diff options
Diffstat (limited to 'net/ipv6/udp.c')
-rw-r--r-- | net/ipv6/udp.c | 1065 |
1 files changed, 1065 insertions, 0 deletions
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c new file mode 100644 index 000000000000..53739de829db --- /dev/null +++ b/net/ipv6/udp.c | |||
@@ -0,0 +1,1065 @@ | |||
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 | static inline int udp_v6_get_port(struct sock *sk, unsigned short snum) | ||
55 | { | ||
56 | return udp_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 (sk->sk_net == 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(skb->dev->nd_net, 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 (s->sk_hash == num && s->sk_family == PF_INET6) { | ||
327 | struct ipv6_pinfo *np = inet6_sk(s); | ||
328 | if (inet->dport) { | ||
329 | if (inet->dport != rmt_port) | ||
330 | continue; | ||
331 | } | ||
332 | if (!ipv6_addr_any(&np->daddr) && | ||
333 | !ipv6_addr_equal(&np->daddr, rmt_addr)) | ||
334 | continue; | ||
335 | |||
336 | if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif) | ||
337 | continue; | ||
338 | |||
339 | if (!ipv6_addr_any(&np->rcv_saddr)) { | ||
340 | if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr)) | ||
341 | continue; | ||
342 | } | ||
343 | if (!inet6_mc_check(s, loc_addr, rmt_addr)) | ||
344 | continue; | ||
345 | return s; | ||
346 | } | ||
347 | } | ||
348 | return NULL; | ||
349 | } | ||
350 | |||
351 | /* | ||
352 | * Note: called only from the BH handler context, | ||
353 | * so we don't need to lock the hashes. | ||
354 | */ | ||
355 | static int __udp6_lib_mcast_deliver(struct sk_buff *skb, struct in6_addr *saddr, | ||
356 | struct in6_addr *daddr, struct hlist_head udptable[]) | ||
357 | { | ||
358 | struct sock *sk, *sk2; | ||
359 | const struct udphdr *uh = udp_hdr(skb); | ||
360 | int dif; | ||
361 | |||
362 | read_lock(&udp_hash_lock); | ||
363 | sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]); | ||
364 | dif = inet6_iif(skb); | ||
365 | sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); | ||
366 | if (!sk) { | ||
367 | kfree_skb(skb); | ||
368 | goto out; | ||
369 | } | ||
370 | |||
371 | sk2 = sk; | ||
372 | while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr, | ||
373 | uh->source, saddr, dif))) { | ||
374 | struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC); | ||
375 | if (buff) { | ||
376 | bh_lock_sock_nested(sk2); | ||
377 | if (!sock_owned_by_user(sk2)) | ||
378 | udpv6_queue_rcv_skb(sk2, buff); | ||
379 | else | ||
380 | sk_add_backlog(sk2, buff); | ||
381 | bh_unlock_sock(sk2); | ||
382 | } | ||
383 | } | ||
384 | bh_lock_sock_nested(sk); | ||
385 | if (!sock_owned_by_user(sk)) | ||
386 | udpv6_queue_rcv_skb(sk, skb); | ||
387 | else | ||
388 | sk_add_backlog(sk, skb); | ||
389 | bh_unlock_sock(sk); | ||
390 | out: | ||
391 | read_unlock(&udp_hash_lock); | ||
392 | return 0; | ||
393 | } | ||
394 | |||
395 | static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh, | ||
396 | int proto) | ||
397 | { | ||
398 | int err; | ||
399 | |||
400 | UDP_SKB_CB(skb)->partial_cov = 0; | ||
401 | UDP_SKB_CB(skb)->cscov = skb->len; | ||
402 | |||
403 | if (proto == IPPROTO_UDPLITE) { | ||
404 | err = udplite_checksum_init(skb, uh); | ||
405 | if (err) | ||
406 | return err; | ||
407 | } | ||
408 | |||
409 | if (uh->check == 0) { | ||
410 | /* RFC 2460 section 8.1 says that we SHOULD log | ||
411 | this error. Well, it is reasonable. | ||
412 | */ | ||
413 | LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n"); | ||
414 | return 1; | ||
415 | } | ||
416 | if (skb->ip_summed == CHECKSUM_COMPLETE && | ||
417 | !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, | ||
418 | skb->len, proto, skb->csum)) | ||
419 | skb->ip_summed = CHECKSUM_UNNECESSARY; | ||
420 | |||
421 | if (!skb_csum_unnecessary(skb)) | ||
422 | skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, | ||
423 | &ipv6_hdr(skb)->daddr, | ||
424 | skb->len, proto, 0)); | ||
425 | |||
426 | return 0; | ||
427 | } | ||
428 | |||
429 | int __udp6_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[], | ||
430 | int proto) | ||
431 | { | ||
432 | struct sock *sk; | ||
433 | struct udphdr *uh; | ||
434 | struct net_device *dev = skb->dev; | ||
435 | struct in6_addr *saddr, *daddr; | ||
436 | u32 ulen = 0; | ||
437 | |||
438 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | ||
439 | goto short_packet; | ||
440 | |||
441 | saddr = &ipv6_hdr(skb)->saddr; | ||
442 | daddr = &ipv6_hdr(skb)->daddr; | ||
443 | uh = udp_hdr(skb); | ||
444 | |||
445 | ulen = ntohs(uh->len); | ||
446 | if (ulen > skb->len) | ||
447 | goto short_packet; | ||
448 | |||
449 | if (proto == IPPROTO_UDP) { | ||
450 | /* UDP validates ulen. */ | ||
451 | |||
452 | /* Check for jumbo payload */ | ||
453 | if (ulen == 0) | ||
454 | ulen = skb->len; | ||
455 | |||
456 | if (ulen < sizeof(*uh)) | ||
457 | goto short_packet; | ||
458 | |||
459 | if (ulen < skb->len) { | ||
460 | if (pskb_trim_rcsum(skb, ulen)) | ||
461 | goto short_packet; | ||
462 | saddr = &ipv6_hdr(skb)->saddr; | ||
463 | daddr = &ipv6_hdr(skb)->daddr; | ||
464 | uh = udp_hdr(skb); | ||
465 | } | ||
466 | } | ||
467 | |||
468 | if (udp6_csum_init(skb, uh, proto)) | ||
469 | goto discard; | ||
470 | |||
471 | /* | ||
472 | * Multicast receive code | ||
473 | */ | ||
474 | if (ipv6_addr_is_multicast(daddr)) | ||
475 | return __udp6_lib_mcast_deliver(skb, saddr, daddr, udptable); | ||
476 | |||
477 | /* Unicast */ | ||
478 | |||
479 | /* | ||
480 | * check socket cache ... must talk to Alan about his plans | ||
481 | * for sock caches... i'll skip this for now. | ||
482 | */ | ||
483 | sk = __udp6_lib_lookup(skb->dev->nd_net, saddr, uh->source, | ||
484 | daddr, uh->dest, inet6_iif(skb), udptable); | ||
485 | |||
486 | if (sk == NULL) { | ||
487 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) | ||
488 | goto discard; | ||
489 | |||
490 | if (udp_lib_checksum_complete(skb)) | ||
491 | goto discard; | ||
492 | UDP6_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); | ||
493 | |||
494 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev); | ||
495 | |||
496 | kfree_skb(skb); | ||
497 | return 0; | ||
498 | } | ||
499 | |||
500 | /* deliver */ | ||
501 | |||
502 | bh_lock_sock_nested(sk); | ||
503 | if (!sock_owned_by_user(sk)) | ||
504 | udpv6_queue_rcv_skb(sk, skb); | ||
505 | else | ||
506 | sk_add_backlog(sk, skb); | ||
507 | bh_unlock_sock(sk); | ||
508 | sock_put(sk); | ||
509 | return 0; | ||
510 | |||
511 | short_packet: | ||
512 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: %d/%u\n", | ||
513 | proto == IPPROTO_UDPLITE ? "-Lite" : "", | ||
514 | ulen, skb->len); | ||
515 | |||
516 | discard: | ||
517 | UDP6_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); | ||
518 | kfree_skb(skb); | ||
519 | return 0; | ||
520 | } | ||
521 | |||
522 | static __inline__ int udpv6_rcv(struct sk_buff *skb) | ||
523 | { | ||
524 | return __udp6_lib_rcv(skb, udp_hash, IPPROTO_UDP); | ||
525 | } | ||
526 | |||
527 | /* | ||
528 | * Throw away all pending data and cancel the corking. Socket is locked. | ||
529 | */ | ||
530 | static void udp_v6_flush_pending_frames(struct sock *sk) | ||
531 | { | ||
532 | struct udp_sock *up = udp_sk(sk); | ||
533 | |||
534 | if (up->pending) { | ||
535 | up->len = 0; | ||
536 | up->pending = 0; | ||
537 | ip6_flush_pending_frames(sk); | ||
538 | } | ||
539 | } | ||
540 | |||
541 | /* | ||
542 | * Sending | ||
543 | */ | ||
544 | |||
545 | static int udp_v6_push_pending_frames(struct sock *sk) | ||
546 | { | ||
547 | struct sk_buff *skb; | ||
548 | struct udphdr *uh; | ||
549 | struct udp_sock *up = udp_sk(sk); | ||
550 | struct inet_sock *inet = inet_sk(sk); | ||
551 | struct flowi *fl = &inet->cork.fl; | ||
552 | int err = 0; | ||
553 | int is_udplite = IS_UDPLITE(sk); | ||
554 | __wsum csum = 0; | ||
555 | |||
556 | /* Grab the skbuff where UDP header space exists. */ | ||
557 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | ||
558 | goto out; | ||
559 | |||
560 | /* | ||
561 | * Create a UDP header | ||
562 | */ | ||
563 | uh = udp_hdr(skb); | ||
564 | uh->source = fl->fl_ip_sport; | ||
565 | uh->dest = fl->fl_ip_dport; | ||
566 | uh->len = htons(up->len); | ||
567 | uh->check = 0; | ||
568 | |||
569 | if (is_udplite) | ||
570 | csum = udplite_csum_outgoing(sk, skb); | ||
571 | else | ||
572 | csum = udp_csum_outgoing(sk, skb); | ||
573 | |||
574 | /* add protocol-dependent pseudo-header */ | ||
575 | uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst, | ||
576 | up->len, fl->proto, csum ); | ||
577 | if (uh->check == 0) | ||
578 | uh->check = CSUM_MANGLED_0; | ||
579 | |||
580 | err = ip6_push_pending_frames(sk); | ||
581 | out: | ||
582 | up->len = 0; | ||
583 | up->pending = 0; | ||
584 | if (!err) | ||
585 | UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite); | ||
586 | return err; | ||
587 | } | ||
588 | |||
589 | int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, | ||
590 | struct msghdr *msg, size_t len) | ||
591 | { | ||
592 | struct ipv6_txoptions opt_space; | ||
593 | struct udp_sock *up = udp_sk(sk); | ||
594 | struct inet_sock *inet = inet_sk(sk); | ||
595 | struct ipv6_pinfo *np = inet6_sk(sk); | ||
596 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name; | ||
597 | struct in6_addr *daddr, *final_p = NULL, final; | ||
598 | struct ipv6_txoptions *opt = NULL; | ||
599 | struct ip6_flowlabel *flowlabel = NULL; | ||
600 | struct flowi fl; | ||
601 | struct dst_entry *dst; | ||
602 | int addr_len = msg->msg_namelen; | ||
603 | int ulen = len; | ||
604 | int hlimit = -1; | ||
605 | int tclass = -1; | ||
606 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | ||
607 | int err; | ||
608 | int connected = 0; | ||
609 | int is_udplite = IS_UDPLITE(sk); | ||
610 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | ||
611 | |||
612 | /* destination address check */ | ||
613 | if (sin6) { | ||
614 | if (addr_len < offsetof(struct sockaddr, sa_data)) | ||
615 | return -EINVAL; | ||
616 | |||
617 | switch (sin6->sin6_family) { | ||
618 | case AF_INET6: | ||
619 | if (addr_len < SIN6_LEN_RFC2133) | ||
620 | return -EINVAL; | ||
621 | daddr = &sin6->sin6_addr; | ||
622 | break; | ||
623 | case AF_INET: | ||
624 | goto do_udp_sendmsg; | ||
625 | case AF_UNSPEC: | ||
626 | msg->msg_name = sin6 = NULL; | ||
627 | msg->msg_namelen = addr_len = 0; | ||
628 | daddr = NULL; | ||
629 | break; | ||
630 | default: | ||
631 | return -EINVAL; | ||
632 | } | ||
633 | } else if (!up->pending) { | ||
634 | if (sk->sk_state != TCP_ESTABLISHED) | ||
635 | return -EDESTADDRREQ; | ||
636 | daddr = &np->daddr; | ||
637 | } else | ||
638 | daddr = NULL; | ||
639 | |||
640 | if (daddr) { | ||
641 | if (ipv6_addr_v4mapped(daddr)) { | ||
642 | struct sockaddr_in sin; | ||
643 | sin.sin_family = AF_INET; | ||
644 | sin.sin_port = sin6 ? sin6->sin6_port : inet->dport; | ||
645 | sin.sin_addr.s_addr = daddr->s6_addr32[3]; | ||
646 | msg->msg_name = &sin; | ||
647 | msg->msg_namelen = sizeof(sin); | ||
648 | do_udp_sendmsg: | ||
649 | if (__ipv6_only_sock(sk)) | ||
650 | return -ENETUNREACH; | ||
651 | return udp_sendmsg(iocb, sk, msg, len); | ||
652 | } | ||
653 | } | ||
654 | |||
655 | if (up->pending == AF_INET) | ||
656 | return udp_sendmsg(iocb, sk, msg, len); | ||
657 | |||
658 | /* Rough check on arithmetic overflow, | ||
659 | better check is made in ip6_append_data(). | ||
660 | */ | ||
661 | if (len > INT_MAX - sizeof(struct udphdr)) | ||
662 | return -EMSGSIZE; | ||
663 | |||
664 | if (up->pending) { | ||
665 | /* | ||
666 | * There are pending frames. | ||
667 | * The socket lock must be held while it's corked. | ||
668 | */ | ||
669 | lock_sock(sk); | ||
670 | if (likely(up->pending)) { | ||
671 | if (unlikely(up->pending != AF_INET6)) { | ||
672 | release_sock(sk); | ||
673 | return -EAFNOSUPPORT; | ||
674 | } | ||
675 | dst = NULL; | ||
676 | goto do_append_data; | ||
677 | } | ||
678 | release_sock(sk); | ||
679 | } | ||
680 | ulen += sizeof(struct udphdr); | ||
681 | |||
682 | memset(&fl, 0, sizeof(fl)); | ||
683 | |||
684 | if (sin6) { | ||
685 | if (sin6->sin6_port == 0) | ||
686 | return -EINVAL; | ||
687 | |||
688 | fl.fl_ip_dport = sin6->sin6_port; | ||
689 | daddr = &sin6->sin6_addr; | ||
690 | |||
691 | if (np->sndflow) { | ||
692 | fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; | ||
693 | if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { | ||
694 | flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); | ||
695 | if (flowlabel == NULL) | ||
696 | return -EINVAL; | ||
697 | daddr = &flowlabel->dst; | ||
698 | } | ||
699 | } | ||
700 | |||
701 | /* | ||
702 | * Otherwise it will be difficult to maintain | ||
703 | * sk->sk_dst_cache. | ||
704 | */ | ||
705 | if (sk->sk_state == TCP_ESTABLISHED && | ||
706 | ipv6_addr_equal(daddr, &np->daddr)) | ||
707 | daddr = &np->daddr; | ||
708 | |||
709 | if (addr_len >= sizeof(struct sockaddr_in6) && | ||
710 | sin6->sin6_scope_id && | ||
711 | ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) | ||
712 | fl.oif = sin6->sin6_scope_id; | ||
713 | } else { | ||
714 | if (sk->sk_state != TCP_ESTABLISHED) | ||
715 | return -EDESTADDRREQ; | ||
716 | |||
717 | fl.fl_ip_dport = inet->dport; | ||
718 | daddr = &np->daddr; | ||
719 | fl.fl6_flowlabel = np->flow_label; | ||
720 | connected = 1; | ||
721 | } | ||
722 | |||
723 | if (!fl.oif) | ||
724 | fl.oif = sk->sk_bound_dev_if; | ||
725 | |||
726 | if (msg->msg_controllen) { | ||
727 | opt = &opt_space; | ||
728 | memset(opt, 0, sizeof(struct ipv6_txoptions)); | ||
729 | opt->tot_len = sizeof(*opt); | ||
730 | |||
731 | err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass); | ||
732 | if (err < 0) { | ||
733 | fl6_sock_release(flowlabel); | ||
734 | return err; | ||
735 | } | ||
736 | if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { | ||
737 | flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); | ||
738 | if (flowlabel == NULL) | ||
739 | return -EINVAL; | ||
740 | } | ||
741 | if (!(opt->opt_nflen|opt->opt_flen)) | ||
742 | opt = NULL; | ||
743 | connected = 0; | ||
744 | } | ||
745 | if (opt == NULL) | ||
746 | opt = np->opt; | ||
747 | if (flowlabel) | ||
748 | opt = fl6_merge_options(&opt_space, flowlabel, opt); | ||
749 | opt = ipv6_fixup_options(&opt_space, opt); | ||
750 | |||
751 | fl.proto = sk->sk_protocol; | ||
752 | ipv6_addr_copy(&fl.fl6_dst, daddr); | ||
753 | if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) | ||
754 | ipv6_addr_copy(&fl.fl6_src, &np->saddr); | ||
755 | fl.fl_ip_sport = inet->sport; | ||
756 | |||
757 | /* merge ip6_build_xmit from ip6_output */ | ||
758 | if (opt && opt->srcrt) { | ||
759 | struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; | ||
760 | ipv6_addr_copy(&final, &fl.fl6_dst); | ||
761 | ipv6_addr_copy(&fl.fl6_dst, rt0->addr); | ||
762 | final_p = &final; | ||
763 | connected = 0; | ||
764 | } | ||
765 | |||
766 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) { | ||
767 | fl.oif = np->mcast_oif; | ||
768 | connected = 0; | ||
769 | } | ||
770 | |||
771 | security_sk_classify_flow(sk, &fl); | ||
772 | |||
773 | err = ip6_sk_dst_lookup(sk, &dst, &fl); | ||
774 | if (err) | ||
775 | goto out; | ||
776 | if (final_p) | ||
777 | ipv6_addr_copy(&fl.fl6_dst, final_p); | ||
778 | |||
779 | if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) { | ||
780 | if (err == -EREMOTE) | ||
781 | err = ip6_dst_blackhole(sk, &dst, &fl); | ||
782 | if (err < 0) | ||
783 | goto out; | ||
784 | } | ||
785 | |||
786 | if (hlimit < 0) { | ||
787 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | ||
788 | hlimit = np->mcast_hops; | ||
789 | else | ||
790 | hlimit = np->hop_limit; | ||
791 | if (hlimit < 0) | ||
792 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | ||
793 | if (hlimit < 0) | ||
794 | hlimit = ipv6_get_hoplimit(dst->dev); | ||
795 | } | ||
796 | |||
797 | if (tclass < 0) { | ||
798 | tclass = np->tclass; | ||
799 | if (tclass < 0) | ||
800 | tclass = 0; | ||
801 | } | ||
802 | |||
803 | if (msg->msg_flags&MSG_CONFIRM) | ||
804 | goto do_confirm; | ||
805 | back_from_confirm: | ||
806 | |||
807 | lock_sock(sk); | ||
808 | if (unlikely(up->pending)) { | ||
809 | /* The socket is already corked while preparing it. */ | ||
810 | /* ... which is an evident application bug. --ANK */ | ||
811 | release_sock(sk); | ||
812 | |||
813 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); | ||
814 | err = -EINVAL; | ||
815 | goto out; | ||
816 | } | ||
817 | |||
818 | up->pending = AF_INET6; | ||
819 | |||
820 | do_append_data: | ||
821 | up->len += ulen; | ||
822 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; | ||
823 | err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen, | ||
824 | sizeof(struct udphdr), hlimit, tclass, opt, &fl, | ||
825 | (struct rt6_info*)dst, | ||
826 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | ||
827 | if (err) | ||
828 | udp_v6_flush_pending_frames(sk); | ||
829 | else if (!corkreq) | ||
830 | err = udp_v6_push_pending_frames(sk); | ||
831 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | ||
832 | up->pending = 0; | ||
833 | |||
834 | if (dst) { | ||
835 | if (connected) { | ||
836 | ip6_dst_store(sk, dst, | ||
837 | ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ? | ||
838 | &np->daddr : NULL, | ||
839 | #ifdef CONFIG_IPV6_SUBTREES | ||
840 | ipv6_addr_equal(&fl.fl6_src, &np->saddr) ? | ||
841 | &np->saddr : | ||
842 | #endif | ||
843 | NULL); | ||
844 | } else { | ||
845 | dst_release(dst); | ||
846 | } | ||
847 | } | ||
848 | |||
849 | if (err > 0) | ||
850 | err = np->recverr ? net_xmit_errno(err) : 0; | ||
851 | release_sock(sk); | ||
852 | out: | ||
853 | fl6_sock_release(flowlabel); | ||
854 | if (!err) | ||
855 | return len; | ||
856 | /* | ||
857 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | ||
858 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | ||
859 | * we don't have a good statistic (IpOutDiscards but it can be too many | ||
860 | * things). We could add another new stat but at least for now that | ||
861 | * seems like overkill. | ||
862 | */ | ||
863 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | ||
864 | UDP6_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite); | ||
865 | } | ||
866 | return err; | ||
867 | |||
868 | do_confirm: | ||
869 | dst_confirm(dst); | ||
870 | if (!(msg->msg_flags&MSG_PROBE) || len) | ||
871 | goto back_from_confirm; | ||
872 | err = 0; | ||
873 | goto out; | ||
874 | } | ||
875 | |||
876 | int udpv6_destroy_sock(struct sock *sk) | ||
877 | { | ||
878 | lock_sock(sk); | ||
879 | udp_v6_flush_pending_frames(sk); | ||
880 | release_sock(sk); | ||
881 | |||
882 | inet6_destroy_sock(sk); | ||
883 | |||
884 | return 0; | ||
885 | } | ||
886 | |||
887 | /* | ||
888 | * Socket option code for UDP | ||
889 | */ | ||
890 | int udpv6_setsockopt(struct sock *sk, int level, int optname, | ||
891 | char __user *optval, int optlen) | ||
892 | { | ||
893 | if (level == SOL_UDP || level == SOL_UDPLITE) | ||
894 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | ||
895 | udp_v6_push_pending_frames); | ||
896 | return ipv6_setsockopt(sk, level, optname, optval, optlen); | ||
897 | } | ||
898 | |||
899 | #ifdef CONFIG_COMPAT | ||
900 | int compat_udpv6_setsockopt(struct sock *sk, int level, int optname, | ||
901 | char __user *optval, int optlen) | ||
902 | { | ||
903 | if (level == SOL_UDP || level == SOL_UDPLITE) | ||
904 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | ||
905 | udp_v6_push_pending_frames); | ||
906 | return compat_ipv6_setsockopt(sk, level, optname, optval, optlen); | ||
907 | } | ||
908 | #endif | ||
909 | |||
910 | int udpv6_getsockopt(struct sock *sk, int level, int optname, | ||
911 | char __user *optval, int __user *optlen) | ||
912 | { | ||
913 | if (level == SOL_UDP || level == SOL_UDPLITE) | ||
914 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | ||
915 | return ipv6_getsockopt(sk, level, optname, optval, optlen); | ||
916 | } | ||
917 | |||
918 | #ifdef CONFIG_COMPAT | ||
919 | int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, | ||
920 | char __user *optval, int __user *optlen) | ||
921 | { | ||
922 | if (level == SOL_UDP || level == SOL_UDPLITE) | ||
923 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | ||
924 | return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); | ||
925 | } | ||
926 | #endif | ||
927 | |||
928 | static struct inet6_protocol udpv6_protocol = { | ||
929 | .handler = udpv6_rcv, | ||
930 | .err_handler = udpv6_err, | ||
931 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | ||
932 | }; | ||
933 | |||
934 | /* ------------------------------------------------------------------------ */ | ||
935 | #ifdef CONFIG_PROC_FS | ||
936 | |||
937 | static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket) | ||
938 | { | ||
939 | struct inet_sock *inet = inet_sk(sp); | ||
940 | struct ipv6_pinfo *np = inet6_sk(sp); | ||
941 | struct in6_addr *dest, *src; | ||
942 | __u16 destp, srcp; | ||
943 | |||
944 | dest = &np->daddr; | ||
945 | src = &np->rcv_saddr; | ||
946 | destp = ntohs(inet->dport); | ||
947 | srcp = ntohs(inet->sport); | ||
948 | seq_printf(seq, | ||
949 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | ||
950 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n", | ||
951 | bucket, | ||
952 | src->s6_addr32[0], src->s6_addr32[1], | ||
953 | src->s6_addr32[2], src->s6_addr32[3], srcp, | ||
954 | dest->s6_addr32[0], dest->s6_addr32[1], | ||
955 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | ||
956 | sp->sk_state, | ||
957 | atomic_read(&sp->sk_wmem_alloc), | ||
958 | atomic_read(&sp->sk_rmem_alloc), | ||
959 | 0, 0L, 0, | ||
960 | sock_i_uid(sp), 0, | ||
961 | sock_i_ino(sp), | ||
962 | atomic_read(&sp->sk_refcnt), sp); | ||
963 | } | ||
964 | |||
965 | int udp6_seq_show(struct seq_file *seq, void *v) | ||
966 | { | ||
967 | if (v == SEQ_START_TOKEN) | ||
968 | seq_printf(seq, | ||
969 | " sl " | ||
970 | "local_address " | ||
971 | "remote_address " | ||
972 | "st tx_queue rx_queue tr tm->when retrnsmt" | ||
973 | " uid timeout inode\n"); | ||
974 | else | ||
975 | udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket); | ||
976 | return 0; | ||
977 | } | ||
978 | |||
979 | static struct file_operations udp6_seq_fops; | ||
980 | static struct udp_seq_afinfo udp6_seq_afinfo = { | ||
981 | .owner = THIS_MODULE, | ||
982 | .name = "udp6", | ||
983 | .family = AF_INET6, | ||
984 | .hashtable = udp_hash, | ||
985 | .seq_show = udp6_seq_show, | ||
986 | .seq_fops = &udp6_seq_fops, | ||
987 | }; | ||
988 | |||
989 | int __init udp6_proc_init(void) | ||
990 | { | ||
991 | return udp_proc_register(&udp6_seq_afinfo); | ||
992 | } | ||
993 | |||
994 | void udp6_proc_exit(void) { | ||
995 | udp_proc_unregister(&udp6_seq_afinfo); | ||
996 | } | ||
997 | #endif /* CONFIG_PROC_FS */ | ||
998 | |||
999 | /* ------------------------------------------------------------------------ */ | ||
1000 | |||
1001 | DEFINE_PROTO_INUSE(udpv6) | ||
1002 | |||
1003 | struct proto udpv6_prot = { | ||
1004 | .name = "UDPv6", | ||
1005 | .owner = THIS_MODULE, | ||
1006 | .close = udp_lib_close, | ||
1007 | .connect = ip6_datagram_connect, | ||
1008 | .disconnect = udp_disconnect, | ||
1009 | .ioctl = udp_ioctl, | ||
1010 | .destroy = udpv6_destroy_sock, | ||
1011 | .setsockopt = udpv6_setsockopt, | ||
1012 | .getsockopt = udpv6_getsockopt, | ||
1013 | .sendmsg = udpv6_sendmsg, | ||
1014 | .recvmsg = udpv6_recvmsg, | ||
1015 | .backlog_rcv = udpv6_queue_rcv_skb, | ||
1016 | .hash = udp_lib_hash, | ||
1017 | .unhash = udp_lib_unhash, | ||
1018 | .get_port = udp_v6_get_port, | ||
1019 | .memory_allocated = &udp_memory_allocated, | ||
1020 | .sysctl_mem = sysctl_udp_mem, | ||
1021 | .sysctl_wmem = &sysctl_udp_wmem_min, | ||
1022 | .sysctl_rmem = &sysctl_udp_rmem_min, | ||
1023 | .obj_size = sizeof(struct udp6_sock), | ||
1024 | #ifdef CONFIG_COMPAT | ||
1025 | .compat_setsockopt = compat_udpv6_setsockopt, | ||
1026 | .compat_getsockopt = compat_udpv6_getsockopt, | ||
1027 | #endif | ||
1028 | REF_PROTO_INUSE(udpv6) | ||
1029 | }; | ||
1030 | |||
1031 | static struct inet_protosw udpv6_protosw = { | ||
1032 | .type = SOCK_DGRAM, | ||
1033 | .protocol = IPPROTO_UDP, | ||
1034 | .prot = &udpv6_prot, | ||
1035 | .ops = &inet6_dgram_ops, | ||
1036 | .capability =-1, | ||
1037 | .no_check = UDP_CSUM_DEFAULT, | ||
1038 | .flags = INET_PROTOSW_PERMANENT, | ||
1039 | }; | ||
1040 | |||
1041 | |||
1042 | int __init udpv6_init(void) | ||
1043 | { | ||
1044 | int ret; | ||
1045 | |||
1046 | ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); | ||
1047 | if (ret) | ||
1048 | goto out; | ||
1049 | |||
1050 | ret = inet6_register_protosw(&udpv6_protosw); | ||
1051 | if (ret) | ||
1052 | goto out_udpv6_protocol; | ||
1053 | out: | ||
1054 | return ret; | ||
1055 | |||
1056 | out_udpv6_protocol: | ||
1057 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | ||
1058 | goto out; | ||
1059 | } | ||
1060 | |||
1061 | void udpv6_exit(void) | ||
1062 | { | ||
1063 | inet6_unregister_protosw(&udpv6_protosw); | ||
1064 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | ||
1065 | } | ||