?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ACPI: make acpi_pci_bind() static
[linux-2.6/linux-acpi-2.6.git] / net / ipv6 / udp.c
blob8905712cfbb81970acabb56c8a6673713e3e1574
1 /*
2 * UDP over IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Based on linux/ipv4/udp.c
9 *
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
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
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>
35 #include <linux/module.h>
36 #include <linux/skbuff.h>
37 #include <asm/uaccess.h>
38
39 #include <net/ndisc.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_route.h>
43 #include <net/raw.h>
44 #include <net/tcp_states.h>
45 #include <net/ip6_checksum.h>
46 #include <net/xfrm.h>
47
48 #include <linux/proc_fs.h>
49 #include <linux/seq_file.h>
50 #include "udp_impl.h"
51
52 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
53 {
54 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
55 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
56 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
57 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
58 int sk_ipv6only = ipv6_only_sock(sk);
59 int sk2_ipv6only = inet_v6_ipv6only(sk2);
60 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
61 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
62
63 /* if both are mapped, treat as IPv4 */
64 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
65 return (!sk2_ipv6only &&
66 (!sk_rcv_saddr || !sk2_rcv_saddr ||
67 sk_rcv_saddr == sk2_rcv_saddr));
68
69 if (addr_type2 == IPV6_ADDR_ANY &&
70 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
71 return 1;
72
73 if (addr_type == IPV6_ADDR_ANY &&
74 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
75 return 1;
76
77 if (sk2_rcv_saddr6 &&
78 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
79 return 1;
80
81 return 0;
82 }
83
84 int udp_v6_get_port(struct sock *sk, unsigned short snum)
85 {
86 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal);
87 }
88
89 static inline int compute_score(struct sock *sk, struct net *net,
90 unsigned short hnum,
91 struct in6_addr *saddr, __be16 sport,
92 struct in6_addr *daddr, __be16 dport,
93 int dif)
94 {
95 int score = -1;
96
97 if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum &&
98 sk->sk_family == PF_INET6) {
99 struct ipv6_pinfo *np = inet6_sk(sk);
100 struct inet_sock *inet = inet_sk(sk);
101
102 score = 0;
103 if (inet->dport) {
104 if (inet->dport != sport)
105 return -1;
106 score++;
107 }
108 if (!ipv6_addr_any(&np->rcv_saddr)) {
109 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
110 return -1;
111 score++;
112 }
113 if (!ipv6_addr_any(&np->daddr)) {
114 if (!ipv6_addr_equal(&np->daddr, saddr))
115 return -1;
116 score++;
117 }
118 if (sk->sk_bound_dev_if) {
119 if (sk->sk_bound_dev_if != dif)
120 return -1;
121 score++;
122 }
123 }
124 return score;
125 }
126
127 static struct sock *__udp6_lib_lookup(struct net *net,
128 struct in6_addr *saddr, __be16 sport,
129 struct in6_addr *daddr, __be16 dport,
130 int dif, struct udp_table *udptable)
131 {
132 struct sock *sk, *result;
133 struct hlist_nulls_node *node;
134 unsigned short hnum = ntohs(dport);
135 unsigned int hash = udp_hashfn(net, hnum);
136 struct udp_hslot *hslot = &udptable->hash[hash];
137 int score, badness;
138
139 rcu_read_lock();
140 begin:
141 result = NULL;
142 badness = -1;
143 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
144 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
145 if (score > badness) {
146 result = sk;
147 badness = score;
148 }
149 }
150 /*
151 * if the nulls value we got at the end of this lookup is
152 * not the expected one, we must restart lookup.
153 * We probably met an item that was moved to another chain.
154 */
155 if (get_nulls_value(node) != hash)
156 goto begin;
157
158 if (result) {
159 if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt)))
160 result = NULL;
161 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
162 daddr, dport, dif) < badness)) {
163 sock_put(result);
164 goto begin;
165 }
166 }
167 rcu_read_unlock();
168 return result;
169 }
170
171 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
172 __be16 sport, __be16 dport,
173 struct udp_table *udptable)
174 {
175 struct sock *sk;
176 struct ipv6hdr *iph = ipv6_hdr(skb);
177
178 if (unlikely(sk = skb_steal_sock(skb)))
179 return sk;
180 else
181 return __udp6_lib_lookup(dev_net(skb->dst->dev), &iph->saddr, sport,
182 &iph->daddr, dport, inet6_iif(skb),
183 udptable);
184 }
185
186 /*
187 * This should be easy, if there is something there we
188 * return it, otherwise we block.
189 */
190
191 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
192 struct msghdr *msg, size_t len,
193 int noblock, int flags, int *addr_len)
194 {
195 struct ipv6_pinfo *np = inet6_sk(sk);
196 struct inet_sock *inet = inet_sk(sk);
197 struct sk_buff *skb;
198 unsigned int ulen, copied;
199 int peeked;
200 int err;
201 int is_udplite = IS_UDPLITE(sk);
202 int is_udp4;
203
204 if (addr_len)
205 *addr_len=sizeof(struct sockaddr_in6);
206
207 if (flags & MSG_ERRQUEUE)
208 return ipv6_recv_error(sk, msg, len);
209
210 try_again:
211 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
212 &peeked, &err);
213 if (!skb)
214 goto out;
215
216 ulen = skb->len - sizeof(struct udphdr);
217 copied = len;
218 if (copied > ulen)
219 copied = ulen;
220 else if (copied < ulen)
221 msg->msg_flags |= MSG_TRUNC;
222
223 is_udp4 = (skb->protocol == htons(ETH_P_IP));
224
225 /*
226 * If checksum is needed at all, try to do it while copying the
227 * data. If the data is truncated, or if we only want a partial
228 * coverage checksum (UDP-Lite), do it before the copy.
229 */
230
231 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
232 if (udp_lib_checksum_complete(skb))
233 goto csum_copy_err;
234 }
235
236 if (skb_csum_unnecessary(skb))
237 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
238 msg->msg_iov, copied );
239 else {
240 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
241 if (err == -EINVAL)
242 goto csum_copy_err;
243 }
244 if (err)
245 goto out_free;
246
247 if (!peeked) {
248 if (is_udp4)
249 UDP_INC_STATS_USER(sock_net(sk),
250 UDP_MIB_INDATAGRAMS, is_udplite);
251 else
252 UDP6_INC_STATS_USER(sock_net(sk),
253 UDP_MIB_INDATAGRAMS, is_udplite);
254 }
255
256 sock_recv_timestamp(msg, sk, skb);
257
258 /* Copy the address. */
259 if (msg->msg_name) {
260 struct sockaddr_in6 *sin6;
261
262 sin6 = (struct sockaddr_in6 *) msg->msg_name;
263 sin6->sin6_family = AF_INET6;
264 sin6->sin6_port = udp_hdr(skb)->source;
265 sin6->sin6_flowinfo = 0;
266 sin6->sin6_scope_id = 0;
267
268 if (is_udp4)
269 ipv6_addr_set(&sin6->sin6_addr, 0, 0,
270 htonl(0xffff), ip_hdr(skb)->saddr);
271 else {
272 ipv6_addr_copy(&sin6->sin6_addr,
273 &ipv6_hdr(skb)->saddr);
274 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
275 sin6->sin6_scope_id = IP6CB(skb)->iif;
276 }
277
278 }
279 if (is_udp4) {
280 if (inet->cmsg_flags)
281 ip_cmsg_recv(msg, skb);
282 } else {
283 if (np->rxopt.all)
284 datagram_recv_ctl(sk, msg, skb);
285 }
286
287 err = copied;
288 if (flags & MSG_TRUNC)
289 err = ulen;
290
291 out_free:
292 lock_sock(sk);
293 skb_free_datagram(sk, skb);
294 release_sock(sk);
295 out:
296 return err;
297
298 csum_copy_err:
299 lock_sock(sk);
300 if (!skb_kill_datagram(sk, skb, flags)) {
301 if (is_udp4)
302 UDP_INC_STATS_USER(sock_net(sk),
303 UDP_MIB_INERRORS, is_udplite);
304 else
305 UDP6_INC_STATS_USER(sock_net(sk),
306 UDP_MIB_INERRORS, is_udplite);
307 }
308 release_sock(sk);
309
310 if (flags & MSG_DONTWAIT)
311 return -EAGAIN;
312 goto try_again;
313 }
314
315 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
316 int type, int code, int offset, __be32 info,
317 struct udp_table *udptable)
318 {
319 struct ipv6_pinfo *np;
320 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
321 struct in6_addr *saddr = &hdr->saddr;
322 struct in6_addr *daddr = &hdr->daddr;
323 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
324 struct sock *sk;
325 int err;
326
327 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
328 saddr, uh->source, inet6_iif(skb), udptable);
329 if (sk == NULL)
330 return;
331
332 np = inet6_sk(sk);
333
334 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
335 goto out;
336
337 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
338 goto out;
339
340 if (np->recverr)
341 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
342
343 sk->sk_err = err;
344 sk->sk_error_report(sk);
345 out:
346 sock_put(sk);
347 }
348
349 static __inline__ void udpv6_err(struct sk_buff *skb,
350 struct inet6_skb_parm *opt, int type,
351 int code, int offset, __be32 info )
352 {
353 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
354 }
355
356 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
357 {
358 struct udp_sock *up = udp_sk(sk);
359 int rc;
360 int is_udplite = IS_UDPLITE(sk);
361
362 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
363 goto drop;
364
365 /*
366 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
367 */
368 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
369
370 if (up->pcrlen == 0) { /* full coverage was set */
371 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
372 " %d while full coverage %d requested\n",
373 UDP_SKB_CB(skb)->cscov, skb->len);
374 goto drop;
375 }
376 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
377 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
378 "too small, need min %d\n",
379 UDP_SKB_CB(skb)->cscov, up->pcrlen);
380 goto drop;
381 }
382 }
383
384 if (sk->sk_filter) {
385 if (udp_lib_checksum_complete(skb))
386 goto drop;
387 }
388
389 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
390 /* Note that an ENOMEM error is charged twice */
391 if (rc == -ENOMEM) {
392 UDP6_INC_STATS_BH(sock_net(sk),
393 UDP_MIB_RCVBUFERRORS, is_udplite);
394 atomic_inc(&sk->sk_drops);
395 }
396 goto drop;
397 }
398
399 return 0;
400 drop:
401 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
402 kfree_skb(skb);
403 return -1;
404 }
405
406 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
407 __be16 loc_port, struct in6_addr *loc_addr,
408 __be16 rmt_port, struct in6_addr *rmt_addr,
409 int dif)
410 {
411 struct hlist_nulls_node *node;
412 struct sock *s = sk;
413 unsigned short num = ntohs(loc_port);
414
415 sk_nulls_for_each_from(s, node) {
416 struct inet_sock *inet = inet_sk(s);
417
418 if (!net_eq(sock_net(s), net))
419 continue;
420
421 if (s->sk_hash == num && s->sk_family == PF_INET6) {
422 struct ipv6_pinfo *np = inet6_sk(s);
423 if (inet->dport) {
424 if (inet->dport != rmt_port)
425 continue;
426 }
427 if (!ipv6_addr_any(&np->daddr) &&
428 !ipv6_addr_equal(&np->daddr, rmt_addr))
429 continue;
430
431 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
432 continue;
433
434 if (!ipv6_addr_any(&np->rcv_saddr)) {
435 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
436 continue;
437 }
438 if (!inet6_mc_check(s, loc_addr, rmt_addr))
439 continue;
440 return s;
441 }
442 }
443 return NULL;
444 }
445
446 /*
447 * Note: called only from the BH handler context,
448 * so we don't need to lock the hashes.
449 */
450 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
451 struct in6_addr *saddr, struct in6_addr *daddr,
452 struct udp_table *udptable)
453 {
454 struct sock *sk, *sk2;
455 const struct udphdr *uh = udp_hdr(skb);
456 struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))];
457 int dif;
458
459 spin_lock(&hslot->lock);
460 sk = sk_nulls_head(&hslot->head);
461 dif = inet6_iif(skb);
462 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
463 if (!sk) {
464 kfree_skb(skb);
465 goto out;
466 }
467
468 sk2 = sk;
469 while ((sk2 = udp_v6_mcast_next(net, sk_nulls_next(sk2), uh->dest, daddr,
470 uh->source, saddr, dif))) {
471 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
472 if (buff) {
473 bh_lock_sock(sk2);
474 if (!sock_owned_by_user(sk2))
475 udpv6_queue_rcv_skb(sk2, buff);
476 else
477 sk_add_backlog(sk2, buff);
478 bh_unlock_sock(sk2);
479 }
480 }
481 bh_lock_sock(sk);
482 if (!sock_owned_by_user(sk))
483 udpv6_queue_rcv_skb(sk, skb);
484 else
485 sk_add_backlog(sk, skb);
486 bh_unlock_sock(sk);
487 out:
488 spin_unlock(&hslot->lock);
489 return 0;
490 }
491
492 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
493 int proto)
494 {
495 int err;
496
497 UDP_SKB_CB(skb)->partial_cov = 0;
498 UDP_SKB_CB(skb)->cscov = skb->len;
499
500 if (proto == IPPROTO_UDPLITE) {
501 err = udplite_checksum_init(skb, uh);
502 if (err)
503 return err;
504 }
505
506 if (uh->check == 0) {
507 /* RFC 2460 section 8.1 says that we SHOULD log
508 this error. Well, it is reasonable.
509 */
510 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
511 return 1;
512 }
513 if (skb->ip_summed == CHECKSUM_COMPLETE &&
514 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
515 skb->len, proto, skb->csum))
516 skb->ip_summed = CHECKSUM_UNNECESSARY;
517
518 if (!skb_csum_unnecessary(skb))
519 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
520 &ipv6_hdr(skb)->daddr,
521 skb->len, proto, 0));
522
523 return 0;
524 }
525
526 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
527 int proto)
528 {
529 struct sock *sk;
530 struct udphdr *uh;
531 struct net_device *dev = skb->dev;
532 struct in6_addr *saddr, *daddr;
533 u32 ulen = 0;
534 struct net *net = dev_net(skb->dev);
535
536 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
537 goto short_packet;
538
539 saddr = &ipv6_hdr(skb)->saddr;
540 daddr = &ipv6_hdr(skb)->daddr;
541 uh = udp_hdr(skb);
542
543 ulen = ntohs(uh->len);
544 if (ulen > skb->len)
545 goto short_packet;
546
547 if (proto == IPPROTO_UDP) {
548 /* UDP validates ulen. */
549
550 /* Check for jumbo payload */
551 if (ulen == 0)
552 ulen = skb->len;
553
554 if (ulen < sizeof(*uh))
555 goto short_packet;
556
557 if (ulen < skb->len) {
558 if (pskb_trim_rcsum(skb, ulen))
559 goto short_packet;
560 saddr = &ipv6_hdr(skb)->saddr;
561 daddr = &ipv6_hdr(skb)->daddr;
562 uh = udp_hdr(skb);
563 }
564 }
565
566 if (udp6_csum_init(skb, uh, proto))
567 goto discard;
568
569 /*
570 * Multicast receive code
571 */
572 if (ipv6_addr_is_multicast(daddr))
573 return __udp6_lib_mcast_deliver(net, skb,
574 saddr, daddr, udptable);
575
576 /* Unicast */
577
578 /*
579 * check socket cache ... must talk to Alan about his plans
580 * for sock caches... i'll skip this for now.
581 */
582 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
583
584 if (sk == NULL) {
585 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
586 goto discard;
587
588 if (udp_lib_checksum_complete(skb))
589 goto discard;
590 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS,
591 proto == IPPROTO_UDPLITE);
592
593 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
594
595 kfree_skb(skb);
596 return 0;
597 }
598
599 /* deliver */
600
601 bh_lock_sock(sk);
602 if (!sock_owned_by_user(sk))
603 udpv6_queue_rcv_skb(sk, skb);
604 else
605 sk_add_backlog(sk, skb);
606 bh_unlock_sock(sk);
607 sock_put(sk);
608 return 0;
609
610 short_packet:
611 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: %d/%u\n",
612 proto == IPPROTO_UDPLITE ? "-Lite" : "",
613 ulen, skb->len);
614
615 discard:
616 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
617 kfree_skb(skb);
618 return 0;
619 }
620
621 static __inline__ int udpv6_rcv(struct sk_buff *skb)
622 {
623 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
624 }
625
626 /*
627 * Throw away all pending data and cancel the corking. Socket is locked.
628 */
629 static void udp_v6_flush_pending_frames(struct sock *sk)
630 {
631 struct udp_sock *up = udp_sk(sk);
632
633 if (up->pending == AF_INET)
634 udp_flush_pending_frames(sk);
635 else if (up->pending) {
636 up->len = 0;
637 up->pending = 0;
638 ip6_flush_pending_frames(sk);
639 }
640 }
641
642 /*
643 * Sending
644 */
645
646 static int udp_v6_push_pending_frames(struct sock *sk)
647 {
648 struct sk_buff *skb;
649 struct udphdr *uh;
650 struct udp_sock *up = udp_sk(sk);
651 struct inet_sock *inet = inet_sk(sk);
652 struct flowi *fl = &inet->cork.fl;
653 int err = 0;
654 int is_udplite = IS_UDPLITE(sk);
655 __wsum csum = 0;
656
657 /* Grab the skbuff where UDP header space exists. */
658 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
659 goto out;
660
661 /*
662 * Create a UDP header
663 */
664 uh = udp_hdr(skb);
665 uh->source = fl->fl_ip_sport;
666 uh->dest = fl->fl_ip_dport;
667 uh->len = htons(up->len);
668 uh->check = 0;
669
670 if (is_udplite)
671 csum = udplite_csum_outgoing(sk, skb);
672 else
673 csum = udp_csum_outgoing(sk, skb);
674
675 /* add protocol-dependent pseudo-header */
676 uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst,
677 up->len, fl->proto, csum );
678 if (uh->check == 0)
679 uh->check = CSUM_MANGLED_0;
680
681 err = ip6_push_pending_frames(sk);
682 out:
683 up->len = 0;
684 up->pending = 0;
685 if (!err)
686 UDP6_INC_STATS_USER(sock_net(sk),
687 UDP_MIB_OUTDATAGRAMS, is_udplite);
688 return err;
689 }
690
691 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
692 struct msghdr *msg, size_t len)
693 {
694 struct ipv6_txoptions opt_space;
695 struct udp_sock *up = udp_sk(sk);
696 struct inet_sock *inet = inet_sk(sk);
697 struct ipv6_pinfo *np = inet6_sk(sk);
698 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
699 struct in6_addr *daddr, *final_p = NULL, final;
700 struct ipv6_txoptions *opt = NULL;
701 struct ip6_flowlabel *flowlabel = NULL;
702 struct flowi fl;
703 struct dst_entry *dst;
704 int addr_len = msg->msg_namelen;
705 int ulen = len;
706 int hlimit = -1;
707 int tclass = -1;
708 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
709 int err;
710 int connected = 0;
711 int is_udplite = IS_UDPLITE(sk);
712 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
713
714 /* destination address check */
715 if (sin6) {
716 if (addr_len < offsetof(struct sockaddr, sa_data))
717 return -EINVAL;
718
719 switch (sin6->sin6_family) {
720 case AF_INET6:
721 if (addr_len < SIN6_LEN_RFC2133)
722 return -EINVAL;
723 daddr = &sin6->sin6_addr;
724 break;
725 case AF_INET:
726 goto do_udp_sendmsg;
727 case AF_UNSPEC:
728 msg->msg_name = sin6 = NULL;
729 msg->msg_namelen = addr_len = 0;
730 daddr = NULL;
731 break;
732 default:
733 return -EINVAL;
734 }
735 } else if (!up->pending) {
736 if (sk->sk_state != TCP_ESTABLISHED)
737 return -EDESTADDRREQ;
738 daddr = &np->daddr;
739 } else
740 daddr = NULL;
741
742 if (daddr) {
743 if (ipv6_addr_v4mapped(daddr)) {
744 struct sockaddr_in sin;
745 sin.sin_family = AF_INET;
746 sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
747 sin.sin_addr.s_addr = daddr->s6_addr32[3];
748 msg->msg_name = &sin;
749 msg->msg_namelen = sizeof(sin);
750 do_udp_sendmsg:
751 if (__ipv6_only_sock(sk))
752 return -ENETUNREACH;
753 return udp_sendmsg(iocb, sk, msg, len);
754 }
755 }
756
757 if (up->pending == AF_INET)
758 return udp_sendmsg(iocb, sk, msg, len);
759
760 /* Rough check on arithmetic overflow,
761 better check is made in ip6_append_data().
762 */
763 if (len > INT_MAX - sizeof(struct udphdr))
764 return -EMSGSIZE;
765
766 if (up->pending) {
767 /*
768 * There are pending frames.
769 * The socket lock must be held while it's corked.
770 */
771 lock_sock(sk);
772 if (likely(up->pending)) {
773 if (unlikely(up->pending != AF_INET6)) {
774 release_sock(sk);
775 return -EAFNOSUPPORT;
776 }
777 dst = NULL;
778 goto do_append_data;
779 }
780 release_sock(sk);
781 }
782 ulen += sizeof(struct udphdr);
783
784 memset(&fl, 0, sizeof(fl));
785
786 if (sin6) {
787 if (sin6->sin6_port == 0)
788 return -EINVAL;
789
790 fl.fl_ip_dport = sin6->sin6_port;
791 daddr = &sin6->sin6_addr;
792
793 if (np->sndflow) {
794 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
795 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
796 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
797 if (flowlabel == NULL)
798 return -EINVAL;
799 daddr = &flowlabel->dst;
800 }
801 }
802
803 /*
804 * Otherwise it will be difficult to maintain
805 * sk->sk_dst_cache.
806 */
807 if (sk->sk_state == TCP_ESTABLISHED &&
808 ipv6_addr_equal(daddr, &np->daddr))
809 daddr = &np->daddr;
810
811 if (addr_len >= sizeof(struct sockaddr_in6) &&
812 sin6->sin6_scope_id &&
813 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
814 fl.oif = sin6->sin6_scope_id;
815 } else {
816 if (sk->sk_state != TCP_ESTABLISHED)
817 return -EDESTADDRREQ;
818
819 fl.fl_ip_dport = inet->dport;
820 daddr = &np->daddr;
821 fl.fl6_flowlabel = np->flow_label;
822 connected = 1;
823 }
824
825 if (!fl.oif)
826 fl.oif = sk->sk_bound_dev_if;
827
828 if (!fl.oif)
829 fl.oif = np->sticky_pktinfo.ipi6_ifindex;
830
831 if (msg->msg_controllen) {
832 opt = &opt_space;
833 memset(opt, 0, sizeof(struct ipv6_txoptions));
834 opt->tot_len = sizeof(*opt);
835
836 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, &tclass);
837 if (err < 0) {
838 fl6_sock_release(flowlabel);
839 return err;
840 }
841 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
842 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
843 if (flowlabel == NULL)
844 return -EINVAL;
845 }
846 if (!(opt->opt_nflen|opt->opt_flen))
847 opt = NULL;
848 connected = 0;
849 }
850 if (opt == NULL)
851 opt = np->opt;
852 if (flowlabel)
853 opt = fl6_merge_options(&opt_space, flowlabel, opt);
854 opt = ipv6_fixup_options(&opt_space, opt);
855
856 fl.proto = sk->sk_protocol;
857 if (!ipv6_addr_any(daddr))
858 ipv6_addr_copy(&fl.fl6_dst, daddr);
859 else
860 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
861 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
862 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
863 fl.fl_ip_sport = inet->sport;
864
865 /* merge ip6_build_xmit from ip6_output */
866 if (opt && opt->srcrt) {
867 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
868 ipv6_addr_copy(&final, &fl.fl6_dst);
869 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
870 final_p = &final;
871 connected = 0;
872 }
873
874 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) {
875 fl.oif = np->mcast_oif;
876 connected = 0;
877 }
878
879 security_sk_classify_flow(sk, &fl);
880
881 err = ip6_sk_dst_lookup(sk, &dst, &fl);
882 if (err)
883 goto out;
884 if (final_p)
885 ipv6_addr_copy(&fl.fl6_dst, final_p);
886
887 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
888 if (err < 0) {
889 if (err == -EREMOTE)
890 err = ip6_dst_blackhole(sk, &dst, &fl);
891 if (err < 0)
892 goto out;
893 }
894
895 if (hlimit < 0) {
896 if (ipv6_addr_is_multicast(&fl.fl6_dst))
897 hlimit = np->mcast_hops;
898 else
899 hlimit = np->hop_limit;
900 if (hlimit < 0)
901 hlimit = ip6_dst_hoplimit(dst);
902 }
903
904 if (tclass < 0) {
905 tclass = np->tclass;
906 if (tclass < 0)
907 tclass = 0;
908 }
909
910 if (msg->msg_flags&MSG_CONFIRM)
911 goto do_confirm;
912 back_from_confirm:
913
914 lock_sock(sk);
915 if (unlikely(up->pending)) {
916 /* The socket is already corked while preparing it. */
917 /* ... which is an evident application bug. --ANK */
918 release_sock(sk);
919
920 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
921 err = -EINVAL;
922 goto out;
923 }
924
925 up->pending = AF_INET6;
926
927 do_append_data:
928 up->len += ulen;
929 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
930 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
931 sizeof(struct udphdr), hlimit, tclass, opt, &fl,
932 (struct rt6_info*)dst,
933 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
934 if (err)
935 udp_v6_flush_pending_frames(sk);
936 else if (!corkreq)
937 err = udp_v6_push_pending_frames(sk);
938 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
939 up->pending = 0;
940
941 if (dst) {
942 if (connected) {
943 ip6_dst_store(sk, dst,
944 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
945 &np->daddr : NULL,
946 #ifdef CONFIG_IPV6_SUBTREES
947 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ?
948 &np->saddr :
949 #endif
950 NULL);
951 } else {
952 dst_release(dst);
953 }
954 dst = NULL;
955 }
956
957 if (err > 0)
958 err = np->recverr ? net_xmit_errno(err) : 0;
959 release_sock(sk);
960 out:
961 dst_release(dst);
962 fl6_sock_release(flowlabel);
963 if (!err)
964 return len;
965 /*
966 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
967 * ENOBUFS might not be good (it's not tunable per se), but otherwise
968 * we don't have a good statistic (IpOutDiscards but it can be too many
969 * things). We could add another new stat but at least for now that
970 * seems like overkill.
971 */
972 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
973 UDP6_INC_STATS_USER(sock_net(sk),
974 UDP_MIB_SNDBUFERRORS, is_udplite);
975 }
976 return err;
977
978 do_confirm:
979 dst_confirm(dst);
980 if (!(msg->msg_flags&MSG_PROBE) || len)
981 goto back_from_confirm;
982 err = 0;
983 goto out;
984 }
985
986 void udpv6_destroy_sock(struct sock *sk)
987 {
988 lock_sock(sk);
989 udp_v6_flush_pending_frames(sk);
990 release_sock(sk);
991
992 inet6_destroy_sock(sk);
993 }
994
995 /*
996 * Socket option code for UDP
997 */
998 int udpv6_setsockopt(struct sock *sk, int level, int optname,
999 char __user *optval, int optlen)
1000 {
1001 if (level == SOL_UDP || level == SOL_UDPLITE)
1002 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1003 udp_v6_push_pending_frames);
1004 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1005 }
1006
1007 #ifdef CONFIG_COMPAT
1008 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1009 char __user *optval, int optlen)
1010 {
1011 if (level == SOL_UDP || level == SOL_UDPLITE)
1012 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1013 udp_v6_push_pending_frames);
1014 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1015 }
1016 #endif
1017
1018 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1019 char __user *optval, int __user *optlen)
1020 {
1021 if (level == SOL_UDP || level == SOL_UDPLITE)
1022 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1023 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1024 }
1025
1026 #ifdef CONFIG_COMPAT
1027 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1028 char __user *optval, int __user *optlen)
1029 {
1030 if (level == SOL_UDP || level == SOL_UDPLITE)
1031 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1032 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1033 }
1034 #endif
1035
1036 static struct inet6_protocol udpv6_protocol = {
1037 .handler = udpv6_rcv,
1038 .err_handler = udpv6_err,
1039 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1040 };
1041
1042 /* ------------------------------------------------------------------------ */
1043 #ifdef CONFIG_PROC_FS
1044
1045 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1046 {
1047 struct inet_sock *inet = inet_sk(sp);
1048 struct ipv6_pinfo *np = inet6_sk(sp);
1049 struct in6_addr *dest, *src;
1050 __u16 destp, srcp;
1051
1052 dest = &np->daddr;
1053 src = &np->rcv_saddr;
1054 destp = ntohs(inet->dport);
1055 srcp = ntohs(inet->sport);
1056 seq_printf(seq,
1057 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1058 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1059 bucket,
1060 src->s6_addr32[0], src->s6_addr32[1],
1061 src->s6_addr32[2], src->s6_addr32[3], srcp,
1062 dest->s6_addr32[0], dest->s6_addr32[1],
1063 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1064 sp->sk_state,
1065 atomic_read(&sp->sk_wmem_alloc),
1066 atomic_read(&sp->sk_rmem_alloc),
1067 0, 0L, 0,
1068 sock_i_uid(sp), 0,
1069 sock_i_ino(sp),
1070 atomic_read(&sp->sk_refcnt), sp,
1071 atomic_read(&sp->sk_drops));
1072 }
1073
1074 int udp6_seq_show(struct seq_file *seq, void *v)
1075 {
1076 if (v == SEQ_START_TOKEN)
1077 seq_printf(seq,
1078 " sl "
1079 "local_address "
1080 "remote_address "
1081 "st tx_queue rx_queue tr tm->when retrnsmt"
1082 " uid timeout inode ref pointer drops\n");
1083 else
1084 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1085 return 0;
1086 }
1087
1088 static struct udp_seq_afinfo udp6_seq_afinfo = {
1089 .name = "udp6",
1090 .family = AF_INET6,
1091 .udp_table = &udp_table,
1092 .seq_fops = {
1093 .owner = THIS_MODULE,
1094 },
1095 .seq_ops = {
1096 .show = udp6_seq_show,
1097 },
1098 };
1099
1100 int udp6_proc_init(struct net *net)
1101 {
1102 return udp_proc_register(net, &udp6_seq_afinfo);
1103 }
1104
1105 void udp6_proc_exit(struct net *net) {
1106 udp_proc_unregister(net, &udp6_seq_afinfo);
1107 }
1108 #endif /* CONFIG_PROC_FS */
1109
1110 /* ------------------------------------------------------------------------ */
1111
1112 struct proto udpv6_prot = {
1113 .name = "UDPv6",
1114 .owner = THIS_MODULE,
1115 .close = udp_lib_close,
1116 .connect = ip6_datagram_connect,
1117 .disconnect = udp_disconnect,
1118 .ioctl = udp_ioctl,
1119 .destroy = udpv6_destroy_sock,
1120 .setsockopt = udpv6_setsockopt,
1121 .getsockopt = udpv6_getsockopt,
1122 .sendmsg = udpv6_sendmsg,
1123 .recvmsg = udpv6_recvmsg,
1124 .backlog_rcv = udpv6_queue_rcv_skb,
1125 .hash = udp_lib_hash,
1126 .unhash = udp_lib_unhash,
1127 .get_port = udp_v6_get_port,
1128 .memory_allocated = &udp_memory_allocated,
1129 .sysctl_mem = sysctl_udp_mem,
1130 .sysctl_wmem = &sysctl_udp_wmem_min,
1131 .sysctl_rmem = &sysctl_udp_rmem_min,
1132 .obj_size = sizeof(struct udp6_sock),
1133 .slab_flags = SLAB_DESTROY_BY_RCU,
1134 .h.udp_table = &udp_table,
1135 #ifdef CONFIG_COMPAT
1136 .compat_setsockopt = compat_udpv6_setsockopt,
1137 .compat_getsockopt = compat_udpv6_getsockopt,
1138 #endif
1139 };
1140
1141 static struct inet_protosw udpv6_protosw = {
1142 .type = SOCK_DGRAM,
1143 .protocol = IPPROTO_UDP,
1144 .prot = &udpv6_prot,
1145 .ops = &inet6_dgram_ops,
1146 .capability =-1,
1147 .no_check = UDP_CSUM_DEFAULT,
1148 .flags = INET_PROTOSW_PERMANENT,
1149 };
1150
1151
1152 int __init udpv6_init(void)
1153 {
1154 int ret;
1155
1156 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1157 if (ret)
1158 goto out;
1159
1160 ret = inet6_register_protosw(&udpv6_protosw);
1161 if (ret)
1162 goto out_udpv6_protocol;
1163 out:
1164 return ret;
1165
1166 out_udpv6_protocol:
1167 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1168 goto out;
1169 }
1170
1171 void udpv6_exit(void)
1172 {
1173 inet6_unregister_protosw(&udpv6_protosw);
1174 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1175 }
Linux ACPI git tree (mirror)