]> 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
[NET]: Supporting UDP-Lite (RFC 3828) in Linux
[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
1da177e4
LT		 41#include <net/ndisc.h>
42#include <net/protocol.h>
43#include <net/transp_v6.h>
44#include <net/ip6_route.h>
1da177e4 45#include <net/raw.h>
c752f073 46#include <net/tcp_states.h>
1da177e4
LT		 47#include <net/ip6_checksum.h>
48#include <net/xfrm.h>
49
50#include <linux/proc_fs.h>
51#include <linux/seq_file.h>
ba4e58ec 52#include "udp_impl.h"
1da177e4 53
ba89966c 54DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6) __read_mostly;
1da177e4 55
25030a7f 56static inline int udp_v6_get_port(struct sock *sk, unsigned short snum)
1da177e4 57{
25030a7f 58 return udp_get_port(sk, snum, ipv6_rcv_saddr_equal);
1da177e4
LT		 59}
60
ba4e58ec
GR		 61static struct sock *__udp6_lib_lookup(struct in6_addr *saddr, __be16 sport,
62 struct in6_addr *daddr, __be16 dport,
63 int dif, struct hlist_head udptable[])
1da177e4
LT		 64{
65 struct sock *sk, *result = NULL;
66 struct hlist_node *node;
67 unsigned short hnum = ntohs(dport);
68 int badness = -1;
69
70 read_lock(&udp_hash_lock);
ba4e58ec 71 sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) {
1da177e4
LT		 72 struct inet_sock *inet = inet_sk(sk);
73
74 if (inet->num == hnum && 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
1da177e4
LT		 112/*
113 * This should be easy, if there is something there we
114 * return it, otherwise we block.
115 */
116
ba4e58ec 117int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT		 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 size_t copied;
ba4e58ec 125 int err, copy_only, is_udplite = IS_UDPLITE(sk);
1da177e4
LT		 126
127 if (addr_len)
128 *addr_len=sizeof(struct sockaddr_in6);
129
130 if (flags & MSG_ERRQUEUE)
131 return ipv6_recv_error(sk, msg, len);
132
133try_again:
134 skb = skb_recv_datagram(sk, flags, noblock, &err);
135 if (!skb)
136 goto out;
137
138 copied = skb->len - sizeof(struct udphdr);
139 if (copied > len) {
140 copied = len;
141 msg->msg_flags |= MSG_TRUNC;
142 }
143
ba4e58ec
GR		 144 /*
145 * Decide whether to checksum and/or copy data.
146 */
147 copy_only = (skb->ip_summed==CHECKSUM_UNNECESSARY);
148
149 if (is_udplite || (!copy_only && msg->msg_flags&MSG_TRUNC)) {
150 if (__udp_lib_checksum_complete(skb))
1da177e4 151 goto csum_copy_err;
ba4e58ec
GR		 152 copy_only = 1;
153 }
154
155 if (copy_only)
156 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
157 msg->msg_iov, copied );
158 else {
1da177e4
LT		 159 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
160 if (err == -EINVAL)
161 goto csum_copy_err;
162 }
163 if (err)
164 goto out_free;
165
166 sock_recv_timestamp(msg, sk, skb);
167
168 /* Copy the address. */
169 if (msg->msg_name) {
170 struct sockaddr_in6 *sin6;
171
172 sin6 = (struct sockaddr_in6 *) msg->msg_name;
173 sin6->sin6_family = AF_INET6;
174 sin6->sin6_port = skb->h.uh->source;
175 sin6->sin6_flowinfo = 0;
176 sin6->sin6_scope_id = 0;
177
178 if (skb->protocol == htons(ETH_P_IP))
179 ipv6_addr_set(&sin6->sin6_addr, 0, 0,
180 htonl(0xffff), skb->nh.iph->saddr);
181 else {
182 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
183 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
184 sin6->sin6_scope_id = IP6CB(skb)->iif;
185 }
186
187 }
188 if (skb->protocol == htons(ETH_P_IP)) {
189 if (inet->cmsg_flags)
190 ip_cmsg_recv(msg, skb);
191 } else {
192 if (np->rxopt.all)
193 datagram_recv_ctl(sk, msg, skb);
194 }
195
196 err = copied;
197 if (flags & MSG_TRUNC)
198 err = skb->len - sizeof(struct udphdr);
199
200out_free:
201 skb_free_datagram(sk, skb);
202out:
203 return err;
204
205csum_copy_err:
3305b80c 206 skb_kill_datagram(sk, skb, flags);
1da177e4
LT		 207
208 if (flags & MSG_DONTWAIT) {
ba4e58ec 209 UDP6_INC_STATS_USER(UDP_MIB_INERRORS, is_udplite);
1da177e4
LT		 210 return -EAGAIN;
211 }
212 goto try_again;
213}
214
ba4e58ec
GR		 215void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
216 int type, int code, int offset, __be32 info,
217 struct hlist_head udptable[] )
1da177e4
LT		 218{
219 struct ipv6_pinfo *np;
220 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
1da177e4
LT		 221 struct in6_addr *saddr = &hdr->saddr;
222 struct in6_addr *daddr = &hdr->daddr;
223 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
224 struct sock *sk;
225 int err;
226
ba4e58ec
GR		 227 sk = __udp6_lib_lookup(daddr, uh->dest,
228 saddr, uh->source, inet6_iif(skb), udptable);
1da177e4
LT		 229 if (sk == NULL)
230 return;
231
232 np = inet6_sk(sk);
233
234 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
235 goto out;
236
237 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
238 goto out;
239
240 if (np->recverr)
241 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
242
243 sk->sk_err = err;
244 sk->sk_error_report(sk);
245out:
246 sock_put(sk);
247}
248
ba4e58ec
GR		 249static __inline__ void udpv6_err(struct sk_buff *skb,
250 struct inet6_skb_parm *opt, int type,
251 int code, int offset, __u32 info )
252{
253 return __udp6_lib_err(skb, opt, type, code, offset, info, udp_hash);
254}
255
256int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1da177e4 257{
ba4e58ec 258 struct udp_sock *up = udp_sk(sk);
a18135eb
DM		 259 int rc;
260
ba4e58ec
GR		 261 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
262 goto drop;
1da177e4 263
ba4e58ec
GR		 264 /*
265 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
266 */
267 if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
268
269 if (up->pcrlen == 0) { /* full coverage was set */
270 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
271 " %d while full coverage %d requested\n",
272 UDP_SKB_CB(skb)->cscov, skb->len);
273 goto drop;
274 }
275 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
276 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
277 "too small, need min %d\n",
278 UDP_SKB_CB(skb)->cscov, up->pcrlen);
279 goto drop;
280 }
1da177e4
LT		 281 }
282
ba4e58ec
GR		 283 if (udp_lib_checksum_complete(skb))
284 goto drop;
285
a18135eb
DM		 286 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
287 /* Note that an ENOMEM error is charged twice */
288 if (rc == -ENOMEM)
ba4e58ec
GR		 289 UDP6_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, up->pcflag);
290 goto drop;
1da177e4 291 }
ba4e58ec 292 UDP6_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1da177e4 293 return 0;
ba4e58ec
GR		 294drop:
295 UDP6_INC_STATS_BH(UDP_MIB_INERRORS, up->pcflag);
296 kfree_skb(skb);
297 return -1;
1da177e4
LT		 298}
299
300static struct sock *udp_v6_mcast_next(struct sock *sk,
301 u16 loc_port, struct in6_addr *loc_addr,
302 u16 rmt_port, struct in6_addr *rmt_addr,
303 int dif)
304{
305 struct hlist_node *node;
306 struct sock *s = sk;
307 unsigned short num = ntohs(loc_port);
308
309 sk_for_each_from(s, node) {
310 struct inet_sock *inet = inet_sk(s);
311
312 if (inet->num == num && s->sk_family == PF_INET6) {
313 struct ipv6_pinfo *np = inet6_sk(s);
314 if (inet->dport) {
315 if (inet->dport != rmt_port)
316 continue;
317 }
318 if (!ipv6_addr_any(&np->daddr) &&
319 !ipv6_addr_equal(&np->daddr, rmt_addr))
320 continue;
321
322 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
323 continue;
324
325 if (!ipv6_addr_any(&np->rcv_saddr)) {
40796c5e
DS		 326 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
327 continue;
1da177e4
LT		 328 }
329 if(!inet6_mc_check(s, loc_addr, rmt_addr))
330 continue;
331 return s;
332 }
333 }
334 return NULL;
335}
336
337/*
338 * Note: called only from the BH handler context,
339 * so we don't need to lock the hashes.
340 */
ba4e58ec
GR		 341static int __udp6_lib_mcast_deliver(struct sk_buff *skb, struct in6_addr *saddr,
342 struct in6_addr *daddr, struct hlist_head udptable[])
1da177e4
LT		 343{
344 struct sock *sk, *sk2;
ba4e58ec 345 const struct udphdr *uh = skb->h.uh;
1da177e4
LT		 346 int dif;
347
348 read_lock(&udp_hash_lock);
ba4e58ec 349 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
f2776ff0 350 dif = inet6_iif(skb);
1da177e4
LT		 351 sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
352 if (!sk) {
353 kfree_skb(skb);
354 goto out;
355 }
356
357 sk2 = sk;
358 while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
359 uh->source, saddr, dif))) {
360 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
361 if (buff)
362 udpv6_queue_rcv_skb(sk2, buff);
363 }
364 udpv6_queue_rcv_skb(sk, skb);
365out:
366 read_unlock(&udp_hash_lock);
ba4e58ec 367 return 0;
1da177e4
LT		 368}
369
ba4e58ec
GR		 370static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh)
371
372{
373 if (uh->check == 0) {
374 /* RFC 2460 section 8.1 says that we SHOULD log
375 this error. Well, it is reasonable.
376 */
377 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
378 return 1;
379 }
380 if (skb->ip_summed == CHECKSUM_COMPLETE &&
381 !csum_ipv6_magic(&skb->nh.ipv6h->saddr, &skb->nh.ipv6h->daddr,
382 skb->len, IPPROTO_UDP, skb->csum ))
383 skb->ip_summed = CHECKSUM_UNNECESSARY;
384
385 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
386 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
387 &skb->nh.ipv6h->daddr,
388 skb->len, IPPROTO_UDP, 0);
389
390 return (UDP_SKB_CB(skb)->partial_cov = 0);
391}
392
393int __udp6_lib_rcv(struct sk_buff **pskb, struct hlist_head udptable[],
394 int is_udplite)
1da177e4
LT		 395{
396 struct sk_buff *skb = *pskb;
397 struct sock *sk;
398 struct udphdr *uh;
399 struct net_device *dev = skb->dev;
400 struct in6_addr *saddr, *daddr;
401 u32 ulen = 0;
402
403 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
404 goto short_packet;
405
406 saddr = &skb->nh.ipv6h->saddr;
407 daddr = &skb->nh.ipv6h->daddr;
408 uh = skb->h.uh;
409
410 ulen = ntohs(uh->len);
ba4e58ec
GR		 411 if (ulen > skb->len)
412 goto short_packet;
1da177e4 413
ba4e58ec 414 if(! is_udplite ) { /* UDP validates ulen. */
1da177e4 415
ba4e58ec
GR		 416 /* Check for jumbo payload */
417 if (ulen == 0)
418 ulen = skb->len;
1da177e4 419
ba4e58ec
GR		 420 if (ulen < sizeof(*uh))
421 goto short_packet;
1da177e4 422
ba4e58ec
GR		 423 if (ulen < skb->len) {
424 if (pskb_trim_rcsum(skb, ulen))
425 goto short_packet;
426 saddr = &skb->nh.ipv6h->saddr;
427 daddr = &skb->nh.ipv6h->daddr;
428 uh = skb->h.uh;
429 }
1da177e4 430
ba4e58ec
GR		 431 if (udp6_csum_init(skb, uh))
432 goto discard;
fb286bb2 433
ba4e58ec
GR		 434 } else { /* UDP-Lite validates cscov. */
435 if (udplite6_csum_init(skb, uh))
436 goto discard;
437 }
1da177e4
LT		 438
439 /*
440 * Multicast receive code
441 */
ba4e58ec
GR		 442 if (ipv6_addr_is_multicast(daddr))
443 return __udp6_lib_mcast_deliver(skb, saddr, daddr, udptable);
1da177e4
LT		 444
445 /* Unicast */
446
447 /*
448 * check socket cache ... must talk to Alan about his plans
449 * for sock caches... i'll skip this for now.
450 */
ba4e58ec
GR		 451 sk = __udp6_lib_lookup(saddr, uh->source,
452 daddr, uh->dest, inet6_iif(skb), udptable);
1da177e4
LT		 453
454 if (sk == NULL) {
455 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
456 goto discard;
457
ba4e58ec 458 if (udp_lib_checksum_complete(skb))
1da177e4 459 goto discard;
ba4e58ec 460 UDP6_INC_STATS_BH(UDP_MIB_NOPORTS, is_udplite);
1da177e4
LT		 461
462 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
463
464 kfree_skb(skb);
465 return(0);
466 }
467
468 /* deliver */
469
470 udpv6_queue_rcv_skb(sk, skb);
471 sock_put(sk);
472 return(0);
473
474short_packet:
ba4e58ec
GR		 475 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: %d/%u\n",
476 is_udplite? "-Lite" : "", ulen, skb->len);
1da177e4
LT		 477
478discard:
ba4e58ec 479 UDP6_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
1da177e4
LT		 480 kfree_skb(skb);
481 return(0);
482}
ba4e58ec
GR		 483
484static __inline__ int udpv6_rcv(struct sk_buff **pskb)
485{
486 return __udp6_lib_rcv(pskb, udp_hash, 0);
487}
488
1da177e4
LT		 489/*
490 * Throw away all pending data and cancel the corking. Socket is locked.
491 */
492static void udp_v6_flush_pending_frames(struct sock *sk)
493{
494 struct udp_sock *up = udp_sk(sk);
495
496 if (up->pending) {
497 up->len = 0;
498 up->pending = 0;
499 ip6_flush_pending_frames(sk);
500 }
501}
502
503/*
504 * Sending
505 */
506
507static int udp_v6_push_pending_frames(struct sock *sk, struct udp_sock *up)
508{
509 struct sk_buff *skb;
510 struct udphdr *uh;
511 struct inet_sock *inet = inet_sk(sk);
512 struct flowi *fl = &inet->cork.fl;
513 int err = 0;
ba4e58ec 514 u32 csum = 0;
1da177e4
LT		 515
516 /* Grab the skbuff where UDP header space exists. */
517 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
518 goto out;
519
520 /*
521 * Create a UDP header
522 */
523 uh = skb->h.uh;
524 uh->source = fl->fl_ip_sport;
525 uh->dest = fl->fl_ip_dport;
526 uh->len = htons(up->len);
527 uh->check = 0;
528
ba4e58ec
GR		 529 if (up->pcflag)
530 csum = udplite_csum_outgoing(sk, skb);
531 else
532 csum = udp_csum_outgoing(sk, skb);
1da177e4 533
ba4e58ec
GR		 534 /* add protocol-dependent pseudo-header */
535 uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst,
536 up->len, fl->proto, csum );
1da177e4
LT		 537 if (uh->check == 0)
538 uh->check = -1;
539
1da177e4
LT		 540 err = ip6_push_pending_frames(sk);
541out:
542 up->len = 0;
543 up->pending = 0;
544 return err;
545}
546
ba4e58ec 547int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT		 548 struct msghdr *msg, size_t len)
549{
550 struct ipv6_txoptions opt_space;
551 struct udp_sock *up = udp_sk(sk);
552 struct inet_sock *inet = inet_sk(sk);
553 struct ipv6_pinfo *np = inet6_sk(sk);
554 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
555 struct in6_addr *daddr, *final_p = NULL, final;
556 struct ipv6_txoptions *opt = NULL;
557 struct ip6_flowlabel *flowlabel = NULL;
132a55f3 558 struct flowi fl;
1da177e4
LT		 559 struct dst_entry *dst;
560 int addr_len = msg->msg_namelen;
561 int ulen = len;
562 int hlimit = -1;
41a1f8ea 563 int tclass = -1;
1da177e4
LT		 564 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
565 int err;
987905de 566 int connected = 0;
ba4e58ec
GR		 567 int is_udplite = up->pcflag;
568 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1da177e4
LT		 569
570 /* destination address check */
571 if (sin6) {
572 if (addr_len < offsetof(struct sockaddr, sa_data))
573 return -EINVAL;
574
575 switch (sin6->sin6_family) {
576 case AF_INET6:
577 if (addr_len < SIN6_LEN_RFC2133)
578 return -EINVAL;
579 daddr = &sin6->sin6_addr;
580 break;
581 case AF_INET:
582 goto do_udp_sendmsg;
583 case AF_UNSPEC:
584 msg->msg_name = sin6 = NULL;
585 msg->msg_namelen = addr_len = 0;
586 daddr = NULL;
587 break;
588 default:
589 return -EINVAL;
590 }
591 } else if (!up->pending) {
592 if (sk->sk_state != TCP_ESTABLISHED)
593 return -EDESTADDRREQ;
594 daddr = &np->daddr;
595 } else
596 daddr = NULL;
597
598 if (daddr) {
599 if (ipv6_addr_type(daddr) == IPV6_ADDR_MAPPED) {
600 struct sockaddr_in sin;
601 sin.sin_family = AF_INET;
602 sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
603 sin.sin_addr.s_addr = daddr->s6_addr32[3];
604 msg->msg_name = &sin;
605 msg->msg_namelen = sizeof(sin);
606do_udp_sendmsg:
607 if (__ipv6_only_sock(sk))
608 return -ENETUNREACH;
609 return udp_sendmsg(iocb, sk, msg, len);
610 }
611 }
612
613 if (up->pending == AF_INET)
614 return udp_sendmsg(iocb, sk, msg, len);
615
616 /* Rough check on arithmetic overflow,
617 better check is made in ip6_build_xmit
618 */
619 if (len > INT_MAX - sizeof(struct udphdr))
620 return -EMSGSIZE;
621
622 if (up->pending) {
623 /*
624 * There are pending frames.
625 * The socket lock must be held while it's corked.
626 */
627 lock_sock(sk);
628 if (likely(up->pending)) {
629 if (unlikely(up->pending != AF_INET6)) {
630 release_sock(sk);
631 return -EAFNOSUPPORT;
632 }
633 dst = NULL;
634 goto do_append_data;
635 }
636 release_sock(sk);
637 }
638 ulen += sizeof(struct udphdr);
639
132a55f3 640 memset(&fl, 0, sizeof(fl));
1da177e4
LT		 641
642 if (sin6) {
643 if (sin6->sin6_port == 0)
644 return -EINVAL;
645
132a55f3 646 fl.fl_ip_dport = sin6->sin6_port;
1da177e4
LT		 647 daddr = &sin6->sin6_addr;
648
649 if (np->sndflow) {
132a55f3
HX		 650 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
651 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
652 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
1da177e4
LT		 653 if (flowlabel == NULL)
654 return -EINVAL;
655 daddr = &flowlabel->dst;
656 }
657 }
658
659 /*
660 * Otherwise it will be difficult to maintain
661 * sk->sk_dst_cache.
662 */
663 if (sk->sk_state == TCP_ESTABLISHED &&
664 ipv6_addr_equal(daddr, &np->daddr))
665 daddr = &np->daddr;
666
667 if (addr_len >= sizeof(struct sockaddr_in6) &&
668 sin6->sin6_scope_id &&
669 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
132a55f3 670 fl.oif = sin6->sin6_scope_id;
1da177e4
LT		 671 } else {
672 if (sk->sk_state != TCP_ESTABLISHED)
673 return -EDESTADDRREQ;
674
132a55f3 675 fl.fl_ip_dport = inet->dport;
1da177e4 676 daddr = &np->daddr;
132a55f3 677 fl.fl6_flowlabel = np->flow_label;
987905de 678 connected = 1;
1da177e4
LT		 679 }
680
132a55f3
HX		 681 if (!fl.oif)
682 fl.oif = sk->sk_bound_dev_if;
1da177e4
LT		 683
684 if (msg->msg_controllen) {
685 opt = &opt_space;
686 memset(opt, 0, sizeof(struct ipv6_txoptions));
687 opt->tot_len = sizeof(*opt);
688
132a55f3 689 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
1da177e4
LT		 690 if (err < 0) {
691 fl6_sock_release(flowlabel);
692 return err;
693 }
132a55f3
HX		 694 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
695 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
1da177e4
LT		 696 if (flowlabel == NULL)
697 return -EINVAL;
698 }
699 if (!(opt->opt_nflen|opt->opt_flen))
700 opt = NULL;
987905de 701 connected = 0;
1da177e4
LT		 702 }
703 if (opt == NULL)
704 opt = np->opt;
df9890c3
YH		 705 if (flowlabel)
706 opt = fl6_merge_options(&opt_space, flowlabel, opt);
707 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 708
ba4e58ec 709 fl.proto = sk->sk_protocol;
132a55f3
HX		 710 ipv6_addr_copy(&fl.fl6_dst, daddr);
711 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
712 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
713 fl.fl_ip_sport = inet->sport;
1da177e4
LT		 714
715 /* merge ip6_build_xmit from ip6_output */
716 if (opt && opt->srcrt) {
717 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
132a55f3
HX		 718 ipv6_addr_copy(&final, &fl.fl6_dst);
719 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1da177e4 720 final_p = &final;
987905de 721 connected = 0;
1da177e4
LT		 722 }
723
132a55f3
HX		 724 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) {
725 fl.oif = np->mcast_oif;
987905de
MK		 726 connected = 0;
727 }
1da177e4 728
132a55f3 729 security_sk_classify_flow(sk, &fl);
beb8d13b 730
132a55f3 731 err = ip6_sk_dst_lookup(sk, &dst, &fl);
1da177e4
LT		 732 if (err)
733 goto out;
734 if (final_p)
132a55f3 735 ipv6_addr_copy(&fl.fl6_dst, final_p);
1da177e4 736
132a55f3 737 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1da177e4 738 goto out;
1da177e4
LT		 739
740 if (hlimit < 0) {
132a55f3 741 if (ipv6_addr_is_multicast(&fl.fl6_dst))
1da177e4
LT		 742 hlimit = np->mcast_hops;
743 else
744 hlimit = np->hop_limit;
745 if (hlimit < 0)
746 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
747 if (hlimit < 0)
748 hlimit = ipv6_get_hoplimit(dst->dev);
749 }
750
41a1f8ea
YH		 751 if (tclass < 0) {
752 tclass = np->tclass;
753 if (tclass < 0)
754 tclass = 0;
755 }
756
1da177e4
LT		 757 if (msg->msg_flags&MSG_CONFIRM)
758 goto do_confirm;
759back_from_confirm:
760
761 lock_sock(sk);
762 if (unlikely(up->pending)) {
763 /* The socket is already corked while preparing it. */
764 /* ... which is an evident application bug. --ANK */
765 release_sock(sk);
766
64ce2073 767 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1da177e4
LT		 768 err = -EINVAL;
769 goto out;
770 }
771
772 up->pending = AF_INET6;
773
774do_append_data:
775 up->len += ulen;
ba4e58ec
GR		 776 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
777 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
132a55f3 778 sizeof(struct udphdr), hlimit, tclass, opt, &fl,
41a1f8ea
YH		 779 (struct rt6_info*)dst,
780 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
1da177e4
LT		 781 if (err)
782 udp_v6_flush_pending_frames(sk);
783 else if (!corkreq)
784 err = udp_v6_push_pending_frames(sk, up);
1e0c14f4
HX		 785 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
786 up->pending = 0;
1da177e4 787
a5e7c210
DM		 788 if (dst) {
789 if (connected) {
790 ip6_dst_store(sk, dst,
132a55f3 791 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
8e1ef0a9
YH		 792 &np->daddr : NULL,
793#ifdef CONFIG_IPV6_SUBTREES
132a55f3 794 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ?
8e1ef0a9
YH		 795 &np->saddr :
796#endif
797 NULL);
a5e7c210
DM		 798 } else {
799 dst_release(dst);
800 }
801 }
802
1da177e4
LT		 803 if (err > 0)
804 err = np->recverr ? net_xmit_errno(err) : 0;
805 release_sock(sk);
806out:
807 fl6_sock_release(flowlabel);
808 if (!err) {
ba4e58ec 809 UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite);
1da177e4
LT		 810 return len;
811 }
a18135eb
DM		 812 /*
813 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
814 * ENOBUFS might not be good (it's not tunable per se), but otherwise
815 * we don't have a good statistic (IpOutDiscards but it can be too many
816 * things). We could add another new stat but at least for now that
817 * seems like overkill.
818 */
819 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
ba4e58ec 820 UDP6_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 821 }
1da177e4
LT		 822 return err;
823
824do_confirm:
825 dst_confirm(dst);
826 if (!(msg->msg_flags&MSG_PROBE) || len)
827 goto back_from_confirm;
828 err = 0;
829 goto out;
830}
831
ba4e58ec 832int udpv6_destroy_sock(struct sock *sk)
1da177e4
LT		 833{
834 lock_sock(sk);
835 udp_v6_flush_pending_frames(sk);
836 release_sock(sk);
837
838 inet6_destroy_sock(sk);
839
840 return 0;
841}
842
843/*
844 * Socket option code for UDP
845 */
3fdadf7d 846static int do_udpv6_setsockopt(struct sock *sk, int level, int optname,
1da177e4
LT		 847 char __user *optval, int optlen)
848{
849 struct udp_sock *up = udp_sk(sk);
850 int val;
851 int err = 0;
852
1da177e4
LT		 853 if(optlen<sizeof(int))
854 return -EINVAL;
855
856 if (get_user(val, (int __user *)optval))
857 return -EFAULT;
858
859 switch(optname) {
860 case UDP_CORK:
861 if (val != 0) {
862 up->corkflag = 1;
863 } else {
864 up->corkflag = 0;
865 lock_sock(sk);
866 udp_v6_push_pending_frames(sk, up);
867 release_sock(sk);
868 }
869 break;
1da177e4
LT		 870 case UDP_ENCAP:
871 switch (val) {
872 case 0:
873 up->encap_type = val;
874 break;
875 default:
876 err = -ENOPROTOOPT;
877 break;
878 }
879 break;
880
ba4e58ec
GR		 881 case UDPLITE_SEND_CSCOV:
882 if (!up->pcflag) /* Disable the option on UDP sockets */
883 return -ENOPROTOOPT;
884 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
885 val = 8;
886 up->pcslen = val;
887 up->pcflag |= UDPLITE_SEND_CC;
888 break;
889
890 case UDPLITE_RECV_CSCOV:
891 if (!up->pcflag) /* Disable the option on UDP sockets */
892 return -ENOPROTOOPT;
893 if (val != 0 && val < 8) /* Avoid silly minimal values. */
894 val = 8;
895 up->pcrlen = val;
896 up->pcflag |= UDPLITE_RECV_CC;
897 break;
898
1da177e4
LT		 899 default:
900 err = -ENOPROTOOPT;
901 break;
902 };
903
904 return err;
905}
906
ba4e58ec
GR		 907int udpv6_setsockopt(struct sock *sk, int level, int optname,
908 char __user *optval, int optlen)
3fdadf7d 909{
ba4e58ec
GR		 910 if (level == SOL_UDP || level == SOL_UDPLITE)
911 return do_udpv6_setsockopt(sk, level, optname, optval, optlen);
912 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 913}
914
915#ifdef CONFIG_COMPAT
ba4e58ec
GR		 916int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
917 char __user *optval, int optlen)
3fdadf7d 918{
ba4e58ec
GR		 919 if (level == SOL_UDP || level == SOL_UDPLITE)
920 return do_udpv6_setsockopt(sk, level, optname, optval, optlen);
921 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 922}
923#endif
924
925static int do_udpv6_getsockopt(struct sock *sk, int level, int optname,
1da177e4
LT		 926 char __user *optval, int __user *optlen)
927{
928 struct udp_sock *up = udp_sk(sk);
929 int val, len;
930
1da177e4
LT		 931 if(get_user(len,optlen))
932 return -EFAULT;
933
934 len = min_t(unsigned int, len, sizeof(int));
935
936 if(len < 0)
937 return -EINVAL;
938
939 switch(optname) {
940 case UDP_CORK:
941 val = up->corkflag;
942 break;
943
944 case UDP_ENCAP:
945 val = up->encap_type;
946 break;
947
ba4e58ec
GR		 948 case UDPLITE_SEND_CSCOV:
949 val = up->pcslen;
950 break;
951
952 case UDPLITE_RECV_CSCOV:
953 val = up->pcrlen;
954 break;
955
1da177e4
LT		 956 default:
957 return -ENOPROTOOPT;
958 };
959
960 if(put_user(len, optlen))
961 return -EFAULT;
962 if(copy_to_user(optval, &val,len))
963 return -EFAULT;
964 return 0;
965}
966
ba4e58ec
GR		 967int udpv6_getsockopt(struct sock *sk, int level, int optname,
968 char __user *optval, int __user *optlen)
3fdadf7d 969{
ba4e58ec
GR		 970 if (level == SOL_UDP || level == SOL_UDPLITE)
971 return do_udpv6_getsockopt(sk, level, optname, optval, optlen);
972 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 973}
974
975#ifdef CONFIG_COMPAT
ba4e58ec
GR		 976int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
977 char __user *optval, int __user *optlen)
3fdadf7d 978{
ba4e58ec
GR		 979 if (level == SOL_UDP || level == SOL_UDPLITE)
980 return do_udpv6_getsockopt(sk, level, optname, optval, optlen);
981 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 982}
983#endif
984
1da177e4
LT		 985static struct inet6_protocol udpv6_protocol = {
986 .handler = udpv6_rcv,
987 .err_handler = udpv6_err,
988 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
989};
990
991/* ------------------------------------------------------------------------ */
992#ifdef CONFIG_PROC_FS
993
994static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
995{
996 struct inet_sock *inet = inet_sk(sp);
997 struct ipv6_pinfo *np = inet6_sk(sp);
998 struct in6_addr *dest, *src;
999 __u16 destp, srcp;
1000
1001 dest = &np->daddr;
1002 src = &np->rcv_saddr;
1003 destp = ntohs(inet->dport);
1004 srcp = ntohs(inet->sport);
1005 seq_printf(seq,
1006 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1007 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1008 bucket,
1009 src->s6_addr32[0], src->s6_addr32[1],
1010 src->s6_addr32[2], src->s6_addr32[3], srcp,
1011 dest->s6_addr32[0], dest->s6_addr32[1],
1012 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1013 sp->sk_state,
1014 atomic_read(&sp->sk_wmem_alloc),
1015 atomic_read(&sp->sk_rmem_alloc),
1016 0, 0L, 0,
1017 sock_i_uid(sp), 0,
1018 sock_i_ino(sp),
1019 atomic_read(&sp->sk_refcnt), sp);
1020}
1021
ba4e58ec 1022int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4
LT		 1023{
1024 if (v == SEQ_START_TOKEN)
1025 seq_printf(seq,
1026 " sl "
1027 "local_address "
1028 "remote_address "
1029 "st tx_queue rx_queue tr tm->when retrnsmt"
1030 " uid timeout inode\n");
1031 else
1032 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1033 return 0;
1034}
1035
1036static struct file_operations udp6_seq_fops;
1037static struct udp_seq_afinfo udp6_seq_afinfo = {
1038 .owner = THIS_MODULE,
1039 .name = "udp6",
1040 .family = AF_INET6,
ba4e58ec 1041 .hashtable = udp_hash,
1da177e4
LT		 1042 .seq_show = udp6_seq_show,
1043 .seq_fops = &udp6_seq_fops,
1044};
1045
1046int __init udp6_proc_init(void)
1047{
1048 return udp_proc_register(&udp6_seq_afinfo);
1049}
1050
1051void udp6_proc_exit(void) {
1052 udp_proc_unregister(&udp6_seq_afinfo);
1053}
1054#endif /* CONFIG_PROC_FS */
1055
1056/* ------------------------------------------------------------------------ */
1057
1058struct proto udpv6_prot = {
543d9cfe
ACM		 1059 .name = "UDPv6",
1060 .owner = THIS_MODULE,
ba4e58ec 1061 .close = udp_lib_close,
543d9cfe
ACM		 1062 .connect = ip6_datagram_connect,
1063 .disconnect = udp_disconnect,
1064 .ioctl = udp_ioctl,
1065 .destroy = udpv6_destroy_sock,
1066 .setsockopt = udpv6_setsockopt,
1067 .getsockopt = udpv6_getsockopt,
1068 .sendmsg = udpv6_sendmsg,
1069 .recvmsg = udpv6_recvmsg,
1070 .backlog_rcv = udpv6_queue_rcv_skb,
ba4e58ec
GR		 1071 .hash = udp_lib_hash,
1072 .unhash = udp_lib_unhash,
543d9cfe
ACM		 1073 .get_port = udp_v6_get_port,
1074 .obj_size = sizeof(struct udp6_sock),
3fdadf7d 1075#ifdef CONFIG_COMPAT
543d9cfe
ACM		 1076 .compat_setsockopt = compat_udpv6_setsockopt,
1077 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1078#endif
1da177e4
LT		 1079};
1080
1da177e4
LT		 1081static struct inet_protosw udpv6_protosw = {
1082 .type = SOCK_DGRAM,
1083 .protocol = IPPROTO_UDP,
1084 .prot = &udpv6_prot,
1085 .ops = &inet6_dgram_ops,
1086 .capability =-1,
1087 .no_check = UDP_CSUM_DEFAULT,
1088 .flags = INET_PROTOSW_PERMANENT,
1089};
1090
1091
1092void __init udpv6_init(void)
1093{
1094 if (inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP) < 0)
1095 printk(KERN_ERR "udpv6_init: Could not register protocol\n");
1096 inet6_register_protosw(&udpv6_protosw);
1097}
Clone of davem's net-next-2.6 tree