]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/ipv6/raw.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
[NETNS][IPV6]: inet6_addr - check ipv6 address per namespace
[mirror_ubuntu-bionic-kernel.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 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
11 *
12 * Fixes:
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 *
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
21 */
22
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv6.h>
34 #include <linux/skbuff.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
57 #include <net/raw.h>
58 #include <net/rawv6.h>
59 #include <net/xfrm.h>
60
61 #include <linux/proc_fs.h>
62 #include <linux/seq_file.h>
63
64 static struct raw_hashinfo raw_v6_hashinfo = {
65 .lock = __RW_LOCK_UNLOCKED(),
66 };
67
68 static void raw_v6_hash(struct sock *sk)
69 {
70 raw_hash_sk(sk, &raw_v6_hashinfo);
71 }
72
73 static void raw_v6_unhash(struct sock *sk)
74 {
75 raw_unhash_sk(sk, &raw_v6_hashinfo);
76 }
77
78
79 static struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
80 struct in6_addr *loc_addr, struct in6_addr *rmt_addr, int dif)
81 {
82 struct hlist_node *node;
83 int is_multicast = ipv6_addr_is_multicast(loc_addr);
84
85 sk_for_each_from(sk, node)
86 if (inet_sk(sk)->num == num) {
87 struct ipv6_pinfo *np = inet6_sk(sk);
88
89 if (!ipv6_addr_any(&np->daddr) &&
90 !ipv6_addr_equal(&np->daddr, rmt_addr))
91 continue;
92
93 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
94 continue;
95
96 if (!ipv6_addr_any(&np->rcv_saddr)) {
97 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
98 goto found;
99 if (is_multicast &&
100 inet6_mc_check(sk, loc_addr, rmt_addr))
101 goto found;
102 continue;
103 }
104 goto found;
105 }
106 sk = NULL;
107 found:
108 return sk;
109 }
110
111 /*
112 * 0 - deliver
113 * 1 - block
114 */
115 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
116 {
117 struct icmp6hdr *icmph;
118 struct raw6_sock *rp = raw6_sk(sk);
119
120 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
121 __u32 *data = &rp->filter.data[0];
122 int bit_nr;
123
124 icmph = (struct icmp6hdr *) skb->data;
125 bit_nr = icmph->icmp6_type;
126
127 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
128 }
129 return 0;
130 }
131
132 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
133 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
134
135 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
136 struct sk_buff *skb))
137 {
138 rcu_assign_pointer(mh_filter, filter);
139 return 0;
140 }
141 EXPORT_SYMBOL(rawv6_mh_filter_register);
142
143 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
144 struct sk_buff *skb))
145 {
146 rcu_assign_pointer(mh_filter, NULL);
147 synchronize_rcu();
148 return 0;
149 }
150 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
151
152 #endif
153
154 /*
155 * demultiplex raw sockets.
156 * (should consider queueing the skb in the sock receive_queue
157 * without calling rawv6.c)
158 *
159 * Caller owns SKB so we must make clones.
160 */
161 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
162 {
163 struct in6_addr *saddr;
164 struct in6_addr *daddr;
165 struct sock *sk;
166 int delivered = 0;
167 __u8 hash;
168
169 saddr = &ipv6_hdr(skb)->saddr;
170 daddr = saddr + 1;
171
172 hash = nexthdr & (MAX_INET_PROTOS - 1);
173
174 read_lock(&raw_v6_hashinfo.lock);
175 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
176
177 /*
178 * The first socket found will be delivered after
179 * delivery to transport protocols.
180 */
181
182 if (sk == NULL)
183 goto out;
184
185 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
186
187 while (sk) {
188 int filtered;
189
190 delivered = 1;
191 switch (nexthdr) {
192 case IPPROTO_ICMPV6:
193 filtered = icmpv6_filter(sk, skb);
194 break;
195
196 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
197 case IPPROTO_MH:
198 {
199 /* XXX: To validate MH only once for each packet,
200 * this is placed here. It should be after checking
201 * xfrm policy, however it doesn't. The checking xfrm
202 * policy is placed in rawv6_rcv() because it is
203 * required for each socket.
204 */
205 int (*filter)(struct sock *sock, struct sk_buff *skb);
206
207 filter = rcu_dereference(mh_filter);
208 filtered = filter ? filter(sk, skb) : 0;
209 break;
210 }
211 #endif
212 default:
213 filtered = 0;
214 break;
215 }
216
217 if (filtered < 0)
218 break;
219 if (filtered == 0) {
220 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
221
222 /* Not releasing hash table! */
223 if (clone) {
224 nf_reset(clone);
225 rawv6_rcv(sk, clone);
226 }
227 }
228 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
229 IP6CB(skb)->iif);
230 }
231 out:
232 read_unlock(&raw_v6_hashinfo.lock);
233 return delivered;
234 }
235
236 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
237 {
238 struct sock *raw_sk;
239
240 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
241 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
242 raw_sk = NULL;
243
244 return raw_sk != NULL;
245 }
246
247 /* This cleans up af_inet6 a bit. -DaveM */
248 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
249 {
250 struct inet_sock *inet = inet_sk(sk);
251 struct ipv6_pinfo *np = inet6_sk(sk);
252 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
253 __be32 v4addr = 0;
254 int addr_type;
255 int err;
256
257 if (addr_len < SIN6_LEN_RFC2133)
258 return -EINVAL;
259 addr_type = ipv6_addr_type(&addr->sin6_addr);
260
261 /* Raw sockets are IPv6 only */
262 if (addr_type == IPV6_ADDR_MAPPED)
263 return(-EADDRNOTAVAIL);
264
265 lock_sock(sk);
266
267 err = -EINVAL;
268 if (sk->sk_state != TCP_CLOSE)
269 goto out;
270
271 /* Check if the address belongs to the host. */
272 if (addr_type != IPV6_ADDR_ANY) {
273 struct net_device *dev = NULL;
274
275 if (addr_type & IPV6_ADDR_LINKLOCAL) {
276 if (addr_len >= sizeof(struct sockaddr_in6) &&
277 addr->sin6_scope_id) {
278 /* Override any existing binding, if another
279 * one is supplied by user.
280 */
281 sk->sk_bound_dev_if = addr->sin6_scope_id;
282 }
283
284 /* Binding to link-local address requires an interface */
285 if (!sk->sk_bound_dev_if)
286 goto out;
287
288 dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if);
289 if (!dev) {
290 err = -ENODEV;
291 goto out;
292 }
293 }
294
295 /* ipv4 addr of the socket is invalid. Only the
296 * unspecified and mapped address have a v4 equivalent.
297 */
298 v4addr = LOOPBACK4_IPV6;
299 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
300 err = -EADDRNOTAVAIL;
301 if (!ipv6_chk_addr(&init_net, &addr->sin6_addr,
302 dev, 0)) {
303 if (dev)
304 dev_put(dev);
305 goto out;
306 }
307 }
308 if (dev)
309 dev_put(dev);
310 }
311
312 inet->rcv_saddr = inet->saddr = v4addr;
313 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
314 if (!(addr_type & IPV6_ADDR_MULTICAST))
315 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
316 err = 0;
317 out:
318 release_sock(sk);
319 return err;
320 }
321
322 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
323 struct inet6_skb_parm *opt,
324 int type, int code, int offset, __be32 info)
325 {
326 struct inet_sock *inet = inet_sk(sk);
327 struct ipv6_pinfo *np = inet6_sk(sk);
328 int err;
329 int harderr;
330
331 /* Report error on raw socket, if:
332 1. User requested recverr.
333 2. Socket is connected (otherwise the error indication
334 is useless without recverr and error is hard.
335 */
336 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
337 return;
338
339 harderr = icmpv6_err_convert(type, code, &err);
340 if (type == ICMPV6_PKT_TOOBIG)
341 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
342
343 if (np->recverr) {
344 u8 *payload = skb->data;
345 if (!inet->hdrincl)
346 payload += offset;
347 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
348 }
349
350 if (np->recverr || harderr) {
351 sk->sk_err = err;
352 sk->sk_error_report(sk);
353 }
354 }
355
356 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
357 int type, int code, int inner_offset, __be32 info)
358 {
359 struct sock *sk;
360 int hash;
361 struct in6_addr *saddr, *daddr;
362
363 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
364
365 read_lock(&raw_v6_hashinfo.lock);
366 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
367 if (sk != NULL) {
368 saddr = &ipv6_hdr(skb)->saddr;
369 daddr = &ipv6_hdr(skb)->daddr;
370
371 while ((sk = __raw_v6_lookup(sk, nexthdr, saddr, daddr,
372 IP6CB(skb)->iif))) {
373 rawv6_err(sk, skb, NULL, type, code,
374 inner_offset, info);
375 sk = sk_next(sk);
376 }
377 }
378 read_unlock(&raw_v6_hashinfo.lock);
379 }
380
381 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
382 {
383 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
384 skb_checksum_complete(skb)) {
385 atomic_inc(&sk->sk_drops);
386 kfree_skb(skb);
387 return 0;
388 }
389
390 /* Charge it to the socket. */
391 if (sock_queue_rcv_skb(sk,skb)<0) {
392 atomic_inc(&sk->sk_drops);
393 kfree_skb(skb);
394 return 0;
395 }
396
397 return 0;
398 }
399
400 /*
401 * This is next to useless...
402 * if we demultiplex in network layer we don't need the extra call
403 * just to queue the skb...
404 * maybe we could have the network decide upon a hint if it
405 * should call raw_rcv for demultiplexing
406 */
407 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
408 {
409 struct inet_sock *inet = inet_sk(sk);
410 struct raw6_sock *rp = raw6_sk(sk);
411
412 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
413 atomic_inc(&sk->sk_drops);
414 kfree_skb(skb);
415 return NET_RX_DROP;
416 }
417
418 if (!rp->checksum)
419 skb->ip_summed = CHECKSUM_UNNECESSARY;
420
421 if (skb->ip_summed == CHECKSUM_COMPLETE) {
422 skb_postpull_rcsum(skb, skb_network_header(skb),
423 skb_network_header_len(skb));
424 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
425 &ipv6_hdr(skb)->daddr,
426 skb->len, inet->num, skb->csum))
427 skb->ip_summed = CHECKSUM_UNNECESSARY;
428 }
429 if (!skb_csum_unnecessary(skb))
430 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
431 &ipv6_hdr(skb)->daddr,
432 skb->len,
433 inet->num, 0));
434
435 if (inet->hdrincl) {
436 if (skb_checksum_complete(skb)) {
437 atomic_inc(&sk->sk_drops);
438 kfree_skb(skb);
439 return 0;
440 }
441 }
442
443 rawv6_rcv_skb(sk, skb);
444 return 0;
445 }
446
447
448 /*
449 * This should be easy, if there is something there
450 * we return it, otherwise we block.
451 */
452
453 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
454 struct msghdr *msg, size_t len,
455 int noblock, int flags, int *addr_len)
456 {
457 struct ipv6_pinfo *np = inet6_sk(sk);
458 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
459 struct sk_buff *skb;
460 size_t copied;
461 int err;
462
463 if (flags & MSG_OOB)
464 return -EOPNOTSUPP;
465
466 if (addr_len)
467 *addr_len=sizeof(*sin6);
468
469 if (flags & MSG_ERRQUEUE)
470 return ipv6_recv_error(sk, msg, len);
471
472 skb = skb_recv_datagram(sk, flags, noblock, &err);
473 if (!skb)
474 goto out;
475
476 copied = skb->len;
477 if (copied > len) {
478 copied = len;
479 msg->msg_flags |= MSG_TRUNC;
480 }
481
482 if (skb_csum_unnecessary(skb)) {
483 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484 } else if (msg->msg_flags&MSG_TRUNC) {
485 if (__skb_checksum_complete(skb))
486 goto csum_copy_err;
487 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
488 } else {
489 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
490 if (err == -EINVAL)
491 goto csum_copy_err;
492 }
493 if (err)
494 goto out_free;
495
496 /* Copy the address. */
497 if (sin6) {
498 sin6->sin6_family = AF_INET6;
499 sin6->sin6_port = 0;
500 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
501 sin6->sin6_flowinfo = 0;
502 sin6->sin6_scope_id = 0;
503 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
504 sin6->sin6_scope_id = IP6CB(skb)->iif;
505 }
506
507 sock_recv_timestamp(msg, sk, skb);
508
509 if (np->rxopt.all)
510 datagram_recv_ctl(sk, msg, skb);
511
512 err = copied;
513 if (flags & MSG_TRUNC)
514 err = skb->len;
515
516 out_free:
517 skb_free_datagram(sk, skb);
518 out:
519 return err;
520
521 csum_copy_err:
522 skb_kill_datagram(sk, skb, flags);
523
524 /* Error for blocking case is chosen to masquerade
525 as some normal condition.
526 */
527 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
528 atomic_inc(&sk->sk_drops);
529 goto out;
530 }
531
532 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
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.length - (skb_network_header(skb) -
551 skb->data);
552 if (offset >= total_len - 1) {
553 err = -EINVAL;
554 ip6_flush_pending_frames(sk);
555 goto out;
556 }
557
558 /* should be check HW csum miyazawa */
559 if (skb_queue_len(&sk->sk_write_queue) == 1) {
560 /*
561 * Only one fragment on the socket.
562 */
563 tmp_csum = skb->csum;
564 } else {
565 struct sk_buff *csum_skb = NULL;
566 tmp_csum = 0;
567
568 skb_queue_walk(&sk->sk_write_queue, skb) {
569 tmp_csum = csum_add(tmp_csum, skb->csum);
570
571 if (csum_skb)
572 continue;
573
574 len = skb->len - skb_transport_offset(skb);
575 if (offset >= len) {
576 offset -= len;
577 continue;
578 }
579
580 csum_skb = skb;
581 }
582
583 skb = csum_skb;
584 }
585
586 offset += skb_transport_offset(skb);
587 if (skb_copy_bits(skb, offset, &csum, 2))
588 BUG();
589
590 /* in case cksum was not initialized */
591 if (unlikely(csum))
592 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
593
594 csum = csum_ipv6_magic(&fl->fl6_src,
595 &fl->fl6_dst,
596 total_len, fl->proto, tmp_csum);
597
598 if (csum == 0 && fl->proto == IPPROTO_UDP)
599 csum = CSUM_MANGLED_0;
600
601 if (skb_store_bits(skb, offset, &csum, 2))
602 BUG();
603
604 send:
605 err = ip6_push_pending_frames(sk);
606 out:
607 return err;
608 }
609
610 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
611 struct flowi *fl, struct rt6_info *rt,
612 unsigned int flags)
613 {
614 struct ipv6_pinfo *np = inet6_sk(sk);
615 struct ipv6hdr *iph;
616 struct sk_buff *skb;
617 unsigned int hh_len;
618 int err;
619
620 if (length > rt->u.dst.dev->mtu) {
621 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
622 return -EMSGSIZE;
623 }
624 if (flags&MSG_PROBE)
625 goto out;
626
627 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
628
629 skb = sock_alloc_send_skb(sk, length+hh_len+15,
630 flags&MSG_DONTWAIT, &err);
631 if (skb == NULL)
632 goto error;
633 skb_reserve(skb, hh_len);
634
635 skb->priority = sk->sk_priority;
636 skb->dst = dst_clone(&rt->u.dst);
637
638 skb_put(skb, length);
639 skb_reset_network_header(skb);
640 iph = ipv6_hdr(skb);
641
642 skb->ip_summed = CHECKSUM_NONE;
643
644 skb->transport_header = skb->network_header;
645 err = memcpy_fromiovecend((void *)iph, from, 0, length);
646 if (err)
647 goto error_fault;
648
649 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
650 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
651 dst_output);
652 if (err > 0)
653 err = np->recverr ? net_xmit_errno(err) : 0;
654 if (err)
655 goto error;
656 out:
657 return 0;
658
659 error_fault:
660 err = -EFAULT;
661 kfree_skb(skb);
662 error:
663 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
664 return err;
665 }
666
667 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
668 {
669 struct iovec *iov;
670 u8 __user *type = NULL;
671 u8 __user *code = NULL;
672 u8 len = 0;
673 int probed = 0;
674 int i;
675
676 if (!msg->msg_iov)
677 return 0;
678
679 for (i = 0; i < msg->msg_iovlen; i++) {
680 iov = &msg->msg_iov[i];
681 if (!iov)
682 continue;
683
684 switch (fl->proto) {
685 case IPPROTO_ICMPV6:
686 /* check if one-byte field is readable or not. */
687 if (iov->iov_base && iov->iov_len < 1)
688 break;
689
690 if (!type) {
691 type = iov->iov_base;
692 /* check if code field is readable or not. */
693 if (iov->iov_len > 1)
694 code = type + 1;
695 } else if (!code)
696 code = iov->iov_base;
697
698 if (type && code) {
699 if (get_user(fl->fl_icmp_type, type) ||
700 get_user(fl->fl_icmp_code, code))
701 return -EFAULT;
702 probed = 1;
703 }
704 break;
705 case IPPROTO_MH:
706 if (iov->iov_base && iov->iov_len < 1)
707 break;
708 /* check if type field is readable or not. */
709 if (iov->iov_len > 2 - len) {
710 u8 __user *p = iov->iov_base;
711 if (get_user(fl->fl_mh_type, &p[2 - len]))
712 return -EFAULT;
713 probed = 1;
714 } else
715 len += iov->iov_len;
716
717 break;
718 default:
719 probed = 1;
720 break;
721 }
722 if (probed)
723 break;
724 }
725 return 0;
726 }
727
728 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
729 struct msghdr *msg, size_t len)
730 {
731 struct ipv6_txoptions opt_space;
732 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
733 struct in6_addr *daddr, *final_p = NULL, final;
734 struct inet_sock *inet = inet_sk(sk);
735 struct ipv6_pinfo *np = inet6_sk(sk);
736 struct raw6_sock *rp = raw6_sk(sk);
737 struct ipv6_txoptions *opt = NULL;
738 struct ip6_flowlabel *flowlabel = NULL;
739 struct dst_entry *dst = NULL;
740 struct flowi fl;
741 int addr_len = msg->msg_namelen;
742 int hlimit = -1;
743 int tclass = -1;
744 u16 proto;
745 int err;
746
747 /* Rough check on arithmetic overflow,
748 better check is made in ip6_append_data().
749 */
750 if (len > INT_MAX)
751 return -EMSGSIZE;
752
753 /* Mirror BSD error message compatibility */
754 if (msg->msg_flags & MSG_OOB)
755 return -EOPNOTSUPP;
756
757 /*
758 * Get and verify the address.
759 */
760 memset(&fl, 0, sizeof(fl));
761
762 if (sin6) {
763 if (addr_len < SIN6_LEN_RFC2133)
764 return -EINVAL;
765
766 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
767 return(-EAFNOSUPPORT);
768
769 /* port is the proto value [0..255] carried in nexthdr */
770 proto = ntohs(sin6->sin6_port);
771
772 if (!proto)
773 proto = inet->num;
774 else if (proto != inet->num)
775 return(-EINVAL);
776
777 if (proto > 255)
778 return(-EINVAL);
779
780 daddr = &sin6->sin6_addr;
781 if (np->sndflow) {
782 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
783 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
784 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
785 if (flowlabel == NULL)
786 return -EINVAL;
787 daddr = &flowlabel->dst;
788 }
789 }
790
791 /*
792 * Otherwise it will be difficult to maintain
793 * sk->sk_dst_cache.
794 */
795 if (sk->sk_state == TCP_ESTABLISHED &&
796 ipv6_addr_equal(daddr, &np->daddr))
797 daddr = &np->daddr;
798
799 if (addr_len >= sizeof(struct sockaddr_in6) &&
800 sin6->sin6_scope_id &&
801 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
802 fl.oif = sin6->sin6_scope_id;
803 } else {
804 if (sk->sk_state != TCP_ESTABLISHED)
805 return -EDESTADDRREQ;
806
807 proto = inet->num;
808 daddr = &np->daddr;
809 fl.fl6_flowlabel = np->flow_label;
810 }
811
812 if (ipv6_addr_any(daddr)) {
813 /*
814 * unspecified destination address
815 * treated as error... is this correct ?
816 */
817 fl6_sock_release(flowlabel);
818 return(-EINVAL);
819 }
820
821 if (fl.oif == 0)
822 fl.oif = sk->sk_bound_dev_if;
823
824 if (msg->msg_controllen) {
825 opt = &opt_space;
826 memset(opt, 0, sizeof(struct ipv6_txoptions));
827 opt->tot_len = sizeof(struct ipv6_txoptions);
828
829 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
830 if (err < 0) {
831 fl6_sock_release(flowlabel);
832 return err;
833 }
834 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
835 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
836 if (flowlabel == NULL)
837 return -EINVAL;
838 }
839 if (!(opt->opt_nflen|opt->opt_flen))
840 opt = NULL;
841 }
842 if (opt == NULL)
843 opt = np->opt;
844 if (flowlabel)
845 opt = fl6_merge_options(&opt_space, flowlabel, opt);
846 opt = ipv6_fixup_options(&opt_space, opt);
847
848 fl.proto = proto;
849 err = rawv6_probe_proto_opt(&fl, msg);
850 if (err)
851 goto out;
852
853 ipv6_addr_copy(&fl.fl6_dst, daddr);
854 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
855 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
856
857 /* merge ip6_build_xmit from ip6_output */
858 if (opt && opt->srcrt) {
859 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
860 ipv6_addr_copy(&final, &fl.fl6_dst);
861 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
862 final_p = &final;
863 }
864
865 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
866 fl.oif = np->mcast_oif;
867 security_sk_classify_flow(sk, &fl);
868
869 err = ip6_dst_lookup(sk, &dst, &fl);
870 if (err)
871 goto out;
872 if (final_p)
873 ipv6_addr_copy(&fl.fl6_dst, final_p);
874
875 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
876 if (err == -EREMOTE)
877 err = ip6_dst_blackhole(sk, &dst, &fl);
878 if (err < 0)
879 goto out;
880 }
881
882 if (hlimit < 0) {
883 if (ipv6_addr_is_multicast(&fl.fl6_dst))
884 hlimit = np->mcast_hops;
885 else
886 hlimit = np->hop_limit;
887 if (hlimit < 0)
888 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
889 if (hlimit < 0)
890 hlimit = ipv6_get_hoplimit(dst->dev);
891 }
892
893 if (tclass < 0) {
894 tclass = np->tclass;
895 if (tclass < 0)
896 tclass = 0;
897 }
898
899 if (msg->msg_flags&MSG_CONFIRM)
900 goto do_confirm;
901
902 back_from_confirm:
903 if (inet->hdrincl) {
904 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
905 } else {
906 lock_sock(sk);
907 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
908 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
909 msg->msg_flags);
910
911 if (err)
912 ip6_flush_pending_frames(sk);
913 else if (!(msg->msg_flags & MSG_MORE))
914 err = rawv6_push_pending_frames(sk, &fl, rp);
915 release_sock(sk);
916 }
917 done:
918 dst_release(dst);
919 out:
920 fl6_sock_release(flowlabel);
921 return err<0?err:len;
922 do_confirm:
923 dst_confirm(dst);
924 if (!(msg->msg_flags & MSG_PROBE) || len)
925 goto back_from_confirm;
926 err = 0;
927 goto done;
928 }
929
930 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
931 char __user *optval, int optlen)
932 {
933 switch (optname) {
934 case ICMPV6_FILTER:
935 if (optlen > sizeof(struct icmp6_filter))
936 optlen = sizeof(struct icmp6_filter);
937 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
938 return -EFAULT;
939 return 0;
940 default:
941 return -ENOPROTOOPT;
942 }
943
944 return 0;
945 }
946
947 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
948 char __user *optval, int __user *optlen)
949 {
950 int len;
951
952 switch (optname) {
953 case ICMPV6_FILTER:
954 if (get_user(len, optlen))
955 return -EFAULT;
956 if (len < 0)
957 return -EINVAL;
958 if (len > sizeof(struct icmp6_filter))
959 len = sizeof(struct icmp6_filter);
960 if (put_user(len, optlen))
961 return -EFAULT;
962 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
963 return -EFAULT;
964 return 0;
965 default:
966 return -ENOPROTOOPT;
967 }
968
969 return 0;
970 }
971
972
973 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
974 char __user *optval, int optlen)
975 {
976 struct raw6_sock *rp = raw6_sk(sk);
977 int val;
978
979 if (get_user(val, (int __user *)optval))
980 return -EFAULT;
981
982 switch (optname) {
983 case IPV6_CHECKSUM:
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 break;
997
998 default:
999 return(-ENOPROTOOPT);
1000 }
1001 }
1002
1003 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1004 char __user *optval, int optlen)
1005 {
1006 switch(level) {
1007 case SOL_RAW:
1008 break;
1009
1010 case SOL_ICMPV6:
1011 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1012 return -EOPNOTSUPP;
1013 return rawv6_seticmpfilter(sk, level, optname, optval,
1014 optlen);
1015 case SOL_IPV6:
1016 if (optname == IPV6_CHECKSUM)
1017 break;
1018 default:
1019 return ipv6_setsockopt(sk, level, optname, optval,
1020 optlen);
1021 }
1022
1023 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1024 }
1025
1026 #ifdef CONFIG_COMPAT
1027 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1028 char __user *optval, int optlen)
1029 {
1030 switch (level) {
1031 case SOL_RAW:
1032 break;
1033 case SOL_ICMPV6:
1034 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1035 return -EOPNOTSUPP;
1036 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1037 case SOL_IPV6:
1038 if (optname == IPV6_CHECKSUM)
1039 break;
1040 default:
1041 return compat_ipv6_setsockopt(sk, level, optname,
1042 optval, optlen);
1043 }
1044 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1045 }
1046 #endif
1047
1048 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1049 char __user *optval, int __user *optlen)
1050 {
1051 struct raw6_sock *rp = raw6_sk(sk);
1052 int val, len;
1053
1054 if (get_user(len,optlen))
1055 return -EFAULT;
1056
1057 switch (optname) {
1058 case IPV6_CHECKSUM:
1059 if (rp->checksum == 0)
1060 val = -1;
1061 else
1062 val = rp->offset;
1063 break;
1064
1065 default:
1066 return -ENOPROTOOPT;
1067 }
1068
1069 len = min_t(unsigned int, sizeof(int), len);
1070
1071 if (put_user(len, optlen))
1072 return -EFAULT;
1073 if (copy_to_user(optval,&val,len))
1074 return -EFAULT;
1075 return 0;
1076 }
1077
1078 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1079 char __user *optval, int __user *optlen)
1080 {
1081 switch(level) {
1082 case SOL_RAW:
1083 break;
1084
1085 case SOL_ICMPV6:
1086 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1087 return -EOPNOTSUPP;
1088 return rawv6_geticmpfilter(sk, level, optname, optval,
1089 optlen);
1090 case SOL_IPV6:
1091 if (optname == IPV6_CHECKSUM)
1092 break;
1093 default:
1094 return ipv6_getsockopt(sk, level, optname, optval,
1095 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)->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 {
1128 int amount = atomic_read(&sk->sk_wmem_alloc);
1129 return put_user(amount, (int __user *)arg);
1130 }
1131 case SIOCINQ:
1132 {
1133 struct sk_buff *skb;
1134 int amount = 0;
1135
1136 spin_lock_bh(&sk->sk_receive_queue.lock);
1137 skb = skb_peek(&sk->sk_receive_queue);
1138 if (skb != NULL)
1139 amount = skb->tail - skb->transport_header;
1140 spin_unlock_bh(&sk->sk_receive_queue.lock);
1141 return put_user(amount, (int __user *)arg);
1142 }
1143
1144 default:
1145 return -ENOIOCTLCMD;
1146 }
1147 }
1148
1149 static void rawv6_close(struct sock *sk, long timeout)
1150 {
1151 if (inet_sk(sk)->num == IPPROTO_RAW)
1152 ip6_ra_control(sk, -1, NULL);
1153
1154 sk_common_release(sk);
1155 }
1156
1157 static int rawv6_init_sk(struct sock *sk)
1158 {
1159 struct raw6_sock *rp = raw6_sk(sk);
1160
1161 switch (inet_sk(sk)->num) {
1162 case IPPROTO_ICMPV6:
1163 rp->checksum = 1;
1164 rp->offset = 2;
1165 break;
1166 case IPPROTO_MH:
1167 rp->checksum = 1;
1168 rp->offset = 4;
1169 break;
1170 default:
1171 break;
1172 }
1173 return(0);
1174 }
1175
1176 DEFINE_PROTO_INUSE(rawv6)
1177
1178 struct proto rawv6_prot = {
1179 .name = "RAWv6",
1180 .owner = THIS_MODULE,
1181 .close = rawv6_close,
1182 .connect = ip6_datagram_connect,
1183 .disconnect = udp_disconnect,
1184 .ioctl = rawv6_ioctl,
1185 .init = rawv6_init_sk,
1186 .destroy = inet6_destroy_sock,
1187 .setsockopt = rawv6_setsockopt,
1188 .getsockopt = rawv6_getsockopt,
1189 .sendmsg = rawv6_sendmsg,
1190 .recvmsg = rawv6_recvmsg,
1191 .bind = rawv6_bind,
1192 .backlog_rcv = rawv6_rcv_skb,
1193 .hash = raw_v6_hash,
1194 .unhash = raw_v6_unhash,
1195 .obj_size = sizeof(struct raw6_sock),
1196 #ifdef CONFIG_COMPAT
1197 .compat_setsockopt = compat_rawv6_setsockopt,
1198 .compat_getsockopt = compat_rawv6_getsockopt,
1199 #endif
1200 REF_PROTO_INUSE(rawv6)
1201 };
1202
1203 #ifdef CONFIG_PROC_FS
1204 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1205 {
1206 struct ipv6_pinfo *np = inet6_sk(sp);
1207 struct in6_addr *dest, *src;
1208 __u16 destp, srcp;
1209
1210 dest = &np->daddr;
1211 src = &np->rcv_saddr;
1212 destp = 0;
1213 srcp = inet_sk(sp)->num;
1214 seq_printf(seq,
1215 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1216 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1217 i,
1218 src->s6_addr32[0], src->s6_addr32[1],
1219 src->s6_addr32[2], src->s6_addr32[3], srcp,
1220 dest->s6_addr32[0], dest->s6_addr32[1],
1221 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1222 sp->sk_state,
1223 atomic_read(&sp->sk_wmem_alloc),
1224 atomic_read(&sp->sk_rmem_alloc),
1225 0, 0L, 0,
1226 sock_i_uid(sp), 0,
1227 sock_i_ino(sp),
1228 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1229 }
1230
1231 static int raw6_seq_show(struct seq_file *seq, void *v)
1232 {
1233 if (v == SEQ_START_TOKEN)
1234 seq_printf(seq,
1235 " sl "
1236 "local_address "
1237 "remote_address "
1238 "st tx_queue rx_queue tr tm->when retrnsmt"
1239 " uid timeout inode drops\n");
1240 else
1241 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1242 return 0;
1243 }
1244
1245 static const struct seq_operations raw6_seq_ops = {
1246 .start = raw_seq_start,
1247 .next = raw_seq_next,
1248 .stop = raw_seq_stop,
1249 .show = raw6_seq_show,
1250 };
1251
1252 static int raw6_seq_open(struct inode *inode, struct file *file)
1253 {
1254 return raw_seq_open(file, &raw_v6_hashinfo, PF_INET6);
1255 }
1256
1257 static const struct file_operations raw6_seq_fops = {
1258 .owner = THIS_MODULE,
1259 .open = raw6_seq_open,
1260 .read = seq_read,
1261 .llseek = seq_lseek,
1262 .release = seq_release_private,
1263 };
1264
1265 int __init raw6_proc_init(void)
1266 {
1267 if (!proc_net_fops_create(&init_net, "raw6", S_IRUGO, &raw6_seq_fops))
1268 return -ENOMEM;
1269 return 0;
1270 }
1271
1272 void raw6_proc_exit(void)
1273 {
1274 proc_net_remove(&init_net, "raw6");
1275 }
1276 #endif /* CONFIG_PROC_FS */
1277
1278 /* Same as inet6_dgram_ops, sans udp_poll. */
1279 static const struct proto_ops inet6_sockraw_ops = {
1280 .family = PF_INET6,
1281 .owner = THIS_MODULE,
1282 .release = inet6_release,
1283 .bind = inet6_bind,
1284 .connect = inet_dgram_connect, /* ok */
1285 .socketpair = sock_no_socketpair, /* a do nothing */
1286 .accept = sock_no_accept, /* a do nothing */
1287 .getname = inet6_getname,
1288 .poll = datagram_poll, /* ok */
1289 .ioctl = inet6_ioctl, /* must change */
1290 .listen = sock_no_listen, /* ok */
1291 .shutdown = inet_shutdown, /* ok */
1292 .setsockopt = sock_common_setsockopt, /* ok */
1293 .getsockopt = sock_common_getsockopt, /* ok */
1294 .sendmsg = inet_sendmsg, /* ok */
1295 .recvmsg = sock_common_recvmsg, /* ok */
1296 .mmap = sock_no_mmap,
1297 .sendpage = sock_no_sendpage,
1298 #ifdef CONFIG_COMPAT
1299 .compat_setsockopt = compat_sock_common_setsockopt,
1300 .compat_getsockopt = compat_sock_common_getsockopt,
1301 #endif
1302 };
1303
1304 static struct inet_protosw rawv6_protosw = {
1305 .type = SOCK_RAW,
1306 .protocol = IPPROTO_IP, /* wild card */
1307 .prot = &rawv6_prot,
1308 .ops = &inet6_sockraw_ops,
1309 .capability = CAP_NET_RAW,
1310 .no_check = UDP_CSUM_DEFAULT,
1311 .flags = INET_PROTOSW_REUSE,
1312 };
1313
1314 int __init rawv6_init(void)
1315 {
1316 int ret;
1317
1318 ret = inet6_register_protosw(&rawv6_protosw);
1319 if (ret)
1320 goto out;
1321 out:
1322 return ret;
1323 }
1324
1325 void rawv6_exit(void)
1326 {
1327 inet6_unregister_protosw(&rawv6_protosw);
1328 }
ubuntu-bionic-kernel mirror