ipv4: Implement IP_UNICAST_IF socket option.
[linux-2.6-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>
48
49#include <linux/proc_fs.h>
50#include <linux/seq_file.h>
ba4e58ec 51#include "udp_impl.h"
1da177e4 52
b2f5e7cd
VY
53int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
54{
55 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
56 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
68835aba
ED
57 __be32 sk1_rcv_saddr = sk_rcv_saddr(sk);
58 __be32 sk2_rcv_saddr = sk_rcv_saddr(sk2);
b2f5e7cd
VY
59 int sk_ipv6only = ipv6_only_sock(sk);
60 int sk2_ipv6only = inet_v6_ipv6only(sk2);
61 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
62 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
63
64 /* if both are mapped, treat as IPv4 */
65 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
499923c7 66 return (!sk2_ipv6only &&
c720c7e8
ED
67 (!sk1_rcv_saddr || !sk2_rcv_saddr ||
68 sk1_rcv_saddr == sk2_rcv_saddr));
b2f5e7cd
VY
69
70 if (addr_type2 == IPV6_ADDR_ANY &&
71 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
72 return 1;
73
74 if (addr_type == IPV6_ADDR_ANY &&
75 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
76 return 1;
77
78 if (sk2_rcv_saddr6 &&
79 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
80 return 1;
81
82 return 0;
83}
84
d4cada4a
ED
85static unsigned int udp6_portaddr_hash(struct net *net,
86 const struct in6_addr *addr6,
87 unsigned int port)
88{
89 unsigned int hash, mix = net_hash_mix(net);
90
91 if (ipv6_addr_any(addr6))
92 hash = jhash_1word(0, mix);
856540ee 93 else if (ipv6_addr_v4mapped(addr6))
0eae88f3 94 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
d4cada4a 95 else
0eae88f3 96 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
d4cada4a
ED
97
98 return hash ^ port;
99}
100
101
6ba5a3c5 102int udp_v6_get_port(struct sock *sk, unsigned short snum)
1da177e4 103{
30fff923
ED
104 unsigned int hash2_nulladdr =
105 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
106 unsigned int hash2_partial =
107 udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0);
108
d4cada4a 109 /* precompute partial secondary hash */
30fff923
ED
110 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
111 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
1da177e4
LT
112}
113
719f8358
ED
114static void udp_v6_rehash(struct sock *sk)
115{
116 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
117 &inet6_sk(sk)->rcv_saddr,
118 inet_sk(sk)->inet_num);
119
120 udp_lib_rehash(sk, new_hash);
121}
122
645ca708
ED
123static inline int compute_score(struct sock *sk, struct net *net,
124 unsigned short hnum,
88440ae7
BS
125 const struct in6_addr *saddr, __be16 sport,
126 const struct in6_addr *daddr, __be16 dport,
645ca708
ED
127 int dif)
128{
129 int score = -1;
130
d4cada4a 131 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
645ca708
ED
132 sk->sk_family == PF_INET6) {
133 struct ipv6_pinfo *np = inet6_sk(sk);
134 struct inet_sock *inet = inet_sk(sk);
135
136 score = 0;
c720c7e8
ED
137 if (inet->inet_dport) {
138 if (inet->inet_dport != sport)
645ca708
ED
139 return -1;
140 score++;
141 }
142 if (!ipv6_addr_any(&np->rcv_saddr)) {
143 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
144 return -1;
145 score++;
146 }
147 if (!ipv6_addr_any(&np->daddr)) {
148 if (!ipv6_addr_equal(&np->daddr, saddr))
149 return -1;
150 score++;
151 }
152 if (sk->sk_bound_dev_if) {
153 if (sk->sk_bound_dev_if != dif)
154 return -1;
155 score++;
156 }
157 }
158 return score;
159}
160
fddc17de
ED
161#define SCORE2_MAX (1 + 1 + 1)
162static inline int compute_score2(struct sock *sk, struct net *net,
163 const struct in6_addr *saddr, __be16 sport,
164 const struct in6_addr *daddr, unsigned short hnum,
165 int dif)
166{
167 int score = -1;
168
169 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
170 sk->sk_family == PF_INET6) {
171 struct ipv6_pinfo *np = inet6_sk(sk);
172 struct inet_sock *inet = inet_sk(sk);
173
174 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
175 return -1;
176 score = 0;
177 if (inet->inet_dport) {
178 if (inet->inet_dport != sport)
179 return -1;
180 score++;
181 }
182 if (!ipv6_addr_any(&np->daddr)) {
183 if (!ipv6_addr_equal(&np->daddr, saddr))
184 return -1;
185 score++;
186 }
187 if (sk->sk_bound_dev_if) {
188 if (sk->sk_bound_dev_if != dif)
189 return -1;
190 score++;
191 }
192 }
193 return score;
194}
195
fddc17de
ED
196
197/* called with read_rcu_lock() */
198static struct sock *udp6_lib_lookup2(struct net *net,
199 const struct in6_addr *saddr, __be16 sport,
200 const struct in6_addr *daddr, unsigned int hnum, int dif,
201 struct udp_hslot *hslot2, unsigned int slot2)
202{
203 struct sock *sk, *result;
204 struct hlist_nulls_node *node;
205 int score, badness;
206
207begin:
208 result = NULL;
209 badness = -1;
210 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
211 score = compute_score2(sk, net, saddr, sport,
212 daddr, hnum, dif);
213 if (score > badness) {
214 result = sk;
215 badness = score;
216 if (score == SCORE2_MAX)
217 goto exact_match;
218 }
219 }
220 /*
221 * if the nulls value we got at the end of this lookup is
222 * not the expected one, we must restart lookup.
223 * We probably met an item that was moved to another chain.
224 */
225 if (get_nulls_value(node) != slot2)
226 goto begin;
227
228 if (result) {
229exact_match:
c31504dc 230 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
fddc17de
ED
231 result = NULL;
232 else if (unlikely(compute_score2(result, net, saddr, sport,
233 daddr, hnum, dif) < badness)) {
234 sock_put(result);
235 goto begin;
236 }
237 }
238 return result;
239}
240
fce82338 241struct sock *__udp6_lib_lookup(struct net *net,
88440ae7
BS
242 const struct in6_addr *saddr, __be16 sport,
243 const struct in6_addr *daddr, __be16 dport,
645ca708 244 int dif, struct udp_table *udptable)
1da177e4 245{
271b72c7 246 struct sock *sk, *result;
88ab1932 247 struct hlist_nulls_node *node;
1da177e4 248 unsigned short hnum = ntohs(dport);
fddc17de
ED
249 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
250 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
271b72c7 251 int score, badness;
645ca708 252
271b72c7 253 rcu_read_lock();
fddc17de
ED
254 if (hslot->count > 10) {
255 hash2 = udp6_portaddr_hash(net, daddr, hnum);
256 slot2 = hash2 & udptable->mask;
257 hslot2 = &udptable->hash2[slot2];
258 if (hslot->count < hslot2->count)
259 goto begin;
260
261 result = udp6_lib_lookup2(net, saddr, sport,
262 daddr, hnum, dif,
263 hslot2, slot2);
264 if (!result) {
265 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
266 slot2 = hash2 & udptable->mask;
267 hslot2 = &udptable->hash2[slot2];
268 if (hslot->count < hslot2->count)
269 goto begin;
270
1223c67c
JBD
271 result = udp6_lib_lookup2(net, saddr, sport,
272 &in6addr_any, hnum, dif,
fddc17de
ED
273 hslot2, slot2);
274 }
275 rcu_read_unlock();
276 return result;
277 }
271b72c7
ED
278begin:
279 result = NULL;
280 badness = -1;
88ab1932 281 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
645ca708
ED
282 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
283 if (score > badness) {
284 result = sk;
285 badness = score;
1da177e4
LT
286 }
287 }
88ab1932
ED
288 /*
289 * if the nulls value we got at the end of this lookup is
290 * not the expected one, we must restart lookup.
291 * We probably met an item that was moved to another chain.
292 */
fddc17de 293 if (get_nulls_value(node) != slot)
88ab1932
ED
294 goto begin;
295
271b72c7 296 if (result) {
c31504dc 297 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
271b72c7
ED
298 result = NULL;
299 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
300 daddr, dport, dif) < badness)) {
301 sock_put(result);
302 goto begin;
303 }
304 }
305 rcu_read_unlock();
1da177e4
LT
306 return result;
307}
fce82338 308EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
1da177e4 309
607c4aaf
KK
310static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
311 __be16 sport, __be16 dport,
645ca708 312 struct udp_table *udptable)
607c4aaf 313{
23542618 314 struct sock *sk;
b71d1d42 315 const struct ipv6hdr *iph = ipv6_hdr(skb);
607c4aaf 316
23542618
KK
317 if (unlikely(sk = skb_steal_sock(skb)))
318 return sk;
adf30907
ED
319 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
320 &iph->daddr, dport, inet6_iif(skb),
321 udptable);
607c4aaf
KK
322}
323
aa976fc0
BS
324struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
325 const struct in6_addr *daddr, __be16 dport, int dif)
326{
327 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
328}
329EXPORT_SYMBOL_GPL(udp6_lib_lookup);
330
331
1da177e4
LT
332/*
333 * This should be easy, if there is something there we
334 * return it, otherwise we block.
335 */
336
ba4e58ec 337int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
338 struct msghdr *msg, size_t len,
339 int noblock, int flags, int *addr_len)
340{
341 struct ipv6_pinfo *np = inet6_sk(sk);
342 struct inet_sock *inet = inet_sk(sk);
1ab1457c 343 struct sk_buff *skb;
59c2cdae 344 unsigned int ulen, copied;
a59322be 345 int peeked;
759e5d00
HX
346 int err;
347 int is_udplite = IS_UDPLITE(sk);
f26ba175 348 int is_udp4;
8a74ad60 349 bool slow;
1da177e4 350
1ab1457c
YH
351 if (addr_len)
352 *addr_len=sizeof(struct sockaddr_in6);
353
1da177e4
LT
354 if (flags & MSG_ERRQUEUE)
355 return ipv6_recv_error(sk, msg, len);
356
4b340ae2
BH
357 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
358 return ipv6_recv_rxpmtu(sk, msg, len);
359
1da177e4 360try_again:
a59322be
HX
361 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
362 &peeked, &err);
1da177e4
LT
363 if (!skb)
364 goto out;
365
759e5d00 366 ulen = skb->len - sizeof(struct udphdr);
59c2cdae
DM
367 copied = len;
368 if (copied > ulen)
369 copied = ulen;
370 else if (copied < ulen)
1ab1457c 371 msg->msg_flags |= MSG_TRUNC;
1da177e4 372
f26ba175
WY
373 is_udp4 = (skb->protocol == htons(ETH_P_IP));
374
ba4e58ec 375 /*
759e5d00
HX
376 * If checksum is needed at all, try to do it while copying the
377 * data. If the data is truncated, or if we only want a partial
378 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 379 */
ba4e58ec 380
59c2cdae 381 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
759e5d00 382 if (udp_lib_checksum_complete(skb))
1da177e4 383 goto csum_copy_err;
ba4e58ec
GR
384 }
385
60476372 386 if (skb_csum_unnecessary(skb))
ba4e58ec 387 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
59c2cdae 388 msg->msg_iov, copied );
ba4e58ec 389 else {
1da177e4
LT
390 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
391 if (err == -EINVAL)
392 goto csum_copy_err;
393 }
394 if (err)
395 goto out_free;
396
f26ba175
WY
397 if (!peeked) {
398 if (is_udp4)
399 UDP_INC_STATS_USER(sock_net(sk),
400 UDP_MIB_INDATAGRAMS, is_udplite);
401 else
402 UDP6_INC_STATS_USER(sock_net(sk),
403 UDP_MIB_INDATAGRAMS, is_udplite);
404 }
cb75994e 405
3b885787 406 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
407
408 /* Copy the address. */
409 if (msg->msg_name) {
410 struct sockaddr_in6 *sin6;
1ab1457c 411
1da177e4
LT
412 sin6 = (struct sockaddr_in6 *) msg->msg_name;
413 sin6->sin6_family = AF_INET6;
4bedb452 414 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4
LT
415 sin6->sin6_flowinfo = 0;
416 sin6->sin6_scope_id = 0;
417
f26ba175 418 if (is_udp4)
b301e82c
BH
419 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
420 &sin6->sin6_addr);
1da177e4 421 else {
4e3fd7a0 422 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
1da177e4
LT
423 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
424 sin6->sin6_scope_id = IP6CB(skb)->iif;
425 }
426
427 }
f26ba175 428 if (is_udp4) {
1da177e4
LT
429 if (inet->cmsg_flags)
430 ip_cmsg_recv(msg, skb);
431 } else {
432 if (np->rxopt.all)
433 datagram_recv_ctl(sk, msg, skb);
1ab1457c 434 }
1da177e4 435
59c2cdae 436 err = copied;
1da177e4 437 if (flags & MSG_TRUNC)
759e5d00 438 err = ulen;
1da177e4
LT
439
440out_free:
9d410c79 441 skb_free_datagram_locked(sk, skb);
1da177e4
LT
442out:
443 return err;
444
445csum_copy_err:
8a74ad60 446 slow = lock_sock_fast(sk);
0856f939
WY
447 if (!skb_kill_datagram(sk, skb, flags)) {
448 if (is_udp4)
449 UDP_INC_STATS_USER(sock_net(sk),
450 UDP_MIB_INERRORS, is_udplite);
451 else
452 UDP6_INC_STATS_USER(sock_net(sk),
453 UDP_MIB_INERRORS, is_udplite);
454 }
8a74ad60 455 unlock_sock_fast(sk, slow);
1da177e4 456
32c90254 457 if (noblock)
1da177e4 458 return -EAGAIN;
9cfaa8de
XZ
459
460 /* starting over for a new packet */
461 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
462 goto try_again;
463}
464
ba4e58ec 465void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 466 u8 type, u8 code, int offset, __be32 info,
645ca708 467 struct udp_table *udptable)
1da177e4
LT
468{
469 struct ipv6_pinfo *np;
b71d1d42
ED
470 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
471 const struct in6_addr *saddr = &hdr->saddr;
472 const struct in6_addr *daddr = &hdr->daddr;
1da177e4
LT
473 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
474 struct sock *sk;
475 int err;
476
c346dca1 477 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
ba4e58ec 478 saddr, uh->source, inet6_iif(skb), udptable);
1da177e4
LT
479 if (sk == NULL)
480 return;
481
482 np = inet6_sk(sk);
483
484 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
485 goto out;
486
487 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
488 goto out;
489
b1faf566 490 if (np->recverr)
1da177e4 491 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
b1faf566 492
1da177e4
LT
493 sk->sk_err = err;
494 sk->sk_error_report(sk);
495out:
496 sock_put(sk);
497}
498
ba4e58ec 499static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba
BH
500 struct inet6_skb_parm *opt, u8 type,
501 u8 code, int offset, __be32 info )
ba4e58ec 502{
645ca708 503 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
504}
505
506int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1da177e4 507{
ba4e58ec 508 struct udp_sock *up = udp_sk(sk);
a18135eb 509 int rc;
b2bf1e26 510 int is_udplite = IS_UDPLITE(sk);
a18135eb 511
47482f13 512 if (!ipv6_addr_any(&inet6_sk(sk)->daddr))
bdeab991 513 sock_rps_save_rxhash(sk, skb);
47482f13 514
ba4e58ec
GR
515 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
516 goto drop;
1da177e4 517
ba4e58ec
GR
518 /*
519 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
520 */
b2bf1e26 521 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
522
523 if (up->pcrlen == 0) { /* full coverage was set */
524 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
525 " %d while full coverage %d requested\n",
526 UDP_SKB_CB(skb)->cscov, skb->len);
527 goto drop;
528 }
529 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
530 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
531 "too small, need min %d\n",
532 UDP_SKB_CB(skb)->cscov, up->pcrlen);
533 goto drop;
534 }
1da177e4
LT
535 }
536
33d480ce 537 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62
HX
538 if (udp_lib_checksum_complete(skb))
539 goto drop;
540 }
ba4e58ec 541
d826eb14
ED
542 skb_dst_drop(skb);
543 rc = sock_queue_rcv_skb(sk, skb);
544 if (rc < 0) {
a18135eb 545 /* Note that an ENOMEM error is charged twice */
766e9037 546 if (rc == -ENOMEM)
ef28d1a2
PE
547 UDP6_INC_STATS_BH(sock_net(sk),
548 UDP_MIB_RCVBUFERRORS, is_udplite);
8edf19c2 549 goto drop_no_sk_drops_inc;
1da177e4 550 }
cb75994e 551
1da177e4 552 return 0;
ba4e58ec 553drop:
8edf19c2
ED
554 atomic_inc(&sk->sk_drops);
555drop_no_sk_drops_inc:
ef28d1a2 556 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
ba4e58ec
GR
557 kfree_skb(skb);
558 return -1;
1da177e4
LT
559}
560
920a4611 561static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
b71d1d42
ED
562 __be16 loc_port, const struct in6_addr *loc_addr,
563 __be16 rmt_port, const struct in6_addr *rmt_addr,
1da177e4
LT
564 int dif)
565{
88ab1932 566 struct hlist_nulls_node *node;
1da177e4
LT
567 struct sock *s = sk;
568 unsigned short num = ntohs(loc_port);
569
88ab1932 570 sk_nulls_for_each_from(s, node) {
1da177e4
LT
571 struct inet_sock *inet = inet_sk(s);
572
920a4611 573 if (!net_eq(sock_net(s), net))
b8ad0cbc
DL
574 continue;
575
d4cada4a
ED
576 if (udp_sk(s)->udp_port_hash == num &&
577 s->sk_family == PF_INET6) {
1da177e4 578 struct ipv6_pinfo *np = inet6_sk(s);
c720c7e8
ED
579 if (inet->inet_dport) {
580 if (inet->inet_dport != rmt_port)
1da177e4
LT
581 continue;
582 }
583 if (!ipv6_addr_any(&np->daddr) &&
584 !ipv6_addr_equal(&np->daddr, rmt_addr))
585 continue;
586
587 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
588 continue;
589
590 if (!ipv6_addr_any(&np->rcv_saddr)) {
40796c5e
DS
591 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
592 continue;
1da177e4 593 }
add459aa 594 if (!inet6_mc_check(s, loc_addr, rmt_addr))
1da177e4
LT
595 continue;
596 return s;
597 }
598 }
599 return NULL;
600}
601
a1ab77f9
ED
602static void flush_stack(struct sock **stack, unsigned int count,
603 struct sk_buff *skb, unsigned int final)
604{
605 unsigned int i;
606 struct sock *sk;
607 struct sk_buff *skb1;
608
609 for (i = 0; i < count; i++) {
610 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
611
f6b8f32c 612 sk = stack[i];
a1ab77f9 613 if (skb1) {
c0722400 614 if (sk_rcvqueues_full(sk, skb1)) {
c377411f
ED
615 kfree_skb(skb1);
616 goto drop;
617 }
a1ab77f9
ED
618 bh_lock_sock(sk);
619 if (!sock_owned_by_user(sk))
620 udpv6_queue_rcv_skb(sk, skb1);
a3a858ff 621 else if (sk_add_backlog(sk, skb1)) {
55349790
ZY
622 kfree_skb(skb1);
623 bh_unlock_sock(sk);
624 goto drop;
625 }
a1ab77f9 626 bh_unlock_sock(sk);
55349790 627 continue;
a1ab77f9 628 }
55349790
ZY
629drop:
630 atomic_inc(&sk->sk_drops);
631 UDP6_INC_STATS_BH(sock_net(sk),
632 UDP_MIB_RCVBUFERRORS, IS_UDPLITE(sk));
633 UDP6_INC_STATS_BH(sock_net(sk),
634 UDP_MIB_INERRORS, IS_UDPLITE(sk));
a1ab77f9
ED
635 }
636}
1da177e4
LT
637/*
638 * Note: called only from the BH handler context,
639 * so we don't need to lock the hashes.
640 */
e3163493 641static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 642 const struct in6_addr *saddr, const struct in6_addr *daddr,
645ca708 643 struct udp_table *udptable)
1da177e4 644{
a1ab77f9 645 struct sock *sk, *stack[256 / sizeof(struct sock *)];
4bedb452 646 const struct udphdr *uh = udp_hdr(skb);
f86dcc5a 647 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
1da177e4 648 int dif;
a1ab77f9 649 unsigned int i, count = 0;
1da177e4 650
645ca708 651 spin_lock(&hslot->lock);
88ab1932 652 sk = sk_nulls_head(&hslot->head);
f2776ff0 653 dif = inet6_iif(skb);
920a4611 654 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
a1ab77f9
ED
655 while (sk) {
656 stack[count++] = sk;
657 sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
658 uh->source, saddr, dif);
659 if (unlikely(count == ARRAY_SIZE(stack))) {
660 if (!sk)
661 break;
662 flush_stack(stack, count, skb, ~0);
663 count = 0;
95766fff 664 }
1da177e4 665 }
a1ab77f9
ED
666 /*
667 * before releasing the lock, we must take reference on sockets
668 */
669 for (i = 0; i < count; i++)
670 sock_hold(stack[i]);
671
645ca708 672 spin_unlock(&hslot->lock);
a1ab77f9
ED
673
674 if (count) {
675 flush_stack(stack, count, skb, count - 1);
676
677 for (i = 0; i < count; i++)
678 sock_put(stack[i]);
679 } else {
680 kfree_skb(skb);
681 }
ba4e58ec 682 return 0;
1da177e4
LT
683}
684
759e5d00
HX
685static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
686 int proto)
ba4e58ec 687{
759e5d00
HX
688 int err;
689
690 UDP_SKB_CB(skb)->partial_cov = 0;
691 UDP_SKB_CB(skb)->cscov = skb->len;
692
db8dac20 693 if (proto == IPPROTO_UDPLITE) {
759e5d00
HX
694 err = udplite_checksum_init(skb, uh);
695 if (err)
696 return err;
697 }
698
ba4e58ec
GR
699 if (uh->check == 0) {
700 /* RFC 2460 section 8.1 says that we SHOULD log
701 this error. Well, it is reasonable.
702 */
703 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
704 return 1;
705 }
706 if (skb->ip_summed == CHECKSUM_COMPLETE &&
0660e03f 707 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
759e5d00 708 skb->len, proto, skb->csum))
ba4e58ec
GR
709 skb->ip_summed = CHECKSUM_UNNECESSARY;
710
60476372 711 if (!skb_csum_unnecessary(skb))
0660e03f
ACM
712 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
713 &ipv6_hdr(skb)->daddr,
759e5d00 714 skb->len, proto, 0));
ba4e58ec 715
759e5d00 716 return 0;
ba4e58ec
GR
717}
718
645ca708 719int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 720 int proto)
1da177e4 721{
3ffe533c 722 struct net *net = dev_net(skb->dev);
1da177e4 723 struct sock *sk;
1ab1457c 724 struct udphdr *uh;
b71d1d42 725 const struct in6_addr *saddr, *daddr;
1da177e4
LT
726 u32 ulen = 0;
727
728 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 729 goto discard;
1da177e4 730
0660e03f
ACM
731 saddr = &ipv6_hdr(skb)->saddr;
732 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 733 uh = udp_hdr(skb);
1da177e4
LT
734
735 ulen = ntohs(uh->len);
ba4e58ec
GR
736 if (ulen > skb->len)
737 goto short_packet;
1da177e4 738
759e5d00
HX
739 if (proto == IPPROTO_UDP) {
740 /* UDP validates ulen. */
1da177e4 741
ba4e58ec
GR
742 /* Check for jumbo payload */
743 if (ulen == 0)
744 ulen = skb->len;
1da177e4 745
ba4e58ec
GR
746 if (ulen < sizeof(*uh))
747 goto short_packet;
1da177e4 748
ba4e58ec
GR
749 if (ulen < skb->len) {
750 if (pskb_trim_rcsum(skb, ulen))
751 goto short_packet;
0660e03f
ACM
752 saddr = &ipv6_hdr(skb)->saddr;
753 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 754 uh = udp_hdr(skb);
ba4e58ec 755 }
ba4e58ec 756 }
1da177e4 757
759e5d00
HX
758 if (udp6_csum_init(skb, uh, proto))
759 goto discard;
760
1ab1457c
YH
761 /*
762 * Multicast receive code
1da177e4 763 */
ba4e58ec 764 if (ipv6_addr_is_multicast(daddr))
e3163493
PE
765 return __udp6_lib_mcast_deliver(net, skb,
766 saddr, daddr, udptable);
1da177e4
LT
767
768 /* Unicast */
769
1ab1457c 770 /*
1da177e4
LT
771 * check socket cache ... must talk to Alan about his plans
772 * for sock caches... i'll skip this for now.
773 */
607c4aaf 774 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
1da177e4
LT
775
776 if (sk == NULL) {
777 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
778 goto discard;
779
ba4e58ec 780 if (udp_lib_checksum_complete(skb))
1da177e4 781 goto discard;
ef28d1a2
PE
782 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS,
783 proto == IPPROTO_UDPLITE);
1da177e4 784
3ffe533c 785 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
1da177e4
LT
786
787 kfree_skb(skb);
add459aa 788 return 0;
1da177e4 789 }
1ab1457c 790
1da177e4 791 /* deliver */
1ab1457c 792
c377411f
ED
793 if (sk_rcvqueues_full(sk, skb)) {
794 sock_put(sk);
795 goto discard;
796 }
d97106ea 797 bh_lock_sock(sk);
95766fff
HA
798 if (!sock_owned_by_user(sk))
799 udpv6_queue_rcv_skb(sk, skb);
a3a858ff 800 else if (sk_add_backlog(sk, skb)) {
55349790
ZY
801 atomic_inc(&sk->sk_drops);
802 bh_unlock_sock(sk);
803 sock_put(sk);
804 goto discard;
805 }
95766fff 806 bh_unlock_sock(sk);
1da177e4 807 sock_put(sk);
add459aa 808 return 0;
1da177e4 809
1ab1457c 810short_packet:
d6bc0149 811 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
db8dac20 812 proto == IPPROTO_UDPLITE ? "-Lite" : "",
d6bc0149
BM
813 saddr,
814 ntohs(uh->source),
815 ulen,
816 skb->len,
817 daddr,
818 ntohs(uh->dest));
1da177e4
LT
819
820discard:
ef28d1a2 821 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 822 kfree_skb(skb);
add459aa 823 return 0;
1da177e4 824}
ba4e58ec 825
e5bbef20 826static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 827{
645ca708 828 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
829}
830
1da177e4
LT
831/*
832 * Throw away all pending data and cancel the corking. Socket is locked.
833 */
834static void udp_v6_flush_pending_frames(struct sock *sk)
835{
836 struct udp_sock *up = udp_sk(sk);
837
36d926b9
DL
838 if (up->pending == AF_INET)
839 udp_flush_pending_frames(sk);
840 else if (up->pending) {
1da177e4
LT
841 up->len = 0;
842 up->pending = 0;
843 ip6_flush_pending_frames(sk);
1ab1457c 844 }
1da177e4
LT
845}
846
493c6be3
SS
847/**
848 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
849 * @sk: socket we are sending on
850 * @skb: sk_buff containing the filled-in UDP header
851 * (checksum field must be zeroed out)
852 */
853static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
854 const struct in6_addr *saddr,
855 const struct in6_addr *daddr, int len)
856{
857 unsigned int offset;
858 struct udphdr *uh = udp_hdr(skb);
859 __wsum csum = 0;
860
861 if (skb_queue_len(&sk->sk_write_queue) == 1) {
862 /* Only one fragment on the socket. */
863 skb->csum_start = skb_transport_header(skb) - skb->head;
864 skb->csum_offset = offsetof(struct udphdr, check);
865 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
866 } else {
867 /*
868 * HW-checksum won't work as there are two or more
869 * fragments on the socket so that all csums of sk_buffs
870 * should be together
871 */
872 offset = skb_transport_offset(skb);
873 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
874
875 skb->ip_summed = CHECKSUM_NONE;
876
877 skb_queue_walk(&sk->sk_write_queue, skb) {
878 csum = csum_add(csum, skb->csum);
879 }
880
881 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
882 csum);
883 if (uh->check == 0)
884 uh->check = CSUM_MANGLED_0;
885 }
886}
887
1da177e4
LT
888/*
889 * Sending
890 */
891
4c0a6cb0 892static int udp_v6_push_pending_frames(struct sock *sk)
1da177e4
LT
893{
894 struct sk_buff *skb;
895 struct udphdr *uh;
4c0a6cb0 896 struct udp_sock *up = udp_sk(sk);
1da177e4 897 struct inet_sock *inet = inet_sk(sk);
4c9483b2 898 struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
1da177e4 899 int err = 0;
b2bf1e26 900 int is_udplite = IS_UDPLITE(sk);
868c86bc 901 __wsum csum = 0;
1da177e4
LT
902
903 /* Grab the skbuff where UDP header space exists. */
904 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
905 goto out;
906
907 /*
908 * Create a UDP header
909 */
4bedb452 910 uh = udp_hdr(skb);
1958b856
DM
911 uh->source = fl6->fl6_sport;
912 uh->dest = fl6->fl6_dport;
1da177e4
LT
913 uh->len = htons(up->len);
914 uh->check = 0;
915
b2bf1e26 916 if (is_udplite)
ba4e58ec 917 csum = udplite_csum_outgoing(sk, skb);
493c6be3 918 else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
4c9483b2 919 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
493c6be3
SS
920 up->len);
921 goto send;
922 } else
ba4e58ec 923 csum = udp_csum_outgoing(sk, skb);
1da177e4 924
ba4e58ec 925 /* add protocol-dependent pseudo-header */
4c9483b2
DM
926 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
927 up->len, fl6->flowi6_proto, csum);
1da177e4 928 if (uh->check == 0)
f6ab0288 929 uh->check = CSUM_MANGLED_0;
1da177e4 930
493c6be3 931send:
1da177e4 932 err = ip6_push_pending_frames(sk);
6ce9e7b5
ED
933 if (err) {
934 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
935 UDP6_INC_STATS_USER(sock_net(sk),
936 UDP_MIB_SNDBUFERRORS, is_udplite);
937 err = 0;
938 }
939 } else
940 UDP6_INC_STATS_USER(sock_net(sk),
941 UDP_MIB_OUTDATAGRAMS, is_udplite);
1da177e4
LT
942out:
943 up->len = 0;
944 up->pending = 0;
945 return err;
946}
947
ba4e58ec 948int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
949 struct msghdr *msg, size_t len)
950{
951 struct ipv6_txoptions opt_space;
952 struct udp_sock *up = udp_sk(sk);
953 struct inet_sock *inet = inet_sk(sk);
954 struct ipv6_pinfo *np = inet6_sk(sk);
955 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
20c59de2 956 struct in6_addr *daddr, *final_p, final;
1da177e4
LT
957 struct ipv6_txoptions *opt = NULL;
958 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 959 struct flowi6 fl6;
1da177e4
LT
960 struct dst_entry *dst;
961 int addr_len = msg->msg_namelen;
962 int ulen = len;
963 int hlimit = -1;
41a1f8ea 964 int tclass = -1;
13b52cd4 965 int dontfrag = -1;
1da177e4
LT
966 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
967 int err;
987905de 968 int connected = 0;
b2bf1e26 969 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 970 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1da177e4
LT
971
972 /* destination address check */
973 if (sin6) {
974 if (addr_len < offsetof(struct sockaddr, sa_data))
975 return -EINVAL;
976
977 switch (sin6->sin6_family) {
978 case AF_INET6:
979 if (addr_len < SIN6_LEN_RFC2133)
980 return -EINVAL;
981 daddr = &sin6->sin6_addr;
982 break;
983 case AF_INET:
984 goto do_udp_sendmsg;
985 case AF_UNSPEC:
986 msg->msg_name = sin6 = NULL;
987 msg->msg_namelen = addr_len = 0;
988 daddr = NULL;
989 break;
990 default:
991 return -EINVAL;
992 }
993 } else if (!up->pending) {
994 if (sk->sk_state != TCP_ESTABLISHED)
995 return -EDESTADDRREQ;
996 daddr = &np->daddr;
1ab1457c 997 } else
1da177e4
LT
998 daddr = NULL;
999
1000 if (daddr) {
e773e4fa 1001 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1002 struct sockaddr_in sin;
1003 sin.sin_family = AF_INET;
c720c7e8 1004 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1005 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1006 msg->msg_name = &sin;
1007 msg->msg_namelen = sizeof(sin);
1008do_udp_sendmsg:
1009 if (__ipv6_only_sock(sk))
1010 return -ENETUNREACH;
1011 return udp_sendmsg(iocb, sk, msg, len);
1012 }
1013 }
1014
1015 if (up->pending == AF_INET)
1016 return udp_sendmsg(iocb, sk, msg, len);
1017
1018 /* Rough check on arithmetic overflow,
b59e139b 1019 better check is made in ip6_append_data().
1da177e4
LT
1020 */
1021 if (len > INT_MAX - sizeof(struct udphdr))
1022 return -EMSGSIZE;
1ab1457c 1023
1da177e4
LT
1024 if (up->pending) {
1025 /*
1026 * There are pending frames.
1027 * The socket lock must be held while it's corked.
1028 */
1029 lock_sock(sk);
1030 if (likely(up->pending)) {
1031 if (unlikely(up->pending != AF_INET6)) {
1032 release_sock(sk);
1033 return -EAFNOSUPPORT;
1034 }
1035 dst = NULL;
1036 goto do_append_data;
1037 }
1038 release_sock(sk);
1039 }
1040 ulen += sizeof(struct udphdr);
1041
4c9483b2 1042 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1043
1044 if (sin6) {
1045 if (sin6->sin6_port == 0)
1046 return -EINVAL;
1047
1958b856 1048 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1049 daddr = &sin6->sin6_addr;
1050
1051 if (np->sndflow) {
4c9483b2
DM
1052 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1053 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1054 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1055 if (flowlabel == NULL)
1056 return -EINVAL;
1057 daddr = &flowlabel->dst;
1058 }
1059 }
1060
1061 /*
1062 * Otherwise it will be difficult to maintain
1063 * sk->sk_dst_cache.
1064 */
1065 if (sk->sk_state == TCP_ESTABLISHED &&
1066 ipv6_addr_equal(daddr, &np->daddr))
1067 daddr = &np->daddr;
1068
1069 if (addr_len >= sizeof(struct sockaddr_in6) &&
1070 sin6->sin6_scope_id &&
1071 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
4c9483b2 1072 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1073 } else {
1074 if (sk->sk_state != TCP_ESTABLISHED)
1075 return -EDESTADDRREQ;
1076
1958b856 1077 fl6.fl6_dport = inet->inet_dport;
1da177e4 1078 daddr = &np->daddr;
4c9483b2 1079 fl6.flowlabel = np->flow_label;
987905de 1080 connected = 1;
1da177e4
LT
1081 }
1082
4c9483b2
DM
1083 if (!fl6.flowi6_oif)
1084 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1085
4c9483b2
DM
1086 if (!fl6.flowi6_oif)
1087 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1088
4c9483b2 1089 fl6.flowi6_mark = sk->sk_mark;
51953d5b 1090
1da177e4
LT
1091 if (msg->msg_controllen) {
1092 opt = &opt_space;
1093 memset(opt, 0, sizeof(struct ipv6_txoptions));
1094 opt->tot_len = sizeof(*opt);
1095
ec0506db
MÅ»
1096 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1097 &hlimit, &tclass, &dontfrag);
1da177e4
LT
1098 if (err < 0) {
1099 fl6_sock_release(flowlabel);
1100 return err;
1101 }
4c9483b2
DM
1102 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1103 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1104 if (flowlabel == NULL)
1105 return -EINVAL;
1106 }
1107 if (!(opt->opt_nflen|opt->opt_flen))
1108 opt = NULL;
987905de 1109 connected = 0;
1da177e4
LT
1110 }
1111 if (opt == NULL)
1112 opt = np->opt;
df9890c3
YH
1113 if (flowlabel)
1114 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1115 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1116
4c9483b2 1117 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1118 if (!ipv6_addr_any(daddr))
4e3fd7a0 1119 fl6.daddr = *daddr;
876c7f41 1120 else
4c9483b2
DM
1121 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1122 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1123 fl6.saddr = np->saddr;
1958b856 1124 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1125
4c9483b2 1126 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1127 if (final_p)
987905de 1128 connected = 0;
1da177e4 1129
4c9483b2
DM
1130 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1131 fl6.flowi6_oif = np->mcast_oif;
987905de
MK
1132 connected = 0;
1133 }
1da177e4 1134
4c9483b2 1135 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1136
4c9483b2 1137 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
68d0c6d3
DM
1138 if (IS_ERR(dst)) {
1139 err = PTR_ERR(dst);
1140 dst = NULL;
1da177e4 1141 goto out;
14e50e57 1142 }
1da177e4
LT
1143
1144 if (hlimit < 0) {
4c9483b2 1145 if (ipv6_addr_is_multicast(&fl6.daddr))
1da177e4
LT
1146 hlimit = np->mcast_hops;
1147 else
1148 hlimit = np->hop_limit;
1149 if (hlimit < 0)
6b75d090 1150 hlimit = ip6_dst_hoplimit(dst);
1da177e4
LT
1151 }
1152
e651f03a 1153 if (tclass < 0)
41a1f8ea 1154 tclass = np->tclass;
41a1f8ea 1155
13b52cd4
BH
1156 if (dontfrag < 0)
1157 dontfrag = np->dontfrag;
1158
1da177e4
LT
1159 if (msg->msg_flags&MSG_CONFIRM)
1160 goto do_confirm;
1161back_from_confirm:
1162
1163 lock_sock(sk);
1164 if (unlikely(up->pending)) {
1165 /* The socket is already corked while preparing it. */
1166 /* ... which is an evident application bug. --ANK */
1167 release_sock(sk);
1168
64ce2073 1169 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1da177e4
LT
1170 err = -EINVAL;
1171 goto out;
1172 }
1173
1174 up->pending = AF_INET6;
1175
1176do_append_data:
1177 up->len += ulen;
ba4e58ec
GR
1178 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1179 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
4c9483b2 1180 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
41a1f8ea 1181 (struct rt6_info*)dst,
13b52cd4 1182 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1da177e4
LT
1183 if (err)
1184 udp_v6_flush_pending_frames(sk);
1185 else if (!corkreq)
4c0a6cb0 1186 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1187 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1188 up->pending = 0;
1da177e4 1189
a5e7c210
DM
1190 if (dst) {
1191 if (connected) {
1192 ip6_dst_store(sk, dst,
4c9483b2 1193 ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
8e1ef0a9
YH
1194 &np->daddr : NULL,
1195#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1196 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1197 &np->saddr :
1198#endif
1199 NULL);
a5e7c210
DM
1200 } else {
1201 dst_release(dst);
1202 }
a3c96089 1203 dst = NULL;
a5e7c210
DM
1204 }
1205
1da177e4
LT
1206 if (err > 0)
1207 err = np->recverr ? net_xmit_errno(err) : 0;
1208 release_sock(sk);
1209out:
a3c96089 1210 dst_release(dst);
1da177e4 1211 fl6_sock_release(flowlabel);
cd562c98 1212 if (!err)
1da177e4 1213 return len;
a18135eb
DM
1214 /*
1215 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1216 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1217 * we don't have a good statistic (IpOutDiscards but it can be too many
1218 * things). We could add another new stat but at least for now that
1219 * seems like overkill.
1220 */
1221 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
235b9f7a
PE
1222 UDP6_INC_STATS_USER(sock_net(sk),
1223 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1224 }
1da177e4
LT
1225 return err;
1226
1227do_confirm:
1228 dst_confirm(dst);
1229 if (!(msg->msg_flags&MSG_PROBE) || len)
1230 goto back_from_confirm;
1231 err = 0;
1232 goto out;
1233}
1234
7d06b2e0 1235void udpv6_destroy_sock(struct sock *sk)
1da177e4
LT
1236{
1237 lock_sock(sk);
1238 udp_v6_flush_pending_frames(sk);
1239 release_sock(sk);
1240
1241 inet6_destroy_sock(sk);
1da177e4
LT
1242}
1243
1244/*
1245 * Socket option code for UDP
1246 */
ba4e58ec 1247int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1248 char __user *optval, unsigned int optlen)
3fdadf7d 1249{
db8dac20 1250 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1251 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1252 udp_v6_push_pending_frames);
ba4e58ec 1253 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1254}
1255
1256#ifdef CONFIG_COMPAT
ba4e58ec 1257int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1258 char __user *optval, unsigned int optlen)
3fdadf7d 1259{
db8dac20 1260 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1261 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1262 udp_v6_push_pending_frames);
ba4e58ec 1263 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1264}
1265#endif
1266
ba4e58ec
GR
1267int udpv6_getsockopt(struct sock *sk, int level, int optname,
1268 char __user *optval, int __user *optlen)
3fdadf7d 1269{
db8dac20 1270 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1271 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1272 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1273}
1274
1275#ifdef CONFIG_COMPAT
ba4e58ec
GR
1276int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1277 char __user *optval, int __user *optlen)
3fdadf7d 1278{
db8dac20 1279 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1280 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1281 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1282}
1283#endif
1284
ba735425
SS
1285static int udp6_ufo_send_check(struct sk_buff *skb)
1286{
b71d1d42 1287 const struct ipv6hdr *ipv6h;
ba735425
SS
1288 struct udphdr *uh;
1289
1290 if (!pskb_may_pull(skb, sizeof(*uh)))
1291 return -EINVAL;
1292
1293 ipv6h = ipv6_hdr(skb);
1294 uh = udp_hdr(skb);
1295
1296 uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
1297 IPPROTO_UDP, 0);
1298 skb->csum_start = skb_transport_header(skb) - skb->head;
1299 skb->csum_offset = offsetof(struct udphdr, check);
1300 skb->ip_summed = CHECKSUM_PARTIAL;
1301 return 0;
1302}
1303
c8f44aff
MM
1304static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
1305 netdev_features_t features)
ba735425
SS
1306{
1307 struct sk_buff *segs = ERR_PTR(-EINVAL);
1308 unsigned int mss;
1309 unsigned int unfrag_ip6hlen, unfrag_len;
1310 struct frag_hdr *fptr;
1311 u8 *mac_start, *prevhdr;
1312 u8 nexthdr;
1313 u8 frag_hdr_sz = sizeof(struct frag_hdr);
1314 int offset;
1315 __wsum csum;
1316
1317 mss = skb_shinfo(skb)->gso_size;
1318 if (unlikely(skb->len <= mss))
1319 goto out;
1320
1321 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) {
1322 /* Packet is from an untrusted source, reset gso_segs. */
1323 int type = skb_shinfo(skb)->gso_type;
1324
1325 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) ||
1326 !(type & (SKB_GSO_UDP))))
1327 goto out;
1328
1329 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss);
1330
1331 segs = NULL;
1332 goto out;
1333 }
1334
1335 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot
1336 * do checksum of UDP packets sent as multiple IP fragments.
1337 */
96339d6c 1338 offset = skb_checksum_start_offset(skb);
ba735425
SS
1339 csum = skb_checksum(skb, offset, skb->len- offset, 0);
1340 offset += skb->csum_offset;
1341 *(__sum16 *)(skb->data + offset) = csum_fold(csum);
1342 skb->ip_summed = CHECKSUM_NONE;
1343
1344 /* Check if there is enough headroom to insert fragment header. */
a9cf73ea 1345 if ((skb_mac_header(skb) < skb->head + frag_hdr_sz) &&
ba735425
SS
1346 pskb_expand_head(skb, frag_hdr_sz, 0, GFP_ATOMIC))
1347 goto out;
1348
1349 /* Find the unfragmentable header and shift it left by frag_hdr_sz
1350 * bytes to insert fragment header.
1351 */
1352 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
1353 nexthdr = *prevhdr;
1354 *prevhdr = NEXTHDR_FRAGMENT;
1355 unfrag_len = skb_network_header(skb) - skb_mac_header(skb) +
1356 unfrag_ip6hlen;
1357 mac_start = skb_mac_header(skb);
1358 memmove(mac_start-frag_hdr_sz, mac_start, unfrag_len);
1359
1360 skb->mac_header -= frag_hdr_sz;
1361 skb->network_header -= frag_hdr_sz;
1362
1363 fptr = (struct frag_hdr *)(skb_network_header(skb) + unfrag_ip6hlen);
1364 fptr->nexthdr = nexthdr;
1365 fptr->reserved = 0;
87c48fa3 1366 ipv6_select_ident(fptr, (struct rt6_info *)skb_dst(skb));
ba735425
SS
1367
1368 /* Fragment the skb. ipv6 header and the remaining fields of the
1369 * fragment header are updated in ipv6_gso_segment()
1370 */
1371 segs = skb_segment(skb, features);
1372
1373out:
1374 return segs;
1375}
1376
41135cc8 1377static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1378 .handler = udpv6_rcv,
1379 .err_handler = udpv6_err,
ba735425
SS
1380 .gso_send_check = udp6_ufo_send_check,
1381 .gso_segment = udp6_ufo_fragment,
1da177e4
LT
1382 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1383};
1384
1385/* ------------------------------------------------------------------------ */
1386#ifdef CONFIG_PROC_FS
1387
1388static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1389{
1390 struct inet_sock *inet = inet_sk(sp);
1391 struct ipv6_pinfo *np = inet6_sk(sp);
b71d1d42 1392 const struct in6_addr *dest, *src;
1da177e4
LT
1393 __u16 destp, srcp;
1394
1395 dest = &np->daddr;
1396 src = &np->rcv_saddr;
c720c7e8
ED
1397 destp = ntohs(inet->inet_dport);
1398 srcp = ntohs(inet->inet_sport);
1da177e4 1399 seq_printf(seq,
f86dcc5a 1400 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1401 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1da177e4
LT
1402 bucket,
1403 src->s6_addr32[0], src->s6_addr32[1],
1404 src->s6_addr32[2], src->s6_addr32[3], srcp,
1405 dest->s6_addr32[0], dest->s6_addr32[1],
1406 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 1407 sp->sk_state,
31e6d363
ED
1408 sk_wmem_alloc_get(sp),
1409 sk_rmem_alloc_get(sp),
1da177e4
LT
1410 0, 0L, 0,
1411 sock_i_uid(sp), 0,
1412 sock_i_ino(sp),
cb61cb9b
ED
1413 atomic_read(&sp->sk_refcnt), sp,
1414 atomic_read(&sp->sk_drops));
1da177e4
LT
1415}
1416
ba4e58ec 1417int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4
LT
1418{
1419 if (v == SEQ_START_TOKEN)
1420 seq_printf(seq,
1421 " sl "
1422 "local_address "
1423 "remote_address "
1424 "st tx_queue rx_queue tr tm->when retrnsmt"
cb61cb9b 1425 " uid timeout inode ref pointer drops\n");
1da177e4
LT
1426 else
1427 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1428 return 0;
1429}
1430
73cb88ec
AV
1431static const struct file_operations udp6_afinfo_seq_fops = {
1432 .owner = THIS_MODULE,
1433 .open = udp_seq_open,
1434 .read = seq_read,
1435 .llseek = seq_lseek,
1436 .release = seq_release_net
1437};
1438
1da177e4 1439static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1440 .name = "udp6",
1441 .family = AF_INET6,
645ca708 1442 .udp_table = &udp_table,
73cb88ec 1443 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1444 .seq_ops = {
1445 .show = udp6_seq_show,
1446 },
1da177e4
LT
1447};
1448
2c8c1e72 1449int __net_init udp6_proc_init(struct net *net)
1da177e4 1450{
0c96d8c5 1451 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1452}
1453
0c96d8c5
DL
1454void udp6_proc_exit(struct net *net) {
1455 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1456}
1457#endif /* CONFIG_PROC_FS */
1458
1459/* ------------------------------------------------------------------------ */
1460
1461struct proto udpv6_prot = {
543d9cfe
ACM
1462 .name = "UDPv6",
1463 .owner = THIS_MODULE,
ba4e58ec 1464 .close = udp_lib_close,
543d9cfe
ACM
1465 .connect = ip6_datagram_connect,
1466 .disconnect = udp_disconnect,
1467 .ioctl = udp_ioctl,
1468 .destroy = udpv6_destroy_sock,
1469 .setsockopt = udpv6_setsockopt,
1470 .getsockopt = udpv6_getsockopt,
1471 .sendmsg = udpv6_sendmsg,
1472 .recvmsg = udpv6_recvmsg,
1473 .backlog_rcv = udpv6_queue_rcv_skb,
ba4e58ec
GR
1474 .hash = udp_lib_hash,
1475 .unhash = udp_lib_unhash,
719f8358 1476 .rehash = udp_v6_rehash,
543d9cfe 1477 .get_port = udp_v6_get_port,
95766fff
HA
1478 .memory_allocated = &udp_memory_allocated,
1479 .sysctl_mem = sysctl_udp_mem,
1480 .sysctl_wmem = &sysctl_udp_wmem_min,
1481 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1482 .obj_size = sizeof(struct udp6_sock),
271b72c7 1483 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1484 .h.udp_table = &udp_table,
3fdadf7d 1485#ifdef CONFIG_COMPAT
543d9cfe
ACM
1486 .compat_setsockopt = compat_udpv6_setsockopt,
1487 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1488#endif
fcbdf09d 1489 .clear_sk = sk_prot_clear_portaddr_nulls,
1da177e4
LT
1490};
1491
1da177e4
LT
1492static struct inet_protosw udpv6_protosw = {
1493 .type = SOCK_DGRAM,
1494 .protocol = IPPROTO_UDP,
1495 .prot = &udpv6_prot,
1496 .ops = &inet6_dgram_ops,
1da177e4
LT
1497 .no_check = UDP_CSUM_DEFAULT,
1498 .flags = INET_PROTOSW_PERMANENT,
1499};
1500
1501
7f4e4868 1502int __init udpv6_init(void)
1da177e4 1503{
7f4e4868
DL
1504 int ret;
1505
1506 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1507 if (ret)
1508 goto out;
1509
1510 ret = inet6_register_protosw(&udpv6_protosw);
1511 if (ret)
1512 goto out_udpv6_protocol;
1513out:
1514 return ret;
1515
1516out_udpv6_protocol:
1517 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1518 goto out;
1519}
1520
09f7709f 1521void udpv6_exit(void)
7f4e4868
DL
1522{
1523 inet6_unregister_protosw(&udpv6_protosw);
1524 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1525}