Merge branch 'bnxt_en-next'
[linux/fpc-iii.git] / net / ipv4 / arp.c
blob89a8cac4726a5e354371bb0c76a7e60cd2d7026e
1 /* linux/net/ipv4/arp.c
3 * Copyright (C) 1994 by Florian La Roche
5 * This module implements the Address Resolution Protocol ARP (RFC 826),
6 * which is used to convert IP addresses (or in the future maybe other
7 * high-level addresses) into a low-level hardware address (like an Ethernet
8 * address).
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
15 * Fixes:
16 * Alan Cox : Removed the Ethernet assumptions in
17 * Florian's code
18 * Alan Cox : Fixed some small errors in the ARP
19 * logic
20 * Alan Cox : Allow >4K in /proc
21 * Alan Cox : Make ARP add its own protocol entry
22 * Ross Martin : Rewrote arp_rcv() and arp_get_info()
23 * Stephen Henson : Add AX25 support to arp_get_info()
24 * Alan Cox : Drop data when a device is downed.
25 * Alan Cox : Use init_timer().
26 * Alan Cox : Double lock fixes.
27 * Martin Seine : Move the arphdr structure
28 * to if_arp.h for compatibility.
29 * with BSD based programs.
30 * Andrew Tridgell : Added ARP netmask code and
31 * re-arranged proxy handling.
32 * Alan Cox : Changed to use notifiers.
33 * Niibe Yutaka : Reply for this device or proxies only.
34 * Alan Cox : Don't proxy across hardware types!
35 * Jonathan Naylor : Added support for NET/ROM.
36 * Mike Shaver : RFC1122 checks.
37 * Jonathan Naylor : Only lookup the hardware address for
38 * the correct hardware type.
39 * Germano Caronni : Assorted subtle races.
40 * Craig Schlenter : Don't modify permanent entry
41 * during arp_rcv.
42 * Russ Nelson : Tidied up a few bits.
43 * Alexey Kuznetsov: Major changes to caching and behaviour,
44 * eg intelligent arp probing and
45 * generation
46 * of host down events.
47 * Alan Cox : Missing unlock in device events.
48 * Eckes : ARP ioctl control errors.
49 * Alexey Kuznetsov: Arp free fix.
50 * Manuel Rodriguez: Gratuitous ARP.
51 * Jonathan Layes : Added arpd support through kerneld
52 * message queue (960314)
53 * Mike Shaver : /proc/sys/net/ipv4/arp_* support
54 * Mike McLagan : Routing by source
55 * Stuart Cheshire : Metricom and grat arp fixes
56 * *** FOR 2.1 clean this up ***
57 * Lawrence V. Stefani: (08/12/96) Added FDDI support.
58 * Alan Cox : Took the AP1000 nasty FDDI hack and
59 * folded into the mainstream FDDI code.
60 * Ack spit, Linus how did you allow that
61 * one in...
62 * Jes Sorensen : Make FDDI work again in 2.1.x and
63 * clean up the APFDDI & gen. FDDI bits.
64 * Alexey Kuznetsov: new arp state machine;
65 * now it is in net/core/neighbour.c.
66 * Krzysztof Halasa: Added Frame Relay ARP support.
67 * Arnaldo C. Melo : convert /proc/net/arp to seq_file
68 * Shmulik Hen: Split arp_send to arp_create and
69 * arp_xmit so intermediate drivers like
70 * bonding can change the skb before
71 * sending (e.g. insert 8021q tag).
72 * Harald Welte : convert to make use of jenkins hash
73 * Jesper D. Brouer: Proxy ARP PVLAN RFC 3069 support.
76 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
78 #include <linux/module.h>
79 #include <linux/types.h>
80 #include <linux/string.h>
81 #include <linux/kernel.h>
82 #include <linux/capability.h>
83 #include <linux/socket.h>
84 #include <linux/sockios.h>
85 #include <linux/errno.h>
86 #include <linux/in.h>
87 #include <linux/mm.h>
88 #include <linux/inet.h>
89 #include <linux/inetdevice.h>
90 #include <linux/netdevice.h>
91 #include <linux/etherdevice.h>
92 #include <linux/fddidevice.h>
93 #include <linux/if_arp.h>
94 #include <linux/skbuff.h>
95 #include <linux/proc_fs.h>
96 #include <linux/seq_file.h>
97 #include <linux/stat.h>
98 #include <linux/init.h>
99 #include <linux/net.h>
100 #include <linux/rcupdate.h>
101 #include <linux/slab.h>
102 #ifdef CONFIG_SYSCTL
103 #include <linux/sysctl.h>
104 #endif
106 #include <net/net_namespace.h>
107 #include <net/ip.h>
108 #include <net/icmp.h>
109 #include <net/route.h>
110 #include <net/protocol.h>
111 #include <net/tcp.h>
112 #include <net/sock.h>
113 #include <net/arp.h>
114 #include <net/ax25.h>
115 #include <net/netrom.h>
116 #include <net/dst_metadata.h>
117 #include <net/ip_tunnels.h>
119 #include <linux/uaccess.h>
121 #include <linux/netfilter_arp.h>
124 * Interface to generic neighbour cache.
126 static u32 arp_hash(const void *pkey, const struct net_device *dev, __u32 *hash_rnd);
127 static bool arp_key_eq(const struct neighbour *n, const void *pkey);
128 static int arp_constructor(struct neighbour *neigh);
129 static void arp_solicit(struct neighbour *neigh, struct sk_buff *skb);
130 static void arp_error_report(struct neighbour *neigh, struct sk_buff *skb);
131 static void parp_redo(struct sk_buff *skb);
133 static const struct neigh_ops arp_generic_ops = {
134 .family = AF_INET,
135 .solicit = arp_solicit,
136 .error_report = arp_error_report,
137 .output = neigh_resolve_output,
138 .connected_output = neigh_connected_output,
141 static const struct neigh_ops arp_hh_ops = {
142 .family = AF_INET,
143 .solicit = arp_solicit,
144 .error_report = arp_error_report,
145 .output = neigh_resolve_output,
146 .connected_output = neigh_resolve_output,
149 static const struct neigh_ops arp_direct_ops = {
150 .family = AF_INET,
151 .output = neigh_direct_output,
152 .connected_output = neigh_direct_output,
155 struct neigh_table arp_tbl = {
156 .family = AF_INET,
157 .key_len = 4,
158 .protocol = cpu_to_be16(ETH_P_IP),
159 .hash = arp_hash,
160 .key_eq = arp_key_eq,
161 .constructor = arp_constructor,
162 .proxy_redo = parp_redo,
163 .id = "arp_cache",
164 .parms = {
165 .tbl = &arp_tbl,
166 .reachable_time = 30 * HZ,
167 .data = {
168 [NEIGH_VAR_MCAST_PROBES] = 3,
169 [NEIGH_VAR_UCAST_PROBES] = 3,
170 [NEIGH_VAR_RETRANS_TIME] = 1 * HZ,
171 [NEIGH_VAR_BASE_REACHABLE_TIME] = 30 * HZ,
172 [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
173 [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
174 [NEIGH_VAR_QUEUE_LEN_BYTES] = 64 * 1024,
175 [NEIGH_VAR_PROXY_QLEN] = 64,
176 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
177 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
178 [NEIGH_VAR_LOCKTIME] = 1 * HZ,
181 .gc_interval = 30 * HZ,
182 .gc_thresh1 = 128,
183 .gc_thresh2 = 512,
184 .gc_thresh3 = 1024,
186 EXPORT_SYMBOL(arp_tbl);
188 int arp_mc_map(__be32 addr, u8 *haddr, struct net_device *dev, int dir)
190 switch (dev->type) {
191 case ARPHRD_ETHER:
192 case ARPHRD_FDDI:
193 case ARPHRD_IEEE802:
194 ip_eth_mc_map(addr, haddr);
195 return 0;
196 case ARPHRD_INFINIBAND:
197 ip_ib_mc_map(addr, dev->broadcast, haddr);
198 return 0;
199 case ARPHRD_IPGRE:
200 ip_ipgre_mc_map(addr, dev->broadcast, haddr);
201 return 0;
202 default:
203 if (dir) {
204 memcpy(haddr, dev->broadcast, dev->addr_len);
205 return 0;
208 return -EINVAL;
212 static u32 arp_hash(const void *pkey,
213 const struct net_device *dev,
214 __u32 *hash_rnd)
216 return arp_hashfn(pkey, dev, hash_rnd);
219 static bool arp_key_eq(const struct neighbour *neigh, const void *pkey)
221 return neigh_key_eq32(neigh, pkey);
224 static int arp_constructor(struct neighbour *neigh)
226 __be32 addr = *(__be32 *)neigh->primary_key;
227 struct net_device *dev = neigh->dev;
228 struct in_device *in_dev;
229 struct neigh_parms *parms;
231 rcu_read_lock();
232 in_dev = __in_dev_get_rcu(dev);
233 if (!in_dev) {
234 rcu_read_unlock();
235 return -EINVAL;
238 neigh->type = inet_addr_type_dev_table(dev_net(dev), dev, addr);
240 parms = in_dev->arp_parms;
241 __neigh_parms_put(neigh->parms);
242 neigh->parms = neigh_parms_clone(parms);
243 rcu_read_unlock();
245 if (!dev->header_ops) {
246 neigh->nud_state = NUD_NOARP;
247 neigh->ops = &arp_direct_ops;
248 neigh->output = neigh_direct_output;
249 } else {
250 /* Good devices (checked by reading texts, but only Ethernet is
251 tested)
253 ARPHRD_ETHER: (ethernet, apfddi)
254 ARPHRD_FDDI: (fddi)
255 ARPHRD_IEEE802: (tr)
256 ARPHRD_METRICOM: (strip)
257 ARPHRD_ARCNET:
258 etc. etc. etc.
260 ARPHRD_IPDDP will also work, if author repairs it.
261 I did not it, because this driver does not work even
262 in old paradigm.
265 if (neigh->type == RTN_MULTICAST) {
266 neigh->nud_state = NUD_NOARP;
267 arp_mc_map(addr, neigh->ha, dev, 1);
268 } else if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) {
269 neigh->nud_state = NUD_NOARP;
270 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
271 } else if (neigh->type == RTN_BROADCAST ||
272 (dev->flags & IFF_POINTOPOINT)) {
273 neigh->nud_state = NUD_NOARP;
274 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
277 if (dev->header_ops->cache)
278 neigh->ops = &arp_hh_ops;
279 else
280 neigh->ops = &arp_generic_ops;
282 if (neigh->nud_state & NUD_VALID)
283 neigh->output = neigh->ops->connected_output;
284 else
285 neigh->output = neigh->ops->output;
287 return 0;
290 static void arp_error_report(struct neighbour *neigh, struct sk_buff *skb)
292 dst_link_failure(skb);
293 kfree_skb(skb);
296 /* Create and send an arp packet. */
297 static void arp_send_dst(int type, int ptype, __be32 dest_ip,
298 struct net_device *dev, __be32 src_ip,
299 const unsigned char *dest_hw,
300 const unsigned char *src_hw,
301 const unsigned char *target_hw,
302 struct dst_entry *dst)
304 struct sk_buff *skb;
306 /* arp on this interface. */
307 if (dev->flags & IFF_NOARP)
308 return;
310 skb = arp_create(type, ptype, dest_ip, dev, src_ip,
311 dest_hw, src_hw, target_hw);
312 if (!skb)
313 return;
315 skb_dst_set(skb, dst_clone(dst));
316 arp_xmit(skb);
319 void arp_send(int type, int ptype, __be32 dest_ip,
320 struct net_device *dev, __be32 src_ip,
321 const unsigned char *dest_hw, const unsigned char *src_hw,
322 const unsigned char *target_hw)
324 arp_send_dst(type, ptype, dest_ip, dev, src_ip, dest_hw, src_hw,
325 target_hw, NULL);
327 EXPORT_SYMBOL(arp_send);
329 static void arp_solicit(struct neighbour *neigh, struct sk_buff *skb)
331 __be32 saddr = 0;
332 u8 dst_ha[MAX_ADDR_LEN], *dst_hw = NULL;
333 struct net_device *dev = neigh->dev;
334 __be32 target = *(__be32 *)neigh->primary_key;
335 int probes = atomic_read(&neigh->probes);
336 struct in_device *in_dev;
337 struct dst_entry *dst = NULL;
339 rcu_read_lock();
340 in_dev = __in_dev_get_rcu(dev);
341 if (!in_dev) {
342 rcu_read_unlock();
343 return;
345 switch (IN_DEV_ARP_ANNOUNCE(in_dev)) {
346 default:
347 case 0: /* By default announce any local IP */
348 if (skb && inet_addr_type_dev_table(dev_net(dev), dev,
349 ip_hdr(skb)->saddr) == RTN_LOCAL)
350 saddr = ip_hdr(skb)->saddr;
351 break;
352 case 1: /* Restrict announcements of saddr in same subnet */
353 if (!skb)
354 break;
355 saddr = ip_hdr(skb)->saddr;
356 if (inet_addr_type_dev_table(dev_net(dev), dev,
357 saddr) == RTN_LOCAL) {
358 /* saddr should be known to target */
359 if (inet_addr_onlink(in_dev, target, saddr))
360 break;
362 saddr = 0;
363 break;
364 case 2: /* Avoid secondary IPs, get a primary/preferred one */
365 break;
367 rcu_read_unlock();
369 if (!saddr)
370 saddr = inet_select_addr(dev, target, RT_SCOPE_LINK);
372 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
373 if (probes < 0) {
374 if (!(neigh->nud_state & NUD_VALID))
375 pr_debug("trying to ucast probe in NUD_INVALID\n");
376 neigh_ha_snapshot(dst_ha, neigh, dev);
377 dst_hw = dst_ha;
378 } else {
379 probes -= NEIGH_VAR(neigh->parms, APP_PROBES);
380 if (probes < 0) {
381 neigh_app_ns(neigh);
382 return;
386 if (skb && !(dev->priv_flags & IFF_XMIT_DST_RELEASE))
387 dst = skb_dst(skb);
388 arp_send_dst(ARPOP_REQUEST, ETH_P_ARP, target, dev, saddr,
389 dst_hw, dev->dev_addr, NULL, dst);
392 static int arp_ignore(struct in_device *in_dev, __be32 sip, __be32 tip)
394 struct net *net = dev_net(in_dev->dev);
395 int scope;
397 switch (IN_DEV_ARP_IGNORE(in_dev)) {
398 case 0: /* Reply, the tip is already validated */
399 return 0;
400 case 1: /* Reply only if tip is configured on the incoming interface */
401 sip = 0;
402 scope = RT_SCOPE_HOST;
403 break;
404 case 2: /*
405 * Reply only if tip is configured on the incoming interface
406 * and is in same subnet as sip
408 scope = RT_SCOPE_HOST;
409 break;
410 case 3: /* Do not reply for scope host addresses */
411 sip = 0;
412 scope = RT_SCOPE_LINK;
413 in_dev = NULL;
414 break;
415 case 4: /* Reserved */
416 case 5:
417 case 6:
418 case 7:
419 return 0;
420 case 8: /* Do not reply */
421 return 1;
422 default:
423 return 0;
425 return !inet_confirm_addr(net, in_dev, sip, tip, scope);
428 static int arp_filter(__be32 sip, __be32 tip, struct net_device *dev)
430 struct rtable *rt;
431 int flag = 0;
432 /*unsigned long now; */
433 struct net *net = dev_net(dev);
435 rt = ip_route_output(net, sip, tip, 0, 0);
436 if (IS_ERR(rt))
437 return 1;
438 if (rt->dst.dev != dev) {
439 __NET_INC_STATS(net, LINUX_MIB_ARPFILTER);
440 flag = 1;
442 ip_rt_put(rt);
443 return flag;
447 * Check if we can use proxy ARP for this path
449 static inline int arp_fwd_proxy(struct in_device *in_dev,
450 struct net_device *dev, struct rtable *rt)
452 struct in_device *out_dev;
453 int imi, omi = -1;
455 if (rt->dst.dev == dev)
456 return 0;
458 if (!IN_DEV_PROXY_ARP(in_dev))
459 return 0;
460 imi = IN_DEV_MEDIUM_ID(in_dev);
461 if (imi == 0)
462 return 1;
463 if (imi == -1)
464 return 0;
466 /* place to check for proxy_arp for routes */
468 out_dev = __in_dev_get_rcu(rt->dst.dev);
469 if (out_dev)
470 omi = IN_DEV_MEDIUM_ID(out_dev);
472 return omi != imi && omi != -1;
476 * Check for RFC3069 proxy arp private VLAN (allow to send back to same dev)
478 * RFC3069 supports proxy arp replies back to the same interface. This
479 * is done to support (ethernet) switch features, like RFC 3069, where
480 * the individual ports are not allowed to communicate with each
481 * other, BUT they are allowed to talk to the upstream router. As
482 * described in RFC 3069, it is possible to allow these hosts to
483 * communicate through the upstream router, by proxy_arp'ing.
485 * RFC 3069: "VLAN Aggregation for Efficient IP Address Allocation"
487 * This technology is known by different names:
488 * In RFC 3069 it is called VLAN Aggregation.
489 * Cisco and Allied Telesyn call it Private VLAN.
490 * Hewlett-Packard call it Source-Port filtering or port-isolation.
491 * Ericsson call it MAC-Forced Forwarding (RFC Draft).
494 static inline int arp_fwd_pvlan(struct in_device *in_dev,
495 struct net_device *dev, struct rtable *rt,
496 __be32 sip, __be32 tip)
498 /* Private VLAN is only concerned about the same ethernet segment */
499 if (rt->dst.dev != dev)
500 return 0;
502 /* Don't reply on self probes (often done by windowz boxes)*/
503 if (sip == tip)
504 return 0;
506 if (IN_DEV_PROXY_ARP_PVLAN(in_dev))
507 return 1;
508 else
509 return 0;
513 * Interface to link layer: send routine and receive handler.
517 * Create an arp packet. If dest_hw is not set, we create a broadcast
518 * message.
520 struct sk_buff *arp_create(int type, int ptype, __be32 dest_ip,
521 struct net_device *dev, __be32 src_ip,
522 const unsigned char *dest_hw,
523 const unsigned char *src_hw,
524 const unsigned char *target_hw)
526 struct sk_buff *skb;
527 struct arphdr *arp;
528 unsigned char *arp_ptr;
529 int hlen = LL_RESERVED_SPACE(dev);
530 int tlen = dev->needed_tailroom;
533 * Allocate a buffer
536 skb = alloc_skb(arp_hdr_len(dev) + hlen + tlen, GFP_ATOMIC);
537 if (!skb)
538 return NULL;
540 skb_reserve(skb, hlen);
541 skb_reset_network_header(skb);
542 arp = (struct arphdr *) skb_put(skb, arp_hdr_len(dev));
543 skb->dev = dev;
544 skb->protocol = htons(ETH_P_ARP);
545 if (!src_hw)
546 src_hw = dev->dev_addr;
547 if (!dest_hw)
548 dest_hw = dev->broadcast;
551 * Fill the device header for the ARP frame
553 if (dev_hard_header(skb, dev, ptype, dest_hw, src_hw, skb->len) < 0)
554 goto out;
557 * Fill out the arp protocol part.
559 * The arp hardware type should match the device type, except for FDDI,
560 * which (according to RFC 1390) should always equal 1 (Ethernet).
563 * Exceptions everywhere. AX.25 uses the AX.25 PID value not the
564 * DIX code for the protocol. Make these device structure fields.
566 switch (dev->type) {
567 default:
568 arp->ar_hrd = htons(dev->type);
569 arp->ar_pro = htons(ETH_P_IP);
570 break;
572 #if IS_ENABLED(CONFIG_AX25)
573 case ARPHRD_AX25:
574 arp->ar_hrd = htons(ARPHRD_AX25);
575 arp->ar_pro = htons(AX25_P_IP);
576 break;
578 #if IS_ENABLED(CONFIG_NETROM)
579 case ARPHRD_NETROM:
580 arp->ar_hrd = htons(ARPHRD_NETROM);
581 arp->ar_pro = htons(AX25_P_IP);
582 break;
583 #endif
584 #endif
586 #if IS_ENABLED(CONFIG_FDDI)
587 case ARPHRD_FDDI:
588 arp->ar_hrd = htons(ARPHRD_ETHER);
589 arp->ar_pro = htons(ETH_P_IP);
590 break;
591 #endif
594 arp->ar_hln = dev->addr_len;
595 arp->ar_pln = 4;
596 arp->ar_op = htons(type);
598 arp_ptr = (unsigned char *)(arp + 1);
600 memcpy(arp_ptr, src_hw, dev->addr_len);
601 arp_ptr += dev->addr_len;
602 memcpy(arp_ptr, &src_ip, 4);
603 arp_ptr += 4;
605 switch (dev->type) {
606 #if IS_ENABLED(CONFIG_FIREWIRE_NET)
607 case ARPHRD_IEEE1394:
608 break;
609 #endif
610 default:
611 if (target_hw)
612 memcpy(arp_ptr, target_hw, dev->addr_len);
613 else
614 memset(arp_ptr, 0, dev->addr_len);
615 arp_ptr += dev->addr_len;
617 memcpy(arp_ptr, &dest_ip, 4);
619 return skb;
621 out:
622 kfree_skb(skb);
623 return NULL;
625 EXPORT_SYMBOL(arp_create);
627 static int arp_xmit_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
629 return dev_queue_xmit(skb);
633 * Send an arp packet.
635 void arp_xmit(struct sk_buff *skb)
637 /* Send it off, maybe filter it using firewalling first. */
638 NF_HOOK(NFPROTO_ARP, NF_ARP_OUT,
639 dev_net(skb->dev), NULL, skb, NULL, skb->dev,
640 arp_xmit_finish);
642 EXPORT_SYMBOL(arp_xmit);
645 * Process an arp request.
648 static int arp_process(struct net *net, struct sock *sk, struct sk_buff *skb)
650 struct net_device *dev = skb->dev;
651 struct in_device *in_dev = __in_dev_get_rcu(dev);
652 struct arphdr *arp;
653 unsigned char *arp_ptr;
654 struct rtable *rt;
655 unsigned char *sha;
656 __be32 sip, tip;
657 u16 dev_type = dev->type;
658 int addr_type;
659 struct neighbour *n;
660 struct dst_entry *reply_dst = NULL;
661 bool is_garp = false;
663 /* arp_rcv below verifies the ARP header and verifies the device
664 * is ARP'able.
667 if (!in_dev)
668 goto out_free_skb;
670 arp = arp_hdr(skb);
672 switch (dev_type) {
673 default:
674 if (arp->ar_pro != htons(ETH_P_IP) ||
675 htons(dev_type) != arp->ar_hrd)
676 goto out_free_skb;
677 break;
678 case ARPHRD_ETHER:
679 case ARPHRD_FDDI:
680 case ARPHRD_IEEE802:
682 * ETHERNET, and Fibre Channel (which are IEEE 802
683 * devices, according to RFC 2625) devices will accept ARP
684 * hardware types of either 1 (Ethernet) or 6 (IEEE 802.2).
685 * This is the case also of FDDI, where the RFC 1390 says that
686 * FDDI devices should accept ARP hardware of (1) Ethernet,
687 * however, to be more robust, we'll accept both 1 (Ethernet)
688 * or 6 (IEEE 802.2)
690 if ((arp->ar_hrd != htons(ARPHRD_ETHER) &&
691 arp->ar_hrd != htons(ARPHRD_IEEE802)) ||
692 arp->ar_pro != htons(ETH_P_IP))
693 goto out_free_skb;
694 break;
695 case ARPHRD_AX25:
696 if (arp->ar_pro != htons(AX25_P_IP) ||
697 arp->ar_hrd != htons(ARPHRD_AX25))
698 goto out_free_skb;
699 break;
700 case ARPHRD_NETROM:
701 if (arp->ar_pro != htons(AX25_P_IP) ||
702 arp->ar_hrd != htons(ARPHRD_NETROM))
703 goto out_free_skb;
704 break;
707 /* Understand only these message types */
709 if (arp->ar_op != htons(ARPOP_REPLY) &&
710 arp->ar_op != htons(ARPOP_REQUEST))
711 goto out_free_skb;
714 * Extract fields
716 arp_ptr = (unsigned char *)(arp + 1);
717 sha = arp_ptr;
718 arp_ptr += dev->addr_len;
719 memcpy(&sip, arp_ptr, 4);
720 arp_ptr += 4;
721 switch (dev_type) {
722 #if IS_ENABLED(CONFIG_FIREWIRE_NET)
723 case ARPHRD_IEEE1394:
724 break;
725 #endif
726 default:
727 arp_ptr += dev->addr_len;
729 memcpy(&tip, arp_ptr, 4);
731 * Check for bad requests for 127.x.x.x and requests for multicast
732 * addresses. If this is one such, delete it.
734 if (ipv4_is_multicast(tip) ||
735 (!IN_DEV_ROUTE_LOCALNET(in_dev) && ipv4_is_loopback(tip)))
736 goto out_free_skb;
739 * For some 802.11 wireless deployments (and possibly other networks),
740 * there will be an ARP proxy and gratuitous ARP frames are attacks
741 * and thus should not be accepted.
743 if (sip == tip && IN_DEV_ORCONF(in_dev, DROP_GRATUITOUS_ARP))
744 goto out_free_skb;
747 * Special case: We must set Frame Relay source Q.922 address
749 if (dev_type == ARPHRD_DLCI)
750 sha = dev->broadcast;
753 * Process entry. The idea here is we want to send a reply if it is a
754 * request for us or if it is a request for someone else that we hold
755 * a proxy for. We want to add an entry to our cache if it is a reply
756 * to us or if it is a request for our address.
757 * (The assumption for this last is that if someone is requesting our
758 * address, they are probably intending to talk to us, so it saves time
759 * if we cache their address. Their address is also probably not in
760 * our cache, since ours is not in their cache.)
762 * Putting this another way, we only care about replies if they are to
763 * us, in which case we add them to the cache. For requests, we care
764 * about those for us and those for our proxies. We reply to both,
765 * and in the case of requests for us we add the requester to the arp
766 * cache.
769 if (arp->ar_op == htons(ARPOP_REQUEST) && skb_metadata_dst(skb))
770 reply_dst = (struct dst_entry *)
771 iptunnel_metadata_reply(skb_metadata_dst(skb),
772 GFP_ATOMIC);
774 /* Special case: IPv4 duplicate address detection packet (RFC2131) */
775 if (sip == 0) {
776 if (arp->ar_op == htons(ARPOP_REQUEST) &&
777 inet_addr_type_dev_table(net, dev, tip) == RTN_LOCAL &&
778 !arp_ignore(in_dev, sip, tip))
779 arp_send_dst(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip,
780 sha, dev->dev_addr, sha, reply_dst);
781 goto out_consume_skb;
784 if (arp->ar_op == htons(ARPOP_REQUEST) &&
785 ip_route_input_noref(skb, tip, sip, 0, dev) == 0) {
787 rt = skb_rtable(skb);
788 addr_type = rt->rt_type;
790 if (addr_type == RTN_LOCAL) {
791 int dont_send;
793 dont_send = arp_ignore(in_dev, sip, tip);
794 if (!dont_send && IN_DEV_ARPFILTER(in_dev))
795 dont_send = arp_filter(sip, tip, dev);
796 if (!dont_send) {
797 n = neigh_event_ns(&arp_tbl, sha, &sip, dev);
798 if (n) {
799 arp_send_dst(ARPOP_REPLY, ETH_P_ARP,
800 sip, dev, tip, sha,
801 dev->dev_addr, sha,
802 reply_dst);
803 neigh_release(n);
806 goto out_consume_skb;
807 } else if (IN_DEV_FORWARD(in_dev)) {
808 if (addr_type == RTN_UNICAST &&
809 (arp_fwd_proxy(in_dev, dev, rt) ||
810 arp_fwd_pvlan(in_dev, dev, rt, sip, tip) ||
811 (rt->dst.dev != dev &&
812 pneigh_lookup(&arp_tbl, net, &tip, dev, 0)))) {
813 n = neigh_event_ns(&arp_tbl, sha, &sip, dev);
814 if (n)
815 neigh_release(n);
817 if (NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED ||
818 skb->pkt_type == PACKET_HOST ||
819 NEIGH_VAR(in_dev->arp_parms, PROXY_DELAY) == 0) {
820 arp_send_dst(ARPOP_REPLY, ETH_P_ARP,
821 sip, dev, tip, sha,
822 dev->dev_addr, sha,
823 reply_dst);
824 } else {
825 pneigh_enqueue(&arp_tbl,
826 in_dev->arp_parms, skb);
827 goto out_free_dst;
829 goto out_consume_skb;
834 /* Update our ARP tables */
836 n = __neigh_lookup(&arp_tbl, &sip, dev, 0);
838 if (IN_DEV_ARP_ACCEPT(in_dev)) {
839 unsigned int addr_type = inet_addr_type_dev_table(net, dev, sip);
841 /* Unsolicited ARP is not accepted by default.
842 It is possible, that this option should be enabled for some
843 devices (strip is candidate)
845 is_garp = arp->ar_op == htons(ARPOP_REQUEST) && tip == sip &&
846 addr_type == RTN_UNICAST;
848 if (!n &&
849 ((arp->ar_op == htons(ARPOP_REPLY) &&
850 addr_type == RTN_UNICAST) || is_garp))
851 n = __neigh_lookup(&arp_tbl, &sip, dev, 1);
854 if (n) {
855 int state = NUD_REACHABLE;
856 int override;
858 /* If several different ARP replies follows back-to-back,
859 use the FIRST one. It is possible, if several proxy
860 agents are active. Taking the first reply prevents
861 arp trashing and chooses the fastest router.
863 override = time_after(jiffies,
864 n->updated +
865 NEIGH_VAR(n->parms, LOCKTIME)) ||
866 is_garp;
868 /* Broadcast replies and request packets
869 do not assert neighbour reachability.
871 if (arp->ar_op != htons(ARPOP_REPLY) ||
872 skb->pkt_type != PACKET_HOST)
873 state = NUD_STALE;
874 neigh_update(n, sha, state,
875 override ? NEIGH_UPDATE_F_OVERRIDE : 0);
876 neigh_release(n);
879 out_consume_skb:
880 consume_skb(skb);
882 out_free_dst:
883 dst_release(reply_dst);
884 return NET_RX_SUCCESS;
886 out_free_skb:
887 kfree_skb(skb);
888 return NET_RX_DROP;
891 static void parp_redo(struct sk_buff *skb)
893 arp_process(dev_net(skb->dev), NULL, skb);
898 * Receive an arp request from the device layer.
901 static int arp_rcv(struct sk_buff *skb, struct net_device *dev,
902 struct packet_type *pt, struct net_device *orig_dev)
904 const struct arphdr *arp;
906 /* do not tweak dropwatch on an ARP we will ignore */
907 if (dev->flags & IFF_NOARP ||
908 skb->pkt_type == PACKET_OTHERHOST ||
909 skb->pkt_type == PACKET_LOOPBACK)
910 goto consumeskb;
912 skb = skb_share_check(skb, GFP_ATOMIC);
913 if (!skb)
914 goto out_of_mem;
916 /* ARP header, plus 2 device addresses, plus 2 IP addresses. */
917 if (!pskb_may_pull(skb, arp_hdr_len(dev)))
918 goto freeskb;
920 arp = arp_hdr(skb);
921 if (arp->ar_hln != dev->addr_len || arp->ar_pln != 4)
922 goto freeskb;
924 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
926 return NF_HOOK(NFPROTO_ARP, NF_ARP_IN,
927 dev_net(dev), NULL, skb, dev, NULL,
928 arp_process);
930 consumeskb:
931 consume_skb(skb);
932 return NET_RX_SUCCESS;
933 freeskb:
934 kfree_skb(skb);
935 out_of_mem:
936 return NET_RX_DROP;
940 * User level interface (ioctl)
944 * Set (create) an ARP cache entry.
947 static int arp_req_set_proxy(struct net *net, struct net_device *dev, int on)
949 if (!dev) {
950 IPV4_DEVCONF_ALL(net, PROXY_ARP) = on;
951 return 0;
953 if (__in_dev_get_rtnl(dev)) {
954 IN_DEV_CONF_SET(__in_dev_get_rtnl(dev), PROXY_ARP, on);
955 return 0;
957 return -ENXIO;
960 static int arp_req_set_public(struct net *net, struct arpreq *r,
961 struct net_device *dev)
963 __be32 ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
964 __be32 mask = ((struct sockaddr_in *)&r->arp_netmask)->sin_addr.s_addr;
966 if (mask && mask != htonl(0xFFFFFFFF))
967 return -EINVAL;
968 if (!dev && (r->arp_flags & ATF_COM)) {
969 dev = dev_getbyhwaddr_rcu(net, r->arp_ha.sa_family,
970 r->arp_ha.sa_data);
971 if (!dev)
972 return -ENODEV;
974 if (mask) {
975 if (!pneigh_lookup(&arp_tbl, net, &ip, dev, 1))
976 return -ENOBUFS;
977 return 0;
980 return arp_req_set_proxy(net, dev, 1);
983 static int arp_req_set(struct net *net, struct arpreq *r,
984 struct net_device *dev)
986 __be32 ip;
987 struct neighbour *neigh;
988 int err;
990 if (r->arp_flags & ATF_PUBL)
991 return arp_req_set_public(net, r, dev);
993 ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
994 if (r->arp_flags & ATF_PERM)
995 r->arp_flags |= ATF_COM;
996 if (!dev) {
997 struct rtable *rt = ip_route_output(net, ip, 0, RTO_ONLINK, 0);
999 if (IS_ERR(rt))
1000 return PTR_ERR(rt);
1001 dev = rt->dst.dev;
1002 ip_rt_put(rt);
1003 if (!dev)
1004 return -EINVAL;
1006 switch (dev->type) {
1007 #if IS_ENABLED(CONFIG_FDDI)
1008 case ARPHRD_FDDI:
1010 * According to RFC 1390, FDDI devices should accept ARP
1011 * hardware types of 1 (Ethernet). However, to be more
1012 * robust, we'll accept hardware types of either 1 (Ethernet)
1013 * or 6 (IEEE 802.2).
1015 if (r->arp_ha.sa_family != ARPHRD_FDDI &&
1016 r->arp_ha.sa_family != ARPHRD_ETHER &&
1017 r->arp_ha.sa_family != ARPHRD_IEEE802)
1018 return -EINVAL;
1019 break;
1020 #endif
1021 default:
1022 if (r->arp_ha.sa_family != dev->type)
1023 return -EINVAL;
1024 break;
1027 neigh = __neigh_lookup_errno(&arp_tbl, &ip, dev);
1028 err = PTR_ERR(neigh);
1029 if (!IS_ERR(neigh)) {
1030 unsigned int state = NUD_STALE;
1031 if (r->arp_flags & ATF_PERM)
1032 state = NUD_PERMANENT;
1033 err = neigh_update(neigh, (r->arp_flags & ATF_COM) ?
1034 r->arp_ha.sa_data : NULL, state,
1035 NEIGH_UPDATE_F_OVERRIDE |
1036 NEIGH_UPDATE_F_ADMIN);
1037 neigh_release(neigh);
1039 return err;
1042 static unsigned int arp_state_to_flags(struct neighbour *neigh)
1044 if (neigh->nud_state&NUD_PERMANENT)
1045 return ATF_PERM | ATF_COM;
1046 else if (neigh->nud_state&NUD_VALID)
1047 return ATF_COM;
1048 else
1049 return 0;
1053 * Get an ARP cache entry.
1056 static int arp_req_get(struct arpreq *r, struct net_device *dev)
1058 __be32 ip = ((struct sockaddr_in *) &r->arp_pa)->sin_addr.s_addr;
1059 struct neighbour *neigh;
1060 int err = -ENXIO;
1062 neigh = neigh_lookup(&arp_tbl, &ip, dev);
1063 if (neigh) {
1064 if (!(neigh->nud_state & NUD_NOARP)) {
1065 read_lock_bh(&neigh->lock);
1066 memcpy(r->arp_ha.sa_data, neigh->ha, dev->addr_len);
1067 r->arp_flags = arp_state_to_flags(neigh);
1068 read_unlock_bh(&neigh->lock);
1069 r->arp_ha.sa_family = dev->type;
1070 strlcpy(r->arp_dev, dev->name, sizeof(r->arp_dev));
1071 err = 0;
1073 neigh_release(neigh);
1075 return err;
1078 static int arp_invalidate(struct net_device *dev, __be32 ip)
1080 struct neighbour *neigh = neigh_lookup(&arp_tbl, &ip, dev);
1081 int err = -ENXIO;
1083 if (neigh) {
1084 if (neigh->nud_state & ~NUD_NOARP)
1085 err = neigh_update(neigh, NULL, NUD_FAILED,
1086 NEIGH_UPDATE_F_OVERRIDE|
1087 NEIGH_UPDATE_F_ADMIN);
1088 neigh_release(neigh);
1091 return err;
1094 static int arp_req_delete_public(struct net *net, struct arpreq *r,
1095 struct net_device *dev)
1097 __be32 ip = ((struct sockaddr_in *) &r->arp_pa)->sin_addr.s_addr;
1098 __be32 mask = ((struct sockaddr_in *)&r->arp_netmask)->sin_addr.s_addr;
1100 if (mask == htonl(0xFFFFFFFF))
1101 return pneigh_delete(&arp_tbl, net, &ip, dev);
1103 if (mask)
1104 return -EINVAL;
1106 return arp_req_set_proxy(net, dev, 0);
1109 static int arp_req_delete(struct net *net, struct arpreq *r,
1110 struct net_device *dev)
1112 __be32 ip;
1114 if (r->arp_flags & ATF_PUBL)
1115 return arp_req_delete_public(net, r, dev);
1117 ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
1118 if (!dev) {
1119 struct rtable *rt = ip_route_output(net, ip, 0, RTO_ONLINK, 0);
1120 if (IS_ERR(rt))
1121 return PTR_ERR(rt);
1122 dev = rt->dst.dev;
1123 ip_rt_put(rt);
1124 if (!dev)
1125 return -EINVAL;
1127 return arp_invalidate(dev, ip);
1131 * Handle an ARP layer I/O control request.
1134 int arp_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1136 int err;
1137 struct arpreq r;
1138 struct net_device *dev = NULL;
1140 switch (cmd) {
1141 case SIOCDARP:
1142 case SIOCSARP:
1143 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1144 return -EPERM;
1145 case SIOCGARP:
1146 err = copy_from_user(&r, arg, sizeof(struct arpreq));
1147 if (err)
1148 return -EFAULT;
1149 break;
1150 default:
1151 return -EINVAL;
1154 if (r.arp_pa.sa_family != AF_INET)
1155 return -EPFNOSUPPORT;
1157 if (!(r.arp_flags & ATF_PUBL) &&
1158 (r.arp_flags & (ATF_NETMASK | ATF_DONTPUB)))
1159 return -EINVAL;
1160 if (!(r.arp_flags & ATF_NETMASK))
1161 ((struct sockaddr_in *)&r.arp_netmask)->sin_addr.s_addr =
1162 htonl(0xFFFFFFFFUL);
1163 rtnl_lock();
1164 if (r.arp_dev[0]) {
1165 err = -ENODEV;
1166 dev = __dev_get_by_name(net, r.arp_dev);
1167 if (!dev)
1168 goto out;
1170 /* Mmmm... It is wrong... ARPHRD_NETROM==0 */
1171 if (!r.arp_ha.sa_family)
1172 r.arp_ha.sa_family = dev->type;
1173 err = -EINVAL;
1174 if ((r.arp_flags & ATF_COM) && r.arp_ha.sa_family != dev->type)
1175 goto out;
1176 } else if (cmd == SIOCGARP) {
1177 err = -ENODEV;
1178 goto out;
1181 switch (cmd) {
1182 case SIOCDARP:
1183 err = arp_req_delete(net, &r, dev);
1184 break;
1185 case SIOCSARP:
1186 err = arp_req_set(net, &r, dev);
1187 break;
1188 case SIOCGARP:
1189 err = arp_req_get(&r, dev);
1190 break;
1192 out:
1193 rtnl_unlock();
1194 if (cmd == SIOCGARP && !err && copy_to_user(arg, &r, sizeof(r)))
1195 err = -EFAULT;
1196 return err;
1199 static int arp_netdev_event(struct notifier_block *this, unsigned long event,
1200 void *ptr)
1202 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1203 struct netdev_notifier_change_info *change_info;
1205 switch (event) {
1206 case NETDEV_CHANGEADDR:
1207 neigh_changeaddr(&arp_tbl, dev);
1208 rt_cache_flush(dev_net(dev));
1209 break;
1210 case NETDEV_CHANGE:
1211 change_info = ptr;
1212 if (change_info->flags_changed & IFF_NOARP)
1213 neigh_changeaddr(&arp_tbl, dev);
1214 break;
1215 default:
1216 break;
1219 return NOTIFY_DONE;
1222 static struct notifier_block arp_netdev_notifier = {
1223 .notifier_call = arp_netdev_event,
1226 /* Note, that it is not on notifier chain.
1227 It is necessary, that this routine was called after route cache will be
1228 flushed.
1230 void arp_ifdown(struct net_device *dev)
1232 neigh_ifdown(&arp_tbl, dev);
1237 * Called once on startup.
1240 static struct packet_type arp_packet_type __read_mostly = {
1241 .type = cpu_to_be16(ETH_P_ARP),
1242 .func = arp_rcv,
1245 static int arp_proc_init(void);
1247 void __init arp_init(void)
1249 neigh_table_init(NEIGH_ARP_TABLE, &arp_tbl);
1251 dev_add_pack(&arp_packet_type);
1252 arp_proc_init();
1253 #ifdef CONFIG_SYSCTL
1254 neigh_sysctl_register(NULL, &arp_tbl.parms, NULL);
1255 #endif
1256 register_netdevice_notifier(&arp_netdev_notifier);
1259 #ifdef CONFIG_PROC_FS
1260 #if IS_ENABLED(CONFIG_AX25)
1262 /* ------------------------------------------------------------------------ */
1264 * ax25 -> ASCII conversion
1266 static char *ax2asc2(ax25_address *a, char *buf)
1268 char c, *s;
1269 int n;
1271 for (n = 0, s = buf; n < 6; n++) {
1272 c = (a->ax25_call[n] >> 1) & 0x7F;
1274 if (c != ' ')
1275 *s++ = c;
1278 *s++ = '-';
1279 n = (a->ax25_call[6] >> 1) & 0x0F;
1280 if (n > 9) {
1281 *s++ = '1';
1282 n -= 10;
1285 *s++ = n + '0';
1286 *s++ = '\0';
1288 if (*buf == '\0' || *buf == '-')
1289 return "*";
1291 return buf;
1293 #endif /* CONFIG_AX25 */
1295 #define HBUFFERLEN 30
1297 static void arp_format_neigh_entry(struct seq_file *seq,
1298 struct neighbour *n)
1300 char hbuffer[HBUFFERLEN];
1301 int k, j;
1302 char tbuf[16];
1303 struct net_device *dev = n->dev;
1304 int hatype = dev->type;
1306 read_lock(&n->lock);
1307 /* Convert hardware address to XX:XX:XX:XX ... form. */
1308 #if IS_ENABLED(CONFIG_AX25)
1309 if (hatype == ARPHRD_AX25 || hatype == ARPHRD_NETROM)
1310 ax2asc2((ax25_address *)n->ha, hbuffer);
1311 else {
1312 #endif
1313 for (k = 0, j = 0; k < HBUFFERLEN - 3 && j < dev->addr_len; j++) {
1314 hbuffer[k++] = hex_asc_hi(n->ha[j]);
1315 hbuffer[k++] = hex_asc_lo(n->ha[j]);
1316 hbuffer[k++] = ':';
1318 if (k != 0)
1319 --k;
1320 hbuffer[k] = 0;
1321 #if IS_ENABLED(CONFIG_AX25)
1323 #endif
1324 sprintf(tbuf, "%pI4", n->primary_key);
1325 seq_printf(seq, "%-16s 0x%-10x0x%-10x%s * %s\n",
1326 tbuf, hatype, arp_state_to_flags(n), hbuffer, dev->name);
1327 read_unlock(&n->lock);
1330 static void arp_format_pneigh_entry(struct seq_file *seq,
1331 struct pneigh_entry *n)
1333 struct net_device *dev = n->dev;
1334 int hatype = dev ? dev->type : 0;
1335 char tbuf[16];
1337 sprintf(tbuf, "%pI4", n->key);
1338 seq_printf(seq, "%-16s 0x%-10x0x%-10x%s * %s\n",
1339 tbuf, hatype, ATF_PUBL | ATF_PERM, "00:00:00:00:00:00",
1340 dev ? dev->name : "*");
1343 static int arp_seq_show(struct seq_file *seq, void *v)
1345 if (v == SEQ_START_TOKEN) {
1346 seq_puts(seq, "IP address HW type Flags "
1347 "HW address Mask Device\n");
1348 } else {
1349 struct neigh_seq_state *state = seq->private;
1351 if (state->flags & NEIGH_SEQ_IS_PNEIGH)
1352 arp_format_pneigh_entry(seq, v);
1353 else
1354 arp_format_neigh_entry(seq, v);
1357 return 0;
1360 static void *arp_seq_start(struct seq_file *seq, loff_t *pos)
1362 /* Don't want to confuse "arp -a" w/ magic entries,
1363 * so we tell the generic iterator to skip NUD_NOARP.
1365 return neigh_seq_start(seq, pos, &arp_tbl, NEIGH_SEQ_SKIP_NOARP);
1368 /* ------------------------------------------------------------------------ */
1370 static const struct seq_operations arp_seq_ops = {
1371 .start = arp_seq_start,
1372 .next = neigh_seq_next,
1373 .stop = neigh_seq_stop,
1374 .show = arp_seq_show,
1377 static int arp_seq_open(struct inode *inode, struct file *file)
1379 return seq_open_net(inode, file, &arp_seq_ops,
1380 sizeof(struct neigh_seq_state));
1383 static const struct file_operations arp_seq_fops = {
1384 .owner = THIS_MODULE,
1385 .open = arp_seq_open,
1386 .read = seq_read,
1387 .llseek = seq_lseek,
1388 .release = seq_release_net,
1392 static int __net_init arp_net_init(struct net *net)
1394 if (!proc_create("arp", S_IRUGO, net->proc_net, &arp_seq_fops))
1395 return -ENOMEM;
1396 return 0;
1399 static void __net_exit arp_net_exit(struct net *net)
1401 remove_proc_entry("arp", net->proc_net);
1404 static struct pernet_operations arp_net_ops = {
1405 .init = arp_net_init,
1406 .exit = arp_net_exit,
1409 static int __init arp_proc_init(void)
1411 return register_pernet_subsys(&arp_net_ops);
1414 #else /* CONFIG_PROC_FS */
1416 static int __init arp_proc_init(void)
1418 return 0;
1421 #endif /* CONFIG_PROC_FS */