udp: remove headers from UDP packets before queueing
[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
ca065d0c
ED
329/* Must be called under rcu_read_lock().
330 * Does increment socket refcount.
331 */
332#if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
333 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
aa976fc0
BS
334struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
335 const struct in6_addr *daddr, __be16 dport, int dif)
336{
ca065d0c
ED
337 struct sock *sk;
338
339 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport,
340 dif, &udp_table, NULL);
341 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
342 sk = NULL;
343 return sk;
aa976fc0
BS
344}
345EXPORT_SYMBOL_GPL(udp6_lib_lookup);
ca065d0c 346#endif
aa976fc0 347
1da177e4 348/*
67ba4152
IM
349 * This should be easy, if there is something there we
350 * return it, otherwise we block.
1da177e4
LT
351 */
352
1b784140 353int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
1da177e4
LT
354 int noblock, int flags, int *addr_len)
355{
356 struct ipv6_pinfo *np = inet6_sk(sk);
357 struct inet_sock *inet = inet_sk(sk);
1ab1457c 358 struct sk_buff *skb;
59c2cdae 359 unsigned int ulen, copied;
3f518bf7 360 int peeked, off = 0;
759e5d00
HX
361 int err;
362 int is_udplite = IS_UDPLITE(sk);
197c949e 363 bool checksum_valid = false;
f26ba175 364 int is_udp4;
8a74ad60 365 bool slow;
1da177e4 366
1da177e4 367 if (flags & MSG_ERRQUEUE)
85fbaa75 368 return ipv6_recv_error(sk, msg, len, addr_len);
1da177e4 369
4b340ae2 370 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
85fbaa75 371 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
4b340ae2 372
1da177e4 373try_again:
a59322be 374 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 375 &peeked, &off, &err);
1da177e4
LT
376 if (!skb)
377 goto out;
378
e6afc8ac 379 ulen = skb->len;
59c2cdae
DM
380 copied = len;
381 if (copied > ulen)
382 copied = ulen;
383 else if (copied < ulen)
1ab1457c 384 msg->msg_flags |= MSG_TRUNC;
1da177e4 385
f26ba175
WY
386 is_udp4 = (skb->protocol == htons(ETH_P_IP));
387
ba4e58ec 388 /*
759e5d00
HX
389 * If checksum is needed at all, try to do it while copying the
390 * data. If the data is truncated, or if we only want a partial
391 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 392 */
ba4e58ec 393
59c2cdae 394 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
197c949e
ED
395 checksum_valid = !udp_lib_checksum_complete(skb);
396 if (!checksum_valid)
1da177e4 397 goto csum_copy_err;
ba4e58ec
GR
398 }
399
197c949e 400 if (checksum_valid || skb_csum_unnecessary(skb))
e6afc8ac 401 err = skb_copy_datagram_msg(skb, 0, msg, copied);
ba4e58ec 402 else {
e6afc8ac 403 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
1da177e4
LT
404 if (err == -EINVAL)
405 goto csum_copy_err;
406 }
22911fc5
ED
407 if (unlikely(err)) {
408 trace_kfree_skb(skb, udpv6_recvmsg);
979402b1
ED
409 if (!peeked) {
410 atomic_inc(&sk->sk_drops);
411 if (is_udp4)
412 UDP_INC_STATS_USER(sock_net(sk),
413 UDP_MIB_INERRORS,
414 is_udplite);
415 else
416 UDP6_INC_STATS_USER(sock_net(sk),
417 UDP_MIB_INERRORS,
418 is_udplite);
419 }
1da177e4 420 goto out_free;
22911fc5 421 }
f26ba175
WY
422 if (!peeked) {
423 if (is_udp4)
424 UDP_INC_STATS_USER(sock_net(sk),
425 UDP_MIB_INDATAGRAMS, is_udplite);
426 else
427 UDP6_INC_STATS_USER(sock_net(sk),
428 UDP_MIB_INDATAGRAMS, is_udplite);
429 }
cb75994e 430
3b885787 431 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
432
433 /* Copy the address. */
434 if (msg->msg_name) {
342dfc30 435 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 436 sin6->sin6_family = AF_INET6;
4bedb452 437 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 438 sin6->sin6_flowinfo = 0;
1da177e4 439
842df073 440 if (is_udp4) {
b301e82c
BH
441 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
442 &sin6->sin6_addr);
842df073
HFS
443 sin6->sin6_scope_id = 0;
444 } else {
4e3fd7a0 445 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
446 sin6->sin6_scope_id =
447 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 448 inet6_iif(skb));
1da177e4 449 }
bceaa902 450 *addr_len = sizeof(*sin6);
1da177e4 451 }
4b261c75
HFS
452
453 if (np->rxopt.all)
454 ip6_datagram_recv_common_ctl(sk, msg, skb);
455
f26ba175 456 if (is_udp4) {
1da177e4
LT
457 if (inet->cmsg_flags)
458 ip_cmsg_recv(msg, skb);
459 } else {
460 if (np->rxopt.all)
4b261c75 461 ip6_datagram_recv_specific_ctl(sk, msg, skb);
1ab1457c 462 }
1da177e4 463
59c2cdae 464 err = copied;
1da177e4 465 if (flags & MSG_TRUNC)
759e5d00 466 err = ulen;
1da177e4
LT
467
468out_free:
9d410c79 469 skb_free_datagram_locked(sk, skb);
1da177e4
LT
470out:
471 return err;
472
473csum_copy_err:
8a74ad60 474 slow = lock_sock_fast(sk);
0856f939 475 if (!skb_kill_datagram(sk, skb, flags)) {
6a5dc9e5
ED
476 if (is_udp4) {
477 UDP_INC_STATS_USER(sock_net(sk),
478 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
479 UDP_INC_STATS_USER(sock_net(sk),
480 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5
ED
481 } else {
482 UDP6_INC_STATS_USER(sock_net(sk),
483 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
484 UDP6_INC_STATS_USER(sock_net(sk),
485 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 486 }
0856f939 487 }
8a74ad60 488 unlock_sock_fast(sk, slow);
1da177e4 489
beb39db5
ED
490 /* starting over for a new packet, but check if we need to yield */
491 cond_resched();
9cfaa8de 492 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
493 goto try_again;
494}
495
ba4e58ec 496void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 497 u8 type, u8 code, int offset, __be32 info,
645ca708 498 struct udp_table *udptable)
1da177e4
LT
499{
500 struct ipv6_pinfo *np;
b71d1d42
ED
501 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
502 const struct in6_addr *saddr = &hdr->saddr;
503 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 504 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4 505 struct sock *sk;
e0d8c1b7 506 int harderr;
1da177e4 507 int err;
7304fe46 508 struct net *net = dev_net(skb->dev);
1da177e4 509
e32ea7e7 510 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
538950a1 511 inet6_iif(skb), udptable, skb);
63159f29 512 if (!sk) {
7304fe46
DJ
513 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
514 ICMP6_MIB_INERRORS);
1da177e4 515 return;
7304fe46 516 }
1da177e4 517
e0d8c1b7
WW
518 harderr = icmpv6_err_convert(type, code, &err);
519 np = inet6_sk(sk);
520
93b36cf3
HFS
521 if (type == ICMPV6_PKT_TOOBIG) {
522 if (!ip6_sk_accept_pmtu(sk))
523 goto out;
81aded24 524 ip6_sk_update_pmtu(skb, sk, info);
e0d8c1b7
WW
525 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
526 harderr = 1;
93b36cf3 527 }
1a462d18 528 if (type == NDISC_REDIRECT) {
ec18d9a2 529 ip6_sk_redirect(skb, sk);
1a462d18
DJ
530 goto out;
531 }
81aded24 532
e0d8c1b7
WW
533 if (!np->recverr) {
534 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
535 goto out;
536 } else {
1da177e4 537 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
e0d8c1b7 538 }
b1faf566 539
1da177e4
LT
540 sk->sk_err = err;
541 sk->sk_error_report(sk);
542out:
ca065d0c 543 return;
1da177e4
LT
544}
545
f7ad74fe
BL
546static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
547{
548 int rc;
549
efe4208f 550 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 551 sock_rps_save_rxhash(sk, skb);
005ec974 552 sk_mark_napi_id(sk, skb);
2c8c56e1 553 sk_incoming_cpu_update(sk);
005ec974 554 }
f7ad74fe 555
e6afc8ac 556 rc = __sock_queue_rcv_skb(sk, skb);
f7ad74fe
BL
557 if (rc < 0) {
558 int is_udplite = IS_UDPLITE(sk);
559
560 /* Note that an ENOMEM error is charged twice */
561 if (rc == -ENOMEM)
562 UDP6_INC_STATS_BH(sock_net(sk),
563 UDP_MIB_RCVBUFERRORS, is_udplite);
564 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
565 kfree_skb(skb);
566 return -1;
567 }
568 return 0;
569}
570
ba4e58ec 571static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 572 struct inet6_skb_parm *opt, u8 type,
67ba4152 573 u8 code, int offset, __be32 info)
ba4e58ec 574{
645ca708 575 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
576}
577
d7f3f621
BL
578static struct static_key udpv6_encap_needed __read_mostly;
579void udpv6_encap_enable(void)
580{
581 if (!static_key_enabled(&udpv6_encap_needed))
582 static_key_slow_inc(&udpv6_encap_needed);
583}
584EXPORT_SYMBOL(udpv6_encap_enable);
585
f7ad74fe 586int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 587{
ba4e58ec 588 struct udp_sock *up = udp_sk(sk);
a18135eb 589 int rc;
b2bf1e26 590 int is_udplite = IS_UDPLITE(sk);
a18135eb 591
ba4e58ec
GR
592 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
593 goto drop;
1da177e4 594
d7f3f621
BL
595 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
596 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
597
598 /*
599 * This is an encapsulation socket so pass the skb to
600 * the socket's udp_encap_rcv() hook. Otherwise, just
601 * fall through and pass this up the UDP socket.
602 * up->encap_rcv() returns the following value:
603 * =0 if skb was successfully passed to the encap
604 * handler or was discarded by it.
605 * >0 if skb should be passed on to UDP.
606 * <0 if skb should be resubmitted as proto -N
607 */
608
609 /* if we're overly short, let UDP handle it */
610 encap_rcv = ACCESS_ONCE(up->encap_rcv);
53b24b8f 611 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
d7f3f621
BL
612 int ret;
613
0a80966b
TH
614 /* Verify checksum before giving to encap */
615 if (udp_lib_checksum_complete(skb))
616 goto csum_error;
617
d7f3f621
BL
618 ret = encap_rcv(sk, skb);
619 if (ret <= 0) {
620 UDP_INC_STATS_BH(sock_net(sk),
621 UDP_MIB_INDATAGRAMS,
622 is_udplite);
623 return -ret;
624 }
625 }
626
627 /* FALLTHROUGH -- it's a UDP Packet */
628 }
629
ba4e58ec
GR
630 /*
631 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
632 */
b2bf1e26 633 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
634
635 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
636 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
637 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
638 goto drop;
639 }
640 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
641 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
642 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
643 goto drop;
644 }
1da177e4
LT
645 }
646
33d480ce 647 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62 648 if (udp_lib_checksum_complete(skb))
6a5dc9e5 649 goto csum_error;
e6afc8ac 650 if (sk_filter(sk, skb))
651 goto drop;
1ab6eb62 652 }
ba4e58ec 653
e6afc8ac 654 udp_csum_pull_header(skb);
274f482d 655 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
3e215c8d
JL
656 UDP6_INC_STATS_BH(sock_net(sk),
657 UDP_MIB_RCVBUFERRORS, is_udplite);
cb80ef46 658 goto drop;
3e215c8d 659 }
cb80ef46 660
d826eb14 661 skb_dst_drop(skb);
cb75994e 662
cb80ef46
BL
663 bh_lock_sock(sk);
664 rc = 0;
665 if (!sock_owned_by_user(sk))
666 rc = __udpv6_queue_rcv_skb(sk, skb);
667 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
668 bh_unlock_sock(sk);
669 goto drop;
670 }
671 bh_unlock_sock(sk);
f7ad74fe
BL
672
673 return rc;
3e215c8d 674
6a5dc9e5
ED
675csum_error:
676 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 677drop:
ef28d1a2 678 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 679 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
680 kfree_skb(skb);
681 return -1;
1da177e4
LT
682}
683
5cf3d461
DH
684static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
685 __be16 loc_port, const struct in6_addr *loc_addr,
686 __be16 rmt_port, const struct in6_addr *rmt_addr,
687 int dif, unsigned short hnum)
1da177e4 688{
5cf3d461 689 struct inet_sock *inet = inet_sk(sk);
1da177e4 690
5cf3d461
DH
691 if (!net_eq(sock_net(sk), net))
692 return false;
693
694 if (udp_sk(sk)->udp_port_hash != hnum ||
695 sk->sk_family != PF_INET6 ||
696 (inet->inet_dport && inet->inet_dport != rmt_port) ||
697 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
698 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
33b4b015
HR
699 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
700 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
701 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
5cf3d461
DH
702 return false;
703 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
704 return false;
705 return true;
1da177e4
LT
706}
707
4068579e
TH
708static void udp6_csum_zero_error(struct sk_buff *skb)
709{
710 /* RFC 2460 section 8.1 says that we SHOULD log
711 * this error. Well, it is reasonable.
712 */
ba7a46f1
JP
713 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
714 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
715 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
716}
717
1da177e4
LT
718/*
719 * Note: called only from the BH handler context,
720 * so we don't need to lock the hashes.
721 */
e3163493 722static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 723 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 724 struct udp_table *udptable, int proto)
1da177e4 725{
ca065d0c 726 struct sock *sk, *first = NULL;
4bedb452 727 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
728 unsigned short hnum = ntohs(uh->dest);
729 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
ca065d0c 730 unsigned int offset = offsetof(typeof(*sk), sk_node);
2dc41cff 731 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
ca065d0c
ED
732 int dif = inet6_iif(skb);
733 struct hlist_node *node;
734 struct sk_buff *nskb;
2dc41cff
DH
735
736 if (use_hash2) {
737 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
738 udp_table.mask;
739 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
740start_lookup:
741 hslot = &udp_table.hash2[hash2];
742 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
743 }
1da177e4 744
ca065d0c
ED
745 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
746 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
747 uh->source, saddr, dif, hnum))
748 continue;
749 /* If zero checksum and no_check is not on for
750 * the socket then skip it.
751 */
752 if (!uh->check && !udp_sk(sk)->no_check6_rx)
753 continue;
754 if (!first) {
755 first = sk;
756 continue;
757 }
758 nskb = skb_clone(skb, GFP_ATOMIC);
759 if (unlikely(!nskb)) {
760 atomic_inc(&sk->sk_drops);
761 UDP6_INC_STATS_BH(net, UDP_MIB_RCVBUFERRORS,
762 IS_UDPLITE(sk));
763 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS,
764 IS_UDPLITE(sk));
765 continue;
95766fff 766 }
a1ab77f9 767
ca065d0c
ED
768 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
769 consume_skb(nskb);
770 }
a1ab77f9 771
2dc41cff
DH
772 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
773 if (use_hash2 && hash2 != hash2_any) {
774 hash2 = hash2_any;
775 goto start_lookup;
776 }
777
ca065d0c
ED
778 if (first) {
779 if (udpv6_queue_rcv_skb(first, skb) > 0)
780 consume_skb(skb);
a1ab77f9 781 } else {
ca065d0c
ED
782 kfree_skb(skb);
783 UDP6_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
784 proto == IPPROTO_UDPLITE);
a1ab77f9 785 }
ba4e58ec 786 return 0;
1da177e4
LT
787}
788
645ca708 789int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 790 int proto)
1da177e4 791{
ca065d0c 792 const struct in6_addr *saddr, *daddr;
3ffe533c 793 struct net *net = dev_net(skb->dev);
1ab1457c 794 struct udphdr *uh;
ca065d0c 795 struct sock *sk;
1da177e4
LT
796 u32 ulen = 0;
797
798 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 799 goto discard;
1da177e4 800
0660e03f
ACM
801 saddr = &ipv6_hdr(skb)->saddr;
802 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 803 uh = udp_hdr(skb);
1da177e4
LT
804
805 ulen = ntohs(uh->len);
ba4e58ec
GR
806 if (ulen > skb->len)
807 goto short_packet;
1da177e4 808
759e5d00
HX
809 if (proto == IPPROTO_UDP) {
810 /* UDP validates ulen. */
1da177e4 811
ba4e58ec
GR
812 /* Check for jumbo payload */
813 if (ulen == 0)
814 ulen = skb->len;
1da177e4 815
ba4e58ec
GR
816 if (ulen < sizeof(*uh))
817 goto short_packet;
1da177e4 818
ba4e58ec
GR
819 if (ulen < skb->len) {
820 if (pskb_trim_rcsum(skb, ulen))
821 goto short_packet;
0660e03f
ACM
822 saddr = &ipv6_hdr(skb)->saddr;
823 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 824 uh = udp_hdr(skb);
ba4e58ec 825 }
ba4e58ec 826 }
1da177e4 827
759e5d00 828 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 829 goto csum_error;
759e5d00 830
1ab1457c
YH
831 /*
832 * Multicast receive code
1da177e4 833 */
ba4e58ec 834 if (ipv6_addr_is_multicast(daddr))
e3163493 835 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 836 saddr, daddr, udptable, proto);
1da177e4
LT
837
838 /* Unicast */
839
1ab1457c 840 /*
1da177e4
LT
841 * check socket cache ... must talk to Alan about his plans
842 * for sock caches... i'll skip this for now.
843 */
607c4aaf 844 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
53b24b8f 845 if (sk) {
a5b50476
ET
846 int ret;
847
1c19448c 848 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
4068579e
TH
849 udp6_csum_zero_error(skb);
850 goto csum_error;
851 }
852
224d019c 853 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
2abb7cdc
TH
854 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
855 ip6_compute_pseudo);
856
a5b50476 857 ret = udpv6_queue_rcv_skb(sk, skb);
1da177e4 858
59dca1d8 859 /* a return value > 0 means to resubmit the input */
cb80ef46 860 if (ret > 0)
59dca1d8 861 return ret;
1da177e4 862
add459aa 863 return 0;
1da177e4 864 }
1ab1457c 865
4068579e
TH
866 if (!uh->check) {
867 udp6_csum_zero_error(skb);
868 goto csum_error;
869 }
870
cb80ef46 871 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 872 goto discard;
cb80ef46
BL
873
874 if (udp_lib_checksum_complete(skb))
6a5dc9e5 875 goto csum_error;
cb80ef46
BL
876
877 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
878 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
879
880 kfree_skb(skb);
add459aa 881 return 0;
1da177e4 882
1ab1457c 883short_packet:
ba7a46f1
JP
884 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
885 proto == IPPROTO_UDPLITE ? "-Lite" : "",
886 saddr, ntohs(uh->source),
887 ulen, skb->len,
888 daddr, ntohs(uh->dest));
6a5dc9e5
ED
889 goto discard;
890csum_error:
891 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 892discard:
ef28d1a2 893 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 894 kfree_skb(skb);
add459aa 895 return 0;
1da177e4 896}
ba4e58ec 897
e5bbef20 898static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 899{
645ca708 900 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
901}
902
1da177e4
LT
903/*
904 * Throw away all pending data and cancel the corking. Socket is locked.
905 */
906static void udp_v6_flush_pending_frames(struct sock *sk)
907{
908 struct udp_sock *up = udp_sk(sk);
909
36d926b9
DL
910 if (up->pending == AF_INET)
911 udp_flush_pending_frames(sk);
912 else if (up->pending) {
1da177e4
LT
913 up->len = 0;
914 up->pending = 0;
915 ip6_flush_pending_frames(sk);
1ab1457c 916 }
1da177e4
LT
917}
918
493c6be3 919/**
67ba4152
IM
920 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
921 * @sk: socket we are sending on
922 * @skb: sk_buff containing the filled-in UDP header
923 * (checksum field must be zeroed out)
493c6be3
SS
924 */
925static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
926 const struct in6_addr *saddr,
927 const struct in6_addr *daddr, int len)
928{
929 unsigned int offset;
930 struct udphdr *uh = udp_hdr(skb);
d39d938c 931 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
493c6be3
SS
932 __wsum csum = 0;
933
d39d938c 934 if (!frags) {
493c6be3
SS
935 /* Only one fragment on the socket. */
936 skb->csum_start = skb_transport_header(skb) - skb->head;
937 skb->csum_offset = offsetof(struct udphdr, check);
938 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
939 } else {
940 /*
941 * HW-checksum won't work as there are two or more
942 * fragments on the socket so that all csums of sk_buffs
943 * should be together
944 */
945 offset = skb_transport_offset(skb);
946 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
947
948 skb->ip_summed = CHECKSUM_NONE;
949
d39d938c
VY
950 do {
951 csum = csum_add(csum, frags->csum);
952 } while ((frags = frags->next));
493c6be3
SS
953
954 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
955 csum);
956 if (uh->check == 0)
957 uh->check = CSUM_MANGLED_0;
958 }
959}
960
1da177e4
LT
961/*
962 * Sending
963 */
964
d39d938c 965static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1da177e4 966{
d39d938c 967 struct sock *sk = skb->sk;
1da177e4 968 struct udphdr *uh;
1da177e4 969 int err = 0;
b2bf1e26 970 int is_udplite = IS_UDPLITE(sk);
868c86bc 971 __wsum csum = 0;
d39d938c
VY
972 int offset = skb_transport_offset(skb);
973 int len = skb->len - offset;
1da177e4
LT
974
975 /*
976 * Create a UDP header
977 */
4bedb452 978 uh = udp_hdr(skb);
1958b856
DM
979 uh->source = fl6->fl6_sport;
980 uh->dest = fl6->fl6_dport;
d39d938c 981 uh->len = htons(len);
1da177e4
LT
982 uh->check = 0;
983
b2bf1e26 984 if (is_udplite)
d39d938c
VY
985 csum = udplite_csum(skb);
986 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
987 skb->ip_summed = CHECKSUM_NONE;
988 goto send;
989 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
d39d938c 990 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
493c6be3
SS
991 goto send;
992 } else
d39d938c 993 csum = udp_csum(skb);
1da177e4 994
ba4e58ec 995 /* add protocol-dependent pseudo-header */
4c9483b2 996 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
d39d938c 997 len, fl6->flowi6_proto, csum);
1da177e4 998 if (uh->check == 0)
f6ab0288 999 uh->check = CSUM_MANGLED_0;
1da177e4 1000
493c6be3 1001send:
d39d938c 1002 err = ip6_send_skb(skb);
6ce9e7b5
ED
1003 if (err) {
1004 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1005 UDP6_INC_STATS_USER(sock_net(sk),
1006 UDP_MIB_SNDBUFERRORS, is_udplite);
1007 err = 0;
1008 }
1009 } else
1010 UDP6_INC_STATS_USER(sock_net(sk),
1011 UDP_MIB_OUTDATAGRAMS, is_udplite);
d39d938c
VY
1012 return err;
1013}
1014
1015static int udp_v6_push_pending_frames(struct sock *sk)
1016{
1017 struct sk_buff *skb;
1018 struct udp_sock *up = udp_sk(sk);
1019 struct flowi6 fl6;
1020 int err = 0;
1021
1022 if (up->pending == AF_INET)
1023 return udp_push_pending_frames(sk);
1024
1025 /* ip6_finish_skb will release the cork, so make a copy of
1026 * fl6 here.
1027 */
1028 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1029
1030 skb = ip6_finish_skb(sk);
1031 if (!skb)
1032 goto out;
1033
1034 err = udp_v6_send_skb(skb, &fl6);
1035
1da177e4
LT
1036out:
1037 up->len = 0;
1038 up->pending = 0;
1039 return err;
1040}
1041
1b784140 1042int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1da177e4
LT
1043{
1044 struct ipv6_txoptions opt_space;
1045 struct udp_sock *up = udp_sk(sk);
1046 struct inet_sock *inet = inet_sk(sk);
1047 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1048 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1049 struct in6_addr *daddr, *final_p, final;
1da177e4 1050 struct ipv6_txoptions *opt = NULL;
45f6fad8 1051 struct ipv6_txoptions *opt_to_free = NULL;
1da177e4 1052 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1053 struct flowi6 fl6;
1da177e4
LT
1054 struct dst_entry *dst;
1055 int addr_len = msg->msg_namelen;
1056 int ulen = len;
1057 int hlimit = -1;
41a1f8ea 1058 int tclass = -1;
13b52cd4 1059 int dontfrag = -1;
1da177e4
LT
1060 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1061 int err;
987905de 1062 int connected = 0;
b2bf1e26 1063 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1064 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
ad1e46a8 1065 struct sockcm_cookie sockc;
1da177e4
LT
1066
1067 /* destination address check */
1068 if (sin6) {
1069 if (addr_len < offsetof(struct sockaddr, sa_data))
1070 return -EINVAL;
1071
1072 switch (sin6->sin6_family) {
1073 case AF_INET6:
1074 if (addr_len < SIN6_LEN_RFC2133)
1075 return -EINVAL;
1076 daddr = &sin6->sin6_addr;
1077 break;
1078 case AF_INET:
1079 goto do_udp_sendmsg;
1080 case AF_UNSPEC:
1081 msg->msg_name = sin6 = NULL;
1082 msg->msg_namelen = addr_len = 0;
1083 daddr = NULL;
1084 break;
1085 default:
1086 return -EINVAL;
1087 }
1088 } else if (!up->pending) {
1089 if (sk->sk_state != TCP_ESTABLISHED)
1090 return -EDESTADDRREQ;
efe4208f 1091 daddr = &sk->sk_v6_daddr;
1ab1457c 1092 } else
1da177e4
LT
1093 daddr = NULL;
1094
1095 if (daddr) {
e773e4fa 1096 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1097 struct sockaddr_in sin;
1098 sin.sin_family = AF_INET;
c720c7e8 1099 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1100 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1101 msg->msg_name = &sin;
1102 msg->msg_namelen = sizeof(sin);
1103do_udp_sendmsg:
1104 if (__ipv6_only_sock(sk))
1105 return -ENETUNREACH;
1b784140 1106 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1107 }
1108 }
1109
1110 if (up->pending == AF_INET)
1b784140 1111 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1112
1113 /* Rough check on arithmetic overflow,
b59e139b 1114 better check is made in ip6_append_data().
1da177e4
LT
1115 */
1116 if (len > INT_MAX - sizeof(struct udphdr))
1117 return -EMSGSIZE;
1ab1457c 1118
03485f2a 1119 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1da177e4
LT
1120 if (up->pending) {
1121 /*
1122 * There are pending frames.
1123 * The socket lock must be held while it's corked.
1124 */
1125 lock_sock(sk);
1126 if (likely(up->pending)) {
1127 if (unlikely(up->pending != AF_INET6)) {
1128 release_sock(sk);
1129 return -EAFNOSUPPORT;
1130 }
1131 dst = NULL;
1132 goto do_append_data;
1133 }
1134 release_sock(sk);
1135 }
1136 ulen += sizeof(struct udphdr);
1137
4c9483b2 1138 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1139
1140 if (sin6) {
1141 if (sin6->sin6_port == 0)
1142 return -EINVAL;
1143
1958b856 1144 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1145 daddr = &sin6->sin6_addr;
1146
1147 if (np->sndflow) {
4c9483b2
DM
1148 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1149 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1150 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1151 if (!flowlabel)
1da177e4 1152 return -EINVAL;
1da177e4
LT
1153 }
1154 }
1155
1156 /*
1157 * Otherwise it will be difficult to maintain
1158 * sk->sk_dst_cache.
1159 */
1160 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1161 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1162 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1163
1164 if (addr_len >= sizeof(struct sockaddr_in6) &&
1165 sin6->sin6_scope_id &&
842df073 1166 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1167 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1168 } else {
1169 if (sk->sk_state != TCP_ESTABLISHED)
1170 return -EDESTADDRREQ;
1171
1958b856 1172 fl6.fl6_dport = inet->inet_dport;
efe4208f 1173 daddr = &sk->sk_v6_daddr;
4c9483b2 1174 fl6.flowlabel = np->flow_label;
987905de 1175 connected = 1;
1da177e4
LT
1176 }
1177
4c9483b2
DM
1178 if (!fl6.flowi6_oif)
1179 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1180
4c9483b2
DM
1181 if (!fl6.flowi6_oif)
1182 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1183
4c9483b2 1184 fl6.flowi6_mark = sk->sk_mark;
c14ac945 1185 sockc.tsflags = sk->sk_tsflags;
51953d5b 1186
1da177e4
LT
1187 if (msg->msg_controllen) {
1188 opt = &opt_space;
1189 memset(opt, 0, sizeof(struct ipv6_txoptions));
1190 opt->tot_len = sizeof(*opt);
1191
73df66f8 1192 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
ad1e46a8
SHY
1193 &hlimit, &tclass, &dontfrag,
1194 &sockc);
1da177e4
LT
1195 if (err < 0) {
1196 fl6_sock_release(flowlabel);
1197 return err;
1198 }
4c9483b2
DM
1199 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1200 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1201 if (!flowlabel)
1da177e4
LT
1202 return -EINVAL;
1203 }
1204 if (!(opt->opt_nflen|opt->opt_flen))
1205 opt = NULL;
987905de 1206 connected = 0;
1da177e4 1207 }
45f6fad8
ED
1208 if (!opt) {
1209 opt = txopt_get(np);
1210 opt_to_free = opt;
1211 }
df9890c3
YH
1212 if (flowlabel)
1213 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1214 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1215
4c9483b2 1216 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1217 if (!ipv6_addr_any(daddr))
4e3fd7a0 1218 fl6.daddr = *daddr;
876c7f41 1219 else
4c9483b2
DM
1220 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1221 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1222 fl6.saddr = np->saddr;
1958b856 1223 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1224
4c9483b2 1225 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1226 if (final_p)
987905de 1227 connected = 0;
1da177e4 1228
4c9483b2
DM
1229 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1230 fl6.flowi6_oif = np->mcast_oif;
987905de 1231 connected = 0;
c4062dfc
EH
1232 } else if (!fl6.flowi6_oif)
1233 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1234
4c9483b2 1235 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1236
0e0d44ab 1237 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1238 if (IS_ERR(dst)) {
1239 err = PTR_ERR(dst);
1240 dst = NULL;
1da177e4 1241 goto out;
14e50e57 1242 }
1da177e4 1243
5c98631c
LC
1244 if (hlimit < 0)
1245 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4 1246
e651f03a 1247 if (tclass < 0)
41a1f8ea 1248 tclass = np->tclass;
41a1f8ea 1249
1da177e4
LT
1250 if (msg->msg_flags&MSG_CONFIRM)
1251 goto do_confirm;
1252back_from_confirm:
1253
03485f2a
VY
1254 /* Lockless fast path for the non-corking case */
1255 if (!corkreq) {
1256 struct sk_buff *skb;
1257
1258 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1259 sizeof(struct udphdr), hlimit, tclass, opt,
1260 &fl6, (struct rt6_info *)dst,
c14ac945 1261 msg->msg_flags, dontfrag, &sockc);
03485f2a
VY
1262 err = PTR_ERR(skb);
1263 if (!IS_ERR_OR_NULL(skb))
1264 err = udp_v6_send_skb(skb, &fl6);
1265 goto release_dst;
1266 }
1267
1da177e4
LT
1268 lock_sock(sk);
1269 if (unlikely(up->pending)) {
1270 /* The socket is already corked while preparing it. */
1271 /* ... which is an evident application bug. --ANK */
1272 release_sock(sk);
1273
ba7a46f1 1274 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1275 err = -EINVAL;
1276 goto out;
1277 }
1278
1279 up->pending = AF_INET6;
1280
1281do_append_data:
e36d3ff9
JP
1282 if (dontfrag < 0)
1283 dontfrag = np->dontfrag;
1da177e4 1284 up->len += ulen;
f69e6d13 1285 err = ip6_append_data(sk, getfrag, msg, ulen,
4c9483b2 1286 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
67ba4152 1287 (struct rt6_info *)dst,
c14ac945
SHY
1288 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag,
1289 &sockc);
1da177e4
LT
1290 if (err)
1291 udp_v6_flush_pending_frames(sk);
1292 else if (!corkreq)
4c0a6cb0 1293 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1294 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1295 up->pending = 0;
1da177e4 1296
03485f2a
VY
1297 if (err > 0)
1298 err = np->recverr ? net_xmit_errno(err) : 0;
1299 release_sock(sk);
1300
1301release_dst:
a5e7c210
DM
1302 if (dst) {
1303 if (connected) {
1304 ip6_dst_store(sk, dst,
efe4208f
ED
1305 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1306 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1307#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1308 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1309 &np->saddr :
1310#endif
1311 NULL);
a5e7c210
DM
1312 } else {
1313 dst_release(dst);
1314 }
a3c96089 1315 dst = NULL;
a5e7c210
DM
1316 }
1317
1da177e4 1318out:
a3c96089 1319 dst_release(dst);
1da177e4 1320 fl6_sock_release(flowlabel);
45f6fad8 1321 txopt_put(opt_to_free);
cd562c98 1322 if (!err)
1da177e4 1323 return len;
a18135eb
DM
1324 /*
1325 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1326 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1327 * we don't have a good statistic (IpOutDiscards but it can be too many
1328 * things). We could add another new stat but at least for now that
1329 * seems like overkill.
1330 */
1331 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
235b9f7a
PE
1332 UDP6_INC_STATS_USER(sock_net(sk),
1333 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1334 }
1da177e4
LT
1335 return err;
1336
1337do_confirm:
1338 dst_confirm(dst);
1339 if (!(msg->msg_flags&MSG_PROBE) || len)
1340 goto back_from_confirm;
1341 err = 0;
1342 goto out;
1343}
1344
7d06b2e0 1345void udpv6_destroy_sock(struct sock *sk)
1da177e4 1346{
44046a59 1347 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1348 lock_sock(sk);
1349 udp_v6_flush_pending_frames(sk);
1350 release_sock(sk);
1351
44046a59
TP
1352 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1353 void (*encap_destroy)(struct sock *sk);
1354 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1355 if (encap_destroy)
1356 encap_destroy(sk);
1357 }
1358
1da177e4 1359 inet6_destroy_sock(sk);
1da177e4
LT
1360}
1361
1362/*
1363 * Socket option code for UDP
1364 */
ba4e58ec 1365int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1366 char __user *optval, unsigned int optlen)
3fdadf7d 1367{
db8dac20 1368 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1369 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1370 udp_v6_push_pending_frames);
ba4e58ec 1371 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1372}
1373
1374#ifdef CONFIG_COMPAT
ba4e58ec 1375int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1376 char __user *optval, unsigned int optlen)
3fdadf7d 1377{
db8dac20 1378 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1379 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1380 udp_v6_push_pending_frames);
ba4e58ec 1381 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1382}
1383#endif
1384
ba4e58ec
GR
1385int udpv6_getsockopt(struct sock *sk, int level, int optname,
1386 char __user *optval, int __user *optlen)
3fdadf7d 1387{
db8dac20 1388 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1389 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1390 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1391}
1392
1393#ifdef CONFIG_COMPAT
ba4e58ec
GR
1394int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1395 char __user *optval, int __user *optlen)
3fdadf7d 1396{
db8dac20 1397 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1398 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1399 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1400}
1401#endif
1402
41135cc8 1403static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1404 .handler = udpv6_rcv,
1405 .err_handler = udpv6_err,
1406 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1407};
1408
1409/* ------------------------------------------------------------------------ */
1410#ifdef CONFIG_PROC_FS
ba4e58ec 1411int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1412{
17ef66af
LC
1413 if (v == SEQ_START_TOKEN) {
1414 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1415 } else {
1416 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1417 struct inet_sock *inet = inet_sk(v);
1418 __u16 srcp = ntohs(inet->inet_sport);
1419 __u16 destp = ntohs(inet->inet_dport);
1420 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1421 }
1da177e4
LT
1422 return 0;
1423}
1424
73cb88ec
AV
1425static const struct file_operations udp6_afinfo_seq_fops = {
1426 .owner = THIS_MODULE,
1427 .open = udp_seq_open,
1428 .read = seq_read,
1429 .llseek = seq_lseek,
1430 .release = seq_release_net
1431};
1432
1da177e4 1433static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1434 .name = "udp6",
1435 .family = AF_INET6,
645ca708 1436 .udp_table = &udp_table,
73cb88ec 1437 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1438 .seq_ops = {
1439 .show = udp6_seq_show,
1440 },
1da177e4
LT
1441};
1442
2c8c1e72 1443int __net_init udp6_proc_init(struct net *net)
1da177e4 1444{
0c96d8c5 1445 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1446}
1447
ec120da6
IM
1448void udp6_proc_exit(struct net *net)
1449{
0c96d8c5 1450 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1451}
1452#endif /* CONFIG_PROC_FS */
1453
f77d6021
ED
1454void udp_v6_clear_sk(struct sock *sk, int size)
1455{
1456 struct inet_sock *inet = inet_sk(sk);
1457
1458 /* we do not want to clear pinet6 field, because of RCU lookups */
1459 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1460
1461 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1462 memset(&inet->pinet6 + 1, 0, size);
1463}
1464
1da177e4
LT
1465/* ------------------------------------------------------------------------ */
1466
1467struct proto udpv6_prot = {
543d9cfe
ACM
1468 .name = "UDPv6",
1469 .owner = THIS_MODULE,
ba4e58ec 1470 .close = udp_lib_close,
543d9cfe
ACM
1471 .connect = ip6_datagram_connect,
1472 .disconnect = udp_disconnect,
1473 .ioctl = udp_ioctl,
1474 .destroy = udpv6_destroy_sock,
1475 .setsockopt = udpv6_setsockopt,
1476 .getsockopt = udpv6_getsockopt,
1477 .sendmsg = udpv6_sendmsg,
1478 .recvmsg = udpv6_recvmsg,
f7ad74fe 1479 .backlog_rcv = __udpv6_queue_rcv_skb,
ba4e58ec
GR
1480 .hash = udp_lib_hash,
1481 .unhash = udp_lib_unhash,
719f8358 1482 .rehash = udp_v6_rehash,
543d9cfe 1483 .get_port = udp_v6_get_port,
95766fff
HA
1484 .memory_allocated = &udp_memory_allocated,
1485 .sysctl_mem = sysctl_udp_mem,
1486 .sysctl_wmem = &sysctl_udp_wmem_min,
1487 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1488 .obj_size = sizeof(struct udp6_sock),
271b72c7 1489 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1490 .h.udp_table = &udp_table,
3fdadf7d 1491#ifdef CONFIG_COMPAT
543d9cfe
ACM
1492 .compat_setsockopt = compat_udpv6_setsockopt,
1493 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1494#endif
f77d6021 1495 .clear_sk = udp_v6_clear_sk,
1da177e4
LT
1496};
1497
1da177e4
LT
1498static struct inet_protosw udpv6_protosw = {
1499 .type = SOCK_DGRAM,
1500 .protocol = IPPROTO_UDP,
1501 .prot = &udpv6_prot,
1502 .ops = &inet6_dgram_ops,
1da177e4
LT
1503 .flags = INET_PROTOSW_PERMANENT,
1504};
1505
7f4e4868 1506int __init udpv6_init(void)
1da177e4 1507{
7f4e4868
DL
1508 int ret;
1509
3336288a
VY
1510 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1511 if (ret)
c6b641a4 1512 goto out;
3336288a 1513
7f4e4868
DL
1514 ret = inet6_register_protosw(&udpv6_protosw);
1515 if (ret)
1516 goto out_udpv6_protocol;
1517out:
1518 return ret;
1519
1520out_udpv6_protocol:
1521 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1522 goto out;
1523}
1524
09f7709f 1525void udpv6_exit(void)
7f4e4868
DL
1526{
1527 inet6_unregister_protosw(&udpv6_protosw);
1528 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1529}