?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ipv6: only static routes qualify for equal cost multipathing
[linux-2.6.git] / net / ipv6 / udp.c
blobf4058150262b111d316a423e1c9ddc7ee8bb29a0
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 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
57 {
58 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
59 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
60 __be32 sk1_rcv_saddr = sk_rcv_saddr(sk);
61 __be32 sk2_rcv_saddr = sk_rcv_saddr(sk2);
62 int sk_ipv6only = ipv6_only_sock(sk);
63 int sk2_ipv6only = inet_v6_ipv6only(sk2);
64 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
65 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
66
67 /* if both are mapped, treat as IPv4 */
68 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
69 return (!sk2_ipv6only &&
70 (!sk1_rcv_saddr || !sk2_rcv_saddr ||
71 sk1_rcv_saddr == sk2_rcv_saddr));
72
73 if (addr_type2 == IPV6_ADDR_ANY &&
74 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
75 return 1;
76
77 if (addr_type == IPV6_ADDR_ANY &&
78 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
79 return 1;
80
81 if (sk2_rcv_saddr6 &&
82 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
83 return 1;
84
85 return 0;
86 }
87
88 static unsigned int udp6_portaddr_hash(struct net *net,
89 const struct in6_addr *addr6,
90 unsigned int port)
91 {
92 unsigned int hash, mix = net_hash_mix(net);
93
94 if (ipv6_addr_any(addr6))
95 hash = jhash_1word(0, mix);
96 else if (ipv6_addr_v4mapped(addr6))
97 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
98 else
99 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
100
101 return hash ^ port;
102 }
103
104
105 int udp_v6_get_port(struct sock *sk, unsigned short snum)
106 {
107 unsigned int hash2_nulladdr =
108 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
109 unsigned int hash2_partial =
110 udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0);
111
112 /* precompute partial secondary hash */
113 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
114 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
115 }
116
117 static void udp_v6_rehash(struct sock *sk)
118 {
119 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
120 &inet6_sk(sk)->rcv_saddr,
121 inet_sk(sk)->inet_num);
122
123 udp_lib_rehash(sk, new_hash);
124 }
125
126 static inline int compute_score(struct sock *sk, struct net *net,
127 unsigned short hnum,
128 const struct in6_addr *saddr, __be16 sport,
129 const struct in6_addr *daddr, __be16 dport,
130 int dif)
131 {
132 int score = -1;
133
134 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
135 sk->sk_family == PF_INET6) {
136 struct ipv6_pinfo *np = inet6_sk(sk);
137 struct inet_sock *inet = inet_sk(sk);
138
139 score = 0;
140 if (inet->inet_dport) {
141 if (inet->inet_dport != sport)
142 return -1;
143 score++;
144 }
145 if (!ipv6_addr_any(&np->rcv_saddr)) {
146 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
147 return -1;
148 score++;
149 }
150 if (!ipv6_addr_any(&np->daddr)) {
151 if (!ipv6_addr_equal(&np->daddr, saddr))
152 return -1;
153 score++;
154 }
155 if (sk->sk_bound_dev_if) {
156 if (sk->sk_bound_dev_if != dif)
157 return -1;
158 score++;
159 }
160 }
161 return score;
162 }
163
164 #define SCORE2_MAX (1 + 1 + 1)
165 static inline int compute_score2(struct sock *sk, struct net *net,
166 const struct in6_addr *saddr, __be16 sport,
167 const struct in6_addr *daddr, unsigned short hnum,
168 int dif)
169 {
170 int score = -1;
171
172 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
173 sk->sk_family == PF_INET6) {
174 struct ipv6_pinfo *np = inet6_sk(sk);
175 struct inet_sock *inet = inet_sk(sk);
176
177 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
178 return -1;
179 score = 0;
180 if (inet->inet_dport) {
181 if (inet->inet_dport != sport)
182 return -1;
183 score++;
184 }
185 if (!ipv6_addr_any(&np->daddr)) {
186 if (!ipv6_addr_equal(&np->daddr, saddr))
187 return -1;
188 score++;
189 }
190 if (sk->sk_bound_dev_if) {
191 if (sk->sk_bound_dev_if != dif)
192 return -1;
193 score++;
194 }
195 }
196 return score;
197 }
198
199
200 /* called with read_rcu_lock() */
201 static struct sock *udp6_lib_lookup2(struct net *net,
202 const struct in6_addr *saddr, __be16 sport,
203 const struct in6_addr *daddr, unsigned int hnum, int dif,
204 struct udp_hslot *hslot2, unsigned int slot2)
205 {
206 struct sock *sk, *result;
207 struct hlist_nulls_node *node;
208 int score, badness, matches = 0, reuseport = 0;
209 u32 hash = 0;
210
211 begin:
212 result = NULL;
213 badness = -1;
214 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
215 score = compute_score2(sk, net, saddr, sport,
216 daddr, hnum, dif);
217 if (score > badness) {
218 result = sk;
219 badness = score;
220 reuseport = sk->sk_reuseport;
221 if (reuseport) {
222 hash = inet6_ehashfn(net, daddr, hnum,
223 saddr, sport);
224 matches = 1;
225 } else if (score == SCORE2_MAX)
226 goto exact_match;
227 } else if (score == badness && reuseport) {
228 matches++;
229 if (((u64)hash * matches) >> 32 == 0)
230 result = sk;
231 hash = next_pseudo_random32(hash);
232 }
233 }
234 /*
235 * if the nulls value we got at the end of this lookup is
236 * not the expected one, we must restart lookup.
237 * We probably met an item that was moved to another chain.
238 */
239 if (get_nulls_value(node) != slot2)
240 goto begin;
241
242 if (result) {
243 exact_match:
244 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
245 result = NULL;
246 else if (unlikely(compute_score2(result, net, saddr, sport,
247 daddr, hnum, dif) < badness)) {
248 sock_put(result);
249 goto begin;
250 }
251 }
252 return result;
253 }
254
255 struct sock *__udp6_lib_lookup(struct net *net,
256 const struct in6_addr *saddr, __be16 sport,
257 const struct in6_addr *daddr, __be16 dport,
258 int dif, struct udp_table *udptable)
259 {
260 struct sock *sk, *result;
261 struct hlist_nulls_node *node;
262 unsigned short hnum = ntohs(dport);
263 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
264 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
265 int score, badness, matches = 0, reuseport = 0;
266 u32 hash = 0;
267
268 rcu_read_lock();
269 if (hslot->count > 10) {
270 hash2 = udp6_portaddr_hash(net, daddr, hnum);
271 slot2 = hash2 & udptable->mask;
272 hslot2 = &udptable->hash2[slot2];
273 if (hslot->count < hslot2->count)
274 goto begin;
275
276 result = udp6_lib_lookup2(net, saddr, sport,
277 daddr, hnum, dif,
278 hslot2, slot2);
279 if (!result) {
280 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
281 slot2 = hash2 & udptable->mask;
282 hslot2 = &udptable->hash2[slot2];
283 if (hslot->count < hslot2->count)
284 goto begin;
285
286 result = udp6_lib_lookup2(net, saddr, sport,
287 &in6addr_any, hnum, dif,
288 hslot2, slot2);
289 }
290 rcu_read_unlock();
291 return result;
292 }
293 begin:
294 result = NULL;
295 badness = -1;
296 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
297 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
298 if (score > badness) {
299 result = sk;
300 badness = score;
301 reuseport = sk->sk_reuseport;
302 if (reuseport) {
303 hash = inet6_ehashfn(net, daddr, hnum,
304 saddr, sport);
305 matches = 1;
306 }
307 } else if (score == badness && reuseport) {
308 matches++;
309 if (((u64)hash * matches) >> 32 == 0)
310 result = sk;
311 hash = next_pseudo_random32(hash);
312 }
313 }
314 /*
315 * if the nulls value we got at the end of this lookup is
316 * not the expected one, we must restart lookup.
317 * We probably met an item that was moved to another chain.
318 */
319 if (get_nulls_value(node) != slot)
320 goto begin;
321
322 if (result) {
323 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
324 result = NULL;
325 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
326 daddr, dport, dif) < badness)) {
327 sock_put(result);
328 goto begin;
329 }
330 }
331 rcu_read_unlock();
332 return result;
333 }
334 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
335
336 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
337 __be16 sport, __be16 dport,
338 struct udp_table *udptable)
339 {
340 struct sock *sk;
341 const struct ipv6hdr *iph = ipv6_hdr(skb);
342
343 if (unlikely(sk = skb_steal_sock(skb)))
344 return sk;
345 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
346 &iph->daddr, dport, inet6_iif(skb),
347 udptable);
348 }
349
350 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
351 const struct in6_addr *daddr, __be16 dport, int dif)
352 {
353 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
354 }
355 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
356
357
358 /*
359 * This should be easy, if there is something there we
360 * return it, otherwise we block.
361 */
362
363 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
364 struct msghdr *msg, size_t len,
365 int noblock, int flags, int *addr_len)
366 {
367 struct ipv6_pinfo *np = inet6_sk(sk);
368 struct inet_sock *inet = inet_sk(sk);
369 struct sk_buff *skb;
370 unsigned int ulen, copied;
371 int peeked, off = 0;
372 int err;
373 int is_udplite = IS_UDPLITE(sk);
374 int is_udp4;
375 bool slow;
376
377 if (addr_len)
378 *addr_len = sizeof(struct sockaddr_in6);
379
380 if (flags & MSG_ERRQUEUE)
381 return ipv6_recv_error(sk, msg, len);
382
383 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
384 return ipv6_recv_rxpmtu(sk, msg, len);
385
386 try_again:
387 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
388 &peeked, &off, &err);
389 if (!skb)
390 goto out;
391
392 ulen = skb->len - sizeof(struct udphdr);
393 copied = len;
394 if (copied > ulen)
395 copied = ulen;
396 else if (copied < ulen)
397 msg->msg_flags |= MSG_TRUNC;
398
399 is_udp4 = (skb->protocol == htons(ETH_P_IP));
400
401 /*
402 * If checksum is needed at all, try to do it while copying the
403 * data. If the data is truncated, or if we only want a partial
404 * coverage checksum (UDP-Lite), do it before the copy.
405 */
406
407 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
408 if (udp_lib_checksum_complete(skb))
409 goto csum_copy_err;
410 }
411
412 if (skb_csum_unnecessary(skb))
413 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
414 msg->msg_iov, copied);
415 else {
416 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
417 if (err == -EINVAL)
418 goto csum_copy_err;
419 }
420 if (unlikely(err)) {
421 trace_kfree_skb(skb, udpv6_recvmsg);
422 if (!peeked) {
423 atomic_inc(&sk->sk_drops);
424 if (is_udp4)
425 UDP_INC_STATS_USER(sock_net(sk),
426 UDP_MIB_INERRORS,
427 is_udplite);
428 else
429 UDP6_INC_STATS_USER(sock_net(sk),
430 UDP_MIB_INERRORS,
431 is_udplite);
432 }
433 goto out_free;
434 }
435 if (!peeked) {
436 if (is_udp4)
437 UDP_INC_STATS_USER(sock_net(sk),
438 UDP_MIB_INDATAGRAMS, is_udplite);
439 else
440 UDP6_INC_STATS_USER(sock_net(sk),
441 UDP_MIB_INDATAGRAMS, is_udplite);
442 }
443
444 sock_recv_ts_and_drops(msg, sk, skb);
445
446 /* Copy the address. */
447 if (msg->msg_name) {
448 struct sockaddr_in6 *sin6;
449
450 sin6 = (struct sockaddr_in6 *) msg->msg_name;
451 sin6->sin6_family = AF_INET6;
452 sin6->sin6_port = udp_hdr(skb)->source;
453 sin6->sin6_flowinfo = 0;
454
455 if (is_udp4) {
456 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
457 &sin6->sin6_addr);
458 sin6->sin6_scope_id = 0;
459 } else {
460 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
461 sin6->sin6_scope_id =
462 ipv6_iface_scope_id(&sin6->sin6_addr,
463 IP6CB(skb)->iif);
464 }
465
466 }
467 if (is_udp4) {
468 if (inet->cmsg_flags)
469 ip_cmsg_recv(msg, skb);
470 } else {
471 if (np->rxopt.all)
472 ip6_datagram_recv_ctl(sk, msg, skb);
473 }
474
475 err = copied;
476 if (flags & MSG_TRUNC)
477 err = ulen;
478
479 out_free:
480 skb_free_datagram_locked(sk, skb);
481 out:
482 return err;
483
484 csum_copy_err:
485 slow = lock_sock_fast(sk);
486 if (!skb_kill_datagram(sk, skb, flags)) {
487 if (is_udp4) {
488 UDP_INC_STATS_USER(sock_net(sk),
489 UDP_MIB_CSUMERRORS, is_udplite);
490 UDP_INC_STATS_USER(sock_net(sk),
491 UDP_MIB_INERRORS, is_udplite);
492 } else {
493 UDP6_INC_STATS_USER(sock_net(sk),
494 UDP_MIB_CSUMERRORS, is_udplite);
495 UDP6_INC_STATS_USER(sock_net(sk),
496 UDP_MIB_INERRORS, is_udplite);
497 }
498 }
499 unlock_sock_fast(sk, slow);
500
501 if (noblock)
502 return -EAGAIN;
503
504 /* starting over for a new packet */
505 msg->msg_flags &= ~MSG_TRUNC;
506 goto try_again;
507 }
508
509 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
510 u8 type, u8 code, int offset, __be32 info,
511 struct udp_table *udptable)
512 {
513 struct ipv6_pinfo *np;
514 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
515 const struct in6_addr *saddr = &hdr->saddr;
516 const struct in6_addr *daddr = &hdr->daddr;
517 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
518 struct sock *sk;
519 int err;
520
521 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
522 saddr, uh->source, inet6_iif(skb), udptable);
523 if (sk == NULL)
524 return;
525
526 if (type == ICMPV6_PKT_TOOBIG)
527 ip6_sk_update_pmtu(skb, sk, info);
528 if (type == NDISC_REDIRECT)
529 ip6_sk_redirect(skb, sk);
530
531 np = inet6_sk(sk);
532
533 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
534 goto out;
535
536 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
537 goto out;
538
539 if (np->recverr)
540 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
541
542 sk->sk_err = err;
543 sk->sk_error_report(sk);
544 out:
545 sock_put(sk);
546 }
547
548 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
549 {
550 int rc;
551
552 if (!ipv6_addr_any(&inet6_sk(sk)->daddr))
553 sock_rps_save_rxhash(sk, skb);
554
555 rc = sock_queue_rcv_skb(sk, skb);
556 if (rc < 0) {
557 int is_udplite = IS_UDPLITE(sk);
558
559 /* Note that an ENOMEM error is charged twice */
560 if (rc == -ENOMEM)
561 UDP6_INC_STATS_BH(sock_net(sk),
562 UDP_MIB_RCVBUFERRORS, is_udplite);
563 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
564 kfree_skb(skb);
565 return -1;
566 }
567 return 0;
568 }
569
570 static __inline__ void udpv6_err(struct sk_buff *skb,
571 struct inet6_skb_parm *opt, u8 type,
572 u8 code, int offset, __be32 info )
573 {
574 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
575 }
576
577 static struct static_key udpv6_encap_needed __read_mostly;
578 void udpv6_encap_enable(void)
579 {
580 if (!static_key_enabled(&udpv6_encap_needed))
581 static_key_slow_inc(&udpv6_encap_needed);
582 }
583 EXPORT_SYMBOL(udpv6_encap_enable);
584
585 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
586 {
587 struct udp_sock *up = udp_sk(sk);
588 int rc;
589 int is_udplite = IS_UDPLITE(sk);
590
591 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
592 goto drop;
593
594 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
595 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
596
597 /*
598 * This is an encapsulation socket so pass the skb to
599 * the socket's udp_encap_rcv() hook. Otherwise, just
600 * fall through and pass this up the UDP socket.
601 * up->encap_rcv() returns the following value:
602 * =0 if skb was successfully passed to the encap
603 * handler or was discarded by it.
604 * >0 if skb should be passed on to UDP.
605 * <0 if skb should be resubmitted as proto -N
606 */
607
608 /* if we're overly short, let UDP handle it */
609 encap_rcv = ACCESS_ONCE(up->encap_rcv);
610 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
611 int ret;
612
613 ret = encap_rcv(sk, skb);
614 if (ret <= 0) {
615 UDP_INC_STATS_BH(sock_net(sk),
616 UDP_MIB_INDATAGRAMS,
617 is_udplite);
618 return -ret;
619 }
620 }
621
622 /* FALLTHROUGH -- it's a UDP Packet */
623 }
624
625 /*
626 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
627 */
628 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
629
630 if (up->pcrlen == 0) { /* full coverage was set */
631 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
632 " %d while full coverage %d requested\n",
633 UDP_SKB_CB(skb)->cscov, skb->len);
634 goto drop;
635 }
636 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
637 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
638 "too small, need min %d\n",
639 UDP_SKB_CB(skb)->cscov, up->pcrlen);
640 goto drop;
641 }
642 }
643
644 if (rcu_access_pointer(sk->sk_filter)) {
645 if (udp_lib_checksum_complete(skb))
646 goto csum_error;
647 }
648
649 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
650 goto drop;
651
652 skb_dst_drop(skb);
653
654 bh_lock_sock(sk);
655 rc = 0;
656 if (!sock_owned_by_user(sk))
657 rc = __udpv6_queue_rcv_skb(sk, skb);
658 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
659 bh_unlock_sock(sk);
660 goto drop;
661 }
662 bh_unlock_sock(sk);
663
664 return rc;
665 csum_error:
666 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
667 drop:
668 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
669 atomic_inc(&sk->sk_drops);
670 kfree_skb(skb);
671 return -1;
672 }
673
674 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
675 __be16 loc_port, const struct in6_addr *loc_addr,
676 __be16 rmt_port, const struct in6_addr *rmt_addr,
677 int dif)
678 {
679 struct hlist_nulls_node *node;
680 struct sock *s = sk;
681 unsigned short num = ntohs(loc_port);
682
683 sk_nulls_for_each_from(s, node) {
684 struct inet_sock *inet = inet_sk(s);
685
686 if (!net_eq(sock_net(s), net))
687 continue;
688
689 if (udp_sk(s)->udp_port_hash == num &&
690 s->sk_family == PF_INET6) {
691 struct ipv6_pinfo *np = inet6_sk(s);
692 if (inet->inet_dport) {
693 if (inet->inet_dport != rmt_port)
694 continue;
695 }
696 if (!ipv6_addr_any(&np->daddr) &&
697 !ipv6_addr_equal(&np->daddr, rmt_addr))
698 continue;
699
700 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
701 continue;
702
703 if (!ipv6_addr_any(&np->rcv_saddr)) {
704 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
705 continue;
706 }
707 if (!inet6_mc_check(s, loc_addr, rmt_addr))
708 continue;
709 return s;
710 }
711 }
712 return NULL;
713 }
714
715 static void flush_stack(struct sock **stack, unsigned int count,
716 struct sk_buff *skb, unsigned int final)
717 {
718 struct sk_buff *skb1 = NULL;
719 struct sock *sk;
720 unsigned int i;
721
722 for (i = 0; i < count; i++) {
723 sk = stack[i];
724 if (likely(skb1 == NULL))
725 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
726 if (!skb1) {
727 atomic_inc(&sk->sk_drops);
728 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
729 IS_UDPLITE(sk));
730 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
731 IS_UDPLITE(sk));
732 }
733
734 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
735 skb1 = NULL;
736 }
737 if (unlikely(skb1))
738 kfree_skb(skb1);
739 }
740 /*
741 * Note: called only from the BH handler context,
742 * so we don't need to lock the hashes.
743 */
744 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
745 const struct in6_addr *saddr, const struct in6_addr *daddr,
746 struct udp_table *udptable)
747 {
748 struct sock *sk, *stack[256 / sizeof(struct sock *)];
749 const struct udphdr *uh = udp_hdr(skb);
750 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
751 int dif;
752 unsigned int i, count = 0;
753
754 spin_lock(&hslot->lock);
755 sk = sk_nulls_head(&hslot->head);
756 dif = inet6_iif(skb);
757 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
758 while (sk) {
759 stack[count++] = sk;
760 sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
761 uh->source, saddr, dif);
762 if (unlikely(count == ARRAY_SIZE(stack))) {
763 if (!sk)
764 break;
765 flush_stack(stack, count, skb, ~0);
766 count = 0;
767 }
768 }
769 /*
770 * before releasing the lock, we must take reference on sockets
771 */
772 for (i = 0; i < count; i++)
773 sock_hold(stack[i]);
774
775 spin_unlock(&hslot->lock);
776
777 if (count) {
778 flush_stack(stack, count, skb, count - 1);
779
780 for (i = 0; i < count; i++)
781 sock_put(stack[i]);
782 } else {
783 kfree_skb(skb);
784 }
785 return 0;
786 }
787
788 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
789 int proto)
790 {
791 struct net *net = dev_net(skb->dev);
792 struct sock *sk;
793 struct udphdr *uh;
794 const struct in6_addr *saddr, *daddr;
795 u32 ulen = 0;
796
797 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
798 goto discard;
799
800 saddr = &ipv6_hdr(skb)->saddr;
801 daddr = &ipv6_hdr(skb)->daddr;
802 uh = udp_hdr(skb);
803
804 ulen = ntohs(uh->len);
805 if (ulen > skb->len)
806 goto short_packet;
807
808 if (proto == IPPROTO_UDP) {
809 /* UDP validates ulen. */
810
811 /* Check for jumbo payload */
812 if (ulen == 0)
813 ulen = skb->len;
814
815 if (ulen < sizeof(*uh))
816 goto short_packet;
817
818 if (ulen < skb->len) {
819 if (pskb_trim_rcsum(skb, ulen))
820 goto short_packet;
821 saddr = &ipv6_hdr(skb)->saddr;
822 daddr = &ipv6_hdr(skb)->daddr;
823 uh = udp_hdr(skb);
824 }
825 }
826
827 if (udp6_csum_init(skb, uh, proto))
828 goto csum_error;
829
830 /*
831 * Multicast receive code
832 */
833 if (ipv6_addr_is_multicast(daddr))
834 return __udp6_lib_mcast_deliver(net, skb,
835 saddr, daddr, udptable);
836
837 /* Unicast */
838
839 /*
840 * check socket cache ... must talk to Alan about his plans
841 * for sock caches... i'll skip this for now.
842 */
843 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
844 if (sk != NULL) {
845 int ret;
846
847 sk_mark_napi_id(sk, skb);
848 ret = udpv6_queue_rcv_skb(sk, skb);
849 sock_put(sk);
850
851 /* a return value > 0 means to resubmit the input, but
852 * it wants the return to be -protocol, or 0
853 */
854 if (ret > 0)
855 return -ret;
856
857 return 0;
858 }
859
860 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
861 goto discard;
862
863 if (udp_lib_checksum_complete(skb))
864 goto csum_error;
865
866 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
867 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
868
869 kfree_skb(skb);
870 return 0;
871
872 short_packet:
873 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
874 proto == IPPROTO_UDPLITE ? "-Lite" : "",
875 saddr,
876 ntohs(uh->source),
877 ulen,
878 skb->len,
879 daddr,
880 ntohs(uh->dest));
881 goto discard;
882 csum_error:
883 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
884 discard:
885 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
886 kfree_skb(skb);
887 return 0;
888 }
889
890 static __inline__ int udpv6_rcv(struct sk_buff *skb)
891 {
892 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
893 }
894
895 /*
896 * Throw away all pending data and cancel the corking. Socket is locked.
897 */
898 static void udp_v6_flush_pending_frames(struct sock *sk)
899 {
900 struct udp_sock *up = udp_sk(sk);
901
902 if (up->pending == AF_INET)
903 udp_flush_pending_frames(sk);
904 else if (up->pending) {
905 up->len = 0;
906 up->pending = 0;
907 ip6_flush_pending_frames(sk);
908 }
909 }
910
911 /**
912 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
913 * @sk: socket we are sending on
914 * @skb: sk_buff containing the filled-in UDP header
915 * (checksum field must be zeroed out)
916 */
917 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
918 const struct in6_addr *saddr,
919 const struct in6_addr *daddr, int len)
920 {
921 unsigned int offset;
922 struct udphdr *uh = udp_hdr(skb);
923 __wsum csum = 0;
924
925 if (skb_queue_len(&sk->sk_write_queue) == 1) {
926 /* Only one fragment on the socket. */
927 skb->csum_start = skb_transport_header(skb) - skb->head;
928 skb->csum_offset = offsetof(struct udphdr, check);
929 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
930 } else {
931 /*
932 * HW-checksum won't work as there are two or more
933 * fragments on the socket so that all csums of sk_buffs
934 * should be together
935 */
936 offset = skb_transport_offset(skb);
937 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
938
939 skb->ip_summed = CHECKSUM_NONE;
940
941 skb_queue_walk(&sk->sk_write_queue, skb) {
942 csum = csum_add(csum, skb->csum);
943 }
944
945 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
946 csum);
947 if (uh->check == 0)
948 uh->check = CSUM_MANGLED_0;
949 }
950 }
951
952 /*
953 * Sending
954 */
955
956 static int udp_v6_push_pending_frames(struct sock *sk)
957 {
958 struct sk_buff *skb;
959 struct udphdr *uh;
960 struct udp_sock *up = udp_sk(sk);
961 struct inet_sock *inet = inet_sk(sk);
962 struct flowi6 *fl6;
963 int err = 0;
964 int is_udplite = IS_UDPLITE(sk);
965 __wsum csum = 0;
966
967 if (up->pending == AF_INET)
968 return udp_push_pending_frames(sk);
969
970 fl6 = &inet->cork.fl.u.ip6;
971
972 /* Grab the skbuff where UDP header space exists. */
973 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
974 goto out;
975
976 /*
977 * Create a UDP header
978 */
979 uh = udp_hdr(skb);
980 uh->source = fl6->fl6_sport;
981 uh->dest = fl6->fl6_dport;
982 uh->len = htons(up->len);
983 uh->check = 0;
984
985 if (is_udplite)
986 csum = udplite_csum_outgoing(sk, skb);
987 else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
988 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
989 up->len);
990 goto send;
991 } else
992 csum = udp_csum_outgoing(sk, skb);
993
994 /* add protocol-dependent pseudo-header */
995 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
996 up->len, fl6->flowi6_proto, csum);
997 if (uh->check == 0)
998 uh->check = CSUM_MANGLED_0;
999
1000 send:
1001 err = ip6_push_pending_frames(sk);
1002 if (err) {
1003 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1004 UDP6_INC_STATS_USER(sock_net(sk),
1005 UDP_MIB_SNDBUFERRORS, is_udplite);
1006 err = 0;
1007 }
1008 } else
1009 UDP6_INC_STATS_USER(sock_net(sk),
1010 UDP_MIB_OUTDATAGRAMS, is_udplite);
1011 out:
1012 up->len = 0;
1013 up->pending = 0;
1014 return err;
1015 }
1016
1017 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1018 struct msghdr *msg, size_t len)
1019 {
1020 struct ipv6_txoptions opt_space;
1021 struct udp_sock *up = udp_sk(sk);
1022 struct inet_sock *inet = inet_sk(sk);
1023 struct ipv6_pinfo *np = inet6_sk(sk);
1024 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
1025 struct in6_addr *daddr, *final_p, final;
1026 struct ipv6_txoptions *opt = NULL;
1027 struct ip6_flowlabel *flowlabel = NULL;
1028 struct flowi6 fl6;
1029 struct dst_entry *dst;
1030 int addr_len = msg->msg_namelen;
1031 int ulen = len;
1032 int hlimit = -1;
1033 int tclass = -1;
1034 int dontfrag = -1;
1035 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1036 int err;
1037 int connected = 0;
1038 int is_udplite = IS_UDPLITE(sk);
1039 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1040
1041 /* destination address check */
1042 if (sin6) {
1043 if (addr_len < offsetof(struct sockaddr, sa_data))
1044 return -EINVAL;
1045
1046 switch (sin6->sin6_family) {
1047 case AF_INET6:
1048 if (addr_len < SIN6_LEN_RFC2133)
1049 return -EINVAL;
1050 daddr = &sin6->sin6_addr;
1051 break;
1052 case AF_INET:
1053 goto do_udp_sendmsg;
1054 case AF_UNSPEC:
1055 msg->msg_name = sin6 = NULL;
1056 msg->msg_namelen = addr_len = 0;
1057 daddr = NULL;
1058 break;
1059 default:
1060 return -EINVAL;
1061 }
1062 } else if (!up->pending) {
1063 if (sk->sk_state != TCP_ESTABLISHED)
1064 return -EDESTADDRREQ;
1065 daddr = &np->daddr;
1066 } else
1067 daddr = NULL;
1068
1069 if (daddr) {
1070 if (ipv6_addr_v4mapped(daddr)) {
1071 struct sockaddr_in sin;
1072 sin.sin_family = AF_INET;
1073 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1074 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1075 msg->msg_name = &sin;
1076 msg->msg_namelen = sizeof(sin);
1077 do_udp_sendmsg:
1078 if (__ipv6_only_sock(sk))
1079 return -ENETUNREACH;
1080 return udp_sendmsg(iocb, sk, msg, len);
1081 }
1082 }
1083
1084 if (up->pending == AF_INET)
1085 return udp_sendmsg(iocb, sk, msg, len);
1086
1087 /* Rough check on arithmetic overflow,
1088 better check is made in ip6_append_data().
1089 */
1090 if (len > INT_MAX - sizeof(struct udphdr))
1091 return -EMSGSIZE;
1092
1093 if (up->pending) {
1094 /*
1095 * There are pending frames.
1096 * The socket lock must be held while it's corked.
1097 */
1098 lock_sock(sk);
1099 if (likely(up->pending)) {
1100 if (unlikely(up->pending != AF_INET6)) {
1101 release_sock(sk);
1102 return -EAFNOSUPPORT;
1103 }
1104 dst = NULL;
1105 goto do_append_data;
1106 }
1107 release_sock(sk);
1108 }
1109 ulen += sizeof(struct udphdr);
1110
1111 memset(&fl6, 0, sizeof(fl6));
1112
1113 if (sin6) {
1114 if (sin6->sin6_port == 0)
1115 return -EINVAL;
1116
1117 fl6.fl6_dport = sin6->sin6_port;
1118 daddr = &sin6->sin6_addr;
1119
1120 if (np->sndflow) {
1121 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1122 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1123 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1124 if (flowlabel == NULL)
1125 return -EINVAL;
1126 daddr = &flowlabel->dst;
1127 }
1128 }
1129
1130 /*
1131 * Otherwise it will be difficult to maintain
1132 * sk->sk_dst_cache.
1133 */
1134 if (sk->sk_state == TCP_ESTABLISHED &&
1135 ipv6_addr_equal(daddr, &np->daddr))
1136 daddr = &np->daddr;
1137
1138 if (addr_len >= sizeof(struct sockaddr_in6) &&
1139 sin6->sin6_scope_id &&
1140 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1141 fl6.flowi6_oif = sin6->sin6_scope_id;
1142 } else {
1143 if (sk->sk_state != TCP_ESTABLISHED)
1144 return -EDESTADDRREQ;
1145
1146 fl6.fl6_dport = inet->inet_dport;
1147 daddr = &np->daddr;
1148 fl6.flowlabel = np->flow_label;
1149 connected = 1;
1150 }
1151
1152 if (!fl6.flowi6_oif)
1153 fl6.flowi6_oif = sk->sk_bound_dev_if;
1154
1155 if (!fl6.flowi6_oif)
1156 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1157
1158 fl6.flowi6_mark = sk->sk_mark;
1159
1160 if (msg->msg_controllen) {
1161 opt = &opt_space;
1162 memset(opt, 0, sizeof(struct ipv6_txoptions));
1163 opt->tot_len = sizeof(*opt);
1164
1165 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1166 &hlimit, &tclass, &dontfrag);
1167 if (err < 0) {
1168 fl6_sock_release(flowlabel);
1169 return err;
1170 }
1171 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1172 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1173 if (flowlabel == NULL)
1174 return -EINVAL;
1175 }
1176 if (!(opt->opt_nflen|opt->opt_flen))
1177 opt = NULL;
1178 connected = 0;
1179 }
1180 if (opt == NULL)
1181 opt = np->opt;
1182 if (flowlabel)
1183 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1184 opt = ipv6_fixup_options(&opt_space, opt);
1185
1186 fl6.flowi6_proto = sk->sk_protocol;
1187 if (!ipv6_addr_any(daddr))
1188 fl6.daddr = *daddr;
1189 else
1190 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1191 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1192 fl6.saddr = np->saddr;
1193 fl6.fl6_sport = inet->inet_sport;
1194
1195 final_p = fl6_update_dst(&fl6, opt, &final);
1196 if (final_p)
1197 connected = 0;
1198
1199 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1200 fl6.flowi6_oif = np->mcast_oif;
1201 connected = 0;
1202 } else if (!fl6.flowi6_oif)
1203 fl6.flowi6_oif = np->ucast_oif;
1204
1205 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1206
1207 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
1208 if (IS_ERR(dst)) {
1209 err = PTR_ERR(dst);
1210 dst = NULL;
1211 goto out;
1212 }
1213
1214 if (hlimit < 0) {
1215 if (ipv6_addr_is_multicast(&fl6.daddr))
1216 hlimit = np->mcast_hops;
1217 else
1218 hlimit = np->hop_limit;
1219 if (hlimit < 0)
1220 hlimit = ip6_dst_hoplimit(dst);
1221 }
1222
1223 if (tclass < 0)
1224 tclass = np->tclass;
1225
1226 if (dontfrag < 0)
1227 dontfrag = np->dontfrag;
1228
1229 if (msg->msg_flags&MSG_CONFIRM)
1230 goto do_confirm;
1231 back_from_confirm:
1232
1233 lock_sock(sk);
1234 if (unlikely(up->pending)) {
1235 /* The socket is already corked while preparing it. */
1236 /* ... which is an evident application bug. --ANK */
1237 release_sock(sk);
1238
1239 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1240 err = -EINVAL;
1241 goto out;
1242 }
1243
1244 up->pending = AF_INET6;
1245
1246 do_append_data:
1247 up->len += ulen;
1248 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1249 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
1250 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1251 (struct rt6_info*)dst,
1252 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1253 if (err)
1254 udp_v6_flush_pending_frames(sk);
1255 else if (!corkreq)
1256 err = udp_v6_push_pending_frames(sk);
1257 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1258 up->pending = 0;
1259
1260 if (dst) {
1261 if (connected) {
1262 ip6_dst_store(sk, dst,
1263 ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
1264 &np->daddr : NULL,
1265 #ifdef CONFIG_IPV6_SUBTREES
1266 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1267 &np->saddr :
1268 #endif
1269 NULL);
1270 } else {
1271 dst_release(dst);
1272 }
1273 dst = NULL;
1274 }
1275
1276 if (err > 0)
1277 err = np->recverr ? net_xmit_errno(err) : 0;
1278 release_sock(sk);
1279 out:
1280 dst_release(dst);
1281 fl6_sock_release(flowlabel);
1282 if (!err)
1283 return len;
1284 /*
1285 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1286 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1287 * we don't have a good statistic (IpOutDiscards but it can be too many
1288 * things). We could add another new stat but at least for now that
1289 * seems like overkill.
1290 */
1291 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1292 UDP6_INC_STATS_USER(sock_net(sk),
1293 UDP_MIB_SNDBUFERRORS, is_udplite);
1294 }
1295 return err;
1296
1297 do_confirm:
1298 dst_confirm(dst);
1299 if (!(msg->msg_flags&MSG_PROBE) || len)
1300 goto back_from_confirm;
1301 err = 0;
1302 goto out;
1303 }
1304
1305 void udpv6_destroy_sock(struct sock *sk)
1306 {
1307 struct udp_sock *up = udp_sk(sk);
1308 lock_sock(sk);
1309 udp_v6_flush_pending_frames(sk);
1310 release_sock(sk);
1311
1312 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1313 void (*encap_destroy)(struct sock *sk);
1314 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1315 if (encap_destroy)
1316 encap_destroy(sk);
1317 }
1318
1319 inet6_destroy_sock(sk);
1320 }
1321
1322 /*
1323 * Socket option code for UDP
1324 */
1325 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1326 char __user *optval, unsigned int optlen)
1327 {
1328 if (level == SOL_UDP || level == SOL_UDPLITE)
1329 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1330 udp_v6_push_pending_frames);
1331 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1332 }
1333
1334 #ifdef CONFIG_COMPAT
1335 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1336 char __user *optval, unsigned int optlen)
1337 {
1338 if (level == SOL_UDP || level == SOL_UDPLITE)
1339 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1340 udp_v6_push_pending_frames);
1341 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1342 }
1343 #endif
1344
1345 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1346 char __user *optval, int __user *optlen)
1347 {
1348 if (level == SOL_UDP || level == SOL_UDPLITE)
1349 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1350 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1351 }
1352
1353 #ifdef CONFIG_COMPAT
1354 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1355 char __user *optval, int __user *optlen)
1356 {
1357 if (level == SOL_UDP || level == SOL_UDPLITE)
1358 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1359 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1360 }
1361 #endif
1362
1363 static const struct inet6_protocol udpv6_protocol = {
1364 .handler = udpv6_rcv,
1365 .err_handler = udpv6_err,
1366 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1367 };
1368
1369 /* ------------------------------------------------------------------------ */
1370 #ifdef CONFIG_PROC_FS
1371 int udp6_seq_show(struct seq_file *seq, void *v)
1372 {
1373 if (v == SEQ_START_TOKEN) {
1374 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1375 } else {
1376 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1377 struct inet_sock *inet = inet_sk(v);
1378 __u16 srcp = ntohs(inet->inet_sport);
1379 __u16 destp = ntohs(inet->inet_dport);
1380 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1381 }
1382 return 0;
1383 }
1384
1385 static const struct file_operations udp6_afinfo_seq_fops = {
1386 .owner = THIS_MODULE,
1387 .open = udp_seq_open,
1388 .read = seq_read,
1389 .llseek = seq_lseek,
1390 .release = seq_release_net
1391 };
1392
1393 static struct udp_seq_afinfo udp6_seq_afinfo = {
1394 .name = "udp6",
1395 .family = AF_INET6,
1396 .udp_table = &udp_table,
1397 .seq_fops = &udp6_afinfo_seq_fops,
1398 .seq_ops = {
1399 .show = udp6_seq_show,
1400 },
1401 };
1402
1403 int __net_init udp6_proc_init(struct net *net)
1404 {
1405 return udp_proc_register(net, &udp6_seq_afinfo);
1406 }
1407
1408 void udp6_proc_exit(struct net *net) {
1409 udp_proc_unregister(net, &udp6_seq_afinfo);
1410 }
1411 #endif /* CONFIG_PROC_FS */
1412
1413 void udp_v6_clear_sk(struct sock *sk, int size)
1414 {
1415 struct inet_sock *inet = inet_sk(sk);
1416
1417 /* we do not want to clear pinet6 field, because of RCU lookups */
1418 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1419
1420 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1421 memset(&inet->pinet6 + 1, 0, size);
1422 }
1423
1424 /* ------------------------------------------------------------------------ */
1425
1426 struct proto udpv6_prot = {
1427 .name = "UDPv6",
1428 .owner = THIS_MODULE,
1429 .close = udp_lib_close,
1430 .connect = ip6_datagram_connect,
1431 .disconnect = udp_disconnect,
1432 .ioctl = udp_ioctl,
1433 .destroy = udpv6_destroy_sock,
1434 .setsockopt = udpv6_setsockopt,
1435 .getsockopt = udpv6_getsockopt,
1436 .sendmsg = udpv6_sendmsg,
1437 .recvmsg = udpv6_recvmsg,
1438 .backlog_rcv = __udpv6_queue_rcv_skb,
1439 .hash = udp_lib_hash,
1440 .unhash = udp_lib_unhash,
1441 .rehash = udp_v6_rehash,
1442 .get_port = udp_v6_get_port,
1443 .memory_allocated = &udp_memory_allocated,
1444 .sysctl_mem = sysctl_udp_mem,
1445 .sysctl_wmem = &sysctl_udp_wmem_min,
1446 .sysctl_rmem = &sysctl_udp_rmem_min,
1447 .obj_size = sizeof(struct udp6_sock),
1448 .slab_flags = SLAB_DESTROY_BY_RCU,
1449 .h.udp_table = &udp_table,
1450 #ifdef CONFIG_COMPAT
1451 .compat_setsockopt = compat_udpv6_setsockopt,
1452 .compat_getsockopt = compat_udpv6_getsockopt,
1453 #endif
1454 .clear_sk = udp_v6_clear_sk,
1455 };
1456
1457 static struct inet_protosw udpv6_protosw = {
1458 .type = SOCK_DGRAM,
1459 .protocol = IPPROTO_UDP,
1460 .prot = &udpv6_prot,
1461 .ops = &inet6_dgram_ops,
1462 .no_check = UDP_CSUM_DEFAULT,
1463 .flags = INET_PROTOSW_PERMANENT,
1464 };
1465
1466
1467 int __init udpv6_init(void)
1468 {
1469 int ret;
1470
1471 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1472 if (ret)
1473 goto out;
1474
1475 ret = inet6_register_protosw(&udpv6_protosw);
1476 if (ret)
1477 goto out_udpv6_protocol;
1478 out:
1479 return ret;
1480
1481 out_udpv6_protocol:
1482 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1483 goto out;
1484 }
1485
1486 void udpv6_exit(void)
1487 {
1488 inet6_unregister_protosw(&udpv6_protosw);
1489 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1490 }
Linus' kernel tree