• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 
50 #include <linux/proc_fs.h>
51 #include <linux/seq_file.h>
52 #include <trace/events/skb.h>
53 #include "udp_impl.h"
54 
ipv6_rcv_saddr_equal(const struct sock * sk,const struct sock * sk2)55 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
56 {
57 	const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
58 	const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
59 	__be32 sk1_rcv_saddr = sk_rcv_saddr(sk);
60 	__be32 sk2_rcv_saddr = sk_rcv_saddr(sk2);
61 	int sk_ipv6only = ipv6_only_sock(sk);
62 	int sk2_ipv6only = inet_v6_ipv6only(sk2);
63 	int addr_type = ipv6_addr_type(sk_rcv_saddr6);
64 	int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
65 
66 	/* if both are mapped, treat as IPv4 */
67 	if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
68 		return (!sk2_ipv6only &&
69 			(!sk1_rcv_saddr || !sk2_rcv_saddr ||
70 			  sk1_rcv_saddr == sk2_rcv_saddr));
71 
72 	if (addr_type2 == IPV6_ADDR_ANY &&
73 	    !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
74 		return 1;
75 
76 	if (addr_type == IPV6_ADDR_ANY &&
77 	    !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
78 		return 1;
79 
80 	if (sk2_rcv_saddr6 &&
81 	    ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
82 		return 1;
83 
84 	return 0;
85 }
86 
udp6_portaddr_hash(struct net * net,const struct in6_addr * addr6,unsigned int port)87 static unsigned int udp6_portaddr_hash(struct net *net,
88 				       const struct in6_addr *addr6,
89 				       unsigned int port)
90 {
91 	unsigned int hash, mix = net_hash_mix(net);
92 
93 	if (ipv6_addr_any(addr6))
94 		hash = jhash_1word(0, mix);
95 	else if (ipv6_addr_v4mapped(addr6))
96 		hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
97 	else
98 		hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
99 
100 	return hash ^ port;
101 }
102 
103 
udp_v6_get_port(struct sock * sk,unsigned short snum)104 int udp_v6_get_port(struct sock *sk, unsigned short snum)
105 {
106 	unsigned int hash2_nulladdr =
107 		udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
108 	unsigned int hash2_partial =
109 		udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0);
110 
111 	/* precompute partial secondary hash */
112 	udp_sk(sk)->udp_portaddr_hash = hash2_partial;
113 	return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
114 }
115 
udp_v6_rehash(struct sock * sk)116 static void udp_v6_rehash(struct sock *sk)
117 {
118 	u16 new_hash = udp6_portaddr_hash(sock_net(sk),
119 					  &inet6_sk(sk)->rcv_saddr,
120 					  inet_sk(sk)->inet_num);
121 
122 	udp_lib_rehash(sk, new_hash);
123 }
124 
compute_score(struct sock * sk,struct net * net,unsigned short hnum,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)125 static inline int compute_score(struct sock *sk, struct net *net,
126 				unsigned short hnum,
127 				const struct in6_addr *saddr, __be16 sport,
128 				const struct in6_addr *daddr, __be16 dport,
129 				int dif)
130 {
131 	int score = -1;
132 
133 	if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
134 			sk->sk_family == PF_INET6) {
135 		struct ipv6_pinfo *np = inet6_sk(sk);
136 		struct inet_sock *inet = inet_sk(sk);
137 
138 		score = 0;
139 		if (inet->inet_dport) {
140 			if (inet->inet_dport != sport)
141 				return -1;
142 			score++;
143 		}
144 		if (!ipv6_addr_any(&np->rcv_saddr)) {
145 			if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
146 				return -1;
147 			score++;
148 		}
149 		if (!ipv6_addr_any(&np->daddr)) {
150 			if (!ipv6_addr_equal(&np->daddr, saddr))
151 				return -1;
152 			score++;
153 		}
154 		if (sk->sk_bound_dev_if) {
155 			if (sk->sk_bound_dev_if != dif)
156 				return -1;
157 			score++;
158 		}
159 	}
160 	return score;
161 }
162 
163 #define SCORE2_MAX (1 + 1 + 1)
compute_score2(struct sock * sk,struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned short hnum,int dif)164 static inline int compute_score2(struct sock *sk, struct net *net,
165 				const struct in6_addr *saddr, __be16 sport,
166 				const struct in6_addr *daddr, unsigned short hnum,
167 				int dif)
168 {
169 	int score = -1;
170 
171 	if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
172 			sk->sk_family == PF_INET6) {
173 		struct ipv6_pinfo *np = inet6_sk(sk);
174 		struct inet_sock *inet = inet_sk(sk);
175 
176 		if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
177 			return -1;
178 		score = 0;
179 		if (inet->inet_dport) {
180 			if (inet->inet_dport != sport)
181 				return -1;
182 			score++;
183 		}
184 		if (!ipv6_addr_any(&np->daddr)) {
185 			if (!ipv6_addr_equal(&np->daddr, saddr))
186 				return -1;
187 			score++;
188 		}
189 		if (sk->sk_bound_dev_if) {
190 			if (sk->sk_bound_dev_if != dif)
191 				return -1;
192 			score++;
193 		}
194 	}
195 	return score;
196 }
197 
198 
199 /* called with read_rcu_lock() */
udp6_lib_lookup2(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,struct udp_hslot * hslot2,unsigned int slot2)200 static struct sock *udp6_lib_lookup2(struct net *net,
201 		const struct in6_addr *saddr, __be16 sport,
202 		const struct in6_addr *daddr, unsigned int hnum, int dif,
203 		struct udp_hslot *hslot2, unsigned int slot2)
204 {
205 	struct sock *sk, *result;
206 	struct hlist_nulls_node *node;
207 	int score, badness, matches = 0, reuseport = 0;
208 	u32 hash = 0;
209 
210 begin:
211 	result = NULL;
212 	badness = -1;
213 	udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
214 		score = compute_score2(sk, net, saddr, sport,
215 				      daddr, hnum, dif);
216 		if (score > badness) {
217 			result = sk;
218 			badness = score;
219 			reuseport = sk->sk_reuseport;
220 			if (reuseport) {
221 				hash = inet6_ehashfn(net, daddr, hnum,
222 						     saddr, sport);
223 				matches = 1;
224 			} else if (score == SCORE2_MAX)
225 				goto exact_match;
226 		} else if (score == badness && reuseport) {
227 			matches++;
228 			if (((u64)hash * matches) >> 32 == 0)
229 				result = sk;
230 			hash = next_pseudo_random32(hash);
231 		}
232 	}
233 	/*
234 	 * if the nulls value we got at the end of this lookup is
235 	 * not the expected one, we must restart lookup.
236 	 * We probably met an item that was moved to another chain.
237 	 */
238 	if (get_nulls_value(node) != slot2)
239 		goto begin;
240 
241 	if (result) {
242 exact_match:
243 		if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
244 			result = NULL;
245 		else if (unlikely(compute_score2(result, net, saddr, sport,
246 				  daddr, hnum, dif) < badness)) {
247 			sock_put(result);
248 			goto begin;
249 		}
250 	}
251 	return result;
252 }
253 
__udp6_lib_lookup(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif,struct udp_table * udptable)254 struct sock *__udp6_lib_lookup(struct net *net,
255 				      const struct in6_addr *saddr, __be16 sport,
256 				      const struct in6_addr *daddr, __be16 dport,
257 				      int dif, struct udp_table *udptable)
258 {
259 	struct sock *sk, *result;
260 	struct hlist_nulls_node *node;
261 	unsigned short hnum = ntohs(dport);
262 	unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
263 	struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
264 	int score, badness, matches = 0, reuseport = 0;
265 	u32 hash = 0;
266 
267 	rcu_read_lock();
268 	if (hslot->count > 10) {
269 		hash2 = udp6_portaddr_hash(net, daddr, hnum);
270 		slot2 = hash2 & udptable->mask;
271 		hslot2 = &udptable->hash2[slot2];
272 		if (hslot->count < hslot2->count)
273 			goto begin;
274 
275 		result = udp6_lib_lookup2(net, saddr, sport,
276 					  daddr, hnum, dif,
277 					  hslot2, slot2);
278 		if (!result) {
279 			hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
280 			slot2 = hash2 & udptable->mask;
281 			hslot2 = &udptable->hash2[slot2];
282 			if (hslot->count < hslot2->count)
283 				goto begin;
284 
285 			result = udp6_lib_lookup2(net, saddr, sport,
286 						  &in6addr_any, hnum, dif,
287 						  hslot2, slot2);
288 		}
289 		rcu_read_unlock();
290 		return result;
291 	}
292 begin:
293 	result = NULL;
294 	badness = -1;
295 	sk_nulls_for_each_rcu(sk, node, &hslot->head) {
296 		score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
297 		if (score > badness) {
298 			result = sk;
299 			badness = score;
300 			reuseport = sk->sk_reuseport;
301 			if (reuseport) {
302 				hash = inet6_ehashfn(net, daddr, hnum,
303 						     saddr, sport);
304 				matches = 1;
305 			}
306 		} else if (score == badness && reuseport) {
307 			matches++;
308 			if (((u64)hash * matches) >> 32 == 0)
309 				result = sk;
310 			hash = next_pseudo_random32(hash);
311 		}
312 	}
313 	/*
314 	 * if the nulls value we got at the end of this lookup is
315 	 * not the expected one, we must restart lookup.
316 	 * We probably met an item that was moved to another chain.
317 	 */
318 	if (get_nulls_value(node) != slot)
319 		goto begin;
320 
321 	if (result) {
322 		if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
323 			result = NULL;
324 		else if (unlikely(compute_score(result, net, hnum, saddr, sport,
325 					daddr, dport, dif) < badness)) {
326 			sock_put(result);
327 			goto begin;
328 		}
329 	}
330 	rcu_read_unlock();
331 	return result;
332 }
333 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
334 
__udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport,struct udp_table * udptable)335 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
336 					  __be16 sport, __be16 dport,
337 					  struct udp_table *udptable)
338 {
339 	struct sock *sk;
340 	const struct ipv6hdr *iph = ipv6_hdr(skb);
341 
342 	if (unlikely(sk = skb_steal_sock(skb)))
343 		return sk;
344 	return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
345 				 &iph->daddr, dport, inet6_iif(skb),
346 				 udptable);
347 }
348 
udp6_lib_lookup(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)349 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
350 			     const struct in6_addr *daddr, __be16 dport, int dif)
351 {
352 	return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
353 }
354 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
355 
356 
357 /*
358  * 	This should be easy, if there is something there we
359  * 	return it, otherwise we block.
360  */
361 
udpv6_recvmsg(struct kiocb * iocb,struct sock * sk,struct msghdr * msg,size_t len,int noblock,int flags,int * addr_len)362 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
363 		  struct msghdr *msg, size_t len,
364 		  int noblock, int flags, int *addr_len)
365 {
366 	struct ipv6_pinfo *np = inet6_sk(sk);
367 	struct inet_sock *inet = inet_sk(sk);
368 	struct sk_buff *skb;
369 	unsigned int ulen, copied;
370 	int peeked, off = 0;
371 	int err;
372 	int is_udplite = IS_UDPLITE(sk);
373 	bool checksum_valid = false;
374 	int is_udp4;
375 	bool slow;
376 
377 	if (flags & MSG_ERRQUEUE)
378 		return ipv6_recv_error(sk, msg, len, addr_len);
379 
380 	if (np->rxpmtu && np->rxopt.bits.rxpmtu)
381 		return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
382 
383 try_again:
384 	skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
385 				  &peeked, &off, &err);
386 	if (!skb)
387 		goto out;
388 
389 	ulen = skb->len - sizeof(struct udphdr);
390 	copied = len;
391 	if (copied > ulen)
392 		copied = ulen;
393 	else if (copied < ulen)
394 		msg->msg_flags |= MSG_TRUNC;
395 
396 	is_udp4 = (skb->protocol == htons(ETH_P_IP));
397 
398 	/*
399 	 * If checksum is needed at all, try to do it while copying the
400 	 * data.  If the data is truncated, or if we only want a partial
401 	 * coverage checksum (UDP-Lite), do it before the copy.
402 	 */
403 
404 	if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
405 		checksum_valid = !udp_lib_checksum_complete(skb);
406 		if (!checksum_valid)
407 			goto csum_copy_err;
408 	}
409 
410 	if (checksum_valid || skb_csum_unnecessary(skb))
411 		err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
412 					      msg->msg_iov, copied);
413 	else {
414 		err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
415 		if (err == -EINVAL)
416 			goto csum_copy_err;
417 	}
418 	if (unlikely(err)) {
419 		trace_kfree_skb(skb, udpv6_recvmsg);
420 		if (!peeked) {
421 			atomic_inc(&sk->sk_drops);
422 			if (is_udp4)
423 				UDP_INC_STATS_USER(sock_net(sk),
424 						   UDP_MIB_INERRORS,
425 						   is_udplite);
426 			else
427 				UDP6_INC_STATS_USER(sock_net(sk),
428 						    UDP_MIB_INERRORS,
429 						    is_udplite);
430 		}
431 		goto out_free;
432 	}
433 	if (!peeked) {
434 		if (is_udp4)
435 			UDP_INC_STATS_USER(sock_net(sk),
436 					UDP_MIB_INDATAGRAMS, is_udplite);
437 		else
438 			UDP6_INC_STATS_USER(sock_net(sk),
439 					UDP_MIB_INDATAGRAMS, is_udplite);
440 	}
441 
442 	sock_recv_ts_and_drops(msg, sk, skb);
443 
444 	/* Copy the address. */
445 	if (msg->msg_name) {
446 		struct sockaddr_in6 *sin6;
447 
448 		sin6 = (struct sockaddr_in6 *) msg->msg_name;
449 		sin6->sin6_family = AF_INET6;
450 		sin6->sin6_port = udp_hdr(skb)->source;
451 		sin6->sin6_flowinfo = 0;
452 
453 		if (is_udp4) {
454 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
455 					       &sin6->sin6_addr);
456 			sin6->sin6_scope_id = 0;
457 		} else {
458 			sin6->sin6_addr = ipv6_hdr(skb)->saddr;
459 			sin6->sin6_scope_id =
460 				ipv6_iface_scope_id(&sin6->sin6_addr,
461 						    IP6CB(skb)->iif);
462 		}
463 		*addr_len = sizeof(*sin6);
464 	}
465 	if (is_udp4) {
466 		if (inet->cmsg_flags)
467 			ip_cmsg_recv(msg, skb);
468 	} else {
469 		if (np->rxopt.all)
470 			ip6_datagram_recv_ctl(sk, msg, skb);
471 	}
472 
473 	err = copied;
474 	if (flags & MSG_TRUNC)
475 		err = ulen;
476 
477 out_free:
478 	skb_free_datagram_locked(sk, skb);
479 out:
480 	return err;
481 
482 csum_copy_err:
483 	slow = lock_sock_fast(sk);
484 	if (!skb_kill_datagram(sk, skb, flags)) {
485 		if (is_udp4) {
486 			UDP_INC_STATS_USER(sock_net(sk),
487 					UDP_MIB_CSUMERRORS, is_udplite);
488 			UDP_INC_STATS_USER(sock_net(sk),
489 					UDP_MIB_INERRORS, is_udplite);
490 		} else {
491 			UDP6_INC_STATS_USER(sock_net(sk),
492 					UDP_MIB_CSUMERRORS, is_udplite);
493 			UDP6_INC_STATS_USER(sock_net(sk),
494 					UDP_MIB_INERRORS, is_udplite);
495 		}
496 	}
497 	unlock_sock_fast(sk, slow);
498 
499 	/* starting over for a new packet, but check if we need to yield */
500 	cond_resched();
501 	msg->msg_flags &= ~MSG_TRUNC;
502 	goto try_again;
503 }
504 
__udp6_lib_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info,struct udp_table * udptable)505 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
506 		    u8 type, u8 code, int offset, __be32 info,
507 		    struct udp_table *udptable)
508 {
509 	struct ipv6_pinfo *np;
510 	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
511 	const struct in6_addr *saddr = &hdr->saddr;
512 	const struct in6_addr *daddr = &hdr->daddr;
513 	struct udphdr *uh = (struct udphdr*)(skb->data+offset);
514 	struct sock *sk;
515 	int err;
516 
517 	sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
518 			       saddr, uh->source, inet6_iif(skb), udptable);
519 	if (sk == NULL)
520 		return;
521 
522 	if (type == ICMPV6_PKT_TOOBIG)
523 		ip6_sk_update_pmtu(skb, sk, info);
524 	if (type == NDISC_REDIRECT)
525 		ip6_sk_redirect(skb, sk);
526 
527 	np = inet6_sk(sk);
528 
529 	if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
530 		goto out;
531 
532 	if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
533 		goto out;
534 
535 	if (np->recverr)
536 		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
537 
538 	sk->sk_err = err;
539 	sk->sk_error_report(sk);
540 out:
541 	sock_put(sk);
542 }
543 
__udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)544 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
545 {
546 	int rc;
547 
548 	if (!ipv6_addr_any(&inet6_sk(sk)->daddr))
549 		sock_rps_save_rxhash(sk, skb);
550 
551 	rc = sock_queue_rcv_skb(sk, skb);
552 	if (rc < 0) {
553 		int is_udplite = IS_UDPLITE(sk);
554 
555 		/* Note that an ENOMEM error is charged twice */
556 		if (rc == -ENOMEM)
557 			UDP6_INC_STATS_BH(sock_net(sk),
558 					UDP_MIB_RCVBUFERRORS, is_udplite);
559 		UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
560 		kfree_skb(skb);
561 		return -1;
562 	}
563 	return 0;
564 }
565 
udpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)566 static __inline__ void udpv6_err(struct sk_buff *skb,
567 				 struct inet6_skb_parm *opt, u8 type,
568 				 u8 code, int offset, __be32 info     )
569 {
570 	__udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
571 }
572 
573 static struct static_key udpv6_encap_needed __read_mostly;
udpv6_encap_enable(void)574 void udpv6_encap_enable(void)
575 {
576 	if (!static_key_enabled(&udpv6_encap_needed))
577 		static_key_slow_inc(&udpv6_encap_needed);
578 }
579 EXPORT_SYMBOL(udpv6_encap_enable);
580 
udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)581 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
582 {
583 	struct udp_sock *up = udp_sk(sk);
584 	int rc;
585 	int is_udplite = IS_UDPLITE(sk);
586 
587 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
588 		goto drop;
589 
590 	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
591 		int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
592 
593 		/*
594 		 * This is an encapsulation socket so pass the skb to
595 		 * the socket's udp_encap_rcv() hook. Otherwise, just
596 		 * fall through and pass this up the UDP socket.
597 		 * up->encap_rcv() returns the following value:
598 		 * =0 if skb was successfully passed to the encap
599 		 *    handler or was discarded by it.
600 		 * >0 if skb should be passed on to UDP.
601 		 * <0 if skb should be resubmitted as proto -N
602 		 */
603 
604 		/* if we're overly short, let UDP handle it */
605 		encap_rcv = ACCESS_ONCE(up->encap_rcv);
606 		if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
607 			int ret;
608 
609 			ret = encap_rcv(sk, skb);
610 			if (ret <= 0) {
611 				UDP_INC_STATS_BH(sock_net(sk),
612 						 UDP_MIB_INDATAGRAMS,
613 						 is_udplite);
614 				return -ret;
615 			}
616 		}
617 
618 		/* FALLTHROUGH -- it's a UDP Packet */
619 	}
620 
621 	/*
622 	 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
623 	 */
624 	if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
625 
626 		if (up->pcrlen == 0) {          /* full coverage was set  */
627 			LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
628 				" %d while full coverage %d requested\n",
629 				UDP_SKB_CB(skb)->cscov, skb->len);
630 			goto drop;
631 		}
632 		if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
633 			LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
634 						    "too small, need min %d\n",
635 				       UDP_SKB_CB(skb)->cscov, up->pcrlen);
636 			goto drop;
637 		}
638 	}
639 
640 	if (rcu_access_pointer(sk->sk_filter)) {
641 		if (udp_lib_checksum_complete(skb))
642 			goto csum_error;
643 	}
644 
645 	if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
646 		goto drop;
647 
648 	skb_dst_drop(skb);
649 
650 	bh_lock_sock(sk);
651 	rc = 0;
652 	if (!sock_owned_by_user(sk))
653 		rc = __udpv6_queue_rcv_skb(sk, skb);
654 	else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
655 		bh_unlock_sock(sk);
656 		goto drop;
657 	}
658 	bh_unlock_sock(sk);
659 
660 	return rc;
661 csum_error:
662 	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
663 drop:
664 	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
665 	atomic_inc(&sk->sk_drops);
666 	kfree_skb(skb);
667 	return -1;
668 }
669 
udp_v6_mcast_next(struct net * net,struct sock * sk,__be16 loc_port,const struct in6_addr * loc_addr,__be16 rmt_port,const struct in6_addr * rmt_addr,int dif)670 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
671 				      __be16 loc_port, const struct in6_addr *loc_addr,
672 				      __be16 rmt_port, const struct in6_addr *rmt_addr,
673 				      int dif)
674 {
675 	struct hlist_nulls_node *node;
676 	struct sock *s = sk;
677 	unsigned short num = ntohs(loc_port);
678 
679 	sk_nulls_for_each_from(s, node) {
680 		struct inet_sock *inet = inet_sk(s);
681 
682 		if (!net_eq(sock_net(s), net))
683 			continue;
684 
685 		if (udp_sk(s)->udp_port_hash == num &&
686 		    s->sk_family == PF_INET6) {
687 			struct ipv6_pinfo *np = inet6_sk(s);
688 			if (inet->inet_dport) {
689 				if (inet->inet_dport != rmt_port)
690 					continue;
691 			}
692 			if (!ipv6_addr_any(&np->daddr) &&
693 			    !ipv6_addr_equal(&np->daddr, rmt_addr))
694 				continue;
695 
696 			if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
697 				continue;
698 
699 			if (!ipv6_addr_any(&np->rcv_saddr)) {
700 				if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
701 					continue;
702 			}
703 			if (!inet6_mc_check(s, loc_addr, rmt_addr))
704 				continue;
705 			return s;
706 		}
707 	}
708 	return NULL;
709 }
710 
flush_stack(struct sock ** stack,unsigned int count,struct sk_buff * skb,unsigned int final)711 static void flush_stack(struct sock **stack, unsigned int count,
712 			struct sk_buff *skb, unsigned int final)
713 {
714 	struct sk_buff *skb1 = NULL;
715 	struct sock *sk;
716 	unsigned int i;
717 
718 	for (i = 0; i < count; i++) {
719 		sk = stack[i];
720 		if (likely(skb1 == NULL))
721 			skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
722 		if (!skb1) {
723 			atomic_inc(&sk->sk_drops);
724 			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
725 					  IS_UDPLITE(sk));
726 			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
727 					  IS_UDPLITE(sk));
728 		}
729 
730 		if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
731 			skb1 = NULL;
732 	}
733 	if (unlikely(skb1))
734 		kfree_skb(skb1);
735 }
736 /*
737  * Note: called only from the BH handler context,
738  * so we don't need to lock the hashes.
739  */
__udp6_lib_mcast_deliver(struct net * net,struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,struct udp_table * udptable)740 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
741 		const struct in6_addr *saddr, const struct in6_addr *daddr,
742 		struct udp_table *udptable)
743 {
744 	struct sock *sk, *stack[256 / sizeof(struct sock *)];
745 	const struct udphdr *uh = udp_hdr(skb);
746 	struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
747 	int dif;
748 	unsigned int i, count = 0;
749 
750 	spin_lock(&hslot->lock);
751 	sk = sk_nulls_head(&hslot->head);
752 	dif = inet6_iif(skb);
753 	sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
754 	while (sk) {
755 		stack[count++] = sk;
756 		sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
757 				       uh->source, saddr, dif);
758 		if (unlikely(count == ARRAY_SIZE(stack))) {
759 			if (!sk)
760 				break;
761 			flush_stack(stack, count, skb, ~0);
762 			count = 0;
763 		}
764 	}
765 	/*
766 	 * before releasing the lock, we must take reference on sockets
767 	 */
768 	for (i = 0; i < count; i++)
769 		sock_hold(stack[i]);
770 
771 	spin_unlock(&hslot->lock);
772 
773 	if (count) {
774 		flush_stack(stack, count, skb, count - 1);
775 
776 		for (i = 0; i < count; i++)
777 			sock_put(stack[i]);
778 	} else {
779 		kfree_skb(skb);
780 	}
781 	return 0;
782 }
783 
__udp6_lib_rcv(struct sk_buff * skb,struct udp_table * udptable,int proto)784 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
785 		   int proto)
786 {
787 	struct net *net = dev_net(skb->dev);
788 	struct sock *sk;
789 	struct udphdr *uh;
790 	const struct in6_addr *saddr, *daddr;
791 	u32 ulen = 0;
792 
793 	if (!pskb_may_pull(skb, sizeof(struct udphdr)))
794 		goto discard;
795 
796 	saddr = &ipv6_hdr(skb)->saddr;
797 	daddr = &ipv6_hdr(skb)->daddr;
798 	uh = udp_hdr(skb);
799 
800 	ulen = ntohs(uh->len);
801 	if (ulen > skb->len)
802 		goto short_packet;
803 
804 	if (proto == IPPROTO_UDP) {
805 		/* UDP validates ulen. */
806 
807 		/* Check for jumbo payload */
808 		if (ulen == 0)
809 			ulen = skb->len;
810 
811 		if (ulen < sizeof(*uh))
812 			goto short_packet;
813 
814 		if (ulen < skb->len) {
815 			if (pskb_trim_rcsum(skb, ulen))
816 				goto short_packet;
817 			saddr = &ipv6_hdr(skb)->saddr;
818 			daddr = &ipv6_hdr(skb)->daddr;
819 			uh = udp_hdr(skb);
820 		}
821 	}
822 
823 	if (udp6_csum_init(skb, uh, proto))
824 		goto csum_error;
825 
826 	/*
827 	 *	Multicast receive code
828 	 */
829 	if (ipv6_addr_is_multicast(daddr))
830 		return __udp6_lib_mcast_deliver(net, skb,
831 				saddr, daddr, udptable);
832 
833 	/* Unicast */
834 
835 	/*
836 	 * check socket cache ... must talk to Alan about his plans
837 	 * for sock caches... i'll skip this for now.
838 	 */
839 	sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
840 	if (sk != NULL) {
841 		int ret = udpv6_queue_rcv_skb(sk, skb);
842 		sock_put(sk);
843 
844 		/* a return value > 0 means to resubmit the input, but
845 		 * it wants the return to be -protocol, or 0
846 		 */
847 		if (ret > 0)
848 			return -ret;
849 
850 		return 0;
851 	}
852 
853 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
854 		goto discard;
855 
856 	if (udp_lib_checksum_complete(skb))
857 		goto csum_error;
858 
859 	UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
860 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
861 
862 	kfree_skb(skb);
863 	return 0;
864 
865 short_packet:
866 	LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
867 		       proto == IPPROTO_UDPLITE ? "-Lite" : "",
868 		       saddr,
869 		       ntohs(uh->source),
870 		       ulen,
871 		       skb->len,
872 		       daddr,
873 		       ntohs(uh->dest));
874 	goto discard;
875 csum_error:
876 	UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
877 discard:
878 	UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
879 	kfree_skb(skb);
880 	return 0;
881 }
882 
udpv6_rcv(struct sk_buff * skb)883 static __inline__ int udpv6_rcv(struct sk_buff *skb)
884 {
885 	return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
886 }
887 
888 /*
889  * Throw away all pending data and cancel the corking. Socket is locked.
890  */
udp_v6_flush_pending_frames(struct sock * sk)891 static void udp_v6_flush_pending_frames(struct sock *sk)
892 {
893 	struct udp_sock *up = udp_sk(sk);
894 
895 	if (up->pending == AF_INET)
896 		udp_flush_pending_frames(sk);
897 	else if (up->pending) {
898 		up->len = 0;
899 		up->pending = 0;
900 		ip6_flush_pending_frames(sk);
901 	}
902 }
903 
904 /**
905  * 	udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
906  * 	@sk: 	socket we are sending on
907  * 	@skb: 	sk_buff containing the filled-in UDP header
908  * 	        (checksum field must be zeroed out)
909  */
udp6_hwcsum_outgoing(struct sock * sk,struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int len)910 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
911 				 const struct in6_addr *saddr,
912 				 const struct in6_addr *daddr, int len)
913 {
914 	unsigned int offset;
915 	struct udphdr *uh = udp_hdr(skb);
916 	__wsum csum = 0;
917 
918 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
919 		/* Only one fragment on the socket.  */
920 		skb->csum_start = skb_transport_header(skb) - skb->head;
921 		skb->csum_offset = offsetof(struct udphdr, check);
922 		uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
923 	} else {
924 		/*
925 		 * HW-checksum won't work as there are two or more
926 		 * fragments on the socket so that all csums of sk_buffs
927 		 * should be together
928 		 */
929 		offset = skb_transport_offset(skb);
930 		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
931 
932 		skb->ip_summed = CHECKSUM_NONE;
933 
934 		skb_queue_walk(&sk->sk_write_queue, skb) {
935 			csum = csum_add(csum, skb->csum);
936 		}
937 
938 		uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
939 					    csum);
940 		if (uh->check == 0)
941 			uh->check = CSUM_MANGLED_0;
942 	}
943 }
944 
945 /*
946  *	Sending
947  */
948 
udp_v6_push_pending_frames(struct sock * sk)949 static int udp_v6_push_pending_frames(struct sock *sk)
950 {
951 	struct sk_buff *skb;
952 	struct udphdr *uh;
953 	struct udp_sock  *up = udp_sk(sk);
954 	struct inet_sock *inet = inet_sk(sk);
955 	struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
956 	int err = 0;
957 	int is_udplite = IS_UDPLITE(sk);
958 	__wsum csum = 0;
959 
960 	/* Grab the skbuff where UDP header space exists. */
961 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
962 		goto out;
963 
964 	/*
965 	 * Create a UDP header
966 	 */
967 	uh = udp_hdr(skb);
968 	uh->source = fl6->fl6_sport;
969 	uh->dest = fl6->fl6_dport;
970 	uh->len = htons(up->len);
971 	uh->check = 0;
972 
973 	if (is_udplite)
974 		csum = udplite_csum_outgoing(sk, skb);
975 	else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
976 		udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
977 				     up->len);
978 		goto send;
979 	} else
980 		csum = udp_csum_outgoing(sk, skb);
981 
982 	/* add protocol-dependent pseudo-header */
983 	uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
984 				    up->len, fl6->flowi6_proto, csum);
985 	if (uh->check == 0)
986 		uh->check = CSUM_MANGLED_0;
987 
988 send:
989 	err = ip6_push_pending_frames(sk);
990 	if (err) {
991 		if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
992 			UDP6_INC_STATS_USER(sock_net(sk),
993 					    UDP_MIB_SNDBUFERRORS, is_udplite);
994 			err = 0;
995 		}
996 	} else
997 		UDP6_INC_STATS_USER(sock_net(sk),
998 				    UDP_MIB_OUTDATAGRAMS, is_udplite);
999 out:
1000 	up->len = 0;
1001 	up->pending = 0;
1002 	return err;
1003 }
1004 
udpv6_sendmsg(struct kiocb * iocb,struct sock * sk,struct msghdr * msg,size_t len)1005 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1006 		  struct msghdr *msg, size_t len)
1007 {
1008 	struct ipv6_txoptions opt_space;
1009 	struct udp_sock *up = udp_sk(sk);
1010 	struct inet_sock *inet = inet_sk(sk);
1011 	struct ipv6_pinfo *np = inet6_sk(sk);
1012 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
1013 	struct in6_addr *daddr, *final_p, final;
1014 	struct ipv6_txoptions *opt = NULL;
1015 	struct ipv6_txoptions *opt_to_free = NULL;
1016 	struct ip6_flowlabel *flowlabel = NULL;
1017 	struct flowi6 fl6;
1018 	struct dst_entry *dst;
1019 	int addr_len = msg->msg_namelen;
1020 	int ulen = len;
1021 	int hlimit = -1;
1022 	int tclass = -1;
1023 	int dontfrag = -1;
1024 	int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1025 	int err;
1026 	int connected = 0;
1027 	int is_udplite = IS_UDPLITE(sk);
1028 	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1029 
1030 	/* destination address check */
1031 	if (sin6) {
1032 		if (addr_len < offsetof(struct sockaddr, sa_data))
1033 			return -EINVAL;
1034 
1035 		switch (sin6->sin6_family) {
1036 		case AF_INET6:
1037 			if (addr_len < SIN6_LEN_RFC2133)
1038 				return -EINVAL;
1039 			daddr = &sin6->sin6_addr;
1040 			break;
1041 		case AF_INET:
1042 			goto do_udp_sendmsg;
1043 		case AF_UNSPEC:
1044 			msg->msg_name = sin6 = NULL;
1045 			msg->msg_namelen = addr_len = 0;
1046 			daddr = NULL;
1047 			break;
1048 		default:
1049 			return -EINVAL;
1050 		}
1051 	} else if (!up->pending) {
1052 		if (sk->sk_state != TCP_ESTABLISHED)
1053 			return -EDESTADDRREQ;
1054 		daddr = &np->daddr;
1055 	} else
1056 		daddr = NULL;
1057 
1058 	if (daddr) {
1059 		if (ipv6_addr_v4mapped(daddr)) {
1060 			struct sockaddr_in sin;
1061 			sin.sin_family = AF_INET;
1062 			sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1063 			sin.sin_addr.s_addr = daddr->s6_addr32[3];
1064 			msg->msg_name = &sin;
1065 			msg->msg_namelen = sizeof(sin);
1066 do_udp_sendmsg:
1067 			if (__ipv6_only_sock(sk))
1068 				return -ENETUNREACH;
1069 			return udp_sendmsg(iocb, sk, msg, len);
1070 		}
1071 	}
1072 
1073 	if (up->pending == AF_INET)
1074 		return udp_sendmsg(iocb, sk, msg, len);
1075 
1076 	/* Rough check on arithmetic overflow,
1077 	   better check is made in ip6_append_data().
1078 	   */
1079 	if (len > INT_MAX - sizeof(struct udphdr))
1080 		return -EMSGSIZE;
1081 
1082 	if (up->pending) {
1083 		/*
1084 		 * There are pending frames.
1085 		 * The socket lock must be held while it's corked.
1086 		 */
1087 		lock_sock(sk);
1088 		if (likely(up->pending)) {
1089 			if (unlikely(up->pending != AF_INET6)) {
1090 				release_sock(sk);
1091 				return -EAFNOSUPPORT;
1092 			}
1093 			dst = NULL;
1094 			goto do_append_data;
1095 		}
1096 		release_sock(sk);
1097 	}
1098 	ulen += sizeof(struct udphdr);
1099 
1100 	memset(&fl6, 0, sizeof(fl6));
1101 
1102 	if (sin6) {
1103 		if (sin6->sin6_port == 0)
1104 			return -EINVAL;
1105 
1106 		fl6.fl6_dport = sin6->sin6_port;
1107 		daddr = &sin6->sin6_addr;
1108 
1109 		if (np->sndflow) {
1110 			fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1111 			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1112 				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1113 				if (flowlabel == NULL)
1114 					return -EINVAL;
1115 				daddr = &flowlabel->dst;
1116 			}
1117 		}
1118 
1119 		/*
1120 		 * Otherwise it will be difficult to maintain
1121 		 * sk->sk_dst_cache.
1122 		 */
1123 		if (sk->sk_state == TCP_ESTABLISHED &&
1124 		    ipv6_addr_equal(daddr, &np->daddr))
1125 			daddr = &np->daddr;
1126 
1127 		if (addr_len >= sizeof(struct sockaddr_in6) &&
1128 		    sin6->sin6_scope_id &&
1129 		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1130 			fl6.flowi6_oif = sin6->sin6_scope_id;
1131 	} else {
1132 		if (sk->sk_state != TCP_ESTABLISHED)
1133 			return -EDESTADDRREQ;
1134 
1135 		fl6.fl6_dport = inet->inet_dport;
1136 		daddr = &np->daddr;
1137 		fl6.flowlabel = np->flow_label;
1138 		connected = 1;
1139 	}
1140 
1141 	if (!fl6.flowi6_oif)
1142 		fl6.flowi6_oif = sk->sk_bound_dev_if;
1143 
1144 	if (!fl6.flowi6_oif)
1145 		fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1146 
1147 	fl6.flowi6_mark = sk->sk_mark;
1148 	fl6.flowi6_uid = sk->sk_uid;
1149 
1150 	if (msg->msg_controllen) {
1151 		opt = &opt_space;
1152 		memset(opt, 0, sizeof(struct ipv6_txoptions));
1153 		opt->tot_len = sizeof(*opt);
1154 
1155 		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1156 					    &hlimit, &tclass, &dontfrag);
1157 		if (err < 0) {
1158 			fl6_sock_release(flowlabel);
1159 			return err;
1160 		}
1161 		if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1162 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1163 			if (flowlabel == NULL)
1164 				return -EINVAL;
1165 		}
1166 		if (!(opt->opt_nflen|opt->opt_flen))
1167 			opt = NULL;
1168 		connected = 0;
1169 	}
1170 	if (!opt) {
1171 		opt = txopt_get(np);
1172 		opt_to_free = opt;
1173 	}
1174 	if (flowlabel)
1175 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
1176 	opt = ipv6_fixup_options(&opt_space, opt);
1177 
1178 	fl6.flowi6_proto = sk->sk_protocol;
1179 	if (!ipv6_addr_any(daddr))
1180 		fl6.daddr = *daddr;
1181 	else
1182 		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1183 	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1184 		fl6.saddr = np->saddr;
1185 	fl6.fl6_sport = inet->inet_sport;
1186 
1187 	final_p = fl6_update_dst(&fl6, opt, &final);
1188 	if (final_p)
1189 		connected = 0;
1190 
1191 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1192 		fl6.flowi6_oif = np->mcast_oif;
1193 		connected = 0;
1194 	} else if (!fl6.flowi6_oif)
1195 		fl6.flowi6_oif = np->ucast_oif;
1196 
1197 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1198 
1199 	dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
1200 	if (IS_ERR(dst)) {
1201 		err = PTR_ERR(dst);
1202 		dst = NULL;
1203 		goto out;
1204 	}
1205 
1206 	if (hlimit < 0) {
1207 		if (ipv6_addr_is_multicast(&fl6.daddr))
1208 			hlimit = np->mcast_hops;
1209 		else
1210 			hlimit = np->hop_limit;
1211 		if (hlimit < 0)
1212 			hlimit = ip6_dst_hoplimit(dst);
1213 	}
1214 
1215 	if (tclass < 0)
1216 		tclass = np->tclass;
1217 
1218 	if (dontfrag < 0)
1219 		dontfrag = np->dontfrag;
1220 
1221 	if (msg->msg_flags&MSG_CONFIRM)
1222 		goto do_confirm;
1223 back_from_confirm:
1224 
1225 	lock_sock(sk);
1226 	if (unlikely(up->pending)) {
1227 		/* The socket is already corked while preparing it. */
1228 		/* ... which is an evident application bug. --ANK */
1229 		release_sock(sk);
1230 
1231 		LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1232 		err = -EINVAL;
1233 		goto out;
1234 	}
1235 
1236 	up->pending = AF_INET6;
1237 
1238 do_append_data:
1239 	up->len += ulen;
1240 	getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1241 	err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
1242 		sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1243 		(struct rt6_info*)dst,
1244 		corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1245 	if (err)
1246 		udp_v6_flush_pending_frames(sk);
1247 	else if (!corkreq)
1248 		err = udp_v6_push_pending_frames(sk);
1249 	else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1250 		up->pending = 0;
1251 
1252 	if (dst) {
1253 		if (connected) {
1254 			ip6_dst_store(sk, dst,
1255 				      ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
1256 				      &np->daddr : NULL,
1257 #ifdef CONFIG_IPV6_SUBTREES
1258 				      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1259 				      &np->saddr :
1260 #endif
1261 				      NULL);
1262 		} else {
1263 			dst_release(dst);
1264 		}
1265 		dst = NULL;
1266 	}
1267 
1268 	if (err > 0)
1269 		err = np->recverr ? net_xmit_errno(err) : 0;
1270 	release_sock(sk);
1271 out:
1272 	dst_release(dst);
1273 	fl6_sock_release(flowlabel);
1274 	txopt_put(opt_to_free);
1275 	if (!err)
1276 		return len;
1277 	/*
1278 	 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1279 	 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1280 	 * we don't have a good statistic (IpOutDiscards but it can be too many
1281 	 * things).  We could add another new stat but at least for now that
1282 	 * seems like overkill.
1283 	 */
1284 	if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1285 		UDP6_INC_STATS_USER(sock_net(sk),
1286 				UDP_MIB_SNDBUFERRORS, is_udplite);
1287 	}
1288 	return err;
1289 
1290 do_confirm:
1291 	dst_confirm(dst);
1292 	if (!(msg->msg_flags&MSG_PROBE) || len)
1293 		goto back_from_confirm;
1294 	err = 0;
1295 	goto out;
1296 }
1297 
udpv6_destroy_sock(struct sock * sk)1298 void udpv6_destroy_sock(struct sock *sk)
1299 {
1300 	struct udp_sock *up = udp_sk(sk);
1301 	lock_sock(sk);
1302 	udp_v6_flush_pending_frames(sk);
1303 	release_sock(sk);
1304 
1305 	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1306 		void (*encap_destroy)(struct sock *sk);
1307 		encap_destroy = ACCESS_ONCE(up->encap_destroy);
1308 		if (encap_destroy)
1309 			encap_destroy(sk);
1310 	}
1311 
1312 	inet6_destroy_sock(sk);
1313 }
1314 
1315 /*
1316  *	Socket option code for UDP
1317  */
udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1318 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1319 		     char __user *optval, unsigned int optlen)
1320 {
1321 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1322 		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1323 					  udp_v6_push_pending_frames);
1324 	return ipv6_setsockopt(sk, level, optname, optval, optlen);
1325 }
1326 
1327 #ifdef CONFIG_COMPAT
compat_udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1328 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1329 			    char __user *optval, unsigned int optlen)
1330 {
1331 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1332 		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1333 					  udp_v6_push_pending_frames);
1334 	return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1335 }
1336 #endif
1337 
udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1338 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1339 		     char __user *optval, int __user *optlen)
1340 {
1341 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1342 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1343 	return ipv6_getsockopt(sk, level, optname, optval, optlen);
1344 }
1345 
1346 #ifdef CONFIG_COMPAT
compat_udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1347 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1348 			    char __user *optval, int __user *optlen)
1349 {
1350 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1351 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1352 	return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1353 }
1354 #endif
1355 
1356 static const struct inet6_protocol udpv6_protocol = {
1357 	.handler	=	udpv6_rcv,
1358 	.err_handler	=	udpv6_err,
1359 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1360 };
1361 
1362 /* ------------------------------------------------------------------------ */
1363 #ifdef CONFIG_PROC_FS
1364 
udp6_sock_seq_show(struct seq_file * seq,struct sock * sp,int bucket)1365 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1366 {
1367 	struct inet_sock *inet = inet_sk(sp);
1368 	struct ipv6_pinfo *np = inet6_sk(sp);
1369 	const struct in6_addr *dest, *src;
1370 	__u16 destp, srcp;
1371 
1372 	dest  = &np->daddr;
1373 	src   = &np->rcv_saddr;
1374 	destp = ntohs(inet->inet_dport);
1375 	srcp  = ntohs(inet->inet_sport);
1376 	seq_printf(seq,
1377 		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1378 		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1379 		   bucket,
1380 		   src->s6_addr32[0], src->s6_addr32[1],
1381 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1382 		   dest->s6_addr32[0], dest->s6_addr32[1],
1383 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1384 		   sp->sk_state,
1385 		   sk_wmem_alloc_get(sp),
1386 		   sk_rmem_alloc_get(sp),
1387 		   0, 0L, 0,
1388 		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1389 		   0,
1390 		   sock_i_ino(sp),
1391 		   atomic_read(&sp->sk_refcnt), sp,
1392 		   atomic_read(&sp->sk_drops));
1393 }
1394 
udp6_seq_show(struct seq_file * seq,void * v)1395 int udp6_seq_show(struct seq_file *seq, void *v)
1396 {
1397 	if (v == SEQ_START_TOKEN)
1398 		seq_printf(seq,
1399 			   "  sl  "
1400 			   "local_address                         "
1401 			   "remote_address                        "
1402 			   "st tx_queue rx_queue tr tm->when retrnsmt"
1403 			   "   uid  timeout inode ref pointer drops\n");
1404 	else
1405 		udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1406 	return 0;
1407 }
1408 
1409 static const struct file_operations udp6_afinfo_seq_fops = {
1410 	.owner    = THIS_MODULE,
1411 	.open     = udp_seq_open,
1412 	.read     = seq_read,
1413 	.llseek   = seq_lseek,
1414 	.release  = seq_release_net
1415 };
1416 
1417 static struct udp_seq_afinfo udp6_seq_afinfo = {
1418 	.name		= "udp6",
1419 	.family		= AF_INET6,
1420 	.udp_table	= &udp_table,
1421 	.seq_fops	= &udp6_afinfo_seq_fops,
1422 	.seq_ops	= {
1423 		.show		= udp6_seq_show,
1424 	},
1425 };
1426 
udp6_proc_init(struct net * net)1427 int __net_init udp6_proc_init(struct net *net)
1428 {
1429 	return udp_proc_register(net, &udp6_seq_afinfo);
1430 }
1431 
udp6_proc_exit(struct net * net)1432 void udp6_proc_exit(struct net *net) {
1433 	udp_proc_unregister(net, &udp6_seq_afinfo);
1434 }
1435 #endif /* CONFIG_PROC_FS */
1436 
udp_v6_clear_sk(struct sock * sk,int size)1437 void udp_v6_clear_sk(struct sock *sk, int size)
1438 {
1439 	struct inet_sock *inet = inet_sk(sk);
1440 
1441 	/* we do not want to clear pinet6 field, because of RCU lookups */
1442 	sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1443 
1444 	size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1445 	memset(&inet->pinet6 + 1, 0, size);
1446 }
1447 
1448 /* ------------------------------------------------------------------------ */
1449 
1450 struct proto udpv6_prot = {
1451 	.name		   = "UDPv6",
1452 	.owner		   = THIS_MODULE,
1453 	.close		   = udp_lib_close,
1454 	.connect	   = ip6_datagram_connect,
1455 	.disconnect	   = udp_disconnect,
1456 	.ioctl		   = udp_ioctl,
1457 	.destroy	   = udpv6_destroy_sock,
1458 	.setsockopt	   = udpv6_setsockopt,
1459 	.getsockopt	   = udpv6_getsockopt,
1460 	.sendmsg	   = udpv6_sendmsg,
1461 	.recvmsg	   = udpv6_recvmsg,
1462 	.backlog_rcv	   = __udpv6_queue_rcv_skb,
1463 	.hash		   = udp_lib_hash,
1464 	.unhash		   = udp_lib_unhash,
1465 	.rehash		   = udp_v6_rehash,
1466 	.get_port	   = udp_v6_get_port,
1467 	.memory_allocated  = &udp_memory_allocated,
1468 	.sysctl_mem	   = sysctl_udp_mem,
1469 	.sysctl_wmem	   = &sysctl_udp_wmem_min,
1470 	.sysctl_rmem	   = &sysctl_udp_rmem_min,
1471 	.obj_size	   = sizeof(struct udp6_sock),
1472 	.slab_flags	   = SLAB_DESTROY_BY_RCU,
1473 	.h.udp_table	   = &udp_table,
1474 #ifdef CONFIG_COMPAT
1475 	.compat_setsockopt = compat_udpv6_setsockopt,
1476 	.compat_getsockopt = compat_udpv6_getsockopt,
1477 #endif
1478 	.clear_sk	   = udp_v6_clear_sk,
1479 	.diag_destroy      = udp_abort,
1480 };
1481 
1482 static struct inet_protosw udpv6_protosw = {
1483 	.type =      SOCK_DGRAM,
1484 	.protocol =  IPPROTO_UDP,
1485 	.prot =      &udpv6_prot,
1486 	.ops =       &inet6_dgram_ops,
1487 	.no_check =  UDP_CSUM_DEFAULT,
1488 	.flags =     INET_PROTOSW_PERMANENT,
1489 };
1490 
1491 
udpv6_init(void)1492 int __init udpv6_init(void)
1493 {
1494 	int ret;
1495 
1496 	ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1497 	if (ret)
1498 		goto out;
1499 
1500 	ret = inet6_register_protosw(&udpv6_protosw);
1501 	if (ret)
1502 		goto out_udpv6_protocol;
1503 out:
1504 	return ret;
1505 
1506 out_udpv6_protocol:
1507 	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1508 	goto out;
1509 }
1510 
udpv6_exit(void)1511 void udpv6_exit(void)
1512 {
1513 	inet6_unregister_protosw(&udpv6_protosw);
1514 	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1515 }
1516