projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
inet: prevent leakage of uninitialized memory to user in recv syscalls
[GitHub/mt8127/android_kernel_alcatel_ttab.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 IS_ENABLED(CONFIG_IPV6_MIP6)
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 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
75
76 sk_for_each_from(sk)
77 if (inet_sk(sk)->inet_num == num) {
78 struct ipv6_pinfo *np = inet6_sk(sk);
79
80 if (!net_eq(sock_net(sk), net))
81 continue;
82
83 if (!ipv6_addr_any(&np->daddr) &&
84 !ipv6_addr_equal(&np->daddr, rmt_addr))
85 continue;
86
87 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
88 continue;
89
90 if (!ipv6_addr_any(&np->rcv_saddr)) {
91 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
92 goto found;
93 if (is_multicast &&
94 inet6_mc_check(sk, loc_addr, rmt_addr))
95 goto found;
96 continue;
97 }
98 goto found;
99 }
100 sk = NULL;
101 found:
102 return sk;
103 }
104
105 /*
106 * 0 - deliver
107 * 1 - block
108 */
109 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
110 {
111 struct icmp6hdr *_hdr;
112 const struct icmp6hdr *hdr;
113
114 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
115 sizeof(_hdr), &_hdr);
116 if (hdr) {
117 const __u32 *data = &raw6_sk(sk)->filter.data[0];
118 unsigned int type = hdr->icmp6_type;
119
120 return (data[type >> 5] & (1U << (type & 31))) != 0;
121 }
122 return 1;
123 }
124
125 #if IS_ENABLED(CONFIG_IPV6_MIP6)
126 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
127
128 static mh_filter_t __rcu *mh_filter __read_mostly;
129
130 int rawv6_mh_filter_register(mh_filter_t filter)
131 {
132 rcu_assign_pointer(mh_filter, filter);
133 return 0;
134 }
135 EXPORT_SYMBOL(rawv6_mh_filter_register);
136
137 int rawv6_mh_filter_unregister(mh_filter_t filter)
138 {
139 RCU_INIT_POINTER(mh_filter, NULL);
140 synchronize_rcu();
141 return 0;
142 }
143 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
144
145 #endif
146
147 /*
148 * demultiplex raw sockets.
149 * (should consider queueing the skb in the sock receive_queue
150 * without calling rawv6.c)
151 *
152 * Caller owns SKB so we must make clones.
153 */
154 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
155 {
156 const struct in6_addr *saddr;
157 const struct in6_addr *daddr;
158 struct sock *sk;
159 bool delivered = false;
160 __u8 hash;
161 struct net *net;
162
163 saddr = &ipv6_hdr(skb)->saddr;
164 daddr = saddr + 1;
165
166 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
167
168 read_lock(&raw_v6_hashinfo.lock);
169 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
170
171 if (sk == NULL)
172 goto out;
173
174 net = dev_net(skb->dev);
175 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
176
177 while (sk) {
178 int filtered;
179
180 delivered = true;
181 switch (nexthdr) {
182 case IPPROTO_ICMPV6:
183 filtered = icmpv6_filter(sk, skb);
184 break;
185
186 #if IS_ENABLED(CONFIG_IPV6_MIP6)
187 case IPPROTO_MH:
188 {
189 /* XXX: To validate MH only once for each packet,
190 * this is placed here. It should be after checking
191 * xfrm policy, however it doesn't. The checking xfrm
192 * policy is placed in rawv6_rcv() because it is
193 * required for each socket.
194 */
195 mh_filter_t *filter;
196
197 filter = rcu_dereference(mh_filter);
198 filtered = filter ? (*filter)(sk, skb) : 0;
199 break;
200 }
201 #endif
202 default:
203 filtered = 0;
204 break;
205 }
206
207 if (filtered < 0)
208 break;
209 if (filtered == 0) {
210 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
211
212 /* Not releasing hash table! */
213 if (clone) {
214 nf_reset(clone);
215 rawv6_rcv(sk, clone);
216 }
217 }
218 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
219 IP6CB(skb)->iif);
220 }
221 out:
222 read_unlock(&raw_v6_hashinfo.lock);
223 return delivered;
224 }
225
226 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
227 {
228 struct sock *raw_sk;
229
230 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
231 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
232 raw_sk = NULL;
233
234 return raw_sk != NULL;
235 }
236
237 /* This cleans up af_inet6 a bit. -DaveM */
238 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
239 {
240 struct inet_sock *inet = inet_sk(sk);
241 struct ipv6_pinfo *np = inet6_sk(sk);
242 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
243 __be32 v4addr = 0;
244 int addr_type;
245 int err;
246
247 if (addr_len < SIN6_LEN_RFC2133)
248 return -EINVAL;
249 addr_type = ipv6_addr_type(&addr->sin6_addr);
250
251 /* Raw sockets are IPv6 only */
252 if (addr_type == IPV6_ADDR_MAPPED)
253 return -EADDRNOTAVAIL;
254
255 lock_sock(sk);
256
257 err = -EINVAL;
258 if (sk->sk_state != TCP_CLOSE)
259 goto out;
260
261 rcu_read_lock();
262 /* Check if the address belongs to the host. */
263 if (addr_type != IPV6_ADDR_ANY) {
264 struct net_device *dev = NULL;
265
266 if (__ipv6_addr_needs_scope_id(addr_type)) {
267 if (addr_len >= sizeof(struct sockaddr_in6) &&
268 addr->sin6_scope_id) {
269 /* Override any existing binding, if another
270 * one is supplied by user.
271 */
272 sk->sk_bound_dev_if = addr->sin6_scope_id;
273 }
274
275 /* Binding to link-local address requires an interface */
276 if (!sk->sk_bound_dev_if)
277 goto out_unlock;
278
279 err = -ENODEV;
280 dev = dev_get_by_index_rcu(sock_net(sk),
281 sk->sk_bound_dev_if);
282 if (!dev)
283 goto out_unlock;
284 }
285
286 /* ipv4 addr of the socket is invalid. Only the
287 * unspecified and mapped address have a v4 equivalent.
288 */
289 v4addr = LOOPBACK4_IPV6;
290 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
291 err = -EADDRNOTAVAIL;
292 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
293 dev, 0)) {
294 goto out_unlock;
295 }
296 }
297 }
298
299 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
300 np->rcv_saddr = addr->sin6_addr;
301 if (!(addr_type & IPV6_ADDR_MULTICAST))
302 np->saddr = addr->sin6_addr;
303 err = 0;
304 out_unlock:
305 rcu_read_unlock();
306 out:
307 release_sock(sk);
308 return err;
309 }
310
311 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
312 struct inet6_skb_parm *opt,
313 u8 type, u8 code, int offset, __be32 info)
314 {
315 struct inet_sock *inet = inet_sk(sk);
316 struct ipv6_pinfo *np = inet6_sk(sk);
317 int err;
318 int harderr;
319
320 /* Report error on raw socket, if:
321 1. User requested recverr.
322 2. Socket is connected (otherwise the error indication
323 is useless without recverr and error is hard.
324 */
325 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
326 return;
327
328 harderr = icmpv6_err_convert(type, code, &err);
329 if (type == ICMPV6_PKT_TOOBIG) {
330 ip6_sk_update_pmtu(skb, sk, info);
331 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
332 }
333 if (type == NDISC_REDIRECT)
334 ip6_sk_redirect(skb, sk);
335 if (np->recverr) {
336 u8 *payload = skb->data;
337 if (!inet->hdrincl)
338 payload += offset;
339 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
340 }
341
342 if (np->recverr || harderr) {
343 sk->sk_err = err;
344 sk->sk_error_report(sk);
345 }
346 }
347
348 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
349 u8 type, u8 code, int inner_offset, __be32 info)
350 {
351 struct sock *sk;
352 int hash;
353 const struct in6_addr *saddr, *daddr;
354 struct net *net;
355
356 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
357
358 read_lock(&raw_v6_hashinfo.lock);
359 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
360 if (sk != NULL) {
361 /* Note: ipv6_hdr(skb) != skb->data */
362 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
363 saddr = &ip6h->saddr;
364 daddr = &ip6h->daddr;
365 net = dev_net(skb->dev);
366
367 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
368 IP6CB(skb)->iif))) {
369 rawv6_err(sk, skb, NULL, type, code,
370 inner_offset, info);
371 sk = sk_next(sk);
372 }
373 }
374 read_unlock(&raw_v6_hashinfo.lock);
375 }
376
377 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
378 {
379 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
380 skb_checksum_complete(skb)) {
381 atomic_inc(&sk->sk_drops);
382 kfree_skb(skb);
383 return NET_RX_DROP;
384 }
385
386 /* Charge it to the socket. */
387 skb_dst_drop(skb);
388 if (sock_queue_rcv_skb(sk, skb) < 0) {
389 kfree_skb(skb);
390 return NET_RX_DROP;
391 }
392
393 return 0;
394 }
395
396 /*
397 * This is next to useless...
398 * if we demultiplex in network layer we don't need the extra call
399 * just to queue the skb...
400 * maybe we could have the network decide upon a hint if it
401 * should call raw_rcv for demultiplexing
402 */
403 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
404 {
405 struct inet_sock *inet = inet_sk(sk);
406 struct raw6_sock *rp = raw6_sk(sk);
407
408 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
409 atomic_inc(&sk->sk_drops);
410 kfree_skb(skb);
411 return NET_RX_DROP;
412 }
413
414 if (!rp->checksum)
415 skb->ip_summed = CHECKSUM_UNNECESSARY;
416
417 if (skb->ip_summed == CHECKSUM_COMPLETE) {
418 skb_postpull_rcsum(skb, skb_network_header(skb),
419 skb_network_header_len(skb));
420 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
421 &ipv6_hdr(skb)->daddr,
422 skb->len, inet->inet_num, skb->csum))
423 skb->ip_summed = CHECKSUM_UNNECESSARY;
424 }
425 if (!skb_csum_unnecessary(skb))
426 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
427 &ipv6_hdr(skb)->daddr,
428 skb->len,
429 inet->inet_num, 0));
430
431 if (inet->hdrincl) {
432 if (skb_checksum_complete(skb)) {
433 atomic_inc(&sk->sk_drops);
434 kfree_skb(skb);
435 return NET_RX_DROP;
436 }
437 }
438
439 rawv6_rcv_skb(sk, skb);
440 return 0;
441 }
442
443
444 /*
445 * This should be easy, if there is something there
446 * we return it, otherwise we block.
447 */
448
449 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
450 struct msghdr *msg, size_t len,
451 int noblock, int flags, int *addr_len)
452 {
453 struct ipv6_pinfo *np = inet6_sk(sk);
454 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
455 struct sk_buff *skb;
456 size_t copied;
457 int err;
458
459 if (flags & MSG_OOB)
460 return -EOPNOTSUPP;
461
462 if (flags & MSG_ERRQUEUE)
463 return ipv6_recv_error(sk, msg, len);
464
465 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
466 return ipv6_recv_rxpmtu(sk, msg, len);
467
468 skb = skb_recv_datagram(sk, flags, noblock, &err);
469 if (!skb)
470 goto out;
471
472 copied = skb->len;
473 if (copied > len) {
474 copied = len;
475 msg->msg_flags |= MSG_TRUNC;
476 }
477
478 if (skb_csum_unnecessary(skb)) {
479 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
480 } else if (msg->msg_flags&MSG_TRUNC) {
481 if (__skb_checksum_complete(skb))
482 goto csum_copy_err;
483 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484 } else {
485 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
486 if (err == -EINVAL)
487 goto csum_copy_err;
488 }
489 if (err)
490 goto out_free;
491
492 /* Copy the address. */
493 if (sin6) {
494 sin6->sin6_family = AF_INET6;
495 sin6->sin6_port = 0;
496 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
497 sin6->sin6_flowinfo = 0;
498 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
499 IP6CB(skb)->iif);
500 *addr_len = sizeof(*sin6);
501 }
502
503 sock_recv_ts_and_drops(msg, sk, skb);
504
505 if (np->rxopt.all)
506 ip6_datagram_recv_ctl(sk, msg, skb);
507
508 err = copied;
509 if (flags & MSG_TRUNC)
510 err = skb->len;
511
512 out_free:
513 skb_free_datagram(sk, skb);
514 out:
515 return err;
516
517 csum_copy_err:
518 skb_kill_datagram(sk, skb, flags);
519
520 /* Error for blocking case is chosen to masquerade
521 as some normal condition.
522 */
523 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
524 goto out;
525 }
526
527 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
528 struct raw6_sock *rp)
529 {
530 struct sk_buff *skb;
531 int err = 0;
532 int offset;
533 int len;
534 int total_len;
535 __wsum tmp_csum;
536 __sum16 csum;
537
538 if (!rp->checksum)
539 goto send;
540
541 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
542 goto out;
543
544 offset = rp->offset;
545 total_len = inet_sk(sk)->cork.base.length;
546 if (offset >= total_len - 1) {
547 err = -EINVAL;
548 ip6_flush_pending_frames(sk);
549 goto out;
550 }
551
552 /* should be check HW csum miyazawa */
553 if (skb_queue_len(&sk->sk_write_queue) == 1) {
554 /*
555 * Only one fragment on the socket.
556 */
557 tmp_csum = skb->csum;
558 } else {
559 struct sk_buff *csum_skb = NULL;
560 tmp_csum = 0;
561
562 skb_queue_walk(&sk->sk_write_queue, skb) {
563 tmp_csum = csum_add(tmp_csum, skb->csum);
564
565 if (csum_skb)
566 continue;
567
568 len = skb->len - skb_transport_offset(skb);
569 if (offset >= len) {
570 offset -= len;
571 continue;
572 }
573
574 csum_skb = skb;
575 }
576
577 skb = csum_skb;
578 }
579
580 offset += skb_transport_offset(skb);
581 if (skb_copy_bits(skb, offset, &csum, 2))
582 BUG();
583
584 /* in case cksum was not initialized */
585 if (unlikely(csum))
586 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
587
588 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
589 total_len, fl6->flowi6_proto, tmp_csum);
590
591 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
592 csum = CSUM_MANGLED_0;
593
594 if (skb_store_bits(skb, offset, &csum, 2))
595 BUG();
596
597 send:
598 err = ip6_push_pending_frames(sk);
599 out:
600 return err;
601 }
602
603 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
604 struct flowi6 *fl6, struct dst_entry **dstp,
605 unsigned int flags)
606 {
607 struct ipv6_pinfo *np = inet6_sk(sk);
608 struct ipv6hdr *iph;
609 struct sk_buff *skb;
610 int err;
611 struct rt6_info *rt = (struct rt6_info *)*dstp;
612 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
613 int tlen = rt->dst.dev->needed_tailroom;
614
615 if (length > rt->dst.dev->mtu) {
616 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
617 return -EMSGSIZE;
618 }
619 if (flags&MSG_PROBE)
620 goto out;
621
622 skb = sock_alloc_send_skb(sk,
623 length + hlen + tlen + 15,
624 flags & MSG_DONTWAIT, &err);
625 if (skb == NULL)
626 goto error;
627 skb_reserve(skb, hlen);
628
629 skb->priority = sk->sk_priority;
630 skb->mark = sk->sk_mark;
631 skb_dst_set(skb, &rt->dst);
632 *dstp = NULL;
633
634 skb_put(skb, length);
635 skb_reset_network_header(skb);
636 iph = ipv6_hdr(skb);
637
638 skb->ip_summed = CHECKSUM_NONE;
639
640 skb->transport_header = skb->network_header;
641 err = memcpy_fromiovecend((void *)iph, from, 0, length);
642 if (err)
643 goto error_fault;
644
645 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
646 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
647 rt->dst.dev, dst_output);
648 if (err > 0)
649 err = net_xmit_errno(err);
650 if (err)
651 goto error;
652 out:
653 return 0;
654
655 error_fault:
656 err = -EFAULT;
657 kfree_skb(skb);
658 error:
659 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
660 if (err == -ENOBUFS && !np->recverr)
661 err = 0;
662 return err;
663 }
664
665 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
666 {
667 struct iovec *iov;
668 u8 __user *type = NULL;
669 u8 __user *code = NULL;
670 u8 len = 0;
671 int probed = 0;
672 int i;
673
674 if (!msg->msg_iov)
675 return 0;
676
677 for (i = 0; i < msg->msg_iovlen; i++) {
678 iov = &msg->msg_iov[i];
679 if (!iov)
680 continue;
681
682 switch (fl6->flowi6_proto) {
683 case IPPROTO_ICMPV6:
684 /* check if one-byte field is readable or not. */
685 if (iov->iov_base && iov->iov_len < 1)
686 break;
687
688 if (!type) {
689 type = iov->iov_base;
690 /* check if code field is readable or not. */
691 if (iov->iov_len > 1)
692 code = type + 1;
693 } else if (!code)
694 code = iov->iov_base;
695
696 if (type && code) {
697 if (get_user(fl6->fl6_icmp_type, type) ||
698 get_user(fl6->fl6_icmp_code, code))
699 return -EFAULT;
700 probed = 1;
701 }
702 break;
703 case IPPROTO_MH:
704 if (iov->iov_base && iov->iov_len < 1)
705 break;
706 /* check if type field is readable or not. */
707 if (iov->iov_len > 2 - len) {
708 u8 __user *p = iov->iov_base;
709 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
710 return -EFAULT;
711 probed = 1;
712 } else
713 len += iov->iov_len;
714
715 break;
716 default:
717 probed = 1;
718 break;
719 }
720 if (probed)
721 break;
722 }
723 return 0;
724 }
725
726 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
727 struct msghdr *msg, size_t len)
728 {
729 struct ipv6_txoptions opt_space;
730 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
731 struct in6_addr *daddr, *final_p, final;
732 struct inet_sock *inet = inet_sk(sk);
733 struct ipv6_pinfo *np = inet6_sk(sk);
734 struct raw6_sock *rp = raw6_sk(sk);
735 struct ipv6_txoptions *opt = NULL;
736 struct ip6_flowlabel *flowlabel = NULL;
737 struct dst_entry *dst = NULL;
738 struct flowi6 fl6;
739 int addr_len = msg->msg_namelen;
740 int hlimit = -1;
741 int tclass = -1;
742 int dontfrag = -1;
743 u16 proto;
744 int err;
745
746 /* Rough check on arithmetic overflow,
747 better check is made in ip6_append_data().
748 */
749 if (len > INT_MAX)
750 return -EMSGSIZE;
751
752 /* Mirror BSD error message compatibility */
753 if (msg->msg_flags & MSG_OOB)
754 return -EOPNOTSUPP;
755
756 /*
757 * Get and verify the address.
758 */
759 memset(&fl6, 0, sizeof(fl6));
760
761 fl6.flowi6_mark = sk->sk_mark;
762
763 if (sin6) {
764 if (addr_len < SIN6_LEN_RFC2133)
765 return -EINVAL;
766
767 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
768 return -EAFNOSUPPORT;
769
770 /* port is the proto value [0..255] carried in nexthdr */
771 proto = ntohs(sin6->sin6_port);
772
773 if (!proto)
774 proto = inet->inet_num;
775 else if (proto != inet->inet_num)
776 return -EINVAL;
777
778 if (proto > 255)
779 return -EINVAL;
780
781 daddr = &sin6->sin6_addr;
782 if (np->sndflow) {
783 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
784 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
785 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
786 if (flowlabel == NULL)
787 return -EINVAL;
788 daddr = &flowlabel->dst;
789 }
790 }
791
792 /*
793 * Otherwise it will be difficult to maintain
794 * sk->sk_dst_cache.
795 */
796 if (sk->sk_state == TCP_ESTABLISHED &&
797 ipv6_addr_equal(daddr, &np->daddr))
798 daddr = &np->daddr;
799
800 if (addr_len >= sizeof(struct sockaddr_in6) &&
801 sin6->sin6_scope_id &&
802 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
803 fl6.flowi6_oif = sin6->sin6_scope_id;
804 } else {
805 if (sk->sk_state != TCP_ESTABLISHED)
806 return -EDESTADDRREQ;
807
808 proto = inet->inet_num;
809 daddr = &np->daddr;
810 fl6.flowlabel = np->flow_label;
811 }
812
813 if (fl6.flowi6_oif == 0)
814 fl6.flowi6_oif = sk->sk_bound_dev_if;
815
816 if (msg->msg_controllen) {
817 opt = &opt_space;
818 memset(opt, 0, sizeof(struct ipv6_txoptions));
819 opt->tot_len = sizeof(struct ipv6_txoptions);
820
821 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
822 &hlimit, &tclass, &dontfrag);
823 if (err < 0) {
824 fl6_sock_release(flowlabel);
825 return err;
826 }
827 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
828 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
829 if (flowlabel == NULL)
830 return -EINVAL;
831 }
832 if (!(opt->opt_nflen|opt->opt_flen))
833 opt = NULL;
834 }
835 if (opt == NULL)
836 opt = np->opt;
837 if (flowlabel)
838 opt = fl6_merge_options(&opt_space, flowlabel, opt);
839 opt = ipv6_fixup_options(&opt_space, opt);
840
841 fl6.flowi6_proto = proto;
842 err = rawv6_probe_proto_opt(&fl6, msg);
843 if (err)
844 goto out;
845
846 if (!ipv6_addr_any(daddr))
847 fl6.daddr = *daddr;
848 else
849 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
850 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
851 fl6.saddr = np->saddr;
852
853 final_p = fl6_update_dst(&fl6, opt, &final);
854
855 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
856 fl6.flowi6_oif = np->mcast_oif;
857 else if (!fl6.flowi6_oif)
858 fl6.flowi6_oif = np->ucast_oif;
859 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
860
861 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
862 if (IS_ERR(dst)) {
863 err = PTR_ERR(dst);
864 goto out;
865 }
866 if (hlimit < 0) {
867 if (ipv6_addr_is_multicast(&fl6.daddr))
868 hlimit = np->mcast_hops;
869 else
870 hlimit = np->hop_limit;
871 if (hlimit < 0)
872 hlimit = ip6_dst_hoplimit(dst);
873 }
874
875 if (tclass < 0)
876 tclass = np->tclass;
877
878 if (dontfrag < 0)
879 dontfrag = np->dontfrag;
880
881 if (msg->msg_flags&MSG_CONFIRM)
882 goto do_confirm;
883
884 back_from_confirm:
885 if (inet->hdrincl)
886 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
887 else {
888 lock_sock(sk);
889 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
890 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
891 msg->msg_flags, dontfrag);
892
893 if (err)
894 ip6_flush_pending_frames(sk);
895 else if (!(msg->msg_flags & MSG_MORE))
896 err = rawv6_push_pending_frames(sk, &fl6, rp);
897 release_sock(sk);
898 }
899 done:
900 dst_release(dst);
901 out:
902 fl6_sock_release(flowlabel);
903 return err<0?err:len;
904 do_confirm:
905 dst_confirm(dst);
906 if (!(msg->msg_flags & MSG_PROBE) || len)
907 goto back_from_confirm;
908 err = 0;
909 goto done;
910 }
911
912 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
913 char __user *optval, int optlen)
914 {
915 switch (optname) {
916 case ICMPV6_FILTER:
917 if (optlen > sizeof(struct icmp6_filter))
918 optlen = sizeof(struct icmp6_filter);
919 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
920 return -EFAULT;
921 return 0;
922 default:
923 return -ENOPROTOOPT;
924 }
925
926 return 0;
927 }
928
929 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
930 char __user *optval, int __user *optlen)
931 {
932 int len;
933
934 switch (optname) {
935 case ICMPV6_FILTER:
936 if (get_user(len, optlen))
937 return -EFAULT;
938 if (len < 0)
939 return -EINVAL;
940 if (len > sizeof(struct icmp6_filter))
941 len = sizeof(struct icmp6_filter);
942 if (put_user(len, optlen))
943 return -EFAULT;
944 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
945 return -EFAULT;
946 return 0;
947 default:
948 return -ENOPROTOOPT;
949 }
950
951 return 0;
952 }
953
954
955 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
956 char __user *optval, unsigned int optlen)
957 {
958 struct raw6_sock *rp = raw6_sk(sk);
959 int val;
960
961 if (get_user(val, (int __user *)optval))
962 return -EFAULT;
963
964 switch (optname) {
965 case IPV6_CHECKSUM:
966 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
967 level == IPPROTO_IPV6) {
968 /*
969 * RFC3542 tells that IPV6_CHECKSUM socket
970 * option in the IPPROTO_IPV6 level is not
971 * allowed on ICMPv6 sockets.
972 * If you want to set it, use IPPROTO_RAW
973 * level IPV6_CHECKSUM socket option
974 * (Linux extension).
975 */
976 return -EINVAL;
977 }
978
979 /* You may get strange result with a positive odd offset;
980 RFC2292bis agrees with me. */
981 if (val > 0 && (val&1))
982 return -EINVAL;
983 if (val < 0) {
984 rp->checksum = 0;
985 } else {
986 rp->checksum = 1;
987 rp->offset = val;
988 }
989
990 return 0;
991
992 default:
993 return -ENOPROTOOPT;
994 }
995 }
996
997 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
998 char __user *optval, unsigned int optlen)
999 {
1000 switch (level) {
1001 case SOL_RAW:
1002 break;
1003
1004 case SOL_ICMPV6:
1005 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1006 return -EOPNOTSUPP;
1007 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1008 case SOL_IPV6:
1009 if (optname == IPV6_CHECKSUM)
1010 break;
1011 default:
1012 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1013 }
1014
1015 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1016 }
1017
1018 #ifdef CONFIG_COMPAT
1019 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1020 char __user *optval, unsigned int optlen)
1021 {
1022 switch (level) {
1023 case SOL_RAW:
1024 break;
1025 case SOL_ICMPV6:
1026 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1027 return -EOPNOTSUPP;
1028 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1029 case SOL_IPV6:
1030 if (optname == IPV6_CHECKSUM)
1031 break;
1032 default:
1033 return compat_ipv6_setsockopt(sk, level, optname,
1034 optval, optlen);
1035 }
1036 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1037 }
1038 #endif
1039
1040 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1041 char __user *optval, int __user *optlen)
1042 {
1043 struct raw6_sock *rp = raw6_sk(sk);
1044 int val, len;
1045
1046 if (get_user(len,optlen))
1047 return -EFAULT;
1048
1049 switch (optname) {
1050 case IPV6_CHECKSUM:
1051 /*
1052 * We allow getsockopt() for IPPROTO_IPV6-level
1053 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1054 * since RFC3542 is silent about it.
1055 */
1056 if (rp->checksum == 0)
1057 val = -1;
1058 else
1059 val = rp->offset;
1060 break;
1061
1062 default:
1063 return -ENOPROTOOPT;
1064 }
1065
1066 len = min_t(unsigned int, sizeof(int), len);
1067
1068 if (put_user(len, optlen))
1069 return -EFAULT;
1070 if (copy_to_user(optval,&val,len))
1071 return -EFAULT;
1072 return 0;
1073 }
1074
1075 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1076 char __user *optval, int __user *optlen)
1077 {
1078 switch (level) {
1079 case SOL_RAW:
1080 break;
1081
1082 case SOL_ICMPV6:
1083 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1084 return -EOPNOTSUPP;
1085 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1086 case SOL_IPV6:
1087 if (optname == IPV6_CHECKSUM)
1088 break;
1089 default:
1090 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1091 }
1092
1093 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1094 }
1095
1096 #ifdef CONFIG_COMPAT
1097 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1098 char __user *optval, int __user *optlen)
1099 {
1100 switch (level) {
1101 case SOL_RAW:
1102 break;
1103 case SOL_ICMPV6:
1104 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1105 return -EOPNOTSUPP;
1106 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1107 case SOL_IPV6:
1108 if (optname == IPV6_CHECKSUM)
1109 break;
1110 default:
1111 return compat_ipv6_getsockopt(sk, level, optname,
1112 optval, optlen);
1113 }
1114 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1115 }
1116 #endif
1117
1118 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1119 {
1120 switch (cmd) {
1121 case SIOCOUTQ: {
1122 int amount = sk_wmem_alloc_get(sk);
1123
1124 return put_user(amount, (int __user *)arg);
1125 }
1126 case SIOCINQ: {
1127 struct sk_buff *skb;
1128 int amount = 0;
1129
1130 spin_lock_bh(&sk->sk_receive_queue.lock);
1131 skb = skb_peek(&sk->sk_receive_queue);
1132 if (skb != NULL)
1133 amount = skb->tail - skb->transport_header;
1134 spin_unlock_bh(&sk->sk_receive_queue.lock);
1135 return put_user(amount, (int __user *)arg);
1136 }
1137
1138 default:
1139 #ifdef CONFIG_IPV6_MROUTE
1140 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1141 #else
1142 return -ENOIOCTLCMD;
1143 #endif
1144 }
1145 }
1146
1147 #ifdef CONFIG_COMPAT
1148 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1149 {
1150 switch (cmd) {
1151 case SIOCOUTQ:
1152 case SIOCINQ:
1153 return -ENOIOCTLCMD;
1154 default:
1155 #ifdef CONFIG_IPV6_MROUTE
1156 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1157 #else
1158 return -ENOIOCTLCMD;
1159 #endif
1160 }
1161 }
1162 #endif
1163
1164 static void rawv6_close(struct sock *sk, long timeout)
1165 {
1166 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1167 ip6_ra_control(sk, -1);
1168 ip6mr_sk_done(sk);
1169 sk_common_release(sk);
1170 }
1171
1172 static void raw6_destroy(struct sock *sk)
1173 {
1174 lock_sock(sk);
1175 ip6_flush_pending_frames(sk);
1176 release_sock(sk);
1177
1178 inet6_destroy_sock(sk);
1179 }
1180
1181 static int rawv6_init_sk(struct sock *sk)
1182 {
1183 struct raw6_sock *rp = raw6_sk(sk);
1184
1185 switch (inet_sk(sk)->inet_num) {
1186 case IPPROTO_ICMPV6:
1187 rp->checksum = 1;
1188 rp->offset = 2;
1189 break;
1190 case IPPROTO_MH:
1191 rp->checksum = 1;
1192 rp->offset = 4;
1193 break;
1194 default:
1195 break;
1196 }
1197 return 0;
1198 }
1199
1200 struct proto rawv6_prot = {
1201 .name = "RAWv6",
1202 .owner = THIS_MODULE,
1203 .close = rawv6_close,
1204 .destroy = raw6_destroy,
1205 .connect = ip6_datagram_connect,
1206 .disconnect = udp_disconnect,
1207 .ioctl = rawv6_ioctl,
1208 .init = rawv6_init_sk,
1209 .setsockopt = rawv6_setsockopt,
1210 .getsockopt = rawv6_getsockopt,
1211 .sendmsg = rawv6_sendmsg,
1212 .recvmsg = rawv6_recvmsg,
1213 .bind = rawv6_bind,
1214 .backlog_rcv = rawv6_rcv_skb,
1215 .hash = raw_hash_sk,
1216 .unhash = raw_unhash_sk,
1217 .obj_size = sizeof(struct raw6_sock),
1218 .h.raw_hash = &raw_v6_hashinfo,
1219 #ifdef CONFIG_COMPAT
1220 .compat_setsockopt = compat_rawv6_setsockopt,
1221 .compat_getsockopt = compat_rawv6_getsockopt,
1222 .compat_ioctl = compat_rawv6_ioctl,
1223 #endif
1224 };
1225
1226 #ifdef CONFIG_PROC_FS
1227 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1228 {
1229 struct ipv6_pinfo *np = inet6_sk(sp);
1230 const struct in6_addr *dest, *src;
1231 __u16 destp, srcp;
1232
1233 dest = &np->daddr;
1234 src = &np->rcv_saddr;
1235 destp = 0;
1236 srcp = inet_sk(sp)->inet_num;
1237 seq_printf(seq,
1238 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1239 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1240 i,
1241 src->s6_addr32[0], src->s6_addr32[1],
1242 src->s6_addr32[2], src->s6_addr32[3], srcp,
1243 dest->s6_addr32[0], dest->s6_addr32[1],
1244 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1245 sp->sk_state,
1246 sk_wmem_alloc_get(sp),
1247 sk_rmem_alloc_get(sp),
1248 0, 0L, 0,
1249 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1250 0,
1251 sock_i_ino(sp),
1252 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1253 }
1254
1255 static int raw6_seq_show(struct seq_file *seq, void *v)
1256 {
1257 if (v == SEQ_START_TOKEN)
1258 seq_printf(seq,
1259 " sl "
1260 "local_address "
1261 "remote_address "
1262 "st tx_queue rx_queue tr tm->when retrnsmt"
1263 " uid timeout inode ref pointer drops\n");
1264 else
1265 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1266 return 0;
1267 }
1268
1269 static const struct seq_operations raw6_seq_ops = {
1270 .start = raw_seq_start,
1271 .next = raw_seq_next,
1272 .stop = raw_seq_stop,
1273 .show = raw6_seq_show,
1274 };
1275
1276 static int raw6_seq_open(struct inode *inode, struct file *file)
1277 {
1278 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1279 }
1280
1281 static const struct file_operations raw6_seq_fops = {
1282 .owner = THIS_MODULE,
1283 .open = raw6_seq_open,
1284 .read = seq_read,
1285 .llseek = seq_lseek,
1286 .release = seq_release_net,
1287 };
1288
1289 static int __net_init raw6_init_net(struct net *net)
1290 {
1291 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1292 return -ENOMEM;
1293
1294 return 0;
1295 }
1296
1297 static void __net_exit raw6_exit_net(struct net *net)
1298 {
1299 remove_proc_entry("raw6", net->proc_net);
1300 }
1301
1302 static struct pernet_operations raw6_net_ops = {
1303 .init = raw6_init_net,
1304 .exit = raw6_exit_net,
1305 };
1306
1307 int __init raw6_proc_init(void)
1308 {
1309 return register_pernet_subsys(&raw6_net_ops);
1310 }
1311
1312 void raw6_proc_exit(void)
1313 {
1314 unregister_pernet_subsys(&raw6_net_ops);
1315 }
1316 #endif /* CONFIG_PROC_FS */
1317
1318 /* Same as inet6_dgram_ops, sans udp_poll. */
1319 static const struct proto_ops inet6_sockraw_ops = {
1320 .family = PF_INET6,
1321 .owner = THIS_MODULE,
1322 .release = inet6_release,
1323 .bind = inet6_bind,
1324 .connect = inet_dgram_connect, /* ok */
1325 .socketpair = sock_no_socketpair, /* a do nothing */
1326 .accept = sock_no_accept, /* a do nothing */
1327 .getname = inet6_getname,
1328 .poll = datagram_poll, /* ok */
1329 .ioctl = inet6_ioctl, /* must change */
1330 .listen = sock_no_listen, /* ok */
1331 .shutdown = inet_shutdown, /* ok */
1332 .setsockopt = sock_common_setsockopt, /* ok */
1333 .getsockopt = sock_common_getsockopt, /* ok */
1334 .sendmsg = inet_sendmsg, /* ok */
1335 .recvmsg = sock_common_recvmsg, /* ok */
1336 .mmap = sock_no_mmap,
1337 .sendpage = sock_no_sendpage,
1338 #ifdef CONFIG_COMPAT
1339 .compat_setsockopt = compat_sock_common_setsockopt,
1340 .compat_getsockopt = compat_sock_common_getsockopt,
1341 #endif
1342 };
1343
1344 static struct inet_protosw rawv6_protosw = {
1345 .type = SOCK_RAW,
1346 .protocol = IPPROTO_IP, /* wild card */
1347 .prot = &rawv6_prot,
1348 .ops = &inet6_sockraw_ops,
1349 .no_check = UDP_CSUM_DEFAULT,
1350 .flags = INET_PROTOSW_REUSE,
1351 };
1352
1353 int __init rawv6_init(void)
1354 {
1355 int ret;
1356
1357 ret = inet6_register_protosw(&rawv6_protosw);
1358 if (ret)
1359 goto out;
1360 out:
1361 return ret;
1362 }
1363
1364 void rawv6_exit(void)
1365 {
1366 inet6_unregister_protosw(&rawv6_protosw);
1367 }
kernel source for telekom puls (alcatel/mediatek)