1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * RAW - implementation of IP "raw" sockets.
8 *
9 * Authors: Ross Biro
10 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 *
12 * Fixes:
13 * Alan Cox : verify_area() fixed up
14 * Alan Cox : ICMP error handling
15 * Alan Cox : EMSGSIZE if you send too big a packet
16 * Alan Cox : Now uses generic datagrams and shared
17 * skbuff library. No more peek crashes,
18 * no more backlogs
19 * Alan Cox : Checks sk->broadcast.
20 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
21 * Alan Cox : Raw passes ip options too
22 * Alan Cox : Setsocketopt added
23 * Alan Cox : Fixed error return for broadcasts
24 * Alan Cox : Removed wake_up calls
25 * Alan Cox : Use ttl/tos
26 * Alan Cox : Cleaned up old debugging
27 * Alan Cox : Use new kernel side addresses
28 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
29 * Alan Cox : BSD style RAW socket demultiplexing.
30 * Alan Cox : Beginnings of mrouted support.
31 * Alan Cox : Added IP_HDRINCL option.
32 * Alan Cox : Skip broadcast check if BSDism set.
33 * David S. Miller : New socket lookup architecture.
34 */
35
36#include <linux/types.h>
37#include <linux/atomic.h>
38#include <asm/byteorder.h>
39#include <asm/current.h>
40#include <linux/uaccess.h>
41#include <asm/ioctls.h>
42#include <linux/stddef.h>
43#include <linux/slab.h>
44#include <linux/errno.h>
45#include <linux/kernel.h>
46#include <linux/export.h>
47#include <linux/spinlock.h>
48#include <linux/sockios.h>
49#include <linux/socket.h>
50#include <linux/in.h>
51#include <linux/mroute.h>
52#include <linux/netdevice.h>
53#include <linux/in_route.h>
54#include <linux/route.h>
55#include <linux/skbuff.h>
56#include <linux/igmp.h>
57#include <net/net_namespace.h>
58#include <net/dst.h>
59#include <net/sock.h>
60#include <linux/ip.h>
61#include <linux/net.h>
62#include <net/ip.h>
63#include <net/icmp.h>
64#include <net/udp.h>
65#include <net/raw.h>
66#include <net/snmp.h>
67#include <net/tcp_states.h>
68#include <net/inet_common.h>
69#include <net/checksum.h>
70#include <net/xfrm.h>
71#include <linux/rtnetlink.h>
72#include <linux/proc_fs.h>
73#include <linux/seq_file.h>
74#include <linux/netfilter.h>
75#include <linux/netfilter_ipv4.h>
76#include <linux/compat.h>
77#include <linux/uio.h>
78
79struct raw_frag_vec {
80 struct msghdr *msg;
81 union {
82 struct icmphdr icmph;
83 char c[1];
84 } hdr;
85 int hlen;
86};
87
88struct raw_hashinfo raw_v4_hashinfo;
89EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
90
91int raw_hash_sk(struct sock *sk)
92{
93 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
94 struct hlist_head *hlist;
95
96 hlist = &h->ht[raw_hashfunc(sock_net(sk), inet_sk(sk)->inet_num)];
97
98 spin_lock(lock: &h->lock);
99 sk_add_node_rcu(sk, list: hlist);
100 sock_set_flag(sk, flag: SOCK_RCU_FREE);
101 spin_unlock(lock: &h->lock);
102 sock_prot_inuse_add(net: sock_net(sk), prot: sk->sk_prot, val: 1);
103
104 return 0;
105}
106EXPORT_SYMBOL_GPL(raw_hash_sk);
107
108void raw_unhash_sk(struct sock *sk)
109{
110 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
111
112 spin_lock(lock: &h->lock);
113 if (sk_del_node_init_rcu(sk))
114 sock_prot_inuse_add(net: sock_net(sk), prot: sk->sk_prot, val: -1);
115 spin_unlock(lock: &h->lock);
116}
117EXPORT_SYMBOL_GPL(raw_unhash_sk);
118
119bool raw_v4_match(struct net *net, const struct sock *sk, unsigned short num,
120 __be32 raddr, __be32 laddr, int dif, int sdif)
121{
122 const struct inet_sock *inet = inet_sk(sk);
123
124 if (net_eq(net1: sock_net(sk), net2: net) && inet->inet_num == num &&
125 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
126 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
127 raw_sk_bound_dev_eq(net, bound_dev_if: sk->sk_bound_dev_if, dif, sdif))
128 return true;
129 return false;
130}
131EXPORT_SYMBOL_GPL(raw_v4_match);
132
133/*
134 * 0 - deliver
135 * 1 - block
136 */
137static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
138{
139 struct icmphdr _hdr;
140 const struct icmphdr *hdr;
141
142 hdr = skb_header_pointer(skb, offset: skb_transport_offset(skb),
143 len: sizeof(_hdr), buffer: &_hdr);
144 if (!hdr)
145 return 1;
146
147 if (hdr->type < 32) {
148 __u32 data = raw_sk(sk)->filter.data;
149
150 return ((1U << hdr->type) & data) != 0;
151 }
152
153 /* Do not block unknown ICMP types */
154 return 0;
155}
156
157/* IP input processing comes here for RAW socket delivery.
158 * Caller owns SKB, so we must make clones.
159 *
160 * RFC 1122: SHOULD pass TOS value up to the transport layer.
161 * -> It does. And not only TOS, but all IP header.
162 */
163static int raw_v4_input(struct net *net, struct sk_buff *skb,
164 const struct iphdr *iph, int hash)
165{
166 int sdif = inet_sdif(skb);
167 struct hlist_head *hlist;
168 int dif = inet_iif(skb);
169 int delivered = 0;
170 struct sock *sk;
171
172 hlist = &raw_v4_hashinfo.ht[hash];
173 rcu_read_lock();
174 sk_for_each_rcu(sk, hlist) {
175 if (!raw_v4_match(net, sk, num: iph->protocol,
176 raddr: iph->saddr, laddr: iph->daddr, dif, sdif))
177 continue;
178 delivered = 1;
179 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
180 ip_mc_sf_allow(sk, local: iph->daddr, rmt: iph->saddr,
181 dif: skb->dev->ifindex, sdif)) {
182 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
183
184 /* Not releasing hash table! */
185 if (clone)
186 raw_rcv(sk, clone);
187 }
188 }
189 rcu_read_unlock();
190 return delivered;
191}
192
193int raw_local_deliver(struct sk_buff *skb, int protocol)
194{
195 struct net *net = dev_net(dev: skb->dev);
196
197 return raw_v4_input(net, skb, ip_hdr(skb),
198 raw_hashfunc(net, protocol));
199}
200
201static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
202{
203 struct inet_sock *inet = inet_sk(sk);
204 const int type = icmp_hdr(skb)->type;
205 const int code = icmp_hdr(skb)->code;
206 int err = 0;
207 int harderr = 0;
208
209 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
210 ipv4_sk_update_pmtu(skb, sk, mtu: info);
211 else if (type == ICMP_REDIRECT) {
212 ipv4_sk_redirect(skb, sk);
213 return;
214 }
215
216 /* Report error on raw socket, if:
217 1. User requested ip_recverr.
218 2. Socket is connected (otherwise the error indication
219 is useless without ip_recverr and error is hard.
220 */
221 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
222 return;
223
224 switch (type) {
225 default:
226 case ICMP_TIME_EXCEEDED:
227 err = EHOSTUNREACH;
228 break;
229 case ICMP_SOURCE_QUENCH:
230 return;
231 case ICMP_PARAMETERPROB:
232 err = EPROTO;
233 harderr = 1;
234 break;
235 case ICMP_DEST_UNREACH:
236 err = EHOSTUNREACH;
237 if (code > NR_ICMP_UNREACH)
238 break;
239 if (code == ICMP_FRAG_NEEDED) {
240 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
241 err = EMSGSIZE;
242 } else {
243 err = icmp_err_convert[code].errno;
244 harderr = icmp_err_convert[code].fatal;
245 }
246 }
247
248 if (inet->recverr) {
249 const struct iphdr *iph = (const struct iphdr *)skb->data;
250 u8 *payload = skb->data + (iph->ihl << 2);
251
252 if (inet->hdrincl)
253 payload = skb->data;
254 ip_icmp_error(sk, skb, err, 0, info, payload);
255 }
256
257 if (inet->recverr || harderr) {
258 sk->sk_err = err;
259 sk_error_report(sk);
260 }
261}
262
263void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
264{
265 struct net *net = dev_net(dev: skb->dev);
266 int dif = skb->dev->ifindex;
267 int sdif = inet_sdif(skb);
268 struct hlist_head *hlist;
269 const struct iphdr *iph;
270 struct sock *sk;
271 int hash;
272
273 hash = raw_hashfunc(net, protocol);
274 hlist = &raw_v4_hashinfo.ht[hash];
275
276 rcu_read_lock();
277 sk_for_each_rcu(sk, hlist) {
278 iph = (const struct iphdr *)skb->data;
279 if (!raw_v4_match(net, sk, num: iph->protocol,
280 raddr: iph->daddr, laddr: iph->saddr, dif, sdif))
281 continue;
282 raw_err(sk, skb, info);
283 }
284 rcu_read_unlock();
285}
286
287static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
288{
289 enum skb_drop_reason reason;
290
291 /* Charge it to the socket. */
292
293 ipv4_pktinfo_prepare(sk, skb);
294 if (sock_queue_rcv_skb_reason(sk, skb, reason: &reason) < 0) {
295 kfree_skb_reason(skb, reason);
296 return NET_RX_DROP;
297 }
298
299 return NET_RX_SUCCESS;
300}
301
302int raw_rcv(struct sock *sk, struct sk_buff *skb)
303{
304 if (!xfrm4_policy_check(sk, dir: XFRM_POLICY_IN, skb)) {
305 atomic_inc(v: &sk->sk_drops);
306 kfree_skb_reason(skb, reason: SKB_DROP_REASON_XFRM_POLICY);
307 return NET_RX_DROP;
308 }
309 nf_reset_ct(skb);
310
311 skb_push(skb, len: skb->data - skb_network_header(skb));
312
313 raw_rcv_skb(sk, skb);
314 return 0;
315}
316
317static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
318 struct msghdr *msg, size_t length,
319 struct rtable **rtp, unsigned int flags,
320 const struct sockcm_cookie *sockc)
321{
322 struct inet_sock *inet = inet_sk(sk);
323 struct net *net = sock_net(sk);
324 struct iphdr *iph;
325 struct sk_buff *skb;
326 unsigned int iphlen;
327 int err;
328 struct rtable *rt = *rtp;
329 int hlen, tlen;
330
331 if (length > rt->dst.dev->mtu) {
332 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
333 rt->dst.dev->mtu);
334 return -EMSGSIZE;
335 }
336 if (length < sizeof(struct iphdr))
337 return -EINVAL;
338
339 if (flags&MSG_PROBE)
340 goto out;
341
342 hlen = LL_RESERVED_SPACE(rt->dst.dev);
343 tlen = rt->dst.dev->needed_tailroom;
344 skb = sock_alloc_send_skb(sk,
345 size: length + hlen + tlen + 15,
346 noblock: flags & MSG_DONTWAIT, errcode: &err);
347 if (!skb)
348 goto error;
349 skb_reserve(skb, len: hlen);
350
351 skb->priority = sk->sk_priority;
352 skb->mark = sockc->mark;
353 skb->tstamp = sockc->transmit_time;
354 skb_dst_set(skb, dst: &rt->dst);
355 *rtp = NULL;
356
357 skb_reset_network_header(skb);
358 iph = ip_hdr(skb);
359 skb_put(skb, len: length);
360
361 skb->ip_summed = CHECKSUM_NONE;
362
363 skb_setup_tx_timestamp(skb, sockc->tsflags);
364
365 if (flags & MSG_CONFIRM)
366 skb_set_dst_pending_confirm(skb, val: 1);
367
368 skb->transport_header = skb->network_header;
369 err = -EFAULT;
370 if (memcpy_from_msg(data: iph, msg, len: length))
371 goto error_free;
372
373 iphlen = iph->ihl * 4;
374
375 /*
376 * We don't want to modify the ip header, but we do need to
377 * be sure that it won't cause problems later along the network
378 * stack. Specifically we want to make sure that iph->ihl is a
379 * sane value. If ihl points beyond the length of the buffer passed
380 * in, reject the frame as invalid
381 */
382 err = -EINVAL;
383 if (iphlen > length)
384 goto error_free;
385
386 if (iphlen >= sizeof(*iph)) {
387 if (!iph->saddr)
388 iph->saddr = fl4->saddr;
389 iph->check = 0;
390 iph->tot_len = htons(length);
391 if (!iph->id)
392 ip_select_ident(net, skb, NULL);
393
394 iph->check = ip_fast_csum(iph: (unsigned char *)iph, ihl: iph->ihl);
395 skb->transport_header += iphlen;
396 if (iph->protocol == IPPROTO_ICMP &&
397 length >= iphlen + sizeof(struct icmphdr))
398 icmp_out_count(net, type: ((struct icmphdr *)
399 skb_transport_header(skb))->type);
400 }
401
402 err = NF_HOOK(pf: NFPROTO_IPV4, hook: NF_INET_LOCAL_OUT,
403 net, sk, skb, NULL, out: rt->dst.dev,
404 okfn: dst_output);
405 if (err > 0)
406 err = net_xmit_errno(err);
407 if (err)
408 goto error;
409out:
410 return 0;
411
412error_free:
413 kfree_skb(skb);
414error:
415 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
416 if (err == -ENOBUFS && !inet->recverr)
417 err = 0;
418 return err;
419}
420
421static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
422{
423 int err;
424
425 if (fl4->flowi4_proto != IPPROTO_ICMP)
426 return 0;
427
428 /* We only need the first two bytes. */
429 rfv->hlen = 2;
430
431 err = memcpy_from_msg(data: rfv->hdr.c, msg: rfv->msg, len: rfv->hlen);
432 if (err)
433 return err;
434
435 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
436 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
437
438 return 0;
439}
440
441static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
442 struct sk_buff *skb)
443{
444 struct raw_frag_vec *rfv = from;
445
446 if (offset < rfv->hlen) {
447 int copy = min(rfv->hlen - offset, len);
448
449 if (skb->ip_summed == CHECKSUM_PARTIAL)
450 memcpy(to, from: rfv->hdr.c + offset, len: copy);
451 else
452 skb->csum = csum_block_add(
453 csum: skb->csum,
454 csum2: csum_partial_copy_nocheck(src: rfv->hdr.c + offset,
455 dst: to, len: copy),
456 offset: odd);
457
458 odd = 0;
459 offset += copy;
460 to += copy;
461 len -= copy;
462
463 if (!len)
464 return 0;
465 }
466
467 offset -= rfv->hlen;
468
469 return ip_generic_getfrag(from: rfv->msg, to, offset, len, odd, skb);
470}
471
472static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
473{
474 struct inet_sock *inet = inet_sk(sk);
475 struct net *net = sock_net(sk);
476 struct ipcm_cookie ipc;
477 struct rtable *rt = NULL;
478 struct flowi4 fl4;
479 int free = 0;
480 __be32 daddr;
481 __be32 saddr;
482 u8 tos;
483 int err;
484 struct ip_options_data opt_copy;
485 struct raw_frag_vec rfv;
486 int hdrincl;
487
488 err = -EMSGSIZE;
489 if (len > 0xFFFF)
490 goto out;
491
492 /* hdrincl should be READ_ONCE(inet->hdrincl)
493 * but READ_ONCE() doesn't work with bit fields.
494 * Doing this indirectly yields the same result.
495 */
496 hdrincl = inet->hdrincl;
497 hdrincl = READ_ONCE(hdrincl);
498 /*
499 * Check the flags.
500 */
501
502 err = -EOPNOTSUPP;
503 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
504 goto out; /* compatibility */
505
506 /*
507 * Get and verify the address.
508 */
509
510 if (msg->msg_namelen) {
511 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
512 err = -EINVAL;
513 if (msg->msg_namelen < sizeof(*usin))
514 goto out;
515 if (usin->sin_family != AF_INET) {
516 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
517 __func__, current->comm);
518 err = -EAFNOSUPPORT;
519 if (usin->sin_family)
520 goto out;
521 }
522 daddr = usin->sin_addr.s_addr;
523 /* ANK: I did not forget to get protocol from port field.
524 * I just do not know, who uses this weirdness.
525 * IP_HDRINCL is much more convenient.
526 */
527 } else {
528 err = -EDESTADDRREQ;
529 if (sk->sk_state != TCP_ESTABLISHED)
530 goto out;
531 daddr = inet->inet_daddr;
532 }
533
534 ipcm_init_sk(ipcm: &ipc, inet);
535
536 if (msg->msg_controllen) {
537 err = ip_cmsg_send(sk, msg, ipc: &ipc, allow_ipv6: false);
538 if (unlikely(err)) {
539 kfree(objp: ipc.opt);
540 goto out;
541 }
542 if (ipc.opt)
543 free = 1;
544 }
545
546 saddr = ipc.addr;
547 ipc.addr = daddr;
548
549 if (!ipc.opt) {
550 struct ip_options_rcu *inet_opt;
551
552 rcu_read_lock();
553 inet_opt = rcu_dereference(inet->inet_opt);
554 if (inet_opt) {
555 memcpy(to: &opt_copy, from: inet_opt,
556 len: sizeof(*inet_opt) + inet_opt->opt.optlen);
557 ipc.opt = &opt_copy.opt;
558 }
559 rcu_read_unlock();
560 }
561
562 if (ipc.opt) {
563 err = -EINVAL;
564 /* Linux does not mangle headers on raw sockets,
565 * so that IP options + IP_HDRINCL is non-sense.
566 */
567 if (hdrincl)
568 goto done;
569 if (ipc.opt->opt.srr) {
570 if (!daddr)
571 goto done;
572 daddr = ipc.opt->opt.faddr;
573 }
574 }
575 tos = get_rtconn_flags(&ipc, sk);
576 if (msg->msg_flags & MSG_DONTROUTE)
577 tos |= RTO_ONLINK;
578
579 if (ipv4_is_multicast(addr: daddr)) {
580 if (!ipc.oif || netif_index_is_l3_master(net: sock_net(sk), ifindex: ipc.oif))
581 ipc.oif = inet->mc_index;
582 if (!saddr)
583 saddr = inet->mc_addr;
584 } else if (!ipc.oif) {
585 ipc.oif = inet->uc_index;
586 } else if (ipv4_is_lbcast(addr: daddr) && inet->uc_index) {
587 /* oif is set, packet is to local broadcast
588 * and uc_index is set. oif is most likely set
589 * by sk_bound_dev_if. If uc_index != oif check if the
590 * oif is an L3 master and uc_index is an L3 slave.
591 * If so, we want to allow the send using the uc_index.
592 */
593 if (ipc.oif != inet->uc_index &&
594 ipc.oif == l3mdev_master_ifindex_by_index(net: sock_net(sk),
595 ifindex: inet->uc_index)) {
596 ipc.oif = inet->uc_index;
597 }
598 }
599
600 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
601 RT_SCOPE_UNIVERSE,
602 hdrincl ? IPPROTO_RAW : sk->sk_protocol,
603 inet_sk_flowi_flags(sk) |
604 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
605 daddr, saddr, 0, 0, sk->sk_uid);
606
607 if (!hdrincl) {
608 rfv.msg = msg;
609 rfv.hlen = 0;
610
611 err = raw_probe_proto_opt(rfv: &rfv, fl4: &fl4);
612 if (err)
613 goto done;
614 }
615
616 security_sk_classify_flow(sk, flic: flowi4_to_flowi_common(fl4: &fl4));
617 rt = ip_route_output_flow(net, flp: &fl4, sk);
618 if (IS_ERR(ptr: rt)) {
619 err = PTR_ERR(ptr: rt);
620 rt = NULL;
621 goto done;
622 }
623
624 err = -EACCES;
625 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, flag: SOCK_BROADCAST))
626 goto done;
627
628 if (msg->msg_flags & MSG_CONFIRM)
629 goto do_confirm;
630back_from_confirm:
631
632 if (hdrincl)
633 err = raw_send_hdrinc(sk, fl4: &fl4, msg, length: len,
634 rtp: &rt, flags: msg->msg_flags, sockc: &ipc.sockc);
635
636 else {
637 if (!ipc.addr)
638 ipc.addr = fl4.daddr;
639 lock_sock(sk);
640 err = ip_append_data(sk, fl4: &fl4, getfrag: raw_getfrag,
641 from: &rfv, len, protolen: 0,
642 ipc: &ipc, rt: &rt, flags: msg->msg_flags);
643 if (err)
644 ip_flush_pending_frames(sk);
645 else if (!(msg->msg_flags & MSG_MORE)) {
646 err = ip_push_pending_frames(sk, fl4: &fl4);
647 if (err == -ENOBUFS && !inet->recverr)
648 err = 0;
649 }
650 release_sock(sk);
651 }
652done:
653 if (free)
654 kfree(objp: ipc.opt);
655 ip_rt_put(rt);
656
657out:
658 if (err < 0)
659 return err;
660 return len;
661
662do_confirm:
663 if (msg->msg_flags & MSG_PROBE)
664 dst_confirm_neigh(dst: &rt->dst, daddr: &fl4.daddr);
665 if (!(msg->msg_flags & MSG_PROBE) || len)
666 goto back_from_confirm;
667 err = 0;
668 goto done;
669}
670
671static void raw_close(struct sock *sk, long timeout)
672{
673 /*
674 * Raw sockets may have direct kernel references. Kill them.
675 */
676 ip_ra_control(sk, on: 0, NULL);
677
678 sk_common_release(sk);
679}
680
681static void raw_destroy(struct sock *sk)
682{
683 lock_sock(sk);
684 ip_flush_pending_frames(sk);
685 release_sock(sk);
686}
687
688/* This gets rid of all the nasties in af_inet. -DaveM */
689static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
690{
691 struct inet_sock *inet = inet_sk(sk);
692 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
693 struct net *net = sock_net(sk);
694 u32 tb_id = RT_TABLE_LOCAL;
695 int ret = -EINVAL;
696 int chk_addr_ret;
697
698 lock_sock(sk);
699 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
700 goto out;
701
702 if (sk->sk_bound_dev_if)
703 tb_id = l3mdev_fib_table_by_index(net,
704 sk->sk_bound_dev_if) ? : tb_id;
705
706 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
707
708 ret = -EADDRNOTAVAIL;
709 if (!inet_addr_valid_or_nonlocal(net, inet, addr: addr->sin_addr.s_addr,
710 addr_type: chk_addr_ret))
711 goto out;
712
713 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
714 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
715 inet->inet_saddr = 0; /* Use device */
716 sk_dst_reset(sk);
717 ret = 0;
718out:
719 release_sock(sk);
720 return ret;
721}
722
723/*
724 * This should be easy, if there is something there
725 * we return it, otherwise we block.
726 */
727
728static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
729 int flags, int *addr_len)
730{
731 struct inet_sock *inet = inet_sk(sk);
732 size_t copied = 0;
733 int err = -EOPNOTSUPP;
734 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
735 struct sk_buff *skb;
736
737 if (flags & MSG_OOB)
738 goto out;
739
740 if (flags & MSG_ERRQUEUE) {
741 err = ip_recv_error(sk, msg, len, addr_len);
742 goto out;
743 }
744
745 skb = skb_recv_datagram(sk, flags, err: &err);
746 if (!skb)
747 goto out;
748
749 copied = skb->len;
750 if (len < copied) {
751 msg->msg_flags |= MSG_TRUNC;
752 copied = len;
753 }
754
755 err = skb_copy_datagram_msg(from: skb, offset: 0, msg, size: copied);
756 if (err)
757 goto done;
758
759 sock_recv_cmsgs(msg, sk, skb);
760
761 /* Copy the address. */
762 if (sin) {
763 sin->sin_family = AF_INET;
764 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
765 sin->sin_port = 0;
766 memset(s: &sin->sin_zero, c: 0, n: sizeof(sin->sin_zero));
767 *addr_len = sizeof(*sin);
768 }
769 if (inet->cmsg_flags)
770 ip_cmsg_recv(msg, skb);
771 if (flags & MSG_TRUNC)
772 copied = skb->len;
773done:
774 skb_free_datagram(sk, skb);
775out:
776 if (err)
777 return err;
778 return copied;
779}
780
781static int raw_sk_init(struct sock *sk)
782{
783 struct raw_sock *rp = raw_sk(sk);
784
785 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
786 memset(s: &rp->filter, c: 0, n: sizeof(rp->filter));
787 return 0;
788}
789
790static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
791{
792 if (optlen > sizeof(struct icmp_filter))
793 optlen = sizeof(struct icmp_filter);
794 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen))
795 return -EFAULT;
796 return 0;
797}
798
799static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
800{
801 int len, ret = -EFAULT;
802
803 if (get_user(len, optlen))
804 goto out;
805 ret = -EINVAL;
806 if (len < 0)
807 goto out;
808 if (len > sizeof(struct icmp_filter))
809 len = sizeof(struct icmp_filter);
810 ret = -EFAULT;
811 if (put_user(len, optlen) ||
812 copy_to_user(to: optval, from: &raw_sk(sk)->filter, n: len))
813 goto out;
814 ret = 0;
815out: return ret;
816}
817
818static int do_raw_setsockopt(struct sock *sk, int level, int optname,
819 sockptr_t optval, unsigned int optlen)
820{
821 if (optname == ICMP_FILTER) {
822 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
823 return -EOPNOTSUPP;
824 else
825 return raw_seticmpfilter(sk, optval, optlen);
826 }
827 return -ENOPROTOOPT;
828}
829
830static int raw_setsockopt(struct sock *sk, int level, int optname,
831 sockptr_t optval, unsigned int optlen)
832{
833 if (level != SOL_RAW)
834 return ip_setsockopt(sk, level, optname, optval, optlen);
835 return do_raw_setsockopt(sk, level, optname, optval, optlen);
836}
837
838static int do_raw_getsockopt(struct sock *sk, int level, int optname,
839 char __user *optval, int __user *optlen)
840{
841 if (optname == ICMP_FILTER) {
842 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
843 return -EOPNOTSUPP;
844 else
845 return raw_geticmpfilter(sk, optval, optlen);
846 }
847 return -ENOPROTOOPT;
848}
849
850static int raw_getsockopt(struct sock *sk, int level, int optname,
851 char __user *optval, int __user *optlen)
852{
853 if (level != SOL_RAW)
854 return ip_getsockopt(sk, level, optname, optval, optlen);
855 return do_raw_getsockopt(sk, level, optname, optval, optlen);
856}
857
858static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
859{
860 switch (cmd) {
861 case SIOCOUTQ: {
862 int amount = sk_wmem_alloc_get(sk);
863
864 return put_user(amount, (int __user *)arg);
865 }
866 case SIOCINQ: {
867 struct sk_buff *skb;
868 int amount = 0;
869
870 spin_lock_bh(lock: &sk->sk_receive_queue.lock);
871 skb = skb_peek(list_: &sk->sk_receive_queue);
872 if (skb)
873 amount = skb->len;
874 spin_unlock_bh(lock: &sk->sk_receive_queue.lock);
875 return put_user(amount, (int __user *)arg);
876 }
877
878 default:
879#ifdef CONFIG_IP_MROUTE
880 return ipmr_ioctl(sk, cmd, (void __user *)arg);
881#else
882 return -ENOIOCTLCMD;
883#endif
884 }
885}
886
887#ifdef CONFIG_COMPAT
888static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
889{
890 switch (cmd) {
891 case SIOCOUTQ:
892 case SIOCINQ:
893 return -ENOIOCTLCMD;
894 default:
895#ifdef CONFIG_IP_MROUTE
896 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
897#else
898 return -ENOIOCTLCMD;
899#endif
900 }
901}
902#endif
903
904int raw_abort(struct sock *sk, int err)
905{
906 lock_sock(sk);
907
908 sk->sk_err = err;
909 sk_error_report(sk);
910 __udp_disconnect(sk, flags: 0);
911
912 release_sock(sk);
913
914 return 0;
915}
916EXPORT_SYMBOL_GPL(raw_abort);
917
918struct proto raw_prot = {
919 .name = "RAW",
920 .owner = THIS_MODULE,
921 .close = raw_close,
922 .destroy = raw_destroy,
923 .connect = ip4_datagram_connect,
924 .disconnect = __udp_disconnect,
925 .ioctl = raw_ioctl,
926 .init = raw_sk_init,
927 .setsockopt = raw_setsockopt,
928 .getsockopt = raw_getsockopt,
929 .sendmsg = raw_sendmsg,
930 .recvmsg = raw_recvmsg,
931 .bind = raw_bind,
932 .backlog_rcv = raw_rcv_skb,
933 .release_cb = ip4_datagram_release_cb,
934 .hash = raw_hash_sk,
935 .unhash = raw_unhash_sk,
936 .obj_size = sizeof(struct raw_sock),
937 .useroffset = offsetof(struct raw_sock, filter),
938 .usersize = sizeof_field(struct raw_sock, filter),
939 .h.raw_hash = &raw_v4_hashinfo,
940#ifdef CONFIG_COMPAT
941 .compat_ioctl = compat_raw_ioctl,
942#endif
943 .diag_destroy = raw_abort,
944};
945
946#ifdef CONFIG_PROC_FS
947static struct sock *raw_get_first(struct seq_file *seq, int bucket)
948{
949 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
950 struct raw_iter_state *state = raw_seq_private(seq);
951 struct hlist_head *hlist;
952 struct sock *sk;
953
954 for (state->bucket = bucket; state->bucket < RAW_HTABLE_SIZE;
955 ++state->bucket) {
956 hlist = &h->ht[state->bucket];
957 sk_for_each(sk, hlist) {
958 if (sock_net(sk) == seq_file_net(seq))
959 return sk;
960 }
961 }
962 return NULL;
963}
964
965static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
966{
967 struct raw_iter_state *state = raw_seq_private(seq);
968
969 do {
970 sk = sk_next(sk);
971 } while (sk && sock_net(sk) != seq_file_net(seq));
972
973 if (!sk)
974 return raw_get_first(seq, state->bucket + 1);
975 return sk;
976}
977
978static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
979{
980 struct sock *sk = raw_get_first(seq, 0);
981
982 if (sk)
983 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
984 --pos;
985 return pos ? NULL : sk;
986}
987
988void *raw_seq_start(struct seq_file *seq, loff_t *pos)
989 __acquires(&h->lock)
990{
991 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
992
993 spin_lock(&h->lock);
994
995 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
996}
997EXPORT_SYMBOL_GPL(raw_seq_start);
998
999void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1000{
1001 struct sock *sk;
1002
1003 if (v == SEQ_START_TOKEN)
1004 sk = raw_get_first(seq, 0);
1005 else
1006 sk = raw_get_next(seq, v);
1007 ++*pos;
1008 return sk;
1009}
1010EXPORT_SYMBOL_GPL(raw_seq_next);
1011
1012void raw_seq_stop(struct seq_file *seq, void *v)
1013 __releases(&h->lock)
1014{
1015 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
1016
1017 spin_unlock(&h->lock);
1018}
1019EXPORT_SYMBOL_GPL(raw_seq_stop);
1020
1021static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1022{
1023 struct inet_sock *inet = inet_sk(sp);
1024 __be32 dest = inet->inet_daddr,
1025 src = inet->inet_rcv_saddr;
1026 __u16 destp = 0,
1027 srcp = inet->inet_num;
1028
1029 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1030 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1031 i, src, srcp, dest, destp, sp->sk_state,
1032 sk_wmem_alloc_get(sp),
1033 sk_rmem_alloc_get(sp),
1034 0, 0L, 0,
1035 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1036 0, sock_i_ino(sp),
1037 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1038}
1039
1040static int raw_seq_show(struct seq_file *seq, void *v)
1041{
1042 if (v == SEQ_START_TOKEN)
1043 seq_printf(seq, " sl local_address rem_address st tx_queue "
1044 "rx_queue tr tm->when retrnsmt uid timeout "
1045 "inode ref pointer drops\n");
1046 else
1047 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1048 return 0;
1049}
1050
1051static const struct seq_operations raw_seq_ops = {
1052 .start = raw_seq_start,
1053 .next = raw_seq_next,
1054 .stop = raw_seq_stop,
1055 .show = raw_seq_show,
1056};
1057
1058static __net_init int raw_init_net(struct net *net)
1059{
1060 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
1061 sizeof(struct raw_iter_state), &raw_v4_hashinfo))
1062 return -ENOMEM;
1063
1064 return 0;
1065}
1066
1067static __net_exit void raw_exit_net(struct net *net)
1068{
1069 remove_proc_entry("raw", net->proc_net);
1070}
1071
1072static __net_initdata struct pernet_operations raw_net_ops = {
1073 .init = raw_init_net,
1074 .exit = raw_exit_net,
1075};
1076
1077int __init raw_proc_init(void)
1078{
1079
1080 return register_pernet_subsys(&raw_net_ops);
1081}
1082
1083void __init raw_proc_exit(void)
1084{
1085 unregister_pernet_subsys(&raw_net_ops);
1086}
1087#endif /* CONFIG_PROC_FS */
1088
1089static void raw_sysctl_init_net(struct net *net)
1090{
1091#ifdef CONFIG_NET_L3_MASTER_DEV
1092 net->ipv4.sysctl_raw_l3mdev_accept = 1;
1093#endif
1094}
1095
1096static int __net_init raw_sysctl_init(struct net *net)
1097{
1098 raw_sysctl_init_net(net);
1099 return 0;
1100}
1101
1102static struct pernet_operations __net_initdata raw_sysctl_ops = {
1103 .init = raw_sysctl_init,
1104};
1105
1106void __init raw_init(void)
1107{
1108 raw_sysctl_init_net(net: &init_net);
1109 if (register_pernet_subsys(&raw_sysctl_ops))
1110 panic(fmt: "RAW: failed to init sysctl parameters.\n");
1111}
1112

source code of linux/net/ipv4/raw.c