projects / GitHub / LineageOS / android_kernel_samsung_universal7580.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ipv4: sendto/hdrincl: don't use destination address found in header
[GitHub/LineageOS/android_kernel_samsung_universal7580.git] / net / ipv4 / raw.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * RAW - implementation of IP "raw" sockets.
7 *
8 * Authors: Ross Biro
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 *
11 * Fixes:
12 * Alan Cox : verify_area() fixed up
13 * Alan Cox : ICMP error handling
14 * Alan Cox : EMSGSIZE if you send too big a packet
15 * Alan Cox : Now uses generic datagrams and shared
16 * skbuff library. No more peek crashes,
17 * no more backlogs
18 * Alan Cox : Checks sk->broadcast.
19 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
20 * Alan Cox : Raw passes ip options too
21 * Alan Cox : Setsocketopt added
22 * Alan Cox : Fixed error return for broadcasts
23 * Alan Cox : Removed wake_up calls
24 * Alan Cox : Use ttl/tos
25 * Alan Cox : Cleaned up old debugging
26 * Alan Cox : Use new kernel side addresses
27 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
28 * Alan Cox : BSD style RAW socket demultiplexing.
29 * Alan Cox : Beginnings of mrouted support.
30 * Alan Cox : Added IP_HDRINCL option.
31 * Alan Cox : Skip broadcast check if BSDism set.
32 * David S. Miller : New socket lookup architecture.
33 *
34 * This program is free software; you can redistribute it and/or
35 * modify it under the terms of the GNU General Public License
36 * as published by the Free Software Foundation; either version
37 * 2 of the License, or (at your option) any later version.
38 */
39
40 #include <linux/types.h>
41 #include <linux/atomic.h>
42 #include <asm/byteorder.h>
43 #include <asm/current.h>
44 #include <asm/uaccess.h>
45 #include <asm/ioctls.h>
46 #include <linux/stddef.h>
47 #include <linux/slab.h>
48 #include <linux/errno.h>
49 #include <linux/aio.h>
50 #include <linux/kernel.h>
51 #include <linux/export.h>
52 #include <linux/spinlock.h>
53 #include <linux/sockios.h>
54 #include <linux/socket.h>
55 #include <linux/in.h>
56 #include <linux/mroute.h>
57 #include <linux/netdevice.h>
58 #include <linux/in_route.h>
59 #include <linux/route.h>
60 #include <linux/skbuff.h>
61 #include <net/net_namespace.h>
62 #include <net/dst.h>
63 #include <net/sock.h>
64 #include <linux/ip.h>
65 #include <linux/net.h>
66 #include <net/ip.h>
67 #include <net/icmp.h>
68 #include <net/udp.h>
69 #include <net/raw.h>
70 #include <net/snmp.h>
71 #include <net/tcp_states.h>
72 #include <net/inet_common.h>
73 #include <net/checksum.h>
74 #include <net/xfrm.h>
75 #include <linux/rtnetlink.h>
76 #include <linux/proc_fs.h>
77 #include <linux/seq_file.h>
78 #include <linux/netfilter.h>
79 #include <linux/netfilter_ipv4.h>
80 #include <linux/compat.h>
81
82 static struct raw_hashinfo raw_v4_hashinfo = {
83 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
84 };
85
86 void raw_hash_sk(struct sock *sk)
87 {
88 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
89 struct hlist_head *head;
90
91 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
92
93 write_lock_bh(&h->lock);
94 sk_add_node(sk, head);
95 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
96 write_unlock_bh(&h->lock);
97 }
98 EXPORT_SYMBOL_GPL(raw_hash_sk);
99
100 void raw_unhash_sk(struct sock *sk)
101 {
102 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
103
104 write_lock_bh(&h->lock);
105 if (sk_del_node_init(sk))
106 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
107 write_unlock_bh(&h->lock);
108 }
109 EXPORT_SYMBOL_GPL(raw_unhash_sk);
110
111 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
112 unsigned short num, __be32 raddr, __be32 laddr, int dif)
113 {
114 sk_for_each_from(sk) {
115 struct inet_sock *inet = inet_sk(sk);
116
117 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
118 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
119 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
120 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
121 goto found; /* gotcha */
122 }
123 sk = NULL;
124 found:
125 return sk;
126 }
127
128 /*
129 * 0 - deliver
130 * 1 - block
131 */
132 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
133 {
134 struct icmphdr _hdr;
135 const struct icmphdr *hdr;
136
137 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
138 sizeof(_hdr), &_hdr);
139 if (!hdr)
140 return 1;
141
142 if (hdr->type < 32) {
143 __u32 data = raw_sk(sk)->filter.data;
144
145 return ((1U << hdr->type) & data) != 0;
146 }
147
148 /* Do not block unknown ICMP types */
149 return 0;
150 }
151
152 /* IP input processing comes here for RAW socket delivery.
153 * Caller owns SKB, so we must make clones.
154 *
155 * RFC 1122: SHOULD pass TOS value up to the transport layer.
156 * -> It does. And not only TOS, but all IP header.
157 */
158 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
159 {
160 struct sock *sk;
161 struct hlist_head *head;
162 int delivered = 0;
163 struct net *net;
164
165 read_lock(&raw_v4_hashinfo.lock);
166 head = &raw_v4_hashinfo.ht[hash];
167 if (hlist_empty(head))
168 goto out;
169
170 net = dev_net(skb->dev);
171 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
172 iph->saddr, iph->daddr,
173 skb->dev->ifindex);
174
175 while (sk) {
176 delivered = 1;
177 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
178 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
179
180 /* Not releasing hash table! */
181 if (clone)
182 raw_rcv(sk, clone);
183 }
184 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
185 iph->saddr, iph->daddr,
186 skb->dev->ifindex);
187 }
188 out:
189 read_unlock(&raw_v4_hashinfo.lock);
190 return delivered;
191 }
192
193 int raw_local_deliver(struct sk_buff *skb, int protocol)
194 {
195 int hash;
196 struct sock *raw_sk;
197
198 hash = protocol & (RAW_HTABLE_SIZE - 1);
199 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
200
201 /* If there maybe a raw socket we must check - if not we
202 * don't care less
203 */
204 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
205 raw_sk = NULL;
206
207 return raw_sk != NULL;
208
209 }
210
211 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
212 {
213 struct inet_sock *inet = inet_sk(sk);
214 const int type = icmp_hdr(skb)->type;
215 const int code = icmp_hdr(skb)->code;
216 int err = 0;
217 int harderr = 0;
218
219 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
220 ipv4_sk_update_pmtu(skb, sk, info);
221 else if (type == ICMP_REDIRECT)
222 ipv4_sk_redirect(skb, sk);
223
224 /* Report error on raw socket, if:
225 1. User requested ip_recverr.
226 2. Socket is connected (otherwise the error indication
227 is useless without ip_recverr and error is hard.
228 */
229 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
230 return;
231
232 switch (type) {
233 default:
234 case ICMP_TIME_EXCEEDED:
235 err = EHOSTUNREACH;
236 break;
237 case ICMP_SOURCE_QUENCH:
238 return;
239 case ICMP_PARAMETERPROB:
240 err = EPROTO;
241 harderr = 1;
242 break;
243 case ICMP_DEST_UNREACH:
244 err = EHOSTUNREACH;
245 if (code > NR_ICMP_UNREACH)
246 break;
247 err = icmp_err_convert[code].errno;
248 harderr = icmp_err_convert[code].fatal;
249 if (code == ICMP_FRAG_NEEDED) {
250 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
251 err = EMSGSIZE;
252 }
253 }
254
255 if (inet->recverr) {
256 const struct iphdr *iph = (const struct iphdr *)skb->data;
257 u8 *payload = skb->data + (iph->ihl << 2);
258
259 if (inet->hdrincl)
260 payload = skb->data;
261 ip_icmp_error(sk, skb, err, 0, info, payload);
262 }
263
264 if (inet->recverr || harderr) {
265 sk->sk_err = err;
266 sk->sk_error_report(sk);
267 }
268 }
269
270 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
271 {
272 int hash;
273 struct sock *raw_sk;
274 const struct iphdr *iph;
275 struct net *net;
276
277 hash = protocol & (RAW_HTABLE_SIZE - 1);
278
279 read_lock(&raw_v4_hashinfo.lock);
280 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
281 if (raw_sk != NULL) {
282 iph = (const struct iphdr *)skb->data;
283 net = dev_net(skb->dev);
284
285 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
286 iph->daddr, iph->saddr,
287 skb->dev->ifindex)) != NULL) {
288 raw_err(raw_sk, skb, info);
289 raw_sk = sk_next(raw_sk);
290 iph = (const struct iphdr *)skb->data;
291 }
292 }
293 read_unlock(&raw_v4_hashinfo.lock);
294 }
295
296 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
297 {
298 /* Charge it to the socket. */
299
300 ipv4_pktinfo_prepare(skb);
301 if (sock_queue_rcv_skb(sk, skb) < 0) {
302 kfree_skb(skb);
303 return NET_RX_DROP;
304 }
305
306 return NET_RX_SUCCESS;
307 }
308
309 int raw_rcv(struct sock *sk, struct sk_buff *skb)
310 {
311 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
312 atomic_inc(&sk->sk_drops);
313 kfree_skb(skb);
314 return NET_RX_DROP;
315 }
316 nf_reset(skb);
317
318 skb_push(skb, skb->data - skb_network_header(skb));
319
320 raw_rcv_skb(sk, skb);
321 return 0;
322 }
323
324 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
325 void *from, size_t length,
326 struct rtable **rtp,
327 unsigned int flags)
328 {
329 struct inet_sock *inet = inet_sk(sk);
330 struct net *net = sock_net(sk);
331 struct iphdr *iph;
332 struct sk_buff *skb;
333 unsigned int iphlen;
334 int err;
335 struct rtable *rt = *rtp;
336 int hlen, tlen;
337
338 if (length > rt->dst.dev->mtu) {
339 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
340 rt->dst.dev->mtu);
341 return -EMSGSIZE;
342 }
343 if (length < sizeof(struct iphdr))
344 return -EINVAL;
345
346 if (flags&MSG_PROBE)
347 goto out;
348
349 hlen = LL_RESERVED_SPACE(rt->dst.dev);
350 tlen = rt->dst.dev->needed_tailroom;
351 skb = sock_alloc_send_skb(sk,
352 length + hlen + tlen + 15,
353 flags & MSG_DONTWAIT, &err);
354 if (skb == NULL)
355 goto error;
356 skb_reserve(skb, hlen);
357
358 skb->priority = sk->sk_priority;
359 skb->mark = sk->sk_mark;
360 skb_dst_set(skb, &rt->dst);
361 *rtp = NULL;
362
363 skb_reset_network_header(skb);
364 iph = ip_hdr(skb);
365 skb_put(skb, length);
366
367 skb->ip_summed = CHECKSUM_NONE;
368
369 skb->transport_header = skb->network_header;
370 err = -EFAULT;
371 if (memcpy_fromiovecend((void *)iph, from, 0, length))
372 goto error_free;
373
374 iphlen = iph->ihl * 4;
375
376 /*
377 * We don't want to modify the ip header, but we do need to
378 * be sure that it won't cause problems later along the network
379 * stack. Specifically we want to make sure that iph->ihl is a
380 * sane value. If ihl points beyond the length of the buffer passed
381 * in, reject the frame as invalid
382 */
383 err = -EINVAL;
384 if (iphlen > length)
385 goto error_free;
386
387 if (iphlen >= sizeof(*iph)) {
388 if (!iph->saddr)
389 iph->saddr = fl4->saddr;
390 iph->check = 0;
391 iph->tot_len = htons(length);
392 if (!iph->id)
393 ip_select_ident(skb, NULL);
394
395 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
396 }
397 if (iph->protocol == IPPROTO_ICMP)
398 icmp_out_count(net, ((struct icmphdr *)
399 skb_transport_header(skb))->type);
400
401 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
402 rt->dst.dev, dst_output);
403 if (err > 0)
404 err = net_xmit_errno(err);
405 if (err)
406 goto error;
407 out:
408 return 0;
409
410 error_free:
411 kfree_skb(skb);
412 error:
413 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
414 if (err == -ENOBUFS && !inet->recverr)
415 err = 0;
416 return err;
417 }
418
419 static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg)
420 {
421 struct iovec *iov;
422 u8 __user *type = NULL;
423 u8 __user *code = NULL;
424 int probed = 0;
425 unsigned int i;
426
427 if (!msg->msg_iov)
428 return 0;
429
430 for (i = 0; i < msg->msg_iovlen; i++) {
431 iov = &msg->msg_iov[i];
432 if (!iov)
433 continue;
434
435 switch (fl4->flowi4_proto) {
436 case IPPROTO_ICMP:
437 /* check if one-byte field is readable or not. */
438 if (iov->iov_base && iov->iov_len < 1)
439 break;
440
441 if (!type) {
442 type = iov->iov_base;
443 /* check if code field is readable or not. */
444 if (iov->iov_len > 1)
445 code = type + 1;
446 } else if (!code)
447 code = iov->iov_base;
448
449 if (type && code) {
450 if (get_user(fl4->fl4_icmp_type, type) ||
451 get_user(fl4->fl4_icmp_code, code))
452 return -EFAULT;
453 probed = 1;
454 }
455 break;
456 default:
457 probed = 1;
458 break;
459 }
460 if (probed)
461 break;
462 }
463 return 0;
464 }
465
466 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
467 size_t len)
468 {
469 struct inet_sock *inet = inet_sk(sk);
470 struct ipcm_cookie ipc;
471 struct rtable *rt = NULL;
472 struct flowi4 fl4;
473 int free = 0;
474 __be32 daddr;
475 __be32 saddr;
476 u8 tos;
477 int err;
478 struct ip_options_data opt_copy;
479
480 err = -EMSGSIZE;
481 if (len > 0xFFFF)
482 goto out;
483
484 /*
485 * Check the flags.
486 */
487
488 err = -EOPNOTSUPP;
489 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
490 goto out; /* compatibility */
491
492 /*
493 * Get and verify the address.
494 */
495
496 if (msg->msg_namelen) {
497 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
498 err = -EINVAL;
499 if (msg->msg_namelen < sizeof(*usin))
500 goto out;
501 if (usin->sin_family != AF_INET) {
502 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
503 __func__, current->comm);
504 err = -EAFNOSUPPORT;
505 if (usin->sin_family)
506 goto out;
507 }
508 daddr = usin->sin_addr.s_addr;
509 /* ANK: I did not forget to get protocol from port field.
510 * I just do not know, who uses this weirdness.
511 * IP_HDRINCL is much more convenient.
512 */
513 } else {
514 err = -EDESTADDRREQ;
515 if (sk->sk_state != TCP_ESTABLISHED)
516 goto out;
517 daddr = inet->inet_daddr;
518 }
519
520 ipc.addr = inet->inet_saddr;
521 ipc.opt = NULL;
522 ipc.tx_flags = 0;
523 ipc.oif = sk->sk_bound_dev_if;
524
525 if (msg->msg_controllen) {
526 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
527 if (err)
528 goto out;
529 if (ipc.opt)
530 free = 1;
531 }
532
533 saddr = ipc.addr;
534 ipc.addr = daddr;
535
536 if (!ipc.opt) {
537 struct ip_options_rcu *inet_opt;
538
539 rcu_read_lock();
540 inet_opt = rcu_dereference(inet->inet_opt);
541 if (inet_opt) {
542 memcpy(&opt_copy, inet_opt,
543 sizeof(*inet_opt) + inet_opt->opt.optlen);
544 ipc.opt = &opt_copy.opt;
545 }
546 rcu_read_unlock();
547 }
548
549 if (ipc.opt) {
550 err = -EINVAL;
551 /* Linux does not mangle headers on raw sockets,
552 * so that IP options + IP_HDRINCL is non-sense.
553 */
554 if (inet->hdrincl)
555 goto done;
556 if (ipc.opt->opt.srr) {
557 if (!daddr)
558 goto done;
559 daddr = ipc.opt->opt.faddr;
560 }
561 }
562 tos = RT_CONN_FLAGS(sk);
563 if (msg->msg_flags & MSG_DONTROUTE)
564 tos |= RTO_ONLINK;
565
566 if (ipv4_is_multicast(daddr)) {
567 if (!ipc.oif)
568 ipc.oif = inet->mc_index;
569 if (!saddr)
570 saddr = inet->mc_addr;
571 } else if (!ipc.oif)
572 ipc.oif = inet->uc_index;
573
574 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
575 RT_SCOPE_UNIVERSE,
576 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
577 inet_sk_flowi_flags(sk) | FLOWI_FLAG_CAN_SLEEP |
578 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
579 daddr, saddr, 0, 0);
580
581 if (!inet->hdrincl) {
582 err = raw_probe_proto_opt(&fl4, msg);
583 if (err)
584 goto done;
585 }
586
587 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
588 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
589 if (IS_ERR(rt)) {
590 err = PTR_ERR(rt);
591 rt = NULL;
592 goto done;
593 }
594
595 err = -EACCES;
596 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
597 goto done;
598
599 if (msg->msg_flags & MSG_CONFIRM)
600 goto do_confirm;
601 back_from_confirm:
602
603 if (inet->hdrincl)
604 err = raw_send_hdrinc(sk, &fl4, msg->msg_iov, len,
605 &rt, msg->msg_flags);
606
607 else {
608 if (!ipc.addr)
609 ipc.addr = fl4.daddr;
610 lock_sock(sk);
611 err = ip_append_data(sk, &fl4, ip_generic_getfrag,
612 msg->msg_iov, len, 0,
613 &ipc, &rt, msg->msg_flags);
614 if (err)
615 ip_flush_pending_frames(sk);
616 else if (!(msg->msg_flags & MSG_MORE)) {
617 err = ip_push_pending_frames(sk, &fl4);
618 if (err == -ENOBUFS && !inet->recverr)
619 err = 0;
620 }
621 release_sock(sk);
622 }
623 done:
624 if (free)
625 kfree(ipc.opt);
626 ip_rt_put(rt);
627
628 out:
629 if (err < 0)
630 return err;
631 return len;
632
633 do_confirm:
634 dst_confirm(&rt->dst);
635 if (!(msg->msg_flags & MSG_PROBE) || len)
636 goto back_from_confirm;
637 err = 0;
638 goto done;
639 }
640
641 static void raw_close(struct sock *sk, long timeout)
642 {
643 /*
644 * Raw sockets may have direct kernel references. Kill them.
645 */
646 ip_ra_control(sk, 0, NULL);
647
648 sk_common_release(sk);
649 }
650
651 static void raw_destroy(struct sock *sk)
652 {
653 lock_sock(sk);
654 ip_flush_pending_frames(sk);
655 release_sock(sk);
656 }
657
658 /* This gets rid of all the nasties in af_inet. -DaveM */
659 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
660 {
661 struct inet_sock *inet = inet_sk(sk);
662 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
663 int ret = -EINVAL;
664 int chk_addr_ret;
665
666 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
667 goto out;
668 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
669 ret = -EADDRNOTAVAIL;
670 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
671 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
672 goto out;
673 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
674 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
675 inet->inet_saddr = 0; /* Use device */
676 sk_dst_reset(sk);
677 ret = 0;
678 out: return ret;
679 }
680
681 /*
682 * This should be easy, if there is something there
683 * we return it, otherwise we block.
684 */
685
686 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
687 size_t len, int noblock, int flags, int *addr_len)
688 {
689 struct inet_sock *inet = inet_sk(sk);
690 size_t copied = 0;
691 int err = -EOPNOTSUPP;
692 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
693 struct sk_buff *skb;
694
695 if (flags & MSG_OOB)
696 goto out;
697
698 if (flags & MSG_ERRQUEUE) {
699 err = ip_recv_error(sk, msg, len, addr_len);
700 goto out;
701 }
702
703 skb = skb_recv_datagram(sk, flags, noblock, &err);
704 if (!skb)
705 goto out;
706
707 copied = skb->len;
708 if (len < copied) {
709 msg->msg_flags |= MSG_TRUNC;
710 copied = len;
711 }
712
713 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
714 if (err)
715 goto done;
716
717 sock_recv_ts_and_drops(msg, sk, skb);
718
719 /* Copy the address. */
720 if (sin) {
721 sin->sin_family = AF_INET;
722 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
723 sin->sin_port = 0;
724 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
725 *addr_len = sizeof(*sin);
726 }
727 if (inet->cmsg_flags)
728 ip_cmsg_recv(msg, skb);
729 if (flags & MSG_TRUNC)
730 copied = skb->len;
731 done:
732 skb_free_datagram(sk, skb);
733 out:
734 if (err)
735 return err;
736 return copied;
737 }
738
739 static int raw_init(struct sock *sk)
740 {
741 struct raw_sock *rp = raw_sk(sk);
742
743 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
744 memset(&rp->filter, 0, sizeof(rp->filter));
745 return 0;
746 }
747
748 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
749 {
750 if (optlen > sizeof(struct icmp_filter))
751 optlen = sizeof(struct icmp_filter);
752 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
753 return -EFAULT;
754 return 0;
755 }
756
757 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
758 {
759 int len, ret = -EFAULT;
760
761 if (get_user(len, optlen))
762 goto out;
763 ret = -EINVAL;
764 if (len < 0)
765 goto out;
766 if (len > sizeof(struct icmp_filter))
767 len = sizeof(struct icmp_filter);
768 ret = -EFAULT;
769 if (put_user(len, optlen) ||
770 copy_to_user(optval, &raw_sk(sk)->filter, len))
771 goto out;
772 ret = 0;
773 out: return ret;
774 }
775
776 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
777 char __user *optval, unsigned int optlen)
778 {
779 if (optname == ICMP_FILTER) {
780 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
781 return -EOPNOTSUPP;
782 else
783 return raw_seticmpfilter(sk, optval, optlen);
784 }
785 return -ENOPROTOOPT;
786 }
787
788 static int raw_setsockopt(struct sock *sk, int level, int optname,
789 char __user *optval, unsigned int optlen)
790 {
791 if (level != SOL_RAW)
792 return ip_setsockopt(sk, level, optname, optval, optlen);
793 return do_raw_setsockopt(sk, level, optname, optval, optlen);
794 }
795
796 #ifdef CONFIG_COMPAT
797 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
798 char __user *optval, unsigned int optlen)
799 {
800 if (level != SOL_RAW)
801 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
802 return do_raw_setsockopt(sk, level, optname, optval, optlen);
803 }
804 #endif
805
806 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
807 char __user *optval, int __user *optlen)
808 {
809 if (optname == ICMP_FILTER) {
810 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
811 return -EOPNOTSUPP;
812 else
813 return raw_geticmpfilter(sk, optval, optlen);
814 }
815 return -ENOPROTOOPT;
816 }
817
818 static int raw_getsockopt(struct sock *sk, int level, int optname,
819 char __user *optval, int __user *optlen)
820 {
821 if (level != SOL_RAW)
822 return ip_getsockopt(sk, level, optname, optval, optlen);
823 return do_raw_getsockopt(sk, level, optname, optval, optlen);
824 }
825
826 #ifdef CONFIG_COMPAT
827 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
828 char __user *optval, int __user *optlen)
829 {
830 if (level != SOL_RAW)
831 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
832 return do_raw_getsockopt(sk, level, optname, optval, optlen);
833 }
834 #endif
835
836 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
837 {
838 switch (cmd) {
839 case SIOCOUTQ: {
840 int amount = sk_wmem_alloc_get(sk);
841
842 return put_user(amount, (int __user *)arg);
843 }
844 case SIOCINQ: {
845 struct sk_buff *skb;
846 int amount = 0;
847
848 spin_lock_bh(&sk->sk_receive_queue.lock);
849 skb = skb_peek(&sk->sk_receive_queue);
850 if (skb != NULL)
851 amount = skb->len;
852 spin_unlock_bh(&sk->sk_receive_queue.lock);
853 return put_user(amount, (int __user *)arg);
854 }
855
856 default:
857 #ifdef CONFIG_IP_MROUTE
858 return ipmr_ioctl(sk, cmd, (void __user *)arg);
859 #else
860 return -ENOIOCTLCMD;
861 #endif
862 }
863 }
864
865 #ifdef CONFIG_COMPAT
866 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
867 {
868 switch (cmd) {
869 case SIOCOUTQ:
870 case SIOCINQ:
871 return -ENOIOCTLCMD;
872 default:
873 #ifdef CONFIG_IP_MROUTE
874 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
875 #else
876 return -ENOIOCTLCMD;
877 #endif
878 }
879 }
880 #endif
881
882 struct proto raw_prot = {
883 .name = "RAW",
884 .owner = THIS_MODULE,
885 .close = raw_close,
886 .destroy = raw_destroy,
887 .connect = ip4_datagram_connect,
888 .disconnect = udp_disconnect,
889 .ioctl = raw_ioctl,
890 .init = raw_init,
891 .setsockopt = raw_setsockopt,
892 .getsockopt = raw_getsockopt,
893 .sendmsg = raw_sendmsg,
894 .recvmsg = raw_recvmsg,
895 .bind = raw_bind,
896 .backlog_rcv = raw_rcv_skb,
897 .release_cb = ip4_datagram_release_cb,
898 .hash = raw_hash_sk,
899 .unhash = raw_unhash_sk,
900 .obj_size = sizeof(struct raw_sock),
901 .h.raw_hash = &raw_v4_hashinfo,
902 #ifdef CONFIG_COMPAT
903 .compat_setsockopt = compat_raw_setsockopt,
904 .compat_getsockopt = compat_raw_getsockopt,
905 .compat_ioctl = compat_raw_ioctl,
906 #endif
907 };
908
909 #ifdef CONFIG_PROC_FS
910 static struct sock *raw_get_first(struct seq_file *seq)
911 {
912 struct sock *sk;
913 struct raw_iter_state *state = raw_seq_private(seq);
914
915 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
916 ++state->bucket) {
917 sk_for_each(sk, &state->h->ht[state->bucket])
918 if (sock_net(sk) == seq_file_net(seq))
919 goto found;
920 }
921 sk = NULL;
922 found:
923 return sk;
924 }
925
926 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
927 {
928 struct raw_iter_state *state = raw_seq_private(seq);
929
930 do {
931 sk = sk_next(sk);
932 try_again:
933 ;
934 } while (sk && sock_net(sk) != seq_file_net(seq));
935
936 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
937 sk = sk_head(&state->h->ht[state->bucket]);
938 goto try_again;
939 }
940 return sk;
941 }
942
943 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
944 {
945 struct sock *sk = raw_get_first(seq);
946
947 if (sk)
948 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
949 --pos;
950 return pos ? NULL : sk;
951 }
952
953 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
954 {
955 struct raw_iter_state *state = raw_seq_private(seq);
956
957 read_lock(&state->h->lock);
958 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
959 }
960 EXPORT_SYMBOL_GPL(raw_seq_start);
961
962 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
963 {
964 struct sock *sk;
965
966 if (v == SEQ_START_TOKEN)
967 sk = raw_get_first(seq);
968 else
969 sk = raw_get_next(seq, v);
970 ++*pos;
971 return sk;
972 }
973 EXPORT_SYMBOL_GPL(raw_seq_next);
974
975 void raw_seq_stop(struct seq_file *seq, void *v)
976 {
977 struct raw_iter_state *state = raw_seq_private(seq);
978
979 read_unlock(&state->h->lock);
980 }
981 EXPORT_SYMBOL_GPL(raw_seq_stop);
982
983 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
984 {
985 struct inet_sock *inet = inet_sk(sp);
986 __be32 dest = inet->inet_daddr,
987 src = inet->inet_rcv_saddr;
988 __u16 destp = 0,
989 srcp = inet->inet_num;
990
991 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
992 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
993 i, src, srcp, dest, destp, sp->sk_state,
994 sk_wmem_alloc_get(sp),
995 sk_rmem_alloc_get(sp),
996 0, 0L, 0,
997 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
998 0, sock_i_ino(sp),
999 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1000 }
1001
1002 static int raw_seq_show(struct seq_file *seq, void *v)
1003 {
1004 if (v == SEQ_START_TOKEN)
1005 seq_printf(seq, " sl local_address rem_address st tx_queue "
1006 "rx_queue tr tm->when retrnsmt uid timeout "
1007 "inode ref pointer drops\n");
1008 else
1009 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1010 return 0;
1011 }
1012
1013 static const struct seq_operations raw_seq_ops = {
1014 .start = raw_seq_start,
1015 .next = raw_seq_next,
1016 .stop = raw_seq_stop,
1017 .show = raw_seq_show,
1018 };
1019
1020 int raw_seq_open(struct inode *ino, struct file *file,
1021 struct raw_hashinfo *h, const struct seq_operations *ops)
1022 {
1023 int err;
1024 struct raw_iter_state *i;
1025
1026 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1027 if (err < 0)
1028 return err;
1029
1030 i = raw_seq_private((struct seq_file *)file->private_data);
1031 i->h = h;
1032 return 0;
1033 }
1034 EXPORT_SYMBOL_GPL(raw_seq_open);
1035
1036 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1037 {
1038 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1039 }
1040
1041 static const struct file_operations raw_seq_fops = {
1042 .owner = THIS_MODULE,
1043 .open = raw_v4_seq_open,
1044 .read = seq_read,
1045 .llseek = seq_lseek,
1046 .release = seq_release_net,
1047 };
1048
1049 static __net_init int raw_init_net(struct net *net)
1050 {
1051 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1052 return -ENOMEM;
1053
1054 return 0;
1055 }
1056
1057 static __net_exit void raw_exit_net(struct net *net)
1058 {
1059 remove_proc_entry("raw", net->proc_net);
1060 }
1061
1062 static __net_initdata struct pernet_operations raw_net_ops = {
1063 .init = raw_init_net,
1064 .exit = raw_exit_net,
1065 };
1066
1067 int __init raw_proc_init(void)
1068 {
1069 return register_pernet_subsys(&raw_net_ops);
1070 }
1071
1072 void __init raw_proc_exit(void)
1073 {
1074 unregister_pernet_subsys(&raw_net_ops);
1075 }
1076 #endif /* CONFIG_PROC_FS */