projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
userns: Print out socket uids in a user namespace aware fashion.
[deliverable/linux.git] / net / ipv6 / raw.c
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/raw.c
9 *
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
14 *
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
19 */
20
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <linux/compat.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
37
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
57
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
61
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64 #include <linux/export.h>
65
66 static struct raw_hashinfo raw_v6_hashinfo = {
67 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
68 };
69
70 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
71 unsigned short num, const struct in6_addr *loc_addr,
72 const struct in6_addr *rmt_addr, int dif)
73 {
74 struct hlist_node *node;
75 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
76
77 sk_for_each_from(sk, node)
78 if (inet_sk(sk)->inet_num == num) {
79 struct ipv6_pinfo *np = inet6_sk(sk);
80
81 if (!net_eq(sock_net(sk), net))
82 continue;
83
84 if (!ipv6_addr_any(&np->daddr) &&
85 !ipv6_addr_equal(&np->daddr, rmt_addr))
86 continue;
87
88 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
89 continue;
90
91 if (!ipv6_addr_any(&np->rcv_saddr)) {
92 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
93 goto found;
94 if (is_multicast &&
95 inet6_mc_check(sk, loc_addr, rmt_addr))
96 goto found;
97 continue;
98 }
99 goto found;
100 }
101 sk = NULL;
102 found:
103 return sk;
104 }
105
106 /*
107 * 0 - deliver
108 * 1 - block
109 */
110 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
111 {
112 struct icmp6hdr *icmph;
113 struct raw6_sock *rp = raw6_sk(sk);
114
115 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
116 __u32 *data = &rp->filter.data[0];
117 int bit_nr;
118
119 icmph = (struct icmp6hdr *) skb->data;
120 bit_nr = icmph->icmp6_type;
121
122 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
123 }
124 return 0;
125 }
126
127 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
128 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
129
130 static mh_filter_t __rcu *mh_filter __read_mostly;
131
132 int rawv6_mh_filter_register(mh_filter_t filter)
133 {
134 rcu_assign_pointer(mh_filter, filter);
135 return 0;
136 }
137 EXPORT_SYMBOL(rawv6_mh_filter_register);
138
139 int rawv6_mh_filter_unregister(mh_filter_t filter)
140 {
141 RCU_INIT_POINTER(mh_filter, NULL);
142 synchronize_rcu();
143 return 0;
144 }
145 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
146
147 #endif
148
149 /*
150 * demultiplex raw sockets.
151 * (should consider queueing the skb in the sock receive_queue
152 * without calling rawv6.c)
153 *
154 * Caller owns SKB so we must make clones.
155 */
156 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
157 {
158 const struct in6_addr *saddr;
159 const struct in6_addr *daddr;
160 struct sock *sk;
161 bool delivered = false;
162 __u8 hash;
163 struct net *net;
164
165 saddr = &ipv6_hdr(skb)->saddr;
166 daddr = saddr + 1;
167
168 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
169
170 read_lock(&raw_v6_hashinfo.lock);
171 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
172
173 if (sk == NULL)
174 goto out;
175
176 net = dev_net(skb->dev);
177 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
178
179 while (sk) {
180 int filtered;
181
182 delivered = true;
183 switch (nexthdr) {
184 case IPPROTO_ICMPV6:
185 filtered = icmpv6_filter(sk, skb);
186 break;
187
188 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
189 case IPPROTO_MH:
190 {
191 /* XXX: To validate MH only once for each packet,
192 * this is placed here. It should be after checking
193 * xfrm policy, however it doesn't. The checking xfrm
194 * policy is placed in rawv6_rcv() because it is
195 * required for each socket.
196 */
197 mh_filter_t *filter;
198
199 filter = rcu_dereference(mh_filter);
200 filtered = filter ? (*filter)(sk, skb) : 0;
201 break;
202 }
203 #endif
204 default:
205 filtered = 0;
206 break;
207 }
208
209 if (filtered < 0)
210 break;
211 if (filtered == 0) {
212 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
213
214 /* Not releasing hash table! */
215 if (clone) {
216 nf_reset(clone);
217 rawv6_rcv(sk, clone);
218 }
219 }
220 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
221 IP6CB(skb)->iif);
222 }
223 out:
224 read_unlock(&raw_v6_hashinfo.lock);
225 return delivered;
226 }
227
228 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
229 {
230 struct sock *raw_sk;
231
232 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
233 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
234 raw_sk = NULL;
235
236 return raw_sk != NULL;
237 }
238
239 /* This cleans up af_inet6 a bit. -DaveM */
240 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
241 {
242 struct inet_sock *inet = inet_sk(sk);
243 struct ipv6_pinfo *np = inet6_sk(sk);
244 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
245 __be32 v4addr = 0;
246 int addr_type;
247 int err;
248
249 if (addr_len < SIN6_LEN_RFC2133)
250 return -EINVAL;
251 addr_type = ipv6_addr_type(&addr->sin6_addr);
252
253 /* Raw sockets are IPv6 only */
254 if (addr_type == IPV6_ADDR_MAPPED)
255 return -EADDRNOTAVAIL;
256
257 lock_sock(sk);
258
259 err = -EINVAL;
260 if (sk->sk_state != TCP_CLOSE)
261 goto out;
262
263 rcu_read_lock();
264 /* Check if the address belongs to the host. */
265 if (addr_type != IPV6_ADDR_ANY) {
266 struct net_device *dev = NULL;
267
268 if (addr_type & IPV6_ADDR_LINKLOCAL) {
269 if (addr_len >= sizeof(struct sockaddr_in6) &&
270 addr->sin6_scope_id) {
271 /* Override any existing binding, if another
272 * one is supplied by user.
273 */
274 sk->sk_bound_dev_if = addr->sin6_scope_id;
275 }
276
277 /* Binding to link-local address requires an interface */
278 if (!sk->sk_bound_dev_if)
279 goto out_unlock;
280
281 err = -ENODEV;
282 dev = dev_get_by_index_rcu(sock_net(sk),
283 sk->sk_bound_dev_if);
284 if (!dev)
285 goto out_unlock;
286 }
287
288 /* ipv4 addr of the socket is invalid. Only the
289 * unspecified and mapped address have a v4 equivalent.
290 */
291 v4addr = LOOPBACK4_IPV6;
292 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
293 err = -EADDRNOTAVAIL;
294 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
295 dev, 0)) {
296 goto out_unlock;
297 }
298 }
299 }
300
301 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
302 np->rcv_saddr = addr->sin6_addr;
303 if (!(addr_type & IPV6_ADDR_MULTICAST))
304 np->saddr = addr->sin6_addr;
305 err = 0;
306 out_unlock:
307 rcu_read_unlock();
308 out:
309 release_sock(sk);
310 return err;
311 }
312
313 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
314 struct inet6_skb_parm *opt,
315 u8 type, u8 code, int offset, __be32 info)
316 {
317 struct inet_sock *inet = inet_sk(sk);
318 struct ipv6_pinfo *np = inet6_sk(sk);
319 int err;
320 int harderr;
321
322 /* Report error on raw socket, if:
323 1. User requested recverr.
324 2. Socket is connected (otherwise the error indication
325 is useless without recverr and error is hard.
326 */
327 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
328 return;
329
330 harderr = icmpv6_err_convert(type, code, &err);
331 if (type == ICMPV6_PKT_TOOBIG) {
332 ip6_sk_update_pmtu(skb, sk, info);
333 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
334 }
335 if (type == NDISC_REDIRECT)
336 ip6_sk_redirect(skb, sk);
337 if (np->recverr) {
338 u8 *payload = skb->data;
339 if (!inet->hdrincl)
340 payload += offset;
341 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
342 }
343
344 if (np->recverr || harderr) {
345 sk->sk_err = err;
346 sk->sk_error_report(sk);
347 }
348 }
349
350 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
351 u8 type, u8 code, int inner_offset, __be32 info)
352 {
353 struct sock *sk;
354 int hash;
355 const struct in6_addr *saddr, *daddr;
356 struct net *net;
357
358 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
359
360 read_lock(&raw_v6_hashinfo.lock);
361 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
362 if (sk != NULL) {
363 /* Note: ipv6_hdr(skb) != skb->data */
364 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
365 saddr = &ip6h->saddr;
366 daddr = &ip6h->daddr;
367 net = dev_net(skb->dev);
368
369 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
370 IP6CB(skb)->iif))) {
371 rawv6_err(sk, skb, NULL, type, code,
372 inner_offset, info);
373 sk = sk_next(sk);
374 }
375 }
376 read_unlock(&raw_v6_hashinfo.lock);
377 }
378
379 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
380 {
381 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
382 skb_checksum_complete(skb)) {
383 atomic_inc(&sk->sk_drops);
384 kfree_skb(skb);
385 return NET_RX_DROP;
386 }
387
388 /* Charge it to the socket. */
389 skb_dst_drop(skb);
390 if (sock_queue_rcv_skb(sk, skb) < 0) {
391 kfree_skb(skb);
392 return NET_RX_DROP;
393 }
394
395 return 0;
396 }
397
398 /*
399 * This is next to useless...
400 * if we demultiplex in network layer we don't need the extra call
401 * just to queue the skb...
402 * maybe we could have the network decide upon a hint if it
403 * should call raw_rcv for demultiplexing
404 */
405 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
406 {
407 struct inet_sock *inet = inet_sk(sk);
408 struct raw6_sock *rp = raw6_sk(sk);
409
410 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
411 atomic_inc(&sk->sk_drops);
412 kfree_skb(skb);
413 return NET_RX_DROP;
414 }
415
416 if (!rp->checksum)
417 skb->ip_summed = CHECKSUM_UNNECESSARY;
418
419 if (skb->ip_summed == CHECKSUM_COMPLETE) {
420 skb_postpull_rcsum(skb, skb_network_header(skb),
421 skb_network_header_len(skb));
422 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
423 &ipv6_hdr(skb)->daddr,
424 skb->len, inet->inet_num, skb->csum))
425 skb->ip_summed = CHECKSUM_UNNECESSARY;
426 }
427 if (!skb_csum_unnecessary(skb))
428 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
429 &ipv6_hdr(skb)->daddr,
430 skb->len,
431 inet->inet_num, 0));
432
433 if (inet->hdrincl) {
434 if (skb_checksum_complete(skb)) {
435 atomic_inc(&sk->sk_drops);
436 kfree_skb(skb);
437 return NET_RX_DROP;
438 }
439 }
440
441 rawv6_rcv_skb(sk, skb);
442 return 0;
443 }
444
445
446 /*
447 * This should be easy, if there is something there
448 * we return it, otherwise we block.
449 */
450
451 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
452 struct msghdr *msg, size_t len,
453 int noblock, int flags, int *addr_len)
454 {
455 struct ipv6_pinfo *np = inet6_sk(sk);
456 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
457 struct sk_buff *skb;
458 size_t copied;
459 int err;
460
461 if (flags & MSG_OOB)
462 return -EOPNOTSUPP;
463
464 if (addr_len)
465 *addr_len=sizeof(*sin6);
466
467 if (flags & MSG_ERRQUEUE)
468 return ipv6_recv_error(sk, msg, len);
469
470 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
471 return ipv6_recv_rxpmtu(sk, msg, len);
472
473 skb = skb_recv_datagram(sk, flags, noblock, &err);
474 if (!skb)
475 goto out;
476
477 copied = skb->len;
478 if (copied > len) {
479 copied = len;
480 msg->msg_flags |= MSG_TRUNC;
481 }
482
483 if (skb_csum_unnecessary(skb)) {
484 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
485 } else if (msg->msg_flags&MSG_TRUNC) {
486 if (__skb_checksum_complete(skb))
487 goto csum_copy_err;
488 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
489 } else {
490 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
491 if (err == -EINVAL)
492 goto csum_copy_err;
493 }
494 if (err)
495 goto out_free;
496
497 /* Copy the address. */
498 if (sin6) {
499 sin6->sin6_family = AF_INET6;
500 sin6->sin6_port = 0;
501 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
502 sin6->sin6_flowinfo = 0;
503 sin6->sin6_scope_id = 0;
504 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
505 sin6->sin6_scope_id = IP6CB(skb)->iif;
506 }
507
508 sock_recv_ts_and_drops(msg, sk, skb);
509
510 if (np->rxopt.all)
511 datagram_recv_ctl(sk, msg, skb);
512
513 err = copied;
514 if (flags & MSG_TRUNC)
515 err = skb->len;
516
517 out_free:
518 skb_free_datagram(sk, skb);
519 out:
520 return err;
521
522 csum_copy_err:
523 skb_kill_datagram(sk, skb, flags);
524
525 /* Error for blocking case is chosen to masquerade
526 as some normal condition.
527 */
528 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
529 goto out;
530 }
531
532 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
533 struct raw6_sock *rp)
534 {
535 struct sk_buff *skb;
536 int err = 0;
537 int offset;
538 int len;
539 int total_len;
540 __wsum tmp_csum;
541 __sum16 csum;
542
543 if (!rp->checksum)
544 goto send;
545
546 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
547 goto out;
548
549 offset = rp->offset;
550 total_len = inet_sk(sk)->cork.base.length;
551 if (offset >= total_len - 1) {
552 err = -EINVAL;
553 ip6_flush_pending_frames(sk);
554 goto out;
555 }
556
557 /* should be check HW csum miyazawa */
558 if (skb_queue_len(&sk->sk_write_queue) == 1) {
559 /*
560 * Only one fragment on the socket.
561 */
562 tmp_csum = skb->csum;
563 } else {
564 struct sk_buff *csum_skb = NULL;
565 tmp_csum = 0;
566
567 skb_queue_walk(&sk->sk_write_queue, skb) {
568 tmp_csum = csum_add(tmp_csum, skb->csum);
569
570 if (csum_skb)
571 continue;
572
573 len = skb->len - skb_transport_offset(skb);
574 if (offset >= len) {
575 offset -= len;
576 continue;
577 }
578
579 csum_skb = skb;
580 }
581
582 skb = csum_skb;
583 }
584
585 offset += skb_transport_offset(skb);
586 if (skb_copy_bits(skb, offset, &csum, 2))
587 BUG();
588
589 /* in case cksum was not initialized */
590 if (unlikely(csum))
591 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
592
593 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
594 total_len, fl6->flowi6_proto, tmp_csum);
595
596 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
597 csum = CSUM_MANGLED_0;
598
599 if (skb_store_bits(skb, offset, &csum, 2))
600 BUG();
601
602 send:
603 err = ip6_push_pending_frames(sk);
604 out:
605 return err;
606 }
607
608 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
609 struct flowi6 *fl6, struct dst_entry **dstp,
610 unsigned int flags)
611 {
612 struct ipv6_pinfo *np = inet6_sk(sk);
613 struct ipv6hdr *iph;
614 struct sk_buff *skb;
615 int err;
616 struct rt6_info *rt = (struct rt6_info *)*dstp;
617 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
618 int tlen = rt->dst.dev->needed_tailroom;
619
620 if (length > rt->dst.dev->mtu) {
621 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
622 return -EMSGSIZE;
623 }
624 if (flags&MSG_PROBE)
625 goto out;
626
627 skb = sock_alloc_send_skb(sk,
628 length + hlen + tlen + 15,
629 flags & MSG_DONTWAIT, &err);
630 if (skb == NULL)
631 goto error;
632 skb_reserve(skb, hlen);
633
634 skb->priority = sk->sk_priority;
635 skb->mark = sk->sk_mark;
636 skb_dst_set(skb, &rt->dst);
637 *dstp = NULL;
638
639 skb_put(skb, length);
640 skb_reset_network_header(skb);
641 iph = ipv6_hdr(skb);
642
643 skb->ip_summed = CHECKSUM_NONE;
644
645 skb->transport_header = skb->network_header;
646 err = memcpy_fromiovecend((void *)iph, from, 0, length);
647 if (err)
648 goto error_fault;
649
650 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
651 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
652 rt->dst.dev, dst_output);
653 if (err > 0)
654 err = net_xmit_errno(err);
655 if (err)
656 goto error;
657 out:
658 return 0;
659
660 error_fault:
661 err = -EFAULT;
662 kfree_skb(skb);
663 error:
664 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
665 if (err == -ENOBUFS && !np->recverr)
666 err = 0;
667 return err;
668 }
669
670 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
671 {
672 struct iovec *iov;
673 u8 __user *type = NULL;
674 u8 __user *code = NULL;
675 u8 len = 0;
676 int probed = 0;
677 int i;
678
679 if (!msg->msg_iov)
680 return 0;
681
682 for (i = 0; i < msg->msg_iovlen; i++) {
683 iov = &msg->msg_iov[i];
684 if (!iov)
685 continue;
686
687 switch (fl6->flowi6_proto) {
688 case IPPROTO_ICMPV6:
689 /* check if one-byte field is readable or not. */
690 if (iov->iov_base && iov->iov_len < 1)
691 break;
692
693 if (!type) {
694 type = iov->iov_base;
695 /* check if code field is readable or not. */
696 if (iov->iov_len > 1)
697 code = type + 1;
698 } else if (!code)
699 code = iov->iov_base;
700
701 if (type && code) {
702 if (get_user(fl6->fl6_icmp_type, type) ||
703 get_user(fl6->fl6_icmp_code, code))
704 return -EFAULT;
705 probed = 1;
706 }
707 break;
708 case IPPROTO_MH:
709 if (iov->iov_base && iov->iov_len < 1)
710 break;
711 /* check if type field is readable or not. */
712 if (iov->iov_len > 2 - len) {
713 u8 __user *p = iov->iov_base;
714 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
715 return -EFAULT;
716 probed = 1;
717 } else
718 len += iov->iov_len;
719
720 break;
721 default:
722 probed = 1;
723 break;
724 }
725 if (probed)
726 break;
727 }
728 return 0;
729 }
730
731 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
732 struct msghdr *msg, size_t len)
733 {
734 struct ipv6_txoptions opt_space;
735 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
736 struct in6_addr *daddr, *final_p, final;
737 struct inet_sock *inet = inet_sk(sk);
738 struct ipv6_pinfo *np = inet6_sk(sk);
739 struct raw6_sock *rp = raw6_sk(sk);
740 struct ipv6_txoptions *opt = NULL;
741 struct ip6_flowlabel *flowlabel = NULL;
742 struct dst_entry *dst = NULL;
743 struct flowi6 fl6;
744 int addr_len = msg->msg_namelen;
745 int hlimit = -1;
746 int tclass = -1;
747 int dontfrag = -1;
748 u16 proto;
749 int err;
750
751 /* Rough check on arithmetic overflow,
752 better check is made in ip6_append_data().
753 */
754 if (len > INT_MAX)
755 return -EMSGSIZE;
756
757 /* Mirror BSD error message compatibility */
758 if (msg->msg_flags & MSG_OOB)
759 return -EOPNOTSUPP;
760
761 /*
762 * Get and verify the address.
763 */
764 memset(&fl6, 0, sizeof(fl6));
765
766 fl6.flowi6_mark = sk->sk_mark;
767
768 if (sin6) {
769 if (addr_len < SIN6_LEN_RFC2133)
770 return -EINVAL;
771
772 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
773 return -EAFNOSUPPORT;
774
775 /* port is the proto value [0..255] carried in nexthdr */
776 proto = ntohs(sin6->sin6_port);
777
778 if (!proto)
779 proto = inet->inet_num;
780 else if (proto != inet->inet_num)
781 return -EINVAL;
782
783 if (proto > 255)
784 return -EINVAL;
785
786 daddr = &sin6->sin6_addr;
787 if (np->sndflow) {
788 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
789 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
790 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
791 if (flowlabel == NULL)
792 return -EINVAL;
793 daddr = &flowlabel->dst;
794 }
795 }
796
797 /*
798 * Otherwise it will be difficult to maintain
799 * sk->sk_dst_cache.
800 */
801 if (sk->sk_state == TCP_ESTABLISHED &&
802 ipv6_addr_equal(daddr, &np->daddr))
803 daddr = &np->daddr;
804
805 if (addr_len >= sizeof(struct sockaddr_in6) &&
806 sin6->sin6_scope_id &&
807 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
808 fl6.flowi6_oif = sin6->sin6_scope_id;
809 } else {
810 if (sk->sk_state != TCP_ESTABLISHED)
811 return -EDESTADDRREQ;
812
813 proto = inet->inet_num;
814 daddr = &np->daddr;
815 fl6.flowlabel = np->flow_label;
816 }
817
818 if (fl6.flowi6_oif == 0)
819 fl6.flowi6_oif = sk->sk_bound_dev_if;
820
821 if (msg->msg_controllen) {
822 opt = &opt_space;
823 memset(opt, 0, sizeof(struct ipv6_txoptions));
824 opt->tot_len = sizeof(struct ipv6_txoptions);
825
826 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
827 &hlimit, &tclass, &dontfrag);
828 if (err < 0) {
829 fl6_sock_release(flowlabel);
830 return err;
831 }
832 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
833 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
834 if (flowlabel == NULL)
835 return -EINVAL;
836 }
837 if (!(opt->opt_nflen|opt->opt_flen))
838 opt = NULL;
839 }
840 if (opt == NULL)
841 opt = np->opt;
842 if (flowlabel)
843 opt = fl6_merge_options(&opt_space, flowlabel, opt);
844 opt = ipv6_fixup_options(&opt_space, opt);
845
846 fl6.flowi6_proto = proto;
847 err = rawv6_probe_proto_opt(&fl6, msg);
848 if (err)
849 goto out;
850
851 if (!ipv6_addr_any(daddr))
852 fl6.daddr = *daddr;
853 else
854 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
855 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
856 fl6.saddr = np->saddr;
857
858 final_p = fl6_update_dst(&fl6, opt, &final);
859
860 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
861 fl6.flowi6_oif = np->mcast_oif;
862 else if (!fl6.flowi6_oif)
863 fl6.flowi6_oif = np->ucast_oif;
864 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
865
866 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
867 if (IS_ERR(dst)) {
868 err = PTR_ERR(dst);
869 goto out;
870 }
871 if (hlimit < 0) {
872 if (ipv6_addr_is_multicast(&fl6.daddr))
873 hlimit = np->mcast_hops;
874 else
875 hlimit = np->hop_limit;
876 if (hlimit < 0)
877 hlimit = ip6_dst_hoplimit(dst);
878 }
879
880 if (tclass < 0)
881 tclass = np->tclass;
882
883 if (dontfrag < 0)
884 dontfrag = np->dontfrag;
885
886 if (msg->msg_flags&MSG_CONFIRM)
887 goto do_confirm;
888
889 back_from_confirm:
890 if (inet->hdrincl)
891 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
892 else {
893 lock_sock(sk);
894 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
895 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
896 msg->msg_flags, dontfrag);
897
898 if (err)
899 ip6_flush_pending_frames(sk);
900 else if (!(msg->msg_flags & MSG_MORE))
901 err = rawv6_push_pending_frames(sk, &fl6, rp);
902 release_sock(sk);
903 }
904 done:
905 dst_release(dst);
906 out:
907 fl6_sock_release(flowlabel);
908 return err<0?err:len;
909 do_confirm:
910 dst_confirm(dst);
911 if (!(msg->msg_flags & MSG_PROBE) || len)
912 goto back_from_confirm;
913 err = 0;
914 goto done;
915 }
916
917 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
918 char __user *optval, int optlen)
919 {
920 switch (optname) {
921 case ICMPV6_FILTER:
922 if (optlen > sizeof(struct icmp6_filter))
923 optlen = sizeof(struct icmp6_filter);
924 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
925 return -EFAULT;
926 return 0;
927 default:
928 return -ENOPROTOOPT;
929 }
930
931 return 0;
932 }
933
934 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
935 char __user *optval, int __user *optlen)
936 {
937 int len;
938
939 switch (optname) {
940 case ICMPV6_FILTER:
941 if (get_user(len, optlen))
942 return -EFAULT;
943 if (len < 0)
944 return -EINVAL;
945 if (len > sizeof(struct icmp6_filter))
946 len = sizeof(struct icmp6_filter);
947 if (put_user(len, optlen))
948 return -EFAULT;
949 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
950 return -EFAULT;
951 return 0;
952 default:
953 return -ENOPROTOOPT;
954 }
955
956 return 0;
957 }
958
959
960 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
961 char __user *optval, unsigned int optlen)
962 {
963 struct raw6_sock *rp = raw6_sk(sk);
964 int val;
965
966 if (get_user(val, (int __user *)optval))
967 return -EFAULT;
968
969 switch (optname) {
970 case IPV6_CHECKSUM:
971 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
972 level == IPPROTO_IPV6) {
973 /*
974 * RFC3542 tells that IPV6_CHECKSUM socket
975 * option in the IPPROTO_IPV6 level is not
976 * allowed on ICMPv6 sockets.
977 * If you want to set it, use IPPROTO_RAW
978 * level IPV6_CHECKSUM socket option
979 * (Linux extension).
980 */
981 return -EINVAL;
982 }
983
984 /* You may get strange result with a positive odd offset;
985 RFC2292bis agrees with me. */
986 if (val > 0 && (val&1))
987 return -EINVAL;
988 if (val < 0) {
989 rp->checksum = 0;
990 } else {
991 rp->checksum = 1;
992 rp->offset = val;
993 }
994
995 return 0;
996
997 default:
998 return -ENOPROTOOPT;
999 }
1000 }
1001
1002 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1003 char __user *optval, unsigned int optlen)
1004 {
1005 switch (level) {
1006 case SOL_RAW:
1007 break;
1008
1009 case SOL_ICMPV6:
1010 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1011 return -EOPNOTSUPP;
1012 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1013 case SOL_IPV6:
1014 if (optname == IPV6_CHECKSUM)
1015 break;
1016 default:
1017 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1018 }
1019
1020 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1021 }
1022
1023 #ifdef CONFIG_COMPAT
1024 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1025 char __user *optval, unsigned int optlen)
1026 {
1027 switch (level) {
1028 case SOL_RAW:
1029 break;
1030 case SOL_ICMPV6:
1031 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1032 return -EOPNOTSUPP;
1033 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1034 case SOL_IPV6:
1035 if (optname == IPV6_CHECKSUM)
1036 break;
1037 default:
1038 return compat_ipv6_setsockopt(sk, level, optname,
1039 optval, optlen);
1040 }
1041 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1042 }
1043 #endif
1044
1045 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1046 char __user *optval, int __user *optlen)
1047 {
1048 struct raw6_sock *rp = raw6_sk(sk);
1049 int val, len;
1050
1051 if (get_user(len,optlen))
1052 return -EFAULT;
1053
1054 switch (optname) {
1055 case IPV6_CHECKSUM:
1056 /*
1057 * We allow getsockopt() for IPPROTO_IPV6-level
1058 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1059 * since RFC3542 is silent about it.
1060 */
1061 if (rp->checksum == 0)
1062 val = -1;
1063 else
1064 val = rp->offset;
1065 break;
1066
1067 default:
1068 return -ENOPROTOOPT;
1069 }
1070
1071 len = min_t(unsigned int, sizeof(int), len);
1072
1073 if (put_user(len, optlen))
1074 return -EFAULT;
1075 if (copy_to_user(optval,&val,len))
1076 return -EFAULT;
1077 return 0;
1078 }
1079
1080 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1081 char __user *optval, int __user *optlen)
1082 {
1083 switch (level) {
1084 case SOL_RAW:
1085 break;
1086
1087 case SOL_ICMPV6:
1088 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1089 return -EOPNOTSUPP;
1090 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1091 case SOL_IPV6:
1092 if (optname == IPV6_CHECKSUM)
1093 break;
1094 default:
1095 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1096 }
1097
1098 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1099 }
1100
1101 #ifdef CONFIG_COMPAT
1102 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1103 char __user *optval, int __user *optlen)
1104 {
1105 switch (level) {
1106 case SOL_RAW:
1107 break;
1108 case SOL_ICMPV6:
1109 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1110 return -EOPNOTSUPP;
1111 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1112 case SOL_IPV6:
1113 if (optname == IPV6_CHECKSUM)
1114 break;
1115 default:
1116 return compat_ipv6_getsockopt(sk, level, optname,
1117 optval, optlen);
1118 }
1119 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1120 }
1121 #endif
1122
1123 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1124 {
1125 switch (cmd) {
1126 case SIOCOUTQ: {
1127 int amount = sk_wmem_alloc_get(sk);
1128
1129 return put_user(amount, (int __user *)arg);
1130 }
1131 case SIOCINQ: {
1132 struct sk_buff *skb;
1133 int amount = 0;
1134
1135 spin_lock_bh(&sk->sk_receive_queue.lock);
1136 skb = skb_peek(&sk->sk_receive_queue);
1137 if (skb != NULL)
1138 amount = skb->tail - skb->transport_header;
1139 spin_unlock_bh(&sk->sk_receive_queue.lock);
1140 return put_user(amount, (int __user *)arg);
1141 }
1142
1143 default:
1144 #ifdef CONFIG_IPV6_MROUTE
1145 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1146 #else
1147 return -ENOIOCTLCMD;
1148 #endif
1149 }
1150 }
1151
1152 #ifdef CONFIG_COMPAT
1153 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1154 {
1155 switch (cmd) {
1156 case SIOCOUTQ:
1157 case SIOCINQ:
1158 return -ENOIOCTLCMD;
1159 default:
1160 #ifdef CONFIG_IPV6_MROUTE
1161 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1162 #else
1163 return -ENOIOCTLCMD;
1164 #endif
1165 }
1166 }
1167 #endif
1168
1169 static void rawv6_close(struct sock *sk, long timeout)
1170 {
1171 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1172 ip6_ra_control(sk, -1);
1173 ip6mr_sk_done(sk);
1174 sk_common_release(sk);
1175 }
1176
1177 static void raw6_destroy(struct sock *sk)
1178 {
1179 lock_sock(sk);
1180 ip6_flush_pending_frames(sk);
1181 release_sock(sk);
1182
1183 inet6_destroy_sock(sk);
1184 }
1185
1186 static int rawv6_init_sk(struct sock *sk)
1187 {
1188 struct raw6_sock *rp = raw6_sk(sk);
1189
1190 switch (inet_sk(sk)->inet_num) {
1191 case IPPROTO_ICMPV6:
1192 rp->checksum = 1;
1193 rp->offset = 2;
1194 break;
1195 case IPPROTO_MH:
1196 rp->checksum = 1;
1197 rp->offset = 4;
1198 break;
1199 default:
1200 break;
1201 }
1202 return 0;
1203 }
1204
1205 struct proto rawv6_prot = {
1206 .name = "RAWv6",
1207 .owner = THIS_MODULE,
1208 .close = rawv6_close,
1209 .destroy = raw6_destroy,
1210 .connect = ip6_datagram_connect,
1211 .disconnect = udp_disconnect,
1212 .ioctl = rawv6_ioctl,
1213 .init = rawv6_init_sk,
1214 .setsockopt = rawv6_setsockopt,
1215 .getsockopt = rawv6_getsockopt,
1216 .sendmsg = rawv6_sendmsg,
1217 .recvmsg = rawv6_recvmsg,
1218 .bind = rawv6_bind,
1219 .backlog_rcv = rawv6_rcv_skb,
1220 .hash = raw_hash_sk,
1221 .unhash = raw_unhash_sk,
1222 .obj_size = sizeof(struct raw6_sock),
1223 .h.raw_hash = &raw_v6_hashinfo,
1224 #ifdef CONFIG_COMPAT
1225 .compat_setsockopt = compat_rawv6_setsockopt,
1226 .compat_getsockopt = compat_rawv6_getsockopt,
1227 .compat_ioctl = compat_rawv6_ioctl,
1228 #endif
1229 };
1230
1231 #ifdef CONFIG_PROC_FS
1232 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1233 {
1234 struct ipv6_pinfo *np = inet6_sk(sp);
1235 const struct in6_addr *dest, *src;
1236 __u16 destp, srcp;
1237
1238 dest = &np->daddr;
1239 src = &np->rcv_saddr;
1240 destp = 0;
1241 srcp = inet_sk(sp)->inet_num;
1242 seq_printf(seq,
1243 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1244 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1245 i,
1246 src->s6_addr32[0], src->s6_addr32[1],
1247 src->s6_addr32[2], src->s6_addr32[3], srcp,
1248 dest->s6_addr32[0], dest->s6_addr32[1],
1249 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1250 sp->sk_state,
1251 sk_wmem_alloc_get(sp),
1252 sk_rmem_alloc_get(sp),
1253 0, 0L, 0,
1254 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1255 0,
1256 sock_i_ino(sp),
1257 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1258 }
1259
1260 static int raw6_seq_show(struct seq_file *seq, void *v)
1261 {
1262 if (v == SEQ_START_TOKEN)
1263 seq_printf(seq,
1264 " sl "
1265 "local_address "
1266 "remote_address "
1267 "st tx_queue rx_queue tr tm->when retrnsmt"
1268 " uid timeout inode ref pointer drops\n");
1269 else
1270 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1271 return 0;
1272 }
1273
1274 static const struct seq_operations raw6_seq_ops = {
1275 .start = raw_seq_start,
1276 .next = raw_seq_next,
1277 .stop = raw_seq_stop,
1278 .show = raw6_seq_show,
1279 };
1280
1281 static int raw6_seq_open(struct inode *inode, struct file *file)
1282 {
1283 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1284 }
1285
1286 static const struct file_operations raw6_seq_fops = {
1287 .owner = THIS_MODULE,
1288 .open = raw6_seq_open,
1289 .read = seq_read,
1290 .llseek = seq_lseek,
1291 .release = seq_release_net,
1292 };
1293
1294 static int __net_init raw6_init_net(struct net *net)
1295 {
1296 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1297 return -ENOMEM;
1298
1299 return 0;
1300 }
1301
1302 static void __net_exit raw6_exit_net(struct net *net)
1303 {
1304 proc_net_remove(net, "raw6");
1305 }
1306
1307 static struct pernet_operations raw6_net_ops = {
1308 .init = raw6_init_net,
1309 .exit = raw6_exit_net,
1310 };
1311
1312 int __init raw6_proc_init(void)
1313 {
1314 return register_pernet_subsys(&raw6_net_ops);
1315 }
1316
1317 void raw6_proc_exit(void)
1318 {
1319 unregister_pernet_subsys(&raw6_net_ops);
1320 }
1321 #endif /* CONFIG_PROC_FS */
1322
1323 /* Same as inet6_dgram_ops, sans udp_poll. */
1324 static const struct proto_ops inet6_sockraw_ops = {
1325 .family = PF_INET6,
1326 .owner = THIS_MODULE,
1327 .release = inet6_release,
1328 .bind = inet6_bind,
1329 .connect = inet_dgram_connect, /* ok */
1330 .socketpair = sock_no_socketpair, /* a do nothing */
1331 .accept = sock_no_accept, /* a do nothing */
1332 .getname = inet6_getname,
1333 .poll = datagram_poll, /* ok */
1334 .ioctl = inet6_ioctl, /* must change */
1335 .listen = sock_no_listen, /* ok */
1336 .shutdown = inet_shutdown, /* ok */
1337 .setsockopt = sock_common_setsockopt, /* ok */
1338 .getsockopt = sock_common_getsockopt, /* ok */
1339 .sendmsg = inet_sendmsg, /* ok */
1340 .recvmsg = sock_common_recvmsg, /* ok */
1341 .mmap = sock_no_mmap,
1342 .sendpage = sock_no_sendpage,
1343 #ifdef CONFIG_COMPAT
1344 .compat_setsockopt = compat_sock_common_setsockopt,
1345 .compat_getsockopt = compat_sock_common_getsockopt,
1346 #endif
1347 };
1348
1349 static struct inet_protosw rawv6_protosw = {
1350 .type = SOCK_RAW,
1351 .protocol = IPPROTO_IP, /* wild card */
1352 .prot = &rawv6_prot,
1353 .ops = &inet6_sockraw_ops,
1354 .no_check = UDP_CSUM_DEFAULT,
1355 .flags = INET_PROTOSW_REUSE,
1356 };
1357
1358 int __init rawv6_init(void)
1359 {
1360 int ret;
1361
1362 ret = inet6_register_protosw(&rawv6_protosw);
1363 if (ret)
1364 goto out;
1365 out:
1366 return ret;
1367 }
1368
1369 void rawv6_exit(void)
1370 {
1371 inet6_unregister_protosw(&rawv6_protosw);
1372 }
Linux kernel - Deliverables
This page took 0.060358 seconds and 5 git commands to generate.