Merge branch 'for-davem' of git://gitorious.org/linux-can/linux-can-next
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / udp.c
CommitLineData
1da177e4
LT
1/*
2 * UDP over IPv6
1ab1457c 3 * Linux INET6 implementation
1da177e4
LT
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT
7 *
8 * Based on linux/ipv4/udp.c
9 *
1da177e4
LT
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
1da177e4
LT
24#include <linux/errno.h>
25#include <linux/types.h>
26#include <linux/socket.h>
27#include <linux/sockios.h>
1da177e4
LT
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>
1781f7f5 35#include <linux/module.h>
3305b80c 36#include <linux/skbuff.h>
5a0e3ad6 37#include <linux/slab.h>
1da177e4
LT
38#include <asm/uaccess.h>
39
1da177e4
LT
40#include <net/ndisc.h>
41#include <net/protocol.h>
42#include <net/transp_v6.h>
43#include <net/ip6_route.h>
1da177e4 44#include <net/raw.h>
c752f073 45#include <net/tcp_states.h>
1da177e4
LT
46#include <net/ip6_checksum.h>
47#include <net/xfrm.h>
48
49#include <linux/proc_fs.h>
50#include <linux/seq_file.h>
22911fc5 51#include <trace/events/skb.h>
ba4e58ec 52#include "udp_impl.h"
1da177e4 53
b2f5e7cd
VY
54int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
55{
56 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
57 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
68835aba
ED
58 __be32 sk1_rcv_saddr = sk_rcv_saddr(sk);
59 __be32 sk2_rcv_saddr = sk_rcv_saddr(sk2);
b2f5e7cd
VY
60 int sk_ipv6only = ipv6_only_sock(sk);
61 int sk2_ipv6only = inet_v6_ipv6only(sk2);
62 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
63 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
64
65 /* if both are mapped, treat as IPv4 */
66 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
499923c7 67 return (!sk2_ipv6only &&
c720c7e8
ED
68 (!sk1_rcv_saddr || !sk2_rcv_saddr ||
69 sk1_rcv_saddr == sk2_rcv_saddr));
b2f5e7cd
VY
70
71 if (addr_type2 == IPV6_ADDR_ANY &&
72 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
73 return 1;
74
75 if (addr_type == IPV6_ADDR_ANY &&
76 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
77 return 1;
78
79 if (sk2_rcv_saddr6 &&
80 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
81 return 1;
82
83 return 0;
84}
85
d4cada4a
ED
86static unsigned int udp6_portaddr_hash(struct net *net,
87 const struct in6_addr *addr6,
88 unsigned int port)
89{
90 unsigned int hash, mix = net_hash_mix(net);
91
92 if (ipv6_addr_any(addr6))
93 hash = jhash_1word(0, mix);
856540ee 94 else if (ipv6_addr_v4mapped(addr6))
0eae88f3 95 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
d4cada4a 96 else
0eae88f3 97 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
d4cada4a
ED
98
99 return hash ^ port;
100}
101
102
6ba5a3c5 103int udp_v6_get_port(struct sock *sk, unsigned short snum)
1da177e4 104{
30fff923
ED
105 unsigned int hash2_nulladdr =
106 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
9a52e97e 107 unsigned int hash2_partial =
30fff923
ED
108 udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0);
109
d4cada4a 110 /* precompute partial secondary hash */
30fff923
ED
111 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
112 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
1da177e4
LT
113}
114
719f8358
ED
115static void udp_v6_rehash(struct sock *sk)
116{
117 u16 new_hash = udp6_portaddr_hash(sock_net(sk),
118 &inet6_sk(sk)->rcv_saddr,
119 inet_sk(sk)->inet_num);
120
121 udp_lib_rehash(sk, new_hash);
122}
123
645ca708
ED
124static inline int compute_score(struct sock *sk, struct net *net,
125 unsigned short hnum,
88440ae7
BS
126 const struct in6_addr *saddr, __be16 sport,
127 const struct in6_addr *daddr, __be16 dport,
645ca708
ED
128 int dif)
129{
130 int score = -1;
131
d4cada4a 132 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
645ca708
ED
133 sk->sk_family == PF_INET6) {
134 struct ipv6_pinfo *np = inet6_sk(sk);
135 struct inet_sock *inet = inet_sk(sk);
136
137 score = 0;
c720c7e8
ED
138 if (inet->inet_dport) {
139 if (inet->inet_dport != sport)
645ca708
ED
140 return -1;
141 score++;
142 }
143 if (!ipv6_addr_any(&np->rcv_saddr)) {
144 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
145 return -1;
146 score++;
147 }
148 if (!ipv6_addr_any(&np->daddr)) {
149 if (!ipv6_addr_equal(&np->daddr, saddr))
150 return -1;
151 score++;
152 }
153 if (sk->sk_bound_dev_if) {
154 if (sk->sk_bound_dev_if != dif)
155 return -1;
156 score++;
157 }
158 }
159 return score;
160}
161
fddc17de
ED
162#define SCORE2_MAX (1 + 1 + 1)
163static inline int compute_score2(struct sock *sk, struct net *net,
164 const struct in6_addr *saddr, __be16 sport,
165 const struct in6_addr *daddr, unsigned short hnum,
166 int dif)
167{
168 int score = -1;
169
170 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
171 sk->sk_family == PF_INET6) {
172 struct ipv6_pinfo *np = inet6_sk(sk);
173 struct inet_sock *inet = inet_sk(sk);
174
175 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
176 return -1;
177 score = 0;
178 if (inet->inet_dport) {
179 if (inet->inet_dport != sport)
180 return -1;
181 score++;
182 }
183 if (!ipv6_addr_any(&np->daddr)) {
184 if (!ipv6_addr_equal(&np->daddr, saddr))
185 return -1;
186 score++;
187 }
188 if (sk->sk_bound_dev_if) {
189 if (sk->sk_bound_dev_if != dif)
190 return -1;
191 score++;
192 }
193 }
194 return score;
195}
196
fddc17de
ED
197
198/* called with read_rcu_lock() */
199static struct sock *udp6_lib_lookup2(struct net *net,
200 const struct in6_addr *saddr, __be16 sport,
201 const struct in6_addr *daddr, unsigned int hnum, int dif,
202 struct udp_hslot *hslot2, unsigned int slot2)
203{
204 struct sock *sk, *result;
205 struct hlist_nulls_node *node;
206 int score, badness;
207
208begin:
209 result = NULL;
210 badness = -1;
211 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
212 score = compute_score2(sk, net, saddr, sport,
213 daddr, hnum, dif);
214 if (score > badness) {
215 result = sk;
216 badness = score;
217 if (score == SCORE2_MAX)
218 goto exact_match;
219 }
220 }
221 /*
222 * if the nulls value we got at the end of this lookup is
223 * not the expected one, we must restart lookup.
224 * We probably met an item that was moved to another chain.
225 */
226 if (get_nulls_value(node) != slot2)
227 goto begin;
228
229 if (result) {
230exact_match:
c31504dc 231 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
fddc17de
ED
232 result = NULL;
233 else if (unlikely(compute_score2(result, net, saddr, sport,
234 daddr, hnum, dif) < badness)) {
235 sock_put(result);
236 goto begin;
237 }
238 }
239 return result;
240}
241
fce82338 242struct sock *__udp6_lib_lookup(struct net *net,
88440ae7
BS
243 const struct in6_addr *saddr, __be16 sport,
244 const struct in6_addr *daddr, __be16 dport,
645ca708 245 int dif, struct udp_table *udptable)
1da177e4 246{
271b72c7 247 struct sock *sk, *result;
88ab1932 248 struct hlist_nulls_node *node;
1da177e4 249 unsigned short hnum = ntohs(dport);
fddc17de
ED
250 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
251 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
271b72c7 252 int score, badness;
645ca708 253
271b72c7 254 rcu_read_lock();
fddc17de
ED
255 if (hslot->count > 10) {
256 hash2 = udp6_portaddr_hash(net, daddr, hnum);
257 slot2 = hash2 & udptable->mask;
258 hslot2 = &udptable->hash2[slot2];
259 if (hslot->count < hslot2->count)
260 goto begin;
261
262 result = udp6_lib_lookup2(net, saddr, sport,
263 daddr, hnum, dif,
264 hslot2, slot2);
265 if (!result) {
266 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
267 slot2 = hash2 & udptable->mask;
268 hslot2 = &udptable->hash2[slot2];
269 if (hslot->count < hslot2->count)
270 goto begin;
271
1223c67c
JBD
272 result = udp6_lib_lookup2(net, saddr, sport,
273 &in6addr_any, hnum, dif,
fddc17de
ED
274 hslot2, slot2);
275 }
276 rcu_read_unlock();
277 return result;
278 }
271b72c7
ED
279begin:
280 result = NULL;
281 badness = -1;
88ab1932 282 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
645ca708
ED
283 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
284 if (score > badness) {
285 result = sk;
286 badness = score;
1da177e4
LT
287 }
288 }
88ab1932
ED
289 /*
290 * if the nulls value we got at the end of this lookup is
291 * not the expected one, we must restart lookup.
292 * We probably met an item that was moved to another chain.
293 */
fddc17de 294 if (get_nulls_value(node) != slot)
88ab1932
ED
295 goto begin;
296
271b72c7 297 if (result) {
c31504dc 298 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
271b72c7
ED
299 result = NULL;
300 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
301 daddr, dport, dif) < badness)) {
302 sock_put(result);
303 goto begin;
304 }
305 }
306 rcu_read_unlock();
1da177e4
LT
307 return result;
308}
fce82338 309EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
1da177e4 310
607c4aaf
KK
311static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
312 __be16 sport, __be16 dport,
645ca708 313 struct udp_table *udptable)
607c4aaf 314{
23542618 315 struct sock *sk;
b71d1d42 316 const struct ipv6hdr *iph = ipv6_hdr(skb);
607c4aaf 317
23542618
KK
318 if (unlikely(sk = skb_steal_sock(skb)))
319 return sk;
adf30907
ED
320 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
321 &iph->daddr, dport, inet6_iif(skb),
322 udptable);
607c4aaf
KK
323}
324
aa976fc0
BS
325struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
326 const struct in6_addr *daddr, __be16 dport, int dif)
327{
328 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
329}
330EXPORT_SYMBOL_GPL(udp6_lib_lookup);
331
332
1da177e4
LT
333/*
334 * This should be easy, if there is something there we
335 * return it, otherwise we block.
336 */
337
ba4e58ec 338int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
339 struct msghdr *msg, size_t len,
340 int noblock, int flags, int *addr_len)
341{
342 struct ipv6_pinfo *np = inet6_sk(sk);
343 struct inet_sock *inet = inet_sk(sk);
1ab1457c 344 struct sk_buff *skb;
59c2cdae 345 unsigned int ulen, copied;
3f518bf7 346 int peeked, off = 0;
759e5d00
HX
347 int err;
348 int is_udplite = IS_UDPLITE(sk);
f26ba175 349 int is_udp4;
8a74ad60 350 bool slow;
1da177e4 351
1ab1457c 352 if (addr_len)
3dde2598 353 *addr_len = sizeof(struct sockaddr_in6);
1ab1457c 354
1da177e4
LT
355 if (flags & MSG_ERRQUEUE)
356 return ipv6_recv_error(sk, msg, len);
357
4b340ae2
BH
358 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
359 return ipv6_recv_rxpmtu(sk, msg, len);
360
1da177e4 361try_again:
a59322be 362 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
3f518bf7 363 &peeked, &off, &err);
1da177e4
LT
364 if (!skb)
365 goto out;
366
759e5d00 367 ulen = skb->len - sizeof(struct udphdr);
59c2cdae
DM
368 copied = len;
369 if (copied > ulen)
370 copied = ulen;
371 else if (copied < ulen)
1ab1457c 372 msg->msg_flags |= MSG_TRUNC;
1da177e4 373
f26ba175
WY
374 is_udp4 = (skb->protocol == htons(ETH_P_IP));
375
ba4e58ec 376 /*
759e5d00
HX
377 * If checksum is needed at all, try to do it while copying the
378 * data. If the data is truncated, or if we only want a partial
379 * coverage checksum (UDP-Lite), do it before the copy.
ba4e58ec 380 */
ba4e58ec 381
59c2cdae 382 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
759e5d00 383 if (udp_lib_checksum_complete(skb))
1da177e4 384 goto csum_copy_err;
ba4e58ec
GR
385 }
386
60476372 387 if (skb_csum_unnecessary(skb))
ba4e58ec 388 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
22911fc5 389 msg->msg_iov, copied);
ba4e58ec 390 else {
1da177e4
LT
391 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
392 if (err == -EINVAL)
393 goto csum_copy_err;
394 }
22911fc5
ED
395 if (unlikely(err)) {
396 trace_kfree_skb(skb, udpv6_recvmsg);
1da177e4 397 goto out_free;
22911fc5 398 }
f26ba175
WY
399 if (!peeked) {
400 if (is_udp4)
401 UDP_INC_STATS_USER(sock_net(sk),
402 UDP_MIB_INDATAGRAMS, is_udplite);
403 else
404 UDP6_INC_STATS_USER(sock_net(sk),
405 UDP_MIB_INDATAGRAMS, is_udplite);
406 }
cb75994e 407
3b885787 408 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
409
410 /* Copy the address. */
411 if (msg->msg_name) {
412 struct sockaddr_in6 *sin6;
1ab1457c 413
1da177e4
LT
414 sin6 = (struct sockaddr_in6 *) msg->msg_name;
415 sin6->sin6_family = AF_INET6;
4bedb452 416 sin6->sin6_port = udp_hdr(skb)->source;
1da177e4
LT
417 sin6->sin6_flowinfo = 0;
418 sin6->sin6_scope_id = 0;
419
f26ba175 420 if (is_udp4)
b301e82c
BH
421 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
422 &sin6->sin6_addr);
1da177e4 423 else {
4e3fd7a0 424 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
1da177e4
LT
425 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
426 sin6->sin6_scope_id = IP6CB(skb)->iif;
427 }
428
429 }
f26ba175 430 if (is_udp4) {
1da177e4
LT
431 if (inet->cmsg_flags)
432 ip_cmsg_recv(msg, skb);
433 } else {
434 if (np->rxopt.all)
435 datagram_recv_ctl(sk, msg, skb);
1ab1457c 436 }
1da177e4 437
59c2cdae 438 err = copied;
1da177e4 439 if (flags & MSG_TRUNC)
759e5d00 440 err = ulen;
1da177e4
LT
441
442out_free:
9d410c79 443 skb_free_datagram_locked(sk, skb);
1da177e4
LT
444out:
445 return err;
446
447csum_copy_err:
8a74ad60 448 slow = lock_sock_fast(sk);
0856f939
WY
449 if (!skb_kill_datagram(sk, skb, flags)) {
450 if (is_udp4)
451 UDP_INC_STATS_USER(sock_net(sk),
452 UDP_MIB_INERRORS, is_udplite);
453 else
454 UDP6_INC_STATS_USER(sock_net(sk),
455 UDP_MIB_INERRORS, is_udplite);
456 }
8a74ad60 457 unlock_sock_fast(sk, slow);
1da177e4 458
32c90254 459 if (noblock)
1da177e4 460 return -EAGAIN;
9cfaa8de
XZ
461
462 /* starting over for a new packet */
463 msg->msg_flags &= ~MSG_TRUNC;
1da177e4
LT
464 goto try_again;
465}
466
ba4e58ec 467void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 468 u8 type, u8 code, int offset, __be32 info,
645ca708 469 struct udp_table *udptable)
1da177e4
LT
470{
471 struct ipv6_pinfo *np;
b71d1d42
ED
472 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
473 const struct in6_addr *saddr = &hdr->saddr;
474 const struct in6_addr *daddr = &hdr->daddr;
1da177e4
LT
475 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
476 struct sock *sk;
477 int err;
478
c346dca1 479 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
ba4e58ec 480 saddr, uh->source, inet6_iif(skb), udptable);
1da177e4
LT
481 if (sk == NULL)
482 return;
483
81aded24
DM
484 if (type == ICMPV6_PKT_TOOBIG)
485 ip6_sk_update_pmtu(skb, sk, info);
486
1da177e4
LT
487 np = inet6_sk(sk);
488
489 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
490 goto out;
491
492 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
493 goto out;
494
b1faf566 495 if (np->recverr)
1da177e4 496 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
b1faf566 497
1da177e4
LT
498 sk->sk_err = err;
499 sk->sk_error_report(sk);
500out:
501 sock_put(sk);
502}
503
f7ad74fe
BL
504static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
505{
506 int rc;
507
508 if (!ipv6_addr_any(&inet6_sk(sk)->daddr))
509 sock_rps_save_rxhash(sk, skb);
510
511 rc = sock_queue_rcv_skb(sk, skb);
512 if (rc < 0) {
513 int is_udplite = IS_UDPLITE(sk);
514
515 /* Note that an ENOMEM error is charged twice */
516 if (rc == -ENOMEM)
517 UDP6_INC_STATS_BH(sock_net(sk),
518 UDP_MIB_RCVBUFERRORS, is_udplite);
519 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
520 kfree_skb(skb);
521 return -1;
522 }
523 return 0;
524}
525
ba4e58ec 526static __inline__ void udpv6_err(struct sk_buff *skb,
d5fdd6ba
BH
527 struct inet6_skb_parm *opt, u8 type,
528 u8 code, int offset, __be32 info )
ba4e58ec 529{
645ca708 530 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
ba4e58ec
GR
531}
532
d7f3f621
BL
533static struct static_key udpv6_encap_needed __read_mostly;
534void udpv6_encap_enable(void)
535{
536 if (!static_key_enabled(&udpv6_encap_needed))
537 static_key_slow_inc(&udpv6_encap_needed);
538}
539EXPORT_SYMBOL(udpv6_encap_enable);
540
f7ad74fe 541int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1da177e4 542{
ba4e58ec 543 struct udp_sock *up = udp_sk(sk);
a18135eb 544 int rc;
b2bf1e26 545 int is_udplite = IS_UDPLITE(sk);
a18135eb 546
ba4e58ec
GR
547 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
548 goto drop;
1da177e4 549
d7f3f621
BL
550 if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
551 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
552
553 /*
554 * This is an encapsulation socket so pass the skb to
555 * the socket's udp_encap_rcv() hook. Otherwise, just
556 * fall through and pass this up the UDP socket.
557 * up->encap_rcv() returns the following value:
558 * =0 if skb was successfully passed to the encap
559 * handler or was discarded by it.
560 * >0 if skb should be passed on to UDP.
561 * <0 if skb should be resubmitted as proto -N
562 */
563
564 /* if we're overly short, let UDP handle it */
565 encap_rcv = ACCESS_ONCE(up->encap_rcv);
566 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
567 int ret;
568
569 ret = encap_rcv(sk, skb);
570 if (ret <= 0) {
571 UDP_INC_STATS_BH(sock_net(sk),
572 UDP_MIB_INDATAGRAMS,
573 is_udplite);
574 return -ret;
575 }
576 }
577
578 /* FALLTHROUGH -- it's a UDP Packet */
579 }
580
ba4e58ec
GR
581 /*
582 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
583 */
b2bf1e26 584 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
ba4e58ec
GR
585
586 if (up->pcrlen == 0) { /* full coverage was set */
587 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
588 " %d while full coverage %d requested\n",
589 UDP_SKB_CB(skb)->cscov, skb->len);
590 goto drop;
591 }
592 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
593 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
594 "too small, need min %d\n",
595 UDP_SKB_CB(skb)->cscov, up->pcrlen);
596 goto drop;
597 }
1da177e4
LT
598 }
599
33d480ce 600 if (rcu_access_pointer(sk->sk_filter)) {
1ab6eb62
HX
601 if (udp_lib_checksum_complete(skb))
602 goto drop;
603 }
ba4e58ec 604
cb80ef46
BL
605 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
606 goto drop;
607
d826eb14 608 skb_dst_drop(skb);
cb75994e 609
cb80ef46
BL
610 bh_lock_sock(sk);
611 rc = 0;
612 if (!sock_owned_by_user(sk))
613 rc = __udpv6_queue_rcv_skb(sk, skb);
614 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
615 bh_unlock_sock(sk);
616 goto drop;
617 }
618 bh_unlock_sock(sk);
f7ad74fe
BL
619
620 return rc;
ba4e58ec 621drop:
ef28d1a2 622 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
cb80ef46 623 atomic_inc(&sk->sk_drops);
ba4e58ec
GR
624 kfree_skb(skb);
625 return -1;
1da177e4
LT
626}
627
920a4611 628static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
b71d1d42
ED
629 __be16 loc_port, const struct in6_addr *loc_addr,
630 __be16 rmt_port, const struct in6_addr *rmt_addr,
1da177e4
LT
631 int dif)
632{
88ab1932 633 struct hlist_nulls_node *node;
1da177e4
LT
634 struct sock *s = sk;
635 unsigned short num = ntohs(loc_port);
636
88ab1932 637 sk_nulls_for_each_from(s, node) {
1da177e4
LT
638 struct inet_sock *inet = inet_sk(s);
639
920a4611 640 if (!net_eq(sock_net(s), net))
b8ad0cbc
DL
641 continue;
642
d4cada4a
ED
643 if (udp_sk(s)->udp_port_hash == num &&
644 s->sk_family == PF_INET6) {
1da177e4 645 struct ipv6_pinfo *np = inet6_sk(s);
c720c7e8
ED
646 if (inet->inet_dport) {
647 if (inet->inet_dport != rmt_port)
1da177e4
LT
648 continue;
649 }
650 if (!ipv6_addr_any(&np->daddr) &&
651 !ipv6_addr_equal(&np->daddr, rmt_addr))
652 continue;
653
654 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
655 continue;
656
657 if (!ipv6_addr_any(&np->rcv_saddr)) {
40796c5e
DS
658 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
659 continue;
1da177e4 660 }
add459aa 661 if (!inet6_mc_check(s, loc_addr, rmt_addr))
1da177e4
LT
662 continue;
663 return s;
664 }
665 }
666 return NULL;
667}
668
a1ab77f9
ED
669static void flush_stack(struct sock **stack, unsigned int count,
670 struct sk_buff *skb, unsigned int final)
671{
cb80ef46 672 struct sk_buff *skb1 = NULL;
a1ab77f9 673 struct sock *sk;
cb80ef46 674 unsigned int i;
a1ab77f9
ED
675
676 for (i = 0; i < count; i++) {
f6b8f32c 677 sk = stack[i];
cb80ef46
BL
678 if (likely(skb1 == NULL))
679 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
680 if (!skb1) {
681 atomic_inc(&sk->sk_drops);
682 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
683 IS_UDPLITE(sk));
684 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
685 IS_UDPLITE(sk));
a1ab77f9 686 }
cb80ef46
BL
687
688 if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
689 skb1 = NULL;
a1ab77f9 690 }
cb80ef46
BL
691 if (unlikely(skb1))
692 kfree_skb(skb1);
a1ab77f9 693}
1da177e4
LT
694/*
695 * Note: called only from the BH handler context,
696 * so we don't need to lock the hashes.
697 */
e3163493 698static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
b71d1d42 699 const struct in6_addr *saddr, const struct in6_addr *daddr,
645ca708 700 struct udp_table *udptable)
1da177e4 701{
a1ab77f9 702 struct sock *sk, *stack[256 / sizeof(struct sock *)];
4bedb452 703 const struct udphdr *uh = udp_hdr(skb);
f86dcc5a 704 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
1da177e4 705 int dif;
a1ab77f9 706 unsigned int i, count = 0;
1da177e4 707
645ca708 708 spin_lock(&hslot->lock);
88ab1932 709 sk = sk_nulls_head(&hslot->head);
f2776ff0 710 dif = inet6_iif(skb);
920a4611 711 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
a1ab77f9
ED
712 while (sk) {
713 stack[count++] = sk;
714 sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
715 uh->source, saddr, dif);
716 if (unlikely(count == ARRAY_SIZE(stack))) {
717 if (!sk)
718 break;
719 flush_stack(stack, count, skb, ~0);
720 count = 0;
95766fff 721 }
1da177e4 722 }
a1ab77f9
ED
723 /*
724 * before releasing the lock, we must take reference on sockets
725 */
726 for (i = 0; i < count; i++)
727 sock_hold(stack[i]);
728
645ca708 729 spin_unlock(&hslot->lock);
a1ab77f9
ED
730
731 if (count) {
732 flush_stack(stack, count, skb, count - 1);
733
734 for (i = 0; i < count; i++)
735 sock_put(stack[i]);
736 } else {
737 kfree_skb(skb);
738 }
ba4e58ec 739 return 0;
1da177e4
LT
740}
741
759e5d00
HX
742static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
743 int proto)
ba4e58ec 744{
759e5d00
HX
745 int err;
746
747 UDP_SKB_CB(skb)->partial_cov = 0;
748 UDP_SKB_CB(skb)->cscov = skb->len;
749
db8dac20 750 if (proto == IPPROTO_UDPLITE) {
759e5d00
HX
751 err = udplite_checksum_init(skb, uh);
752 if (err)
753 return err;
754 }
755
ba4e58ec
GR
756 if (uh->check == 0) {
757 /* RFC 2460 section 8.1 says that we SHOULD log
758 this error. Well, it is reasonable.
759 */
760 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
761 return 1;
762 }
763 if (skb->ip_summed == CHECKSUM_COMPLETE &&
0660e03f 764 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
759e5d00 765 skb->len, proto, skb->csum))
ba4e58ec
GR
766 skb->ip_summed = CHECKSUM_UNNECESSARY;
767
60476372 768 if (!skb_csum_unnecessary(skb))
0660e03f
ACM
769 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
770 &ipv6_hdr(skb)->daddr,
759e5d00 771 skb->len, proto, 0));
ba4e58ec 772
759e5d00 773 return 0;
ba4e58ec
GR
774}
775
645ca708 776int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
759e5d00 777 int proto)
1da177e4 778{
3ffe533c 779 struct net *net = dev_net(skb->dev);
1da177e4 780 struct sock *sk;
1ab1457c 781 struct udphdr *uh;
b71d1d42 782 const struct in6_addr *saddr, *daddr;
1da177e4
LT
783 u32 ulen = 0;
784
785 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
d6bc0149 786 goto discard;
1da177e4 787
0660e03f
ACM
788 saddr = &ipv6_hdr(skb)->saddr;
789 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 790 uh = udp_hdr(skb);
1da177e4
LT
791
792 ulen = ntohs(uh->len);
ba4e58ec
GR
793 if (ulen > skb->len)
794 goto short_packet;
1da177e4 795
759e5d00
HX
796 if (proto == IPPROTO_UDP) {
797 /* UDP validates ulen. */
1da177e4 798
ba4e58ec
GR
799 /* Check for jumbo payload */
800 if (ulen == 0)
801 ulen = skb->len;
1da177e4 802
ba4e58ec
GR
803 if (ulen < sizeof(*uh))
804 goto short_packet;
1da177e4 805
ba4e58ec
GR
806 if (ulen < skb->len) {
807 if (pskb_trim_rcsum(skb, ulen))
808 goto short_packet;
0660e03f
ACM
809 saddr = &ipv6_hdr(skb)->saddr;
810 daddr = &ipv6_hdr(skb)->daddr;
4bedb452 811 uh = udp_hdr(skb);
ba4e58ec 812 }
ba4e58ec 813 }
1da177e4 814
759e5d00
HX
815 if (udp6_csum_init(skb, uh, proto))
816 goto discard;
817
1ab1457c
YH
818 /*
819 * Multicast receive code
1da177e4 820 */
ba4e58ec 821 if (ipv6_addr_is_multicast(daddr))
e3163493
PE
822 return __udp6_lib_mcast_deliver(net, skb,
823 saddr, daddr, udptable);
1da177e4
LT
824
825 /* Unicast */
826
1ab1457c 827 /*
1da177e4
LT
828 * check socket cache ... must talk to Alan about his plans
829 * for sock caches... i'll skip this for now.
830 */
607c4aaf 831 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
cb80ef46
BL
832 if (sk != NULL) {
833 int ret = udpv6_queue_rcv_skb(sk, skb);
834 sock_put(sk);
1da177e4 835
cb80ef46
BL
836 /* a return value > 0 means to resubmit the input, but
837 * it wants the return to be -protocol, or 0
838 */
839 if (ret > 0)
840 return -ret;
1da177e4 841
add459aa 842 return 0;
1da177e4 843 }
1ab1457c 844
cb80ef46 845 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
c377411f 846 goto discard;
cb80ef46
BL
847
848 if (udp_lib_checksum_complete(skb))
55349790 849 goto discard;
cb80ef46
BL
850
851 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
852 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
853
854 kfree_skb(skb);
add459aa 855 return 0;
1da177e4 856
1ab1457c 857short_packet:
d6bc0149 858 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
db8dac20 859 proto == IPPROTO_UDPLITE ? "-Lite" : "",
d6bc0149
BM
860 saddr,
861 ntohs(uh->source),
862 ulen,
863 skb->len,
864 daddr,
865 ntohs(uh->dest));
1da177e4
LT
866
867discard:
ef28d1a2 868 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1da177e4 869 kfree_skb(skb);
add459aa 870 return 0;
1da177e4 871}
ba4e58ec 872
e5bbef20 873static __inline__ int udpv6_rcv(struct sk_buff *skb)
ba4e58ec 874{
645ca708 875 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
ba4e58ec
GR
876}
877
1da177e4
LT
878/*
879 * Throw away all pending data and cancel the corking. Socket is locked.
880 */
881static void udp_v6_flush_pending_frames(struct sock *sk)
882{
883 struct udp_sock *up = udp_sk(sk);
884
36d926b9
DL
885 if (up->pending == AF_INET)
886 udp_flush_pending_frames(sk);
887 else if (up->pending) {
1da177e4
LT
888 up->len = 0;
889 up->pending = 0;
890 ip6_flush_pending_frames(sk);
1ab1457c 891 }
1da177e4
LT
892}
893
493c6be3
SS
894/**
895 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
896 * @sk: socket we are sending on
897 * @skb: sk_buff containing the filled-in UDP header
898 * (checksum field must be zeroed out)
899 */
900static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
901 const struct in6_addr *saddr,
902 const struct in6_addr *daddr, int len)
903{
904 unsigned int offset;
905 struct udphdr *uh = udp_hdr(skb);
906 __wsum csum = 0;
907
908 if (skb_queue_len(&sk->sk_write_queue) == 1) {
909 /* Only one fragment on the socket. */
910 skb->csum_start = skb_transport_header(skb) - skb->head;
911 skb->csum_offset = offsetof(struct udphdr, check);
912 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
913 } else {
914 /*
915 * HW-checksum won't work as there are two or more
916 * fragments on the socket so that all csums of sk_buffs
917 * should be together
918 */
919 offset = skb_transport_offset(skb);
920 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
921
922 skb->ip_summed = CHECKSUM_NONE;
923
924 skb_queue_walk(&sk->sk_write_queue, skb) {
925 csum = csum_add(csum, skb->csum);
926 }
927
928 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
929 csum);
930 if (uh->check == 0)
931 uh->check = CSUM_MANGLED_0;
932 }
933}
934
1da177e4
LT
935/*
936 * Sending
937 */
938
4c0a6cb0 939static int udp_v6_push_pending_frames(struct sock *sk)
1da177e4
LT
940{
941 struct sk_buff *skb;
942 struct udphdr *uh;
4c0a6cb0 943 struct udp_sock *up = udp_sk(sk);
1da177e4 944 struct inet_sock *inet = inet_sk(sk);
4c9483b2 945 struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
1da177e4 946 int err = 0;
b2bf1e26 947 int is_udplite = IS_UDPLITE(sk);
868c86bc 948 __wsum csum = 0;
1da177e4
LT
949
950 /* Grab the skbuff where UDP header space exists. */
951 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
952 goto out;
953
954 /*
955 * Create a UDP header
956 */
4bedb452 957 uh = udp_hdr(skb);
1958b856
DM
958 uh->source = fl6->fl6_sport;
959 uh->dest = fl6->fl6_dport;
1da177e4
LT
960 uh->len = htons(up->len);
961 uh->check = 0;
962
b2bf1e26 963 if (is_udplite)
ba4e58ec 964 csum = udplite_csum_outgoing(sk, skb);
493c6be3 965 else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
4c9483b2 966 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
493c6be3
SS
967 up->len);
968 goto send;
969 } else
ba4e58ec 970 csum = udp_csum_outgoing(sk, skb);
1da177e4 971
ba4e58ec 972 /* add protocol-dependent pseudo-header */
4c9483b2
DM
973 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
974 up->len, fl6->flowi6_proto, csum);
1da177e4 975 if (uh->check == 0)
f6ab0288 976 uh->check = CSUM_MANGLED_0;
1da177e4 977
493c6be3 978send:
1da177e4 979 err = ip6_push_pending_frames(sk);
6ce9e7b5
ED
980 if (err) {
981 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
982 UDP6_INC_STATS_USER(sock_net(sk),
983 UDP_MIB_SNDBUFERRORS, is_udplite);
984 err = 0;
985 }
986 } else
987 UDP6_INC_STATS_USER(sock_net(sk),
988 UDP_MIB_OUTDATAGRAMS, is_udplite);
1da177e4
LT
989out:
990 up->len = 0;
991 up->pending = 0;
992 return err;
993}
994
ba4e58ec 995int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
1da177e4
LT
996 struct msghdr *msg, size_t len)
997{
998 struct ipv6_txoptions opt_space;
999 struct udp_sock *up = udp_sk(sk);
1000 struct inet_sock *inet = inet_sk(sk);
1001 struct ipv6_pinfo *np = inet6_sk(sk);
1002 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
20c59de2 1003 struct in6_addr *daddr, *final_p, final;
1da177e4
LT
1004 struct ipv6_txoptions *opt = NULL;
1005 struct ip6_flowlabel *flowlabel = NULL;
4c9483b2 1006 struct flowi6 fl6;
1da177e4
LT
1007 struct dst_entry *dst;
1008 int addr_len = msg->msg_namelen;
1009 int ulen = len;
1010 int hlimit = -1;
41a1f8ea 1011 int tclass = -1;
13b52cd4 1012 int dontfrag = -1;
1da177e4
LT
1013 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1014 int err;
987905de 1015 int connected = 0;
b2bf1e26 1016 int is_udplite = IS_UDPLITE(sk);
ba4e58ec 1017 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1da177e4
LT
1018
1019 /* destination address check */
1020 if (sin6) {
1021 if (addr_len < offsetof(struct sockaddr, sa_data))
1022 return -EINVAL;
1023
1024 switch (sin6->sin6_family) {
1025 case AF_INET6:
1026 if (addr_len < SIN6_LEN_RFC2133)
1027 return -EINVAL;
1028 daddr = &sin6->sin6_addr;
1029 break;
1030 case AF_INET:
1031 goto do_udp_sendmsg;
1032 case AF_UNSPEC:
1033 msg->msg_name = sin6 = NULL;
1034 msg->msg_namelen = addr_len = 0;
1035 daddr = NULL;
1036 break;
1037 default:
1038 return -EINVAL;
1039 }
1040 } else if (!up->pending) {
1041 if (sk->sk_state != TCP_ESTABLISHED)
1042 return -EDESTADDRREQ;
1043 daddr = &np->daddr;
1ab1457c 1044 } else
1da177e4
LT
1045 daddr = NULL;
1046
1047 if (daddr) {
e773e4fa 1048 if (ipv6_addr_v4mapped(daddr)) {
1da177e4
LT
1049 struct sockaddr_in sin;
1050 sin.sin_family = AF_INET;
c720c7e8 1051 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1da177e4
LT
1052 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1053 msg->msg_name = &sin;
1054 msg->msg_namelen = sizeof(sin);
1055do_udp_sendmsg:
1056 if (__ipv6_only_sock(sk))
1057 return -ENETUNREACH;
1058 return udp_sendmsg(iocb, sk, msg, len);
1059 }
1060 }
1061
1062 if (up->pending == AF_INET)
1063 return udp_sendmsg(iocb, sk, msg, len);
1064
1065 /* Rough check on arithmetic overflow,
b59e139b 1066 better check is made in ip6_append_data().
1da177e4
LT
1067 */
1068 if (len > INT_MAX - sizeof(struct udphdr))
1069 return -EMSGSIZE;
1ab1457c 1070
1da177e4
LT
1071 if (up->pending) {
1072 /*
1073 * There are pending frames.
1074 * The socket lock must be held while it's corked.
1075 */
1076 lock_sock(sk);
1077 if (likely(up->pending)) {
1078 if (unlikely(up->pending != AF_INET6)) {
1079 release_sock(sk);
1080 return -EAFNOSUPPORT;
1081 }
1082 dst = NULL;
1083 goto do_append_data;
1084 }
1085 release_sock(sk);
1086 }
1087 ulen += sizeof(struct udphdr);
1088
4c9483b2 1089 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
1090
1091 if (sin6) {
1092 if (sin6->sin6_port == 0)
1093 return -EINVAL;
1094
1958b856 1095 fl6.fl6_dport = sin6->sin6_port;
1da177e4
LT
1096 daddr = &sin6->sin6_addr;
1097
1098 if (np->sndflow) {
4c9483b2
DM
1099 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1100 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1101 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1102 if (flowlabel == NULL)
1103 return -EINVAL;
1104 daddr = &flowlabel->dst;
1105 }
1106 }
1107
1108 /*
1109 * Otherwise it will be difficult to maintain
1110 * sk->sk_dst_cache.
1111 */
1112 if (sk->sk_state == TCP_ESTABLISHED &&
1113 ipv6_addr_equal(daddr, &np->daddr))
1114 daddr = &np->daddr;
1115
1116 if (addr_len >= sizeof(struct sockaddr_in6) &&
1117 sin6->sin6_scope_id &&
1118 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
4c9483b2 1119 fl6.flowi6_oif = sin6->sin6_scope_id;
1da177e4
LT
1120 } else {
1121 if (sk->sk_state != TCP_ESTABLISHED)
1122 return -EDESTADDRREQ;
1123
1958b856 1124 fl6.fl6_dport = inet->inet_dport;
1da177e4 1125 daddr = &np->daddr;
4c9483b2 1126 fl6.flowlabel = np->flow_label;
987905de 1127 connected = 1;
1da177e4
LT
1128 }
1129
4c9483b2
DM
1130 if (!fl6.flowi6_oif)
1131 fl6.flowi6_oif = sk->sk_bound_dev_if;
1da177e4 1132
4c9483b2
DM
1133 if (!fl6.flowi6_oif)
1134 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
9f690db7 1135
4c9483b2 1136 fl6.flowi6_mark = sk->sk_mark;
51953d5b 1137
1da177e4
LT
1138 if (msg->msg_controllen) {
1139 opt = &opt_space;
1140 memset(opt, 0, sizeof(struct ipv6_txoptions));
1141 opt->tot_len = sizeof(*opt);
1142
ec0506db
MÅ»
1143 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1144 &hlimit, &tclass, &dontfrag);
1da177e4
LT
1145 if (err < 0) {
1146 fl6_sock_release(flowlabel);
1147 return err;
1148 }
4c9483b2
DM
1149 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1150 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
1151 if (flowlabel == NULL)
1152 return -EINVAL;
1153 }
1154 if (!(opt->opt_nflen|opt->opt_flen))
1155 opt = NULL;
987905de 1156 connected = 0;
1da177e4
LT
1157 }
1158 if (opt == NULL)
1159 opt = np->opt;
df9890c3
YH
1160 if (flowlabel)
1161 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1162 opt = ipv6_fixup_options(&opt_space, opt);
1da177e4 1163
4c9483b2 1164 fl6.flowi6_proto = sk->sk_protocol;
876c7f41 1165 if (!ipv6_addr_any(daddr))
4e3fd7a0 1166 fl6.daddr = *daddr;
876c7f41 1167 else
4c9483b2
DM
1168 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1169 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
4e3fd7a0 1170 fl6.saddr = np->saddr;
1958b856 1171 fl6.fl6_sport = inet->inet_sport;
1ab1457c 1172
4c9483b2 1173 final_p = fl6_update_dst(&fl6, opt, &final);
20c59de2 1174 if (final_p)
987905de 1175 connected = 0;
1da177e4 1176
4c9483b2
DM
1177 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1178 fl6.flowi6_oif = np->mcast_oif;
987905de 1179 connected = 0;
c4062dfc
EH
1180 } else if (!fl6.flowi6_oif)
1181 fl6.flowi6_oif = np->ucast_oif;
1da177e4 1182
4c9483b2 1183 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 1184
4c9483b2 1185 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
68d0c6d3
DM
1186 if (IS_ERR(dst)) {
1187 err = PTR_ERR(dst);
1188 dst = NULL;
1da177e4 1189 goto out;
14e50e57 1190 }
1da177e4
LT
1191
1192 if (hlimit < 0) {
4c9483b2 1193 if (ipv6_addr_is_multicast(&fl6.daddr))
1da177e4
LT
1194 hlimit = np->mcast_hops;
1195 else
1196 hlimit = np->hop_limit;
1197 if (hlimit < 0)
6b75d090 1198 hlimit = ip6_dst_hoplimit(dst);
1da177e4
LT
1199 }
1200
e651f03a 1201 if (tclass < 0)
41a1f8ea 1202 tclass = np->tclass;
41a1f8ea 1203
13b52cd4
BH
1204 if (dontfrag < 0)
1205 dontfrag = np->dontfrag;
1206
1da177e4
LT
1207 if (msg->msg_flags&MSG_CONFIRM)
1208 goto do_confirm;
1209back_from_confirm:
1210
1211 lock_sock(sk);
1212 if (unlikely(up->pending)) {
1213 /* The socket is already corked while preparing it. */
1214 /* ... which is an evident application bug. --ANK */
1215 release_sock(sk);
1216
64ce2073 1217 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1da177e4
LT
1218 err = -EINVAL;
1219 goto out;
1220 }
1221
1222 up->pending = AF_INET6;
1223
1224do_append_data:
1225 up->len += ulen;
ba4e58ec
GR
1226 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1227 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
4c9483b2 1228 sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
41a1f8ea 1229 (struct rt6_info*)dst,
13b52cd4 1230 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1da177e4
LT
1231 if (err)
1232 udp_v6_flush_pending_frames(sk);
1233 else if (!corkreq)
4c0a6cb0 1234 err = udp_v6_push_pending_frames(sk);
1e0c14f4
HX
1235 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1236 up->pending = 0;
1da177e4 1237
a5e7c210
DM
1238 if (dst) {
1239 if (connected) {
1240 ip6_dst_store(sk, dst,
4c9483b2 1241 ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
8e1ef0a9
YH
1242 &np->daddr : NULL,
1243#ifdef CONFIG_IPV6_SUBTREES
4c9483b2 1244 ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
8e1ef0a9
YH
1245 &np->saddr :
1246#endif
1247 NULL);
a5e7c210
DM
1248 } else {
1249 dst_release(dst);
1250 }
a3c96089 1251 dst = NULL;
a5e7c210
DM
1252 }
1253
1da177e4
LT
1254 if (err > 0)
1255 err = np->recverr ? net_xmit_errno(err) : 0;
1256 release_sock(sk);
1257out:
a3c96089 1258 dst_release(dst);
1da177e4 1259 fl6_sock_release(flowlabel);
cd562c98 1260 if (!err)
1da177e4 1261 return len;
a18135eb
DM
1262 /*
1263 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1264 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1265 * we don't have a good statistic (IpOutDiscards but it can be too many
1266 * things). We could add another new stat but at least for now that
1267 * seems like overkill.
1268 */
1269 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
235b9f7a
PE
1270 UDP6_INC_STATS_USER(sock_net(sk),
1271 UDP_MIB_SNDBUFERRORS, is_udplite);
a18135eb 1272 }
1da177e4
LT
1273 return err;
1274
1275do_confirm:
1276 dst_confirm(dst);
1277 if (!(msg->msg_flags&MSG_PROBE) || len)
1278 goto back_from_confirm;
1279 err = 0;
1280 goto out;
1281}
1282
7d06b2e0 1283void udpv6_destroy_sock(struct sock *sk)
1da177e4
LT
1284{
1285 lock_sock(sk);
1286 udp_v6_flush_pending_frames(sk);
1287 release_sock(sk);
1288
1289 inet6_destroy_sock(sk);
1da177e4
LT
1290}
1291
1292/*
1293 * Socket option code for UDP
1294 */
ba4e58ec 1295int udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1296 char __user *optval, unsigned int optlen)
3fdadf7d 1297{
db8dac20 1298 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1299 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1300 udp_v6_push_pending_frames);
ba4e58ec 1301 return ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1302}
1303
1304#ifdef CONFIG_COMPAT
ba4e58ec 1305int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
b7058842 1306 char __user *optval, unsigned int optlen)
3fdadf7d 1307{
db8dac20 1308 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0
GR
1309 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1310 udp_v6_push_pending_frames);
ba4e58ec 1311 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1312}
1313#endif
1314
ba4e58ec
GR
1315int udpv6_getsockopt(struct sock *sk, int level, int optname,
1316 char __user *optval, int __user *optlen)
3fdadf7d 1317{
db8dac20 1318 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1319 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1320 return ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1321}
1322
1323#ifdef CONFIG_COMPAT
ba4e58ec
GR
1324int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1325 char __user *optval, int __user *optlen)
3fdadf7d 1326{
db8dac20 1327 if (level == SOL_UDP || level == SOL_UDPLITE)
4c0a6cb0 1328 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
ba4e58ec 1329 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1330}
1331#endif
1332
ba735425
SS
1333static int udp6_ufo_send_check(struct sk_buff *skb)
1334{
b71d1d42 1335 const struct ipv6hdr *ipv6h;
ba735425
SS
1336 struct udphdr *uh;
1337
1338 if (!pskb_may_pull(skb, sizeof(*uh)))
1339 return -EINVAL;
1340
1341 ipv6h = ipv6_hdr(skb);
1342 uh = udp_hdr(skb);
1343
1344 uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
1345 IPPROTO_UDP, 0);
1346 skb->csum_start = skb_transport_header(skb) - skb->head;
1347 skb->csum_offset = offsetof(struct udphdr, check);
1348 skb->ip_summed = CHECKSUM_PARTIAL;
1349 return 0;
1350}
1351
c8f44aff
MM
1352static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
1353 netdev_features_t features)
ba735425
SS
1354{
1355 struct sk_buff *segs = ERR_PTR(-EINVAL);
1356 unsigned int mss;
1357 unsigned int unfrag_ip6hlen, unfrag_len;
1358 struct frag_hdr *fptr;
1359 u8 *mac_start, *prevhdr;
1360 u8 nexthdr;
1361 u8 frag_hdr_sz = sizeof(struct frag_hdr);
1362 int offset;
1363 __wsum csum;
1364
1365 mss = skb_shinfo(skb)->gso_size;
1366 if (unlikely(skb->len <= mss))
1367 goto out;
1368
1369 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) {
1370 /* Packet is from an untrusted source, reset gso_segs. */
1371 int type = skb_shinfo(skb)->gso_type;
1372
1373 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) ||
1374 !(type & (SKB_GSO_UDP))))
1375 goto out;
1376
1377 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss);
1378
1379 segs = NULL;
1380 goto out;
1381 }
1382
1383 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot
1384 * do checksum of UDP packets sent as multiple IP fragments.
1385 */
96339d6c 1386 offset = skb_checksum_start_offset(skb);
3dde2598 1387 csum = skb_checksum(skb, offset, skb->len - offset, 0);
ba735425
SS
1388 offset += skb->csum_offset;
1389 *(__sum16 *)(skb->data + offset) = csum_fold(csum);
1390 skb->ip_summed = CHECKSUM_NONE;
1391
1392 /* Check if there is enough headroom to insert fragment header. */
a9cf73ea 1393 if ((skb_mac_header(skb) < skb->head + frag_hdr_sz) &&
ba735425
SS
1394 pskb_expand_head(skb, frag_hdr_sz, 0, GFP_ATOMIC))
1395 goto out;
1396
1397 /* Find the unfragmentable header and shift it left by frag_hdr_sz
1398 * bytes to insert fragment header.
1399 */
1400 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
1401 nexthdr = *prevhdr;
1402 *prevhdr = NEXTHDR_FRAGMENT;
1403 unfrag_len = skb_network_header(skb) - skb_mac_header(skb) +
1404 unfrag_ip6hlen;
1405 mac_start = skb_mac_header(skb);
1406 memmove(mac_start-frag_hdr_sz, mac_start, unfrag_len);
1407
1408 skb->mac_header -= frag_hdr_sz;
1409 skb->network_header -= frag_hdr_sz;
1410
1411 fptr = (struct frag_hdr *)(skb_network_header(skb) + unfrag_ip6hlen);
1412 fptr->nexthdr = nexthdr;
1413 fptr->reserved = 0;
87c48fa3 1414 ipv6_select_ident(fptr, (struct rt6_info *)skb_dst(skb));
ba735425
SS
1415
1416 /* Fragment the skb. ipv6 header and the remaining fields of the
1417 * fragment header are updated in ipv6_gso_segment()
1418 */
1419 segs = skb_segment(skb, features);
1420
1421out:
1422 return segs;
1423}
1424
41135cc8 1425static const struct inet6_protocol udpv6_protocol = {
1da177e4
LT
1426 .handler = udpv6_rcv,
1427 .err_handler = udpv6_err,
ba735425
SS
1428 .gso_send_check = udp6_ufo_send_check,
1429 .gso_segment = udp6_ufo_fragment,
1da177e4
LT
1430 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1431};
1432
1433/* ------------------------------------------------------------------------ */
1434#ifdef CONFIG_PROC_FS
1435
1436static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1437{
1438 struct inet_sock *inet = inet_sk(sp);
1439 struct ipv6_pinfo *np = inet6_sk(sp);
b71d1d42 1440 const struct in6_addr *dest, *src;
1da177e4
LT
1441 __u16 destp, srcp;
1442
1443 dest = &np->daddr;
1444 src = &np->rcv_saddr;
c720c7e8
ED
1445 destp = ntohs(inet->inet_dport);
1446 srcp = ntohs(inet->inet_sport);
1da177e4 1447 seq_printf(seq,
f86dcc5a 1448 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1449 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1da177e4
LT
1450 bucket,
1451 src->s6_addr32[0], src->s6_addr32[1],
1452 src->s6_addr32[2], src->s6_addr32[3], srcp,
1453 dest->s6_addr32[0], dest->s6_addr32[1],
1454 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 1455 sp->sk_state,
31e6d363
ED
1456 sk_wmem_alloc_get(sp),
1457 sk_rmem_alloc_get(sp),
1da177e4
LT
1458 0, 0L, 0,
1459 sock_i_uid(sp), 0,
1460 sock_i_ino(sp),
cb61cb9b
ED
1461 atomic_read(&sp->sk_refcnt), sp,
1462 atomic_read(&sp->sk_drops));
1da177e4
LT
1463}
1464
ba4e58ec 1465int udp6_seq_show(struct seq_file *seq, void *v)
1da177e4
LT
1466{
1467 if (v == SEQ_START_TOKEN)
1468 seq_printf(seq,
1469 " sl "
1470 "local_address "
1471 "remote_address "
1472 "st tx_queue rx_queue tr tm->when retrnsmt"
cb61cb9b 1473 " uid timeout inode ref pointer drops\n");
1da177e4
LT
1474 else
1475 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1476 return 0;
1477}
1478
73cb88ec
AV
1479static const struct file_operations udp6_afinfo_seq_fops = {
1480 .owner = THIS_MODULE,
1481 .open = udp_seq_open,
1482 .read = seq_read,
1483 .llseek = seq_lseek,
1484 .release = seq_release_net
1485};
1486
1da177e4 1487static struct udp_seq_afinfo udp6_seq_afinfo = {
1da177e4
LT
1488 .name = "udp6",
1489 .family = AF_INET6,
645ca708 1490 .udp_table = &udp_table,
73cb88ec 1491 .seq_fops = &udp6_afinfo_seq_fops,
dda61925
DL
1492 .seq_ops = {
1493 .show = udp6_seq_show,
1494 },
1da177e4
LT
1495};
1496
2c8c1e72 1497int __net_init udp6_proc_init(struct net *net)
1da177e4 1498{
0c96d8c5 1499 return udp_proc_register(net, &udp6_seq_afinfo);
1da177e4
LT
1500}
1501
0c96d8c5
DL
1502void udp6_proc_exit(struct net *net) {
1503 udp_proc_unregister(net, &udp6_seq_afinfo);
1da177e4
LT
1504}
1505#endif /* CONFIG_PROC_FS */
1506
1507/* ------------------------------------------------------------------------ */
1508
1509struct proto udpv6_prot = {
543d9cfe
ACM
1510 .name = "UDPv6",
1511 .owner = THIS_MODULE,
ba4e58ec 1512 .close = udp_lib_close,
543d9cfe
ACM
1513 .connect = ip6_datagram_connect,
1514 .disconnect = udp_disconnect,
1515 .ioctl = udp_ioctl,
1516 .destroy = udpv6_destroy_sock,
1517 .setsockopt = udpv6_setsockopt,
1518 .getsockopt = udpv6_getsockopt,
1519 .sendmsg = udpv6_sendmsg,
1520 .recvmsg = udpv6_recvmsg,
f7ad74fe 1521 .backlog_rcv = __udpv6_queue_rcv_skb,
ba4e58ec
GR
1522 .hash = udp_lib_hash,
1523 .unhash = udp_lib_unhash,
719f8358 1524 .rehash = udp_v6_rehash,
543d9cfe 1525 .get_port = udp_v6_get_port,
95766fff
HA
1526 .memory_allocated = &udp_memory_allocated,
1527 .sysctl_mem = sysctl_udp_mem,
1528 .sysctl_wmem = &sysctl_udp_wmem_min,
1529 .sysctl_rmem = &sysctl_udp_rmem_min,
543d9cfe 1530 .obj_size = sizeof(struct udp6_sock),
271b72c7 1531 .slab_flags = SLAB_DESTROY_BY_RCU,
645ca708 1532 .h.udp_table = &udp_table,
3fdadf7d 1533#ifdef CONFIG_COMPAT
543d9cfe
ACM
1534 .compat_setsockopt = compat_udpv6_setsockopt,
1535 .compat_getsockopt = compat_udpv6_getsockopt,
3fdadf7d 1536#endif
fcbdf09d 1537 .clear_sk = sk_prot_clear_portaddr_nulls,
1da177e4
LT
1538};
1539
1da177e4
LT
1540static struct inet_protosw udpv6_protosw = {
1541 .type = SOCK_DGRAM,
1542 .protocol = IPPROTO_UDP,
1543 .prot = &udpv6_prot,
1544 .ops = &inet6_dgram_ops,
1da177e4
LT
1545 .no_check = UDP_CSUM_DEFAULT,
1546 .flags = INET_PROTOSW_PERMANENT,
1547};
1548
1549
7f4e4868 1550int __init udpv6_init(void)
1da177e4 1551{
7f4e4868
DL
1552 int ret;
1553
1554 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1555 if (ret)
1556 goto out;
1557
1558 ret = inet6_register_protosw(&udpv6_protosw);
1559 if (ret)
1560 goto out_udpv6_protocol;
1561out:
1562 return ret;
1563
1564out_udpv6_protocol:
1565 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1566 goto out;
1567}
1568
09f7709f 1569void udpv6_exit(void)
7f4e4868
DL
1570{
1571 inet6_unregister_protosw(&udpv6_protosw);
1572 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1da177e4 1573}