]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - net/ipv6/raw.c
projects / mirror_ubuntu-zesty-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[IPV6]: Support several new sockopt / ancillary data in Advanced API (RFC3542).
[mirror_ubuntu-zesty-kernel.git] / net / ipv6 / raw.c
CommitLineData
1da177e4
LT		 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/sched.h>
28#include <linux/net.h>
29#include <linux/in6.h>
30#include <linux/netdevice.h>
31#include <linux/if_arp.h>
32#include <linux/icmpv6.h>
33#include <linux/netfilter.h>
34#include <linux/netfilter_ipv6.h>
35#include <asm/uaccess.h>
36#include <asm/ioctls.h>
357b40a1 37#include <asm/bug.h>
1da177e4
LT		 38
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>
c752f073 52#include <net/tcp_states.h>
1da177e4
LT		 53
54#include <net/rawv6.h>
55#include <net/xfrm.h>
56
57#include <linux/proc_fs.h>
58#include <linux/seq_file.h>
59
60struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
61DEFINE_RWLOCK(raw_v6_lock);
62
63static void raw_v6_hash(struct sock *sk)
64{
65 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
66 (RAWV6_HTABLE_SIZE - 1)];
67
68 write_lock_bh(&raw_v6_lock);
69 sk_add_node(sk, list);
70 sock_prot_inc_use(sk->sk_prot);
71 write_unlock_bh(&raw_v6_lock);
72}
73
74static void raw_v6_unhash(struct sock *sk)
75{
76 write_lock_bh(&raw_v6_lock);
77 if (sk_del_node_init(sk))
78 sock_prot_dec_use(sk->sk_prot);
79 write_unlock_bh(&raw_v6_lock);
80}
81
82
83/* Grumble... icmp and ip_input want to get at this... */
84struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
0bd1b59b
AM		 85 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
86 int dif)
1da177e4
LT		 87{
88 struct hlist_node *node;
89 int is_multicast = ipv6_addr_is_multicast(loc_addr);
90
91 sk_for_each_from(sk, node)
92 if (inet_sk(sk)->num == num) {
93 struct ipv6_pinfo *np = inet6_sk(sk);
94
95 if (!ipv6_addr_any(&np->daddr) &&
96 !ipv6_addr_equal(&np->daddr, rmt_addr))
97 continue;
98
0bd1b59b
AM		 99 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
100 continue;
101
1da177e4
LT		 102 if (!ipv6_addr_any(&np->rcv_saddr)) {
103 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
104 goto found;
105 if (is_multicast &&
106 inet6_mc_check(sk, loc_addr, rmt_addr))
107 goto found;
108 continue;
109 }
110 goto found;
111 }
112 sk = NULL;
113found:
114 return sk;
115}
116
117/*
118 * 0 - deliver
119 * 1 - block
120 */
121static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
122{
123 struct icmp6hdr *icmph;
124 struct raw6_sock *rp = raw6_sk(sk);
125
126 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
127 __u32 *data = &rp->filter.data[0];
128 int bit_nr;
129
130 icmph = (struct icmp6hdr *) skb->data;
131 bit_nr = icmph->icmp6_type;
132
133 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
134 }
135 return 0;
136}
137
138/*
139 * demultiplex raw sockets.
140 * (should consider queueing the skb in the sock receive_queue
141 * without calling rawv6.c)
142 *
143 * Caller owns SKB so we must make clones.
144 */
d13964f4 145int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
1da177e4
LT		 146{
147 struct in6_addr *saddr;
148 struct in6_addr *daddr;
149 struct sock *sk;
d13964f4 150 int delivered = 0;
1da177e4
LT		 151 __u8 hash;
152
153 saddr = &skb->nh.ipv6h->saddr;
154 daddr = saddr + 1;
155
156 hash = nexthdr & (MAX_INET_PROTOS - 1);
157
158 read_lock(&raw_v6_lock);
159 sk = sk_head(&raw_v6_htable[hash]);
160
161 /*
162 * The first socket found will be delivered after
163 * delivery to transport protocols.
164 */
165
166 if (sk == NULL)
167 goto out;
168
2dac4b96 169 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
1da177e4
LT		 170
171 while (sk) {
d13964f4 172 delivered = 1;
1da177e4
LT		 173 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
174 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
175
176 /* Not releasing hash table! */
177 if (clone)
178 rawv6_rcv(sk, clone);
179 }
0bd1b59b 180 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
2dac4b96 181 IP6CB(skb)->iif);
1da177e4
LT		 182 }
183out:
184 read_unlock(&raw_v6_lock);
d13964f4 185 return delivered;
1da177e4
LT		 186}
187
188/* This cleans up af_inet6 a bit. -DaveM */
189static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
190{
191 struct inet_sock *inet = inet_sk(sk);
192 struct ipv6_pinfo *np = inet6_sk(sk);
193 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
194 __u32 v4addr = 0;
195 int addr_type;
196 int err;
197
198 if (addr_len < SIN6_LEN_RFC2133)
199 return -EINVAL;
200 addr_type = ipv6_addr_type(&addr->sin6_addr);
201
202 /* Raw sockets are IPv6 only */
203 if (addr_type == IPV6_ADDR_MAPPED)
204 return(-EADDRNOTAVAIL);
205
206 lock_sock(sk);
207
208 err = -EINVAL;
209 if (sk->sk_state != TCP_CLOSE)
210 goto out;
211
212 /* Check if the address belongs to the host. */
213 if (addr_type != IPV6_ADDR_ANY) {
214 struct net_device *dev = NULL;
215
216 if (addr_type & IPV6_ADDR_LINKLOCAL) {
217 if (addr_len >= sizeof(struct sockaddr_in6) &&
218 addr->sin6_scope_id) {
219 /* Override any existing binding, if another
220 * one is supplied by user.
221 */
222 sk->sk_bound_dev_if = addr->sin6_scope_id;
223 }
224
225 /* Binding to link-local address requires an interface */
226 if (!sk->sk_bound_dev_if)
227 goto out;
228
229 dev = dev_get_by_index(sk->sk_bound_dev_if);
230 if (!dev) {
231 err = -ENODEV;
232 goto out;
233 }
234 }
235
236 /* ipv4 addr of the socket is invalid. Only the
237 * unspecified and mapped address have a v4 equivalent.
238 */
239 v4addr = LOOPBACK4_IPV6;
240 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
241 err = -EADDRNOTAVAIL;
242 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
243 if (dev)
244 dev_put(dev);
245 goto out;
246 }
247 }
248 if (dev)
249 dev_put(dev);
250 }
251
252 inet->rcv_saddr = inet->saddr = v4addr;
253 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
254 if (!(addr_type & IPV6_ADDR_MULTICAST))
255 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
256 err = 0;
257out:
258 release_sock(sk);
259 return err;
260}
261
262void rawv6_err(struct sock *sk, struct sk_buff *skb,
263 struct inet6_skb_parm *opt,
264 int type, int code, int offset, u32 info)
265{
266 struct inet_sock *inet = inet_sk(sk);
267 struct ipv6_pinfo *np = inet6_sk(sk);
268 int err;
269 int harderr;
270
271 /* Report error on raw socket, if:
272 1. User requested recverr.
273 2. Socket is connected (otherwise the error indication
274 is useless without recverr and error is hard.
275 */
276 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
277 return;
278
279 harderr = icmpv6_err_convert(type, code, &err);
280 if (type == ICMPV6_PKT_TOOBIG)
281 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
282
283 if (np->recverr) {
284 u8 *payload = skb->data;
285 if (!inet->hdrincl)
286 payload += offset;
287 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
288 }
289
290 if (np->recverr || harderr) {
291 sk->sk_err = err;
292 sk->sk_error_report(sk);
293 }
294}
295
296static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
297{
298 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
299 skb->ip_summed != CHECKSUM_UNNECESSARY) {
300 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
301 /* FIXME: increment a raw6 drops counter here */
302 kfree_skb(skb);
303 return 0;
304 }
305 skb->ip_summed = CHECKSUM_UNNECESSARY;
306 }
307
308 /* Charge it to the socket. */
309 if (sock_queue_rcv_skb(sk,skb)<0) {
310 /* FIXME: increment a raw6 drops counter here */
311 kfree_skb(skb);
312 return 0;
313 }
314
315 return 0;
316}
317
318/*
319 * This is next to useless...
320 * if we demultiplex in network layer we don't need the extra call
321 * just to queue the skb...
322 * maybe we could have the network decide upon a hint if it
323 * should call raw_rcv for demultiplexing
324 */
325int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
326{
327 struct inet_sock *inet = inet_sk(sk);
328 struct raw6_sock *rp = raw6_sk(sk);
329
330 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
331 kfree_skb(skb);
332 return NET_RX_DROP;
333 }
334
335 if (!rp->checksum)
336 skb->ip_summed = CHECKSUM_UNNECESSARY;
337
338 if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
339 if (skb->ip_summed == CHECKSUM_HW) {
793245ee
PM		 340 skb_postpull_rcsum(skb, skb->nh.raw,
341 skb->h.raw - skb->nh.raw);
1da177e4
LT		 342 skb->ip_summed = CHECKSUM_UNNECESSARY;
343 if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
344 &skb->nh.ipv6h->daddr,
345 skb->len, inet->num, skb->csum)) {
64ce2073 346 LIMIT_NETDEBUG(KERN_DEBUG "raw v6 hw csum failure.\n");
1da177e4
LT		 347 skb->ip_summed = CHECKSUM_NONE;
348 }
349 }
350 if (skb->ip_summed == CHECKSUM_NONE)
351 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
352 &skb->nh.ipv6h->daddr,
353 skb->len, inet->num, 0);
354 }
355
356 if (inet->hdrincl) {
357 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
358 (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
359 /* FIXME: increment a raw6 drops counter here */
360 kfree_skb(skb);
361 return 0;
362 }
363 skb->ip_summed = CHECKSUM_UNNECESSARY;
364 }
365
366 rawv6_rcv_skb(sk, skb);
367 return 0;
368}
369
370
371/*
372 * This should be easy, if there is something there
373 * we return it, otherwise we block.
374 */
375
376static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
377 struct msghdr *msg, size_t len,
378 int noblock, int flags, int *addr_len)
379{
380 struct ipv6_pinfo *np = inet6_sk(sk);
381 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
382 struct sk_buff *skb;
383 size_t copied;
384 int err;
385
386 if (flags & MSG_OOB)
387 return -EOPNOTSUPP;
388
389 if (addr_len)
390 *addr_len=sizeof(*sin6);
391
392 if (flags & MSG_ERRQUEUE)
393 return ipv6_recv_error(sk, msg, len);
394
395 skb = skb_recv_datagram(sk, flags, noblock, &err);
396 if (!skb)
397 goto out;
398
399 copied = skb->len;
400 if (copied > len) {
401 copied = len;
402 msg->msg_flags |= MSG_TRUNC;
403 }
404
405 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
406 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
407 } else if (msg->msg_flags&MSG_TRUNC) {
408 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
409 goto csum_copy_err;
410 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
411 } else {
412 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
413 if (err == -EINVAL)
414 goto csum_copy_err;
415 }
416 if (err)
417 goto out_free;
418
419 /* Copy the address. */
420 if (sin6) {
421 sin6->sin6_family = AF_INET6;
422 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
423 sin6->sin6_flowinfo = 0;
424 sin6->sin6_scope_id = 0;
425 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
426 sin6->sin6_scope_id = IP6CB(skb)->iif;
427 }
428
429 sock_recv_timestamp(msg, sk, skb);
430
431 if (np->rxopt.all)
432 datagram_recv_ctl(sk, msg, skb);
433
434 err = copied;
435 if (flags & MSG_TRUNC)
436 err = skb->len;
437
438out_free:
439 skb_free_datagram(sk, skb);
440out:
441 return err;
442
443csum_copy_err:
444 /* Clear queue. */
445 if (flags&MSG_PEEK) {
446 int clear = 0;
e0f9f858 447 spin_lock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 448 if (skb == skb_peek(&sk->sk_receive_queue)) {
449 __skb_unlink(skb, &sk->sk_receive_queue);
450 clear = 1;
451 }
e0f9f858 452 spin_unlock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 453 if (clear)
454 kfree_skb(skb);
455 }
456
457 /* Error for blocking case is chosen to masquerade
458 as some normal condition.
459 */
460 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
461 /* FIXME: increment a raw6 drops counter here */
462 goto out_free;
463}
464
465static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
357b40a1 466 struct raw6_sock *rp)
1da177e4
LT		 467{
468 struct sk_buff *skb;
469 int err = 0;
357b40a1
HX		 470 int offset;
471 int len;
679a8738 472 int total_len;
1da177e4 473 u32 tmp_csum;
357b40a1 474 u16 csum;
1da177e4
LT		 475
476 if (!rp->checksum)
477 goto send;
478
479 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
480 goto out;
481
357b40a1 482 offset = rp->offset;
679a8738
HX		 483 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
484 if (offset >= total_len - 1) {
1da177e4 485 err = -EINVAL;
357b40a1 486 ip6_flush_pending_frames(sk);
1da177e4
LT		 487 goto out;
488 }
489
490 /* should be check HW csum miyazawa */
491 if (skb_queue_len(&sk->sk_write_queue) == 1) {
492 /*
493 * Only one fragment on the socket.
494 */
495 tmp_csum = skb->csum;
496 } else {
357b40a1 497 struct sk_buff *csum_skb = NULL;
1da177e4
LT		 498 tmp_csum = 0;
499
500 skb_queue_walk(&sk->sk_write_queue, skb) {
501 tmp_csum = csum_add(tmp_csum, skb->csum);
357b40a1
HX		 502
503 if (csum_skb)
504 continue;
505
506 len = skb->len - (skb->h.raw - skb->data);
507 if (offset >= len) {
508 offset -= len;
509 continue;
510 }
511
512 csum_skb = skb;
1da177e4 513 }
357b40a1
HX		 514
515 skb = csum_skb;
1da177e4
LT		 516 }
517
357b40a1
HX		 518 offset += skb->h.raw - skb->data;
519 if (skb_copy_bits(skb, offset, &csum, 2))
520 BUG();
521
1da177e4 522 /* in case cksum was not initialized */
357b40a1
HX		 523 if (unlikely(csum))
524 tmp_csum = csum_sub(tmp_csum, csum);
525
526 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
527 &fl->fl6_dst,
679a8738 528 total_len, fl->proto, tmp_csum);
357b40a1
HX		 529
530 if (tmp_csum == 0)
531 tmp_csum = -1;
1da177e4 532
357b40a1
HX		 533 csum = tmp_csum;
534 if (skb_store_bits(skb, offset, &csum, 2))
535 BUG();
1da177e4 536
1da177e4
LT		 537send:
538 err = ip6_push_pending_frames(sk);
539out:
540 return err;
541}
542
543static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
544 struct flowi *fl, struct rt6_info *rt,
545 unsigned int flags)
546{
3320da89 547 struct ipv6_pinfo *np = inet6_sk(sk);
1da177e4
LT		 548 struct ipv6hdr *iph;
549 struct sk_buff *skb;
550 unsigned int hh_len;
551 int err;
552
553 if (length > rt->u.dst.dev->mtu) {
554 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
555 return -EMSGSIZE;
556 }
557 if (flags&MSG_PROBE)
558 goto out;
559
560 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
561
562 skb = sock_alloc_send_skb(sk, length+hh_len+15,
563 flags&MSG_DONTWAIT, &err);
564 if (skb == NULL)
565 goto error;
566 skb_reserve(skb, hh_len);
567
568 skb->priority = sk->sk_priority;
569 skb->dst = dst_clone(&rt->u.dst);
570
571 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
572
573 skb->ip_summed = CHECKSUM_NONE;
574
575 skb->h.raw = skb->nh.raw;
576 err = memcpy_fromiovecend((void *)iph, from, 0, length);
577 if (err)
578 goto error_fault;
579
580 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
581 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
582 dst_output);
583 if (err > 0)
3320da89 584 err = np->recverr ? net_xmit_errno(err) : 0;
1da177e4
LT		 585 if (err)
586 goto error;
587out:
588 return 0;
589
590error_fault:
591 err = -EFAULT;
592 kfree_skb(skb);
593error:
594 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
595 return err;
596}
597
598static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
599{
600 struct iovec *iov;
601 u8 __user *type = NULL;
602 u8 __user *code = NULL;
603 int probed = 0;
604 int i;
605
606 if (!msg->msg_iov)
607 return;
608
609 for (i = 0; i < msg->msg_iovlen; i++) {
610 iov = &msg->msg_iov[i];
611 if (!iov)
612 continue;
613
614 switch (fl->proto) {
615 case IPPROTO_ICMPV6:
616 /* check if one-byte field is readable or not. */
617 if (iov->iov_base && iov->iov_len < 1)
618 break;
619
620 if (!type) {
621 type = iov->iov_base;
622 /* check if code field is readable or not. */
623 if (iov->iov_len > 1)
624 code = type + 1;
625 } else if (!code)
626 code = iov->iov_base;
627
628 if (type && code) {
629 get_user(fl->fl_icmp_type, type);
630 __get_user(fl->fl_icmp_code, code);
631 probed = 1;
632 }
633 break;
634 default:
635 probed = 1;
636 break;
637 }
638 if (probed)
639 break;
640 }
641}
642
643static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
644 struct msghdr *msg, size_t len)
645{
646 struct ipv6_txoptions opt_space;
647 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
648 struct in6_addr *daddr, *final_p = NULL, final;
649 struct inet_sock *inet = inet_sk(sk);
650 struct ipv6_pinfo *np = inet6_sk(sk);
651 struct raw6_sock *rp = raw6_sk(sk);
652 struct ipv6_txoptions *opt = NULL;
653 struct ip6_flowlabel *flowlabel = NULL;
654 struct dst_entry *dst = NULL;
655 struct flowi fl;
656 int addr_len = msg->msg_namelen;
657 int hlimit = -1;
658 u16 proto;
659 int err;
660
661 /* Rough check on arithmetic overflow,
662 better check is made in ip6_build_xmit
663 */
664 if (len < 0)
665 return -EMSGSIZE;
666
667 /* Mirror BSD error message compatibility */
668 if (msg->msg_flags & MSG_OOB)
669 return -EOPNOTSUPP;
670
671 /*
672 * Get and verify the address.
673 */
674 memset(&fl, 0, sizeof(fl));
675
676 if (sin6) {
677 if (addr_len < SIN6_LEN_RFC2133)
678 return -EINVAL;
679
680 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
681 return(-EAFNOSUPPORT);
682
683 /* port is the proto value [0..255] carried in nexthdr */
684 proto = ntohs(sin6->sin6_port);
685
686 if (!proto)
687 proto = inet->num;
688 else if (proto != inet->num)
689 return(-EINVAL);
690
691 if (proto > 255)
692 return(-EINVAL);
693
694 daddr = &sin6->sin6_addr;
695 if (np->sndflow) {
696 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
697 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
698 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
699 if (flowlabel == NULL)
700 return -EINVAL;
701 daddr = &flowlabel->dst;
702 }
703 }
704
705 /*
706 * Otherwise it will be difficult to maintain
707 * sk->sk_dst_cache.
708 */
709 if (sk->sk_state == TCP_ESTABLISHED &&
710 ipv6_addr_equal(daddr, &np->daddr))
711 daddr = &np->daddr;
712
713 if (addr_len >= sizeof(struct sockaddr_in6) &&
714 sin6->sin6_scope_id &&
715 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
716 fl.oif = sin6->sin6_scope_id;
717 } else {
718 if (sk->sk_state != TCP_ESTABLISHED)
719 return -EDESTADDRREQ;
720
721 proto = inet->num;
722 daddr = &np->daddr;
723 fl.fl6_flowlabel = np->flow_label;
724 }
725
726 if (ipv6_addr_any(daddr)) {
727 /*
728 * unspecified destination address
729 * treated as error... is this correct ?
730 */
731 fl6_sock_release(flowlabel);
732 return(-EINVAL);
733 }
734
735 if (fl.oif == 0)
736 fl.oif = sk->sk_bound_dev_if;
737
738 if (msg->msg_controllen) {
739 opt = &opt_space;
740 memset(opt, 0, sizeof(struct ipv6_txoptions));
741 opt->tot_len = sizeof(struct ipv6_txoptions);
742
743 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
744 if (err < 0) {
745 fl6_sock_release(flowlabel);
746 return err;
747 }
748 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
749 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
750 if (flowlabel == NULL)
751 return -EINVAL;
752 }
753 if (!(opt->opt_nflen|opt->opt_flen))
754 opt = NULL;
755 }
756 if (opt == NULL)
757 opt = np->opt;
333fad53 758 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1da177e4
LT		 759
760 fl.proto = proto;
761 rawv6_probe_proto_opt(&fl, msg);
762
763 ipv6_addr_copy(&fl.fl6_dst, daddr);
764 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
765 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
766
767 /* merge ip6_build_xmit from ip6_output */
768 if (opt && opt->srcrt) {
769 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
770 ipv6_addr_copy(&final, &fl.fl6_dst);
771 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
772 final_p = &final;
773 }
774
775 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
776 fl.oif = np->mcast_oif;
777
778 err = ip6_dst_lookup(sk, &dst, &fl);
779 if (err)
780 goto out;
781 if (final_p)
782 ipv6_addr_copy(&fl.fl6_dst, final_p);
783
784 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
785 dst_release(dst);
786 goto out;
787 }
788
789 if (hlimit < 0) {
790 if (ipv6_addr_is_multicast(&fl.fl6_dst))
791 hlimit = np->mcast_hops;
792 else
793 hlimit = np->hop_limit;
794 if (hlimit < 0)
795 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
796 if (hlimit < 0)
797 hlimit = ipv6_get_hoplimit(dst->dev);
798 }
799
800 if (msg->msg_flags&MSG_CONFIRM)
801 goto do_confirm;
802
803back_from_confirm:
804 if (inet->hdrincl) {
805 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
806 } else {
807 lock_sock(sk);
808 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
809 hlimit, opt, &fl, (struct rt6_info*)dst, msg->msg_flags);
810
811 if (err)
812 ip6_flush_pending_frames(sk);
813 else if (!(msg->msg_flags & MSG_MORE))
357b40a1 814 err = rawv6_push_pending_frames(sk, &fl, rp);
1da177e4
LT		 815 }
816done:
817 ip6_dst_store(sk, dst,
818 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
819 &np->daddr : NULL);
1da177e4
LT		 820
821 release_sock(sk);
822out:
823 fl6_sock_release(flowlabel);
824 return err<0?err:len;
825do_confirm:
826 dst_confirm(dst);
827 if (!(msg->msg_flags & MSG_PROBE) || len)
828 goto back_from_confirm;
829 err = 0;
830 goto done;
831}
832
833static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
834 char __user *optval, int optlen)
835{
836 switch (optname) {
837 case ICMPV6_FILTER:
838 if (optlen > sizeof(struct icmp6_filter))
839 optlen = sizeof(struct icmp6_filter);
840 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
841 return -EFAULT;
842 return 0;
843 default:
844 return -ENOPROTOOPT;
845 };
846
847 return 0;
848}
849
850static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
851 char __user *optval, int __user *optlen)
852{
853 int len;
854
855 switch (optname) {
856 case ICMPV6_FILTER:
857 if (get_user(len, optlen))
858 return -EFAULT;
859 if (len < 0)
860 return -EINVAL;
861 if (len > sizeof(struct icmp6_filter))
862 len = sizeof(struct icmp6_filter);
863 if (put_user(len, optlen))
864 return -EFAULT;
865 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
866 return -EFAULT;
867 return 0;
868 default:
869 return -ENOPROTOOPT;
870 };
871
872 return 0;
873}
874
875
876static int rawv6_setsockopt(struct sock *sk, int level, int optname,
877 char __user *optval, int optlen)
878{
879 struct raw6_sock *rp = raw6_sk(sk);
880 int val;
881
882 switch(level) {
883 case SOL_RAW:
884 break;
885
886 case SOL_ICMPV6:
887 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
888 return -EOPNOTSUPP;
889 return rawv6_seticmpfilter(sk, level, optname, optval,
890 optlen);
891 case SOL_IPV6:
892 if (optname == IPV6_CHECKSUM)
893 break;
894 default:
895 return ipv6_setsockopt(sk, level, optname, optval,
896 optlen);
897 };
898
899 if (get_user(val, (int __user *)optval))
900 return -EFAULT;
901
902 switch (optname) {
903 case IPV6_CHECKSUM:
904 /* You may get strange result with a positive odd offset;
905 RFC2292bis agrees with me. */
906 if (val > 0 && (val&1))
907 return(-EINVAL);
908 if (val < 0) {
909 rp->checksum = 0;
910 } else {
911 rp->checksum = 1;
912 rp->offset = val;
913 }
914
915 return 0;
916 break;
917
918 default:
919 return(-ENOPROTOOPT);
920 }
921}
922
923static int rawv6_getsockopt(struct sock *sk, int level, int optname,
924 char __user *optval, int __user *optlen)
925{
926 struct raw6_sock *rp = raw6_sk(sk);
927 int val, len;
928
929 switch(level) {
930 case SOL_RAW:
931 break;
932
933 case SOL_ICMPV6:
934 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
935 return -EOPNOTSUPP;
936 return rawv6_geticmpfilter(sk, level, optname, optval,
937 optlen);
938 case SOL_IPV6:
939 if (optname == IPV6_CHECKSUM)
940 break;
941 default:
942 return ipv6_getsockopt(sk, level, optname, optval,
943 optlen);
944 };
945
946 if (get_user(len,optlen))
947 return -EFAULT;
948
949 switch (optname) {
950 case IPV6_CHECKSUM:
951 if (rp->checksum == 0)
952 val = -1;
953 else
954 val = rp->offset;
955 break;
956
957 default:
958 return -ENOPROTOOPT;
959 }
960
961 len = min_t(unsigned int, sizeof(int), len);
962
963 if (put_user(len, optlen))
964 return -EFAULT;
965 if (copy_to_user(optval,&val,len))
966 return -EFAULT;
967 return 0;
968}
969
970static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
971{
972 switch(cmd) {
973 case SIOCOUTQ:
974 {
975 int amount = atomic_read(&sk->sk_wmem_alloc);
976 return put_user(amount, (int __user *)arg);
977 }
978 case SIOCINQ:
979 {
980 struct sk_buff *skb;
981 int amount = 0;
982
e0f9f858 983 spin_lock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 984 skb = skb_peek(&sk->sk_receive_queue);
985 if (skb != NULL)
986 amount = skb->tail - skb->h.raw;
e0f9f858 987 spin_unlock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 988 return put_user(amount, (int __user *)arg);
989 }
990
991 default:
992 return -ENOIOCTLCMD;
993 }
994}
995
996static void rawv6_close(struct sock *sk, long timeout)
997{
998 if (inet_sk(sk)->num == IPPROTO_RAW)
999 ip6_ra_control(sk, -1, NULL);
1000
1001 sk_common_release(sk);
1002}
1003
1004static int rawv6_init_sk(struct sock *sk)
1005{
1006 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1007 struct raw6_sock *rp = raw6_sk(sk);
1008 rp->checksum = 1;
1009 rp->offset = 2;
1010 }
1011 return(0);
1012}
1013
1014struct proto rawv6_prot = {
1015 .name = "RAWv6",
1016 .owner = THIS_MODULE,
1017 .close = rawv6_close,
1018 .connect = ip6_datagram_connect,
1019 .disconnect = udp_disconnect,
1020 .ioctl = rawv6_ioctl,
1021 .init = rawv6_init_sk,
1022 .destroy = inet6_destroy_sock,
1023 .setsockopt = rawv6_setsockopt,
1024 .getsockopt = rawv6_getsockopt,
1025 .sendmsg = rawv6_sendmsg,
1026 .recvmsg = rawv6_recvmsg,
1027 .bind = rawv6_bind,
1028 .backlog_rcv = rawv6_rcv_skb,
1029 .hash = raw_v6_hash,
1030 .unhash = raw_v6_unhash,
1031 .obj_size = sizeof(struct raw6_sock),
1032};
1033
1034#ifdef CONFIG_PROC_FS
1035struct raw6_iter_state {
1036 int bucket;
1037};
1038
1039#define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1040
1041static struct sock *raw6_get_first(struct seq_file *seq)
1042{
1043 struct sock *sk;
1044 struct hlist_node *node;
1045 struct raw6_iter_state* state = raw6_seq_private(seq);
1046
1047 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1048 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1049 if (sk->sk_family == PF_INET6)
1050 goto out;
1051 sk = NULL;
1052out:
1053 return sk;
1054}
1055
1056static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1057{
1058 struct raw6_iter_state* state = raw6_seq_private(seq);
1059
1060 do {
1061 sk = sk_next(sk);
1062try_again:
1063 ;
1064 } while (sk && sk->sk_family != PF_INET6);
1065
1066 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1067 sk = sk_head(&raw_v6_htable[state->bucket]);
1068 goto try_again;
1069 }
1070 return sk;
1071}
1072
1073static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1074{
1075 struct sock *sk = raw6_get_first(seq);
1076 if (sk)
1077 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1078 --pos;
1079 return pos ? NULL : sk;
1080}
1081
1082static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1083{
1084 read_lock(&raw_v6_lock);
1085 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1086}
1087
1088static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1089{
1090 struct sock *sk;
1091
1092 if (v == SEQ_START_TOKEN)
1093 sk = raw6_get_first(seq);
1094 else
1095 sk = raw6_get_next(seq, v);
1096 ++*pos;
1097 return sk;
1098}
1099
1100static void raw6_seq_stop(struct seq_file *seq, void *v)
1101{
1102 read_unlock(&raw_v6_lock);
1103}
1104
1105static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1106{
1107 struct ipv6_pinfo *np = inet6_sk(sp);
1108 struct in6_addr *dest, *src;
1109 __u16 destp, srcp;
1110
1111 dest = &np->daddr;
1112 src = &np->rcv_saddr;
1113 destp = 0;
1114 srcp = inet_sk(sp)->num;
1115 seq_printf(seq,
1116 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1117 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1118 i,
1119 src->s6_addr32[0], src->s6_addr32[1],
1120 src->s6_addr32[2], src->s6_addr32[3], srcp,
1121 dest->s6_addr32[0], dest->s6_addr32[1],
1122 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1123 sp->sk_state,
1124 atomic_read(&sp->sk_wmem_alloc),
1125 atomic_read(&sp->sk_rmem_alloc),
1126 0, 0L, 0,
1127 sock_i_uid(sp), 0,
1128 sock_i_ino(sp),
1129 atomic_read(&sp->sk_refcnt), sp);
1130}
1131
1132static int raw6_seq_show(struct seq_file *seq, void *v)
1133{
1134 if (v == SEQ_START_TOKEN)
1135 seq_printf(seq,
1136 " sl "
1137 "local_address "
1138 "remote_address "
1139 "st tx_queue rx_queue tr tm->when retrnsmt"
1140 " uid timeout inode\n");
1141 else
1142 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1143 return 0;
1144}
1145
1146static struct seq_operations raw6_seq_ops = {
1147 .start = raw6_seq_start,
1148 .next = raw6_seq_next,
1149 .stop = raw6_seq_stop,
1150 .show = raw6_seq_show,
1151};
1152
1153static int raw6_seq_open(struct inode *inode, struct file *file)
1154{
1155 struct seq_file *seq;
1156 int rc = -ENOMEM;
1157 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1158 if (!s)
1159 goto out;
1160 rc = seq_open(file, &raw6_seq_ops);
1161 if (rc)
1162 goto out_kfree;
1163 seq = file->private_data;
1164 seq->private = s;
1165 memset(s, 0, sizeof(*s));
1166out:
1167 return rc;
1168out_kfree:
1169 kfree(s);
1170 goto out;
1171}
1172
1173static struct file_operations raw6_seq_fops = {
1174 .owner = THIS_MODULE,
1175 .open = raw6_seq_open,
1176 .read = seq_read,
1177 .llseek = seq_lseek,
1178 .release = seq_release_private,
1179};
1180
1181int __init raw6_proc_init(void)
1182{
1183 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1184 return -ENOMEM;
1185 return 0;
1186}
1187
1188void raw6_proc_exit(void)
1189{
1190 proc_net_remove("raw6");
1191}
1192#endif /* CONFIG_PROC_FS */
ubuntu-zesty-kernel mirror