]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - net/ipv6/udp.c
be2net: Fix pcie error recovery in case of NIC+RoCE adapters
[mirror_ubuntu-artful-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
496611d7 40#include <net/addrconf.h>
1da177e4
LT
41#include <net/ndisc.h>
42#include <net/protocol.h>
43#include <net/transp_v6.h>
44#include <net/ip6_route.h>
1da177e4 45#include <net/raw.h>
c752f073 46#include <net/tcp_states.h>
1da177e4
LT
47#include <net/ip6_checksum.h>
48#include <net/xfrm.h>
72289b96 49#include <net/inet6_hashtables.h>
076bb0c8 50#include <net/busy_poll.h>
e32ea7e7 51#include <net/sock_reuseport.h>
1da177e4
LT
52
53#include <linux/proc_fs.h>
54#include <linux/seq_file.h>
22911fc5 55#include <trace/events/skb.h>
ba4e58ec 56#include "udp_impl.h"
1da177e4 57
6eada011
ED
58static u32 udp6_ehashfn(const struct net *net,
59 const struct in6_addr *laddr,
60 const u16 lport,
61 const struct in6_addr *faddr,
62 const __be16 fport)
b50026b5 63{
1bbdceef
HFS
64 static u32 udp6_ehash_secret __read_mostly;
65 static u32 udp_ipv6_hash_secret __read_mostly;
66
67 u32 lhash, fhash;
68
69 net_get_random_once(&udp6_ehash_secret,
70 sizeof(udp6_ehash_secret));
71 net_get_random_once(&udp_ipv6_hash_secret,
72 sizeof(udp_ipv6_hash_secret));
73
74 lhash = (__force u32)laddr->s6_addr32[3];
75 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
76
b50026b5 77 return __inet6_ehashfn(lhash, lport, fhash, fport,
1bbdceef 78 udp_ipv6_hash_secret + net_hash_mix(net));
b50026b5
HFS
79}
80
6eada011
ED
81static u32 udp6_portaddr_hash(const struct net *net,
82 const struct in6_addr *addr6,
83 unsigned int port)
d4cada4a
ED
84{
85 unsigned int hash, mix = net_hash_mix(net);
86
87 if (ipv6_addr_any(addr6))
88 hash = jhash_1word(0, mix);
856540ee 89 else if (ipv6_addr_v4mapped(addr6))
0eae88f3 90 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
d4cada4a 91 else
0eae88f3 92 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
d4cada4a
ED
93
94 return hash ^ port;
95}
96
6ba5a3c5 97int udp_v6_get_port(struct sock *sk, unsigned short snum)
1da177e4 98{
30fff923
ED
99 unsigned int hash2_nulladdr =
100 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
9a52e97e 101 unsigned int hash2_partial =
efe4208f 102 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
30fff923 103
d4cada4a 104 /* precompute partial secondary hash */
30fff923
ED
105 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
106 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
1da177e4
LT
107}
108
719f8358
ED
109static void udp_v6_rehash(struct sock *sk)
110{
111 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
efe4208f 112 &sk->sk_v6_rcv_saddr,
719f8358
ED
113 inet_sk(sk)->inet_num);
114
115 udp_lib_rehash(sk, new_hash);
116}
117
645ca708
ED
118static inline int compute_score(struct sock *sk, struct net *net,
119 unsigned short hnum,
88440ae7
BS
120 const struct in6_addr *saddr, __be16 sport,
121 const struct in6_addr *daddr, __be16 dport,
645ca708
ED
122 int dif)
123{
60c04aec
JP
124 int score;
125 struct inet_sock *inet;
126
127 if (!net_eq(sock_net(sk), net) ||
128 udp_sk(sk)->udp_port_hash != hnum ||
129 sk->sk_family != PF_INET6)
130 return -1;
131
132 score = 0;
133 inet = inet_sk(sk);
134
135 if (inet->inet_dport) {
136 if (inet->inet_dport != sport)
137 return -1;
138 score++;
139 }
140
141 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
142 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
143 return -1;
144 score++;
145 }
146
147 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
148 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
149 return -1;
150 score++;
151 }
152
153 if (sk->sk_bound_dev_if) {
154 if (sk->sk_bound_dev_if != dif)
155 return -1;
156 score++;
645ca708 157 }
60c04aec 158
70da268b
ED
159 if (sk->sk_incoming_cpu == raw_smp_processor_id())
160 score++;
161
645ca708
ED
162 return score;
163}
164
fddc17de 165static inline int compute_score2(struct sock *sk, struct net *net,
60c04aec
JP
166 const struct in6_addr *saddr, __be16 sport,
167 const struct in6_addr *daddr,
168 unsigned short hnum, int dif)
fddc17de 169{
60c04aec
JP
170 int score;
171 struct inet_sock *inet;
fddc17de 172
60c04aec
JP
173 if (!net_eq(sock_net(sk), net) ||
174 udp_sk(sk)->udp_port_hash != hnum ||
175 sk->sk_family != PF_INET6)
176 return -1;
fddc17de 177
60c04aec
JP
178 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
179 return -1;
180
181 score = 0;
182 inet = inet_sk(sk);
183
184 if (inet->inet_dport) {
185 if (inet->inet_dport != sport)
186 return -1;
187 score++;
188 }
189
190 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
191 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
fddc17de 192 return -1;
60c04aec 193 score++;
fddc17de 194 }
60c04aec
JP
195
196 if (sk->sk_bound_dev_if) {
197 if (sk->sk_bound_dev_if != dif)
198 return -1;
199 score++;
200 }
201
70da268b
ED
202 if (sk->sk_incoming_cpu == raw_smp_processor_id())
203 score++;
204
fddc17de
ED
205 return score;
206}
207
fddc17de
ED
208/* called with read_rcu_lock() */
209static struct sock *udp6_lib_lookup2(struct net *net,
210 const struct in6_addr *saddr, __be16 sport,
211 const struct in6_addr *daddr, unsigned int hnum, int dif,
1134158b
CG
212 struct udp_hslot *hslot2, unsigned int slot2,
213 struct sk_buff *skb)
fddc17de
ED
214{
215 struct sock *sk, *result;
216 struct hlist_nulls_node *node;
72289b96 217 int score, badness, matches = 0, reuseport = 0;
ed0dfffd 218 bool select_ok = true;
72289b96 219 u32 hash = 0;
fddc17de
ED
220
221begin:
222 result = NULL;
223 badness = -1;
224 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
225 score = compute_score2(sk, net, saddr, sport,
226 daddr, hnum, dif);
227 if (score > badness) {
228 result = sk;
229 badness = score;
72289b96
TH
230 reuseport = sk->sk_reuseport;
231 if (reuseport) {
b50026b5
HFS
232 hash = udp6_ehashfn(net, daddr, hnum,
233 saddr, sport);
ed0dfffd
ED
234 if (select_ok) {
235 struct sock *sk2;
236
237 sk2 = reuseport_select_sock(sk, hash, skb,
238 sizeof(struct udphdr));
239 if (sk2) {
240 result = sk2;
241 select_ok = false;
242 goto found;
243 }
e32ea7e7 244 }
72289b96 245 matches = 1;
70da268b 246 }
72289b96
TH
247 } else if (score == badness && reuseport) {
248 matches++;
8fc54f68 249 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
250 result = sk;
251 hash = next_pseudo_random32(hash);
fddc17de
ED
252 }
253 }
254 /*
255 * if the nulls value we got at the end of this lookup is
256 * not the expected one, we must restart lookup.
257 * We probably met an item that was moved to another chain.
258 */
259 if (get_nulls_value(node) != slot2)
260 goto begin;
261
262 if (result) {
e32ea7e7 263found:
c31504dc 264 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
fddc17de
ED
265 result = NULL;
266 else if (unlikely(compute_score2(result, net, saddr, sport,
267 daddr, hnum, dif) < badness)) {
268 sock_put(result);
269 goto begin;
270 }
271 }
272 return result;
273}
274
fce82338 275struct sock *__udp6_lib_lookup(struct net *net,
88440ae7
BS
276 const struct in6_addr *saddr, __be16 sport,
277 const struct in6_addr *daddr, __be16 dport,
538950a1
CG
278 int dif, struct udp_table *udptable,
279 struct sk_buff *skb)
1da177e4 280{
271b72c7 281 struct sock *sk, *result;
88ab1932 282 struct hlist_nulls_node *node;
1da177e4 283 unsigned short hnum = ntohs(dport);
fddc17de
ED
284 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
285 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
72289b96 286 int score, badness, matches = 0, reuseport = 0;
ed0dfffd 287 bool select_ok = true;
72289b96 288 u32 hash = 0;
645ca708 289
271b72c7 290 rcu_read_lock();
fddc17de
ED
291 if (hslot->count > 10) {
292 hash2 = udp6_portaddr_hash(net, daddr, hnum);
293 slot2 = hash2 & udptable->mask;
294 hslot2 = &udptable->hash2[slot2];
295 if (hslot->count < hslot2->count)
296 goto begin;
297
298 result = udp6_lib_lookup2(net, saddr, sport,
299 daddr, hnum, dif,
1134158b 300 hslot2, slot2, skb);
fddc17de
ED
301 if (!result) {
302 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
303 slot2 = hash2 & udptable->mask;
304 hslot2 = &udptable->hash2[slot2];
305 if (hslot->count < hslot2->count)
306 goto begin;
307
1223c67c
JBD
308 result = udp6_lib_lookup2(net, saddr, sport,
309 &in6addr_any, hnum, dif,
1134158b 310 hslot2, slot2, skb);
fddc17de
ED
311 }
312 rcu_read_unlock();
313 return result;
314 }
271b72c7
ED
315begin:
316 result = NULL;
317 badness = -1;
88ab1932 318 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
645ca708
ED
319 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
320 if (score > badness) {
321 result = sk;
322 badness = score;
72289b96
TH
323 reuseport = sk->sk_reuseport;
324 if (reuseport) {
b50026b5
HFS
325 hash = udp6_ehashfn(net, daddr, hnum,
326 saddr, sport);
ed0dfffd
ED
327 if (select_ok) {
328 struct sock *sk2;
329
330 sk2 = reuseport_select_sock(sk, hash, skb,
538950a1 331 sizeof(struct udphdr));
ed0dfffd
ED
332 if (sk2) {
333 result = sk2;
334 select_ok = false;
335 goto found;
336 }
e32ea7e7 337 }
72289b96
TH
338 matches = 1;
339 }
340 } else if (score == badness && reuseport) {
341 matches++;
8fc54f68 342 if (reciprocal_scale(hash, matches) == 0)
72289b96
TH
343 result = sk;
344 hash = next_pseudo_random32(hash);
1da177e4
LT
345 }
346 }
88ab1932
ED
347 /*
348 * if the nulls value we got at the end of this lookup is
349 * not the expected one, we must restart lookup.
350 * We probably met an item that was moved to another chain.
351 */
fddc17de 352 if (get_nulls_value(node) != slot)
88ab1932
ED
353 goto begin;
354
271b72c7 355 if (result) {
e32ea7e7 356found:
c31504dc 357 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
271b72c7
ED
358 result = NULL;
359 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
360 daddr, dport, dif) < badness)) {
361 sock_put(result);
362 goto begin;
363 }
364 }
365 rcu_read_unlock();
1da177e4
LT
366 return result;
367}
fce82338 368EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
1da177e4 369
607c4aaf
KK
370static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
371 __be16 sport, __be16 dport,
645ca708 372 struct udp_table *udptable)
607c4aaf 373{
23542618 374 struct sock *sk;
b71d1d42 375 const struct ipv6hdr *iph = ipv6_hdr(skb);
607c4aaf 376
e5d08d71
IM
377 sk = skb_steal_sock(skb);
378 if (unlikely(sk))
23542618 379 return sk;
adf30907
ED
380 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
381 &iph->daddr, dport, inet6_iif(skb),
538950a1 382 udptable, skb);
607c4aaf
KK
383}
384
aa976fc0
BS
385struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
386 const struct in6_addr *daddr, __be16 dport, int dif)
387{
538950a1 388 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table, NULL);
aa976fc0
BS
389}
390EXPORT_SYMBOL_GPL(udp6_lib_lookup);
391
1da177e4 392/*
67ba4152
IM
393 * This should be easy, if there is something there we
394 * return it, otherwise we block.
1da177e4
LT
395 */
396
1b784140 397int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
1da177e4
LT
398 int noblock, int flags, int *addr_len)
399{
400 struct ipv6_pinfo *np = inet6_sk(sk);
401 struct inet_sock *inet = inet_sk(sk);
1ab1457c 402 struct sk_buff *skb;
59c2cdae 403 unsigned int ulen, copied;
3f518bf7 404 int peeked, off = 0;
759e5d00
HX
405 int err;
406 int is_udplite = IS_UDPLITE(sk);
197c949e 407 bool checksum_valid = false;
f26ba175 408 int is_udp4;
8a74ad60 409 bool slow;
1da177e4 410
1da177e4 411 if (flags & MSG_ERRQUEUE)
85fbaa75 412 return ipv6_recv_error(sk, msg, len, addr_len);
1da177e4 413
4b340ae2 414 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
85fbaa75 415 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
4b340ae2 416
1da177e4 417try_again:
a59322be 418 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 419 &peeked, &off, &err);
1da177e4
LT
420 if (!skb)
421 goto out;
422
759e5d00 423 ulen = skb->len - sizeof(struct udphdr);
59c2cdae
DM
424 copied = len;
425 if (copied > ulen)
426 copied = ulen;
427 else if (copied < ulen)
1ab1457c 428 msg->msg_flags |= MSG_TRUNC;
1da177e4 429
f26ba175
WY
430 is_udp4 = (skb->protocol == htons(ETH_P_IP));
431
ba4e58ec 432 /*
759e5d00
HX
433 * If checksum is needed at all, try to do it while copying the
434 * data. If the data is truncated, or if we only want a partial
435 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 436 */
ba4e58ec 437
59c2cdae 438 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
197c949e
ED
439 checksum_valid = !udp_lib_checksum_complete(skb);
440 if (!checksum_valid)
1da177e4 441 goto csum_copy_err;
ba4e58ec
GR
442 }
443
197c949e 444 if (checksum_valid || skb_csum_unnecessary(skb))
51f3d02b
DM
445 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
446 msg, copied);
ba4e58ec 447 else {
227158db 448 err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), msg);
1da177e4
LT
449 if (err == -EINVAL)
450 goto csum_copy_err;
451 }
22911fc5
ED
452 if (unlikely(err)) {
453 trace_kfree_skb(skb, udpv6_recvmsg);
979402b1
ED
454 if (!peeked) {
455 atomic_inc(&sk->sk_drops);
456 if (is_udp4)
457 UDP_INC_STATS_USER(sock_net(sk),
458 UDP_MIB_INERRORS,
459 is_udplite);
460 else
461 UDP6_INC_STATS_USER(sock_net(sk),
462 UDP_MIB_INERRORS,
463 is_udplite);
464 }
1da177e4 465 goto out_free;
22911fc5 466 }
f26ba175
WY
467 if (!peeked) {
468 if (is_udp4)
469 UDP_INC_STATS_USER(sock_net(sk),
470 UDP_MIB_INDATAGRAMS, is_udplite);
471 else
472 UDP6_INC_STATS_USER(sock_net(sk),
473 UDP_MIB_INDATAGRAMS, is_udplite);
474 }
cb75994e 475
3b885787 476 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
477
478 /* Copy the address. */
479 if (msg->msg_name) {
342dfc30 480 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1da177e4 481 sin6->sin6_family = AF_INET6;
4bedb452 482 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4 483 sin6->sin6_flowinfo = 0;
1da177e4 484
842df073 485 if (is_udp4) {
b301e82c
BH
486 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
487 &sin6->sin6_addr);
842df073
HFS
488 sin6->sin6_scope_id = 0;
489 } else {
4e3fd7a0 490 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
842df073
HFS
491 sin6->sin6_scope_id =
492 ipv6_iface_scope_id(&sin6->sin6_addr,
4330487a 493 inet6_iif(skb));
1da177e4 494 }
bceaa902 495 *addr_len = sizeof(*sin6);
1da177e4 496 }
4b261c75
HFS
497
498 if (np->rxopt.all)
499 ip6_datagram_recv_common_ctl(sk, msg, skb);
500
f26ba175 501 if (is_udp4) {
1da177e4
LT
502 if (inet->cmsg_flags)
503 ip_cmsg_recv(msg, skb);
504 } else {
505 if (np->rxopt.all)
4b261c75 506 ip6_datagram_recv_specific_ctl(sk, msg, skb);
1ab1457c 507 }
1da177e4 508
59c2cdae 509 err = copied;
1da177e4 510 if (flags & MSG_TRUNC)
759e5d00 511 err = ulen;
1da177e4
LT
512
513out_free:
9d410c79 514 skb_free_datagram_locked(sk, skb);
1da177e4
LT
515out:
516 return err;
517
518csum_copy_err:
8a74ad60 519 slow = lock_sock_fast(sk);
0856f939 520 if (!skb_kill_datagram(sk, skb, flags)) {
6a5dc9e5
ED
521 if (is_udp4) {
522 UDP_INC_STATS_USER(sock_net(sk),
523 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
524 UDP_INC_STATS_USER(sock_net(sk),
525 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5
ED
526 } else {
527 UDP6_INC_STATS_USER(sock_net(sk),
528 UDP_MIB_CSUMERRORS, is_udplite);
0856f939
WY
529 UDP6_INC_STATS_USER(sock_net(sk),
530 UDP_MIB_INERRORS, is_udplite);
6a5dc9e5 531 }
0856f939 532 }
8a74ad60 533 unlock_sock_fast(sk, slow);
1da177e4 534
beb39db5
ED
535 /* starting over for a new packet, but check if we need to yield */
536 cond_resched();
9cfaa8de 537 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
538 goto try_again;
539}
540
ba4e58ec 541void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 542 u8 type, u8 code, int offset, __be32 info,
645ca708 543 struct udp_table *udptable)
1da177e4
LT
544{
545 struct ipv6_pinfo *np;
b71d1d42
ED
546 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
547 const struct in6_addr *saddr = &hdr->saddr;
548 const struct in6_addr *daddr = &hdr->daddr;
67ba4152 549 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
1da177e4
LT
550 struct sock *sk;
551 int err;
7304fe46 552 struct net *net = dev_net(skb->dev);
1da177e4 553
e32ea7e7 554 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
538950a1 555 inet6_iif(skb), udptable, skb);
63159f29 556 if (!sk) {
7304fe46
DJ
557 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
558 ICMP6_MIB_INERRORS);
1da177e4 559 return;
7304fe46 560 }
1da177e4 561
93b36cf3
HFS
562 if (type == ICMPV6_PKT_TOOBIG) {
563 if (!ip6_sk_accept_pmtu(sk))
564 goto out;
81aded24 565 ip6_sk_update_pmtu(skb, sk, info);
93b36cf3 566 }
1a462d18 567 if (type == NDISC_REDIRECT) {
ec18d9a2 568 ip6_sk_redirect(skb, sk);
1a462d18
DJ
569 goto out;
570 }
81aded24 571
1da177e4
LT
572 np = inet6_sk(sk);
573
574 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
575 goto out;
576
577 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
578 goto out;
579
b1faf566 580 if (np->recverr)
1da177e4 581 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
b1faf566 582
1da177e4
LT
583 sk->sk_err = err;
584 sk->sk_error_report(sk);
585out:
586 sock_put(sk);
587}
588
f7ad74fe
BL
589static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
590{
591 int rc;
592
efe4208f 593 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
f7ad74fe 594 sock_rps_save_rxhash(sk, skb);
005ec974 595 sk_mark_napi_id(sk, skb);
2c8c56e1 596 sk_incoming_cpu_update(sk);
005ec974 597 }
f7ad74fe
BL
598
599 rc = sock_queue_rcv_skb(sk, skb);
600 if (rc < 0) {
601 int is_udplite = IS_UDPLITE(sk);
602
603 /* Note that an ENOMEM error is charged twice */
604 if (rc == -ENOMEM)
605 UDP6_INC_STATS_BH(sock_net(sk),
606 UDP_MIB_RCVBUFERRORS, is_udplite);
607 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
608 kfree_skb(skb);
609 return -1;
610 }
611 return 0;
612}
613
ba4e58ec 614static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba 615 struct inet6_skb_parm *opt, u8 type,
67ba4152 616 u8 code, int offset, __be32 info)
ba4e58ec 617{
645ca708 618 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
619}
620
d7f3f621
BL
621static struct static_key udpv6_encap_needed __read_mostly;
622void udpv6_encap_enable(void)
623{
624 if (!static_key_enabled(&udpv6_encap_needed))
625 static_key_slow_inc(&udpv6_encap_needed);
626}
627EXPORT_SYMBOL(udpv6_encap_enable);
628
f7ad74fe 629int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 630{
ba4e58ec 631 struct udp_sock *up = udp_sk(sk);
a18135eb 632 int rc;
b2bf1e26 633 int is_udplite = IS_UDPLITE(sk);
a18135eb 634
ba4e58ec
GR
635 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
636 goto drop;
1da177e4 637
d7f3f621
BL
638 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
639 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
640
641 /*
642 * This is an encapsulation socket so pass the skb to
643 * the socket's udp_encap_rcv() hook. Otherwise, just
644 * fall through and pass this up the UDP socket.
645 * up->encap_rcv() returns the following value:
646 * =0 if skb was successfully passed to the encap
647 * handler or was discarded by it.
648 * >0 if skb should be passed on to UDP.
649 * <0 if skb should be resubmitted as proto -N
650 */
651
652 /* if we're overly short, let UDP handle it */
653 encap_rcv = ACCESS_ONCE(up->encap_rcv);
53b24b8f 654 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
d7f3f621
BL
655 int ret;
656
0a80966b
TH
657 /* Verify checksum before giving to encap */
658 if (udp_lib_checksum_complete(skb))
659 goto csum_error;
660
d7f3f621
BL
661 ret = encap_rcv(sk, skb);
662 if (ret <= 0) {
663 UDP_INC_STATS_BH(sock_net(sk),
664 UDP_MIB_INDATAGRAMS,
665 is_udplite);
666 return -ret;
667 }
668 }
669
670 /* FALLTHROUGH -- it's a UDP Packet */
671 }
672
ba4e58ec
GR
673 /*
674 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
675 */
b2bf1e26 676 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
677
678 if (up->pcrlen == 0) { /* full coverage was set */
ba7a46f1
JP
679 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
680 UDP_SKB_CB(skb)->cscov, skb->len);
ba4e58ec
GR
681 goto drop;
682 }
683 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
ba7a46f1
JP
684 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
685 UDP_SKB_CB(skb)->cscov, up->pcrlen);
ba4e58ec
GR
686 goto drop;
687 }
1da177e4
LT
688 }
689
33d480ce 690 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62 691 if (udp_lib_checksum_complete(skb))
6a5dc9e5 692 goto csum_error;
1ab6eb62 693 }
ba4e58ec 694
274f482d 695 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
3e215c8d
JL
696 UDP6_INC_STATS_BH(sock_net(sk),
697 UDP_MIB_RCVBUFERRORS, is_udplite);
cb80ef46 698 goto drop;
3e215c8d 699 }
cb80ef46 700
d826eb14 701 skb_dst_drop(skb);
cb75994e 702
cb80ef46
BL
703 bh_lock_sock(sk);
704 rc = 0;
705 if (!sock_owned_by_user(sk))
706 rc = __udpv6_queue_rcv_skb(sk, skb);
707 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
708 bh_unlock_sock(sk);
709 goto drop;
710 }
711 bh_unlock_sock(sk);
f7ad74fe
BL
712
713 return rc;
3e215c8d 714
6a5dc9e5
ED
715csum_error:
716 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
ba4e58ec 717drop:
ef28d1a2 718 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 719 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
720 kfree_skb(skb);
721 return -1;
1da177e4
LT
722}
723
5cf3d461
DH
724static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
725 __be16 loc_port, const struct in6_addr *loc_addr,
726 __be16 rmt_port, const struct in6_addr *rmt_addr,
727 int dif, unsigned short hnum)
1da177e4 728{
5cf3d461 729 struct inet_sock *inet = inet_sk(sk);
1da177e4 730
5cf3d461
DH
731 if (!net_eq(sock_net(sk), net))
732 return false;
733
734 if (udp_sk(sk)->udp_port_hash != hnum ||
735 sk->sk_family != PF_INET6 ||
736 (inet->inet_dport && inet->inet_dport != rmt_port) ||
737 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
738 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
33b4b015
HR
739 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
740 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
741 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
5cf3d461
DH
742 return false;
743 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
744 return false;
745 return true;
1da177e4
LT
746}
747
a1ab77f9
ED
748static void flush_stack(struct sock **stack, unsigned int count,
749 struct sk_buff *skb, unsigned int final)
750{
cb80ef46 751 struct sk_buff *skb1 = NULL;
a1ab77f9 752 struct sock *sk;
cb80ef46 753 unsigned int i;
a1ab77f9
ED
754
755 for (i = 0; i < count; i++) {
f6b8f32c 756 sk = stack[i];
63159f29 757 if (likely(!skb1))
cb80ef46
BL
758 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
759 if (!skb1) {
760 atomic_inc(&sk->sk_drops);
761 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
762 IS_UDPLITE(sk));
763 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
764 IS_UDPLITE(sk));
a1ab77f9 765 }
cb80ef46
BL
766
767 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
768 skb1 = NULL;
2dc41cff 769 sock_put(sk);
a1ab77f9 770 }
cb80ef46
BL
771 if (unlikely(skb1))
772 kfree_skb(skb1);
a1ab77f9 773}
4068579e
TH
774
775static void udp6_csum_zero_error(struct sk_buff *skb)
776{
777 /* RFC 2460 section 8.1 says that we SHOULD log
778 * this error. Well, it is reasonable.
779 */
ba7a46f1
JP
780 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
781 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
782 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
4068579e
TH
783}
784
1da177e4
LT
785/*
786 * Note: called only from the BH handler context,
787 * so we don't need to lock the hashes.
788 */
e3163493 789static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 790 const struct in6_addr *saddr, const struct in6_addr *daddr,
36cbb245 791 struct udp_table *udptable, int proto)
1da177e4 792{
a1ab77f9 793 struct sock *sk, *stack[256 / sizeof(struct sock *)];
4bedb452 794 const struct udphdr *uh = udp_hdr(skb);
5cf3d461
DH
795 struct hlist_nulls_node *node;
796 unsigned short hnum = ntohs(uh->dest);
797 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
798 int dif = inet6_iif(skb);
2dc41cff
DH
799 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
800 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
36cbb245 801 bool inner_flushed = false;
2dc41cff
DH
802
803 if (use_hash2) {
804 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
805 udp_table.mask;
806 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
807start_lookup:
808 hslot = &udp_table.hash2[hash2];
809 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
810 }
1da177e4 811
645ca708 812 spin_lock(&hslot->lock);
2dc41cff 813 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
5cf3d461
DH
814 if (__udp_v6_is_mcast_sock(net, sk,
815 uh->dest, daddr,
816 uh->source, saddr,
817 dif, hnum) &&
818 /* If zero checksum and no_check is not on for
819 * the socket then skip it.
820 */
821 (uh->check || udp_sk(sk)->no_check6_rx)) {
822 if (unlikely(count == ARRAY_SIZE(stack))) {
823 flush_stack(stack, count, skb, ~0);
36cbb245 824 inner_flushed = true;
5cf3d461
DH
825 count = 0;
826 }
4068579e 827 stack[count++] = sk;
2dc41cff 828 sock_hold(sk);
95766fff 829 }
1da177e4 830 }
a1ab77f9 831
645ca708 832 spin_unlock(&hslot->lock);
a1ab77f9 833
2dc41cff
DH
834 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
835 if (use_hash2 && hash2 != hash2_any) {
836 hash2 = hash2_any;
837 goto start_lookup;
838 }
839
a1ab77f9
ED
840 if (count) {
841 flush_stack(stack, count, skb, count - 1);
a1ab77f9 842 } else {
36cbb245
RJ
843 if (!inner_flushed)
844 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
845 proto == IPPROTO_UDPLITE);
846 consume_skb(skb);
a1ab77f9 847 }
ba4e58ec 848 return 0;
1da177e4
LT
849}
850
645ca708 851int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 852 int proto)
1da177e4 853{
3ffe533c 854 struct net *net = dev_net(skb->dev);
1da177e4 855 struct sock *sk;
1ab1457c 856 struct udphdr *uh;
b71d1d42 857 const struct in6_addr *saddr, *daddr;
1da177e4
LT
858 u32 ulen = 0;
859
860 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 861 goto discard;
1da177e4 862
0660e03f
ACM
863 saddr = &ipv6_hdr(skb)->saddr;
864 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 865 uh = udp_hdr(skb);
1da177e4
LT
866
867 ulen = ntohs(uh->len);
ba4e58ec
GR
868 if (ulen > skb->len)
869 goto short_packet;
1da177e4 870
759e5d00
HX
871 if (proto == IPPROTO_UDP) {
872 /* UDP validates ulen. */
1da177e4 873
ba4e58ec
GR
874 /* Check for jumbo payload */
875 if (ulen == 0)
876 ulen = skb->len;
1da177e4 877
ba4e58ec
GR
878 if (ulen < sizeof(*uh))
879 goto short_packet;
1da177e4 880
ba4e58ec
GR
881 if (ulen < skb->len) {
882 if (pskb_trim_rcsum(skb, ulen))
883 goto short_packet;
0660e03f
ACM
884 saddr = &ipv6_hdr(skb)->saddr;
885 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 886 uh = udp_hdr(skb);
ba4e58ec 887 }
ba4e58ec 888 }
1da177e4 889
759e5d00 890 if (udp6_csum_init(skb, uh, proto))
6a5dc9e5 891 goto csum_error;
759e5d00 892
1ab1457c
YH
893 /*
894 * Multicast receive code
1da177e4 895 */
ba4e58ec 896 if (ipv6_addr_is_multicast(daddr))
e3163493 897 return __udp6_lib_mcast_deliver(net, skb,
36cbb245 898 saddr, daddr, udptable, proto);
1da177e4
LT
899
900 /* Unicast */
901
1ab1457c 902 /*
1da177e4
LT
903 * check socket cache ... must talk to Alan about his plans
904 * for sock caches... i'll skip this for now.
905 */
607c4aaf 906 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
53b24b8f 907 if (sk) {
a5b50476
ET
908 int ret;
909
1c19448c 910 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
79e0f1c9 911 sock_put(sk);
4068579e
TH
912 udp6_csum_zero_error(skb);
913 goto csum_error;
914 }
915
224d019c 916 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
2abb7cdc
TH
917 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
918 ip6_compute_pseudo);
919
a5b50476 920 ret = udpv6_queue_rcv_skb(sk, skb);
cb80ef46 921 sock_put(sk);
1da177e4 922
cb80ef46
BL
923 /* a return value > 0 means to resubmit the input, but
924 * it wants the return to be -protocol, or 0
925 */
926 if (ret > 0)
927 return -ret;
1da177e4 928
add459aa 929 return 0;
1da177e4 930 }
1ab1457c 931
4068579e
TH
932 if (!uh->check) {
933 udp6_csum_zero_error(skb);
934 goto csum_error;
935 }
936
cb80ef46 937 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 938 goto discard;
cb80ef46
BL
939
940 if (udp_lib_checksum_complete(skb))
6a5dc9e5 941 goto csum_error;
cb80ef46
BL
942
943 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
944 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
945
946 kfree_skb(skb);
add459aa 947 return 0;
1da177e4 948
1ab1457c 949short_packet:
ba7a46f1
JP
950 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
951 proto == IPPROTO_UDPLITE ? "-Lite" : "",
952 saddr, ntohs(uh->source),
953 ulen, skb->len,
954 daddr, ntohs(uh->dest));
6a5dc9e5
ED
955 goto discard;
956csum_error:
957 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1da177e4 958discard:
ef28d1a2 959 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 960 kfree_skb(skb);
add459aa 961 return 0;
1da177e4 962}
ba4e58ec 963
e5bbef20 964static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 965{
645ca708 966 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
967}
968
1da177e4
LT
969/*
970 * Throw away all pending data and cancel the corking. Socket is locked.
971 */
972static void udp_v6_flush_pending_frames(struct sock *sk)
973{
974 struct udp_sock *up = udp_sk(sk);
975
36d926b9
DL
976 if (up->pending == AF_INET)
977 udp_flush_pending_frames(sk);
978 else if (up->pending) {
1da177e4
LT
979 up->len = 0;
980 up->pending = 0;
981 ip6_flush_pending_frames(sk);
1ab1457c 982 }
1da177e4
LT
983}
984
493c6be3 985/**
67ba4152
IM
986 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
987 * @sk: socket we are sending on
988 * @skb: sk_buff containing the filled-in UDP header
989 * (checksum field must be zeroed out)
493c6be3
SS
990 */
991static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
992 const struct in6_addr *saddr,
993 const struct in6_addr *daddr, int len)
994{
995 unsigned int offset;
996 struct udphdr *uh = udp_hdr(skb);
d39d938c 997 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
493c6be3
SS
998 __wsum csum = 0;
999
d39d938c 1000 if (!frags) {
493c6be3
SS
1001 /* Only one fragment on the socket. */
1002 skb->csum_start = skb_transport_header(skb) - skb->head;
1003 skb->csum_offset = offsetof(struct udphdr, check);
1004 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1005 } else {
1006 /*
1007 * HW-checksum won't work as there are two or more
1008 * fragments on the socket so that all csums of sk_buffs
1009 * should be together
1010 */
1011 offset = skb_transport_offset(skb);
1012 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1013
1014 skb->ip_summed = CHECKSUM_NONE;
1015
d39d938c
VY
1016 do {
1017 csum = csum_add(csum, frags->csum);
1018 } while ((frags = frags->next));
493c6be3
SS
1019
1020 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1021 csum);
1022 if (uh->check == 0)
1023 uh->check = CSUM_MANGLED_0;
1024 }
1025}
1026
1da177e4
LT
1027/*
1028 * Sending
1029 */
1030
d39d938c 1031static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1da177e4 1032{
d39d938c 1033 struct sock *sk = skb->sk;
1da177e4 1034 struct udphdr *uh;
1da177e4 1035 int err = 0;
b2bf1e26 1036 int is_udplite = IS_UDPLITE(sk);
868c86bc 1037 __wsum csum = 0;
d39d938c
VY
1038 int offset = skb_transport_offset(skb);
1039 int len = skb->len - offset;
1da177e4
LT
1040
1041 /*
1042 * Create a UDP header
1043 */
4bedb452 1044 uh = udp_hdr(skb);
1958b856
DM
1045 uh->source = fl6->fl6_sport;
1046 uh->dest = fl6->fl6_dport;
d39d938c 1047 uh->len = htons(len);
1da177e4
LT
1048 uh->check = 0;
1049
b2bf1e26 1050 if (is_udplite)
d39d938c
VY
1051 csum = udplite_csum(skb);
1052 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
4068579e
TH
1053 skb->ip_summed = CHECKSUM_NONE;
1054 goto send;
1055 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
d39d938c 1056 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
493c6be3
SS
1057 goto send;
1058 } else
d39d938c 1059 csum = udp_csum(skb);
1da177e4 1060
ba4e58ec 1061 /* add protocol-dependent pseudo-header */
4c9483b2 1062 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
d39d938c 1063 len, fl6->flowi6_proto, csum);
1da177e4 1064 if (uh->check == 0)
f6ab0288 1065 uh->check = CSUM_MANGLED_0;
1da177e4 1066
493c6be3 1067send:
d39d938c 1068 err = ip6_send_skb(skb);
6ce9e7b5
ED
1069 if (err) {
1070 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1071 UDP6_INC_STATS_USER(sock_net(sk),
1072 UDP_MIB_SNDBUFERRORS, is_udplite);
1073 err = 0;
1074 }
1075 } else
1076 UDP6_INC_STATS_USER(sock_net(sk),
1077 UDP_MIB_OUTDATAGRAMS, is_udplite);
d39d938c
VY
1078 return err;
1079}
1080
1081static int udp_v6_push_pending_frames(struct sock *sk)
1082{
1083 struct sk_buff *skb;
1084 struct udp_sock *up = udp_sk(sk);
1085 struct flowi6 fl6;
1086 int err = 0;
1087
1088 if (up->pending == AF_INET)
1089 return udp_push_pending_frames(sk);
1090
1091 /* ip6_finish_skb will release the cork, so make a copy of
1092 * fl6 here.
1093 */
1094 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1095
1096 skb = ip6_finish_skb(sk);
1097 if (!skb)
1098 goto out;
1099
1100 err = udp_v6_send_skb(skb, &fl6);
1101
1da177e4
LT
1102out:
1103 up->len = 0;
1104 up->pending = 0;
1105 return err;
1106}
1107
1b784140 1108int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1da177e4
LT
1109{
1110 struct ipv6_txoptions opt_space;
1111 struct udp_sock *up = udp_sk(sk);
1112 struct inet_sock *inet = inet_sk(sk);
1113 struct ipv6_pinfo *np = inet6_sk(sk);
342dfc30 1114 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
20c59de2 1115 struct in6_addr *daddr, *final_p, final;
1da177e4 1116 struct ipv6_txoptions *opt = NULL;
45f6fad8 1117 struct ipv6_txoptions *opt_to_free = NULL;
1da177e4 1118 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1119 struct flowi6 fl6;
1da177e4
LT
1120 struct dst_entry *dst;
1121 int addr_len = msg->msg_namelen;
1122 int ulen = len;
1123 int hlimit = -1;
41a1f8ea 1124 int tclass = -1;
13b52cd4 1125 int dontfrag = -1;
1da177e4
LT
1126 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1127 int err;
987905de 1128 int connected = 0;
b2bf1e26 1129 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1130 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1da177e4
LT
1131
1132 /* destination address check */
1133 if (sin6) {
1134 if (addr_len < offsetof(struct sockaddr, sa_data))
1135 return -EINVAL;
1136
1137 switch (sin6->sin6_family) {
1138 case AF_INET6:
1139 if (addr_len < SIN6_LEN_RFC2133)
1140 return -EINVAL;
1141 daddr = &sin6->sin6_addr;
1142 break;
1143 case AF_INET:
1144 goto do_udp_sendmsg;
1145 case AF_UNSPEC:
1146 msg->msg_name = sin6 = NULL;
1147 msg->msg_namelen = addr_len = 0;
1148 daddr = NULL;
1149 break;
1150 default:
1151 return -EINVAL;
1152 }
1153 } else if (!up->pending) {
1154 if (sk->sk_state != TCP_ESTABLISHED)
1155 return -EDESTADDRREQ;
efe4208f 1156 daddr = &sk->sk_v6_daddr;
1ab1457c 1157 } else
1da177e4
LT
1158 daddr = NULL;
1159
1160 if (daddr) {
e773e4fa 1161 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1162 struct sockaddr_in sin;
1163 sin.sin_family = AF_INET;
c720c7e8 1164 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1165 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1166 msg->msg_name = &sin;
1167 msg->msg_namelen = sizeof(sin);
1168do_udp_sendmsg:
1169 if (__ipv6_only_sock(sk))
1170 return -ENETUNREACH;
1b784140 1171 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1172 }
1173 }
1174
1175 if (up->pending == AF_INET)
1b784140 1176 return udp_sendmsg(sk, msg, len);
1da177e4
LT
1177
1178 /* Rough check on arithmetic overflow,
b59e139b 1179 better check is made in ip6_append_data().
1da177e4
LT
1180 */
1181 if (len > INT_MAX - sizeof(struct udphdr))
1182 return -EMSGSIZE;
1ab1457c 1183
03485f2a 1184 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1da177e4
LT
1185 if (up->pending) {
1186 /*
1187 * There are pending frames.
1188 * The socket lock must be held while it's corked.
1189 */
1190 lock_sock(sk);
1191 if (likely(up->pending)) {
1192 if (unlikely(up->pending != AF_INET6)) {
1193 release_sock(sk);
1194 return -EAFNOSUPPORT;
1195 }
1196 dst = NULL;
1197 goto do_append_data;
1198 }
1199 release_sock(sk);
1200 }
1201 ulen += sizeof(struct udphdr);
1202
4c9483b2 1203 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1204
1205 if (sin6) {
1206 if (sin6->sin6_port == 0)
1207 return -EINVAL;
1208
1958b856 1209 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1210 daddr = &sin6->sin6_addr;
1211
1212 if (np->sndflow) {
4c9483b2
DM
1213 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1214 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1215 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1216 if (!flowlabel)
1da177e4 1217 return -EINVAL;
1da177e4
LT
1218 }
1219 }
1220
1221 /*
1222 * Otherwise it will be difficult to maintain
1223 * sk->sk_dst_cache.
1224 */
1225 if (sk->sk_state == TCP_ESTABLISHED &&
efe4208f
ED
1226 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1227 daddr = &sk->sk_v6_daddr;
1da177e4
LT
1228
1229 if (addr_len >= sizeof(struct sockaddr_in6) &&
1230 sin6->sin6_scope_id &&
842df073 1231 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
4c9483b2 1232 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1233 } else {
1234 if (sk->sk_state != TCP_ESTABLISHED)
1235 return -EDESTADDRREQ;
1236
1958b856 1237 fl6.fl6_dport = inet->inet_dport;
efe4208f 1238 daddr = &sk->sk_v6_daddr;
4c9483b2 1239 fl6.flowlabel = np->flow_label;
987905de 1240 connected = 1;
1da177e4
LT
1241 }
1242
4c9483b2
DM
1243 if (!fl6.flowi6_oif)
1244 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1245
4c9483b2
DM
1246 if (!fl6.flowi6_oif)
1247 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1248
4c9483b2 1249 fl6.flowi6_mark = sk->sk_mark;
51953d5b 1250
1da177e4
LT
1251 if (msg->msg_controllen) {
1252 opt = &opt_space;
1253 memset(opt, 0, sizeof(struct ipv6_txoptions));
1254 opt->tot_len = sizeof(*opt);
1255
73df66f8
TP
1256 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1257 &hlimit, &tclass, &dontfrag);
1da177e4
LT
1258 if (err < 0) {
1259 fl6_sock_release(flowlabel);
1260 return err;
1261 }
4c9483b2
DM
1262 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1263 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 1264 if (!flowlabel)
1da177e4
LT
1265 return -EINVAL;
1266 }
1267 if (!(opt->opt_nflen|opt->opt_flen))
1268 opt = NULL;
987905de 1269 connected = 0;
1da177e4 1270 }
45f6fad8
ED
1271 if (!opt) {
1272 opt = txopt_get(np);
1273 opt_to_free = opt;
1274 }
df9890c3
YH
1275 if (flowlabel)
1276 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1277 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1278
4c9483b2 1279 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1280 if (!ipv6_addr_any(daddr))
4e3fd7a0 1281 fl6.daddr = *daddr;
876c7f41 1282 else
4c9483b2
DM
1283 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1284 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1285 fl6.saddr = np->saddr;
1958b856 1286 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1287
4c9483b2 1288 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1289 if (final_p)
987905de 1290 connected = 0;
1da177e4 1291
4c9483b2
DM
1292 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1293 fl6.flowi6_oif = np->mcast_oif;
987905de 1294 connected = 0;
c4062dfc
EH
1295 } else if (!fl6.flowi6_oif)
1296 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1297
4c9483b2 1298 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1299
0e0d44ab 1300 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
1301 if (IS_ERR(dst)) {
1302 err = PTR_ERR(dst);
1303 dst = NULL;
1da177e4 1304 goto out;
14e50e57 1305 }
1da177e4 1306
5c98631c
LC
1307 if (hlimit < 0)
1308 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1da177e4 1309
e651f03a 1310 if (tclass < 0)
41a1f8ea 1311 tclass = np->tclass;
41a1f8ea 1312
1da177e4
LT
1313 if (msg->msg_flags&MSG_CONFIRM)
1314 goto do_confirm;
1315back_from_confirm:
1316
03485f2a
VY
1317 /* Lockless fast path for the non-corking case */
1318 if (!corkreq) {
1319 struct sk_buff *skb;
1320
1321 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1322 sizeof(struct udphdr), hlimit, tclass, opt,
1323 &fl6, (struct rt6_info *)dst,
1324 msg->msg_flags, dontfrag);
1325 err = PTR_ERR(skb);
1326 if (!IS_ERR_OR_NULL(skb))
1327 err = udp_v6_send_skb(skb, &fl6);
1328 goto release_dst;
1329 }
1330
1da177e4
LT
1331 lock_sock(sk);
1332 if (unlikely(up->pending)) {
1333 /* The socket is already corked while preparing it. */
1334 /* ... which is an evident application bug. --ANK */
1335 release_sock(sk);
1336
ba7a46f1 1337 net_dbg_ratelimited("udp cork app bug 2\n");
1da177e4
LT
1338 err = -EINVAL;
1339 goto out;
1340 }
1341
1342 up->pending = AF_INET6;
1343
1344do_append_data:
e36d3ff9
JP
1345 if (dontfrag < 0)
1346 dontfrag = np->dontfrag;
1da177e4 1347 up->len += ulen;
f69e6d13 1348 err = ip6_append_data(sk, getfrag, msg, ulen,
4c9483b2 1349 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
67ba4152 1350 (struct rt6_info *)dst,
13b52cd4 1351 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1da177e4
LT
1352 if (err)
1353 udp_v6_flush_pending_frames(sk);
1354 else if (!corkreq)
4c0a6cb0 1355 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1356 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1357 up->pending = 0;
1da177e4 1358
03485f2a
VY
1359 if (err > 0)
1360 err = np->recverr ? net_xmit_errno(err) : 0;
1361 release_sock(sk);
1362
1363release_dst:
a5e7c210
DM
1364 if (dst) {
1365 if (connected) {
1366 ip6_dst_store(sk, dst,
efe4208f
ED
1367 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1368 &sk->sk_v6_daddr : NULL,
8e1ef0a9 1369#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1370 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1371 &np->saddr :
1372#endif
1373 NULL);
a5e7c210
DM
1374 } else {
1375 dst_release(dst);
1376 }
a3c96089 1377 dst = NULL;
a5e7c210
DM
1378 }
1379
1da177e4 1380out:
a3c96089 1381 dst_release(dst);
1da177e4 1382 fl6_sock_release(flowlabel);
45f6fad8 1383 txopt_put(opt_to_free);
cd562c98 1384 if (!err)
1da177e4 1385 return len;
a18135eb
DM
1386 /*
1387 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1388 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1389 * we don't have a good statistic (IpOutDiscards but it can be too many
1390 * things). We could add another new stat but at least for now that
1391 * seems like overkill.
1392 */
1393 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
235b9f7a
PE
1394 UDP6_INC_STATS_USER(sock_net(sk),
1395 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1396 }
1da177e4
LT
1397 return err;
1398
1399do_confirm:
1400 dst_confirm(dst);
1401 if (!(msg->msg_flags&MSG_PROBE) || len)
1402 goto back_from_confirm;
1403 err = 0;
1404 goto out;
1405}
1406
7d06b2e0 1407void udpv6_destroy_sock(struct sock *sk)
1da177e4 1408{
44046a59 1409 struct udp_sock *up = udp_sk(sk);
1da177e4
LT
1410 lock_sock(sk);
1411 udp_v6_flush_pending_frames(sk);
1412 release_sock(sk);
1413
44046a59
TP
1414 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1415 void (*encap_destroy)(struct sock *sk);
1416 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1417 if (encap_destroy)
1418 encap_destroy(sk);
1419 }
1420
1da177e4 1421 inet6_destroy_sock(sk);
1da177e4
LT
1422}
1423
1424/*
1425 * Socket option code for UDP
1426 */
ba4e58ec 1427int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1428 char __user *optval, unsigned int optlen)
3fdadf7d 1429{
db8dac20 1430 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1431 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1432 udp_v6_push_pending_frames);
ba4e58ec 1433 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1434}
1435
1436#ifdef CONFIG_COMPAT
ba4e58ec 1437int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1438 char __user *optval, unsigned int optlen)
3fdadf7d 1439{
db8dac20 1440 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1441 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1442 udp_v6_push_pending_frames);
ba4e58ec 1443 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1444}
1445#endif
1446
ba4e58ec
GR
1447int udpv6_getsockopt(struct sock *sk, int level, int optname,
1448 char __user *optval, int __user *optlen)
3fdadf7d 1449{
db8dac20 1450 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1451 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1452 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1453}
1454
1455#ifdef CONFIG_COMPAT
ba4e58ec
GR
1456int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1457 char __user *optval, int __user *optlen)
3fdadf7d 1458{
db8dac20 1459 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1460 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1461 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1462}
1463#endif
1464
41135cc8 1465static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1466 .handler = udpv6_rcv,
1467 .err_handler = udpv6_err,
1468 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1469};
1470
1471/* ------------------------------------------------------------------------ */
1472#ifdef CONFIG_PROC_FS
ba4e58ec 1473int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4 1474{
17ef66af
LC
1475 if (v == SEQ_START_TOKEN) {
1476 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1477 } else {
1478 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1479 struct inet_sock *inet = inet_sk(v);
1480 __u16 srcp = ntohs(inet->inet_sport);
1481 __u16 destp = ntohs(inet->inet_dport);
1482 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1483 }
1da177e4
LT
1484 return 0;
1485}
1486
73cb88ec
AV
1487static const struct file_operations udp6_afinfo_seq_fops = {
1488 .owner = THIS_MODULE,
1489 .open = udp_seq_open,
1490 .read = seq_read,
1491 .llseek = seq_lseek,
1492 .release = seq_release_net
1493};
1494
1da177e4 1495static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1496 .name = "udp6",
1497 .family = AF_INET6,
645ca708 1498 .udp_table = &udp_table,
73cb88ec 1499 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1500 .seq_ops = {
1501 .show = udp6_seq_show,
1502 },
1da177e4
LT
1503};
1504
2c8c1e72 1505int __net_init udp6_proc_init(struct net *net)
1da177e4 1506{
0c96d8c5 1507 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1508}
1509
ec120da6
IM
1510void udp6_proc_exit(struct net *net)
1511{
0c96d8c5 1512 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1513}
1514#endif /* CONFIG_PROC_FS */
1515
f77d6021
ED
1516void udp_v6_clear_sk(struct sock *sk, int size)
1517{
1518 struct inet_sock *inet = inet_sk(sk);
1519
1520 /* we do not want to clear pinet6 field, because of RCU lookups */
1521 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1522
1523 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1524 memset(&inet->pinet6 + 1, 0, size);
1525}
1526
1da177e4
LT
1527/* ------------------------------------------------------------------------ */
1528
1529struct proto udpv6_prot = {
543d9cfe
ACM
1530 .name = "UDPv6",
1531 .owner = THIS_MODULE,
ba4e58ec 1532 .close = udp_lib_close,
543d9cfe
ACM
1533 .connect = ip6_datagram_connect,
1534 .disconnect = udp_disconnect,
1535 .ioctl = udp_ioctl,
1536 .destroy = udpv6_destroy_sock,
1537 .setsockopt = udpv6_setsockopt,
1538 .getsockopt = udpv6_getsockopt,
1539 .sendmsg = udpv6_sendmsg,
1540 .recvmsg = udpv6_recvmsg,
f7ad74fe 1541 .backlog_rcv = __udpv6_queue_rcv_skb,
ba4e58ec
GR
1542 .hash = udp_lib_hash,
1543 .unhash = udp_lib_unhash,
719f8358 1544 .rehash = udp_v6_rehash,
543d9cfe 1545 .get_port = udp_v6_get_port,
95766fff
HA
1546 .memory_allocated = &udp_memory_allocated,
1547 .sysctl_mem = sysctl_udp_mem,
1548 .sysctl_wmem = &sysctl_udp_wmem_min,
1549 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1550 .obj_size = sizeof(struct udp6_sock),
271b72c7 1551 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1552 .h.udp_table = &udp_table,
3fdadf7d 1553#ifdef CONFIG_COMPAT
543d9cfe
ACM
1554 .compat_setsockopt = compat_udpv6_setsockopt,
1555 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1556#endif
f77d6021 1557 .clear_sk = udp_v6_clear_sk,
1da177e4
LT
1558};
1559
1da177e4
LT
1560static struct inet_protosw udpv6_protosw = {
1561 .type = SOCK_DGRAM,
1562 .protocol = IPPROTO_UDP,
1563 .prot = &udpv6_prot,
1564 .ops = &inet6_dgram_ops,
1da177e4
LT
1565 .flags = INET_PROTOSW_PERMANENT,
1566};
1567
7f4e4868 1568int __init udpv6_init(void)
1da177e4 1569{
7f4e4868
DL
1570 int ret;
1571
3336288a
VY
1572 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1573 if (ret)
c6b641a4 1574 goto out;
3336288a 1575
7f4e4868
DL
1576 ret = inet6_register_protosw(&udpv6_protosw);
1577 if (ret)
1578 goto out_udpv6_protocol;
1579out:
1580 return ret;
1581
1582out_udpv6_protocol:
1583 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1584 goto out;
1585}
1586
09f7709f 1587void udpv6_exit(void)
7f4e4868
DL
1588{
1589 inet6_unregister_protosw(&udpv6_protosw);
1590 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1591}