Merge branch 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszer...
[linux-2.6-block.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>
1da177e4
LT
38#include <asm/uaccess.h>
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) || \
305 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
aa976fc0
BS
306struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
307 const struct in6_addr *daddr, __be16 dport, int dif)
308{
ca065d0c
ED
309 struct sock *sk;
310
311 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport,
312 dif, &udp_table, NULL);
313 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
314 sk = NULL;
315 return sk;
aa976fc0
BS
316}
317EXPORT_SYMBOL_GPL(udp6_lib_lookup);
ca065d0c 318#endif
aa976fc0 319
1da177e4 320/*
67ba4152
IM
321 * This should be easy, if there is something there we
322 * return it, otherwise we block.
1da177e4
LT
323 */
324
1b784140 325int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
1da177e4
LT
326 int noblock, int flags, int *addr_len)
327{
328 struct ipv6_pinfo *np = inet6_sk(sk);
329 struct inet_sock *inet = inet_sk(sk);
1ab1457c 330 struct sk_buff *skb;
59c2cdae 331 unsigned int ulen, copied;
627d2d6b 332 int peeked, peeking, off;
759e5d00
HX
333 int err;
334 int is_udplite = IS_UDPLITE(sk);
197c949e 335 bool checksum_valid = false;
f26ba175 336 int is_udp4;
8a74ad60 337 bool slow;
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);
a59322be 347 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 348 &peeked, &off, &err);
1da177e4 349 if (!skb)
627d2d6b 350 return err;
1da177e4 351
e6afc8ac 352 ulen = skb->len;
59c2cdae 353 copied = len;
627d2d6b 354 if (copied > ulen - off)
355 copied = ulen - off;
59c2cdae 356 else if (copied < ulen)
1ab1457c 357 msg->msg_flags |= MSG_TRUNC;
1da177e4 358
f26ba175
WY
359 is_udp4 = (skb->protocol == htons(ETH_P_IP));
360
ba4e58ec 361 /*
759e5d00
HX
362 * If checksum is needed at all, try to do it while copying the
363 * data. If the data is truncated, or if we only want a partial
364 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 365 */
ba4e58ec 366
627d2d6b 367 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
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
ED
380 if (unlikely(err)) {
381 trace_kfree_skb(skb, udpv6_recvmsg);
979402b1
ED
382 if (!peeked) {
383 atomic_inc(&sk->sk_drops);
384 if (is_udp4)
6aef70a8
ED
385 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
386 is_udplite);
979402b1 387 else
6aef70a8
ED
388 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
389 is_udplite);
979402b1 390 }
627d2d6b 391 skb_free_datagram_locked(sk, skb);
392 return err;
22911fc5 393 }
f26ba175
WY
394 if (!peeked) {
395 if (is_udp4)
6aef70a8
ED
396 UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
397 is_udplite);
f26ba175 398 else
6aef70a8
ED
399 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
400 is_udplite);
f26ba175 401 }
cb75994e 402
3b885787 403 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
404
405 /* Copy the address. */
406 if (msg->msg_name) {
342dfc30 407 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 408 sin6->sin6_family = AF_INET6;
4bedb452 409 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 410 sin6->sin6_flowinfo = 0;
1da177e4 411
842df073 412 if (is_udp4) {
b301e82c
BH
413 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
414 &sin6->sin6_addr);
842df073
HFS
415 sin6->sin6_scope_id = 0;
416 } else {
4e3fd7a0 417 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
418 sin6->sin6_scope_id =
419 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 420 inet6_iif(skb));
1da177e4 421 }
bceaa902 422 *addr_len = sizeof(*sin6);
1da177e4 423 }
4b261c75
HFS
424
425 if (np->rxopt.all)
426 ip6_datagram_recv_common_ctl(sk, msg, skb);
427
f26ba175 428 if (is_udp4) {
1da177e4
LT
429 if (inet->cmsg_flags)
430 ip_cmsg_recv(msg, skb);
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
627d2d6b 440 __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
1da177e4
LT
441 return err;
442
443csum_copy_err:
8a74ad60 444 slow = lock_sock_fast(sk);
0856f939 445 if (!skb_kill_datagram(sk, skb, flags)) {
6a5dc9e5 446 if (is_udp4) {
6aef70a8
ED
447 UDP_INC_STATS(sock_net(sk),
448 UDP_MIB_CSUMERRORS, is_udplite);
449 UDP_INC_STATS(sock_net(sk),
450 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 451 } else {
6aef70a8
ED
452 UDP6_INC_STATS(sock_net(sk),
453 UDP_MIB_CSUMERRORS, is_udplite);
454 UDP6_INC_STATS(sock_net(sk),
455 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 456 }
0856f939 457 }
8a74ad60 458 unlock_sock_fast(sk, slow);
1da177e4 459
beb39db5
ED
460 /* starting over for a new packet, but check if we need to yield */
461 cond_resched();
9cfaa8de 462 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
463 goto try_again;
464}
465
ba4e58ec 466void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 467 u8 type, u8 code, int offset, __be32 info,
645ca708 468 struct udp_table *udptable)
1da177e4
LT
469{
470 struct ipv6_pinfo *np;
b71d1d42
ED
471 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
472 const struct in6_addr *saddr = &hdr->saddr;
473 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 474 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4 475 struct sock *sk;
e0d8c1b7 476 int harderr;
1da177e4 477 int err;
7304fe46 478 struct net *net = dev_net(skb->dev);
1da177e4 479
e32ea7e7 480 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
538950a1 481 inet6_iif(skb), udptable, skb);
63159f29 482 if (!sk) {
a16292a0
ED
483 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
484 ICMP6_MIB_INERRORS);
1da177e4 485 return;
7304fe46 486 }
1da177e4 487
e0d8c1b7
WW
488 harderr = icmpv6_err_convert(type, code, &err);
489 np = inet6_sk(sk);
490
93b36cf3
HFS
491 if (type == ICMPV6_PKT_TOOBIG) {
492 if (!ip6_sk_accept_pmtu(sk))
493 goto out;
81aded24 494 ip6_sk_update_pmtu(skb, sk, info);
e0d8c1b7
WW
495 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
496 harderr = 1;
93b36cf3 497 }
1a462d18 498 if (type == NDISC_REDIRECT) {
ec18d9a2 499 ip6_sk_redirect(skb, sk);
1a462d18
DJ
500 goto out;
501 }
81aded24 502
e0d8c1b7
WW
503 if (!np->recverr) {
504 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
505 goto out;
506 } else {
1da177e4 507 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
e0d8c1b7 508 }
b1faf566 509
1da177e4
LT
510 sk->sk_err = err;
511 sk->sk_error_report(sk);
512out:
ca065d0c 513 return;
1da177e4
LT
514}
515
f7ad74fe
BL
516static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
517{
518 int rc;
519
efe4208f 520 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 521 sock_rps_save_rxhash(sk, skb);
005ec974 522 sk_mark_napi_id(sk, skb);
2c8c56e1 523 sk_incoming_cpu_update(sk);
005ec974 524 }
f7ad74fe 525
e6afc8ac 526 rc = __sock_queue_rcv_skb(sk, skb);
f7ad74fe
BL
527 if (rc < 0) {
528 int is_udplite = IS_UDPLITE(sk);
529
530 /* Note that an ENOMEM error is charged twice */
531 if (rc == -ENOMEM)
e61da9e2 532 UDP6_INC_STATS(sock_net(sk),
02c22347 533 UDP_MIB_RCVBUFERRORS, is_udplite);
e61da9e2 534 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
f7ad74fe
BL
535 kfree_skb(skb);
536 return -1;
537 }
538 return 0;
539}
540
ba4e58ec 541static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 542 struct inet6_skb_parm *opt, u8 type,
67ba4152 543 u8 code, int offset, __be32 info)
ba4e58ec 544{
645ca708 545 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
546}
547
d7f3f621
BL
548static struct static_key udpv6_encap_needed __read_mostly;
549void udpv6_encap_enable(void)
550{
551 if (!static_key_enabled(&udpv6_encap_needed))
552 static_key_slow_inc(&udpv6_encap_needed);
553}
554EXPORT_SYMBOL(udpv6_encap_enable);
555
f7ad74fe 556int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 557{
ba4e58ec 558 struct udp_sock *up = udp_sk(sk);
a18135eb 559 int rc;
b2bf1e26 560 int is_udplite = IS_UDPLITE(sk);
a18135eb 561
ba4e58ec
GR
562 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
563 goto drop;
1da177e4 564
d7f3f621
BL
565 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
566 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
567
568 /*
569 * This is an encapsulation socket so pass the skb to
570 * the socket's udp_encap_rcv() hook. Otherwise, just
571 * fall through and pass this up the UDP socket.
572 * up->encap_rcv() returns the following value:
573 * =0 if skb was successfully passed to the encap
574 * handler or was discarded by it.
575 * >0 if skb should be passed on to UDP.
576 * <0 if skb should be resubmitted as proto -N
577 */
578
579 /* if we're overly short, let UDP handle it */
580 encap_rcv = ACCESS_ONCE(up->encap_rcv);
e5aed006 581 if (encap_rcv) {
d7f3f621
BL
582 int ret;
583
0a80966b
TH
584 /* Verify checksum before giving to encap */
585 if (udp_lib_checksum_complete(skb))
586 goto csum_error;
587
d7f3f621
BL
588 ret = encap_rcv(sk, skb);
589 if (ret <= 0) {
02c22347
ED
590 __UDP_INC_STATS(sock_net(sk),
591 UDP_MIB_INDATAGRAMS,
592 is_udplite);
d7f3f621
BL
593 return -ret;
594 }
595 }
596
597 /* FALLTHROUGH -- it's a UDP Packet */
598 }
599
ba4e58ec
GR
600 /*
601 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
602 */
b2bf1e26 603 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
604
605 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
606 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
607 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
608 goto drop;
609 }
610 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
611 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
612 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
613 goto drop;
614 }
1da177e4
LT
615 }
616
ce25d66a
ED
617 if (rcu_access_pointer(sk->sk_filter) &&
618 udp_lib_checksum_complete(skb))
619 goto csum_error;
620
621 if (sk_filter(sk, skb))
622 goto drop;
ba4e58ec 623
e6afc8ac 624 udp_csum_pull_header(skb);
274f482d 625 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
02c22347
ED
626 __UDP6_INC_STATS(sock_net(sk),
627 UDP_MIB_RCVBUFERRORS, is_udplite);
cb80ef46 628 goto drop;
3e215c8d 629 }
cb80ef46 630
d826eb14 631 skb_dst_drop(skb);
cb75994e 632
cb80ef46
BL
633 bh_lock_sock(sk);
634 rc = 0;
635 if (!sock_owned_by_user(sk))
636 rc = __udpv6_queue_rcv_skb(sk, skb);
637 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
638 bh_unlock_sock(sk);
639 goto drop;
640 }
641 bh_unlock_sock(sk);
f7ad74fe
BL
642
643 return rc;
3e215c8d 644
6a5dc9e5 645csum_error:
02c22347 646 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 647drop:
02c22347 648 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 649 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
650 kfree_skb(skb);
651 return -1;
1da177e4
LT
652}
653
5cf3d461
DH
654static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
655 __be16 loc_port, const struct in6_addr *loc_addr,
656 __be16 rmt_port, const struct in6_addr *rmt_addr,
657 int dif, unsigned short hnum)
1da177e4 658{
5cf3d461 659 struct inet_sock *inet = inet_sk(sk);
1da177e4 660
5cf3d461
DH
661 if (!net_eq(sock_net(sk), net))
662 return false;
663
664 if (udp_sk(sk)->udp_port_hash != hnum ||
665 sk->sk_family != PF_INET6 ||
666 (inet->inet_dport && inet->inet_dport != rmt_port) ||
667 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
668 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
33b4b015
HR
669 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
670 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
671 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
5cf3d461
DH
672 return false;
673 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
674 return false;
675 return true;
1da177e4
LT
676}
677
4068579e
TH
678static void udp6_csum_zero_error(struct sk_buff *skb)
679{
680 /* RFC 2460 section 8.1 says that we SHOULD log
681 * this error. Well, it is reasonable.
682 */
ba7a46f1
JP
683 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
684 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
685 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
686}
687
1da177e4
LT
688/*
689 * Note: called only from the BH handler context,
690 * so we don't need to lock the hashes.
691 */
e3163493 692static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 693 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 694 struct udp_table *udptable, int proto)
1da177e4 695{
ca065d0c 696 struct sock *sk, *first = NULL;
4bedb452 697 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
698 unsigned short hnum = ntohs(uh->dest);
699 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
ca065d0c 700 unsigned int offset = offsetof(typeof(*sk), sk_node);
2dc41cff 701 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
ca065d0c
ED
702 int dif = inet6_iif(skb);
703 struct hlist_node *node;
704 struct sk_buff *nskb;
2dc41cff
DH
705
706 if (use_hash2) {
707 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
708 udp_table.mask;
709 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
710start_lookup:
711 hslot = &udp_table.hash2[hash2];
712 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
713 }
1da177e4 714
ca065d0c
ED
715 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
716 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
717 uh->source, saddr, dif, hnum))
718 continue;
719 /* If zero checksum and no_check is not on for
720 * the socket then skip it.
721 */
722 if (!uh->check && !udp_sk(sk)->no_check6_rx)
723 continue;
724 if (!first) {
725 first = sk;
726 continue;
727 }
728 nskb = skb_clone(skb, GFP_ATOMIC);
729 if (unlikely(!nskb)) {
730 atomic_inc(&sk->sk_drops);
02c22347
ED
731 __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
732 IS_UDPLITE(sk));
733 __UDP6_INC_STATS(net, UDP_MIB_INERRORS,
734 IS_UDPLITE(sk));
ca065d0c 735 continue;
95766fff 736 }
a1ab77f9 737
ca065d0c
ED
738 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
739 consume_skb(nskb);
740 }
a1ab77f9 741
2dc41cff
DH
742 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
743 if (use_hash2 && hash2 != hash2_any) {
744 hash2 = hash2_any;
745 goto start_lookup;
746 }
747
ca065d0c
ED
748 if (first) {
749 if (udpv6_queue_rcv_skb(first, skb) > 0)
750 consume_skb(skb);
a1ab77f9 751 } else {
ca065d0c 752 kfree_skb(skb);
02c22347
ED
753 __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
754 proto == IPPROTO_UDPLITE);
a1ab77f9 755 }
ba4e58ec 756 return 0;
1da177e4
LT
757}
758
645ca708 759int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 760 int proto)
1da177e4 761{
ca065d0c 762 const struct in6_addr *saddr, *daddr;
3ffe533c 763 struct net *net = dev_net(skb->dev);
1ab1457c 764 struct udphdr *uh;
ca065d0c 765 struct sock *sk;
1da177e4
LT
766 u32 ulen = 0;
767
768 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 769 goto discard;
1da177e4 770
0660e03f
ACM
771 saddr = &ipv6_hdr(skb)->saddr;
772 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 773 uh = udp_hdr(skb);
1da177e4
LT
774
775 ulen = ntohs(uh->len);
ba4e58ec
GR
776 if (ulen > skb->len)
777 goto short_packet;
1da177e4 778
759e5d00
HX
779 if (proto == IPPROTO_UDP) {
780 /* UDP validates ulen. */
1da177e4 781
ba4e58ec
GR
782 /* Check for jumbo payload */
783 if (ulen == 0)
784 ulen = skb->len;
1da177e4 785
ba4e58ec
GR
786 if (ulen < sizeof(*uh))
787 goto short_packet;
1da177e4 788
ba4e58ec
GR
789 if (ulen < skb->len) {
790 if (pskb_trim_rcsum(skb, ulen))
791 goto short_packet;
0660e03f
ACM
792 saddr = &ipv6_hdr(skb)->saddr;
793 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 794 uh = udp_hdr(skb);
ba4e58ec 795 }
ba4e58ec 796 }
1da177e4 797
759e5d00 798 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 799 goto csum_error;
759e5d00 800
1ab1457c
YH
801 /*
802 * Multicast receive code
1da177e4 803 */
ba4e58ec 804 if (ipv6_addr_is_multicast(daddr))
e3163493 805 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 806 saddr, daddr, udptable, proto);
1da177e4
LT
807
808 /* Unicast */
809
1ab1457c 810 /*
1da177e4
LT
811 * check socket cache ... must talk to Alan about his plans
812 * for sock caches... i'll skip this for now.
813 */
607c4aaf 814 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
53b24b8f 815 if (sk) {
a5b50476
ET
816 int ret;
817
1c19448c 818 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
4068579e
TH
819 udp6_csum_zero_error(skb);
820 goto csum_error;
821 }
822
224d019c 823 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
2abb7cdc
TH
824 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
825 ip6_compute_pseudo);
826
a5b50476 827 ret = udpv6_queue_rcv_skb(sk, skb);
1da177e4 828
59dca1d8 829 /* a return value > 0 means to resubmit the input */
cb80ef46 830 if (ret > 0)
59dca1d8 831 return ret;
1da177e4 832
add459aa 833 return 0;
1da177e4 834 }
1ab1457c 835
4068579e
TH
836 if (!uh->check) {
837 udp6_csum_zero_error(skb);
838 goto csum_error;
839 }
840
cb80ef46 841 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 842 goto discard;
cb80ef46
BL
843
844 if (udp_lib_checksum_complete(skb))
6a5dc9e5 845 goto csum_error;
cb80ef46 846
02c22347 847 __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
cb80ef46
BL
848 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
849
850 kfree_skb(skb);
add459aa 851 return 0;
1da177e4 852
1ab1457c 853short_packet:
ba7a46f1
JP
854 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
855 proto == IPPROTO_UDPLITE ? "-Lite" : "",
856 saddr, ntohs(uh->source),
857 ulen, skb->len,
858 daddr, ntohs(uh->dest));
6a5dc9e5
ED
859 goto discard;
860csum_error:
02c22347 861 __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 862discard:
02c22347 863 __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 864 kfree_skb(skb);
add459aa 865 return 0;
1da177e4 866}
ba4e58ec 867
e5bbef20 868static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 869{
645ca708 870 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
871}
872
1da177e4
LT
873/*
874 * Throw away all pending data and cancel the corking. Socket is locked.
875 */
876static void udp_v6_flush_pending_frames(struct sock *sk)
877{
878 struct udp_sock *up = udp_sk(sk);
879
36d926b9
DL
880 if (up->pending == AF_INET)
881 udp_flush_pending_frames(sk);
882 else if (up->pending) {
1da177e4
LT
883 up->len = 0;
884 up->pending = 0;
885 ip6_flush_pending_frames(sk);
1ab1457c 886 }
1da177e4
LT
887}
888
493c6be3 889/**
67ba4152
IM
890 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
891 * @sk: socket we are sending on
892 * @skb: sk_buff containing the filled-in UDP header
893 * (checksum field must be zeroed out)
493c6be3
SS
894 */
895static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
896 const struct in6_addr *saddr,
897 const struct in6_addr *daddr, int len)
898{
899 unsigned int offset;
900 struct udphdr *uh = udp_hdr(skb);
d39d938c 901 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
493c6be3
SS
902 __wsum csum = 0;
903
d39d938c 904 if (!frags) {
493c6be3
SS
905 /* Only one fragment on the socket. */
906 skb->csum_start = skb_transport_header(skb) - skb->head;
907 skb->csum_offset = offsetof(struct udphdr, check);
908 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
909 } else {
910 /*
911 * HW-checksum won't work as there are two or more
912 * fragments on the socket so that all csums of sk_buffs
913 * should be together
914 */
915 offset = skb_transport_offset(skb);
916 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
917
918 skb->ip_summed = CHECKSUM_NONE;
919
d39d938c
VY
920 do {
921 csum = csum_add(csum, frags->csum);
922 } while ((frags = frags->next));
493c6be3
SS
923
924 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
925 csum);
926 if (uh->check == 0)
927 uh->check = CSUM_MANGLED_0;
928 }
929}
930
1da177e4
LT
931/*
932 * Sending
933 */
934
d39d938c 935static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1da177e4 936{
d39d938c 937 struct sock *sk = skb->sk;
1da177e4 938 struct udphdr *uh;
1da177e4 939 int err = 0;
b2bf1e26 940 int is_udplite = IS_UDPLITE(sk);
868c86bc 941 __wsum csum = 0;
d39d938c
VY
942 int offset = skb_transport_offset(skb);
943 int len = skb->len - offset;
1da177e4
LT
944
945 /*
946 * Create a UDP header
947 */
4bedb452 948 uh = udp_hdr(skb);
1958b856
DM
949 uh->source = fl6->fl6_sport;
950 uh->dest = fl6->fl6_dport;
d39d938c 951 uh->len = htons(len);
1da177e4
LT
952 uh->check = 0;
953
b2bf1e26 954 if (is_udplite)
d39d938c
VY
955 csum = udplite_csum(skb);
956 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
957 skb->ip_summed = CHECKSUM_NONE;
958 goto send;
959 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
d39d938c 960 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
493c6be3
SS
961 goto send;
962 } else
d39d938c 963 csum = udp_csum(skb);
1da177e4 964
ba4e58ec 965 /* add protocol-dependent pseudo-header */
4c9483b2 966 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
d39d938c 967 len, fl6->flowi6_proto, csum);
1da177e4 968 if (uh->check == 0)
f6ab0288 969 uh->check = CSUM_MANGLED_0;
1da177e4 970
493c6be3 971send:
d39d938c 972 err = ip6_send_skb(skb);
6ce9e7b5
ED
973 if (err) {
974 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
6aef70a8
ED
975 UDP6_INC_STATS(sock_net(sk),
976 UDP_MIB_SNDBUFERRORS, is_udplite);
6ce9e7b5
ED
977 err = 0;
978 }
6aef70a8
ED
979 } else {
980 UDP6_INC_STATS(sock_net(sk),
981 UDP_MIB_OUTDATAGRAMS, is_udplite);
982 }
d39d938c
VY
983 return err;
984}
985
986static int udp_v6_push_pending_frames(struct sock *sk)
987{
988 struct sk_buff *skb;
989 struct udp_sock *up = udp_sk(sk);
990 struct flowi6 fl6;
991 int err = 0;
992
993 if (up->pending == AF_INET)
994 return udp_push_pending_frames(sk);
995
996 /* ip6_finish_skb will release the cork, so make a copy of
997 * fl6 here.
998 */
999 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1000
1001 skb = ip6_finish_skb(sk);
1002 if (!skb)
1003 goto out;
1004
1005 err = udp_v6_send_skb(skb, &fl6);
1006
1da177e4
LT
1007out:
1008 up->len = 0;
1009 up->pending = 0;
1010 return err;
1011}
1012
1b784140 1013int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1da177e4
LT
1014{
1015 struct ipv6_txoptions opt_space;
1016 struct udp_sock *up = udp_sk(sk);
1017 struct inet_sock *inet = inet_sk(sk);
1018 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1019 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1020 struct in6_addr *daddr, *final_p, final;
1da177e4 1021 struct ipv6_txoptions *opt = NULL;
45f6fad8 1022 struct ipv6_txoptions *opt_to_free = NULL;
1da177e4 1023 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1024 struct flowi6 fl6;
1da177e4 1025 struct dst_entry *dst;
26879da5 1026 struct ipcm6_cookie ipc6;
1da177e4
LT
1027 int addr_len = msg->msg_namelen;
1028 int ulen = len;
1da177e4
LT
1029 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1030 int err;
987905de 1031 int connected = 0;
b2bf1e26 1032 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1033 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
ad1e46a8 1034 struct sockcm_cookie sockc;
1da177e4 1035
26879da5
WW
1036 ipc6.hlimit = -1;
1037 ipc6.tclass = -1;
1038 ipc6.dontfrag = -1;
1039
1da177e4
LT
1040 /* destination address check */
1041 if (sin6) {
1042 if (addr_len < offsetof(struct sockaddr, sa_data))
1043 return -EINVAL;
1044
1045 switch (sin6->sin6_family) {
1046 case AF_INET6:
1047 if (addr_len < SIN6_LEN_RFC2133)
1048 return -EINVAL;
1049 daddr = &sin6->sin6_addr;
1050 break;
1051 case AF_INET:
1052 goto do_udp_sendmsg;
1053 case AF_UNSPEC:
1054 msg->msg_name = sin6 = NULL;
1055 msg->msg_namelen = addr_len = 0;
1056 daddr = NULL;
1057 break;
1058 default:
1059 return -EINVAL;
1060 }
1061 } else if (!up->pending) {
1062 if (sk->sk_state != TCP_ESTABLISHED)
1063 return -EDESTADDRREQ;
efe4208f 1064 daddr = &sk->sk_v6_daddr;
1ab1457c 1065 } else
1da177e4
LT
1066 daddr = NULL;
1067
1068 if (daddr) {
e773e4fa 1069 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1070 struct sockaddr_in sin;
1071 sin.sin_family = AF_INET;
c720c7e8 1072 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1073 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1074 msg->msg_name = &sin;
1075 msg->msg_namelen = sizeof(sin);
1076do_udp_sendmsg:
1077 if (__ipv6_only_sock(sk))
1078 return -ENETUNREACH;
1b784140 1079 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1080 }
1081 }
1082
1083 if (up->pending == AF_INET)
1b784140 1084 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1085
1086 /* Rough check on arithmetic overflow,
b59e139b 1087 better check is made in ip6_append_data().
1da177e4
LT
1088 */
1089 if (len > INT_MAX - sizeof(struct udphdr))
1090 return -EMSGSIZE;
1ab1457c 1091
03485f2a 1092 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1da177e4
LT
1093 if (up->pending) {
1094 /*
1095 * There are pending frames.
1096 * The socket lock must be held while it's corked.
1097 */
1098 lock_sock(sk);
1099 if (likely(up->pending)) {
1100 if (unlikely(up->pending != AF_INET6)) {
1101 release_sock(sk);
1102 return -EAFNOSUPPORT;
1103 }
1104 dst = NULL;
1105 goto do_append_data;
1106 }
1107 release_sock(sk);
1108 }
1109 ulen += sizeof(struct udphdr);
1110
4c9483b2 1111 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1112
1113 if (sin6) {
1114 if (sin6->sin6_port == 0)
1115 return -EINVAL;
1116
1958b856 1117 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1118 daddr = &sin6->sin6_addr;
1119
1120 if (np->sndflow) {
4c9483b2
DM
1121 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1122 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1123 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1124 if (!flowlabel)
1da177e4 1125 return -EINVAL;
1da177e4
LT
1126 }
1127 }
1128
1129 /*
1130 * Otherwise it will be difficult to maintain
1131 * sk->sk_dst_cache.
1132 */
1133 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1134 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1135 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1136
1137 if (addr_len >= sizeof(struct sockaddr_in6) &&
1138 sin6->sin6_scope_id &&
842df073 1139 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1140 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1141 } else {
1142 if (sk->sk_state != TCP_ESTABLISHED)
1143 return -EDESTADDRREQ;
1144
1958b856 1145 fl6.fl6_dport = inet->inet_dport;
efe4208f 1146 daddr = &sk->sk_v6_daddr;
4c9483b2 1147 fl6.flowlabel = np->flow_label;
987905de 1148 connected = 1;
1da177e4
LT
1149 }
1150
4c9483b2
DM
1151 if (!fl6.flowi6_oif)
1152 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1153
4c9483b2
DM
1154 if (!fl6.flowi6_oif)
1155 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1156
4c9483b2 1157 fl6.flowi6_mark = sk->sk_mark;
c14ac945 1158 sockc.tsflags = sk->sk_tsflags;
51953d5b 1159
1da177e4
LT
1160 if (msg->msg_controllen) {
1161 opt = &opt_space;
1162 memset(opt, 0, sizeof(struct ipv6_txoptions));
1163 opt->tot_len = sizeof(*opt);
26879da5 1164 ipc6.opt = opt;
1da177e4 1165
26879da5 1166 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
1da177e4
LT
1167 if (err < 0) {
1168 fl6_sock_release(flowlabel);
1169 return err;
1170 }
4c9483b2
DM
1171 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1172 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1173 if (!flowlabel)
1da177e4
LT
1174 return -EINVAL;
1175 }
1176 if (!(opt->opt_nflen|opt->opt_flen))
1177 opt = NULL;
987905de 1178 connected = 0;
1da177e4 1179 }
45f6fad8
ED
1180 if (!opt) {
1181 opt = txopt_get(np);
1182 opt_to_free = opt;
1183 }
df9890c3
YH
1184 if (flowlabel)
1185 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1186 opt = ipv6_fixup_options(&opt_space, opt);
26879da5 1187 ipc6.opt = opt;
1da177e4 1188
4c9483b2 1189 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1190 if (!ipv6_addr_any(daddr))
4e3fd7a0 1191 fl6.daddr = *daddr;
876c7f41 1192 else
4c9483b2
DM
1193 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1194 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1195 fl6.saddr = np->saddr;
1958b856 1196 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1197
4c9483b2 1198 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1199 if (final_p)
987905de 1200 connected = 0;
1da177e4 1201
4c9483b2
DM
1202 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1203 fl6.flowi6_oif = np->mcast_oif;
987905de 1204 connected = 0;
c4062dfc
EH
1205 } else if (!fl6.flowi6_oif)
1206 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1207
4c9483b2 1208 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1209
0e0d44ab 1210 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1211 if (IS_ERR(dst)) {
1212 err = PTR_ERR(dst);
1213 dst = NULL;
1da177e4 1214 goto out;
14e50e57 1215 }
1da177e4 1216
26879da5
WW
1217 if (ipc6.hlimit < 0)
1218 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4 1219
26879da5
WW
1220 if (ipc6.tclass < 0)
1221 ipc6.tclass = np->tclass;
41a1f8ea 1222
1da177e4
LT
1223 if (msg->msg_flags&MSG_CONFIRM)
1224 goto do_confirm;
1225back_from_confirm:
1226
03485f2a
VY
1227 /* Lockless fast path for the non-corking case */
1228 if (!corkreq) {
1229 struct sk_buff *skb;
1230
1231 skb = ip6_make_skb(sk, getfrag, msg, ulen,
26879da5 1232 sizeof(struct udphdr), &ipc6,
03485f2a 1233 &fl6, (struct rt6_info *)dst,
26879da5 1234 msg->msg_flags, &sockc);
03485f2a
VY
1235 err = PTR_ERR(skb);
1236 if (!IS_ERR_OR_NULL(skb))
1237 err = udp_v6_send_skb(skb, &fl6);
1238 goto release_dst;
1239 }
1240
1da177e4
LT
1241 lock_sock(sk);
1242 if (unlikely(up->pending)) {
1243 /* The socket is already corked while preparing it. */
1244 /* ... which is an evident application bug. --ANK */
1245 release_sock(sk);
1246
ba7a46f1 1247 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1248 err = -EINVAL;
1249 goto out;
1250 }
1251
1252 up->pending = AF_INET6;
1253
1254do_append_data:
26879da5
WW
1255 if (ipc6.dontfrag < 0)
1256 ipc6.dontfrag = np->dontfrag;
1da177e4 1257 up->len += ulen;
26879da5
WW
1258 err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1259 &ipc6, &fl6, (struct rt6_info *)dst,
1260 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, &sockc);
1da177e4
LT
1261 if (err)
1262 udp_v6_flush_pending_frames(sk);
1263 else if (!corkreq)
4c0a6cb0 1264 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1265 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1266 up->pending = 0;
1da177e4 1267
03485f2a
VY
1268 if (err > 0)
1269 err = np->recverr ? net_xmit_errno(err) : 0;
1270 release_sock(sk);
1271
1272release_dst:
a5e7c210
DM
1273 if (dst) {
1274 if (connected) {
1275 ip6_dst_store(sk, dst,
efe4208f
ED
1276 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1277 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1278#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1279 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1280 &np->saddr :
1281#endif
1282 NULL);
a5e7c210
DM
1283 } else {
1284 dst_release(dst);
1285 }
a3c96089 1286 dst = NULL;
a5e7c210
DM
1287 }
1288
1da177e4 1289out:
a3c96089 1290 dst_release(dst);
1da177e4 1291 fl6_sock_release(flowlabel);
45f6fad8 1292 txopt_put(opt_to_free);
cd562c98 1293 if (!err)
1da177e4 1294 return len;
a18135eb
DM
1295 /*
1296 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1297 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1298 * we don't have a good statistic (IpOutDiscards but it can be too many
1299 * things). We could add another new stat but at least for now that
1300 * seems like overkill.
1301 */
1302 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
6aef70a8
ED
1303 UDP6_INC_STATS(sock_net(sk),
1304 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1305 }
1da177e4
LT
1306 return err;
1307
1308do_confirm:
1309 dst_confirm(dst);
1310 if (!(msg->msg_flags&MSG_PROBE) || len)
1311 goto back_from_confirm;
1312 err = 0;
1313 goto out;
1314}
1315
7d06b2e0 1316void udpv6_destroy_sock(struct sock *sk)
1da177e4 1317{
44046a59 1318 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1319 lock_sock(sk);
1320 udp_v6_flush_pending_frames(sk);
1321 release_sock(sk);
1322
44046a59
TP
1323 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1324 void (*encap_destroy)(struct sock *sk);
1325 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1326 if (encap_destroy)
1327 encap_destroy(sk);
1328 }
1329
1da177e4 1330 inet6_destroy_sock(sk);
1da177e4
LT
1331}
1332
1333/*
1334 * Socket option code for UDP
1335 */
ba4e58ec 1336int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1337 char __user *optval, unsigned int optlen)
3fdadf7d 1338{
db8dac20 1339 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1340 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1341 udp_v6_push_pending_frames);
ba4e58ec 1342 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1343}
1344
1345#ifdef CONFIG_COMPAT
ba4e58ec 1346int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1347 char __user *optval, unsigned int optlen)
3fdadf7d 1348{
db8dac20 1349 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1350 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1351 udp_v6_push_pending_frames);
ba4e58ec 1352 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1353}
1354#endif
1355
ba4e58ec
GR
1356int udpv6_getsockopt(struct sock *sk, int level, int optname,
1357 char __user *optval, int __user *optlen)
3fdadf7d 1358{
db8dac20 1359 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1360 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1361 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1362}
1363
1364#ifdef CONFIG_COMPAT
ba4e58ec
GR
1365int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1366 char __user *optval, int __user *optlen)
3fdadf7d 1367{
db8dac20 1368 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1369 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1370 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1371}
1372#endif
1373
41135cc8 1374static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1375 .handler = udpv6_rcv,
1376 .err_handler = udpv6_err,
1377 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1378};
1379
1380/* ------------------------------------------------------------------------ */
1381#ifdef CONFIG_PROC_FS
ba4e58ec 1382int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1383{
17ef66af
LC
1384 if (v == SEQ_START_TOKEN) {
1385 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1386 } else {
1387 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1388 struct inet_sock *inet = inet_sk(v);
1389 __u16 srcp = ntohs(inet->inet_sport);
1390 __u16 destp = ntohs(inet->inet_dport);
1391 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1392 }
1da177e4
LT
1393 return 0;
1394}
1395
73cb88ec
AV
1396static const struct file_operations udp6_afinfo_seq_fops = {
1397 .owner = THIS_MODULE,
1398 .open = udp_seq_open,
1399 .read = seq_read,
1400 .llseek = seq_lseek,
1401 .release = seq_release_net
1402};
1403
1da177e4 1404static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1405 .name = "udp6",
1406 .family = AF_INET6,
645ca708 1407 .udp_table = &udp_table,
73cb88ec 1408 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1409 .seq_ops = {
1410 .show = udp6_seq_show,
1411 },
1da177e4
LT
1412};
1413
2c8c1e72 1414int __net_init udp6_proc_init(struct net *net)
1da177e4 1415{
0c96d8c5 1416 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1417}
1418
ec120da6
IM
1419void udp6_proc_exit(struct net *net)
1420{
0c96d8c5 1421 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1422}
1423#endif /* CONFIG_PROC_FS */
1424
f77d6021
ED
1425void udp_v6_clear_sk(struct sock *sk, int size)
1426{
1427 struct inet_sock *inet = inet_sk(sk);
1428
1429 /* we do not want to clear pinet6 field, because of RCU lookups */
1430 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1431
1432 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1433 memset(&inet->pinet6 + 1, 0, size);
1434}
1435
1da177e4
LT
1436/* ------------------------------------------------------------------------ */
1437
1438struct proto udpv6_prot = {
543d9cfe
ACM
1439 .name = "UDPv6",
1440 .owner = THIS_MODULE,
ba4e58ec 1441 .close = udp_lib_close,
543d9cfe
ACM
1442 .connect = ip6_datagram_connect,
1443 .disconnect = udp_disconnect,
1444 .ioctl = udp_ioctl,
1445 .destroy = udpv6_destroy_sock,
1446 .setsockopt = udpv6_setsockopt,
1447 .getsockopt = udpv6_getsockopt,
1448 .sendmsg = udpv6_sendmsg,
1449 .recvmsg = udpv6_recvmsg,
f7ad74fe 1450 .backlog_rcv = __udpv6_queue_rcv_skb,
e646b657 1451 .release_cb = ip6_datagram_release_cb,
ba4e58ec
GR
1452 .hash = udp_lib_hash,
1453 .unhash = udp_lib_unhash,
719f8358 1454 .rehash = udp_v6_rehash,
543d9cfe 1455 .get_port = udp_v6_get_port,
95766fff
HA
1456 .memory_allocated = &udp_memory_allocated,
1457 .sysctl_mem = sysctl_udp_mem,
1458 .sysctl_wmem = &sysctl_udp_wmem_min,
1459 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1460 .obj_size = sizeof(struct udp6_sock),
271b72c7 1461 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1462 .h.udp_table = &udp_table,
3fdadf7d 1463#ifdef CONFIG_COMPAT
543d9cfe
ACM
1464 .compat_setsockopt = compat_udpv6_setsockopt,
1465 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1466#endif
f77d6021 1467 .clear_sk = udp_v6_clear_sk,
1da177e4
LT
1468};
1469
1da177e4
LT
1470static struct inet_protosw udpv6_protosw = {
1471 .type = SOCK_DGRAM,
1472 .protocol = IPPROTO_UDP,
1473 .prot = &udpv6_prot,
1474 .ops = &inet6_dgram_ops,
1da177e4
LT
1475 .flags = INET_PROTOSW_PERMANENT,
1476};
1477
7f4e4868 1478int __init udpv6_init(void)
1da177e4 1479{
7f4e4868
DL
1480 int ret;
1481
3336288a
VY
1482 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1483 if (ret)
c6b641a4 1484 goto out;
3336288a 1485
7f4e4868
DL
1486 ret = inet6_register_protosw(&udpv6_protosw);
1487 if (ret)
1488 goto out_udpv6_protocol;
1489out:
1490 return ret;
1491
1492out_udpv6_protocol:
1493 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1494 goto out;
1495}
1496
09f7709f 1497void udpv6_exit(void)
7f4e4868
DL
1498{
1499 inet6_unregister_protosw(&udpv6_protosw);
1500 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1501}