]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - net/ipv6/udp.c
net: use dst_confirm_neigh for UDP, RAW, ICMP, L2TP
[mirror_ubuntu-zesty-kernel.git] / net / ipv6 / udp.c
CommitLineData
1da177e4
LT
1/*
2 * UDP over IPv6
1ab1457c 3 * Linux INET6 implementation
1da177e4
LT
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT
7 *
8 * Based on linux/ipv4/udp.c
9 *
1da177e4
LT
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
14 * a single port at the same time.
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 */
23
1da177e4
LT
24#include <linux/errno.h>
25#include <linux/types.h>
26#include <linux/socket.h>
27#include <linux/sockios.h>
1da177e4
LT
28#include <linux/net.h>
29#include <linux/in6.h>
30#include <linux/netdevice.h>
31#include <linux/if_arp.h>
32#include <linux/ipv6.h>
33#include <linux/icmpv6.h>
34#include <linux/init.h>
1781f7f5 35#include <linux/module.h>
3305b80c 36#include <linux/skbuff.h>
5a0e3ad6 37#include <linux/slab.h>
7c0f6ba6 38#include <linux/uaccess.h>
1da177e4 39
496611d7 40#include <net/addrconf.h>
1da177e4
LT
41#include <net/ndisc.h>
42#include <net/protocol.h>
43#include <net/transp_v6.h>
44#include <net/ip6_route.h>
1da177e4 45#include <net/raw.h>
c752f073 46#include <net/tcp_states.h>
1da177e4
LT
47#include <net/ip6_checksum.h>
48#include <net/xfrm.h>
72289b96 49#include <net/inet6_hashtables.h>
076bb0c8 50#include <net/busy_poll.h>
e32ea7e7 51#include <net/sock_reuseport.h>
1da177e4
LT
52
53#include <linux/proc_fs.h>
54#include <linux/seq_file.h>
22911fc5 55#include <trace/events/skb.h>
ba4e58ec 56#include "udp_impl.h"
1da177e4 57
6eada011
ED
58static u32 udp6_ehashfn(const struct net *net,
59 const struct in6_addr *laddr,
60 const u16 lport,
61 const struct in6_addr *faddr,
62 const __be16 fport)
b50026b5 63{
1bbdceef
HFS
64 static u32 udp6_ehash_secret __read_mostly;
65 static u32 udp_ipv6_hash_secret __read_mostly;
66
67 u32 lhash, fhash;
68
69 net_get_random_once(&udp6_ehash_secret,
70 sizeof(udp6_ehash_secret));
71 net_get_random_once(&udp_ipv6_hash_secret,
72 sizeof(udp_ipv6_hash_secret));
73
74 lhash = (__force u32)laddr->s6_addr32[3];
75 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
76
b50026b5 77 return __inet6_ehashfn(lhash, lport, fhash, fport,
1bbdceef 78 udp_ipv6_hash_secret + net_hash_mix(net));
b50026b5
HFS
79}
80
6eada011
ED
81static u32 udp6_portaddr_hash(const struct net *net,
82 const struct in6_addr *addr6,
83 unsigned int port)
d4cada4a
ED
84{
85 unsigned int hash, mix = net_hash_mix(net);
86
87 if (ipv6_addr_any(addr6))
88 hash = jhash_1word(0, mix);
856540ee 89 else if (ipv6_addr_v4mapped(addr6))
0eae88f3 90 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
d4cada4a 91 else
0eae88f3 92 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
d4cada4a
ED
93
94 return hash ^ port;
95}
96
6ba5a3c5 97int udp_v6_get_port(struct sock *sk, unsigned short snum)
1da177e4 98{
30fff923
ED
99 unsigned int hash2_nulladdr =
100 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
9a52e97e 101 unsigned int hash2_partial =
efe4208f 102 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
30fff923 103
d4cada4a 104 /* precompute partial secondary hash */
30fff923
ED
105 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
106 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
1da177e4
LT
107}
108
719f8358
ED
109static void udp_v6_rehash(struct sock *sk)
110{
111 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
efe4208f 112 &sk->sk_v6_rcv_saddr,
719f8358
ED
113 inet_sk(sk)->inet_num);
114
115 udp_lib_rehash(sk, new_hash);
116}
117
d1e37288
SX
118static int compute_score(struct sock *sk, struct net *net,
119 const struct in6_addr *saddr, __be16 sport,
120 const struct in6_addr *daddr, unsigned short hnum,
121 int dif)
645ca708 122{
60c04aec
JP
123 int score;
124 struct inet_sock *inet;
125
126 if (!net_eq(sock_net(sk), net) ||
127 udp_sk(sk)->udp_port_hash != hnum ||
128 sk->sk_family != PF_INET6)
129 return -1;
130
131 score = 0;
132 inet = inet_sk(sk);
133
134 if (inet->inet_dport) {
135 if (inet->inet_dport != sport)
136 return -1;
137 score++;
138 }
139
140 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
141 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
142 return -1;
143 score++;
144 }
145
146 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
147 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
148 return -1;
149 score++;
150 }
151
152 if (sk->sk_bound_dev_if) {
153 if (sk->sk_bound_dev_if != dif)
154 return -1;
155 score++;
645ca708 156 }
60c04aec 157
70da268b
ED
158 if (sk->sk_incoming_cpu == raw_smp_processor_id())
159 score++;
160
645ca708
ED
161 return score;
162}
163
d1e37288 164/* called with rcu_read_lock() */
fddc17de
ED
165static struct sock *udp6_lib_lookup2(struct net *net,
166 const struct in6_addr *saddr, __be16 sport,
167 const struct in6_addr *daddr, unsigned int hnum, int dif,
d1e37288 168 struct udp_hslot *hslot2,
1134158b 169 struct sk_buff *skb)
fddc17de
ED
170{
171 struct sock *sk, *result;
72289b96
TH
172 int score, badness, matches = 0, reuseport = 0;
173 u32 hash = 0;
fddc17de 174
fddc17de
ED
175 result = NULL;
176 badness = -1;
ca065d0c 177 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
d1e37288 178 score = compute_score(sk, net, saddr, sport,
fddc17de
ED
179 daddr, hnum, dif);
180 if (score > badness) {
72289b96
TH
181 reuseport = sk->sk_reuseport;
182 if (reuseport) {
b50026b5
HFS
183 hash = udp6_ehashfn(net, daddr, hnum,
184 saddr, sport);
ed0dfffd 185
ca065d0c 186 result = reuseport_select_sock(sk, hash, skb,
ed0dfffd 187 sizeof(struct udphdr));
ca065d0c
ED
188 if (result)
189 return result;
72289b96 190 matches = 1;
70da268b 191 }
ca065d0c
ED
192 result = sk;
193 badness = score;
72289b96
TH
194 } else if (score == badness && reuseport) {
195 matches++;
8fc54f68 196 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
197 result = sk;
198 hash = next_pseudo_random32(hash);
fddc17de
ED
199 }
200 }
fddc17de
ED
201 return result;
202}
203
ca065d0c 204/* rcu_read_lock() must be held */
fce82338 205struct sock *__udp6_lib_lookup(struct net *net,
88440ae7
BS
206 const struct in6_addr *saddr, __be16 sport,
207 const struct in6_addr *daddr, __be16 dport,
538950a1
CG
208 int dif, struct udp_table *udptable,
209 struct sk_buff *skb)
1da177e4 210{
271b72c7 211 struct sock *sk, *result;
1da177e4 212 unsigned short hnum = ntohs(dport);
fddc17de
ED
213 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
214 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
72289b96
TH
215 int score, badness, matches = 0, reuseport = 0;
216 u32 hash = 0;
645ca708 217
fddc17de
ED
218 if (hslot->count > 10) {
219 hash2 = udp6_portaddr_hash(net, daddr, hnum);
220 slot2 = hash2 & udptable->mask;
221 hslot2 = &udptable->hash2[slot2];
222 if (hslot->count < hslot2->count)
223 goto begin;
224
225 result = udp6_lib_lookup2(net, saddr, sport,
226 daddr, hnum, dif,
d1e37288 227 hslot2, skb);
fddc17de 228 if (!result) {
d1e37288 229 unsigned int old_slot2 = slot2;
fddc17de
ED
230 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
231 slot2 = hash2 & udptable->mask;
d1e37288
SX
232 /* avoid searching the same slot again. */
233 if (unlikely(slot2 == old_slot2))
234 return result;
235
fddc17de
ED
236 hslot2 = &udptable->hash2[slot2];
237 if (hslot->count < hslot2->count)
238 goto begin;
239
1223c67c 240 result = udp6_lib_lookup2(net, saddr, sport,
d1e37288
SX
241 daddr, hnum, dif,
242 hslot2, skb);
fddc17de 243 }
fddc17de
ED
244 return result;
245 }
271b72c7
ED
246begin:
247 result = NULL;
248 badness = -1;
ca065d0c 249 sk_for_each_rcu(sk, &hslot->head) {
d1e37288 250 score = compute_score(sk, net, saddr, sport, daddr, hnum, dif);
645ca708 251 if (score > badness) {
72289b96
TH
252 reuseport = sk->sk_reuseport;
253 if (reuseport) {
b50026b5
HFS
254 hash = udp6_ehashfn(net, daddr, hnum,
255 saddr, sport);
ca065d0c 256 result = reuseport_select_sock(sk, hash, skb,
538950a1 257 sizeof(struct udphdr));
ca065d0c
ED
258 if (result)
259 return result;
72289b96
TH
260 matches = 1;
261 }
ca065d0c
ED
262 result = sk;
263 badness = score;
72289b96
TH
264 } else if (score == badness && reuseport) {
265 matches++;
8fc54f68 266 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
267 result = sk;
268 hash = next_pseudo_random32(hash);
1da177e4
LT
269 }
270 }
1da177e4
LT
271 return result;
272}
fce82338 273EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
1da177e4 274
607c4aaf
KK
275static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
276 __be16 sport, __be16 dport,
645ca708 277 struct udp_table *udptable)
607c4aaf 278{
b71d1d42 279 const struct ipv6hdr *iph = ipv6_hdr(skb);
ed7cbbce 280 struct sock *sk;
607c4aaf 281
e5d08d71
IM
282 sk = skb_steal_sock(skb);
283 if (unlikely(sk))
23542618 284 return sk;
ed7cbbce 285 return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
adf30907 286 &iph->daddr, dport, inet6_iif(skb),
538950a1 287 udptable, skb);
607c4aaf
KK
288}
289
63058308
TH
290struct sock *udp6_lib_lookup_skb(struct sk_buff *skb,
291 __be16 sport, __be16 dport)
292{
293 const struct ipv6hdr *iph = ipv6_hdr(skb);
63058308 294
ed7cbbce 295 return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
63058308
TH
296 &iph->daddr, dport, inet6_iif(skb),
297 &udp_table, skb);
298}
299EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb);
300
ca065d0c
ED
301/* Must be called under rcu_read_lock().
302 * Does increment socket refcount.
303 */
304#if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
30f58158
AB
305 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY) || \
306 IS_ENABLED(CONFIG_NF_SOCKET_IPV6)
aa976fc0
BS
307struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
308 const struct in6_addr *daddr, __be16 dport, int dif)
309{
ca065d0c
ED
310 struct sock *sk;
311
312 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport,
313 dif, &udp_table, NULL);
314 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
315 sk = NULL;
316 return sk;
aa976fc0
BS
317}
318EXPORT_SYMBOL_GPL(udp6_lib_lookup);
ca065d0c 319#endif
aa976fc0 320
1da177e4 321/*
67ba4152
IM
322 * This should be easy, if there is something there we
323 * return it, otherwise we block.
1da177e4
LT
324 */
325
1b784140 326int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
1da177e4
LT
327 int noblock, int flags, int *addr_len)
328{
329 struct ipv6_pinfo *np = inet6_sk(sk);
330 struct inet_sock *inet = inet_sk(sk);
1ab1457c 331 struct sk_buff *skb;
59c2cdae 332 unsigned int ulen, copied;
627d2d6b 333 int peeked, peeking, off;
759e5d00
HX
334 int err;
335 int is_udplite = IS_UDPLITE(sk);
197c949e 336 bool checksum_valid = false;
f26ba175 337 int is_udp4;
1da177e4 338
1da177e4 339 if (flags & MSG_ERRQUEUE)
85fbaa75 340 return ipv6_recv_error(sk, msg, len, addr_len);
1da177e4 341
4b340ae2 342 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
85fbaa75 343 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
4b340ae2 344
1da177e4 345try_again:
627d2d6b 346 peeking = off = sk_peek_offset(sk, flags);
7c13f97f 347 skb = __skb_recv_udp(sk, flags, noblock, &peeked, &off, &err);
1da177e4 348 if (!skb)
627d2d6b 349 return err;
1da177e4 350
e6afc8ac 351 ulen = skb->len;
59c2cdae 352 copied = len;
627d2d6b 353 if (copied > ulen - off)
354 copied = ulen - off;
59c2cdae 355 else if (copied < ulen)
1ab1457c 356 msg->msg_flags |= MSG_TRUNC;
1da177e4 357
f26ba175
WY
358 is_udp4 = (skb->protocol == htons(ETH_P_IP));
359
ba4e58ec 360 /*
759e5d00
HX
361 * If checksum is needed at all, try to do it while copying the
362 * data. If the data is truncated, or if we only want a partial
363 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 364 */
ba4e58ec 365
d21dbdfe
ED
366 if (copied < ulen || peeking ||
367 (is_udplite && UDP_SKB_CB(skb)->partial_cov)) {
197c949e
ED
368 checksum_valid = !udp_lib_checksum_complete(skb);
369 if (!checksum_valid)
1da177e4 370 goto csum_copy_err;
ba4e58ec
GR
371 }
372
197c949e 373 if (checksum_valid || skb_csum_unnecessary(skb))
627d2d6b 374 err = skb_copy_datagram_msg(skb, off, msg, copied);
ba4e58ec 375 else {
627d2d6b 376 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
1da177e4
LT
377 if (err == -EINVAL)
378 goto csum_copy_err;
379 }
22911fc5 380 if (unlikely(err)) {
979402b1
ED
381 if (!peeked) {
382 atomic_inc(&sk->sk_drops);
383 if (is_udp4)
6aef70a8
ED
384 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
385 is_udplite);
979402b1 386 else
6aef70a8
ED
387 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
388 is_udplite);
979402b1 389 }
850cbadd 390 kfree_skb(skb);
627d2d6b 391 return err;
22911fc5 392 }
f26ba175
WY
393 if (!peeked) {
394 if (is_udp4)
6aef70a8
ED
395 UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
396 is_udplite);
f26ba175 397 else
6aef70a8
ED
398 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
399 is_udplite);
f26ba175 400 }
cb75994e 401
3b885787 402 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
403
404 /* Copy the address. */
405 if (msg->msg_name) {
342dfc30 406 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 407 sin6->sin6_family = AF_INET6;
4bedb452 408 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 409 sin6->sin6_flowinfo = 0;
1da177e4 410
842df073 411 if (is_udp4) {
b301e82c
BH
412 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
413 &sin6->sin6_addr);
842df073
HFS
414 sin6->sin6_scope_id = 0;
415 } else {
4e3fd7a0 416 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
417 sin6->sin6_scope_id =
418 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 419 inet6_iif(skb));
1da177e4 420 }
bceaa902 421 *addr_len = sizeof(*sin6);
1da177e4 422 }
4b261c75
HFS
423
424 if (np->rxopt.all)
425 ip6_datagram_recv_common_ctl(sk, msg, skb);
426
f26ba175 427 if (is_udp4) {
1da177e4 428 if (inet->cmsg_flags)
ad959036 429 ip_cmsg_recv_offset(msg, sk, skb,
10df8e61 430 sizeof(struct udphdr), off);
1da177e4
LT
431 } else {
432 if (np->rxopt.all)
4b261c75 433 ip6_datagram_recv_specific_ctl(sk, msg, skb);
1ab1457c 434 }
1da177e4 435
59c2cdae 436 err = copied;
1da177e4 437 if (flags & MSG_TRUNC)
759e5d00 438 err = ulen;
1da177e4 439
850cbadd 440 skb_consume_udp(sk, skb, peeking ? -err : err);
1da177e4
LT
441 return err;
442
443csum_copy_err:
69629464 444 if (!__sk_queue_drop_skb(sk, skb, flags, udp_skb_destructor)) {
6a5dc9e5 445 if (is_udp4) {
6aef70a8
ED
446 UDP_INC_STATS(sock_net(sk),
447 UDP_MIB_CSUMERRORS, is_udplite);
448 UDP_INC_STATS(sock_net(sk),
449 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 450 } else {
6aef70a8
ED
451 UDP6_INC_STATS(sock_net(sk),
452 UDP_MIB_CSUMERRORS, is_udplite);
453 UDP6_INC_STATS(sock_net(sk),
454 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 455 }
0856f939 456 }
850cbadd 457 kfree_skb(skb);
1da177e4 458
beb39db5
ED
459 /* starting over for a new packet, but check if we need to yield */
460 cond_resched();
9cfaa8de 461 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
462 goto try_again;
463}
464
ba4e58ec 465void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 466 u8 type, u8 code, int offset, __be32 info,
645ca708 467 struct udp_table *udptable)
1da177e4
LT
468{
469 struct ipv6_pinfo *np;
b71d1d42
ED
470 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
471 const struct in6_addr *saddr = &hdr->saddr;
472 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 473 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4 474 struct sock *sk;
e0d8c1b7 475 int harderr;
1da177e4 476 int err;
7304fe46 477 struct net *net = dev_net(skb->dev);
1da177e4 478
e32ea7e7 479 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
538950a1 480 inet6_iif(skb), udptable, skb);
63159f29 481 if (!sk) {
a16292a0
ED
482 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
483 ICMP6_MIB_INERRORS);
1da177e4 484 return;
7304fe46 485 }
1da177e4 486
e0d8c1b7
WW
487 harderr = icmpv6_err_convert(type, code, &err);
488 np = inet6_sk(sk);
489
93b36cf3
HFS
490 if (type == ICMPV6_PKT_TOOBIG) {
491 if (!ip6_sk_accept_pmtu(sk))
492 goto out;
81aded24 493 ip6_sk_update_pmtu(skb, sk, info);
e0d8c1b7
WW
494 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
495 harderr = 1;
93b36cf3 496 }
1a462d18 497 if (type == NDISC_REDIRECT) {
ec18d9a2 498 ip6_sk_redirect(skb, sk);
1a462d18
DJ
499 goto out;
500 }
81aded24 501
e0d8c1b7
WW
502 if (!np->recverr) {
503 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
504 goto out;
505 } else {
1da177e4 506 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
e0d8c1b7 507 }
b1faf566 508
1da177e4
LT
509 sk->sk_err = err;
510 sk->sk_error_report(sk);
511out:
ca065d0c 512 return;
1da177e4
LT
513}
514
30c7be26 515int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
f7ad74fe
BL
516{
517 int rc;
518
efe4208f 519 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 520 sock_rps_save_rxhash(sk, skb);
005ec974 521 sk_mark_napi_id(sk, skb);
2c8c56e1 522 sk_incoming_cpu_update(sk);
e68b6e50
ED
523 } else {
524 sk_mark_napi_id_once(sk, skb);
005ec974 525 }
f7ad74fe 526
850cbadd 527 rc = __udp_enqueue_schedule_skb(sk, skb);
f7ad74fe
BL
528 if (rc < 0) {
529 int is_udplite = IS_UDPLITE(sk);
530
531 /* Note that an ENOMEM error is charged twice */
532 if (rc == -ENOMEM)
e61da9e2 533 UDP6_INC_STATS(sock_net(sk),
02c22347 534 UDP_MIB_RCVBUFERRORS, is_udplite);
e61da9e2 535 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
f7ad74fe
BL
536 kfree_skb(skb);
537 return -1;
538 }
850cbadd 539
f7ad74fe
BL
540 return 0;
541}
542
ba4e58ec 543static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 544 struct inet6_skb_parm *opt, u8 type,
67ba4152 545 u8 code, int offset, __be32 info)
ba4e58ec 546{
645ca708 547 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
548}
549
d7f3f621
BL
550static struct static_key udpv6_encap_needed __read_mostly;
551void udpv6_encap_enable(void)
552{
553 if (!static_key_enabled(&udpv6_encap_needed))
554 static_key_slow_inc(&udpv6_encap_needed);
555}
556EXPORT_SYMBOL(udpv6_encap_enable);
557
f7ad74fe 558int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 559{
ba4e58ec 560 struct udp_sock *up = udp_sk(sk);
b2bf1e26 561 int is_udplite = IS_UDPLITE(sk);
a18135eb 562
ba4e58ec
GR
563 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
564 goto drop;
1da177e4 565
d7f3f621
BL
566 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
567 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
568
569 /*
570 * This is an encapsulation socket so pass the skb to
571 * the socket's udp_encap_rcv() hook. Otherwise, just
572 * fall through and pass this up the UDP socket.
573 * up->encap_rcv() returns the following value:
574 * =0 if skb was successfully passed to the encap
575 * handler or was discarded by it.
576 * >0 if skb should be passed on to UDP.
577 * <0 if skb should be resubmitted as proto -N
578 */
579
580 /* if we're overly short, let UDP handle it */
581 encap_rcv = ACCESS_ONCE(up->encap_rcv);
e5aed006 582 if (encap_rcv) {
d7f3f621
BL
583 int ret;
584
0a80966b
TH
585 /* Verify checksum before giving to encap */
586 if (udp_lib_checksum_complete(skb))
587 goto csum_error;
588
d7f3f621
BL
589 ret = encap_rcv(sk, skb);
590 if (ret <= 0) {
02c22347
ED
591 __UDP_INC_STATS(sock_net(sk),
592 UDP_MIB_INDATAGRAMS,
593 is_udplite);
d7f3f621
BL
594 return -ret;
595 }
596 }
597
598 /* FALLTHROUGH -- it's a UDP Packet */
599 }
600
ba4e58ec
GR
601 /*
602 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
603 */
b2bf1e26 604 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
605
606 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
607 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
608 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
609 goto drop;
610 }
611 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
612 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
613 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
614 goto drop;
615 }
1da177e4
LT
616 }
617
ce25d66a
ED
618 if (rcu_access_pointer(sk->sk_filter) &&
619 udp_lib_checksum_complete(skb))
620 goto csum_error;
621
ba66bbe5 622 if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
a6127697 623 goto drop;
ba4e58ec 624
e6afc8ac 625 udp_csum_pull_header(skb);
cb80ef46 626
d826eb14 627 skb_dst_drop(skb);
cb75994e 628
850cbadd 629 return __udpv6_queue_rcv_skb(sk, skb);
3e215c8d 630
6a5dc9e5 631csum_error:
02c22347 632 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 633drop:
02c22347 634 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 635 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
636 kfree_skb(skb);
637 return -1;
1da177e4
LT
638}
639
5cf3d461
DH
640static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
641 __be16 loc_port, const struct in6_addr *loc_addr,
642 __be16 rmt_port, const struct in6_addr *rmt_addr,
643 int dif, unsigned short hnum)
1da177e4 644{
5cf3d461 645 struct inet_sock *inet = inet_sk(sk);
1da177e4 646
5cf3d461
DH
647 if (!net_eq(sock_net(sk), net))
648 return false;
649
650 if (udp_sk(sk)->udp_port_hash != hnum ||
651 sk->sk_family != PF_INET6 ||
652 (inet->inet_dport && inet->inet_dport != rmt_port) ||
653 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
654 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
33b4b015
HR
655 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
656 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
657 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
5cf3d461
DH
658 return false;
659 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
660 return false;
661 return true;
1da177e4
LT
662}
663
4068579e
TH
664static void udp6_csum_zero_error(struct sk_buff *skb)
665{
666 /* RFC 2460 section 8.1 says that we SHOULD log
667 * this error. Well, it is reasonable.
668 */
ba7a46f1
JP
669 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
670 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
671 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
672}
673
1da177e4
LT
674/*
675 * Note: called only from the BH handler context,
676 * so we don't need to lock the hashes.
677 */
e3163493 678static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 679 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 680 struct udp_table *udptable, int proto)
1da177e4 681{
ca065d0c 682 struct sock *sk, *first = NULL;
4bedb452 683 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
684 unsigned short hnum = ntohs(uh->dest);
685 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
ca065d0c 686 unsigned int offset = offsetof(typeof(*sk), sk_node);
2dc41cff 687 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
ca065d0c
ED
688 int dif = inet6_iif(skb);
689 struct hlist_node *node;
690 struct sk_buff *nskb;
2dc41cff
DH
691
692 if (use_hash2) {
693 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
73e2d5e3
PN
694 udptable->mask;
695 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udptable->mask;
2dc41cff 696start_lookup:
73e2d5e3 697 hslot = &udptable->hash2[hash2];
2dc41cff
DH
698 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
699 }
1da177e4 700
ca065d0c
ED
701 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
702 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
703 uh->source, saddr, dif, hnum))
704 continue;
705 /* If zero checksum and no_check is not on for
706 * the socket then skip it.
707 */
708 if (!uh->check && !udp_sk(sk)->no_check6_rx)
709 continue;
710 if (!first) {
711 first = sk;
712 continue;
713 }
714 nskb = skb_clone(skb, GFP_ATOMIC);
715 if (unlikely(!nskb)) {
716 atomic_inc(&sk->sk_drops);
02c22347
ED
717 __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
718 IS_UDPLITE(sk));
719 __UDP6_INC_STATS(net, UDP_MIB_INERRORS,
720 IS_UDPLITE(sk));
ca065d0c 721 continue;
95766fff 722 }
a1ab77f9 723
ca065d0c
ED
724 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
725 consume_skb(nskb);
726 }
a1ab77f9 727
2dc41cff
DH
728 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
729 if (use_hash2 && hash2 != hash2_any) {
730 hash2 = hash2_any;
731 goto start_lookup;
732 }
733
ca065d0c
ED
734 if (first) {
735 if (udpv6_queue_rcv_skb(first, skb) > 0)
736 consume_skb(skb);
a1ab77f9 737 } else {
ca065d0c 738 kfree_skb(skb);
02c22347
ED
739 __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
740 proto == IPPROTO_UDPLITE);
a1ab77f9 741 }
ba4e58ec 742 return 0;
1da177e4
LT
743}
744
645ca708 745int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 746 int proto)
1da177e4 747{
ca065d0c 748 const struct in6_addr *saddr, *daddr;
3ffe533c 749 struct net *net = dev_net(skb->dev);
1ab1457c 750 struct udphdr *uh;
ca065d0c 751 struct sock *sk;
1da177e4
LT
752 u32 ulen = 0;
753
754 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 755 goto discard;
1da177e4 756
0660e03f
ACM
757 saddr = &ipv6_hdr(skb)->saddr;
758 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 759 uh = udp_hdr(skb);
1da177e4
LT
760
761 ulen = ntohs(uh->len);
ba4e58ec
GR
762 if (ulen > skb->len)
763 goto short_packet;
1da177e4 764
759e5d00
HX
765 if (proto == IPPROTO_UDP) {
766 /* UDP validates ulen. */
1da177e4 767
ba4e58ec
GR
768 /* Check for jumbo payload */
769 if (ulen == 0)
770 ulen = skb->len;
1da177e4 771
ba4e58ec
GR
772 if (ulen < sizeof(*uh))
773 goto short_packet;
1da177e4 774
ba4e58ec
GR
775 if (ulen < skb->len) {
776 if (pskb_trim_rcsum(skb, ulen))
777 goto short_packet;
0660e03f
ACM
778 saddr = &ipv6_hdr(skb)->saddr;
779 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 780 uh = udp_hdr(skb);
ba4e58ec 781 }
ba4e58ec 782 }
1da177e4 783
759e5d00 784 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 785 goto csum_error;
759e5d00 786
1ab1457c
YH
787 /*
788 * Multicast receive code
1da177e4 789 */
ba4e58ec 790 if (ipv6_addr_is_multicast(daddr))
e3163493 791 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 792 saddr, daddr, udptable, proto);
1da177e4
LT
793
794 /* Unicast */
795
1ab1457c 796 /*
1da177e4
LT
797 * check socket cache ... must talk to Alan about his plans
798 * for sock caches... i'll skip this for now.
799 */
607c4aaf 800 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
53b24b8f 801 if (sk) {
a5b50476
ET
802 int ret;
803
1c19448c 804 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
4068579e
TH
805 udp6_csum_zero_error(skb);
806 goto csum_error;
807 }
808
224d019c 809 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
2abb7cdc
TH
810 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
811 ip6_compute_pseudo);
812
a5b50476 813 ret = udpv6_queue_rcv_skb(sk, skb);
1da177e4 814
59dca1d8 815 /* a return value > 0 means to resubmit the input */
cb80ef46 816 if (ret > 0)
59dca1d8 817 return ret;
1da177e4 818
add459aa 819 return 0;
1da177e4 820 }
1ab1457c 821
4068579e
TH
822 if (!uh->check) {
823 udp6_csum_zero_error(skb);
824 goto csum_error;
825 }
826
cb80ef46 827 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 828 goto discard;
cb80ef46
BL
829
830 if (udp_lib_checksum_complete(skb))
6a5dc9e5 831 goto csum_error;
cb80ef46 832
02c22347 833 __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
cb80ef46
BL
834 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
835
836 kfree_skb(skb);
add459aa 837 return 0;
1da177e4 838
1ab1457c 839short_packet:
ba7a46f1
JP
840 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
841 proto == IPPROTO_UDPLITE ? "-Lite" : "",
842 saddr, ntohs(uh->source),
843 ulen, skb->len,
844 daddr, ntohs(uh->dest));
6a5dc9e5
ED
845 goto discard;
846csum_error:
02c22347 847 __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 848discard:
02c22347 849 __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 850 kfree_skb(skb);
add459aa 851 return 0;
1da177e4 852}
ba4e58ec 853
e5bbef20 854static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 855{
645ca708 856 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
857}
858
1da177e4
LT
859/*
860 * Throw away all pending data and cancel the corking. Socket is locked.
861 */
862static void udp_v6_flush_pending_frames(struct sock *sk)
863{
864 struct udp_sock *up = udp_sk(sk);
865
36d926b9
DL
866 if (up->pending == AF_INET)
867 udp_flush_pending_frames(sk);
868 else if (up->pending) {
1da177e4
LT
869 up->len = 0;
870 up->pending = 0;
871 ip6_flush_pending_frames(sk);
1ab1457c 872 }
1da177e4
LT
873}
874
493c6be3 875/**
67ba4152
IM
876 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
877 * @sk: socket we are sending on
878 * @skb: sk_buff containing the filled-in UDP header
879 * (checksum field must be zeroed out)
493c6be3
SS
880 */
881static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
882 const struct in6_addr *saddr,
883 const struct in6_addr *daddr, int len)
884{
885 unsigned int offset;
886 struct udphdr *uh = udp_hdr(skb);
d39d938c 887 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
493c6be3
SS
888 __wsum csum = 0;
889
d39d938c 890 if (!frags) {
493c6be3
SS
891 /* Only one fragment on the socket. */
892 skb->csum_start = skb_transport_header(skb) - skb->head;
893 skb->csum_offset = offsetof(struct udphdr, check);
894 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
895 } else {
896 /*
897 * HW-checksum won't work as there are two or more
898 * fragments on the socket so that all csums of sk_buffs
899 * should be together
900 */
901 offset = skb_transport_offset(skb);
902 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
903
904 skb->ip_summed = CHECKSUM_NONE;
905
d39d938c
VY
906 do {
907 csum = csum_add(csum, frags->csum);
908 } while ((frags = frags->next));
493c6be3
SS
909
910 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
911 csum);
912 if (uh->check == 0)
913 uh->check = CSUM_MANGLED_0;
914 }
915}
916
1da177e4
LT
917/*
918 * Sending
919 */
920
d39d938c 921static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1da177e4 922{
d39d938c 923 struct sock *sk = skb->sk;
1da177e4 924 struct udphdr *uh;
1da177e4 925 int err = 0;
b2bf1e26 926 int is_udplite = IS_UDPLITE(sk);
868c86bc 927 __wsum csum = 0;
d39d938c
VY
928 int offset = skb_transport_offset(skb);
929 int len = skb->len - offset;
1da177e4
LT
930
931 /*
932 * Create a UDP header
933 */
4bedb452 934 uh = udp_hdr(skb);
1958b856
DM
935 uh->source = fl6->fl6_sport;
936 uh->dest = fl6->fl6_dport;
d39d938c 937 uh->len = htons(len);
1da177e4
LT
938 uh->check = 0;
939
b2bf1e26 940 if (is_udplite)
d39d938c
VY
941 csum = udplite_csum(skb);
942 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
943 skb->ip_summed = CHECKSUM_NONE;
944 goto send;
945 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
d39d938c 946 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
493c6be3
SS
947 goto send;
948 } else
d39d938c 949 csum = udp_csum(skb);
1da177e4 950
ba4e58ec 951 /* add protocol-dependent pseudo-header */
4c9483b2 952 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
d39d938c 953 len, fl6->flowi6_proto, csum);
1da177e4 954 if (uh->check == 0)
f6ab0288 955 uh->check = CSUM_MANGLED_0;
1da177e4 956
493c6be3 957send:
d39d938c 958 err = ip6_send_skb(skb);
6ce9e7b5
ED
959 if (err) {
960 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
6aef70a8
ED
961 UDP6_INC_STATS(sock_net(sk),
962 UDP_MIB_SNDBUFERRORS, is_udplite);
6ce9e7b5
ED
963 err = 0;
964 }
6aef70a8
ED
965 } else {
966 UDP6_INC_STATS(sock_net(sk),
967 UDP_MIB_OUTDATAGRAMS, is_udplite);
968 }
d39d938c
VY
969 return err;
970}
971
972static int udp_v6_push_pending_frames(struct sock *sk)
973{
974 struct sk_buff *skb;
975 struct udp_sock *up = udp_sk(sk);
976 struct flowi6 fl6;
977 int err = 0;
978
979 if (up->pending == AF_INET)
980 return udp_push_pending_frames(sk);
981
982 /* ip6_finish_skb will release the cork, so make a copy of
983 * fl6 here.
984 */
985 fl6 = inet_sk(sk)->cork.fl.u.ip6;
986
987 skb = ip6_finish_skb(sk);
988 if (!skb)
989 goto out;
990
991 err = udp_v6_send_skb(skb, &fl6);
992
1da177e4
LT
993out:
994 up->len = 0;
995 up->pending = 0;
996 return err;
997}
998
1b784140 999int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1da177e4
LT
1000{
1001 struct ipv6_txoptions opt_space;
1002 struct udp_sock *up = udp_sk(sk);
1003 struct inet_sock *inet = inet_sk(sk);
1004 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1005 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1006 struct in6_addr *daddr, *final_p, final;
1da177e4 1007 struct ipv6_txoptions *opt = NULL;
45f6fad8 1008 struct ipv6_txoptions *opt_to_free = NULL;
1da177e4 1009 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1010 struct flowi6 fl6;
1da177e4 1011 struct dst_entry *dst;
26879da5 1012 struct ipcm6_cookie ipc6;
1da177e4
LT
1013 int addr_len = msg->msg_namelen;
1014 int ulen = len;
1da177e4
LT
1015 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1016 int err;
987905de 1017 int connected = 0;
b2bf1e26 1018 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1019 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
ad1e46a8 1020 struct sockcm_cookie sockc;
1da177e4 1021
26879da5
WW
1022 ipc6.hlimit = -1;
1023 ipc6.tclass = -1;
1024 ipc6.dontfrag = -1;
f7f7cd33 1025 sockc.tsflags = sk->sk_tsflags;
26879da5 1026
1da177e4
LT
1027 /* destination address check */
1028 if (sin6) {
1029 if (addr_len < offsetof(struct sockaddr, sa_data))
1030 return -EINVAL;
1031
1032 switch (sin6->sin6_family) {
1033 case AF_INET6:
1034 if (addr_len < SIN6_LEN_RFC2133)
1035 return -EINVAL;
1036 daddr = &sin6->sin6_addr;
052d2369
JL
1037 if (ipv6_addr_any(daddr) &&
1038 ipv6_addr_v4mapped(&np->saddr))
1039 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
1040 daddr);
1da177e4
LT
1041 break;
1042 case AF_INET:
1043 goto do_udp_sendmsg;
1044 case AF_UNSPEC:
1045 msg->msg_name = sin6 = NULL;
1046 msg->msg_namelen = addr_len = 0;
1047 daddr = NULL;
1048 break;
1049 default:
1050 return -EINVAL;
1051 }
1052 } else if (!up->pending) {
1053 if (sk->sk_state != TCP_ESTABLISHED)
1054 return -EDESTADDRREQ;
efe4208f 1055 daddr = &sk->sk_v6_daddr;
1ab1457c 1056 } else
1da177e4
LT
1057 daddr = NULL;
1058
1059 if (daddr) {
e773e4fa 1060 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1061 struct sockaddr_in sin;
1062 sin.sin_family = AF_INET;
c720c7e8 1063 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1064 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1065 msg->msg_name = &sin;
1066 msg->msg_namelen = sizeof(sin);
1067do_udp_sendmsg:
1068 if (__ipv6_only_sock(sk))
1069 return -ENETUNREACH;
1b784140 1070 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1071 }
1072 }
1073
1074 if (up->pending == AF_INET)
1b784140 1075 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1076
1077 /* Rough check on arithmetic overflow,
b59e139b 1078 better check is made in ip6_append_data().
1da177e4
LT
1079 */
1080 if (len > INT_MAX - sizeof(struct udphdr))
1081 return -EMSGSIZE;
1ab1457c 1082
03485f2a 1083 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1da177e4
LT
1084 if (up->pending) {
1085 /*
1086 * There are pending frames.
1087 * The socket lock must be held while it's corked.
1088 */
1089 lock_sock(sk);
1090 if (likely(up->pending)) {
1091 if (unlikely(up->pending != AF_INET6)) {
1092 release_sock(sk);
1093 return -EAFNOSUPPORT;
1094 }
1095 dst = NULL;
1096 goto do_append_data;
1097 }
1098 release_sock(sk);
1099 }
1100 ulen += sizeof(struct udphdr);
1101
4c9483b2 1102 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1103
1104 if (sin6) {
1105 if (sin6->sin6_port == 0)
1106 return -EINVAL;
1107
1958b856 1108 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1109 daddr = &sin6->sin6_addr;
1110
1111 if (np->sndflow) {
4c9483b2
DM
1112 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1113 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1114 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1115 if (!flowlabel)
1da177e4 1116 return -EINVAL;
1da177e4
LT
1117 }
1118 }
1119
1120 /*
1121 * Otherwise it will be difficult to maintain
1122 * sk->sk_dst_cache.
1123 */
1124 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1125 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1126 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1127
1128 if (addr_len >= sizeof(struct sockaddr_in6) &&
1129 sin6->sin6_scope_id &&
842df073 1130 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1131 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1132 } else {
1133 if (sk->sk_state != TCP_ESTABLISHED)
1134 return -EDESTADDRREQ;
1135
1958b856 1136 fl6.fl6_dport = inet->inet_dport;
efe4208f 1137 daddr = &sk->sk_v6_daddr;
4c9483b2 1138 fl6.flowlabel = np->flow_label;
987905de 1139 connected = 1;
1da177e4
LT
1140 }
1141
4c9483b2
DM
1142 if (!fl6.flowi6_oif)
1143 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1144
4c9483b2
DM
1145 if (!fl6.flowi6_oif)
1146 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1147
4c9483b2 1148 fl6.flowi6_mark = sk->sk_mark;
e2d118a1 1149 fl6.flowi6_uid = sk->sk_uid;
51953d5b 1150
1da177e4
LT
1151 if (msg->msg_controllen) {
1152 opt = &opt_space;
1153 memset(opt, 0, sizeof(struct ipv6_txoptions));
1154 opt->tot_len = sizeof(*opt);
26879da5 1155 ipc6.opt = opt;
1da177e4 1156
26879da5 1157 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
1da177e4
LT
1158 if (err < 0) {
1159 fl6_sock_release(flowlabel);
1160 return err;
1161 }
4c9483b2
DM
1162 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1163 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1164 if (!flowlabel)
1da177e4
LT
1165 return -EINVAL;
1166 }
1167 if (!(opt->opt_nflen|opt->opt_flen))
1168 opt = NULL;
987905de 1169 connected = 0;
1da177e4 1170 }
45f6fad8
ED
1171 if (!opt) {
1172 opt = txopt_get(np);
1173 opt_to_free = opt;
1174 }
df9890c3
YH
1175 if (flowlabel)
1176 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1177 opt = ipv6_fixup_options(&opt_space, opt);
26879da5 1178 ipc6.opt = opt;
1da177e4 1179
4c9483b2 1180 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1181 if (!ipv6_addr_any(daddr))
4e3fd7a0 1182 fl6.daddr = *daddr;
876c7f41 1183 else
4c9483b2
DM
1184 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1185 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1186 fl6.saddr = np->saddr;
1958b856 1187 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1188
4c9483b2 1189 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1190 if (final_p)
987905de 1191 connected = 0;
1da177e4 1192
4c9483b2
DM
1193 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1194 fl6.flowi6_oif = np->mcast_oif;
987905de 1195 connected = 0;
c4062dfc
EH
1196 } else if (!fl6.flowi6_oif)
1197 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1198
4c9483b2 1199 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1200
38b7097b
HFS
1201 if (ipc6.tclass < 0)
1202 ipc6.tclass = np->tclass;
1203
1204 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
1205
0e0d44ab 1206 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1207 if (IS_ERR(dst)) {
1208 err = PTR_ERR(dst);
1209 dst = NULL;
1da177e4 1210 goto out;
14e50e57 1211 }
1da177e4 1212
26879da5
WW
1213 if (ipc6.hlimit < 0)
1214 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4
LT
1215
1216 if (msg->msg_flags&MSG_CONFIRM)
1217 goto do_confirm;
1218back_from_confirm:
1219
03485f2a
VY
1220 /* Lockless fast path for the non-corking case */
1221 if (!corkreq) {
1222 struct sk_buff *skb;
1223
1224 skb = ip6_make_skb(sk, getfrag, msg, ulen,
26879da5 1225 sizeof(struct udphdr), &ipc6,
03485f2a 1226 &fl6, (struct rt6_info *)dst,
26879da5 1227 msg->msg_flags, &sockc);
03485f2a
VY
1228 err = PTR_ERR(skb);
1229 if (!IS_ERR_OR_NULL(skb))
1230 err = udp_v6_send_skb(skb, &fl6);
1231 goto release_dst;
1232 }
1233
1da177e4
LT
1234 lock_sock(sk);
1235 if (unlikely(up->pending)) {
1236 /* The socket is already corked while preparing it. */
1237 /* ... which is an evident application bug. --ANK */
1238 release_sock(sk);
1239
ba7a46f1 1240 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1241 err = -EINVAL;
1242 goto out;
1243 }
1244
1245 up->pending = AF_INET6;
1246
1247do_append_data:
26879da5
WW
1248 if (ipc6.dontfrag < 0)
1249 ipc6.dontfrag = np->dontfrag;
1da177e4 1250 up->len += ulen;
26879da5
WW
1251 err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1252 &ipc6, &fl6, (struct rt6_info *)dst,
1253 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, &sockc);
1da177e4
LT
1254 if (err)
1255 udp_v6_flush_pending_frames(sk);
1256 else if (!corkreq)
4c0a6cb0 1257 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1258 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1259 up->pending = 0;
1da177e4 1260
03485f2a
VY
1261 if (err > 0)
1262 err = np->recverr ? net_xmit_errno(err) : 0;
1263 release_sock(sk);
1264
1265release_dst:
a5e7c210
DM
1266 if (dst) {
1267 if (connected) {
1268 ip6_dst_store(sk, dst,
efe4208f
ED
1269 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1270 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1271#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1272 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1273 &np->saddr :
1274#endif
1275 NULL);
a5e7c210
DM
1276 } else {
1277 dst_release(dst);
1278 }
a3c96089 1279 dst = NULL;
a5e7c210
DM
1280 }
1281
1da177e4 1282out:
a3c96089 1283 dst_release(dst);
1da177e4 1284 fl6_sock_release(flowlabel);
45f6fad8 1285 txopt_put(opt_to_free);
cd562c98 1286 if (!err)
1da177e4 1287 return len;
a18135eb
DM
1288 /*
1289 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1290 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1291 * we don't have a good statistic (IpOutDiscards but it can be too many
1292 * things). We could add another new stat but at least for now that
1293 * seems like overkill.
1294 */
1295 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
6aef70a8
ED
1296 UDP6_INC_STATS(sock_net(sk),
1297 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1298 }
1da177e4
LT
1299 return err;
1300
1301do_confirm:
78556423
JA
1302 if (msg->msg_flags & MSG_PROBE)
1303 dst_confirm_neigh(dst, &fl6.daddr);
1da177e4
LT
1304 if (!(msg->msg_flags&MSG_PROBE) || len)
1305 goto back_from_confirm;
1306 err = 0;
1307 goto out;
1308}
1309
7d06b2e0 1310void udpv6_destroy_sock(struct sock *sk)
1da177e4 1311{
44046a59 1312 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1313 lock_sock(sk);
1314 udp_v6_flush_pending_frames(sk);
1315 release_sock(sk);
1316
44046a59
TP
1317 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1318 void (*encap_destroy)(struct sock *sk);
1319 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1320 if (encap_destroy)
1321 encap_destroy(sk);
1322 }
1323
1da177e4 1324 inet6_destroy_sock(sk);
1da177e4
LT
1325}
1326
1327/*
1328 * Socket option code for UDP
1329 */
ba4e58ec 1330int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1331 char __user *optval, unsigned int optlen)
3fdadf7d 1332{
db8dac20 1333 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1334 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1335 udp_v6_push_pending_frames);
ba4e58ec 1336 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1337}
1338
1339#ifdef CONFIG_COMPAT
ba4e58ec 1340int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1341 char __user *optval, unsigned int optlen)
3fdadf7d 1342{
db8dac20 1343 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1344 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1345 udp_v6_push_pending_frames);
ba4e58ec 1346 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1347}
1348#endif
1349
ba4e58ec
GR
1350int udpv6_getsockopt(struct sock *sk, int level, int optname,
1351 char __user *optval, int __user *optlen)
3fdadf7d 1352{
db8dac20 1353 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1354 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1355 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1356}
1357
1358#ifdef CONFIG_COMPAT
ba4e58ec
GR
1359int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1360 char __user *optval, int __user *optlen)
3fdadf7d 1361{
db8dac20 1362 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1363 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1364 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1365}
1366#endif
1367
41135cc8 1368static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1369 .handler = udpv6_rcv,
1370 .err_handler = udpv6_err,
1371 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1372};
1373
1374/* ------------------------------------------------------------------------ */
1375#ifdef CONFIG_PROC_FS
ba4e58ec 1376int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1377{
17ef66af
LC
1378 if (v == SEQ_START_TOKEN) {
1379 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1380 } else {
1381 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1382 struct inet_sock *inet = inet_sk(v);
1383 __u16 srcp = ntohs(inet->inet_sport);
1384 __u16 destp = ntohs(inet->inet_dport);
1385 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1386 }
1da177e4
LT
1387 return 0;
1388}
1389
73cb88ec
AV
1390static const struct file_operations udp6_afinfo_seq_fops = {
1391 .owner = THIS_MODULE,
1392 .open = udp_seq_open,
1393 .read = seq_read,
1394 .llseek = seq_lseek,
1395 .release = seq_release_net
1396};
1397
1da177e4 1398static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1399 .name = "udp6",
1400 .family = AF_INET6,
645ca708 1401 .udp_table = &udp_table,
73cb88ec 1402 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1403 .seq_ops = {
1404 .show = udp6_seq_show,
1405 },
1da177e4
LT
1406};
1407
2c8c1e72 1408int __net_init udp6_proc_init(struct net *net)
1da177e4 1409{
0c96d8c5 1410 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1411}
1412
ec120da6
IM
1413void udp6_proc_exit(struct net *net)
1414{
0c96d8c5 1415 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1416}
1417#endif /* CONFIG_PROC_FS */
1418
1419/* ------------------------------------------------------------------------ */
1420
1421struct proto udpv6_prot = {
543d9cfe
ACM
1422 .name = "UDPv6",
1423 .owner = THIS_MODULE,
ba4e58ec 1424 .close = udp_lib_close,
543d9cfe
ACM
1425 .connect = ip6_datagram_connect,
1426 .disconnect = udp_disconnect,
1427 .ioctl = udp_ioctl,
850cbadd 1428 .init = udp_init_sock,
543d9cfe
ACM
1429 .destroy = udpv6_destroy_sock,
1430 .setsockopt = udpv6_setsockopt,
1431 .getsockopt = udpv6_getsockopt,
1432 .sendmsg = udpv6_sendmsg,
1433 .recvmsg = udpv6_recvmsg,
e646b657 1434 .release_cb = ip6_datagram_release_cb,
ba4e58ec
GR
1435 .hash = udp_lib_hash,
1436 .unhash = udp_lib_unhash,
719f8358 1437 .rehash = udp_v6_rehash,
543d9cfe 1438 .get_port = udp_v6_get_port,
95766fff
HA
1439 .memory_allocated = &udp_memory_allocated,
1440 .sysctl_mem = sysctl_udp_mem,
1441 .sysctl_wmem = &sysctl_udp_wmem_min,
1442 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1443 .obj_size = sizeof(struct udp6_sock),
645ca708 1444 .h.udp_table = &udp_table,
3fdadf7d 1445#ifdef CONFIG_COMPAT
543d9cfe
ACM
1446 .compat_setsockopt = compat_udpv6_setsockopt,
1447 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1448#endif
5d77dca8 1449 .diag_destroy = udp_abort,
1da177e4
LT
1450};
1451
1da177e4
LT
1452static struct inet_protosw udpv6_protosw = {
1453 .type = SOCK_DGRAM,
1454 .protocol = IPPROTO_UDP,
1455 .prot = &udpv6_prot,
1456 .ops = &inet6_dgram_ops,
1da177e4
LT
1457 .flags = INET_PROTOSW_PERMANENT,
1458};
1459
7f4e4868 1460int __init udpv6_init(void)
1da177e4 1461{
7f4e4868
DL
1462 int ret;
1463
3336288a
VY
1464 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1465 if (ret)
c6b641a4 1466 goto out;
3336288a 1467
7f4e4868
DL
1468 ret = inet6_register_protosw(&udpv6_protosw);
1469 if (ret)
1470 goto out_udpv6_protocol;
1471out:
1472 return ret;
1473
1474out_udpv6_protocol:
1475 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1476 goto out;
1477}
1478
09f7709f 1479void udpv6_exit(void)
7f4e4868
DL
1480{
1481 inet6_unregister_protosw(&udpv6_protosw);
1482 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1483}