]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv6/udp.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[IPV6]: Add UDP_MIB_{SND,RCV}BUFERRORS handling.
[net-next-2.6.git] / net / ipv6 / udp.c
CommitLineData
1da177e4
LT		 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 * $Id: udp.c,v 1.65 2002/02/01 22:01:04 davem Exp $
11 *
12 * Fixes:
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
15 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
16 * a single port at the same time.
17 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
18 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
19 *
20 * This program is free software; you can redistribute it and/or
21 * modify it under the terms of the GNU General Public License
22 * as published by the Free Software Foundation; either version
23 * 2 of the License, or (at your option) any later version.
24 */
25
1da177e4
LT		 26#include <linux/errno.h>
27#include <linux/types.h>
28#include <linux/socket.h>
29#include <linux/sockios.h>
30#include <linux/sched.h>
31#include <linux/net.h>
32#include <linux/in6.h>
33#include <linux/netdevice.h>
34#include <linux/if_arp.h>
35#include <linux/ipv6.h>
36#include <linux/icmpv6.h>
37#include <linux/init.h>
3305b80c 38#include <linux/skbuff.h>
1da177e4
LT		 39#include <asm/uaccess.h>
40
41#include <net/sock.h>
42#include <net/snmp.h>
43
44#include <net/ipv6.h>
45#include <net/ndisc.h>
46#include <net/protocol.h>
47#include <net/transp_v6.h>
48#include <net/ip6_route.h>
49#include <net/addrconf.h>
50#include <net/ip.h>
51#include <net/udp.h>
52#include <net/raw.h>
53#include <net/inet_common.h>
c752f073 54#include <net/tcp_states.h>
1da177e4
LT		 55
56#include <net/ip6_checksum.h>
57#include <net/xfrm.h>
58
59#include <linux/proc_fs.h>
60#include <linux/seq_file.h>
61
ba89966c 62DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6) __read_mostly;
1da177e4
LT		 63
64/* Grrr, addr_type already calculated by caller, but I don't want
65 * to add some silly "cookie" argument to this method just for that.
66 */
67static int udp_v6_get_port(struct sock *sk, unsigned short snum)
68{
69 struct sock *sk2;
70 struct hlist_node *node;
71
72 write_lock_bh(&udp_hash_lock);
73 if (snum == 0) {
74 int best_size_so_far, best, result, i;
75
76 if (udp_port_rover > sysctl_local_port_range[1] ||
77 udp_port_rover < sysctl_local_port_range[0])
78 udp_port_rover = sysctl_local_port_range[0];
79 best_size_so_far = 32767;
80 best = result = udp_port_rover;
81 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
82 int size;
83 struct hlist_head *list;
84
85 list = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
86 if (hlist_empty(list)) {
87 if (result > sysctl_local_port_range[1])
88 result = sysctl_local_port_range[0] +
89 ((result - sysctl_local_port_range[0]) &
90 (UDP_HTABLE_SIZE - 1));
91 goto gotit;
92 }
93 size = 0;
94 sk_for_each(sk2, node, list)
95 if (++size >= best_size_so_far)
96 goto next;
97 best_size_so_far = size;
98 best = result;
99 next:;
100 }
101 result = best;
87bf9c97 102 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) {
1da177e4
LT		 103 if (result > sysctl_local_port_range[1])
104 result = sysctl_local_port_range[0]
105 + ((result - sysctl_local_port_range[0]) &
106 (UDP_HTABLE_SIZE - 1));
107 if (!udp_lport_inuse(result))
108 break;
109 }
87bf9c97
YH		 110 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
111 goto fail;
1da177e4
LT		 112gotit:
113 udp_port_rover = snum = result;
114 } else {
115 sk_for_each(sk2, node,
116 &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) {
117 if (inet_sk(sk2)->num == snum &&
118 sk2 != sk &&
119 (!sk2->sk_bound_dev_if ||
120 !sk->sk_bound_dev_if ||
121 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
122 (!sk2->sk_reuse || !sk->sk_reuse) &&
123 ipv6_rcv_saddr_equal(sk, sk2))
124 goto fail;
125 }
126 }
127
128 inet_sk(sk)->num = snum;
129 if (sk_unhashed(sk)) {
130 sk_add_node(sk, &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]);
131 sock_prot_inc_use(sk->sk_prot);
132 }
133 write_unlock_bh(&udp_hash_lock);
134 return 0;
135
136fail:
137 write_unlock_bh(&udp_hash_lock);
138 return 1;
139}
140
141static void udp_v6_hash(struct sock *sk)
142{
143 BUG();
144}
145
146static void udp_v6_unhash(struct sock *sk)
147{
148 write_lock_bh(&udp_hash_lock);
149 if (sk_del_node_init(sk)) {
150 inet_sk(sk)->num = 0;
151 sock_prot_dec_use(sk->sk_prot);
152 }
153 write_unlock_bh(&udp_hash_lock);
154}
155
156static struct sock *udp_v6_lookup(struct in6_addr *saddr, u16 sport,
157 struct in6_addr *daddr, u16 dport, int dif)
158{
159 struct sock *sk, *result = NULL;
160 struct hlist_node *node;
161 unsigned short hnum = ntohs(dport);
162 int badness = -1;
163
164 read_lock(&udp_hash_lock);
165 sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
166 struct inet_sock *inet = inet_sk(sk);
167
168 if (inet->num == hnum && sk->sk_family == PF_INET6) {
169 struct ipv6_pinfo *np = inet6_sk(sk);
170 int score = 0;
171 if (inet->dport) {
172 if (inet->dport != sport)
173 continue;
174 score++;
175 }
176 if (!ipv6_addr_any(&np->rcv_saddr)) {
177 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
178 continue;
179 score++;
180 }
181 if (!ipv6_addr_any(&np->daddr)) {
182 if (!ipv6_addr_equal(&np->daddr, saddr))
183 continue;
184 score++;
185 }
186 if (sk->sk_bound_dev_if) {
187 if (sk->sk_bound_dev_if != dif)
188 continue;
189 score++;
190 }
191 if(score == 4) {
192 result = sk;
193 break;
194 } else if(score > badness) {
195 result = sk;
196 badness = score;
197 }
198 }
199 }
200 if (result)
201 sock_hold(result);
202 read_unlock(&udp_hash_lock);
203 return result;
204}
205
206/*
207 *
208 */
209
210static void udpv6_close(struct sock *sk, long timeout)
211{
212 sk_common_release(sk);
213}
214
215/*
216 * This should be easy, if there is something there we
217 * return it, otherwise we block.
218 */
219
220static int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
221 struct msghdr *msg, size_t len,
222 int noblock, int flags, int *addr_len)
223{
224 struct ipv6_pinfo *np = inet6_sk(sk);
225 struct inet_sock *inet = inet_sk(sk);
226 struct sk_buff *skb;
227 size_t copied;
228 int err;
229
230 if (addr_len)
231 *addr_len=sizeof(struct sockaddr_in6);
232
233 if (flags & MSG_ERRQUEUE)
234 return ipv6_recv_error(sk, msg, len);
235
236try_again:
237 skb = skb_recv_datagram(sk, flags, noblock, &err);
238 if (!skb)
239 goto out;
240
241 copied = skb->len - sizeof(struct udphdr);
242 if (copied > len) {
243 copied = len;
244 msg->msg_flags |= MSG_TRUNC;
245 }
246
247 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
248 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
249 copied);
250 } else if (msg->msg_flags&MSG_TRUNC) {
fb286bb2 251 if (__skb_checksum_complete(skb))
1da177e4
LT		 252 goto csum_copy_err;
253 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
254 copied);
255 } else {
256 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
257 if (err == -EINVAL)
258 goto csum_copy_err;
259 }
260 if (err)
261 goto out_free;
262
263 sock_recv_timestamp(msg, sk, skb);
264
265 /* Copy the address. */
266 if (msg->msg_name) {
267 struct sockaddr_in6 *sin6;
268
269 sin6 = (struct sockaddr_in6 *) msg->msg_name;
270 sin6->sin6_family = AF_INET6;
271 sin6->sin6_port = skb->h.uh->source;
272 sin6->sin6_flowinfo = 0;
273 sin6->sin6_scope_id = 0;
274
275 if (skb->protocol == htons(ETH_P_IP))
276 ipv6_addr_set(&sin6->sin6_addr, 0, 0,
277 htonl(0xffff), skb->nh.iph->saddr);
278 else {
279 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
280 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
281 sin6->sin6_scope_id = IP6CB(skb)->iif;
282 }
283
284 }
285 if (skb->protocol == htons(ETH_P_IP)) {
286 if (inet->cmsg_flags)
287 ip_cmsg_recv(msg, skb);
288 } else {
289 if (np->rxopt.all)
290 datagram_recv_ctl(sk, msg, skb);
291 }
292
293 err = copied;
294 if (flags & MSG_TRUNC)
295 err = skb->len - sizeof(struct udphdr);
296
297out_free:
298 skb_free_datagram(sk, skb);
299out:
300 return err;
301
302csum_copy_err:
3305b80c 303 skb_kill_datagram(sk, skb, flags);
1da177e4
LT		 304
305 if (flags & MSG_DONTWAIT) {
306 UDP6_INC_STATS_USER(UDP_MIB_INERRORS);
307 return -EAGAIN;
308 }
309 goto try_again;
310}
311
312static void udpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
313 int type, int code, int offset, __u32 info)
314{
315 struct ipv6_pinfo *np;
316 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
317 struct net_device *dev = skb->dev;
318 struct in6_addr *saddr = &hdr->saddr;
319 struct in6_addr *daddr = &hdr->daddr;
320 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
321 struct sock *sk;
322 int err;
323
324 sk = udp_v6_lookup(daddr, uh->dest, saddr, uh->source, dev->ifindex);
325
326 if (sk == NULL)
327 return;
328
329 np = inet6_sk(sk);
330
331 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
332 goto out;
333
334 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
335 goto out;
336
337 if (np->recverr)
338 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
339
340 sk->sk_err = err;
341 sk->sk_error_report(sk);
342out:
343 sock_put(sk);
344}
345
346static inline int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
347{
a18135eb
DM		 348 int rc;
349
1da177e4
LT		 350 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
351 kfree_skb(skb);
352 return -1;
353 }
354
fb286bb2
HX		 355 if (skb_checksum_complete(skb)) {
356 UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
357 kfree_skb(skb);
358 return 0;
1da177e4
LT		 359 }
360
a18135eb
DM		 361 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
362 /* Note that an ENOMEM error is charged twice */
363 if (rc == -ENOMEM)
364 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS);
1da177e4
LT		 365 UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
366 kfree_skb(skb);
367 return 0;
368 }
369 UDP6_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
370 return 0;
371}
372
373static struct sock *udp_v6_mcast_next(struct sock *sk,
374 u16 loc_port, struct in6_addr *loc_addr,
375 u16 rmt_port, struct in6_addr *rmt_addr,
376 int dif)
377{
378 struct hlist_node *node;
379 struct sock *s = sk;
380 unsigned short num = ntohs(loc_port);
381
382 sk_for_each_from(s, node) {
383 struct inet_sock *inet = inet_sk(s);
384
385 if (inet->num == num && s->sk_family == PF_INET6) {
386 struct ipv6_pinfo *np = inet6_sk(s);
387 if (inet->dport) {
388 if (inet->dport != rmt_port)
389 continue;
390 }
391 if (!ipv6_addr_any(&np->daddr) &&
392 !ipv6_addr_equal(&np->daddr, rmt_addr))
393 continue;
394
395 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
396 continue;
397
398 if (!ipv6_addr_any(&np->rcv_saddr)) {
40796c5e
DS		 399 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
400 continue;
1da177e4
LT		 401 }
402 if(!inet6_mc_check(s, loc_addr, rmt_addr))
403 continue;
404 return s;
405 }
406 }
407 return NULL;
408}
409
410/*
411 * Note: called only from the BH handler context,
412 * so we don't need to lock the hashes.
413 */
414static void udpv6_mcast_deliver(struct udphdr *uh,
415 struct in6_addr *saddr, struct in6_addr *daddr,
416 struct sk_buff *skb)
417{
418 struct sock *sk, *sk2;
419 int dif;
420
421 read_lock(&udp_hash_lock);
422 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
423 dif = skb->dev->ifindex;
424 sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
425 if (!sk) {
426 kfree_skb(skb);
427 goto out;
428 }
429
430 sk2 = sk;
431 while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
432 uh->source, saddr, dif))) {
433 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
434 if (buff)
435 udpv6_queue_rcv_skb(sk2, buff);
436 }
437 udpv6_queue_rcv_skb(sk, skb);
438out:
439 read_unlock(&udp_hash_lock);
440}
441
951dbc8a 442static int udpv6_rcv(struct sk_buff **pskb)
1da177e4
LT		 443{
444 struct sk_buff *skb = *pskb;
445 struct sock *sk;
446 struct udphdr *uh;
447 struct net_device *dev = skb->dev;
448 struct in6_addr *saddr, *daddr;
449 u32 ulen = 0;
450
451 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
452 goto short_packet;
453
454 saddr = &skb->nh.ipv6h->saddr;
455 daddr = &skb->nh.ipv6h->daddr;
456 uh = skb->h.uh;
457
458 ulen = ntohs(uh->len);
459
460 /* Check for jumbo payload */
461 if (ulen == 0)
462 ulen = skb->len;
463
464 if (ulen > skb->len || ulen < sizeof(*uh))
465 goto short_packet;
466
467 if (uh->check == 0) {
468 /* RFC 2460 section 8.1 says that we SHOULD log
469 this error. Well, it is reasonable.
470 */
64ce2073 471 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
1da177e4
LT		 472 goto discard;
473 }
474
475 if (ulen < skb->len) {
42ca89c1 476 if (pskb_trim_rcsum(skb, ulen))
1da177e4
LT		 477 goto discard;
478 saddr = &skb->nh.ipv6h->saddr;
479 daddr = &skb->nh.ipv6h->daddr;
480 uh = skb->h.uh;
481 }
482
84fa7933 483 if (skb->ip_summed == CHECKSUM_COMPLETE &&
fb286bb2 484 !csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, skb->csum))
1da177e4 485 skb->ip_summed = CHECKSUM_UNNECESSARY;
fb286bb2 486
1da177e4
LT		 487 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
488 skb->csum = ~csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, 0);
489
490 /*
491 * Multicast receive code
492 */
493 if (ipv6_addr_is_multicast(daddr)) {
494 udpv6_mcast_deliver(uh, saddr, daddr, skb);
495 return 0;
496 }
497
498 /* Unicast */
499
500 /*
501 * check socket cache ... must talk to Alan about his plans
502 * for sock caches... i'll skip this for now.
503 */
504 sk = udp_v6_lookup(saddr, uh->source, daddr, uh->dest, dev->ifindex);
505
506 if (sk == NULL) {
507 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
508 goto discard;
509
fb286bb2 510 if (skb_checksum_complete(skb))
1da177e4
LT		 511 goto discard;
512 UDP6_INC_STATS_BH(UDP_MIB_NOPORTS);
513
514 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
515
516 kfree_skb(skb);
517 return(0);
518 }
519
520 /* deliver */
521
522 udpv6_queue_rcv_skb(sk, skb);
523 sock_put(sk);
524 return(0);
525
526short_packet:
527 if (net_ratelimit())
528 printk(KERN_DEBUG "UDP: short packet: %d/%u\n", ulen, skb->len);
529
530discard:
531 UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
532 kfree_skb(skb);
533 return(0);
534}
535/*
536 * Throw away all pending data and cancel the corking. Socket is locked.
537 */
538static void udp_v6_flush_pending_frames(struct sock *sk)
539{
540 struct udp_sock *up = udp_sk(sk);
541
542 if (up->pending) {
543 up->len = 0;
544 up->pending = 0;
545 ip6_flush_pending_frames(sk);
546 }
547}
548
549/*
550 * Sending
551 */
552
553static int udp_v6_push_pending_frames(struct sock *sk, struct udp_sock *up)
554{
555 struct sk_buff *skb;
556 struct udphdr *uh;
557 struct inet_sock *inet = inet_sk(sk);
558 struct flowi *fl = &inet->cork.fl;
559 int err = 0;
560
561 /* Grab the skbuff where UDP header space exists. */
562 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
563 goto out;
564
565 /*
566 * Create a UDP header
567 */
568 uh = skb->h.uh;
569 uh->source = fl->fl_ip_sport;
570 uh->dest = fl->fl_ip_dport;
571 uh->len = htons(up->len);
572 uh->check = 0;
573
574 if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
575 skb->ip_summed = CHECKSUM_NONE;
576 goto send;
577 }
578
579 if (skb_queue_len(&sk->sk_write_queue) == 1) {
580 skb->csum = csum_partial((char *)uh,
581 sizeof(struct udphdr), skb->csum);
582 uh->check = csum_ipv6_magic(&fl->fl6_src,
583 &fl->fl6_dst,
584 up->len, fl->proto, skb->csum);
585 } else {
586 u32 tmp_csum = 0;
587
588 skb_queue_walk(&sk->sk_write_queue, skb) {
589 tmp_csum = csum_add(tmp_csum, skb->csum);
590 }
591 tmp_csum = csum_partial((char *)uh,
592 sizeof(struct udphdr), tmp_csum);
593 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
594 &fl->fl6_dst,
595 up->len, fl->proto, tmp_csum);
596 uh->check = tmp_csum;
597
598 }
599 if (uh->check == 0)
600 uh->check = -1;
601
602send:
603 err = ip6_push_pending_frames(sk);
604out:
605 up->len = 0;
606 up->pending = 0;
607 return err;
608}
609
610static int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
611 struct msghdr *msg, size_t len)
612{
613 struct ipv6_txoptions opt_space;
614 struct udp_sock *up = udp_sk(sk);
615 struct inet_sock *inet = inet_sk(sk);
616 struct ipv6_pinfo *np = inet6_sk(sk);
617 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
618 struct in6_addr *daddr, *final_p = NULL, final;
619 struct ipv6_txoptions *opt = NULL;
620 struct ip6_flowlabel *flowlabel = NULL;
621 struct flowi *fl = &inet->cork.fl;
622 struct dst_entry *dst;
623 int addr_len = msg->msg_namelen;
624 int ulen = len;
625 int hlimit = -1;
41a1f8ea 626 int tclass = -1;
1da177e4
LT		 627 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
628 int err;
987905de 629 int connected = 0;
1da177e4
LT		 630
631 /* destination address check */
632 if (sin6) {
633 if (addr_len < offsetof(struct sockaddr, sa_data))
634 return -EINVAL;
635
636 switch (sin6->sin6_family) {
637 case AF_INET6:
638 if (addr_len < SIN6_LEN_RFC2133)
639 return -EINVAL;
640 daddr = &sin6->sin6_addr;
641 break;
642 case AF_INET:
643 goto do_udp_sendmsg;
644 case AF_UNSPEC:
645 msg->msg_name = sin6 = NULL;
646 msg->msg_namelen = addr_len = 0;
647 daddr = NULL;
648 break;
649 default:
650 return -EINVAL;
651 }
652 } else if (!up->pending) {
653 if (sk->sk_state != TCP_ESTABLISHED)
654 return -EDESTADDRREQ;
655 daddr = &np->daddr;
656 } else
657 daddr = NULL;
658
659 if (daddr) {
660 if (ipv6_addr_type(daddr) == IPV6_ADDR_MAPPED) {
661 struct sockaddr_in sin;
662 sin.sin_family = AF_INET;
663 sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
664 sin.sin_addr.s_addr = daddr->s6_addr32[3];
665 msg->msg_name = &sin;
666 msg->msg_namelen = sizeof(sin);
667do_udp_sendmsg:
668 if (__ipv6_only_sock(sk))
669 return -ENETUNREACH;
670 return udp_sendmsg(iocb, sk, msg, len);
671 }
672 }
673
674 if (up->pending == AF_INET)
675 return udp_sendmsg(iocb, sk, msg, len);
676
677 /* Rough check on arithmetic overflow,
678 better check is made in ip6_build_xmit
679 */
680 if (len > INT_MAX - sizeof(struct udphdr))
681 return -EMSGSIZE;
682
683 if (up->pending) {
684 /*
685 * There are pending frames.
686 * The socket lock must be held while it's corked.
687 */
688 lock_sock(sk);
689 if (likely(up->pending)) {
690 if (unlikely(up->pending != AF_INET6)) {
691 release_sock(sk);
692 return -EAFNOSUPPORT;
693 }
694 dst = NULL;
695 goto do_append_data;
696 }
697 release_sock(sk);
698 }
699 ulen += sizeof(struct udphdr);
700
701 memset(fl, 0, sizeof(*fl));
702
703 if (sin6) {
704 if (sin6->sin6_port == 0)
705 return -EINVAL;
706
707 fl->fl_ip_dport = sin6->sin6_port;
708 daddr = &sin6->sin6_addr;
709
710 if (np->sndflow) {
711 fl->fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
712 if (fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
713 flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
714 if (flowlabel == NULL)
715 return -EINVAL;
716 daddr = &flowlabel->dst;
717 }
718 }
719
720 /*
721 * Otherwise it will be difficult to maintain
722 * sk->sk_dst_cache.
723 */
724 if (sk->sk_state == TCP_ESTABLISHED &&
725 ipv6_addr_equal(daddr, &np->daddr))
726 daddr = &np->daddr;
727
728 if (addr_len >= sizeof(struct sockaddr_in6) &&
729 sin6->sin6_scope_id &&
730 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
731 fl->oif = sin6->sin6_scope_id;
732 } else {
733 if (sk->sk_state != TCP_ESTABLISHED)
734 return -EDESTADDRREQ;
735
736 fl->fl_ip_dport = inet->dport;
737 daddr = &np->daddr;
738 fl->fl6_flowlabel = np->flow_label;
987905de 739 connected = 1;
1da177e4
LT		 740 }
741
742 if (!fl->oif)
743 fl->oif = sk->sk_bound_dev_if;
744
745 if (msg->msg_controllen) {
746 opt = &opt_space;
747 memset(opt, 0, sizeof(struct ipv6_txoptions));
748 opt->tot_len = sizeof(*opt);
749
41a1f8ea 750 err = datagram_send_ctl(msg, fl, opt, &hlimit, &tclass);
1da177e4
LT		 751 if (err < 0) {
752 fl6_sock_release(flowlabel);
753 return err;
754 }
755 if ((fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
756 flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
757 if (flowlabel == NULL)
758 return -EINVAL;
759 }
760 if (!(opt->opt_nflen|opt->opt_flen))
761 opt = NULL;
987905de 762 connected = 0;
1da177e4
LT		 763 }
764 if (opt == NULL)
765 opt = np->opt;
df9890c3
YH		 766 if (flowlabel)
767 opt = fl6_merge_options(&opt_space, flowlabel, opt);
768 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4
LT		 769
770 fl->proto = IPPROTO_UDP;
771 ipv6_addr_copy(&fl->fl6_dst, daddr);
772 if (ipv6_addr_any(&fl->fl6_src) && !ipv6_addr_any(&np->saddr))
773 ipv6_addr_copy(&fl->fl6_src, &np->saddr);
774 fl->fl_ip_sport = inet->sport;
775
776 /* merge ip6_build_xmit from ip6_output */
777 if (opt && opt->srcrt) {
778 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
779 ipv6_addr_copy(&final, &fl->fl6_dst);
780 ipv6_addr_copy(&fl->fl6_dst, rt0->addr);
781 final_p = &final;
987905de 782 connected = 0;
1da177e4
LT		 783 }
784
987905de 785 if (!fl->oif && ipv6_addr_is_multicast(&fl->fl6_dst)) {
1da177e4 786 fl->oif = np->mcast_oif;
987905de
MK		 787 connected = 0;
788 }
1da177e4 789
beb8d13b
VY		 790 security_sk_classify_flow(sk, fl);
791
497c615a 792 err = ip6_sk_dst_lookup(sk, &dst, fl);
1da177e4
LT		 793 if (err)
794 goto out;
795 if (final_p)
796 ipv6_addr_copy(&fl->fl6_dst, final_p);
797
e104411b 798 if ((err = xfrm_lookup(&dst, fl, sk, 0)) < 0)
1da177e4 799 goto out;
1da177e4
LT		 800
801 if (hlimit < 0) {
802 if (ipv6_addr_is_multicast(&fl->fl6_dst))
803 hlimit = np->mcast_hops;
804 else
805 hlimit = np->hop_limit;
806 if (hlimit < 0)
807 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
808 if (hlimit < 0)
809 hlimit = ipv6_get_hoplimit(dst->dev);
810 }
811
41a1f8ea
YH		 812 if (tclass < 0) {
813 tclass = np->tclass;
814 if (tclass < 0)
815 tclass = 0;
816 }
817
1da177e4
LT		 818 if (msg->msg_flags&MSG_CONFIRM)
819 goto do_confirm;
820back_from_confirm:
821
822 lock_sock(sk);
823 if (unlikely(up->pending)) {
824 /* The socket is already corked while preparing it. */
825 /* ... which is an evident application bug. --ANK */
826 release_sock(sk);
827
64ce2073 828 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1da177e4
LT		 829 err = -EINVAL;
830 goto out;
831 }
832
833 up->pending = AF_INET6;
834
835do_append_data:
836 up->len += ulen;
41a1f8ea
YH		 837 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen,
838 sizeof(struct udphdr), hlimit, tclass, opt, fl,
839 (struct rt6_info*)dst,
840 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
1da177e4
LT		 841 if (err)
842 udp_v6_flush_pending_frames(sk);
843 else if (!corkreq)
844 err = udp_v6_push_pending_frames(sk, up);
845
a5e7c210
DM		 846 if (dst) {
847 if (connected) {
848 ip6_dst_store(sk, dst,
849 ipv6_addr_equal(&fl->fl6_dst, &np->daddr) ?
850 &np->daddr : NULL);
851 } else {
852 dst_release(dst);
853 }
854 }
855
1da177e4
LT		 856 if (err > 0)
857 err = np->recverr ? net_xmit_errno(err) : 0;
858 release_sock(sk);
859out:
860 fl6_sock_release(flowlabel);
861 if (!err) {
862 UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS);
863 return len;
864 }
a18135eb
DM		 865 /*
866 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
867 * ENOBUFS might not be good (it's not tunable per se), but otherwise
868 * we don't have a good statistic (IpOutDiscards but it can be too many
869 * things). We could add another new stat but at least for now that
870 * seems like overkill.
871 */
872 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
873 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS);
874 }
1da177e4
LT		 875 return err;
876
877do_confirm:
878 dst_confirm(dst);
879 if (!(msg->msg_flags&MSG_PROBE) || len)
880 goto back_from_confirm;
881 err = 0;
882 goto out;
883}
884
885static int udpv6_destroy_sock(struct sock *sk)
886{
887 lock_sock(sk);
888 udp_v6_flush_pending_frames(sk);
889 release_sock(sk);
890
891 inet6_destroy_sock(sk);
892
893 return 0;
894}
895
896/*
897 * Socket option code for UDP
898 */
3fdadf7d 899static int do_udpv6_setsockopt(struct sock *sk, int level, int optname,
1da177e4
LT		 900 char __user *optval, int optlen)
901{
902 struct udp_sock *up = udp_sk(sk);
903 int val;
904 int err = 0;
905
1da177e4
LT		 906 if(optlen<sizeof(int))
907 return -EINVAL;
908
909 if (get_user(val, (int __user *)optval))
910 return -EFAULT;
911
912 switch(optname) {
913 case UDP_CORK:
914 if (val != 0) {
915 up->corkflag = 1;
916 } else {
917 up->corkflag = 0;
918 lock_sock(sk);
919 udp_v6_push_pending_frames(sk, up);
920 release_sock(sk);
921 }
922 break;
923
924 case UDP_ENCAP:
925 switch (val) {
926 case 0:
927 up->encap_type = val;
928 break;
929 default:
930 err = -ENOPROTOOPT;
931 break;
932 }
933 break;
934
935 default:
936 err = -ENOPROTOOPT;
937 break;
938 };
939
940 return err;
941}
942
3fdadf7d
DM		 943static int udpv6_setsockopt(struct sock *sk, int level, int optname,
944 char __user *optval, int optlen)
945{
946 if (level != SOL_UDP)
947 return ipv6_setsockopt(sk, level, optname, optval, optlen);
948 return do_udpv6_setsockopt(sk, level, optname, optval, optlen);
949}
950
951#ifdef CONFIG_COMPAT
952static int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
543d9cfe 953 char __user *optval, int optlen)
3fdadf7d
DM		 954{
955 if (level != SOL_UDP)
543d9cfe
ACM		 956 return compat_ipv6_setsockopt(sk, level, optname,
957 optval, optlen);
3fdadf7d
DM		 958 return do_udpv6_setsockopt(sk, level, optname, optval, optlen);
959}
960#endif
961
962static int do_udpv6_getsockopt(struct sock *sk, int level, int optname,
1da177e4
LT		 963 char __user *optval, int __user *optlen)
964{
965 struct udp_sock *up = udp_sk(sk);
966 int val, len;
967
1da177e4
LT		 968 if(get_user(len,optlen))
969 return -EFAULT;
970
971 len = min_t(unsigned int, len, sizeof(int));
972
973 if(len < 0)
974 return -EINVAL;
975
976 switch(optname) {
977 case UDP_CORK:
978 val = up->corkflag;
979 break;
980
981 case UDP_ENCAP:
982 val = up->encap_type;
983 break;
984
985 default:
986 return -ENOPROTOOPT;
987 };
988
989 if(put_user(len, optlen))
990 return -EFAULT;
991 if(copy_to_user(optval, &val,len))
992 return -EFAULT;
993 return 0;
994}
995
3fdadf7d
DM		 996static int udpv6_getsockopt(struct sock *sk, int level, int optname,
997 char __user *optval, int __user *optlen)
998{
999 if (level != SOL_UDP)
1000 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1001 return do_udpv6_getsockopt(sk, level, optname, optval, optlen);
1002}
1003
1004#ifdef CONFIG_COMPAT
1005static int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
543d9cfe 1006 char __user *optval, int __user *optlen)
3fdadf7d
DM		 1007{
1008 if (level != SOL_UDP)
543d9cfe
ACM		 1009 return compat_ipv6_getsockopt(sk, level, optname,
1010 optval, optlen);
3fdadf7d
DM		 1011 return do_udpv6_getsockopt(sk, level, optname, optval, optlen);
1012}
1013#endif
1014
1da177e4
LT		 1015static struct inet6_protocol udpv6_protocol = {
1016 .handler = udpv6_rcv,
1017 .err_handler = udpv6_err,
1018 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1019};
1020
1021/* ------------------------------------------------------------------------ */
1022#ifdef CONFIG_PROC_FS
1023
1024static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1025{
1026 struct inet_sock *inet = inet_sk(sp);
1027 struct ipv6_pinfo *np = inet6_sk(sp);
1028 struct in6_addr *dest, *src;
1029 __u16 destp, srcp;
1030
1031 dest = &np->daddr;
1032 src = &np->rcv_saddr;
1033 destp = ntohs(inet->dport);
1034 srcp = ntohs(inet->sport);
1035 seq_printf(seq,
1036 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1037 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1038 bucket,
1039 src->s6_addr32[0], src->s6_addr32[1],
1040 src->s6_addr32[2], src->s6_addr32[3], srcp,
1041 dest->s6_addr32[0], dest->s6_addr32[1],
1042 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1043 sp->sk_state,
1044 atomic_read(&sp->sk_wmem_alloc),
1045 atomic_read(&sp->sk_rmem_alloc),
1046 0, 0L, 0,
1047 sock_i_uid(sp), 0,
1048 sock_i_ino(sp),
1049 atomic_read(&sp->sk_refcnt), sp);
1050}
1051
1052static int udp6_seq_show(struct seq_file *seq, void *v)
1053{
1054 if (v == SEQ_START_TOKEN)
1055 seq_printf(seq,
1056 " sl "
1057 "local_address "
1058 "remote_address "
1059 "st tx_queue rx_queue tr tm->when retrnsmt"
1060 " uid timeout inode\n");
1061 else
1062 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1063 return 0;
1064}
1065
1066static struct file_operations udp6_seq_fops;
1067static struct udp_seq_afinfo udp6_seq_afinfo = {
1068 .owner = THIS_MODULE,
1069 .name = "udp6",
1070 .family = AF_INET6,
1071 .seq_show = udp6_seq_show,
1072 .seq_fops = &udp6_seq_fops,
1073};
1074
1075int __init udp6_proc_init(void)
1076{
1077 return udp_proc_register(&udp6_seq_afinfo);
1078}
1079
1080void udp6_proc_exit(void) {
1081 udp_proc_unregister(&udp6_seq_afinfo);
1082}
1083#endif /* CONFIG_PROC_FS */
1084
1085/* ------------------------------------------------------------------------ */
1086
1087struct proto udpv6_prot = {
543d9cfe
ACM		 1088 .name = "UDPv6",
1089 .owner = THIS_MODULE,
1090 .close = udpv6_close,
1091 .connect = ip6_datagram_connect,
1092 .disconnect = udp_disconnect,
1093 .ioctl = udp_ioctl,
1094 .destroy = udpv6_destroy_sock,
1095 .setsockopt = udpv6_setsockopt,
1096 .getsockopt = udpv6_getsockopt,
1097 .sendmsg = udpv6_sendmsg,
1098 .recvmsg = udpv6_recvmsg,
1099 .backlog_rcv = udpv6_queue_rcv_skb,
1100 .hash = udp_v6_hash,
1101 .unhash = udp_v6_unhash,
1102 .get_port = udp_v6_get_port,
1103 .obj_size = sizeof(struct udp6_sock),
3fdadf7d 1104#ifdef CONFIG_COMPAT
543d9cfe
ACM		 1105 .compat_setsockopt = compat_udpv6_setsockopt,
1106 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1107#endif
1da177e4
LT		 1108};
1109
1da177e4
LT		 1110static struct inet_protosw udpv6_protosw = {
1111 .type = SOCK_DGRAM,
1112 .protocol = IPPROTO_UDP,
1113 .prot = &udpv6_prot,
1114 .ops = &inet6_dgram_ops,
1115 .capability =-1,
1116 .no_check = UDP_CSUM_DEFAULT,
1117 .flags = INET_PROTOSW_PERMANENT,
1118};
1119
1120
1121void __init udpv6_init(void)
1122{
1123 if (inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP) < 0)
1124 printk(KERN_ERR "udpv6_init: Could not register protocol\n");
1125 inet6_register_protosw(&udpv6_protosw);
1126}
Clone of davem's net-next-2.6 tree