xref: /linux/net/ipv6/udp.c (revision 3cf0a98fea776adb09087e521fe150c295a4b031)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	UDP over IPv6
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *
9  *	Based on linux/ipv4/udp.c
10  *
11  *	Fixes:
12  *	Hideaki YOSHIFUJI	:	sin6_scope_id support
13  *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
14  *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
15  *					a single port at the same time.
16  *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
17  *      YOSHIFUJI Hideaki @USAGI:	convert /proc/net/udp6 to seq_file.
18  */
19 
20 #include <linux/bpf-cgroup.h>
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/sockios.h>
25 #include <linux/net.h>
26 #include <linux/in6.h>
27 #include <linux/netdevice.h>
28 #include <linux/if_arp.h>
29 #include <linux/ipv6.h>
30 #include <linux/icmpv6.h>
31 #include <linux/init.h>
32 #include <linux/module.h>
33 #include <linux/skbuff.h>
34 #include <linux/slab.h>
35 #include <linux/uaccess.h>
36 #include <linux/indirect_call_wrapper.h>
37 #include <trace/events/udp.h>
38 
39 #include <net/addrconf.h>
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/seg6.h>
46 #include <net/tcp_states.h>
47 #include <net/ip6_checksum.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/inet_hashtables.h>
51 #include <net/inet6_hashtables.h>
52 #include <net/busy_poll.h>
53 #include <net/sock_reuseport.h>
54 #include <net/gro.h>
55 
56 #include <linux/proc_fs.h>
57 #include <linux/seq_file.h>
58 #include <trace/events/skb.h>
59 #include "udp_impl.h"
60 
udpv6_destruct_sock(struct sock * sk)61 static void udpv6_destruct_sock(struct sock *sk)
62 {
63 	udp_destruct_common(sk);
64 	inet6_sock_destruct(sk);
65 }
66 
udpv6_init_sock(struct sock * sk)67 int udpv6_init_sock(struct sock *sk)
68 {
69 	udp_lib_init_sock(sk);
70 	sk->sk_destruct = udpv6_destruct_sock;
71 	set_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags);
72 	return 0;
73 }
74 
75 INDIRECT_CALLABLE_SCOPE
udp6_ehashfn(const struct net * net,const struct in6_addr * laddr,const u16 lport,const struct in6_addr * faddr,const __be16 fport)76 u32 udp6_ehashfn(const struct net *net,
77 		 const struct in6_addr *laddr,
78 		 const u16 lport,
79 		 const struct in6_addr *faddr,
80 		 const __be16 fport)
81 {
82 	u32 lhash, fhash;
83 
84 	net_get_random_once(&udp6_ehash_secret,
85 			    sizeof(udp6_ehash_secret));
86 	net_get_random_once(&udp_ipv6_hash_secret,
87 			    sizeof(udp_ipv6_hash_secret));
88 
89 	lhash = (__force u32)laddr->s6_addr32[3];
90 	fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
91 
92 	return __inet6_ehashfn(lhash, lport, fhash, fport,
93 			       udp6_ehash_secret + net_hash_mix(net));
94 }
95 
udp_v6_get_port(struct sock * sk,unsigned short snum)96 int udp_v6_get_port(struct sock *sk, unsigned short snum)
97 {
98 	unsigned int hash2_nulladdr =
99 		ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
100 	unsigned int hash2_partial =
101 		ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
102 
103 	/* precompute partial secondary hash */
104 	udp_sk(sk)->udp_portaddr_hash = hash2_partial;
105 	return udp_lib_get_port(sk, snum, hash2_nulladdr);
106 }
107 
udp_v6_rehash(struct sock * sk)108 void udp_v6_rehash(struct sock *sk)
109 {
110 	u16 new_hash = ipv6_portaddr_hash(sock_net(sk),
111 					  &sk->sk_v6_rcv_saddr,
112 					  inet_sk(sk)->inet_num);
113 	u16 new_hash4;
114 
115 	if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) {
116 		new_hash4 = udp_ehashfn(sock_net(sk),
117 					sk->sk_rcv_saddr, sk->sk_num,
118 					sk->sk_daddr, sk->sk_dport);
119 	} else {
120 		new_hash4 = udp6_ehashfn(sock_net(sk),
121 					 &sk->sk_v6_rcv_saddr, sk->sk_num,
122 					 &sk->sk_v6_daddr, sk->sk_dport);
123 	}
124 
125 	udp_lib_rehash(sk, new_hash, new_hash4);
126 }
127 
compute_score(struct sock * sk,const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned short hnum,int dif,int sdif)128 static int compute_score(struct sock *sk, const struct net *net,
129 			 const struct in6_addr *saddr, __be16 sport,
130 			 const struct in6_addr *daddr, unsigned short hnum,
131 			 int dif, int sdif)
132 {
133 	int bound_dev_if, score;
134 	struct inet_sock *inet;
135 	bool dev_match;
136 
137 	if (!net_eq(sock_net(sk), net) ||
138 	    udp_sk(sk)->udp_port_hash != hnum ||
139 	    sk->sk_family != PF_INET6)
140 		return -1;
141 
142 	if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
143 		return -1;
144 
145 	score = 0;
146 	inet = inet_sk(sk);
147 
148 	if (inet->inet_dport) {
149 		if (inet->inet_dport != sport)
150 			return -1;
151 		score++;
152 	}
153 
154 	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
155 		if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
156 			return -1;
157 		score++;
158 	}
159 
160 	bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
161 	dev_match = udp_sk_bound_dev_eq(net, bound_dev_if, dif, sdif);
162 	if (!dev_match)
163 		return -1;
164 	if (bound_dev_if)
165 		score++;
166 
167 	if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id())
168 		score++;
169 
170 	return score;
171 }
172 
173 /**
174  * udp6_lib_lookup1() - Simplified lookup using primary hash (destination port)
175  * @net:	Network namespace
176  * @saddr:	Source address, network order
177  * @sport:	Source port, network order
178  * @daddr:	Destination address, network order
179  * @hnum:	Destination port, host order
180  * @dif:	Destination interface index
181  * @sdif:	Destination bridge port index, if relevant
182  * @udptable:	Set of UDP hash tables
183  *
184  * Simplified lookup to be used as fallback if no sockets are found due to a
185  * potential race between (receive) address change, and lookup happening before
186  * the rehash operation. This function ignores SO_REUSEPORT groups while scoring
187  * result sockets, because if we have one, we don't need the fallback at all.
188  *
189  * Called under rcu_read_lock().
190  *
191  * Return: socket with highest matching score if any, NULL if none
192  */
udp6_lib_lookup1(const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,int sdif,const struct udp_table * udptable)193 static struct sock *udp6_lib_lookup1(const struct net *net,
194 				     const struct in6_addr *saddr, __be16 sport,
195 				     const struct in6_addr *daddr,
196 				     unsigned int hnum, int dif, int sdif,
197 				     const struct udp_table *udptable)
198 {
199 	unsigned int slot = udp_hashfn(net, hnum, udptable->mask);
200 	struct udp_hslot *hslot = &udptable->hash[slot];
201 	struct sock *sk, *result = NULL;
202 	int score, badness = 0;
203 
204 	sk_for_each_rcu(sk, &hslot->head) {
205 		score = compute_score(sk, net,
206 				      saddr, sport, daddr, hnum, dif, sdif);
207 		if (score > badness) {
208 			result = sk;
209 			badness = score;
210 		}
211 	}
212 
213 	return result;
214 }
215 
216 /* called with rcu_read_lock() */
udp6_lib_lookup2(const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,int sdif,struct udp_hslot * hslot2,struct sk_buff * skb)217 static struct sock *udp6_lib_lookup2(const struct net *net,
218 		const struct in6_addr *saddr, __be16 sport,
219 		const struct in6_addr *daddr, unsigned int hnum,
220 		int dif, int sdif, struct udp_hslot *hslot2,
221 		struct sk_buff *skb)
222 {
223 	struct sock *sk, *result;
224 	int score, badness;
225 	bool need_rescore;
226 
227 	result = NULL;
228 	badness = -1;
229 	udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
230 		need_rescore = false;
231 rescore:
232 		score = compute_score(need_rescore ? result : sk, net, saddr,
233 				      sport, daddr, hnum, dif, sdif);
234 		if (score > badness) {
235 			badness = score;
236 
237 			if (need_rescore)
238 				continue;
239 
240 			if (sk->sk_state == TCP_ESTABLISHED) {
241 				result = sk;
242 				continue;
243 			}
244 
245 			result = inet6_lookup_reuseport(net, sk, skb, sizeof(struct udphdr),
246 							saddr, sport, daddr, hnum, udp6_ehashfn);
247 			if (!result) {
248 				result = sk;
249 				continue;
250 			}
251 
252 			/* Fall back to scoring if group has connections */
253 			if (!reuseport_has_conns(sk))
254 				return result;
255 
256 			/* Reuseport logic returned an error, keep original score. */
257 			if (IS_ERR(result))
258 				continue;
259 
260 			/* compute_score is too long of a function to be
261 			 * inlined, and calling it again here yields
262 			 * measureable overhead for some
263 			 * workloads. Work around it by jumping
264 			 * backwards to rescore 'result'.
265 			 */
266 			need_rescore = true;
267 			goto rescore;
268 		}
269 	}
270 	return result;
271 }
272 
273 #if IS_ENABLED(CONFIG_BASE_SMALL)
udp6_lib_lookup4(const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,int sdif,struct udp_table * udptable)274 static struct sock *udp6_lib_lookup4(const struct net *net,
275 				     const struct in6_addr *saddr, __be16 sport,
276 				     const struct in6_addr *daddr,
277 				     unsigned int hnum, int dif, int sdif,
278 				     struct udp_table *udptable)
279 {
280 	return NULL;
281 }
282 
udp6_hash4(struct sock * sk)283 static void udp6_hash4(struct sock *sk)
284 {
285 }
286 #else /* !CONFIG_BASE_SMALL */
udp6_lib_lookup4(const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,int sdif,struct udp_table * udptable)287 static struct sock *udp6_lib_lookup4(const struct net *net,
288 				     const struct in6_addr *saddr, __be16 sport,
289 				     const struct in6_addr *daddr,
290 				     unsigned int hnum, int dif, int sdif,
291 				     struct udp_table *udptable)
292 {
293 	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
294 	const struct hlist_nulls_node *node;
295 	struct udp_hslot *hslot4;
296 	unsigned int hash4, slot;
297 	struct udp_sock *up;
298 	struct sock *sk;
299 
300 	hash4 = udp6_ehashfn(net, daddr, hnum, saddr, sport);
301 	slot = hash4 & udptable->mask;
302 	hslot4 = &udptable->hash4[slot];
303 
304 begin:
305 	udp_lrpa_for_each_entry_rcu(up, node, &hslot4->nulls_head) {
306 		sk = (struct sock *)up;
307 		if (inet6_match(net, sk, saddr, daddr, ports, dif, sdif))
308 			return sk;
309 	}
310 
311 	/* if the nulls value we got at the end of this lookup is not the
312 	 * expected one, we must restart lookup. We probably met an item that
313 	 * was moved to another chain due to rehash.
314 	 */
315 	if (get_nulls_value(node) != slot)
316 		goto begin;
317 
318 	return NULL;
319 }
320 
udp6_hash4(struct sock * sk)321 static void udp6_hash4(struct sock *sk)
322 {
323 	struct net *net = sock_net(sk);
324 	unsigned int hash;
325 
326 	if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) {
327 		udp4_hash4(sk);
328 		return;
329 	}
330 
331 	if (sk_unhashed(sk) || ipv6_addr_any(&sk->sk_v6_rcv_saddr))
332 		return;
333 
334 	hash = udp6_ehashfn(net, &sk->sk_v6_rcv_saddr, sk->sk_num,
335 			    &sk->sk_v6_daddr, sk->sk_dport);
336 
337 	udp_lib_hash4(sk, hash);
338 }
339 #endif /* CONFIG_BASE_SMALL */
340 
341 /* rcu_read_lock() must be held */
__udp6_lib_lookup(const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif,int sdif,struct udp_table * udptable,struct sk_buff * skb)342 struct sock *__udp6_lib_lookup(const struct net *net,
343 			       const struct in6_addr *saddr, __be16 sport,
344 			       const struct in6_addr *daddr, __be16 dport,
345 			       int dif, int sdif, struct udp_table *udptable,
346 			       struct sk_buff *skb)
347 {
348 	unsigned short hnum = ntohs(dport);
349 	struct udp_hslot *hslot2;
350 	struct sock *result, *sk;
351 	unsigned int hash2;
352 
353 	hash2 = ipv6_portaddr_hash(net, daddr, hnum);
354 	hslot2 = udp_hashslot2(udptable, hash2);
355 
356 	if (udp_has_hash4(hslot2)) {
357 		result = udp6_lib_lookup4(net, saddr, sport, daddr, hnum,
358 					  dif, sdif, udptable);
359 		if (result) /* udp6_lib_lookup4 return sk or NULL */
360 			return result;
361 	}
362 
363 	/* Lookup connected or non-wildcard sockets */
364 	result = udp6_lib_lookup2(net, saddr, sport,
365 				  daddr, hnum, dif, sdif,
366 				  hslot2, skb);
367 	if (!IS_ERR_OR_NULL(result) && result->sk_state == TCP_ESTABLISHED)
368 		goto done;
369 
370 	/* Lookup redirect from BPF */
371 	if (static_branch_unlikely(&bpf_sk_lookup_enabled) &&
372 	    udptable == net->ipv4.udp_table) {
373 		sk = inet6_lookup_run_sk_lookup(net, IPPROTO_UDP, skb, sizeof(struct udphdr),
374 						saddr, sport, daddr, hnum, dif,
375 						udp6_ehashfn);
376 		if (sk) {
377 			result = sk;
378 			goto done;
379 		}
380 	}
381 
382 	/* Got non-wildcard socket or error on first lookup */
383 	if (result)
384 		goto done;
385 
386 	/* Lookup wildcard sockets */
387 	hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum);
388 	hslot2 = udp_hashslot2(udptable, hash2);
389 
390 	result = udp6_lib_lookup2(net, saddr, sport,
391 				  &in6addr_any, hnum, dif, sdif,
392 				  hslot2, skb);
393 	if (!IS_ERR_OR_NULL(result))
394 		goto done;
395 
396 	/* Cover address change/lookup/rehash race: see __udp4_lib_lookup() */
397 	result = udp6_lib_lookup1(net, saddr, sport, daddr, hnum, dif, sdif,
398 				  udptable);
399 
400 done:
401 	if (IS_ERR(result))
402 		return NULL;
403 	return result;
404 }
405 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
406 
__udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport,struct udp_table * udptable)407 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
408 					  __be16 sport, __be16 dport,
409 					  struct udp_table *udptable)
410 {
411 	const struct ipv6hdr *iph = ipv6_hdr(skb);
412 
413 	return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
414 				 &iph->daddr, dport, inet6_iif(skb),
415 				 inet6_sdif(skb), udptable, skb);
416 }
417 
udp6_lib_lookup_skb(const struct sk_buff * skb,__be16 sport,__be16 dport)418 struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb,
419 				 __be16 sport, __be16 dport)
420 {
421 	const u16 offset = NAPI_GRO_CB(skb)->network_offsets[skb->encapsulation];
422 	const struct ipv6hdr *iph = (struct ipv6hdr *)(skb->data + offset);
423 	struct net *net = dev_net(skb->dev);
424 	int iif, sdif;
425 
426 	inet6_get_iif_sdif(skb, &iif, &sdif);
427 
428 	return __udp6_lib_lookup(net, &iph->saddr, sport,
429 				 &iph->daddr, dport, iif,
430 				 sdif, net->ipv4.udp_table, NULL);
431 }
432 
433 /* Must be called under rcu_read_lock().
434  * Does increment socket refcount.
435  */
436 #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6)
udp6_lib_lookup(const struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)437 struct sock *udp6_lib_lookup(const struct net *net, const struct in6_addr *saddr, __be16 sport,
438 			     const struct in6_addr *daddr, __be16 dport, int dif)
439 {
440 	struct sock *sk;
441 
442 	sk =  __udp6_lib_lookup(net, saddr, sport, daddr, dport,
443 				dif, 0, net->ipv4.udp_table, NULL);
444 	if (sk && !refcount_inc_not_zero(&sk->sk_refcnt))
445 		sk = NULL;
446 	return sk;
447 }
448 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
449 #endif
450 
451 /* do not use the scratch area len for jumbogram: their length execeeds the
452  * scratch area space; note that the IP6CB flags is still in the first
453  * cacheline, so checking for jumbograms is cheap
454  */
udp6_skb_len(struct sk_buff * skb)455 static int udp6_skb_len(struct sk_buff *skb)
456 {
457 	return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb);
458 }
459 
460 /*
461  *	This should be easy, if there is something there we
462  *	return it, otherwise we block.
463  */
464 
udpv6_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int flags,int * addr_len)465 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
466 		  int flags, int *addr_len)
467 {
468 	struct ipv6_pinfo *np = inet6_sk(sk);
469 	struct inet_sock *inet = inet_sk(sk);
470 	struct sk_buff *skb;
471 	unsigned int ulen, copied;
472 	int off, err, peeking = flags & MSG_PEEK;
473 	int is_udplite = IS_UDPLITE(sk);
474 	struct udp_mib __percpu *mib;
475 	bool checksum_valid = false;
476 	int is_udp4;
477 
478 	if (flags & MSG_ERRQUEUE)
479 		return ipv6_recv_error(sk, msg, len, addr_len);
480 
481 	if (np->rxpmtu && np->rxopt.bits.rxpmtu)
482 		return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
483 
484 try_again:
485 	off = sk_peek_offset(sk, flags);
486 	skb = __skb_recv_udp(sk, flags, &off, &err);
487 	if (!skb)
488 		return err;
489 
490 	ulen = udp6_skb_len(skb);
491 	copied = len;
492 	if (copied > ulen - off)
493 		copied = ulen - off;
494 	else if (copied < ulen)
495 		msg->msg_flags |= MSG_TRUNC;
496 
497 	is_udp4 = (skb->protocol == htons(ETH_P_IP));
498 	mib = __UDPX_MIB(sk, is_udp4);
499 
500 	/*
501 	 * If checksum is needed at all, try to do it while copying the
502 	 * data.  If the data is truncated, or if we only want a partial
503 	 * coverage checksum (UDP-Lite), do it before the copy.
504 	 */
505 
506 	if (copied < ulen || peeking ||
507 	    (is_udplite && UDP_SKB_CB(skb)->partial_cov)) {
508 		checksum_valid = udp_skb_csum_unnecessary(skb) ||
509 				!__udp_lib_checksum_complete(skb);
510 		if (!checksum_valid)
511 			goto csum_copy_err;
512 	}
513 
514 	if (checksum_valid || udp_skb_csum_unnecessary(skb)) {
515 		if (udp_skb_is_linear(skb))
516 			err = copy_linear_skb(skb, copied, off, &msg->msg_iter);
517 		else
518 			err = skb_copy_datagram_msg(skb, off, msg, copied);
519 	} else {
520 		err = skb_copy_and_csum_datagram_msg(skb, off, msg);
521 		if (err == -EINVAL)
522 			goto csum_copy_err;
523 	}
524 	if (unlikely(err)) {
525 		if (!peeking) {
526 			atomic_inc(&sk->sk_drops);
527 			SNMP_INC_STATS(mib, UDP_MIB_INERRORS);
528 		}
529 		kfree_skb(skb);
530 		return err;
531 	}
532 	if (!peeking)
533 		SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS);
534 
535 	sock_recv_cmsgs(msg, sk, skb);
536 
537 	/* Copy the address. */
538 	if (msg->msg_name) {
539 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
540 		sin6->sin6_family = AF_INET6;
541 		sin6->sin6_port = udp_hdr(skb)->source;
542 		sin6->sin6_flowinfo = 0;
543 
544 		if (is_udp4) {
545 			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
546 					       &sin6->sin6_addr);
547 			sin6->sin6_scope_id = 0;
548 		} else {
549 			sin6->sin6_addr = ipv6_hdr(skb)->saddr;
550 			sin6->sin6_scope_id =
551 				ipv6_iface_scope_id(&sin6->sin6_addr,
552 						    inet6_iif(skb));
553 		}
554 		*addr_len = sizeof(*sin6);
555 
556 		BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk,
557 						      (struct sockaddr *)sin6,
558 						      addr_len);
559 	}
560 
561 	if (udp_test_bit(GRO_ENABLED, sk))
562 		udp_cmsg_recv(msg, sk, skb);
563 
564 	if (np->rxopt.all)
565 		ip6_datagram_recv_common_ctl(sk, msg, skb);
566 
567 	if (is_udp4) {
568 		if (inet_cmsg_flags(inet))
569 			ip_cmsg_recv_offset(msg, sk, skb,
570 					    sizeof(struct udphdr), off);
571 	} else {
572 		if (np->rxopt.all)
573 			ip6_datagram_recv_specific_ctl(sk, msg, skb);
574 	}
575 
576 	err = copied;
577 	if (flags & MSG_TRUNC)
578 		err = ulen;
579 
580 	skb_consume_udp(sk, skb, peeking ? -err : err);
581 	return err;
582 
583 csum_copy_err:
584 	if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags,
585 				 udp_skb_destructor)) {
586 		SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS);
587 		SNMP_INC_STATS(mib, UDP_MIB_INERRORS);
588 	}
589 	kfree_skb(skb);
590 
591 	/* starting over for a new packet, but check if we need to yield */
592 	cond_resched();
593 	msg->msg_flags &= ~MSG_TRUNC;
594 	goto try_again;
595 }
596 
597 DECLARE_STATIC_KEY_FALSE(udpv6_encap_needed_key);
udpv6_encap_enable(void)598 void udpv6_encap_enable(void)
599 {
600 	static_branch_inc(&udpv6_encap_needed_key);
601 }
602 EXPORT_SYMBOL(udpv6_encap_enable);
603 
604 /* Handler for tunnels with arbitrary destination ports: no socket lookup, go
605  * through error handlers in encapsulations looking for a match.
606  */
__udp6_lib_err_encap_no_sk(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)607 static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb,
608 				      struct inet6_skb_parm *opt,
609 				      u8 type, u8 code, int offset, __be32 info)
610 {
611 	int i;
612 
613 	for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) {
614 		int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt,
615 			       u8 type, u8 code, int offset, __be32 info);
616 		const struct ip6_tnl_encap_ops *encap;
617 
618 		encap = rcu_dereference(ip6tun_encaps[i]);
619 		if (!encap)
620 			continue;
621 		handler = encap->err_handler;
622 		if (handler && !handler(skb, opt, type, code, offset, info))
623 			return 0;
624 	}
625 
626 	return -ENOENT;
627 }
628 
629 /* Try to match ICMP errors to UDP tunnels by looking up a socket without
630  * reversing source and destination port: this will match tunnels that force the
631  * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that
632  * lwtunnels might actually break this assumption by being configured with
633  * different destination ports on endpoints, in this case we won't be able to
634  * trace ICMP messages back to them.
635  *
636  * If this doesn't match any socket, probe tunnels with arbitrary destination
637  * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port
638  * we've sent packets to won't necessarily match the local destination port.
639  *
640  * Then ask the tunnel implementation to match the error against a valid
641  * association.
642  *
643  * Return an error if we can't find a match, the socket if we need further
644  * processing, zero otherwise.
645  */
__udp6_lib_err_encap(struct net * net,const struct ipv6hdr * hdr,int offset,struct udphdr * uh,struct udp_table * udptable,struct sock * sk,struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,__be32 info)646 static struct sock *__udp6_lib_err_encap(struct net *net,
647 					 const struct ipv6hdr *hdr, int offset,
648 					 struct udphdr *uh,
649 					 struct udp_table *udptable,
650 					 struct sock *sk,
651 					 struct sk_buff *skb,
652 					 struct inet6_skb_parm *opt,
653 					 u8 type, u8 code, __be32 info)
654 {
655 	int (*lookup)(struct sock *sk, struct sk_buff *skb);
656 	int network_offset, transport_offset;
657 	struct udp_sock *up;
658 
659 	network_offset = skb_network_offset(skb);
660 	transport_offset = skb_transport_offset(skb);
661 
662 	/* Network header needs to point to the outer IPv6 header inside ICMP */
663 	skb_reset_network_header(skb);
664 
665 	/* Transport header needs to point to the UDP header */
666 	skb_set_transport_header(skb, offset);
667 
668 	if (sk) {
669 		up = udp_sk(sk);
670 
671 		lookup = READ_ONCE(up->encap_err_lookup);
672 		if (lookup && lookup(sk, skb))
673 			sk = NULL;
674 
675 		goto out;
676 	}
677 
678 	sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source,
679 			       &hdr->saddr, uh->dest,
680 			       inet6_iif(skb), 0, udptable, skb);
681 	if (sk) {
682 		up = udp_sk(sk);
683 
684 		lookup = READ_ONCE(up->encap_err_lookup);
685 		if (!lookup || lookup(sk, skb))
686 			sk = NULL;
687 	}
688 
689 out:
690 	if (!sk) {
691 		sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code,
692 							offset, info));
693 	}
694 
695 	skb_set_transport_header(skb, transport_offset);
696 	skb_set_network_header(skb, network_offset);
697 
698 	return sk;
699 }
700 
__udp6_lib_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info,struct udp_table * udptable)701 int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
702 		   u8 type, u8 code, int offset, __be32 info,
703 		   struct udp_table *udptable)
704 {
705 	struct ipv6_pinfo *np;
706 	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
707 	const struct in6_addr *saddr = &hdr->saddr;
708 	const struct in6_addr *daddr = seg6_get_daddr(skb, opt) ? : &hdr->daddr;
709 	struct udphdr *uh = (struct udphdr *)(skb->data+offset);
710 	bool tunnel = false;
711 	struct sock *sk;
712 	int harderr;
713 	int err;
714 	struct net *net = dev_net(skb->dev);
715 
716 	sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
717 			       inet6_iif(skb), inet6_sdif(skb), udptable, NULL);
718 
719 	if (!sk || READ_ONCE(udp_sk(sk)->encap_type)) {
720 		/* No socket for error: try tunnels before discarding */
721 		if (static_branch_unlikely(&udpv6_encap_needed_key)) {
722 			sk = __udp6_lib_err_encap(net, hdr, offset, uh,
723 						  udptable, sk, skb,
724 						  opt, type, code, info);
725 			if (!sk)
726 				return 0;
727 		} else
728 			sk = ERR_PTR(-ENOENT);
729 
730 		if (IS_ERR(sk)) {
731 			__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
732 					  ICMP6_MIB_INERRORS);
733 			return PTR_ERR(sk);
734 		}
735 
736 		tunnel = true;
737 	}
738 
739 	harderr = icmpv6_err_convert(type, code, &err);
740 	np = inet6_sk(sk);
741 
742 	if (type == ICMPV6_PKT_TOOBIG) {
743 		if (!ip6_sk_accept_pmtu(sk))
744 			goto out;
745 		ip6_sk_update_pmtu(skb, sk, info);
746 		if (READ_ONCE(np->pmtudisc) != IPV6_PMTUDISC_DONT)
747 			harderr = 1;
748 	}
749 	if (type == NDISC_REDIRECT) {
750 		if (tunnel) {
751 			ip6_redirect(skb, sock_net(sk), inet6_iif(skb),
752 				     READ_ONCE(sk->sk_mark), sk->sk_uid);
753 		} else {
754 			ip6_sk_redirect(skb, sk);
755 		}
756 		goto out;
757 	}
758 
759 	/* Tunnels don't have an application socket: don't pass errors back */
760 	if (tunnel) {
761 		if (udp_sk(sk)->encap_err_rcv)
762 			udp_sk(sk)->encap_err_rcv(sk, skb, err, uh->dest,
763 						  ntohl(info), (u8 *)(uh+1));
764 		goto out;
765 	}
766 
767 	if (!inet6_test_bit(RECVERR6, sk)) {
768 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
769 			goto out;
770 	} else {
771 		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
772 	}
773 
774 	sk->sk_err = err;
775 	sk_error_report(sk);
776 out:
777 	return 0;
778 }
779 
__udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)780 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
781 {
782 	int rc;
783 
784 	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
785 		sock_rps_save_rxhash(sk, skb);
786 		sk_mark_napi_id(sk, skb);
787 		sk_incoming_cpu_update(sk);
788 	} else {
789 		sk_mark_napi_id_once(sk, skb);
790 	}
791 
792 	rc = __udp_enqueue_schedule_skb(sk, skb);
793 	if (rc < 0) {
794 		int is_udplite = IS_UDPLITE(sk);
795 		enum skb_drop_reason drop_reason;
796 
797 		/* Note that an ENOMEM error is charged twice */
798 		if (rc == -ENOMEM) {
799 			UDP6_INC_STATS(sock_net(sk),
800 					 UDP_MIB_RCVBUFERRORS, is_udplite);
801 			drop_reason = SKB_DROP_REASON_SOCKET_RCVBUFF;
802 		} else {
803 			UDP6_INC_STATS(sock_net(sk),
804 				       UDP_MIB_MEMERRORS, is_udplite);
805 			drop_reason = SKB_DROP_REASON_PROTO_MEM;
806 		}
807 		UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
808 		trace_udp_fail_queue_rcv_skb(rc, sk, skb);
809 		sk_skb_reason_drop(sk, skb, drop_reason);
810 		return -1;
811 	}
812 
813 	return 0;
814 }
815 
udpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)816 static __inline__ int udpv6_err(struct sk_buff *skb,
817 				struct inet6_skb_parm *opt, u8 type,
818 				u8 code, int offset, __be32 info)
819 {
820 	return __udp6_lib_err(skb, opt, type, code, offset, info,
821 			      dev_net(skb->dev)->ipv4.udp_table);
822 }
823 
udpv6_queue_rcv_one_skb(struct sock * sk,struct sk_buff * skb)824 static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb)
825 {
826 	enum skb_drop_reason drop_reason = SKB_DROP_REASON_NOT_SPECIFIED;
827 	struct udp_sock *up = udp_sk(sk);
828 	int is_udplite = IS_UDPLITE(sk);
829 
830 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
831 		drop_reason = SKB_DROP_REASON_XFRM_POLICY;
832 		goto drop;
833 	}
834 	nf_reset_ct(skb);
835 
836 	if (static_branch_unlikely(&udpv6_encap_needed_key) &&
837 	    READ_ONCE(up->encap_type)) {
838 		int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
839 
840 		/*
841 		 * This is an encapsulation socket so pass the skb to
842 		 * the socket's udp_encap_rcv() hook. Otherwise, just
843 		 * fall through and pass this up the UDP socket.
844 		 * up->encap_rcv() returns the following value:
845 		 * =0 if skb was successfully passed to the encap
846 		 *    handler or was discarded by it.
847 		 * >0 if skb should be passed on to UDP.
848 		 * <0 if skb should be resubmitted as proto -N
849 		 */
850 
851 		/* if we're overly short, let UDP handle it */
852 		encap_rcv = READ_ONCE(up->encap_rcv);
853 		if (encap_rcv) {
854 			int ret;
855 
856 			/* Verify checksum before giving to encap */
857 			if (udp_lib_checksum_complete(skb))
858 				goto csum_error;
859 
860 			ret = encap_rcv(sk, skb);
861 			if (ret <= 0) {
862 				__UDP6_INC_STATS(sock_net(sk),
863 						 UDP_MIB_INDATAGRAMS,
864 						 is_udplite);
865 				return -ret;
866 			}
867 		}
868 
869 		/* FALLTHROUGH -- it's a UDP Packet */
870 	}
871 
872 	/*
873 	 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
874 	 */
875 	if (udp_test_bit(UDPLITE_RECV_CC, sk) && UDP_SKB_CB(skb)->partial_cov) {
876 		u16 pcrlen = READ_ONCE(up->pcrlen);
877 
878 		if (pcrlen == 0) {          /* full coverage was set  */
879 			net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
880 					    UDP_SKB_CB(skb)->cscov, skb->len);
881 			goto drop;
882 		}
883 		if (UDP_SKB_CB(skb)->cscov < pcrlen) {
884 			net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
885 					    UDP_SKB_CB(skb)->cscov, pcrlen);
886 			goto drop;
887 		}
888 	}
889 
890 	prefetch(&sk->sk_rmem_alloc);
891 	if (rcu_access_pointer(sk->sk_filter) &&
892 	    udp_lib_checksum_complete(skb))
893 		goto csum_error;
894 
895 	if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) {
896 		drop_reason = SKB_DROP_REASON_SOCKET_FILTER;
897 		goto drop;
898 	}
899 
900 	udp_csum_pull_header(skb);
901 
902 	skb_dst_drop(skb);
903 
904 	return __udpv6_queue_rcv_skb(sk, skb);
905 
906 csum_error:
907 	drop_reason = SKB_DROP_REASON_UDP_CSUM;
908 	__UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
909 drop:
910 	__UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
911 	atomic_inc(&sk->sk_drops);
912 	sk_skb_reason_drop(sk, skb, drop_reason);
913 	return -1;
914 }
915 
udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)916 static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
917 {
918 	struct sk_buff *next, *segs;
919 	int ret;
920 
921 	if (likely(!udp_unexpected_gso(sk, skb)))
922 		return udpv6_queue_rcv_one_skb(sk, skb);
923 
924 	__skb_push(skb, -skb_mac_offset(skb));
925 	segs = udp_rcv_segment(sk, skb, false);
926 	skb_list_walk_safe(segs, skb, next) {
927 		__skb_pull(skb, skb_transport_offset(skb));
928 
929 		udp_post_segment_fix_csum(skb);
930 		ret = udpv6_queue_rcv_one_skb(sk, skb);
931 		if (ret > 0)
932 			ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret,
933 						 true);
934 	}
935 	return 0;
936 }
937 
__udp_v6_is_mcast_sock(struct net * net,const struct sock * sk,__be16 loc_port,const struct in6_addr * loc_addr,__be16 rmt_port,const struct in6_addr * rmt_addr,int dif,int sdif,unsigned short hnum)938 static bool __udp_v6_is_mcast_sock(struct net *net, const struct sock *sk,
939 				   __be16 loc_port, const struct in6_addr *loc_addr,
940 				   __be16 rmt_port, const struct in6_addr *rmt_addr,
941 				   int dif, int sdif, unsigned short hnum)
942 {
943 	const struct inet_sock *inet = inet_sk(sk);
944 
945 	if (!net_eq(sock_net(sk), net))
946 		return false;
947 
948 	if (udp_sk(sk)->udp_port_hash != hnum ||
949 	    sk->sk_family != PF_INET6 ||
950 	    (inet->inet_dport && inet->inet_dport != rmt_port) ||
951 	    (!ipv6_addr_any(&sk->sk_v6_daddr) &&
952 		    !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
953 	    !udp_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif, sdif) ||
954 	    (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
955 		    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
956 		return false;
957 	if (!inet6_mc_check(sk, loc_addr, rmt_addr))
958 		return false;
959 	return true;
960 }
961 
udp6_csum_zero_error(struct sk_buff * skb)962 static void udp6_csum_zero_error(struct sk_buff *skb)
963 {
964 	/* RFC 2460 section 8.1 says that we SHOULD log
965 	 * this error. Well, it is reasonable.
966 	 */
967 	net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
968 			    &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
969 			    &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
970 }
971 
972 /*
973  * Note: called only from the BH handler context,
974  * so we don't need to lock the hashes.
975  */
__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,int proto)976 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
977 		const struct in6_addr *saddr, const struct in6_addr *daddr,
978 		struct udp_table *udptable, int proto)
979 {
980 	struct sock *sk, *first = NULL;
981 	const struct udphdr *uh = udp_hdr(skb);
982 	unsigned short hnum = ntohs(uh->dest);
983 	struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
984 	unsigned int offset = offsetof(typeof(*sk), sk_node);
985 	unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
986 	int dif = inet6_iif(skb);
987 	int sdif = inet6_sdif(skb);
988 	struct hlist_node *node;
989 	struct sk_buff *nskb;
990 
991 	if (use_hash2) {
992 		hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) &
993 			    udptable->mask;
994 		hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask;
995 start_lookup:
996 		hslot = &udptable->hash2[hash2].hslot;
997 		offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
998 	}
999 
1000 	sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
1001 		if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
1002 					    uh->source, saddr, dif, sdif,
1003 					    hnum))
1004 			continue;
1005 		/* If zero checksum and no_check is not on for
1006 		 * the socket then skip it.
1007 		 */
1008 		if (!uh->check && !udp_get_no_check6_rx(sk))
1009 			continue;
1010 		if (!first) {
1011 			first = sk;
1012 			continue;
1013 		}
1014 		nskb = skb_clone(skb, GFP_ATOMIC);
1015 		if (unlikely(!nskb)) {
1016 			atomic_inc(&sk->sk_drops);
1017 			__UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
1018 					 IS_UDPLITE(sk));
1019 			__UDP6_INC_STATS(net, UDP_MIB_INERRORS,
1020 					 IS_UDPLITE(sk));
1021 			continue;
1022 		}
1023 
1024 		if (udpv6_queue_rcv_skb(sk, nskb) > 0)
1025 			consume_skb(nskb);
1026 	}
1027 
1028 	/* Also lookup *:port if we are using hash2 and haven't done so yet. */
1029 	if (use_hash2 && hash2 != hash2_any) {
1030 		hash2 = hash2_any;
1031 		goto start_lookup;
1032 	}
1033 
1034 	if (first) {
1035 		if (udpv6_queue_rcv_skb(first, skb) > 0)
1036 			consume_skb(skb);
1037 	} else {
1038 		kfree_skb(skb);
1039 		__UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
1040 				 proto == IPPROTO_UDPLITE);
1041 	}
1042 	return 0;
1043 }
1044 
udp6_sk_rx_dst_set(struct sock * sk,struct dst_entry * dst)1045 static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
1046 {
1047 	if (udp_sk_rx_dst_set(sk, dst))
1048 		sk->sk_rx_dst_cookie = rt6_get_cookie(dst_rt6_info(dst));
1049 }
1050 
1051 /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and
1052  * return code conversion for ip layer consumption
1053  */
udp6_unicast_rcv_skb(struct sock * sk,struct sk_buff * skb,struct udphdr * uh)1054 static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb,
1055 				struct udphdr *uh)
1056 {
1057 	int ret;
1058 
1059 	if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
1060 		skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo);
1061 
1062 	ret = udpv6_queue_rcv_skb(sk, skb);
1063 
1064 	/* a return value > 0 means to resubmit the input */
1065 	if (ret > 0)
1066 		return ret;
1067 	return 0;
1068 }
1069 
__udp6_lib_rcv(struct sk_buff * skb,struct udp_table * udptable,int proto)1070 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
1071 		   int proto)
1072 {
1073 	enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED;
1074 	const struct in6_addr *saddr, *daddr;
1075 	struct net *net = dev_net(skb->dev);
1076 	struct sock *sk = NULL;
1077 	struct udphdr *uh;
1078 	bool refcounted;
1079 	u32 ulen = 0;
1080 
1081 	if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1082 		goto discard;
1083 
1084 	saddr = &ipv6_hdr(skb)->saddr;
1085 	daddr = &ipv6_hdr(skb)->daddr;
1086 	uh = udp_hdr(skb);
1087 
1088 	ulen = ntohs(uh->len);
1089 	if (ulen > skb->len)
1090 		goto short_packet;
1091 
1092 	if (proto == IPPROTO_UDP) {
1093 		/* UDP validates ulen. */
1094 
1095 		/* Check for jumbo payload */
1096 		if (ulen == 0)
1097 			ulen = skb->len;
1098 
1099 		if (ulen < sizeof(*uh))
1100 			goto short_packet;
1101 
1102 		if (ulen < skb->len) {
1103 			if (pskb_trim_rcsum(skb, ulen))
1104 				goto short_packet;
1105 			saddr = &ipv6_hdr(skb)->saddr;
1106 			daddr = &ipv6_hdr(skb)->daddr;
1107 			uh = udp_hdr(skb);
1108 		}
1109 	}
1110 
1111 	if (udp6_csum_init(skb, uh, proto))
1112 		goto csum_error;
1113 
1114 	/* Check if the socket is already available, e.g. due to early demux */
1115 	sk = inet6_steal_sock(net, skb, sizeof(struct udphdr), saddr, uh->source, daddr, uh->dest,
1116 			      &refcounted, udp6_ehashfn);
1117 	if (IS_ERR(sk))
1118 		goto no_sk;
1119 
1120 	if (sk) {
1121 		struct dst_entry *dst = skb_dst(skb);
1122 		int ret;
1123 
1124 		if (unlikely(rcu_dereference(sk->sk_rx_dst) != dst))
1125 			udp6_sk_rx_dst_set(sk, dst);
1126 
1127 		if (!uh->check && !udp_get_no_check6_rx(sk)) {
1128 			if (refcounted)
1129 				sock_put(sk);
1130 			goto report_csum_error;
1131 		}
1132 
1133 		ret = udp6_unicast_rcv_skb(sk, skb, uh);
1134 		if (refcounted)
1135 			sock_put(sk);
1136 		return ret;
1137 	}
1138 
1139 	/*
1140 	 *	Multicast receive code
1141 	 */
1142 	if (ipv6_addr_is_multicast(daddr))
1143 		return __udp6_lib_mcast_deliver(net, skb,
1144 				saddr, daddr, udptable, proto);
1145 
1146 	/* Unicast */
1147 	sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
1148 	if (sk) {
1149 		if (!uh->check && !udp_get_no_check6_rx(sk))
1150 			goto report_csum_error;
1151 		return udp6_unicast_rcv_skb(sk, skb, uh);
1152 	}
1153 no_sk:
1154 	reason = SKB_DROP_REASON_NO_SOCKET;
1155 
1156 	if (!uh->check)
1157 		goto report_csum_error;
1158 
1159 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1160 		goto discard;
1161 	nf_reset_ct(skb);
1162 
1163 	if (udp_lib_checksum_complete(skb))
1164 		goto csum_error;
1165 
1166 	__UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1167 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
1168 
1169 	sk_skb_reason_drop(sk, skb, reason);
1170 	return 0;
1171 
1172 short_packet:
1173 	if (reason == SKB_DROP_REASON_NOT_SPECIFIED)
1174 		reason = SKB_DROP_REASON_PKT_TOO_SMALL;
1175 	net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
1176 			    proto == IPPROTO_UDPLITE ? "-Lite" : "",
1177 			    saddr, ntohs(uh->source),
1178 			    ulen, skb->len,
1179 			    daddr, ntohs(uh->dest));
1180 	goto discard;
1181 
1182 report_csum_error:
1183 	udp6_csum_zero_error(skb);
1184 csum_error:
1185 	if (reason == SKB_DROP_REASON_NOT_SPECIFIED)
1186 		reason = SKB_DROP_REASON_UDP_CSUM;
1187 	__UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1188 discard:
1189 	__UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1190 	sk_skb_reason_drop(sk, skb, reason);
1191 	return 0;
1192 }
1193 
1194 
__udp6_lib_demux_lookup(struct net * net,__be16 loc_port,const struct in6_addr * loc_addr,__be16 rmt_port,const struct in6_addr * rmt_addr,int dif,int sdif)1195 static struct sock *__udp6_lib_demux_lookup(struct net *net,
1196 			__be16 loc_port, const struct in6_addr *loc_addr,
1197 			__be16 rmt_port, const struct in6_addr *rmt_addr,
1198 			int dif, int sdif)
1199 {
1200 	struct udp_table *udptable = net->ipv4.udp_table;
1201 	unsigned short hnum = ntohs(loc_port);
1202 	struct udp_hslot *hslot2;
1203 	unsigned int hash2;
1204 	__portpair ports;
1205 	struct sock *sk;
1206 
1207 	hash2 = ipv6_portaddr_hash(net, loc_addr, hnum);
1208 	hslot2 = udp_hashslot2(udptable, hash2);
1209 	ports = INET_COMBINED_PORTS(rmt_port, hnum);
1210 
1211 	udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
1212 		if (sk->sk_state == TCP_ESTABLISHED &&
1213 		    inet6_match(net, sk, rmt_addr, loc_addr, ports, dif, sdif))
1214 			return sk;
1215 		/* Only check first socket in chain */
1216 		break;
1217 	}
1218 	return NULL;
1219 }
1220 
udp_v6_early_demux(struct sk_buff * skb)1221 void udp_v6_early_demux(struct sk_buff *skb)
1222 {
1223 	struct net *net = dev_net(skb->dev);
1224 	const struct udphdr *uh;
1225 	struct sock *sk;
1226 	struct dst_entry *dst;
1227 	int dif = skb->dev->ifindex;
1228 	int sdif = inet6_sdif(skb);
1229 
1230 	if (!pskb_may_pull(skb, skb_transport_offset(skb) +
1231 	    sizeof(struct udphdr)))
1232 		return;
1233 
1234 	uh = udp_hdr(skb);
1235 
1236 	if (skb->pkt_type == PACKET_HOST)
1237 		sk = __udp6_lib_demux_lookup(net, uh->dest,
1238 					     &ipv6_hdr(skb)->daddr,
1239 					     uh->source, &ipv6_hdr(skb)->saddr,
1240 					     dif, sdif);
1241 	else
1242 		return;
1243 
1244 	if (!sk)
1245 		return;
1246 
1247 	skb->sk = sk;
1248 	DEBUG_NET_WARN_ON_ONCE(sk_is_refcounted(sk));
1249 	skb->destructor = sock_pfree;
1250 	dst = rcu_dereference(sk->sk_rx_dst);
1251 
1252 	if (dst)
1253 		dst = dst_check(dst, sk->sk_rx_dst_cookie);
1254 	if (dst) {
1255 		/* set noref for now.
1256 		 * any place which wants to hold dst has to call
1257 		 * dst_hold_safe()
1258 		 */
1259 		skb_dst_set_noref(skb, dst);
1260 	}
1261 }
1262 
udpv6_rcv(struct sk_buff * skb)1263 INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb)
1264 {
1265 	return __udp6_lib_rcv(skb, dev_net(skb->dev)->ipv4.udp_table, IPPROTO_UDP);
1266 }
1267 
1268 /*
1269  * Throw away all pending data and cancel the corking. Socket is locked.
1270  */
udp_v6_flush_pending_frames(struct sock * sk)1271 static void udp_v6_flush_pending_frames(struct sock *sk)
1272 {
1273 	struct udp_sock *up = udp_sk(sk);
1274 
1275 	if (up->pending == AF_INET)
1276 		udp_flush_pending_frames(sk);
1277 	else if (up->pending) {
1278 		up->len = 0;
1279 		WRITE_ONCE(up->pending, 0);
1280 		ip6_flush_pending_frames(sk);
1281 	}
1282 }
1283 
udpv6_pre_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)1284 static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr,
1285 			     int addr_len)
1286 {
1287 	if (addr_len < offsetofend(struct sockaddr, sa_family))
1288 		return -EINVAL;
1289 	/* The following checks are replicated from __ip6_datagram_connect()
1290 	 * and intended to prevent BPF program called below from accessing
1291 	 * bytes that are out of the bound specified by user in addr_len.
1292 	 */
1293 	if (uaddr->sa_family == AF_INET) {
1294 		if (ipv6_only_sock(sk))
1295 			return -EAFNOSUPPORT;
1296 		return udp_pre_connect(sk, uaddr, addr_len);
1297 	}
1298 
1299 	if (addr_len < SIN6_LEN_RFC2133)
1300 		return -EINVAL;
1301 
1302 	return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr, &addr_len);
1303 }
1304 
udpv6_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)1305 static int udpv6_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
1306 {
1307 	int res;
1308 
1309 	lock_sock(sk);
1310 	res = __ip6_datagram_connect(sk, uaddr, addr_len);
1311 	if (!res)
1312 		udp6_hash4(sk);
1313 	release_sock(sk);
1314 	return res;
1315 }
1316 
1317 /**
1318  *	udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
1319  *	@sk:	socket we are sending on
1320  *	@skb:	sk_buff containing the filled-in UDP header
1321  *		(checksum field must be zeroed out)
1322  *	@saddr: source address
1323  *	@daddr: destination address
1324  *	@len:	length of packet
1325  */
udp6_hwcsum_outgoing(struct sock * sk,struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int len)1326 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
1327 				 const struct in6_addr *saddr,
1328 				 const struct in6_addr *daddr, int len)
1329 {
1330 	unsigned int offset;
1331 	struct udphdr *uh = udp_hdr(skb);
1332 	struct sk_buff *frags = skb_shinfo(skb)->frag_list;
1333 	__wsum csum = 0;
1334 
1335 	if (!frags) {
1336 		/* Only one fragment on the socket.  */
1337 		skb->csum_start = skb_transport_header(skb) - skb->head;
1338 		skb->csum_offset = offsetof(struct udphdr, check);
1339 		uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1340 	} else {
1341 		/*
1342 		 * HW-checksum won't work as there are two or more
1343 		 * fragments on the socket so that all csums of sk_buffs
1344 		 * should be together
1345 		 */
1346 		offset = skb_transport_offset(skb);
1347 		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1348 		csum = skb->csum;
1349 
1350 		skb->ip_summed = CHECKSUM_NONE;
1351 
1352 		do {
1353 			csum = csum_add(csum, frags->csum);
1354 		} while ((frags = frags->next));
1355 
1356 		uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1357 					    csum);
1358 		if (uh->check == 0)
1359 			uh->check = CSUM_MANGLED_0;
1360 	}
1361 }
1362 
1363 /*
1364  *	Sending
1365  */
1366 
udp_v6_send_skb(struct sk_buff * skb,struct flowi6 * fl6,struct inet_cork * cork)1367 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6,
1368 			   struct inet_cork *cork)
1369 {
1370 	struct sock *sk = skb->sk;
1371 	struct udphdr *uh;
1372 	int err = 0;
1373 	int is_udplite = IS_UDPLITE(sk);
1374 	__wsum csum = 0;
1375 	int offset = skb_transport_offset(skb);
1376 	int len = skb->len - offset;
1377 	int datalen = len - sizeof(*uh);
1378 
1379 	/*
1380 	 * Create a UDP header
1381 	 */
1382 	uh = udp_hdr(skb);
1383 	uh->source = fl6->fl6_sport;
1384 	uh->dest = fl6->fl6_dport;
1385 	uh->len = htons(len);
1386 	uh->check = 0;
1387 
1388 	if (cork->gso_size) {
1389 		const int hlen = skb_network_header_len(skb) +
1390 				 sizeof(struct udphdr);
1391 
1392 		if (hlen + min(datalen, cork->gso_size) > cork->fragsize) {
1393 			kfree_skb(skb);
1394 			return -EMSGSIZE;
1395 		}
1396 		if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) {
1397 			kfree_skb(skb);
1398 			return -EINVAL;
1399 		}
1400 		if (udp_get_no_check6_tx(sk)) {
1401 			kfree_skb(skb);
1402 			return -EINVAL;
1403 		}
1404 		if (is_udplite || dst_xfrm(skb_dst(skb))) {
1405 			kfree_skb(skb);
1406 			return -EIO;
1407 		}
1408 
1409 		if (datalen > cork->gso_size) {
1410 			skb_shinfo(skb)->gso_size = cork->gso_size;
1411 			skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4;
1412 			skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(datalen,
1413 								 cork->gso_size);
1414 
1415 			/* Don't checksum the payload, skb will get segmented */
1416 			goto csum_partial;
1417 		}
1418 	}
1419 
1420 	if (is_udplite)
1421 		csum = udplite_csum(skb);
1422 	else if (udp_get_no_check6_tx(sk)) {   /* UDP csum disabled */
1423 		skb->ip_summed = CHECKSUM_NONE;
1424 		goto send;
1425 	} else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1426 csum_partial:
1427 		udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1428 		goto send;
1429 	} else
1430 		csum = udp_csum(skb);
1431 
1432 	/* add protocol-dependent pseudo-header */
1433 	uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1434 				    len, fl6->flowi6_proto, csum);
1435 	if (uh->check == 0)
1436 		uh->check = CSUM_MANGLED_0;
1437 
1438 send:
1439 	err = ip6_send_skb(skb);
1440 	if (err) {
1441 		if (err == -ENOBUFS && !inet6_test_bit(RECVERR6, sk)) {
1442 			UDP6_INC_STATS(sock_net(sk),
1443 				       UDP_MIB_SNDBUFERRORS, is_udplite);
1444 			err = 0;
1445 		}
1446 	} else {
1447 		UDP6_INC_STATS(sock_net(sk),
1448 			       UDP_MIB_OUTDATAGRAMS, is_udplite);
1449 	}
1450 	return err;
1451 }
1452 
udp_v6_push_pending_frames(struct sock * sk)1453 static int udp_v6_push_pending_frames(struct sock *sk)
1454 {
1455 	struct sk_buff *skb;
1456 	struct udp_sock  *up = udp_sk(sk);
1457 	int err = 0;
1458 
1459 	if (up->pending == AF_INET)
1460 		return udp_push_pending_frames(sk);
1461 
1462 	skb = ip6_finish_skb(sk);
1463 	if (!skb)
1464 		goto out;
1465 
1466 	err = udp_v6_send_skb(skb, &inet_sk(sk)->cork.fl.u.ip6,
1467 			      &inet_sk(sk)->cork.base);
1468 out:
1469 	up->len = 0;
1470 	WRITE_ONCE(up->pending, 0);
1471 	return err;
1472 }
1473 
udpv6_sendmsg(struct sock * sk,struct msghdr * msg,size_t len)1474 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1475 {
1476 	struct ipv6_txoptions opt_space;
1477 	struct udp_sock *up = udp_sk(sk);
1478 	struct inet_sock *inet = inet_sk(sk);
1479 	struct ipv6_pinfo *np = inet6_sk(sk);
1480 	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1481 	struct in6_addr *daddr, *final_p, final;
1482 	struct ipv6_txoptions *opt = NULL;
1483 	struct ipv6_txoptions *opt_to_free = NULL;
1484 	struct ip6_flowlabel *flowlabel = NULL;
1485 	struct inet_cork_full cork;
1486 	struct flowi6 *fl6 = &cork.fl.u.ip6;
1487 	struct dst_entry *dst;
1488 	struct ipcm6_cookie ipc6;
1489 	int addr_len = msg->msg_namelen;
1490 	bool connected = false;
1491 	int ulen = len;
1492 	int corkreq = udp_test_bit(CORK, sk) || msg->msg_flags & MSG_MORE;
1493 	int err;
1494 	int is_udplite = IS_UDPLITE(sk);
1495 	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1496 
1497 	ipcm6_init(&ipc6);
1498 	ipc6.gso_size = READ_ONCE(up->gso_size);
1499 	ipc6.sockc.tsflags = READ_ONCE(sk->sk_tsflags);
1500 	ipc6.sockc.mark = READ_ONCE(sk->sk_mark);
1501 	ipc6.sockc.priority = READ_ONCE(sk->sk_priority);
1502 
1503 	/* destination address check */
1504 	if (sin6) {
1505 		if (addr_len < offsetof(struct sockaddr, sa_data))
1506 			return -EINVAL;
1507 
1508 		switch (sin6->sin6_family) {
1509 		case AF_INET6:
1510 			if (addr_len < SIN6_LEN_RFC2133)
1511 				return -EINVAL;
1512 			daddr = &sin6->sin6_addr;
1513 			if (ipv6_addr_any(daddr) &&
1514 			    ipv6_addr_v4mapped(&np->saddr))
1515 				ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
1516 						       daddr);
1517 			break;
1518 		case AF_INET:
1519 			goto do_udp_sendmsg;
1520 		case AF_UNSPEC:
1521 			msg->msg_name = sin6 = NULL;
1522 			msg->msg_namelen = addr_len = 0;
1523 			daddr = NULL;
1524 			break;
1525 		default:
1526 			return -EINVAL;
1527 		}
1528 	} else if (!READ_ONCE(up->pending)) {
1529 		if (sk->sk_state != TCP_ESTABLISHED)
1530 			return -EDESTADDRREQ;
1531 		daddr = &sk->sk_v6_daddr;
1532 	} else
1533 		daddr = NULL;
1534 
1535 	if (daddr) {
1536 		if (ipv6_addr_v4mapped(daddr)) {
1537 			struct sockaddr_in sin;
1538 			sin.sin_family = AF_INET;
1539 			sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1540 			sin.sin_addr.s_addr = daddr->s6_addr32[3];
1541 			msg->msg_name = &sin;
1542 			msg->msg_namelen = sizeof(sin);
1543 do_udp_sendmsg:
1544 			err = ipv6_only_sock(sk) ?
1545 				-ENETUNREACH : udp_sendmsg(sk, msg, len);
1546 			msg->msg_name = sin6;
1547 			msg->msg_namelen = addr_len;
1548 			return err;
1549 		}
1550 	}
1551 
1552 	/* Rough check on arithmetic overflow,
1553 	   better check is made in ip6_append_data().
1554 	   */
1555 	if (len > INT_MAX - sizeof(struct udphdr))
1556 		return -EMSGSIZE;
1557 
1558 	getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1559 	if (READ_ONCE(up->pending)) {
1560 		if (READ_ONCE(up->pending) == AF_INET)
1561 			return udp_sendmsg(sk, msg, len);
1562 		/*
1563 		 * There are pending frames.
1564 		 * The socket lock must be held while it's corked.
1565 		 */
1566 		lock_sock(sk);
1567 		if (likely(up->pending)) {
1568 			if (unlikely(up->pending != AF_INET6)) {
1569 				release_sock(sk);
1570 				return -EAFNOSUPPORT;
1571 			}
1572 			dst = NULL;
1573 			goto do_append_data;
1574 		}
1575 		release_sock(sk);
1576 	}
1577 	ulen += sizeof(struct udphdr);
1578 
1579 	memset(fl6, 0, sizeof(*fl6));
1580 
1581 	if (sin6) {
1582 		if (sin6->sin6_port == 0)
1583 			return -EINVAL;
1584 
1585 		fl6->fl6_dport = sin6->sin6_port;
1586 		daddr = &sin6->sin6_addr;
1587 
1588 		if (inet6_test_bit(SNDFLOW, sk)) {
1589 			fl6->flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1590 			if (fl6->flowlabel & IPV6_FLOWLABEL_MASK) {
1591 				flowlabel = fl6_sock_lookup(sk, fl6->flowlabel);
1592 				if (IS_ERR(flowlabel))
1593 					return -EINVAL;
1594 			}
1595 		}
1596 
1597 		/*
1598 		 * Otherwise it will be difficult to maintain
1599 		 * sk->sk_dst_cache.
1600 		 */
1601 		if (sk->sk_state == TCP_ESTABLISHED &&
1602 		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1603 			daddr = &sk->sk_v6_daddr;
1604 
1605 		if (addr_len >= sizeof(struct sockaddr_in6) &&
1606 		    sin6->sin6_scope_id &&
1607 		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1608 			fl6->flowi6_oif = sin6->sin6_scope_id;
1609 	} else {
1610 		if (sk->sk_state != TCP_ESTABLISHED)
1611 			return -EDESTADDRREQ;
1612 
1613 		fl6->fl6_dport = inet->inet_dport;
1614 		daddr = &sk->sk_v6_daddr;
1615 		fl6->flowlabel = np->flow_label;
1616 		connected = true;
1617 	}
1618 
1619 	if (!fl6->flowi6_oif)
1620 		fl6->flowi6_oif = READ_ONCE(sk->sk_bound_dev_if);
1621 
1622 	if (!fl6->flowi6_oif)
1623 		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1624 
1625 	fl6->flowi6_uid = sk->sk_uid;
1626 
1627 	if (msg->msg_controllen) {
1628 		opt = &opt_space;
1629 		memset(opt, 0, sizeof(struct ipv6_txoptions));
1630 		opt->tot_len = sizeof(*opt);
1631 		ipc6.opt = opt;
1632 
1633 		err = udp_cmsg_send(sk, msg, &ipc6.gso_size);
1634 		if (err > 0) {
1635 			err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, fl6,
1636 						    &ipc6);
1637 			connected = false;
1638 		}
1639 		if (err < 0) {
1640 			fl6_sock_release(flowlabel);
1641 			return err;
1642 		}
1643 		if ((fl6->flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1644 			flowlabel = fl6_sock_lookup(sk, fl6->flowlabel);
1645 			if (IS_ERR(flowlabel))
1646 				return -EINVAL;
1647 		}
1648 		if (!(opt->opt_nflen|opt->opt_flen))
1649 			opt = NULL;
1650 	}
1651 	if (!opt) {
1652 		opt = txopt_get(np);
1653 		opt_to_free = opt;
1654 	}
1655 	if (flowlabel)
1656 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
1657 	opt = ipv6_fixup_options(&opt_space, opt);
1658 	ipc6.opt = opt;
1659 
1660 	fl6->flowi6_proto = sk->sk_protocol;
1661 	fl6->flowi6_mark = ipc6.sockc.mark;
1662 	fl6->daddr = *daddr;
1663 	if (ipv6_addr_any(&fl6->saddr) && !ipv6_addr_any(&np->saddr))
1664 		fl6->saddr = np->saddr;
1665 	fl6->fl6_sport = inet->inet_sport;
1666 
1667 	if (cgroup_bpf_enabled(CGROUP_UDP6_SENDMSG) && !connected) {
1668 		err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk,
1669 					   (struct sockaddr *)sin6,
1670 					   &addr_len,
1671 					   &fl6->saddr);
1672 		if (err)
1673 			goto out_no_dst;
1674 		if (sin6) {
1675 			if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
1676 				/* BPF program rewrote IPv6-only by IPv4-mapped
1677 				 * IPv6. It's currently unsupported.
1678 				 */
1679 				err = -ENOTSUPP;
1680 				goto out_no_dst;
1681 			}
1682 			if (sin6->sin6_port == 0) {
1683 				/* BPF program set invalid port. Reject it. */
1684 				err = -EINVAL;
1685 				goto out_no_dst;
1686 			}
1687 			fl6->fl6_dport = sin6->sin6_port;
1688 			fl6->daddr = sin6->sin6_addr;
1689 		}
1690 	}
1691 
1692 	if (ipv6_addr_any(&fl6->daddr))
1693 		fl6->daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1694 
1695 	final_p = fl6_update_dst(fl6, opt, &final);
1696 	if (final_p)
1697 		connected = false;
1698 
1699 	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr)) {
1700 		fl6->flowi6_oif = READ_ONCE(np->mcast_oif);
1701 		connected = false;
1702 	} else if (!fl6->flowi6_oif)
1703 		fl6->flowi6_oif = READ_ONCE(np->ucast_oif);
1704 
1705 	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
1706 
1707 	if (ipc6.tclass < 0)
1708 		ipc6.tclass = np->tclass;
1709 
1710 	fl6->flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6->flowlabel);
1711 
1712 	dst = ip6_sk_dst_lookup_flow(sk, fl6, final_p, connected);
1713 	if (IS_ERR(dst)) {
1714 		err = PTR_ERR(dst);
1715 		dst = NULL;
1716 		goto out;
1717 	}
1718 
1719 	if (ipc6.hlimit < 0)
1720 		ipc6.hlimit = ip6_sk_dst_hoplimit(np, fl6, dst);
1721 
1722 	if (msg->msg_flags&MSG_CONFIRM)
1723 		goto do_confirm;
1724 back_from_confirm:
1725 
1726 	/* Lockless fast path for the non-corking case */
1727 	if (!corkreq) {
1728 		struct sk_buff *skb;
1729 
1730 		skb = ip6_make_skb(sk, getfrag, msg, ulen,
1731 				   sizeof(struct udphdr), &ipc6,
1732 				   dst_rt6_info(dst),
1733 				   msg->msg_flags, &cork);
1734 		err = PTR_ERR(skb);
1735 		if (!IS_ERR_OR_NULL(skb))
1736 			err = udp_v6_send_skb(skb, fl6, &cork.base);
1737 		/* ip6_make_skb steals dst reference */
1738 		goto out_no_dst;
1739 	}
1740 
1741 	lock_sock(sk);
1742 	if (unlikely(up->pending)) {
1743 		/* The socket is already corked while preparing it. */
1744 		/* ... which is an evident application bug. --ANK */
1745 		release_sock(sk);
1746 
1747 		net_dbg_ratelimited("udp cork app bug 2\n");
1748 		err = -EINVAL;
1749 		goto out;
1750 	}
1751 
1752 	WRITE_ONCE(up->pending, AF_INET6);
1753 
1754 do_append_data:
1755 	if (ipc6.dontfrag < 0)
1756 		ipc6.dontfrag = inet6_test_bit(DONTFRAG, sk);
1757 	up->len += ulen;
1758 	err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1759 			      &ipc6, fl6, dst_rt6_info(dst),
1760 			      corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
1761 	if (err)
1762 		udp_v6_flush_pending_frames(sk);
1763 	else if (!corkreq)
1764 		err = udp_v6_push_pending_frames(sk);
1765 	else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1766 		WRITE_ONCE(up->pending, 0);
1767 
1768 	if (err > 0)
1769 		err = inet6_test_bit(RECVERR6, sk) ? net_xmit_errno(err) : 0;
1770 	release_sock(sk);
1771 
1772 out:
1773 	dst_release(dst);
1774 out_no_dst:
1775 	fl6_sock_release(flowlabel);
1776 	txopt_put(opt_to_free);
1777 	if (!err)
1778 		return len;
1779 	/*
1780 	 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1781 	 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1782 	 * we don't have a good statistic (IpOutDiscards but it can be too many
1783 	 * things).  We could add another new stat but at least for now that
1784 	 * seems like overkill.
1785 	 */
1786 	if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1787 		UDP6_INC_STATS(sock_net(sk),
1788 			       UDP_MIB_SNDBUFERRORS, is_udplite);
1789 	}
1790 	return err;
1791 
1792 do_confirm:
1793 	if (msg->msg_flags & MSG_PROBE)
1794 		dst_confirm_neigh(dst, &fl6->daddr);
1795 	if (!(msg->msg_flags&MSG_PROBE) || len)
1796 		goto back_from_confirm;
1797 	err = 0;
1798 	goto out;
1799 }
1800 EXPORT_SYMBOL(udpv6_sendmsg);
1801 
udpv6_splice_eof(struct socket * sock)1802 static void udpv6_splice_eof(struct socket *sock)
1803 {
1804 	struct sock *sk = sock->sk;
1805 	struct udp_sock *up = udp_sk(sk);
1806 
1807 	if (!READ_ONCE(up->pending) || udp_test_bit(CORK, sk))
1808 		return;
1809 
1810 	lock_sock(sk);
1811 	if (up->pending && !udp_test_bit(CORK, sk))
1812 		udp_v6_push_pending_frames(sk);
1813 	release_sock(sk);
1814 }
1815 
udpv6_destroy_sock(struct sock * sk)1816 void udpv6_destroy_sock(struct sock *sk)
1817 {
1818 	struct udp_sock *up = udp_sk(sk);
1819 	lock_sock(sk);
1820 
1821 	/* protects from races with udp_abort() */
1822 	sock_set_flag(sk, SOCK_DEAD);
1823 	udp_v6_flush_pending_frames(sk);
1824 	release_sock(sk);
1825 
1826 	if (static_branch_unlikely(&udpv6_encap_needed_key)) {
1827 		if (up->encap_type) {
1828 			void (*encap_destroy)(struct sock *sk);
1829 			encap_destroy = READ_ONCE(up->encap_destroy);
1830 			if (encap_destroy)
1831 				encap_destroy(sk);
1832 		}
1833 		if (udp_test_bit(ENCAP_ENABLED, sk)) {
1834 			static_branch_dec(&udpv6_encap_needed_key);
1835 			udp_encap_disable();
1836 		}
1837 	}
1838 }
1839 
1840 /*
1841  *	Socket option code for UDP
1842  */
udpv6_setsockopt(struct sock * sk,int level,int optname,sockptr_t optval,unsigned int optlen)1843 int udpv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1844 		     unsigned int optlen)
1845 {
1846 	if (level == SOL_UDP  ||  level == SOL_UDPLITE || level == SOL_SOCKET)
1847 		return udp_lib_setsockopt(sk, level, optname,
1848 					  optval, optlen,
1849 					  udp_v6_push_pending_frames);
1850 	return ipv6_setsockopt(sk, level, optname, optval, optlen);
1851 }
1852 
udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1853 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1854 		     char __user *optval, int __user *optlen)
1855 {
1856 	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1857 		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1858 	return ipv6_getsockopt(sk, level, optname, optval, optlen);
1859 }
1860 
1861 
1862 /* ------------------------------------------------------------------------ */
1863 #ifdef CONFIG_PROC_FS
udp6_seq_show(struct seq_file * seq,void * v)1864 int udp6_seq_show(struct seq_file *seq, void *v)
1865 {
1866 	if (v == SEQ_START_TOKEN) {
1867 		seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1868 	} else {
1869 		int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1870 		const struct inet_sock *inet = inet_sk((const struct sock *)v);
1871 		__u16 srcp = ntohs(inet->inet_sport);
1872 		__u16 destp = ntohs(inet->inet_dport);
1873 		__ip6_dgram_sock_seq_show(seq, v, srcp, destp,
1874 					  udp_rqueue_get(v), bucket);
1875 	}
1876 	return 0;
1877 }
1878 
1879 const struct seq_operations udp6_seq_ops = {
1880 	.start		= udp_seq_start,
1881 	.next		= udp_seq_next,
1882 	.stop		= udp_seq_stop,
1883 	.show		= udp6_seq_show,
1884 };
1885 EXPORT_SYMBOL(udp6_seq_ops);
1886 
1887 static struct udp_seq_afinfo udp6_seq_afinfo = {
1888 	.family		= AF_INET6,
1889 	.udp_table	= NULL,
1890 };
1891 
udp6_proc_init(struct net * net)1892 int __net_init udp6_proc_init(struct net *net)
1893 {
1894 	if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops,
1895 			sizeof(struct udp_iter_state), &udp6_seq_afinfo))
1896 		return -ENOMEM;
1897 	return 0;
1898 }
1899 
udp6_proc_exit(struct net * net)1900 void udp6_proc_exit(struct net *net)
1901 {
1902 	remove_proc_entry("udp6", net->proc_net);
1903 }
1904 #endif /* CONFIG_PROC_FS */
1905 
1906 /* ------------------------------------------------------------------------ */
1907 
1908 struct proto udpv6_prot = {
1909 	.name			= "UDPv6",
1910 	.owner			= THIS_MODULE,
1911 	.close			= udp_lib_close,
1912 	.pre_connect		= udpv6_pre_connect,
1913 	.connect		= udpv6_connect,
1914 	.disconnect		= udp_disconnect,
1915 	.ioctl			= udp_ioctl,
1916 	.init			= udpv6_init_sock,
1917 	.destroy		= udpv6_destroy_sock,
1918 	.setsockopt		= udpv6_setsockopt,
1919 	.getsockopt		= udpv6_getsockopt,
1920 	.sendmsg		= udpv6_sendmsg,
1921 	.recvmsg		= udpv6_recvmsg,
1922 	.splice_eof		= udpv6_splice_eof,
1923 	.release_cb		= ip6_datagram_release_cb,
1924 	.hash			= udp_lib_hash,
1925 	.unhash			= udp_lib_unhash,
1926 	.rehash			= udp_v6_rehash,
1927 	.get_port		= udp_v6_get_port,
1928 	.put_port		= udp_lib_unhash,
1929 #ifdef CONFIG_BPF_SYSCALL
1930 	.psock_update_sk_prot	= udp_bpf_update_proto,
1931 #endif
1932 
1933 	.memory_allocated	= &udp_memory_allocated,
1934 	.per_cpu_fw_alloc	= &udp_memory_per_cpu_fw_alloc,
1935 
1936 	.sysctl_mem		= sysctl_udp_mem,
1937 	.sysctl_wmem_offset     = offsetof(struct net, ipv4.sysctl_udp_wmem_min),
1938 	.sysctl_rmem_offset     = offsetof(struct net, ipv4.sysctl_udp_rmem_min),
1939 	.obj_size		= sizeof(struct udp6_sock),
1940 	.ipv6_pinfo_offset = offsetof(struct udp6_sock, inet6),
1941 	.h.udp_table		= NULL,
1942 	.diag_destroy		= udp_abort,
1943 };
1944 
1945 static struct inet_protosw udpv6_protosw = {
1946 	.type =      SOCK_DGRAM,
1947 	.protocol =  IPPROTO_UDP,
1948 	.prot =      &udpv6_prot,
1949 	.ops =       &inet6_dgram_ops,
1950 	.flags =     INET_PROTOSW_PERMANENT,
1951 };
1952 
udpv6_init(void)1953 int __init udpv6_init(void)
1954 {
1955 	int ret;
1956 
1957 	net_hotdata.udpv6_protocol = (struct inet6_protocol) {
1958 		.handler     = udpv6_rcv,
1959 		.err_handler = udpv6_err,
1960 		.flags	     = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1961 	};
1962 	ret = inet6_add_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP);
1963 	if (ret)
1964 		goto out;
1965 
1966 	ret = inet6_register_protosw(&udpv6_protosw);
1967 	if (ret)
1968 		goto out_udpv6_protocol;
1969 out:
1970 	return ret;
1971 
1972 out_udpv6_protocol:
1973 	inet6_del_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP);
1974 	goto out;
1975 }
1976 
udpv6_exit(void)1977 void udpv6_exit(void)
1978 {
1979 	inet6_unregister_protosw(&udpv6_protosw);
1980 	inet6_del_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP);
1981 }
1982