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, void *arg)
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;
593 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
596 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
597 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
600 if (ifa->ifa_flags & IFA_F_PERMANENT)
603 /* We try to batch several events at once. */
604 age = (now - ifa->ifa_tstamp +
605 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
607 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
608 age >= ifa->ifa_valid_lft) {
609 struct in_ifaddr **ifap ;
612 for (ifap = &ifa->ifa_dev->ifa_list;
613 *ifap != NULL; ifap = &ifa->ifa_next) {
615 inet_del_ifa(ifa->ifa_dev,
619 } else if (ifa->ifa_preferred_lft ==
620 INFINITY_LIFE_TIME) {
622 } else if (age >= ifa->ifa_preferred_lft) {
623 if (time_before(ifa->ifa_tstamp +
624 ifa->ifa_valid_lft * HZ, next))
625 next = ifa->ifa_tstamp +
626 ifa->ifa_valid_lft * HZ;
628 if (!(ifa->ifa_flags & IFA_F_DEPRECATED)) {
629 ifa->ifa_flags |= IFA_F_DEPRECATED;
630 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
632 } else if (time_before(ifa->ifa_tstamp +
633 ifa->ifa_preferred_lft * HZ,
635 next = ifa->ifa_tstamp +
636 ifa->ifa_preferred_lft * HZ;
642 next_sec = round_jiffies_up(next);
645 /* If rounded timeout is accurate enough, accept it. */
646 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
647 next_sched = next_sec;
650 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
651 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
652 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
654 schedule_delayed_work(&check_lifetime_work, next_sched - now);
657 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
660 unsigned long timeout;
662 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
664 timeout = addrconf_timeout_fixup(valid_lft, HZ);
665 if (addrconf_finite_timeout(timeout))
666 ifa->ifa_valid_lft = timeout;
668 ifa->ifa_flags |= IFA_F_PERMANENT;
670 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
671 if (addrconf_finite_timeout(timeout)) {
673 ifa->ifa_flags |= IFA_F_DEPRECATED;
674 ifa->ifa_preferred_lft = timeout;
676 ifa->ifa_tstamp = jiffies;
677 if (!ifa->ifa_cstamp)
678 ifa->ifa_cstamp = ifa->ifa_tstamp;
681 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
682 __u32 *pvalid_lft, __u32 *pprefered_lft)
684 struct nlattr *tb[IFA_MAX+1];
685 struct in_ifaddr *ifa;
686 struct ifaddrmsg *ifm;
687 struct net_device *dev;
688 struct in_device *in_dev;
691 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
695 ifm = nlmsg_data(nlh);
697 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
700 dev = __dev_get_by_index(net, ifm->ifa_index);
705 in_dev = __in_dev_get_rtnl(dev);
710 ifa = inet_alloc_ifa();
713 * A potential indev allocation can be left alive, it stays
714 * assigned to its device and is destroy with it.
718 ipv4_devconf_setall(in_dev);
721 if (tb[IFA_ADDRESS] == NULL)
722 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
724 INIT_HLIST_NODE(&ifa->hash);
725 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
726 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
727 ifa->ifa_flags = ifm->ifa_flags;
728 ifa->ifa_scope = ifm->ifa_scope;
729 ifa->ifa_dev = in_dev;
731 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
732 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
734 if (tb[IFA_BROADCAST])
735 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
738 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
740 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
742 if (tb[IFA_CACHEINFO]) {
743 struct ifa_cacheinfo *ci;
745 ci = nla_data(tb[IFA_CACHEINFO]);
746 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
750 *pvalid_lft = ci->ifa_valid;
751 *pprefered_lft = ci->ifa_prefered;
760 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
762 struct in_device *in_dev = ifa->ifa_dev;
763 struct in_ifaddr *ifa1, **ifap;
768 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
769 ifap = &ifa1->ifa_next) {
770 if (ifa1->ifa_mask == ifa->ifa_mask &&
771 inet_ifa_match(ifa1->ifa_address, ifa) &&
772 ifa1->ifa_local == ifa->ifa_local)
778 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
780 struct net *net = sock_net(skb->sk);
781 struct in_ifaddr *ifa;
782 struct in_ifaddr *ifa_existing;
783 __u32 valid_lft = INFINITY_LIFE_TIME;
784 __u32 prefered_lft = INFINITY_LIFE_TIME;
788 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
792 ifa_existing = find_matching_ifa(ifa);
794 /* It would be best to check for !NLM_F_CREATE here but
795 * userspace alreay relies on not having to provide this.
797 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
798 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
802 if (nlh->nlmsg_flags & NLM_F_EXCL ||
803 !(nlh->nlmsg_flags & NLM_F_REPLACE))
806 set_ifa_lifetime(ifa_existing, valid_lft, prefered_lft);
812 * Determine a default network mask, based on the IP address.
815 static int inet_abc_len(__be32 addr)
817 int rc = -1; /* Something else, probably a multicast. */
819 if (ipv4_is_zeronet(addr))
822 __u32 haddr = ntohl(addr);
824 if (IN_CLASSA(haddr))
826 else if (IN_CLASSB(haddr))
828 else if (IN_CLASSC(haddr))
836 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
839 struct sockaddr_in sin_orig;
840 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
841 struct in_device *in_dev;
842 struct in_ifaddr **ifap = NULL;
843 struct in_ifaddr *ifa = NULL;
844 struct net_device *dev;
847 int tryaddrmatch = 0;
850 * Fetch the caller's info block into kernel space
853 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
855 ifr.ifr_name[IFNAMSIZ - 1] = 0;
857 /* save original address for comparison */
858 memcpy(&sin_orig, sin, sizeof(*sin));
860 colon = strchr(ifr.ifr_name, ':');
864 dev_load(net, ifr.ifr_name);
867 case SIOCGIFADDR: /* Get interface address */
868 case SIOCGIFBRDADDR: /* Get the broadcast address */
869 case SIOCGIFDSTADDR: /* Get the destination address */
870 case SIOCGIFNETMASK: /* Get the netmask for the interface */
871 /* Note that these ioctls will not sleep,
872 so that we do not impose a lock.
873 One day we will be forced to put shlock here (I mean SMP)
875 tryaddrmatch = (sin_orig.sin_family == AF_INET);
876 memset(sin, 0, sizeof(*sin));
877 sin->sin_family = AF_INET;
882 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
885 case SIOCSIFADDR: /* Set interface address (and family) */
886 case SIOCSIFBRDADDR: /* Set the broadcast address */
887 case SIOCSIFDSTADDR: /* Set the destination address */
888 case SIOCSIFNETMASK: /* Set the netmask for the interface */
890 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
893 if (sin->sin_family != AF_INET)
904 dev = __dev_get_by_name(net, ifr.ifr_name);
911 in_dev = __in_dev_get_rtnl(dev);
914 /* Matthias Andree */
915 /* compare label and address (4.4BSD style) */
916 /* note: we only do this for a limited set of ioctls
917 and only if the original address family was AF_INET.
918 This is checked above. */
919 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
920 ifap = &ifa->ifa_next) {
921 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
922 sin_orig.sin_addr.s_addr ==
928 /* we didn't get a match, maybe the application is
929 4.3BSD-style and passed in junk so we fall back to
930 comparing just the label */
932 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
933 ifap = &ifa->ifa_next)
934 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
939 ret = -EADDRNOTAVAIL;
940 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
944 case SIOCGIFADDR: /* Get interface address */
945 sin->sin_addr.s_addr = ifa->ifa_local;
948 case SIOCGIFBRDADDR: /* Get the broadcast address */
949 sin->sin_addr.s_addr = ifa->ifa_broadcast;
952 case SIOCGIFDSTADDR: /* Get the destination address */
953 sin->sin_addr.s_addr = ifa->ifa_address;
956 case SIOCGIFNETMASK: /* Get the netmask for the interface */
957 sin->sin_addr.s_addr = ifa->ifa_mask;
962 ret = -EADDRNOTAVAIL;
966 if (!(ifr.ifr_flags & IFF_UP))
967 inet_del_ifa(in_dev, ifap, 1);
970 ret = dev_change_flags(dev, ifr.ifr_flags);
973 case SIOCSIFADDR: /* Set interface address (and family) */
975 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
980 ifa = inet_alloc_ifa();
983 INIT_HLIST_NODE(&ifa->hash);
985 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
987 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
990 if (ifa->ifa_local == sin->sin_addr.s_addr)
992 inet_del_ifa(in_dev, ifap, 0);
993 ifa->ifa_broadcast = 0;
997 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
999 if (!(dev->flags & IFF_POINTOPOINT)) {
1000 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1001 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1002 if ((dev->flags & IFF_BROADCAST) &&
1003 ifa->ifa_prefixlen < 31)
1004 ifa->ifa_broadcast = ifa->ifa_address |
1007 ifa->ifa_prefixlen = 32;
1008 ifa->ifa_mask = inet_make_mask(32);
1010 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1011 ret = inet_set_ifa(dev, ifa);
1014 case SIOCSIFBRDADDR: /* Set the broadcast address */
1016 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1017 inet_del_ifa(in_dev, ifap, 0);
1018 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1019 inet_insert_ifa(ifa);
1023 case SIOCSIFDSTADDR: /* Set the destination address */
1025 if (ifa->ifa_address == sin->sin_addr.s_addr)
1028 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1031 inet_del_ifa(in_dev, ifap, 0);
1032 ifa->ifa_address = sin->sin_addr.s_addr;
1033 inet_insert_ifa(ifa);
1036 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1039 * The mask we set must be legal.
1042 if (bad_mask(sin->sin_addr.s_addr, 0))
1045 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1046 __be32 old_mask = ifa->ifa_mask;
1047 inet_del_ifa(in_dev, ifap, 0);
1048 ifa->ifa_mask = sin->sin_addr.s_addr;
1049 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1051 /* See if current broadcast address matches
1052 * with current netmask, then recalculate
1053 * the broadcast address. Otherwise it's a
1054 * funny address, so don't touch it since
1055 * the user seems to know what (s)he's doing...
1057 if ((dev->flags & IFF_BROADCAST) &&
1058 (ifa->ifa_prefixlen < 31) &&
1059 (ifa->ifa_broadcast ==
1060 (ifa->ifa_local|~old_mask))) {
1061 ifa->ifa_broadcast = (ifa->ifa_local |
1062 ~sin->sin_addr.s_addr);
1064 inet_insert_ifa(ifa);
1074 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
1078 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
1080 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1081 struct in_ifaddr *ifa;
1088 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1090 done += sizeof(ifr);
1093 if (len < (int) sizeof(ifr))
1095 memset(&ifr, 0, sizeof(struct ifreq));
1097 strcpy(ifr.ifr_name, ifa->ifa_label);
1099 strcpy(ifr.ifr_name, dev->name);
1101 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1102 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1105 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
1109 buf += sizeof(struct ifreq);
1110 len -= sizeof(struct ifreq);
1111 done += sizeof(struct ifreq);
1117 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1120 struct in_device *in_dev;
1121 struct net *net = dev_net(dev);
1124 in_dev = __in_dev_get_rcu(dev);
1128 for_primary_ifa(in_dev) {
1129 if (ifa->ifa_scope > scope)
1131 if (!dst || inet_ifa_match(dst, ifa)) {
1132 addr = ifa->ifa_local;
1136 addr = ifa->ifa_local;
1137 } endfor_ifa(in_dev);
1143 /* Not loopback addresses on loopback should be preferred
1144 in this case. It is importnat that lo is the first interface
1147 for_each_netdev_rcu(net, dev) {
1148 in_dev = __in_dev_get_rcu(dev);
1152 for_primary_ifa(in_dev) {
1153 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1154 ifa->ifa_scope <= scope) {
1155 addr = ifa->ifa_local;
1158 } endfor_ifa(in_dev);
1164 EXPORT_SYMBOL(inet_select_addr);
1166 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1167 __be32 local, int scope)
1174 (local == ifa->ifa_local || !local) &&
1175 ifa->ifa_scope <= scope) {
1176 addr = ifa->ifa_local;
1181 same = (!local || inet_ifa_match(local, ifa)) &&
1182 (!dst || inet_ifa_match(dst, ifa));
1186 /* Is the selected addr into dst subnet? */
1187 if (inet_ifa_match(addr, ifa))
1189 /* No, then can we use new local src? */
1190 if (ifa->ifa_scope <= scope) {
1191 addr = ifa->ifa_local;
1194 /* search for large dst subnet for addr */
1198 } endfor_ifa(in_dev);
1200 return same ? addr : 0;
1204 * Confirm that local IP address exists using wildcards:
1205 * - in_dev: only on this interface, 0=any interface
1206 * - dst: only in the same subnet as dst, 0=any dst
1207 * - local: address, 0=autoselect the local address
1208 * - scope: maximum allowed scope value for the local address
1210 __be32 inet_confirm_addr(struct in_device *in_dev,
1211 __be32 dst, __be32 local, int scope)
1214 struct net_device *dev;
1217 if (scope != RT_SCOPE_LINK)
1218 return confirm_addr_indev(in_dev, dst, local, scope);
1220 net = dev_net(in_dev->dev);
1222 for_each_netdev_rcu(net, dev) {
1223 in_dev = __in_dev_get_rcu(dev);
1225 addr = confirm_addr_indev(in_dev, dst, local, scope);
1234 EXPORT_SYMBOL(inet_confirm_addr);
1240 int register_inetaddr_notifier(struct notifier_block *nb)
1242 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1244 EXPORT_SYMBOL(register_inetaddr_notifier);
1246 int unregister_inetaddr_notifier(struct notifier_block *nb)
1248 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1250 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1252 /* Rename ifa_labels for a device name change. Make some effort to preserve
1253 * existing alias numbering and to create unique labels if possible.
1255 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1257 struct in_ifaddr *ifa;
1260 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1261 char old[IFNAMSIZ], *dot;
1263 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1264 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1267 dot = strchr(old, ':');
1269 sprintf(old, ":%d", named);
1272 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1273 strcat(ifa->ifa_label, dot);
1275 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1277 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1281 static bool inetdev_valid_mtu(unsigned int mtu)
1286 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1287 struct in_device *in_dev)
1290 struct in_ifaddr *ifa;
1292 for (ifa = in_dev->ifa_list; ifa;
1293 ifa = ifa->ifa_next) {
1294 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1295 ifa->ifa_local, dev,
1296 ifa->ifa_local, NULL,
1297 dev->dev_addr, NULL);
1301 /* Called only under RTNL semaphore */
1303 static int inetdev_event(struct notifier_block *this, unsigned long event,
1306 struct net_device *dev = ptr;
1307 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1312 if (event == NETDEV_REGISTER) {
1313 in_dev = inetdev_init(dev);
1315 return notifier_from_errno(-ENOMEM);
1316 if (dev->flags & IFF_LOOPBACK) {
1317 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1318 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1320 } else if (event == NETDEV_CHANGEMTU) {
1321 /* Re-enabling IP */
1322 if (inetdev_valid_mtu(dev->mtu))
1323 in_dev = inetdev_init(dev);
1329 case NETDEV_REGISTER:
1330 pr_debug("%s: bug\n", __func__);
1331 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1334 if (!inetdev_valid_mtu(dev->mtu))
1336 if (dev->flags & IFF_LOOPBACK) {
1337 struct in_ifaddr *ifa = inet_alloc_ifa();
1340 INIT_HLIST_NODE(&ifa->hash);
1342 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1343 ifa->ifa_prefixlen = 8;
1344 ifa->ifa_mask = inet_make_mask(8);
1345 in_dev_hold(in_dev);
1346 ifa->ifa_dev = in_dev;
1347 ifa->ifa_scope = RT_SCOPE_HOST;
1348 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1349 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1350 INFINITY_LIFE_TIME);
1351 inet_insert_ifa(ifa);
1356 case NETDEV_CHANGEADDR:
1357 if (!IN_DEV_ARP_NOTIFY(in_dev))
1360 case NETDEV_NOTIFY_PEERS:
1361 /* Send gratuitous ARP to notify of link change */
1362 inetdev_send_gratuitous_arp(dev, in_dev);
1367 case NETDEV_PRE_TYPE_CHANGE:
1368 ip_mc_unmap(in_dev);
1370 case NETDEV_POST_TYPE_CHANGE:
1371 ip_mc_remap(in_dev);
1373 case NETDEV_CHANGEMTU:
1374 if (inetdev_valid_mtu(dev->mtu))
1376 /* disable IP when MTU is not enough */
1377 case NETDEV_UNREGISTER:
1378 inetdev_destroy(in_dev);
1380 case NETDEV_CHANGENAME:
1381 /* Do not notify about label change, this event is
1382 * not interesting to applications using netlink.
1384 inetdev_changename(dev, in_dev);
1386 devinet_sysctl_unregister(in_dev);
1387 devinet_sysctl_register(in_dev);
1394 static struct notifier_block ip_netdev_notifier = {
1395 .notifier_call = inetdev_event,
1398 static size_t inet_nlmsg_size(void)
1400 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1401 + nla_total_size(4) /* IFA_ADDRESS */
1402 + nla_total_size(4) /* IFA_LOCAL */
1403 + nla_total_size(4) /* IFA_BROADCAST */
1404 + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
1407 static inline u32 cstamp_delta(unsigned long cstamp)
1409 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1412 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1413 unsigned long tstamp, u32 preferred, u32 valid)
1415 struct ifa_cacheinfo ci;
1417 ci.cstamp = cstamp_delta(cstamp);
1418 ci.tstamp = cstamp_delta(tstamp);
1419 ci.ifa_prefered = preferred;
1420 ci.ifa_valid = valid;
1422 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1425 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1426 u32 portid, u32 seq, int event, unsigned int flags)
1428 struct ifaddrmsg *ifm;
1429 struct nlmsghdr *nlh;
1430 u32 preferred, valid;
1432 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
1436 ifm = nlmsg_data(nlh);
1437 ifm->ifa_family = AF_INET;
1438 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1439 ifm->ifa_flags = ifa->ifa_flags;
1440 ifm->ifa_scope = ifa->ifa_scope;
1441 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1443 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1444 preferred = ifa->ifa_preferred_lft;
1445 valid = ifa->ifa_valid_lft;
1446 if (preferred != INFINITY_LIFE_TIME) {
1447 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1449 if (preferred > tval)
1453 if (valid != INFINITY_LIFE_TIME) {
1461 preferred = INFINITY_LIFE_TIME;
1462 valid = INFINITY_LIFE_TIME;
1464 if ((ifa->ifa_address &&
1465 nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1467 nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1468 (ifa->ifa_broadcast &&
1469 nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1470 (ifa->ifa_label[0] &&
1471 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1472 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1474 goto nla_put_failure;
1476 return nlmsg_end(skb, nlh);
1479 nlmsg_cancel(skb, nlh);
1483 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1485 struct net *net = sock_net(skb->sk);
1488 int ip_idx, s_ip_idx;
1489 struct net_device *dev;
1490 struct in_device *in_dev;
1491 struct in_ifaddr *ifa;
1492 struct hlist_head *head;
1495 s_idx = idx = cb->args[1];
1496 s_ip_idx = ip_idx = cb->args[2];
1498 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1500 head = &net->dev_index_head[h];
1502 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1505 if (h > s_h || idx > s_idx)
1507 in_dev = __in_dev_get_rcu(dev);
1511 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1512 ifa = ifa->ifa_next, ip_idx++) {
1513 if (ip_idx < s_ip_idx)
1515 if (inet_fill_ifaddr(skb, ifa,
1516 NETLINK_CB(cb->skb).portid,
1518 RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1532 cb->args[2] = ip_idx;
1537 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1540 struct sk_buff *skb;
1541 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1545 net = dev_net(ifa->ifa_dev->dev);
1546 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1550 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
1552 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1553 WARN_ON(err == -EMSGSIZE);
1557 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1561 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1564 static size_t inet_get_link_af_size(const struct net_device *dev)
1566 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1571 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1574 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
1576 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1583 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1587 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1588 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1593 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1594 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1597 static int inet_validate_link_af(const struct net_device *dev,
1598 const struct nlattr *nla)
1600 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1603 if (dev && !__in_dev_get_rtnl(dev))
1604 return -EAFNOSUPPORT;
1606 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1610 if (tb[IFLA_INET_CONF]) {
1611 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1612 int cfgid = nla_type(a);
1617 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1625 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1627 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1628 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1632 return -EAFNOSUPPORT;
1634 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1637 if (tb[IFLA_INET_CONF]) {
1638 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1639 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1645 static int inet_netconf_msgsize_devconf(int type)
1647 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1648 + nla_total_size(4); /* NETCONFA_IFINDEX */
1650 /* type -1 is used for ALL */
1651 if (type == -1 || type == NETCONFA_FORWARDING)
1652 size += nla_total_size(4);
1653 if (type == -1 || type == NETCONFA_RP_FILTER)
1654 size += nla_total_size(4);
1655 if (type == -1 || type == NETCONFA_MC_FORWARDING)
1656 size += nla_total_size(4);
1661 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1662 struct ipv4_devconf *devconf, u32 portid,
1663 u32 seq, int event, unsigned int flags,
1666 struct nlmsghdr *nlh;
1667 struct netconfmsg *ncm;
1669 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1674 ncm = nlmsg_data(nlh);
1675 ncm->ncm_family = AF_INET;
1677 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1678 goto nla_put_failure;
1680 /* type -1 is used for ALL */
1681 if ((type == -1 || type == NETCONFA_FORWARDING) &&
1682 nla_put_s32(skb, NETCONFA_FORWARDING,
1683 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1684 goto nla_put_failure;
1685 if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1686 nla_put_s32(skb, NETCONFA_RP_FILTER,
1687 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1688 goto nla_put_failure;
1689 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1690 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1691 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1692 goto nla_put_failure;
1694 return nlmsg_end(skb, nlh);
1697 nlmsg_cancel(skb, nlh);
1701 void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1702 struct ipv4_devconf *devconf)
1704 struct sk_buff *skb;
1707 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
1711 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1712 RTM_NEWNETCONF, 0, type);
1714 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1715 WARN_ON(err == -EMSGSIZE);
1719 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1723 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1726 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1727 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1728 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
1729 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
1732 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
1733 struct nlmsghdr *nlh,
1736 struct net *net = sock_net(in_skb->sk);
1737 struct nlattr *tb[NETCONFA_MAX+1];
1738 struct netconfmsg *ncm;
1739 struct sk_buff *skb;
1740 struct ipv4_devconf *devconf;
1741 struct in_device *in_dev;
1742 struct net_device *dev;
1746 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1747 devconf_ipv4_policy);
1752 if (!tb[NETCONFA_IFINDEX])
1755 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1757 case NETCONFA_IFINDEX_ALL:
1758 devconf = net->ipv4.devconf_all;
1760 case NETCONFA_IFINDEX_DEFAULT:
1761 devconf = net->ipv4.devconf_dflt;
1764 dev = __dev_get_by_index(net, ifindex);
1767 in_dev = __in_dev_get_rtnl(dev);
1770 devconf = &in_dev->cnf;
1775 skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
1779 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1780 NETLINK_CB(in_skb).portid,
1781 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1784 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1785 WARN_ON(err == -EMSGSIZE);
1789 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1794 #ifdef CONFIG_SYSCTL
1796 static void devinet_copy_dflt_conf(struct net *net, int i)
1798 struct net_device *dev;
1801 for_each_netdev_rcu(net, dev) {
1802 struct in_device *in_dev;
1804 in_dev = __in_dev_get_rcu(dev);
1805 if (in_dev && !test_bit(i, in_dev->cnf.state))
1806 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
1811 /* called with RTNL locked */
1812 static void inet_forward_change(struct net *net)
1814 struct net_device *dev;
1815 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1817 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1818 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
1819 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1820 NETCONFA_IFINDEX_ALL,
1821 net->ipv4.devconf_all);
1822 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1823 NETCONFA_IFINDEX_DEFAULT,
1824 net->ipv4.devconf_dflt);
1826 for_each_netdev(net, dev) {
1827 struct in_device *in_dev;
1829 dev_disable_lro(dev);
1831 in_dev = __in_dev_get_rcu(dev);
1833 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1834 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1835 dev->ifindex, &in_dev->cnf);
1841 static int devinet_conf_proc(ctl_table *ctl, int write,
1842 void __user *buffer,
1843 size_t *lenp, loff_t *ppos)
1845 int old_value = *(int *)ctl->data;
1846 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1847 int new_value = *(int *)ctl->data;
1850 struct ipv4_devconf *cnf = ctl->extra1;
1851 struct net *net = ctl->extra2;
1852 int i = (int *)ctl->data - cnf->data;
1854 set_bit(i, cnf->state);
1856 if (cnf == net->ipv4.devconf_dflt)
1857 devinet_copy_dflt_conf(net, i);
1858 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
1859 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
1860 if ((new_value == 0) && (old_value != 0))
1861 rt_cache_flush(net);
1862 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
1863 new_value != old_value) {
1866 if (cnf == net->ipv4.devconf_dflt)
1867 ifindex = NETCONFA_IFINDEX_DEFAULT;
1868 else if (cnf == net->ipv4.devconf_all)
1869 ifindex = NETCONFA_IFINDEX_ALL;
1871 struct in_device *idev =
1872 container_of(cnf, struct in_device,
1874 ifindex = idev->dev->ifindex;
1876 inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
1884 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1885 void __user *buffer,
1886 size_t *lenp, loff_t *ppos)
1888 int *valp = ctl->data;
1891 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1893 if (write && *valp != val) {
1894 struct net *net = ctl->extra2;
1896 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
1897 if (!rtnl_trylock()) {
1898 /* Restore the original values before restarting */
1901 return restart_syscall();
1903 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
1904 inet_forward_change(net);
1906 struct ipv4_devconf *cnf = ctl->extra1;
1907 struct in_device *idev =
1908 container_of(cnf, struct in_device, cnf);
1910 dev_disable_lro(idev->dev);
1911 inet_netconf_notify_devconf(net,
1912 NETCONFA_FORWARDING,
1917 rt_cache_flush(net);
1919 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1920 NETCONFA_IFINDEX_DEFAULT,
1921 net->ipv4.devconf_dflt);
1927 static int ipv4_doint_and_flush(ctl_table *ctl, int write,
1928 void __user *buffer,
1929 size_t *lenp, loff_t *ppos)
1931 int *valp = ctl->data;
1933 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1934 struct net *net = ctl->extra2;
1936 if (write && *valp != val)
1937 rt_cache_flush(net);
1942 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
1945 .data = ipv4_devconf.data + \
1946 IPV4_DEVCONF_ ## attr - 1, \
1947 .maxlen = sizeof(int), \
1949 .proc_handler = proc, \
1950 .extra1 = &ipv4_devconf, \
1953 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
1954 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
1956 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
1957 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
1959 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
1960 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
1962 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
1963 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
1965 static struct devinet_sysctl_table {
1966 struct ctl_table_header *sysctl_header;
1967 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
1968 } devinet_sysctl = {
1970 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
1971 devinet_sysctl_forward),
1972 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
1974 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
1975 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
1976 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
1977 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
1978 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
1979 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
1980 "accept_source_route"),
1981 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
1982 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
1983 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
1984 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
1985 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
1986 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
1987 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
1988 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
1989 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
1990 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
1991 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
1992 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
1993 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
1995 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
1996 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
1997 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
1998 "force_igmp_version"),
1999 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2000 "promote_secondaries"),
2001 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2006 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2007 struct ipv4_devconf *p)
2010 struct devinet_sysctl_table *t;
2011 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2013 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2017 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2018 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2019 t->devinet_vars[i].extra1 = p;
2020 t->devinet_vars[i].extra2 = net;
2023 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2025 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2026 if (!t->sysctl_header)
2038 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
2040 struct devinet_sysctl_table *t = cnf->sysctl;
2046 unregister_net_sysctl_table(t->sysctl_header);
2050 static void devinet_sysctl_register(struct in_device *idev)
2052 neigh_sysctl_register(idev->dev, idev->arp_parms, "ipv4", NULL);
2053 __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2057 static void devinet_sysctl_unregister(struct in_device *idev)
2059 __devinet_sysctl_unregister(&idev->cnf);
2060 neigh_sysctl_unregister(idev->arp_parms);
2063 static struct ctl_table ctl_forward_entry[] = {
2065 .procname = "ip_forward",
2066 .data = &ipv4_devconf.data[
2067 IPV4_DEVCONF_FORWARDING - 1],
2068 .maxlen = sizeof(int),
2070 .proc_handler = devinet_sysctl_forward,
2071 .extra1 = &ipv4_devconf,
2072 .extra2 = &init_net,
2078 static __net_init int devinet_init_net(struct net *net)
2081 struct ipv4_devconf *all, *dflt;
2082 #ifdef CONFIG_SYSCTL
2083 struct ctl_table *tbl = ctl_forward_entry;
2084 struct ctl_table_header *forw_hdr;
2088 all = &ipv4_devconf;
2089 dflt = &ipv4_devconf_dflt;
2091 if (!net_eq(net, &init_net)) {
2092 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
2096 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2098 goto err_alloc_dflt;
2100 #ifdef CONFIG_SYSCTL
2101 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
2105 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2106 tbl[0].extra1 = all;
2107 tbl[0].extra2 = net;
2111 #ifdef CONFIG_SYSCTL
2112 err = __devinet_sysctl_register(net, "all", all);
2116 err = __devinet_sysctl_register(net, "default", dflt);
2121 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2122 if (forw_hdr == NULL)
2124 net->ipv4.forw_hdr = forw_hdr;
2127 net->ipv4.devconf_all = all;
2128 net->ipv4.devconf_dflt = dflt;
2131 #ifdef CONFIG_SYSCTL
2133 __devinet_sysctl_unregister(dflt);
2135 __devinet_sysctl_unregister(all);
2137 if (tbl != ctl_forward_entry)
2141 if (dflt != &ipv4_devconf_dflt)
2144 if (all != &ipv4_devconf)
2150 static __net_exit void devinet_exit_net(struct net *net)
2152 #ifdef CONFIG_SYSCTL
2153 struct ctl_table *tbl;
2155 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2156 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2157 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
2158 __devinet_sysctl_unregister(net->ipv4.devconf_all);
2161 kfree(net->ipv4.devconf_dflt);
2162 kfree(net->ipv4.devconf_all);
2165 static __net_initdata struct pernet_operations devinet_ops = {
2166 .init = devinet_init_net,
2167 .exit = devinet_exit_net,
2170 static struct rtnl_af_ops inet_af_ops = {
2172 .fill_link_af = inet_fill_link_af,
2173 .get_link_af_size = inet_get_link_af_size,
2174 .validate_link_af = inet_validate_link_af,
2175 .set_link_af = inet_set_link_af,
2178 void __init devinet_init(void)
2182 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2183 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2185 register_pernet_subsys(&devinet_ops);
2187 register_gifconf(PF_INET, inet_gifconf);
2188 register_netdevice_notifier(&ip_netdev_notifier);
2190 schedule_delayed_work(&check_lifetime_work, 0);
2192 rtnl_af_register(&inet_af_ops);
2194 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
2195 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
2196 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
2197 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,