2 * NET3 IP device support routines.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
9 * Derived from the IP parts of dev.c 1.0.19
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Mark Evans, <evansmp@uhura.aston.ac.uk>
15 * Alan Cox, <gw4pts@gw4pts.ampr.org>
16 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
19 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
21 * Cyrus Durgin: updated for kmod
22 * Matthias Andree: in devinet_ioctl, compare label and
23 * address (4.4BSD alias style support),
24 * fall back to comparing just the label
29 #include <asm/uaccess.h>
30 #include <linux/bitops.h>
31 #include <linux/capability.h>
32 #include <linux/module.h>
33 #include <linux/types.h>
34 #include <linux/kernel.h>
35 #include <linux/string.h>
37 #include <linux/socket.h>
38 #include <linux/sockios.h>
40 #include <linux/errno.h>
41 #include <linux/interrupt.h>
42 #include <linux/if_addr.h>
43 #include <linux/if_ether.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <linux/etherdevice.h>
47 #include <linux/skbuff.h>
48 #include <linux/init.h>
49 #include <linux/notifier.h>
50 #include <linux/inetdevice.h>
51 #include <linux/igmp.h>
52 #include <linux/slab.h>
53 #include <linux/hash.h>
55 #include <linux/sysctl.h>
57 #include <linux/kmod.h>
58 #include <linux/netconf.h>
62 #include <net/route.h>
63 #include <net/ip_fib.h>
64 #include <net/rtnetlink.h>
65 #include <net/net_namespace.h>
66 #include <net/addrconf.h>
68 #include "fib_lookup.h"
70 static struct ipv4_devconf ipv4_devconf = {
72 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
73 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
74 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
75 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
79 static struct ipv4_devconf ipv4_devconf_dflt = {
81 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
82 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
83 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
85 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
89 #define IPV4_DEVCONF_DFLT(net, attr) \
90 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
92 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
93 [IFA_LOCAL] = { .type = NLA_U32 },
94 [IFA_ADDRESS] = { .type = NLA_U32 },
95 [IFA_BROADCAST] = { .type = NLA_U32 },
96 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
97 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
100 #define IN4_ADDR_HSIZE_SHIFT 8
101 #define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
103 static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
104 static DEFINE_SPINLOCK(inet_addr_hash_lock);
106 static u32 inet_addr_hash(struct net *net, __be32 addr)
108 u32 val = (__force u32) addr ^ net_hash_mix(net);
110 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
113 static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
115 u32 hash = inet_addr_hash(net, ifa->ifa_local);
117 spin_lock(&inet_addr_hash_lock);
118 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
119 spin_unlock(&inet_addr_hash_lock);
122 static void inet_hash_remove(struct in_ifaddr *ifa)
124 spin_lock(&inet_addr_hash_lock);
125 hlist_del_init_rcu(&ifa->hash);
126 spin_unlock(&inet_addr_hash_lock);
130 * __ip_dev_find - find the first device with a given source address.
131 * @net: the net namespace
132 * @addr: the source address
133 * @devref: if true, take a reference on the found device
135 * If a caller uses devref=false, it should be protected by RCU, or RTNL
137 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
139 u32 hash = inet_addr_hash(net, addr);
140 struct net_device *result = NULL;
141 struct in_ifaddr *ifa;
144 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash) {
145 if (ifa->ifa_local == addr) {
146 struct net_device *dev = ifa->ifa_dev->dev;
148 if (!net_eq(dev_net(dev), net))
155 struct flowi4 fl4 = { .daddr = addr };
156 struct fib_result res = { 0 };
157 struct fib_table *local;
159 /* Fallback to FIB local table so that communication
160 * over loopback subnets work.
162 local = fib_get_table(net, RT_TABLE_LOCAL);
164 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
165 res.type == RTN_LOCAL)
166 result = FIB_RES_DEV(res);
168 if (result && devref)
173 EXPORT_SYMBOL(__ip_dev_find);
175 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
177 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
178 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
181 static void devinet_sysctl_register(struct in_device *idev);
182 static void devinet_sysctl_unregister(struct in_device *idev);
184 static void devinet_sysctl_register(struct in_device *idev)
187 static void devinet_sysctl_unregister(struct in_device *idev)
192 /* Locks all the inet devices. */
194 static struct in_ifaddr *inet_alloc_ifa(void)
196 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
199 static void inet_rcu_free_ifa(struct rcu_head *head)
201 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
203 in_dev_put(ifa->ifa_dev);
207 static void inet_free_ifa(struct in_ifaddr *ifa)
209 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
212 void in_dev_finish_destroy(struct in_device *idev)
214 struct net_device *dev = idev->dev;
216 WARN_ON(idev->ifa_list);
217 WARN_ON(idev->mc_list);
218 #ifdef NET_REFCNT_DEBUG
219 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
223 pr_err("Freeing alive in_device %p\n", idev);
227 EXPORT_SYMBOL(in_dev_finish_destroy);
229 static struct in_device *inetdev_init(struct net_device *dev)
231 struct in_device *in_dev;
235 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
238 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
239 sizeof(in_dev->cnf));
240 in_dev->cnf.sysctl = NULL;
242 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
243 if (!in_dev->arp_parms)
245 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
246 dev_disable_lro(dev);
247 /* Reference in_dev->dev */
249 /* Account for reference dev->ip_ptr (below) */
252 devinet_sysctl_register(in_dev);
253 ip_mc_init_dev(in_dev);
254 if (dev->flags & IFF_UP)
257 /* we can receive as soon as ip_ptr is set -- do this last */
258 rcu_assign_pointer(dev->ip_ptr, in_dev);
267 static void in_dev_rcu_put(struct rcu_head *head)
269 struct in_device *idev = container_of(head, struct in_device, rcu_head);
273 static void inetdev_destroy(struct in_device *in_dev)
275 struct in_ifaddr *ifa;
276 struct net_device *dev;
284 ip_mc_destroy_dev(in_dev);
286 while ((ifa = in_dev->ifa_list) != NULL) {
287 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
291 RCU_INIT_POINTER(dev->ip_ptr, NULL);
293 devinet_sysctl_unregister(in_dev);
294 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
297 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
300 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
303 for_primary_ifa(in_dev) {
304 if (inet_ifa_match(a, ifa)) {
305 if (!b || inet_ifa_match(b, ifa)) {
310 } endfor_ifa(in_dev);
315 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
316 int destroy, struct nlmsghdr *nlh, u32 portid)
318 struct in_ifaddr *promote = NULL;
319 struct in_ifaddr *ifa, *ifa1 = *ifap;
320 struct in_ifaddr *last_prim = in_dev->ifa_list;
321 struct in_ifaddr *prev_prom = NULL;
322 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
326 /* 1. Deleting primary ifaddr forces deletion all secondaries
327 * unless alias promotion is set
330 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
331 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
333 while ((ifa = *ifap1) != NULL) {
334 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
335 ifa1->ifa_scope <= ifa->ifa_scope)
338 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
339 ifa1->ifa_mask != ifa->ifa_mask ||
340 !inet_ifa_match(ifa1->ifa_address, ifa)) {
341 ifap1 = &ifa->ifa_next;
347 inet_hash_remove(ifa);
348 *ifap1 = ifa->ifa_next;
350 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
351 blocking_notifier_call_chain(&inetaddr_chain,
361 /* On promotion all secondaries from subnet are changing
362 * the primary IP, we must remove all their routes silently
363 * and later to add them back with new prefsrc. Do this
364 * while all addresses are on the device list.
366 for (ifa = promote; ifa; ifa = ifa->ifa_next) {
367 if (ifa1->ifa_mask == ifa->ifa_mask &&
368 inet_ifa_match(ifa1->ifa_address, ifa))
369 fib_del_ifaddr(ifa, ifa1);
374 *ifap = ifa1->ifa_next;
375 inet_hash_remove(ifa1);
377 /* 3. Announce address deletion */
379 /* Send message first, then call notifier.
380 At first sight, FIB update triggered by notifier
381 will refer to already deleted ifaddr, that could confuse
382 netlink listeners. It is not true: look, gated sees
383 that route deleted and if it still thinks that ifaddr
384 is valid, it will try to restore deleted routes... Grr.
385 So that, this order is correct.
387 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
388 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
391 struct in_ifaddr *next_sec = promote->ifa_next;
394 prev_prom->ifa_next = promote->ifa_next;
395 promote->ifa_next = last_prim->ifa_next;
396 last_prim->ifa_next = promote;
399 promote->ifa_flags &= ~IFA_F_SECONDARY;
400 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
401 blocking_notifier_call_chain(&inetaddr_chain,
403 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
404 if (ifa1->ifa_mask != ifa->ifa_mask ||
405 !inet_ifa_match(ifa1->ifa_address, ifa))
415 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
418 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
421 static void check_lifetime(struct work_struct *work);
423 static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
425 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
428 struct in_device *in_dev = ifa->ifa_dev;
429 struct in_ifaddr *ifa1, **ifap, **last_primary;
433 if (!ifa->ifa_local) {
438 ifa->ifa_flags &= ~IFA_F_SECONDARY;
439 last_primary = &in_dev->ifa_list;
441 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
442 ifap = &ifa1->ifa_next) {
443 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
444 ifa->ifa_scope <= ifa1->ifa_scope)
445 last_primary = &ifa1->ifa_next;
446 if (ifa1->ifa_mask == ifa->ifa_mask &&
447 inet_ifa_match(ifa1->ifa_address, ifa)) {
448 if (ifa1->ifa_local == ifa->ifa_local) {
452 if (ifa1->ifa_scope != ifa->ifa_scope) {
456 ifa->ifa_flags |= IFA_F_SECONDARY;
460 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
461 net_srandom(ifa->ifa_local);
465 ifa->ifa_next = *ifap;
468 inet_hash_insert(dev_net(in_dev->dev), ifa);
470 cancel_delayed_work(&check_lifetime_work);
471 schedule_delayed_work(&check_lifetime_work, 0);
473 /* Send message first, then call notifier.
474 Notifier will trigger FIB update, so that
475 listeners of netlink will know about new ifaddr */
476 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
477 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
482 static int inet_insert_ifa(struct in_ifaddr *ifa)
484 return __inet_insert_ifa(ifa, NULL, 0);
487 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
489 struct in_device *in_dev = __in_dev_get_rtnl(dev);
497 ipv4_devconf_setall(in_dev);
498 if (ifa->ifa_dev != in_dev) {
499 WARN_ON(ifa->ifa_dev);
501 ifa->ifa_dev = in_dev;
503 if (ipv4_is_loopback(ifa->ifa_local))
504 ifa->ifa_scope = RT_SCOPE_HOST;
505 return inet_insert_ifa(ifa);
508 /* Caller must hold RCU or RTNL :
509 * We dont take a reference on found in_device
511 struct in_device *inetdev_by_index(struct net *net, int ifindex)
513 struct net_device *dev;
514 struct in_device *in_dev = NULL;
517 dev = dev_get_by_index_rcu(net, ifindex);
519 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
523 EXPORT_SYMBOL(inetdev_by_index);
525 /* Called only from RTNL semaphored context. No locks. */
527 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
532 for_primary_ifa(in_dev) {
533 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
535 } endfor_ifa(in_dev);
539 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
541 struct net *net = sock_net(skb->sk);
542 struct nlattr *tb[IFA_MAX+1];
543 struct in_device *in_dev;
544 struct ifaddrmsg *ifm;
545 struct in_ifaddr *ifa, **ifap;
550 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
554 ifm = nlmsg_data(nlh);
555 in_dev = inetdev_by_index(net, ifm->ifa_index);
556 if (in_dev == NULL) {
561 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
562 ifap = &ifa->ifa_next) {
564 ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
567 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
570 if (tb[IFA_ADDRESS] &&
571 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
572 !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
575 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
579 err = -EADDRNOTAVAIL;
584 #define INFINITY_LIFE_TIME 0xFFFFFFFF
586 static void check_lifetime(struct work_struct *work)
588 unsigned long now, next, next_sec, next_sched;
589 struct in_ifaddr *ifa;
590 struct hlist_node *n;
594 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
596 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
597 bool change_needed = false;
600 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
603 if (ifa->ifa_flags & IFA_F_PERMANENT)
606 /* We try to batch several events at once. */
607 age = (now - ifa->ifa_tstamp +
608 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
610 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
611 age >= ifa->ifa_valid_lft) {
612 change_needed = true;
613 } else if (ifa->ifa_preferred_lft ==
614 INFINITY_LIFE_TIME) {
616 } else if (age >= ifa->ifa_preferred_lft) {
617 if (time_before(ifa->ifa_tstamp +
618 ifa->ifa_valid_lft * HZ, next))
619 next = ifa->ifa_tstamp +
620 ifa->ifa_valid_lft * HZ;
622 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
623 change_needed = true;
624 } else if (time_before(ifa->ifa_tstamp +
625 ifa->ifa_preferred_lft * HZ,
627 next = ifa->ifa_tstamp +
628 ifa->ifa_preferred_lft * HZ;
635 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
638 if (ifa->ifa_flags & IFA_F_PERMANENT)
641 /* We try to batch several events at once. */
642 age = (now - ifa->ifa_tstamp +
643 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
645 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
646 age >= ifa->ifa_valid_lft) {
647 struct in_ifaddr **ifap;
649 for (ifap = &ifa->ifa_dev->ifa_list;
650 *ifap != NULL; ifap = &(*ifap)->ifa_next) {
652 inet_del_ifa(ifa->ifa_dev,
657 } else if (ifa->ifa_preferred_lft !=
658 INFINITY_LIFE_TIME &&
659 age >= ifa->ifa_preferred_lft &&
660 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
661 ifa->ifa_flags |= IFA_F_DEPRECATED;
662 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
668 next_sec = round_jiffies_up(next);
671 /* If rounded timeout is accurate enough, accept it. */
672 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
673 next_sched = next_sec;
676 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
677 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
678 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
680 schedule_delayed_work(&check_lifetime_work, next_sched - now);
683 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
686 unsigned long timeout;
688 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
690 timeout = addrconf_timeout_fixup(valid_lft, HZ);
691 if (addrconf_finite_timeout(timeout))
692 ifa->ifa_valid_lft = timeout;
694 ifa->ifa_flags |= IFA_F_PERMANENT;
696 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
697 if (addrconf_finite_timeout(timeout)) {
699 ifa->ifa_flags |= IFA_F_DEPRECATED;
700 ifa->ifa_preferred_lft = timeout;
702 ifa->ifa_tstamp = jiffies;
703 if (!ifa->ifa_cstamp)
704 ifa->ifa_cstamp = ifa->ifa_tstamp;
707 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
708 __u32 *pvalid_lft, __u32 *pprefered_lft)
710 struct nlattr *tb[IFA_MAX+1];
711 struct in_ifaddr *ifa;
712 struct ifaddrmsg *ifm;
713 struct net_device *dev;
714 struct in_device *in_dev;
717 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
721 ifm = nlmsg_data(nlh);
723 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
726 dev = __dev_get_by_index(net, ifm->ifa_index);
731 in_dev = __in_dev_get_rtnl(dev);
736 ifa = inet_alloc_ifa();
739 * A potential indev allocation can be left alive, it stays
740 * assigned to its device and is destroy with it.
744 ipv4_devconf_setall(in_dev);
747 if (tb[IFA_ADDRESS] == NULL)
748 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
750 INIT_HLIST_NODE(&ifa->hash);
751 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
752 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
753 ifa->ifa_flags = ifm->ifa_flags;
754 ifa->ifa_scope = ifm->ifa_scope;
755 ifa->ifa_dev = in_dev;
757 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
758 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
760 if (tb[IFA_BROADCAST])
761 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
764 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
766 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
768 if (tb[IFA_CACHEINFO]) {
769 struct ifa_cacheinfo *ci;
771 ci = nla_data(tb[IFA_CACHEINFO]);
772 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
776 *pvalid_lft = ci->ifa_valid;
777 *pprefered_lft = ci->ifa_prefered;
788 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
790 struct in_device *in_dev = ifa->ifa_dev;
791 struct in_ifaddr *ifa1, **ifap;
796 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
797 ifap = &ifa1->ifa_next) {
798 if (ifa1->ifa_mask == ifa->ifa_mask &&
799 inet_ifa_match(ifa1->ifa_address, ifa) &&
800 ifa1->ifa_local == ifa->ifa_local)
806 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
808 struct net *net = sock_net(skb->sk);
809 struct in_ifaddr *ifa;
810 struct in_ifaddr *ifa_existing;
811 __u32 valid_lft = INFINITY_LIFE_TIME;
812 __u32 prefered_lft = INFINITY_LIFE_TIME;
816 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
820 ifa_existing = find_matching_ifa(ifa);
822 /* It would be best to check for !NLM_F_CREATE here but
823 * userspace alreay relies on not having to provide this.
825 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
826 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
830 if (nlh->nlmsg_flags & NLM_F_EXCL ||
831 !(nlh->nlmsg_flags & NLM_F_REPLACE))
834 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
835 cancel_delayed_work(&check_lifetime_work);
836 schedule_delayed_work(&check_lifetime_work, 0);
837 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
838 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
844 * Determine a default network mask, based on the IP address.
847 static int inet_abc_len(__be32 addr)
849 int rc = -1; /* Something else, probably a multicast. */
851 if (ipv4_is_zeronet(addr))
854 __u32 haddr = ntohl(addr);
856 if (IN_CLASSA(haddr))
858 else if (IN_CLASSB(haddr))
860 else if (IN_CLASSC(haddr))
868 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
871 struct sockaddr_in sin_orig;
872 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
873 struct in_device *in_dev;
874 struct in_ifaddr **ifap = NULL;
875 struct in_ifaddr *ifa = NULL;
876 struct net_device *dev;
879 int tryaddrmatch = 0;
882 * Fetch the caller's info block into kernel space
885 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
887 ifr.ifr_name[IFNAMSIZ - 1] = 0;
889 /* save original address for comparison */
890 memcpy(&sin_orig, sin, sizeof(*sin));
892 colon = strchr(ifr.ifr_name, ':');
896 dev_load(net, ifr.ifr_name);
899 case SIOCGIFADDR: /* Get interface address */
900 case SIOCGIFBRDADDR: /* Get the broadcast address */
901 case SIOCGIFDSTADDR: /* Get the destination address */
902 case SIOCGIFNETMASK: /* Get the netmask for the interface */
903 /* Note that these ioctls will not sleep,
904 so that we do not impose a lock.
905 One day we will be forced to put shlock here (I mean SMP)
907 tryaddrmatch = (sin_orig.sin_family == AF_INET);
908 memset(sin, 0, sizeof(*sin));
909 sin->sin_family = AF_INET;
914 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
917 case SIOCSIFADDR: /* Set interface address (and family) */
918 case SIOCSIFBRDADDR: /* Set the broadcast address */
919 case SIOCSIFDSTADDR: /* Set the destination address */
920 case SIOCSIFNETMASK: /* Set the netmask for the interface */
922 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
925 if (sin->sin_family != AF_INET)
936 dev = __dev_get_by_name(net, ifr.ifr_name);
943 in_dev = __in_dev_get_rtnl(dev);
946 /* Matthias Andree */
947 /* compare label and address (4.4BSD style) */
948 /* note: we only do this for a limited set of ioctls
949 and only if the original address family was AF_INET.
950 This is checked above. */
951 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
952 ifap = &ifa->ifa_next) {
953 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
954 sin_orig.sin_addr.s_addr ==
960 /* we didn't get a match, maybe the application is
961 4.3BSD-style and passed in junk so we fall back to
962 comparing just the label */
964 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
965 ifap = &ifa->ifa_next)
966 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
971 ret = -EADDRNOTAVAIL;
972 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
976 case SIOCGIFADDR: /* Get interface address */
977 sin->sin_addr.s_addr = ifa->ifa_local;
980 case SIOCGIFBRDADDR: /* Get the broadcast address */
981 sin->sin_addr.s_addr = ifa->ifa_broadcast;
984 case SIOCGIFDSTADDR: /* Get the destination address */
985 sin->sin_addr.s_addr = ifa->ifa_address;
988 case SIOCGIFNETMASK: /* Get the netmask for the interface */
989 sin->sin_addr.s_addr = ifa->ifa_mask;
994 ret = -EADDRNOTAVAIL;
998 if (!(ifr.ifr_flags & IFF_UP))
999 inet_del_ifa(in_dev, ifap, 1);
1002 ret = dev_change_flags(dev, ifr.ifr_flags);
1005 case SIOCSIFADDR: /* Set interface address (and family) */
1007 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1012 ifa = inet_alloc_ifa();
1015 INIT_HLIST_NODE(&ifa->hash);
1017 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
1019 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1022 if (ifa->ifa_local == sin->sin_addr.s_addr)
1024 inet_del_ifa(in_dev, ifap, 0);
1025 ifa->ifa_broadcast = 0;
1029 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1031 if (!(dev->flags & IFF_POINTOPOINT)) {
1032 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1033 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1034 if ((dev->flags & IFF_BROADCAST) &&
1035 ifa->ifa_prefixlen < 31)
1036 ifa->ifa_broadcast = ifa->ifa_address |
1039 ifa->ifa_prefixlen = 32;
1040 ifa->ifa_mask = inet_make_mask(32);
1042 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1043 ret = inet_set_ifa(dev, ifa);
1046 case SIOCSIFBRDADDR: /* Set the broadcast address */
1048 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1049 inet_del_ifa(in_dev, ifap, 0);
1050 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1051 inet_insert_ifa(ifa);
1055 case SIOCSIFDSTADDR: /* Set the destination address */
1057 if (ifa->ifa_address == sin->sin_addr.s_addr)
1060 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1063 inet_del_ifa(in_dev, ifap, 0);
1064 ifa->ifa_address = sin->sin_addr.s_addr;
1065 inet_insert_ifa(ifa);
1068 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1071 * The mask we set must be legal.
1074 if (bad_mask(sin->sin_addr.s_addr, 0))
1077 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1078 __be32 old_mask = ifa->ifa_mask;
1079 inet_del_ifa(in_dev, ifap, 0);
1080 ifa->ifa_mask = sin->sin_addr.s_addr;
1081 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1083 /* See if current broadcast address matches
1084 * with current netmask, then recalculate
1085 * the broadcast address. Otherwise it's a
1086 * funny address, so don't touch it since
1087 * the user seems to know what (s)he's doing...
1089 if ((dev->flags & IFF_BROADCAST) &&
1090 (ifa->ifa_prefixlen < 31) &&
1091 (ifa->ifa_broadcast ==
1092 (ifa->ifa_local|~old_mask))) {
1093 ifa->ifa_broadcast = (ifa->ifa_local |
1094 ~sin->sin_addr.s_addr);
1096 inet_insert_ifa(ifa);
1106 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
1110 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
1112 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1113 struct in_ifaddr *ifa;
1120 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1122 done += sizeof(ifr);
1125 if (len < (int) sizeof(ifr))
1127 memset(&ifr, 0, sizeof(struct ifreq));
1129 strcpy(ifr.ifr_name, ifa->ifa_label);
1131 strcpy(ifr.ifr_name, dev->name);
1133 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1134 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1137 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
1141 buf += sizeof(struct ifreq);
1142 len -= sizeof(struct ifreq);
1143 done += sizeof(struct ifreq);
1149 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1152 struct in_device *in_dev;
1153 struct net *net = dev_net(dev);
1156 in_dev = __in_dev_get_rcu(dev);
1160 for_primary_ifa(in_dev) {
1161 if (ifa->ifa_scope > scope)
1163 if (!dst || inet_ifa_match(dst, ifa)) {
1164 addr = ifa->ifa_local;
1168 addr = ifa->ifa_local;
1169 } endfor_ifa(in_dev);
1175 /* Not loopback addresses on loopback should be preferred
1176 in this case. It is importnat that lo is the first interface
1179 for_each_netdev_rcu(net, dev) {
1180 in_dev = __in_dev_get_rcu(dev);
1184 for_primary_ifa(in_dev) {
1185 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1186 ifa->ifa_scope <= scope) {
1187 addr = ifa->ifa_local;
1190 } endfor_ifa(in_dev);
1196 EXPORT_SYMBOL(inet_select_addr);
1198 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1199 __be32 local, int scope)
1206 (local == ifa->ifa_local || !local) &&
1207 ifa->ifa_scope <= scope) {
1208 addr = ifa->ifa_local;
1213 same = (!local || inet_ifa_match(local, ifa)) &&
1214 (!dst || inet_ifa_match(dst, ifa));
1218 /* Is the selected addr into dst subnet? */
1219 if (inet_ifa_match(addr, ifa))
1221 /* No, then can we use new local src? */
1222 if (ifa->ifa_scope <= scope) {
1223 addr = ifa->ifa_local;
1226 /* search for large dst subnet for addr */
1230 } endfor_ifa(in_dev);
1232 return same ? addr : 0;
1236 * Confirm that local IP address exists using wildcards:
1237 * - in_dev: only on this interface, 0=any interface
1238 * - dst: only in the same subnet as dst, 0=any dst
1239 * - local: address, 0=autoselect the local address
1240 * - scope: maximum allowed scope value for the local address
1242 __be32 inet_confirm_addr(struct in_device *in_dev,
1243 __be32 dst, __be32 local, int scope)
1246 struct net_device *dev;
1249 if (scope != RT_SCOPE_LINK)
1250 return confirm_addr_indev(in_dev, dst, local, scope);
1252 net = dev_net(in_dev->dev);
1254 for_each_netdev_rcu(net, dev) {
1255 in_dev = __in_dev_get_rcu(dev);
1257 addr = confirm_addr_indev(in_dev, dst, local, scope);
1266 EXPORT_SYMBOL(inet_confirm_addr);
1272 int register_inetaddr_notifier(struct notifier_block *nb)
1274 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1276 EXPORT_SYMBOL(register_inetaddr_notifier);
1278 int unregister_inetaddr_notifier(struct notifier_block *nb)
1280 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1282 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1284 /* Rename ifa_labels for a device name change. Make some effort to preserve
1285 * existing alias numbering and to create unique labels if possible.
1287 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1289 struct in_ifaddr *ifa;
1292 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1293 char old[IFNAMSIZ], *dot;
1295 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1296 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1299 dot = strchr(old, ':');
1301 sprintf(old, ":%d", named);
1304 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1305 strcat(ifa->ifa_label, dot);
1307 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1309 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1313 static bool inetdev_valid_mtu(unsigned int mtu)
1318 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1319 struct in_device *in_dev)
1322 struct in_ifaddr *ifa;
1324 for (ifa = in_dev->ifa_list; ifa;
1325 ifa = ifa->ifa_next) {
1326 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1327 ifa->ifa_local, dev,
1328 ifa->ifa_local, NULL,
1329 dev->dev_addr, NULL);
1333 /* Called only under RTNL semaphore */
1335 static int inetdev_event(struct notifier_block *this, unsigned long event,
1338 struct net_device *dev = ptr;
1339 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1344 if (event == NETDEV_REGISTER) {
1345 in_dev = inetdev_init(dev);
1347 return notifier_from_errno(-ENOMEM);
1348 if (dev->flags & IFF_LOOPBACK) {
1349 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1350 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1352 } else if (event == NETDEV_CHANGEMTU) {
1353 /* Re-enabling IP */
1354 if (inetdev_valid_mtu(dev->mtu))
1355 in_dev = inetdev_init(dev);
1361 case NETDEV_REGISTER:
1362 pr_debug("%s: bug\n", __func__);
1363 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1366 if (!inetdev_valid_mtu(dev->mtu))
1368 if (dev->flags & IFF_LOOPBACK) {
1369 struct in_ifaddr *ifa = inet_alloc_ifa();
1372 INIT_HLIST_NODE(&ifa->hash);
1374 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1375 ifa->ifa_prefixlen = 8;
1376 ifa->ifa_mask = inet_make_mask(8);
1377 in_dev_hold(in_dev);
1378 ifa->ifa_dev = in_dev;
1379 ifa->ifa_scope = RT_SCOPE_HOST;
1380 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1381 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1382 INFINITY_LIFE_TIME);
1383 inet_insert_ifa(ifa);
1388 case NETDEV_CHANGEADDR:
1389 if (!IN_DEV_ARP_NOTIFY(in_dev))
1392 case NETDEV_NOTIFY_PEERS:
1393 /* Send gratuitous ARP to notify of link change */
1394 inetdev_send_gratuitous_arp(dev, in_dev);
1399 case NETDEV_PRE_TYPE_CHANGE:
1400 ip_mc_unmap(in_dev);
1402 case NETDEV_POST_TYPE_CHANGE:
1403 ip_mc_remap(in_dev);
1405 case NETDEV_CHANGEMTU:
1406 if (inetdev_valid_mtu(dev->mtu))
1408 /* disable IP when MTU is not enough */
1409 case NETDEV_UNREGISTER:
1410 inetdev_destroy(in_dev);
1412 case NETDEV_CHANGENAME:
1413 /* Do not notify about label change, this event is
1414 * not interesting to applications using netlink.
1416 inetdev_changename(dev, in_dev);
1418 devinet_sysctl_unregister(in_dev);
1419 devinet_sysctl_register(in_dev);
1426 static struct notifier_block ip_netdev_notifier = {
1427 .notifier_call = inetdev_event,
1430 static size_t inet_nlmsg_size(void)
1432 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1433 + nla_total_size(4) /* IFA_ADDRESS */
1434 + nla_total_size(4) /* IFA_LOCAL */
1435 + nla_total_size(4) /* IFA_BROADCAST */
1436 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1437 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1440 static inline u32 cstamp_delta(unsigned long cstamp)
1442 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1445 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1446 unsigned long tstamp, u32 preferred, u32 valid)
1448 struct ifa_cacheinfo ci;
1450 ci.cstamp = cstamp_delta(cstamp);
1451 ci.tstamp = cstamp_delta(tstamp);
1452 ci.ifa_prefered = preferred;
1453 ci.ifa_valid = valid;
1455 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1458 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1459 u32 portid, u32 seq, int event, unsigned int flags)
1461 struct ifaddrmsg *ifm;
1462 struct nlmsghdr *nlh;
1463 u32 preferred, valid;
1465 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
1469 ifm = nlmsg_data(nlh);
1470 ifm->ifa_family = AF_INET;
1471 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1472 ifm->ifa_flags = ifa->ifa_flags;
1473 ifm->ifa_scope = ifa->ifa_scope;
1474 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1476 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1477 preferred = ifa->ifa_preferred_lft;
1478 valid = ifa->ifa_valid_lft;
1479 if (preferred != INFINITY_LIFE_TIME) {
1480 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1482 if (preferred > tval)
1486 if (valid != INFINITY_LIFE_TIME) {
1494 preferred = INFINITY_LIFE_TIME;
1495 valid = INFINITY_LIFE_TIME;
1497 if ((ifa->ifa_address &&
1498 nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1500 nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1501 (ifa->ifa_broadcast &&
1502 nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1503 (ifa->ifa_label[0] &&
1504 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1505 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1507 goto nla_put_failure;
1509 return nlmsg_end(skb, nlh);
1512 nlmsg_cancel(skb, nlh);
1516 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1518 struct net *net = sock_net(skb->sk);
1521 int ip_idx, s_ip_idx;
1522 struct net_device *dev;
1523 struct in_device *in_dev;
1524 struct in_ifaddr *ifa;
1525 struct hlist_head *head;
1528 s_idx = idx = cb->args[1];
1529 s_ip_idx = ip_idx = cb->args[2];
1531 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1533 head = &net->dev_index_head[h];
1535 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1537 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1540 if (h > s_h || idx > s_idx)
1542 in_dev = __in_dev_get_rcu(dev);
1546 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1547 ifa = ifa->ifa_next, ip_idx++) {
1548 if (ip_idx < s_ip_idx)
1550 if (inet_fill_ifaddr(skb, ifa,
1551 NETLINK_CB(cb->skb).portid,
1553 RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1557 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1568 cb->args[2] = ip_idx;
1573 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1576 struct sk_buff *skb;
1577 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1581 net = dev_net(ifa->ifa_dev->dev);
1582 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1586 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
1588 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1589 WARN_ON(err == -EMSGSIZE);
1593 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1597 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1600 static size_t inet_get_link_af_size(const struct net_device *dev)
1602 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1607 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1610 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
1612 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1619 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1623 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1624 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1629 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1630 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1633 static int inet_validate_link_af(const struct net_device *dev,
1634 const struct nlattr *nla)
1636 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1639 if (dev && !__in_dev_get_rtnl(dev))
1640 return -EAFNOSUPPORT;
1642 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1646 if (tb[IFLA_INET_CONF]) {
1647 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1648 int cfgid = nla_type(a);
1653 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1661 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1663 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1664 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1668 return -EAFNOSUPPORT;
1670 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1673 if (tb[IFLA_INET_CONF]) {
1674 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1675 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1681 static int inet_netconf_msgsize_devconf(int type)
1683 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1684 + nla_total_size(4); /* NETCONFA_IFINDEX */
1686 /* type -1 is used for ALL */
1687 if (type == -1 || type == NETCONFA_FORWARDING)
1688 size += nla_total_size(4);
1689 if (type == -1 || type == NETCONFA_RP_FILTER)
1690 size += nla_total_size(4);
1691 if (type == -1 || type == NETCONFA_MC_FORWARDING)
1692 size += nla_total_size(4);
1697 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1698 struct ipv4_devconf *devconf, u32 portid,
1699 u32 seq, int event, unsigned int flags,
1702 struct nlmsghdr *nlh;
1703 struct netconfmsg *ncm;
1705 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1710 ncm = nlmsg_data(nlh);
1711 ncm->ncm_family = AF_INET;
1713 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1714 goto nla_put_failure;
1716 /* type -1 is used for ALL */
1717 if ((type == -1 || type == NETCONFA_FORWARDING) &&
1718 nla_put_s32(skb, NETCONFA_FORWARDING,
1719 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1720 goto nla_put_failure;
1721 if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1722 nla_put_s32(skb, NETCONFA_RP_FILTER,
1723 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1724 goto nla_put_failure;
1725 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1726 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1727 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1728 goto nla_put_failure;
1730 return nlmsg_end(skb, nlh);
1733 nlmsg_cancel(skb, nlh);
1737 void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1738 struct ipv4_devconf *devconf)
1740 struct sk_buff *skb;
1743 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
1747 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1748 RTM_NEWNETCONF, 0, type);
1750 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1751 WARN_ON(err == -EMSGSIZE);
1755 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1759 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1762 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1763 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1764 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
1765 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
1768 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
1769 struct nlmsghdr *nlh)
1771 struct net *net = sock_net(in_skb->sk);
1772 struct nlattr *tb[NETCONFA_MAX+1];
1773 struct netconfmsg *ncm;
1774 struct sk_buff *skb;
1775 struct ipv4_devconf *devconf;
1776 struct in_device *in_dev;
1777 struct net_device *dev;
1781 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1782 devconf_ipv4_policy);
1787 if (!tb[NETCONFA_IFINDEX])
1790 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1792 case NETCONFA_IFINDEX_ALL:
1793 devconf = net->ipv4.devconf_all;
1795 case NETCONFA_IFINDEX_DEFAULT:
1796 devconf = net->ipv4.devconf_dflt;
1799 dev = __dev_get_by_index(net, ifindex);
1802 in_dev = __in_dev_get_rtnl(dev);
1805 devconf = &in_dev->cnf;
1810 skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
1814 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1815 NETLINK_CB(in_skb).portid,
1816 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1819 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1820 WARN_ON(err == -EMSGSIZE);
1824 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1829 static int inet_netconf_dump_devconf(struct sk_buff *skb,
1830 struct netlink_callback *cb)
1832 struct net *net = sock_net(skb->sk);
1835 struct net_device *dev;
1836 struct in_device *in_dev;
1837 struct hlist_head *head;
1840 s_idx = idx = cb->args[1];
1842 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1844 head = &net->dev_index_head[h];
1846 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1848 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1851 in_dev = __in_dev_get_rcu(dev);
1855 if (inet_netconf_fill_devconf(skb, dev->ifindex,
1857 NETLINK_CB(cb->skb).portid,
1865 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1871 if (h == NETDEV_HASHENTRIES) {
1872 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
1873 net->ipv4.devconf_all,
1874 NETLINK_CB(cb->skb).portid,
1876 RTM_NEWNETCONF, NLM_F_MULTI,
1882 if (h == NETDEV_HASHENTRIES + 1) {
1883 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
1884 net->ipv4.devconf_dflt,
1885 NETLINK_CB(cb->skb).portid,
1887 RTM_NEWNETCONF, NLM_F_MULTI,
1900 #ifdef CONFIG_SYSCTL
1902 static void devinet_copy_dflt_conf(struct net *net, int i)
1904 struct net_device *dev;
1907 for_each_netdev_rcu(net, dev) {
1908 struct in_device *in_dev;
1910 in_dev = __in_dev_get_rcu(dev);
1911 if (in_dev && !test_bit(i, in_dev->cnf.state))
1912 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
1917 /* called with RTNL locked */
1918 static void inet_forward_change(struct net *net)
1920 struct net_device *dev;
1921 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1923 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1924 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
1925 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1926 NETCONFA_IFINDEX_ALL,
1927 net->ipv4.devconf_all);
1928 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1929 NETCONFA_IFINDEX_DEFAULT,
1930 net->ipv4.devconf_dflt);
1932 for_each_netdev(net, dev) {
1933 struct in_device *in_dev;
1935 dev_disable_lro(dev);
1937 in_dev = __in_dev_get_rcu(dev);
1939 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1940 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1941 dev->ifindex, &in_dev->cnf);
1947 static int devinet_conf_proc(ctl_table *ctl, int write,
1948 void __user *buffer,
1949 size_t *lenp, loff_t *ppos)
1951 int old_value = *(int *)ctl->data;
1952 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1953 int new_value = *(int *)ctl->data;
1956 struct ipv4_devconf *cnf = ctl->extra1;
1957 struct net *net = ctl->extra2;
1958 int i = (int *)ctl->data - cnf->data;
1960 set_bit(i, cnf->state);
1962 if (cnf == net->ipv4.devconf_dflt)
1963 devinet_copy_dflt_conf(net, i);
1964 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
1965 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
1966 if ((new_value == 0) && (old_value != 0))
1967 rt_cache_flush(net);
1968 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
1969 new_value != old_value) {
1972 if (cnf == net->ipv4.devconf_dflt)
1973 ifindex = NETCONFA_IFINDEX_DEFAULT;
1974 else if (cnf == net->ipv4.devconf_all)
1975 ifindex = NETCONFA_IFINDEX_ALL;
1977 struct in_device *idev =
1978 container_of(cnf, struct in_device,
1980 ifindex = idev->dev->ifindex;
1982 inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
1990 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1991 void __user *buffer,
1992 size_t *lenp, loff_t *ppos)
1994 int *valp = ctl->data;
1997 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1999 if (write && *valp != val) {
2000 struct net *net = ctl->extra2;
2002 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2003 if (!rtnl_trylock()) {
2004 /* Restore the original values before restarting */
2007 return restart_syscall();
2009 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2010 inet_forward_change(net);
2012 struct ipv4_devconf *cnf = ctl->extra1;
2013 struct in_device *idev =
2014 container_of(cnf, struct in_device, cnf);
2016 dev_disable_lro(idev->dev);
2017 inet_netconf_notify_devconf(net,
2018 NETCONFA_FORWARDING,
2023 rt_cache_flush(net);
2025 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
2026 NETCONFA_IFINDEX_DEFAULT,
2027 net->ipv4.devconf_dflt);
2033 static int ipv4_doint_and_flush(ctl_table *ctl, int write,
2034 void __user *buffer,
2035 size_t *lenp, loff_t *ppos)
2037 int *valp = ctl->data;
2039 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2040 struct net *net = ctl->extra2;
2042 if (write && *valp != val)
2043 rt_cache_flush(net);
2048 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2051 .data = ipv4_devconf.data + \
2052 IPV4_DEVCONF_ ## attr - 1, \
2053 .maxlen = sizeof(int), \
2055 .proc_handler = proc, \
2056 .extra1 = &ipv4_devconf, \
2059 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2060 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2062 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2063 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2065 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2066 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2068 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2069 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2071 static struct devinet_sysctl_table {
2072 struct ctl_table_header *sysctl_header;
2073 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2074 } devinet_sysctl = {
2076 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2077 devinet_sysctl_forward),
2078 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2080 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2081 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2082 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2083 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2084 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2085 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2086 "accept_source_route"),
2087 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2088 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2089 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2090 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2091 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2092 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2093 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2094 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2095 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2096 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2097 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2098 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2099 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2101 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2102 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2103 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
2104 "force_igmp_version"),
2105 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2106 "promote_secondaries"),
2107 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2112 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2113 struct ipv4_devconf *p)
2116 struct devinet_sysctl_table *t;
2117 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2119 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2123 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2124 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2125 t->devinet_vars[i].extra1 = p;
2126 t->devinet_vars[i].extra2 = net;
2129 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2131 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2132 if (!t->sysctl_header)
2144 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
2146 struct devinet_sysctl_table *t = cnf->sysctl;
2152 unregister_net_sysctl_table(t->sysctl_header);
2156 static void devinet_sysctl_register(struct in_device *idev)
2158 neigh_sysctl_register(idev->dev, idev->arp_parms, "ipv4", NULL);
2159 __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2163 static void devinet_sysctl_unregister(struct in_device *idev)
2165 __devinet_sysctl_unregister(&idev->cnf);
2166 neigh_sysctl_unregister(idev->arp_parms);
2169 static struct ctl_table ctl_forward_entry[] = {
2171 .procname = "ip_forward",
2172 .data = &ipv4_devconf.data[
2173 IPV4_DEVCONF_FORWARDING - 1],
2174 .maxlen = sizeof(int),
2176 .proc_handler = devinet_sysctl_forward,
2177 .extra1 = &ipv4_devconf,
2178 .extra2 = &init_net,
2184 static __net_init int devinet_init_net(struct net *net)
2187 struct ipv4_devconf *all, *dflt;
2188 #ifdef CONFIG_SYSCTL
2189 struct ctl_table *tbl = ctl_forward_entry;
2190 struct ctl_table_header *forw_hdr;
2194 all = &ipv4_devconf;
2195 dflt = &ipv4_devconf_dflt;
2197 if (!net_eq(net, &init_net)) {
2198 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
2202 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2204 goto err_alloc_dflt;
2206 #ifdef CONFIG_SYSCTL
2207 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
2211 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2212 tbl[0].extra1 = all;
2213 tbl[0].extra2 = net;
2217 #ifdef CONFIG_SYSCTL
2218 err = __devinet_sysctl_register(net, "all", all);
2222 err = __devinet_sysctl_register(net, "default", dflt);
2227 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2228 if (forw_hdr == NULL)
2230 net->ipv4.forw_hdr = forw_hdr;
2233 net->ipv4.devconf_all = all;
2234 net->ipv4.devconf_dflt = dflt;
2237 #ifdef CONFIG_SYSCTL
2239 __devinet_sysctl_unregister(dflt);
2241 __devinet_sysctl_unregister(all);
2243 if (tbl != ctl_forward_entry)
2247 if (dflt != &ipv4_devconf_dflt)
2250 if (all != &ipv4_devconf)
2256 static __net_exit void devinet_exit_net(struct net *net)
2258 #ifdef CONFIG_SYSCTL
2259 struct ctl_table *tbl;
2261 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2262 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2263 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
2264 __devinet_sysctl_unregister(net->ipv4.devconf_all);
2267 kfree(net->ipv4.devconf_dflt);
2268 kfree(net->ipv4.devconf_all);
2271 static __net_initdata struct pernet_operations devinet_ops = {
2272 .init = devinet_init_net,
2273 .exit = devinet_exit_net,
2276 static struct rtnl_af_ops inet_af_ops = {
2278 .fill_link_af = inet_fill_link_af,
2279 .get_link_af_size = inet_get_link_af_size,
2280 .validate_link_af = inet_validate_link_af,
2281 .set_link_af = inet_set_link_af,
2284 void __init devinet_init(void)
2288 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2289 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2291 register_pernet_subsys(&devinet_ops);
2293 register_gifconf(PF_INET, inet_gifconf);
2294 register_netdevice_notifier(&ip_netdev_notifier);
2296 schedule_delayed_work(&check_lifetime_work, 0);
2298 rtnl_af_register(&inet_af_ops);
2300 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
2301 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
2302 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
2303 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2304 inet_netconf_dump_devconf, NULL);