?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
rxrpc: Show some more information through /proc files
[linux/fpc-iii.git] / net / ipv4 / raw.c
blob33df4d76db2d948d620ffc809574b364ae24ad4b
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 <linux/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/kernel.h>
50 #include <linux/export.h>
51 #include <linux/spinlock.h>
52 #include <linux/sockios.h>
53 #include <linux/socket.h>
54 #include <linux/in.h>
55 #include <linux/mroute.h>
56 #include <linux/netdevice.h>
57 #include <linux/in_route.h>
58 #include <linux/route.h>
59 #include <linux/skbuff.h>
60 #include <linux/igmp.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 #include <linux/uio.h>
82
83 struct raw_frag_vec {
84 struct msghdr *msg;
85 union {
86 struct icmphdr icmph;
87 char c[1];
88 } hdr;
89 int hlen;
90 };
91
92 struct raw_hashinfo raw_v4_hashinfo = {
93 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
94 };
95 EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
96
97 int raw_hash_sk(struct sock *sk)
98 {
99 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
100 struct hlist_head *head;
101
102 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
103
104 write_lock_bh(&h->lock);
105 sk_add_node(sk, head);
106 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
107 write_unlock_bh(&h->lock);
108
109 return 0;
110 }
111 EXPORT_SYMBOL_GPL(raw_hash_sk);
112
113 void raw_unhash_sk(struct sock *sk)
114 {
115 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
116
117 write_lock_bh(&h->lock);
118 if (sk_del_node_init(sk))
119 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
120 write_unlock_bh(&h->lock);
121 }
122 EXPORT_SYMBOL_GPL(raw_unhash_sk);
123
124 struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
125 unsigned short num, __be32 raddr, __be32 laddr,
126 int dif, int sdif)
127 {
128 sk_for_each_from(sk) {
129 struct inet_sock *inet = inet_sk(sk);
130
131 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
132 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
133 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
134 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
135 sk->sk_bound_dev_if != sdif))
136 goto found; /* gotcha */
137 }
138 sk = NULL;
139 found:
140 return sk;
141 }
142 EXPORT_SYMBOL_GPL(__raw_v4_lookup);
143
144 /*
145 * 0 - deliver
146 * 1 - block
147 */
148 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
149 {
150 struct icmphdr _hdr;
151 const struct icmphdr *hdr;
152
153 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
154 sizeof(_hdr), &_hdr);
155 if (!hdr)
156 return 1;
157
158 if (hdr->type < 32) {
159 __u32 data = raw_sk(sk)->filter.data;
160
161 return ((1U << hdr->type) & data) != 0;
162 }
163
164 /* Do not block unknown ICMP types */
165 return 0;
166 }
167
168 /* IP input processing comes here for RAW socket delivery.
169 * Caller owns SKB, so we must make clones.
170 *
171 * RFC 1122: SHOULD pass TOS value up to the transport layer.
172 * -> It does. And not only TOS, but all IP header.
173 */
174 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
175 {
176 int sdif = inet_sdif(skb);
177 struct sock *sk;
178 struct hlist_head *head;
179 int delivered = 0;
180 struct net *net;
181
182 read_lock(&raw_v4_hashinfo.lock);
183 head = &raw_v4_hashinfo.ht[hash];
184 if (hlist_empty(head))
185 goto out;
186
187 net = dev_net(skb->dev);
188 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
189 iph->saddr, iph->daddr,
190 skb->dev->ifindex, sdif);
191
192 while (sk) {
193 delivered = 1;
194 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
195 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
196 skb->dev->ifindex, sdif)) {
197 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
198
199 /* Not releasing hash table! */
200 if (clone)
201 raw_rcv(sk, clone);
202 }
203 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
204 iph->saddr, iph->daddr,
205 skb->dev->ifindex, sdif);
206 }
207 out:
208 read_unlock(&raw_v4_hashinfo.lock);
209 return delivered;
210 }
211
212 int raw_local_deliver(struct sk_buff *skb, int protocol)
213 {
214 int hash;
215 struct sock *raw_sk;
216
217 hash = protocol & (RAW_HTABLE_SIZE - 1);
218 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
219
220 /* If there maybe a raw socket we must check - if not we
221 * don't care less
222 */
223 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
224 raw_sk = NULL;
225
226 return raw_sk != NULL;
227
228 }
229
230 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
231 {
232 struct inet_sock *inet = inet_sk(sk);
233 const int type = icmp_hdr(skb)->type;
234 const int code = icmp_hdr(skb)->code;
235 int err = 0;
236 int harderr = 0;
237
238 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
239 ipv4_sk_update_pmtu(skb, sk, info);
240 else if (type == ICMP_REDIRECT) {
241 ipv4_sk_redirect(skb, sk);
242 return;
243 }
244
245 /* Report error on raw socket, if:
246 1. User requested ip_recverr.
247 2. Socket is connected (otherwise the error indication
248 is useless without ip_recverr and error is hard.
249 */
250 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
251 return;
252
253 switch (type) {
254 default:
255 case ICMP_TIME_EXCEEDED:
256 err = EHOSTUNREACH;
257 break;
258 case ICMP_SOURCE_QUENCH:
259 return;
260 case ICMP_PARAMETERPROB:
261 err = EPROTO;
262 harderr = 1;
263 break;
264 case ICMP_DEST_UNREACH:
265 err = EHOSTUNREACH;
266 if (code > NR_ICMP_UNREACH)
267 break;
268 err = icmp_err_convert[code].errno;
269 harderr = icmp_err_convert[code].fatal;
270 if (code == ICMP_FRAG_NEEDED) {
271 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
272 err = EMSGSIZE;
273 }
274 }
275
276 if (inet->recverr) {
277 const struct iphdr *iph = (const struct iphdr *)skb->data;
278 u8 *payload = skb->data + (iph->ihl << 2);
279
280 if (inet->hdrincl)
281 payload = skb->data;
282 ip_icmp_error(sk, skb, err, 0, info, payload);
283 }
284
285 if (inet->recverr || harderr) {
286 sk->sk_err = err;
287 sk->sk_error_report(sk);
288 }
289 }
290
291 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
292 {
293 int hash;
294 struct sock *raw_sk;
295 const struct iphdr *iph;
296 struct net *net;
297
298 hash = protocol & (RAW_HTABLE_SIZE - 1);
299
300 read_lock(&raw_v4_hashinfo.lock);
301 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
302 if (raw_sk) {
303 int dif = skb->dev->ifindex;
304 int sdif = inet_sdif(skb);
305
306 iph = (const struct iphdr *)skb->data;
307 net = dev_net(skb->dev);
308
309 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
310 iph->daddr, iph->saddr,
311 dif, sdif)) != NULL) {
312 raw_err(raw_sk, skb, info);
313 raw_sk = sk_next(raw_sk);
314 iph = (const struct iphdr *)skb->data;
315 }
316 }
317 read_unlock(&raw_v4_hashinfo.lock);
318 }
319
320 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
321 {
322 /* Charge it to the socket. */
323
324 ipv4_pktinfo_prepare(sk, skb);
325 if (sock_queue_rcv_skb(sk, skb) < 0) {
326 kfree_skb(skb);
327 return NET_RX_DROP;
328 }
329
330 return NET_RX_SUCCESS;
331 }
332
333 int raw_rcv(struct sock *sk, struct sk_buff *skb)
334 {
335 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
336 atomic_inc(&sk->sk_drops);
337 kfree_skb(skb);
338 return NET_RX_DROP;
339 }
340 nf_reset(skb);
341
342 skb_push(skb, skb->data - skb_network_header(skb));
343
344 raw_rcv_skb(sk, skb);
345 return 0;
346 }
347
348 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
349 struct msghdr *msg, size_t length,
350 struct rtable **rtp, unsigned int flags,
351 const struct sockcm_cookie *sockc)
352 {
353 struct inet_sock *inet = inet_sk(sk);
354 struct net *net = sock_net(sk);
355 struct iphdr *iph;
356 struct sk_buff *skb;
357 unsigned int iphlen;
358 int err;
359 struct rtable *rt = *rtp;
360 int hlen, tlen;
361
362 if (length > rt->dst.dev->mtu) {
363 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
364 rt->dst.dev->mtu);
365 return -EMSGSIZE;
366 }
367 if (length < sizeof(struct iphdr))
368 return -EINVAL;
369
370 if (flags&MSG_PROBE)
371 goto out;
372
373 hlen = LL_RESERVED_SPACE(rt->dst.dev);
374 tlen = rt->dst.dev->needed_tailroom;
375 skb = sock_alloc_send_skb(sk,
376 length + hlen + tlen + 15,
377 flags & MSG_DONTWAIT, &err);
378 if (!skb)
379 goto error;
380 skb_reserve(skb, hlen);
381
382 skb->priority = sk->sk_priority;
383 skb->mark = sk->sk_mark;
384 skb->tstamp = sockc->transmit_time;
385 skb_dst_set(skb, &rt->dst);
386 *rtp = NULL;
387
388 skb_reset_network_header(skb);
389 iph = ip_hdr(skb);
390 skb_put(skb, length);
391
392 skb->ip_summed = CHECKSUM_NONE;
393
394 sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
395
396 if (flags & MSG_CONFIRM)
397 skb_set_dst_pending_confirm(skb, 1);
398
399 skb->transport_header = skb->network_header;
400 err = -EFAULT;
401 if (memcpy_from_msg(iph, msg, length))
402 goto error_free;
403
404 iphlen = iph->ihl * 4;
405
406 /*
407 * We don't want to modify the ip header, but we do need to
408 * be sure that it won't cause problems later along the network
409 * stack. Specifically we want to make sure that iph->ihl is a
410 * sane value. If ihl points beyond the length of the buffer passed
411 * in, reject the frame as invalid
412 */
413 err = -EINVAL;
414 if (iphlen > length)
415 goto error_free;
416
417 if (iphlen >= sizeof(*iph)) {
418 if (!iph->saddr)
419 iph->saddr = fl4->saddr;
420 iph->check = 0;
421 iph->tot_len = htons(length);
422 if (!iph->id)
423 ip_select_ident(net, skb, NULL);
424
425 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
426 skb->transport_header += iphlen;
427 if (iph->protocol == IPPROTO_ICMP &&
428 length >= iphlen + sizeof(struct icmphdr))
429 icmp_out_count(net, ((struct icmphdr *)
430 skb_transport_header(skb))->type);
431 }
432
433 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
434 net, sk, skb, NULL, rt->dst.dev,
435 dst_output);
436 if (err > 0)
437 err = net_xmit_errno(err);
438 if (err)
439 goto error;
440 out:
441 return 0;
442
443 error_free:
444 kfree_skb(skb);
445 error:
446 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
447 if (err == -ENOBUFS && !inet->recverr)
448 err = 0;
449 return err;
450 }
451
452 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
453 {
454 int err;
455
456 if (fl4->flowi4_proto != IPPROTO_ICMP)
457 return 0;
458
459 /* We only need the first two bytes. */
460 rfv->hlen = 2;
461
462 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
463 if (err)
464 return err;
465
466 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
467 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
468
469 return 0;
470 }
471
472 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
473 struct sk_buff *skb)
474 {
475 struct raw_frag_vec *rfv = from;
476
477 if (offset < rfv->hlen) {
478 int copy = min(rfv->hlen - offset, len);
479
480 if (skb->ip_summed == CHECKSUM_PARTIAL)
481 memcpy(to, rfv->hdr.c + offset, copy);
482 else
483 skb->csum = csum_block_add(
484 skb->csum,
485 csum_partial_copy_nocheck(rfv->hdr.c + offset,
486 to, copy, 0),
487 odd);
488
489 odd = 0;
490 offset += copy;
491 to += copy;
492 len -= copy;
493
494 if (!len)
495 return 0;
496 }
497
498 offset -= rfv->hlen;
499
500 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
501 }
502
503 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
504 {
505 struct inet_sock *inet = inet_sk(sk);
506 struct net *net = sock_net(sk);
507 struct ipcm_cookie ipc;
508 struct rtable *rt = NULL;
509 struct flowi4 fl4;
510 int free = 0;
511 __be32 daddr;
512 __be32 saddr;
513 u8 tos;
514 int err;
515 struct ip_options_data opt_copy;
516 struct raw_frag_vec rfv;
517 int hdrincl;
518
519 err = -EMSGSIZE;
520 if (len > 0xFFFF)
521 goto out;
522
523 /* hdrincl should be READ_ONCE(inet->hdrincl)
524 * but READ_ONCE() doesn't work with bit fields.
525 * Doing this indirectly yields the same result.
526 */
527 hdrincl = inet->hdrincl;
528 hdrincl = READ_ONCE(hdrincl);
529 /*
530 * Check the flags.
531 */
532
533 err = -EOPNOTSUPP;
534 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
535 goto out; /* compatibility */
536
537 /*
538 * Get and verify the address.
539 */
540
541 if (msg->msg_namelen) {
542 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
543 err = -EINVAL;
544 if (msg->msg_namelen < sizeof(*usin))
545 goto out;
546 if (usin->sin_family != AF_INET) {
547 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
548 __func__, current->comm);
549 err = -EAFNOSUPPORT;
550 if (usin->sin_family)
551 goto out;
552 }
553 daddr = usin->sin_addr.s_addr;
554 /* ANK: I did not forget to get protocol from port field.
555 * I just do not know, who uses this weirdness.
556 * IP_HDRINCL is much more convenient.
557 */
558 } else {
559 err = -EDESTADDRREQ;
560 if (sk->sk_state != TCP_ESTABLISHED)
561 goto out;
562 daddr = inet->inet_daddr;
563 }
564
565 ipcm_init_sk(&ipc, inet);
566
567 if (msg->msg_controllen) {
568 err = ip_cmsg_send(sk, msg, &ipc, false);
569 if (unlikely(err)) {
570 kfree(ipc.opt);
571 goto out;
572 }
573 if (ipc.opt)
574 free = 1;
575 }
576
577 saddr = ipc.addr;
578 ipc.addr = daddr;
579
580 if (!ipc.opt) {
581 struct ip_options_rcu *inet_opt;
582
583 rcu_read_lock();
584 inet_opt = rcu_dereference(inet->inet_opt);
585 if (inet_opt) {
586 memcpy(&opt_copy, inet_opt,
587 sizeof(*inet_opt) + inet_opt->opt.optlen);
588 ipc.opt = &opt_copy.opt;
589 }
590 rcu_read_unlock();
591 }
592
593 if (ipc.opt) {
594 err = -EINVAL;
595 /* Linux does not mangle headers on raw sockets,
596 * so that IP options + IP_HDRINCL is non-sense.
597 */
598 if (hdrincl)
599 goto done;
600 if (ipc.opt->opt.srr) {
601 if (!daddr)
602 goto done;
603 daddr = ipc.opt->opt.faddr;
604 }
605 }
606 tos = get_rtconn_flags(&ipc, sk);
607 if (msg->msg_flags & MSG_DONTROUTE)
608 tos |= RTO_ONLINK;
609
610 if (ipv4_is_multicast(daddr)) {
611 if (!ipc.oif)
612 ipc.oif = inet->mc_index;
613 if (!saddr)
614 saddr = inet->mc_addr;
615 } else if (!ipc.oif) {
616 ipc.oif = inet->uc_index;
617 } else if (ipv4_is_lbcast(daddr) && inet->uc_index) {
618 /* oif is set, packet is to local broadcast and
619 * and uc_index is set. oif is most likely set
620 * by sk_bound_dev_if. If uc_index != oif check if the
621 * oif is an L3 master and uc_index is an L3 slave.
622 * If so, we want to allow the send using the uc_index.
623 */
624 if (ipc.oif != inet->uc_index &&
625 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk),
626 inet->uc_index)) {
627 ipc.oif = inet->uc_index;
628 }
629 }
630
631 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
632 RT_SCOPE_UNIVERSE,
633 hdrincl ? IPPROTO_RAW : sk->sk_protocol,
634 inet_sk_flowi_flags(sk) |
635 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
636 daddr, saddr, 0, 0, sk->sk_uid);
637
638 if (!hdrincl) {
639 rfv.msg = msg;
640 rfv.hlen = 0;
641
642 err = raw_probe_proto_opt(&rfv, &fl4);
643 if (err)
644 goto done;
645 }
646
647 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
648 rt = ip_route_output_flow(net, &fl4, sk);
649 if (IS_ERR(rt)) {
650 err = PTR_ERR(rt);
651 rt = NULL;
652 goto done;
653 }
654
655 err = -EACCES;
656 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
657 goto done;
658
659 if (msg->msg_flags & MSG_CONFIRM)
660 goto do_confirm;
661 back_from_confirm:
662
663 if (hdrincl)
664 err = raw_send_hdrinc(sk, &fl4, msg, len,
665 &rt, msg->msg_flags, &ipc.sockc);
666
667 else {
668 if (!ipc.addr)
669 ipc.addr = fl4.daddr;
670 lock_sock(sk);
671 err = ip_append_data(sk, &fl4, raw_getfrag,
672 &rfv, len, 0,
673 &ipc, &rt, msg->msg_flags);
674 if (err)
675 ip_flush_pending_frames(sk);
676 else if (!(msg->msg_flags & MSG_MORE)) {
677 err = ip_push_pending_frames(sk, &fl4);
678 if (err == -ENOBUFS && !inet->recverr)
679 err = 0;
680 }
681 release_sock(sk);
682 }
683 done:
684 if (free)
685 kfree(ipc.opt);
686 ip_rt_put(rt);
687
688 out:
689 if (err < 0)
690 return err;
691 return len;
692
693 do_confirm:
694 if (msg->msg_flags & MSG_PROBE)
695 dst_confirm_neigh(&rt->dst, &fl4.daddr);
696 if (!(msg->msg_flags & MSG_PROBE) || len)
697 goto back_from_confirm;
698 err = 0;
699 goto done;
700 }
701
702 static void raw_close(struct sock *sk, long timeout)
703 {
704 /*
705 * Raw sockets may have direct kernel references. Kill them.
706 */
707 ip_ra_control(sk, 0, NULL);
708
709 sk_common_release(sk);
710 }
711
712 static void raw_destroy(struct sock *sk)
713 {
714 lock_sock(sk);
715 ip_flush_pending_frames(sk);
716 release_sock(sk);
717 }
718
719 /* This gets rid of all the nasties in af_inet. -DaveM */
720 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
721 {
722 struct inet_sock *inet = inet_sk(sk);
723 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
724 u32 tb_id = RT_TABLE_LOCAL;
725 int ret = -EINVAL;
726 int chk_addr_ret;
727
728 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
729 goto out;
730
731 if (sk->sk_bound_dev_if)
732 tb_id = l3mdev_fib_table_by_index(sock_net(sk),
733 sk->sk_bound_dev_if) ? : tb_id;
734
735 chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
736 tb_id);
737
738 ret = -EADDRNOTAVAIL;
739 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
740 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
741 goto out;
742 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
743 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
744 inet->inet_saddr = 0; /* Use device */
745 sk_dst_reset(sk);
746 ret = 0;
747 out: return ret;
748 }
749
750 /*
751 * This should be easy, if there is something there
752 * we return it, otherwise we block.
753 */
754
755 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
756 int noblock, int flags, int *addr_len)
757 {
758 struct inet_sock *inet = inet_sk(sk);
759 size_t copied = 0;
760 int err = -EOPNOTSUPP;
761 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
762 struct sk_buff *skb;
763
764 if (flags & MSG_OOB)
765 goto out;
766
767 if (flags & MSG_ERRQUEUE) {
768 err = ip_recv_error(sk, msg, len, addr_len);
769 goto out;
770 }
771
772 skb = skb_recv_datagram(sk, flags, noblock, &err);
773 if (!skb)
774 goto out;
775
776 copied = skb->len;
777 if (len < copied) {
778 msg->msg_flags |= MSG_TRUNC;
779 copied = len;
780 }
781
782 err = skb_copy_datagram_msg(skb, 0, msg, copied);
783 if (err)
784 goto done;
785
786 sock_recv_ts_and_drops(msg, sk, skb);
787
788 /* Copy the address. */
789 if (sin) {
790 sin->sin_family = AF_INET;
791 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
792 sin->sin_port = 0;
793 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
794 *addr_len = sizeof(*sin);
795 }
796 if (inet->cmsg_flags)
797 ip_cmsg_recv(msg, skb);
798 if (flags & MSG_TRUNC)
799 copied = skb->len;
800 done:
801 skb_free_datagram(sk, skb);
802 out:
803 if (err)
804 return err;
805 return copied;
806 }
807
808 static int raw_init(struct sock *sk)
809 {
810 struct raw_sock *rp = raw_sk(sk);
811
812 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
813 memset(&rp->filter, 0, sizeof(rp->filter));
814 return 0;
815 }
816
817 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
818 {
819 if (optlen > sizeof(struct icmp_filter))
820 optlen = sizeof(struct icmp_filter);
821 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
822 return -EFAULT;
823 return 0;
824 }
825
826 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
827 {
828 int len, ret = -EFAULT;
829
830 if (get_user(len, optlen))
831 goto out;
832 ret = -EINVAL;
833 if (len < 0)
834 goto out;
835 if (len > sizeof(struct icmp_filter))
836 len = sizeof(struct icmp_filter);
837 ret = -EFAULT;
838 if (put_user(len, optlen) ||
839 copy_to_user(optval, &raw_sk(sk)->filter, len))
840 goto out;
841 ret = 0;
842 out: return ret;
843 }
844
845 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
846 char __user *optval, unsigned int optlen)
847 {
848 if (optname == ICMP_FILTER) {
849 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
850 return -EOPNOTSUPP;
851 else
852 return raw_seticmpfilter(sk, optval, optlen);
853 }
854 return -ENOPROTOOPT;
855 }
856
857 static int raw_setsockopt(struct sock *sk, int level, int optname,
858 char __user *optval, unsigned int optlen)
859 {
860 if (level != SOL_RAW)
861 return ip_setsockopt(sk, level, optname, optval, optlen);
862 return do_raw_setsockopt(sk, level, optname, optval, optlen);
863 }
864
865 #ifdef CONFIG_COMPAT
866 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
867 char __user *optval, unsigned int optlen)
868 {
869 if (level != SOL_RAW)
870 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
871 return do_raw_setsockopt(sk, level, optname, optval, optlen);
872 }
873 #endif
874
875 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
876 char __user *optval, int __user *optlen)
877 {
878 if (optname == ICMP_FILTER) {
879 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
880 return -EOPNOTSUPP;
881 else
882 return raw_geticmpfilter(sk, optval, optlen);
883 }
884 return -ENOPROTOOPT;
885 }
886
887 static int raw_getsockopt(struct sock *sk, int level, int optname,
888 char __user *optval, int __user *optlen)
889 {
890 if (level != SOL_RAW)
891 return ip_getsockopt(sk, level, optname, optval, optlen);
892 return do_raw_getsockopt(sk, level, optname, optval, optlen);
893 }
894
895 #ifdef CONFIG_COMPAT
896 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
897 char __user *optval, int __user *optlen)
898 {
899 if (level != SOL_RAW)
900 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
901 return do_raw_getsockopt(sk, level, optname, optval, optlen);
902 }
903 #endif
904
905 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
906 {
907 switch (cmd) {
908 case SIOCOUTQ: {
909 int amount = sk_wmem_alloc_get(sk);
910
911 return put_user(amount, (int __user *)arg);
912 }
913 case SIOCINQ: {
914 struct sk_buff *skb;
915 int amount = 0;
916
917 spin_lock_bh(&sk->sk_receive_queue.lock);
918 skb = skb_peek(&sk->sk_receive_queue);
919 if (skb)
920 amount = skb->len;
921 spin_unlock_bh(&sk->sk_receive_queue.lock);
922 return put_user(amount, (int __user *)arg);
923 }
924
925 default:
926 #ifdef CONFIG_IP_MROUTE
927 return ipmr_ioctl(sk, cmd, (void __user *)arg);
928 #else
929 return -ENOIOCTLCMD;
930 #endif
931 }
932 }
933
934 #ifdef CONFIG_COMPAT
935 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
936 {
937 switch (cmd) {
938 case SIOCOUTQ:
939 case SIOCINQ:
940 return -ENOIOCTLCMD;
941 default:
942 #ifdef CONFIG_IP_MROUTE
943 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
944 #else
945 return -ENOIOCTLCMD;
946 #endif
947 }
948 }
949 #endif
950
951 int raw_abort(struct sock *sk, int err)
952 {
953 lock_sock(sk);
954
955 sk->sk_err = err;
956 sk->sk_error_report(sk);
957 __udp_disconnect(sk, 0);
958
959 release_sock(sk);
960
961 return 0;
962 }
963 EXPORT_SYMBOL_GPL(raw_abort);
964
965 struct proto raw_prot = {
966 .name = "RAW",
967 .owner = THIS_MODULE,
968 .close = raw_close,
969 .destroy = raw_destroy,
970 .connect = ip4_datagram_connect,
971 .disconnect = __udp_disconnect,
972 .ioctl = raw_ioctl,
973 .init = raw_init,
974 .setsockopt = raw_setsockopt,
975 .getsockopt = raw_getsockopt,
976 .sendmsg = raw_sendmsg,
977 .recvmsg = raw_recvmsg,
978 .bind = raw_bind,
979 .backlog_rcv = raw_rcv_skb,
980 .release_cb = ip4_datagram_release_cb,
981 .hash = raw_hash_sk,
982 .unhash = raw_unhash_sk,
983 .obj_size = sizeof(struct raw_sock),
984 .useroffset = offsetof(struct raw_sock, filter),
985 .usersize = sizeof_field(struct raw_sock, filter),
986 .h.raw_hash = &raw_v4_hashinfo,
987 #ifdef CONFIG_COMPAT
988 .compat_setsockopt = compat_raw_setsockopt,
989 .compat_getsockopt = compat_raw_getsockopt,
990 .compat_ioctl = compat_raw_ioctl,
991 #endif
992 .diag_destroy = raw_abort,
993 };
994
995 #ifdef CONFIG_PROC_FS
996 static struct sock *raw_get_first(struct seq_file *seq)
997 {
998 struct sock *sk;
999 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
1000 struct raw_iter_state *state = raw_seq_private(seq);
1001
1002 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
1003 ++state->bucket) {
1004 sk_for_each(sk, &h->ht[state->bucket])
1005 if (sock_net(sk) == seq_file_net(seq))
1006 goto found;
1007 }
1008 sk = NULL;
1009 found:
1010 return sk;
1011 }
1012
1013 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
1014 {
1015 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
1016 struct raw_iter_state *state = raw_seq_private(seq);
1017
1018 do {
1019 sk = sk_next(sk);
1020 try_again:
1021 ;
1022 } while (sk && sock_net(sk) != seq_file_net(seq));
1023
1024 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
1025 sk = sk_head(&h->ht[state->bucket]);
1026 goto try_again;
1027 }
1028 return sk;
1029 }
1030
1031 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
1032 {
1033 struct sock *sk = raw_get_first(seq);
1034
1035 if (sk)
1036 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
1037 --pos;
1038 return pos ? NULL : sk;
1039 }
1040
1041 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1042 {
1043 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
1044
1045 read_lock(&h->lock);
1046 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1047 }
1048 EXPORT_SYMBOL_GPL(raw_seq_start);
1049
1050 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1051 {
1052 struct sock *sk;
1053
1054 if (v == SEQ_START_TOKEN)
1055 sk = raw_get_first(seq);
1056 else
1057 sk = raw_get_next(seq, v);
1058 ++*pos;
1059 return sk;
1060 }
1061 EXPORT_SYMBOL_GPL(raw_seq_next);
1062
1063 void raw_seq_stop(struct seq_file *seq, void *v)
1064 {
1065 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
1066
1067 read_unlock(&h->lock);
1068 }
1069 EXPORT_SYMBOL_GPL(raw_seq_stop);
1070
1071 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1072 {
1073 struct inet_sock *inet = inet_sk(sp);
1074 __be32 dest = inet->inet_daddr,
1075 src = inet->inet_rcv_saddr;
1076 __u16 destp = 0,
1077 srcp = inet->inet_num;
1078
1079 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1080 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1081 i, src, srcp, dest, destp, sp->sk_state,
1082 sk_wmem_alloc_get(sp),
1083 sk_rmem_alloc_get(sp),
1084 0, 0L, 0,
1085 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1086 0, sock_i_ino(sp),
1087 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1088 }
1089
1090 static int raw_seq_show(struct seq_file *seq, void *v)
1091 {
1092 if (v == SEQ_START_TOKEN)
1093 seq_printf(seq, " sl local_address rem_address st tx_queue "
1094 "rx_queue tr tm->when retrnsmt uid timeout "
1095 "inode ref pointer drops\n");
1096 else
1097 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1098 return 0;
1099 }
1100
1101 static const struct seq_operations raw_seq_ops = {
1102 .start = raw_seq_start,
1103 .next = raw_seq_next,
1104 .stop = raw_seq_stop,
1105 .show = raw_seq_show,
1106 };
1107
1108 static __net_init int raw_init_net(struct net *net)
1109 {
1110 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
1111 sizeof(struct raw_iter_state), &raw_v4_hashinfo))
1112 return -ENOMEM;
1113
1114 return 0;
1115 }
1116
1117 static __net_exit void raw_exit_net(struct net *net)
1118 {
1119 remove_proc_entry("raw", net->proc_net);
1120 }
1121
1122 static __net_initdata struct pernet_operations raw_net_ops = {
1123 .init = raw_init_net,
1124 .exit = raw_exit_net,
1125 };
1126
1127 int __init raw_proc_init(void)
1128 {
1129 return register_pernet_subsys(&raw_net_ops);
1130 }
1131
1132 void __init raw_proc_exit(void)
1133 {
1134 unregister_pernet_subsys(&raw_net_ops);
1135 }
1136 #endif /* CONFIG_PROC_FS */
Linux with added FPC-III support