xref: /linux/net/ipv6/udp.c (revision 2d87650a3bf1b80f7d0d150ee1af3f8a89e5b7aa)
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 unsigned int udp6_ehashfn(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 sk_ipv6only = ipv6_only_sock(sk);
83 	int sk2_ipv6only = inet_v6_ipv6only(sk2);
84 	int addr_type = ipv6_addr_type(&sk->sk_v6_rcv_saddr);
85 	int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
86 
87 	/* if both are mapped, treat as IPv4 */
88 	if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
89 		return (!sk2_ipv6only &&
90 			(!sk->sk_rcv_saddr || !sk2->sk_rcv_saddr ||
91 			  sk->sk_rcv_saddr == sk2->sk_rcv_saddr));
92 
93 	if (addr_type2 == IPV6_ADDR_ANY &&
94 	    !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
95 		return 1;
96 
97 	if (addr_type == IPV6_ADDR_ANY &&
98 	    !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
99 		return 1;
100 
101 	if (sk2_rcv_saddr6 &&
102 	    ipv6_addr_equal(&sk->sk_v6_rcv_saddr, sk2_rcv_saddr6))
103 		return 1;
104 
105 	return 0;
106 }
107 
108 static unsigned int udp6_portaddr_hash(struct net *net,
109 				       const struct in6_addr *addr6,
110 				       unsigned int port)
111 {
112 	unsigned int hash, mix = net_hash_mix(net);
113 
114 	if (ipv6_addr_any(addr6))
115 		hash = jhash_1word(0, mix);
116 	else if (ipv6_addr_v4mapped(addr6))
117 		hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
118 	else
119 		hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
120 
121 	return hash ^ port;
122 }
123 
124 
125 int udp_v6_get_port(struct sock *sk, unsigned short snum)
126 {
127 	unsigned int hash2_nulladdr =
128 		udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
129 	unsigned int hash2_partial =
130 		udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
131 
132 	/* precompute partial secondary hash */
133 	udp_sk(sk)->udp_portaddr_hash = hash2_partial;
134 	return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
135 }
136 
137 static void udp_v6_rehash(struct sock *sk)
138 {
139 	u16 new_hash = udp6_portaddr_hash(sock_net(sk),
140 					  &sk->sk_v6_rcv_saddr,
141 					  inet_sk(sk)->inet_num);
142 
143 	udp_lib_rehash(sk, new_hash);
144 }
145 
146 static inline int compute_score(struct sock *sk, struct net *net,
147 				unsigned short hnum,
148 				const struct in6_addr *saddr, __be16 sport,
149 				const struct in6_addr *daddr, __be16 dport,
150 				int dif)
151 {
152 	int score = -1;
153 
154 	if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
155 			sk->sk_family == PF_INET6) {
156 		struct inet_sock *inet = inet_sk(sk);
157 
158 		score = 0;
159 		if (inet->inet_dport) {
160 			if (inet->inet_dport != sport)
161 				return -1;
162 			score++;
163 		}
164 		if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
165 			if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
166 				return -1;
167 			score++;
168 		}
169 		if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
170 			if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
171 				return -1;
172 			score++;
173 		}
174 		if (sk->sk_bound_dev_if) {
175 			if (sk->sk_bound_dev_if != dif)
176 				return -1;
177 			score++;
178 		}
179 	}
180 	return score;
181 }
182 
183 #define SCORE2_MAX (1 + 1 + 1)
184 static inline int compute_score2(struct sock *sk, struct net *net,
185 				const struct in6_addr *saddr, __be16 sport,
186 				const struct in6_addr *daddr, unsigned short hnum,
187 				int dif)
188 {
189 	int score = -1;
190 
191 	if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
192 			sk->sk_family == PF_INET6) {
193 		struct inet_sock *inet = inet_sk(sk);
194 
195 		if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
196 			return -1;
197 		score = 0;
198 		if (inet->inet_dport) {
199 			if (inet->inet_dport != sport)
200 				return -1;
201 			score++;
202 		}
203 		if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
204 			if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
205 				return -1;
206 			score++;
207 		}
208 		if (sk->sk_bound_dev_if) {
209 			if (sk->sk_bound_dev_if != dif)
210 				return -1;
211 			score++;
212 		}
213 	}
214 	return score;
215 }
216 
217 
218 /* called with read_rcu_lock() */
219 static struct sock *udp6_lib_lookup2(struct net *net,
220 		const struct in6_addr *saddr, __be16 sport,
221 		const struct in6_addr *daddr, unsigned int hnum, int dif,
222 		struct udp_hslot *hslot2, unsigned int slot2)
223 {
224 	struct sock *sk, *result;
225 	struct hlist_nulls_node *node;
226 	int score, badness, matches = 0, reuseport = 0;
227 	u32 hash = 0;
228 
229 begin:
230 	result = NULL;
231 	badness = -1;
232 	udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
233 		score = compute_score2(sk, net, saddr, sport,
234 				      daddr, hnum, dif);
235 		if (score > badness) {
236 			result = sk;
237 			badness = score;
238 			reuseport = sk->sk_reuseport;
239 			if (reuseport) {
240 				hash = udp6_ehashfn(net, daddr, hnum,
241 						    saddr, sport);
242 				matches = 1;
243 			} else if (score == SCORE2_MAX)
244 				goto exact_match;
245 		} else if (score == badness && reuseport) {
246 			matches++;
247 			if (((u64)hash * matches) >> 32 == 0)
248 				result = sk;
249 			hash = next_pseudo_random32(hash);
250 		}
251 	}
252 	/*
253 	 * if the nulls value we got at the end of this lookup is
254 	 * not the expected one, we must restart lookup.
255 	 * We probably met an item that was moved to another chain.
256 	 */
257 	if (get_nulls_value(node) != slot2)
258 		goto begin;
259 
260 	if (result) {
261 exact_match:
262 		if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
263 			result = NULL;
264 		else if (unlikely(compute_score2(result, net, saddr, sport,
265 				  daddr, hnum, dif) < badness)) {
266 			sock_put(result);
267 			goto begin;
268 		}
269 	}
270 	return result;
271 }
272 
273 struct sock *__udp6_lib_lookup(struct net *net,
274 				      const struct in6_addr *saddr, __be16 sport,
275 				      const struct in6_addr *daddr, __be16 dport,
276 				      int dif, struct udp_table *udptable)
277 {
278 	struct sock *sk, *result;
279 	struct hlist_nulls_node *node;
280 	unsigned short hnum = ntohs(dport);
281 	unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
282 	struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
283 	int score, badness, matches = 0, reuseport = 0;
284 	u32 hash = 0;
285 
286 	rcu_read_lock();
287 	if (hslot->count > 10) {
288 		hash2 = udp6_portaddr_hash(net, daddr, hnum);
289 		slot2 = hash2 & udptable->mask;
290 		hslot2 = &udptable->hash2[slot2];
291 		if (hslot->count < hslot2->count)
292 			goto begin;
293 
294 		result = udp6_lib_lookup2(net, saddr, sport,
295 					  daddr, hnum, dif,
296 					  hslot2, slot2);
297 		if (!result) {
298 			hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
299 			slot2 = hash2 & udptable->mask;
300 			hslot2 = &udptable->hash2[slot2];
301 			if (hslot->count < hslot2->count)
302 				goto begin;
303 
304 			result = udp6_lib_lookup2(net, saddr, sport,
305 						  &in6addr_any, hnum, dif,
306 						  hslot2, slot2);
307 		}
308 		rcu_read_unlock();
309 		return result;
310 	}
311 begin:
312 	result = NULL;
313 	badness = -1;
314 	sk_nulls_for_each_rcu(sk, node, &hslot->head) {
315 		score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
316 		if (score > badness) {
317 			result = sk;
318 			badness = score;
319 			reuseport = sk->sk_reuseport;
320 			if (reuseport) {
321 				hash = udp6_ehashfn(net, daddr, hnum,
322 						    saddr, sport);
323 				matches = 1;
324 			}
325 		} else if (score == badness && reuseport) {
326 			matches++;
327 			if (((u64)hash * matches) >> 32 == 0)
328 				result = sk;
329 			hash = next_pseudo_random32(hash);
330 		}
331 	}
332 	/*
333 	 * if the nulls value we got at the end of this lookup is
334 	 * not the expected one, we must restart lookup.
335 	 * We probably met an item that was moved to another chain.
336 	 */
337 	if (get_nulls_value(node) != slot)
338 		goto begin;
339 
340 	if (result) {
341 		if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
342 			result = NULL;
343 		else if (unlikely(compute_score(result, net, hnum, saddr, sport,
344 					daddr, dport, dif) < badness)) {
345 			sock_put(result);
346 			goto begin;
347 		}
348 	}
349 	rcu_read_unlock();
350 	return result;
351 }
352 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
353 
354 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
355 					  __be16 sport, __be16 dport,
356 					  struct udp_table *udptable)
357 {
358 	struct sock *sk;
359 	const struct ipv6hdr *iph = ipv6_hdr(skb);
360 
361 	if (unlikely(sk = skb_steal_sock(skb)))
362 		return sk;
363 	return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
364 				 &iph->daddr, dport, inet6_iif(skb),
365 				 udptable);
366 }
367 
368 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
369 			     const struct in6_addr *daddr, __be16 dport, int dif)
370 {
371 	return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
372 }
373 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
374 
375 
376 /*
377  * 	This should be easy, if there is something there we
378  * 	return it, otherwise we block.
379  */
380 
381 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
382 		  struct msghdr *msg, size_t len,
383 		  int noblock, int flags, int *addr_len)
384 {
385 	struct ipv6_pinfo *np = inet6_sk(sk);
386 	struct inet_sock *inet = inet_sk(sk);
387 	struct sk_buff *skb;
388 	unsigned int ulen, copied;
389 	int peeked, off = 0;
390 	int err;
391 	int is_udplite = IS_UDPLITE(sk);
392 	int is_udp4;
393 	bool slow;
394 
395 	if (flags & MSG_ERRQUEUE)
396 		return ipv6_recv_error(sk, msg, len, addr_len);
397 
398 	if (np->rxpmtu && np->rxopt.bits.rxpmtu)
399 		return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
400 
401 try_again:
402 	skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
403 				  &peeked, &off, &err);
404 	if (!skb)
405 		goto out;
406 
407 	ulen = skb->len - sizeof(struct udphdr);
408 	copied = len;
409 	if (copied > ulen)
410 		copied = ulen;
411 	else if (copied < ulen)
412 		msg->msg_flags |= MSG_TRUNC;
413 
414 	is_udp4 = (skb->protocol == htons(ETH_P_IP));
415 
416 	/*
417 	 * If checksum is needed at all, try to do it while copying the
418 	 * data.  If the data is truncated, or if we only want a partial
419 	 * coverage checksum (UDP-Lite), do it before the copy.
420 	 */
421 
422 	if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
423 		if (udp_lib_checksum_complete(skb))
424 			goto csum_copy_err;
425 	}
426 
427 	if (skb_csum_unnecessary(skb))
428 		err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
429 					      msg->msg_iov, copied);
430 	else {
431 		err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
432 		if (err == -EINVAL)
433 			goto csum_copy_err;
434 	}
435 	if (unlikely(err)) {
436 		trace_kfree_skb(skb, udpv6_recvmsg);
437 		if (!peeked) {
438 			atomic_inc(&sk->sk_drops);
439 			if (is_udp4)
440 				UDP_INC_STATS_USER(sock_net(sk),
441 						   UDP_MIB_INERRORS,
442 						   is_udplite);
443 			else
444 				UDP6_INC_STATS_USER(sock_net(sk),
445 						    UDP_MIB_INERRORS,
446 						    is_udplite);
447 		}
448 		goto out_free;
449 	}
450 	if (!peeked) {
451 		if (is_udp4)
452 			UDP_INC_STATS_USER(sock_net(sk),
453 					UDP_MIB_INDATAGRAMS, is_udplite);
454 		else
455 			UDP6_INC_STATS_USER(sock_net(sk),
456 					UDP_MIB_INDATAGRAMS, is_udplite);
457 	}
458 
459 	sock_recv_ts_and_drops(msg, sk, skb);
460 
461 	/* Copy the address. */
462 	if (msg->msg_name) {
463 		struct sockaddr_in6 *sin6;
464 
465 		sin6 = (struct sockaddr_in6 *) msg->msg_name;
466 		sin6->sin6_family = AF_INET6;
467 		sin6->sin6_port = udp_hdr(skb)->source;
468 		sin6->sin6_flowinfo = 0;
469 
470 		if (is_udp4) {
471 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
472 					       &sin6->sin6_addr);
473 			sin6->sin6_scope_id = 0;
474 		} else {
475 			sin6->sin6_addr = ipv6_hdr(skb)->saddr;
476 			sin6->sin6_scope_id =
477 				ipv6_iface_scope_id(&sin6->sin6_addr,
478 						    IP6CB(skb)->iif);
479 		}
480 		*addr_len = sizeof(*sin6);
481 	}
482 	if (is_udp4) {
483 		if (inet->cmsg_flags)
484 			ip_cmsg_recv(msg, skb);
485 	} else {
486 		if (np->rxopt.all)
487 			ip6_datagram_recv_ctl(sk, msg, skb);
488 	}
489 
490 	err = copied;
491 	if (flags & MSG_TRUNC)
492 		err = ulen;
493 
494 out_free:
495 	skb_free_datagram_locked(sk, skb);
496 out:
497 	return err;
498 
499 csum_copy_err:
500 	slow = lock_sock_fast(sk);
501 	if (!skb_kill_datagram(sk, skb, flags)) {
502 		if (is_udp4) {
503 			UDP_INC_STATS_USER(sock_net(sk),
504 					UDP_MIB_CSUMERRORS, is_udplite);
505 			UDP_INC_STATS_USER(sock_net(sk),
506 					UDP_MIB_INERRORS, is_udplite);
507 		} else {
508 			UDP6_INC_STATS_USER(sock_net(sk),
509 					UDP_MIB_CSUMERRORS, is_udplite);
510 			UDP6_INC_STATS_USER(sock_net(sk),
511 					UDP_MIB_INERRORS, is_udplite);
512 		}
513 	}
514 	unlock_sock_fast(sk, slow);
515 
516 	if (noblock)
517 		return -EAGAIN;
518 
519 	/* starting over for a new packet */
520 	msg->msg_flags &= ~MSG_TRUNC;
521 	goto try_again;
522 }
523 
524 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
525 		    u8 type, u8 code, int offset, __be32 info,
526 		    struct udp_table *udptable)
527 {
528 	struct ipv6_pinfo *np;
529 	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
530 	const struct in6_addr *saddr = &hdr->saddr;
531 	const struct in6_addr *daddr = &hdr->daddr;
532 	struct udphdr *uh = (struct udphdr*)(skb->data+offset);
533 	struct sock *sk;
534 	int err;
535 
536 	sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
537 			       saddr, uh->source, inet6_iif(skb), udptable);
538 	if (sk == NULL)
539 		return;
540 
541 	if (type == ICMPV6_PKT_TOOBIG) {
542 		if (!ip6_sk_accept_pmtu(sk))
543 			goto out;
544 		ip6_sk_update_pmtu(skb, sk, info);
545 	}
546 	if (type == NDISC_REDIRECT) {
547 		ip6_sk_redirect(skb, sk);
548 		goto out;
549 	}
550 
551 	np = inet6_sk(sk);
552 
553 	if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
554 		goto out;
555 
556 	if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
557 		goto out;
558 
559 	if (np->recverr)
560 		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
561 
562 	sk->sk_err = err;
563 	sk->sk_error_report(sk);
564 out:
565 	sock_put(sk);
566 }
567 
568 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
569 {
570 	int rc;
571 
572 	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
573 		sock_rps_save_rxhash(sk, skb);
574 		sk_mark_napi_id(sk, skb);
575 	}
576 
577 	rc = sock_queue_rcv_skb(sk, skb);
578 	if (rc < 0) {
579 		int is_udplite = IS_UDPLITE(sk);
580 
581 		/* Note that an ENOMEM error is charged twice */
582 		if (rc == -ENOMEM)
583 			UDP6_INC_STATS_BH(sock_net(sk),
584 					UDP_MIB_RCVBUFERRORS, is_udplite);
585 		UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
586 		kfree_skb(skb);
587 		return -1;
588 	}
589 	return 0;
590 }
591 
592 static __inline__ void udpv6_err(struct sk_buff *skb,
593 				 struct inet6_skb_parm *opt, u8 type,
594 				 u8 code, int offset, __be32 info     )
595 {
596 	__udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
597 }
598 
599 static struct static_key udpv6_encap_needed __read_mostly;
600 void udpv6_encap_enable(void)
601 {
602 	if (!static_key_enabled(&udpv6_encap_needed))
603 		static_key_slow_inc(&udpv6_encap_needed);
604 }
605 EXPORT_SYMBOL(udpv6_encap_enable);
606 
607 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
608 {
609 	struct udp_sock *up = udp_sk(sk);
610 	int rc;
611 	int is_udplite = IS_UDPLITE(sk);
612 
613 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
614 		goto drop;
615 
616 	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
617 		int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
618 
619 		/*
620 		 * This is an encapsulation socket so pass the skb to
621 		 * the socket's udp_encap_rcv() hook. Otherwise, just
622 		 * fall through and pass this up the UDP socket.
623 		 * up->encap_rcv() returns the following value:
624 		 * =0 if skb was successfully passed to the encap
625 		 *    handler or was discarded by it.
626 		 * >0 if skb should be passed on to UDP.
627 		 * <0 if skb should be resubmitted as proto -N
628 		 */
629 
630 		/* if we're overly short, let UDP handle it */
631 		encap_rcv = ACCESS_ONCE(up->encap_rcv);
632 		if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
633 			int ret;
634 
635 			ret = encap_rcv(sk, skb);
636 			if (ret <= 0) {
637 				UDP_INC_STATS_BH(sock_net(sk),
638 						 UDP_MIB_INDATAGRAMS,
639 						 is_udplite);
640 				return -ret;
641 			}
642 		}
643 
644 		/* FALLTHROUGH -- it's a UDP Packet */
645 	}
646 
647 	/*
648 	 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
649 	 */
650 	if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
651 
652 		if (up->pcrlen == 0) {          /* full coverage was set  */
653 			LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
654 				" %d while full coverage %d requested\n",
655 				UDP_SKB_CB(skb)->cscov, skb->len);
656 			goto drop;
657 		}
658 		if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
659 			LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
660 						    "too small, need min %d\n",
661 				       UDP_SKB_CB(skb)->cscov, up->pcrlen);
662 			goto drop;
663 		}
664 	}
665 
666 	if (rcu_access_pointer(sk->sk_filter)) {
667 		if (udp_lib_checksum_complete(skb))
668 			goto csum_error;
669 	}
670 
671 	if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
672 		goto drop;
673 
674 	skb_dst_drop(skb);
675 
676 	bh_lock_sock(sk);
677 	rc = 0;
678 	if (!sock_owned_by_user(sk))
679 		rc = __udpv6_queue_rcv_skb(sk, skb);
680 	else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
681 		bh_unlock_sock(sk);
682 		goto drop;
683 	}
684 	bh_unlock_sock(sk);
685 
686 	return rc;
687 csum_error:
688 	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
689 drop:
690 	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
691 	atomic_inc(&sk->sk_drops);
692 	kfree_skb(skb);
693 	return -1;
694 }
695 
696 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
697 				      __be16 loc_port, const struct in6_addr *loc_addr,
698 				      __be16 rmt_port, const struct in6_addr *rmt_addr,
699 				      int dif)
700 {
701 	struct hlist_nulls_node *node;
702 	struct sock *s = sk;
703 	unsigned short num = ntohs(loc_port);
704 
705 	sk_nulls_for_each_from(s, node) {
706 		struct inet_sock *inet = inet_sk(s);
707 
708 		if (!net_eq(sock_net(s), net))
709 			continue;
710 
711 		if (udp_sk(s)->udp_port_hash == num &&
712 		    s->sk_family == PF_INET6) {
713 			if (inet->inet_dport) {
714 				if (inet->inet_dport != rmt_port)
715 					continue;
716 			}
717 			if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
718 			    !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
719 				continue;
720 
721 			if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
722 				continue;
723 
724 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
725 				if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
726 					continue;
727 			}
728 			if (!inet6_mc_check(s, loc_addr, rmt_addr))
729 				continue;
730 			return s;
731 		}
732 	}
733 	return NULL;
734 }
735 
736 static void flush_stack(struct sock **stack, unsigned int count,
737 			struct sk_buff *skb, unsigned int final)
738 {
739 	struct sk_buff *skb1 = NULL;
740 	struct sock *sk;
741 	unsigned int i;
742 
743 	for (i = 0; i < count; i++) {
744 		sk = stack[i];
745 		if (likely(skb1 == NULL))
746 			skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
747 		if (!skb1) {
748 			atomic_inc(&sk->sk_drops);
749 			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
750 					  IS_UDPLITE(sk));
751 			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
752 					  IS_UDPLITE(sk));
753 		}
754 
755 		if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
756 			skb1 = NULL;
757 	}
758 	if (unlikely(skb1))
759 		kfree_skb(skb1);
760 }
761 /*
762  * Note: called only from the BH handler context,
763  * so we don't need to lock the hashes.
764  */
765 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
766 		const struct in6_addr *saddr, const struct in6_addr *daddr,
767 		struct udp_table *udptable)
768 {
769 	struct sock *sk, *stack[256 / sizeof(struct sock *)];
770 	const struct udphdr *uh = udp_hdr(skb);
771 	struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
772 	int dif;
773 	unsigned int i, count = 0;
774 
775 	spin_lock(&hslot->lock);
776 	sk = sk_nulls_head(&hslot->head);
777 	dif = inet6_iif(skb);
778 	sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
779 	while (sk) {
780 		stack[count++] = sk;
781 		sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
782 				       uh->source, saddr, dif);
783 		if (unlikely(count == ARRAY_SIZE(stack))) {
784 			if (!sk)
785 				break;
786 			flush_stack(stack, count, skb, ~0);
787 			count = 0;
788 		}
789 	}
790 	/*
791 	 * before releasing the lock, we must take reference on sockets
792 	 */
793 	for (i = 0; i < count; i++)
794 		sock_hold(stack[i]);
795 
796 	spin_unlock(&hslot->lock);
797 
798 	if (count) {
799 		flush_stack(stack, count, skb, count - 1);
800 
801 		for (i = 0; i < count; i++)
802 			sock_put(stack[i]);
803 	} else {
804 		kfree_skb(skb);
805 	}
806 	return 0;
807 }
808 
809 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
810 		   int proto)
811 {
812 	struct net *net = dev_net(skb->dev);
813 	struct sock *sk;
814 	struct udphdr *uh;
815 	const struct in6_addr *saddr, *daddr;
816 	u32 ulen = 0;
817 
818 	if (!pskb_may_pull(skb, sizeof(struct udphdr)))
819 		goto discard;
820 
821 	saddr = &ipv6_hdr(skb)->saddr;
822 	daddr = &ipv6_hdr(skb)->daddr;
823 	uh = udp_hdr(skb);
824 
825 	ulen = ntohs(uh->len);
826 	if (ulen > skb->len)
827 		goto short_packet;
828 
829 	if (proto == IPPROTO_UDP) {
830 		/* UDP validates ulen. */
831 
832 		/* Check for jumbo payload */
833 		if (ulen == 0)
834 			ulen = skb->len;
835 
836 		if (ulen < sizeof(*uh))
837 			goto short_packet;
838 
839 		if (ulen < skb->len) {
840 			if (pskb_trim_rcsum(skb, ulen))
841 				goto short_packet;
842 			saddr = &ipv6_hdr(skb)->saddr;
843 			daddr = &ipv6_hdr(skb)->daddr;
844 			uh = udp_hdr(skb);
845 		}
846 	}
847 
848 	if (udp6_csum_init(skb, uh, proto))
849 		goto csum_error;
850 
851 	/*
852 	 *	Multicast receive code
853 	 */
854 	if (ipv6_addr_is_multicast(daddr))
855 		return __udp6_lib_mcast_deliver(net, skb,
856 				saddr, daddr, udptable);
857 
858 	/* Unicast */
859 
860 	/*
861 	 * check socket cache ... must talk to Alan about his plans
862 	 * for sock caches... i'll skip this for now.
863 	 */
864 	sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
865 	if (sk != NULL) {
866 		int ret;
867 
868 		ret = udpv6_queue_rcv_skb(sk, skb);
869 		sock_put(sk);
870 
871 		/* a return value > 0 means to resubmit the input, but
872 		 * it wants the return to be -protocol, or 0
873 		 */
874 		if (ret > 0)
875 			return -ret;
876 
877 		return 0;
878 	}
879 
880 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
881 		goto discard;
882 
883 	if (udp_lib_checksum_complete(skb))
884 		goto csum_error;
885 
886 	UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
887 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
888 
889 	kfree_skb(skb);
890 	return 0;
891 
892 short_packet:
893 	LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
894 		       proto == IPPROTO_UDPLITE ? "-Lite" : "",
895 		       saddr,
896 		       ntohs(uh->source),
897 		       ulen,
898 		       skb->len,
899 		       daddr,
900 		       ntohs(uh->dest));
901 	goto discard;
902 csum_error:
903 	UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
904 discard:
905 	UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
906 	kfree_skb(skb);
907 	return 0;
908 }
909 
910 static __inline__ int udpv6_rcv(struct sk_buff *skb)
911 {
912 	return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
913 }
914 
915 /*
916  * Throw away all pending data and cancel the corking. Socket is locked.
917  */
918 static void udp_v6_flush_pending_frames(struct sock *sk)
919 {
920 	struct udp_sock *up = udp_sk(sk);
921 
922 	if (up->pending == AF_INET)
923 		udp_flush_pending_frames(sk);
924 	else if (up->pending) {
925 		up->len = 0;
926 		up->pending = 0;
927 		ip6_flush_pending_frames(sk);
928 	}
929 }
930 
931 /**
932  * 	udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
933  * 	@sk: 	socket we are sending on
934  * 	@skb: 	sk_buff containing the filled-in UDP header
935  * 	        (checksum field must be zeroed out)
936  */
937 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
938 				 const struct in6_addr *saddr,
939 				 const struct in6_addr *daddr, int len)
940 {
941 	unsigned int offset;
942 	struct udphdr *uh = udp_hdr(skb);
943 	__wsum csum = 0;
944 
945 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
946 		/* Only one fragment on the socket.  */
947 		skb->csum_start = skb_transport_header(skb) - skb->head;
948 		skb->csum_offset = offsetof(struct udphdr, check);
949 		uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
950 	} else {
951 		/*
952 		 * HW-checksum won't work as there are two or more
953 		 * fragments on the socket so that all csums of sk_buffs
954 		 * should be together
955 		 */
956 		offset = skb_transport_offset(skb);
957 		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
958 
959 		skb->ip_summed = CHECKSUM_NONE;
960 
961 		skb_queue_walk(&sk->sk_write_queue, skb) {
962 			csum = csum_add(csum, skb->csum);
963 		}
964 
965 		uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
966 					    csum);
967 		if (uh->check == 0)
968 			uh->check = CSUM_MANGLED_0;
969 	}
970 }
971 
972 /*
973  *	Sending
974  */
975 
976 static int udp_v6_push_pending_frames(struct sock *sk)
977 {
978 	struct sk_buff *skb;
979 	struct udphdr *uh;
980 	struct udp_sock  *up = udp_sk(sk);
981 	struct inet_sock *inet = inet_sk(sk);
982 	struct flowi6 *fl6;
983 	int err = 0;
984 	int is_udplite = IS_UDPLITE(sk);
985 	__wsum csum = 0;
986 
987 	if (up->pending == AF_INET)
988 		return udp_push_pending_frames(sk);
989 
990 	fl6 = &inet->cork.fl.u.ip6;
991 
992 	/* Grab the skbuff where UDP header space exists. */
993 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
994 		goto out;
995 
996 	/*
997 	 * Create a UDP header
998 	 */
999 	uh = udp_hdr(skb);
1000 	uh->source = fl6->fl6_sport;
1001 	uh->dest = fl6->fl6_dport;
1002 	uh->len = htons(up->len);
1003 	uh->check = 0;
1004 
1005 	if (is_udplite)
1006 		csum = udplite_csum_outgoing(sk, skb);
1007 	else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1008 		udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
1009 				     up->len);
1010 		goto send;
1011 	} else
1012 		csum = udp_csum_outgoing(sk, skb);
1013 
1014 	/* add protocol-dependent pseudo-header */
1015 	uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1016 				    up->len, fl6->flowi6_proto, csum);
1017 	if (uh->check == 0)
1018 		uh->check = CSUM_MANGLED_0;
1019 
1020 send:
1021 	err = ip6_push_pending_frames(sk);
1022 	if (err) {
1023 		if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1024 			UDP6_INC_STATS_USER(sock_net(sk),
1025 					    UDP_MIB_SNDBUFERRORS, is_udplite);
1026 			err = 0;
1027 		}
1028 	} else
1029 		UDP6_INC_STATS_USER(sock_net(sk),
1030 				    UDP_MIB_OUTDATAGRAMS, is_udplite);
1031 out:
1032 	up->len = 0;
1033 	up->pending = 0;
1034 	return err;
1035 }
1036 
1037 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1038 		  struct msghdr *msg, size_t len)
1039 {
1040 	struct ipv6_txoptions opt_space;
1041 	struct udp_sock *up = udp_sk(sk);
1042 	struct inet_sock *inet = inet_sk(sk);
1043 	struct ipv6_pinfo *np = inet6_sk(sk);
1044 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
1045 	struct in6_addr *daddr, *final_p, final;
1046 	struct ipv6_txoptions *opt = NULL;
1047 	struct ip6_flowlabel *flowlabel = NULL;
1048 	struct flowi6 fl6;
1049 	struct dst_entry *dst;
1050 	int addr_len = msg->msg_namelen;
1051 	int ulen = len;
1052 	int hlimit = -1;
1053 	int tclass = -1;
1054 	int dontfrag = -1;
1055 	int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1056 	int err;
1057 	int connected = 0;
1058 	int is_udplite = IS_UDPLITE(sk);
1059 	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1060 
1061 	/* destination address check */
1062 	if (sin6) {
1063 		if (addr_len < offsetof(struct sockaddr, sa_data))
1064 			return -EINVAL;
1065 
1066 		switch (sin6->sin6_family) {
1067 		case AF_INET6:
1068 			if (addr_len < SIN6_LEN_RFC2133)
1069 				return -EINVAL;
1070 			daddr = &sin6->sin6_addr;
1071 			break;
1072 		case AF_INET:
1073 			goto do_udp_sendmsg;
1074 		case AF_UNSPEC:
1075 			msg->msg_name = sin6 = NULL;
1076 			msg->msg_namelen = addr_len = 0;
1077 			daddr = NULL;
1078 			break;
1079 		default:
1080 			return -EINVAL;
1081 		}
1082 	} else if (!up->pending) {
1083 		if (sk->sk_state != TCP_ESTABLISHED)
1084 			return -EDESTADDRREQ;
1085 		daddr = &sk->sk_v6_daddr;
1086 	} else
1087 		daddr = NULL;
1088 
1089 	if (daddr) {
1090 		if (ipv6_addr_v4mapped(daddr)) {
1091 			struct sockaddr_in sin;
1092 			sin.sin_family = AF_INET;
1093 			sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1094 			sin.sin_addr.s_addr = daddr->s6_addr32[3];
1095 			msg->msg_name = &sin;
1096 			msg->msg_namelen = sizeof(sin);
1097 do_udp_sendmsg:
1098 			if (__ipv6_only_sock(sk))
1099 				return -ENETUNREACH;
1100 			return udp_sendmsg(iocb, sk, msg, len);
1101 		}
1102 	}
1103 
1104 	if (up->pending == AF_INET)
1105 		return udp_sendmsg(iocb, sk, msg, len);
1106 
1107 	/* Rough check on arithmetic overflow,
1108 	   better check is made in ip6_append_data().
1109 	   */
1110 	if (len > INT_MAX - sizeof(struct udphdr))
1111 		return -EMSGSIZE;
1112 
1113 	if (up->pending) {
1114 		/*
1115 		 * There are pending frames.
1116 		 * The socket lock must be held while it's corked.
1117 		 */
1118 		lock_sock(sk);
1119 		if (likely(up->pending)) {
1120 			if (unlikely(up->pending != AF_INET6)) {
1121 				release_sock(sk);
1122 				return -EAFNOSUPPORT;
1123 			}
1124 			dst = NULL;
1125 			goto do_append_data;
1126 		}
1127 		release_sock(sk);
1128 	}
1129 	ulen += sizeof(struct udphdr);
1130 
1131 	memset(&fl6, 0, sizeof(fl6));
1132 
1133 	if (sin6) {
1134 		if (sin6->sin6_port == 0)
1135 			return -EINVAL;
1136 
1137 		fl6.fl6_dport = sin6->sin6_port;
1138 		daddr = &sin6->sin6_addr;
1139 
1140 		if (np->sndflow) {
1141 			fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1142 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1143 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1144 				if (flowlabel == NULL)
1145 					return -EINVAL;
1146 			}
1147 		}
1148 
1149 		/*
1150 		 * Otherwise it will be difficult to maintain
1151 		 * sk->sk_dst_cache.
1152 		 */
1153 		if (sk->sk_state == TCP_ESTABLISHED &&
1154 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1155 			daddr = &sk->sk_v6_daddr;
1156 
1157 		if (addr_len >= sizeof(struct sockaddr_in6) &&
1158 		    sin6->sin6_scope_id &&
1159 		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1160 			fl6.flowi6_oif = sin6->sin6_scope_id;
1161 	} else {
1162 		if (sk->sk_state != TCP_ESTABLISHED)
1163 			return -EDESTADDRREQ;
1164 
1165 		fl6.fl6_dport = inet->inet_dport;
1166 		daddr = &sk->sk_v6_daddr;
1167 		fl6.flowlabel = np->flow_label;
1168 		connected = 1;
1169 	}
1170 
1171 	if (!fl6.flowi6_oif)
1172 		fl6.flowi6_oif = sk->sk_bound_dev_if;
1173 
1174 	if (!fl6.flowi6_oif)
1175 		fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1176 
1177 	fl6.flowi6_mark = sk->sk_mark;
1178 
1179 	if (msg->msg_controllen) {
1180 		opt = &opt_space;
1181 		memset(opt, 0, sizeof(struct ipv6_txoptions));
1182 		opt->tot_len = sizeof(*opt);
1183 
1184 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1185 					    &hlimit, &tclass, &dontfrag);
1186 		if (err < 0) {
1187 			fl6_sock_release(flowlabel);
1188 			return err;
1189 		}
1190 		if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1191 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1192 			if (flowlabel == NULL)
1193 				return -EINVAL;
1194 		}
1195 		if (!(opt->opt_nflen|opt->opt_flen))
1196 			opt = NULL;
1197 		connected = 0;
1198 	}
1199 	if (opt == NULL)
1200 		opt = np->opt;
1201 	if (flowlabel)
1202 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
1203 	opt = ipv6_fixup_options(&opt_space, opt);
1204 
1205 	fl6.flowi6_proto = sk->sk_protocol;
1206 	if (!ipv6_addr_any(daddr))
1207 		fl6.daddr = *daddr;
1208 	else
1209 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1210 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1211 		fl6.saddr = np->saddr;
1212 	fl6.fl6_sport = inet->inet_sport;
1213 
1214 	final_p = fl6_update_dst(&fl6, opt, &final);
1215 	if (final_p)
1216 		connected = 0;
1217 
1218 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1219 		fl6.flowi6_oif = np->mcast_oif;
1220 		connected = 0;
1221 	} else if (!fl6.flowi6_oif)
1222 		fl6.flowi6_oif = np->ucast_oif;
1223 
1224 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1225 
1226 	dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1227 	if (IS_ERR(dst)) {
1228 		err = PTR_ERR(dst);
1229 		dst = NULL;
1230 		goto out;
1231 	}
1232 
1233 	if (hlimit < 0) {
1234 		if (ipv6_addr_is_multicast(&fl6.daddr))
1235 			hlimit = np->mcast_hops;
1236 		else
1237 			hlimit = np->hop_limit;
1238 		if (hlimit < 0)
1239 			hlimit = ip6_dst_hoplimit(dst);
1240 	}
1241 
1242 	if (tclass < 0)
1243 		tclass = np->tclass;
1244 
1245 	if (msg->msg_flags&MSG_CONFIRM)
1246 		goto do_confirm;
1247 back_from_confirm:
1248 
1249 	lock_sock(sk);
1250 	if (unlikely(up->pending)) {
1251 		/* The socket is already corked while preparing it. */
1252 		/* ... which is an evident application bug. --ANK */
1253 		release_sock(sk);
1254 
1255 		LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1256 		err = -EINVAL;
1257 		goto out;
1258 	}
1259 
1260 	up->pending = AF_INET6;
1261 
1262 do_append_data:
1263 	if (dontfrag < 0)
1264 		dontfrag = np->dontfrag;
1265 	up->len += ulen;
1266 	getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1267 	err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
1268 		sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1269 		(struct rt6_info*)dst,
1270 		corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1271 	if (err)
1272 		udp_v6_flush_pending_frames(sk);
1273 	else if (!corkreq)
1274 		err = udp_v6_push_pending_frames(sk);
1275 	else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1276 		up->pending = 0;
1277 
1278 	if (dst) {
1279 		if (connected) {
1280 			ip6_dst_store(sk, dst,
1281 				      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1282 				      &sk->sk_v6_daddr : NULL,
1283 #ifdef CONFIG_IPV6_SUBTREES
1284 				      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1285 				      &np->saddr :
1286 #endif
1287 				      NULL);
1288 		} else {
1289 			dst_release(dst);
1290 		}
1291 		dst = NULL;
1292 	}
1293 
1294 	if (err > 0)
1295 		err = np->recverr ? net_xmit_errno(err) : 0;
1296 	release_sock(sk);
1297 out:
1298 	dst_release(dst);
1299 	fl6_sock_release(flowlabel);
1300 	if (!err)
1301 		return len;
1302 	/*
1303 	 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1304 	 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1305 	 * we don't have a good statistic (IpOutDiscards but it can be too many
1306 	 * things).  We could add another new stat but at least for now that
1307 	 * seems like overkill.
1308 	 */
1309 	if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1310 		UDP6_INC_STATS_USER(sock_net(sk),
1311 				UDP_MIB_SNDBUFERRORS, is_udplite);
1312 	}
1313 	return err;
1314 
1315 do_confirm:
1316 	dst_confirm(dst);
1317 	if (!(msg->msg_flags&MSG_PROBE) || len)
1318 		goto back_from_confirm;
1319 	err = 0;
1320 	goto out;
1321 }
1322 
1323 void udpv6_destroy_sock(struct sock *sk)
1324 {
1325 	struct udp_sock *up = udp_sk(sk);
1326 	lock_sock(sk);
1327 	udp_v6_flush_pending_frames(sk);
1328 	release_sock(sk);
1329 
1330 	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1331 		void (*encap_destroy)(struct sock *sk);
1332 		encap_destroy = ACCESS_ONCE(up->encap_destroy);
1333 		if (encap_destroy)
1334 			encap_destroy(sk);
1335 	}
1336 
1337 	inet6_destroy_sock(sk);
1338 }
1339 
1340 /*
1341  *	Socket option code for UDP
1342  */
1343 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1344 		     char __user *optval, unsigned int optlen)
1345 {
1346 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1347 		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1348 					  udp_v6_push_pending_frames);
1349 	return ipv6_setsockopt(sk, level, optname, optval, optlen);
1350 }
1351 
1352 #ifdef CONFIG_COMPAT
1353 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1354 			    char __user *optval, unsigned int optlen)
1355 {
1356 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1357 		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1358 					  udp_v6_push_pending_frames);
1359 	return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1360 }
1361 #endif
1362 
1363 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1364 		     char __user *optval, int __user *optlen)
1365 {
1366 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1367 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1368 	return ipv6_getsockopt(sk, level, optname, optval, optlen);
1369 }
1370 
1371 #ifdef CONFIG_COMPAT
1372 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1373 			    char __user *optval, int __user *optlen)
1374 {
1375 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1376 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1377 	return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1378 }
1379 #endif
1380 
1381 static const struct inet6_protocol udpv6_protocol = {
1382 	.handler	=	udpv6_rcv,
1383 	.err_handler	=	udpv6_err,
1384 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1385 };
1386 
1387 /* ------------------------------------------------------------------------ */
1388 #ifdef CONFIG_PROC_FS
1389 int udp6_seq_show(struct seq_file *seq, void *v)
1390 {
1391 	if (v == SEQ_START_TOKEN) {
1392 		seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1393 	} else {
1394 		int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1395 		struct inet_sock *inet = inet_sk(v);
1396 		__u16 srcp = ntohs(inet->inet_sport);
1397 		__u16 destp = ntohs(inet->inet_dport);
1398 		ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1399 	}
1400 	return 0;
1401 }
1402 
1403 static const struct file_operations udp6_afinfo_seq_fops = {
1404 	.owner    = THIS_MODULE,
1405 	.open     = udp_seq_open,
1406 	.read     = seq_read,
1407 	.llseek   = seq_lseek,
1408 	.release  = seq_release_net
1409 };
1410 
1411 static struct udp_seq_afinfo udp6_seq_afinfo = {
1412 	.name		= "udp6",
1413 	.family		= AF_INET6,
1414 	.udp_table	= &udp_table,
1415 	.seq_fops	= &udp6_afinfo_seq_fops,
1416 	.seq_ops	= {
1417 		.show		= udp6_seq_show,
1418 	},
1419 };
1420 
1421 int __net_init udp6_proc_init(struct net *net)
1422 {
1423 	return udp_proc_register(net, &udp6_seq_afinfo);
1424 }
1425 
1426 void udp6_proc_exit(struct net *net) {
1427 	udp_proc_unregister(net, &udp6_seq_afinfo);
1428 }
1429 #endif /* CONFIG_PROC_FS */
1430 
1431 void udp_v6_clear_sk(struct sock *sk, int size)
1432 {
1433 	struct inet_sock *inet = inet_sk(sk);
1434 
1435 	/* we do not want to clear pinet6 field, because of RCU lookups */
1436 	sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1437 
1438 	size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1439 	memset(&inet->pinet6 + 1, 0, size);
1440 }
1441 
1442 /* ------------------------------------------------------------------------ */
1443 
1444 struct proto udpv6_prot = {
1445 	.name		   = "UDPv6",
1446 	.owner		   = THIS_MODULE,
1447 	.close		   = udp_lib_close,
1448 	.connect	   = ip6_datagram_connect,
1449 	.disconnect	   = udp_disconnect,
1450 	.ioctl		   = udp_ioctl,
1451 	.destroy	   = udpv6_destroy_sock,
1452 	.setsockopt	   = udpv6_setsockopt,
1453 	.getsockopt	   = udpv6_getsockopt,
1454 	.sendmsg	   = udpv6_sendmsg,
1455 	.recvmsg	   = udpv6_recvmsg,
1456 	.backlog_rcv	   = __udpv6_queue_rcv_skb,
1457 	.hash		   = udp_lib_hash,
1458 	.unhash		   = udp_lib_unhash,
1459 	.rehash		   = udp_v6_rehash,
1460 	.get_port	   = udp_v6_get_port,
1461 	.memory_allocated  = &udp_memory_allocated,
1462 	.sysctl_mem	   = sysctl_udp_mem,
1463 	.sysctl_wmem	   = &sysctl_udp_wmem_min,
1464 	.sysctl_rmem	   = &sysctl_udp_rmem_min,
1465 	.obj_size	   = sizeof(struct udp6_sock),
1466 	.slab_flags	   = SLAB_DESTROY_BY_RCU,
1467 	.h.udp_table	   = &udp_table,
1468 #ifdef CONFIG_COMPAT
1469 	.compat_setsockopt = compat_udpv6_setsockopt,
1470 	.compat_getsockopt = compat_udpv6_getsockopt,
1471 #endif
1472 	.clear_sk	   = udp_v6_clear_sk,
1473 };
1474 
1475 static struct inet_protosw udpv6_protosw = {
1476 	.type =      SOCK_DGRAM,
1477 	.protocol =  IPPROTO_UDP,
1478 	.prot =      &udpv6_prot,
1479 	.ops =       &inet6_dgram_ops,
1480 	.no_check =  UDP_CSUM_DEFAULT,
1481 	.flags =     INET_PROTOSW_PERMANENT,
1482 };
1483 
1484 
1485 int __init udpv6_init(void)
1486 {
1487 	int ret;
1488 
1489 	ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1490 	if (ret)
1491 		goto out;
1492 
1493 	ret = inet6_register_protosw(&udpv6_protosw);
1494 	if (ret)
1495 		goto out_udpv6_protocol;
1496 out:
1497 	return ret;
1498 
1499 out_udpv6_protocol:
1500 	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1501 	goto out;
1502 }
1503 
1504 void udpv6_exit(void)
1505 {
1506 	inet6_unregister_protosw(&udpv6_protosw);
1507 	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1508 }
1509