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 <linux/uaccess.h>
39
40 #include <net/addrconf.h>
41 #include <net/ndisc.h>
42 #include <net/protocol.h>
43 #include <net/transp_v6.h>
44 #include <net/ip6_route.h>
45 #include <net/raw.h>
46 #include <net/tcp_states.h>
47 #include <net/ip6_checksum.h>
48 #include <net/xfrm.h>
49 #include <net/inet_hashtables.h>
50 #include <net/inet6_hashtables.h>
51 #include <net/busy_poll.h>
52 #include <net/sock_reuseport.h>
53
54 #include <linux/proc_fs.h>
55 #include <linux/seq_file.h>
56 #include <trace/events/skb.h>
57 #include "udp_impl.h"
58
udp6_lib_exact_dif_match(struct net * net,struct sk_buff * skb)59 static bool udp6_lib_exact_dif_match(struct net *net, struct sk_buff *skb)
60 {
61 #if defined(CONFIG_NET_L3_MASTER_DEV)
62 if (!net->ipv4.sysctl_udp_l3mdev_accept &&
63 skb && ipv6_l3mdev_skb(IP6CB(skb)->flags))
64 return true;
65 #endif
66 return false;
67 }
68
udp6_ehashfn(const struct net * net,const struct in6_addr * laddr,const u16 lport,const struct in6_addr * faddr,const __be16 fport)69 static u32 udp6_ehashfn(const struct net *net,
70 const struct in6_addr *laddr,
71 const u16 lport,
72 const struct in6_addr *faddr,
73 const __be16 fport)
74 {
75 static u32 udp6_ehash_secret __read_mostly;
76 static u32 udp_ipv6_hash_secret __read_mostly;
77
78 u32 lhash, fhash;
79
80 net_get_random_once(&udp6_ehash_secret,
81 sizeof(udp6_ehash_secret));
82 net_get_random_once(&udp_ipv6_hash_secret,
83 sizeof(udp_ipv6_hash_secret));
84
85 lhash = (__force u32)laddr->s6_addr32[3];
86 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
87
88 return __inet6_ehashfn(lhash, lport, fhash, fport,
89 udp_ipv6_hash_secret + net_hash_mix(net));
90 }
91
udp6_portaddr_hash(const struct net * net,const struct in6_addr * addr6,unsigned int port)92 static u32 udp6_portaddr_hash(const struct net *net,
93 const struct in6_addr *addr6,
94 unsigned int port)
95 {
96 unsigned int hash, mix = net_hash_mix(net);
97
98 if (ipv6_addr_any(addr6))
99 hash = jhash_1word(0, mix);
100 else if (ipv6_addr_v4mapped(addr6))
101 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
102 else
103 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
104
105 return hash ^ port;
106 }
107
udp_v6_get_port(struct sock * sk,unsigned short snum)108 int udp_v6_get_port(struct sock *sk, unsigned short snum)
109 {
110 unsigned int hash2_nulladdr =
111 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
112 unsigned int hash2_partial =
113 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
114
115 /* precompute partial secondary hash */
116 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
117 return udp_lib_get_port(sk, snum, hash2_nulladdr);
118 }
119
udp_v6_rehash(struct sock * sk)120 static void udp_v6_rehash(struct sock *sk)
121 {
122 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
123 &sk->sk_v6_rcv_saddr,
124 inet_sk(sk)->inet_num);
125
126 udp_lib_rehash(sk, new_hash);
127 }
128
compute_score(struct sock * sk,struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned short hnum,int dif,int sdif,bool exact_dif)129 static int compute_score(struct sock *sk, struct net *net,
130 const struct in6_addr *saddr, __be16 sport,
131 const struct in6_addr *daddr, unsigned short hnum,
132 int dif, int sdif, bool exact_dif)
133 {
134 int score;
135 struct inet_sock *inet;
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 score = 0;
143 inet = inet_sk(sk);
144
145 if (inet->inet_dport) {
146 if (inet->inet_dport != sport)
147 return -1;
148 score++;
149 }
150
151 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
152 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
153 return -1;
154 score++;
155 }
156
157 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
158 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
159 return -1;
160 score++;
161 }
162
163 if (sk->sk_bound_dev_if || exact_dif) {
164 bool dev_match = (sk->sk_bound_dev_if == dif ||
165 sk->sk_bound_dev_if == sdif);
166
167 if (!dev_match)
168 return -1;
169 if (sk->sk_bound_dev_if)
170 score++;
171 }
172
173 if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id())
174 score++;
175
176 return score;
177 }
178
179 /* called with rcu_read_lock() */
udp6_lib_lookup2(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,int sdif,bool exact_dif,struct udp_hslot * hslot2,struct sk_buff * skb)180 static struct sock *udp6_lib_lookup2(struct net *net,
181 const struct in6_addr *saddr, __be16 sport,
182 const struct in6_addr *daddr, unsigned int hnum,
183 int dif, int sdif, bool exact_dif,
184 struct udp_hslot *hslot2, struct sk_buff *skb)
185 {
186 struct sock *sk, *result;
187 int score, badness, matches = 0, reuseport = 0;
188 u32 hash = 0;
189
190 result = NULL;
191 badness = -1;
192 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
193 score = compute_score(sk, net, saddr, sport,
194 daddr, hnum, dif, sdif, exact_dif);
195 if (score > badness) {
196 reuseport = sk->sk_reuseport;
197 if (reuseport) {
198 hash = udp6_ehashfn(net, daddr, hnum,
199 saddr, sport);
200
201 result = reuseport_select_sock(sk, hash, skb,
202 sizeof(struct udphdr));
203 if (result)
204 return result;
205 matches = 1;
206 }
207 result = sk;
208 badness = score;
209 } else if (score == badness && reuseport) {
210 matches++;
211 if (reciprocal_scale(hash, matches) == 0)
212 result = sk;
213 hash = next_pseudo_random32(hash);
214 }
215 }
216 return result;
217 }
218
219 /* rcu_read_lock() must be held */
__udp6_lib_lookup(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)220 struct sock *__udp6_lib_lookup(struct net *net,
221 const struct in6_addr *saddr, __be16 sport,
222 const struct in6_addr *daddr, __be16 dport,
223 int dif, int sdif, struct udp_table *udptable,
224 struct sk_buff *skb)
225 {
226 struct sock *sk, *result;
227 unsigned short hnum = ntohs(dport);
228 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
229 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
230 bool exact_dif = udp6_lib_exact_dif_match(net, skb);
231 int score, badness, matches = 0, reuseport = 0;
232 u32 hash = 0;
233
234 if (hslot->count > 10) {
235 hash2 = udp6_portaddr_hash(net, daddr, hnum);
236 slot2 = hash2 & udptable->mask;
237 hslot2 = &udptable->hash2[slot2];
238 if (hslot->count < hslot2->count)
239 goto begin;
240
241 result = udp6_lib_lookup2(net, saddr, sport,
242 daddr, hnum, dif, sdif, exact_dif,
243 hslot2, skb);
244 if (!result) {
245 unsigned int old_slot2 = slot2;
246 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
247 slot2 = hash2 & udptable->mask;
248 /* avoid searching the same slot again. */
249 if (unlikely(slot2 == old_slot2))
250 return result;
251
252 hslot2 = &udptable->hash2[slot2];
253 if (hslot->count < hslot2->count)
254 goto begin;
255
256 result = udp6_lib_lookup2(net, saddr, sport,
257 daddr, hnum, dif, sdif,
258 exact_dif, hslot2,
259 skb);
260 }
261 return result;
262 }
263 begin:
264 result = NULL;
265 badness = -1;
266 sk_for_each_rcu(sk, &hslot->head) {
267 score = compute_score(sk, net, saddr, sport, daddr, hnum, dif,
268 sdif, exact_dif);
269 if (score > badness) {
270 reuseport = sk->sk_reuseport;
271 if (reuseport) {
272 hash = udp6_ehashfn(net, daddr, hnum,
273 saddr, sport);
274 result = reuseport_select_sock(sk, hash, skb,
275 sizeof(struct udphdr));
276 if (result)
277 return result;
278 matches = 1;
279 }
280 result = sk;
281 badness = score;
282 } else if (score == badness && reuseport) {
283 matches++;
284 if (reciprocal_scale(hash, matches) == 0)
285 result = sk;
286 hash = next_pseudo_random32(hash);
287 }
288 }
289 return result;
290 }
291 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
292
__udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport,struct udp_table * udptable)293 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
294 __be16 sport, __be16 dport,
295 struct udp_table *udptable)
296 {
297 const struct ipv6hdr *iph = ipv6_hdr(skb);
298
299 return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
300 &iph->daddr, dport, inet6_iif(skb),
301 inet6_sdif(skb), udptable, skb);
302 }
303
udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport)304 struct sock *udp6_lib_lookup_skb(struct sk_buff *skb,
305 __be16 sport, __be16 dport)
306 {
307 const struct ipv6hdr *iph = ipv6_hdr(skb);
308
309 return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
310 &iph->daddr, dport, inet6_iif(skb),
311 inet6_sdif(skb), &udp_table, NULL);
312 }
313 EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb);
314
315 /* Must be called under rcu_read_lock().
316 * Does increment socket refcount.
317 */
318 #if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
319 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY) || \
320 IS_ENABLED(CONFIG_NF_SOCKET_IPV6)
udp6_lib_lookup(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)321 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
322 const struct in6_addr *daddr, __be16 dport, int dif)
323 {
324 struct sock *sk;
325
326 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport,
327 dif, 0, &udp_table, NULL);
328 if (sk && !refcount_inc_not_zero(&sk->sk_refcnt))
329 sk = NULL;
330 return sk;
331 }
332 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
333 #endif
334
335 /* do not use the scratch area len for jumbogram: their length execeeds the
336 * scratch area space; note that the IP6CB flags is still in the first
337 * cacheline, so checking for jumbograms is cheap
338 */
udp6_skb_len(struct sk_buff * skb)339 static int udp6_skb_len(struct sk_buff *skb)
340 {
341 return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb);
342 }
343
344 /*
345 * This should be easy, if there is something there we
346 * return it, otherwise we block.
347 */
348
udpv6_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int noblock,int flags,int * addr_len)349 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
350 int noblock, int flags, int *addr_len)
351 {
352 struct ipv6_pinfo *np = inet6_sk(sk);
353 struct inet_sock *inet = inet_sk(sk);
354 struct sk_buff *skb;
355 unsigned int ulen, copied;
356 int peeked, peeking, off;
357 int err;
358 int is_udplite = IS_UDPLITE(sk);
359 bool checksum_valid = false;
360 int is_udp4;
361
362 if (flags & MSG_ERRQUEUE)
363 return ipv6_recv_error(sk, msg, len, addr_len);
364
365 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
366 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
367
368 try_again:
369 peeking = flags & MSG_PEEK;
370 off = sk_peek_offset(sk, flags);
371 skb = __skb_recv_udp(sk, flags, noblock, &peeked, &off, &err);
372 if (!skb)
373 return err;
374
375 ulen = udp6_skb_len(skb);
376 copied = len;
377 if (copied > ulen - off)
378 copied = ulen - off;
379 else if (copied < ulen)
380 msg->msg_flags |= MSG_TRUNC;
381
382 is_udp4 = (skb->protocol == htons(ETH_P_IP));
383
384 /*
385 * If checksum is needed at all, try to do it while copying the
386 * data. If the data is truncated, or if we only want a partial
387 * coverage checksum (UDP-Lite), do it before the copy.
388 */
389
390 if (copied < ulen || peeking ||
391 (is_udplite && UDP_SKB_CB(skb)->partial_cov)) {
392 checksum_valid = udp_skb_csum_unnecessary(skb) ||
393 !__udp_lib_checksum_complete(skb);
394 if (!checksum_valid)
395 goto csum_copy_err;
396 }
397
398 if (checksum_valid || udp_skb_csum_unnecessary(skb)) {
399 if (udp_skb_is_linear(skb))
400 err = copy_linear_skb(skb, copied, off, &msg->msg_iter);
401 else
402 err = skb_copy_datagram_msg(skb, off, msg, copied);
403 } else {
404 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
405 if (err == -EINVAL)
406 goto csum_copy_err;
407 }
408 if (unlikely(err)) {
409 if (!peeked) {
410 atomic_inc(&sk->sk_drops);
411 if (is_udp4)
412 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
413 is_udplite);
414 else
415 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
416 is_udplite);
417 }
418 kfree_skb(skb);
419 return err;
420 }
421 if (!peeked) {
422 if (is_udp4)
423 UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
424 is_udplite);
425 else
426 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
427 is_udplite);
428 }
429
430 sock_recv_ts_and_drops(msg, sk, skb);
431
432 /* Copy the address. */
433 if (msg->msg_name) {
434 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
435 sin6->sin6_family = AF_INET6;
436 sin6->sin6_port = udp_hdr(skb)->source;
437 sin6->sin6_flowinfo = 0;
438
439 if (is_udp4) {
440 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
441 &sin6->sin6_addr);
442 sin6->sin6_scope_id = 0;
443 } else {
444 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
445 sin6->sin6_scope_id =
446 ipv6_iface_scope_id(&sin6->sin6_addr,
447 inet6_iif(skb));
448 }
449 *addr_len = sizeof(*sin6);
450 }
451
452 if (np->rxopt.all)
453 ip6_datagram_recv_common_ctl(sk, msg, skb);
454
455 if (is_udp4) {
456 if (inet->cmsg_flags)
457 ip_cmsg_recv_offset(msg, sk, skb,
458 sizeof(struct udphdr), off);
459 } else {
460 if (np->rxopt.all)
461 ip6_datagram_recv_specific_ctl(sk, msg, skb);
462 }
463
464 err = copied;
465 if (flags & MSG_TRUNC)
466 err = ulen;
467
468 skb_consume_udp(sk, skb, peeking ? -err : err);
469 return err;
470
471 csum_copy_err:
472 if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags,
473 udp_skb_destructor)) {
474 if (is_udp4) {
475 UDP_INC_STATS(sock_net(sk),
476 UDP_MIB_CSUMERRORS, is_udplite);
477 UDP_INC_STATS(sock_net(sk),
478 UDP_MIB_INERRORS, is_udplite);
479 } else {
480 UDP6_INC_STATS(sock_net(sk),
481 UDP_MIB_CSUMERRORS, is_udplite);
482 UDP6_INC_STATS(sock_net(sk),
483 UDP_MIB_INERRORS, is_udplite);
484 }
485 }
486 kfree_skb(skb);
487
488 /* starting over for a new packet, but check if we need to yield */
489 cond_resched();
490 msg->msg_flags &= ~MSG_TRUNC;
491 goto try_again;
492 }
493
__udp6_lib_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info,struct udp_table * udptable)494 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
495 u8 type, u8 code, int offset, __be32 info,
496 struct udp_table *udptable)
497 {
498 struct ipv6_pinfo *np;
499 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
500 const struct in6_addr *saddr = &hdr->saddr;
501 const struct in6_addr *daddr = &hdr->daddr;
502 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
503 struct sock *sk;
504 int harderr;
505 int err;
506 struct net *net = dev_net(skb->dev);
507
508 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
509 inet6_iif(skb), 0, udptable, NULL);
510 if (!sk) {
511 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
512 ICMP6_MIB_INERRORS);
513 return;
514 }
515
516 harderr = icmpv6_err_convert(type, code, &err);
517 np = inet6_sk(sk);
518
519 if (type == ICMPV6_PKT_TOOBIG) {
520 if (!ip6_sk_accept_pmtu(sk))
521 goto out;
522 ip6_sk_update_pmtu(skb, sk, info);
523 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
524 harderr = 1;
525 }
526 if (type == NDISC_REDIRECT) {
527 ip6_sk_redirect(skb, sk);
528 goto out;
529 }
530
531 if (!np->recverr) {
532 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
533 goto out;
534 } else {
535 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
536 }
537
538 sk->sk_err = err;
539 sk->sk_error_report(sk);
540 out:
541 return;
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(&sk->sk_v6_daddr)) {
549 sock_rps_save_rxhash(sk, skb);
550 sk_mark_napi_id(sk, skb);
551 sk_incoming_cpu_update(sk);
552 } else {
553 sk_mark_napi_id_once(sk, skb);
554 }
555
556 rc = __udp_enqueue_schedule_skb(sk, skb);
557 if (rc < 0) {
558 int is_udplite = IS_UDPLITE(sk);
559
560 /* Note that an ENOMEM error is charged twice */
561 if (rc == -ENOMEM)
562 UDP6_INC_STATS(sock_net(sk),
563 UDP_MIB_RCVBUFERRORS, is_udplite);
564 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
565 kfree_skb(skb);
566 return -1;
567 }
568
569 return 0;
570 }
571
udpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)572 static __inline__ void udpv6_err(struct sk_buff *skb,
573 struct inet6_skb_parm *opt, u8 type,
574 u8 code, int offset, __be32 info)
575 {
576 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
577 }
578
579 static struct static_key udpv6_encap_needed __read_mostly;
udpv6_encap_enable(void)580 void udpv6_encap_enable(void)
581 {
582 static_key_enable(&udpv6_encap_needed);
583 }
584 EXPORT_SYMBOL(udpv6_encap_enable);
585
udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)586 static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
587 {
588 struct udp_sock *up = udp_sk(sk);
589 int is_udplite = IS_UDPLITE(sk);
590
591 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
592 goto drop;
593
594 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
595 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
596
597 /*
598 * This is an encapsulation socket so pass the skb to
599 * the socket's udp_encap_rcv() hook. Otherwise, just
600 * fall through and pass this up the UDP socket.
601 * up->encap_rcv() returns the following value:
602 * =0 if skb was successfully passed to the encap
603 * handler or was discarded by it.
604 * >0 if skb should be passed on to UDP.
605 * <0 if skb should be resubmitted as proto -N
606 */
607
608 /* if we're overly short, let UDP handle it */
609 encap_rcv = ACCESS_ONCE(up->encap_rcv);
610 if (encap_rcv) {
611 int ret;
612
613 /* Verify checksum before giving to encap */
614 if (udp_lib_checksum_complete(skb))
615 goto csum_error;
616
617 ret = encap_rcv(sk, skb);
618 if (ret <= 0) {
619 __UDP_INC_STATS(sock_net(sk),
620 UDP_MIB_INDATAGRAMS,
621 is_udplite);
622 return -ret;
623 }
624 }
625
626 /* FALLTHROUGH -- it's a UDP Packet */
627 }
628
629 /*
630 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
631 */
632 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
633
634 if (up->pcrlen == 0) { /* full coverage was set */
635 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
636 UDP_SKB_CB(skb)->cscov, skb->len);
637 goto drop;
638 }
639 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
640 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
641 UDP_SKB_CB(skb)->cscov, up->pcrlen);
642 goto drop;
643 }
644 }
645
646 prefetch(&sk->sk_rmem_alloc);
647 if (rcu_access_pointer(sk->sk_filter) &&
648 udp_lib_checksum_complete(skb))
649 goto csum_error;
650
651 if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
652 goto drop;
653
654 udp_csum_pull_header(skb);
655
656 skb_dst_drop(skb);
657
658 return __udpv6_queue_rcv_skb(sk, skb);
659
660 csum_error:
661 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
662 drop:
663 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
664 atomic_inc(&sk->sk_drops);
665 kfree_skb(skb);
666 return -1;
667 }
668
__udp_v6_is_mcast_sock(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,unsigned short hnum)669 static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
670 __be16 loc_port, const struct in6_addr *loc_addr,
671 __be16 rmt_port, const struct in6_addr *rmt_addr,
672 int dif, unsigned short hnum)
673 {
674 struct inet_sock *inet = inet_sk(sk);
675
676 if (!net_eq(sock_net(sk), net))
677 return false;
678
679 if (udp_sk(sk)->udp_port_hash != hnum ||
680 sk->sk_family != PF_INET6 ||
681 (inet->inet_dport && inet->inet_dport != rmt_port) ||
682 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
683 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
684 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
685 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
686 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
687 return false;
688 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
689 return false;
690 return true;
691 }
692
udp6_csum_zero_error(struct sk_buff * skb)693 static void udp6_csum_zero_error(struct sk_buff *skb)
694 {
695 /* RFC 2460 section 8.1 says that we SHOULD log
696 * this error. Well, it is reasonable.
697 */
698 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
699 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
700 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
701 }
702
703 /*
704 * Note: called only from the BH handler context,
705 * so we don't need to lock the hashes.
706 */
__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)707 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
708 const struct in6_addr *saddr, const struct in6_addr *daddr,
709 struct udp_table *udptable, int proto)
710 {
711 struct sock *sk, *first = NULL;
712 const struct udphdr *uh = udp_hdr(skb);
713 unsigned short hnum = ntohs(uh->dest);
714 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
715 unsigned int offset = offsetof(typeof(*sk), sk_node);
716 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
717 int dif = inet6_iif(skb);
718 struct hlist_node *node;
719 struct sk_buff *nskb;
720
721 if (use_hash2) {
722 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
723 udptable->mask;
724 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udptable->mask;
725 start_lookup:
726 hslot = &udptable->hash2[hash2];
727 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
728 }
729
730 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
731 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
732 uh->source, saddr, dif, hnum))
733 continue;
734 /* If zero checksum and no_check is not on for
735 * the socket then skip it.
736 */
737 if (!uh->check && !udp_sk(sk)->no_check6_rx)
738 continue;
739 if (!first) {
740 first = sk;
741 continue;
742 }
743 nskb = skb_clone(skb, GFP_ATOMIC);
744 if (unlikely(!nskb)) {
745 atomic_inc(&sk->sk_drops);
746 __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
747 IS_UDPLITE(sk));
748 __UDP6_INC_STATS(net, UDP_MIB_INERRORS,
749 IS_UDPLITE(sk));
750 continue;
751 }
752
753 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
754 consume_skb(nskb);
755 }
756
757 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
758 if (use_hash2 && hash2 != hash2_any) {
759 hash2 = hash2_any;
760 goto start_lookup;
761 }
762
763 if (first) {
764 if (udpv6_queue_rcv_skb(first, skb) > 0)
765 consume_skb(skb);
766 } else {
767 kfree_skb(skb);
768 __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
769 proto == IPPROTO_UDPLITE);
770 }
771 return 0;
772 }
773
udp6_sk_rx_dst_set(struct sock * sk,struct dst_entry * dst)774 static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
775 {
776 if (udp_sk_rx_dst_set(sk, dst)) {
777 const struct rt6_info *rt = (const struct rt6_info *)dst;
778
779 inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
780 }
781 }
782
783 /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and
784 * return code conversion for ip layer consumption
785 */
udp6_unicast_rcv_skb(struct sock * sk,struct sk_buff * skb,struct udphdr * uh)786 static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb,
787 struct udphdr *uh)
788 {
789 int ret;
790
791 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
792 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
793 ip6_compute_pseudo);
794
795 ret = udpv6_queue_rcv_skb(sk, skb);
796
797 /* a return value > 0 means to resubmit the input */
798 if (ret > 0)
799 return ret;
800 return 0;
801 }
802
__udp6_lib_rcv(struct sk_buff * skb,struct udp_table * udptable,int proto)803 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
804 int proto)
805 {
806 const struct in6_addr *saddr, *daddr;
807 struct net *net = dev_net(skb->dev);
808 struct udphdr *uh;
809 struct sock *sk;
810 u32 ulen = 0;
811
812 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
813 goto discard;
814
815 saddr = &ipv6_hdr(skb)->saddr;
816 daddr = &ipv6_hdr(skb)->daddr;
817 uh = udp_hdr(skb);
818
819 ulen = ntohs(uh->len);
820 if (ulen > skb->len)
821 goto short_packet;
822
823 if (proto == IPPROTO_UDP) {
824 /* UDP validates ulen. */
825
826 /* Check for jumbo payload */
827 if (ulen == 0)
828 ulen = skb->len;
829
830 if (ulen < sizeof(*uh))
831 goto short_packet;
832
833 if (ulen < skb->len) {
834 if (pskb_trim_rcsum(skb, ulen))
835 goto short_packet;
836 saddr = &ipv6_hdr(skb)->saddr;
837 daddr = &ipv6_hdr(skb)->daddr;
838 uh = udp_hdr(skb);
839 }
840 }
841
842 if (udp6_csum_init(skb, uh, proto))
843 goto csum_error;
844
845 /* Check if the socket is already available, e.g. due to early demux */
846 sk = skb_steal_sock(skb);
847 if (sk) {
848 struct dst_entry *dst = skb_dst(skb);
849 int ret;
850
851 if (unlikely(sk->sk_rx_dst != dst))
852 udp6_sk_rx_dst_set(sk, dst);
853
854 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
855 sock_put(sk);
856 goto report_csum_error;
857 }
858
859 ret = udp6_unicast_rcv_skb(sk, skb, uh);
860 sock_put(sk);
861 return ret;
862 }
863
864 /*
865 * Multicast receive code
866 */
867 if (ipv6_addr_is_multicast(daddr))
868 return __udp6_lib_mcast_deliver(net, skb,
869 saddr, daddr, udptable, proto);
870
871 /* Unicast */
872 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
873 if (sk) {
874 if (!uh->check && !udp_sk(sk)->no_check6_rx)
875 goto report_csum_error;
876 return udp6_unicast_rcv_skb(sk, skb, uh);
877 }
878
879 if (!uh->check)
880 goto report_csum_error;
881
882 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
883 goto discard;
884
885 if (udp_lib_checksum_complete(skb))
886 goto csum_error;
887
888 __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
889 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
890
891 kfree_skb(skb);
892 return 0;
893
894 short_packet:
895 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
896 proto == IPPROTO_UDPLITE ? "-Lite" : "",
897 saddr, ntohs(uh->source),
898 ulen, skb->len,
899 daddr, ntohs(uh->dest));
900 goto discard;
901
902 report_csum_error:
903 udp6_csum_zero_error(skb);
904 csum_error:
905 __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
906 discard:
907 __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
908 kfree_skb(skb);
909 return 0;
910 }
911
912
__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)913 static struct sock *__udp6_lib_demux_lookup(struct net *net,
914 __be16 loc_port, const struct in6_addr *loc_addr,
915 __be16 rmt_port, const struct in6_addr *rmt_addr,
916 int dif, int sdif)
917 {
918 unsigned short hnum = ntohs(loc_port);
919 unsigned int hash2 = udp6_portaddr_hash(net, loc_addr, hnum);
920 unsigned int slot2 = hash2 & udp_table.mask;
921 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
922 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
923 struct sock *sk;
924
925 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
926 if (sk->sk_state == TCP_ESTABLISHED &&
927 INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif, sdif))
928 return sk;
929 /* Only check first socket in chain */
930 break;
931 }
932 return NULL;
933 }
934
udp_v6_early_demux(struct sk_buff * skb)935 static void udp_v6_early_demux(struct sk_buff *skb)
936 {
937 struct net *net = dev_net(skb->dev);
938 const struct udphdr *uh;
939 struct sock *sk;
940 struct dst_entry *dst;
941 int dif = skb->dev->ifindex;
942 int sdif = inet6_sdif(skb);
943
944 if (!pskb_may_pull(skb, skb_transport_offset(skb) +
945 sizeof(struct udphdr)))
946 return;
947
948 uh = udp_hdr(skb);
949
950 if (skb->pkt_type == PACKET_HOST)
951 sk = __udp6_lib_demux_lookup(net, uh->dest,
952 &ipv6_hdr(skb)->daddr,
953 uh->source, &ipv6_hdr(skb)->saddr,
954 dif, sdif);
955 else
956 return;
957
958 if (!sk || !refcount_inc_not_zero(&sk->sk_refcnt))
959 return;
960
961 skb->sk = sk;
962 skb->destructor = sock_efree;
963 dst = READ_ONCE(sk->sk_rx_dst);
964
965 if (dst)
966 dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
967 if (dst) {
968 /* set noref for now.
969 * any place which wants to hold dst has to call
970 * dst_hold_safe()
971 */
972 skb_dst_set_noref(skb, dst);
973 }
974 }
975
udpv6_rcv(struct sk_buff * skb)976 static __inline__ int udpv6_rcv(struct sk_buff *skb)
977 {
978 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
979 }
980
981 /*
982 * Throw away all pending data and cancel the corking. Socket is locked.
983 */
udp_v6_flush_pending_frames(struct sock * sk)984 static void udp_v6_flush_pending_frames(struct sock *sk)
985 {
986 struct udp_sock *up = udp_sk(sk);
987
988 if (up->pending == AF_INET)
989 udp_flush_pending_frames(sk);
990 else if (up->pending) {
991 up->len = 0;
992 up->pending = 0;
993 ip6_flush_pending_frames(sk);
994 }
995 }
996
997 /**
998 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
999 * @sk: socket we are sending on
1000 * @skb: sk_buff containing the filled-in UDP header
1001 * (checksum field must be zeroed out)
1002 */
udp6_hwcsum_outgoing(struct sock * sk,struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int len)1003 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
1004 const struct in6_addr *saddr,
1005 const struct in6_addr *daddr, int len)
1006 {
1007 unsigned int offset;
1008 struct udphdr *uh = udp_hdr(skb);
1009 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
1010 __wsum csum = 0;
1011
1012 if (!frags) {
1013 /* Only one fragment on the socket. */
1014 skb->csum_start = skb_transport_header(skb) - skb->head;
1015 skb->csum_offset = offsetof(struct udphdr, check);
1016 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1017 } else {
1018 /*
1019 * HW-checksum won't work as there are two or more
1020 * fragments on the socket so that all csums of sk_buffs
1021 * should be together
1022 */
1023 offset = skb_transport_offset(skb);
1024 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1025 csum = skb->csum;
1026
1027 skb->ip_summed = CHECKSUM_NONE;
1028
1029 do {
1030 csum = csum_add(csum, frags->csum);
1031 } while ((frags = frags->next));
1032
1033 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1034 csum);
1035 if (uh->check == 0)
1036 uh->check = CSUM_MANGLED_0;
1037 }
1038 }
1039
1040 /*
1041 * Sending
1042 */
1043
udp_v6_send_skb(struct sk_buff * skb,struct flowi6 * fl6)1044 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1045 {
1046 struct sock *sk = skb->sk;
1047 struct udphdr *uh;
1048 int err = 0;
1049 int is_udplite = IS_UDPLITE(sk);
1050 __wsum csum = 0;
1051 int offset = skb_transport_offset(skb);
1052 int len = skb->len - offset;
1053
1054 /*
1055 * Create a UDP header
1056 */
1057 uh = udp_hdr(skb);
1058 uh->source = fl6->fl6_sport;
1059 uh->dest = fl6->fl6_dport;
1060 uh->len = htons(len);
1061 uh->check = 0;
1062
1063 if (is_udplite)
1064 csum = udplite_csum(skb);
1065 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
1066 skb->ip_summed = CHECKSUM_NONE;
1067 goto send;
1068 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1069 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1070 goto send;
1071 } else
1072 csum = udp_csum(skb);
1073
1074 /* add protocol-dependent pseudo-header */
1075 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1076 len, fl6->flowi6_proto, csum);
1077 if (uh->check == 0)
1078 uh->check = CSUM_MANGLED_0;
1079
1080 send:
1081 err = ip6_send_skb(skb);
1082 if (err) {
1083 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1084 UDP6_INC_STATS(sock_net(sk),
1085 UDP_MIB_SNDBUFERRORS, is_udplite);
1086 err = 0;
1087 }
1088 } else {
1089 UDP6_INC_STATS(sock_net(sk),
1090 UDP_MIB_OUTDATAGRAMS, is_udplite);
1091 }
1092 return err;
1093 }
1094
udp_v6_push_pending_frames(struct sock * sk)1095 static int udp_v6_push_pending_frames(struct sock *sk)
1096 {
1097 struct sk_buff *skb;
1098 struct udp_sock *up = udp_sk(sk);
1099 struct flowi6 fl6;
1100 int err = 0;
1101
1102 if (up->pending == AF_INET)
1103 return udp_push_pending_frames(sk);
1104
1105 /* ip6_finish_skb will release the cork, so make a copy of
1106 * fl6 here.
1107 */
1108 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1109
1110 skb = ip6_finish_skb(sk);
1111 if (!skb)
1112 goto out;
1113
1114 err = udp_v6_send_skb(skb, &fl6);
1115
1116 out:
1117 up->len = 0;
1118 up->pending = 0;
1119 return err;
1120 }
1121
udpv6_sendmsg(struct sock * sk,struct msghdr * msg,size_t len)1122 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1123 {
1124 struct ipv6_txoptions opt_space;
1125 struct udp_sock *up = udp_sk(sk);
1126 struct inet_sock *inet = inet_sk(sk);
1127 struct ipv6_pinfo *np = inet6_sk(sk);
1128 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1129 struct in6_addr *daddr, *final_p, final;
1130 struct ipv6_txoptions *opt = NULL;
1131 struct ipv6_txoptions *opt_to_free = NULL;
1132 struct ip6_flowlabel *flowlabel = NULL;
1133 struct flowi6 fl6;
1134 struct dst_entry *dst;
1135 struct ipcm6_cookie ipc6;
1136 int addr_len = msg->msg_namelen;
1137 int ulen = len;
1138 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1139 int err;
1140 int connected = 0;
1141 int is_udplite = IS_UDPLITE(sk);
1142 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1143 struct sockcm_cookie sockc;
1144
1145 ipc6.hlimit = -1;
1146 ipc6.tclass = -1;
1147 ipc6.dontfrag = -1;
1148 sockc.tsflags = sk->sk_tsflags;
1149
1150 /* destination address check */
1151 if (sin6) {
1152 if (addr_len < offsetof(struct sockaddr, sa_data))
1153 return -EINVAL;
1154
1155 switch (sin6->sin6_family) {
1156 case AF_INET6:
1157 if (addr_len < SIN6_LEN_RFC2133)
1158 return -EINVAL;
1159 daddr = &sin6->sin6_addr;
1160 if (ipv6_addr_any(daddr) &&
1161 ipv6_addr_v4mapped(&np->saddr))
1162 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
1163 daddr);
1164 break;
1165 case AF_INET:
1166 goto do_udp_sendmsg;
1167 case AF_UNSPEC:
1168 msg->msg_name = sin6 = NULL;
1169 msg->msg_namelen = addr_len = 0;
1170 daddr = NULL;
1171 break;
1172 default:
1173 return -EINVAL;
1174 }
1175 } else if (!up->pending) {
1176 if (sk->sk_state != TCP_ESTABLISHED)
1177 return -EDESTADDRREQ;
1178 daddr = &sk->sk_v6_daddr;
1179 } else
1180 daddr = NULL;
1181
1182 if (daddr) {
1183 if (ipv6_addr_v4mapped(daddr)) {
1184 struct sockaddr_in sin;
1185 sin.sin_family = AF_INET;
1186 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1187 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1188 msg->msg_name = &sin;
1189 msg->msg_namelen = sizeof(sin);
1190 do_udp_sendmsg:
1191 if (__ipv6_only_sock(sk))
1192 return -ENETUNREACH;
1193 return udp_sendmsg(sk, msg, len);
1194 }
1195 }
1196
1197 if (up->pending == AF_INET)
1198 return udp_sendmsg(sk, msg, len);
1199
1200 /* Rough check on arithmetic overflow,
1201 better check is made in ip6_append_data().
1202 */
1203 if (len > INT_MAX - sizeof(struct udphdr))
1204 return -EMSGSIZE;
1205
1206 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1207 if (up->pending) {
1208 /*
1209 * There are pending frames.
1210 * The socket lock must be held while it's corked.
1211 */
1212 lock_sock(sk);
1213 if (likely(up->pending)) {
1214 if (unlikely(up->pending != AF_INET6)) {
1215 release_sock(sk);
1216 return -EAFNOSUPPORT;
1217 }
1218 dst = NULL;
1219 goto do_append_data;
1220 }
1221 release_sock(sk);
1222 }
1223 ulen += sizeof(struct udphdr);
1224
1225 memset(&fl6, 0, sizeof(fl6));
1226
1227 if (sin6) {
1228 if (sin6->sin6_port == 0)
1229 return -EINVAL;
1230
1231 fl6.fl6_dport = sin6->sin6_port;
1232 daddr = &sin6->sin6_addr;
1233
1234 if (np->sndflow) {
1235 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1236 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1237 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1238 if (!flowlabel)
1239 return -EINVAL;
1240 }
1241 }
1242
1243 /*
1244 * Otherwise it will be difficult to maintain
1245 * sk->sk_dst_cache.
1246 */
1247 if (sk->sk_state == TCP_ESTABLISHED &&
1248 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1249 daddr = &sk->sk_v6_daddr;
1250
1251 if (addr_len >= sizeof(struct sockaddr_in6) &&
1252 sin6->sin6_scope_id &&
1253 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1254 fl6.flowi6_oif = sin6->sin6_scope_id;
1255 } else {
1256 if (sk->sk_state != TCP_ESTABLISHED)
1257 return -EDESTADDRREQ;
1258
1259 fl6.fl6_dport = inet->inet_dport;
1260 daddr = &sk->sk_v6_daddr;
1261 fl6.flowlabel = np->flow_label;
1262 connected = 1;
1263 }
1264
1265 if (!fl6.flowi6_oif)
1266 fl6.flowi6_oif = sk->sk_bound_dev_if;
1267
1268 if (!fl6.flowi6_oif)
1269 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1270
1271 fl6.flowi6_mark = sk->sk_mark;
1272 fl6.flowi6_uid = sk->sk_uid;
1273
1274 if (msg->msg_controllen) {
1275 opt = &opt_space;
1276 memset(opt, 0, sizeof(struct ipv6_txoptions));
1277 opt->tot_len = sizeof(*opt);
1278 ipc6.opt = opt;
1279
1280 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
1281 if (err < 0) {
1282 fl6_sock_release(flowlabel);
1283 return err;
1284 }
1285 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1286 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1287 if (!flowlabel)
1288 return -EINVAL;
1289 }
1290 if (!(opt->opt_nflen|opt->opt_flen))
1291 opt = NULL;
1292 connected = 0;
1293 }
1294 if (!opt) {
1295 opt = txopt_get(np);
1296 opt_to_free = opt;
1297 }
1298 if (flowlabel)
1299 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1300 opt = ipv6_fixup_options(&opt_space, opt);
1301 ipc6.opt = opt;
1302
1303 fl6.flowi6_proto = sk->sk_protocol;
1304 if (!ipv6_addr_any(daddr))
1305 fl6.daddr = *daddr;
1306 else
1307 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1308 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1309 fl6.saddr = np->saddr;
1310 fl6.fl6_sport = inet->inet_sport;
1311
1312 final_p = fl6_update_dst(&fl6, opt, &final);
1313 if (final_p)
1314 connected = 0;
1315
1316 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1317 fl6.flowi6_oif = np->mcast_oif;
1318 connected = 0;
1319 } else if (!fl6.flowi6_oif)
1320 fl6.flowi6_oif = np->ucast_oif;
1321
1322 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1323
1324 if (ipc6.tclass < 0)
1325 ipc6.tclass = np->tclass;
1326
1327 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
1328
1329 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1330 if (IS_ERR(dst)) {
1331 err = PTR_ERR(dst);
1332 dst = NULL;
1333 goto out;
1334 }
1335
1336 if (ipc6.hlimit < 0)
1337 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1338
1339 if (msg->msg_flags&MSG_CONFIRM)
1340 goto do_confirm;
1341 back_from_confirm:
1342
1343 /* Lockless fast path for the non-corking case */
1344 if (!corkreq) {
1345 struct sk_buff *skb;
1346
1347 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1348 sizeof(struct udphdr), &ipc6,
1349 &fl6, (struct rt6_info *)dst,
1350 msg->msg_flags, &sockc);
1351 err = PTR_ERR(skb);
1352 if (!IS_ERR_OR_NULL(skb))
1353 err = udp_v6_send_skb(skb, &fl6);
1354 goto release_dst;
1355 }
1356
1357 lock_sock(sk);
1358 if (unlikely(up->pending)) {
1359 /* The socket is already corked while preparing it. */
1360 /* ... which is an evident application bug. --ANK */
1361 release_sock(sk);
1362
1363 net_dbg_ratelimited("udp cork app bug 2\n");
1364 err = -EINVAL;
1365 goto out;
1366 }
1367
1368 up->pending = AF_INET6;
1369
1370 do_append_data:
1371 if (ipc6.dontfrag < 0)
1372 ipc6.dontfrag = np->dontfrag;
1373 up->len += ulen;
1374 err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1375 &ipc6, &fl6, (struct rt6_info *)dst,
1376 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, &sockc);
1377 if (err)
1378 udp_v6_flush_pending_frames(sk);
1379 else if (!corkreq)
1380 err = udp_v6_push_pending_frames(sk);
1381 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1382 up->pending = 0;
1383
1384 if (err > 0)
1385 err = np->recverr ? net_xmit_errno(err) : 0;
1386 release_sock(sk);
1387
1388 release_dst:
1389 if (dst) {
1390 if (connected) {
1391 ip6_dst_store(sk, dst,
1392 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1393 &sk->sk_v6_daddr : NULL,
1394 #ifdef CONFIG_IPV6_SUBTREES
1395 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1396 &np->saddr :
1397 #endif
1398 NULL);
1399 } else {
1400 dst_release(dst);
1401 }
1402 dst = NULL;
1403 }
1404
1405 out:
1406 dst_release(dst);
1407 fl6_sock_release(flowlabel);
1408 txopt_put(opt_to_free);
1409 if (!err)
1410 return len;
1411 /*
1412 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1413 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1414 * we don't have a good statistic (IpOutDiscards but it can be too many
1415 * things). We could add another new stat but at least for now that
1416 * seems like overkill.
1417 */
1418 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1419 UDP6_INC_STATS(sock_net(sk),
1420 UDP_MIB_SNDBUFERRORS, is_udplite);
1421 }
1422 return err;
1423
1424 do_confirm:
1425 if (msg->msg_flags & MSG_PROBE)
1426 dst_confirm_neigh(dst, &fl6.daddr);
1427 if (!(msg->msg_flags&MSG_PROBE) || len)
1428 goto back_from_confirm;
1429 err = 0;
1430 goto out;
1431 }
1432
udpv6_destroy_sock(struct sock * sk)1433 void udpv6_destroy_sock(struct sock *sk)
1434 {
1435 struct udp_sock *up = udp_sk(sk);
1436 lock_sock(sk);
1437 udp_v6_flush_pending_frames(sk);
1438 release_sock(sk);
1439
1440 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1441 void (*encap_destroy)(struct sock *sk);
1442 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1443 if (encap_destroy)
1444 encap_destroy(sk);
1445 }
1446
1447 inet6_destroy_sock(sk);
1448 }
1449
1450 /*
1451 * Socket option code for UDP
1452 */
udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1453 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1454 char __user *optval, unsigned int optlen)
1455 {
1456 if (level == SOL_UDP || level == SOL_UDPLITE)
1457 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1458 udp_v6_push_pending_frames);
1459 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1460 }
1461
1462 #ifdef CONFIG_COMPAT
compat_udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1463 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1464 char __user *optval, unsigned int optlen)
1465 {
1466 if (level == SOL_UDP || level == SOL_UDPLITE)
1467 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1468 udp_v6_push_pending_frames);
1469 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1470 }
1471 #endif
1472
udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1473 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1474 char __user *optval, int __user *optlen)
1475 {
1476 if (level == SOL_UDP || level == SOL_UDPLITE)
1477 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1478 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1479 }
1480
1481 #ifdef CONFIG_COMPAT
compat_udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1482 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1483 char __user *optval, int __user *optlen)
1484 {
1485 if (level == SOL_UDP || level == SOL_UDPLITE)
1486 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1487 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1488 }
1489 #endif
1490
1491 /* thinking of making this const? Don't.
1492 * early_demux can change based on sysctl.
1493 */
1494 static struct inet6_protocol udpv6_protocol = {
1495 .early_demux = udp_v6_early_demux,
1496 .early_demux_handler = udp_v6_early_demux,
1497 .handler = udpv6_rcv,
1498 .err_handler = udpv6_err,
1499 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1500 };
1501
1502 /* ------------------------------------------------------------------------ */
1503 #ifdef CONFIG_PROC_FS
udp6_seq_show(struct seq_file * seq,void * v)1504 int udp6_seq_show(struct seq_file *seq, void *v)
1505 {
1506 if (v == SEQ_START_TOKEN) {
1507 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1508 } else {
1509 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1510 struct inet_sock *inet = inet_sk(v);
1511 __u16 srcp = ntohs(inet->inet_sport);
1512 __u16 destp = ntohs(inet->inet_dport);
1513 __ip6_dgram_sock_seq_show(seq, v, srcp, destp,
1514 udp_rqueue_get(v), bucket);
1515 }
1516 return 0;
1517 }
1518
1519 static const struct file_operations udp6_afinfo_seq_fops = {
1520 .owner = THIS_MODULE,
1521 .open = udp_seq_open,
1522 .read = seq_read,
1523 .llseek = seq_lseek,
1524 .release = seq_release_net
1525 };
1526
1527 static struct udp_seq_afinfo udp6_seq_afinfo = {
1528 .name = "udp6",
1529 .family = AF_INET6,
1530 .udp_table = &udp_table,
1531 .seq_fops = &udp6_afinfo_seq_fops,
1532 .seq_ops = {
1533 .show = udp6_seq_show,
1534 },
1535 };
1536
udp6_proc_init(struct net * net)1537 int __net_init udp6_proc_init(struct net *net)
1538 {
1539 return udp_proc_register(net, &udp6_seq_afinfo);
1540 }
1541
udp6_proc_exit(struct net * net)1542 void udp6_proc_exit(struct net *net)
1543 {
1544 udp_proc_unregister(net, &udp6_seq_afinfo);
1545 }
1546 #endif /* CONFIG_PROC_FS */
1547
1548 /* ------------------------------------------------------------------------ */
1549
1550 struct proto udpv6_prot = {
1551 .name = "UDPv6",
1552 .owner = THIS_MODULE,
1553 .close = udp_lib_close,
1554 .connect = ip6_datagram_connect,
1555 .disconnect = udp_disconnect,
1556 .ioctl = udp_ioctl,
1557 .init = udp_init_sock,
1558 .destroy = udpv6_destroy_sock,
1559 .setsockopt = udpv6_setsockopt,
1560 .getsockopt = udpv6_getsockopt,
1561 .sendmsg = udpv6_sendmsg,
1562 .recvmsg = udpv6_recvmsg,
1563 .release_cb = ip6_datagram_release_cb,
1564 .hash = udp_lib_hash,
1565 .unhash = udp_lib_unhash,
1566 .rehash = udp_v6_rehash,
1567 .get_port = udp_v6_get_port,
1568 .memory_allocated = &udp_memory_allocated,
1569 .sysctl_mem = sysctl_udp_mem,
1570 .sysctl_wmem = &sysctl_udp_wmem_min,
1571 .sysctl_rmem = &sysctl_udp_rmem_min,
1572 .obj_size = sizeof(struct udp6_sock),
1573 .h.udp_table = &udp_table,
1574 #ifdef CONFIG_COMPAT
1575 .compat_setsockopt = compat_udpv6_setsockopt,
1576 .compat_getsockopt = compat_udpv6_getsockopt,
1577 #endif
1578 .diag_destroy = udp_abort,
1579 };
1580
1581 static struct inet_protosw udpv6_protosw = {
1582 .type = SOCK_DGRAM,
1583 .protocol = IPPROTO_UDP,
1584 .prot = &udpv6_prot,
1585 .ops = &inet6_dgram_ops,
1586 .flags = INET_PROTOSW_PERMANENT,
1587 };
1588
udpv6_init(void)1589 int __init udpv6_init(void)
1590 {
1591 int ret;
1592
1593 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1594 if (ret)
1595 goto out;
1596
1597 ret = inet6_register_protosw(&udpv6_protosw);
1598 if (ret)
1599 goto out_udpv6_protocol;
1600 out:
1601 return ret;
1602
1603 out_udpv6_protocol:
1604 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1605 goto out;
1606 }
1607
udpv6_exit(void)1608 void udpv6_exit(void)
1609 {
1610 inet6_unregister_protosw(&udpv6_protosw);
1611 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1612 }
1613