?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sched/fair: Fix comments
[linux/fpc-iii.git] / net / ipv4 / raw.c
blob8119e1f66e036ad2a8372bf24dd943c7d9631d8e
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, int dif)
126 {
127 sk_for_each_from(sk) {
128 struct inet_sock *inet = inet_sk(sk);
129
130 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
131 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
132 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
133 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
134 goto found; /* gotcha */
135 }
136 sk = NULL;
137 found:
138 return sk;
139 }
140 EXPORT_SYMBOL_GPL(__raw_v4_lookup);
141
142 /*
143 * 0 - deliver
144 * 1 - block
145 */
146 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
147 {
148 struct icmphdr _hdr;
149 const struct icmphdr *hdr;
150
151 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
152 sizeof(_hdr), &_hdr);
153 if (!hdr)
154 return 1;
155
156 if (hdr->type < 32) {
157 __u32 data = raw_sk(sk)->filter.data;
158
159 return ((1U << hdr->type) & data) != 0;
160 }
161
162 /* Do not block unknown ICMP types */
163 return 0;
164 }
165
166 /* IP input processing comes here for RAW socket delivery.
167 * Caller owns SKB, so we must make clones.
168 *
169 * RFC 1122: SHOULD pass TOS value up to the transport layer.
170 * -> It does. And not only TOS, but all IP header.
171 */
172 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
173 {
174 struct sock *sk;
175 struct hlist_head *head;
176 int delivered = 0;
177 struct net *net;
178
179 read_lock(&raw_v4_hashinfo.lock);
180 head = &raw_v4_hashinfo.ht[hash];
181 if (hlist_empty(head))
182 goto out;
183
184 net = dev_net(skb->dev);
185 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
186 iph->saddr, iph->daddr,
187 skb->dev->ifindex);
188
189 while (sk) {
190 delivered = 1;
191 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
192 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
193 skb->dev->ifindex)) {
194 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
195
196 /* Not releasing hash table! */
197 if (clone)
198 raw_rcv(sk, clone);
199 }
200 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
201 iph->saddr, iph->daddr,
202 skb->dev->ifindex);
203 }
204 out:
205 read_unlock(&raw_v4_hashinfo.lock);
206 return delivered;
207 }
208
209 int raw_local_deliver(struct sk_buff *skb, int protocol)
210 {
211 int hash;
212 struct sock *raw_sk;
213
214 hash = protocol & (RAW_HTABLE_SIZE - 1);
215 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
216
217 /* If there maybe a raw socket we must check - if not we
218 * don't care less
219 */
220 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
221 raw_sk = NULL;
222
223 return raw_sk != NULL;
224
225 }
226
227 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
228 {
229 struct inet_sock *inet = inet_sk(sk);
230 const int type = icmp_hdr(skb)->type;
231 const int code = icmp_hdr(skb)->code;
232 int err = 0;
233 int harderr = 0;
234
235 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
236 ipv4_sk_update_pmtu(skb, sk, info);
237 else if (type == ICMP_REDIRECT) {
238 ipv4_sk_redirect(skb, sk);
239 return;
240 }
241
242 /* Report error on raw socket, if:
243 1. User requested ip_recverr.
244 2. Socket is connected (otherwise the error indication
245 is useless without ip_recverr and error is hard.
246 */
247 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
248 return;
249
250 switch (type) {
251 default:
252 case ICMP_TIME_EXCEEDED:
253 err = EHOSTUNREACH;
254 break;
255 case ICMP_SOURCE_QUENCH:
256 return;
257 case ICMP_PARAMETERPROB:
258 err = EPROTO;
259 harderr = 1;
260 break;
261 case ICMP_DEST_UNREACH:
262 err = EHOSTUNREACH;
263 if (code > NR_ICMP_UNREACH)
264 break;
265 err = icmp_err_convert[code].errno;
266 harderr = icmp_err_convert[code].fatal;
267 if (code == ICMP_FRAG_NEEDED) {
268 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
269 err = EMSGSIZE;
270 }
271 }
272
273 if (inet->recverr) {
274 const struct iphdr *iph = (const struct iphdr *)skb->data;
275 u8 *payload = skb->data + (iph->ihl << 2);
276
277 if (inet->hdrincl)
278 payload = skb->data;
279 ip_icmp_error(sk, skb, err, 0, info, payload);
280 }
281
282 if (inet->recverr || harderr) {
283 sk->sk_err = err;
284 sk->sk_error_report(sk);
285 }
286 }
287
288 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
289 {
290 int hash;
291 struct sock *raw_sk;
292 const struct iphdr *iph;
293 struct net *net;
294
295 hash = protocol & (RAW_HTABLE_SIZE - 1);
296
297 read_lock(&raw_v4_hashinfo.lock);
298 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
299 if (raw_sk) {
300 iph = (const struct iphdr *)skb->data;
301 net = dev_net(skb->dev);
302
303 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
304 iph->daddr, iph->saddr,
305 skb->dev->ifindex)) != NULL) {
306 raw_err(raw_sk, skb, info);
307 raw_sk = sk_next(raw_sk);
308 iph = (const struct iphdr *)skb->data;
309 }
310 }
311 read_unlock(&raw_v4_hashinfo.lock);
312 }
313
314 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
315 {
316 /* Charge it to the socket. */
317
318 ipv4_pktinfo_prepare(sk, skb);
319 if (sock_queue_rcv_skb(sk, skb) < 0) {
320 kfree_skb(skb);
321 return NET_RX_DROP;
322 }
323
324 return NET_RX_SUCCESS;
325 }
326
327 int raw_rcv(struct sock *sk, struct sk_buff *skb)
328 {
329 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
330 atomic_inc(&sk->sk_drops);
331 kfree_skb(skb);
332 return NET_RX_DROP;
333 }
334 nf_reset(skb);
335
336 skb_push(skb, skb->data - skb_network_header(skb));
337
338 raw_rcv_skb(sk, skb);
339 return 0;
340 }
341
342 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
343 struct msghdr *msg, size_t length,
344 struct rtable **rtp, unsigned int flags,
345 const struct sockcm_cookie *sockc)
346 {
347 struct inet_sock *inet = inet_sk(sk);
348 struct net *net = sock_net(sk);
349 struct iphdr *iph;
350 struct sk_buff *skb;
351 unsigned int iphlen;
352 int err;
353 struct rtable *rt = *rtp;
354 int hlen, tlen;
355
356 if (length > rt->dst.dev->mtu) {
357 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
358 rt->dst.dev->mtu);
359 return -EMSGSIZE;
360 }
361 if (flags&MSG_PROBE)
362 goto out;
363
364 hlen = LL_RESERVED_SPACE(rt->dst.dev);
365 tlen = rt->dst.dev->needed_tailroom;
366 skb = sock_alloc_send_skb(sk,
367 length + hlen + tlen + 15,
368 flags & MSG_DONTWAIT, &err);
369 if (!skb)
370 goto error;
371 skb_reserve(skb, hlen);
372
373 skb->priority = sk->sk_priority;
374 skb->mark = sk->sk_mark;
375 skb_dst_set(skb, &rt->dst);
376 *rtp = NULL;
377
378 skb_reset_network_header(skb);
379 iph = ip_hdr(skb);
380 skb_put(skb, length);
381
382 skb->ip_summed = CHECKSUM_NONE;
383
384 sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
385
386 if (flags & MSG_CONFIRM)
387 skb_set_dst_pending_confirm(skb, 1);
388
389 skb->transport_header = skb->network_header;
390 err = -EFAULT;
391 if (memcpy_from_msg(iph, msg, length))
392 goto error_free;
393
394 iphlen = iph->ihl * 4;
395
396 /*
397 * We don't want to modify the ip header, but we do need to
398 * be sure that it won't cause problems later along the network
399 * stack. Specifically we want to make sure that iph->ihl is a
400 * sane value. If ihl points beyond the length of the buffer passed
401 * in, reject the frame as invalid
402 */
403 err = -EINVAL;
404 if (iphlen > length)
405 goto error_free;
406
407 if (iphlen >= sizeof(*iph)) {
408 if (!iph->saddr)
409 iph->saddr = fl4->saddr;
410 iph->check = 0;
411 iph->tot_len = htons(length);
412 if (!iph->id)
413 ip_select_ident(net, skb, NULL);
414
415 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
416 skb->transport_header += iphlen;
417 if (iph->protocol == IPPROTO_ICMP &&
418 length >= iphlen + sizeof(struct icmphdr))
419 icmp_out_count(net, ((struct icmphdr *)
420 skb_transport_header(skb))->type);
421 }
422
423 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
424 net, sk, skb, NULL, rt->dst.dev,
425 dst_output);
426 if (err > 0)
427 err = net_xmit_errno(err);
428 if (err)
429 goto error;
430 out:
431 return 0;
432
433 error_free:
434 kfree_skb(skb);
435 error:
436 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
437 if (err == -ENOBUFS && !inet->recverr)
438 err = 0;
439 return err;
440 }
441
442 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
443 {
444 int err;
445
446 if (fl4->flowi4_proto != IPPROTO_ICMP)
447 return 0;
448
449 /* We only need the first two bytes. */
450 rfv->hlen = 2;
451
452 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
453 if (err)
454 return err;
455
456 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
457 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
458
459 return 0;
460 }
461
462 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
463 struct sk_buff *skb)
464 {
465 struct raw_frag_vec *rfv = from;
466
467 if (offset < rfv->hlen) {
468 int copy = min(rfv->hlen - offset, len);
469
470 if (skb->ip_summed == CHECKSUM_PARTIAL)
471 memcpy(to, rfv->hdr.c + offset, copy);
472 else
473 skb->csum = csum_block_add(
474 skb->csum,
475 csum_partial_copy_nocheck(rfv->hdr.c + offset,
476 to, copy, 0),
477 odd);
478
479 odd = 0;
480 offset += copy;
481 to += copy;
482 len -= copy;
483
484 if (!len)
485 return 0;
486 }
487
488 offset -= rfv->hlen;
489
490 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
491 }
492
493 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
494 {
495 struct inet_sock *inet = inet_sk(sk);
496 struct net *net = sock_net(sk);
497 struct ipcm_cookie ipc;
498 struct rtable *rt = NULL;
499 struct flowi4 fl4;
500 int free = 0;
501 __be32 daddr;
502 __be32 saddr;
503 u8 tos;
504 int err;
505 struct ip_options_data opt_copy;
506 struct raw_frag_vec rfv;
507
508 err = -EMSGSIZE;
509 if (len > 0xFFFF)
510 goto out;
511
512 /*
513 * Check the flags.
514 */
515
516 err = -EOPNOTSUPP;
517 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
518 goto out; /* compatibility */
519
520 /*
521 * Get and verify the address.
522 */
523
524 if (msg->msg_namelen) {
525 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
526 err = -EINVAL;
527 if (msg->msg_namelen < sizeof(*usin))
528 goto out;
529 if (usin->sin_family != AF_INET) {
530 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
531 __func__, current->comm);
532 err = -EAFNOSUPPORT;
533 if (usin->sin_family)
534 goto out;
535 }
536 daddr = usin->sin_addr.s_addr;
537 /* ANK: I did not forget to get protocol from port field.
538 * I just do not know, who uses this weirdness.
539 * IP_HDRINCL is much more convenient.
540 */
541 } else {
542 err = -EDESTADDRREQ;
543 if (sk->sk_state != TCP_ESTABLISHED)
544 goto out;
545 daddr = inet->inet_daddr;
546 }
547
548 ipc.sockc.tsflags = sk->sk_tsflags;
549 ipc.addr = inet->inet_saddr;
550 ipc.opt = NULL;
551 ipc.tx_flags = 0;
552 ipc.ttl = 0;
553 ipc.tos = -1;
554 ipc.oif = sk->sk_bound_dev_if;
555
556 if (msg->msg_controllen) {
557 err = ip_cmsg_send(sk, msg, &ipc, false);
558 if (unlikely(err)) {
559 kfree(ipc.opt);
560 goto out;
561 }
562 if (ipc.opt)
563 free = 1;
564 }
565
566 saddr = ipc.addr;
567 ipc.addr = daddr;
568
569 if (!ipc.opt) {
570 struct ip_options_rcu *inet_opt;
571
572 rcu_read_lock();
573 inet_opt = rcu_dereference(inet->inet_opt);
574 if (inet_opt) {
575 memcpy(&opt_copy, inet_opt,
576 sizeof(*inet_opt) + inet_opt->opt.optlen);
577 ipc.opt = &opt_copy.opt;
578 }
579 rcu_read_unlock();
580 }
581
582 if (ipc.opt) {
583 err = -EINVAL;
584 /* Linux does not mangle headers on raw sockets,
585 * so that IP options + IP_HDRINCL is non-sense.
586 */
587 if (inet->hdrincl)
588 goto done;
589 if (ipc.opt->opt.srr) {
590 if (!daddr)
591 goto done;
592 daddr = ipc.opt->opt.faddr;
593 }
594 }
595 tos = get_rtconn_flags(&ipc, sk);
596 if (msg->msg_flags & MSG_DONTROUTE)
597 tos |= RTO_ONLINK;
598
599 if (ipv4_is_multicast(daddr)) {
600 if (!ipc.oif)
601 ipc.oif = inet->mc_index;
602 if (!saddr)
603 saddr = inet->mc_addr;
604 } else if (!ipc.oif)
605 ipc.oif = inet->uc_index;
606
607 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
608 RT_SCOPE_UNIVERSE,
609 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
610 inet_sk_flowi_flags(sk) |
611 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
612 daddr, saddr, 0, 0, sk->sk_uid);
613
614 if (!inet->hdrincl) {
615 rfv.msg = msg;
616 rfv.hlen = 0;
617
618 err = raw_probe_proto_opt(&rfv, &fl4);
619 if (err)
620 goto done;
621 }
622
623 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
624 rt = ip_route_output_flow(net, &fl4, sk);
625 if (IS_ERR(rt)) {
626 err = PTR_ERR(rt);
627 rt = NULL;
628 goto done;
629 }
630
631 err = -EACCES;
632 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
633 goto done;
634
635 if (msg->msg_flags & MSG_CONFIRM)
636 goto do_confirm;
637 back_from_confirm:
638
639 if (inet->hdrincl)
640 err = raw_send_hdrinc(sk, &fl4, msg, len,
641 &rt, msg->msg_flags, &ipc.sockc);
642
643 else {
644 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
645
646 if (!ipc.addr)
647 ipc.addr = fl4.daddr;
648 lock_sock(sk);
649 err = ip_append_data(sk, &fl4, raw_getfrag,
650 &rfv, len, 0,
651 &ipc, &rt, msg->msg_flags);
652 if (err)
653 ip_flush_pending_frames(sk);
654 else if (!(msg->msg_flags & MSG_MORE)) {
655 err = ip_push_pending_frames(sk, &fl4);
656 if (err == -ENOBUFS && !inet->recverr)
657 err = 0;
658 }
659 release_sock(sk);
660 }
661 done:
662 if (free)
663 kfree(ipc.opt);
664 ip_rt_put(rt);
665
666 out:
667 if (err < 0)
668 return err;
669 return len;
670
671 do_confirm:
672 if (msg->msg_flags & MSG_PROBE)
673 dst_confirm_neigh(&rt->dst, &fl4.daddr);
674 if (!(msg->msg_flags & MSG_PROBE) || len)
675 goto back_from_confirm;
676 err = 0;
677 goto done;
678 }
679
680 static void raw_close(struct sock *sk, long timeout)
681 {
682 /*
683 * Raw sockets may have direct kernel references. Kill them.
684 */
685 ip_ra_control(sk, 0, NULL);
686
687 sk_common_release(sk);
688 }
689
690 static void raw_destroy(struct sock *sk)
691 {
692 lock_sock(sk);
693 ip_flush_pending_frames(sk);
694 release_sock(sk);
695 }
696
697 /* This gets rid of all the nasties in af_inet. -DaveM */
698 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
699 {
700 struct inet_sock *inet = inet_sk(sk);
701 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
702 u32 tb_id = RT_TABLE_LOCAL;
703 int ret = -EINVAL;
704 int chk_addr_ret;
705
706 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
707 goto out;
708
709 if (sk->sk_bound_dev_if)
710 tb_id = l3mdev_fib_table_by_index(sock_net(sk),
711 sk->sk_bound_dev_if) ? : tb_id;
712
713 chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
714 tb_id);
715
716 ret = -EADDRNOTAVAIL;
717 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
718 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
719 goto out;
720 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
721 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
722 inet->inet_saddr = 0; /* Use device */
723 sk_dst_reset(sk);
724 ret = 0;
725 out: return ret;
726 }
727
728 /*
729 * This should be easy, if there is something there
730 * we return it, otherwise we block.
731 */
732
733 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
734 int noblock, int flags, int *addr_len)
735 {
736 struct inet_sock *inet = inet_sk(sk);
737 size_t copied = 0;
738 int err = -EOPNOTSUPP;
739 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
740 struct sk_buff *skb;
741
742 if (flags & MSG_OOB)
743 goto out;
744
745 if (flags & MSG_ERRQUEUE) {
746 err = ip_recv_error(sk, msg, len, addr_len);
747 goto out;
748 }
749
750 skb = skb_recv_datagram(sk, flags, noblock, &err);
751 if (!skb)
752 goto out;
753
754 copied = skb->len;
755 if (len < copied) {
756 msg->msg_flags |= MSG_TRUNC;
757 copied = len;
758 }
759
760 err = skb_copy_datagram_msg(skb, 0, msg, copied);
761 if (err)
762 goto done;
763
764 sock_recv_ts_and_drops(msg, sk, skb);
765
766 /* Copy the address. */
767 if (sin) {
768 sin->sin_family = AF_INET;
769 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
770 sin->sin_port = 0;
771 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
772 *addr_len = sizeof(*sin);
773 }
774 if (inet->cmsg_flags)
775 ip_cmsg_recv(msg, skb);
776 if (flags & MSG_TRUNC)
777 copied = skb->len;
778 done:
779 skb_free_datagram(sk, skb);
780 out:
781 if (err)
782 return err;
783 return copied;
784 }
785
786 static int raw_init(struct sock *sk)
787 {
788 struct raw_sock *rp = raw_sk(sk);
789
790 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
791 memset(&rp->filter, 0, sizeof(rp->filter));
792 return 0;
793 }
794
795 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
796 {
797 if (optlen > sizeof(struct icmp_filter))
798 optlen = sizeof(struct icmp_filter);
799 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
800 return -EFAULT;
801 return 0;
802 }
803
804 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
805 {
806 int len, ret = -EFAULT;
807
808 if (get_user(len, optlen))
809 goto out;
810 ret = -EINVAL;
811 if (len < 0)
812 goto out;
813 if (len > sizeof(struct icmp_filter))
814 len = sizeof(struct icmp_filter);
815 ret = -EFAULT;
816 if (put_user(len, optlen) ||
817 copy_to_user(optval, &raw_sk(sk)->filter, len))
818 goto out;
819 ret = 0;
820 out: return ret;
821 }
822
823 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
824 char __user *optval, unsigned int optlen)
825 {
826 if (optname == ICMP_FILTER) {
827 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
828 return -EOPNOTSUPP;
829 else
830 return raw_seticmpfilter(sk, optval, optlen);
831 }
832 return -ENOPROTOOPT;
833 }
834
835 static int raw_setsockopt(struct sock *sk, int level, int optname,
836 char __user *optval, unsigned int optlen)
837 {
838 if (level != SOL_RAW)
839 return ip_setsockopt(sk, level, optname, optval, optlen);
840 return do_raw_setsockopt(sk, level, optname, optval, optlen);
841 }
842
843 #ifdef CONFIG_COMPAT
844 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
845 char __user *optval, unsigned int optlen)
846 {
847 if (level != SOL_RAW)
848 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
849 return do_raw_setsockopt(sk, level, optname, optval, optlen);
850 }
851 #endif
852
853 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
854 char __user *optval, int __user *optlen)
855 {
856 if (optname == ICMP_FILTER) {
857 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
858 return -EOPNOTSUPP;
859 else
860 return raw_geticmpfilter(sk, optval, optlen);
861 }
862 return -ENOPROTOOPT;
863 }
864
865 static int raw_getsockopt(struct sock *sk, int level, int optname,
866 char __user *optval, int __user *optlen)
867 {
868 if (level != SOL_RAW)
869 return ip_getsockopt(sk, level, optname, optval, optlen);
870 return do_raw_getsockopt(sk, level, optname, optval, optlen);
871 }
872
873 #ifdef CONFIG_COMPAT
874 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
875 char __user *optval, int __user *optlen)
876 {
877 if (level != SOL_RAW)
878 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
879 return do_raw_getsockopt(sk, level, optname, optval, optlen);
880 }
881 #endif
882
883 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
884 {
885 switch (cmd) {
886 case SIOCOUTQ: {
887 int amount = sk_wmem_alloc_get(sk);
888
889 return put_user(amount, (int __user *)arg);
890 }
891 case SIOCINQ: {
892 struct sk_buff *skb;
893 int amount = 0;
894
895 spin_lock_bh(&sk->sk_receive_queue.lock);
896 skb = skb_peek(&sk->sk_receive_queue);
897 if (skb)
898 amount = skb->len;
899 spin_unlock_bh(&sk->sk_receive_queue.lock);
900 return put_user(amount, (int __user *)arg);
901 }
902
903 default:
904 #ifdef CONFIG_IP_MROUTE
905 return ipmr_ioctl(sk, cmd, (void __user *)arg);
906 #else
907 return -ENOIOCTLCMD;
908 #endif
909 }
910 }
911
912 #ifdef CONFIG_COMPAT
913 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
914 {
915 switch (cmd) {
916 case SIOCOUTQ:
917 case SIOCINQ:
918 return -ENOIOCTLCMD;
919 default:
920 #ifdef CONFIG_IP_MROUTE
921 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
922 #else
923 return -ENOIOCTLCMD;
924 #endif
925 }
926 }
927 #endif
928
929 int raw_abort(struct sock *sk, int err)
930 {
931 lock_sock(sk);
932
933 sk->sk_err = err;
934 sk->sk_error_report(sk);
935 __udp_disconnect(sk, 0);
936
937 release_sock(sk);
938
939 return 0;
940 }
941 EXPORT_SYMBOL_GPL(raw_abort);
942
943 struct proto raw_prot = {
944 .name = "RAW",
945 .owner = THIS_MODULE,
946 .close = raw_close,
947 .destroy = raw_destroy,
948 .connect = ip4_datagram_connect,
949 .disconnect = __udp_disconnect,
950 .ioctl = raw_ioctl,
951 .init = raw_init,
952 .setsockopt = raw_setsockopt,
953 .getsockopt = raw_getsockopt,
954 .sendmsg = raw_sendmsg,
955 .recvmsg = raw_recvmsg,
956 .bind = raw_bind,
957 .backlog_rcv = raw_rcv_skb,
958 .release_cb = ip4_datagram_release_cb,
959 .hash = raw_hash_sk,
960 .unhash = raw_unhash_sk,
961 .obj_size = sizeof(struct raw_sock),
962 .h.raw_hash = &raw_v4_hashinfo,
963 #ifdef CONFIG_COMPAT
964 .compat_setsockopt = compat_raw_setsockopt,
965 .compat_getsockopt = compat_raw_getsockopt,
966 .compat_ioctl = compat_raw_ioctl,
967 #endif
968 .diag_destroy = raw_abort,
969 };
970
971 #ifdef CONFIG_PROC_FS
972 static struct sock *raw_get_first(struct seq_file *seq)
973 {
974 struct sock *sk;
975 struct raw_iter_state *state = raw_seq_private(seq);
976
977 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
978 ++state->bucket) {
979 sk_for_each(sk, &state->h->ht[state->bucket])
980 if (sock_net(sk) == seq_file_net(seq))
981 goto found;
982 }
983 sk = NULL;
984 found:
985 return sk;
986 }
987
988 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
989 {
990 struct raw_iter_state *state = raw_seq_private(seq);
991
992 do {
993 sk = sk_next(sk);
994 try_again:
995 ;
996 } while (sk && sock_net(sk) != seq_file_net(seq));
997
998 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
999 sk = sk_head(&state->h->ht[state->bucket]);
1000 goto try_again;
1001 }
1002 return sk;
1003 }
1004
1005 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
1006 {
1007 struct sock *sk = raw_get_first(seq);
1008
1009 if (sk)
1010 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
1011 --pos;
1012 return pos ? NULL : sk;
1013 }
1014
1015 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1016 {
1017 struct raw_iter_state *state = raw_seq_private(seq);
1018
1019 read_lock(&state->h->lock);
1020 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1021 }
1022 EXPORT_SYMBOL_GPL(raw_seq_start);
1023
1024 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1025 {
1026 struct sock *sk;
1027
1028 if (v == SEQ_START_TOKEN)
1029 sk = raw_get_first(seq);
1030 else
1031 sk = raw_get_next(seq, v);
1032 ++*pos;
1033 return sk;
1034 }
1035 EXPORT_SYMBOL_GPL(raw_seq_next);
1036
1037 void raw_seq_stop(struct seq_file *seq, void *v)
1038 {
1039 struct raw_iter_state *state = raw_seq_private(seq);
1040
1041 read_unlock(&state->h->lock);
1042 }
1043 EXPORT_SYMBOL_GPL(raw_seq_stop);
1044
1045 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1046 {
1047 struct inet_sock *inet = inet_sk(sp);
1048 __be32 dest = inet->inet_daddr,
1049 src = inet->inet_rcv_saddr;
1050 __u16 destp = 0,
1051 srcp = inet->inet_num;
1052
1053 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1054 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1055 i, src, srcp, dest, destp, sp->sk_state,
1056 sk_wmem_alloc_get(sp),
1057 sk_rmem_alloc_get(sp),
1058 0, 0L, 0,
1059 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1060 0, sock_i_ino(sp),
1061 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1062 }
1063
1064 static int raw_seq_show(struct seq_file *seq, void *v)
1065 {
1066 if (v == SEQ_START_TOKEN)
1067 seq_printf(seq, " sl local_address rem_address st tx_queue "
1068 "rx_queue tr tm->when retrnsmt uid timeout "
1069 "inode ref pointer drops\n");
1070 else
1071 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1072 return 0;
1073 }
1074
1075 static const struct seq_operations raw_seq_ops = {
1076 .start = raw_seq_start,
1077 .next = raw_seq_next,
1078 .stop = raw_seq_stop,
1079 .show = raw_seq_show,
1080 };
1081
1082 int raw_seq_open(struct inode *ino, struct file *file,
1083 struct raw_hashinfo *h, const struct seq_operations *ops)
1084 {
1085 int err;
1086 struct raw_iter_state *i;
1087
1088 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1089 if (err < 0)
1090 return err;
1091
1092 i = raw_seq_private((struct seq_file *)file->private_data);
1093 i->h = h;
1094 return 0;
1095 }
1096 EXPORT_SYMBOL_GPL(raw_seq_open);
1097
1098 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1099 {
1100 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1101 }
1102
1103 static const struct file_operations raw_seq_fops = {
1104 .owner = THIS_MODULE,
1105 .open = raw_v4_seq_open,
1106 .read = seq_read,
1107 .llseek = seq_lseek,
1108 .release = seq_release_net,
1109 };
1110
1111 static __net_init int raw_init_net(struct net *net)
1112 {
1113 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1114 return -ENOMEM;
1115
1116 return 0;
1117 }
1118
1119 static __net_exit void raw_exit_net(struct net *net)
1120 {
1121 remove_proc_entry("raw", net->proc_net);
1122 }
1123
1124 static __net_initdata struct pernet_operations raw_net_ops = {
1125 .init = raw_init_net,
1126 .exit = raw_exit_net,
1127 };
1128
1129 int __init raw_proc_init(void)
1130 {
1131 return register_pernet_subsys(&raw_net_ops);
1132 }
1133
1134 void __init raw_proc_exit(void)
1135 {
1136 unregister_pernet_subsys(&raw_net_ops);
1137 }
1138 #endif /* CONFIG_PROC_FS */
Linux with added FPC-III support