2 * NET3 IP device support routines.
4 * Version: $Id: devinet.c,v 1.44 2001/10/31 21:55:54 davem Exp $
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
11 * Derived from the IP parts of dev.c 1.0.19
13 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
14 * Mark Evans, <evansmp@uhura.aston.ac.uk>
17 * Alan Cox, <gw4pts@gw4pts.ampr.org>
18 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
21 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
23 * Cyrus Durgin: updated for kmod
24 * Matthias Andree: in devinet_ioctl, compare label and
25 * address (4.4BSD alias style support),
26 * fall back to comparing just the label
31 #include <asm/uaccess.h>
32 #include <asm/system.h>
33 #include <linux/bitops.h>
34 #include <linux/capability.h>
35 #include <linux/module.h>
36 #include <linux/types.h>
37 #include <linux/kernel.h>
38 #include <linux/string.h>
40 #include <linux/socket.h>
41 #include <linux/sockios.h>
43 #include <linux/errno.h>
44 #include <linux/interrupt.h>
45 #include <linux/if_addr.h>
46 #include <linux/if_ether.h>
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <linux/etherdevice.h>
50 #include <linux/skbuff.h>
51 #include <linux/init.h>
52 #include <linux/notifier.h>
53 #include <linux/inetdevice.h>
54 #include <linux/igmp.h>
56 #include <linux/sysctl.h>
58 #include <linux/kmod.h>
62 #include <net/route.h>
63 #include <net/ip_fib.h>
64 #include <net/rtnetlink.h>
65 #include <net/net_namespace.h>
67 static struct ipv4_devconf ipv4_devconf = {
69 [NET_IPV4_CONF_ACCEPT_REDIRECTS - 1] = 1,
70 [NET_IPV4_CONF_SEND_REDIRECTS - 1] = 1,
71 [NET_IPV4_CONF_SECURE_REDIRECTS - 1] = 1,
72 [NET_IPV4_CONF_SHARED_MEDIA - 1] = 1,
76 static struct ipv4_devconf ipv4_devconf_dflt = {
78 [NET_IPV4_CONF_ACCEPT_REDIRECTS - 1] = 1,
79 [NET_IPV4_CONF_SEND_REDIRECTS - 1] = 1,
80 [NET_IPV4_CONF_SECURE_REDIRECTS - 1] = 1,
81 [NET_IPV4_CONF_SHARED_MEDIA - 1] = 1,
82 [NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
86 #define IPV4_DEVCONF_DFLT(net, attr) \
87 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
89 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
90 [IFA_LOCAL] = { .type = NLA_U32 },
91 [IFA_ADDRESS] = { .type = NLA_U32 },
92 [IFA_BROADCAST] = { .type = NLA_U32 },
93 [IFA_ANYCAST] = { .type = NLA_U32 },
94 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
97 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
99 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
100 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
103 static void devinet_sysctl_register(struct in_device *idev);
104 static void devinet_sysctl_unregister(struct in_device *idev);
106 static inline void devinet_sysctl_register(struct in_device *idev)
109 static inline void devinet_sysctl_unregister(struct in_device *idev)
114 /* Locks all the inet devices. */
116 static struct in_ifaddr *inet_alloc_ifa(void)
118 struct in_ifaddr *ifa = kzalloc(sizeof(*ifa), GFP_KERNEL);
121 INIT_RCU_HEAD(&ifa->rcu_head);
127 static void inet_rcu_free_ifa(struct rcu_head *head)
129 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
131 in_dev_put(ifa->ifa_dev);
135 static inline void inet_free_ifa(struct in_ifaddr *ifa)
137 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
140 void in_dev_finish_destroy(struct in_device *idev)
142 struct net_device *dev = idev->dev;
144 BUG_TRAP(!idev->ifa_list);
145 BUG_TRAP(!idev->mc_list);
146 #ifdef NET_REFCNT_DEBUG
147 printk(KERN_DEBUG "in_dev_finish_destroy: %p=%s\n",
148 idev, dev ? dev->name : "NIL");
152 printk("Freeing alive in_device %p\n", idev);
158 static struct in_device *inetdev_init(struct net_device *dev)
160 struct in_device *in_dev;
164 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
167 INIT_RCU_HEAD(&in_dev->rcu_head);
168 memcpy(&in_dev->cnf, dev->nd_net->ipv4.devconf_dflt,
169 sizeof(in_dev->cnf));
170 in_dev->cnf.sysctl = NULL;
172 if ((in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl)) == NULL)
174 /* Reference in_dev->dev */
176 /* Account for reference dev->ip_ptr (below) */
179 devinet_sysctl_register(in_dev);
180 ip_mc_init_dev(in_dev);
181 if (dev->flags & IFF_UP)
184 /* we can receive as soon as ip_ptr is set -- do this last */
185 rcu_assign_pointer(dev->ip_ptr, in_dev);
194 static void in_dev_rcu_put(struct rcu_head *head)
196 struct in_device *idev = container_of(head, struct in_device, rcu_head);
200 static void inetdev_destroy(struct in_device *in_dev)
202 struct in_ifaddr *ifa;
203 struct net_device *dev;
211 ip_mc_destroy_dev(in_dev);
213 while ((ifa = in_dev->ifa_list) != NULL) {
214 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
220 devinet_sysctl_unregister(in_dev);
221 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
224 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
227 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
230 for_primary_ifa(in_dev) {
231 if (inet_ifa_match(a, ifa)) {
232 if (!b || inet_ifa_match(b, ifa)) {
237 } endfor_ifa(in_dev);
242 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
243 int destroy, struct nlmsghdr *nlh, u32 pid)
245 struct in_ifaddr *promote = NULL;
246 struct in_ifaddr *ifa, *ifa1 = *ifap;
247 struct in_ifaddr *last_prim = in_dev->ifa_list;
248 struct in_ifaddr *prev_prom = NULL;
249 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
253 /* 1. Deleting primary ifaddr forces deletion all secondaries
254 * unless alias promotion is set
257 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
258 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
260 while ((ifa = *ifap1) != NULL) {
261 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
262 ifa1->ifa_scope <= ifa->ifa_scope)
265 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
266 ifa1->ifa_mask != ifa->ifa_mask ||
267 !inet_ifa_match(ifa1->ifa_address, ifa)) {
268 ifap1 = &ifa->ifa_next;
274 *ifap1 = ifa->ifa_next;
276 rtmsg_ifa(RTM_DELADDR, ifa, nlh, pid);
277 blocking_notifier_call_chain(&inetaddr_chain,
289 *ifap = ifa1->ifa_next;
291 /* 3. Announce address deletion */
293 /* Send message first, then call notifier.
294 At first sight, FIB update triggered by notifier
295 will refer to already deleted ifaddr, that could confuse
296 netlink listeners. It is not true: look, gated sees
297 that route deleted and if it still thinks that ifaddr
298 is valid, it will try to restore deleted routes... Grr.
299 So that, this order is correct.
301 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, pid);
302 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
307 prev_prom->ifa_next = promote->ifa_next;
308 promote->ifa_next = last_prim->ifa_next;
309 last_prim->ifa_next = promote;
312 promote->ifa_flags &= ~IFA_F_SECONDARY;
313 rtmsg_ifa(RTM_NEWADDR, promote, nlh, pid);
314 blocking_notifier_call_chain(&inetaddr_chain,
316 for (ifa = promote->ifa_next; ifa; ifa = ifa->ifa_next) {
317 if (ifa1->ifa_mask != ifa->ifa_mask ||
318 !inet_ifa_match(ifa1->ifa_address, ifa))
328 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
331 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
334 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
337 struct in_device *in_dev = ifa->ifa_dev;
338 struct in_ifaddr *ifa1, **ifap, **last_primary;
342 if (!ifa->ifa_local) {
347 ifa->ifa_flags &= ~IFA_F_SECONDARY;
348 last_primary = &in_dev->ifa_list;
350 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
351 ifap = &ifa1->ifa_next) {
352 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
353 ifa->ifa_scope <= ifa1->ifa_scope)
354 last_primary = &ifa1->ifa_next;
355 if (ifa1->ifa_mask == ifa->ifa_mask &&
356 inet_ifa_match(ifa1->ifa_address, ifa)) {
357 if (ifa1->ifa_local == ifa->ifa_local) {
361 if (ifa1->ifa_scope != ifa->ifa_scope) {
365 ifa->ifa_flags |= IFA_F_SECONDARY;
369 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
370 net_srandom(ifa->ifa_local);
374 ifa->ifa_next = *ifap;
377 /* Send message first, then call notifier.
378 Notifier will trigger FIB update, so that
379 listeners of netlink will know about new ifaddr */
380 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, pid);
381 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
386 static int inet_insert_ifa(struct in_ifaddr *ifa)
388 return __inet_insert_ifa(ifa, NULL, 0);
391 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
393 struct in_device *in_dev = __in_dev_get_rtnl(dev);
401 ipv4_devconf_setall(in_dev);
402 if (ifa->ifa_dev != in_dev) {
403 BUG_TRAP(!ifa->ifa_dev);
405 ifa->ifa_dev = in_dev;
407 if (ipv4_is_loopback(ifa->ifa_local))
408 ifa->ifa_scope = RT_SCOPE_HOST;
409 return inet_insert_ifa(ifa);
412 struct in_device *inetdev_by_index(struct net *net, int ifindex)
414 struct net_device *dev;
415 struct in_device *in_dev = NULL;
416 read_lock(&dev_base_lock);
417 dev = __dev_get_by_index(net, ifindex);
419 in_dev = in_dev_get(dev);
420 read_unlock(&dev_base_lock);
424 /* Called only from RTNL semaphored context. No locks. */
426 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
431 for_primary_ifa(in_dev) {
432 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
434 } endfor_ifa(in_dev);
438 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
440 struct net *net = skb->sk->sk_net;
441 struct nlattr *tb[IFA_MAX+1];
442 struct in_device *in_dev;
443 struct ifaddrmsg *ifm;
444 struct in_ifaddr *ifa, **ifap;
449 if (net != &init_net)
452 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
456 ifm = nlmsg_data(nlh);
457 in_dev = inetdev_by_index(net, ifm->ifa_index);
458 if (in_dev == NULL) {
463 __in_dev_put(in_dev);
465 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
466 ifap = &ifa->ifa_next) {
468 ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
471 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
474 if (tb[IFA_ADDRESS] &&
475 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
476 !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
479 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).pid);
483 err = -EADDRNOTAVAIL;
488 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh)
490 struct nlattr *tb[IFA_MAX+1];
491 struct in_ifaddr *ifa;
492 struct ifaddrmsg *ifm;
493 struct net_device *dev;
494 struct in_device *in_dev;
497 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
501 ifm = nlmsg_data(nlh);
503 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
506 dev = __dev_get_by_index(net, ifm->ifa_index);
511 in_dev = __in_dev_get_rtnl(dev);
516 ifa = inet_alloc_ifa();
519 * A potential indev allocation can be left alive, it stays
520 * assigned to its device and is destroy with it.
524 ipv4_devconf_setall(in_dev);
527 if (tb[IFA_ADDRESS] == NULL)
528 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
530 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
531 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
532 ifa->ifa_flags = ifm->ifa_flags;
533 ifa->ifa_scope = ifm->ifa_scope;
534 ifa->ifa_dev = in_dev;
536 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
537 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
539 if (tb[IFA_BROADCAST])
540 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
543 ifa->ifa_anycast = nla_get_be32(tb[IFA_ANYCAST]);
546 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
548 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
556 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
558 struct net *net = skb->sk->sk_net;
559 struct in_ifaddr *ifa;
563 if (net != &init_net)
566 ifa = rtm_to_ifaddr(net, nlh);
570 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).pid);
574 * Determine a default network mask, based on the IP address.
577 static __inline__ int inet_abc_len(__be32 addr)
579 int rc = -1; /* Something else, probably a multicast. */
581 if (ipv4_is_zeronet(addr))
584 __u32 haddr = ntohl(addr);
586 if (IN_CLASSA(haddr))
588 else if (IN_CLASSB(haddr))
590 else if (IN_CLASSC(haddr))
598 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
601 struct sockaddr_in sin_orig;
602 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
603 struct in_device *in_dev;
604 struct in_ifaddr **ifap = NULL;
605 struct in_ifaddr *ifa = NULL;
606 struct net_device *dev;
609 int tryaddrmatch = 0;
612 * Fetch the caller's info block into kernel space
615 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
617 ifr.ifr_name[IFNAMSIZ - 1] = 0;
619 /* save original address for comparison */
620 memcpy(&sin_orig, sin, sizeof(*sin));
622 colon = strchr(ifr.ifr_name, ':');
627 dev_load(net, ifr.ifr_name);
631 case SIOCGIFADDR: /* Get interface address */
632 case SIOCGIFBRDADDR: /* Get the broadcast address */
633 case SIOCGIFDSTADDR: /* Get the destination address */
634 case SIOCGIFNETMASK: /* Get the netmask for the interface */
635 /* Note that these ioctls will not sleep,
636 so that we do not impose a lock.
637 One day we will be forced to put shlock here (I mean SMP)
639 tryaddrmatch = (sin_orig.sin_family == AF_INET);
640 memset(sin, 0, sizeof(*sin));
641 sin->sin_family = AF_INET;
646 if (!capable(CAP_NET_ADMIN))
649 case SIOCSIFADDR: /* Set interface address (and family) */
650 case SIOCSIFBRDADDR: /* Set the broadcast address */
651 case SIOCSIFDSTADDR: /* Set the destination address */
652 case SIOCSIFNETMASK: /* Set the netmask for the interface */
654 if (!capable(CAP_NET_ADMIN))
657 if (sin->sin_family != AF_INET)
668 if ((dev = __dev_get_by_name(net, ifr.ifr_name)) == NULL)
674 if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
676 /* Matthias Andree */
677 /* compare label and address (4.4BSD style) */
678 /* note: we only do this for a limited set of ioctls
679 and only if the original address family was AF_INET.
680 This is checked above. */
681 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
682 ifap = &ifa->ifa_next) {
683 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
684 sin_orig.sin_addr.s_addr ==
690 /* we didn't get a match, maybe the application is
691 4.3BSD-style and passed in junk so we fall back to
692 comparing just the label */
694 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
695 ifap = &ifa->ifa_next)
696 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
701 ret = -EADDRNOTAVAIL;
702 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
706 case SIOCGIFADDR: /* Get interface address */
707 sin->sin_addr.s_addr = ifa->ifa_local;
710 case SIOCGIFBRDADDR: /* Get the broadcast address */
711 sin->sin_addr.s_addr = ifa->ifa_broadcast;
714 case SIOCGIFDSTADDR: /* Get the destination address */
715 sin->sin_addr.s_addr = ifa->ifa_address;
718 case SIOCGIFNETMASK: /* Get the netmask for the interface */
719 sin->sin_addr.s_addr = ifa->ifa_mask;
724 ret = -EADDRNOTAVAIL;
728 if (!(ifr.ifr_flags & IFF_UP))
729 inet_del_ifa(in_dev, ifap, 1);
732 ret = dev_change_flags(dev, ifr.ifr_flags);
735 case SIOCSIFADDR: /* Set interface address (and family) */
737 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
742 if ((ifa = inet_alloc_ifa()) == NULL)
745 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
747 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
750 if (ifa->ifa_local == sin->sin_addr.s_addr)
752 inet_del_ifa(in_dev, ifap, 0);
753 ifa->ifa_broadcast = 0;
754 ifa->ifa_anycast = 0;
758 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
760 if (!(dev->flags & IFF_POINTOPOINT)) {
761 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
762 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
763 if ((dev->flags & IFF_BROADCAST) &&
764 ifa->ifa_prefixlen < 31)
765 ifa->ifa_broadcast = ifa->ifa_address |
768 ifa->ifa_prefixlen = 32;
769 ifa->ifa_mask = inet_make_mask(32);
771 ret = inet_set_ifa(dev, ifa);
774 case SIOCSIFBRDADDR: /* Set the broadcast address */
776 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
777 inet_del_ifa(in_dev, ifap, 0);
778 ifa->ifa_broadcast = sin->sin_addr.s_addr;
779 inet_insert_ifa(ifa);
783 case SIOCSIFDSTADDR: /* Set the destination address */
785 if (ifa->ifa_address == sin->sin_addr.s_addr)
788 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
791 inet_del_ifa(in_dev, ifap, 0);
792 ifa->ifa_address = sin->sin_addr.s_addr;
793 inet_insert_ifa(ifa);
796 case SIOCSIFNETMASK: /* Set the netmask for the interface */
799 * The mask we set must be legal.
802 if (bad_mask(sin->sin_addr.s_addr, 0))
805 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
806 __be32 old_mask = ifa->ifa_mask;
807 inet_del_ifa(in_dev, ifap, 0);
808 ifa->ifa_mask = sin->sin_addr.s_addr;
809 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
811 /* See if current broadcast address matches
812 * with current netmask, then recalculate
813 * the broadcast address. Otherwise it's a
814 * funny address, so don't touch it since
815 * the user seems to know what (s)he's doing...
817 if ((dev->flags & IFF_BROADCAST) &&
818 (ifa->ifa_prefixlen < 31) &&
819 (ifa->ifa_broadcast ==
820 (ifa->ifa_local|~old_mask))) {
821 ifa->ifa_broadcast = (ifa->ifa_local |
822 ~sin->sin_addr.s_addr);
824 inet_insert_ifa(ifa);
834 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
838 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
840 struct in_device *in_dev = __in_dev_get_rtnl(dev);
841 struct in_ifaddr *ifa;
845 if (!in_dev || (ifa = in_dev->ifa_list) == NULL)
848 for (; ifa; ifa = ifa->ifa_next) {
853 if (len < (int) sizeof(ifr))
855 memset(&ifr, 0, sizeof(struct ifreq));
857 strcpy(ifr.ifr_name, ifa->ifa_label);
859 strcpy(ifr.ifr_name, dev->name);
861 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
862 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
865 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
869 buf += sizeof(struct ifreq);
870 len -= sizeof(struct ifreq);
871 done += sizeof(struct ifreq);
877 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
880 struct in_device *in_dev;
883 in_dev = __in_dev_get_rcu(dev);
887 for_primary_ifa(in_dev) {
888 if (ifa->ifa_scope > scope)
890 if (!dst || inet_ifa_match(dst, ifa)) {
891 addr = ifa->ifa_local;
895 addr = ifa->ifa_local;
896 } endfor_ifa(in_dev);
903 /* Not loopback addresses on loopback should be preferred
904 in this case. It is importnat that lo is the first interface
907 read_lock(&dev_base_lock);
909 for_each_netdev(&init_net, dev) {
910 if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
913 for_primary_ifa(in_dev) {
914 if (ifa->ifa_scope != RT_SCOPE_LINK &&
915 ifa->ifa_scope <= scope) {
916 addr = ifa->ifa_local;
917 goto out_unlock_both;
919 } endfor_ifa(in_dev);
922 read_unlock(&dev_base_lock);
928 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
929 __be32 local, int scope)
936 (local == ifa->ifa_local || !local) &&
937 ifa->ifa_scope <= scope) {
938 addr = ifa->ifa_local;
943 same = (!local || inet_ifa_match(local, ifa)) &&
944 (!dst || inet_ifa_match(dst, ifa));
948 /* Is the selected addr into dst subnet? */
949 if (inet_ifa_match(addr, ifa))
951 /* No, then can we use new local src? */
952 if (ifa->ifa_scope <= scope) {
953 addr = ifa->ifa_local;
956 /* search for large dst subnet for addr */
960 } endfor_ifa(in_dev);
962 return same? addr : 0;
966 * Confirm that local IP address exists using wildcards:
967 * - in_dev: only on this interface, 0=any interface
968 * - dst: only in the same subnet as dst, 0=any dst
969 * - local: address, 0=autoselect the local address
970 * - scope: maximum allowed scope value for the local address
972 __be32 inet_confirm_addr(struct in_device *in_dev,
973 __be32 dst, __be32 local, int scope)
976 struct net_device *dev;
979 if (scope != RT_SCOPE_LINK)
980 return confirm_addr_indev(in_dev, dst, local, scope);
982 net = in_dev->dev->nd_net;
983 read_lock(&dev_base_lock);
985 for_each_netdev(net, dev) {
986 if ((in_dev = __in_dev_get_rcu(dev))) {
987 addr = confirm_addr_indev(in_dev, dst, local, scope);
993 read_unlock(&dev_base_lock);
1002 int register_inetaddr_notifier(struct notifier_block *nb)
1004 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1007 int unregister_inetaddr_notifier(struct notifier_block *nb)
1009 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1012 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
1013 * alias numbering and to create unique labels if possible.
1015 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1017 struct in_ifaddr *ifa;
1020 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1021 char old[IFNAMSIZ], *dot;
1023 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1024 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1027 dot = strchr(old, ':');
1029 sprintf(old, ":%d", named);
1032 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) {
1033 strcat(ifa->ifa_label, dot);
1035 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1040 /* Called only under RTNL semaphore */
1042 static int inetdev_event(struct notifier_block *this, unsigned long event,
1045 struct net_device *dev = ptr;
1046 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1051 if (event == NETDEV_REGISTER) {
1052 in_dev = inetdev_init(dev);
1054 return notifier_from_errno(-ENOMEM);
1055 if (dev->flags & IFF_LOOPBACK) {
1056 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1057 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1064 case NETDEV_REGISTER:
1065 printk(KERN_DEBUG "inetdev_event: bug\n");
1071 if (dev->flags & IFF_LOOPBACK) {
1072 struct in_ifaddr *ifa;
1073 if ((ifa = inet_alloc_ifa()) != NULL) {
1075 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1076 ifa->ifa_prefixlen = 8;
1077 ifa->ifa_mask = inet_make_mask(8);
1078 in_dev_hold(in_dev);
1079 ifa->ifa_dev = in_dev;
1080 ifa->ifa_scope = RT_SCOPE_HOST;
1081 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1082 inet_insert_ifa(ifa);
1090 case NETDEV_CHANGEMTU:
1093 /* MTU falled under 68, disable IP */
1094 case NETDEV_UNREGISTER:
1095 inetdev_destroy(in_dev);
1097 case NETDEV_CHANGENAME:
1098 /* Do not notify about label change, this event is
1099 * not interesting to applications using netlink.
1101 inetdev_changename(dev, in_dev);
1103 devinet_sysctl_unregister(in_dev);
1104 devinet_sysctl_register(in_dev);
1111 static struct notifier_block ip_netdev_notifier = {
1112 .notifier_call =inetdev_event,
1115 static inline size_t inet_nlmsg_size(void)
1117 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1118 + nla_total_size(4) /* IFA_ADDRESS */
1119 + nla_total_size(4) /* IFA_LOCAL */
1120 + nla_total_size(4) /* IFA_BROADCAST */
1121 + nla_total_size(4) /* IFA_ANYCAST */
1122 + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
1125 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1126 u32 pid, u32 seq, int event, unsigned int flags)
1128 struct ifaddrmsg *ifm;
1129 struct nlmsghdr *nlh;
1131 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*ifm), flags);
1135 ifm = nlmsg_data(nlh);
1136 ifm->ifa_family = AF_INET;
1137 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1138 ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
1139 ifm->ifa_scope = ifa->ifa_scope;
1140 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1142 if (ifa->ifa_address)
1143 NLA_PUT_BE32(skb, IFA_ADDRESS, ifa->ifa_address);
1146 NLA_PUT_BE32(skb, IFA_LOCAL, ifa->ifa_local);
1148 if (ifa->ifa_broadcast)
1149 NLA_PUT_BE32(skb, IFA_BROADCAST, ifa->ifa_broadcast);
1151 if (ifa->ifa_anycast)
1152 NLA_PUT_BE32(skb, IFA_ANYCAST, ifa->ifa_anycast);
1154 if (ifa->ifa_label[0])
1155 NLA_PUT_STRING(skb, IFA_LABEL, ifa->ifa_label);
1157 return nlmsg_end(skb, nlh);
1160 nlmsg_cancel(skb, nlh);
1164 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1166 struct net *net = skb->sk->sk_net;
1168 struct net_device *dev;
1169 struct in_device *in_dev;
1170 struct in_ifaddr *ifa;
1171 int s_ip_idx, s_idx = cb->args[0];
1173 if (net != &init_net)
1176 s_ip_idx = ip_idx = cb->args[1];
1178 for_each_netdev(net, dev) {
1183 if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
1186 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1187 ifa = ifa->ifa_next, ip_idx++) {
1188 if (ip_idx < s_ip_idx)
1190 if (inet_fill_ifaddr(skb, ifa, NETLINK_CB(cb->skb).pid,
1192 RTM_NEWADDR, NLM_F_MULTI) <= 0)
1201 cb->args[1] = ip_idx;
1206 static void rtmsg_ifa(int event, struct in_ifaddr* ifa, struct nlmsghdr *nlh,
1209 struct sk_buff *skb;
1210 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1214 net = ifa->ifa_dev->dev->nd_net;
1215 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1219 err = inet_fill_ifaddr(skb, ifa, pid, seq, event, 0);
1221 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1222 WARN_ON(err == -EMSGSIZE);
1226 err = rtnl_notify(skb, net, pid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1229 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1232 #ifdef CONFIG_SYSCTL
1234 static void devinet_copy_dflt_conf(struct net *net, int i)
1236 struct net_device *dev;
1238 read_lock(&dev_base_lock);
1239 for_each_netdev(net, dev) {
1240 struct in_device *in_dev;
1242 in_dev = __in_dev_get_rcu(dev);
1243 if (in_dev && !test_bit(i, in_dev->cnf.state))
1244 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
1247 read_unlock(&dev_base_lock);
1250 static void inet_forward_change(struct net *net)
1252 struct net_device *dev;
1253 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1255 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1256 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
1258 read_lock(&dev_base_lock);
1259 for_each_netdev(net, dev) {
1260 struct in_device *in_dev;
1262 in_dev = __in_dev_get_rcu(dev);
1264 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1267 read_unlock(&dev_base_lock);
1272 static int devinet_conf_proc(ctl_table *ctl, int write,
1273 struct file* filp, void __user *buffer,
1274 size_t *lenp, loff_t *ppos)
1276 int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1279 struct ipv4_devconf *cnf = ctl->extra1;
1280 struct net *net = ctl->extra2;
1281 int i = (int *)ctl->data - cnf->data;
1283 set_bit(i, cnf->state);
1285 if (cnf == net->ipv4.devconf_dflt)
1286 devinet_copy_dflt_conf(net, i);
1292 static int devinet_conf_sysctl(ctl_table *table, int __user *name, int nlen,
1293 void __user *oldval, size_t __user *oldlenp,
1294 void __user *newval, size_t newlen)
1296 struct ipv4_devconf *cnf;
1298 int *valp = table->data;
1302 if (!newval || !newlen)
1305 if (newlen != sizeof(int))
1308 if (get_user(new, (int __user *)newval))
1314 if (oldval && oldlenp) {
1317 if (get_user(len, oldlenp))
1321 if (len > table->maxlen)
1322 len = table->maxlen;
1323 if (copy_to_user(oldval, valp, len))
1325 if (put_user(len, oldlenp))
1332 cnf = table->extra1;
1333 net = table->extra2;
1334 i = (int *)table->data - cnf->data;
1336 set_bit(i, cnf->state);
1338 if (cnf == net->ipv4.devconf_dflt)
1339 devinet_copy_dflt_conf(net, i);
1344 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1345 struct file* filp, void __user *buffer,
1346 size_t *lenp, loff_t *ppos)
1348 int *valp = ctl->data;
1350 int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1352 if (write && *valp != val) {
1353 struct net *net = ctl->extra2;
1355 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING))
1356 inet_forward_change(net);
1357 else if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING))
1364 int ipv4_doint_and_flush(ctl_table *ctl, int write,
1365 struct file* filp, void __user *buffer,
1366 size_t *lenp, loff_t *ppos)
1368 int *valp = ctl->data;
1370 int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1372 if (write && *valp != val)
1378 int ipv4_doint_and_flush_strategy(ctl_table *table, int __user *name, int nlen,
1379 void __user *oldval, size_t __user *oldlenp,
1380 void __user *newval, size_t newlen)
1382 int ret = devinet_conf_sysctl(table, name, nlen, oldval, oldlenp,
1392 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc, sysctl) \
1394 .ctl_name = NET_IPV4_CONF_ ## attr, \
1396 .data = ipv4_devconf.data + \
1397 NET_IPV4_CONF_ ## attr - 1, \
1398 .maxlen = sizeof(int), \
1400 .proc_handler = proc, \
1401 .strategy = sysctl, \
1402 .extra1 = &ipv4_devconf, \
1405 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
1406 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc, \
1407 devinet_conf_sysctl)
1409 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
1410 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc, \
1411 devinet_conf_sysctl)
1413 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc, sysctl) \
1414 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc, sysctl)
1416 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
1417 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush, \
1418 ipv4_doint_and_flush_strategy)
1420 static struct devinet_sysctl_table {
1421 struct ctl_table_header *sysctl_header;
1422 struct ctl_table devinet_vars[__NET_IPV4_CONF_MAX];
1424 } devinet_sysctl = {
1426 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
1427 devinet_sysctl_forward,
1428 devinet_conf_sysctl),
1429 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
1431 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
1432 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
1433 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
1434 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
1435 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
1436 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
1437 "accept_source_route"),
1438 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
1439 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
1440 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
1441 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
1442 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
1443 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
1444 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
1445 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
1446 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
1448 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
1449 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
1450 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
1451 "force_igmp_version"),
1452 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
1453 "promote_secondaries"),
1457 static int __devinet_sysctl_register(struct net *net, char *dev_name,
1458 int ctl_name, struct ipv4_devconf *p)
1461 struct devinet_sysctl_table *t;
1463 #define DEVINET_CTL_PATH_DEV 3
1465 struct ctl_path devinet_ctl_path[] = {
1466 { .procname = "net", .ctl_name = CTL_NET, },
1467 { .procname = "ipv4", .ctl_name = NET_IPV4, },
1468 { .procname = "conf", .ctl_name = NET_IPV4_CONF, },
1469 { /* to be set */ },
1473 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1477 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
1478 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
1479 t->devinet_vars[i].extra1 = p;
1480 t->devinet_vars[i].extra2 = net;
1484 * Make a copy of dev_name, because '.procname' is regarded as const
1485 * by sysctl and we wouldn't want anyone to change it under our feet
1486 * (see SIOCSIFNAME).
1488 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
1492 devinet_ctl_path[DEVINET_CTL_PATH_DEV].procname = t->dev_name;
1493 devinet_ctl_path[DEVINET_CTL_PATH_DEV].ctl_name = ctl_name;
1495 t->sysctl_header = register_net_sysctl_table(net, devinet_ctl_path,
1497 if (!t->sysctl_header)
1511 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
1513 struct devinet_sysctl_table *t = cnf->sysctl;
1519 unregister_sysctl_table(t->sysctl_header);
1524 static void devinet_sysctl_register(struct in_device *idev)
1526 neigh_sysctl_register(idev->dev, idev->arp_parms, NET_IPV4,
1527 NET_IPV4_NEIGH, "ipv4", NULL, NULL);
1528 __devinet_sysctl_register(idev->dev->nd_net, idev->dev->name,
1529 idev->dev->ifindex, &idev->cnf);
1532 static void devinet_sysctl_unregister(struct in_device *idev)
1534 __devinet_sysctl_unregister(&idev->cnf);
1535 neigh_sysctl_unregister(idev->arp_parms);
1538 static struct ctl_table ctl_forward_entry[] = {
1540 .ctl_name = NET_IPV4_FORWARD,
1541 .procname = "ip_forward",
1542 .data = &ipv4_devconf.data[
1543 NET_IPV4_CONF_FORWARDING - 1],
1544 .maxlen = sizeof(int),
1546 .proc_handler = devinet_sysctl_forward,
1547 .strategy = devinet_conf_sysctl,
1548 .extra1 = &ipv4_devconf,
1549 .extra2 = &init_net,
1554 static __net_initdata struct ctl_path net_ipv4_path[] = {
1555 { .procname = "net", .ctl_name = CTL_NET, },
1556 { .procname = "ipv4", .ctl_name = NET_IPV4, },
1561 static __net_init int devinet_init_net(struct net *net)
1564 struct ipv4_devconf *all, *dflt;
1565 #ifdef CONFIG_SYSCTL
1566 struct ctl_table *tbl = ctl_forward_entry;
1567 struct ctl_table_header *forw_hdr;
1571 all = &ipv4_devconf;
1572 dflt = &ipv4_devconf_dflt;
1574 if (net != &init_net) {
1575 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
1579 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
1581 goto err_alloc_dflt;
1583 #ifdef CONFIG_SYSCTL
1584 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
1588 tbl[0].data = &all->data[NET_IPV4_CONF_FORWARDING - 1];
1589 tbl[0].extra1 = all;
1590 tbl[0].extra2 = net;
1594 #ifdef CONFIG_SYSCTL
1595 err = __devinet_sysctl_register(net, "all",
1596 NET_PROTO_CONF_ALL, all);
1600 err = __devinet_sysctl_register(net, "default",
1601 NET_PROTO_CONF_DEFAULT, dflt);
1606 forw_hdr = register_net_sysctl_table(net, net_ipv4_path, tbl);
1607 if (forw_hdr == NULL)
1609 net->ipv4.forw_hdr = forw_hdr;
1612 net->ipv4.devconf_all = all;
1613 net->ipv4.devconf_dflt = dflt;
1616 #ifdef CONFIG_SYSCTL
1618 __devinet_sysctl_unregister(dflt);
1620 __devinet_sysctl_unregister(all);
1622 if (tbl != ctl_forward_entry)
1626 if (dflt != &ipv4_devconf_dflt)
1629 if (all != &ipv4_devconf)
1635 static __net_exit void devinet_exit_net(struct net *net)
1637 #ifdef CONFIG_SYSCTL
1638 struct ctl_table *tbl;
1640 tbl = net->ipv4.forw_hdr->ctl_table_arg;
1641 unregister_net_sysctl_table(net->ipv4.forw_hdr);
1642 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
1643 __devinet_sysctl_unregister(net->ipv4.devconf_all);
1646 kfree(net->ipv4.devconf_dflt);
1647 kfree(net->ipv4.devconf_all);
1650 static __net_initdata struct pernet_operations devinet_ops = {
1651 .init = devinet_init_net,
1652 .exit = devinet_exit_net,
1655 void __init devinet_init(void)
1657 register_pernet_subsys(&devinet_ops);
1659 register_gifconf(PF_INET, inet_gifconf);
1660 register_netdevice_notifier(&ip_netdev_notifier);
1662 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL);
1663 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL);
1664 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr);
1667 EXPORT_SYMBOL(in_dev_finish_destroy);
1668 EXPORT_SYMBOL(inet_select_addr);
1669 EXPORT_SYMBOL(inetdev_by_index);
1670 EXPORT_SYMBOL(register_inetaddr_notifier);
1671 EXPORT_SYMBOL(unregister_inetaddr_notifier);