1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET3 IP device support routines.
5 * Derived from the IP parts of dev.c 1.0.19
7 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
8 * Mark Evans, <evansmp@uhura.aston.ac.uk>
11 * Alan Cox, <gw4pts@gw4pts.ampr.org>
12 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
15 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
17 * Cyrus Durgin: updated for kmod
18 * Matthias Andree: in devinet_ioctl, compare label and
19 * address (4.4BSD alias style support),
20 * fall back to comparing just the label
25 #include <linux/uaccess.h>
26 #include <linux/bitops.h>
27 #include <linux/capability.h>
28 #include <linux/module.h>
29 #include <linux/types.h>
30 #include <linux/kernel.h>
31 #include <linux/sched/signal.h>
32 #include <linux/string.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
37 #include <linux/errno.h>
38 #include <linux/interrupt.h>
39 #include <linux/if_addr.h>
40 #include <linux/if_ether.h>
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/etherdevice.h>
44 #include <linux/skbuff.h>
45 #include <linux/init.h>
46 #include <linux/notifier.h>
47 #include <linux/inetdevice.h>
48 #include <linux/igmp.h>
49 #include <linux/slab.h>
50 #include <linux/hash.h>
52 #include <linux/sysctl.h>
54 #include <linux/kmod.h>
55 #include <linux/netconf.h>
59 #include <net/route.h>
60 #include <net/ip_fib.h>
61 #include <net/rtnetlink.h>
62 #include <net/net_namespace.h>
63 #include <net/addrconf.h>
65 #define IPV6ONLY_FLAGS \
66 (IFA_F_NODAD | IFA_F_OPTIMISTIC | IFA_F_DADFAILED | \
67 IFA_F_HOMEADDRESS | IFA_F_TENTATIVE | \
68 IFA_F_MANAGETEMPADDR | IFA_F_STABLE_PRIVACY)
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,
76 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
77 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
81 static struct ipv4_devconf ipv4_devconf_dflt = {
83 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
85 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
86 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
87 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
88 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
89 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
93 #define IPV4_DEVCONF_DFLT(net, attr) \
94 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
96 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
97 [IFA_LOCAL] = { .type = NLA_U32 },
98 [IFA_ADDRESS] = { .type = NLA_U32 },
99 [IFA_BROADCAST] = { .type = NLA_U32 },
100 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
101 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
102 [IFA_FLAGS] = { .type = NLA_U32 },
103 [IFA_RT_PRIORITY] = { .type = NLA_U32 },
104 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
107 struct inet_fill_args {
116 #define IN4_ADDR_HSIZE_SHIFT 8
117 #define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
119 static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
121 static u32 inet_addr_hash(const struct net *net, __be32 addr)
123 u32 val = (__force u32) addr ^ net_hash_mix(net);
125 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
128 static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
130 u32 hash = inet_addr_hash(net, ifa->ifa_local);
133 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
136 static void inet_hash_remove(struct in_ifaddr *ifa)
139 hlist_del_init_rcu(&ifa->hash);
143 * __ip_dev_find - find the first device with a given source address.
144 * @net: the net namespace
145 * @addr: the source address
146 * @devref: if true, take a reference on the found device
148 * If a caller uses devref=false, it should be protected by RCU, or RTNL
150 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
152 struct net_device *result = NULL;
153 struct in_ifaddr *ifa;
156 ifa = inet_lookup_ifaddr_rcu(net, addr);
158 struct flowi4 fl4 = { .daddr = addr };
159 struct fib_result res = { 0 };
160 struct fib_table *local;
162 /* Fallback to FIB local table so that communication
163 * over loopback subnets work.
165 local = fib_get_table(net, RT_TABLE_LOCAL);
167 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
168 res.type == RTN_LOCAL)
169 result = FIB_RES_DEV(res);
171 result = ifa->ifa_dev->dev;
173 if (result && devref)
178 EXPORT_SYMBOL(__ip_dev_find);
180 /* called under RCU lock */
181 struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
183 u32 hash = inet_addr_hash(net, addr);
184 struct in_ifaddr *ifa;
186 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash)
187 if (ifa->ifa_local == addr &&
188 net_eq(dev_net(ifa->ifa_dev->dev), net))
194 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
196 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
197 static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
198 static void inet_del_ifa(struct in_device *in_dev,
199 struct in_ifaddr __rcu **ifap,
202 static int devinet_sysctl_register(struct in_device *idev);
203 static void devinet_sysctl_unregister(struct in_device *idev);
205 static int devinet_sysctl_register(struct in_device *idev)
209 static void devinet_sysctl_unregister(struct in_device *idev)
214 /* Locks all the inet devices. */
216 static struct in_ifaddr *inet_alloc_ifa(void)
218 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
221 static void inet_rcu_free_ifa(struct rcu_head *head)
223 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
225 in_dev_put(ifa->ifa_dev);
229 static void inet_free_ifa(struct in_ifaddr *ifa)
231 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
234 void in_dev_finish_destroy(struct in_device *idev)
236 struct net_device *dev = idev->dev;
238 WARN_ON(idev->ifa_list);
239 WARN_ON(idev->mc_list);
240 kfree(rcu_dereference_protected(idev->mc_hash, 1));
241 #ifdef NET_REFCNT_DEBUG
242 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
246 pr_err("Freeing alive in_device %p\n", idev);
250 EXPORT_SYMBOL(in_dev_finish_destroy);
252 static struct in_device *inetdev_init(struct net_device *dev)
254 struct in_device *in_dev;
259 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
262 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
263 sizeof(in_dev->cnf));
264 in_dev->cnf.sysctl = NULL;
266 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
267 if (!in_dev->arp_parms)
269 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
270 dev_disable_lro(dev);
271 /* Reference in_dev->dev */
273 /* Account for reference dev->ip_ptr (below) */
274 refcount_set(&in_dev->refcnt, 1);
276 err = devinet_sysctl_register(in_dev);
279 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
284 ip_mc_init_dev(in_dev);
285 if (dev->flags & IFF_UP)
288 /* we can receive as soon as ip_ptr is set -- do this last */
289 rcu_assign_pointer(dev->ip_ptr, in_dev);
291 return in_dev ?: ERR_PTR(err);
298 static void in_dev_rcu_put(struct rcu_head *head)
300 struct in_device *idev = container_of(head, struct in_device, rcu_head);
304 static void inetdev_destroy(struct in_device *in_dev)
306 struct net_device *dev;
307 struct in_ifaddr *ifa;
315 ip_mc_destroy_dev(in_dev);
317 while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
318 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
322 RCU_INIT_POINTER(dev->ip_ptr, NULL);
324 devinet_sysctl_unregister(in_dev);
325 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
328 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
331 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
333 const struct in_ifaddr *ifa;
336 in_dev_for_each_ifa_rcu(ifa, in_dev) {
337 if (inet_ifa_match(a, ifa)) {
338 if (!b || inet_ifa_match(b, ifa)) {
348 static void __inet_del_ifa(struct in_device *in_dev,
349 struct in_ifaddr __rcu **ifap,
350 int destroy, struct nlmsghdr *nlh, u32 portid)
352 struct in_ifaddr *promote = NULL;
353 struct in_ifaddr *ifa, *ifa1;
354 struct in_ifaddr *last_prim;
355 struct in_ifaddr *prev_prom = NULL;
356 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
360 ifa1 = rtnl_dereference(*ifap);
361 last_prim = rtnl_dereference(in_dev->ifa_list);
365 /* 1. Deleting primary ifaddr forces deletion all secondaries
366 * unless alias promotion is set
369 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
370 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
372 while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
373 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
374 ifa1->ifa_scope <= ifa->ifa_scope)
377 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
378 ifa1->ifa_mask != ifa->ifa_mask ||
379 !inet_ifa_match(ifa1->ifa_address, ifa)) {
380 ifap1 = &ifa->ifa_next;
386 inet_hash_remove(ifa);
387 *ifap1 = ifa->ifa_next;
389 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
390 blocking_notifier_call_chain(&inetaddr_chain,
400 /* On promotion all secondaries from subnet are changing
401 * the primary IP, we must remove all their routes silently
402 * and later to add them back with new prefsrc. Do this
403 * while all addresses are on the device list.
405 for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
406 if (ifa1->ifa_mask == ifa->ifa_mask &&
407 inet_ifa_match(ifa1->ifa_address, ifa))
408 fib_del_ifaddr(ifa, ifa1);
414 *ifap = ifa1->ifa_next;
415 inet_hash_remove(ifa1);
417 /* 3. Announce address deletion */
419 /* Send message first, then call notifier.
420 At first sight, FIB update triggered by notifier
421 will refer to already deleted ifaddr, that could confuse
422 netlink listeners. It is not true: look, gated sees
423 that route deleted and if it still thinks that ifaddr
424 is valid, it will try to restore deleted routes... Grr.
425 So that, this order is correct.
427 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
428 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
431 struct in_ifaddr *next_sec;
433 next_sec = rtnl_dereference(promote->ifa_next);
435 struct in_ifaddr *last_sec;
437 rcu_assign_pointer(prev_prom->ifa_next, next_sec);
439 last_sec = rtnl_dereference(last_prim->ifa_next);
440 rcu_assign_pointer(promote->ifa_next, last_sec);
441 rcu_assign_pointer(last_prim->ifa_next, promote);
444 promote->ifa_flags &= ~IFA_F_SECONDARY;
445 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
446 blocking_notifier_call_chain(&inetaddr_chain,
448 for (ifa = next_sec; ifa;
449 ifa = rtnl_dereference(ifa->ifa_next)) {
450 if (ifa1->ifa_mask != ifa->ifa_mask ||
451 !inet_ifa_match(ifa1->ifa_address, ifa))
461 static void inet_del_ifa(struct in_device *in_dev,
462 struct in_ifaddr __rcu **ifap,
465 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
468 static void check_lifetime(struct work_struct *work);
470 static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
472 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
473 u32 portid, struct netlink_ext_ack *extack)
475 struct in_ifaddr __rcu **last_primary, **ifap;
476 struct in_device *in_dev = ifa->ifa_dev;
477 struct in_validator_info ivi;
478 struct in_ifaddr *ifa1;
483 if (!ifa->ifa_local) {
488 ifa->ifa_flags &= ~IFA_F_SECONDARY;
489 last_primary = &in_dev->ifa_list;
491 /* Don't set IPv6 only flags to IPv4 addresses */
492 ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
494 ifap = &in_dev->ifa_list;
495 ifa1 = rtnl_dereference(*ifap);
498 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
499 ifa->ifa_scope <= ifa1->ifa_scope)
500 last_primary = &ifa1->ifa_next;
501 if (ifa1->ifa_mask == ifa->ifa_mask &&
502 inet_ifa_match(ifa1->ifa_address, ifa)) {
503 if (ifa1->ifa_local == ifa->ifa_local) {
507 if (ifa1->ifa_scope != ifa->ifa_scope) {
511 ifa->ifa_flags |= IFA_F_SECONDARY;
514 ifap = &ifa1->ifa_next;
515 ifa1 = rtnl_dereference(*ifap);
518 /* Allow any devices that wish to register ifaddr validtors to weigh
519 * in now, before changes are committed. The rntl lock is serializing
520 * access here, so the state should not change between a validator call
521 * and a final notify on commit. This isn't invoked on promotion under
522 * the assumption that validators are checking the address itself, and
525 ivi.ivi_addr = ifa->ifa_address;
526 ivi.ivi_dev = ifa->ifa_dev;
528 ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
530 ret = notifier_to_errno(ret);
536 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
537 prandom_seed((__force u32) ifa->ifa_local);
541 rcu_assign_pointer(ifa->ifa_next, *ifap);
542 rcu_assign_pointer(*ifap, ifa);
544 inet_hash_insert(dev_net(in_dev->dev), ifa);
546 cancel_delayed_work(&check_lifetime_work);
547 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
549 /* Send message first, then call notifier.
550 Notifier will trigger FIB update, so that
551 listeners of netlink will know about new ifaddr */
552 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
553 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
558 static int inet_insert_ifa(struct in_ifaddr *ifa)
560 return __inet_insert_ifa(ifa, NULL, 0, NULL);
563 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
565 struct in_device *in_dev = __in_dev_get_rtnl(dev);
573 ipv4_devconf_setall(in_dev);
574 neigh_parms_data_state_setall(in_dev->arp_parms);
575 if (ifa->ifa_dev != in_dev) {
576 WARN_ON(ifa->ifa_dev);
578 ifa->ifa_dev = in_dev;
580 if (ipv4_is_loopback(ifa->ifa_local))
581 ifa->ifa_scope = RT_SCOPE_HOST;
582 return inet_insert_ifa(ifa);
585 /* Caller must hold RCU or RTNL :
586 * We dont take a reference on found in_device
588 struct in_device *inetdev_by_index(struct net *net, int ifindex)
590 struct net_device *dev;
591 struct in_device *in_dev = NULL;
594 dev = dev_get_by_index_rcu(net, ifindex);
596 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
600 EXPORT_SYMBOL(inetdev_by_index);
602 /* Called only from RTNL semaphored context. No locks. */
604 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
607 struct in_ifaddr *ifa;
611 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
612 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
618 static int ip_mc_autojoin_config(struct net *net, bool join,
619 const struct in_ifaddr *ifa)
621 #if defined(CONFIG_IP_MULTICAST)
622 struct ip_mreqn mreq = {
623 .imr_multiaddr.s_addr = ifa->ifa_address,
624 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
626 struct sock *sk = net->ipv4.mc_autojoin_sk;
633 ret = ip_mc_join_group(sk, &mreq);
635 ret = ip_mc_leave_group(sk, &mreq);
644 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
645 struct netlink_ext_ack *extack)
647 struct net *net = sock_net(skb->sk);
648 struct in_ifaddr __rcu **ifap;
649 struct nlattr *tb[IFA_MAX+1];
650 struct in_device *in_dev;
651 struct ifaddrmsg *ifm;
652 struct in_ifaddr *ifa;
657 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
658 ifa_ipv4_policy, extack);
662 ifm = nlmsg_data(nlh);
663 in_dev = inetdev_by_index(net, ifm->ifa_index);
669 for (ifap = &in_dev->ifa_list; (ifa = rtnl_dereference(*ifap)) != NULL;
670 ifap = &ifa->ifa_next) {
672 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
675 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
678 if (tb[IFA_ADDRESS] &&
679 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
680 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
683 if (ipv4_is_multicast(ifa->ifa_address))
684 ip_mc_autojoin_config(net, false, ifa);
685 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
689 err = -EADDRNOTAVAIL;
694 #define INFINITY_LIFE_TIME 0xFFFFFFFF
696 static void check_lifetime(struct work_struct *work)
698 unsigned long now, next, next_sec, next_sched;
699 struct in_ifaddr *ifa;
700 struct hlist_node *n;
704 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
706 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
707 bool change_needed = false;
710 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
713 if (ifa->ifa_flags & IFA_F_PERMANENT)
716 /* We try to batch several events at once. */
717 age = (now - ifa->ifa_tstamp +
718 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
720 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
721 age >= ifa->ifa_valid_lft) {
722 change_needed = true;
723 } else if (ifa->ifa_preferred_lft ==
724 INFINITY_LIFE_TIME) {
726 } else if (age >= ifa->ifa_preferred_lft) {
727 if (time_before(ifa->ifa_tstamp +
728 ifa->ifa_valid_lft * HZ, next))
729 next = ifa->ifa_tstamp +
730 ifa->ifa_valid_lft * HZ;
732 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
733 change_needed = true;
734 } else if (time_before(ifa->ifa_tstamp +
735 ifa->ifa_preferred_lft * HZ,
737 next = ifa->ifa_tstamp +
738 ifa->ifa_preferred_lft * HZ;
745 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
748 if (ifa->ifa_flags & IFA_F_PERMANENT)
751 /* We try to batch several events at once. */
752 age = (now - ifa->ifa_tstamp +
753 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
755 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
756 age >= ifa->ifa_valid_lft) {
757 struct in_ifaddr __rcu **ifap;
758 struct in_ifaddr *tmp;
760 ifap = &ifa->ifa_dev->ifa_list;
761 tmp = rtnl_dereference(*ifap);
764 inet_del_ifa(ifa->ifa_dev,
768 ifap = &tmp->ifa_next;
769 tmp = rtnl_dereference(*ifap);
771 } else if (ifa->ifa_preferred_lft !=
772 INFINITY_LIFE_TIME &&
773 age >= ifa->ifa_preferred_lft &&
774 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
775 ifa->ifa_flags |= IFA_F_DEPRECATED;
776 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
782 next_sec = round_jiffies_up(next);
785 /* If rounded timeout is accurate enough, accept it. */
786 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
787 next_sched = next_sec;
790 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
791 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
792 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
794 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
798 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
801 unsigned long timeout;
803 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
805 timeout = addrconf_timeout_fixup(valid_lft, HZ);
806 if (addrconf_finite_timeout(timeout))
807 ifa->ifa_valid_lft = timeout;
809 ifa->ifa_flags |= IFA_F_PERMANENT;
811 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
812 if (addrconf_finite_timeout(timeout)) {
814 ifa->ifa_flags |= IFA_F_DEPRECATED;
815 ifa->ifa_preferred_lft = timeout;
817 ifa->ifa_tstamp = jiffies;
818 if (!ifa->ifa_cstamp)
819 ifa->ifa_cstamp = ifa->ifa_tstamp;
822 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
823 __u32 *pvalid_lft, __u32 *pprefered_lft,
824 struct netlink_ext_ack *extack)
826 struct nlattr *tb[IFA_MAX+1];
827 struct in_ifaddr *ifa;
828 struct ifaddrmsg *ifm;
829 struct net_device *dev;
830 struct in_device *in_dev;
833 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
834 ifa_ipv4_policy, extack);
838 ifm = nlmsg_data(nlh);
840 if (ifm->ifa_prefixlen > 32 || !tb[IFA_LOCAL])
843 dev = __dev_get_by_index(net, ifm->ifa_index);
848 in_dev = __in_dev_get_rtnl(dev);
853 ifa = inet_alloc_ifa();
856 * A potential indev allocation can be left alive, it stays
857 * assigned to its device and is destroy with it.
861 ipv4_devconf_setall(in_dev);
862 neigh_parms_data_state_setall(in_dev->arp_parms);
865 if (!tb[IFA_ADDRESS])
866 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
868 INIT_HLIST_NODE(&ifa->hash);
869 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
870 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
871 ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
873 ifa->ifa_scope = ifm->ifa_scope;
874 ifa->ifa_dev = in_dev;
876 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
877 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
879 if (tb[IFA_BROADCAST])
880 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
883 nla_strscpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
885 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
887 if (tb[IFA_RT_PRIORITY])
888 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
890 if (tb[IFA_CACHEINFO]) {
891 struct ifa_cacheinfo *ci;
893 ci = nla_data(tb[IFA_CACHEINFO]);
894 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
898 *pvalid_lft = ci->ifa_valid;
899 *pprefered_lft = ci->ifa_prefered;
910 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
912 struct in_device *in_dev = ifa->ifa_dev;
913 struct in_ifaddr *ifa1;
918 in_dev_for_each_ifa_rtnl(ifa1, in_dev) {
919 if (ifa1->ifa_mask == ifa->ifa_mask &&
920 inet_ifa_match(ifa1->ifa_address, ifa) &&
921 ifa1->ifa_local == ifa->ifa_local)
927 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
928 struct netlink_ext_ack *extack)
930 struct net *net = sock_net(skb->sk);
931 struct in_ifaddr *ifa;
932 struct in_ifaddr *ifa_existing;
933 __u32 valid_lft = INFINITY_LIFE_TIME;
934 __u32 prefered_lft = INFINITY_LIFE_TIME;
938 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft, extack);
942 ifa_existing = find_matching_ifa(ifa);
944 /* It would be best to check for !NLM_F_CREATE here but
945 * userspace already relies on not having to provide this.
947 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
948 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
949 int ret = ip_mc_autojoin_config(net, true, ifa);
956 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
959 u32 new_metric = ifa->ifa_rt_priority;
963 if (nlh->nlmsg_flags & NLM_F_EXCL ||
964 !(nlh->nlmsg_flags & NLM_F_REPLACE))
968 if (ifa->ifa_rt_priority != new_metric) {
969 fib_modify_prefix_metric(ifa, new_metric);
970 ifa->ifa_rt_priority = new_metric;
973 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
974 cancel_delayed_work(&check_lifetime_work);
975 queue_delayed_work(system_power_efficient_wq,
976 &check_lifetime_work, 0);
977 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
983 * Determine a default network mask, based on the IP address.
986 static int inet_abc_len(__be32 addr)
988 int rc = -1; /* Something else, probably a multicast. */
990 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
993 __u32 haddr = ntohl(addr);
994 if (IN_CLASSA(haddr))
996 else if (IN_CLASSB(haddr))
998 else if (IN_CLASSC(haddr))
1000 else if (IN_CLASSE(haddr))
1008 int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1010 struct sockaddr_in sin_orig;
1011 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1012 struct in_ifaddr __rcu **ifap = NULL;
1013 struct in_device *in_dev;
1014 struct in_ifaddr *ifa = NULL;
1015 struct net_device *dev;
1018 int tryaddrmatch = 0;
1020 ifr->ifr_name[IFNAMSIZ - 1] = 0;
1022 /* save original address for comparison */
1023 memcpy(&sin_orig, sin, sizeof(*sin));
1025 colon = strchr(ifr->ifr_name, ':');
1029 dev_load(net, ifr->ifr_name);
1032 case SIOCGIFADDR: /* Get interface address */
1033 case SIOCGIFBRDADDR: /* Get the broadcast address */
1034 case SIOCGIFDSTADDR: /* Get the destination address */
1035 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1036 /* Note that these ioctls will not sleep,
1037 so that we do not impose a lock.
1038 One day we will be forced to put shlock here (I mean SMP)
1040 tryaddrmatch = (sin_orig.sin_family == AF_INET);
1041 memset(sin, 0, sizeof(*sin));
1042 sin->sin_family = AF_INET;
1047 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1050 case SIOCSIFADDR: /* Set interface address (and family) */
1051 case SIOCSIFBRDADDR: /* Set the broadcast address */
1052 case SIOCSIFDSTADDR: /* Set the destination address */
1053 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1055 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1058 if (sin->sin_family != AF_INET)
1069 dev = __dev_get_by_name(net, ifr->ifr_name);
1076 in_dev = __in_dev_get_rtnl(dev);
1079 /* Matthias Andree */
1080 /* compare label and address (4.4BSD style) */
1081 /* note: we only do this for a limited set of ioctls
1082 and only if the original address family was AF_INET.
1083 This is checked above. */
1085 for (ifap = &in_dev->ifa_list;
1086 (ifa = rtnl_dereference(*ifap)) != NULL;
1087 ifap = &ifa->ifa_next) {
1088 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1089 sin_orig.sin_addr.s_addr ==
1095 /* we didn't get a match, maybe the application is
1096 4.3BSD-style and passed in junk so we fall back to
1097 comparing just the label */
1099 for (ifap = &in_dev->ifa_list;
1100 (ifa = rtnl_dereference(*ifap)) != NULL;
1101 ifap = &ifa->ifa_next)
1102 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1107 ret = -EADDRNOTAVAIL;
1108 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1112 case SIOCGIFADDR: /* Get interface address */
1114 sin->sin_addr.s_addr = ifa->ifa_local;
1117 case SIOCGIFBRDADDR: /* Get the broadcast address */
1119 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1122 case SIOCGIFDSTADDR: /* Get the destination address */
1124 sin->sin_addr.s_addr = ifa->ifa_address;
1127 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1129 sin->sin_addr.s_addr = ifa->ifa_mask;
1134 ret = -EADDRNOTAVAIL;
1138 if (!(ifr->ifr_flags & IFF_UP))
1139 inet_del_ifa(in_dev, ifap, 1);
1142 ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1145 case SIOCSIFADDR: /* Set interface address (and family) */
1147 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1152 ifa = inet_alloc_ifa();
1155 INIT_HLIST_NODE(&ifa->hash);
1157 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1159 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1162 if (ifa->ifa_local == sin->sin_addr.s_addr)
1164 inet_del_ifa(in_dev, ifap, 0);
1165 ifa->ifa_broadcast = 0;
1169 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1171 if (!(dev->flags & IFF_POINTOPOINT)) {
1172 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1173 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1174 if ((dev->flags & IFF_BROADCAST) &&
1175 ifa->ifa_prefixlen < 31)
1176 ifa->ifa_broadcast = ifa->ifa_address |
1179 ifa->ifa_prefixlen = 32;
1180 ifa->ifa_mask = inet_make_mask(32);
1182 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1183 ret = inet_set_ifa(dev, ifa);
1186 case SIOCSIFBRDADDR: /* Set the broadcast address */
1188 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1189 inet_del_ifa(in_dev, ifap, 0);
1190 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1191 inet_insert_ifa(ifa);
1195 case SIOCSIFDSTADDR: /* Set the destination address */
1197 if (ifa->ifa_address == sin->sin_addr.s_addr)
1200 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1203 inet_del_ifa(in_dev, ifap, 0);
1204 ifa->ifa_address = sin->sin_addr.s_addr;
1205 inet_insert_ifa(ifa);
1208 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1211 * The mask we set must be legal.
1214 if (bad_mask(sin->sin_addr.s_addr, 0))
1217 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1218 __be32 old_mask = ifa->ifa_mask;
1219 inet_del_ifa(in_dev, ifap, 0);
1220 ifa->ifa_mask = sin->sin_addr.s_addr;
1221 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1223 /* See if current broadcast address matches
1224 * with current netmask, then recalculate
1225 * the broadcast address. Otherwise it's a
1226 * funny address, so don't touch it since
1227 * the user seems to know what (s)he's doing...
1229 if ((dev->flags & IFF_BROADCAST) &&
1230 (ifa->ifa_prefixlen < 31) &&
1231 (ifa->ifa_broadcast ==
1232 (ifa->ifa_local|~old_mask))) {
1233 ifa->ifa_broadcast = (ifa->ifa_local |
1234 ~sin->sin_addr.s_addr);
1236 inet_insert_ifa(ifa);
1246 static int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1248 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1249 const struct in_ifaddr *ifa;
1253 if (WARN_ON(size > sizeof(struct ifreq)))
1259 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1266 memset(&ifr, 0, sizeof(struct ifreq));
1267 strcpy(ifr.ifr_name, ifa->ifa_label);
1269 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1270 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1273 if (copy_to_user(buf + done, &ifr, size)) {
1284 static __be32 in_dev_select_addr(const struct in_device *in_dev,
1287 const struct in_ifaddr *ifa;
1289 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1290 if (ifa->ifa_flags & IFA_F_SECONDARY)
1292 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1293 ifa->ifa_scope <= scope)
1294 return ifa->ifa_local;
1300 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1302 const struct in_ifaddr *ifa;
1304 unsigned char localnet_scope = RT_SCOPE_HOST;
1305 struct in_device *in_dev;
1306 struct net *net = dev_net(dev);
1310 in_dev = __in_dev_get_rcu(dev);
1314 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1315 localnet_scope = RT_SCOPE_LINK;
1317 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1318 if (ifa->ifa_flags & IFA_F_SECONDARY)
1320 if (min(ifa->ifa_scope, localnet_scope) > scope)
1322 if (!dst || inet_ifa_match(dst, ifa)) {
1323 addr = ifa->ifa_local;
1327 addr = ifa->ifa_local;
1333 master_idx = l3mdev_master_ifindex_rcu(dev);
1335 /* For VRFs, the VRF device takes the place of the loopback device,
1336 * with addresses on it being preferred. Note in such cases the
1337 * loopback device will be among the devices that fail the master_idx
1338 * equality check in the loop below.
1341 (dev = dev_get_by_index_rcu(net, master_idx)) &&
1342 (in_dev = __in_dev_get_rcu(dev))) {
1343 addr = in_dev_select_addr(in_dev, scope);
1348 /* Not loopback addresses on loopback should be preferred
1349 in this case. It is important that lo is the first interface
1352 for_each_netdev_rcu(net, dev) {
1353 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1356 in_dev = __in_dev_get_rcu(dev);
1360 addr = in_dev_select_addr(in_dev, scope);
1368 EXPORT_SYMBOL(inet_select_addr);
1370 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1371 __be32 local, int scope)
1373 unsigned char localnet_scope = RT_SCOPE_HOST;
1374 const struct in_ifaddr *ifa;
1378 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1379 localnet_scope = RT_SCOPE_LINK;
1381 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1382 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1385 (local == ifa->ifa_local || !local) &&
1386 min_scope <= scope) {
1387 addr = ifa->ifa_local;
1392 same = (!local || inet_ifa_match(local, ifa)) &&
1393 (!dst || inet_ifa_match(dst, ifa));
1397 /* Is the selected addr into dst subnet? */
1398 if (inet_ifa_match(addr, ifa))
1400 /* No, then can we use new local src? */
1401 if (min_scope <= scope) {
1402 addr = ifa->ifa_local;
1405 /* search for large dst subnet for addr */
1411 return same ? addr : 0;
1415 * Confirm that local IP address exists using wildcards:
1416 * - net: netns to check, cannot be NULL
1417 * - in_dev: only on this interface, NULL=any interface
1418 * - dst: only in the same subnet as dst, 0=any dst
1419 * - local: address, 0=autoselect the local address
1420 * - scope: maximum allowed scope value for the local address
1422 __be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1423 __be32 dst, __be32 local, int scope)
1426 struct net_device *dev;
1429 return confirm_addr_indev(in_dev, dst, local, scope);
1432 for_each_netdev_rcu(net, dev) {
1433 in_dev = __in_dev_get_rcu(dev);
1435 addr = confirm_addr_indev(in_dev, dst, local, scope);
1444 EXPORT_SYMBOL(inet_confirm_addr);
1450 int register_inetaddr_notifier(struct notifier_block *nb)
1452 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1454 EXPORT_SYMBOL(register_inetaddr_notifier);
1456 int unregister_inetaddr_notifier(struct notifier_block *nb)
1458 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1460 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1462 int register_inetaddr_validator_notifier(struct notifier_block *nb)
1464 return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1466 EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1468 int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1470 return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1473 EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1475 /* Rename ifa_labels for a device name change. Make some effort to preserve
1476 * existing alias numbering and to create unique labels if possible.
1478 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1480 struct in_ifaddr *ifa;
1483 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1484 char old[IFNAMSIZ], *dot;
1486 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1487 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1490 dot = strchr(old, ':');
1492 sprintf(old, ":%d", named);
1495 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1496 strcat(ifa->ifa_label, dot);
1498 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1500 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1504 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1505 struct in_device *in_dev)
1508 const struct in_ifaddr *ifa;
1510 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1511 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1512 ifa->ifa_local, dev,
1513 ifa->ifa_local, NULL,
1514 dev->dev_addr, NULL);
1518 /* Called only under RTNL semaphore */
1520 static int inetdev_event(struct notifier_block *this, unsigned long event,
1523 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1524 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1529 if (event == NETDEV_REGISTER) {
1530 in_dev = inetdev_init(dev);
1532 return notifier_from_errno(PTR_ERR(in_dev));
1533 if (dev->flags & IFF_LOOPBACK) {
1534 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1535 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1537 } else if (event == NETDEV_CHANGEMTU) {
1538 /* Re-enabling IP */
1539 if (inetdev_valid_mtu(dev->mtu))
1540 in_dev = inetdev_init(dev);
1546 case NETDEV_REGISTER:
1547 pr_debug("%s: bug\n", __func__);
1548 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1551 if (!inetdev_valid_mtu(dev->mtu))
1553 if (dev->flags & IFF_LOOPBACK) {
1554 struct in_ifaddr *ifa = inet_alloc_ifa();
1557 INIT_HLIST_NODE(&ifa->hash);
1559 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1560 ifa->ifa_prefixlen = 8;
1561 ifa->ifa_mask = inet_make_mask(8);
1562 in_dev_hold(in_dev);
1563 ifa->ifa_dev = in_dev;
1564 ifa->ifa_scope = RT_SCOPE_HOST;
1565 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1566 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1567 INFINITY_LIFE_TIME);
1568 ipv4_devconf_setall(in_dev);
1569 neigh_parms_data_state_setall(in_dev->arp_parms);
1570 inet_insert_ifa(ifa);
1575 case NETDEV_CHANGEADDR:
1576 if (!IN_DEV_ARP_NOTIFY(in_dev))
1579 case NETDEV_NOTIFY_PEERS:
1580 /* Send gratuitous ARP to notify of link change */
1581 inetdev_send_gratuitous_arp(dev, in_dev);
1586 case NETDEV_PRE_TYPE_CHANGE:
1587 ip_mc_unmap(in_dev);
1589 case NETDEV_POST_TYPE_CHANGE:
1590 ip_mc_remap(in_dev);
1592 case NETDEV_CHANGEMTU:
1593 if (inetdev_valid_mtu(dev->mtu))
1595 /* disable IP when MTU is not enough */
1597 case NETDEV_UNREGISTER:
1598 inetdev_destroy(in_dev);
1600 case NETDEV_CHANGENAME:
1601 /* Do not notify about label change, this event is
1602 * not interesting to applications using netlink.
1604 inetdev_changename(dev, in_dev);
1606 devinet_sysctl_unregister(in_dev);
1607 devinet_sysctl_register(in_dev);
1614 static struct notifier_block ip_netdev_notifier = {
1615 .notifier_call = inetdev_event,
1618 static size_t inet_nlmsg_size(void)
1620 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1621 + nla_total_size(4) /* IFA_ADDRESS */
1622 + nla_total_size(4) /* IFA_LOCAL */
1623 + nla_total_size(4) /* IFA_BROADCAST */
1624 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1625 + nla_total_size(4) /* IFA_FLAGS */
1626 + nla_total_size(4) /* IFA_RT_PRIORITY */
1627 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1630 static inline u32 cstamp_delta(unsigned long cstamp)
1632 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1635 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1636 unsigned long tstamp, u32 preferred, u32 valid)
1638 struct ifa_cacheinfo ci;
1640 ci.cstamp = cstamp_delta(cstamp);
1641 ci.tstamp = cstamp_delta(tstamp);
1642 ci.ifa_prefered = preferred;
1643 ci.ifa_valid = valid;
1645 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1648 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1649 struct inet_fill_args *args)
1651 struct ifaddrmsg *ifm;
1652 struct nlmsghdr *nlh;
1653 u32 preferred, valid;
1655 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1660 ifm = nlmsg_data(nlh);
1661 ifm->ifa_family = AF_INET;
1662 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1663 ifm->ifa_flags = ifa->ifa_flags;
1664 ifm->ifa_scope = ifa->ifa_scope;
1665 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1667 if (args->netnsid >= 0 &&
1668 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1669 goto nla_put_failure;
1671 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1672 preferred = ifa->ifa_preferred_lft;
1673 valid = ifa->ifa_valid_lft;
1674 if (preferred != INFINITY_LIFE_TIME) {
1675 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1677 if (preferred > tval)
1681 if (valid != INFINITY_LIFE_TIME) {
1689 preferred = INFINITY_LIFE_TIME;
1690 valid = INFINITY_LIFE_TIME;
1692 if ((ifa->ifa_address &&
1693 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1695 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1696 (ifa->ifa_broadcast &&
1697 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1698 (ifa->ifa_label[0] &&
1699 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1700 nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1701 (ifa->ifa_rt_priority &&
1702 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1703 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1705 goto nla_put_failure;
1707 nlmsg_end(skb, nlh);
1711 nlmsg_cancel(skb, nlh);
1715 static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1716 struct inet_fill_args *fillargs,
1717 struct net **tgt_net, struct sock *sk,
1718 struct netlink_callback *cb)
1720 struct netlink_ext_ack *extack = cb->extack;
1721 struct nlattr *tb[IFA_MAX+1];
1722 struct ifaddrmsg *ifm;
1725 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1726 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1730 ifm = nlmsg_data(nlh);
1731 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1732 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1736 fillargs->ifindex = ifm->ifa_index;
1737 if (fillargs->ifindex) {
1738 cb->answer_flags |= NLM_F_DUMP_FILTERED;
1739 fillargs->flags |= NLM_F_DUMP_FILTERED;
1742 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1743 ifa_ipv4_policy, extack);
1747 for (i = 0; i <= IFA_MAX; ++i) {
1751 if (i == IFA_TARGET_NETNSID) {
1754 fillargs->netnsid = nla_get_s32(tb[i]);
1756 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1758 fillargs->netnsid = -1;
1759 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1760 return PTR_ERR(net);
1764 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1772 static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1773 struct netlink_callback *cb, int s_ip_idx,
1774 struct inet_fill_args *fillargs)
1776 struct in_ifaddr *ifa;
1780 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1781 if (ip_idx < s_ip_idx) {
1785 err = inet_fill_ifaddr(skb, ifa, fillargs);
1789 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1795 cb->args[2] = ip_idx;
1800 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1802 const struct nlmsghdr *nlh = cb->nlh;
1803 struct inet_fill_args fillargs = {
1804 .portid = NETLINK_CB(cb->skb).portid,
1805 .seq = nlh->nlmsg_seq,
1806 .event = RTM_NEWADDR,
1807 .flags = NLM_F_MULTI,
1810 struct net *net = sock_net(skb->sk);
1811 struct net *tgt_net = net;
1815 struct net_device *dev;
1816 struct in_device *in_dev;
1817 struct hlist_head *head;
1821 s_idx = idx = cb->args[1];
1822 s_ip_idx = cb->args[2];
1824 if (cb->strict_check) {
1825 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1831 if (fillargs.ifindex) {
1832 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
1838 in_dev = __in_dev_get_rtnl(dev);
1840 err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1847 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1849 head = &tgt_net->dev_index_head[h];
1851 cb->seq = atomic_read(&tgt_net->ipv4.dev_addr_genid) ^
1852 tgt_net->dev_base_seq;
1853 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1856 if (h > s_h || idx > s_idx)
1858 in_dev = __in_dev_get_rcu(dev);
1862 err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1878 if (fillargs.netnsid >= 0)
1881 return skb->len ? : err;
1884 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1887 struct inet_fill_args fillargs = {
1889 .seq = nlh ? nlh->nlmsg_seq : 0,
1894 struct sk_buff *skb;
1898 net = dev_net(ifa->ifa_dev->dev);
1899 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1903 err = inet_fill_ifaddr(skb, ifa, &fillargs);
1905 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1906 WARN_ON(err == -EMSGSIZE);
1910 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1914 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1917 static size_t inet_get_link_af_size(const struct net_device *dev,
1918 u32 ext_filter_mask)
1920 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1925 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1928 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1929 u32 ext_filter_mask)
1931 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1938 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1942 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1943 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1948 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1949 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1952 static int inet_validate_link_af(const struct net_device *dev,
1953 const struct nlattr *nla)
1955 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1958 if (dev && !__in_dev_get_rcu(dev))
1959 return -EAFNOSUPPORT;
1961 err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
1962 inet_af_policy, NULL);
1966 if (tb[IFLA_INET_CONF]) {
1967 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1968 int cfgid = nla_type(a);
1973 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1981 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1983 struct in_device *in_dev = __in_dev_get_rcu(dev);
1984 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1988 return -EAFNOSUPPORT;
1990 if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
1993 if (tb[IFLA_INET_CONF]) {
1994 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1995 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
2001 static int inet_netconf_msgsize_devconf(int type)
2003 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
2004 + nla_total_size(4); /* NETCONFA_IFINDEX */
2007 if (type == NETCONFA_ALL)
2010 if (all || type == NETCONFA_FORWARDING)
2011 size += nla_total_size(4);
2012 if (all || type == NETCONFA_RP_FILTER)
2013 size += nla_total_size(4);
2014 if (all || type == NETCONFA_MC_FORWARDING)
2015 size += nla_total_size(4);
2016 if (all || type == NETCONFA_BC_FORWARDING)
2017 size += nla_total_size(4);
2018 if (all || type == NETCONFA_PROXY_NEIGH)
2019 size += nla_total_size(4);
2020 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2021 size += nla_total_size(4);
2026 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2027 struct ipv4_devconf *devconf, u32 portid,
2028 u32 seq, int event, unsigned int flags,
2031 struct nlmsghdr *nlh;
2032 struct netconfmsg *ncm;
2035 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2040 if (type == NETCONFA_ALL)
2043 ncm = nlmsg_data(nlh);
2044 ncm->ncm_family = AF_INET;
2046 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2047 goto nla_put_failure;
2052 if ((all || type == NETCONFA_FORWARDING) &&
2053 nla_put_s32(skb, NETCONFA_FORWARDING,
2054 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
2055 goto nla_put_failure;
2056 if ((all || type == NETCONFA_RP_FILTER) &&
2057 nla_put_s32(skb, NETCONFA_RP_FILTER,
2058 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
2059 goto nla_put_failure;
2060 if ((all || type == NETCONFA_MC_FORWARDING) &&
2061 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2062 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
2063 goto nla_put_failure;
2064 if ((all || type == NETCONFA_BC_FORWARDING) &&
2065 nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2066 IPV4_DEVCONF(*devconf, BC_FORWARDING)) < 0)
2067 goto nla_put_failure;
2068 if ((all || type == NETCONFA_PROXY_NEIGH) &&
2069 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2070 IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
2071 goto nla_put_failure;
2072 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2073 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2074 IPV4_DEVCONF(*devconf, IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2075 goto nla_put_failure;
2078 nlmsg_end(skb, nlh);
2082 nlmsg_cancel(skb, nlh);
2086 void inet_netconf_notify_devconf(struct net *net, int event, int type,
2087 int ifindex, struct ipv4_devconf *devconf)
2089 struct sk_buff *skb;
2092 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2096 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2099 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2100 WARN_ON(err == -EMSGSIZE);
2104 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2108 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2111 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2112 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
2113 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
2114 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
2115 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
2116 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
2119 static int inet_netconf_valid_get_req(struct sk_buff *skb,
2120 const struct nlmsghdr *nlh,
2122 struct netlink_ext_ack *extack)
2126 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2127 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2131 if (!netlink_strict_get_check(skb))
2132 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2134 devconf_ipv4_policy, extack);
2136 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2138 devconf_ipv4_policy, extack);
2142 for (i = 0; i <= NETCONFA_MAX; i++) {
2147 case NETCONFA_IFINDEX:
2150 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2158 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2159 struct nlmsghdr *nlh,
2160 struct netlink_ext_ack *extack)
2162 struct net *net = sock_net(in_skb->sk);
2163 struct nlattr *tb[NETCONFA_MAX+1];
2164 struct sk_buff *skb;
2165 struct ipv4_devconf *devconf;
2166 struct in_device *in_dev;
2167 struct net_device *dev;
2171 err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2176 if (!tb[NETCONFA_IFINDEX])
2179 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2181 case NETCONFA_IFINDEX_ALL:
2182 devconf = net->ipv4.devconf_all;
2184 case NETCONFA_IFINDEX_DEFAULT:
2185 devconf = net->ipv4.devconf_dflt;
2188 dev = __dev_get_by_index(net, ifindex);
2191 in_dev = __in_dev_get_rtnl(dev);
2194 devconf = &in_dev->cnf;
2199 skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2203 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2204 NETLINK_CB(in_skb).portid,
2205 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2208 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2209 WARN_ON(err == -EMSGSIZE);
2213 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2218 static int inet_netconf_dump_devconf(struct sk_buff *skb,
2219 struct netlink_callback *cb)
2221 const struct nlmsghdr *nlh = cb->nlh;
2222 struct net *net = sock_net(skb->sk);
2225 struct net_device *dev;
2226 struct in_device *in_dev;
2227 struct hlist_head *head;
2229 if (cb->strict_check) {
2230 struct netlink_ext_ack *extack = cb->extack;
2231 struct netconfmsg *ncm;
2233 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2234 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2238 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2239 NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2245 s_idx = idx = cb->args[1];
2247 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2249 head = &net->dev_index_head[h];
2251 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
2253 hlist_for_each_entry_rcu(dev, head, index_hlist) {
2256 in_dev = __in_dev_get_rcu(dev);
2260 if (inet_netconf_fill_devconf(skb, dev->ifindex,
2262 NETLINK_CB(cb->skb).portid,
2266 NETCONFA_ALL) < 0) {
2270 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2276 if (h == NETDEV_HASHENTRIES) {
2277 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2278 net->ipv4.devconf_all,
2279 NETLINK_CB(cb->skb).portid,
2281 RTM_NEWNETCONF, NLM_F_MULTI,
2287 if (h == NETDEV_HASHENTRIES + 1) {
2288 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2289 net->ipv4.devconf_dflt,
2290 NETLINK_CB(cb->skb).portid,
2292 RTM_NEWNETCONF, NLM_F_MULTI,
2305 #ifdef CONFIG_SYSCTL
2307 static void devinet_copy_dflt_conf(struct net *net, int i)
2309 struct net_device *dev;
2312 for_each_netdev_rcu(net, dev) {
2313 struct in_device *in_dev;
2315 in_dev = __in_dev_get_rcu(dev);
2316 if (in_dev && !test_bit(i, in_dev->cnf.state))
2317 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2322 /* called with RTNL locked */
2323 static void inet_forward_change(struct net *net)
2325 struct net_device *dev;
2326 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2328 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2329 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2330 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2331 NETCONFA_FORWARDING,
2332 NETCONFA_IFINDEX_ALL,
2333 net->ipv4.devconf_all);
2334 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2335 NETCONFA_FORWARDING,
2336 NETCONFA_IFINDEX_DEFAULT,
2337 net->ipv4.devconf_dflt);
2339 for_each_netdev(net, dev) {
2340 struct in_device *in_dev;
2343 dev_disable_lro(dev);
2345 in_dev = __in_dev_get_rtnl(dev);
2347 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2348 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2349 NETCONFA_FORWARDING,
2350 dev->ifindex, &in_dev->cnf);
2355 static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2357 if (cnf == net->ipv4.devconf_dflt)
2358 return NETCONFA_IFINDEX_DEFAULT;
2359 else if (cnf == net->ipv4.devconf_all)
2360 return NETCONFA_IFINDEX_ALL;
2362 struct in_device *idev
2363 = container_of(cnf, struct in_device, cnf);
2364 return idev->dev->ifindex;
2368 static int devinet_conf_proc(struct ctl_table *ctl, int write,
2369 void *buffer, size_t *lenp, loff_t *ppos)
2371 int old_value = *(int *)ctl->data;
2372 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2373 int new_value = *(int *)ctl->data;
2376 struct ipv4_devconf *cnf = ctl->extra1;
2377 struct net *net = ctl->extra2;
2378 int i = (int *)ctl->data - cnf->data;
2381 set_bit(i, cnf->state);
2383 if (cnf == net->ipv4.devconf_dflt)
2384 devinet_copy_dflt_conf(net, i);
2385 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2386 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2387 if ((new_value == 0) && (old_value != 0))
2388 rt_cache_flush(net);
2390 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2391 new_value != old_value)
2392 rt_cache_flush(net);
2394 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2395 new_value != old_value) {
2396 ifindex = devinet_conf_ifindex(net, cnf);
2397 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2401 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2402 new_value != old_value) {
2403 ifindex = devinet_conf_ifindex(net, cnf);
2404 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2405 NETCONFA_PROXY_NEIGH,
2408 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2409 new_value != old_value) {
2410 ifindex = devinet_conf_ifindex(net, cnf);
2411 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2412 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2420 static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2421 void *buffer, size_t *lenp, loff_t *ppos)
2423 int *valp = ctl->data;
2426 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2428 if (write && *valp != val) {
2429 struct net *net = ctl->extra2;
2431 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2432 if (!rtnl_trylock()) {
2433 /* Restore the original values before restarting */
2436 return restart_syscall();
2438 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2439 inet_forward_change(net);
2441 struct ipv4_devconf *cnf = ctl->extra1;
2442 struct in_device *idev =
2443 container_of(cnf, struct in_device, cnf);
2445 dev_disable_lro(idev->dev);
2446 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2447 NETCONFA_FORWARDING,
2452 rt_cache_flush(net);
2454 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2455 NETCONFA_FORWARDING,
2456 NETCONFA_IFINDEX_DEFAULT,
2457 net->ipv4.devconf_dflt);
2463 static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2464 void *buffer, size_t *lenp, loff_t *ppos)
2466 int *valp = ctl->data;
2468 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2469 struct net *net = ctl->extra2;
2471 if (write && *valp != val)
2472 rt_cache_flush(net);
2477 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2480 .data = ipv4_devconf.data + \
2481 IPV4_DEVCONF_ ## attr - 1, \
2482 .maxlen = sizeof(int), \
2484 .proc_handler = proc, \
2485 .extra1 = &ipv4_devconf, \
2488 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2489 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2491 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2492 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2494 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2495 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2497 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2498 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2500 static struct devinet_sysctl_table {
2501 struct ctl_table_header *sysctl_header;
2502 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2503 } devinet_sysctl = {
2505 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2506 devinet_sysctl_forward),
2507 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2508 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2510 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2511 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2512 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2513 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2514 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2515 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2516 "accept_source_route"),
2517 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2518 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2519 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2520 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2521 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2522 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2523 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2524 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2525 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2526 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2527 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2528 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2529 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2530 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2531 "force_igmp_version"),
2532 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2533 "igmpv2_unsolicited_report_interval"),
2534 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2535 "igmpv3_unsolicited_report_interval"),
2536 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2537 "ignore_routes_with_linkdown"),
2538 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2539 "drop_gratuitous_arp"),
2541 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2542 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2543 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2544 "promote_secondaries"),
2545 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2547 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2548 "drop_unicast_in_l2_multicast"),
2552 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2553 int ifindex, struct ipv4_devconf *p)
2556 struct devinet_sysctl_table *t;
2557 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2559 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2563 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2564 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2565 t->devinet_vars[i].extra1 = p;
2566 t->devinet_vars[i].extra2 = net;
2569 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2571 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2572 if (!t->sysctl_header)
2577 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2587 static void __devinet_sysctl_unregister(struct net *net,
2588 struct ipv4_devconf *cnf, int ifindex)
2590 struct devinet_sysctl_table *t = cnf->sysctl;
2594 unregister_net_sysctl_table(t->sysctl_header);
2598 inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2601 static int devinet_sysctl_register(struct in_device *idev)
2605 if (!sysctl_dev_name_is_allowed(idev->dev->name))
2608 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2611 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2612 idev->dev->ifindex, &idev->cnf);
2614 neigh_sysctl_unregister(idev->arp_parms);
2618 static void devinet_sysctl_unregister(struct in_device *idev)
2620 struct net *net = dev_net(idev->dev);
2622 __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2623 neigh_sysctl_unregister(idev->arp_parms);
2626 static struct ctl_table ctl_forward_entry[] = {
2628 .procname = "ip_forward",
2629 .data = &ipv4_devconf.data[
2630 IPV4_DEVCONF_FORWARDING - 1],
2631 .maxlen = sizeof(int),
2633 .proc_handler = devinet_sysctl_forward,
2634 .extra1 = &ipv4_devconf,
2635 .extra2 = &init_net,
2641 static __net_init int devinet_init_net(struct net *net)
2644 struct ipv4_devconf *all, *dflt;
2645 #ifdef CONFIG_SYSCTL
2646 struct ctl_table *tbl;
2647 struct ctl_table_header *forw_hdr;
2651 all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2655 dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2657 goto err_alloc_dflt;
2659 #ifdef CONFIG_SYSCTL
2660 tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2664 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2665 tbl[0].extra1 = all;
2666 tbl[0].extra2 = net;
2669 if (!net_eq(net, &init_net)) {
2670 if (IS_ENABLED(CONFIG_SYSCTL) &&
2671 sysctl_devconf_inherit_init_net == 3) {
2672 /* copy from the current netns */
2673 memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all,
2674 sizeof(ipv4_devconf));
2676 current->nsproxy->net_ns->ipv4.devconf_dflt,
2677 sizeof(ipv4_devconf_dflt));
2678 } else if (!IS_ENABLED(CONFIG_SYSCTL) ||
2679 sysctl_devconf_inherit_init_net != 2) {
2680 /* inherit == 0 or 1: copy from init_net */
2681 memcpy(all, init_net.ipv4.devconf_all,
2682 sizeof(ipv4_devconf));
2683 memcpy(dflt, init_net.ipv4.devconf_dflt,
2684 sizeof(ipv4_devconf_dflt));
2686 /* else inherit == 2: use compiled values */
2689 #ifdef CONFIG_SYSCTL
2690 err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2694 err = __devinet_sysctl_register(net, "default",
2695 NETCONFA_IFINDEX_DEFAULT, dflt);
2700 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2703 net->ipv4.forw_hdr = forw_hdr;
2706 net->ipv4.devconf_all = all;
2707 net->ipv4.devconf_dflt = dflt;
2710 #ifdef CONFIG_SYSCTL
2712 __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2714 __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2726 static __net_exit void devinet_exit_net(struct net *net)
2728 #ifdef CONFIG_SYSCTL
2729 struct ctl_table *tbl;
2731 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2732 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2733 __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2734 NETCONFA_IFINDEX_DEFAULT);
2735 __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2736 NETCONFA_IFINDEX_ALL);
2739 kfree(net->ipv4.devconf_dflt);
2740 kfree(net->ipv4.devconf_all);
2743 static __net_initdata struct pernet_operations devinet_ops = {
2744 .init = devinet_init_net,
2745 .exit = devinet_exit_net,
2748 static struct rtnl_af_ops inet_af_ops __read_mostly = {
2750 .fill_link_af = inet_fill_link_af,
2751 .get_link_af_size = inet_get_link_af_size,
2752 .validate_link_af = inet_validate_link_af,
2753 .set_link_af = inet_set_link_af,
2756 void __init devinet_init(void)
2760 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2761 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2763 register_pernet_subsys(&devinet_ops);
2765 register_gifconf(PF_INET, inet_gifconf);
2766 register_netdevice_notifier(&ip_netdev_notifier);
2768 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
2770 rtnl_af_register(&inet_af_ops);
2772 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);
2773 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, 0);
2774 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, 0);
2775 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2776 inet_netconf_dump_devconf, 0);