]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - net/ipv6/udp.c
enic: use spin_lock(wq_lock) instead of spin_lock_irqsave(wq_lock)
[mirror_ubuntu-zesty-kernel.git] / net / ipv6 / udp.c
CommitLineData
1da177e4
LT
1/*
2 * UDP over IPv6
1ab1457c 3 * Linux INET6 implementation
1da177e4
LT
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT
7 *
8 * Based on linux/ipv4/udp.c
9 *
1da177e4
LT
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
14 * a single port at the same time.
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 */
23
1da177e4
LT
24#include <linux/errno.h>
25#include <linux/types.h>
26#include <linux/socket.h>
27#include <linux/sockios.h>
1da177e4
LT
28#include <linux/net.h>
29#include <linux/in6.h>
30#include <linux/netdevice.h>
31#include <linux/if_arp.h>
32#include <linux/ipv6.h>
33#include <linux/icmpv6.h>
34#include <linux/init.h>
1781f7f5 35#include <linux/module.h>
3305b80c 36#include <linux/skbuff.h>
5a0e3ad6 37#include <linux/slab.h>
1da177e4
LT
38#include <asm/uaccess.h>
39
1da177e4
LT
40#include <net/ndisc.h>
41#include <net/protocol.h>
42#include <net/transp_v6.h>
43#include <net/ip6_route.h>
1da177e4 44#include <net/raw.h>
c752f073 45#include <net/tcp_states.h>
1da177e4
LT
46#include <net/ip6_checksum.h>
47#include <net/xfrm.h>
72289b96 48#include <net/inet6_hashtables.h>
076bb0c8 49#include <net/busy_poll.h>
1da177e4
LT
50
51#include <linux/proc_fs.h>
52#include <linux/seq_file.h>
22911fc5 53#include <trace/events/skb.h>
ba4e58ec 54#include "udp_impl.h"
1da177e4 55
b50026b5
HFS
56static unsigned int udp6_ehashfn(struct net *net,
57 const struct in6_addr *laddr,
58 const u16 lport,
59 const struct in6_addr *faddr,
60 const __be16 fport)
61{
1bbdceef
HFS
62 static u32 udp6_ehash_secret __read_mostly;
63 static u32 udp_ipv6_hash_secret __read_mostly;
64
65 u32 lhash, fhash;
66
67 net_get_random_once(&udp6_ehash_secret,
68 sizeof(udp6_ehash_secret));
69 net_get_random_once(&udp_ipv6_hash_secret,
70 sizeof(udp_ipv6_hash_secret));
71
72 lhash = (__force u32)laddr->s6_addr32[3];
73 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
74
b50026b5 75 return __inet6_ehashfn(lhash, lport, fhash, fport,
1bbdceef 76 udp_ipv6_hash_secret + net_hash_mix(net));
b50026b5
HFS
77}
78
b2f5e7cd
VY
79int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
80{
b2f5e7cd 81 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
b2f5e7cd 82 int sk2_ipv6only = inet_v6_ipv6only(sk2);
efe4208f 83 int addr_type = ipv6_addr_type(&sk->sk_v6_rcv_saddr);
b2f5e7cd
VY
84 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
85
86 /* if both are mapped, treat as IPv4 */
87 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
499923c7 88 return (!sk2_ipv6only &&
50805466
ED
89 (!sk->sk_rcv_saddr || !sk2->sk_rcv_saddr ||
90 sk->sk_rcv_saddr == sk2->sk_rcv_saddr));
b2f5e7cd
VY
91
92 if (addr_type2 == IPV6_ADDR_ANY &&
93 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
94 return 1;
95
96 if (addr_type == IPV6_ADDR_ANY &&
9fe516ba 97 !(ipv6_only_sock(sk) && addr_type2 == IPV6_ADDR_MAPPED))
b2f5e7cd
VY
98 return 1;
99
100 if (sk2_rcv_saddr6 &&
efe4208f 101 ipv6_addr_equal(&sk->sk_v6_rcv_saddr, sk2_rcv_saddr6))
b2f5e7cd
VY
102 return 1;
103
104 return 0;
105}
106
d4cada4a
ED
107static unsigned int udp6_portaddr_hash(struct net *net,
108 const struct in6_addr *addr6,
109 unsigned int port)
110{
111 unsigned int hash, mix = net_hash_mix(net);
112
113 if (ipv6_addr_any(addr6))
114 hash = jhash_1word(0, mix);
856540ee 115 else if (ipv6_addr_v4mapped(addr6))
0eae88f3 116 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
d4cada4a 117 else
0eae88f3 118 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
d4cada4a
ED
119
120 return hash ^ port;
121}
122
123
6ba5a3c5 124int udp_v6_get_port(struct sock *sk, unsigned short snum)
1da177e4 125{
30fff923
ED
126 unsigned int hash2_nulladdr =
127 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
9a52e97e 128 unsigned int hash2_partial =
efe4208f 129 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
30fff923 130
d4cada4a 131 /* precompute partial secondary hash */
30fff923
ED
132 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
133 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
1da177e4
LT
134}
135
719f8358
ED
136static void udp_v6_rehash(struct sock *sk)
137{
138 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
efe4208f 139 &sk->sk_v6_rcv_saddr,
719f8358
ED
140 inet_sk(sk)->inet_num);
141
142 udp_lib_rehash(sk, new_hash);
143}
144
645ca708
ED
145static inline int compute_score(struct sock *sk, struct net *net,
146 unsigned short hnum,
88440ae7
BS
147 const struct in6_addr *saddr, __be16 sport,
148 const struct in6_addr *daddr, __be16 dport,
645ca708
ED
149 int dif)
150{
151 int score = -1;
152
d4cada4a 153 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
645ca708 154 sk->sk_family == PF_INET6) {
645ca708
ED
155 struct inet_sock *inet = inet_sk(sk);
156
157 score = 0;
c720c7e8
ED
158 if (inet->inet_dport) {
159 if (inet->inet_dport != sport)
645ca708
ED
160 return -1;
161 score++;
162 }
efe4208f
ED
163 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
164 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
645ca708
ED
165 return -1;
166 score++;
167 }
efe4208f
ED
168 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
169 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
645ca708
ED
170 return -1;
171 score++;
172 }
173 if (sk->sk_bound_dev_if) {
174 if (sk->sk_bound_dev_if != dif)
175 return -1;
176 score++;
177 }
178 }
179 return score;
180}
181
fddc17de
ED
182#define SCORE2_MAX (1 + 1 + 1)
183static inline int compute_score2(struct sock *sk, struct net *net,
184 const struct in6_addr *saddr, __be16 sport,
185 const struct in6_addr *daddr, unsigned short hnum,
186 int dif)
187{
188 int score = -1;
189
190 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
191 sk->sk_family == PF_INET6) {
fddc17de
ED
192 struct inet_sock *inet = inet_sk(sk);
193
efe4208f 194 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
fddc17de
ED
195 return -1;
196 score = 0;
197 if (inet->inet_dport) {
198 if (inet->inet_dport != sport)
199 return -1;
200 score++;
201 }
efe4208f
ED
202 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
203 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
fddc17de
ED
204 return -1;
205 score++;
206 }
207 if (sk->sk_bound_dev_if) {
208 if (sk->sk_bound_dev_if != dif)
209 return -1;
210 score++;
211 }
212 }
213 return score;
214}
215
fddc17de
ED
216
217/* called with read_rcu_lock() */
218static struct sock *udp6_lib_lookup2(struct net *net,
219 const struct in6_addr *saddr, __be16 sport,
220 const struct in6_addr *daddr, unsigned int hnum, int dif,
221 struct udp_hslot *hslot2, unsigned int slot2)
222{
223 struct sock *sk, *result;
224 struct hlist_nulls_node *node;
72289b96
TH
225 int score, badness, matches = 0, reuseport = 0;
226 u32 hash = 0;
fddc17de
ED
227
228begin:
229 result = NULL;
230 badness = -1;
231 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
232 score = compute_score2(sk, net, saddr, sport,
233 daddr, hnum, dif);
234 if (score > badness) {
235 result = sk;
236 badness = score;
72289b96
TH
237 reuseport = sk->sk_reuseport;
238 if (reuseport) {
b50026b5
HFS
239 hash = udp6_ehashfn(net, daddr, hnum,
240 saddr, sport);
72289b96
TH
241 matches = 1;
242 } else if (score == SCORE2_MAX)
fddc17de 243 goto exact_match;
72289b96
TH
244 } else if (score == badness && reuseport) {
245 matches++;
8fc54f68 246 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
247 result = sk;
248 hash = next_pseudo_random32(hash);
fddc17de
ED
249 }
250 }
251 /*
252 * if the nulls value we got at the end of this lookup is
253 * not the expected one, we must restart lookup.
254 * We probably met an item that was moved to another chain.
255 */
256 if (get_nulls_value(node) != slot2)
257 goto begin;
258
259 if (result) {
260exact_match:
c31504dc 261 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
fddc17de
ED
262 result = NULL;
263 else if (unlikely(compute_score2(result, net, saddr, sport,
264 daddr, hnum, dif) < badness)) {
265 sock_put(result);
266 goto begin;
267 }
268 }
269 return result;
270}
271
fce82338 272struct sock *__udp6_lib_lookup(struct net *net,
88440ae7
BS
273 const struct in6_addr *saddr, __be16 sport,
274 const struct in6_addr *daddr, __be16 dport,
645ca708 275 int dif, struct udp_table *udptable)
1da177e4 276{
271b72c7 277 struct sock *sk, *result;
88ab1932 278 struct hlist_nulls_node *node;
1da177e4 279 unsigned short hnum = ntohs(dport);
fddc17de
ED
280 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
281 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
72289b96
TH
282 int score, badness, matches = 0, reuseport = 0;
283 u32 hash = 0;
645ca708 284
271b72c7 285 rcu_read_lock();
fddc17de
ED
286 if (hslot->count > 10) {
287 hash2 = udp6_portaddr_hash(net, daddr, hnum);
288 slot2 = hash2 & udptable->mask;
289 hslot2 = &udptable->hash2[slot2];
290 if (hslot->count < hslot2->count)
291 goto begin;
292
293 result = udp6_lib_lookup2(net, saddr, sport,
294 daddr, hnum, dif,
295 hslot2, slot2);
296 if (!result) {
297 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
298 slot2 = hash2 & udptable->mask;
299 hslot2 = &udptable->hash2[slot2];
300 if (hslot->count < hslot2->count)
301 goto begin;
302
1223c67c
JBD
303 result = udp6_lib_lookup2(net, saddr, sport,
304 &in6addr_any, hnum, dif,
fddc17de
ED
305 hslot2, slot2);
306 }
307 rcu_read_unlock();
308 return result;
309 }
271b72c7
ED
310begin:
311 result = NULL;
312 badness = -1;
88ab1932 313 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
645ca708
ED
314 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
315 if (score > badness) {
316 result = sk;
317 badness = score;
72289b96
TH
318 reuseport = sk->sk_reuseport;
319 if (reuseport) {
b50026b5
HFS
320 hash = udp6_ehashfn(net, daddr, hnum,
321 saddr, sport);
72289b96
TH
322 matches = 1;
323 }
324 } else if (score == badness && reuseport) {
325 matches++;
8fc54f68 326 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
327 result = sk;
328 hash = next_pseudo_random32(hash);
1da177e4
LT
329 }
330 }
88ab1932
ED
331 /*
332 * if the nulls value we got at the end of this lookup is
333 * not the expected one, we must restart lookup.
334 * We probably met an item that was moved to another chain.
335 */
fddc17de 336 if (get_nulls_value(node) != slot)
88ab1932
ED
337 goto begin;
338
271b72c7 339 if (result) {
c31504dc 340 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
271b72c7
ED
341 result = NULL;
342 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
343 daddr, dport, dif) < badness)) {
344 sock_put(result);
345 goto begin;
346 }
347 }
348 rcu_read_unlock();
1da177e4
LT
349 return result;
350}
fce82338 351EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
1da177e4 352
607c4aaf
KK
353static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
354 __be16 sport, __be16 dport,
645ca708 355 struct udp_table *udptable)
607c4aaf 356{
23542618 357 struct sock *sk;
b71d1d42 358 const struct ipv6hdr *iph = ipv6_hdr(skb);
607c4aaf 359
23542618
KK
360 if (unlikely(sk = skb_steal_sock(skb)))
361 return sk;
adf30907
ED
362 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
363 &iph->daddr, dport, inet6_iif(skb),
364 udptable);
607c4aaf
KK
365}
366
aa976fc0
BS
367struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
368 const struct in6_addr *daddr, __be16 dport, int dif)
369{
370 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
371}
372EXPORT_SYMBOL_GPL(udp6_lib_lookup);
373
374
1da177e4 375/*
67ba4152
IM
376 * This should be easy, if there is something there we
377 * return it, otherwise we block.
1da177e4
LT
378 */
379
ba4e58ec 380int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
381 struct msghdr *msg, size_t len,
382 int noblock, int flags, int *addr_len)
383{
384 struct ipv6_pinfo *np = inet6_sk(sk);
385 struct inet_sock *inet = inet_sk(sk);
1ab1457c 386 struct sk_buff *skb;
59c2cdae 387 unsigned int ulen, copied;
3f518bf7 388 int peeked, off = 0;
759e5d00
HX
389 int err;
390 int is_udplite = IS_UDPLITE(sk);
f26ba175 391 int is_udp4;
8a74ad60 392 bool slow;
1da177e4 393
1da177e4 394 if (flags & MSG_ERRQUEUE)
85fbaa75 395 return ipv6_recv_error(sk, msg, len, addr_len);
1da177e4 396
4b340ae2 397 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
85fbaa75 398 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
4b340ae2 399
1da177e4 400try_again:
a59322be 401 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 402 &peeked, &off, &err);
1da177e4
LT
403 if (!skb)
404 goto out;
405
759e5d00 406 ulen = skb->len - sizeof(struct udphdr);
59c2cdae
DM
407 copied = len;
408 if (copied > ulen)
409 copied = ulen;
410 else if (copied < ulen)
1ab1457c 411 msg->msg_flags |= MSG_TRUNC;
1da177e4 412
f26ba175
WY
413 is_udp4 = (skb->protocol == htons(ETH_P_IP));
414
ba4e58ec 415 /*
759e5d00
HX
416 * If checksum is needed at all, try to do it while copying the
417 * data. If the data is truncated, or if we only want a partial
418 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 419 */
ba4e58ec 420
59c2cdae 421 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
759e5d00 422 if (udp_lib_checksum_complete(skb))
1da177e4 423 goto csum_copy_err;
ba4e58ec
GR
424 }
425
60476372 426 if (skb_csum_unnecessary(skb))
51f3d02b
DM
427 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
428 msg, copied);
ba4e58ec 429 else {
1da177e4
LT
430 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
431 if (err == -EINVAL)
432 goto csum_copy_err;
433 }
22911fc5
ED
434 if (unlikely(err)) {
435 trace_kfree_skb(skb, udpv6_recvmsg);
979402b1
ED
436 if (!peeked) {
437 atomic_inc(&sk->sk_drops);
438 if (is_udp4)
439 UDP_INC_STATS_USER(sock_net(sk),
440 UDP_MIB_INERRORS,
441 is_udplite);
442 else
443 UDP6_INC_STATS_USER(sock_net(sk),
444 UDP_MIB_INERRORS,
445 is_udplite);
446 }
1da177e4 447 goto out_free;
22911fc5 448 }
f26ba175
WY
449 if (!peeked) {
450 if (is_udp4)
451 UDP_INC_STATS_USER(sock_net(sk),
452 UDP_MIB_INDATAGRAMS, is_udplite);
453 else
454 UDP6_INC_STATS_USER(sock_net(sk),
455 UDP_MIB_INDATAGRAMS, is_udplite);
456 }
cb75994e 457
3b885787 458 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
459
460 /* Copy the address. */
461 if (msg->msg_name) {
342dfc30 462 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 463 sin6->sin6_family = AF_INET6;
4bedb452 464 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 465 sin6->sin6_flowinfo = 0;
1da177e4 466
842df073 467 if (is_udp4) {
b301e82c
BH
468 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
469 &sin6->sin6_addr);
842df073
HFS
470 sin6->sin6_scope_id = 0;
471 } else {
4e3fd7a0 472 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
473 sin6->sin6_scope_id =
474 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 475 inet6_iif(skb));
1da177e4 476 }
bceaa902 477 *addr_len = sizeof(*sin6);
1da177e4 478 }
4b261c75
HFS
479
480 if (np->rxopt.all)
481 ip6_datagram_recv_common_ctl(sk, msg, skb);
482
f26ba175 483 if (is_udp4) {
1da177e4
LT
484 if (inet->cmsg_flags)
485 ip_cmsg_recv(msg, skb);
486 } else {
487 if (np->rxopt.all)
4b261c75 488 ip6_datagram_recv_specific_ctl(sk, msg, skb);
1ab1457c 489 }
1da177e4 490
59c2cdae 491 err = copied;
1da177e4 492 if (flags & MSG_TRUNC)
759e5d00 493 err = ulen;
1da177e4
LT
494
495out_free:
9d410c79 496 skb_free_datagram_locked(sk, skb);
1da177e4
LT
497out:
498 return err;
499
500csum_copy_err:
8a74ad60 501 slow = lock_sock_fast(sk);
0856f939 502 if (!skb_kill_datagram(sk, skb, flags)) {
6a5dc9e5
ED
503 if (is_udp4) {
504 UDP_INC_STATS_USER(sock_net(sk),
505 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
506 UDP_INC_STATS_USER(sock_net(sk),
507 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5
ED
508 } else {
509 UDP6_INC_STATS_USER(sock_net(sk),
510 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
511 UDP6_INC_STATS_USER(sock_net(sk),
512 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 513 }
0856f939 514 }
8a74ad60 515 unlock_sock_fast(sk, slow);
1da177e4 516
32c90254 517 if (noblock)
1da177e4 518 return -EAGAIN;
9cfaa8de
XZ
519
520 /* starting over for a new packet */
521 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
522 goto try_again;
523}
524
ba4e58ec 525void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 526 u8 type, u8 code, int offset, __be32 info,
645ca708 527 struct udp_table *udptable)
1da177e4
LT
528{
529 struct ipv6_pinfo *np;
b71d1d42
ED
530 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
531 const struct in6_addr *saddr = &hdr->saddr;
532 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 533 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4
LT
534 struct sock *sk;
535 int err;
7304fe46 536 struct net *net = dev_net(skb->dev);
1da177e4 537
7304fe46 538 sk = __udp6_lib_lookup(net, daddr, uh->dest,
ba4e58ec 539 saddr, uh->source, inet6_iif(skb), udptable);
7304fe46
DJ
540 if (sk == NULL) {
541 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
542 ICMP6_MIB_INERRORS);
1da177e4 543 return;
7304fe46 544 }
1da177e4 545
93b36cf3
HFS
546 if (type == ICMPV6_PKT_TOOBIG) {
547 if (!ip6_sk_accept_pmtu(sk))
548 goto out;
81aded24 549 ip6_sk_update_pmtu(skb, sk, info);
93b36cf3 550 }
1a462d18 551 if (type == NDISC_REDIRECT) {
ec18d9a2 552 ip6_sk_redirect(skb, sk);
1a462d18
DJ
553 goto out;
554 }
81aded24 555
1da177e4
LT
556 np = inet6_sk(sk);
557
558 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
559 goto out;
560
561 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
562 goto out;
563
b1faf566 564 if (np->recverr)
1da177e4 565 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
b1faf566 566
1da177e4
LT
567 sk->sk_err = err;
568 sk->sk_error_report(sk);
569out:
570 sock_put(sk);
571}
572
f7ad74fe
BL
573static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
574{
575 int rc;
576
efe4208f 577 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 578 sock_rps_save_rxhash(sk, skb);
005ec974 579 sk_mark_napi_id(sk, skb);
2c8c56e1 580 sk_incoming_cpu_update(sk);
005ec974 581 }
f7ad74fe
BL
582
583 rc = sock_queue_rcv_skb(sk, skb);
584 if (rc < 0) {
585 int is_udplite = IS_UDPLITE(sk);
586
587 /* Note that an ENOMEM error is charged twice */
588 if (rc == -ENOMEM)
589 UDP6_INC_STATS_BH(sock_net(sk),
590 UDP_MIB_RCVBUFERRORS, is_udplite);
591 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
592 kfree_skb(skb);
593 return -1;
594 }
595 return 0;
596}
597
ba4e58ec 598static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 599 struct inet6_skb_parm *opt, u8 type,
67ba4152 600 u8 code, int offset, __be32 info)
ba4e58ec 601{
645ca708 602 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
603}
604
d7f3f621
BL
605static struct static_key udpv6_encap_needed __read_mostly;
606void udpv6_encap_enable(void)
607{
608 if (!static_key_enabled(&udpv6_encap_needed))
609 static_key_slow_inc(&udpv6_encap_needed);
610}
611EXPORT_SYMBOL(udpv6_encap_enable);
612
f7ad74fe 613int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 614{
ba4e58ec 615 struct udp_sock *up = udp_sk(sk);
a18135eb 616 int rc;
b2bf1e26 617 int is_udplite = IS_UDPLITE(sk);
a18135eb 618
ba4e58ec
GR
619 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
620 goto drop;
1da177e4 621
d7f3f621
BL
622 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
623 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
624
625 /*
626 * This is an encapsulation socket so pass the skb to
627 * the socket's udp_encap_rcv() hook. Otherwise, just
628 * fall through and pass this up the UDP socket.
629 * up->encap_rcv() returns the following value:
630 * =0 if skb was successfully passed to the encap
631 * handler or was discarded by it.
632 * >0 if skb should be passed on to UDP.
633 * <0 if skb should be resubmitted as proto -N
634 */
635
636 /* if we're overly short, let UDP handle it */
637 encap_rcv = ACCESS_ONCE(up->encap_rcv);
638 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
639 int ret;
640
0a80966b
TH
641 /* Verify checksum before giving to encap */
642 if (udp_lib_checksum_complete(skb))
643 goto csum_error;
644
d7f3f621
BL
645 ret = encap_rcv(sk, skb);
646 if (ret <= 0) {
647 UDP_INC_STATS_BH(sock_net(sk),
648 UDP_MIB_INDATAGRAMS,
649 is_udplite);
650 return -ret;
651 }
652 }
653
654 /* FALLTHROUGH -- it's a UDP Packet */
655 }
656
ba4e58ec
GR
657 /*
658 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
659 */
b2bf1e26 660 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
661
662 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
663 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
664 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
665 goto drop;
666 }
667 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
668 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
669 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
670 goto drop;
671 }
1da177e4
LT
672 }
673
33d480ce 674 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62 675 if (udp_lib_checksum_complete(skb))
6a5dc9e5 676 goto csum_error;
1ab6eb62 677 }
ba4e58ec 678
274f482d 679 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
3e215c8d
JL
680 UDP6_INC_STATS_BH(sock_net(sk),
681 UDP_MIB_RCVBUFERRORS, is_udplite);
cb80ef46 682 goto drop;
3e215c8d 683 }
cb80ef46 684
d826eb14 685 skb_dst_drop(skb);
cb75994e 686
cb80ef46
BL
687 bh_lock_sock(sk);
688 rc = 0;
689 if (!sock_owned_by_user(sk))
690 rc = __udpv6_queue_rcv_skb(sk, skb);
691 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
692 bh_unlock_sock(sk);
693 goto drop;
694 }
695 bh_unlock_sock(sk);
f7ad74fe
BL
696
697 return rc;
3e215c8d 698
6a5dc9e5
ED
699csum_error:
700 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 701drop:
ef28d1a2 702 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 703 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
704 kfree_skb(skb);
705 return -1;
1da177e4
LT
706}
707
5cf3d461
DH
708static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
709 __be16 loc_port, const struct in6_addr *loc_addr,
710 __be16 rmt_port, const struct in6_addr *rmt_addr,
711 int dif, unsigned short hnum)
1da177e4 712{
5cf3d461 713 struct inet_sock *inet = inet_sk(sk);
1da177e4 714
5cf3d461
DH
715 if (!net_eq(sock_net(sk), net))
716 return false;
717
718 if (udp_sk(sk)->udp_port_hash != hnum ||
719 sk->sk_family != PF_INET6 ||
720 (inet->inet_dport && inet->inet_dport != rmt_port) ||
721 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
722 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
723 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
724 return false;
725 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
726 return false;
727 return true;
1da177e4
LT
728}
729
a1ab77f9
ED
730static void flush_stack(struct sock **stack, unsigned int count,
731 struct sk_buff *skb, unsigned int final)
732{
cb80ef46 733 struct sk_buff *skb1 = NULL;
a1ab77f9 734 struct sock *sk;
cb80ef46 735 unsigned int i;
a1ab77f9
ED
736
737 for (i = 0; i < count; i++) {
f6b8f32c 738 sk = stack[i];
cb80ef46
BL
739 if (likely(skb1 == NULL))
740 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
741 if (!skb1) {
742 atomic_inc(&sk->sk_drops);
743 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
744 IS_UDPLITE(sk));
745 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
746 IS_UDPLITE(sk));
a1ab77f9 747 }
cb80ef46
BL
748
749 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
750 skb1 = NULL;
2dc41cff 751 sock_put(sk);
a1ab77f9 752 }
cb80ef46
BL
753 if (unlikely(skb1))
754 kfree_skb(skb1);
a1ab77f9 755}
4068579e
TH
756
757static void udp6_csum_zero_error(struct sk_buff *skb)
758{
759 /* RFC 2460 section 8.1 says that we SHOULD log
760 * this error. Well, it is reasonable.
761 */
ba7a46f1
JP
762 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
763 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
764 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
765}
766
1da177e4
LT
767/*
768 * Note: called only from the BH handler context,
769 * so we don't need to lock the hashes.
770 */
e3163493 771static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 772 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 773 struct udp_table *udptable, int proto)
1da177e4 774{
a1ab77f9 775 struct sock *sk, *stack[256 / sizeof(struct sock *)];
4bedb452 776 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
777 struct hlist_nulls_node *node;
778 unsigned short hnum = ntohs(uh->dest);
779 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
780 int dif = inet6_iif(skb);
2dc41cff
DH
781 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
782 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
36cbb245 783 bool inner_flushed = false;
2dc41cff
DH
784
785 if (use_hash2) {
786 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
787 udp_table.mask;
788 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
789start_lookup:
790 hslot = &udp_table.hash2[hash2];
791 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
792 }
1da177e4 793
645ca708 794 spin_lock(&hslot->lock);
2dc41cff 795 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
5cf3d461
DH
796 if (__udp_v6_is_mcast_sock(net, sk,
797 uh->dest, daddr,
798 uh->source, saddr,
799 dif, hnum) &&
800 /* If zero checksum and no_check is not on for
801 * the socket then skip it.
802 */
803 (uh->check || udp_sk(sk)->no_check6_rx)) {
804 if (unlikely(count == ARRAY_SIZE(stack))) {
805 flush_stack(stack, count, skb, ~0);
36cbb245 806 inner_flushed = true;
5cf3d461
DH
807 count = 0;
808 }
4068579e 809 stack[count++] = sk;
2dc41cff 810 sock_hold(sk);
95766fff 811 }
1da177e4 812 }
a1ab77f9 813
645ca708 814 spin_unlock(&hslot->lock);
a1ab77f9 815
2dc41cff
DH
816 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
817 if (use_hash2 && hash2 != hash2_any) {
818 hash2 = hash2_any;
819 goto start_lookup;
820 }
821
a1ab77f9
ED
822 if (count) {
823 flush_stack(stack, count, skb, count - 1);
a1ab77f9 824 } else {
36cbb245
RJ
825 if (!inner_flushed)
826 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
827 proto == IPPROTO_UDPLITE);
828 consume_skb(skb);
a1ab77f9 829 }
ba4e58ec 830 return 0;
1da177e4
LT
831}
832
645ca708 833int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 834 int proto)
1da177e4 835{
3ffe533c 836 struct net *net = dev_net(skb->dev);
1da177e4 837 struct sock *sk;
1ab1457c 838 struct udphdr *uh;
b71d1d42 839 const struct in6_addr *saddr, *daddr;
1da177e4
LT
840 u32 ulen = 0;
841
842 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 843 goto discard;
1da177e4 844
0660e03f
ACM
845 saddr = &ipv6_hdr(skb)->saddr;
846 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 847 uh = udp_hdr(skb);
1da177e4
LT
848
849 ulen = ntohs(uh->len);
ba4e58ec
GR
850 if (ulen > skb->len)
851 goto short_packet;
1da177e4 852
759e5d00
HX
853 if (proto == IPPROTO_UDP) {
854 /* UDP validates ulen. */
1da177e4 855
ba4e58ec
GR
856 /* Check for jumbo payload */
857 if (ulen == 0)
858 ulen = skb->len;
1da177e4 859
ba4e58ec
GR
860 if (ulen < sizeof(*uh))
861 goto short_packet;
1da177e4 862
ba4e58ec
GR
863 if (ulen < skb->len) {
864 if (pskb_trim_rcsum(skb, ulen))
865 goto short_packet;
0660e03f
ACM
866 saddr = &ipv6_hdr(skb)->saddr;
867 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 868 uh = udp_hdr(skb);
ba4e58ec 869 }
ba4e58ec 870 }
1da177e4 871
759e5d00 872 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 873 goto csum_error;
759e5d00 874
1ab1457c
YH
875 /*
876 * Multicast receive code
1da177e4 877 */
ba4e58ec 878 if (ipv6_addr_is_multicast(daddr))
e3163493 879 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 880 saddr, daddr, udptable, proto);
1da177e4
LT
881
882 /* Unicast */
883
1ab1457c 884 /*
1da177e4
LT
885 * check socket cache ... must talk to Alan about his plans
886 * for sock caches... i'll skip this for now.
887 */
607c4aaf 888 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
cb80ef46 889 if (sk != NULL) {
a5b50476
ET
890 int ret;
891
1c19448c 892 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
79e0f1c9 893 sock_put(sk);
4068579e
TH
894 udp6_csum_zero_error(skb);
895 goto csum_error;
896 }
897
2abb7cdc
TH
898 if (udp_sk(sk)->convert_csum && uh->check && !IS_UDPLITE(sk))
899 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
900 ip6_compute_pseudo);
901
a5b50476 902 ret = udpv6_queue_rcv_skb(sk, skb);
cb80ef46 903 sock_put(sk);
1da177e4 904
cb80ef46
BL
905 /* a return value > 0 means to resubmit the input, but
906 * it wants the return to be -protocol, or 0
907 */
908 if (ret > 0)
909 return -ret;
1da177e4 910
add459aa 911 return 0;
1da177e4 912 }
1ab1457c 913
4068579e
TH
914 if (!uh->check) {
915 udp6_csum_zero_error(skb);
916 goto csum_error;
917 }
918
cb80ef46 919 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 920 goto discard;
cb80ef46
BL
921
922 if (udp_lib_checksum_complete(skb))
6a5dc9e5 923 goto csum_error;
cb80ef46
BL
924
925 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
926 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
927
928 kfree_skb(skb);
add459aa 929 return 0;
1da177e4 930
1ab1457c 931short_packet:
ba7a46f1
JP
932 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
933 proto == IPPROTO_UDPLITE ? "-Lite" : "",
934 saddr, ntohs(uh->source),
935 ulen, skb->len,
936 daddr, ntohs(uh->dest));
6a5dc9e5
ED
937 goto discard;
938csum_error:
939 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 940discard:
ef28d1a2 941 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 942 kfree_skb(skb);
add459aa 943 return 0;
1da177e4 944}
ba4e58ec 945
e5bbef20 946static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 947{
645ca708 948 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
949}
950
1da177e4
LT
951/*
952 * Throw away all pending data and cancel the corking. Socket is locked.
953 */
954static void udp_v6_flush_pending_frames(struct sock *sk)
955{
956 struct udp_sock *up = udp_sk(sk);
957
36d926b9
DL
958 if (up->pending == AF_INET)
959 udp_flush_pending_frames(sk);
960 else if (up->pending) {
1da177e4
LT
961 up->len = 0;
962 up->pending = 0;
963 ip6_flush_pending_frames(sk);
1ab1457c 964 }
1da177e4
LT
965}
966
493c6be3 967/**
67ba4152
IM
968 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
969 * @sk: socket we are sending on
970 * @skb: sk_buff containing the filled-in UDP header
971 * (checksum field must be zeroed out)
493c6be3
SS
972 */
973static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
974 const struct in6_addr *saddr,
975 const struct in6_addr *daddr, int len)
976{
977 unsigned int offset;
978 struct udphdr *uh = udp_hdr(skb);
979 __wsum csum = 0;
980
981 if (skb_queue_len(&sk->sk_write_queue) == 1) {
982 /* Only one fragment on the socket. */
983 skb->csum_start = skb_transport_header(skb) - skb->head;
984 skb->csum_offset = offsetof(struct udphdr, check);
985 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
986 } else {
987 /*
988 * HW-checksum won't work as there are two or more
989 * fragments on the socket so that all csums of sk_buffs
990 * should be together
991 */
992 offset = skb_transport_offset(skb);
993 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
994
995 skb->ip_summed = CHECKSUM_NONE;
996
997 skb_queue_walk(&sk->sk_write_queue, skb) {
998 csum = csum_add(csum, skb->csum);
999 }
1000
1001 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1002 csum);
1003 if (uh->check == 0)
1004 uh->check = CSUM_MANGLED_0;
1005 }
1006}
1007
1da177e4
LT
1008/*
1009 * Sending
1010 */
1011
4c0a6cb0 1012static int udp_v6_push_pending_frames(struct sock *sk)
1da177e4
LT
1013{
1014 struct sk_buff *skb;
1015 struct udphdr *uh;
4c0a6cb0 1016 struct udp_sock *up = udp_sk(sk);
1da177e4 1017 struct inet_sock *inet = inet_sk(sk);
8822b64a 1018 struct flowi6 *fl6;
1da177e4 1019 int err = 0;
b2bf1e26 1020 int is_udplite = IS_UDPLITE(sk);
868c86bc 1021 __wsum csum = 0;
1da177e4 1022
8822b64a
HFS
1023 if (up->pending == AF_INET)
1024 return udp_push_pending_frames(sk);
1025
1026 fl6 = &inet->cork.fl.u.ip6;
1027
1da177e4
LT
1028 /* Grab the skbuff where UDP header space exists. */
1029 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
1030 goto out;
1031
1032 /*
1033 * Create a UDP header
1034 */
4bedb452 1035 uh = udp_hdr(skb);
1958b856
DM
1036 uh->source = fl6->fl6_sport;
1037 uh->dest = fl6->fl6_dport;
1da177e4
LT
1038 uh->len = htons(up->len);
1039 uh->check = 0;
1040
b2bf1e26 1041 if (is_udplite)
ba4e58ec 1042 csum = udplite_csum_outgoing(sk, skb);
1c19448c 1043 else if (up->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
1044 skb->ip_summed = CHECKSUM_NONE;
1045 goto send;
1046 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
4c9483b2 1047 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
493c6be3
SS
1048 up->len);
1049 goto send;
1050 } else
ba4e58ec 1051 csum = udp_csum_outgoing(sk, skb);
1da177e4 1052
ba4e58ec 1053 /* add protocol-dependent pseudo-header */
4c9483b2
DM
1054 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1055 up->len, fl6->flowi6_proto, csum);
1da177e4 1056 if (uh->check == 0)
f6ab0288 1057 uh->check = CSUM_MANGLED_0;
1da177e4 1058
493c6be3 1059send:
1da177e4 1060 err = ip6_push_pending_frames(sk);
6ce9e7b5
ED
1061 if (err) {
1062 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1063 UDP6_INC_STATS_USER(sock_net(sk),
1064 UDP_MIB_SNDBUFERRORS, is_udplite);
1065 err = 0;
1066 }
1067 } else
1068 UDP6_INC_STATS_USER(sock_net(sk),
1069 UDP_MIB_OUTDATAGRAMS, is_udplite);
1da177e4
LT
1070out:
1071 up->len = 0;
1072 up->pending = 0;
1073 return err;
1074}
1075
ba4e58ec 1076int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
1077 struct msghdr *msg, size_t len)
1078{
1079 struct ipv6_txoptions opt_space;
1080 struct udp_sock *up = udp_sk(sk);
1081 struct inet_sock *inet = inet_sk(sk);
1082 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1083 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1084 struct in6_addr *daddr, *final_p, final;
1da177e4
LT
1085 struct ipv6_txoptions *opt = NULL;
1086 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1087 struct flowi6 fl6;
1da177e4
LT
1088 struct dst_entry *dst;
1089 int addr_len = msg->msg_namelen;
1090 int ulen = len;
1091 int hlimit = -1;
41a1f8ea 1092 int tclass = -1;
13b52cd4 1093 int dontfrag = -1;
1da177e4
LT
1094 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1095 int err;
987905de 1096 int connected = 0;
b2bf1e26 1097 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1098 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1da177e4
LT
1099
1100 /* destination address check */
1101 if (sin6) {
1102 if (addr_len < offsetof(struct sockaddr, sa_data))
1103 return -EINVAL;
1104
1105 switch (sin6->sin6_family) {
1106 case AF_INET6:
1107 if (addr_len < SIN6_LEN_RFC2133)
1108 return -EINVAL;
1109 daddr = &sin6->sin6_addr;
1110 break;
1111 case AF_INET:
1112 goto do_udp_sendmsg;
1113 case AF_UNSPEC:
1114 msg->msg_name = sin6 = NULL;
1115 msg->msg_namelen = addr_len = 0;
1116 daddr = NULL;
1117 break;
1118 default:
1119 return -EINVAL;
1120 }
1121 } else if (!up->pending) {
1122 if (sk->sk_state != TCP_ESTABLISHED)
1123 return -EDESTADDRREQ;
efe4208f 1124 daddr = &sk->sk_v6_daddr;
1ab1457c 1125 } else
1da177e4
LT
1126 daddr = NULL;
1127
1128 if (daddr) {
e773e4fa 1129 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1130 struct sockaddr_in sin;
1131 sin.sin_family = AF_INET;
c720c7e8 1132 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1133 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1134 msg->msg_name = &sin;
1135 msg->msg_namelen = sizeof(sin);
1136do_udp_sendmsg:
1137 if (__ipv6_only_sock(sk))
1138 return -ENETUNREACH;
1139 return udp_sendmsg(iocb, sk, msg, len);
1140 }
1141 }
1142
1143 if (up->pending == AF_INET)
1144 return udp_sendmsg(iocb, sk, msg, len);
1145
1146 /* Rough check on arithmetic overflow,
b59e139b 1147 better check is made in ip6_append_data().
1da177e4
LT
1148 */
1149 if (len > INT_MAX - sizeof(struct udphdr))
1150 return -EMSGSIZE;
1ab1457c 1151
1da177e4
LT
1152 if (up->pending) {
1153 /*
1154 * There are pending frames.
1155 * The socket lock must be held while it's corked.
1156 */
1157 lock_sock(sk);
1158 if (likely(up->pending)) {
1159 if (unlikely(up->pending != AF_INET6)) {
1160 release_sock(sk);
1161 return -EAFNOSUPPORT;
1162 }
1163 dst = NULL;
1164 goto do_append_data;
1165 }
1166 release_sock(sk);
1167 }
1168 ulen += sizeof(struct udphdr);
1169
4c9483b2 1170 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1171
1172 if (sin6) {
1173 if (sin6->sin6_port == 0)
1174 return -EINVAL;
1175
1958b856 1176 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1177 daddr = &sin6->sin6_addr;
1178
1179 if (np->sndflow) {
4c9483b2
DM
1180 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1181 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1182 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1183 if (flowlabel == NULL)
1184 return -EINVAL;
1da177e4
LT
1185 }
1186 }
1187
1188 /*
1189 * Otherwise it will be difficult to maintain
1190 * sk->sk_dst_cache.
1191 */
1192 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1193 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1194 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1195
1196 if (addr_len >= sizeof(struct sockaddr_in6) &&
1197 sin6->sin6_scope_id &&
842df073 1198 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1199 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1200 } else {
1201 if (sk->sk_state != TCP_ESTABLISHED)
1202 return -EDESTADDRREQ;
1203
1958b856 1204 fl6.fl6_dport = inet->inet_dport;
efe4208f 1205 daddr = &sk->sk_v6_daddr;
4c9483b2 1206 fl6.flowlabel = np->flow_label;
987905de 1207 connected = 1;
1da177e4
LT
1208 }
1209
4c9483b2
DM
1210 if (!fl6.flowi6_oif)
1211 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1212
4c9483b2
DM
1213 if (!fl6.flowi6_oif)
1214 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1215
4c9483b2 1216 fl6.flowi6_mark = sk->sk_mark;
51953d5b 1217
1da177e4
LT
1218 if (msg->msg_controllen) {
1219 opt = &opt_space;
1220 memset(opt, 0, sizeof(struct ipv6_txoptions));
1221 opt->tot_len = sizeof(*opt);
1222
73df66f8
TP
1223 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1224 &hlimit, &tclass, &dontfrag);
1da177e4
LT
1225 if (err < 0) {
1226 fl6_sock_release(flowlabel);
1227 return err;
1228 }
4c9483b2
DM
1229 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1230 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1231 if (flowlabel == NULL)
1232 return -EINVAL;
1233 }
1234 if (!(opt->opt_nflen|opt->opt_flen))
1235 opt = NULL;
987905de 1236 connected = 0;
1da177e4
LT
1237 }
1238 if (opt == NULL)
1239 opt = np->opt;
df9890c3
YH
1240 if (flowlabel)
1241 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1242 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1243
4c9483b2 1244 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1245 if (!ipv6_addr_any(daddr))
4e3fd7a0 1246 fl6.daddr = *daddr;
876c7f41 1247 else
4c9483b2
DM
1248 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1249 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1250 fl6.saddr = np->saddr;
1958b856 1251 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1252
4c9483b2 1253 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1254 if (final_p)
987905de 1255 connected = 0;
1da177e4 1256
4c9483b2
DM
1257 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1258 fl6.flowi6_oif = np->mcast_oif;
987905de 1259 connected = 0;
c4062dfc
EH
1260 } else if (!fl6.flowi6_oif)
1261 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1262
4c9483b2 1263 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1264
0e0d44ab 1265 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1266 if (IS_ERR(dst)) {
1267 err = PTR_ERR(dst);
1268 dst = NULL;
1da177e4 1269 goto out;
14e50e57 1270 }
1da177e4 1271
5c98631c
LC
1272 if (hlimit < 0)
1273 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4 1274
e651f03a 1275 if (tclass < 0)
41a1f8ea 1276 tclass = np->tclass;
41a1f8ea 1277
1da177e4
LT
1278 if (msg->msg_flags&MSG_CONFIRM)
1279 goto do_confirm;
1280back_from_confirm:
1281
1282 lock_sock(sk);
1283 if (unlikely(up->pending)) {
1284 /* The socket is already corked while preparing it. */
1285 /* ... which is an evident application bug. --ANK */
1286 release_sock(sk);
1287
ba7a46f1 1288 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1289 err = -EINVAL;
1290 goto out;
1291 }
1292
1293 up->pending = AF_INET6;
1294
1295do_append_data:
e36d3ff9
JP
1296 if (dontfrag < 0)
1297 dontfrag = np->dontfrag;
1da177e4 1298 up->len += ulen;
ba4e58ec
GR
1299 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1300 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
4c9483b2 1301 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
67ba4152 1302 (struct rt6_info *)dst,
13b52cd4 1303 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1da177e4
LT
1304 if (err)
1305 udp_v6_flush_pending_frames(sk);
1306 else if (!corkreq)
4c0a6cb0 1307 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1308 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1309 up->pending = 0;
1da177e4 1310
a5e7c210
DM
1311 if (dst) {
1312 if (connected) {
1313 ip6_dst_store(sk, dst,
efe4208f
ED
1314 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1315 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1316#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1317 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1318 &np->saddr :
1319#endif
1320 NULL);
a5e7c210
DM
1321 } else {
1322 dst_release(dst);
1323 }
a3c96089 1324 dst = NULL;
a5e7c210
DM
1325 }
1326
1da177e4
LT
1327 if (err > 0)
1328 err = np->recverr ? net_xmit_errno(err) : 0;
1329 release_sock(sk);
1330out:
a3c96089 1331 dst_release(dst);
1da177e4 1332 fl6_sock_release(flowlabel);
cd562c98 1333 if (!err)
1da177e4 1334 return len;
a18135eb
DM
1335 /*
1336 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1337 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1338 * we don't have a good statistic (IpOutDiscards but it can be too many
1339 * things). We could add another new stat but at least for now that
1340 * seems like overkill.
1341 */
1342 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
235b9f7a
PE
1343 UDP6_INC_STATS_USER(sock_net(sk),
1344 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1345 }
1da177e4
LT
1346 return err;
1347
1348do_confirm:
1349 dst_confirm(dst);
1350 if (!(msg->msg_flags&MSG_PROBE) || len)
1351 goto back_from_confirm;
1352 err = 0;
1353 goto out;
1354}
1355
7d06b2e0 1356void udpv6_destroy_sock(struct sock *sk)
1da177e4 1357{
44046a59 1358 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1359 lock_sock(sk);
1360 udp_v6_flush_pending_frames(sk);
1361 release_sock(sk);
1362
44046a59
TP
1363 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1364 void (*encap_destroy)(struct sock *sk);
1365 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1366 if (encap_destroy)
1367 encap_destroy(sk);
1368 }
1369
1da177e4 1370 inet6_destroy_sock(sk);
1da177e4
LT
1371}
1372
1373/*
1374 * Socket option code for UDP
1375 */
ba4e58ec 1376int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1377 char __user *optval, unsigned int optlen)
3fdadf7d 1378{
db8dac20 1379 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1380 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1381 udp_v6_push_pending_frames);
ba4e58ec 1382 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1383}
1384
1385#ifdef CONFIG_COMPAT
ba4e58ec 1386int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1387 char __user *optval, unsigned int optlen)
3fdadf7d 1388{
db8dac20 1389 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1390 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1391 udp_v6_push_pending_frames);
ba4e58ec 1392 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1393}
1394#endif
1395
ba4e58ec
GR
1396int udpv6_getsockopt(struct sock *sk, int level, int optname,
1397 char __user *optval, int __user *optlen)
3fdadf7d 1398{
db8dac20 1399 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1400 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1401 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1402}
1403
1404#ifdef CONFIG_COMPAT
ba4e58ec
GR
1405int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1406 char __user *optval, int __user *optlen)
3fdadf7d 1407{
db8dac20 1408 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1409 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1410 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1411}
1412#endif
1413
41135cc8 1414static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1415 .handler = udpv6_rcv,
1416 .err_handler = udpv6_err,
1417 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1418};
1419
1420/* ------------------------------------------------------------------------ */
1421#ifdef CONFIG_PROC_FS
ba4e58ec 1422int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1423{
17ef66af
LC
1424 if (v == SEQ_START_TOKEN) {
1425 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1426 } else {
1427 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1428 struct inet_sock *inet = inet_sk(v);
1429 __u16 srcp = ntohs(inet->inet_sport);
1430 __u16 destp = ntohs(inet->inet_dport);
1431 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1432 }
1da177e4
LT
1433 return 0;
1434}
1435
73cb88ec
AV
1436static const struct file_operations udp6_afinfo_seq_fops = {
1437 .owner = THIS_MODULE,
1438 .open = udp_seq_open,
1439 .read = seq_read,
1440 .llseek = seq_lseek,
1441 .release = seq_release_net
1442};
1443
1da177e4 1444static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1445 .name = "udp6",
1446 .family = AF_INET6,
645ca708 1447 .udp_table = &udp_table,
73cb88ec 1448 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1449 .seq_ops = {
1450 .show = udp6_seq_show,
1451 },
1da177e4
LT
1452};
1453
2c8c1e72 1454int __net_init udp6_proc_init(struct net *net)
1da177e4 1455{
0c96d8c5 1456 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1457}
1458
0c96d8c5
DL
1459void udp6_proc_exit(struct net *net) {
1460 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1461}
1462#endif /* CONFIG_PROC_FS */
1463
f77d6021
ED
1464void udp_v6_clear_sk(struct sock *sk, int size)
1465{
1466 struct inet_sock *inet = inet_sk(sk);
1467
1468 /* we do not want to clear pinet6 field, because of RCU lookups */
1469 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1470
1471 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1472 memset(&inet->pinet6 + 1, 0, size);
1473}
1474
1da177e4
LT
1475/* ------------------------------------------------------------------------ */
1476
1477struct proto udpv6_prot = {
543d9cfe
ACM
1478 .name = "UDPv6",
1479 .owner = THIS_MODULE,
ba4e58ec 1480 .close = udp_lib_close,
543d9cfe
ACM
1481 .connect = ip6_datagram_connect,
1482 .disconnect = udp_disconnect,
1483 .ioctl = udp_ioctl,
1484 .destroy = udpv6_destroy_sock,
1485 .setsockopt = udpv6_setsockopt,
1486 .getsockopt = udpv6_getsockopt,
1487 .sendmsg = udpv6_sendmsg,
1488 .recvmsg = udpv6_recvmsg,
f7ad74fe 1489 .backlog_rcv = __udpv6_queue_rcv_skb,
ba4e58ec
GR
1490 .hash = udp_lib_hash,
1491 .unhash = udp_lib_unhash,
719f8358 1492 .rehash = udp_v6_rehash,
543d9cfe 1493 .get_port = udp_v6_get_port,
95766fff
HA
1494 .memory_allocated = &udp_memory_allocated,
1495 .sysctl_mem = sysctl_udp_mem,
1496 .sysctl_wmem = &sysctl_udp_wmem_min,
1497 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1498 .obj_size = sizeof(struct udp6_sock),
271b72c7 1499 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1500 .h.udp_table = &udp_table,
3fdadf7d 1501#ifdef CONFIG_COMPAT
543d9cfe
ACM
1502 .compat_setsockopt = compat_udpv6_setsockopt,
1503 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1504#endif
f77d6021 1505 .clear_sk = udp_v6_clear_sk,
1da177e4
LT
1506};
1507
1da177e4
LT
1508static struct inet_protosw udpv6_protosw = {
1509 .type = SOCK_DGRAM,
1510 .protocol = IPPROTO_UDP,
1511 .prot = &udpv6_prot,
1512 .ops = &inet6_dgram_ops,
1da177e4
LT
1513 .flags = INET_PROTOSW_PERMANENT,
1514};
1515
1516
7f4e4868 1517int __init udpv6_init(void)
1da177e4 1518{
7f4e4868
DL
1519 int ret;
1520
3336288a
VY
1521 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1522 if (ret)
c6b641a4 1523 goto out;
3336288a 1524
7f4e4868
DL
1525 ret = inet6_register_protosw(&udpv6_protosw);
1526 if (ret)
1527 goto out_udpv6_protocol;
1528out:
1529 return ret;
1530
1531out_udpv6_protocol:
1532 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1533 goto out;
1534}
1535
09f7709f 1536void udpv6_exit(void)
7f4e4868
DL
1537{
1538 inet6_unregister_protosw(&udpv6_protosw);
1539 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1540}