isn && |
b40cf18e | 801 | (ipv6_opt_accepted(sk_listener, skb, &TCP_SKB_CB(skb)->header.h6) || |
a224772d | 802 | np->rxopt.bits.rxinfo || |
16bea70a OP |
803 | np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim || |
804 | np->rxopt.bits.rxohlim || np->repflow)) { | |
63354797 | 805 | refcount_inc(&skb->users); |
16bea70a OP |
806 | ireq->pktopts = skb; |
807 | } | |
808 | } | |
809 | ||
f964629e | 810 | static struct dst_entry *tcp_v6_route_req(const struct sock *sk, |
7ea851d1 | 811 | struct sk_buff *skb, |
f964629e | 812 | struct flowi *fl, |
7ea851d1 | 813 | struct request_sock *req) |
d94e0417 | 814 | { |
7ea851d1 FW |
815 | tcp_v6_init_req(req, sk, skb); |
816 | ||
817 | if (security_inet_conn_request(sk, skb, req)) | |
818 | return NULL; | |
819 | ||
f76b33c3 | 820 | return inet6_csk_route_req(sk, &fl->u.ip6, req, IPPROTO_TCP); |
d94e0417 OP |
821 | } |
822 | ||
c6aefafb | 823 | struct request_sock_ops tcp6_request_sock_ops __read_mostly = { |
1da177e4 | 824 | .family = AF_INET6, |
2e6599cb | 825 | .obj_size = sizeof(struct tcp6_request_sock), |
5db92c99 | 826 | .rtx_syn_ack = tcp_rtx_synack, |
60236fdd ACM |
827 | .send_ack = tcp_v6_reqsk_send_ack, |
828 | .destructor = tcp_v6_reqsk_destructor, | |
72659ecc | 829 | .send_reset = tcp_v6_send_reset, |
4aa956d8 | 830 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
831 | }; |
832 | ||
35b2c321 | 833 | const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = { |
2aec4a29 OP |
834 | .mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - |
835 | sizeof(struct ipv6hdr), | |
16bea70a | 836 | #ifdef CONFIG_TCP_MD5SIG |
fd3a154a | 837 | .req_md5_lookup = tcp_v6_md5_lookup, |
e3afe7b7 | 838 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
b6332e6c | 839 | #endif |
fb7b37a7 OP |
840 | #ifdef CONFIG_SYN_COOKIES |
841 | .cookie_init_seq = cookie_v6_init_sequence, | |
842 | #endif | |
d94e0417 | 843 | .route_req = tcp_v6_route_req, |
84b114b9 ED |
844 | .init_seq = tcp_v6_init_seq, |
845 | .init_ts_off = tcp_v6_init_ts_off, | |
d6274bd8 | 846 | .send_synack = tcp_v6_send_synack, |
16bea70a | 847 | }; |
cfb6eeb4 | 848 | |
a00e7444 | 849 | static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 seq, |
0f85feae ED |
850 | u32 ack, u32 win, u32 tsval, u32 tsecr, |
851 | int oif, struct tcp_md5sig_key *key, int rst, | |
aa51b80e | 852 | u8 tclass, __be32 label, u32 priority, u32 txhash) |
1da177e4 | 853 | { |
cf533ea5 ED |
854 | const struct tcphdr *th = tcp_hdr(skb); |
855 | struct tcphdr *t1; | |
1da177e4 | 856 | struct sk_buff *buff; |
4c9483b2 | 857 | struct flowi6 fl6; |
0f85feae | 858 | struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); |
e5047992 | 859 | struct sock *ctl_sk = net->ipv6.tcp_sk; |
77c676da | 860 | unsigned int tot_len = sizeof(struct tcphdr); |
dc87efdb | 861 | __be32 mrst = 0, *topt; |
adf30907 | 862 | struct dst_entry *dst; |
00483690 | 863 | __u32 mark = 0; |
1da177e4 | 864 | |
ee684b6f | 865 | if (tsecr) |
626e264d | 866 | tot_len += TCPOLEN_TSTAMP_ALIGNED; |
cfb6eeb4 | 867 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
868 | if (key) |
869 | tot_len += TCPOLEN_MD5SIG_ALIGNED; | |
870 | #endif | |
871 | ||
dc87efdb FW |
872 | #ifdef CONFIG_MPTCP |
873 | if (rst && !key) { | |
874 | mrst = mptcp_reset_option(skb); | |
875 | ||
876 | if (mrst) | |
877 | tot_len += sizeof(__be32); | |
878 | } | |
879 | #endif | |
880 | ||
0ebea8f9 | 881 | buff = alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC); |
63159f29 | 882 | if (!buff) |
1ab1457c | 883 | return; |
1da177e4 | 884 | |
0ebea8f9 | 885 | skb_reserve(buff, MAX_TCP_HEADER); |
1da177e4 | 886 | |
d58ff351 | 887 | t1 = skb_push(buff, tot_len); |
6651ffc8 | 888 | skb_reset_transport_header(buff); |
1da177e4 LT |
889 | |
890 | /* Swap the send and the receive. */ | |
891 | memset(t1, 0, sizeof(*t1)); | |
892 | t1->dest = th->source; | |
893 | t1->source = th->dest; | |
cfb6eeb4 | 894 | t1->doff = tot_len / 4; |
626e264d IJ |
895 | t1->seq = htonl(seq); |
896 | t1->ack_seq = htonl(ack); | |
897 | t1->ack = !rst || !th->ack; | |
898 | t1->rst = rst; | |
899 | t1->window = htons(win); | |
1da177e4 | 900 | |
81ada62d IJ |
901 | topt = (__be32 *)(t1 + 1); |
902 | ||
ee684b6f | 903 | if (tsecr) { |
626e264d IJ |
904 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
905 | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
906 | *topt++ = htonl(tsval); |
907 | *topt++ = htonl(tsecr); | |
626e264d IJ |
908 | } |
909 | ||
dc87efdb FW |
910 | if (mrst) |
911 | *topt++ = mrst; | |
912 | ||
cfb6eeb4 YH |
913 | #ifdef CONFIG_TCP_MD5SIG |
914 | if (key) { | |
81ada62d IJ |
915 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
916 | (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG); | |
917 | tcp_v6_md5_hash_hdr((__u8 *)topt, key, | |
78e645cb IJ |
918 | &ipv6_hdr(skb)->saddr, |
919 | &ipv6_hdr(skb)->daddr, t1); | |
cfb6eeb4 YH |
920 | } |
921 | #endif | |
922 | ||
4c9483b2 | 923 | memset(&fl6, 0, sizeof(fl6)); |
4e3fd7a0 AD |
924 | fl6.daddr = ipv6_hdr(skb)->saddr; |
925 | fl6.saddr = ipv6_hdr(skb)->daddr; | |
1d13a96c | 926 | fl6.flowlabel = label; |
1da177e4 | 927 | |
e5700aff | 928 | buff->ip_summed = CHECKSUM_PARTIAL; |
e5700aff | 929 | |
4c9483b2 | 930 | __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr); |
1da177e4 | 931 | |
4c9483b2 | 932 | fl6.flowi6_proto = IPPROTO_TCP; |
a36dbdb2 | 933 | if (rt6_need_strict(&fl6.daddr) && !oif) |
870c3151 | 934 | fl6.flowi6_oif = tcp_v6_iif(skb); |
9b6c14d5 DA |
935 | else { |
936 | if (!oif && netif_index_is_l3_master(net, skb->skb_iif)) | |
937 | oif = skb->skb_iif; | |
938 | ||
939 | fl6.flowi6_oif = oif; | |
940 | } | |
1d2f7b2d | 941 | |
c67b8555 | 942 | if (sk) { |
aa51b80e | 943 | if (sk->sk_state == TCP_TIME_WAIT) |
c67b8555 | 944 | mark = inet_twsk(sk)->tw_mark; |
aa51b80e | 945 | else |
c67b8555 | 946 | mark = sk->sk_mark; |
a1ac9c8a | 947 | skb_set_delivery_time(buff, tcp_transmit_time(sk), true); |
c67b8555 | 948 | } |
aa51b80e ED |
949 | if (txhash) { |
950 | /* autoflowlabel/skb_get_hash_flowi6 rely on buff->hash */ | |
951 | skb_set_hash(buff, txhash, PKT_HASH_TYPE_L4); | |
952 | } | |
00483690 | 953 | fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark) ?: mark; |
1958b856 DM |
954 | fl6.fl6_dport = t1->dest; |
955 | fl6.fl6_sport = t1->source; | |
e2d118a1 | 956 | fl6.flowi6_uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL); |
3df98d79 | 957 | security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6)); |
1da177e4 | 958 | |
c20121ae DL |
959 | /* Pass a socket to ip6_dst_lookup either it is for RST |
960 | * Underlying function will use this to retrieve the network | |
961 | * namespace | |
962 | */ | |
e22aa148 | 963 | if (sk && sk->sk_state != TCP_TIME_WAIT) |
964 | dst = ip6_dst_lookup_flow(net, sk, &fl6, NULL); /*sk's xfrm_policy can be referred*/ | |
965 | else | |
966 | dst = ip6_dst_lookup_flow(net, ctl_sk, &fl6, NULL); | |
68d0c6d3 DM |
967 | if (!IS_ERR(dst)) { |
968 | skb_dst_set(buff, dst); | |
e92dd77e WW |
969 | ip6_xmit(ctl_sk, buff, &fl6, fl6.flowi6_mark, NULL, |
970 | tclass & ~INET_ECN_MASK, priority); | |
c10d9310 | 971 | TCP_INC_STATS(net, TCP_MIB_OUTSEGS); |
68d0c6d3 | 972 | if (rst) |
c10d9310 | 973 | TCP_INC_STATS(net, TCP_MIB_OUTRSTS); |
68d0c6d3 | 974 | return; |
1da177e4 LT |
975 | } |
976 | ||
977 | kfree_skb(buff); | |
978 | } | |
979 | ||
a00e7444 | 980 | static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb) |
1da177e4 | 981 | { |
cf533ea5 | 982 | const struct tcphdr *th = tcp_hdr(skb); |
323a53c4 | 983 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); |
626e264d | 984 | u32 seq = 0, ack_seq = 0; |
fa3e5b4e | 985 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
986 | #ifdef CONFIG_TCP_MD5SIG |
987 | const __u8 *hash_location = NULL; | |
658ddaaf SL |
988 | unsigned char newhash[16]; |
989 | int genhash; | |
990 | struct sock *sk1 = NULL; | |
991 | #endif | |
323a53c4 | 992 | __be32 label = 0; |
e9a5dcee | 993 | u32 priority = 0; |
323a53c4 | 994 | struct net *net; |
9258b8b1 | 995 | u32 txhash = 0; |
c24b14c4 | 996 | int oif = 0; |
1da177e4 | 997 | |
626e264d | 998 | if (th->rst) |
1da177e4 LT |
999 | return; |
1000 | ||
c3658e8d ED |
1001 | /* If sk not NULL, it means we did a successful lookup and incoming |
1002 | * route had to be correct. prequeue might have dropped our dst. | |
1003 | */ | |
1004 | if (!sk && !ipv6_unicast_destination(skb)) | |
626e264d | 1005 | return; |
1da177e4 | 1006 | |
39209673 | 1007 | net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); |
cfb6eeb4 | 1008 | #ifdef CONFIG_TCP_MD5SIG |
3b24d854 | 1009 | rcu_read_lock(); |
658ddaaf | 1010 | hash_location = tcp_parse_md5sig_option(th); |
271c3b9b | 1011 | if (sk && sk_fullsock(sk)) { |
dea53bb8 DA |
1012 | int l3index; |
1013 | ||
1014 | /* sdif set, means packet ingressed via a device | |
1015 | * in an L3 domain and inet_iif is set to it. | |
1016 | */ | |
1017 | l3index = tcp_v6_sdif(skb) ? tcp_v6_iif_l3_slave(skb) : 0; | |
1018 | key = tcp_v6_md5_do_lookup(sk, &ipv6h->saddr, l3index); | |
e46787f0 | 1019 | } else if (hash_location) { |
d14c77e0 DA |
1020 | int dif = tcp_v6_iif_l3_slave(skb); |
1021 | int sdif = tcp_v6_sdif(skb); | |
dea53bb8 | 1022 | int l3index; |
d14c77e0 | 1023 | |
658ddaaf SL |
1024 | /* |
1025 | * active side is lost. Try to find listening socket through | |
1026 | * source port, and then find md5 key through listening socket. | |
1027 | * we are not loose security here: | |
1028 | * Incoming packet is checked with md5 hash with finding key, | |
1029 | * no RST generated if md5 hash doesn't match. | |
1030 | */ | |
4461568a KI |
1031 | sk1 = inet6_lookup_listener(net, net->ipv4.tcp_death_row.hashinfo, |
1032 | NULL, 0, &ipv6h->saddr, th->source, | |
1033 | &ipv6h->daddr, ntohs(th->source), | |
1034 | dif, sdif); | |
658ddaaf | 1035 | if (!sk1) |
3b24d854 | 1036 | goto out; |
658ddaaf | 1037 | |
dea53bb8 DA |
1038 | /* sdif set, means packet ingressed via a device |
1039 | * in an L3 domain and dif is set to it. | |
1040 | */ | |
1041 | l3index = tcp_v6_sdif(skb) ? dif : 0; | |
1042 | ||
1043 | key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr, l3index); | |
658ddaaf | 1044 | if (!key) |
3b24d854 | 1045 | goto out; |
658ddaaf | 1046 | |
39f8e58e | 1047 | genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb); |
658ddaaf | 1048 | if (genhash || memcmp(hash_location, newhash, 16) != 0) |
3b24d854 | 1049 | goto out; |
658ddaaf | 1050 | } |
cfb6eeb4 YH |
1051 | #endif |
1052 | ||
626e264d IJ |
1053 | if (th->ack) |
1054 | seq = ntohl(th->ack_seq); | |
1055 | else | |
1056 | ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len - | |
1057 | (th->doff << 2); | |
1da177e4 | 1058 | |
c24b14c4 SL |
1059 | if (sk) { |
1060 | oif = sk->sk_bound_dev_if; | |
052e0690 ED |
1061 | if (sk_fullsock(sk)) { |
1062 | const struct ipv6_pinfo *np = tcp_inet6_sk(sk); | |
1063 | ||
5c487bb9 | 1064 | trace_tcp_send_reset(sk, skb); |
052e0690 ED |
1065 | if (np->repflow) |
1066 | label = ip6_flowlabel(ipv6h); | |
e9a5dcee | 1067 | priority = sk->sk_priority; |
dc6456e9 | 1068 | txhash = sk->sk_txhash; |
052e0690 | 1069 | } |
f6c0f5d2 | 1070 | if (sk->sk_state == TCP_TIME_WAIT) { |
50a8accf | 1071 | label = cpu_to_be32(inet_twsk(sk)->tw_flowlabel); |
f6c0f5d2 | 1072 | priority = inet_twsk(sk)->tw_priority; |
9258b8b1 | 1073 | txhash = inet_twsk(sk)->tw_txhash; |
f6c0f5d2 | 1074 | } |
323a53c4 | 1075 | } else { |
a346abe0 | 1076 | if (net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_TCP_RESET) |
323a53c4 | 1077 | label = ip6_flowlabel(ipv6h); |
c24b14c4 SL |
1078 | } |
1079 | ||
e92dd77e | 1080 | tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, |
9258b8b1 | 1081 | ipv6_get_dsfield(ipv6h), label, priority, txhash); |
658ddaaf SL |
1082 | |
1083 | #ifdef CONFIG_TCP_MD5SIG | |
3b24d854 ED |
1084 | out: |
1085 | rcu_read_unlock(); | |
658ddaaf | 1086 | #endif |
626e264d | 1087 | } |
1da177e4 | 1088 | |
a00e7444 | 1089 | static void tcp_v6_send_ack(const struct sock *sk, struct sk_buff *skb, u32 seq, |
0f85feae | 1090 | u32 ack, u32 win, u32 tsval, u32 tsecr, int oif, |
1d13a96c | 1091 | struct tcp_md5sig_key *key, u8 tclass, |
aa51b80e | 1092 | __be32 label, u32 priority, u32 txhash) |
626e264d | 1093 | { |
0f85feae | 1094 | tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0, |
aa51b80e | 1095 | tclass, label, priority, txhash); |
1da177e4 LT |
1096 | } |
1097 | ||
1098 | static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
1099 | { | |
8feaf0c0 | 1100 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 1101 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 1102 | |
0f85feae | 1103 | tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
8feaf0c0 | 1104 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
9a568de4 | 1105 | tcp_time_stamp_raw() + tcptw->tw_ts_offset, |
9c76a114 | 1106 | tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw), |
aa51b80e ED |
1107 | tw->tw_tclass, cpu_to_be32(tw->tw_flowlabel), tw->tw_priority, |
1108 | tw->tw_txhash); | |
1da177e4 | 1109 | |
8feaf0c0 | 1110 | inet_twsk_put(tw); |
1da177e4 LT |
1111 | } |
1112 | ||
a00e7444 | 1113 | static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb, |
6edafaaf | 1114 | struct request_sock *req) |
1da177e4 | 1115 | { |
dea53bb8 DA |
1116 | int l3index; |
1117 | ||
1118 | l3index = tcp_v6_sdif(skb) ? tcp_v6_iif_l3_slave(skb) : 0; | |
1119 | ||
3a19ce0e DL |
1120 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
1121 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
1122 | */ | |
20a2b49f ED |
1123 | /* RFC 7323 2.3 |
1124 | * The window field (SEG.WND) of every outgoing segment, with the | |
1125 | * exception of <SYN> segments, MUST be right-shifted by | |
1126 | * Rcv.Wind.Shift bits: | |
1127 | */ | |
0f85feae | 1128 | tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ? |
3a19ce0e | 1129 | tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt, |
20a2b49f ED |
1130 | tcp_rsk(req)->rcv_nxt, |
1131 | req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale, | |
9a568de4 | 1132 | tcp_time_stamp_raw() + tcp_rsk(req)->ts_off, |
95a22cae | 1133 | req->ts_recent, sk->sk_bound_dev_if, |
dea53bb8 | 1134 | tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->saddr, l3index), |
aa51b80e ED |
1135 | ipv6_get_dsfield(ipv6_hdr(skb)), 0, sk->sk_priority, |
1136 | tcp_rsk(req)->txhash); | |
1da177e4 LT |
1137 | } |
1138 | ||
1139 | ||
079096f1 | 1140 | static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1141 | { |
079096f1 | 1142 | #ifdef CONFIG_SYN_COOKIES |
aa8223c7 | 1143 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 1144 | |
af9b4738 | 1145 | if (!th->syn) |
c6aefafb | 1146 | sk = cookie_v6_check(sk, skb); |
1da177e4 LT |
1147 | #endif |
1148 | return sk; | |
1149 | } | |
1150 | ||
9349d600 PP |
1151 | u16 tcp_v6_get_syncookie(struct sock *sk, struct ipv6hdr *iph, |
1152 | struct tcphdr *th, u32 *cookie) | |
1153 | { | |
1154 | u16 mss = 0; | |
1155 | #ifdef CONFIG_SYN_COOKIES | |
1156 | mss = tcp_get_syncookie_mss(&tcp6_request_sock_ops, | |
1157 | &tcp_request_sock_ipv6_ops, sk, th); | |
1158 | if (mss) { | |
1159 | *cookie = __cookie_v6_init_sequence(iph, th, &mss); | |
1160 | tcp_synq_overflow(sk); | |
1161 | } | |
1162 | #endif | |
1163 | return mss; | |
1164 | } | |
1165 | ||
1da177e4 LT |
1166 | static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb) |
1167 | { | |
1da177e4 LT |
1168 | if (skb->protocol == htons(ETH_P_IP)) |
1169 | return tcp_v4_conn_request(sk, skb); | |
1170 | ||
1171 | if (!ipv6_unicast_destination(skb)) | |
1ab1457c | 1172 | goto drop; |
1da177e4 | 1173 | |
dcc32f4f JK |
1174 | if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) { |
1175 | __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS); | |
1176 | return 0; | |
1177 | } | |
1178 | ||
1fb6f159 OP |
1179 | return tcp_conn_request(&tcp6_request_sock_ops, |
1180 | &tcp_request_sock_ipv6_ops, sk, skb); | |
1da177e4 LT |
1181 | |
1182 | drop: | |
9caad864 | 1183 | tcp_listendrop(sk); |
1da177e4 LT |
1184 | return 0; /* don't send reset */ |
1185 | } | |
1186 | ||
ebf6c9cb ED |
1187 | static void tcp_v6_restore_cb(struct sk_buff *skb) |
1188 | { | |
1189 | /* We need to move header back to the beginning if xfrm6_policy_check() | |
1190 | * and tcp_v6_fill_cb() are going to be called again. | |
1191 | * ip6_datagram_recv_specific_ctl() also expects IP6CB to be there. | |
1192 | */ | |
1193 | memmove(IP6CB(skb), &TCP_SKB_CB(skb)->header.h6, | |
1194 | sizeof(struct inet6_skb_parm)); | |
1195 | } | |
1196 | ||
0c27171e | 1197 | static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *skb, |
4c99aa40 | 1198 | struct request_sock *req, |
5e0724d0 ED |
1199 | struct dst_entry *dst, |
1200 | struct request_sock *req_unhash, | |
1201 | bool *own_req) | |
1da177e4 | 1202 | { |
634fb979 | 1203 | struct inet_request_sock *ireq; |
0c27171e | 1204 | struct ipv6_pinfo *newnp; |
93a77c11 | 1205 | const struct ipv6_pinfo *np = tcp_inet6_sk(sk); |
45f6fad8 | 1206 | struct ipv6_txoptions *opt; |
1da177e4 | 1207 | struct inet_sock *newinet; |
01770a16 | 1208 | bool found_dup_sk = false; |
1da177e4 LT |
1209 | struct tcp_sock *newtp; |
1210 | struct sock *newsk; | |
cfb6eeb4 YH |
1211 | #ifdef CONFIG_TCP_MD5SIG |
1212 | struct tcp_md5sig_key *key; | |
dea53bb8 | 1213 | int l3index; |
cfb6eeb4 | 1214 | #endif |
3840a06e | 1215 | struct flowi6 fl6; |
1da177e4 LT |
1216 | |
1217 | if (skb->protocol == htons(ETH_P_IP)) { | |
1218 | /* | |
1219 | * v6 mapped | |
1220 | */ | |
1221 | ||
5e0724d0 ED |
1222 | newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst, |
1223 | req_unhash, own_req); | |
1da177e4 | 1224 | |
63159f29 | 1225 | if (!newsk) |
1da177e4 LT |
1226 | return NULL; |
1227 | ||
93a77c11 | 1228 | inet_sk(newsk)->pinet6 = tcp_inet6_sk(newsk); |
1da177e4 | 1229 | |
93a77c11 | 1230 | newnp = tcp_inet6_sk(newsk); |
1da177e4 LT |
1231 | newtp = tcp_sk(newsk); |
1232 | ||
1233 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1234 | ||
d1e559d0 | 1235 | newnp->saddr = newsk->sk_v6_rcv_saddr; |
1da177e4 | 1236 | |
8292a17a | 1237 | inet_csk(newsk)->icsk_af_ops = &ipv6_mapped; |
cec37a6e | 1238 | if (sk_is_mptcp(newsk)) |
31484d56 | 1239 | mptcpv6_handle_mapped(newsk, true); |
1da177e4 | 1240 | newsk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
1241 | #ifdef CONFIG_TCP_MD5SIG |
1242 | newtp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
1243 | #endif | |
1244 | ||
83eaddab | 1245 | newnp->ipv6_mc_list = NULL; |
676a1184 YZ |
1246 | newnp->ipv6_ac_list = NULL; |
1247 | newnp->ipv6_fl_list = NULL; | |
1da177e4 LT |
1248 | newnp->pktoptions = NULL; |
1249 | newnp->opt = NULL; | |
89e41309 ED |
1250 | newnp->mcast_oif = inet_iif(skb); |
1251 | newnp->mcast_hops = ip_hdr(skb)->ttl; | |
1252 | newnp->rcv_flowinfo = 0; | |
df3687ff | 1253 | if (np->repflow) |
89e41309 | 1254 | newnp->flow_label = 0; |
1da177e4 | 1255 | |
e6848976 ACM |
1256 | /* |
1257 | * No need to charge this sock to the relevant IPv6 refcnt debug socks count | |
1258 | * here, tcp_create_openreq_child now does this for us, see the comment in | |
1259 | * that function for the gory details. -acme | |
1da177e4 | 1260 | */ |
1da177e4 LT |
1261 | |
1262 | /* It is tricky place. Until this moment IPv4 tcp | |
8292a17a | 1263 | worked with IPv6 icsk.icsk_af_ops. |
1da177e4 LT |
1264 | Sync it now. |
1265 | */ | |
d83d8461 | 1266 | tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); |
1da177e4 LT |
1267 | |
1268 | return newsk; | |
1269 | } | |
1270 | ||
634fb979 | 1271 | ireq = inet_rsk(req); |
1da177e4 LT |
1272 | |
1273 | if (sk_acceptq_is_full(sk)) | |
1274 | goto out_overflow; | |
1275 | ||
493f377d | 1276 | if (!dst) { |
f76b33c3 | 1277 | dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_TCP); |
493f377d | 1278 | if (!dst) |
1da177e4 | 1279 | goto out; |
1ab1457c | 1280 | } |
1da177e4 LT |
1281 | |
1282 | newsk = tcp_create_openreq_child(sk, req, skb); | |
63159f29 | 1283 | if (!newsk) |
093d2823 | 1284 | goto out_nonewsk; |
1da177e4 | 1285 | |
e6848976 ACM |
1286 | /* |
1287 | * No need to charge this sock to the relevant IPv6 refcnt debug socks | |
1288 | * count here, tcp_create_openreq_child now does this for us, see the | |
1289 | * comment in that function for the gory details. -acme | |
1290 | */ | |
1da177e4 | 1291 | |
59eed279 | 1292 | newsk->sk_gso_type = SKB_GSO_TCPV6; |
6bd4f355 | 1293 | ip6_dst_store(newsk, dst, NULL, NULL); |
fae6ef87 | 1294 | inet6_sk_rx_dst_set(newsk, skb); |
1da177e4 | 1295 | |
93a77c11 | 1296 | inet_sk(newsk)->pinet6 = tcp_inet6_sk(newsk); |
1da177e4 LT |
1297 | |
1298 | newtp = tcp_sk(newsk); | |
1299 | newinet = inet_sk(newsk); | |
93a77c11 | 1300 | newnp = tcp_inet6_sk(newsk); |
1da177e4 LT |
1301 | |
1302 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1303 | ||
634fb979 ED |
1304 | newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr; |
1305 | newnp->saddr = ireq->ir_v6_loc_addr; | |
1306 | newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr; | |
1307 | newsk->sk_bound_dev_if = ireq->ir_iif; | |
1da177e4 | 1308 | |
1ab1457c | 1309 | /* Now IPv6 options... |
1da177e4 LT |
1310 | |
1311 | First: no IPv4 options. | |
1312 | */ | |
f6d8bd05 | 1313 | newinet->inet_opt = NULL; |
83eaddab | 1314 | newnp->ipv6_mc_list = NULL; |
676a1184 | 1315 | newnp->ipv6_ac_list = NULL; |
d35690be | 1316 | newnp->ipv6_fl_list = NULL; |
1da177e4 LT |
1317 | |
1318 | /* Clone RX bits */ | |
1319 | newnp->rxopt.all = np->rxopt.all; | |
1320 | ||
1da177e4 | 1321 | newnp->pktoptions = NULL; |
1da177e4 | 1322 | newnp->opt = NULL; |
870c3151 | 1323 | newnp->mcast_oif = tcp_v6_iif(skb); |
0660e03f | 1324 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
1397ed35 | 1325 | newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb)); |
df3687ff FF |
1326 | if (np->repflow) |
1327 | newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb)); | |
1da177e4 | 1328 | |
8ef44b6f WW |
1329 | /* Set ToS of the new socket based upon the value of incoming SYN. |
1330 | * ECT bits are set later in tcp_init_transfer(). | |
1331 | */ | |
870e3a63 | 1332 | if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_reflect_tos)) |
ac8f1710 WW |
1333 | newnp->tclass = tcp_rsk(req)->syn_tos & ~INET_ECN_MASK; |
1334 | ||
1da177e4 LT |
1335 | /* Clone native IPv6 options from listening socket (if any) |
1336 | ||
1337 | Yes, keeping reference count would be much more clever, | |
1338 | but we make one more one thing there: reattach optmem | |
1339 | to newsk. | |
1340 | */ | |
56ac42bc HD |
1341 | opt = ireq->ipv6_opt; |
1342 | if (!opt) | |
1343 | opt = rcu_dereference(np->opt); | |
45f6fad8 ED |
1344 | if (opt) { |
1345 | opt = ipv6_dup_options(newsk, opt); | |
1346 | RCU_INIT_POINTER(newnp->opt, opt); | |
1347 | } | |
d83d8461 | 1348 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
45f6fad8 ED |
1349 | if (opt) |
1350 | inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen + | |
1351 | opt->opt_flen; | |
1da177e4 | 1352 | |
81164413 DB |
1353 | tcp_ca_openreq_child(newsk, dst); |
1354 | ||
1da177e4 | 1355 | tcp_sync_mss(newsk, dst_mtu(dst)); |
3541f9e8 | 1356 | newtp->advmss = tcp_mss_clamp(tcp_sk(sk), dst_metric_advmss(dst)); |
d135c522 | 1357 | |
1da177e4 LT |
1358 | tcp_initialize_rcv_mss(newsk); |
1359 | ||
c720c7e8 ED |
1360 | newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; |
1361 | newinet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
1da177e4 | 1362 | |
cfb6eeb4 | 1363 | #ifdef CONFIG_TCP_MD5SIG |
dea53bb8 DA |
1364 | l3index = l3mdev_master_ifindex_by_index(sock_net(sk), ireq->ir_iif); |
1365 | ||
cfb6eeb4 | 1366 | /* Copy over the MD5 key from the original socket */ |
dea53bb8 | 1367 | key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr, l3index); |
53b24b8f | 1368 | if (key) { |
b389d1af DS |
1369 | const union tcp_md5_addr *addr; |
1370 | ||
1371 | addr = (union tcp_md5_addr *)&newsk->sk_v6_daddr; | |
1372 | if (tcp_md5_key_copy(newsk, addr, AF_INET6, 128, l3index, key)) { | |
1373 | inet_csk_prepare_forced_close(newsk); | |
1374 | tcp_done(newsk); | |
1375 | goto out; | |
1376 | } | |
cfb6eeb4 YH |
1377 | } |
1378 | #endif | |
1379 | ||
093d2823 | 1380 | if (__inet_inherit_port(sk, newsk) < 0) { |
e337e24d CP |
1381 | inet_csk_prepare_forced_close(newsk); |
1382 | tcp_done(newsk); | |
093d2823 BS |
1383 | goto out; |
1384 | } | |
01770a16 RD |
1385 | *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), |
1386 | &found_dup_sk); | |
805c4bc0 | 1387 | if (*own_req) { |
49a496c9 | 1388 | tcp_move_syn(newtp, req); |
805c4bc0 ED |
1389 | |
1390 | /* Clone pktoptions received with SYN, if we own the req */ | |
1391 | if (ireq->pktopts) { | |
ca43ccf4 | 1392 | newnp->pktoptions = skb_clone_and_charge_r(ireq->pktopts, newsk); |
805c4bc0 ED |
1393 | consume_skb(ireq->pktopts); |
1394 | ireq->pktopts = NULL; | |
ca43ccf4 | 1395 | if (newnp->pktoptions) |
ebf6c9cb | 1396 | tcp_v6_restore_cb(newnp->pktoptions); |
805c4bc0 | 1397 | } |
01770a16 RD |
1398 | } else { |
1399 | if (!req_unhash && found_dup_sk) { | |
1400 | /* This code path should only be executed in the | |
1401 | * syncookie case only | |
1402 | */ | |
1403 | bh_unlock_sock(newsk); | |
1404 | sock_put(newsk); | |
1405 | newsk = NULL; | |
1406 | } | |
ce105008 | 1407 | } |
1da177e4 LT |
1408 | |
1409 | return newsk; | |
1410 | ||
1411 | out_overflow: | |
02a1d6e7 | 1412 | __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 | 1413 | out_nonewsk: |
1da177e4 | 1414 | dst_release(dst); |
093d2823 | 1415 | out: |
9caad864 | 1416 | tcp_listendrop(sk); |
1da177e4 LT |
1417 | return NULL; |
1418 | } | |
1419 | ||
bbd807df BV |
1420 | INDIRECT_CALLABLE_DECLARE(struct dst_entry *ipv4_dst_check(struct dst_entry *, |
1421 | u32)); | |
1da177e4 | 1422 | /* The socket must have it's spinlock held when we get |
e994b2f0 | 1423 | * here, unless it is a TCP_LISTEN socket. |
1da177e4 LT |
1424 | * |
1425 | * We have a potential double-lock case here, so even when | |
1426 | * doing backlog processing we use the BH locking scheme. | |
1427 | * This is because we cannot sleep with the original spinlock | |
1428 | * held. | |
1429 | */ | |
d2489c7b ED |
1430 | INDIRECT_CALLABLE_SCOPE |
1431 | int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1da177e4 | 1432 | { |
93a77c11 | 1433 | struct ipv6_pinfo *np = tcp_inet6_sk(sk); |
1da177e4 | 1434 | struct sk_buff *opt_skb = NULL; |
8eba65fa | 1435 | enum skb_drop_reason reason; |
93a77c11 | 1436 | struct tcp_sock *tp; |
1da177e4 LT |
1437 | |
1438 | /* Imagine: socket is IPv6. IPv4 packet arrives, | |
1439 | goes to IPv4 receive handler and backlogged. | |
1440 | From backlog it always goes here. Kerboom... | |
1441 | Fortunately, tcp_rcv_established and rcv_established | |
1442 | handle them correctly, but it is not case with | |
1443 | tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK | |
1444 | */ | |
1445 | ||
1446 | if (skb->protocol == htons(ETH_P_IP)) | |
1447 | return tcp_v4_do_rcv(sk, skb); | |
1448 | ||
1da177e4 LT |
1449 | /* |
1450 | * socket locking is here for SMP purposes as backlog rcv | |
1451 | * is currently called with bh processing disabled. | |
1452 | */ | |
1453 | ||
1454 | /* Do Stevens' IPV6_PKTOPTIONS. | |
1455 | ||
1456 | Yes, guys, it is the only place in our code, where we | |
1457 | may make it not affecting IPv4. | |
1458 | The rest of code is protocol independent, | |
1459 | and I do not like idea to uglify IPv4. | |
1460 | ||
1461 | Actually, all the idea behind IPV6_PKTOPTIONS | |
1462 | looks not very well thought. For now we latch | |
1463 | options, received in the last packet, enqueued | |
1464 | by tcp. Feel free to propose better solution. | |
1ab1457c | 1465 | --ANK (980728) |
1da177e4 LT |
1466 | */ |
1467 | if (np->rxopt.all) | |
ca43ccf4 | 1468 | opt_skb = skb_clone_and_charge_r(skb, sk); |
1da177e4 | 1469 | |
8eba65fa | 1470 | reason = SKB_DROP_REASON_NOT_SPECIFIED; |
1da177e4 | 1471 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ |
8f905c0e ED |
1472 | struct dst_entry *dst; |
1473 | ||
1474 | dst = rcu_dereference_protected(sk->sk_rx_dst, | |
1475 | lockdep_sock_is_held(sk)); | |
5d299f3d | 1476 | |
bdeab991 | 1477 | sock_rps_save_rxhash(sk, skb); |
3d97379a | 1478 | sk_mark_napi_id(sk, skb); |
5d299f3d | 1479 | if (dst) { |
0c0a5ef8 | 1480 | if (sk->sk_rx_dst_ifindex != skb->skb_iif || |
bbd807df | 1481 | INDIRECT_CALL_1(dst->ops->check, ip6_dst_check, |
ef57c161 | 1482 | dst, sk->sk_rx_dst_cookie) == NULL) { |
8f905c0e | 1483 | RCU_INIT_POINTER(sk->sk_rx_dst, NULL); |
5d299f3d | 1484 | dst_release(dst); |
5d299f3d ED |
1485 | } |
1486 | } | |
1487 | ||
3d97d88e | 1488 | tcp_rcv_established(sk, skb); |
1da177e4 LT |
1489 | if (opt_skb) |
1490 | goto ipv6_pktoptions; | |
1491 | return 0; | |
1492 | } | |
1493 | ||
12e25e10 | 1494 | if (tcp_checksum_complete(skb)) |
1da177e4 LT |
1495 | goto csum_err; |
1496 | ||
1ab1457c | 1497 | if (sk->sk_state == TCP_LISTEN) { |
079096f1 ED |
1498 | struct sock *nsk = tcp_v6_cookie_check(sk, skb); |
1499 | ||
1da177e4 LT |
1500 | if (!nsk) |
1501 | goto discard; | |
1502 | ||
4c99aa40 | 1503 | if (nsk != sk) { |
1da177e4 LT |
1504 | if (tcp_child_process(sk, nsk, skb)) |
1505 | goto reset; | |
1506 | if (opt_skb) | |
1507 | __kfree_skb(opt_skb); | |
1508 | return 0; | |
1509 | } | |
47482f13 | 1510 | } else |
bdeab991 | 1511 | sock_rps_save_rxhash(sk, skb); |
1da177e4 | 1512 | |
72ab4a86 | 1513 | if (tcp_rcv_state_process(sk, skb)) |
1da177e4 | 1514 | goto reset; |
1da177e4 LT |
1515 | if (opt_skb) |
1516 | goto ipv6_pktoptions; | |
1517 | return 0; | |
1518 | ||
1519 | reset: | |
cfb6eeb4 | 1520 | tcp_v6_send_reset(sk, skb); |
1da177e4 LT |
1521 | discard: |
1522 | if (opt_skb) | |
1523 | __kfree_skb(opt_skb); | |
8eba65fa | 1524 | kfree_skb_reason(skb, reason); |
1da177e4 LT |
1525 | return 0; |
1526 | csum_err: | |
8eba65fa | 1527 | reason = SKB_DROP_REASON_TCP_CSUM; |
709c0314 | 1528 | trace_tcp_bad_csum(skb); |
c10d9310 ED |
1529 | TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS); |
1530 | TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS); | |
1da177e4 LT |
1531 | goto discard; |
1532 | ||
1533 | ||
1534 | ipv6_pktoptions: | |
1535 | /* Do you ask, what is it? | |
1536 | ||
1537 | 1. skb was enqueued by tcp. | |
1538 | 2. skb is added to tail of read queue, rather than out of order. | |
1539 | 3. socket is not in passive state. | |
1540 | 4. Finally, it really contains options, which user wants to receive. | |
1541 | */ | |
1542 | tp = tcp_sk(sk); | |
1543 | if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt && | |
1544 | !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { | |
333fad53 | 1545 | if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) |
870c3151 | 1546 | np->mcast_oif = tcp_v6_iif(opt_skb); |
333fad53 | 1547 | if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) |
0660e03f | 1548 | np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; |
82e9f105 | 1549 | if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass) |
1397ed35 | 1550 | np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb)); |
df3687ff FF |
1551 | if (np->repflow) |
1552 | np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb)); | |
a224772d | 1553 | if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) { |
8ce48623 | 1554 | tcp_v6_restore_cb(opt_skb); |
1da177e4 LT |
1555 | opt_skb = xchg(&np->pktoptions, opt_skb); |
1556 | } else { | |
1557 | __kfree_skb(opt_skb); | |
1558 | opt_skb = xchg(&np->pktoptions, NULL); | |
1559 | } | |
1560 | } | |
1561 | ||
12c8691d | 1562 | consume_skb(opt_skb); |
1da177e4 LT |
1563 | return 0; |
1564 | } | |
1565 | ||
2dc49d16 ND |
1566 | static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr, |
1567 | const struct tcphdr *th) | |
1568 | { | |
1569 | /* This is tricky: we move IP6CB at its correct location into | |
1570 | * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because | |
1571 | * _decode_session6() uses IP6CB(). | |
1572 | * barrier() makes sure compiler won't play aliasing games. | |
1573 | */ | |
1574 | memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb), | |
1575 | sizeof(struct inet6_skb_parm)); | |
1576 | barrier(); | |
1577 | ||
1578 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); | |
1579 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1580 | skb->len - th->doff*4); | |
1581 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1582 | TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th); | |
1583 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; | |
1584 | TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr); | |
1585 | TCP_SKB_CB(skb)->sacked = 0; | |
98aaa913 MM |
1586 | TCP_SKB_CB(skb)->has_rxtstamp = |
1587 | skb->tstamp || skb_hwtstamps(skb)->hwtstamp; | |
2dc49d16 ND |
1588 | } |
1589 | ||
0e219ae4 | 1590 | INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb) |
1da177e4 | 1591 | { |
c0e3154d | 1592 | enum skb_drop_reason drop_reason; |
4297a0ef | 1593 | int sdif = inet6_sdif(skb); |
d14c77e0 | 1594 | int dif = inet6_iif(skb); |
cf533ea5 | 1595 | const struct tcphdr *th; |
b71d1d42 | 1596 | const struct ipv6hdr *hdr; |
3b24d854 | 1597 | bool refcounted; |
1da177e4 LT |
1598 | struct sock *sk; |
1599 | int ret; | |
a86b1e30 | 1600 | struct net *net = dev_net(skb->dev); |
1da177e4 | 1601 | |
c0e3154d | 1602 | drop_reason = SKB_DROP_REASON_NOT_SPECIFIED; |
1da177e4 LT |
1603 | if (skb->pkt_type != PACKET_HOST) |
1604 | goto discard_it; | |
1605 | ||
1606 | /* | |
1607 | * Count it even if it's bad. | |
1608 | */ | |
90bbcc60 | 1609 | __TCP_INC_STATS(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1610 | |
1611 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1612 | goto discard_it; | |
1613 | ||
ea1627c2 | 1614 | th = (const struct tcphdr *)skb->data; |
1da177e4 | 1615 | |
c0e3154d MD |
1616 | if (unlikely(th->doff < sizeof(struct tcphdr) / 4)) { |
1617 | drop_reason = SKB_DROP_REASON_PKT_TOO_SMALL; | |
1da177e4 | 1618 | goto bad_packet; |
c0e3154d | 1619 | } |
1da177e4 LT |
1620 | if (!pskb_may_pull(skb, th->doff*4)) |
1621 | goto discard_it; | |
1622 | ||
e4f45b7f | 1623 | if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo)) |
6a5dc9e5 | 1624 | goto csum_error; |
1da177e4 | 1625 | |
ea1627c2 | 1626 | th = (const struct tcphdr *)skb->data; |
e802af9c | 1627 | hdr = ipv6_hdr(skb); |
1da177e4 | 1628 | |
4bdc3d66 | 1629 | lookup: |
4461568a | 1630 | sk = __inet6_lookup_skb(net->ipv4.tcp_death_row.hashinfo, skb, __tcp_hdrlen(th), |
4297a0ef | 1631 | th->source, th->dest, inet6_iif(skb), sdif, |
3b24d854 | 1632 | &refcounted); |
1da177e4 LT |
1633 | if (!sk) |
1634 | goto no_tcp_socket; | |
1635 | ||
1636 | process: | |
1637 | if (sk->sk_state == TCP_TIME_WAIT) | |
1638 | goto do_time_wait; | |
1639 | ||
079096f1 ED |
1640 | if (sk->sk_state == TCP_NEW_SYN_RECV) { |
1641 | struct request_sock *req = inet_reqsk(sk); | |
e0f9759f | 1642 | bool req_stolen = false; |
7716682c | 1643 | struct sock *nsk; |
079096f1 ED |
1644 | |
1645 | sk = req->rsk_listener; | |
1330b6ef JK |
1646 | drop_reason = tcp_inbound_md5_hash(sk, skb, |
1647 | &hdr->saddr, &hdr->daddr, | |
1648 | AF_INET6, dif, sdif); | |
1649 | if (drop_reason) { | |
e65c332d | 1650 | sk_drops_add(sk, skb); |
079096f1 ED |
1651 | reqsk_put(req); |
1652 | goto discard_it; | |
1653 | } | |
4fd44a98 FL |
1654 | if (tcp_checksum_complete(skb)) { |
1655 | reqsk_put(req); | |
1656 | goto csum_error; | |
1657 | } | |
7716682c | 1658 | if (unlikely(sk->sk_state != TCP_LISTEN)) { |
d4f2c86b KI |
1659 | nsk = reuseport_migrate_sock(sk, req_to_sk(req), skb); |
1660 | if (!nsk) { | |
1661 | inet_csk_reqsk_queue_drop_and_put(sk, req); | |
1662 | goto lookup; | |
1663 | } | |
1664 | sk = nsk; | |
1665 | /* reuseport_migrate_sock() has already held one sk_refcnt | |
1666 | * before returning. | |
1667 | */ | |
1668 | } else { | |
1669 | sock_hold(sk); | |
4bdc3d66 | 1670 | } |
3b24d854 | 1671 | refcounted = true; |
1f3b359f | 1672 | nsk = NULL; |
eeea10b8 ED |
1673 | if (!tcp_filter(sk, skb)) { |
1674 | th = (const struct tcphdr *)skb->data; | |
1675 | hdr = ipv6_hdr(skb); | |
1676 | tcp_v6_fill_cb(skb, hdr, th); | |
e0f9759f | 1677 | nsk = tcp_check_req(sk, skb, req, false, &req_stolen); |
c0e3154d MD |
1678 | } else { |
1679 | drop_reason = SKB_DROP_REASON_SOCKET_FILTER; | |
eeea10b8 | 1680 | } |
079096f1 ED |
1681 | if (!nsk) { |
1682 | reqsk_put(req); | |
e0f9759f ED |
1683 | if (req_stolen) { |
1684 | /* Another cpu got exclusive access to req | |
1685 | * and created a full blown socket. | |
1686 | * Try to feed this packet to this socket | |
1687 | * instead of discarding it. | |
1688 | */ | |
1689 | tcp_v6_restore_cb(skb); | |
1690 | sock_put(sk); | |
1691 | goto lookup; | |
1692 | } | |
7716682c | 1693 | goto discard_and_relse; |
079096f1 ED |
1694 | } |
1695 | if (nsk == sk) { | |
079096f1 ED |
1696 | reqsk_put(req); |
1697 | tcp_v6_restore_cb(skb); | |
1698 | } else if (tcp_child_process(sk, nsk, skb)) { | |
1699 | tcp_v6_send_reset(nsk, skb); | |
7716682c | 1700 | goto discard_and_relse; |
079096f1 | 1701 | } else { |
7716682c | 1702 | sock_put(sk); |
079096f1 ED |
1703 | return 0; |
1704 | } | |
1705 | } | |
790eb673 ED |
1706 | |
1707 | if (static_branch_unlikely(&ip6_min_hopcount)) { | |
1708 | /* min_hopcount can be changed concurrently from do_ipv6_setsockopt() */ | |
2798e36d | 1709 | if (unlikely(hdr->hop_limit < READ_ONCE(tcp_inet6_sk(sk)->min_hopcount))) { |
790eb673 | 1710 | __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP); |
2798e36d | 1711 | drop_reason = SKB_DROP_REASON_TCP_MINTTL; |
790eb673 ED |
1712 | goto discard_and_relse; |
1713 | } | |
e802af9c SH |
1714 | } |
1715 | ||
c0e3154d MD |
1716 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { |
1717 | drop_reason = SKB_DROP_REASON_XFRM_POLICY; | |
1da177e4 | 1718 | goto discard_and_relse; |
c0e3154d | 1719 | } |
1da177e4 | 1720 | |
1330b6ef JK |
1721 | drop_reason = tcp_inbound_md5_hash(sk, skb, &hdr->saddr, &hdr->daddr, |
1722 | AF_INET6, dif, sdif); | |
1723 | if (drop_reason) | |
9ea88a15 | 1724 | goto discard_and_relse; |
9ea88a15 | 1725 | |
b0e214d2 MK |
1726 | nf_reset_ct(skb); |
1727 | ||
c0e3154d MD |
1728 | if (tcp_filter(sk, skb)) { |
1729 | drop_reason = SKB_DROP_REASON_SOCKET_FILTER; | |
1da177e4 | 1730 | goto discard_and_relse; |
c0e3154d | 1731 | } |
ac6e7800 ED |
1732 | th = (const struct tcphdr *)skb->data; |
1733 | hdr = ipv6_hdr(skb); | |
eeea10b8 | 1734 | tcp_v6_fill_cb(skb, hdr, th); |
1da177e4 LT |
1735 | |
1736 | skb->dev = NULL; | |
1737 | ||
e994b2f0 ED |
1738 | if (sk->sk_state == TCP_LISTEN) { |
1739 | ret = tcp_v6_do_rcv(sk, skb); | |
1740 | goto put_and_return; | |
1741 | } | |
1742 | ||
1743 | sk_incoming_cpu_update(sk); | |
1744 | ||
293b9c42 | 1745 | bh_lock_sock_nested(sk); |
a44d6eac | 1746 | tcp_segs_in(tcp_sk(sk), skb); |
1da177e4 LT |
1747 | ret = 0; |
1748 | if (!sock_owned_by_user(sk)) { | |
e7942d06 | 1749 | ret = tcp_v6_do_rcv(sk, skb); |
8b27dae5 | 1750 | } else { |
7a26dc9e | 1751 | if (tcp_add_backlog(sk, skb, &drop_reason)) |
8b27dae5 | 1752 | goto discard_and_relse; |
6b03a53a | 1753 | } |
1da177e4 | 1754 | bh_unlock_sock(sk); |
e994b2f0 | 1755 | put_and_return: |
3b24d854 ED |
1756 | if (refcounted) |
1757 | sock_put(sk); | |
1da177e4 LT |
1758 | return ret ? -1 : 0; |
1759 | ||
1760 | no_tcp_socket: | |
c0e3154d | 1761 | drop_reason = SKB_DROP_REASON_NO_SOCKET; |
1da177e4 LT |
1762 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) |
1763 | goto discard_it; | |
1764 | ||
2dc49d16 ND |
1765 | tcp_v6_fill_cb(skb, hdr, th); |
1766 | ||
12e25e10 | 1767 | if (tcp_checksum_complete(skb)) { |
6a5dc9e5 | 1768 | csum_error: |
c0e3154d | 1769 | drop_reason = SKB_DROP_REASON_TCP_CSUM; |
709c0314 | 1770 | trace_tcp_bad_csum(skb); |
90bbcc60 | 1771 | __TCP_INC_STATS(net, TCP_MIB_CSUMERRORS); |
1da177e4 | 1772 | bad_packet: |
90bbcc60 | 1773 | __TCP_INC_STATS(net, TCP_MIB_INERRS); |
1da177e4 | 1774 | } else { |
cfb6eeb4 | 1775 | tcp_v6_send_reset(NULL, skb); |
1da177e4 LT |
1776 | } |
1777 | ||
1778 | discard_it: | |
dc776924 | 1779 | SKB_DR_OR(drop_reason, NOT_SPECIFIED); |
c0e3154d | 1780 | kfree_skb_reason(skb, drop_reason); |
1da177e4 LT |
1781 | return 0; |
1782 | ||
1783 | discard_and_relse: | |
532182cd | 1784 | sk_drops_add(sk, skb); |
3b24d854 ED |
1785 | if (refcounted) |
1786 | sock_put(sk); | |
1da177e4 LT |
1787 | goto discard_it; |
1788 | ||
1789 | do_time_wait: | |
1790 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
c0e3154d | 1791 | drop_reason = SKB_DROP_REASON_XFRM_POLICY; |
9469c7b4 | 1792 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1793 | goto discard_it; |
1794 | } | |
1795 | ||
2dc49d16 ND |
1796 | tcp_v6_fill_cb(skb, hdr, th); |
1797 | ||
6a5dc9e5 ED |
1798 | if (tcp_checksum_complete(skb)) { |
1799 | inet_twsk_put(inet_twsk(sk)); | |
1800 | goto csum_error; | |
1da177e4 LT |
1801 | } |
1802 | ||
9469c7b4 | 1803 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 LT |
1804 | case TCP_TW_SYN: |
1805 | { | |
1806 | struct sock *sk2; | |
1807 | ||
4461568a | 1808 | sk2 = inet6_lookup_listener(net, net->ipv4.tcp_death_row.hashinfo, |
a583636a | 1809 | skb, __tcp_hdrlen(th), |
5ba24953 | 1810 | &ipv6_hdr(skb)->saddr, th->source, |
0660e03f | 1811 | &ipv6_hdr(skb)->daddr, |
24b711ed DA |
1812 | ntohs(th->dest), |
1813 | tcp_v6_iif_l3_slave(skb), | |
4297a0ef | 1814 | sdif); |
53b24b8f | 1815 | if (sk2) { |
295ff7ed | 1816 | struct inet_timewait_sock *tw = inet_twsk(sk); |
dbe7faa4 | 1817 | inet_twsk_deschedule_put(tw); |
1da177e4 | 1818 | sk = sk2; |
4ad19de8 | 1819 | tcp_v6_restore_cb(skb); |
3b24d854 | 1820 | refcounted = false; |
1da177e4 LT |
1821 | goto process; |
1822 | } | |
1da177e4 | 1823 | } |
275757e6 | 1824 | /* to ACK */ |
a8eceea8 | 1825 | fallthrough; |
1da177e4 LT |
1826 | case TCP_TW_ACK: |
1827 | tcp_v6_timewait_ack(sk, skb); | |
1828 | break; | |
1829 | case TCP_TW_RST: | |
271c3b9b FW |
1830 | tcp_v6_send_reset(sk, skb); |
1831 | inet_twsk_deschedule_put(inet_twsk(sk)); | |
1832 | goto discard_it; | |
4aa956d8 WY |
1833 | case TCP_TW_SUCCESS: |
1834 | ; | |
1da177e4 LT |
1835 | } |
1836 | goto discard_it; | |
1837 | } | |
1838 | ||
11052589 | 1839 | void tcp_v6_early_demux(struct sk_buff *skb) |
c7109986 | 1840 | { |
4461568a | 1841 | struct net *net = dev_net(skb->dev); |
c7109986 ED |
1842 | const struct ipv6hdr *hdr; |
1843 | const struct tcphdr *th; | |
1844 | struct sock *sk; | |
1845 | ||
1846 | if (skb->pkt_type != PACKET_HOST) | |
1847 | return; | |
1848 | ||
1849 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) | |
1850 | return; | |
1851 | ||
1852 | hdr = ipv6_hdr(skb); | |
1853 | th = tcp_hdr(skb); | |
1854 | ||
1855 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1856 | return; | |
1857 | ||
870c3151 | 1858 | /* Note : We use inet6_iif() here, not tcp_v6_iif() */ |
4461568a | 1859 | sk = __inet6_lookup_established(net, net->ipv4.tcp_death_row.hashinfo, |
c7109986 ED |
1860 | &hdr->saddr, th->source, |
1861 | &hdr->daddr, ntohs(th->dest), | |
4297a0ef | 1862 | inet6_iif(skb), inet6_sdif(skb)); |
c7109986 ED |
1863 | if (sk) { |
1864 | skb->sk = sk; | |
1865 | skb->destructor = sock_edemux; | |
f7e4eb03 | 1866 | if (sk_fullsock(sk)) { |
8f905c0e | 1867 | struct dst_entry *dst = rcu_dereference(sk->sk_rx_dst); |
f3f12135 | 1868 | |
c7109986 | 1869 | if (dst) |
ef57c161 | 1870 | dst = dst_check(dst, sk->sk_rx_dst_cookie); |
c7109986 | 1871 | if (dst && |
0c0a5ef8 | 1872 | sk->sk_rx_dst_ifindex == skb->skb_iif) |
c7109986 ED |
1873 | skb_dst_set_noref(skb, dst); |
1874 | } | |
1875 | } | |
1876 | } | |
1877 | ||
ccb7c410 DM |
1878 | static struct timewait_sock_ops tcp6_timewait_sock_ops = { |
1879 | .twsk_obj_size = sizeof(struct tcp6_timewait_sock), | |
1880 | .twsk_unique = tcp_twsk_unique, | |
4aa956d8 | 1881 | .twsk_destructor = tcp_twsk_destructor, |
ccb7c410 DM |
1882 | }; |
1883 | ||
dd2e0b86 ED |
1884 | INDIRECT_CALLABLE_SCOPE void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb) |
1885 | { | |
37354402 | 1886 | __tcp_v6_send_check(skb, &sk->sk_v6_rcv_saddr, &sk->sk_v6_daddr); |
dd2e0b86 ED |
1887 | } |
1888 | ||
35b2c321 | 1889 | const struct inet_connection_sock_af_ops ipv6_specific = { |
543d9cfe ACM |
1890 | .queue_xmit = inet6_csk_xmit, |
1891 | .send_check = tcp_v6_send_check, | |
1892 | .rebuild_header = inet6_sk_rebuild_header, | |
5d299f3d | 1893 | .sk_rx_dst_set = inet6_sk_rx_dst_set, |
543d9cfe ACM |
1894 | .conn_request = tcp_v6_conn_request, |
1895 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
543d9cfe | 1896 | .net_header_len = sizeof(struct ipv6hdr), |
67469601 | 1897 | .net_frag_header_len = sizeof(struct frag_hdr), |
543d9cfe ACM |
1898 | .setsockopt = ipv6_setsockopt, |
1899 | .getsockopt = ipv6_getsockopt, | |
1900 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1901 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
4fab9071 | 1902 | .mtu_reduced = tcp_v6_mtu_reduced, |
1da177e4 LT |
1903 | }; |
1904 | ||
cfb6eeb4 | 1905 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1906 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = { |
cfb6eeb4 | 1907 | .md5_lookup = tcp_v6_md5_lookup, |
49a72dfb | 1908 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 1909 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1910 | }; |
a928630a | 1911 | #endif |
cfb6eeb4 | 1912 | |
1da177e4 LT |
1913 | /* |
1914 | * TCP over IPv4 via INET6 API | |
1915 | */ | |
3b401a81 | 1916 | static const struct inet_connection_sock_af_ops ipv6_mapped = { |
543d9cfe ACM |
1917 | .queue_xmit = ip_queue_xmit, |
1918 | .send_check = tcp_v4_send_check, | |
1919 | .rebuild_header = inet_sk_rebuild_header, | |
63d02d15 | 1920 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1921 | .conn_request = tcp_v6_conn_request, |
1922 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
543d9cfe ACM |
1923 | .net_header_len = sizeof(struct iphdr), |
1924 | .setsockopt = ipv6_setsockopt, | |
1925 | .getsockopt = ipv6_getsockopt, | |
1926 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1927 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
4fab9071 | 1928 | .mtu_reduced = tcp_v4_mtu_reduced, |
1da177e4 LT |
1929 | }; |
1930 | ||
cfb6eeb4 | 1931 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1932 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = { |
cfb6eeb4 | 1933 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1934 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1935 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1936 | }; |
a928630a | 1937 | #endif |
cfb6eeb4 | 1938 | |
1da177e4 LT |
1939 | /* NOTE: A lot of things set to zero explicitly by call to |
1940 | * sk_alloc() so need not be done here. | |
1941 | */ | |
1942 | static int tcp_v6_init_sock(struct sock *sk) | |
1943 | { | |
6687e988 | 1944 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1945 | |
900f65d3 | 1946 | tcp_init_sock(sk); |
1da177e4 | 1947 | |
8292a17a | 1948 | icsk->icsk_af_ops = &ipv6_specific; |
1da177e4 | 1949 | |
cfb6eeb4 | 1950 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1951 | tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific; |
cfb6eeb4 YH |
1952 | #endif |
1953 | ||
1da177e4 LT |
1954 | return 0; |
1955 | } | |
1956 | ||
952a10be | 1957 | #ifdef CONFIG_PROC_FS |
1da177e4 | 1958 | /* Proc filesystem TCPv6 sock list dumping. */ |
1ab1457c | 1959 | static void get_openreq6(struct seq_file *seq, |
aa3a0c8c | 1960 | const struct request_sock *req, int i) |
1da177e4 | 1961 | { |
fa76ce73 | 1962 | long ttd = req->rsk_timer.expires - jiffies; |
634fb979 ED |
1963 | const struct in6_addr *src = &inet_rsk(req)->ir_v6_loc_addr; |
1964 | const struct in6_addr *dest = &inet_rsk(req)->ir_v6_rmt_addr; | |
1da177e4 LT |
1965 | |
1966 | if (ttd < 0) | |
1967 | ttd = 0; | |
1968 | ||
1da177e4 LT |
1969 | seq_printf(seq, |
1970 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
d14c5ab6 | 1971 | "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n", |
1da177e4 LT |
1972 | i, |
1973 | src->s6_addr32[0], src->s6_addr32[1], | |
1974 | src->s6_addr32[2], src->s6_addr32[3], | |
b44084c2 | 1975 | inet_rsk(req)->ir_num, |
1da177e4 LT |
1976 | dest->s6_addr32[0], dest->s6_addr32[1], |
1977 | dest->s6_addr32[2], dest->s6_addr32[3], | |
634fb979 | 1978 | ntohs(inet_rsk(req)->ir_rmt_port), |
1da177e4 | 1979 | TCP_SYN_RECV, |
4c99aa40 | 1980 | 0, 0, /* could print option size, but that is af dependent. */ |
1ab1457c YH |
1981 | 1, /* timers active (only the expire timer) */ |
1982 | jiffies_to_clock_t(ttd), | |
e6c022a4 | 1983 | req->num_timeout, |
aa3a0c8c ED |
1984 | from_kuid_munged(seq_user_ns(seq), |
1985 | sock_i_uid(req->rsk_listener)), | |
1ab1457c | 1986 | 0, /* non standard timer */ |
1da177e4 LT |
1987 | 0, /* open_requests have no inode */ |
1988 | 0, req); | |
1989 | } | |
1990 | ||
1991 | static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i) | |
1992 | { | |
b71d1d42 | 1993 | const struct in6_addr *dest, *src; |
1da177e4 LT |
1994 | __u16 destp, srcp; |
1995 | int timer_active; | |
1996 | unsigned long timer_expires; | |
cf533ea5 ED |
1997 | const struct inet_sock *inet = inet_sk(sp); |
1998 | const struct tcp_sock *tp = tcp_sk(sp); | |
463c84b9 | 1999 | const struct inet_connection_sock *icsk = inet_csk(sp); |
0536fcc0 | 2000 | const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq; |
00fd38d9 ED |
2001 | int rx_queue; |
2002 | int state; | |
1da177e4 | 2003 | |
efe4208f ED |
2004 | dest = &sp->sk_v6_daddr; |
2005 | src = &sp->sk_v6_rcv_saddr; | |
c720c7e8 ED |
2006 | destp = ntohs(inet->inet_dport); |
2007 | srcp = ntohs(inet->inet_sport); | |
463c84b9 | 2008 | |
ce3cf4ec | 2009 | if (icsk->icsk_pending == ICSK_TIME_RETRANS || |
57dde7f7 | 2010 | icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || |
ce3cf4ec | 2011 | icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { |
1da177e4 | 2012 | timer_active = 1; |
463c84b9 ACM |
2013 | timer_expires = icsk->icsk_timeout; |
2014 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 2015 | timer_active = 4; |
463c84b9 | 2016 | timer_expires = icsk->icsk_timeout; |
1da177e4 LT |
2017 | } else if (timer_pending(&sp->sk_timer)) { |
2018 | timer_active = 2; | |
2019 | timer_expires = sp->sk_timer.expires; | |
2020 | } else { | |
2021 | timer_active = 0; | |
2022 | timer_expires = jiffies; | |
2023 | } | |
2024 | ||
986ffdfd | 2025 | state = inet_sk_state_load(sp); |
00fd38d9 | 2026 | if (state == TCP_LISTEN) |
288efe86 | 2027 | rx_queue = READ_ONCE(sp->sk_ack_backlog); |
00fd38d9 ED |
2028 | else |
2029 | /* Because we don't lock the socket, | |
2030 | * we might find a transient negative value. | |
2031 | */ | |
dba7d9b8 | 2032 | rx_queue = max_t(int, READ_ONCE(tp->rcv_nxt) - |
7db48e98 | 2033 | READ_ONCE(tp->copied_seq), 0); |
00fd38d9 | 2034 | |
1da177e4 LT |
2035 | seq_printf(seq, |
2036 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
d14c5ab6 | 2037 | "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n", |
1da177e4 LT |
2038 | i, |
2039 | src->s6_addr32[0], src->s6_addr32[1], | |
2040 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
2041 | dest->s6_addr32[0], dest->s6_addr32[1], | |
2042 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
00fd38d9 | 2043 | state, |
0f317464 | 2044 | READ_ONCE(tp->write_seq) - tp->snd_una, |
00fd38d9 | 2045 | rx_queue, |
1da177e4 | 2046 | timer_active, |
a399a805 | 2047 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 2048 | icsk->icsk_retransmits, |
a7cb5a49 | 2049 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), |
6687e988 | 2050 | icsk->icsk_probes_out, |
1da177e4 | 2051 | sock_i_ino(sp), |
41c6d650 | 2052 | refcount_read(&sp->sk_refcnt), sp, |
7be87351 SH |
2053 | jiffies_to_clock_t(icsk->icsk_rto), |
2054 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
31954cd8 | 2055 | (icsk->icsk_ack.quick << 1) | inet_csk_in_pingpong_mode(sp), |
40570375 | 2056 | tcp_snd_cwnd(tp), |
00fd38d9 | 2057 | state == TCP_LISTEN ? |
0536fcc0 | 2058 | fastopenq->max_qlen : |
0a672f74 | 2059 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh) |
1da177e4 LT |
2060 | ); |
2061 | } | |
2062 | ||
1ab1457c | 2063 | static void get_timewait6_sock(struct seq_file *seq, |
8feaf0c0 | 2064 | struct inet_timewait_sock *tw, int i) |
1da177e4 | 2065 | { |
789f558c | 2066 | long delta = tw->tw_timer.expires - jiffies; |
b71d1d42 | 2067 | const struct in6_addr *dest, *src; |
1da177e4 | 2068 | __u16 destp, srcp; |
1da177e4 | 2069 | |
efe4208f ED |
2070 | dest = &tw->tw_v6_daddr; |
2071 | src = &tw->tw_v6_rcv_saddr; | |
1da177e4 LT |
2072 | destp = ntohs(tw->tw_dport); |
2073 | srcp = ntohs(tw->tw_sport); | |
2074 | ||
2075 | seq_printf(seq, | |
2076 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 2077 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
2078 | i, |
2079 | src->s6_addr32[0], src->s6_addr32[1], | |
2080 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
2081 | dest->s6_addr32[0], dest->s6_addr32[1], | |
2082 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
2083 | tw->tw_substate, 0, 0, | |
a399a805 | 2084 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
41c6d650 | 2085 | refcount_read(&tw->tw_refcnt), tw); |
1da177e4 LT |
2086 | } |
2087 | ||
1da177e4 LT |
2088 | static int tcp6_seq_show(struct seq_file *seq, void *v) |
2089 | { | |
2090 | struct tcp_iter_state *st; | |
05dbc7b5 | 2091 | struct sock *sk = v; |
1da177e4 LT |
2092 | |
2093 | if (v == SEQ_START_TOKEN) { | |
2094 | seq_puts(seq, | |
2095 | " sl " | |
2096 | "local_address " | |
2097 | "remote_address " | |
2098 | "st tx_queue rx_queue tr tm->when retrnsmt" | |
2099 | " uid timeout inode\n"); | |
2100 | goto out; | |
2101 | } | |
2102 | st = seq->private; | |
2103 | ||
079096f1 ED |
2104 | if (sk->sk_state == TCP_TIME_WAIT) |
2105 | get_timewait6_sock(seq, v, st->num); | |
2106 | else if (sk->sk_state == TCP_NEW_SYN_RECV) | |
aa3a0c8c | 2107 | get_openreq6(seq, v, st->num); |
079096f1 ED |
2108 | else |
2109 | get_tcp6_sock(seq, v, st->num); | |
1da177e4 LT |
2110 | out: |
2111 | return 0; | |
2112 | } | |
2113 | ||
37d849bb CH |
2114 | static const struct seq_operations tcp6_seq_ops = { |
2115 | .show = tcp6_seq_show, | |
2116 | .start = tcp_seq_start, | |
2117 | .next = tcp_seq_next, | |
2118 | .stop = tcp_seq_stop, | |
2119 | }; | |
2120 | ||
1da177e4 | 2121 | static struct tcp_seq_afinfo tcp6_seq_afinfo = { |
1da177e4 | 2122 | .family = AF_INET6, |
1da177e4 LT |
2123 | }; |
2124 | ||
2c8c1e72 | 2125 | int __net_init tcp6_proc_init(struct net *net) |
1da177e4 | 2126 | { |
c3506372 CH |
2127 | if (!proc_create_net_data("tcp6", 0444, net->proc_net, &tcp6_seq_ops, |
2128 | sizeof(struct tcp_iter_state), &tcp6_seq_afinfo)) | |
37d849bb CH |
2129 | return -ENOMEM; |
2130 | return 0; | |
1da177e4 LT |
2131 | } |
2132 | ||
6f8b13bc | 2133 | void tcp6_proc_exit(struct net *net) |
1da177e4 | 2134 | { |
37d849bb | 2135 | remove_proc_entry("tcp6", net->proc_net); |
1da177e4 LT |
2136 | } |
2137 | #endif | |
2138 | ||
2139 | struct proto tcpv6_prot = { | |
2140 | .name = "TCPv6", | |
2141 | .owner = THIS_MODULE, | |
2142 | .close = tcp_close, | |
d74bad4e | 2143 | .pre_connect = tcp_v6_pre_connect, |
1da177e4 LT |
2144 | .connect = tcp_v6_connect, |
2145 | .disconnect = tcp_disconnect, | |
463c84b9 | 2146 | .accept = inet_csk_accept, |
1da177e4 LT |
2147 | .ioctl = tcp_ioctl, |
2148 | .init = tcp_v6_init_sock, | |
b5fc2923 | 2149 | .destroy = tcp_v4_destroy_sock, |
1da177e4 LT |
2150 | .shutdown = tcp_shutdown, |
2151 | .setsockopt = tcp_setsockopt, | |
2152 | .getsockopt = tcp_getsockopt, | |
9cacf81f | 2153 | .bpf_bypass_getsockopt = tcp_bpf_bypass_getsockopt, |
4b9d07a4 | 2154 | .keepalive = tcp_set_keepalive, |
1da177e4 | 2155 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2156 | .sendmsg = tcp_sendmsg, |
2157 | .sendpage = tcp_sendpage, | |
1da177e4 | 2158 | .backlog_rcv = tcp_v6_do_rcv, |
46d3ceab | 2159 | .release_cb = tcp_release_cb, |
496611d7 | 2160 | .hash = inet6_hash, |
ab1e0a13 ACM |
2161 | .unhash = inet_unhash, |
2162 | .get_port = inet_csk_get_port, | |
91a760b2 | 2163 | .put_port = inet_put_port, |
8a59f9d1 CW |
2164 | #ifdef CONFIG_BPF_SYSCALL |
2165 | .psock_update_sk_prot = tcp_bpf_update_proto, | |
2166 | #endif | |
1da177e4 | 2167 | .enter_memory_pressure = tcp_enter_memory_pressure, |
06044751 | 2168 | .leave_memory_pressure = tcp_leave_memory_pressure, |
c9bee3b7 | 2169 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 | 2170 | .sockets_allocated = &tcp_sockets_allocated, |
0defbb0a | 2171 | |
1da177e4 | 2172 | .memory_allocated = &tcp_memory_allocated, |
0defbb0a ED |
2173 | .per_cpu_fw_alloc = &tcp_memory_per_cpu_fw_alloc, |
2174 | ||
1da177e4 | 2175 | .memory_pressure = &tcp_memory_pressure, |
0a5578cf | 2176 | .orphan_count = &tcp_orphan_count, |
a4fe34bf | 2177 | .sysctl_mem = sysctl_tcp_mem, |
356d1833 ED |
2178 | .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_tcp_wmem), |
2179 | .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_tcp_rmem), | |
1da177e4 LT |
2180 | .max_header = MAX_TCP_HEADER, |
2181 | .obj_size = sizeof(struct tcp6_sock), | |
5f0d5a3a | 2182 | .slab_flags = SLAB_TYPESAFE_BY_RCU, |
6d6ee43e | 2183 | .twsk_prot = &tcp6_timewait_sock_ops, |
60236fdd | 2184 | .rsk_prot = &tcp6_request_sock_ops, |
429e42c1 | 2185 | .h.hashinfo = NULL, |
7ba42910 | 2186 | .no_autobind = true, |
c1e64e29 | 2187 | .diag_destroy = tcp_abort, |
1da177e4 | 2188 | }; |
6abde0b2 | 2189 | EXPORT_SYMBOL_GPL(tcpv6_prot); |
1da177e4 | 2190 | |
11052589 | 2191 | static const struct inet6_protocol tcpv6_protocol = { |
1da177e4 LT |
2192 | .handler = tcp_v6_rcv, |
2193 | .err_handler = tcp_v6_err, | |
2194 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
2195 | }; | |
2196 | ||
1da177e4 LT |
2197 | static struct inet_protosw tcpv6_protosw = { |
2198 | .type = SOCK_STREAM, | |
2199 | .protocol = IPPROTO_TCP, | |
2200 | .prot = &tcpv6_prot, | |
2201 | .ops = &inet6_stream_ops, | |
d83d8461 ACM |
2202 | .flags = INET_PROTOSW_PERMANENT | |
2203 | INET_PROTOSW_ICSK, | |
1da177e4 LT |
2204 | }; |
2205 | ||
2c8c1e72 | 2206 | static int __net_init tcpv6_net_init(struct net *net) |
93ec926b | 2207 | { |
5677242f DL |
2208 | return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6, |
2209 | SOCK_RAW, IPPROTO_TCP, net); | |
93ec926b DL |
2210 | } |
2211 | ||
2c8c1e72 | 2212 | static void __net_exit tcpv6_net_exit(struct net *net) |
93ec926b | 2213 | { |
5677242f | 2214 | inet_ctl_sock_destroy(net->ipv6.tcp_sk); |
b099ce26 EB |
2215 | } |
2216 | ||
04c494e6 ED |
2217 | static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list) |
2218 | { | |
edc12f03 | 2219 | tcp_twsk_purge(net_exit_list, AF_INET6); |
04c494e6 ED |
2220 | } |
2221 | ||
93ec926b | 2222 | static struct pernet_operations tcpv6_net_ops = { |
b099ce26 EB |
2223 | .init = tcpv6_net_init, |
2224 | .exit = tcpv6_net_exit, | |
04c494e6 | 2225 | .exit_batch = tcpv6_net_exit_batch, |
93ec926b DL |
2226 | }; |
2227 | ||
7f4e4868 | 2228 | int __init tcpv6_init(void) |
1da177e4 | 2229 | { |
7f4e4868 DL |
2230 | int ret; |
2231 | ||
3336288a VY |
2232 | ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP); |
2233 | if (ret) | |
c6b641a4 | 2234 | goto out; |
3336288a | 2235 | |
1da177e4 | 2236 | /* register inet6 protocol */ |
7f4e4868 DL |
2237 | ret = inet6_register_protosw(&tcpv6_protosw); |
2238 | if (ret) | |
2239 | goto out_tcpv6_protocol; | |
2240 | ||
93ec926b | 2241 | ret = register_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2242 | if (ret) |
2243 | goto out_tcpv6_protosw; | |
f870fa0b MM |
2244 | |
2245 | ret = mptcpv6_init(); | |
2246 | if (ret) | |
2247 | goto out_tcpv6_pernet_subsys; | |
2248 | ||
7f4e4868 DL |
2249 | out: |
2250 | return ret; | |
ae0f7d5f | 2251 | |
f870fa0b MM |
2252 | out_tcpv6_pernet_subsys: |
2253 | unregister_pernet_subsys(&tcpv6_net_ops); | |
7f4e4868 DL |
2254 | out_tcpv6_protosw: |
2255 | inet6_unregister_protosw(&tcpv6_protosw); | |
3336288a VY |
2256 | out_tcpv6_protocol: |
2257 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
7f4e4868 DL |
2258 | goto out; |
2259 | } | |
2260 | ||
09f7709f | 2261 | void tcpv6_exit(void) |
7f4e4868 | 2262 | { |
93ec926b | 2263 | unregister_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2264 | inet6_unregister_protosw(&tcpv6_protosw); |
2265 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1da177e4 | 2266 | } |