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 * The Internet Protocol (IP) module.
7 *
8 * Authors: Ross Biro
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Donald Becker, <becker@super.org>
11 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
12 * Richard Underwood
13 * Stefan Becker, <stefanb@yello.ping.de>
14 * Jorge Cwik, <jorge@laser.satlink.net>
15 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
16 *
17 *
18 * Fixes:
19 * Alan Cox : Commented a couple of minor bits of surplus code
20 * Alan Cox : Undefining IP_FORWARD doesn't include the code
21 * (just stops a compiler warning).
22 * Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes
23 * are junked rather than corrupting things.
24 * Alan Cox : Frames to bad broadcast subnets are dumped
25 * We used to process them non broadcast and
26 * boy could that cause havoc.
27 * Alan Cox : ip_forward sets the free flag on the
28 * new frame it queues. Still crap because
29 * it copies the frame but at least it
30 * doesn't eat memory too.
31 * Alan Cox : Generic queue code and memory fixes.
32 * Fred Van Kempen : IP fragment support (borrowed from NET2E)
33 * Gerhard Koerting: Forward fragmented frames correctly.
34 * Gerhard Koerting: Fixes to my fix of the above 8-).
35 * Gerhard Koerting: IP interface addressing fix.
36 * Linus Torvalds : More robustness checks
37 * Alan Cox : Even more checks: Still not as robust as it ought to be
38 * Alan Cox : Save IP header pointer for later
39 * Alan Cox : ip option setting
40 * Alan Cox : Use ip_tos/ip_ttl settings
41 * Alan Cox : Fragmentation bogosity removed
42 * (Thanks to Mark.Bush@prg.ox.ac.uk)
43 * Dmitry Gorodchanin : Send of a raw packet crash fix.
44 * Alan Cox : Silly ip bug when an overlength
45 * fragment turns up. Now frees the
46 * queue.
47 * Linus Torvalds/ : Memory leakage on fragmentation
48 * Alan Cox : handling.
49 * Gerhard Koerting: Forwarding uses IP priority hints
50 * Teemu Rantanen : Fragment problems.
51 * Alan Cox : General cleanup, comments and reformat
52 * Alan Cox : SNMP statistics
53 * Alan Cox : BSD address rule semantics. Also see
54 * UDP as there is a nasty checksum issue
55 * if you do things the wrong way.
56 * Alan Cox : Always defrag, moved IP_FORWARD to the config.in file
57 * Alan Cox : IP options adjust sk->priority.
58 * Pedro Roque : Fix mtu/length error in ip_forward.
59 * Alan Cox : Avoid ip_chk_addr when possible.
60 * Richard Underwood : IP multicasting.
61 * Alan Cox : Cleaned up multicast handlers.
62 * Alan Cox : RAW sockets demultiplex in the BSD style.
63 * Gunther Mayer : Fix the SNMP reporting typo
64 * Alan Cox : Always in group 224.0.0.1
65 * Pauline Middelink : Fast ip_checksum update when forwarding
66 * Masquerading support.
67 * Alan Cox : Multicast loopback error for 224.0.0.1
68 * Alan Cox : IP_MULTICAST_LOOP option.
69 * Alan Cox : Use notifiers.
70 * Bjorn Ekwall : Removed ip_csum (from slhc.c too)
71 * Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!)
72 * Stefan Becker : Send out ICMP HOST REDIRECT
73 * Arnt Gulbrandsen : ip_build_xmit
74 * Alan Cox : Per socket routing cache
75 * Alan Cox : Fixed routing cache, added header cache.
76 * Alan Cox : Loopback didn't work right in original ip_build_xmit - fixed it.
77 * Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net.
78 * Alan Cox : Incoming IP option handling.
79 * Alan Cox : Set saddr on raw output frames as per BSD.
80 * Alan Cox : Stopped broadcast source route explosions.
81 * Alan Cox : Can disable source routing
82 * Takeshi Sone : Masquerading didn't work.
83 * Dave Bonn,Alan Cox : Faster IP forwarding whenever possible.
84 * Alan Cox : Memory leaks, tramples, misc debugging.
85 * Alan Cox : Fixed multicast (by popular demand 8))
86 * Alan Cox : Fixed forwarding (by even more popular demand 8))
87 * Alan Cox : Fixed SNMP statistics [I think]
88 * Gerhard Koerting : IP fragmentation forwarding fix
89 * Alan Cox : Device lock against page fault.
90 * Alan Cox : IP_HDRINCL facility.
91 * Werner Almesberger : Zero fragment bug
92 * Alan Cox : RAW IP frame length bug
93 * Alan Cox : Outgoing firewall on build_xmit
94 * A.N.Kuznetsov : IP_OPTIONS support throughout the kernel
95 * Alan Cox : Multicast routing hooks
96 * Jos Vos : Do accounting *before* call_in_firewall
97 * Willy Konynenberg : Transparent proxying support
98 *
99 *
100 *
101 * To Fix:
102 * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
103 * and could be made very efficient with the addition of some virtual memory hacks to permit
104 * the allocation of a buffer that can then be 'grown' by twiddling page tables.
105 * Output fragmentation wants updating along with the buffer management to use a single
106 * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
107 * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
108 * fragmentation anyway.
109 *
110 * This program is free software; you can redistribute it and/or
111 * modify it under the terms of the GNU General Public License
112 * as published by the Free Software Foundation; either version
113 * 2 of the License, or (at your option) any later version.
114 */
115
116#define pr_fmt(fmt) "IPv4: " fmt
117
118#include <linux/module.h>
119#include <linux/types.h>
120#include <linux/kernel.h>
121#include <linux/string.h>
122#include <linux/errno.h>
123#include <linux/slab.h>
124
125#include <linux/net.h>
126#include <linux/socket.h>
127#include <linux/sockios.h>
128#include <linux/in.h>
129#include <linux/inet.h>
130#include <linux/inetdevice.h>
131#include <linux/netdevice.h>
132#include <linux/etherdevice.h>
133
134#include <net/snmp.h>
135#include <net/ip.h>
136#include <net/protocol.h>
137#include <net/route.h>
138#include <linux/skbuff.h>
139#include <net/sock.h>
140#include <net/arp.h>
141#include <net/icmp.h>
142#include <net/raw.h>
143#include <net/checksum.h>
144#include <net/inet_ecn.h>
145#include <linux/netfilter_ipv4.h>
146#include <net/xfrm.h>
147#include <linux/mroute.h>
148#include <linux/netlink.h>
149#include <net/dst_metadata.h>
150
151/*
152 * Process Router Attention IP option (RFC 2113)
153 */
154bool ip_call_ra_chain(struct sk_buff *skb)
155{
156 struct ip_ra_chain *ra;
157 u8 protocol = ip_hdr(skb)->protocol;
158 struct sock *last = NULL;
159 struct net_device *dev = skb->dev;
160 struct net *net = dev_net(dev);
161
162 for (ra = rcu_dereference(net->ipv4.ra_chain); ra; ra = rcu_dereference(ra->next)) {
163 struct sock *sk = ra->sk;
164
165 /* If socket is bound to an interface, only report
166 * the packet if it came from that interface.
167 */
168 if (sk && inet_sk(sk)->inet_num == protocol &&
169 (!sk->sk_bound_dev_if ||
170 sk->sk_bound_dev_if == dev->ifindex)) {
171 if (ip_is_fragment(ip_hdr(skb))) {
172 if (ip_defrag(net, skb, IP_DEFRAG_CALL_RA_CHAIN))
173 return true;
174 }
175 if (last) {
176 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
177 if (skb2)
178 raw_rcv(last, skb2);
179 }
180 last = sk;
181 }
182 }
183
184 if (last) {
185 raw_rcv(last, skb);
186 return true;
187 }
188 return false;
189}
190
191void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int protocol)
192{
193 const struct net_protocol *ipprot;
194 int raw, ret;
195
196resubmit:
197 raw = raw_local_deliver(skb, protocol);
198
199 ipprot = rcu_dereference(inet_protos[protocol]);
200 if (ipprot) {
201 if (!ipprot->no_policy) {
202 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
203 kfree_skb(skb);
204 return;
205 }
206 nf_reset(skb);
207 }
208 ret = ipprot->handler(skb);
209 if (ret < 0) {
210 protocol = -ret;
211 goto resubmit;
212 }
213 __IP_INC_STATS(net, IPSTATS_MIB_INDELIVERS);
214 } else {
215 if (!raw) {
216 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
217 __IP_INC_STATS(net, IPSTATS_MIB_INUNKNOWNPROTOS);
218 icmp_send(skb, ICMP_DEST_UNREACH,
219 ICMP_PROT_UNREACH, 0);
220 }
221 kfree_skb(skb);
222 } else {
223 __IP_INC_STATS(net, IPSTATS_MIB_INDELIVERS);
224 consume_skb(skb);
225 }
226 }
227}
228
229static int ip_local_deliver_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
230{
231 __skb_pull(skb, skb_network_header_len(skb));
232
233 rcu_read_lock();
234 ip_protocol_deliver_rcu(net, skb, ip_hdr(skb)->protocol);
235 rcu_read_unlock();
236
237 return 0;
238}
239
240/*
241 * Deliver IP Packets to the higher protocol layers.
242 */
243int ip_local_deliver(struct sk_buff *skb)
244{
245 /*
246 * Reassemble IP fragments.
247 */
248 struct net *net = dev_net(skb->dev);
249
250 if (ip_is_fragment(ip_hdr(skb))) {
251 if (ip_defrag(net, skb, IP_DEFRAG_LOCAL_DELIVER))
252 return 0;
253 }
254
255 return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN,
256 net, NULL, skb, skb->dev, NULL,
257 ip_local_deliver_finish);
258}
259
260static inline bool ip_rcv_options(struct sk_buff *skb)
261{
262 struct ip_options *opt;
263 const struct iphdr *iph;
264 struct net_device *dev = skb->dev;
265
266 /* It looks as overkill, because not all
267 IP options require packet mangling.
268 But it is the easiest for now, especially taking
269 into account that combination of IP options
270 and running sniffer is extremely rare condition.
271 --ANK (980813)
272 */
273 if (skb_cow(skb, skb_headroom(skb))) {
274 __IP_INC_STATS(dev_net(dev), IPSTATS_MIB_INDISCARDS);
275 goto drop;
276 }
277
278 iph = ip_hdr(skb);
279 opt = &(IPCB(skb)->opt);
280 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
281
282 if (ip_options_compile(dev_net(dev), opt, skb)) {
283 __IP_INC_STATS(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
284 goto drop;
285 }
286
287 if (unlikely(opt->srr)) {
288 struct in_device *in_dev = __in_dev_get_rcu(dev);
289
290 if (in_dev) {
291 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
292 if (IN_DEV_LOG_MARTIANS(in_dev))
293 net_info_ratelimited("source route option %pI4 -> %pI4\n",
294 &iph->saddr,
295 &iph->daddr);
296 goto drop;
297 }
298 }
299
300 if (ip_options_rcv_srr(skb))
301 goto drop;
302 }
303
304 return false;
305drop:
306 return true;
307}
308
309static int ip_rcv_finish_core(struct net *net, struct sock *sk,
310 struct sk_buff *skb, struct net_device *dev)
311{
312 const struct iphdr *iph = ip_hdr(skb);
313 int (*edemux)(struct sk_buff *skb);
314 struct rtable *rt;
315 int err;
316
317 if (net->ipv4.sysctl_ip_early_demux &&
318 !skb_dst(skb) &&
319 !skb->sk &&
320 !ip_is_fragment(iph)) {
321 const struct net_protocol *ipprot;
322 int protocol = iph->protocol;
323
324 ipprot = rcu_dereference(inet_protos[protocol]);
325 if (ipprot && (edemux = READ_ONCE(ipprot->early_demux))) {
326 err = edemux(skb);
327 if (unlikely(err))
328 goto drop_error;
329 /* must reload iph, skb->head might have changed */
330 iph = ip_hdr(skb);
331 }
332 }
333
334 /*
335 * Initialise the virtual path cache for the packet. It describes
336 * how the packet travels inside Linux networking.
337 */
338 if (!skb_valid_dst(skb)) {
339 err = ip_route_input_noref(skb, iph->daddr, iph->saddr,
340 iph->tos, dev);
341 if (unlikely(err))
342 goto drop_error;
343 }
344
345#ifdef CONFIG_IP_ROUTE_CLASSID
346 if (unlikely(skb_dst(skb)->tclassid)) {
347 struct ip_rt_acct *st = this_cpu_ptr(ip_rt_acct);
348 u32 idx = skb_dst(skb)->tclassid;
349 st[idx&0xFF].o_packets++;
350 st[idx&0xFF].o_bytes += skb->len;
351 st[(idx>>16)&0xFF].i_packets++;
352 st[(idx>>16)&0xFF].i_bytes += skb->len;
353 }
354#endif
355
356 if (iph->ihl > 5 && ip_rcv_options(skb))
357 goto drop;
358
359 rt = skb_rtable(skb);
360 if (rt->rt_type == RTN_MULTICAST) {
361 __IP_UPD_PO_STATS(net, IPSTATS_MIB_INMCAST, skb->len);
362 } else if (rt->rt_type == RTN_BROADCAST) {
363 __IP_UPD_PO_STATS(net, IPSTATS_MIB_INBCAST, skb->len);
364 } else if (skb->pkt_type == PACKET_BROADCAST ||
365 skb->pkt_type == PACKET_MULTICAST) {
366 struct in_device *in_dev = __in_dev_get_rcu(dev);
367
368 /* RFC 1122 3.3.6:
369 *
370 * When a host sends a datagram to a link-layer broadcast
371 * address, the IP destination address MUST be a legal IP
372 * broadcast or IP multicast address.
373 *
374 * A host SHOULD silently discard a datagram that is received
375 * via a link-layer broadcast (see Section 2.4) but does not
376 * specify an IP multicast or broadcast destination address.
377 *
378 * This doesn't explicitly say L2 *broadcast*, but broadcast is
379 * in a way a form of multicast and the most common use case for
380 * this is 802.11 protecting against cross-station spoofing (the
381 * so-called "hole-196" attack) so do it for both.
382 */
383 if (in_dev &&
384 IN_DEV_ORCONF(in_dev, DROP_UNICAST_IN_L2_MULTICAST))
385 goto drop;
386 }
387
388 return NET_RX_SUCCESS;
389
390drop:
391 kfree_skb(skb);
392 return NET_RX_DROP;
393
394drop_error:
395 if (err == -EXDEV)
396 __NET_INC_STATS(net, LINUX_MIB_IPRPFILTER);
397 goto drop;
398}
399
400static int ip_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
401{
402 struct net_device *dev = skb->dev;
403 int ret;
404
405 /* if ingress device is enslaved to an L3 master device pass the
406 * skb to its handler for processing
407 */
408 skb = l3mdev_ip_rcv(skb);
409 if (!skb)
410 return NET_RX_SUCCESS;
411
412 ret = ip_rcv_finish_core(net, sk, skb, dev);
413 if (ret != NET_RX_DROP)
414 ret = dst_input(skb);
415 return ret;
416}
417
418/*
419 * Main IP Receive routine.
420 */
421static struct sk_buff *ip_rcv_core(struct sk_buff *skb, struct net *net)
422{
423 const struct iphdr *iph;
424 u32 len;
425
426 /* When the interface is in promisc. mode, drop all the crap
427 * that it receives, do not try to analyse it.
428 */
429 if (skb->pkt_type == PACKET_OTHERHOST)
430 goto drop;
431
432 __IP_UPD_PO_STATS(net, IPSTATS_MIB_IN, skb->len);
433
434 skb = skb_share_check(skb, GFP_ATOMIC);
435 if (!skb) {
436 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
437 goto out;
438 }
439
440 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
441 goto inhdr_error;
442
443 iph = ip_hdr(skb);
444
445 /*
446 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
447 *
448 * Is the datagram acceptable?
449 *
450 * 1. Length at least the size of an ip header
451 * 2. Version of 4
452 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
453 * 4. Doesn't have a bogus length
454 */
455
456 if (iph->ihl < 5 || iph->version != 4)
457 goto inhdr_error;
458
459 BUILD_BUG_ON(IPSTATS_MIB_ECT1PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_1);
460 BUILD_BUG_ON(IPSTATS_MIB_ECT0PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_0);
461 BUILD_BUG_ON(IPSTATS_MIB_CEPKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_CE);
462 __IP_ADD_STATS(net,
463 IPSTATS_MIB_NOECTPKTS + (iph->tos & INET_ECN_MASK),
464 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
465
466 if (!pskb_may_pull(skb, iph->ihl*4))
467 goto inhdr_error;
468
469 iph = ip_hdr(skb);
470
471 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
472 goto csum_error;
473
474 len = ntohs(iph->tot_len);
475 if (skb->len < len) {
476 __IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS);
477 goto drop;
478 } else if (len < (iph->ihl*4))
479 goto inhdr_error;
480
481 /* Our transport medium may have padded the buffer out. Now we know it
482 * is IP we can trim to the true length of the frame.
483 * Note this now means skb->len holds ntohs(iph->tot_len).
484 */
485 if (pskb_trim_rcsum(skb, len)) {
486 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
487 goto drop;
488 }
489
490 iph = ip_hdr(skb);
491 skb->transport_header = skb->network_header + iph->ihl*4;
492
493 /* Remove any debris in the socket control block */
494 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
495 IPCB(skb)->iif = skb->skb_iif;
496
497 /* Must drop socket now because of tproxy. */
498 skb_orphan(skb);
499
500 return skb;
501
502csum_error:
503 __IP_INC_STATS(net, IPSTATS_MIB_CSUMERRORS);
504inhdr_error:
505 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
506drop:
507 kfree_skb(skb);
508out:
509 return NULL;
510}
511
512/*
513 * IP receive entry point
514 */
515int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
516 struct net_device *orig_dev)
517{
518 struct net *net = dev_net(dev);
519
520 skb = ip_rcv_core(skb, net);
521 if (skb == NULL)
522 return NET_RX_DROP;
523
524 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING,
525 net, NULL, skb, dev, NULL,
526 ip_rcv_finish);
527}
528
529static void ip_sublist_rcv_finish(struct list_head *head)
530{
531 struct sk_buff *skb, *next;
532
533 list_for_each_entry_safe(skb, next, head, list) {
534 skb_list_del_init(skb);
535 dst_input(skb);
536 }
537}
538
539static void ip_list_rcv_finish(struct net *net, struct sock *sk,
540 struct list_head *head)
541{
542 struct dst_entry *curr_dst = NULL;
543 struct sk_buff *skb, *next;
544 struct list_head sublist;
545
546 INIT_LIST_HEAD(&sublist);
547 list_for_each_entry_safe(skb, next, head, list) {
548 struct net_device *dev = skb->dev;
549 struct dst_entry *dst;
550
551 skb_list_del_init(skb);
552 /* if ingress device is enslaved to an L3 master device pass the
553 * skb to its handler for processing
554 */
555 skb = l3mdev_ip_rcv(skb);
556 if (!skb)
557 continue;
558 if (ip_rcv_finish_core(net, sk, skb, dev) == NET_RX_DROP)
559 continue;
560
561 dst = skb_dst(skb);
562 if (curr_dst != dst) {
563 /* dispatch old sublist */
564 if (!list_empty(&sublist))
565 ip_sublist_rcv_finish(&sublist);
566 /* start new sublist */
567 INIT_LIST_HEAD(&sublist);
568 curr_dst = dst;
569 }
570 list_add_tail(&skb->list, &sublist);
571 }
572 /* dispatch final sublist */
573 ip_sublist_rcv_finish(&sublist);
574}
575
576static void ip_sublist_rcv(struct list_head *head, struct net_device *dev,
577 struct net *net)
578{
579 NF_HOOK_LIST(NFPROTO_IPV4, NF_INET_PRE_ROUTING, net, NULL,
580 head, dev, NULL, ip_rcv_finish);
581 ip_list_rcv_finish(net, NULL, head);
582}
583
584/* Receive a list of IP packets */
585void ip_list_rcv(struct list_head *head, struct packet_type *pt,
586 struct net_device *orig_dev)
587{
588 struct net_device *curr_dev = NULL;
589 struct net *curr_net = NULL;
590 struct sk_buff *skb, *next;
591 struct list_head sublist;
592
593 INIT_LIST_HEAD(&sublist);
594 list_for_each_entry_safe(skb, next, head, list) {
595 struct net_device *dev = skb->dev;
596 struct net *net = dev_net(dev);
597
598 skb_list_del_init(skb);
599 skb = ip_rcv_core(skb, net);
600 if (skb == NULL)
601 continue;
602
603 if (curr_dev != dev || curr_net != net) {
604 /* dispatch old sublist */
605 if (!list_empty(&sublist))
606 ip_sublist_rcv(&sublist, curr_dev, curr_net);
607 /* start new sublist */
608 INIT_LIST_HEAD(&sublist);
609 curr_dev = dev;
610 curr_net = net;
611 }
612 list_add_tail(&skb->list, &sublist);
613 }
614 /* dispatch final sublist */
615 ip_sublist_rcv(&sublist, curr_dev, curr_net);
616}
617