?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
udp: properly support MSG_PEEK with truncated buffers
[linux-2.6/btrfs-unstable.git] / net / ipv6 / udp.c
blob00775ee27d86daabb65b8e86bc80959890293d16
1 /*
2 * UDP over IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Based on linux/ipv4/udp.c
9 *
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
14 * a single port at the same time.
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 */
23
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/ipv6.h>
33 #include <linux/icmpv6.h>
34 #include <linux/init.h>
35 #include <linux/module.h>
36 #include <linux/skbuff.h>
37 #include <linux/slab.h>
38 #include <asm/uaccess.h>
39
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_route.h>
44 #include <net/raw.h>
45 #include <net/tcp_states.h>
46 #include <net/ip6_checksum.h>
47 #include <net/xfrm.h>
48 #include <net/inet6_hashtables.h>
49 #include <net/busy_poll.h>
50
51 #include <linux/proc_fs.h>
52 #include <linux/seq_file.h>
53 #include <trace/events/skb.h>
54 #include "udp_impl.h"
55
56 static u32 udp6_ehashfn(const struct net *net,
57 const struct in6_addr *laddr,
58 const u16 lport,
59 const struct in6_addr *faddr,
60 const __be16 fport)
61 {
62 static u32 udp6_ehash_secret __read_mostly;
63 static u32 udp_ipv6_hash_secret __read_mostly;
64
65 u32 lhash, fhash;
66
67 net_get_random_once(&udp6_ehash_secret,
68 sizeof(udp6_ehash_secret));
69 net_get_random_once(&udp_ipv6_hash_secret,
70 sizeof(udp_ipv6_hash_secret));
71
72 lhash = (__force u32)laddr->s6_addr32[3];
73 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
74
75 return __inet6_ehashfn(lhash, lport, fhash, fport,
76 udp_ipv6_hash_secret + net_hash_mix(net));
77 }
78
79 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
80 {
81 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
82 int sk2_ipv6only = inet_v6_ipv6only(sk2);
83 int addr_type = ipv6_addr_type(&sk->sk_v6_rcv_saddr);
84 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
85
86 /* if both are mapped, treat as IPv4 */
87 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
88 return (!sk2_ipv6only &&
89 (!sk->sk_rcv_saddr || !sk2->sk_rcv_saddr ||
90 sk->sk_rcv_saddr == sk2->sk_rcv_saddr));
91
92 if (addr_type2 == IPV6_ADDR_ANY &&
93 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
94 return 1;
95
96 if (addr_type == IPV6_ADDR_ANY &&
97 !(ipv6_only_sock(sk) && addr_type2 == IPV6_ADDR_MAPPED))
98 return 1;
99
100 if (sk2_rcv_saddr6 &&
101 ipv6_addr_equal(&sk->sk_v6_rcv_saddr, sk2_rcv_saddr6))
102 return 1;
103
104 return 0;
105 }
106
107 static u32 udp6_portaddr_hash(const struct net *net,
108 const struct in6_addr *addr6,
109 unsigned int port)
110 {
111 unsigned int hash, mix = net_hash_mix(net);
112
113 if (ipv6_addr_any(addr6))
114 hash = jhash_1word(0, mix);
115 else if (ipv6_addr_v4mapped(addr6))
116 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
117 else
118 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
119
120 return hash ^ port;
121 }
122
123 int udp_v6_get_port(struct sock *sk, unsigned short snum)
124 {
125 unsigned int hash2_nulladdr =
126 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
127 unsigned int hash2_partial =
128 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
129
130 /* precompute partial secondary hash */
131 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
132 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
133 }
134
135 static void udp_v6_rehash(struct sock *sk)
136 {
137 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
138 &sk->sk_v6_rcv_saddr,
139 inet_sk(sk)->inet_num);
140
141 udp_lib_rehash(sk, new_hash);
142 }
143
144 static inline int compute_score(struct sock *sk, struct net *net,
145 unsigned short hnum,
146 const struct in6_addr *saddr, __be16 sport,
147 const struct in6_addr *daddr, __be16 dport,
148 int dif)
149 {
150 int score;
151 struct inet_sock *inet;
152
153 if (!net_eq(sock_net(sk), net) ||
154 udp_sk(sk)->udp_port_hash != hnum ||
155 sk->sk_family != PF_INET6)
156 return -1;
157
158 score = 0;
159 inet = inet_sk(sk);
160
161 if (inet->inet_dport) {
162 if (inet->inet_dport != sport)
163 return -1;
164 score++;
165 }
166
167 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
168 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
169 return -1;
170 score++;
171 }
172
173 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
174 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
175 return -1;
176 score++;
177 }
178
179 if (sk->sk_bound_dev_if) {
180 if (sk->sk_bound_dev_if != dif)
181 return -1;
182 score++;
183 }
184
185 if (sk->sk_incoming_cpu == raw_smp_processor_id())
186 score++;
187
188 return score;
189 }
190
191 static inline int compute_score2(struct sock *sk, struct net *net,
192 const struct in6_addr *saddr, __be16 sport,
193 const struct in6_addr *daddr,
194 unsigned short hnum, int dif)
195 {
196 int score;
197 struct inet_sock *inet;
198
199 if (!net_eq(sock_net(sk), net) ||
200 udp_sk(sk)->udp_port_hash != hnum ||
201 sk->sk_family != PF_INET6)
202 return -1;
203
204 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
205 return -1;
206
207 score = 0;
208 inet = inet_sk(sk);
209
210 if (inet->inet_dport) {
211 if (inet->inet_dport != sport)
212 return -1;
213 score++;
214 }
215
216 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
217 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
218 return -1;
219 score++;
220 }
221
222 if (sk->sk_bound_dev_if) {
223 if (sk->sk_bound_dev_if != dif)
224 return -1;
225 score++;
226 }
227
228 if (sk->sk_incoming_cpu == raw_smp_processor_id())
229 score++;
230
231 return score;
232 }
233
234 /* called with read_rcu_lock() */
235 static struct sock *udp6_lib_lookup2(struct net *net,
236 const struct in6_addr *saddr, __be16 sport,
237 const struct in6_addr *daddr, unsigned int hnum, int dif,
238 struct udp_hslot *hslot2, unsigned int slot2)
239 {
240 struct sock *sk, *result;
241 struct hlist_nulls_node *node;
242 int score, badness, matches = 0, reuseport = 0;
243 u32 hash = 0;
244
245 begin:
246 result = NULL;
247 badness = -1;
248 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
249 score = compute_score2(sk, net, saddr, sport,
250 daddr, hnum, dif);
251 if (score > badness) {
252 result = sk;
253 badness = score;
254 reuseport = sk->sk_reuseport;
255 if (reuseport) {
256 hash = udp6_ehashfn(net, daddr, hnum,
257 saddr, sport);
258 matches = 1;
259 }
260 } else if (score == badness && reuseport) {
261 matches++;
262 if (reciprocal_scale(hash, matches) == 0)
263 result = sk;
264 hash = next_pseudo_random32(hash);
265 }
266 }
267 /*
268 * if the nulls value we got at the end of this lookup is
269 * not the expected one, we must restart lookup.
270 * We probably met an item that was moved to another chain.
271 */
272 if (get_nulls_value(node) != slot2)
273 goto begin;
274
275 if (result) {
276 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
277 result = NULL;
278 else if (unlikely(compute_score2(result, net, saddr, sport,
279 daddr, hnum, dif) < badness)) {
280 sock_put(result);
281 goto begin;
282 }
283 }
284 return result;
285 }
286
287 struct sock *__udp6_lib_lookup(struct net *net,
288 const struct in6_addr *saddr, __be16 sport,
289 const struct in6_addr *daddr, __be16 dport,
290 int dif, struct udp_table *udptable)
291 {
292 struct sock *sk, *result;
293 struct hlist_nulls_node *node;
294 unsigned short hnum = ntohs(dport);
295 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
296 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
297 int score, badness, matches = 0, reuseport = 0;
298 u32 hash = 0;
299
300 rcu_read_lock();
301 if (hslot->count > 10) {
302 hash2 = udp6_portaddr_hash(net, daddr, hnum);
303 slot2 = hash2 & udptable->mask;
304 hslot2 = &udptable->hash2[slot2];
305 if (hslot->count < hslot2->count)
306 goto begin;
307
308 result = udp6_lib_lookup2(net, saddr, sport,
309 daddr, hnum, dif,
310 hslot2, slot2);
311 if (!result) {
312 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
313 slot2 = hash2 & udptable->mask;
314 hslot2 = &udptable->hash2[slot2];
315 if (hslot->count < hslot2->count)
316 goto begin;
317
318 result = udp6_lib_lookup2(net, saddr, sport,
319 &in6addr_any, hnum, dif,
320 hslot2, slot2);
321 }
322 rcu_read_unlock();
323 return result;
324 }
325 begin:
326 result = NULL;
327 badness = -1;
328 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
329 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
330 if (score > badness) {
331 result = sk;
332 badness = score;
333 reuseport = sk->sk_reuseport;
334 if (reuseport) {
335 hash = udp6_ehashfn(net, daddr, hnum,
336 saddr, sport);
337 matches = 1;
338 }
339 } else if (score == badness && reuseport) {
340 matches++;
341 if (reciprocal_scale(hash, matches) == 0)
342 result = sk;
343 hash = next_pseudo_random32(hash);
344 }
345 }
346 /*
347 * if the nulls value we got at the end of this lookup is
348 * not the expected one, we must restart lookup.
349 * We probably met an item that was moved to another chain.
350 */
351 if (get_nulls_value(node) != slot)
352 goto begin;
353
354 if (result) {
355 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
356 result = NULL;
357 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
358 daddr, dport, dif) < badness)) {
359 sock_put(result);
360 goto begin;
361 }
362 }
363 rcu_read_unlock();
364 return result;
365 }
366 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
367
368 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
369 __be16 sport, __be16 dport,
370 struct udp_table *udptable)
371 {
372 struct sock *sk;
373 const struct ipv6hdr *iph = ipv6_hdr(skb);
374
375 sk = skb_steal_sock(skb);
376 if (unlikely(sk))
377 return sk;
378 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
379 &iph->daddr, dport, inet6_iif(skb),
380 udptable);
381 }
382
383 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
384 const struct in6_addr *daddr, __be16 dport, int dif)
385 {
386 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
387 }
388 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
389
390 /*
391 * This should be easy, if there is something there we
392 * return it, otherwise we block.
393 */
394
395 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
396 int noblock, int flags, int *addr_len)
397 {
398 struct ipv6_pinfo *np = inet6_sk(sk);
399 struct inet_sock *inet = inet_sk(sk);
400 struct sk_buff *skb;
401 unsigned int ulen, copied;
402 int peeked, off = 0;
403 int err;
404 int is_udplite = IS_UDPLITE(sk);
405 bool checksum_valid = false;
406 int is_udp4;
407 bool slow;
408
409 if (flags & MSG_ERRQUEUE)
410 return ipv6_recv_error(sk, msg, len, addr_len);
411
412 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
413 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
414
415 try_again:
416 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
417 &peeked, &off, &err);
418 if (!skb)
419 goto out;
420
421 ulen = skb->len - sizeof(struct udphdr);
422 copied = len;
423 if (copied > ulen)
424 copied = ulen;
425 else if (copied < ulen)
426 msg->msg_flags |= MSG_TRUNC;
427
428 is_udp4 = (skb->protocol == htons(ETH_P_IP));
429
430 /*
431 * If checksum is needed at all, try to do it while copying the
432 * data. If the data is truncated, or if we only want a partial
433 * coverage checksum (UDP-Lite), do it before the copy.
434 */
435
436 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
437 checksum_valid = !udp_lib_checksum_complete(skb);
438 if (!checksum_valid)
439 goto csum_copy_err;
440 }
441
442 if (checksum_valid || skb_csum_unnecessary(skb))
443 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
444 msg, copied);
445 else {
446 err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), msg);
447 if (err == -EINVAL)
448 goto csum_copy_err;
449 }
450 if (unlikely(err)) {
451 trace_kfree_skb(skb, udpv6_recvmsg);
452 if (!peeked) {
453 atomic_inc(&sk->sk_drops);
454 if (is_udp4)
455 UDP_INC_STATS_USER(sock_net(sk),
456 UDP_MIB_INERRORS,
457 is_udplite);
458 else
459 UDP6_INC_STATS_USER(sock_net(sk),
460 UDP_MIB_INERRORS,
461 is_udplite);
462 }
463 goto out_free;
464 }
465 if (!peeked) {
466 if (is_udp4)
467 UDP_INC_STATS_USER(sock_net(sk),
468 UDP_MIB_INDATAGRAMS, is_udplite);
469 else
470 UDP6_INC_STATS_USER(sock_net(sk),
471 UDP_MIB_INDATAGRAMS, is_udplite);
472 }
473
474 sock_recv_ts_and_drops(msg, sk, skb);
475
476 /* Copy the address. */
477 if (msg->msg_name) {
478 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
479 sin6->sin6_family = AF_INET6;
480 sin6->sin6_port = udp_hdr(skb)->source;
481 sin6->sin6_flowinfo = 0;
482
483 if (is_udp4) {
484 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
485 &sin6->sin6_addr);
486 sin6->sin6_scope_id = 0;
487 } else {
488 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
489 sin6->sin6_scope_id =
490 ipv6_iface_scope_id(&sin6->sin6_addr,
491 inet6_iif(skb));
492 }
493 *addr_len = sizeof(*sin6);
494 }
495
496 if (np->rxopt.all)
497 ip6_datagram_recv_common_ctl(sk, msg, skb);
498
499 if (is_udp4) {
500 if (inet->cmsg_flags)
501 ip_cmsg_recv(msg, skb);
502 } else {
503 if (np->rxopt.all)
504 ip6_datagram_recv_specific_ctl(sk, msg, skb);
505 }
506
507 err = copied;
508 if (flags & MSG_TRUNC)
509 err = ulen;
510
511 out_free:
512 skb_free_datagram_locked(sk, skb);
513 out:
514 return err;
515
516 csum_copy_err:
517 slow = lock_sock_fast(sk);
518 if (!skb_kill_datagram(sk, skb, flags)) {
519 if (is_udp4) {
520 UDP_INC_STATS_USER(sock_net(sk),
521 UDP_MIB_CSUMERRORS, is_udplite);
522 UDP_INC_STATS_USER(sock_net(sk),
523 UDP_MIB_INERRORS, is_udplite);
524 } else {
525 UDP6_INC_STATS_USER(sock_net(sk),
526 UDP_MIB_CSUMERRORS, is_udplite);
527 UDP6_INC_STATS_USER(sock_net(sk),
528 UDP_MIB_INERRORS, is_udplite);
529 }
530 }
531 unlock_sock_fast(sk, slow);
532
533 /* starting over for a new packet, but check if we need to yield */
534 cond_resched();
535 msg->msg_flags &= ~MSG_TRUNC;
536 goto try_again;
537 }
538
539 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
540 u8 type, u8 code, int offset, __be32 info,
541 struct udp_table *udptable)
542 {
543 struct ipv6_pinfo *np;
544 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
545 const struct in6_addr *saddr = &hdr->saddr;
546 const struct in6_addr *daddr = &hdr->daddr;
547 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
548 struct sock *sk;
549 int err;
550 struct net *net = dev_net(skb->dev);
551
552 sk = __udp6_lib_lookup(net, daddr, uh->dest,
553 saddr, uh->source, inet6_iif(skb), udptable);
554 if (!sk) {
555 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
556 ICMP6_MIB_INERRORS);
557 return;
558 }
559
560 if (type == ICMPV6_PKT_TOOBIG) {
561 if (!ip6_sk_accept_pmtu(sk))
562 goto out;
563 ip6_sk_update_pmtu(skb, sk, info);
564 }
565 if (type == NDISC_REDIRECT) {
566 ip6_sk_redirect(skb, sk);
567 goto out;
568 }
569
570 np = inet6_sk(sk);
571
572 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
573 goto out;
574
575 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
576 goto out;
577
578 if (np->recverr)
579 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
580
581 sk->sk_err = err;
582 sk->sk_error_report(sk);
583 out:
584 sock_put(sk);
585 }
586
587 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
588 {
589 int rc;
590
591 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
592 sock_rps_save_rxhash(sk, skb);
593 sk_mark_napi_id(sk, skb);
594 sk_incoming_cpu_update(sk);
595 }
596
597 rc = sock_queue_rcv_skb(sk, skb);
598 if (rc < 0) {
599 int is_udplite = IS_UDPLITE(sk);
600
601 /* Note that an ENOMEM error is charged twice */
602 if (rc == -ENOMEM)
603 UDP6_INC_STATS_BH(sock_net(sk),
604 UDP_MIB_RCVBUFERRORS, is_udplite);
605 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
606 kfree_skb(skb);
607 return -1;
608 }
609 return 0;
610 }
611
612 static __inline__ void udpv6_err(struct sk_buff *skb,
613 struct inet6_skb_parm *opt, u8 type,
614 u8 code, int offset, __be32 info)
615 {
616 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
617 }
618
619 static struct static_key udpv6_encap_needed __read_mostly;
620 void udpv6_encap_enable(void)
621 {
622 if (!static_key_enabled(&udpv6_encap_needed))
623 static_key_slow_inc(&udpv6_encap_needed);
624 }
625 EXPORT_SYMBOL(udpv6_encap_enable);
626
627 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
628 {
629 struct udp_sock *up = udp_sk(sk);
630 int rc;
631 int is_udplite = IS_UDPLITE(sk);
632
633 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
634 goto drop;
635
636 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
637 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
638
639 /*
640 * This is an encapsulation socket so pass the skb to
641 * the socket's udp_encap_rcv() hook. Otherwise, just
642 * fall through and pass this up the UDP socket.
643 * up->encap_rcv() returns the following value:
644 * =0 if skb was successfully passed to the encap
645 * handler or was discarded by it.
646 * >0 if skb should be passed on to UDP.
647 * <0 if skb should be resubmitted as proto -N
648 */
649
650 /* if we're overly short, let UDP handle it */
651 encap_rcv = ACCESS_ONCE(up->encap_rcv);
652 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
653 int ret;
654
655 /* Verify checksum before giving to encap */
656 if (udp_lib_checksum_complete(skb))
657 goto csum_error;
658
659 ret = encap_rcv(sk, skb);
660 if (ret <= 0) {
661 UDP_INC_STATS_BH(sock_net(sk),
662 UDP_MIB_INDATAGRAMS,
663 is_udplite);
664 return -ret;
665 }
666 }
667
668 /* FALLTHROUGH -- it's a UDP Packet */
669 }
670
671 /*
672 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
673 */
674 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
675
676 if (up->pcrlen == 0) { /* full coverage was set */
677 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
678 UDP_SKB_CB(skb)->cscov, skb->len);
679 goto drop;
680 }
681 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
682 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
683 UDP_SKB_CB(skb)->cscov, up->pcrlen);
684 goto drop;
685 }
686 }
687
688 if (rcu_access_pointer(sk->sk_filter)) {
689 if (udp_lib_checksum_complete(skb))
690 goto csum_error;
691 }
692
693 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
694 UDP6_INC_STATS_BH(sock_net(sk),
695 UDP_MIB_RCVBUFERRORS, is_udplite);
696 goto drop;
697 }
698
699 skb_dst_drop(skb);
700
701 bh_lock_sock(sk);
702 rc = 0;
703 if (!sock_owned_by_user(sk))
704 rc = __udpv6_queue_rcv_skb(sk, skb);
705 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
706 bh_unlock_sock(sk);
707 goto drop;
708 }
709 bh_unlock_sock(sk);
710
711 return rc;
712
713 csum_error:
714 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
715 drop:
716 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
717 atomic_inc(&sk->sk_drops);
718 kfree_skb(skb);
719 return -1;
720 }
721
722 static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
723 __be16 loc_port, const struct in6_addr *loc_addr,
724 __be16 rmt_port, const struct in6_addr *rmt_addr,
725 int dif, unsigned short hnum)
726 {
727 struct inet_sock *inet = inet_sk(sk);
728
729 if (!net_eq(sock_net(sk), net))
730 return false;
731
732 if (udp_sk(sk)->udp_port_hash != hnum ||
733 sk->sk_family != PF_INET6 ||
734 (inet->inet_dport && inet->inet_dport != rmt_port) ||
735 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
736 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
737 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
738 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
739 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
740 return false;
741 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
742 return false;
743 return true;
744 }
745
746 static void flush_stack(struct sock **stack, unsigned int count,
747 struct sk_buff *skb, unsigned int final)
748 {
749 struct sk_buff *skb1 = NULL;
750 struct sock *sk;
751 unsigned int i;
752
753 for (i = 0; i < count; i++) {
754 sk = stack[i];
755 if (likely(!skb1))
756 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
757 if (!skb1) {
758 atomic_inc(&sk->sk_drops);
759 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
760 IS_UDPLITE(sk));
761 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
762 IS_UDPLITE(sk));
763 }
764
765 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
766 skb1 = NULL;
767 sock_put(sk);
768 }
769 if (unlikely(skb1))
770 kfree_skb(skb1);
771 }
772
773 static void udp6_csum_zero_error(struct sk_buff *skb)
774 {
775 /* RFC 2460 section 8.1 says that we SHOULD log
776 * this error. Well, it is reasonable.
777 */
778 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
779 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
780 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
781 }
782
783 /*
784 * Note: called only from the BH handler context,
785 * so we don't need to lock the hashes.
786 */
787 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
788 const struct in6_addr *saddr, const struct in6_addr *daddr,
789 struct udp_table *udptable, int proto)
790 {
791 struct sock *sk, *stack[256 / sizeof(struct sock *)];
792 const struct udphdr *uh = udp_hdr(skb);
793 struct hlist_nulls_node *node;
794 unsigned short hnum = ntohs(uh->dest);
795 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
796 int dif = inet6_iif(skb);
797 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
798 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
799 bool inner_flushed = false;
800
801 if (use_hash2) {
802 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
803 udp_table.mask;
804 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
805 start_lookup:
806 hslot = &udp_table.hash2[hash2];
807 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
808 }
809
810 spin_lock(&hslot->lock);
811 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
812 if (__udp_v6_is_mcast_sock(net, sk,
813 uh->dest, daddr,
814 uh->source, saddr,
815 dif, hnum) &&
816 /* If zero checksum and no_check is not on for
817 * the socket then skip it.
818 */
819 (uh->check || udp_sk(sk)->no_check6_rx)) {
820 if (unlikely(count == ARRAY_SIZE(stack))) {
821 flush_stack(stack, count, skb, ~0);
822 inner_flushed = true;
823 count = 0;
824 }
825 stack[count++] = sk;
826 sock_hold(sk);
827 }
828 }
829
830 spin_unlock(&hslot->lock);
831
832 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
833 if (use_hash2 && hash2 != hash2_any) {
834 hash2 = hash2_any;
835 goto start_lookup;
836 }
837
838 if (count) {
839 flush_stack(stack, count, skb, count - 1);
840 } else {
841 if (!inner_flushed)
842 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
843 proto == IPPROTO_UDPLITE);
844 consume_skb(skb);
845 }
846 return 0;
847 }
848
849 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
850 int proto)
851 {
852 struct net *net = dev_net(skb->dev);
853 struct sock *sk;
854 struct udphdr *uh;
855 const struct in6_addr *saddr, *daddr;
856 u32 ulen = 0;
857
858 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
859 goto discard;
860
861 saddr = &ipv6_hdr(skb)->saddr;
862 daddr = &ipv6_hdr(skb)->daddr;
863 uh = udp_hdr(skb);
864
865 ulen = ntohs(uh->len);
866 if (ulen > skb->len)
867 goto short_packet;
868
869 if (proto == IPPROTO_UDP) {
870 /* UDP validates ulen. */
871
872 /* Check for jumbo payload */
873 if (ulen == 0)
874 ulen = skb->len;
875
876 if (ulen < sizeof(*uh))
877 goto short_packet;
878
879 if (ulen < skb->len) {
880 if (pskb_trim_rcsum(skb, ulen))
881 goto short_packet;
882 saddr = &ipv6_hdr(skb)->saddr;
883 daddr = &ipv6_hdr(skb)->daddr;
884 uh = udp_hdr(skb);
885 }
886 }
887
888 if (udp6_csum_init(skb, uh, proto))
889 goto csum_error;
890
891 /*
892 * Multicast receive code
893 */
894 if (ipv6_addr_is_multicast(daddr))
895 return __udp6_lib_mcast_deliver(net, skb,
896 saddr, daddr, udptable, proto);
897
898 /* Unicast */
899
900 /*
901 * check socket cache ... must talk to Alan about his plans
902 * for sock caches... i'll skip this for now.
903 */
904 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
905 if (sk) {
906 int ret;
907
908 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
909 sock_put(sk);
910 udp6_csum_zero_error(skb);
911 goto csum_error;
912 }
913
914 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
915 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
916 ip6_compute_pseudo);
917
918 ret = udpv6_queue_rcv_skb(sk, skb);
919 sock_put(sk);
920
921 /* a return value > 0 means to resubmit the input, but
922 * it wants the return to be -protocol, or 0
923 */
924 if (ret > 0)
925 return -ret;
926
927 return 0;
928 }
929
930 if (!uh->check) {
931 udp6_csum_zero_error(skb);
932 goto csum_error;
933 }
934
935 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
936 goto discard;
937
938 if (udp_lib_checksum_complete(skb))
939 goto csum_error;
940
941 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
942 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
943
944 kfree_skb(skb);
945 return 0;
946
947 short_packet:
948 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
949 proto == IPPROTO_UDPLITE ? "-Lite" : "",
950 saddr, ntohs(uh->source),
951 ulen, skb->len,
952 daddr, ntohs(uh->dest));
953 goto discard;
954 csum_error:
955 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
956 discard:
957 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
958 kfree_skb(skb);
959 return 0;
960 }
961
962 static __inline__ int udpv6_rcv(struct sk_buff *skb)
963 {
964 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
965 }
966
967 /*
968 * Throw away all pending data and cancel the corking. Socket is locked.
969 */
970 static void udp_v6_flush_pending_frames(struct sock *sk)
971 {
972 struct udp_sock *up = udp_sk(sk);
973
974 if (up->pending == AF_INET)
975 udp_flush_pending_frames(sk);
976 else if (up->pending) {
977 up->len = 0;
978 up->pending = 0;
979 ip6_flush_pending_frames(sk);
980 }
981 }
982
983 /**
984 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
985 * @sk: socket we are sending on
986 * @skb: sk_buff containing the filled-in UDP header
987 * (checksum field must be zeroed out)
988 */
989 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
990 const struct in6_addr *saddr,
991 const struct in6_addr *daddr, int len)
992 {
993 unsigned int offset;
994 struct udphdr *uh = udp_hdr(skb);
995 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
996 __wsum csum = 0;
997
998 if (!frags) {
999 /* Only one fragment on the socket. */
1000 skb->csum_start = skb_transport_header(skb) - skb->head;
1001 skb->csum_offset = offsetof(struct udphdr, check);
1002 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1003 } else {
1004 /*
1005 * HW-checksum won't work as there are two or more
1006 * fragments on the socket so that all csums of sk_buffs
1007 * should be together
1008 */
1009 offset = skb_transport_offset(skb);
1010 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1011
1012 skb->ip_summed = CHECKSUM_NONE;
1013
1014 do {
1015 csum = csum_add(csum, frags->csum);
1016 } while ((frags = frags->next));
1017
1018 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1019 csum);
1020 if (uh->check == 0)
1021 uh->check = CSUM_MANGLED_0;
1022 }
1023 }
1024
1025 /*
1026 * Sending
1027 */
1028
1029 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1030 {
1031 struct sock *sk = skb->sk;
1032 struct udphdr *uh;
1033 int err = 0;
1034 int is_udplite = IS_UDPLITE(sk);
1035 __wsum csum = 0;
1036 int offset = skb_transport_offset(skb);
1037 int len = skb->len - offset;
1038
1039 /*
1040 * Create a UDP header
1041 */
1042 uh = udp_hdr(skb);
1043 uh->source = fl6->fl6_sport;
1044 uh->dest = fl6->fl6_dport;
1045 uh->len = htons(len);
1046 uh->check = 0;
1047
1048 if (is_udplite)
1049 csum = udplite_csum(skb);
1050 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
1051 skb->ip_summed = CHECKSUM_NONE;
1052 goto send;
1053 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1054 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1055 goto send;
1056 } else
1057 csum = udp_csum(skb);
1058
1059 /* add protocol-dependent pseudo-header */
1060 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1061 len, fl6->flowi6_proto, csum);
1062 if (uh->check == 0)
1063 uh->check = CSUM_MANGLED_0;
1064
1065 send:
1066 err = ip6_send_skb(skb);
1067 if (err) {
1068 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1069 UDP6_INC_STATS_USER(sock_net(sk),
1070 UDP_MIB_SNDBUFERRORS, is_udplite);
1071 err = 0;
1072 }
1073 } else
1074 UDP6_INC_STATS_USER(sock_net(sk),
1075 UDP_MIB_OUTDATAGRAMS, is_udplite);
1076 return err;
1077 }
1078
1079 static int udp_v6_push_pending_frames(struct sock *sk)
1080 {
1081 struct sk_buff *skb;
1082 struct udp_sock *up = udp_sk(sk);
1083 struct flowi6 fl6;
1084 int err = 0;
1085
1086 if (up->pending == AF_INET)
1087 return udp_push_pending_frames(sk);
1088
1089 /* ip6_finish_skb will release the cork, so make a copy of
1090 * fl6 here.
1091 */
1092 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1093
1094 skb = ip6_finish_skb(sk);
1095 if (!skb)
1096 goto out;
1097
1098 err = udp_v6_send_skb(skb, &fl6);
1099
1100 out:
1101 up->len = 0;
1102 up->pending = 0;
1103 return err;
1104 }
1105
1106 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1107 {
1108 struct ipv6_txoptions opt_space;
1109 struct udp_sock *up = udp_sk(sk);
1110 struct inet_sock *inet = inet_sk(sk);
1111 struct ipv6_pinfo *np = inet6_sk(sk);
1112 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1113 struct in6_addr *daddr, *final_p, final;
1114 struct ipv6_txoptions *opt = NULL;
1115 struct ipv6_txoptions *opt_to_free = NULL;
1116 struct ip6_flowlabel *flowlabel = NULL;
1117 struct flowi6 fl6;
1118 struct dst_entry *dst;
1119 int addr_len = msg->msg_namelen;
1120 int ulen = len;
1121 int hlimit = -1;
1122 int tclass = -1;
1123 int dontfrag = -1;
1124 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1125 int err;
1126 int connected = 0;
1127 int is_udplite = IS_UDPLITE(sk);
1128 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1129
1130 /* destination address check */
1131 if (sin6) {
1132 if (addr_len < offsetof(struct sockaddr, sa_data))
1133 return -EINVAL;
1134
1135 switch (sin6->sin6_family) {
1136 case AF_INET6:
1137 if (addr_len < SIN6_LEN_RFC2133)
1138 return -EINVAL;
1139 daddr = &sin6->sin6_addr;
1140 break;
1141 case AF_INET:
1142 goto do_udp_sendmsg;
1143 case AF_UNSPEC:
1144 msg->msg_name = sin6 = NULL;
1145 msg->msg_namelen = addr_len = 0;
1146 daddr = NULL;
1147 break;
1148 default:
1149 return -EINVAL;
1150 }
1151 } else if (!up->pending) {
1152 if (sk->sk_state != TCP_ESTABLISHED)
1153 return -EDESTADDRREQ;
1154 daddr = &sk->sk_v6_daddr;
1155 } else
1156 daddr = NULL;
1157
1158 if (daddr) {
1159 if (ipv6_addr_v4mapped(daddr)) {
1160 struct sockaddr_in sin;
1161 sin.sin_family = AF_INET;
1162 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1163 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1164 msg->msg_name = &sin;
1165 msg->msg_namelen = sizeof(sin);
1166 do_udp_sendmsg:
1167 if (__ipv6_only_sock(sk))
1168 return -ENETUNREACH;
1169 return udp_sendmsg(sk, msg, len);
1170 }
1171 }
1172
1173 if (up->pending == AF_INET)
1174 return udp_sendmsg(sk, msg, len);
1175
1176 /* Rough check on arithmetic overflow,
1177 better check is made in ip6_append_data().
1178 */
1179 if (len > INT_MAX - sizeof(struct udphdr))
1180 return -EMSGSIZE;
1181
1182 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1183 if (up->pending) {
1184 /*
1185 * There are pending frames.
1186 * The socket lock must be held while it's corked.
1187 */
1188 lock_sock(sk);
1189 if (likely(up->pending)) {
1190 if (unlikely(up->pending != AF_INET6)) {
1191 release_sock(sk);
1192 return -EAFNOSUPPORT;
1193 }
1194 dst = NULL;
1195 goto do_append_data;
1196 }
1197 release_sock(sk);
1198 }
1199 ulen += sizeof(struct udphdr);
1200
1201 memset(&fl6, 0, sizeof(fl6));
1202
1203 if (sin6) {
1204 if (sin6->sin6_port == 0)
1205 return -EINVAL;
1206
1207 fl6.fl6_dport = sin6->sin6_port;
1208 daddr = &sin6->sin6_addr;
1209
1210 if (np->sndflow) {
1211 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1212 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1213 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1214 if (!flowlabel)
1215 return -EINVAL;
1216 }
1217 }
1218
1219 /*
1220 * Otherwise it will be difficult to maintain
1221 * sk->sk_dst_cache.
1222 */
1223 if (sk->sk_state == TCP_ESTABLISHED &&
1224 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1225 daddr = &sk->sk_v6_daddr;
1226
1227 if (addr_len >= sizeof(struct sockaddr_in6) &&
1228 sin6->sin6_scope_id &&
1229 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1230 fl6.flowi6_oif = sin6->sin6_scope_id;
1231 } else {
1232 if (sk->sk_state != TCP_ESTABLISHED)
1233 return -EDESTADDRREQ;
1234
1235 fl6.fl6_dport = inet->inet_dport;
1236 daddr = &sk->sk_v6_daddr;
1237 fl6.flowlabel = np->flow_label;
1238 connected = 1;
1239 }
1240
1241 if (!fl6.flowi6_oif)
1242 fl6.flowi6_oif = sk->sk_bound_dev_if;
1243
1244 if (!fl6.flowi6_oif)
1245 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1246
1247 fl6.flowi6_mark = sk->sk_mark;
1248
1249 if (msg->msg_controllen) {
1250 opt = &opt_space;
1251 memset(opt, 0, sizeof(struct ipv6_txoptions));
1252 opt->tot_len = sizeof(*opt);
1253
1254 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1255 &hlimit, &tclass, &dontfrag);
1256 if (err < 0) {
1257 fl6_sock_release(flowlabel);
1258 return err;
1259 }
1260 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1261 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1262 if (!flowlabel)
1263 return -EINVAL;
1264 }
1265 if (!(opt->opt_nflen|opt->opt_flen))
1266 opt = NULL;
1267 connected = 0;
1268 }
1269 if (!opt) {
1270 opt = txopt_get(np);
1271 opt_to_free = opt;
1272 }
1273 if (flowlabel)
1274 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1275 opt = ipv6_fixup_options(&opt_space, opt);
1276
1277 fl6.flowi6_proto = sk->sk_protocol;
1278 if (!ipv6_addr_any(daddr))
1279 fl6.daddr = *daddr;
1280 else
1281 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1282 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1283 fl6.saddr = np->saddr;
1284 fl6.fl6_sport = inet->inet_sport;
1285
1286 final_p = fl6_update_dst(&fl6, opt, &final);
1287 if (final_p)
1288 connected = 0;
1289
1290 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1291 fl6.flowi6_oif = np->mcast_oif;
1292 connected = 0;
1293 } else if (!fl6.flowi6_oif)
1294 fl6.flowi6_oif = np->ucast_oif;
1295
1296 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1297
1298 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1299 if (IS_ERR(dst)) {
1300 err = PTR_ERR(dst);
1301 dst = NULL;
1302 goto out;
1303 }
1304
1305 if (hlimit < 0)
1306 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1307
1308 if (tclass < 0)
1309 tclass = np->tclass;
1310
1311 if (msg->msg_flags&MSG_CONFIRM)
1312 goto do_confirm;
1313 back_from_confirm:
1314
1315 /* Lockless fast path for the non-corking case */
1316 if (!corkreq) {
1317 struct sk_buff *skb;
1318
1319 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1320 sizeof(struct udphdr), hlimit, tclass, opt,
1321 &fl6, (struct rt6_info *)dst,
1322 msg->msg_flags, dontfrag);
1323 err = PTR_ERR(skb);
1324 if (!IS_ERR_OR_NULL(skb))
1325 err = udp_v6_send_skb(skb, &fl6);
1326 goto release_dst;
1327 }
1328
1329 lock_sock(sk);
1330 if (unlikely(up->pending)) {
1331 /* The socket is already corked while preparing it. */
1332 /* ... which is an evident application bug. --ANK */
1333 release_sock(sk);
1334
1335 net_dbg_ratelimited("udp cork app bug 2\n");
1336 err = -EINVAL;
1337 goto out;
1338 }
1339
1340 up->pending = AF_INET6;
1341
1342 do_append_data:
1343 if (dontfrag < 0)
1344 dontfrag = np->dontfrag;
1345 up->len += ulen;
1346 err = ip6_append_data(sk, getfrag, msg, ulen,
1347 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1348 (struct rt6_info *)dst,
1349 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1350 if (err)
1351 udp_v6_flush_pending_frames(sk);
1352 else if (!corkreq)
1353 err = udp_v6_push_pending_frames(sk);
1354 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1355 up->pending = 0;
1356
1357 if (err > 0)
1358 err = np->recverr ? net_xmit_errno(err) : 0;
1359 release_sock(sk);
1360
1361 release_dst:
1362 if (dst) {
1363 if (connected) {
1364 ip6_dst_store(sk, dst,
1365 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1366 &sk->sk_v6_daddr : NULL,
1367 #ifdef CONFIG_IPV6_SUBTREES
1368 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1369 &np->saddr :
1370 #endif
1371 NULL);
1372 } else {
1373 dst_release(dst);
1374 }
1375 dst = NULL;
1376 }
1377
1378 out:
1379 dst_release(dst);
1380 fl6_sock_release(flowlabel);
1381 txopt_put(opt_to_free);
1382 if (!err)
1383 return len;
1384 /*
1385 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1386 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1387 * we don't have a good statistic (IpOutDiscards but it can be too many
1388 * things). We could add another new stat but at least for now that
1389 * seems like overkill.
1390 */
1391 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1392 UDP6_INC_STATS_USER(sock_net(sk),
1393 UDP_MIB_SNDBUFERRORS, is_udplite);
1394 }
1395 return err;
1396
1397 do_confirm:
1398 dst_confirm(dst);
1399 if (!(msg->msg_flags&MSG_PROBE) || len)
1400 goto back_from_confirm;
1401 err = 0;
1402 goto out;
1403 }
1404
1405 void udpv6_destroy_sock(struct sock *sk)
1406 {
1407 struct udp_sock *up = udp_sk(sk);
1408 lock_sock(sk);
1409 udp_v6_flush_pending_frames(sk);
1410 release_sock(sk);
1411
1412 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1413 void (*encap_destroy)(struct sock *sk);
1414 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1415 if (encap_destroy)
1416 encap_destroy(sk);
1417 }
1418
1419 inet6_destroy_sock(sk);
1420 }
1421
1422 /*
1423 * Socket option code for UDP
1424 */
1425 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1426 char __user *optval, unsigned int optlen)
1427 {
1428 if (level == SOL_UDP || level == SOL_UDPLITE)
1429 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1430 udp_v6_push_pending_frames);
1431 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1432 }
1433
1434 #ifdef CONFIG_COMPAT
1435 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1436 char __user *optval, unsigned int optlen)
1437 {
1438 if (level == SOL_UDP || level == SOL_UDPLITE)
1439 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1440 udp_v6_push_pending_frames);
1441 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1442 }
1443 #endif
1444
1445 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1446 char __user *optval, int __user *optlen)
1447 {
1448 if (level == SOL_UDP || level == SOL_UDPLITE)
1449 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1450 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1451 }
1452
1453 #ifdef CONFIG_COMPAT
1454 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1455 char __user *optval, int __user *optlen)
1456 {
1457 if (level == SOL_UDP || level == SOL_UDPLITE)
1458 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1459 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1460 }
1461 #endif
1462
1463 static const struct inet6_protocol udpv6_protocol = {
1464 .handler = udpv6_rcv,
1465 .err_handler = udpv6_err,
1466 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1467 };
1468
1469 /* ------------------------------------------------------------------------ */
1470 #ifdef CONFIG_PROC_FS
1471 int udp6_seq_show(struct seq_file *seq, void *v)
1472 {
1473 if (v == SEQ_START_TOKEN) {
1474 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1475 } else {
1476 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1477 struct inet_sock *inet = inet_sk(v);
1478 __u16 srcp = ntohs(inet->inet_sport);
1479 __u16 destp = ntohs(inet->inet_dport);
1480 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1481 }
1482 return 0;
1483 }
1484
1485 static const struct file_operations udp6_afinfo_seq_fops = {
1486 .owner = THIS_MODULE,
1487 .open = udp_seq_open,
1488 .read = seq_read,
1489 .llseek = seq_lseek,
1490 .release = seq_release_net
1491 };
1492
1493 static struct udp_seq_afinfo udp6_seq_afinfo = {
1494 .name = "udp6",
1495 .family = AF_INET6,
1496 .udp_table = &udp_table,
1497 .seq_fops = &udp6_afinfo_seq_fops,
1498 .seq_ops = {
1499 .show = udp6_seq_show,
1500 },
1501 };
1502
1503 int __net_init udp6_proc_init(struct net *net)
1504 {
1505 return udp_proc_register(net, &udp6_seq_afinfo);
1506 }
1507
1508 void udp6_proc_exit(struct net *net)
1509 {
1510 udp_proc_unregister(net, &udp6_seq_afinfo);
1511 }
1512 #endif /* CONFIG_PROC_FS */
1513
1514 void udp_v6_clear_sk(struct sock *sk, int size)
1515 {
1516 struct inet_sock *inet = inet_sk(sk);
1517
1518 /* we do not want to clear pinet6 field, because of RCU lookups */
1519 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1520
1521 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1522 memset(&inet->pinet6 + 1, 0, size);
1523 }
1524
1525 /* ------------------------------------------------------------------------ */
1526
1527 struct proto udpv6_prot = {
1528 .name = "UDPv6",
1529 .owner = THIS_MODULE,
1530 .close = udp_lib_close,
1531 .connect = ip6_datagram_connect,
1532 .disconnect = udp_disconnect,
1533 .ioctl = udp_ioctl,
1534 .destroy = udpv6_destroy_sock,
1535 .setsockopt = udpv6_setsockopt,
1536 .getsockopt = udpv6_getsockopt,
1537 .sendmsg = udpv6_sendmsg,
1538 .recvmsg = udpv6_recvmsg,
1539 .backlog_rcv = __udpv6_queue_rcv_skb,
1540 .hash = udp_lib_hash,
1541 .unhash = udp_lib_unhash,
1542 .rehash = udp_v6_rehash,
1543 .get_port = udp_v6_get_port,
1544 .memory_allocated = &udp_memory_allocated,
1545 .sysctl_mem = sysctl_udp_mem,
1546 .sysctl_wmem = &sysctl_udp_wmem_min,
1547 .sysctl_rmem = &sysctl_udp_rmem_min,
1548 .obj_size = sizeof(struct udp6_sock),
1549 .slab_flags = SLAB_DESTROY_BY_RCU,
1550 .h.udp_table = &udp_table,
1551 #ifdef CONFIG_COMPAT
1552 .compat_setsockopt = compat_udpv6_setsockopt,
1553 .compat_getsockopt = compat_udpv6_getsockopt,
1554 #endif
1555 .clear_sk = udp_v6_clear_sk,
1556 };
1557
1558 static struct inet_protosw udpv6_protosw = {
1559 .type = SOCK_DGRAM,
1560 .protocol = IPPROTO_UDP,
1561 .prot = &udpv6_prot,
1562 .ops = &inet6_dgram_ops,
1563 .flags = INET_PROTOSW_PERMANENT,
1564 };
1565
1566 int __init udpv6_init(void)
1567 {
1568 int ret;
1569
1570 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1571 if (ret)
1572 goto out;
1573
1574 ret = inet6_register_protosw(&udpv6_protosw);
1575 if (ret)
1576 goto out_udpv6_protocol;
1577 out:
1578 return ret;
1579
1580 out_udpv6_protocol:
1581 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1582 goto out;
1583 }
1584
1585 void udpv6_exit(void)
1586 {
1587 inet6_unregister_protosw(&udpv6_protosw);
1588 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1589 }
(deprecated) Btrfs devel tree