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
49 #include <linux/proc_fs.h>
50 #include <linux/seq_file.h>
51 #include "udp_impl.h"
52
ipv6_rcv_saddr_equal(const struct sock * sk,const struct sock * sk2)53 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
54 {
55 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
56 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
57 __be32 sk1_rcv_saddr = sk_rcv_saddr(sk);
58 __be32 sk2_rcv_saddr = sk_rcv_saddr(sk2);
59 int sk_ipv6only = ipv6_only_sock(sk);
60 int sk2_ipv6only = inet_v6_ipv6only(sk2);
61 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
62 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
63
64 /* if both are mapped, treat as IPv4 */
65 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
66 return (!sk2_ipv6only &&
67 (!sk1_rcv_saddr || !sk2_rcv_saddr ||
68 sk1_rcv_saddr == sk2_rcv_saddr));
69
70 if (addr_type2 == IPV6_ADDR_ANY &&
71 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
72 return 1;
73
74 if (addr_type == IPV6_ADDR_ANY &&
75 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
76 return 1;
77
78 if (sk2_rcv_saddr6 &&
79 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
80 return 1;
81
82 return 0;
83 }
84
udp6_portaddr_hash(struct net * net,const struct in6_addr * addr6,unsigned int port)85 static unsigned int udp6_portaddr_hash(struct net *net,
86 const struct in6_addr *addr6,
87 unsigned int port)
88 {
89 unsigned int hash, mix = net_hash_mix(net);
90
91 if (ipv6_addr_any(addr6))
92 hash = jhash_1word(0, mix);
93 else if (ipv6_addr_v4mapped(addr6))
94 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
95 else
96 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
97
98 return hash ^ port;
99 }
100
101
udp_v6_get_port(struct sock * sk,unsigned short snum)102 int udp_v6_get_port(struct sock *sk, unsigned short snum)
103 {
104 unsigned int hash2_nulladdr =
105 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
106 unsigned int hash2_partial =
107 udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0);
108
109 /* precompute partial secondary hash */
110 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
111 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
112 }
113
udp_v6_rehash(struct sock * sk)114 static void udp_v6_rehash(struct sock *sk)
115 {
116 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
117 &inet6_sk(sk)->rcv_saddr,
118 inet_sk(sk)->inet_num);
119
120 udp_lib_rehash(sk, new_hash);
121 }
122
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)123 static inline int compute_score(struct sock *sk, struct net *net,
124 unsigned short hnum,
125 const struct in6_addr *saddr, __be16 sport,
126 const struct in6_addr *daddr, __be16 dport,
127 int dif)
128 {
129 int score = -1;
130
131 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
132 sk->sk_family == PF_INET6) {
133 struct ipv6_pinfo *np = inet6_sk(sk);
134 struct inet_sock *inet = inet_sk(sk);
135
136 score = 0;
137 if (inet->inet_dport) {
138 if (inet->inet_dport != sport)
139 return -1;
140 score++;
141 }
142 if (!ipv6_addr_any(&np->rcv_saddr)) {
143 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
144 return -1;
145 score++;
146 }
147 if (!ipv6_addr_any(&np->daddr)) {
148 if (!ipv6_addr_equal(&np->daddr, saddr))
149 return -1;
150 score++;
151 }
152 if (sk->sk_bound_dev_if) {
153 if (sk->sk_bound_dev_if != dif)
154 return -1;
155 score++;
156 }
157 }
158 return score;
159 }
160
161 #define SCORE2_MAX (1 + 1 + 1)
compute_score2(struct sock * sk,struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,unsigned short hnum,int dif)162 static inline int compute_score2(struct sock *sk, struct net *net,
163 const struct in6_addr *saddr, __be16 sport,
164 const struct in6_addr *daddr, unsigned short hnum,
165 int dif)
166 {
167 int score = -1;
168
169 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
170 sk->sk_family == PF_INET6) {
171 struct ipv6_pinfo *np = inet6_sk(sk);
172 struct inet_sock *inet = inet_sk(sk);
173
174 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
175 return -1;
176 score = 0;
177 if (inet->inet_dport) {
178 if (inet->inet_dport != sport)
179 return -1;
180 score++;
181 }
182 if (!ipv6_addr_any(&np->daddr)) {
183 if (!ipv6_addr_equal(&np->daddr, saddr))
184 return -1;
185 score++;
186 }
187 if (sk->sk_bound_dev_if) {
188 if (sk->sk_bound_dev_if != dif)
189 return -1;
190 score++;
191 }
192 }
193 return score;
194 }
195
196
197 /* 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)198 static struct sock *udp6_lib_lookup2(struct net *net,
199 const struct in6_addr *saddr, __be16 sport,
200 const struct in6_addr *daddr, unsigned int hnum, int dif,
201 struct udp_hslot *hslot2, unsigned int slot2)
202 {
203 struct sock *sk, *result;
204 struct hlist_nulls_node *node;
205 int score, badness;
206
207 begin:
208 result = NULL;
209 badness = -1;
210 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
211 score = compute_score2(sk, net, saddr, sport,
212 daddr, hnum, dif);
213 if (score > badness) {
214 result = sk;
215 badness = score;
216 if (score == SCORE2_MAX)
217 goto exact_match;
218 }
219 }
220 /*
221 * if the nulls value we got at the end of this lookup is
222 * not the expected one, we must restart lookup.
223 * We probably met an item that was moved to another chain.
224 */
225 if (get_nulls_value(node) != slot2)
226 goto begin;
227
228 if (result) {
229 exact_match:
230 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
231 result = NULL;
232 else if (unlikely(compute_score2(result, net, saddr, sport,
233 daddr, hnum, dif) < badness)) {
234 sock_put(result);
235 goto begin;
236 }
237 }
238 return result;
239 }
240
__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)241 struct sock *__udp6_lib_lookup(struct net *net,
242 const struct in6_addr *saddr, __be16 sport,
243 const struct in6_addr *daddr, __be16 dport,
244 int dif, struct udp_table *udptable)
245 {
246 struct sock *sk, *result;
247 struct hlist_nulls_node *node;
248 unsigned short hnum = ntohs(dport);
249 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
250 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
251 int score, badness;
252
253 rcu_read_lock();
254 if (hslot->count > 10) {
255 hash2 = udp6_portaddr_hash(net, daddr, hnum);
256 slot2 = hash2 & udptable->mask;
257 hslot2 = &udptable->hash2[slot2];
258 if (hslot->count < hslot2->count)
259 goto begin;
260
261 result = udp6_lib_lookup2(net, saddr, sport,
262 daddr, hnum, dif,
263 hslot2, slot2);
264 if (!result) {
265 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
266 slot2 = hash2 & udptable->mask;
267 hslot2 = &udptable->hash2[slot2];
268 if (hslot->count < hslot2->count)
269 goto begin;
270
271 result = udp6_lib_lookup2(net, saddr, sport,
272 &in6addr_any, hnum, dif,
273 hslot2, slot2);
274 }
275 rcu_read_unlock();
276 return result;
277 }
278 begin:
279 result = NULL;
280 badness = -1;
281 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
282 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
283 if (score > badness) {
284 result = sk;
285 badness = score;
286 }
287 }
288 /*
289 * if the nulls value we got at the end of this lookup is
290 * not the expected one, we must restart lookup.
291 * We probably met an item that was moved to another chain.
292 */
293 if (get_nulls_value(node) != slot)
294 goto begin;
295
296 if (result) {
297 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
298 result = NULL;
299 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
300 daddr, dport, dif) < badness)) {
301 sock_put(result);
302 goto begin;
303 }
304 }
305 rcu_read_unlock();
306 return result;
307 }
308 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
309
__udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport,struct udp_table * udptable)310 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
311 __be16 sport, __be16 dport,
312 struct udp_table *udptable)
313 {
314 struct sock *sk;
315 const struct ipv6hdr *iph = ipv6_hdr(skb);
316
317 if (unlikely(sk = skb_steal_sock(skb)))
318 return sk;
319 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
320 &iph->daddr, dport, inet6_iif(skb),
321 udptable);
322 }
323
udp6_lib_lookup(struct net * net,const struct in6_addr * saddr,__be16 sport,const struct in6_addr * daddr,__be16 dport,int dif)324 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
325 const struct in6_addr *daddr, __be16 dport, int dif)
326 {
327 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
328 }
329 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
330
331
332 /*
333 * This should be easy, if there is something there we
334 * return it, otherwise we block.
335 */
336
udpv6_recvmsg(struct kiocb * iocb,struct sock * sk,struct msghdr * msg,size_t len,int noblock,int flags,int * addr_len)337 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
338 struct msghdr *msg, size_t len,
339 int noblock, int flags, int *addr_len)
340 {
341 struct ipv6_pinfo *np = inet6_sk(sk);
342 struct inet_sock *inet = inet_sk(sk);
343 struct sk_buff *skb;
344 unsigned int ulen, copied;
345 int peeked, off = 0;
346 int err;
347 int is_udplite = IS_UDPLITE(sk);
348 int is_udp4;
349 bool slow;
350
351 if (flags & MSG_ERRQUEUE)
352 return ipv6_recv_error(sk, msg, len, addr_len);
353
354 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
355 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
356
357 try_again:
358 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
359 &peeked, &off, &err);
360 if (!skb)
361 goto out;
362
363 ulen = skb->len - sizeof(struct udphdr);
364 copied = len;
365 if (copied > ulen)
366 copied = ulen;
367 else if (copied < ulen)
368 msg->msg_flags |= MSG_TRUNC;
369
370 is_udp4 = (skb->protocol == htons(ETH_P_IP));
371
372 /*
373 * If checksum is needed at all, try to do it while copying the
374 * data. If the data is truncated, or if we only want a partial
375 * coverage checksum (UDP-Lite), do it before the copy.
376 */
377
378 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
379 if (udp_lib_checksum_complete(skb))
380 goto csum_copy_err;
381 }
382
383 if (skb_csum_unnecessary(skb))
384 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
385 msg->msg_iov, copied );
386 else {
387 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
388 if (err == -EINVAL)
389 goto csum_copy_err;
390 }
391 if (err)
392 goto out_free;
393
394 if (!peeked) {
395 if (is_udp4)
396 UDP_INC_STATS_USER(sock_net(sk),
397 UDP_MIB_INDATAGRAMS, is_udplite);
398 else
399 UDP6_INC_STATS_USER(sock_net(sk),
400 UDP_MIB_INDATAGRAMS, is_udplite);
401 }
402
403 sock_recv_ts_and_drops(msg, sk, skb);
404
405 /* Copy the address. */
406 if (msg->msg_name) {
407 struct sockaddr_in6 *sin6;
408
409 sin6 = (struct sockaddr_in6 *) msg->msg_name;
410 sin6->sin6_family = AF_INET6;
411 sin6->sin6_port = udp_hdr(skb)->source;
412 sin6->sin6_flowinfo = 0;
413 sin6->sin6_scope_id = 0;
414
415 if (is_udp4)
416 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
417 &sin6->sin6_addr);
418 else {
419 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
420 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
421 sin6->sin6_scope_id = IP6CB(skb)->iif;
422 }
423 *addr_len = sizeof(*sin6);
424 }
425 if (is_udp4) {
426 if (inet->cmsg_flags)
427 ip_cmsg_recv(msg, skb);
428 } else {
429 if (np->rxopt.all)
430 datagram_recv_ctl(sk, msg, skb);
431 }
432
433 err = copied;
434 if (flags & MSG_TRUNC)
435 err = ulen;
436
437 out_free:
438 skb_free_datagram_locked(sk, skb);
439 out:
440 return err;
441
442 csum_copy_err:
443 slow = lock_sock_fast(sk);
444 if (!skb_kill_datagram(sk, skb, flags)) {
445 if (is_udp4)
446 UDP_INC_STATS_USER(sock_net(sk),
447 UDP_MIB_INERRORS, is_udplite);
448 else
449 UDP6_INC_STATS_USER(sock_net(sk),
450 UDP_MIB_INERRORS, is_udplite);
451 }
452 unlock_sock_fast(sk, slow);
453
454 if (noblock)
455 return -EAGAIN;
456
457 /* starting over for a new packet */
458 msg->msg_flags &= ~MSG_TRUNC;
459 goto try_again;
460 }
461
__udp6_lib_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info,struct udp_table * udptable)462 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
463 u8 type, u8 code, int offset, __be32 info,
464 struct udp_table *udptable)
465 {
466 struct ipv6_pinfo *np;
467 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
468 const struct in6_addr *saddr = &hdr->saddr;
469 const struct in6_addr *daddr = &hdr->daddr;
470 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
471 struct sock *sk;
472 int err;
473
474 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
475 saddr, uh->source, inet6_iif(skb), udptable);
476 if (sk == NULL)
477 return;
478
479 np = inet6_sk(sk);
480
481 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
482 goto out;
483
484 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
485 goto out;
486
487 if (np->recverr)
488 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
489
490 sk->sk_err = err;
491 sk->sk_error_report(sk);
492 out:
493 sock_put(sk);
494 }
495
udpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)496 static __inline__ void udpv6_err(struct sk_buff *skb,
497 struct inet6_skb_parm *opt, u8 type,
498 u8 code, int offset, __be32 info )
499 {
500 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
501 }
502
udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)503 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
504 {
505 struct udp_sock *up = udp_sk(sk);
506 int rc;
507 int is_udplite = IS_UDPLITE(sk);
508
509 if (!ipv6_addr_any(&inet6_sk(sk)->daddr))
510 sock_rps_save_rxhash(sk, skb);
511
512 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
513 goto drop;
514
515 /*
516 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
517 */
518 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
519
520 if (up->pcrlen == 0) { /* full coverage was set */
521 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
522 " %d while full coverage %d requested\n",
523 UDP_SKB_CB(skb)->cscov, skb->len);
524 goto drop;
525 }
526 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
527 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
528 "too small, need min %d\n",
529 UDP_SKB_CB(skb)->cscov, up->pcrlen);
530 goto drop;
531 }
532 }
533
534 if (rcu_access_pointer(sk->sk_filter)) {
535 if (udp_lib_checksum_complete(skb))
536 goto drop;
537 }
538
539 skb_dst_drop(skb);
540 rc = sock_queue_rcv_skb(sk, skb);
541 if (rc < 0) {
542 /* Note that an ENOMEM error is charged twice */
543 if (rc == -ENOMEM)
544 UDP6_INC_STATS_BH(sock_net(sk),
545 UDP_MIB_RCVBUFERRORS, is_udplite);
546 goto drop_no_sk_drops_inc;
547 }
548
549 return 0;
550 drop:
551 atomic_inc(&sk->sk_drops);
552 drop_no_sk_drops_inc:
553 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
554 kfree_skb(skb);
555 return -1;
556 }
557
udp_v6_mcast_next(struct net * net,struct sock * sk,__be16 loc_port,const struct in6_addr * loc_addr,__be16 rmt_port,const struct in6_addr * rmt_addr,int dif)558 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
559 __be16 loc_port, const struct in6_addr *loc_addr,
560 __be16 rmt_port, const struct in6_addr *rmt_addr,
561 int dif)
562 {
563 struct hlist_nulls_node *node;
564 struct sock *s = sk;
565 unsigned short num = ntohs(loc_port);
566
567 sk_nulls_for_each_from(s, node) {
568 struct inet_sock *inet = inet_sk(s);
569
570 if (!net_eq(sock_net(s), net))
571 continue;
572
573 if (udp_sk(s)->udp_port_hash == num &&
574 s->sk_family == PF_INET6) {
575 struct ipv6_pinfo *np = inet6_sk(s);
576 if (inet->inet_dport) {
577 if (inet->inet_dport != rmt_port)
578 continue;
579 }
580 if (!ipv6_addr_any(&np->daddr) &&
581 !ipv6_addr_equal(&np->daddr, rmt_addr))
582 continue;
583
584 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
585 continue;
586
587 if (!ipv6_addr_any(&np->rcv_saddr)) {
588 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
589 continue;
590 }
591 if (!inet6_mc_check(s, loc_addr, rmt_addr))
592 continue;
593 return s;
594 }
595 }
596 return NULL;
597 }
598
flush_stack(struct sock ** stack,unsigned int count,struct sk_buff * skb,unsigned int final)599 static void flush_stack(struct sock **stack, unsigned int count,
600 struct sk_buff *skb, unsigned int final)
601 {
602 unsigned int i;
603 struct sock *sk;
604 struct sk_buff *skb1;
605
606 for (i = 0; i < count; i++) {
607 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
608
609 sk = stack[i];
610 if (skb1) {
611 if (sk_rcvqueues_full(sk, skb1)) {
612 kfree_skb(skb1);
613 goto drop;
614 }
615 bh_lock_sock(sk);
616 if (!sock_owned_by_user(sk))
617 udpv6_queue_rcv_skb(sk, skb1);
618 else if (sk_add_backlog(sk, skb1)) {
619 kfree_skb(skb1);
620 bh_unlock_sock(sk);
621 goto drop;
622 }
623 bh_unlock_sock(sk);
624 continue;
625 }
626 drop:
627 atomic_inc(&sk->sk_drops);
628 UDP6_INC_STATS_BH(sock_net(sk),
629 UDP_MIB_RCVBUFERRORS, IS_UDPLITE(sk));
630 UDP6_INC_STATS_BH(sock_net(sk),
631 UDP_MIB_INERRORS, IS_UDPLITE(sk));
632 }
633 }
634 /*
635 * Note: called only from the BH handler context,
636 * so we don't need to lock the hashes.
637 */
__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)638 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
639 const struct in6_addr *saddr, const struct in6_addr *daddr,
640 struct udp_table *udptable)
641 {
642 struct sock *sk, *stack[256 / sizeof(struct sock *)];
643 const struct udphdr *uh = udp_hdr(skb);
644 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
645 int dif;
646 unsigned int i, count = 0;
647
648 spin_lock(&hslot->lock);
649 sk = sk_nulls_head(&hslot->head);
650 dif = inet6_iif(skb);
651 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
652 while (sk) {
653 stack[count++] = sk;
654 sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
655 uh->source, saddr, dif);
656 if (unlikely(count == ARRAY_SIZE(stack))) {
657 if (!sk)
658 break;
659 flush_stack(stack, count, skb, ~0);
660 count = 0;
661 }
662 }
663 /*
664 * before releasing the lock, we must take reference on sockets
665 */
666 for (i = 0; i < count; i++)
667 sock_hold(stack[i]);
668
669 spin_unlock(&hslot->lock);
670
671 if (count) {
672 flush_stack(stack, count, skb, count - 1);
673
674 for (i = 0; i < count; i++)
675 sock_put(stack[i]);
676 } else {
677 kfree_skb(skb);
678 }
679 return 0;
680 }
681
udp6_csum_init(struct sk_buff * skb,struct udphdr * uh,int proto)682 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
683 int proto)
684 {
685 int err;
686
687 UDP_SKB_CB(skb)->partial_cov = 0;
688 UDP_SKB_CB(skb)->cscov = skb->len;
689
690 if (proto == IPPROTO_UDPLITE) {
691 err = udplite_checksum_init(skb, uh);
692 if (err)
693 return err;
694 }
695
696 if (uh->check == 0) {
697 /* RFC 2460 section 8.1 says that we SHOULD log
698 this error. Well, it is reasonable.
699 */
700 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
701 return 1;
702 }
703 if (skb->ip_summed == CHECKSUM_COMPLETE &&
704 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
705 skb->len, proto, skb->csum))
706 skb->ip_summed = CHECKSUM_UNNECESSARY;
707
708 if (!skb_csum_unnecessary(skb))
709 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
710 &ipv6_hdr(skb)->daddr,
711 skb->len, proto, 0));
712
713 return 0;
714 }
715
__udp6_lib_rcv(struct sk_buff * skb,struct udp_table * udptable,int proto)716 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
717 int proto)
718 {
719 struct net *net = dev_net(skb->dev);
720 struct sock *sk;
721 struct udphdr *uh;
722 const struct in6_addr *saddr, *daddr;
723 u32 ulen = 0;
724
725 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
726 goto discard;
727
728 saddr = &ipv6_hdr(skb)->saddr;
729 daddr = &ipv6_hdr(skb)->daddr;
730 uh = udp_hdr(skb);
731
732 ulen = ntohs(uh->len);
733 if (ulen > skb->len)
734 goto short_packet;
735
736 if (proto == IPPROTO_UDP) {
737 /* UDP validates ulen. */
738
739 /* Check for jumbo payload */
740 if (ulen == 0)
741 ulen = skb->len;
742
743 if (ulen < sizeof(*uh))
744 goto short_packet;
745
746 if (ulen < skb->len) {
747 if (pskb_trim_rcsum(skb, ulen))
748 goto short_packet;
749 saddr = &ipv6_hdr(skb)->saddr;
750 daddr = &ipv6_hdr(skb)->daddr;
751 uh = udp_hdr(skb);
752 }
753 }
754
755 if (udp6_csum_init(skb, uh, proto))
756 goto discard;
757
758 /*
759 * Multicast receive code
760 */
761 if (ipv6_addr_is_multicast(daddr))
762 return __udp6_lib_mcast_deliver(net, skb,
763 saddr, daddr, udptable);
764
765 /* Unicast */
766
767 /*
768 * check socket cache ... must talk to Alan about his plans
769 * for sock caches... i'll skip this for now.
770 */
771 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
772
773 if (sk == NULL) {
774 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
775 goto discard;
776
777 if (udp_lib_checksum_complete(skb))
778 goto discard;
779 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS,
780 proto == IPPROTO_UDPLITE);
781
782 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
783
784 kfree_skb(skb);
785 return 0;
786 }
787
788 /* deliver */
789
790 if (sk_rcvqueues_full(sk, skb)) {
791 sock_put(sk);
792 goto discard;
793 }
794 bh_lock_sock(sk);
795 if (!sock_owned_by_user(sk))
796 udpv6_queue_rcv_skb(sk, skb);
797 else if (sk_add_backlog(sk, skb)) {
798 atomic_inc(&sk->sk_drops);
799 bh_unlock_sock(sk);
800 sock_put(sk);
801 goto discard;
802 }
803 bh_unlock_sock(sk);
804 sock_put(sk);
805 return 0;
806
807 short_packet:
808 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
809 proto == IPPROTO_UDPLITE ? "-Lite" : "",
810 saddr,
811 ntohs(uh->source),
812 ulen,
813 skb->len,
814 daddr,
815 ntohs(uh->dest));
816
817 discard:
818 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
819 kfree_skb(skb);
820 return 0;
821 }
822
udpv6_rcv(struct sk_buff * skb)823 static __inline__ int udpv6_rcv(struct sk_buff *skb)
824 {
825 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
826 }
827
828 /*
829 * Throw away all pending data and cancel the corking. Socket is locked.
830 */
udp_v6_flush_pending_frames(struct sock * sk)831 static void udp_v6_flush_pending_frames(struct sock *sk)
832 {
833 struct udp_sock *up = udp_sk(sk);
834
835 if (up->pending == AF_INET)
836 udp_flush_pending_frames(sk);
837 else if (up->pending) {
838 up->len = 0;
839 up->pending = 0;
840 ip6_flush_pending_frames(sk);
841 }
842 }
843
844 /**
845 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
846 * @sk: socket we are sending on
847 * @skb: sk_buff containing the filled-in UDP header
848 * (checksum field must be zeroed out)
849 */
udp6_hwcsum_outgoing(struct sock * sk,struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int len)850 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
851 const struct in6_addr *saddr,
852 const struct in6_addr *daddr, int len)
853 {
854 unsigned int offset;
855 struct udphdr *uh = udp_hdr(skb);
856 __wsum csum = 0;
857
858 if (skb_queue_len(&sk->sk_write_queue) == 1) {
859 /* Only one fragment on the socket. */
860 skb->csum_start = skb_transport_header(skb) - skb->head;
861 skb->csum_offset = offsetof(struct udphdr, check);
862 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
863 } else {
864 /*
865 * HW-checksum won't work as there are two or more
866 * fragments on the socket so that all csums of sk_buffs
867 * should be together
868 */
869 offset = skb_transport_offset(skb);
870 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
871
872 skb->ip_summed = CHECKSUM_NONE;
873
874 skb_queue_walk(&sk->sk_write_queue, skb) {
875 csum = csum_add(csum, skb->csum);
876 }
877
878 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
879 csum);
880 if (uh->check == 0)
881 uh->check = CSUM_MANGLED_0;
882 }
883 }
884
885 /*
886 * Sending
887 */
888
udp_v6_push_pending_frames(struct sock * sk)889 static int udp_v6_push_pending_frames(struct sock *sk)
890 {
891 struct sk_buff *skb;
892 struct udphdr *uh;
893 struct udp_sock *up = udp_sk(sk);
894 struct inet_sock *inet = inet_sk(sk);
895 struct flowi6 *fl6;
896 int err = 0;
897 int is_udplite = IS_UDPLITE(sk);
898 __wsum csum = 0;
899
900 if (up->pending == AF_INET)
901 return udp_push_pending_frames(sk);
902
903 fl6 = &inet->cork.fl.u.ip6;
904
905 /* Grab the skbuff where UDP header space exists. */
906 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
907 goto out;
908
909 /*
910 * Create a UDP header
911 */
912 uh = udp_hdr(skb);
913 uh->source = fl6->fl6_sport;
914 uh->dest = fl6->fl6_dport;
915 uh->len = htons(up->len);
916 uh->check = 0;
917
918 if (is_udplite)
919 csum = udplite_csum_outgoing(sk, skb);
920 else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
921 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
922 up->len);
923 goto send;
924 } else
925 csum = udp_csum_outgoing(sk, skb);
926
927 /* add protocol-dependent pseudo-header */
928 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
929 up->len, fl6->flowi6_proto, csum);
930 if (uh->check == 0)
931 uh->check = CSUM_MANGLED_0;
932
933 send:
934 err = ip6_push_pending_frames(sk);
935 if (err) {
936 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
937 UDP6_INC_STATS_USER(sock_net(sk),
938 UDP_MIB_SNDBUFERRORS, is_udplite);
939 err = 0;
940 }
941 } else
942 UDP6_INC_STATS_USER(sock_net(sk),
943 UDP_MIB_OUTDATAGRAMS, is_udplite);
944 out:
945 up->len = 0;
946 up->pending = 0;
947 return err;
948 }
949
udpv6_sendmsg(struct kiocb * iocb,struct sock * sk,struct msghdr * msg,size_t len)950 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
951 struct msghdr *msg, size_t len)
952 {
953 struct ipv6_txoptions opt_space;
954 struct udp_sock *up = udp_sk(sk);
955 struct inet_sock *inet = inet_sk(sk);
956 struct ipv6_pinfo *np = inet6_sk(sk);
957 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
958 struct in6_addr *daddr, *final_p, final;
959 struct ipv6_txoptions *opt = NULL;
960 struct ip6_flowlabel *flowlabel = NULL;
961 struct flowi6 fl6;
962 struct dst_entry *dst;
963 int addr_len = msg->msg_namelen;
964 int ulen = len;
965 int hlimit = -1;
966 int tclass = -1;
967 int dontfrag = -1;
968 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
969 int err;
970 int connected = 0;
971 int is_udplite = IS_UDPLITE(sk);
972 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
973
974 /* destination address check */
975 if (sin6) {
976 if (addr_len < offsetof(struct sockaddr, sa_data))
977 return -EINVAL;
978
979 switch (sin6->sin6_family) {
980 case AF_INET6:
981 if (addr_len < SIN6_LEN_RFC2133)
982 return -EINVAL;
983 daddr = &sin6->sin6_addr;
984 break;
985 case AF_INET:
986 goto do_udp_sendmsg;
987 case AF_UNSPEC:
988 msg->msg_name = sin6 = NULL;
989 msg->msg_namelen = addr_len = 0;
990 daddr = NULL;
991 break;
992 default:
993 return -EINVAL;
994 }
995 } else if (!up->pending) {
996 if (sk->sk_state != TCP_ESTABLISHED)
997 return -EDESTADDRREQ;
998 daddr = &np->daddr;
999 } else
1000 daddr = NULL;
1001
1002 if (daddr) {
1003 if (ipv6_addr_v4mapped(daddr)) {
1004 struct sockaddr_in sin;
1005 sin.sin_family = AF_INET;
1006 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1007 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1008 msg->msg_name = &sin;
1009 msg->msg_namelen = sizeof(sin);
1010 do_udp_sendmsg:
1011 if (__ipv6_only_sock(sk))
1012 return -ENETUNREACH;
1013 return udp_sendmsg(iocb, sk, msg, len);
1014 }
1015 }
1016
1017 if (up->pending == AF_INET)
1018 return udp_sendmsg(iocb, sk, msg, len);
1019
1020 /* Rough check on arithmetic overflow,
1021 better check is made in ip6_append_data().
1022 */
1023 if (len > INT_MAX - sizeof(struct udphdr))
1024 return -EMSGSIZE;
1025
1026 if (up->pending) {
1027 /*
1028 * There are pending frames.
1029 * The socket lock must be held while it's corked.
1030 */
1031 lock_sock(sk);
1032 if (likely(up->pending)) {
1033 if (unlikely(up->pending != AF_INET6)) {
1034 release_sock(sk);
1035 return -EAFNOSUPPORT;
1036 }
1037 dst = NULL;
1038 goto do_append_data;
1039 }
1040 release_sock(sk);
1041 }
1042 ulen += sizeof(struct udphdr);
1043
1044 memset(&fl6, 0, sizeof(fl6));
1045
1046 if (sin6) {
1047 if (sin6->sin6_port == 0)
1048 return -EINVAL;
1049
1050 fl6.fl6_dport = sin6->sin6_port;
1051 daddr = &sin6->sin6_addr;
1052
1053 if (np->sndflow) {
1054 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1055 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1056 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1057 if (flowlabel == NULL)
1058 return -EINVAL;
1059 daddr = &flowlabel->dst;
1060 }
1061 }
1062
1063 /*
1064 * Otherwise it will be difficult to maintain
1065 * sk->sk_dst_cache.
1066 */
1067 if (sk->sk_state == TCP_ESTABLISHED &&
1068 ipv6_addr_equal(daddr, &np->daddr))
1069 daddr = &np->daddr;
1070
1071 if (addr_len >= sizeof(struct sockaddr_in6) &&
1072 sin6->sin6_scope_id &&
1073 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
1074 fl6.flowi6_oif = sin6->sin6_scope_id;
1075 } else {
1076 if (sk->sk_state != TCP_ESTABLISHED)
1077 return -EDESTADDRREQ;
1078
1079 fl6.fl6_dport = inet->inet_dport;
1080 daddr = &np->daddr;
1081 fl6.flowlabel = np->flow_label;
1082 connected = 1;
1083 }
1084
1085 if (!fl6.flowi6_oif)
1086 fl6.flowi6_oif = sk->sk_bound_dev_if;
1087
1088 if (!fl6.flowi6_oif)
1089 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1090
1091 fl6.flowi6_mark = sk->sk_mark;
1092
1093 if (msg->msg_controllen) {
1094 opt = &opt_space;
1095 memset(opt, 0, sizeof(struct ipv6_txoptions));
1096 opt->tot_len = sizeof(*opt);
1097
1098 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1099 &hlimit, &tclass, &dontfrag);
1100 if (err < 0) {
1101 fl6_sock_release(flowlabel);
1102 return err;
1103 }
1104 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1105 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1106 if (flowlabel == NULL)
1107 return -EINVAL;
1108 }
1109 if (!(opt->opt_nflen|opt->opt_flen))
1110 opt = NULL;
1111 connected = 0;
1112 }
1113 if (opt == NULL)
1114 opt = np->opt;
1115 if (flowlabel)
1116 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1117 opt = ipv6_fixup_options(&opt_space, opt);
1118
1119 fl6.flowi6_proto = sk->sk_protocol;
1120 if (!ipv6_addr_any(daddr))
1121 fl6.daddr = *daddr;
1122 else
1123 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1124 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1125 fl6.saddr = np->saddr;
1126 fl6.fl6_sport = inet->inet_sport;
1127
1128 final_p = fl6_update_dst(&fl6, opt, &final);
1129 if (final_p)
1130 connected = 0;
1131
1132 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1133 fl6.flowi6_oif = np->mcast_oif;
1134 connected = 0;
1135 } else if (!fl6.flowi6_oif)
1136 fl6.flowi6_oif = np->ucast_oif;
1137
1138 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1139
1140 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
1141 if (IS_ERR(dst)) {
1142 err = PTR_ERR(dst);
1143 dst = NULL;
1144 goto out;
1145 }
1146
1147 if (hlimit < 0) {
1148 if (ipv6_addr_is_multicast(&fl6.daddr))
1149 hlimit = np->mcast_hops;
1150 else
1151 hlimit = np->hop_limit;
1152 if (hlimit < 0)
1153 hlimit = ip6_dst_hoplimit(dst);
1154 }
1155
1156 if (tclass < 0)
1157 tclass = np->tclass;
1158
1159 if (dontfrag < 0)
1160 dontfrag = np->dontfrag;
1161
1162 if (msg->msg_flags&MSG_CONFIRM)
1163 goto do_confirm;
1164 back_from_confirm:
1165
1166 lock_sock(sk);
1167 if (unlikely(up->pending)) {
1168 /* The socket is already corked while preparing it. */
1169 /* ... which is an evident application bug. --ANK */
1170 release_sock(sk);
1171
1172 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1173 err = -EINVAL;
1174 goto out;
1175 }
1176
1177 up->pending = AF_INET6;
1178
1179 do_append_data:
1180 up->len += ulen;
1181 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1182 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
1183 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1184 (struct rt6_info*)dst,
1185 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1186 if (err)
1187 udp_v6_flush_pending_frames(sk);
1188 else if (!corkreq)
1189 err = udp_v6_push_pending_frames(sk);
1190 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1191 up->pending = 0;
1192
1193 if (dst) {
1194 if (connected) {
1195 ip6_dst_store(sk, dst,
1196 ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
1197 &np->daddr : NULL,
1198 #ifdef CONFIG_IPV6_SUBTREES
1199 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1200 &np->saddr :
1201 #endif
1202 NULL);
1203 } else {
1204 dst_release(dst);
1205 }
1206 dst = NULL;
1207 }
1208
1209 if (err > 0)
1210 err = np->recverr ? net_xmit_errno(err) : 0;
1211 release_sock(sk);
1212 out:
1213 dst_release(dst);
1214 fl6_sock_release(flowlabel);
1215 if (!err)
1216 return len;
1217 /*
1218 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1219 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1220 * we don't have a good statistic (IpOutDiscards but it can be too many
1221 * things). We could add another new stat but at least for now that
1222 * seems like overkill.
1223 */
1224 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1225 UDP6_INC_STATS_USER(sock_net(sk),
1226 UDP_MIB_SNDBUFERRORS, is_udplite);
1227 }
1228 return err;
1229
1230 do_confirm:
1231 dst_confirm(dst);
1232 if (!(msg->msg_flags&MSG_PROBE) || len)
1233 goto back_from_confirm;
1234 err = 0;
1235 goto out;
1236 }
1237
udpv6_destroy_sock(struct sock * sk)1238 void udpv6_destroy_sock(struct sock *sk)
1239 {
1240 lock_sock(sk);
1241 udp_v6_flush_pending_frames(sk);
1242 release_sock(sk);
1243
1244 inet6_destroy_sock(sk);
1245 }
1246
1247 /*
1248 * Socket option code for UDP
1249 */
udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1250 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1251 char __user *optval, unsigned int optlen)
1252 {
1253 if (level == SOL_UDP || level == SOL_UDPLITE)
1254 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1255 udp_v6_push_pending_frames);
1256 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1257 }
1258
1259 #ifdef CONFIG_COMPAT
compat_udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,unsigned int optlen)1260 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1261 char __user *optval, unsigned int optlen)
1262 {
1263 if (level == SOL_UDP || level == SOL_UDPLITE)
1264 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1265 udp_v6_push_pending_frames);
1266 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1267 }
1268 #endif
1269
udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1270 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1271 char __user *optval, int __user *optlen)
1272 {
1273 if (level == SOL_UDP || level == SOL_UDPLITE)
1274 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1275 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1276 }
1277
1278 #ifdef CONFIG_COMPAT
compat_udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1279 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1280 char __user *optval, int __user *optlen)
1281 {
1282 if (level == SOL_UDP || level == SOL_UDPLITE)
1283 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1284 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1285 }
1286 #endif
1287
udp6_ufo_send_check(struct sk_buff * skb)1288 static int udp6_ufo_send_check(struct sk_buff *skb)
1289 {
1290 const struct ipv6hdr *ipv6h;
1291 struct udphdr *uh;
1292
1293 if (!pskb_may_pull(skb, sizeof(*uh)))
1294 return -EINVAL;
1295
1296 ipv6h = ipv6_hdr(skb);
1297 uh = udp_hdr(skb);
1298
1299 uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
1300 IPPROTO_UDP, 0);
1301 skb->csum_start = skb_transport_header(skb) - skb->head;
1302 skb->csum_offset = offsetof(struct udphdr, check);
1303 skb->ip_summed = CHECKSUM_PARTIAL;
1304 return 0;
1305 }
1306
udp6_ufo_fragment(struct sk_buff * skb,netdev_features_t features)1307 static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
1308 netdev_features_t features)
1309 {
1310 struct sk_buff *segs = ERR_PTR(-EINVAL);
1311 unsigned int mss;
1312 unsigned int unfrag_ip6hlen, unfrag_len;
1313 struct frag_hdr *fptr;
1314 u8 *mac_start, *prevhdr;
1315 u8 nexthdr;
1316 u8 frag_hdr_sz = sizeof(struct frag_hdr);
1317 int offset;
1318 __wsum csum;
1319
1320 mss = skb_shinfo(skb)->gso_size;
1321 if (unlikely(skb->len <= mss))
1322 goto out;
1323
1324 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) {
1325 /* Packet is from an untrusted source, reset gso_segs. */
1326 int type = skb_shinfo(skb)->gso_type;
1327
1328 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) ||
1329 !(type & (SKB_GSO_UDP))))
1330 goto out;
1331
1332 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss);
1333
1334 segs = NULL;
1335 goto out;
1336 }
1337
1338 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot
1339 * do checksum of UDP packets sent as multiple IP fragments.
1340 */
1341 offset = skb_checksum_start_offset(skb);
1342 csum = skb_checksum(skb, offset, skb->len- offset, 0);
1343 offset += skb->csum_offset;
1344 *(__sum16 *)(skb->data + offset) = csum_fold(csum);
1345 skb->ip_summed = CHECKSUM_NONE;
1346
1347 /* Check if there is enough headroom to insert fragment header. */
1348 if ((skb_mac_header(skb) < skb->head + frag_hdr_sz) &&
1349 pskb_expand_head(skb, frag_hdr_sz, 0, GFP_ATOMIC))
1350 goto out;
1351
1352 /* Find the unfragmentable header and shift it left by frag_hdr_sz
1353 * bytes to insert fragment header.
1354 */
1355 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
1356 nexthdr = *prevhdr;
1357 *prevhdr = NEXTHDR_FRAGMENT;
1358 unfrag_len = skb_network_header(skb) - skb_mac_header(skb) +
1359 unfrag_ip6hlen;
1360 mac_start = skb_mac_header(skb);
1361 memmove(mac_start-frag_hdr_sz, mac_start, unfrag_len);
1362
1363 skb->mac_header -= frag_hdr_sz;
1364 skb->network_header -= frag_hdr_sz;
1365
1366 fptr = (struct frag_hdr *)(skb_network_header(skb) + unfrag_ip6hlen);
1367 fptr->nexthdr = nexthdr;
1368 fptr->reserved = 0;
1369 ipv6_select_ident(fptr, (struct rt6_info *)skb_dst(skb));
1370
1371 /* Fragment the skb. ipv6 header and the remaining fields of the
1372 * fragment header are updated in ipv6_gso_segment()
1373 */
1374 segs = skb_segment(skb, features);
1375
1376 out:
1377 return segs;
1378 }
1379
1380 static const struct inet6_protocol udpv6_protocol = {
1381 .handler = udpv6_rcv,
1382 .err_handler = udpv6_err,
1383 .gso_send_check = udp6_ufo_send_check,
1384 .gso_segment = udp6_ufo_fragment,
1385 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1386 };
1387
1388 /* ------------------------------------------------------------------------ */
1389 #ifdef CONFIG_PROC_FS
1390
udp6_sock_seq_show(struct seq_file * seq,struct sock * sp,int bucket)1391 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1392 {
1393 struct inet_sock *inet = inet_sk(sp);
1394 struct ipv6_pinfo *np = inet6_sk(sp);
1395 const struct in6_addr *dest, *src;
1396 __u16 destp, srcp;
1397
1398 dest = &np->daddr;
1399 src = &np->rcv_saddr;
1400 destp = ntohs(inet->inet_dport);
1401 srcp = ntohs(inet->inet_sport);
1402 seq_printf(seq,
1403 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1404 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1405 bucket,
1406 src->s6_addr32[0], src->s6_addr32[1],
1407 src->s6_addr32[2], src->s6_addr32[3], srcp,
1408 dest->s6_addr32[0], dest->s6_addr32[1],
1409 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1410 sp->sk_state,
1411 sk_wmem_alloc_get(sp),
1412 sk_rmem_alloc_get(sp),
1413 0, 0L, 0,
1414 sock_i_uid(sp), 0,
1415 sock_i_ino(sp),
1416 atomic_read(&sp->sk_refcnt), sp,
1417 atomic_read(&sp->sk_drops));
1418 }
1419
udp6_seq_show(struct seq_file * seq,void * v)1420 int udp6_seq_show(struct seq_file *seq, void *v)
1421 {
1422 if (v == SEQ_START_TOKEN)
1423 seq_printf(seq,
1424 " sl "
1425 "local_address "
1426 "remote_address "
1427 "st tx_queue rx_queue tr tm->when retrnsmt"
1428 " uid timeout inode ref pointer drops\n");
1429 else
1430 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1431 return 0;
1432 }
1433
1434 static const struct file_operations udp6_afinfo_seq_fops = {
1435 .owner = THIS_MODULE,
1436 .open = udp_seq_open,
1437 .read = seq_read,
1438 .llseek = seq_lseek,
1439 .release = seq_release_net
1440 };
1441
1442 static struct udp_seq_afinfo udp6_seq_afinfo = {
1443 .name = "udp6",
1444 .family = AF_INET6,
1445 .udp_table = &udp_table,
1446 .seq_fops = &udp6_afinfo_seq_fops,
1447 .seq_ops = {
1448 .show = udp6_seq_show,
1449 },
1450 };
1451
udp6_proc_init(struct net * net)1452 int __net_init udp6_proc_init(struct net *net)
1453 {
1454 return udp_proc_register(net, &udp6_seq_afinfo);
1455 }
1456
udp6_proc_exit(struct net * net)1457 void udp6_proc_exit(struct net *net) {
1458 udp_proc_unregister(net, &udp6_seq_afinfo);
1459 }
1460 #endif /* CONFIG_PROC_FS */
1461
udp_v6_clear_sk(struct sock * sk,int size)1462 void udp_v6_clear_sk(struct sock *sk, int size)
1463 {
1464 struct inet_sock *inet = inet_sk(sk);
1465
1466 /* we do not want to clear pinet6 field, because of RCU lookups */
1467 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1468
1469 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1470 memset(&inet->pinet6 + 1, 0, size);
1471 }
1472
1473 /* ------------------------------------------------------------------------ */
1474
1475 struct proto udpv6_prot = {
1476 .name = "UDPv6",
1477 .owner = THIS_MODULE,
1478 .close = udp_lib_close,
1479 .connect = ip6_datagram_connect,
1480 .disconnect = udp_disconnect,
1481 .ioctl = udp_ioctl,
1482 .destroy = udpv6_destroy_sock,
1483 .setsockopt = udpv6_setsockopt,
1484 .getsockopt = udpv6_getsockopt,
1485 .sendmsg = udpv6_sendmsg,
1486 .recvmsg = udpv6_recvmsg,
1487 .backlog_rcv = udpv6_queue_rcv_skb,
1488 .hash = udp_lib_hash,
1489 .unhash = udp_lib_unhash,
1490 .rehash = udp_v6_rehash,
1491 .get_port = udp_v6_get_port,
1492 .memory_allocated = &udp_memory_allocated,
1493 .sysctl_mem = sysctl_udp_mem,
1494 .sysctl_wmem = &sysctl_udp_wmem_min,
1495 .sysctl_rmem = &sysctl_udp_rmem_min,
1496 .obj_size = sizeof(struct udp6_sock),
1497 .slab_flags = SLAB_DESTROY_BY_RCU,
1498 .h.udp_table = &udp_table,
1499 #ifdef CONFIG_COMPAT
1500 .compat_setsockopt = compat_udpv6_setsockopt,
1501 .compat_getsockopt = compat_udpv6_getsockopt,
1502 #endif
1503 .clear_sk = udp_v6_clear_sk,
1504 };
1505
1506 static struct inet_protosw udpv6_protosw = {
1507 .type = SOCK_DGRAM,
1508 .protocol = IPPROTO_UDP,
1509 .prot = &udpv6_prot,
1510 .ops = &inet6_dgram_ops,
1511 .no_check = UDP_CSUM_DEFAULT,
1512 .flags = INET_PROTOSW_PERMANENT,
1513 };
1514
1515
udpv6_init(void)1516 int __init udpv6_init(void)
1517 {
1518 int ret;
1519
1520 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1521 if (ret)
1522 goto out;
1523
1524 ret = inet6_register_protosw(&udpv6_protosw);
1525 if (ret)
1526 goto out_udpv6_protocol;
1527 out:
1528 return ret;
1529
1530 out_udpv6_protocol:
1531 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1532 goto out;
1533 }
1534
udpv6_exit(void)1535 void udpv6_exit(void)
1536 {
1537 inet6_unregister_protosw(&udpv6_protosw);
1538 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1539 }
1540