?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[SOCK]: Drop inuse pcounter from struct proto (v2).
[linux-2.6/zen-sources.git] / net / ipv6 / udp.c
blob463ae4d448a3dc3243b640a088a993bb0c272200
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
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/socket.h>
29 #include <linux/sockios.h>
30 #include <linux/net.h>
31 #include <linux/in6.h>
32 #include <linux/netdevice.h>
33 #include <linux/if_arp.h>
34 #include <linux/ipv6.h>
35 #include <linux/icmpv6.h>
36 #include <linux/init.h>
37 #include <linux/module.h>
38 #include <linux/skbuff.h>
39 #include <asm/uaccess.h>
40
41 #include <net/ndisc.h>
42 #include <net/protocol.h>
43 #include <net/transp_v6.h>
44 #include <net/ip6_route.h>
45 #include <net/raw.h>
46 #include <net/tcp_states.h>
47 #include <net/ip6_checksum.h>
48 #include <net/xfrm.h>
49
50 #include <linux/proc_fs.h>
51 #include <linux/seq_file.h>
52 #include "udp_impl.h"
53
54 int udp_v6_get_port(struct sock *sk, unsigned short snum)
55 {
56 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal);
57 }
58
59 static struct sock *__udp6_lib_lookup(struct net *net,
60 struct in6_addr *saddr, __be16 sport,
61 struct in6_addr *daddr, __be16 dport,
62 int dif, struct hlist_head udptable[])
63 {
64 struct sock *sk, *result = NULL;
65 struct hlist_node *node;
66 unsigned short hnum = ntohs(dport);
67 int badness = -1;
68
69 read_lock(&udp_hash_lock);
70 sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) {
71 struct inet_sock *inet = inet_sk(sk);
72
73 if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum &&
74 sk->sk_family == PF_INET6) {
75 struct ipv6_pinfo *np = inet6_sk(sk);
76 int score = 0;
77 if (inet->dport) {
78 if (inet->dport != sport)
79 continue;
80 score++;
81 }
82 if (!ipv6_addr_any(&np->rcv_saddr)) {
83 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
84 continue;
85 score++;
86 }
87 if (!ipv6_addr_any(&np->daddr)) {
88 if (!ipv6_addr_equal(&np->daddr, saddr))
89 continue;
90 score++;
91 }
92 if (sk->sk_bound_dev_if) {
93 if (sk->sk_bound_dev_if != dif)
94 continue;
95 score++;
96 }
97 if (score == 4) {
98 result = sk;
99 break;
100 } else if (score > badness) {
101 result = sk;
102 badness = score;
103 }
104 }
105 }
106 if (result)
107 sock_hold(result);
108 read_unlock(&udp_hash_lock);
109 return result;
110 }
111
112 /*
113 * This should be easy, if there is something there we
114 * return it, otherwise we block.
115 */
116
117 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
118 struct msghdr *msg, size_t len,
119 int noblock, int flags, int *addr_len)
120 {
121 struct ipv6_pinfo *np = inet6_sk(sk);
122 struct inet_sock *inet = inet_sk(sk);
123 struct sk_buff *skb;
124 unsigned int ulen, copied;
125 int peeked;
126 int err;
127 int is_udplite = IS_UDPLITE(sk);
128
129 if (addr_len)
130 *addr_len=sizeof(struct sockaddr_in6);
131
132 if (flags & MSG_ERRQUEUE)
133 return ipv6_recv_error(sk, msg, len);
134
135 try_again:
136 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
137 &peeked, &err);
138 if (!skb)
139 goto out;
140
141 ulen = skb->len - sizeof(struct udphdr);
142 copied = len;
143 if (copied > ulen)
144 copied = ulen;
145 else if (copied < ulen)
146 msg->msg_flags |= MSG_TRUNC;
147
148 /*
149 * If checksum is needed at all, try to do it while copying the
150 * data. If the data is truncated, or if we only want a partial
151 * coverage checksum (UDP-Lite), do it before the copy.
152 */
153
154 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
155 if (udp_lib_checksum_complete(skb))
156 goto csum_copy_err;
157 }
158
159 if (skb_csum_unnecessary(skb))
160 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
161 msg->msg_iov, copied );
162 else {
163 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
164 if (err == -EINVAL)
165 goto csum_copy_err;
166 }
167 if (err)
168 goto out_free;
169
170 if (!peeked)
171 UDP6_INC_STATS_USER(UDP_MIB_INDATAGRAMS, is_udplite);
172
173 sock_recv_timestamp(msg, sk, skb);
174
175 /* Copy the address. */
176 if (msg->msg_name) {
177 struct sockaddr_in6 *sin6;
178
179 sin6 = (struct sockaddr_in6 *) msg->msg_name;
180 sin6->sin6_family = AF_INET6;
181 sin6->sin6_port = udp_hdr(skb)->source;
182 sin6->sin6_flowinfo = 0;
183 sin6->sin6_scope_id = 0;
184
185 if (skb->protocol == htons(ETH_P_IP))
186 ipv6_addr_set(&sin6->sin6_addr, 0, 0,
187 htonl(0xffff), ip_hdr(skb)->saddr);
188 else {
189 ipv6_addr_copy(&sin6->sin6_addr,
190 &ipv6_hdr(skb)->saddr);
191 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
192 sin6->sin6_scope_id = IP6CB(skb)->iif;
193 }
194
195 }
196 if (skb->protocol == htons(ETH_P_IP)) {
197 if (inet->cmsg_flags)
198 ip_cmsg_recv(msg, skb);
199 } else {
200 if (np->rxopt.all)
201 datagram_recv_ctl(sk, msg, skb);
202 }
203
204 err = copied;
205 if (flags & MSG_TRUNC)
206 err = ulen;
207
208 out_free:
209 lock_sock(sk);
210 skb_free_datagram(sk, skb);
211 release_sock(sk);
212 out:
213 return err;
214
215 csum_copy_err:
216 lock_sock(sk);
217 if (!skb_kill_datagram(sk, skb, flags))
218 UDP6_INC_STATS_USER(UDP_MIB_INERRORS, is_udplite);
219 release_sock(sk);
220
221 if (flags & MSG_DONTWAIT)
222 return -EAGAIN;
223 goto try_again;
224 }
225
226 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
227 int type, int code, int offset, __be32 info,
228 struct hlist_head udptable[] )
229 {
230 struct ipv6_pinfo *np;
231 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
232 struct in6_addr *saddr = &hdr->saddr;
233 struct in6_addr *daddr = &hdr->daddr;
234 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
235 struct sock *sk;
236 int err;
237
238 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
239 saddr, uh->source, inet6_iif(skb), udptable);
240 if (sk == NULL)
241 return;
242
243 np = inet6_sk(sk);
244
245 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
246 goto out;
247
248 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
249 goto out;
250
251 if (np->recverr)
252 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
253
254 sk->sk_err = err;
255 sk->sk_error_report(sk);
256 out:
257 sock_put(sk);
258 }
259
260 static __inline__ void udpv6_err(struct sk_buff *skb,
261 struct inet6_skb_parm *opt, int type,
262 int code, int offset, __be32 info )
263 {
264 __udp6_lib_err(skb, opt, type, code, offset, info, udp_hash);
265 }
266
267 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
268 {
269 struct udp_sock *up = udp_sk(sk);
270 int rc;
271 int is_udplite = IS_UDPLITE(sk);
272
273 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
274 goto drop;
275
276 /*
277 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
278 */
279 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
280
281 if (up->pcrlen == 0) { /* full coverage was set */
282 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
283 " %d while full coverage %d requested\n",
284 UDP_SKB_CB(skb)->cscov, skb->len);
285 goto drop;
286 }
287 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
288 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
289 "too small, need min %d\n",
290 UDP_SKB_CB(skb)->cscov, up->pcrlen);
291 goto drop;
292 }
293 }
294
295 if (sk->sk_filter) {
296 if (udp_lib_checksum_complete(skb))
297 goto drop;
298 }
299
300 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
301 /* Note that an ENOMEM error is charged twice */
302 if (rc == -ENOMEM)
303 UDP6_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, is_udplite);
304 goto drop;
305 }
306
307 return 0;
308 drop:
309 UDP6_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
310 kfree_skb(skb);
311 return -1;
312 }
313
314 static struct sock *udp_v6_mcast_next(struct sock *sk,
315 __be16 loc_port, struct in6_addr *loc_addr,
316 __be16 rmt_port, struct in6_addr *rmt_addr,
317 int dif)
318 {
319 struct hlist_node *node;
320 struct sock *s = sk;
321 unsigned short num = ntohs(loc_port);
322
323 sk_for_each_from(s, node) {
324 struct inet_sock *inet = inet_sk(s);
325
326 if (sock_net(s) != sock_net(sk))
327 continue;
328
329 if (s->sk_hash == num && s->sk_family == PF_INET6) {
330 struct ipv6_pinfo *np = inet6_sk(s);
331 if (inet->dport) {
332 if (inet->dport != rmt_port)
333 continue;
334 }
335 if (!ipv6_addr_any(&np->daddr) &&
336 !ipv6_addr_equal(&np->daddr, rmt_addr))
337 continue;
338
339 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
340 continue;
341
342 if (!ipv6_addr_any(&np->rcv_saddr)) {
343 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
344 continue;
345 }
346 if (!inet6_mc_check(s, loc_addr, rmt_addr))
347 continue;
348 return s;
349 }
350 }
351 return NULL;
352 }
353
354 /*
355 * Note: called only from the BH handler context,
356 * so we don't need to lock the hashes.
357 */
358 static int __udp6_lib_mcast_deliver(struct sk_buff *skb, struct in6_addr *saddr,
359 struct in6_addr *daddr, struct hlist_head udptable[])
360 {
361 struct sock *sk, *sk2;
362 const struct udphdr *uh = udp_hdr(skb);
363 int dif;
364
365 read_lock(&udp_hash_lock);
366 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
367 dif = inet6_iif(skb);
368 sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
369 if (!sk) {
370 kfree_skb(skb);
371 goto out;
372 }
373
374 sk2 = sk;
375 while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
376 uh->source, saddr, dif))) {
377 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
378 if (buff) {
379 bh_lock_sock_nested(sk2);
380 if (!sock_owned_by_user(sk2))
381 udpv6_queue_rcv_skb(sk2, buff);
382 else
383 sk_add_backlog(sk2, buff);
384 bh_unlock_sock(sk2);
385 }
386 }
387 bh_lock_sock_nested(sk);
388 if (!sock_owned_by_user(sk))
389 udpv6_queue_rcv_skb(sk, skb);
390 else
391 sk_add_backlog(sk, skb);
392 bh_unlock_sock(sk);
393 out:
394 read_unlock(&udp_hash_lock);
395 return 0;
396 }
397
398 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
399 int proto)
400 {
401 int err;
402
403 UDP_SKB_CB(skb)->partial_cov = 0;
404 UDP_SKB_CB(skb)->cscov = skb->len;
405
406 if (proto == IPPROTO_UDPLITE) {
407 err = udplite_checksum_init(skb, uh);
408 if (err)
409 return err;
410 }
411
412 if (uh->check == 0) {
413 /* RFC 2460 section 8.1 says that we SHOULD log
414 this error. Well, it is reasonable.
415 */
416 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
417 return 1;
418 }
419 if (skb->ip_summed == CHECKSUM_COMPLETE &&
420 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
421 skb->len, proto, skb->csum))
422 skb->ip_summed = CHECKSUM_UNNECESSARY;
423
424 if (!skb_csum_unnecessary(skb))
425 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
426 &ipv6_hdr(skb)->daddr,
427 skb->len, proto, 0));
428
429 return 0;
430 }
431
432 int __udp6_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
433 int proto)
434 {
435 struct sock *sk;
436 struct udphdr *uh;
437 struct net_device *dev = skb->dev;
438 struct in6_addr *saddr, *daddr;
439 u32 ulen = 0;
440
441 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
442 goto short_packet;
443
444 saddr = &ipv6_hdr(skb)->saddr;
445 daddr = &ipv6_hdr(skb)->daddr;
446 uh = udp_hdr(skb);
447
448 ulen = ntohs(uh->len);
449 if (ulen > skb->len)
450 goto short_packet;
451
452 if (proto == IPPROTO_UDP) {
453 /* UDP validates ulen. */
454
455 /* Check for jumbo payload */
456 if (ulen == 0)
457 ulen = skb->len;
458
459 if (ulen < sizeof(*uh))
460 goto short_packet;
461
462 if (ulen < skb->len) {
463 if (pskb_trim_rcsum(skb, ulen))
464 goto short_packet;
465 saddr = &ipv6_hdr(skb)->saddr;
466 daddr = &ipv6_hdr(skb)->daddr;
467 uh = udp_hdr(skb);
468 }
469 }
470
471 if (udp6_csum_init(skb, uh, proto))
472 goto discard;
473
474 /*
475 * Multicast receive code
476 */
477 if (ipv6_addr_is_multicast(daddr))
478 return __udp6_lib_mcast_deliver(skb, saddr, daddr, udptable);
479
480 /* Unicast */
481
482 /*
483 * check socket cache ... must talk to Alan about his plans
484 * for sock caches... i'll skip this for now.
485 */
486 sk = __udp6_lib_lookup(dev_net(skb->dev), saddr, uh->source,
487 daddr, uh->dest, inet6_iif(skb), udptable);
488
489 if (sk == NULL) {
490 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
491 goto discard;
492
493 if (udp_lib_checksum_complete(skb))
494 goto discard;
495 UDP6_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
496
497 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
498
499 kfree_skb(skb);
500 return 0;
501 }
502
503 /* deliver */
504
505 bh_lock_sock_nested(sk);
506 if (!sock_owned_by_user(sk))
507 udpv6_queue_rcv_skb(sk, skb);
508 else
509 sk_add_backlog(sk, skb);
510 bh_unlock_sock(sk);
511 sock_put(sk);
512 return 0;
513
514 short_packet:
515 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: %d/%u\n",
516 proto == IPPROTO_UDPLITE ? "-Lite" : "",
517 ulen, skb->len);
518
519 discard:
520 UDP6_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
521 kfree_skb(skb);
522 return 0;
523 }
524
525 static __inline__ int udpv6_rcv(struct sk_buff *skb)
526 {
527 return __udp6_lib_rcv(skb, udp_hash, IPPROTO_UDP);
528 }
529
530 /*
531 * Throw away all pending data and cancel the corking. Socket is locked.
532 */
533 static void udp_v6_flush_pending_frames(struct sock *sk)
534 {
535 struct udp_sock *up = udp_sk(sk);
536
537 if (up->pending) {
538 up->len = 0;
539 up->pending = 0;
540 ip6_flush_pending_frames(sk);
541 }
542 }
543
544 /*
545 * Sending
546 */
547
548 static int udp_v6_push_pending_frames(struct sock *sk)
549 {
550 struct sk_buff *skb;
551 struct udphdr *uh;
552 struct udp_sock *up = udp_sk(sk);
553 struct inet_sock *inet = inet_sk(sk);
554 struct flowi *fl = &inet->cork.fl;
555 int err = 0;
556 int is_udplite = IS_UDPLITE(sk);
557 __wsum csum = 0;
558
559 /* Grab the skbuff where UDP header space exists. */
560 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
561 goto out;
562
563 /*
564 * Create a UDP header
565 */
566 uh = udp_hdr(skb);
567 uh->source = fl->fl_ip_sport;
568 uh->dest = fl->fl_ip_dport;
569 uh->len = htons(up->len);
570 uh->check = 0;
571
572 if (is_udplite)
573 csum = udplite_csum_outgoing(sk, skb);
574 else
575 csum = udp_csum_outgoing(sk, skb);
576
577 /* add protocol-dependent pseudo-header */
578 uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst,
579 up->len, fl->proto, csum );
580 if (uh->check == 0)
581 uh->check = CSUM_MANGLED_0;
582
583 err = ip6_push_pending_frames(sk);
584 out:
585 up->len = 0;
586 up->pending = 0;
587 if (!err)
588 UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite);
589 return err;
590 }
591
592 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
593 struct msghdr *msg, size_t len)
594 {
595 struct ipv6_txoptions opt_space;
596 struct udp_sock *up = udp_sk(sk);
597 struct inet_sock *inet = inet_sk(sk);
598 struct ipv6_pinfo *np = inet6_sk(sk);
599 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
600 struct in6_addr *daddr, *final_p = NULL, final;
601 struct ipv6_txoptions *opt = NULL;
602 struct ip6_flowlabel *flowlabel = NULL;
603 struct flowi fl;
604 struct dst_entry *dst;
605 int addr_len = msg->msg_namelen;
606 int ulen = len;
607 int hlimit = -1;
608 int tclass = -1;
609 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
610 int err;
611 int connected = 0;
612 int is_udplite = IS_UDPLITE(sk);
613 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
614
615 /* destination address check */
616 if (sin6) {
617 if (addr_len < offsetof(struct sockaddr, sa_data))
618 return -EINVAL;
619
620 switch (sin6->sin6_family) {
621 case AF_INET6:
622 if (addr_len < SIN6_LEN_RFC2133)
623 return -EINVAL;
624 daddr = &sin6->sin6_addr;
625 break;
626 case AF_INET:
627 goto do_udp_sendmsg;
628 case AF_UNSPEC:
629 msg->msg_name = sin6 = NULL;
630 msg->msg_namelen = addr_len = 0;
631 daddr = NULL;
632 break;
633 default:
634 return -EINVAL;
635 }
636 } else if (!up->pending) {
637 if (sk->sk_state != TCP_ESTABLISHED)
638 return -EDESTADDRREQ;
639 daddr = &np->daddr;
640 } else
641 daddr = NULL;
642
643 if (daddr) {
644 if (ipv6_addr_v4mapped(daddr)) {
645 struct sockaddr_in sin;
646 sin.sin_family = AF_INET;
647 sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
648 sin.sin_addr.s_addr = daddr->s6_addr32[3];
649 msg->msg_name = &sin;
650 msg->msg_namelen = sizeof(sin);
651 do_udp_sendmsg:
652 if (__ipv6_only_sock(sk))
653 return -ENETUNREACH;
654 return udp_sendmsg(iocb, sk, msg, len);
655 }
656 }
657
658 if (up->pending == AF_INET)
659 return udp_sendmsg(iocb, sk, msg, len);
660
661 /* Rough check on arithmetic overflow,
662 better check is made in ip6_append_data().
663 */
664 if (len > INT_MAX - sizeof(struct udphdr))
665 return -EMSGSIZE;
666
667 if (up->pending) {
668 /*
669 * There are pending frames.
670 * The socket lock must be held while it's corked.
671 */
672 lock_sock(sk);
673 if (likely(up->pending)) {
674 if (unlikely(up->pending != AF_INET6)) {
675 release_sock(sk);
676 return -EAFNOSUPPORT;
677 }
678 dst = NULL;
679 goto do_append_data;
680 }
681 release_sock(sk);
682 }
683 ulen += sizeof(struct udphdr);
684
685 memset(&fl, 0, sizeof(fl));
686
687 if (sin6) {
688 if (sin6->sin6_port == 0)
689 return -EINVAL;
690
691 fl.fl_ip_dport = sin6->sin6_port;
692 daddr = &sin6->sin6_addr;
693
694 if (np->sndflow) {
695 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
696 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
697 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
698 if (flowlabel == NULL)
699 return -EINVAL;
700 daddr = &flowlabel->dst;
701 }
702 }
703
704 /*
705 * Otherwise it will be difficult to maintain
706 * sk->sk_dst_cache.
707 */
708 if (sk->sk_state == TCP_ESTABLISHED &&
709 ipv6_addr_equal(daddr, &np->daddr))
710 daddr = &np->daddr;
711
712 if (addr_len >= sizeof(struct sockaddr_in6) &&
713 sin6->sin6_scope_id &&
714 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
715 fl.oif = sin6->sin6_scope_id;
716 } else {
717 if (sk->sk_state != TCP_ESTABLISHED)
718 return -EDESTADDRREQ;
719
720 fl.fl_ip_dport = inet->dport;
721 daddr = &np->daddr;
722 fl.fl6_flowlabel = np->flow_label;
723 connected = 1;
724 }
725
726 if (!fl.oif)
727 fl.oif = sk->sk_bound_dev_if;
728
729 if (msg->msg_controllen) {
730 opt = &opt_space;
731 memset(opt, 0, sizeof(struct ipv6_txoptions));
732 opt->tot_len = sizeof(*opt);
733
734 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
735 if (err < 0) {
736 fl6_sock_release(flowlabel);
737 return err;
738 }
739 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
740 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
741 if (flowlabel == NULL)
742 return -EINVAL;
743 }
744 if (!(opt->opt_nflen|opt->opt_flen))
745 opt = NULL;
746 connected = 0;
747 }
748 if (opt == NULL)
749 opt = np->opt;
750 if (flowlabel)
751 opt = fl6_merge_options(&opt_space, flowlabel, opt);
752 opt = ipv6_fixup_options(&opt_space, opt);
753
754 fl.proto = sk->sk_protocol;
755 ipv6_addr_copy(&fl.fl6_dst, daddr);
756 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
757 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
758 fl.fl_ip_sport = inet->sport;
759
760 /* merge ip6_build_xmit from ip6_output */
761 if (opt && opt->srcrt) {
762 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
763 ipv6_addr_copy(&final, &fl.fl6_dst);
764 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
765 final_p = &final;
766 connected = 0;
767 }
768
769 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) {
770 fl.oif = np->mcast_oif;
771 connected = 0;
772 }
773
774 security_sk_classify_flow(sk, &fl);
775
776 err = ip6_sk_dst_lookup(sk, &dst, &fl);
777 if (err)
778 goto out;
779 if (final_p)
780 ipv6_addr_copy(&fl.fl6_dst, final_p);
781
782 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
783 if (err == -EREMOTE)
784 err = ip6_dst_blackhole(sk, &dst, &fl);
785 if (err < 0)
786 goto out;
787 }
788
789 if (hlimit < 0) {
790 if (ipv6_addr_is_multicast(&fl.fl6_dst))
791 hlimit = np->mcast_hops;
792 else
793 hlimit = np->hop_limit;
794 if (hlimit < 0)
795 hlimit = ip6_dst_hoplimit(dst);
796 }
797
798 if (tclass < 0) {
799 tclass = np->tclass;
800 if (tclass < 0)
801 tclass = 0;
802 }
803
804 if (msg->msg_flags&MSG_CONFIRM)
805 goto do_confirm;
806 back_from_confirm:
807
808 lock_sock(sk);
809 if (unlikely(up->pending)) {
810 /* The socket is already corked while preparing it. */
811 /* ... which is an evident application bug. --ANK */
812 release_sock(sk);
813
814 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
815 err = -EINVAL;
816 goto out;
817 }
818
819 up->pending = AF_INET6;
820
821 do_append_data:
822 up->len += ulen;
823 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
824 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
825 sizeof(struct udphdr), hlimit, tclass, opt, &fl,
826 (struct rt6_info*)dst,
827 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
828 if (err)
829 udp_v6_flush_pending_frames(sk);
830 else if (!corkreq)
831 err = udp_v6_push_pending_frames(sk);
832 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
833 up->pending = 0;
834
835 if (dst) {
836 if (connected) {
837 ip6_dst_store(sk, dst,
838 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
839 &np->daddr : NULL,
840 #ifdef CONFIG_IPV6_SUBTREES
841 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ?
842 &np->saddr :
843 #endif
844 NULL);
845 } else {
846 dst_release(dst);
847 }
848 }
849
850 if (err > 0)
851 err = np->recverr ? net_xmit_errno(err) : 0;
852 release_sock(sk);
853 out:
854 fl6_sock_release(flowlabel);
855 if (!err)
856 return len;
857 /*
858 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
859 * ENOBUFS might not be good (it's not tunable per se), but otherwise
860 * we don't have a good statistic (IpOutDiscards but it can be too many
861 * things). We could add another new stat but at least for now that
862 * seems like overkill.
863 */
864 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
865 UDP6_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite);
866 }
867 return err;
868
869 do_confirm:
870 dst_confirm(dst);
871 if (!(msg->msg_flags&MSG_PROBE) || len)
872 goto back_from_confirm;
873 err = 0;
874 goto out;
875 }
876
877 int udpv6_destroy_sock(struct sock *sk)
878 {
879 lock_sock(sk);
880 udp_v6_flush_pending_frames(sk);
881 release_sock(sk);
882
883 inet6_destroy_sock(sk);
884
885 return 0;
886 }
887
888 /*
889 * Socket option code for UDP
890 */
891 int udpv6_setsockopt(struct sock *sk, int level, int optname,
892 char __user *optval, int optlen)
893 {
894 if (level == SOL_UDP || level == SOL_UDPLITE)
895 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
896 udp_v6_push_pending_frames);
897 return ipv6_setsockopt(sk, level, optname, optval, optlen);
898 }
899
900 #ifdef CONFIG_COMPAT
901 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
902 char __user *optval, int optlen)
903 {
904 if (level == SOL_UDP || level == SOL_UDPLITE)
905 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
906 udp_v6_push_pending_frames);
907 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
908 }
909 #endif
910
911 int udpv6_getsockopt(struct sock *sk, int level, int optname,
912 char __user *optval, int __user *optlen)
913 {
914 if (level == SOL_UDP || level == SOL_UDPLITE)
915 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
916 return ipv6_getsockopt(sk, level, optname, optval, optlen);
917 }
918
919 #ifdef CONFIG_COMPAT
920 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
921 char __user *optval, int __user *optlen)
922 {
923 if (level == SOL_UDP || level == SOL_UDPLITE)
924 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
925 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
926 }
927 #endif
928
929 static struct inet6_protocol udpv6_protocol = {
930 .handler = udpv6_rcv,
931 .err_handler = udpv6_err,
932 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
933 };
934
935 /* ------------------------------------------------------------------------ */
936 #ifdef CONFIG_PROC_FS
937
938 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
939 {
940 struct inet_sock *inet = inet_sk(sp);
941 struct ipv6_pinfo *np = inet6_sk(sp);
942 struct in6_addr *dest, *src;
943 __u16 destp, srcp;
944
945 dest = &np->daddr;
946 src = &np->rcv_saddr;
947 destp = ntohs(inet->dport);
948 srcp = ntohs(inet->sport);
949 seq_printf(seq,
950 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
951 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
952 bucket,
953 src->s6_addr32[0], src->s6_addr32[1],
954 src->s6_addr32[2], src->s6_addr32[3], srcp,
955 dest->s6_addr32[0], dest->s6_addr32[1],
956 dest->s6_addr32[2], dest->s6_addr32[3], destp,
957 sp->sk_state,
958 atomic_read(&sp->sk_wmem_alloc),
959 atomic_read(&sp->sk_rmem_alloc),
960 0, 0L, 0,
961 sock_i_uid(sp), 0,
962 sock_i_ino(sp),
963 atomic_read(&sp->sk_refcnt), sp);
964 }
965
966 int udp6_seq_show(struct seq_file *seq, void *v)
967 {
968 if (v == SEQ_START_TOKEN)
969 seq_printf(seq,
970 " sl "
971 "local_address "
972 "remote_address "
973 "st tx_queue rx_queue tr tm->when retrnsmt"
974 " uid timeout inode\n");
975 else
976 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
977 return 0;
978 }
979
980 static struct file_operations udp6_seq_fops;
981 static struct udp_seq_afinfo udp6_seq_afinfo = {
982 .owner = THIS_MODULE,
983 .name = "udp6",
984 .family = AF_INET6,
985 .hashtable = udp_hash,
986 .seq_show = udp6_seq_show,
987 .seq_fops = &udp6_seq_fops,
988 };
989
990 int udp6_proc_init(struct net *net)
991 {
992 return udp_proc_register(net, &udp6_seq_afinfo);
993 }
994
995 void udp6_proc_exit(struct net *net) {
996 udp_proc_unregister(net, &udp6_seq_afinfo);
997 }
998 #endif /* CONFIG_PROC_FS */
999
1000 /* ------------------------------------------------------------------------ */
1001
1002 struct proto udpv6_prot = {
1003 .name = "UDPv6",
1004 .owner = THIS_MODULE,
1005 .close = udp_lib_close,
1006 .connect = ip6_datagram_connect,
1007 .disconnect = udp_disconnect,
1008 .ioctl = udp_ioctl,
1009 .destroy = udpv6_destroy_sock,
1010 .setsockopt = udpv6_setsockopt,
1011 .getsockopt = udpv6_getsockopt,
1012 .sendmsg = udpv6_sendmsg,
1013 .recvmsg = udpv6_recvmsg,
1014 .backlog_rcv = udpv6_queue_rcv_skb,
1015 .hash = udp_lib_hash,
1016 .unhash = udp_lib_unhash,
1017 .get_port = udp_v6_get_port,
1018 .memory_allocated = &udp_memory_allocated,
1019 .sysctl_mem = sysctl_udp_mem,
1020 .sysctl_wmem = &sysctl_udp_wmem_min,
1021 .sysctl_rmem = &sysctl_udp_rmem_min,
1022 .obj_size = sizeof(struct udp6_sock),
1023 .h.udp_hash = udp_hash,
1024 #ifdef CONFIG_COMPAT
1025 .compat_setsockopt = compat_udpv6_setsockopt,
1026 .compat_getsockopt = compat_udpv6_getsockopt,
1027 #endif
1028 };
1029
1030 static struct inet_protosw udpv6_protosw = {
1031 .type = SOCK_DGRAM,
1032 .protocol = IPPROTO_UDP,
1033 .prot = &udpv6_prot,
1034 .ops = &inet6_dgram_ops,
1035 .capability =-1,
1036 .no_check = UDP_CSUM_DEFAULT,
1037 .flags = INET_PROTOSW_PERMANENT,
1038 };
1039
1040
1041 int __init udpv6_init(void)
1042 {
1043 int ret;
1044
1045 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1046 if (ret)
1047 goto out;
1048
1049 ret = inet6_register_protosw(&udpv6_protosw);
1050 if (ret)
1051 goto out_udpv6_protocol;
1052 out:
1053 return ret;
1054
1055 out_udpv6_protocol:
1056 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1057 goto out;
1058 }
1059
1060 void udpv6_exit(void)
1061 {
1062 inet6_unregister_protosw(&udpv6_protosw);
1063 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1064 }
Stable & featurefull patchset based on Linus's git branch