net: rename ICMP_INC_STATS_BH()
[linux-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
645ca708
ED
118static inline int compute_score(struct sock *sk, struct net *net,
119 unsigned short hnum,
88440ae7
BS
120 const struct in6_addr *saddr, __be16 sport,
121 const struct in6_addr *daddr, __be16 dport,
645ca708
ED
122 int dif)
123{
60c04aec
JP
124 int score;
125 struct inet_sock *inet;
126
127 if (!net_eq(sock_net(sk), net) ||
128 udp_sk(sk)->udp_port_hash != hnum ||
129 sk->sk_family != PF_INET6)
130 return -1;
131
132 score = 0;
133 inet = inet_sk(sk);
134
135 if (inet->inet_dport) {
136 if (inet->inet_dport != sport)
137 return -1;
138 score++;
139 }
140
141 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
142 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
143 return -1;
144 score++;
145 }
146
147 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
148 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
149 return -1;
150 score++;
151 }
152
153 if (sk->sk_bound_dev_if) {
154 if (sk->sk_bound_dev_if != dif)
155 return -1;
156 score++;
645ca708 157 }
60c04aec 158
70da268b
ED
159 if (sk->sk_incoming_cpu == raw_smp_processor_id())
160 score++;
161
645ca708
ED
162 return score;
163}
164
fddc17de 165static inline int compute_score2(struct sock *sk, struct net *net,
60c04aec
JP
166 const struct in6_addr *saddr, __be16 sport,
167 const struct in6_addr *daddr,
168 unsigned short hnum, int dif)
fddc17de 169{
60c04aec
JP
170 int score;
171 struct inet_sock *inet;
fddc17de 172
60c04aec
JP
173 if (!net_eq(sock_net(sk), net) ||
174 udp_sk(sk)->udp_port_hash != hnum ||
175 sk->sk_family != PF_INET6)
176 return -1;
fddc17de 177
60c04aec
JP
178 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
179 return -1;
180
181 score = 0;
182 inet = inet_sk(sk);
183
184 if (inet->inet_dport) {
185 if (inet->inet_dport != sport)
186 return -1;
187 score++;
188 }
189
190 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
191 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
fddc17de 192 return -1;
60c04aec 193 score++;
fddc17de 194 }
60c04aec
JP
195
196 if (sk->sk_bound_dev_if) {
197 if (sk->sk_bound_dev_if != dif)
198 return -1;
199 score++;
200 }
201
70da268b
ED
202 if (sk->sk_incoming_cpu == raw_smp_processor_id())
203 score++;
204
fddc17de
ED
205 return score;
206}
207
fddc17de
ED
208/* called with read_rcu_lock() */
209static struct sock *udp6_lib_lookup2(struct net *net,
210 const struct in6_addr *saddr, __be16 sport,
211 const struct in6_addr *daddr, unsigned int hnum, int dif,
1134158b
CG
212 struct udp_hslot *hslot2, unsigned int slot2,
213 struct sk_buff *skb)
fddc17de
ED
214{
215 struct sock *sk, *result;
72289b96
TH
216 int score, badness, matches = 0, reuseport = 0;
217 u32 hash = 0;
fddc17de 218
fddc17de
ED
219 result = NULL;
220 badness = -1;
ca065d0c 221 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
fddc17de
ED
222 score = compute_score2(sk, net, saddr, sport,
223 daddr, hnum, dif);
224 if (score > badness) {
72289b96
TH
225 reuseport = sk->sk_reuseport;
226 if (reuseport) {
b50026b5
HFS
227 hash = udp6_ehashfn(net, daddr, hnum,
228 saddr, sport);
ed0dfffd 229
ca065d0c 230 result = reuseport_select_sock(sk, hash, skb,
ed0dfffd 231 sizeof(struct udphdr));
ca065d0c
ED
232 if (result)
233 return result;
72289b96 234 matches = 1;
70da268b 235 }
ca065d0c
ED
236 result = sk;
237 badness = score;
72289b96
TH
238 } else if (score == badness && reuseport) {
239 matches++;
8fc54f68 240 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
241 result = sk;
242 hash = next_pseudo_random32(hash);
fddc17de
ED
243 }
244 }
fddc17de
ED
245 return result;
246}
247
ca065d0c 248/* rcu_read_lock() must be held */
fce82338 249struct sock *__udp6_lib_lookup(struct net *net,
88440ae7
BS
250 const struct in6_addr *saddr, __be16 sport,
251 const struct in6_addr *daddr, __be16 dport,
538950a1
CG
252 int dif, struct udp_table *udptable,
253 struct sk_buff *skb)
1da177e4 254{
271b72c7 255 struct sock *sk, *result;
1da177e4 256 unsigned short hnum = ntohs(dport);
fddc17de
ED
257 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
258 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
72289b96
TH
259 int score, badness, matches = 0, reuseport = 0;
260 u32 hash = 0;
645ca708 261
fddc17de
ED
262 if (hslot->count > 10) {
263 hash2 = udp6_portaddr_hash(net, daddr, hnum);
264 slot2 = hash2 & udptable->mask;
265 hslot2 = &udptable->hash2[slot2];
266 if (hslot->count < hslot2->count)
267 goto begin;
268
269 result = udp6_lib_lookup2(net, saddr, sport,
270 daddr, hnum, dif,
1134158b 271 hslot2, slot2, skb);
fddc17de
ED
272 if (!result) {
273 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
274 slot2 = hash2 & udptable->mask;
275 hslot2 = &udptable->hash2[slot2];
276 if (hslot->count < hslot2->count)
277 goto begin;
278
1223c67c
JBD
279 result = udp6_lib_lookup2(net, saddr, sport,
280 &in6addr_any, hnum, dif,
1134158b 281 hslot2, slot2, skb);
fddc17de 282 }
fddc17de
ED
283 return result;
284 }
271b72c7
ED
285begin:
286 result = NULL;
287 badness = -1;
ca065d0c 288 sk_for_each_rcu(sk, &hslot->head) {
645ca708
ED
289 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
290 if (score > badness) {
72289b96
TH
291 reuseport = sk->sk_reuseport;
292 if (reuseport) {
b50026b5
HFS
293 hash = udp6_ehashfn(net, daddr, hnum,
294 saddr, sport);
ca065d0c 295 result = reuseport_select_sock(sk, hash, skb,
538950a1 296 sizeof(struct udphdr));
ca065d0c
ED
297 if (result)
298 return result;
72289b96
TH
299 matches = 1;
300 }
ca065d0c
ED
301 result = sk;
302 badness = score;
72289b96
TH
303 } else if (score == badness && reuseport) {
304 matches++;
8fc54f68 305 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
306 result = sk;
307 hash = next_pseudo_random32(hash);
1da177e4
LT
308 }
309 }
1da177e4
LT
310 return result;
311}
fce82338 312EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
1da177e4 313
607c4aaf
KK
314static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
315 __be16 sport, __be16 dport,
645ca708 316 struct udp_table *udptable)
607c4aaf 317{
23542618 318 struct sock *sk;
b71d1d42 319 const struct ipv6hdr *iph = ipv6_hdr(skb);
607c4aaf 320
e5d08d71
IM
321 sk = skb_steal_sock(skb);
322 if (unlikely(sk))
23542618 323 return sk;
adf30907
ED
324 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
325 &iph->daddr, dport, inet6_iif(skb),
538950a1 326 udptable, skb);
607c4aaf
KK
327}
328
63058308
TH
329struct sock *udp6_lib_lookup_skb(struct sk_buff *skb,
330 __be16 sport, __be16 dport)
331{
332 const struct ipv6hdr *iph = ipv6_hdr(skb);
333 const struct net_device *dev =
334 skb_dst(skb) ? skb_dst(skb)->dev : skb->dev;
335
336 return __udp6_lib_lookup(dev_net(dev), &iph->saddr, sport,
337 &iph->daddr, dport, inet6_iif(skb),
338 &udp_table, skb);
339}
340EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb);
341
ca065d0c
ED
342/* Must be called under rcu_read_lock().
343 * Does increment socket refcount.
344 */
345#if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
346 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
aa976fc0
BS
347struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
348 const struct in6_addr *daddr, __be16 dport, int dif)
349{
ca065d0c
ED
350 struct sock *sk;
351
352 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport,
353 dif, &udp_table, NULL);
354 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
355 sk = NULL;
356 return sk;
aa976fc0
BS
357}
358EXPORT_SYMBOL_GPL(udp6_lib_lookup);
ca065d0c 359#endif
aa976fc0 360
1da177e4 361/*
67ba4152
IM
362 * This should be easy, if there is something there we
363 * return it, otherwise we block.
1da177e4
LT
364 */
365
1b784140 366int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
1da177e4
LT
367 int noblock, int flags, int *addr_len)
368{
369 struct ipv6_pinfo *np = inet6_sk(sk);
370 struct inet_sock *inet = inet_sk(sk);
1ab1457c 371 struct sk_buff *skb;
59c2cdae 372 unsigned int ulen, copied;
627d2d6b 373 int peeked, peeking, off;
759e5d00
HX
374 int err;
375 int is_udplite = IS_UDPLITE(sk);
197c949e 376 bool checksum_valid = false;
f26ba175 377 int is_udp4;
8a74ad60 378 bool slow;
1da177e4 379
1da177e4 380 if (flags & MSG_ERRQUEUE)
85fbaa75 381 return ipv6_recv_error(sk, msg, len, addr_len);
1da177e4 382
4b340ae2 383 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
85fbaa75 384 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
4b340ae2 385
1da177e4 386try_again:
627d2d6b 387 peeking = off = sk_peek_offset(sk, flags);
a59322be 388 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 389 &peeked, &off, &err);
1da177e4 390 if (!skb)
627d2d6b 391 return err;
1da177e4 392
e6afc8ac 393 ulen = skb->len;
59c2cdae 394 copied = len;
627d2d6b 395 if (copied > ulen - off)
396 copied = ulen - off;
59c2cdae 397 else if (copied < ulen)
1ab1457c 398 msg->msg_flags |= MSG_TRUNC;
1da177e4 399
f26ba175
WY
400 is_udp4 = (skb->protocol == htons(ETH_P_IP));
401
ba4e58ec 402 /*
759e5d00
HX
403 * If checksum is needed at all, try to do it while copying the
404 * data. If the data is truncated, or if we only want a partial
405 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 406 */
ba4e58ec 407
627d2d6b 408 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
197c949e
ED
409 checksum_valid = !udp_lib_checksum_complete(skb);
410 if (!checksum_valid)
1da177e4 411 goto csum_copy_err;
ba4e58ec
GR
412 }
413
197c949e 414 if (checksum_valid || skb_csum_unnecessary(skb))
627d2d6b 415 err = skb_copy_datagram_msg(skb, off, msg, copied);
ba4e58ec 416 else {
627d2d6b 417 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
1da177e4
LT
418 if (err == -EINVAL)
419 goto csum_copy_err;
420 }
22911fc5
ED
421 if (unlikely(err)) {
422 trace_kfree_skb(skb, udpv6_recvmsg);
979402b1
ED
423 if (!peeked) {
424 atomic_inc(&sk->sk_drops);
425 if (is_udp4)
6aef70a8
ED
426 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
427 is_udplite);
979402b1 428 else
6aef70a8
ED
429 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
430 is_udplite);
979402b1 431 }
627d2d6b 432 skb_free_datagram_locked(sk, skb);
433 return err;
22911fc5 434 }
f26ba175
WY
435 if (!peeked) {
436 if (is_udp4)
6aef70a8
ED
437 UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
438 is_udplite);
f26ba175 439 else
6aef70a8
ED
440 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
441 is_udplite);
f26ba175 442 }
cb75994e 443
3b885787 444 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
445
446 /* Copy the address. */
447 if (msg->msg_name) {
342dfc30 448 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 449 sin6->sin6_family = AF_INET6;
4bedb452 450 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 451 sin6->sin6_flowinfo = 0;
1da177e4 452
842df073 453 if (is_udp4) {
b301e82c
BH
454 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
455 &sin6->sin6_addr);
842df073
HFS
456 sin6->sin6_scope_id = 0;
457 } else {
4e3fd7a0 458 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
459 sin6->sin6_scope_id =
460 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 461 inet6_iif(skb));
1da177e4 462 }
bceaa902 463 *addr_len = sizeof(*sin6);
1da177e4 464 }
4b261c75
HFS
465
466 if (np->rxopt.all)
467 ip6_datagram_recv_common_ctl(sk, msg, skb);
468
f26ba175 469 if (is_udp4) {
1da177e4
LT
470 if (inet->cmsg_flags)
471 ip_cmsg_recv(msg, skb);
472 } else {
473 if (np->rxopt.all)
4b261c75 474 ip6_datagram_recv_specific_ctl(sk, msg, skb);
1ab1457c 475 }
1da177e4 476
59c2cdae 477 err = copied;
1da177e4 478 if (flags & MSG_TRUNC)
759e5d00 479 err = ulen;
1da177e4 480
627d2d6b 481 __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
1da177e4
LT
482 return err;
483
484csum_copy_err:
8a74ad60 485 slow = lock_sock_fast(sk);
0856f939 486 if (!skb_kill_datagram(sk, skb, flags)) {
6a5dc9e5 487 if (is_udp4) {
6aef70a8
ED
488 UDP_INC_STATS(sock_net(sk),
489 UDP_MIB_CSUMERRORS, is_udplite);
490 UDP_INC_STATS(sock_net(sk),
491 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 492 } else {
6aef70a8
ED
493 UDP6_INC_STATS(sock_net(sk),
494 UDP_MIB_CSUMERRORS, is_udplite);
495 UDP6_INC_STATS(sock_net(sk),
496 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 497 }
0856f939 498 }
8a74ad60 499 unlock_sock_fast(sk, slow);
1da177e4 500
beb39db5
ED
501 /* starting over for a new packet, but check if we need to yield */
502 cond_resched();
9cfaa8de 503 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
504 goto try_again;
505}
506
ba4e58ec 507void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 508 u8 type, u8 code, int offset, __be32 info,
645ca708 509 struct udp_table *udptable)
1da177e4
LT
510{
511 struct ipv6_pinfo *np;
b71d1d42
ED
512 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
513 const struct in6_addr *saddr = &hdr->saddr;
514 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 515 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4 516 struct sock *sk;
e0d8c1b7 517 int harderr;
1da177e4 518 int err;
7304fe46 519 struct net *net = dev_net(skb->dev);
1da177e4 520
e32ea7e7 521 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
538950a1 522 inet6_iif(skb), udptable, skb);
63159f29 523 if (!sk) {
7304fe46
DJ
524 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
525 ICMP6_MIB_INERRORS);
1da177e4 526 return;
7304fe46 527 }
1da177e4 528
e0d8c1b7
WW
529 harderr = icmpv6_err_convert(type, code, &err);
530 np = inet6_sk(sk);
531
93b36cf3
HFS
532 if (type == ICMPV6_PKT_TOOBIG) {
533 if (!ip6_sk_accept_pmtu(sk))
534 goto out;
81aded24 535 ip6_sk_update_pmtu(skb, sk, info);
e0d8c1b7
WW
536 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
537 harderr = 1;
93b36cf3 538 }
1a462d18 539 if (type == NDISC_REDIRECT) {
ec18d9a2 540 ip6_sk_redirect(skb, sk);
1a462d18
DJ
541 goto out;
542 }
81aded24 543
e0d8c1b7
WW
544 if (!np->recverr) {
545 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
546 goto out;
547 } else {
1da177e4 548 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
e0d8c1b7 549 }
b1faf566 550
1da177e4
LT
551 sk->sk_err = err;
552 sk->sk_error_report(sk);
553out:
ca065d0c 554 return;
1da177e4
LT
555}
556
f7ad74fe
BL
557static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
558{
559 int rc;
560
efe4208f 561 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 562 sock_rps_save_rxhash(sk, skb);
005ec974 563 sk_mark_napi_id(sk, skb);
2c8c56e1 564 sk_incoming_cpu_update(sk);
005ec974 565 }
f7ad74fe 566
e6afc8ac 567 rc = __sock_queue_rcv_skb(sk, skb);
f7ad74fe
BL
568 if (rc < 0) {
569 int is_udplite = IS_UDPLITE(sk);
570
571 /* Note that an ENOMEM error is charged twice */
572 if (rc == -ENOMEM)
573 UDP6_INC_STATS_BH(sock_net(sk),
574 UDP_MIB_RCVBUFERRORS, is_udplite);
575 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
576 kfree_skb(skb);
577 return -1;
578 }
579 return 0;
580}
581
ba4e58ec 582static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 583 struct inet6_skb_parm *opt, u8 type,
67ba4152 584 u8 code, int offset, __be32 info)
ba4e58ec 585{
645ca708 586 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
587}
588
d7f3f621
BL
589static struct static_key udpv6_encap_needed __read_mostly;
590void udpv6_encap_enable(void)
591{
592 if (!static_key_enabled(&udpv6_encap_needed))
593 static_key_slow_inc(&udpv6_encap_needed);
594}
595EXPORT_SYMBOL(udpv6_encap_enable);
596
f7ad74fe 597int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 598{
ba4e58ec 599 struct udp_sock *up = udp_sk(sk);
a18135eb 600 int rc;
b2bf1e26 601 int is_udplite = IS_UDPLITE(sk);
a18135eb 602
ba4e58ec
GR
603 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
604 goto drop;
1da177e4 605
d7f3f621
BL
606 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
607 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
608
609 /*
610 * This is an encapsulation socket so pass the skb to
611 * the socket's udp_encap_rcv() hook. Otherwise, just
612 * fall through and pass this up the UDP socket.
613 * up->encap_rcv() returns the following value:
614 * =0 if skb was successfully passed to the encap
615 * handler or was discarded by it.
616 * >0 if skb should be passed on to UDP.
617 * <0 if skb should be resubmitted as proto -N
618 */
619
620 /* if we're overly short, let UDP handle it */
621 encap_rcv = ACCESS_ONCE(up->encap_rcv);
53b24b8f 622 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
d7f3f621
BL
623 int ret;
624
0a80966b
TH
625 /* Verify checksum before giving to encap */
626 if (udp_lib_checksum_complete(skb))
627 goto csum_error;
628
d7f3f621
BL
629 ret = encap_rcv(sk, skb);
630 if (ret <= 0) {
631 UDP_INC_STATS_BH(sock_net(sk),
632 UDP_MIB_INDATAGRAMS,
633 is_udplite);
634 return -ret;
635 }
636 }
637
638 /* FALLTHROUGH -- it's a UDP Packet */
639 }
640
ba4e58ec
GR
641 /*
642 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
643 */
b2bf1e26 644 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
645
646 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
647 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
648 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
649 goto drop;
650 }
651 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
652 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
653 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
654 goto drop;
655 }
1da177e4
LT
656 }
657
33d480ce 658 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62 659 if (udp_lib_checksum_complete(skb))
6a5dc9e5 660 goto csum_error;
e6afc8ac 661 if (sk_filter(sk, skb))
662 goto drop;
1ab6eb62 663 }
ba4e58ec 664
e6afc8ac 665 udp_csum_pull_header(skb);
274f482d 666 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
3e215c8d
JL
667 UDP6_INC_STATS_BH(sock_net(sk),
668 UDP_MIB_RCVBUFERRORS, is_udplite);
cb80ef46 669 goto drop;
3e215c8d 670 }
cb80ef46 671
d826eb14 672 skb_dst_drop(skb);
cb75994e 673
cb80ef46
BL
674 bh_lock_sock(sk);
675 rc = 0;
676 if (!sock_owned_by_user(sk))
677 rc = __udpv6_queue_rcv_skb(sk, skb);
678 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
679 bh_unlock_sock(sk);
680 goto drop;
681 }
682 bh_unlock_sock(sk);
f7ad74fe
BL
683
684 return rc;
3e215c8d 685
6a5dc9e5
ED
686csum_error:
687 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 688drop:
ef28d1a2 689 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 690 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
691 kfree_skb(skb);
692 return -1;
1da177e4
LT
693}
694
5cf3d461
DH
695static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
696 __be16 loc_port, const struct in6_addr *loc_addr,
697 __be16 rmt_port, const struct in6_addr *rmt_addr,
698 int dif, unsigned short hnum)
1da177e4 699{
5cf3d461 700 struct inet_sock *inet = inet_sk(sk);
1da177e4 701
5cf3d461
DH
702 if (!net_eq(sock_net(sk), net))
703 return false;
704
705 if (udp_sk(sk)->udp_port_hash != hnum ||
706 sk->sk_family != PF_INET6 ||
707 (inet->inet_dport && inet->inet_dport != rmt_port) ||
708 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
709 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
33b4b015
HR
710 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
711 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
712 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
5cf3d461
DH
713 return false;
714 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
715 return false;
716 return true;
1da177e4
LT
717}
718
4068579e
TH
719static void udp6_csum_zero_error(struct sk_buff *skb)
720{
721 /* RFC 2460 section 8.1 says that we SHOULD log
722 * this error. Well, it is reasonable.
723 */
ba7a46f1
JP
724 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
725 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
726 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
727}
728
1da177e4
LT
729/*
730 * Note: called only from the BH handler context,
731 * so we don't need to lock the hashes.
732 */
e3163493 733static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 734 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 735 struct udp_table *udptable, int proto)
1da177e4 736{
ca065d0c 737 struct sock *sk, *first = NULL;
4bedb452 738 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
739 unsigned short hnum = ntohs(uh->dest);
740 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
ca065d0c 741 unsigned int offset = offsetof(typeof(*sk), sk_node);
2dc41cff 742 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
ca065d0c
ED
743 int dif = inet6_iif(skb);
744 struct hlist_node *node;
745 struct sk_buff *nskb;
2dc41cff
DH
746
747 if (use_hash2) {
748 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
749 udp_table.mask;
750 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
751start_lookup:
752 hslot = &udp_table.hash2[hash2];
753 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
754 }
1da177e4 755
ca065d0c
ED
756 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
757 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
758 uh->source, saddr, dif, hnum))
759 continue;
760 /* If zero checksum and no_check is not on for
761 * the socket then skip it.
762 */
763 if (!uh->check && !udp_sk(sk)->no_check6_rx)
764 continue;
765 if (!first) {
766 first = sk;
767 continue;
768 }
769 nskb = skb_clone(skb, GFP_ATOMIC);
770 if (unlikely(!nskb)) {
771 atomic_inc(&sk->sk_drops);
772 UDP6_INC_STATS_BH(net, UDP_MIB_RCVBUFERRORS,
773 IS_UDPLITE(sk));
774 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS,
775 IS_UDPLITE(sk));
776 continue;
95766fff 777 }
a1ab77f9 778
ca065d0c
ED
779 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
780 consume_skb(nskb);
781 }
a1ab77f9 782
2dc41cff
DH
783 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
784 if (use_hash2 && hash2 != hash2_any) {
785 hash2 = hash2_any;
786 goto start_lookup;
787 }
788
ca065d0c
ED
789 if (first) {
790 if (udpv6_queue_rcv_skb(first, skb) > 0)
791 consume_skb(skb);
a1ab77f9 792 } else {
ca065d0c
ED
793 kfree_skb(skb);
794 UDP6_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
795 proto == IPPROTO_UDPLITE);
a1ab77f9 796 }
ba4e58ec 797 return 0;
1da177e4
LT
798}
799
645ca708 800int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 801 int proto)
1da177e4 802{
ca065d0c 803 const struct in6_addr *saddr, *daddr;
3ffe533c 804 struct net *net = dev_net(skb->dev);
1ab1457c 805 struct udphdr *uh;
ca065d0c 806 struct sock *sk;
1da177e4
LT
807 u32 ulen = 0;
808
809 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 810 goto discard;
1da177e4 811
0660e03f
ACM
812 saddr = &ipv6_hdr(skb)->saddr;
813 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 814 uh = udp_hdr(skb);
1da177e4
LT
815
816 ulen = ntohs(uh->len);
ba4e58ec
GR
817 if (ulen > skb->len)
818 goto short_packet;
1da177e4 819
759e5d00
HX
820 if (proto == IPPROTO_UDP) {
821 /* UDP validates ulen. */
1da177e4 822
ba4e58ec
GR
823 /* Check for jumbo payload */
824 if (ulen == 0)
825 ulen = skb->len;
1da177e4 826
ba4e58ec
GR
827 if (ulen < sizeof(*uh))
828 goto short_packet;
1da177e4 829
ba4e58ec
GR
830 if (ulen < skb->len) {
831 if (pskb_trim_rcsum(skb, ulen))
832 goto short_packet;
0660e03f
ACM
833 saddr = &ipv6_hdr(skb)->saddr;
834 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 835 uh = udp_hdr(skb);
ba4e58ec 836 }
ba4e58ec 837 }
1da177e4 838
759e5d00 839 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 840 goto csum_error;
759e5d00 841
1ab1457c
YH
842 /*
843 * Multicast receive code
1da177e4 844 */
ba4e58ec 845 if (ipv6_addr_is_multicast(daddr))
e3163493 846 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 847 saddr, daddr, udptable, proto);
1da177e4
LT
848
849 /* Unicast */
850
1ab1457c 851 /*
1da177e4
LT
852 * check socket cache ... must talk to Alan about his plans
853 * for sock caches... i'll skip this for now.
854 */
607c4aaf 855 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
53b24b8f 856 if (sk) {
a5b50476
ET
857 int ret;
858
1c19448c 859 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
4068579e
TH
860 udp6_csum_zero_error(skb);
861 goto csum_error;
862 }
863
224d019c 864 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
2abb7cdc
TH
865 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
866 ip6_compute_pseudo);
867
a5b50476 868 ret = udpv6_queue_rcv_skb(sk, skb);
1da177e4 869
59dca1d8 870 /* a return value > 0 means to resubmit the input */
cb80ef46 871 if (ret > 0)
59dca1d8 872 return ret;
1da177e4 873
add459aa 874 return 0;
1da177e4 875 }
1ab1457c 876
4068579e
TH
877 if (!uh->check) {
878 udp6_csum_zero_error(skb);
879 goto csum_error;
880 }
881
cb80ef46 882 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 883 goto discard;
cb80ef46
BL
884
885 if (udp_lib_checksum_complete(skb))
6a5dc9e5 886 goto csum_error;
cb80ef46
BL
887
888 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
889 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
890
891 kfree_skb(skb);
add459aa 892 return 0;
1da177e4 893
1ab1457c 894short_packet:
ba7a46f1
JP
895 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
896 proto == IPPROTO_UDPLITE ? "-Lite" : "",
897 saddr, ntohs(uh->source),
898 ulen, skb->len,
899 daddr, ntohs(uh->dest));
6a5dc9e5
ED
900 goto discard;
901csum_error:
902 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 903discard:
ef28d1a2 904 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 905 kfree_skb(skb);
add459aa 906 return 0;
1da177e4 907}
ba4e58ec 908
e5bbef20 909static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 910{
645ca708 911 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
912}
913
1da177e4
LT
914/*
915 * Throw away all pending data and cancel the corking. Socket is locked.
916 */
917static void udp_v6_flush_pending_frames(struct sock *sk)
918{
919 struct udp_sock *up = udp_sk(sk);
920
36d926b9
DL
921 if (up->pending == AF_INET)
922 udp_flush_pending_frames(sk);
923 else if (up->pending) {
1da177e4
LT
924 up->len = 0;
925 up->pending = 0;
926 ip6_flush_pending_frames(sk);
1ab1457c 927 }
1da177e4
LT
928}
929
493c6be3 930/**
67ba4152
IM
931 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
932 * @sk: socket we are sending on
933 * @skb: sk_buff containing the filled-in UDP header
934 * (checksum field must be zeroed out)
493c6be3
SS
935 */
936static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
937 const struct in6_addr *saddr,
938 const struct in6_addr *daddr, int len)
939{
940 unsigned int offset;
941 struct udphdr *uh = udp_hdr(skb);
d39d938c 942 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
493c6be3
SS
943 __wsum csum = 0;
944
d39d938c 945 if (!frags) {
493c6be3
SS
946 /* Only one fragment on the socket. */
947 skb->csum_start = skb_transport_header(skb) - skb->head;
948 skb->csum_offset = offsetof(struct udphdr, check);
949 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
950 } else {
951 /*
952 * HW-checksum won't work as there are two or more
953 * fragments on the socket so that all csums of sk_buffs
954 * should be together
955 */
956 offset = skb_transport_offset(skb);
957 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
958
959 skb->ip_summed = CHECKSUM_NONE;
960
d39d938c
VY
961 do {
962 csum = csum_add(csum, frags->csum);
963 } while ((frags = frags->next));
493c6be3
SS
964
965 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
966 csum);
967 if (uh->check == 0)
968 uh->check = CSUM_MANGLED_0;
969 }
970}
971
1da177e4
LT
972/*
973 * Sending
974 */
975
d39d938c 976static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1da177e4 977{
d39d938c 978 struct sock *sk = skb->sk;
1da177e4 979 struct udphdr *uh;
1da177e4 980 int err = 0;
b2bf1e26 981 int is_udplite = IS_UDPLITE(sk);
868c86bc 982 __wsum csum = 0;
d39d938c
VY
983 int offset = skb_transport_offset(skb);
984 int len = skb->len - offset;
1da177e4
LT
985
986 /*
987 * Create a UDP header
988 */
4bedb452 989 uh = udp_hdr(skb);
1958b856
DM
990 uh->source = fl6->fl6_sport;
991 uh->dest = fl6->fl6_dport;
d39d938c 992 uh->len = htons(len);
1da177e4
LT
993 uh->check = 0;
994
b2bf1e26 995 if (is_udplite)
d39d938c
VY
996 csum = udplite_csum(skb);
997 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
998 skb->ip_summed = CHECKSUM_NONE;
999 goto send;
1000 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
d39d938c 1001 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
493c6be3
SS
1002 goto send;
1003 } else
d39d938c 1004 csum = udp_csum(skb);
1da177e4 1005
ba4e58ec 1006 /* add protocol-dependent pseudo-header */
4c9483b2 1007 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
d39d938c 1008 len, fl6->flowi6_proto, csum);
1da177e4 1009 if (uh->check == 0)
f6ab0288 1010 uh->check = CSUM_MANGLED_0;
1da177e4 1011
493c6be3 1012send:
d39d938c 1013 err = ip6_send_skb(skb);
6ce9e7b5
ED
1014 if (err) {
1015 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
6aef70a8
ED
1016 UDP6_INC_STATS(sock_net(sk),
1017 UDP_MIB_SNDBUFERRORS, is_udplite);
6ce9e7b5
ED
1018 err = 0;
1019 }
6aef70a8
ED
1020 } else {
1021 UDP6_INC_STATS(sock_net(sk),
1022 UDP_MIB_OUTDATAGRAMS, is_udplite);
1023 }
d39d938c
VY
1024 return err;
1025}
1026
1027static int udp_v6_push_pending_frames(struct sock *sk)
1028{
1029 struct sk_buff *skb;
1030 struct udp_sock *up = udp_sk(sk);
1031 struct flowi6 fl6;
1032 int err = 0;
1033
1034 if (up->pending == AF_INET)
1035 return udp_push_pending_frames(sk);
1036
1037 /* ip6_finish_skb will release the cork, so make a copy of
1038 * fl6 here.
1039 */
1040 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1041
1042 skb = ip6_finish_skb(sk);
1043 if (!skb)
1044 goto out;
1045
1046 err = udp_v6_send_skb(skb, &fl6);
1047
1da177e4
LT
1048out:
1049 up->len = 0;
1050 up->pending = 0;
1051 return err;
1052}
1053
1b784140 1054int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1da177e4
LT
1055{
1056 struct ipv6_txoptions opt_space;
1057 struct udp_sock *up = udp_sk(sk);
1058 struct inet_sock *inet = inet_sk(sk);
1059 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1060 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1061 struct in6_addr *daddr, *final_p, final;
1da177e4 1062 struct ipv6_txoptions *opt = NULL;
45f6fad8 1063 struct ipv6_txoptions *opt_to_free = NULL;
1da177e4 1064 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1065 struct flowi6 fl6;
1da177e4
LT
1066 struct dst_entry *dst;
1067 int addr_len = msg->msg_namelen;
1068 int ulen = len;
1069 int hlimit = -1;
41a1f8ea 1070 int tclass = -1;
13b52cd4 1071 int dontfrag = -1;
1da177e4
LT
1072 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1073 int err;
987905de 1074 int connected = 0;
b2bf1e26 1075 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1076 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
ad1e46a8 1077 struct sockcm_cookie sockc;
1da177e4
LT
1078
1079 /* destination address check */
1080 if (sin6) {
1081 if (addr_len < offsetof(struct sockaddr, sa_data))
1082 return -EINVAL;
1083
1084 switch (sin6->sin6_family) {
1085 case AF_INET6:
1086 if (addr_len < SIN6_LEN_RFC2133)
1087 return -EINVAL;
1088 daddr = &sin6->sin6_addr;
1089 break;
1090 case AF_INET:
1091 goto do_udp_sendmsg;
1092 case AF_UNSPEC:
1093 msg->msg_name = sin6 = NULL;
1094 msg->msg_namelen = addr_len = 0;
1095 daddr = NULL;
1096 break;
1097 default:
1098 return -EINVAL;
1099 }
1100 } else if (!up->pending) {
1101 if (sk->sk_state != TCP_ESTABLISHED)
1102 return -EDESTADDRREQ;
efe4208f 1103 daddr = &sk->sk_v6_daddr;
1ab1457c 1104 } else
1da177e4
LT
1105 daddr = NULL;
1106
1107 if (daddr) {
e773e4fa 1108 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1109 struct sockaddr_in sin;
1110 sin.sin_family = AF_INET;
c720c7e8 1111 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1112 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1113 msg->msg_name = &sin;
1114 msg->msg_namelen = sizeof(sin);
1115do_udp_sendmsg:
1116 if (__ipv6_only_sock(sk))
1117 return -ENETUNREACH;
1b784140 1118 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1119 }
1120 }
1121
1122 if (up->pending == AF_INET)
1b784140 1123 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1124
1125 /* Rough check on arithmetic overflow,
b59e139b 1126 better check is made in ip6_append_data().
1da177e4
LT
1127 */
1128 if (len > INT_MAX - sizeof(struct udphdr))
1129 return -EMSGSIZE;
1ab1457c 1130
03485f2a 1131 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1da177e4
LT
1132 if (up->pending) {
1133 /*
1134 * There are pending frames.
1135 * The socket lock must be held while it's corked.
1136 */
1137 lock_sock(sk);
1138 if (likely(up->pending)) {
1139 if (unlikely(up->pending != AF_INET6)) {
1140 release_sock(sk);
1141 return -EAFNOSUPPORT;
1142 }
1143 dst = NULL;
1144 goto do_append_data;
1145 }
1146 release_sock(sk);
1147 }
1148 ulen += sizeof(struct udphdr);
1149
4c9483b2 1150 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1151
1152 if (sin6) {
1153 if (sin6->sin6_port == 0)
1154 return -EINVAL;
1155
1958b856 1156 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1157 daddr = &sin6->sin6_addr;
1158
1159 if (np->sndflow) {
4c9483b2
DM
1160 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1161 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1162 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1163 if (!flowlabel)
1da177e4 1164 return -EINVAL;
1da177e4
LT
1165 }
1166 }
1167
1168 /*
1169 * Otherwise it will be difficult to maintain
1170 * sk->sk_dst_cache.
1171 */
1172 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1173 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1174 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1175
1176 if (addr_len >= sizeof(struct sockaddr_in6) &&
1177 sin6->sin6_scope_id &&
842df073 1178 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1179 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1180 } else {
1181 if (sk->sk_state != TCP_ESTABLISHED)
1182 return -EDESTADDRREQ;
1183
1958b856 1184 fl6.fl6_dport = inet->inet_dport;
efe4208f 1185 daddr = &sk->sk_v6_daddr;
4c9483b2 1186 fl6.flowlabel = np->flow_label;
987905de 1187 connected = 1;
1da177e4
LT
1188 }
1189
4c9483b2
DM
1190 if (!fl6.flowi6_oif)
1191 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1192
4c9483b2
DM
1193 if (!fl6.flowi6_oif)
1194 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1195
4c9483b2 1196 fl6.flowi6_mark = sk->sk_mark;
c14ac945 1197 sockc.tsflags = sk->sk_tsflags;
51953d5b 1198
1da177e4
LT
1199 if (msg->msg_controllen) {
1200 opt = &opt_space;
1201 memset(opt, 0, sizeof(struct ipv6_txoptions));
1202 opt->tot_len = sizeof(*opt);
1203
73df66f8 1204 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
ad1e46a8
SHY
1205 &hlimit, &tclass, &dontfrag,
1206 &sockc);
1da177e4
LT
1207 if (err < 0) {
1208 fl6_sock_release(flowlabel);
1209 return err;
1210 }
4c9483b2
DM
1211 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1212 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1213 if (!flowlabel)
1da177e4
LT
1214 return -EINVAL;
1215 }
1216 if (!(opt->opt_nflen|opt->opt_flen))
1217 opt = NULL;
987905de 1218 connected = 0;
1da177e4 1219 }
45f6fad8
ED
1220 if (!opt) {
1221 opt = txopt_get(np);
1222 opt_to_free = opt;
1223 }
df9890c3
YH
1224 if (flowlabel)
1225 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1226 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1227
4c9483b2 1228 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1229 if (!ipv6_addr_any(daddr))
4e3fd7a0 1230 fl6.daddr = *daddr;
876c7f41 1231 else
4c9483b2
DM
1232 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1233 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1234 fl6.saddr = np->saddr;
1958b856 1235 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1236
4c9483b2 1237 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1238 if (final_p)
987905de 1239 connected = 0;
1da177e4 1240
4c9483b2
DM
1241 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1242 fl6.flowi6_oif = np->mcast_oif;
987905de 1243 connected = 0;
c4062dfc
EH
1244 } else if (!fl6.flowi6_oif)
1245 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1246
4c9483b2 1247 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1248
0e0d44ab 1249 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1250 if (IS_ERR(dst)) {
1251 err = PTR_ERR(dst);
1252 dst = NULL;
1da177e4 1253 goto out;
14e50e57 1254 }
1da177e4 1255
5c98631c
LC
1256 if (hlimit < 0)
1257 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4 1258
e651f03a 1259 if (tclass < 0)
41a1f8ea 1260 tclass = np->tclass;
41a1f8ea 1261
1da177e4
LT
1262 if (msg->msg_flags&MSG_CONFIRM)
1263 goto do_confirm;
1264back_from_confirm:
1265
03485f2a
VY
1266 /* Lockless fast path for the non-corking case */
1267 if (!corkreq) {
1268 struct sk_buff *skb;
1269
1270 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1271 sizeof(struct udphdr), hlimit, tclass, opt,
1272 &fl6, (struct rt6_info *)dst,
c14ac945 1273 msg->msg_flags, dontfrag, &sockc);
03485f2a
VY
1274 err = PTR_ERR(skb);
1275 if (!IS_ERR_OR_NULL(skb))
1276 err = udp_v6_send_skb(skb, &fl6);
1277 goto release_dst;
1278 }
1279
1da177e4
LT
1280 lock_sock(sk);
1281 if (unlikely(up->pending)) {
1282 /* The socket is already corked while preparing it. */
1283 /* ... which is an evident application bug. --ANK */
1284 release_sock(sk);
1285
ba7a46f1 1286 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1287 err = -EINVAL;
1288 goto out;
1289 }
1290
1291 up->pending = AF_INET6;
1292
1293do_append_data:
e36d3ff9
JP
1294 if (dontfrag < 0)
1295 dontfrag = np->dontfrag;
1da177e4 1296 up->len += ulen;
f69e6d13 1297 err = ip6_append_data(sk, getfrag, msg, ulen,
4c9483b2 1298 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
67ba4152 1299 (struct rt6_info *)dst,
c14ac945
SHY
1300 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag,
1301 &sockc);
1da177e4
LT
1302 if (err)
1303 udp_v6_flush_pending_frames(sk);
1304 else if (!corkreq)
4c0a6cb0 1305 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1306 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1307 up->pending = 0;
1da177e4 1308
03485f2a
VY
1309 if (err > 0)
1310 err = np->recverr ? net_xmit_errno(err) : 0;
1311 release_sock(sk);
1312
1313release_dst:
a5e7c210
DM
1314 if (dst) {
1315 if (connected) {
1316 ip6_dst_store(sk, dst,
efe4208f
ED
1317 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1318 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1319#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1320 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1321 &np->saddr :
1322#endif
1323 NULL);
a5e7c210
DM
1324 } else {
1325 dst_release(dst);
1326 }
a3c96089 1327 dst = NULL;
a5e7c210
DM
1328 }
1329
1da177e4 1330out:
a3c96089 1331 dst_release(dst);
1da177e4 1332 fl6_sock_release(flowlabel);
45f6fad8 1333 txopt_put(opt_to_free);
cd562c98 1334 if (!err)
1da177e4 1335 return len;
a18135eb
DM
1336 /*
1337 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1338 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1339 * we don't have a good statistic (IpOutDiscards but it can be too many
1340 * things). We could add another new stat but at least for now that
1341 * seems like overkill.
1342 */
1343 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
6aef70a8
ED
1344 UDP6_INC_STATS(sock_net(sk),
1345 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1346 }
1da177e4
LT
1347 return err;
1348
1349do_confirm:
1350 dst_confirm(dst);
1351 if (!(msg->msg_flags&MSG_PROBE) || len)
1352 goto back_from_confirm;
1353 err = 0;
1354 goto out;
1355}
1356
7d06b2e0 1357void udpv6_destroy_sock(struct sock *sk)
1da177e4 1358{
44046a59 1359 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1360 lock_sock(sk);
1361 udp_v6_flush_pending_frames(sk);
1362 release_sock(sk);
1363
44046a59
TP
1364 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1365 void (*encap_destroy)(struct sock *sk);
1366 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1367 if (encap_destroy)
1368 encap_destroy(sk);
1369 }
1370
1da177e4 1371 inet6_destroy_sock(sk);
1da177e4
LT
1372}
1373
1374/*
1375 * Socket option code for UDP
1376 */
ba4e58ec 1377int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1378 char __user *optval, unsigned int optlen)
3fdadf7d 1379{
db8dac20 1380 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1381 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1382 udp_v6_push_pending_frames);
ba4e58ec 1383 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1384}
1385
1386#ifdef CONFIG_COMPAT
ba4e58ec 1387int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1388 char __user *optval, unsigned int optlen)
3fdadf7d 1389{
db8dac20 1390 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1391 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1392 udp_v6_push_pending_frames);
ba4e58ec 1393 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1394}
1395#endif
1396
ba4e58ec
GR
1397int udpv6_getsockopt(struct sock *sk, int level, int optname,
1398 char __user *optval, int __user *optlen)
3fdadf7d 1399{
db8dac20 1400 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1401 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1402 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1403}
1404
1405#ifdef CONFIG_COMPAT
ba4e58ec
GR
1406int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1407 char __user *optval, int __user *optlen)
3fdadf7d 1408{
db8dac20 1409 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1410 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1411 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1412}
1413#endif
1414
41135cc8 1415static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1416 .handler = udpv6_rcv,
1417 .err_handler = udpv6_err,
1418 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1419};
1420
1421/* ------------------------------------------------------------------------ */
1422#ifdef CONFIG_PROC_FS
ba4e58ec 1423int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1424{
17ef66af
LC
1425 if (v == SEQ_START_TOKEN) {
1426 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1427 } else {
1428 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1429 struct inet_sock *inet = inet_sk(v);
1430 __u16 srcp = ntohs(inet->inet_sport);
1431 __u16 destp = ntohs(inet->inet_dport);
1432 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1433 }
1da177e4
LT
1434 return 0;
1435}
1436
73cb88ec
AV
1437static const struct file_operations udp6_afinfo_seq_fops = {
1438 .owner = THIS_MODULE,
1439 .open = udp_seq_open,
1440 .read = seq_read,
1441 .llseek = seq_lseek,
1442 .release = seq_release_net
1443};
1444
1da177e4 1445static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1446 .name = "udp6",
1447 .family = AF_INET6,
645ca708 1448 .udp_table = &udp_table,
73cb88ec 1449 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1450 .seq_ops = {
1451 .show = udp6_seq_show,
1452 },
1da177e4
LT
1453};
1454
2c8c1e72 1455int __net_init udp6_proc_init(struct net *net)
1da177e4 1456{
0c96d8c5 1457 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1458}
1459
ec120da6
IM
1460void udp6_proc_exit(struct net *net)
1461{
0c96d8c5 1462 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1463}
1464#endif /* CONFIG_PROC_FS */
1465
f77d6021
ED
1466void udp_v6_clear_sk(struct sock *sk, int size)
1467{
1468 struct inet_sock *inet = inet_sk(sk);
1469
1470 /* we do not want to clear pinet6 field, because of RCU lookups */
1471 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1472
1473 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1474 memset(&inet->pinet6 + 1, 0, size);
1475}
1476
1da177e4
LT
1477/* ------------------------------------------------------------------------ */
1478
1479struct proto udpv6_prot = {
543d9cfe
ACM
1480 .name = "UDPv6",
1481 .owner = THIS_MODULE,
ba4e58ec 1482 .close = udp_lib_close,
543d9cfe
ACM
1483 .connect = ip6_datagram_connect,
1484 .disconnect = udp_disconnect,
1485 .ioctl = udp_ioctl,
1486 .destroy = udpv6_destroy_sock,
1487 .setsockopt = udpv6_setsockopt,
1488 .getsockopt = udpv6_getsockopt,
1489 .sendmsg = udpv6_sendmsg,
1490 .recvmsg = udpv6_recvmsg,
f7ad74fe 1491 .backlog_rcv = __udpv6_queue_rcv_skb,
e646b657 1492 .release_cb = ip6_datagram_release_cb,
ba4e58ec
GR
1493 .hash = udp_lib_hash,
1494 .unhash = udp_lib_unhash,
719f8358 1495 .rehash = udp_v6_rehash,
543d9cfe 1496 .get_port = udp_v6_get_port,
95766fff
HA
1497 .memory_allocated = &udp_memory_allocated,
1498 .sysctl_mem = sysctl_udp_mem,
1499 .sysctl_wmem = &sysctl_udp_wmem_min,
1500 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1501 .obj_size = sizeof(struct udp6_sock),
271b72c7 1502 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1503 .h.udp_table = &udp_table,
3fdadf7d 1504#ifdef CONFIG_COMPAT
543d9cfe
ACM
1505 .compat_setsockopt = compat_udpv6_setsockopt,
1506 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1507#endif
f77d6021 1508 .clear_sk = udp_v6_clear_sk,
1da177e4
LT
1509};
1510
1da177e4
LT
1511static struct inet_protosw udpv6_protosw = {
1512 .type = SOCK_DGRAM,
1513 .protocol = IPPROTO_UDP,
1514 .prot = &udpv6_prot,
1515 .ops = &inet6_dgram_ops,
1da177e4
LT
1516 .flags = INET_PROTOSW_PERMANENT,
1517};
1518
7f4e4868 1519int __init udpv6_init(void)
1da177e4 1520{
7f4e4868
DL
1521 int ret;
1522
3336288a
VY
1523 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1524 if (ret)
c6b641a4 1525 goto out;
3336288a 1526
7f4e4868
DL
1527 ret = inet6_register_protosw(&udpv6_protosw);
1528 if (ret)
1529 goto out_udpv6_protocol;
1530out:
1531 return ret;
1532
1533out_udpv6_protocol:
1534 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1535 goto out;
1536}
1537
09f7709f 1538void udpv6_exit(void)
7f4e4868
DL
1539{
1540 inet6_unregister_protosw(&udpv6_protosw);
1541 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1542}