enic: use netdev_rss_key_fill() helper
[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
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
e5d08d71
IM
360 sk = skb_steal_sock(skb);
361 if (unlikely(sk))
23542618 362 return sk;
adf30907
ED
363 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
364 &iph->daddr, dport, inet6_iif(skb),
365 udptable);
607c4aaf
KK
366}
367
aa976fc0
BS
368struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
369 const struct in6_addr *daddr, __be16 dport, int dif)
370{
371 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
372}
373EXPORT_SYMBOL_GPL(udp6_lib_lookup);
374
375
1da177e4 376/*
67ba4152
IM
377 * This should be easy, if there is something there we
378 * return it, otherwise we block.
1da177e4
LT
379 */
380
ba4e58ec 381int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
382 struct msghdr *msg, size_t len,
383 int noblock, int flags, int *addr_len)
384{
385 struct ipv6_pinfo *np = inet6_sk(sk);
386 struct inet_sock *inet = inet_sk(sk);
1ab1457c 387 struct sk_buff *skb;
59c2cdae 388 unsigned int ulen, copied;
3f518bf7 389 int peeked, off = 0;
759e5d00
HX
390 int err;
391 int is_udplite = IS_UDPLITE(sk);
f26ba175 392 int is_udp4;
8a74ad60 393 bool slow;
1da177e4 394
1da177e4 395 if (flags & MSG_ERRQUEUE)
85fbaa75 396 return ipv6_recv_error(sk, msg, len, addr_len);
1da177e4 397
4b340ae2 398 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
85fbaa75 399 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
4b340ae2 400
1da177e4 401try_again:
a59322be 402 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 403 &peeked, &off, &err);
1da177e4
LT
404 if (!skb)
405 goto out;
406
759e5d00 407 ulen = skb->len - sizeof(struct udphdr);
59c2cdae
DM
408 copied = len;
409 if (copied > ulen)
410 copied = ulen;
411 else if (copied < ulen)
1ab1457c 412 msg->msg_flags |= MSG_TRUNC;
1da177e4 413
f26ba175
WY
414 is_udp4 = (skb->protocol == htons(ETH_P_IP));
415
ba4e58ec 416 /*
759e5d00
HX
417 * If checksum is needed at all, try to do it while copying the
418 * data. If the data is truncated, or if we only want a partial
419 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 420 */
ba4e58ec 421
59c2cdae 422 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
759e5d00 423 if (udp_lib_checksum_complete(skb))
1da177e4 424 goto csum_copy_err;
ba4e58ec
GR
425 }
426
60476372 427 if (skb_csum_unnecessary(skb))
51f3d02b
DM
428 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
429 msg, copied);
ba4e58ec 430 else {
1da177e4
LT
431 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
432 if (err == -EINVAL)
433 goto csum_copy_err;
434 }
22911fc5
ED
435 if (unlikely(err)) {
436 trace_kfree_skb(skb, udpv6_recvmsg);
979402b1
ED
437 if (!peeked) {
438 atomic_inc(&sk->sk_drops);
439 if (is_udp4)
440 UDP_INC_STATS_USER(sock_net(sk),
441 UDP_MIB_INERRORS,
442 is_udplite);
443 else
444 UDP6_INC_STATS_USER(sock_net(sk),
445 UDP_MIB_INERRORS,
446 is_udplite);
447 }
1da177e4 448 goto out_free;
22911fc5 449 }
f26ba175
WY
450 if (!peeked) {
451 if (is_udp4)
452 UDP_INC_STATS_USER(sock_net(sk),
453 UDP_MIB_INDATAGRAMS, is_udplite);
454 else
455 UDP6_INC_STATS_USER(sock_net(sk),
456 UDP_MIB_INDATAGRAMS, is_udplite);
457 }
cb75994e 458
3b885787 459 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
460
461 /* Copy the address. */
462 if (msg->msg_name) {
342dfc30 463 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 464 sin6->sin6_family = AF_INET6;
4bedb452 465 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 466 sin6->sin6_flowinfo = 0;
1da177e4 467
842df073 468 if (is_udp4) {
b301e82c
BH
469 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
470 &sin6->sin6_addr);
842df073
HFS
471 sin6->sin6_scope_id = 0;
472 } else {
4e3fd7a0 473 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
474 sin6->sin6_scope_id =
475 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 476 inet6_iif(skb));
1da177e4 477 }
bceaa902 478 *addr_len = sizeof(*sin6);
1da177e4 479 }
4b261c75
HFS
480
481 if (np->rxopt.all)
482 ip6_datagram_recv_common_ctl(sk, msg, skb);
483
f26ba175 484 if (is_udp4) {
1da177e4
LT
485 if (inet->cmsg_flags)
486 ip_cmsg_recv(msg, skb);
487 } else {
488 if (np->rxopt.all)
4b261c75 489 ip6_datagram_recv_specific_ctl(sk, msg, skb);
1ab1457c 490 }
1da177e4 491
59c2cdae 492 err = copied;
1da177e4 493 if (flags & MSG_TRUNC)
759e5d00 494 err = ulen;
1da177e4
LT
495
496out_free:
9d410c79 497 skb_free_datagram_locked(sk, skb);
1da177e4
LT
498out:
499 return err;
500
501csum_copy_err:
8a74ad60 502 slow = lock_sock_fast(sk);
0856f939 503 if (!skb_kill_datagram(sk, skb, flags)) {
6a5dc9e5
ED
504 if (is_udp4) {
505 UDP_INC_STATS_USER(sock_net(sk),
506 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
507 UDP_INC_STATS_USER(sock_net(sk),
508 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5
ED
509 } else {
510 UDP6_INC_STATS_USER(sock_net(sk),
511 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
512 UDP6_INC_STATS_USER(sock_net(sk),
513 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 514 }
0856f939 515 }
8a74ad60 516 unlock_sock_fast(sk, slow);
1da177e4 517
32c90254 518 if (noblock)
1da177e4 519 return -EAGAIN;
9cfaa8de
XZ
520
521 /* starting over for a new packet */
522 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
523 goto try_again;
524}
525
ba4e58ec 526void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 527 u8 type, u8 code, int offset, __be32 info,
645ca708 528 struct udp_table *udptable)
1da177e4
LT
529{
530 struct ipv6_pinfo *np;
b71d1d42
ED
531 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
532 const struct in6_addr *saddr = &hdr->saddr;
533 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 534 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4
LT
535 struct sock *sk;
536 int err;
7304fe46 537 struct net *net = dev_net(skb->dev);
1da177e4 538
7304fe46 539 sk = __udp6_lib_lookup(net, daddr, uh->dest,
ba4e58ec 540 saddr, uh->source, inet6_iif(skb), udptable);
7304fe46
DJ
541 if (sk == NULL) {
542 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
543 ICMP6_MIB_INERRORS);
1da177e4 544 return;
7304fe46 545 }
1da177e4 546
93b36cf3
HFS
547 if (type == ICMPV6_PKT_TOOBIG) {
548 if (!ip6_sk_accept_pmtu(sk))
549 goto out;
81aded24 550 ip6_sk_update_pmtu(skb, sk, info);
93b36cf3 551 }
1a462d18 552 if (type == NDISC_REDIRECT) {
ec18d9a2 553 ip6_sk_redirect(skb, sk);
1a462d18
DJ
554 goto out;
555 }
81aded24 556
1da177e4
LT
557 np = inet6_sk(sk);
558
559 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
560 goto out;
561
562 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
563 goto out;
564
b1faf566 565 if (np->recverr)
1da177e4 566 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
b1faf566 567
1da177e4
LT
568 sk->sk_err = err;
569 sk->sk_error_report(sk);
570out:
571 sock_put(sk);
572}
573
f7ad74fe
BL
574static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
575{
576 int rc;
577
efe4208f 578 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 579 sock_rps_save_rxhash(sk, skb);
005ec974 580 sk_mark_napi_id(sk, skb);
2c8c56e1 581 sk_incoming_cpu_update(sk);
005ec974 582 }
f7ad74fe
BL
583
584 rc = sock_queue_rcv_skb(sk, skb);
585 if (rc < 0) {
586 int is_udplite = IS_UDPLITE(sk);
587
588 /* Note that an ENOMEM error is charged twice */
589 if (rc == -ENOMEM)
590 UDP6_INC_STATS_BH(sock_net(sk),
591 UDP_MIB_RCVBUFERRORS, is_udplite);
592 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
593 kfree_skb(skb);
594 return -1;
595 }
596 return 0;
597}
598
ba4e58ec 599static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 600 struct inet6_skb_parm *opt, u8 type,
67ba4152 601 u8 code, int offset, __be32 info)
ba4e58ec 602{
645ca708 603 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
604}
605
d7f3f621
BL
606static struct static_key udpv6_encap_needed __read_mostly;
607void udpv6_encap_enable(void)
608{
609 if (!static_key_enabled(&udpv6_encap_needed))
610 static_key_slow_inc(&udpv6_encap_needed);
611}
612EXPORT_SYMBOL(udpv6_encap_enable);
613
f7ad74fe 614int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 615{
ba4e58ec 616 struct udp_sock *up = udp_sk(sk);
a18135eb 617 int rc;
b2bf1e26 618 int is_udplite = IS_UDPLITE(sk);
a18135eb 619
ba4e58ec
GR
620 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
621 goto drop;
1da177e4 622
d7f3f621
BL
623 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
624 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
625
626 /*
627 * This is an encapsulation socket so pass the skb to
628 * the socket's udp_encap_rcv() hook. Otherwise, just
629 * fall through and pass this up the UDP socket.
630 * up->encap_rcv() returns the following value:
631 * =0 if skb was successfully passed to the encap
632 * handler or was discarded by it.
633 * >0 if skb should be passed on to UDP.
634 * <0 if skb should be resubmitted as proto -N
635 */
636
637 /* if we're overly short, let UDP handle it */
638 encap_rcv = ACCESS_ONCE(up->encap_rcv);
639 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
640 int ret;
641
0a80966b
TH
642 /* Verify checksum before giving to encap */
643 if (udp_lib_checksum_complete(skb))
644 goto csum_error;
645
d7f3f621
BL
646 ret = encap_rcv(sk, skb);
647 if (ret <= 0) {
648 UDP_INC_STATS_BH(sock_net(sk),
649 UDP_MIB_INDATAGRAMS,
650 is_udplite);
651 return -ret;
652 }
653 }
654
655 /* FALLTHROUGH -- it's a UDP Packet */
656 }
657
ba4e58ec
GR
658 /*
659 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
660 */
b2bf1e26 661 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
662
663 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
664 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
665 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
666 goto drop;
667 }
668 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
669 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
670 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
671 goto drop;
672 }
1da177e4
LT
673 }
674
33d480ce 675 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62 676 if (udp_lib_checksum_complete(skb))
6a5dc9e5 677 goto csum_error;
1ab6eb62 678 }
ba4e58ec 679
274f482d 680 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
3e215c8d
JL
681 UDP6_INC_STATS_BH(sock_net(sk),
682 UDP_MIB_RCVBUFERRORS, is_udplite);
cb80ef46 683 goto drop;
3e215c8d 684 }
cb80ef46 685
d826eb14 686 skb_dst_drop(skb);
cb75994e 687
cb80ef46
BL
688 bh_lock_sock(sk);
689 rc = 0;
690 if (!sock_owned_by_user(sk))
691 rc = __udpv6_queue_rcv_skb(sk, skb);
692 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
693 bh_unlock_sock(sk);
694 goto drop;
695 }
696 bh_unlock_sock(sk);
f7ad74fe
BL
697
698 return rc;
3e215c8d 699
6a5dc9e5
ED
700csum_error:
701 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 702drop:
ef28d1a2 703 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 704 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
705 kfree_skb(skb);
706 return -1;
1da177e4
LT
707}
708
5cf3d461
DH
709static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
710 __be16 loc_port, const struct in6_addr *loc_addr,
711 __be16 rmt_port, const struct in6_addr *rmt_addr,
712 int dif, unsigned short hnum)
1da177e4 713{
5cf3d461 714 struct inet_sock *inet = inet_sk(sk);
1da177e4 715
5cf3d461
DH
716 if (!net_eq(sock_net(sk), net))
717 return false;
718
719 if (udp_sk(sk)->udp_port_hash != hnum ||
720 sk->sk_family != PF_INET6 ||
721 (inet->inet_dport && inet->inet_dport != rmt_port) ||
722 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
723 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
724 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
725 return false;
726 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
727 return false;
728 return true;
1da177e4
LT
729}
730
a1ab77f9
ED
731static void flush_stack(struct sock **stack, unsigned int count,
732 struct sk_buff *skb, unsigned int final)
733{
cb80ef46 734 struct sk_buff *skb1 = NULL;
a1ab77f9 735 struct sock *sk;
cb80ef46 736 unsigned int i;
a1ab77f9
ED
737
738 for (i = 0; i < count; i++) {
f6b8f32c 739 sk = stack[i];
cb80ef46
BL
740 if (likely(skb1 == NULL))
741 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
742 if (!skb1) {
743 atomic_inc(&sk->sk_drops);
744 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
745 IS_UDPLITE(sk));
746 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
747 IS_UDPLITE(sk));
a1ab77f9 748 }
cb80ef46
BL
749
750 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
751 skb1 = NULL;
2dc41cff 752 sock_put(sk);
a1ab77f9 753 }
cb80ef46
BL
754 if (unlikely(skb1))
755 kfree_skb(skb1);
a1ab77f9 756}
4068579e
TH
757
758static void udp6_csum_zero_error(struct sk_buff *skb)
759{
760 /* RFC 2460 section 8.1 says that we SHOULD log
761 * this error. Well, it is reasonable.
762 */
ba7a46f1
JP
763 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
764 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
765 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
766}
767
1da177e4
LT
768/*
769 * Note: called only from the BH handler context,
770 * so we don't need to lock the hashes.
771 */
e3163493 772static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 773 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 774 struct udp_table *udptable, int proto)
1da177e4 775{
a1ab77f9 776 struct sock *sk, *stack[256 / sizeof(struct sock *)];
4bedb452 777 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
778 struct hlist_nulls_node *node;
779 unsigned short hnum = ntohs(uh->dest);
780 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
781 int dif = inet6_iif(skb);
2dc41cff
DH
782 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
783 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
36cbb245 784 bool inner_flushed = false;
2dc41cff
DH
785
786 if (use_hash2) {
787 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
788 udp_table.mask;
789 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
790start_lookup:
791 hslot = &udp_table.hash2[hash2];
792 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
793 }
1da177e4 794
645ca708 795 spin_lock(&hslot->lock);
2dc41cff 796 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
5cf3d461
DH
797 if (__udp_v6_is_mcast_sock(net, sk,
798 uh->dest, daddr,
799 uh->source, saddr,
800 dif, hnum) &&
801 /* If zero checksum and no_check is not on for
802 * the socket then skip it.
803 */
804 (uh->check || udp_sk(sk)->no_check6_rx)) {
805 if (unlikely(count == ARRAY_SIZE(stack))) {
806 flush_stack(stack, count, skb, ~0);
36cbb245 807 inner_flushed = true;
5cf3d461
DH
808 count = 0;
809 }
4068579e 810 stack[count++] = sk;
2dc41cff 811 sock_hold(sk);
95766fff 812 }
1da177e4 813 }
a1ab77f9 814
645ca708 815 spin_unlock(&hslot->lock);
a1ab77f9 816
2dc41cff
DH
817 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
818 if (use_hash2 && hash2 != hash2_any) {
819 hash2 = hash2_any;
820 goto start_lookup;
821 }
822
a1ab77f9
ED
823 if (count) {
824 flush_stack(stack, count, skb, count - 1);
a1ab77f9 825 } else {
36cbb245
RJ
826 if (!inner_flushed)
827 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
828 proto == IPPROTO_UDPLITE);
829 consume_skb(skb);
a1ab77f9 830 }
ba4e58ec 831 return 0;
1da177e4
LT
832}
833
645ca708 834int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 835 int proto)
1da177e4 836{
3ffe533c 837 struct net *net = dev_net(skb->dev);
1da177e4 838 struct sock *sk;
1ab1457c 839 struct udphdr *uh;
b71d1d42 840 const struct in6_addr *saddr, *daddr;
1da177e4
LT
841 u32 ulen = 0;
842
843 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 844 goto discard;
1da177e4 845
0660e03f
ACM
846 saddr = &ipv6_hdr(skb)->saddr;
847 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 848 uh = udp_hdr(skb);
1da177e4
LT
849
850 ulen = ntohs(uh->len);
ba4e58ec
GR
851 if (ulen > skb->len)
852 goto short_packet;
1da177e4 853
759e5d00
HX
854 if (proto == IPPROTO_UDP) {
855 /* UDP validates ulen. */
1da177e4 856
ba4e58ec
GR
857 /* Check for jumbo payload */
858 if (ulen == 0)
859 ulen = skb->len;
1da177e4 860
ba4e58ec
GR
861 if (ulen < sizeof(*uh))
862 goto short_packet;
1da177e4 863
ba4e58ec
GR
864 if (ulen < skb->len) {
865 if (pskb_trim_rcsum(skb, ulen))
866 goto short_packet;
0660e03f
ACM
867 saddr = &ipv6_hdr(skb)->saddr;
868 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 869 uh = udp_hdr(skb);
ba4e58ec 870 }
ba4e58ec 871 }
1da177e4 872
759e5d00 873 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 874 goto csum_error;
759e5d00 875
1ab1457c
YH
876 /*
877 * Multicast receive code
1da177e4 878 */
ba4e58ec 879 if (ipv6_addr_is_multicast(daddr))
e3163493 880 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 881 saddr, daddr, udptable, proto);
1da177e4
LT
882
883 /* Unicast */
884
1ab1457c 885 /*
1da177e4
LT
886 * check socket cache ... must talk to Alan about his plans
887 * for sock caches... i'll skip this for now.
888 */
607c4aaf 889 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
cb80ef46 890 if (sk != NULL) {
a5b50476
ET
891 int ret;
892
1c19448c 893 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
79e0f1c9 894 sock_put(sk);
4068579e
TH
895 udp6_csum_zero_error(skb);
896 goto csum_error;
897 }
898
2abb7cdc
TH
899 if (udp_sk(sk)->convert_csum && uh->check && !IS_UDPLITE(sk))
900 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
901 ip6_compute_pseudo);
902
a5b50476 903 ret = udpv6_queue_rcv_skb(sk, skb);
cb80ef46 904 sock_put(sk);
1da177e4 905
cb80ef46
BL
906 /* a return value > 0 means to resubmit the input, but
907 * it wants the return to be -protocol, or 0
908 */
909 if (ret > 0)
910 return -ret;
1da177e4 911
add459aa 912 return 0;
1da177e4 913 }
1ab1457c 914
4068579e
TH
915 if (!uh->check) {
916 udp6_csum_zero_error(skb);
917 goto csum_error;
918 }
919
cb80ef46 920 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 921 goto discard;
cb80ef46
BL
922
923 if (udp_lib_checksum_complete(skb))
6a5dc9e5 924 goto csum_error;
cb80ef46
BL
925
926 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
927 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
928
929 kfree_skb(skb);
add459aa 930 return 0;
1da177e4 931
1ab1457c 932short_packet:
ba7a46f1
JP
933 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
934 proto == IPPROTO_UDPLITE ? "-Lite" : "",
935 saddr, ntohs(uh->source),
936 ulen, skb->len,
937 daddr, ntohs(uh->dest));
6a5dc9e5
ED
938 goto discard;
939csum_error:
940 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 941discard:
ef28d1a2 942 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 943 kfree_skb(skb);
add459aa 944 return 0;
1da177e4 945}
ba4e58ec 946
e5bbef20 947static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 948{
645ca708 949 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
950}
951
1da177e4
LT
952/*
953 * Throw away all pending data and cancel the corking. Socket is locked.
954 */
955static void udp_v6_flush_pending_frames(struct sock *sk)
956{
957 struct udp_sock *up = udp_sk(sk);
958
36d926b9
DL
959 if (up->pending == AF_INET)
960 udp_flush_pending_frames(sk);
961 else if (up->pending) {
1da177e4
LT
962 up->len = 0;
963 up->pending = 0;
964 ip6_flush_pending_frames(sk);
1ab1457c 965 }
1da177e4
LT
966}
967
493c6be3 968/**
67ba4152
IM
969 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
970 * @sk: socket we are sending on
971 * @skb: sk_buff containing the filled-in UDP header
972 * (checksum field must be zeroed out)
493c6be3
SS
973 */
974static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
975 const struct in6_addr *saddr,
976 const struct in6_addr *daddr, int len)
977{
978 unsigned int offset;
979 struct udphdr *uh = udp_hdr(skb);
980 __wsum csum = 0;
981
982 if (skb_queue_len(&sk->sk_write_queue) == 1) {
983 /* Only one fragment on the socket. */
984 skb->csum_start = skb_transport_header(skb) - skb->head;
985 skb->csum_offset = offsetof(struct udphdr, check);
986 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
987 } else {
988 /*
989 * HW-checksum won't work as there are two or more
990 * fragments on the socket so that all csums of sk_buffs
991 * should be together
992 */
993 offset = skb_transport_offset(skb);
994 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
995
996 skb->ip_summed = CHECKSUM_NONE;
997
998 skb_queue_walk(&sk->sk_write_queue, skb) {
999 csum = csum_add(csum, skb->csum);
1000 }
1001
1002 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1003 csum);
1004 if (uh->check == 0)
1005 uh->check = CSUM_MANGLED_0;
1006 }
1007}
1008
1da177e4
LT
1009/*
1010 * Sending
1011 */
1012
4c0a6cb0 1013static int udp_v6_push_pending_frames(struct sock *sk)
1da177e4
LT
1014{
1015 struct sk_buff *skb;
1016 struct udphdr *uh;
4c0a6cb0 1017 struct udp_sock *up = udp_sk(sk);
1da177e4 1018 struct inet_sock *inet = inet_sk(sk);
8822b64a 1019 struct flowi6 *fl6;
1da177e4 1020 int err = 0;
b2bf1e26 1021 int is_udplite = IS_UDPLITE(sk);
868c86bc 1022 __wsum csum = 0;
1da177e4 1023
8822b64a
HFS
1024 if (up->pending == AF_INET)
1025 return udp_push_pending_frames(sk);
1026
1027 fl6 = &inet->cork.fl.u.ip6;
1028
1da177e4 1029 /* Grab the skbuff where UDP header space exists. */
e5d08d71
IM
1030 skb = skb_peek(&sk->sk_write_queue);
1031 if (skb == NULL)
1da177e4
LT
1032 goto out;
1033
1034 /*
1035 * Create a UDP header
1036 */
4bedb452 1037 uh = udp_hdr(skb);
1958b856
DM
1038 uh->source = fl6->fl6_sport;
1039 uh->dest = fl6->fl6_dport;
1da177e4
LT
1040 uh->len = htons(up->len);
1041 uh->check = 0;
1042
b2bf1e26 1043 if (is_udplite)
ba4e58ec 1044 csum = udplite_csum_outgoing(sk, skb);
1c19448c 1045 else if (up->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
1046 skb->ip_summed = CHECKSUM_NONE;
1047 goto send;
1048 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
4c9483b2 1049 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
493c6be3
SS
1050 up->len);
1051 goto send;
1052 } else
ba4e58ec 1053 csum = udp_csum_outgoing(sk, skb);
1da177e4 1054
ba4e58ec 1055 /* add protocol-dependent pseudo-header */
4c9483b2
DM
1056 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1057 up->len, fl6->flowi6_proto, csum);
1da177e4 1058 if (uh->check == 0)
f6ab0288 1059 uh->check = CSUM_MANGLED_0;
1da177e4 1060
493c6be3 1061send:
1da177e4 1062 err = ip6_push_pending_frames(sk);
6ce9e7b5
ED
1063 if (err) {
1064 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1065 UDP6_INC_STATS_USER(sock_net(sk),
1066 UDP_MIB_SNDBUFERRORS, is_udplite);
1067 err = 0;
1068 }
1069 } else
1070 UDP6_INC_STATS_USER(sock_net(sk),
1071 UDP_MIB_OUTDATAGRAMS, is_udplite);
1da177e4
LT
1072out:
1073 up->len = 0;
1074 up->pending = 0;
1075 return err;
1076}
1077
ba4e58ec 1078int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
1079 struct msghdr *msg, size_t len)
1080{
1081 struct ipv6_txoptions opt_space;
1082 struct udp_sock *up = udp_sk(sk);
1083 struct inet_sock *inet = inet_sk(sk);
1084 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1085 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1086 struct in6_addr *daddr, *final_p, final;
1da177e4
LT
1087 struct ipv6_txoptions *opt = NULL;
1088 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1089 struct flowi6 fl6;
1da177e4
LT
1090 struct dst_entry *dst;
1091 int addr_len = msg->msg_namelen;
1092 int ulen = len;
1093 int hlimit = -1;
41a1f8ea 1094 int tclass = -1;
13b52cd4 1095 int dontfrag = -1;
1da177e4
LT
1096 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1097 int err;
987905de 1098 int connected = 0;
b2bf1e26 1099 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1100 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1da177e4
LT
1101
1102 /* destination address check */
1103 if (sin6) {
1104 if (addr_len < offsetof(struct sockaddr, sa_data))
1105 return -EINVAL;
1106
1107 switch (sin6->sin6_family) {
1108 case AF_INET6:
1109 if (addr_len < SIN6_LEN_RFC2133)
1110 return -EINVAL;
1111 daddr = &sin6->sin6_addr;
1112 break;
1113 case AF_INET:
1114 goto do_udp_sendmsg;
1115 case AF_UNSPEC:
1116 msg->msg_name = sin6 = NULL;
1117 msg->msg_namelen = addr_len = 0;
1118 daddr = NULL;
1119 break;
1120 default:
1121 return -EINVAL;
1122 }
1123 } else if (!up->pending) {
1124 if (sk->sk_state != TCP_ESTABLISHED)
1125 return -EDESTADDRREQ;
efe4208f 1126 daddr = &sk->sk_v6_daddr;
1ab1457c 1127 } else
1da177e4
LT
1128 daddr = NULL;
1129
1130 if (daddr) {
e773e4fa 1131 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1132 struct sockaddr_in sin;
1133 sin.sin_family = AF_INET;
c720c7e8 1134 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1135 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1136 msg->msg_name = &sin;
1137 msg->msg_namelen = sizeof(sin);
1138do_udp_sendmsg:
1139 if (__ipv6_only_sock(sk))
1140 return -ENETUNREACH;
1141 return udp_sendmsg(iocb, sk, msg, len);
1142 }
1143 }
1144
1145 if (up->pending == AF_INET)
1146 return udp_sendmsg(iocb, sk, msg, len);
1147
1148 /* Rough check on arithmetic overflow,
b59e139b 1149 better check is made in ip6_append_data().
1da177e4
LT
1150 */
1151 if (len > INT_MAX - sizeof(struct udphdr))
1152 return -EMSGSIZE;
1ab1457c 1153
1da177e4
LT
1154 if (up->pending) {
1155 /*
1156 * There are pending frames.
1157 * The socket lock must be held while it's corked.
1158 */
1159 lock_sock(sk);
1160 if (likely(up->pending)) {
1161 if (unlikely(up->pending != AF_INET6)) {
1162 release_sock(sk);
1163 return -EAFNOSUPPORT;
1164 }
1165 dst = NULL;
1166 goto do_append_data;
1167 }
1168 release_sock(sk);
1169 }
1170 ulen += sizeof(struct udphdr);
1171
4c9483b2 1172 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1173
1174 if (sin6) {
1175 if (sin6->sin6_port == 0)
1176 return -EINVAL;
1177
1958b856 1178 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1179 daddr = &sin6->sin6_addr;
1180
1181 if (np->sndflow) {
4c9483b2
DM
1182 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1183 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1184 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1185 if (flowlabel == NULL)
1186 return -EINVAL;
1da177e4
LT
1187 }
1188 }
1189
1190 /*
1191 * Otherwise it will be difficult to maintain
1192 * sk->sk_dst_cache.
1193 */
1194 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1195 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1196 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1197
1198 if (addr_len >= sizeof(struct sockaddr_in6) &&
1199 sin6->sin6_scope_id &&
842df073 1200 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1201 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1202 } else {
1203 if (sk->sk_state != TCP_ESTABLISHED)
1204 return -EDESTADDRREQ;
1205
1958b856 1206 fl6.fl6_dport = inet->inet_dport;
efe4208f 1207 daddr = &sk->sk_v6_daddr;
4c9483b2 1208 fl6.flowlabel = np->flow_label;
987905de 1209 connected = 1;
1da177e4
LT
1210 }
1211
4c9483b2
DM
1212 if (!fl6.flowi6_oif)
1213 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1214
4c9483b2
DM
1215 if (!fl6.flowi6_oif)
1216 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1217
4c9483b2 1218 fl6.flowi6_mark = sk->sk_mark;
51953d5b 1219
1da177e4
LT
1220 if (msg->msg_controllen) {
1221 opt = &opt_space;
1222 memset(opt, 0, sizeof(struct ipv6_txoptions));
1223 opt->tot_len = sizeof(*opt);
1224
73df66f8
TP
1225 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1226 &hlimit, &tclass, &dontfrag);
1da177e4
LT
1227 if (err < 0) {
1228 fl6_sock_release(flowlabel);
1229 return err;
1230 }
4c9483b2
DM
1231 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1232 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1233 if (flowlabel == NULL)
1234 return -EINVAL;
1235 }
1236 if (!(opt->opt_nflen|opt->opt_flen))
1237 opt = NULL;
987905de 1238 connected = 0;
1da177e4
LT
1239 }
1240 if (opt == NULL)
1241 opt = np->opt;
df9890c3
YH
1242 if (flowlabel)
1243 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1244 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1245
4c9483b2 1246 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1247 if (!ipv6_addr_any(daddr))
4e3fd7a0 1248 fl6.daddr = *daddr;
876c7f41 1249 else
4c9483b2
DM
1250 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1251 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1252 fl6.saddr = np->saddr;
1958b856 1253 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1254
4c9483b2 1255 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1256 if (final_p)
987905de 1257 connected = 0;
1da177e4 1258
4c9483b2
DM
1259 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1260 fl6.flowi6_oif = np->mcast_oif;
987905de 1261 connected = 0;
c4062dfc
EH
1262 } else if (!fl6.flowi6_oif)
1263 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1264
4c9483b2 1265 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1266
0e0d44ab 1267 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1268 if (IS_ERR(dst)) {
1269 err = PTR_ERR(dst);
1270 dst = NULL;
1da177e4 1271 goto out;
14e50e57 1272 }
1da177e4 1273
5c98631c
LC
1274 if (hlimit < 0)
1275 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4 1276
e651f03a 1277 if (tclass < 0)
41a1f8ea 1278 tclass = np->tclass;
41a1f8ea 1279
1da177e4
LT
1280 if (msg->msg_flags&MSG_CONFIRM)
1281 goto do_confirm;
1282back_from_confirm:
1283
1284 lock_sock(sk);
1285 if (unlikely(up->pending)) {
1286 /* The socket is already corked while preparing it. */
1287 /* ... which is an evident application bug. --ANK */
1288 release_sock(sk);
1289
ba7a46f1 1290 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1291 err = -EINVAL;
1292 goto out;
1293 }
1294
1295 up->pending = AF_INET6;
1296
1297do_append_data:
e36d3ff9
JP
1298 if (dontfrag < 0)
1299 dontfrag = np->dontfrag;
1da177e4 1300 up->len += ulen;
ba4e58ec
GR
1301 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1302 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
4c9483b2 1303 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
67ba4152 1304 (struct rt6_info *)dst,
13b52cd4 1305 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1da177e4
LT
1306 if (err)
1307 udp_v6_flush_pending_frames(sk);
1308 else if (!corkreq)
4c0a6cb0 1309 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1310 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1311 up->pending = 0;
1da177e4 1312
a5e7c210
DM
1313 if (dst) {
1314 if (connected) {
1315 ip6_dst_store(sk, dst,
efe4208f
ED
1316 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1317 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1318#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1319 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1320 &np->saddr :
1321#endif
1322 NULL);
a5e7c210
DM
1323 } else {
1324 dst_release(dst);
1325 }
a3c96089 1326 dst = NULL;
a5e7c210
DM
1327 }
1328
1da177e4
LT
1329 if (err > 0)
1330 err = np->recverr ? net_xmit_errno(err) : 0;
1331 release_sock(sk);
1332out:
a3c96089 1333 dst_release(dst);
1da177e4 1334 fl6_sock_release(flowlabel);
cd562c98 1335 if (!err)
1da177e4 1336 return len;
a18135eb
DM
1337 /*
1338 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1339 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1340 * we don't have a good statistic (IpOutDiscards but it can be too many
1341 * things). We could add another new stat but at least for now that
1342 * seems like overkill.
1343 */
1344 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
235b9f7a
PE
1345 UDP6_INC_STATS_USER(sock_net(sk),
1346 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1347 }
1da177e4
LT
1348 return err;
1349
1350do_confirm:
1351 dst_confirm(dst);
1352 if (!(msg->msg_flags&MSG_PROBE) || len)
1353 goto back_from_confirm;
1354 err = 0;
1355 goto out;
1356}
1357
7d06b2e0 1358void udpv6_destroy_sock(struct sock *sk)
1da177e4 1359{
44046a59 1360 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1361 lock_sock(sk);
1362 udp_v6_flush_pending_frames(sk);
1363 release_sock(sk);
1364
44046a59
TP
1365 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1366 void (*encap_destroy)(struct sock *sk);
1367 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1368 if (encap_destroy)
1369 encap_destroy(sk);
1370 }
1371
1da177e4 1372 inet6_destroy_sock(sk);
1da177e4
LT
1373}
1374
1375/*
1376 * Socket option code for UDP
1377 */
ba4e58ec 1378int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1379 char __user *optval, unsigned int optlen)
3fdadf7d 1380{
db8dac20 1381 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1382 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1383 udp_v6_push_pending_frames);
ba4e58ec 1384 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1385}
1386
1387#ifdef CONFIG_COMPAT
ba4e58ec 1388int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1389 char __user *optval, unsigned int optlen)
3fdadf7d 1390{
db8dac20 1391 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1392 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1393 udp_v6_push_pending_frames);
ba4e58ec 1394 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1395}
1396#endif
1397
ba4e58ec
GR
1398int udpv6_getsockopt(struct sock *sk, int level, int optname,
1399 char __user *optval, int __user *optlen)
3fdadf7d 1400{
db8dac20 1401 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1402 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1403 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1404}
1405
1406#ifdef CONFIG_COMPAT
ba4e58ec
GR
1407int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1408 char __user *optval, int __user *optlen)
3fdadf7d 1409{
db8dac20 1410 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1411 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1412 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1413}
1414#endif
1415
41135cc8 1416static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1417 .handler = udpv6_rcv,
1418 .err_handler = udpv6_err,
1419 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1420};
1421
1422/* ------------------------------------------------------------------------ */
1423#ifdef CONFIG_PROC_FS
ba4e58ec 1424int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1425{
17ef66af
LC
1426 if (v == SEQ_START_TOKEN) {
1427 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1428 } else {
1429 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1430 struct inet_sock *inet = inet_sk(v);
1431 __u16 srcp = ntohs(inet->inet_sport);
1432 __u16 destp = ntohs(inet->inet_dport);
1433 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1434 }
1da177e4
LT
1435 return 0;
1436}
1437
73cb88ec
AV
1438static const struct file_operations udp6_afinfo_seq_fops = {
1439 .owner = THIS_MODULE,
1440 .open = udp_seq_open,
1441 .read = seq_read,
1442 .llseek = seq_lseek,
1443 .release = seq_release_net
1444};
1445
1da177e4 1446static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1447 .name = "udp6",
1448 .family = AF_INET6,
645ca708 1449 .udp_table = &udp_table,
73cb88ec 1450 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1451 .seq_ops = {
1452 .show = udp6_seq_show,
1453 },
1da177e4
LT
1454};
1455
2c8c1e72 1456int __net_init udp6_proc_init(struct net *net)
1da177e4 1457{
0c96d8c5 1458 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1459}
1460
0c96d8c5
DL
1461void udp6_proc_exit(struct net *net) {
1462 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1463}
1464#endif /* CONFIG_PROC_FS */
1465
f77d6021
ED
1466void udp_v6_clear_sk(struct sock *sk, int size)
1467{
1468 struct inet_sock *inet = inet_sk(sk);
1469
1470 /* we do not want to clear pinet6 field, because of RCU lookups */
1471 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1472
1473 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1474 memset(&inet->pinet6 + 1, 0, size);
1475}
1476
1da177e4
LT
1477/* ------------------------------------------------------------------------ */
1478
1479struct proto udpv6_prot = {
543d9cfe
ACM
1480 .name = "UDPv6",
1481 .owner = THIS_MODULE,
ba4e58ec 1482 .close = udp_lib_close,
543d9cfe
ACM
1483 .connect = ip6_datagram_connect,
1484 .disconnect = udp_disconnect,
1485 .ioctl = udp_ioctl,
1486 .destroy = udpv6_destroy_sock,
1487 .setsockopt = udpv6_setsockopt,
1488 .getsockopt = udpv6_getsockopt,
1489 .sendmsg = udpv6_sendmsg,
1490 .recvmsg = udpv6_recvmsg,
f7ad74fe 1491 .backlog_rcv = __udpv6_queue_rcv_skb,
ba4e58ec
GR
1492 .hash = udp_lib_hash,
1493 .unhash = udp_lib_unhash,
719f8358 1494 .rehash = udp_v6_rehash,
543d9cfe 1495 .get_port = udp_v6_get_port,
95766fff
HA
1496 .memory_allocated = &udp_memory_allocated,
1497 .sysctl_mem = sysctl_udp_mem,
1498 .sysctl_wmem = &sysctl_udp_wmem_min,
1499 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1500 .obj_size = sizeof(struct udp6_sock),
271b72c7 1501 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1502 .h.udp_table = &udp_table,
3fdadf7d 1503#ifdef CONFIG_COMPAT
543d9cfe
ACM
1504 .compat_setsockopt = compat_udpv6_setsockopt,
1505 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1506#endif
f77d6021 1507 .clear_sk = udp_v6_clear_sk,
1da177e4
LT
1508};
1509
1da177e4
LT
1510static struct inet_protosw udpv6_protosw = {
1511 .type = SOCK_DGRAM,
1512 .protocol = IPPROTO_UDP,
1513 .prot = &udpv6_prot,
1514 .ops = &inet6_dgram_ops,
1da177e4
LT
1515 .flags = INET_PROTOSW_PERMANENT,
1516};
1517
1518
7f4e4868 1519int __init udpv6_init(void)
1da177e4 1520{
7f4e4868
DL
1521 int ret;
1522
3336288a
VY
1523 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1524 if (ret)
c6b641a4 1525 goto out;
3336288a 1526
7f4e4868
DL
1527 ret = inet6_register_protosw(&udpv6_protosw);
1528 if (ret)
1529 goto out_udpv6_protocol;
1530out:
1531 return ret;
1532
1533out_udpv6_protocol:
1534 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1535 goto out;
1536}
1537
09f7709f 1538void udpv6_exit(void)
7f4e4868
DL
1539{
1540 inet6_unregister_protosw(&udpv6_protosw);
1541 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1542}