?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/speculation/mds: Fix documentation typo
[linux/fpc-iii.git] / net / ipv4 / raw.c
blob125c1eab3eaa6d894804c3aa8918aa7fcc736ca0
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_dst_set(skb, &rt->dst);
385 *rtp = NULL;
386
387 skb_reset_network_header(skb);
388 iph = ip_hdr(skb);
389 skb_put(skb, length);
390
391 skb->ip_summed = CHECKSUM_NONE;
392
393 sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
394
395 if (flags & MSG_CONFIRM)
396 skb_set_dst_pending_confirm(skb, 1);
397
398 skb->transport_header = skb->network_header;
399 err = -EFAULT;
400 if (memcpy_from_msg(iph, msg, length))
401 goto error_free;
402
403 iphlen = iph->ihl * 4;
404
405 /*
406 * We don't want to modify the ip header, but we do need to
407 * be sure that it won't cause problems later along the network
408 * stack. Specifically we want to make sure that iph->ihl is a
409 * sane value. If ihl points beyond the length of the buffer passed
410 * in, reject the frame as invalid
411 */
412 err = -EINVAL;
413 if (iphlen > length)
414 goto error_free;
415
416 if (iphlen >= sizeof(*iph)) {
417 if (!iph->saddr)
418 iph->saddr = fl4->saddr;
419 iph->check = 0;
420 iph->tot_len = htons(length);
421 if (!iph->id)
422 ip_select_ident(net, skb, NULL);
423
424 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
425 skb->transport_header += iphlen;
426 if (iph->protocol == IPPROTO_ICMP &&
427 length >= iphlen + sizeof(struct icmphdr))
428 icmp_out_count(net, ((struct icmphdr *)
429 skb_transport_header(skb))->type);
430 }
431
432 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
433 net, sk, skb, NULL, rt->dst.dev,
434 dst_output);
435 if (err > 0)
436 err = net_xmit_errno(err);
437 if (err)
438 goto error;
439 out:
440 return 0;
441
442 error_free:
443 kfree_skb(skb);
444 error:
445 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
446 if (err == -ENOBUFS && !inet->recverr)
447 err = 0;
448 return err;
449 }
450
451 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
452 {
453 int err;
454
455 if (fl4->flowi4_proto != IPPROTO_ICMP)
456 return 0;
457
458 /* We only need the first two bytes. */
459 rfv->hlen = 2;
460
461 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
462 if (err)
463 return err;
464
465 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
466 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
467
468 return 0;
469 }
470
471 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
472 struct sk_buff *skb)
473 {
474 struct raw_frag_vec *rfv = from;
475
476 if (offset < rfv->hlen) {
477 int copy = min(rfv->hlen - offset, len);
478
479 if (skb->ip_summed == CHECKSUM_PARTIAL)
480 memcpy(to, rfv->hdr.c + offset, copy);
481 else
482 skb->csum = csum_block_add(
483 skb->csum,
484 csum_partial_copy_nocheck(rfv->hdr.c + offset,
485 to, copy, 0),
486 odd);
487
488 odd = 0;
489 offset += copy;
490 to += copy;
491 len -= copy;
492
493 if (!len)
494 return 0;
495 }
496
497 offset -= rfv->hlen;
498
499 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
500 }
501
502 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
503 {
504 struct inet_sock *inet = inet_sk(sk);
505 struct net *net = sock_net(sk);
506 struct ipcm_cookie ipc;
507 struct rtable *rt = NULL;
508 struct flowi4 fl4;
509 int free = 0;
510 __be32 daddr;
511 __be32 saddr;
512 u8 tos;
513 int err;
514 struct ip_options_data opt_copy;
515 struct raw_frag_vec rfv;
516 int hdrincl;
517
518 err = -EMSGSIZE;
519 if (len > 0xFFFF)
520 goto out;
521
522 /* hdrincl should be READ_ONCE(inet->hdrincl)
523 * but READ_ONCE() doesn't work with bit fields
524 */
525 hdrincl = inet->hdrincl;
526 /*
527 * Check the flags.
528 */
529
530 err = -EOPNOTSUPP;
531 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
532 goto out; /* compatibility */
533
534 /*
535 * Get and verify the address.
536 */
537
538 if (msg->msg_namelen) {
539 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
540 err = -EINVAL;
541 if (msg->msg_namelen < sizeof(*usin))
542 goto out;
543 if (usin->sin_family != AF_INET) {
544 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
545 __func__, current->comm);
546 err = -EAFNOSUPPORT;
547 if (usin->sin_family)
548 goto out;
549 }
550 daddr = usin->sin_addr.s_addr;
551 /* ANK: I did not forget to get protocol from port field.
552 * I just do not know, who uses this weirdness.
553 * IP_HDRINCL is much more convenient.
554 */
555 } else {
556 err = -EDESTADDRREQ;
557 if (sk->sk_state != TCP_ESTABLISHED)
558 goto out;
559 daddr = inet->inet_daddr;
560 }
561
562 ipc.sockc.tsflags = sk->sk_tsflags;
563 ipc.addr = inet->inet_saddr;
564 ipc.opt = NULL;
565 ipc.tx_flags = 0;
566 ipc.ttl = 0;
567 ipc.tos = -1;
568 ipc.oif = sk->sk_bound_dev_if;
569
570 if (msg->msg_controllen) {
571 err = ip_cmsg_send(sk, msg, &ipc, false);
572 if (unlikely(err)) {
573 kfree(ipc.opt);
574 goto out;
575 }
576 if (ipc.opt)
577 free = 1;
578 }
579
580 saddr = ipc.addr;
581 ipc.addr = daddr;
582
583 if (!ipc.opt) {
584 struct ip_options_rcu *inet_opt;
585
586 rcu_read_lock();
587 inet_opt = rcu_dereference(inet->inet_opt);
588 if (inet_opt) {
589 memcpy(&opt_copy, inet_opt,
590 sizeof(*inet_opt) + inet_opt->opt.optlen);
591 ipc.opt = &opt_copy.opt;
592 }
593 rcu_read_unlock();
594 }
595
596 if (ipc.opt) {
597 err = -EINVAL;
598 /* Linux does not mangle headers on raw sockets,
599 * so that IP options + IP_HDRINCL is non-sense.
600 */
601 if (hdrincl)
602 goto done;
603 if (ipc.opt->opt.srr) {
604 if (!daddr)
605 goto done;
606 daddr = ipc.opt->opt.faddr;
607 }
608 }
609 tos = get_rtconn_flags(&ipc, sk);
610 if (msg->msg_flags & MSG_DONTROUTE)
611 tos |= RTO_ONLINK;
612
613 if (ipv4_is_multicast(daddr)) {
614 if (!ipc.oif)
615 ipc.oif = inet->mc_index;
616 if (!saddr)
617 saddr = inet->mc_addr;
618 } else if (!ipc.oif)
619 ipc.oif = inet->uc_index;
620
621 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
622 RT_SCOPE_UNIVERSE,
623 hdrincl ? IPPROTO_RAW : sk->sk_protocol,
624 inet_sk_flowi_flags(sk) |
625 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
626 daddr, saddr, 0, 0, sk->sk_uid);
627
628 if (!hdrincl) {
629 rfv.msg = msg;
630 rfv.hlen = 0;
631
632 err = raw_probe_proto_opt(&rfv, &fl4);
633 if (err)
634 goto done;
635 }
636
637 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
638 rt = ip_route_output_flow(net, &fl4, sk);
639 if (IS_ERR(rt)) {
640 err = PTR_ERR(rt);
641 rt = NULL;
642 goto done;
643 }
644
645 err = -EACCES;
646 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
647 goto done;
648
649 if (msg->msg_flags & MSG_CONFIRM)
650 goto do_confirm;
651 back_from_confirm:
652
653 if (hdrincl)
654 err = raw_send_hdrinc(sk, &fl4, msg, len,
655 &rt, msg->msg_flags, &ipc.sockc);
656
657 else {
658 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
659
660 if (!ipc.addr)
661 ipc.addr = fl4.daddr;
662 lock_sock(sk);
663 err = ip_append_data(sk, &fl4, raw_getfrag,
664 &rfv, len, 0,
665 &ipc, &rt, msg->msg_flags);
666 if (err)
667 ip_flush_pending_frames(sk);
668 else if (!(msg->msg_flags & MSG_MORE)) {
669 err = ip_push_pending_frames(sk, &fl4);
670 if (err == -ENOBUFS && !inet->recverr)
671 err = 0;
672 }
673 release_sock(sk);
674 }
675 done:
676 if (free)
677 kfree(ipc.opt);
678 ip_rt_put(rt);
679
680 out:
681 if (err < 0)
682 return err;
683 return len;
684
685 do_confirm:
686 if (msg->msg_flags & MSG_PROBE)
687 dst_confirm_neigh(&rt->dst, &fl4.daddr);
688 if (!(msg->msg_flags & MSG_PROBE) || len)
689 goto back_from_confirm;
690 err = 0;
691 goto done;
692 }
693
694 static void raw_close(struct sock *sk, long timeout)
695 {
696 /*
697 * Raw sockets may have direct kernel references. Kill them.
698 */
699 rtnl_lock();
700 ip_ra_control(sk, 0, NULL);
701 rtnl_unlock();
702
703 sk_common_release(sk);
704 }
705
706 static void raw_destroy(struct sock *sk)
707 {
708 lock_sock(sk);
709 ip_flush_pending_frames(sk);
710 release_sock(sk);
711 }
712
713 /* This gets rid of all the nasties in af_inet. -DaveM */
714 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
715 {
716 struct inet_sock *inet = inet_sk(sk);
717 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
718 u32 tb_id = RT_TABLE_LOCAL;
719 int ret = -EINVAL;
720 int chk_addr_ret;
721
722 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
723 goto out;
724
725 if (sk->sk_bound_dev_if)
726 tb_id = l3mdev_fib_table_by_index(sock_net(sk),
727 sk->sk_bound_dev_if) ? : tb_id;
728
729 chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
730 tb_id);
731
732 ret = -EADDRNOTAVAIL;
733 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
734 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
735 goto out;
736 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
737 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
738 inet->inet_saddr = 0; /* Use device */
739 sk_dst_reset(sk);
740 ret = 0;
741 out: return ret;
742 }
743
744 /*
745 * This should be easy, if there is something there
746 * we return it, otherwise we block.
747 */
748
749 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
750 int noblock, int flags, int *addr_len)
751 {
752 struct inet_sock *inet = inet_sk(sk);
753 size_t copied = 0;
754 int err = -EOPNOTSUPP;
755 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
756 struct sk_buff *skb;
757
758 if (flags & MSG_OOB)
759 goto out;
760
761 if (flags & MSG_ERRQUEUE) {
762 err = ip_recv_error(sk, msg, len, addr_len);
763 goto out;
764 }
765
766 skb = skb_recv_datagram(sk, flags, noblock, &err);
767 if (!skb)
768 goto out;
769
770 copied = skb->len;
771 if (len < copied) {
772 msg->msg_flags |= MSG_TRUNC;
773 copied = len;
774 }
775
776 err = skb_copy_datagram_msg(skb, 0, msg, copied);
777 if (err)
778 goto done;
779
780 sock_recv_ts_and_drops(msg, sk, skb);
781
782 /* Copy the address. */
783 if (sin) {
784 sin->sin_family = AF_INET;
785 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
786 sin->sin_port = 0;
787 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
788 *addr_len = sizeof(*sin);
789 }
790 if (inet->cmsg_flags)
791 ip_cmsg_recv(msg, skb);
792 if (flags & MSG_TRUNC)
793 copied = skb->len;
794 done:
795 skb_free_datagram(sk, skb);
796 out:
797 if (err)
798 return err;
799 return copied;
800 }
801
802 static int raw_init(struct sock *sk)
803 {
804 struct raw_sock *rp = raw_sk(sk);
805
806 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
807 memset(&rp->filter, 0, sizeof(rp->filter));
808 return 0;
809 }
810
811 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
812 {
813 if (optlen > sizeof(struct icmp_filter))
814 optlen = sizeof(struct icmp_filter);
815 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
816 return -EFAULT;
817 return 0;
818 }
819
820 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
821 {
822 int len, ret = -EFAULT;
823
824 if (get_user(len, optlen))
825 goto out;
826 ret = -EINVAL;
827 if (len < 0)
828 goto out;
829 if (len > sizeof(struct icmp_filter))
830 len = sizeof(struct icmp_filter);
831 ret = -EFAULT;
832 if (put_user(len, optlen) ||
833 copy_to_user(optval, &raw_sk(sk)->filter, len))
834 goto out;
835 ret = 0;
836 out: return ret;
837 }
838
839 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
840 char __user *optval, unsigned int optlen)
841 {
842 if (optname == ICMP_FILTER) {
843 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
844 return -EOPNOTSUPP;
845 else
846 return raw_seticmpfilter(sk, optval, optlen);
847 }
848 return -ENOPROTOOPT;
849 }
850
851 static int raw_setsockopt(struct sock *sk, int level, int optname,
852 char __user *optval, unsigned int optlen)
853 {
854 if (level != SOL_RAW)
855 return ip_setsockopt(sk, level, optname, optval, optlen);
856 return do_raw_setsockopt(sk, level, optname, optval, optlen);
857 }
858
859 #ifdef CONFIG_COMPAT
860 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
861 char __user *optval, unsigned int optlen)
862 {
863 if (level != SOL_RAW)
864 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
865 return do_raw_setsockopt(sk, level, optname, optval, optlen);
866 }
867 #endif
868
869 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
870 char __user *optval, int __user *optlen)
871 {
872 if (optname == ICMP_FILTER) {
873 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
874 return -EOPNOTSUPP;
875 else
876 return raw_geticmpfilter(sk, optval, optlen);
877 }
878 return -ENOPROTOOPT;
879 }
880
881 static int raw_getsockopt(struct sock *sk, int level, int optname,
882 char __user *optval, int __user *optlen)
883 {
884 if (level != SOL_RAW)
885 return ip_getsockopt(sk, level, optname, optval, optlen);
886 return do_raw_getsockopt(sk, level, optname, optval, optlen);
887 }
888
889 #ifdef CONFIG_COMPAT
890 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
891 char __user *optval, int __user *optlen)
892 {
893 if (level != SOL_RAW)
894 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
895 return do_raw_getsockopt(sk, level, optname, optval, optlen);
896 }
897 #endif
898
899 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
900 {
901 switch (cmd) {
902 case SIOCOUTQ: {
903 int amount = sk_wmem_alloc_get(sk);
904
905 return put_user(amount, (int __user *)arg);
906 }
907 case SIOCINQ: {
908 struct sk_buff *skb;
909 int amount = 0;
910
911 spin_lock_bh(&sk->sk_receive_queue.lock);
912 skb = skb_peek(&sk->sk_receive_queue);
913 if (skb)
914 amount = skb->len;
915 spin_unlock_bh(&sk->sk_receive_queue.lock);
916 return put_user(amount, (int __user *)arg);
917 }
918
919 default:
920 #ifdef CONFIG_IP_MROUTE
921 return ipmr_ioctl(sk, cmd, (void __user *)arg);
922 #else
923 return -ENOIOCTLCMD;
924 #endif
925 }
926 }
927
928 #ifdef CONFIG_COMPAT
929 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
930 {
931 switch (cmd) {
932 case SIOCOUTQ:
933 case SIOCINQ:
934 return -ENOIOCTLCMD;
935 default:
936 #ifdef CONFIG_IP_MROUTE
937 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
938 #else
939 return -ENOIOCTLCMD;
940 #endif
941 }
942 }
943 #endif
944
945 int raw_abort(struct sock *sk, int err)
946 {
947 lock_sock(sk);
948
949 sk->sk_err = err;
950 sk->sk_error_report(sk);
951 __udp_disconnect(sk, 0);
952
953 release_sock(sk);
954
955 return 0;
956 }
957 EXPORT_SYMBOL_GPL(raw_abort);
958
959 struct proto raw_prot = {
960 .name = "RAW",
961 .owner = THIS_MODULE,
962 .close = raw_close,
963 .destroy = raw_destroy,
964 .connect = ip4_datagram_connect,
965 .disconnect = __udp_disconnect,
966 .ioctl = raw_ioctl,
967 .init = raw_init,
968 .setsockopt = raw_setsockopt,
969 .getsockopt = raw_getsockopt,
970 .sendmsg = raw_sendmsg,
971 .recvmsg = raw_recvmsg,
972 .bind = raw_bind,
973 .backlog_rcv = raw_rcv_skb,
974 .release_cb = ip4_datagram_release_cb,
975 .hash = raw_hash_sk,
976 .unhash = raw_unhash_sk,
977 .obj_size = sizeof(struct raw_sock),
978 .h.raw_hash = &raw_v4_hashinfo,
979 #ifdef CONFIG_COMPAT
980 .compat_setsockopt = compat_raw_setsockopt,
981 .compat_getsockopt = compat_raw_getsockopt,
982 .compat_ioctl = compat_raw_ioctl,
983 #endif
984 .diag_destroy = raw_abort,
985 };
986
987 #ifdef CONFIG_PROC_FS
988 static struct sock *raw_get_first(struct seq_file *seq)
989 {
990 struct sock *sk;
991 struct raw_iter_state *state = raw_seq_private(seq);
992
993 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
994 ++state->bucket) {
995 sk_for_each(sk, &state->h->ht[state->bucket])
996 if (sock_net(sk) == seq_file_net(seq))
997 goto found;
998 }
999 sk = NULL;
1000 found:
1001 return sk;
1002 }
1003
1004 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
1005 {
1006 struct raw_iter_state *state = raw_seq_private(seq);
1007
1008 do {
1009 sk = sk_next(sk);
1010 try_again:
1011 ;
1012 } while (sk && sock_net(sk) != seq_file_net(seq));
1013
1014 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
1015 sk = sk_head(&state->h->ht[state->bucket]);
1016 goto try_again;
1017 }
1018 return sk;
1019 }
1020
1021 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
1022 {
1023 struct sock *sk = raw_get_first(seq);
1024
1025 if (sk)
1026 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
1027 --pos;
1028 return pos ? NULL : sk;
1029 }
1030
1031 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1032 {
1033 struct raw_iter_state *state = raw_seq_private(seq);
1034
1035 read_lock(&state->h->lock);
1036 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1037 }
1038 EXPORT_SYMBOL_GPL(raw_seq_start);
1039
1040 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1041 {
1042 struct sock *sk;
1043
1044 if (v == SEQ_START_TOKEN)
1045 sk = raw_get_first(seq);
1046 else
1047 sk = raw_get_next(seq, v);
1048 ++*pos;
1049 return sk;
1050 }
1051 EXPORT_SYMBOL_GPL(raw_seq_next);
1052
1053 void raw_seq_stop(struct seq_file *seq, void *v)
1054 {
1055 struct raw_iter_state *state = raw_seq_private(seq);
1056
1057 read_unlock(&state->h->lock);
1058 }
1059 EXPORT_SYMBOL_GPL(raw_seq_stop);
1060
1061 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1062 {
1063 struct inet_sock *inet = inet_sk(sp);
1064 __be32 dest = inet->inet_daddr,
1065 src = inet->inet_rcv_saddr;
1066 __u16 destp = 0,
1067 srcp = inet->inet_num;
1068
1069 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1070 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1071 i, src, srcp, dest, destp, sp->sk_state,
1072 sk_wmem_alloc_get(sp),
1073 sk_rmem_alloc_get(sp),
1074 0, 0L, 0,
1075 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1076 0, sock_i_ino(sp),
1077 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1078 }
1079
1080 static int raw_seq_show(struct seq_file *seq, void *v)
1081 {
1082 if (v == SEQ_START_TOKEN)
1083 seq_printf(seq, " sl local_address rem_address st tx_queue "
1084 "rx_queue tr tm->when retrnsmt uid timeout "
1085 "inode ref pointer drops\n");
1086 else
1087 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1088 return 0;
1089 }
1090
1091 static const struct seq_operations raw_seq_ops = {
1092 .start = raw_seq_start,
1093 .next = raw_seq_next,
1094 .stop = raw_seq_stop,
1095 .show = raw_seq_show,
1096 };
1097
1098 int raw_seq_open(struct inode *ino, struct file *file,
1099 struct raw_hashinfo *h, const struct seq_operations *ops)
1100 {
1101 int err;
1102 struct raw_iter_state *i;
1103
1104 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1105 if (err < 0)
1106 return err;
1107
1108 i = raw_seq_private((struct seq_file *)file->private_data);
1109 i->h = h;
1110 return 0;
1111 }
1112 EXPORT_SYMBOL_GPL(raw_seq_open);
1113
1114 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1115 {
1116 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1117 }
1118
1119 static const struct file_operations raw_seq_fops = {
1120 .owner = THIS_MODULE,
1121 .open = raw_v4_seq_open,
1122 .read = seq_read,
1123 .llseek = seq_lseek,
1124 .release = seq_release_net,
1125 };
1126
1127 static __net_init int raw_init_net(struct net *net)
1128 {
1129 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1130 return -ENOMEM;
1131
1132 return 0;
1133 }
1134
1135 static __net_exit void raw_exit_net(struct net *net)
1136 {
1137 remove_proc_entry("raw", net->proc_net);
1138 }
1139
1140 static __net_initdata struct pernet_operations raw_net_ops = {
1141 .init = raw_init_net,
1142 .exit = raw_exit_net,
1143 };
1144
1145 int __init raw_proc_init(void)
1146 {
1147 return register_pernet_subsys(&raw_net_ops);
1148 }
1149
1150 void __init raw_proc_exit(void)
1151 {
1152 unregister_pernet_subsys(&raw_net_ops);
1153 }
1154 #endif /* CONFIG_PROC_FS */
Linux with added FPC-III support