1 /*
2 * UDP over IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Based on linux/ipv4/udp.c
9 *
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
14 * a single port at the same time.
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 */
23
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/ipv6.h>
33 #include <linux/icmpv6.h>
34 #include <linux/init.h>
35 #include <linux/module.h>
36 #include <linux/skbuff.h>
37 #include <linux/slab.h>
38 #include <asm/uaccess.h>
39
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_route.h>
44 #include <net/raw.h>
45 #include <net/tcp_states.h>
46 #include <net/ip6_checksum.h>
47 #include <net/xfrm.h>
48 #include <net/inet6_hashtables.h>
49 #include <net/busy_poll.h>
50
51 #include <linux/proc_fs.h>
52 #include <linux/seq_file.h>
53 #include <trace/events/skb.h>
54 #include "udp_impl.h"
55
udp6_ehashfn(const struct net * net,const struct in6_addr * laddr,const u16 lport,const struct in6_addr * faddr,const __be16 fport)56 static u32 udp6_ehashfn(const struct net *net,
57 const struct in6_addr *laddr,
58 const u16 lport,
59 const struct in6_addr *faddr,
60 const __be16 fport)
61 {
62 static u32 udp6_ehash_secret __read_mostly;
63 static u32 udp_ipv6_hash_secret __read_mostly;
64
65 u32 lhash, fhash;
66
67 net_get_random_once(&udp6_ehash_secret,
68 sizeof(udp6_ehash_secret));
69 net_get_random_once(&udp_ipv6_hash_secret,
70 sizeof(udp_ipv6_hash_secret));
71
72 lhash = (__force u32)laddr->s6_addr32[3];
73 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
74
75 return __inet6_ehashfn(lhash, lport, fhash, fport,
76 udp_ipv6_hash_secret + net_hash_mix(net));
77 }
78
ipv6_rcv_saddr_equal(const struct sock * sk,const struct sock * sk2)79 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
80 {
81 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
82 int sk2_ipv6only = inet_v6_ipv6only(sk2);
83 int addr_type = ipv6_addr_type(&sk->sk_v6_rcv_saddr);
84 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
85
86 /* if both are mapped, treat as IPv4 */
87 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
88 return (!sk2_ipv6only &&
89 (!sk->sk_rcv_saddr || !sk2->sk_rcv_saddr ||
90 sk->sk_rcv_saddr == sk2->sk_rcv_saddr));
91
92 if (addr_type2 == IPV6_ADDR_ANY &&
93 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
94 return 1;
95
96 if (addr_type == IPV6_ADDR_ANY &&
97 !(ipv6_only_sock(sk) && addr_type2 == IPV6_ADDR_MAPPED))
98 return 1;
99
100 if (sk2_rcv_saddr6 &&
101 ipv6_addr_equal(&sk->sk_v6_rcv_saddr, sk2_rcv_saddr6))
102 return 1;
103
104 return 0;
105 }
106
udp6_portaddr_hash(const struct net * net,const struct in6_addr * addr6,unsigned int port)107 static u32 udp6_portaddr_hash(const struct net *net,
108 const struct in6_addr *addr6,
109 unsigned int port)
110 {
111 unsigned int hash, mix = net_hash_mix(net);
112
113 if (ipv6_addr_any(addr6))
114 hash = jhash_1word(0, mix);
115 else if (ipv6_addr_v4mapped(addr6))
116 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
117 else
118 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
119
120 return hash ^ port;
121 }
122
udp_v6_get_port(struct sock * sk,unsigned short snum)123 int udp_v6_get_port(struct sock *sk, unsigned short snum)
124 {
125 unsigned int hash2_nulladdr =
126 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
127 unsigned int hash2_partial =
128 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
129
130 /* precompute partial secondary hash */
131 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
132 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
133 }
134
udp_v6_rehash(struct sock * sk)135 static void udp_v6_rehash(struct sock *sk)
136 {
137 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
138 &sk->sk_v6_rcv_saddr,
139 inet_sk(sk)->inet_num);
140
141 udp_lib_rehash(sk, new_hash);
142 }
143
compute_score(struct sock * sk,struct net * net,unsigned short hnum,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)144 static inline int compute_score(struct sock *sk, struct net *net,
145 unsigned short hnum,
146 const struct in6_addr *saddr, __be16 sport,
147 const struct in6_addr *daddr, __be16 dport,
148 int dif)
149 {
150 int score;
151 struct inet_sock *inet;
152
153 if (!net_eq(sock_net(sk), net) ||
154 udp_sk(sk)->udp_port_hash != hnum ||
155 sk->sk_family != PF_INET6)
156 return -1;
157
158 score = 0;
159 inet = inet_sk(sk);
160
161 if (inet->inet_dport) {
162 if (inet->inet_dport != sport)
163 return -1;
164 score++;
165 }
166
167 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
168 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
169 return -1;
170 score++;
171 }
172
173 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
174 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
175 return -1;
176 score++;
177 }
178
179 if (sk->sk_bound_dev_if) {
180 if (sk->sk_bound_dev_if != dif)
181 return -1;
182 score++;
183 }
184
185 if (sk->sk_incoming_cpu == raw_smp_processor_id())
186 score++;
187
188 return score;
189 }
190
compute_score2(struct sock * sk,struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned short hnum,int dif)191 static inline int compute_score2(struct sock *sk, struct net *net,
192 const struct in6_addr *saddr, __be16 sport,
193 const struct in6_addr *daddr,
194 unsigned short hnum, int dif)
195 {
196 int score;
197 struct inet_sock *inet;
198
199 if (!net_eq(sock_net(sk), net) ||
200 udp_sk(sk)->udp_port_hash != hnum ||
201 sk->sk_family != PF_INET6)
202 return -1;
203
204 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
205 return -1;
206
207 score = 0;
208 inet = inet_sk(sk);
209
210 if (inet->inet_dport) {
211 if (inet->inet_dport != sport)
212 return -1;
213 score++;
214 }
215
216 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
217 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
218 return -1;
219 score++;
220 }
221
222 if (sk->sk_bound_dev_if) {
223 if (sk->sk_bound_dev_if != dif)
224 return -1;
225 score++;
226 }
227
228 if (sk->sk_incoming_cpu == raw_smp_processor_id())
229 score++;
230
231 return score;
232 }
233
234 /* called with read_rcu_lock() */
udp6_lib_lookup2(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned int hnum,int dif,struct udp_hslot * hslot2,unsigned int slot2)235 static struct sock *udp6_lib_lookup2(struct net *net,
236 const struct in6_addr *saddr, __be16 sport,
237 const struct in6_addr *daddr, unsigned int hnum, int dif,
238 struct udp_hslot *hslot2, unsigned int slot2)
239 {
240 struct sock *sk, *result;
241 struct hlist_nulls_node *node;
242 int score, badness, matches = 0, reuseport = 0;
243 u32 hash = 0;
244
245 begin:
246 result = NULL;
247 badness = -1;
248 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
249 score = compute_score2(sk, net, saddr, sport,
250 daddr, hnum, dif);
251 if (score > badness) {
252 result = sk;
253 badness = score;
254 reuseport = sk->sk_reuseport;
255 if (reuseport) {
256 hash = udp6_ehashfn(net, daddr, hnum,
257 saddr, sport);
258 matches = 1;
259 }
260 } else if (score == badness && reuseport) {
261 matches++;
262 if (reciprocal_scale(hash, matches) == 0)
263 result = sk;
264 hash = next_pseudo_random32(hash);
265 }
266 }
267 /*
268 * if the nulls value we got at the end of this lookup is
269 * not the expected one, we must restart lookup.
270 * We probably met an item that was moved to another chain.
271 */
272 if (get_nulls_value(node) != slot2)
273 goto begin;
274
275 if (result) {
276 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
277 result = NULL;
278 else if (unlikely(compute_score2(result, net, saddr, sport,
279 daddr, hnum, dif) < badness)) {
280 sock_put(result);
281 goto begin;
282 }
283 }
284 return result;
285 }
286
__udp6_lib_lookup(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif,struct udp_table * udptable)287 struct sock *__udp6_lib_lookup(struct net *net,
288 const struct in6_addr *saddr, __be16 sport,
289 const struct in6_addr *daddr, __be16 dport,
290 int dif, struct udp_table *udptable)
291 {
292 struct sock *sk, *result;
293 struct hlist_nulls_node *node;
294 unsigned short hnum = ntohs(dport);
295 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
296 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
297 int score, badness, matches = 0, reuseport = 0;
298 u32 hash = 0;
299
300 rcu_read_lock();
301 if (hslot->count > 10) {
302 hash2 = udp6_portaddr_hash(net, daddr, hnum);
303 slot2 = hash2 & udptable->mask;
304 hslot2 = &udptable->hash2[slot2];
305 if (hslot->count < hslot2->count)
306 goto begin;
307
308 result = udp6_lib_lookup2(net, saddr, sport,
309 daddr, hnum, dif,
310 hslot2, slot2);
311 if (!result) {
312 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
313 slot2 = hash2 & udptable->mask;
314 hslot2 = &udptable->hash2[slot2];
315 if (hslot->count < hslot2->count)
316 goto begin;
317
318 result = udp6_lib_lookup2(net, saddr, sport,
319 &in6addr_any, hnum, dif,
320 hslot2, slot2);
321 }
322 rcu_read_unlock();
323 return result;
324 }
325 begin:
326 result = NULL;
327 badness = -1;
328 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
329 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
330 if (score > badness) {
331 result = sk;
332 badness = score;
333 reuseport = sk->sk_reuseport;
334 if (reuseport) {
335 hash = udp6_ehashfn(net, daddr, hnum,
336 saddr, sport);
337 matches = 1;
338 }
339 } else if (score == badness && reuseport) {
340 matches++;
341 if (reciprocal_scale(hash, matches) == 0)
342 result = sk;
343 hash = next_pseudo_random32(hash);
344 }
345 }
346 /*
347 * if the nulls value we got at the end of this lookup is
348 * not the expected one, we must restart lookup.
349 * We probably met an item that was moved to another chain.
350 */
351 if (get_nulls_value(node) != slot)
352 goto begin;
353
354 if (result) {
355 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
356 result = NULL;
357 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
358 daddr, dport, dif) < badness)) {
359 sock_put(result);
360 goto begin;
361 }
362 }
363 rcu_read_unlock();
364 return result;
365 }
366 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
367
__udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport,struct udp_table * udptable)368 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
369 __be16 sport, __be16 dport,
370 struct udp_table *udptable)
371 {
372 struct sock *sk;
373 const struct ipv6hdr *iph = ipv6_hdr(skb);
374
375 sk = skb_steal_sock(skb);
376 if (unlikely(sk))
377 return sk;
378 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
379 &iph->daddr, dport, inet6_iif(skb),
380 udptable);
381 }
382
udp6_lib_lookup(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)383 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
384 const struct in6_addr *daddr, __be16 dport, int dif)
385 {
386 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
387 }
388 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
389
390 /*
391 * This should be easy, if there is something there we
392 * return it, otherwise we block.
393 */
394
udpv6_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int noblock,int flags,int * addr_len)395 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
396 int noblock, int flags, int *addr_len)
397 {
398 struct ipv6_pinfo *np = inet6_sk(sk);
399 struct inet_sock *inet = inet_sk(sk);
400 struct sk_buff *skb;
401 unsigned int ulen, copied;
402 int peeked, off = 0;
403 int err;
404 int is_udplite = IS_UDPLITE(sk);
405 int is_udp4;
406 bool slow;
407
408 if (flags & MSG_ERRQUEUE)
409 return ipv6_recv_error(sk, msg, len, addr_len);
410
411 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
412 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
413
414 try_again:
415 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
416 &peeked, &off, &err);
417 if (!skb)
418 goto out;
419
420 ulen = skb->len - sizeof(struct udphdr);
421 copied = len;
422 if (copied > ulen)
423 copied = ulen;
424 else if (copied < ulen)
425 msg->msg_flags |= MSG_TRUNC;
426
427 is_udp4 = (skb->protocol == htons(ETH_P_IP));
428
429 /*
430 * If checksum is needed at all, try to do it while copying the
431 * data. If the data is truncated, or if we only want a partial
432 * coverage checksum (UDP-Lite), do it before the copy.
433 */
434
435 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
436 if (udp_lib_checksum_complete(skb))
437 goto csum_copy_err;
438 }
439
440 if (skb_csum_unnecessary(skb))
441 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
442 msg, copied);
443 else {
444 err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), msg);
445 if (err == -EINVAL)
446 goto csum_copy_err;
447 }
448 if (unlikely(err)) {
449 trace_kfree_skb(skb, udpv6_recvmsg);
450 if (!peeked) {
451 atomic_inc(&sk->sk_drops);
452 if (is_udp4)
453 UDP_INC_STATS_USER(sock_net(sk),
454 UDP_MIB_INERRORS,
455 is_udplite);
456 else
457 UDP6_INC_STATS_USER(sock_net(sk),
458 UDP_MIB_INERRORS,
459 is_udplite);
460 }
461 goto out_free;
462 }
463 if (!peeked) {
464 if (is_udp4)
465 UDP_INC_STATS_USER(sock_net(sk),
466 UDP_MIB_INDATAGRAMS, is_udplite);
467 else
468 UDP6_INC_STATS_USER(sock_net(sk),
469 UDP_MIB_INDATAGRAMS, is_udplite);
470 }
471
472 sock_recv_ts_and_drops(msg, sk, skb);
473
474 /* Copy the address. */
475 if (msg->msg_name) {
476 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
477 sin6->sin6_family = AF_INET6;
478 sin6->sin6_port = udp_hdr(skb)->source;
479 sin6->sin6_flowinfo = 0;
480
481 if (is_udp4) {
482 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
483 &sin6->sin6_addr);
484 sin6->sin6_scope_id = 0;
485 } else {
486 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
487 sin6->sin6_scope_id =
488 ipv6_iface_scope_id(&sin6->sin6_addr,
489 inet6_iif(skb));
490 }
491 *addr_len = sizeof(*sin6);
492 }
493
494 if (np->rxopt.all)
495 ip6_datagram_recv_common_ctl(sk, msg, skb);
496
497 if (is_udp4) {
498 if (inet->cmsg_flags)
499 ip_cmsg_recv(msg, skb);
500 } else {
501 if (np->rxopt.all)
502 ip6_datagram_recv_specific_ctl(sk, msg, skb);
503 }
504
505 err = copied;
506 if (flags & MSG_TRUNC)
507 err = ulen;
508
509 out_free:
510 skb_free_datagram_locked(sk, skb);
511 out:
512 return err;
513
514 csum_copy_err:
515 slow = lock_sock_fast(sk);
516 if (!skb_kill_datagram(sk, skb, flags)) {
517 if (is_udp4) {
518 UDP_INC_STATS_USER(sock_net(sk),
519 UDP_MIB_CSUMERRORS, is_udplite);
520 UDP_INC_STATS_USER(sock_net(sk),
521 UDP_MIB_INERRORS, is_udplite);
522 } else {
523 UDP6_INC_STATS_USER(sock_net(sk),
524 UDP_MIB_CSUMERRORS, is_udplite);
525 UDP6_INC_STATS_USER(sock_net(sk),
526 UDP_MIB_INERRORS, is_udplite);
527 }
528 }
529 unlock_sock_fast(sk, slow);
530
531 /* starting over for a new packet, but check if we need to yield */
532 cond_resched();
533 msg->msg_flags &= ~MSG_TRUNC;
534 goto try_again;
535 }
536
__udp6_lib_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info,struct udp_table * udptable)537 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
538 u8 type, u8 code, int offset, __be32 info,
539 struct udp_table *udptable)
540 {
541 struct ipv6_pinfo *np;
542 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
543 const struct in6_addr *saddr = &hdr->saddr;
544 const struct in6_addr *daddr = &hdr->daddr;
545 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
546 struct sock *sk;
547 int err;
548 struct net *net = dev_net(skb->dev);
549
550 sk = __udp6_lib_lookup(net, daddr, uh->dest,
551 saddr, uh->source, inet6_iif(skb), udptable);
552 if (!sk) {
553 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
554 ICMP6_MIB_INERRORS);
555 return;
556 }
557
558 if (type == ICMPV6_PKT_TOOBIG) {
559 if (!ip6_sk_accept_pmtu(sk))
560 goto out;
561 ip6_sk_update_pmtu(skb, sk, info);
562 }
563 if (type == NDISC_REDIRECT) {
564 ip6_sk_redirect(skb, sk);
565 goto out;
566 }
567
568 np = inet6_sk(sk);
569
570 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
571 goto out;
572
573 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
574 goto out;
575
576 if (np->recverr)
577 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
578
579 sk->sk_err = err;
580 sk->sk_error_report(sk);
581 out:
582 sock_put(sk);
583 }
584
__udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)585 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
586 {
587 int rc;
588
589 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
590 sock_rps_save_rxhash(sk, skb);
591 sk_mark_napi_id(sk, skb);
592 sk_incoming_cpu_update(sk);
593 }
594
595 rc = sock_queue_rcv_skb(sk, skb);
596 if (rc < 0) {
597 int is_udplite = IS_UDPLITE(sk);
598
599 /* Note that an ENOMEM error is charged twice */
600 if (rc == -ENOMEM)
601 UDP6_INC_STATS_BH(sock_net(sk),
602 UDP_MIB_RCVBUFERRORS, is_udplite);
603 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
604 kfree_skb(skb);
605 return -1;
606 }
607 return 0;
608 }
609
udpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)610 static __inline__ void udpv6_err(struct sk_buff *skb,
611 struct inet6_skb_parm *opt, u8 type,
612 u8 code, int offset, __be32 info)
613 {
614 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
615 }
616
617 static struct static_key udpv6_encap_needed __read_mostly;
udpv6_encap_enable(void)618 void udpv6_encap_enable(void)
619 {
620 if (!static_key_enabled(&udpv6_encap_needed))
621 static_key_slow_inc(&udpv6_encap_needed);
622 }
623 EXPORT_SYMBOL(udpv6_encap_enable);
624
udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)625 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
626 {
627 struct udp_sock *up = udp_sk(sk);
628 int rc;
629 int is_udplite = IS_UDPLITE(sk);
630
631 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
632 goto drop;
633
634 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
635 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
636
637 /*
638 * This is an encapsulation socket so pass the skb to
639 * the socket's udp_encap_rcv() hook. Otherwise, just
640 * fall through and pass this up the UDP socket.
641 * up->encap_rcv() returns the following value:
642 * =0 if skb was successfully passed to the encap
643 * handler or was discarded by it.
644 * >0 if skb should be passed on to UDP.
645 * <0 if skb should be resubmitted as proto -N
646 */
647
648 /* if we're overly short, let UDP handle it */
649 encap_rcv = ACCESS_ONCE(up->encap_rcv);
650 if (encap_rcv) {
651 int ret;
652
653 /* Verify checksum before giving to encap */
654 if (udp_lib_checksum_complete(skb))
655 goto csum_error;
656
657 ret = encap_rcv(sk, skb);
658 if (ret <= 0) {
659 UDP_INC_STATS_BH(sock_net(sk),
660 UDP_MIB_INDATAGRAMS,
661 is_udplite);
662 return -ret;
663 }
664 }
665
666 /* FALLTHROUGH -- it's a UDP Packet */
667 }
668
669 /*
670 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
671 */
672 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
673
674 if (up->pcrlen == 0) { /* full coverage was set */
675 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
676 UDP_SKB_CB(skb)->cscov, skb->len);
677 goto drop;
678 }
679 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
680 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
681 UDP_SKB_CB(skb)->cscov, up->pcrlen);
682 goto drop;
683 }
684 }
685
686 if (rcu_access_pointer(sk->sk_filter)) {
687 if (udp_lib_checksum_complete(skb))
688 goto csum_error;
689 }
690
691 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
692 UDP6_INC_STATS_BH(sock_net(sk),
693 UDP_MIB_RCVBUFERRORS, is_udplite);
694 goto drop;
695 }
696
697 skb_dst_drop(skb);
698
699 bh_lock_sock(sk);
700 rc = 0;
701 if (!sock_owned_by_user(sk))
702 rc = __udpv6_queue_rcv_skb(sk, skb);
703 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
704 bh_unlock_sock(sk);
705 goto drop;
706 }
707 bh_unlock_sock(sk);
708
709 return rc;
710
711 csum_error:
712 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
713 drop:
714 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
715 atomic_inc(&sk->sk_drops);
716 kfree_skb(skb);
717 return -1;
718 }
719
__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)720 static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
721 __be16 loc_port, const struct in6_addr *loc_addr,
722 __be16 rmt_port, const struct in6_addr *rmt_addr,
723 int dif, unsigned short hnum)
724 {
725 struct inet_sock *inet = inet_sk(sk);
726
727 if (!net_eq(sock_net(sk), net))
728 return false;
729
730 if (udp_sk(sk)->udp_port_hash != hnum ||
731 sk->sk_family != PF_INET6 ||
732 (inet->inet_dport && inet->inet_dport != rmt_port) ||
733 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
734 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
735 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
736 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
737 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
738 return false;
739 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
740 return false;
741 return true;
742 }
743
flush_stack(struct sock ** stack,unsigned int count,struct sk_buff * skb,unsigned int final)744 static void flush_stack(struct sock **stack, unsigned int count,
745 struct sk_buff *skb, unsigned int final)
746 {
747 struct sk_buff *skb1 = NULL;
748 struct sock *sk;
749 unsigned int i;
750
751 for (i = 0; i < count; i++) {
752 sk = stack[i];
753 if (likely(!skb1))
754 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
755 if (!skb1) {
756 atomic_inc(&sk->sk_drops);
757 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
758 IS_UDPLITE(sk));
759 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
760 IS_UDPLITE(sk));
761 }
762
763 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
764 skb1 = NULL;
765 sock_put(sk);
766 }
767 if (unlikely(skb1))
768 kfree_skb(skb1);
769 }
770
udp6_csum_zero_error(struct sk_buff * skb)771 static void udp6_csum_zero_error(struct sk_buff *skb)
772 {
773 /* RFC 2460 section 8.1 says that we SHOULD log
774 * this error. Well, it is reasonable.
775 */
776 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
777 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
778 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
779 }
780
781 /*
782 * Note: called only from the BH handler context,
783 * so we don't need to lock the hashes.
784 */
__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)785 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
786 const struct in6_addr *saddr, const struct in6_addr *daddr,
787 struct udp_table *udptable, int proto)
788 {
789 struct sock *sk, *stack[256 / sizeof(struct sock *)];
790 const struct udphdr *uh = udp_hdr(skb);
791 struct hlist_nulls_node *node;
792 unsigned short hnum = ntohs(uh->dest);
793 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
794 int dif = inet6_iif(skb);
795 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
796 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
797 bool inner_flushed = false;
798
799 if (use_hash2) {
800 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
801 udp_table.mask;
802 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
803 start_lookup:
804 hslot = &udp_table.hash2[hash2];
805 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
806 }
807
808 spin_lock(&hslot->lock);
809 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
810 if (__udp_v6_is_mcast_sock(net, sk,
811 uh->dest, daddr,
812 uh->source, saddr,
813 dif, hnum) &&
814 /* If zero checksum and no_check is not on for
815 * the socket then skip it.
816 */
817 (uh->check || udp_sk(sk)->no_check6_rx)) {
818 if (unlikely(count == ARRAY_SIZE(stack))) {
819 flush_stack(stack, count, skb, ~0);
820 inner_flushed = true;
821 count = 0;
822 }
823 stack[count++] = sk;
824 sock_hold(sk);
825 }
826 }
827
828 spin_unlock(&hslot->lock);
829
830 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
831 if (use_hash2 && hash2 != hash2_any) {
832 hash2 = hash2_any;
833 goto start_lookup;
834 }
835
836 if (count) {
837 flush_stack(stack, count, skb, count - 1);
838 } else {
839 if (!inner_flushed)
840 UDP6_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
841 proto == IPPROTO_UDPLITE);
842 consume_skb(skb);
843 }
844 return 0;
845 }
846
__udp6_lib_rcv(struct sk_buff * skb,struct udp_table * udptable,int proto)847 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
848 int proto)
849 {
850 struct net *net = dev_net(skb->dev);
851 struct sock *sk;
852 struct udphdr *uh;
853 const struct in6_addr *saddr, *daddr;
854 u32 ulen = 0;
855
856 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
857 goto discard;
858
859 saddr = &ipv6_hdr(skb)->saddr;
860 daddr = &ipv6_hdr(skb)->daddr;
861 uh = udp_hdr(skb);
862
863 ulen = ntohs(uh->len);
864 if (ulen > skb->len)
865 goto short_packet;
866
867 if (proto == IPPROTO_UDP) {
868 /* UDP validates ulen. */
869
870 /* Check for jumbo payload */
871 if (ulen == 0)
872 ulen = skb->len;
873
874 if (ulen < sizeof(*uh))
875 goto short_packet;
876
877 if (ulen < skb->len) {
878 if (pskb_trim_rcsum(skb, ulen))
879 goto short_packet;
880 saddr = &ipv6_hdr(skb)->saddr;
881 daddr = &ipv6_hdr(skb)->daddr;
882 uh = udp_hdr(skb);
883 }
884 }
885
886 if (udp6_csum_init(skb, uh, proto))
887 goto csum_error;
888
889 /*
890 * Multicast receive code
891 */
892 if (ipv6_addr_is_multicast(daddr))
893 return __udp6_lib_mcast_deliver(net, skb,
894 saddr, daddr, udptable, proto);
895
896 /* Unicast */
897
898 /*
899 * check socket cache ... must talk to Alan about his plans
900 * for sock caches... i'll skip this for now.
901 */
902 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
903 if (sk) {
904 int ret;
905
906 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
907 sock_put(sk);
908 udp6_csum_zero_error(skb);
909 goto csum_error;
910 }
911
912 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
913 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
914 ip6_compute_pseudo);
915
916 ret = udpv6_queue_rcv_skb(sk, skb);
917 sock_put(sk);
918
919 /* a return value > 0 means to resubmit the input */
920 if (ret > 0)
921 return ret;
922
923 return 0;
924 }
925
926 if (!uh->check) {
927 udp6_csum_zero_error(skb);
928 goto csum_error;
929 }
930
931 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
932 goto discard;
933
934 if (udp_lib_checksum_complete(skb))
935 goto csum_error;
936
937 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
938 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
939
940 kfree_skb(skb);
941 return 0;
942
943 short_packet:
944 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
945 proto == IPPROTO_UDPLITE ? "-Lite" : "",
946 saddr, ntohs(uh->source),
947 ulen, skb->len,
948 daddr, ntohs(uh->dest));
949 goto discard;
950 csum_error:
951 UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
952 discard:
953 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
954 kfree_skb(skb);
955 return 0;
956 }
957
udpv6_rcv(struct sk_buff * skb)958 static __inline__ int udpv6_rcv(struct sk_buff *skb)
959 {
960 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
961 }
962
963 /*
964 * Throw away all pending data and cancel the corking. Socket is locked.
965 */
udp_v6_flush_pending_frames(struct sock * sk)966 static void udp_v6_flush_pending_frames(struct sock *sk)
967 {
968 struct udp_sock *up = udp_sk(sk);
969
970 if (up->pending == AF_INET)
971 udp_flush_pending_frames(sk);
972 else if (up->pending) {
973 up->len = 0;
974 up->pending = 0;
975 ip6_flush_pending_frames(sk);
976 }
977 }
978
979 /**
980 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
981 * @sk: socket we are sending on
982 * @skb: sk_buff containing the filled-in UDP header
983 * (checksum field must be zeroed out)
984 */
udp6_hwcsum_outgoing(struct sock * sk,struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int len)985 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
986 const struct in6_addr *saddr,
987 const struct in6_addr *daddr, int len)
988 {
989 unsigned int offset;
990 struct udphdr *uh = udp_hdr(skb);
991 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
992 __wsum csum = 0;
993
994 if (!frags) {
995 /* Only one fragment on the socket. */
996 skb->csum_start = skb_transport_header(skb) - skb->head;
997 skb->csum_offset = offsetof(struct udphdr, check);
998 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
999 } else {
1000 /*
1001 * HW-checksum won't work as there are two or more
1002 * fragments on the socket so that all csums of sk_buffs
1003 * should be together
1004 */
1005 offset = skb_transport_offset(skb);
1006 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1007
1008 skb->ip_summed = CHECKSUM_NONE;
1009
1010 do {
1011 csum = csum_add(csum, frags->csum);
1012 } while ((frags = frags->next));
1013
1014 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1015 csum);
1016 if (uh->check == 0)
1017 uh->check = CSUM_MANGLED_0;
1018 }
1019 }
1020
1021 /*
1022 * Sending
1023 */
1024
udp_v6_send_skb(struct sk_buff * skb,struct flowi6 * fl6)1025 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
1026 {
1027 struct sock *sk = skb->sk;
1028 struct udphdr *uh;
1029 int err = 0;
1030 int is_udplite = IS_UDPLITE(sk);
1031 __wsum csum = 0;
1032 int offset = skb_transport_offset(skb);
1033 int len = skb->len - offset;
1034
1035 /*
1036 * Create a UDP header
1037 */
1038 uh = udp_hdr(skb);
1039 uh->source = fl6->fl6_sport;
1040 uh->dest = fl6->fl6_dport;
1041 uh->len = htons(len);
1042 uh->check = 0;
1043
1044 if (is_udplite)
1045 csum = udplite_csum(skb);
1046 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
1047 skb->ip_summed = CHECKSUM_NONE;
1048 goto send;
1049 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1050 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1051 goto send;
1052 } else
1053 csum = udp_csum(skb);
1054
1055 /* add protocol-dependent pseudo-header */
1056 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1057 len, fl6->flowi6_proto, csum);
1058 if (uh->check == 0)
1059 uh->check = CSUM_MANGLED_0;
1060
1061 send:
1062 err = ip6_send_skb(skb);
1063 if (err) {
1064 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1065 UDP6_INC_STATS_USER(sock_net(sk),
1066 UDP_MIB_SNDBUFERRORS, is_udplite);
1067 err = 0;
1068 }
1069 } else
1070 UDP6_INC_STATS_USER(sock_net(sk),
1071 UDP_MIB_OUTDATAGRAMS, is_udplite);
1072 return err;
1073 }
1074
udp_v6_push_pending_frames(struct sock * sk)1075 static int udp_v6_push_pending_frames(struct sock *sk)
1076 {
1077 struct sk_buff *skb;
1078 struct udp_sock *up = udp_sk(sk);
1079 struct flowi6 fl6;
1080 int err = 0;
1081
1082 if (up->pending == AF_INET)
1083 return udp_push_pending_frames(sk);
1084
1085 /* ip6_finish_skb will release the cork, so make a copy of
1086 * fl6 here.
1087 */
1088 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1089
1090 skb = ip6_finish_skb(sk);
1091 if (!skb)
1092 goto out;
1093
1094 err = udp_v6_send_skb(skb, &fl6);
1095
1096 out:
1097 up->len = 0;
1098 up->pending = 0;
1099 return err;
1100 }
1101
udpv6_sendmsg(struct sock * sk,struct msghdr * msg,size_t len)1102 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1103 {
1104 struct ipv6_txoptions opt_space;
1105 struct udp_sock *up = udp_sk(sk);
1106 struct inet_sock *inet = inet_sk(sk);
1107 struct ipv6_pinfo *np = inet6_sk(sk);
1108 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1109 struct in6_addr *daddr, *final_p, final;
1110 struct ipv6_txoptions *opt = NULL;
1111 struct ipv6_txoptions *opt_to_free = NULL;
1112 struct ip6_flowlabel *flowlabel = NULL;
1113 struct flowi6 fl6;
1114 struct dst_entry *dst;
1115 int addr_len = msg->msg_namelen;
1116 int ulen = len;
1117 int hlimit = -1;
1118 int tclass = -1;
1119 int dontfrag = -1;
1120 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1121 int err;
1122 int connected = 0;
1123 int is_udplite = IS_UDPLITE(sk);
1124 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1125
1126 /* destination address check */
1127 if (sin6) {
1128 if (addr_len < offsetof(struct sockaddr, sa_data))
1129 return -EINVAL;
1130
1131 switch (sin6->sin6_family) {
1132 case AF_INET6:
1133 if (addr_len < SIN6_LEN_RFC2133)
1134 return -EINVAL;
1135 daddr = &sin6->sin6_addr;
1136 break;
1137 case AF_INET:
1138 goto do_udp_sendmsg;
1139 case AF_UNSPEC:
1140 msg->msg_name = sin6 = NULL;
1141 msg->msg_namelen = addr_len = 0;
1142 daddr = NULL;
1143 break;
1144 default:
1145 return -EINVAL;
1146 }
1147 } else if (!up->pending) {
1148 if (sk->sk_state != TCP_ESTABLISHED)
1149 return -EDESTADDRREQ;
1150 daddr = &sk->sk_v6_daddr;
1151 } else
1152 daddr = NULL;
1153
1154 if (daddr) {
1155 if (ipv6_addr_v4mapped(daddr)) {
1156 struct sockaddr_in sin;
1157 sin.sin_family = AF_INET;
1158 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1159 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1160 msg->msg_name = &sin;
1161 msg->msg_namelen = sizeof(sin);
1162 do_udp_sendmsg:
1163 if (__ipv6_only_sock(sk))
1164 return -ENETUNREACH;
1165 return udp_sendmsg(sk, msg, len);
1166 }
1167 }
1168
1169 if (up->pending == AF_INET)
1170 return udp_sendmsg(sk, msg, len);
1171
1172 /* Rough check on arithmetic overflow,
1173 better check is made in ip6_append_data().
1174 */
1175 if (len > INT_MAX - sizeof(struct udphdr))
1176 return -EMSGSIZE;
1177
1178 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1179 if (up->pending) {
1180 /*
1181 * There are pending frames.
1182 * The socket lock must be held while it's corked.
1183 */
1184 lock_sock(sk);
1185 if (likely(up->pending)) {
1186 if (unlikely(up->pending != AF_INET6)) {
1187 release_sock(sk);
1188 return -EAFNOSUPPORT;
1189 }
1190 dst = NULL;
1191 goto do_append_data;
1192 }
1193 release_sock(sk);
1194 }
1195 ulen += sizeof(struct udphdr);
1196
1197 memset(&fl6, 0, sizeof(fl6));
1198
1199 if (sin6) {
1200 if (sin6->sin6_port == 0)
1201 return -EINVAL;
1202
1203 fl6.fl6_dport = sin6->sin6_port;
1204 daddr = &sin6->sin6_addr;
1205
1206 if (np->sndflow) {
1207 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1208 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1209 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1210 if (!flowlabel)
1211 return -EINVAL;
1212 }
1213 }
1214
1215 /*
1216 * Otherwise it will be difficult to maintain
1217 * sk->sk_dst_cache.
1218 */
1219 if (sk->sk_state == TCP_ESTABLISHED &&
1220 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1221 daddr = &sk->sk_v6_daddr;
1222
1223 if (addr_len >= sizeof(struct sockaddr_in6) &&
1224 sin6->sin6_scope_id &&
1225 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1226 fl6.flowi6_oif = sin6->sin6_scope_id;
1227 } else {
1228 if (sk->sk_state != TCP_ESTABLISHED)
1229 return -EDESTADDRREQ;
1230
1231 fl6.fl6_dport = inet->inet_dport;
1232 daddr = &sk->sk_v6_daddr;
1233 fl6.flowlabel = np->flow_label;
1234 connected = 1;
1235 }
1236
1237 if (!fl6.flowi6_oif)
1238 fl6.flowi6_oif = sk->sk_bound_dev_if;
1239
1240 if (!fl6.flowi6_oif)
1241 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1242
1243 fl6.flowi6_mark = sk->sk_mark;
1244
1245 if (msg->msg_controllen) {
1246 opt = &opt_space;
1247 memset(opt, 0, sizeof(struct ipv6_txoptions));
1248 opt->tot_len = sizeof(*opt);
1249
1250 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1251 &hlimit, &tclass, &dontfrag);
1252 if (err < 0) {
1253 fl6_sock_release(flowlabel);
1254 return err;
1255 }
1256 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1257 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1258 if (!flowlabel)
1259 return -EINVAL;
1260 }
1261 if (!(opt->opt_nflen|opt->opt_flen))
1262 opt = NULL;
1263 connected = 0;
1264 }
1265 if (!opt) {
1266 opt = txopt_get(np);
1267 opt_to_free = opt;
1268 }
1269 if (flowlabel)
1270 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1271 opt = ipv6_fixup_options(&opt_space, opt);
1272
1273 fl6.flowi6_proto = sk->sk_protocol;
1274 if (!ipv6_addr_any(daddr))
1275 fl6.daddr = *daddr;
1276 else
1277 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1278 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1279 fl6.saddr = np->saddr;
1280 fl6.fl6_sport = inet->inet_sport;
1281
1282 final_p = fl6_update_dst(&fl6, opt, &final);
1283 if (final_p)
1284 connected = 0;
1285
1286 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1287 fl6.flowi6_oif = np->mcast_oif;
1288 connected = 0;
1289 } else if (!fl6.flowi6_oif)
1290 fl6.flowi6_oif = np->ucast_oif;
1291
1292 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1293
1294 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1295 if (IS_ERR(dst)) {
1296 err = PTR_ERR(dst);
1297 dst = NULL;
1298 goto out;
1299 }
1300
1301 if (hlimit < 0)
1302 hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1303
1304 if (tclass < 0)
1305 tclass = np->tclass;
1306
1307 if (msg->msg_flags&MSG_CONFIRM)
1308 goto do_confirm;
1309 back_from_confirm:
1310
1311 /* Lockless fast path for the non-corking case */
1312 if (!corkreq) {
1313 struct sk_buff *skb;
1314
1315 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1316 sizeof(struct udphdr), hlimit, tclass, opt,
1317 &fl6, (struct rt6_info *)dst,
1318 msg->msg_flags, dontfrag);
1319 err = PTR_ERR(skb);
1320 if (!IS_ERR_OR_NULL(skb))
1321 err = udp_v6_send_skb(skb, &fl6);
1322 goto release_dst;
1323 }
1324
1325 lock_sock(sk);
1326 if (unlikely(up->pending)) {
1327 /* The socket is already corked while preparing it. */
1328 /* ... which is an evident application bug. --ANK */
1329 release_sock(sk);
1330
1331 net_dbg_ratelimited("udp cork app bug 2\n");
1332 err = -EINVAL;
1333 goto out;
1334 }
1335
1336 up->pending = AF_INET6;
1337
1338 do_append_data:
1339 if (dontfrag < 0)
1340 dontfrag = np->dontfrag;
1341 up->len += ulen;
1342 err = ip6_append_data(sk, getfrag, msg, ulen,
1343 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1344 (struct rt6_info *)dst,
1345 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1346 if (err)
1347 udp_v6_flush_pending_frames(sk);
1348 else if (!corkreq)
1349 err = udp_v6_push_pending_frames(sk);
1350 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1351 up->pending = 0;
1352
1353 if (err > 0)
1354 err = np->recverr ? net_xmit_errno(err) : 0;
1355 release_sock(sk);
1356
1357 release_dst:
1358 if (dst) {
1359 if (connected) {
1360 ip6_dst_store(sk, dst,
1361 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1362 &sk->sk_v6_daddr : NULL,
1363 #ifdef CONFIG_IPV6_SUBTREES
1364 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1365 &np->saddr :
1366 #endif
1367 NULL);
1368 } else {
1369 dst_release(dst);
1370 }
1371 dst = NULL;
1372 }
1373
1374 out:
1375 dst_release(dst);
1376 fl6_sock_release(flowlabel);
1377 txopt_put(opt_to_free);
1378 if (!err)
1379 return len;
1380 /*
1381 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1382 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1383 * we don't have a good statistic (IpOutDiscards but it can be too many
1384 * things). We could add another new stat but at least for now that
1385 * seems like overkill.
1386 */
1387 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1388 UDP6_INC_STATS_USER(sock_net(sk),
1389 UDP_MIB_SNDBUFERRORS, is_udplite);
1390 }
1391 return err;
1392
1393 do_confirm:
1394 dst_confirm(dst);
1395 if (!(msg->msg_flags&MSG_PROBE) || len)
1396 goto back_from_confirm;
1397 err = 0;
1398 goto out;
1399 }
1400
udpv6_destroy_sock(struct sock * sk)1401 void udpv6_destroy_sock(struct sock *sk)
1402 {
1403 struct udp_sock *up = udp_sk(sk);
1404 lock_sock(sk);
1405 udp_v6_flush_pending_frames(sk);
1406 release_sock(sk);
1407
1408 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1409 void (*encap_destroy)(struct sock *sk);
1410 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1411 if (encap_destroy)
1412 encap_destroy(sk);
1413 }
1414
1415 inet6_destroy_sock(sk);
1416 }
1417
1418 /*
1419 * Socket option code for UDP
1420 */
udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1421 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1422 char __user *optval, unsigned int optlen)
1423 {
1424 if (level == SOL_UDP || level == SOL_UDPLITE)
1425 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1426 udp_v6_push_pending_frames);
1427 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1428 }
1429
1430 #ifdef CONFIG_COMPAT
compat_udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1431 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1432 char __user *optval, unsigned int optlen)
1433 {
1434 if (level == SOL_UDP || level == SOL_UDPLITE)
1435 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1436 udp_v6_push_pending_frames);
1437 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1438 }
1439 #endif
1440
udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1441 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1442 char __user *optval, int __user *optlen)
1443 {
1444 if (level == SOL_UDP || level == SOL_UDPLITE)
1445 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1446 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1447 }
1448
1449 #ifdef CONFIG_COMPAT
compat_udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1450 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1451 char __user *optval, int __user *optlen)
1452 {
1453 if (level == SOL_UDP || level == SOL_UDPLITE)
1454 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1455 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1456 }
1457 #endif
1458
1459 static const struct inet6_protocol udpv6_protocol = {
1460 .handler = udpv6_rcv,
1461 .err_handler = udpv6_err,
1462 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1463 };
1464
1465 /* ------------------------------------------------------------------------ */
1466 #ifdef CONFIG_PROC_FS
udp6_seq_show(struct seq_file * seq,void * v)1467 int udp6_seq_show(struct seq_file *seq, void *v)
1468 {
1469 if (v == SEQ_START_TOKEN) {
1470 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1471 } else {
1472 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1473 struct inet_sock *inet = inet_sk(v);
1474 __u16 srcp = ntohs(inet->inet_sport);
1475 __u16 destp = ntohs(inet->inet_dport);
1476 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1477 }
1478 return 0;
1479 }
1480
1481 static const struct file_operations udp6_afinfo_seq_fops = {
1482 .owner = THIS_MODULE,
1483 .open = udp_seq_open,
1484 .read = seq_read,
1485 .llseek = seq_lseek,
1486 .release = seq_release_net
1487 };
1488
1489 static struct udp_seq_afinfo udp6_seq_afinfo = {
1490 .name = "udp6",
1491 .family = AF_INET6,
1492 .udp_table = &udp_table,
1493 .seq_fops = &udp6_afinfo_seq_fops,
1494 .seq_ops = {
1495 .show = udp6_seq_show,
1496 },
1497 };
1498
udp6_proc_init(struct net * net)1499 int __net_init udp6_proc_init(struct net *net)
1500 {
1501 return udp_proc_register(net, &udp6_seq_afinfo);
1502 }
1503
udp6_proc_exit(struct net * net)1504 void udp6_proc_exit(struct net *net)
1505 {
1506 udp_proc_unregister(net, &udp6_seq_afinfo);
1507 }
1508 #endif /* CONFIG_PROC_FS */
1509
udp_v6_clear_sk(struct sock * sk,int size)1510 void udp_v6_clear_sk(struct sock *sk, int size)
1511 {
1512 struct inet_sock *inet = inet_sk(sk);
1513
1514 /* we do not want to clear pinet6 field, because of RCU lookups */
1515 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1516
1517 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1518 memset(&inet->pinet6 + 1, 0, size);
1519 }
1520
1521 /* ------------------------------------------------------------------------ */
1522
1523 struct proto udpv6_prot = {
1524 .name = "UDPv6",
1525 .owner = THIS_MODULE,
1526 .close = udp_lib_close,
1527 .connect = ip6_datagram_connect,
1528 .disconnect = udp_disconnect,
1529 .ioctl = udp_ioctl,
1530 .destroy = udpv6_destroy_sock,
1531 .setsockopt = udpv6_setsockopt,
1532 .getsockopt = udpv6_getsockopt,
1533 .sendmsg = udpv6_sendmsg,
1534 .recvmsg = udpv6_recvmsg,
1535 .backlog_rcv = __udpv6_queue_rcv_skb,
1536 .hash = udp_lib_hash,
1537 .unhash = udp_lib_unhash,
1538 .rehash = udp_v6_rehash,
1539 .get_port = udp_v6_get_port,
1540 .memory_allocated = &udp_memory_allocated,
1541 .sysctl_mem = sysctl_udp_mem,
1542 .sysctl_wmem = &sysctl_udp_wmem_min,
1543 .sysctl_rmem = &sysctl_udp_rmem_min,
1544 .obj_size = sizeof(struct udp6_sock),
1545 .slab_flags = SLAB_DESTROY_BY_RCU,
1546 .h.udp_table = &udp_table,
1547 #ifdef CONFIG_COMPAT
1548 .compat_setsockopt = compat_udpv6_setsockopt,
1549 .compat_getsockopt = compat_udpv6_getsockopt,
1550 #endif
1551 .clear_sk = udp_v6_clear_sk,
1552 };
1553
1554 static struct inet_protosw udpv6_protosw = {
1555 .type = SOCK_DGRAM,
1556 .protocol = IPPROTO_UDP,
1557 .prot = &udpv6_prot,
1558 .ops = &inet6_dgram_ops,
1559 .flags = INET_PROTOSW_PERMANENT,
1560 };
1561
udpv6_init(void)1562 int __init udpv6_init(void)
1563 {
1564 int ret;
1565
1566 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1567 if (ret)
1568 goto out;
1569
1570 ret = inet6_register_protosw(&udpv6_protosw);
1571 if (ret)
1572 goto out_udpv6_protocol;
1573 out:
1574 return ret;
1575
1576 out_udpv6_protocol:
1577 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1578 goto out;
1579 }
1580
udpv6_exit(void)1581 void udpv6_exit(void)
1582 {
1583 inet6_unregister_protosw(&udpv6_protosw);
1584 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1585 }
1586