xref: /openbmc/linux/net/ipv6/udp.c (revision c4ee0af3)
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 		ip6_sk_update_pmtu(skb, sk, info);
543 	if (type == NDISC_REDIRECT) {
544 		ip6_sk_redirect(skb, sk);
545 		goto out;
546 	}
547 
548 	np = inet6_sk(sk);
549 
550 	if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
551 		goto out;
552 
553 	if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
554 		goto out;
555 
556 	if (np->recverr)
557 		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
558 
559 	sk->sk_err = err;
560 	sk->sk_error_report(sk);
561 out:
562 	sock_put(sk);
563 }
564 
565 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
566 {
567 	int rc;
568 
569 	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
570 		sock_rps_save_rxhash(sk, skb);
571 		sk_mark_napi_id(sk, skb);
572 	}
573 
574 	rc = sock_queue_rcv_skb(sk, skb);
575 	if (rc < 0) {
576 		int is_udplite = IS_UDPLITE(sk);
577 
578 		/* Note that an ENOMEM error is charged twice */
579 		if (rc == -ENOMEM)
580 			UDP6_INC_STATS_BH(sock_net(sk),
581 					UDP_MIB_RCVBUFERRORS, is_udplite);
582 		UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
583 		kfree_skb(skb);
584 		return -1;
585 	}
586 	return 0;
587 }
588 
589 static __inline__ void udpv6_err(struct sk_buff *skb,
590 				 struct inet6_skb_parm *opt, u8 type,
591 				 u8 code, int offset, __be32 info     )
592 {
593 	__udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
594 }
595 
596 static struct static_key udpv6_encap_needed __read_mostly;
597 void udpv6_encap_enable(void)
598 {
599 	if (!static_key_enabled(&udpv6_encap_needed))
600 		static_key_slow_inc(&udpv6_encap_needed);
601 }
602 EXPORT_SYMBOL(udpv6_encap_enable);
603 
604 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
605 {
606 	struct udp_sock *up = udp_sk(sk);
607 	int rc;
608 	int is_udplite = IS_UDPLITE(sk);
609 
610 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
611 		goto drop;
612 
613 	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
614 		int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
615 
616 		/*
617 		 * This is an encapsulation socket so pass the skb to
618 		 * the socket's udp_encap_rcv() hook. Otherwise, just
619 		 * fall through and pass this up the UDP socket.
620 		 * up->encap_rcv() returns the following value:
621 		 * =0 if skb was successfully passed to the encap
622 		 *    handler or was discarded by it.
623 		 * >0 if skb should be passed on to UDP.
624 		 * <0 if skb should be resubmitted as proto -N
625 		 */
626 
627 		/* if we're overly short, let UDP handle it */
628 		encap_rcv = ACCESS_ONCE(up->encap_rcv);
629 		if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
630 			int ret;
631 
632 			ret = encap_rcv(sk, skb);
633 			if (ret <= 0) {
634 				UDP_INC_STATS_BH(sock_net(sk),
635 						 UDP_MIB_INDATAGRAMS,
636 						 is_udplite);
637 				return -ret;
638 			}
639 		}
640 
641 		/* FALLTHROUGH -- it's a UDP Packet */
642 	}
643 
644 	/*
645 	 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
646 	 */
647 	if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
648 
649 		if (up->pcrlen == 0) {          /* full coverage was set  */
650 			LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
651 				" %d while full coverage %d requested\n",
652 				UDP_SKB_CB(skb)->cscov, skb->len);
653 			goto drop;
654 		}
655 		if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
656 			LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
657 						    "too small, need min %d\n",
658 				       UDP_SKB_CB(skb)->cscov, up->pcrlen);
659 			goto drop;
660 		}
661 	}
662 
663 	if (rcu_access_pointer(sk->sk_filter)) {
664 		if (udp_lib_checksum_complete(skb))
665 			goto csum_error;
666 	}
667 
668 	if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
669 		goto drop;
670 
671 	skb_dst_drop(skb);
672 
673 	bh_lock_sock(sk);
674 	rc = 0;
675 	if (!sock_owned_by_user(sk))
676 		rc = __udpv6_queue_rcv_skb(sk, skb);
677 	else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
678 		bh_unlock_sock(sk);
679 		goto drop;
680 	}
681 	bh_unlock_sock(sk);
682 
683 	return rc;
684 csum_error:
685 	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
686 drop:
687 	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
688 	atomic_inc(&sk->sk_drops);
689 	kfree_skb(skb);
690 	return -1;
691 }
692 
693 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
694 				      __be16 loc_port, const struct in6_addr *loc_addr,
695 				      __be16 rmt_port, const struct in6_addr *rmt_addr,
696 				      int dif)
697 {
698 	struct hlist_nulls_node *node;
699 	struct sock *s = sk;
700 	unsigned short num = ntohs(loc_port);
701 
702 	sk_nulls_for_each_from(s, node) {
703 		struct inet_sock *inet = inet_sk(s);
704 
705 		if (!net_eq(sock_net(s), net))
706 			continue;
707 
708 		if (udp_sk(s)->udp_port_hash == num &&
709 		    s->sk_family == PF_INET6) {
710 			if (inet->inet_dport) {
711 				if (inet->inet_dport != rmt_port)
712 					continue;
713 			}
714 			if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
715 			    !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
716 				continue;
717 
718 			if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
719 				continue;
720 
721 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
722 				if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
723 					continue;
724 			}
725 			if (!inet6_mc_check(s, loc_addr, rmt_addr))
726 				continue;
727 			return s;
728 		}
729 	}
730 	return NULL;
731 }
732 
733 static void flush_stack(struct sock **stack, unsigned int count,
734 			struct sk_buff *skb, unsigned int final)
735 {
736 	struct sk_buff *skb1 = NULL;
737 	struct sock *sk;
738 	unsigned int i;
739 
740 	for (i = 0; i < count; i++) {
741 		sk = stack[i];
742 		if (likely(skb1 == NULL))
743 			skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
744 		if (!skb1) {
745 			atomic_inc(&sk->sk_drops);
746 			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
747 					  IS_UDPLITE(sk));
748 			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
749 					  IS_UDPLITE(sk));
750 		}
751 
752 		if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
753 			skb1 = NULL;
754 	}
755 	if (unlikely(skb1))
756 		kfree_skb(skb1);
757 }
758 /*
759  * Note: called only from the BH handler context,
760  * so we don't need to lock the hashes.
761  */
762 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
763 		const struct in6_addr *saddr, const struct in6_addr *daddr,
764 		struct udp_table *udptable)
765 {
766 	struct sock *sk, *stack[256 / sizeof(struct sock *)];
767 	const struct udphdr *uh = udp_hdr(skb);
768 	struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
769 	int dif;
770 	unsigned int i, count = 0;
771 
772 	spin_lock(&hslot->lock);
773 	sk = sk_nulls_head(&hslot->head);
774 	dif = inet6_iif(skb);
775 	sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
776 	while (sk) {
777 		stack[count++] = sk;
778 		sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
779 				       uh->source, saddr, dif);
780 		if (unlikely(count == ARRAY_SIZE(stack))) {
781 			if (!sk)
782 				break;
783 			flush_stack(stack, count, skb, ~0);
784 			count = 0;
785 		}
786 	}
787 	/*
788 	 * before releasing the lock, we must take reference on sockets
789 	 */
790 	for (i = 0; i < count; i++)
791 		sock_hold(stack[i]);
792 
793 	spin_unlock(&hslot->lock);
794 
795 	if (count) {
796 		flush_stack(stack, count, skb, count - 1);
797 
798 		for (i = 0; i < count; i++)
799 			sock_put(stack[i]);
800 	} else {
801 		kfree_skb(skb);
802 	}
803 	return 0;
804 }
805 
806 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
807 		   int proto)
808 {
809 	struct net *net = dev_net(skb->dev);
810 	struct sock *sk;
811 	struct udphdr *uh;
812 	const struct in6_addr *saddr, *daddr;
813 	u32 ulen = 0;
814 
815 	if (!pskb_may_pull(skb, sizeof(struct udphdr)))
816 		goto discard;
817 
818 	saddr = &ipv6_hdr(skb)->saddr;
819 	daddr = &ipv6_hdr(skb)->daddr;
820 	uh = udp_hdr(skb);
821 
822 	ulen = ntohs(uh->len);
823 	if (ulen > skb->len)
824 		goto short_packet;
825 
826 	if (proto == IPPROTO_UDP) {
827 		/* UDP validates ulen. */
828 
829 		/* Check for jumbo payload */
830 		if (ulen == 0)
831 			ulen = skb->len;
832 
833 		if (ulen < sizeof(*uh))
834 			goto short_packet;
835 
836 		if (ulen < skb->len) {
837 			if (pskb_trim_rcsum(skb, ulen))
838 				goto short_packet;
839 			saddr = &ipv6_hdr(skb)->saddr;
840 			daddr = &ipv6_hdr(skb)->daddr;
841 			uh = udp_hdr(skb);
842 		}
843 	}
844 
845 	if (udp6_csum_init(skb, uh, proto))
846 		goto csum_error;
847 
848 	/*
849 	 *	Multicast receive code
850 	 */
851 	if (ipv6_addr_is_multicast(daddr))
852 		return __udp6_lib_mcast_deliver(net, skb,
853 				saddr, daddr, udptable);
854 
855 	/* Unicast */
856 
857 	/*
858 	 * check socket cache ... must talk to Alan about his plans
859 	 * for sock caches... i'll skip this for now.
860 	 */
861 	sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
862 	if (sk != NULL) {
863 		int ret;
864 
865 		ret = udpv6_queue_rcv_skb(sk, skb);
866 		sock_put(sk);
867 
868 		/* a return value > 0 means to resubmit the input, but
869 		 * it wants the return to be -protocol, or 0
870 		 */
871 		if (ret > 0)
872 			return -ret;
873 
874 		return 0;
875 	}
876 
877 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
878 		goto discard;
879 
880 	if (udp_lib_checksum_complete(skb))
881 		goto csum_error;
882 
883 	UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
884 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
885 
886 	kfree_skb(skb);
887 	return 0;
888 
889 short_packet:
890 	LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
891 		       proto == IPPROTO_UDPLITE ? "-Lite" : "",
892 		       saddr,
893 		       ntohs(uh->source),
894 		       ulen,
895 		       skb->len,
896 		       daddr,
897 		       ntohs(uh->dest));
898 	goto discard;
899 csum_error:
900 	UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
901 discard:
902 	UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
903 	kfree_skb(skb);
904 	return 0;
905 }
906 
907 static __inline__ int udpv6_rcv(struct sk_buff *skb)
908 {
909 	return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
910 }
911 
912 /*
913  * Throw away all pending data and cancel the corking. Socket is locked.
914  */
915 static void udp_v6_flush_pending_frames(struct sock *sk)
916 {
917 	struct udp_sock *up = udp_sk(sk);
918 
919 	if (up->pending == AF_INET)
920 		udp_flush_pending_frames(sk);
921 	else if (up->pending) {
922 		up->len = 0;
923 		up->pending = 0;
924 		ip6_flush_pending_frames(sk);
925 	}
926 }
927 
928 /**
929  * 	udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
930  * 	@sk: 	socket we are sending on
931  * 	@skb: 	sk_buff containing the filled-in UDP header
932  * 	        (checksum field must be zeroed out)
933  */
934 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
935 				 const struct in6_addr *saddr,
936 				 const struct in6_addr *daddr, int len)
937 {
938 	unsigned int offset;
939 	struct udphdr *uh = udp_hdr(skb);
940 	__wsum csum = 0;
941 
942 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
943 		/* Only one fragment on the socket.  */
944 		skb->csum_start = skb_transport_header(skb) - skb->head;
945 		skb->csum_offset = offsetof(struct udphdr, check);
946 		uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
947 	} else {
948 		/*
949 		 * HW-checksum won't work as there are two or more
950 		 * fragments on the socket so that all csums of sk_buffs
951 		 * should be together
952 		 */
953 		offset = skb_transport_offset(skb);
954 		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
955 
956 		skb->ip_summed = CHECKSUM_NONE;
957 
958 		skb_queue_walk(&sk->sk_write_queue, skb) {
959 			csum = csum_add(csum, skb->csum);
960 		}
961 
962 		uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
963 					    csum);
964 		if (uh->check == 0)
965 			uh->check = CSUM_MANGLED_0;
966 	}
967 }
968 
969 /*
970  *	Sending
971  */
972 
973 static int udp_v6_push_pending_frames(struct sock *sk)
974 {
975 	struct sk_buff *skb;
976 	struct udphdr *uh;
977 	struct udp_sock  *up = udp_sk(sk);
978 	struct inet_sock *inet = inet_sk(sk);
979 	struct flowi6 *fl6;
980 	int err = 0;
981 	int is_udplite = IS_UDPLITE(sk);
982 	__wsum csum = 0;
983 
984 	if (up->pending == AF_INET)
985 		return udp_push_pending_frames(sk);
986 
987 	fl6 = &inet->cork.fl.u.ip6;
988 
989 	/* Grab the skbuff where UDP header space exists. */
990 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
991 		goto out;
992 
993 	/*
994 	 * Create a UDP header
995 	 */
996 	uh = udp_hdr(skb);
997 	uh->source = fl6->fl6_sport;
998 	uh->dest = fl6->fl6_dport;
999 	uh->len = htons(up->len);
1000 	uh->check = 0;
1001 
1002 	if (is_udplite)
1003 		csum = udplite_csum_outgoing(sk, skb);
1004 	else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1005 		udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
1006 				     up->len);
1007 		goto send;
1008 	} else
1009 		csum = udp_csum_outgoing(sk, skb);
1010 
1011 	/* add protocol-dependent pseudo-header */
1012 	uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1013 				    up->len, fl6->flowi6_proto, csum);
1014 	if (uh->check == 0)
1015 		uh->check = CSUM_MANGLED_0;
1016 
1017 send:
1018 	err = ip6_push_pending_frames(sk);
1019 	if (err) {
1020 		if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1021 			UDP6_INC_STATS_USER(sock_net(sk),
1022 					    UDP_MIB_SNDBUFERRORS, is_udplite);
1023 			err = 0;
1024 		}
1025 	} else
1026 		UDP6_INC_STATS_USER(sock_net(sk),
1027 				    UDP_MIB_OUTDATAGRAMS, is_udplite);
1028 out:
1029 	up->len = 0;
1030 	up->pending = 0;
1031 	return err;
1032 }
1033 
1034 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1035 		  struct msghdr *msg, size_t len)
1036 {
1037 	struct ipv6_txoptions opt_space;
1038 	struct udp_sock *up = udp_sk(sk);
1039 	struct inet_sock *inet = inet_sk(sk);
1040 	struct ipv6_pinfo *np = inet6_sk(sk);
1041 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
1042 	struct in6_addr *daddr, *final_p, final;
1043 	struct ipv6_txoptions *opt = NULL;
1044 	struct ip6_flowlabel *flowlabel = NULL;
1045 	struct flowi6 fl6;
1046 	struct dst_entry *dst;
1047 	int addr_len = msg->msg_namelen;
1048 	int ulen = len;
1049 	int hlimit = -1;
1050 	int tclass = -1;
1051 	int dontfrag = -1;
1052 	int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1053 	int err;
1054 	int connected = 0;
1055 	int is_udplite = IS_UDPLITE(sk);
1056 	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1057 
1058 	/* destination address check */
1059 	if (sin6) {
1060 		if (addr_len < offsetof(struct sockaddr, sa_data))
1061 			return -EINVAL;
1062 
1063 		switch (sin6->sin6_family) {
1064 		case AF_INET6:
1065 			if (addr_len < SIN6_LEN_RFC2133)
1066 				return -EINVAL;
1067 			daddr = &sin6->sin6_addr;
1068 			break;
1069 		case AF_INET:
1070 			goto do_udp_sendmsg;
1071 		case AF_UNSPEC:
1072 			msg->msg_name = sin6 = NULL;
1073 			msg->msg_namelen = addr_len = 0;
1074 			daddr = NULL;
1075 			break;
1076 		default:
1077 			return -EINVAL;
1078 		}
1079 	} else if (!up->pending) {
1080 		if (sk->sk_state != TCP_ESTABLISHED)
1081 			return -EDESTADDRREQ;
1082 		daddr = &sk->sk_v6_daddr;
1083 	} else
1084 		daddr = NULL;
1085 
1086 	if (daddr) {
1087 		if (ipv6_addr_v4mapped(daddr)) {
1088 			struct sockaddr_in sin;
1089 			sin.sin_family = AF_INET;
1090 			sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1091 			sin.sin_addr.s_addr = daddr->s6_addr32[3];
1092 			msg->msg_name = &sin;
1093 			msg->msg_namelen = sizeof(sin);
1094 do_udp_sendmsg:
1095 			if (__ipv6_only_sock(sk))
1096 				return -ENETUNREACH;
1097 			return udp_sendmsg(iocb, sk, msg, len);
1098 		}
1099 	}
1100 
1101 	if (up->pending == AF_INET)
1102 		return udp_sendmsg(iocb, sk, msg, len);
1103 
1104 	/* Rough check on arithmetic overflow,
1105 	   better check is made in ip6_append_data().
1106 	   */
1107 	if (len > INT_MAX - sizeof(struct udphdr))
1108 		return -EMSGSIZE;
1109 
1110 	if (up->pending) {
1111 		/*
1112 		 * There are pending frames.
1113 		 * The socket lock must be held while it's corked.
1114 		 */
1115 		lock_sock(sk);
1116 		if (likely(up->pending)) {
1117 			if (unlikely(up->pending != AF_INET6)) {
1118 				release_sock(sk);
1119 				return -EAFNOSUPPORT;
1120 			}
1121 			dst = NULL;
1122 			goto do_append_data;
1123 		}
1124 		release_sock(sk);
1125 	}
1126 	ulen += sizeof(struct udphdr);
1127 
1128 	memset(&fl6, 0, sizeof(fl6));
1129 
1130 	if (sin6) {
1131 		if (sin6->sin6_port == 0)
1132 			return -EINVAL;
1133 
1134 		fl6.fl6_dport = sin6->sin6_port;
1135 		daddr = &sin6->sin6_addr;
1136 
1137 		if (np->sndflow) {
1138 			fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1139 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1140 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1141 				if (flowlabel == NULL)
1142 					return -EINVAL;
1143 			}
1144 		}
1145 
1146 		/*
1147 		 * Otherwise it will be difficult to maintain
1148 		 * sk->sk_dst_cache.
1149 		 */
1150 		if (sk->sk_state == TCP_ESTABLISHED &&
1151 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1152 			daddr = &sk->sk_v6_daddr;
1153 
1154 		if (addr_len >= sizeof(struct sockaddr_in6) &&
1155 		    sin6->sin6_scope_id &&
1156 		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1157 			fl6.flowi6_oif = sin6->sin6_scope_id;
1158 	} else {
1159 		if (sk->sk_state != TCP_ESTABLISHED)
1160 			return -EDESTADDRREQ;
1161 
1162 		fl6.fl6_dport = inet->inet_dport;
1163 		daddr = &sk->sk_v6_daddr;
1164 		fl6.flowlabel = np->flow_label;
1165 		connected = 1;
1166 	}
1167 
1168 	if (!fl6.flowi6_oif)
1169 		fl6.flowi6_oif = sk->sk_bound_dev_if;
1170 
1171 	if (!fl6.flowi6_oif)
1172 		fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1173 
1174 	fl6.flowi6_mark = sk->sk_mark;
1175 
1176 	if (msg->msg_controllen) {
1177 		opt = &opt_space;
1178 		memset(opt, 0, sizeof(struct ipv6_txoptions));
1179 		opt->tot_len = sizeof(*opt);
1180 
1181 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1182 					    &hlimit, &tclass, &dontfrag);
1183 		if (err < 0) {
1184 			fl6_sock_release(flowlabel);
1185 			return err;
1186 		}
1187 		if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1188 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1189 			if (flowlabel == NULL)
1190 				return -EINVAL;
1191 		}
1192 		if (!(opt->opt_nflen|opt->opt_flen))
1193 			opt = NULL;
1194 		connected = 0;
1195 	}
1196 	if (opt == NULL)
1197 		opt = np->opt;
1198 	if (flowlabel)
1199 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
1200 	opt = ipv6_fixup_options(&opt_space, opt);
1201 
1202 	fl6.flowi6_proto = sk->sk_protocol;
1203 	if (!ipv6_addr_any(daddr))
1204 		fl6.daddr = *daddr;
1205 	else
1206 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1207 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1208 		fl6.saddr = np->saddr;
1209 	fl6.fl6_sport = inet->inet_sport;
1210 
1211 	final_p = fl6_update_dst(&fl6, opt, &final);
1212 	if (final_p)
1213 		connected = 0;
1214 
1215 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1216 		fl6.flowi6_oif = np->mcast_oif;
1217 		connected = 0;
1218 	} else if (!fl6.flowi6_oif)
1219 		fl6.flowi6_oif = np->ucast_oif;
1220 
1221 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1222 
1223 	dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
1224 	if (IS_ERR(dst)) {
1225 		err = PTR_ERR(dst);
1226 		dst = NULL;
1227 		goto out;
1228 	}
1229 
1230 	if (hlimit < 0) {
1231 		if (ipv6_addr_is_multicast(&fl6.daddr))
1232 			hlimit = np->mcast_hops;
1233 		else
1234 			hlimit = np->hop_limit;
1235 		if (hlimit < 0)
1236 			hlimit = ip6_dst_hoplimit(dst);
1237 	}
1238 
1239 	if (tclass < 0)
1240 		tclass = np->tclass;
1241 
1242 	if (msg->msg_flags&MSG_CONFIRM)
1243 		goto do_confirm;
1244 back_from_confirm:
1245 
1246 	lock_sock(sk);
1247 	if (unlikely(up->pending)) {
1248 		/* The socket is already corked while preparing it. */
1249 		/* ... which is an evident application bug. --ANK */
1250 		release_sock(sk);
1251 
1252 		LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1253 		err = -EINVAL;
1254 		goto out;
1255 	}
1256 
1257 	up->pending = AF_INET6;
1258 
1259 do_append_data:
1260 	if (dontfrag < 0)
1261 		dontfrag = np->dontfrag;
1262 	up->len += ulen;
1263 	getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1264 	err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
1265 		sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1266 		(struct rt6_info*)dst,
1267 		corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1268 	if (err)
1269 		udp_v6_flush_pending_frames(sk);
1270 	else if (!corkreq)
1271 		err = udp_v6_push_pending_frames(sk);
1272 	else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1273 		up->pending = 0;
1274 
1275 	if (dst) {
1276 		if (connected) {
1277 			ip6_dst_store(sk, dst,
1278 				      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1279 				      &sk->sk_v6_daddr : NULL,
1280 #ifdef CONFIG_IPV6_SUBTREES
1281 				      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1282 				      &np->saddr :
1283 #endif
1284 				      NULL);
1285 		} else {
1286 			dst_release(dst);
1287 		}
1288 		dst = NULL;
1289 	}
1290 
1291 	if (err > 0)
1292 		err = np->recverr ? net_xmit_errno(err) : 0;
1293 	release_sock(sk);
1294 out:
1295 	dst_release(dst);
1296 	fl6_sock_release(flowlabel);
1297 	if (!err)
1298 		return len;
1299 	/*
1300 	 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1301 	 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1302 	 * we don't have a good statistic (IpOutDiscards but it can be too many
1303 	 * things).  We could add another new stat but at least for now that
1304 	 * seems like overkill.
1305 	 */
1306 	if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1307 		UDP6_INC_STATS_USER(sock_net(sk),
1308 				UDP_MIB_SNDBUFERRORS, is_udplite);
1309 	}
1310 	return err;
1311 
1312 do_confirm:
1313 	dst_confirm(dst);
1314 	if (!(msg->msg_flags&MSG_PROBE) || len)
1315 		goto back_from_confirm;
1316 	err = 0;
1317 	goto out;
1318 }
1319 
1320 void udpv6_destroy_sock(struct sock *sk)
1321 {
1322 	struct udp_sock *up = udp_sk(sk);
1323 	lock_sock(sk);
1324 	udp_v6_flush_pending_frames(sk);
1325 	release_sock(sk);
1326 
1327 	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1328 		void (*encap_destroy)(struct sock *sk);
1329 		encap_destroy = ACCESS_ONCE(up->encap_destroy);
1330 		if (encap_destroy)
1331 			encap_destroy(sk);
1332 	}
1333 
1334 	inet6_destroy_sock(sk);
1335 }
1336 
1337 /*
1338  *	Socket option code for UDP
1339  */
1340 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1341 		     char __user *optval, unsigned int optlen)
1342 {
1343 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1344 		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1345 					  udp_v6_push_pending_frames);
1346 	return ipv6_setsockopt(sk, level, optname, optval, optlen);
1347 }
1348 
1349 #ifdef CONFIG_COMPAT
1350 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1351 			    char __user *optval, unsigned int optlen)
1352 {
1353 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1354 		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1355 					  udp_v6_push_pending_frames);
1356 	return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1357 }
1358 #endif
1359 
1360 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1361 		     char __user *optval, int __user *optlen)
1362 {
1363 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1364 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1365 	return ipv6_getsockopt(sk, level, optname, optval, optlen);
1366 }
1367 
1368 #ifdef CONFIG_COMPAT
1369 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1370 			    char __user *optval, int __user *optlen)
1371 {
1372 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1373 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1374 	return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1375 }
1376 #endif
1377 
1378 static const struct inet6_protocol udpv6_protocol = {
1379 	.handler	=	udpv6_rcv,
1380 	.err_handler	=	udpv6_err,
1381 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1382 };
1383 
1384 /* ------------------------------------------------------------------------ */
1385 #ifdef CONFIG_PROC_FS
1386 int udp6_seq_show(struct seq_file *seq, void *v)
1387 {
1388 	if (v == SEQ_START_TOKEN) {
1389 		seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1390 	} else {
1391 		int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1392 		struct inet_sock *inet = inet_sk(v);
1393 		__u16 srcp = ntohs(inet->inet_sport);
1394 		__u16 destp = ntohs(inet->inet_dport);
1395 		ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1396 	}
1397 	return 0;
1398 }
1399 
1400 static const struct file_operations udp6_afinfo_seq_fops = {
1401 	.owner    = THIS_MODULE,
1402 	.open     = udp_seq_open,
1403 	.read     = seq_read,
1404 	.llseek   = seq_lseek,
1405 	.release  = seq_release_net
1406 };
1407 
1408 static struct udp_seq_afinfo udp6_seq_afinfo = {
1409 	.name		= "udp6",
1410 	.family		= AF_INET6,
1411 	.udp_table	= &udp_table,
1412 	.seq_fops	= &udp6_afinfo_seq_fops,
1413 	.seq_ops	= {
1414 		.show		= udp6_seq_show,
1415 	},
1416 };
1417 
1418 int __net_init udp6_proc_init(struct net *net)
1419 {
1420 	return udp_proc_register(net, &udp6_seq_afinfo);
1421 }
1422 
1423 void udp6_proc_exit(struct net *net) {
1424 	udp_proc_unregister(net, &udp6_seq_afinfo);
1425 }
1426 #endif /* CONFIG_PROC_FS */
1427 
1428 void udp_v6_clear_sk(struct sock *sk, int size)
1429 {
1430 	struct inet_sock *inet = inet_sk(sk);
1431 
1432 	/* we do not want to clear pinet6 field, because of RCU lookups */
1433 	sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1434 
1435 	size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1436 	memset(&inet->pinet6 + 1, 0, size);
1437 }
1438 
1439 /* ------------------------------------------------------------------------ */
1440 
1441 struct proto udpv6_prot = {
1442 	.name		   = "UDPv6",
1443 	.owner		   = THIS_MODULE,
1444 	.close		   = udp_lib_close,
1445 	.connect	   = ip6_datagram_connect,
1446 	.disconnect	   = udp_disconnect,
1447 	.ioctl		   = udp_ioctl,
1448 	.destroy	   = udpv6_destroy_sock,
1449 	.setsockopt	   = udpv6_setsockopt,
1450 	.getsockopt	   = udpv6_getsockopt,
1451 	.sendmsg	   = udpv6_sendmsg,
1452 	.recvmsg	   = udpv6_recvmsg,
1453 	.backlog_rcv	   = __udpv6_queue_rcv_skb,
1454 	.hash		   = udp_lib_hash,
1455 	.unhash		   = udp_lib_unhash,
1456 	.rehash		   = udp_v6_rehash,
1457 	.get_port	   = udp_v6_get_port,
1458 	.memory_allocated  = &udp_memory_allocated,
1459 	.sysctl_mem	   = sysctl_udp_mem,
1460 	.sysctl_wmem	   = &sysctl_udp_wmem_min,
1461 	.sysctl_rmem	   = &sysctl_udp_rmem_min,
1462 	.obj_size	   = sizeof(struct udp6_sock),
1463 	.slab_flags	   = SLAB_DESTROY_BY_RCU,
1464 	.h.udp_table	   = &udp_table,
1465 #ifdef CONFIG_COMPAT
1466 	.compat_setsockopt = compat_udpv6_setsockopt,
1467 	.compat_getsockopt = compat_udpv6_getsockopt,
1468 #endif
1469 	.clear_sk	   = udp_v6_clear_sk,
1470 };
1471 
1472 static struct inet_protosw udpv6_protosw = {
1473 	.type =      SOCK_DGRAM,
1474 	.protocol =  IPPROTO_UDP,
1475 	.prot =      &udpv6_prot,
1476 	.ops =       &inet6_dgram_ops,
1477 	.no_check =  UDP_CSUM_DEFAULT,
1478 	.flags =     INET_PROTOSW_PERMANENT,
1479 };
1480 
1481 
1482 int __init udpv6_init(void)
1483 {
1484 	int ret;
1485 
1486 	ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1487 	if (ret)
1488 		goto out;
1489 
1490 	ret = inet6_register_protosw(&udpv6_protosw);
1491 	if (ret)
1492 		goto out_udpv6_protocol;
1493 out:
1494 	return ret;
1495 
1496 out_udpv6_protocol:
1497 	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1498 	goto out;
1499 }
1500 
1501 void udpv6_exit(void)
1502 {
1503 	inet6_unregister_protosw(&udpv6_protosw);
1504 	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1505 }
1506