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>
66 struct ipv4_devconf ipv4_devconf
= {
68 [NET_IPV4_CONF_ACCEPT_REDIRECTS
- 1] = 1,
69 [NET_IPV4_CONF_SEND_REDIRECTS
- 1] = 1,
70 [NET_IPV4_CONF_SECURE_REDIRECTS
- 1] = 1,
71 [NET_IPV4_CONF_SHARED_MEDIA
- 1] = 1,
75 static struct ipv4_devconf ipv4_devconf_dflt
= {
77 [NET_IPV4_CONF_ACCEPT_REDIRECTS
- 1] = 1,
78 [NET_IPV4_CONF_SEND_REDIRECTS
- 1] = 1,
79 [NET_IPV4_CONF_SECURE_REDIRECTS
- 1] = 1,
80 [NET_IPV4_CONF_SHARED_MEDIA
- 1] = 1,
81 [NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE
- 1] = 1,
85 #define IPV4_DEVCONF_DFLT(attr) IPV4_DEVCONF(ipv4_devconf_dflt, attr)
87 static const struct nla_policy ifa_ipv4_policy
[IFA_MAX
+1] = {
88 [IFA_LOCAL
] = { .type
= NLA_U32
},
89 [IFA_ADDRESS
] = { .type
= NLA_U32
},
90 [IFA_BROADCAST
] = { .type
= NLA_U32
},
91 [IFA_ANYCAST
] = { .type
= NLA_U32
},
92 [IFA_LABEL
] = { .type
= NLA_STRING
, .len
= IFNAMSIZ
- 1 },
95 static void rtmsg_ifa(int event
, struct in_ifaddr
*, struct nlmsghdr
*, u32
);
97 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain
);
98 static void inet_del_ifa(struct in_device
*in_dev
, struct in_ifaddr
**ifap
,
101 static void devinet_sysctl_register(struct in_device
*in_dev
,
102 struct ipv4_devconf
*p
);
103 static void devinet_sysctl_unregister(struct ipv4_devconf
*p
);
106 /* Locks all the inet devices. */
108 static struct in_ifaddr
*inet_alloc_ifa(void)
110 struct in_ifaddr
*ifa
= kzalloc(sizeof(*ifa
), GFP_KERNEL
);
113 INIT_RCU_HEAD(&ifa
->rcu_head
);
119 static void inet_rcu_free_ifa(struct rcu_head
*head
)
121 struct in_ifaddr
*ifa
= container_of(head
, struct in_ifaddr
, rcu_head
);
123 in_dev_put(ifa
->ifa_dev
);
127 static inline void inet_free_ifa(struct in_ifaddr
*ifa
)
129 call_rcu(&ifa
->rcu_head
, inet_rcu_free_ifa
);
132 void in_dev_finish_destroy(struct in_device
*idev
)
134 struct net_device
*dev
= idev
->dev
;
136 BUG_TRAP(!idev
->ifa_list
);
137 BUG_TRAP(!idev
->mc_list
);
138 #ifdef NET_REFCNT_DEBUG
139 printk(KERN_DEBUG
"in_dev_finish_destroy: %p=%s\n",
140 idev
, dev
? dev
->name
: "NIL");
144 printk("Freeing alive in_device %p\n", idev
);
150 static struct in_device
*inetdev_init(struct net_device
*dev
)
152 struct in_device
*in_dev
;
156 in_dev
= kzalloc(sizeof(*in_dev
), GFP_KERNEL
);
159 INIT_RCU_HEAD(&in_dev
->rcu_head
);
160 memcpy(&in_dev
->cnf
, &ipv4_devconf_dflt
, sizeof(in_dev
->cnf
));
161 in_dev
->cnf
.sysctl
= NULL
;
163 if ((in_dev
->arp_parms
= neigh_parms_alloc(dev
, &arp_tbl
)) == NULL
)
165 /* Reference in_dev->dev */
168 neigh_sysctl_register(dev
, in_dev
->arp_parms
, NET_IPV4
,
169 NET_IPV4_NEIGH
, "ipv4", NULL
, NULL
);
172 /* Account for reference dev->ip_ptr (below) */
176 devinet_sysctl_register(in_dev
, &in_dev
->cnf
);
178 ip_mc_init_dev(in_dev
);
179 if (dev
->flags
& IFF_UP
)
182 /* we can receive as soon as ip_ptr is set -- do this last */
183 rcu_assign_pointer(dev
->ip_ptr
, in_dev
);
192 static void in_dev_rcu_put(struct rcu_head
*head
)
194 struct in_device
*idev
= container_of(head
, struct in_device
, rcu_head
);
198 static void inetdev_destroy(struct in_device
*in_dev
)
200 struct in_ifaddr
*ifa
;
201 struct net_device
*dev
;
206 if (dev
== &loopback_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);
219 devinet_sysctl_unregister(&in_dev
->cnf
);
225 neigh_sysctl_unregister(in_dev
->arp_parms
);
227 neigh_parms_release(&arp_tbl
, in_dev
->arp_parms
);
230 call_rcu(&in_dev
->rcu_head
, in_dev_rcu_put
);
233 int inet_addr_onlink(struct in_device
*in_dev
, __be32 a
, __be32 b
)
236 for_primary_ifa(in_dev
) {
237 if (inet_ifa_match(a
, ifa
)) {
238 if (!b
|| inet_ifa_match(b
, ifa
)) {
243 } endfor_ifa(in_dev
);
248 static void __inet_del_ifa(struct in_device
*in_dev
, struct in_ifaddr
**ifap
,
249 int destroy
, struct nlmsghdr
*nlh
, u32 pid
)
251 struct in_ifaddr
*promote
= NULL
;
252 struct in_ifaddr
*ifa
, *ifa1
= *ifap
;
253 struct in_ifaddr
*last_prim
= in_dev
->ifa_list
;
254 struct in_ifaddr
*prev_prom
= NULL
;
255 int do_promote
= IN_DEV_PROMOTE_SECONDARIES(in_dev
);
259 /* 1. Deleting primary ifaddr forces deletion all secondaries
260 * unless alias promotion is set
263 if (!(ifa1
->ifa_flags
& IFA_F_SECONDARY
)) {
264 struct in_ifaddr
**ifap1
= &ifa1
->ifa_next
;
266 while ((ifa
= *ifap1
) != NULL
) {
267 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
268 ifa1
->ifa_scope
<= ifa
->ifa_scope
)
271 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
) ||
272 ifa1
->ifa_mask
!= ifa
->ifa_mask
||
273 !inet_ifa_match(ifa1
->ifa_address
, ifa
)) {
274 ifap1
= &ifa
->ifa_next
;
280 *ifap1
= ifa
->ifa_next
;
282 rtmsg_ifa(RTM_DELADDR
, ifa
, nlh
, pid
);
283 blocking_notifier_call_chain(&inetaddr_chain
,
295 *ifap
= ifa1
->ifa_next
;
297 /* 3. Announce address deletion */
299 /* Send message first, then call notifier.
300 At first sight, FIB update triggered by notifier
301 will refer to already deleted ifaddr, that could confuse
302 netlink listeners. It is not true: look, gated sees
303 that route deleted and if it still thinks that ifaddr
304 is valid, it will try to restore deleted routes... Grr.
305 So that, this order is correct.
307 rtmsg_ifa(RTM_DELADDR
, ifa1
, nlh
, pid
);
308 blocking_notifier_call_chain(&inetaddr_chain
, NETDEV_DOWN
, ifa1
);
313 prev_prom
->ifa_next
= promote
->ifa_next
;
314 promote
->ifa_next
= last_prim
->ifa_next
;
315 last_prim
->ifa_next
= promote
;
318 promote
->ifa_flags
&= ~IFA_F_SECONDARY
;
319 rtmsg_ifa(RTM_NEWADDR
, promote
, nlh
, pid
);
320 blocking_notifier_call_chain(&inetaddr_chain
,
322 for (ifa
= promote
->ifa_next
; ifa
; ifa
= ifa
->ifa_next
) {
323 if (ifa1
->ifa_mask
!= ifa
->ifa_mask
||
324 !inet_ifa_match(ifa1
->ifa_address
, ifa
))
334 static void inet_del_ifa(struct in_device
*in_dev
, struct in_ifaddr
**ifap
,
337 __inet_del_ifa(in_dev
, ifap
, destroy
, NULL
, 0);
340 static int __inet_insert_ifa(struct in_ifaddr
*ifa
, struct nlmsghdr
*nlh
,
343 struct in_device
*in_dev
= ifa
->ifa_dev
;
344 struct in_ifaddr
*ifa1
, **ifap
, **last_primary
;
348 if (!ifa
->ifa_local
) {
353 ifa
->ifa_flags
&= ~IFA_F_SECONDARY
;
354 last_primary
= &in_dev
->ifa_list
;
356 for (ifap
= &in_dev
->ifa_list
; (ifa1
= *ifap
) != NULL
;
357 ifap
= &ifa1
->ifa_next
) {
358 if (!(ifa1
->ifa_flags
& IFA_F_SECONDARY
) &&
359 ifa
->ifa_scope
<= ifa1
->ifa_scope
)
360 last_primary
= &ifa1
->ifa_next
;
361 if (ifa1
->ifa_mask
== ifa
->ifa_mask
&&
362 inet_ifa_match(ifa1
->ifa_address
, ifa
)) {
363 if (ifa1
->ifa_local
== ifa
->ifa_local
) {
367 if (ifa1
->ifa_scope
!= ifa
->ifa_scope
) {
371 ifa
->ifa_flags
|= IFA_F_SECONDARY
;
375 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
)) {
376 net_srandom(ifa
->ifa_local
);
380 ifa
->ifa_next
= *ifap
;
383 /* Send message first, then call notifier.
384 Notifier will trigger FIB update, so that
385 listeners of netlink will know about new ifaddr */
386 rtmsg_ifa(RTM_NEWADDR
, ifa
, nlh
, pid
);
387 blocking_notifier_call_chain(&inetaddr_chain
, NETDEV_UP
, ifa
);
392 static int inet_insert_ifa(struct in_ifaddr
*ifa
)
394 return __inet_insert_ifa(ifa
, NULL
, 0);
397 static int inet_set_ifa(struct net_device
*dev
, struct in_ifaddr
*ifa
)
399 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
407 ipv4_devconf_setall(in_dev
);
408 if (ifa
->ifa_dev
!= in_dev
) {
409 BUG_TRAP(!ifa
->ifa_dev
);
411 ifa
->ifa_dev
= in_dev
;
413 if (LOOPBACK(ifa
->ifa_local
))
414 ifa
->ifa_scope
= RT_SCOPE_HOST
;
415 return inet_insert_ifa(ifa
);
418 struct in_device
*inetdev_by_index(int ifindex
)
420 struct net_device
*dev
;
421 struct in_device
*in_dev
= NULL
;
422 read_lock(&dev_base_lock
);
423 dev
= __dev_get_by_index(ifindex
);
425 in_dev
= in_dev_get(dev
);
426 read_unlock(&dev_base_lock
);
430 /* Called only from RTNL semaphored context. No locks. */
432 struct in_ifaddr
*inet_ifa_byprefix(struct in_device
*in_dev
, __be32 prefix
,
437 for_primary_ifa(in_dev
) {
438 if (ifa
->ifa_mask
== mask
&& inet_ifa_match(prefix
, ifa
))
440 } endfor_ifa(in_dev
);
444 static int inet_rtm_deladdr(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
446 struct nlattr
*tb
[IFA_MAX
+1];
447 struct in_device
*in_dev
;
448 struct ifaddrmsg
*ifm
;
449 struct in_ifaddr
*ifa
, **ifap
;
454 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFA_MAX
, ifa_ipv4_policy
);
458 ifm
= nlmsg_data(nlh
);
459 in_dev
= inetdev_by_index(ifm
->ifa_index
);
460 if (in_dev
== NULL
) {
465 __in_dev_put(in_dev
);
467 for (ifap
= &in_dev
->ifa_list
; (ifa
= *ifap
) != NULL
;
468 ifap
= &ifa
->ifa_next
) {
470 ifa
->ifa_local
!= nla_get_be32(tb
[IFA_LOCAL
]))
473 if (tb
[IFA_LABEL
] && nla_strcmp(tb
[IFA_LABEL
], ifa
->ifa_label
))
476 if (tb
[IFA_ADDRESS
] &&
477 (ifm
->ifa_prefixlen
!= ifa
->ifa_prefixlen
||
478 !inet_ifa_match(nla_get_be32(tb
[IFA_ADDRESS
]), ifa
)))
481 __inet_del_ifa(in_dev
, ifap
, 1, nlh
, NETLINK_CB(skb
).pid
);
485 err
= -EADDRNOTAVAIL
;
490 static struct in_ifaddr
*rtm_to_ifaddr(struct nlmsghdr
*nlh
)
492 struct nlattr
*tb
[IFA_MAX
+1];
493 struct in_ifaddr
*ifa
;
494 struct ifaddrmsg
*ifm
;
495 struct net_device
*dev
;
496 struct in_device
*in_dev
;
499 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFA_MAX
, ifa_ipv4_policy
);
503 ifm
= nlmsg_data(nlh
);
504 if (ifm
->ifa_prefixlen
> 32 || tb
[IFA_LOCAL
] == NULL
) {
509 dev
= __dev_get_by_index(ifm
->ifa_index
);
515 in_dev
= __in_dev_get_rtnl(dev
);
516 if (in_dev
== NULL
) {
521 ipv4_devconf_setall(in_dev
);
523 ifa
= inet_alloc_ifa();
526 * A potential indev allocation can be left alive, it stays
527 * assigned to its device and is destroy with it.
535 if (tb
[IFA_ADDRESS
] == NULL
)
536 tb
[IFA_ADDRESS
] = tb
[IFA_LOCAL
];
538 ifa
->ifa_prefixlen
= ifm
->ifa_prefixlen
;
539 ifa
->ifa_mask
= inet_make_mask(ifm
->ifa_prefixlen
);
540 ifa
->ifa_flags
= ifm
->ifa_flags
;
541 ifa
->ifa_scope
= ifm
->ifa_scope
;
542 ifa
->ifa_dev
= in_dev
;
544 ifa
->ifa_local
= nla_get_be32(tb
[IFA_LOCAL
]);
545 ifa
->ifa_address
= nla_get_be32(tb
[IFA_ADDRESS
]);
547 if (tb
[IFA_BROADCAST
])
548 ifa
->ifa_broadcast
= nla_get_be32(tb
[IFA_BROADCAST
]);
551 ifa
->ifa_anycast
= nla_get_be32(tb
[IFA_ANYCAST
]);
554 nla_strlcpy(ifa
->ifa_label
, tb
[IFA_LABEL
], IFNAMSIZ
);
556 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
564 static int inet_rtm_newaddr(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
566 struct in_ifaddr
*ifa
;
570 ifa
= rtm_to_ifaddr(nlh
);
574 return __inet_insert_ifa(ifa
, nlh
, NETLINK_CB(skb
).pid
);
578 * Determine a default network mask, based on the IP address.
581 static __inline__
int inet_abc_len(__be32 addr
)
583 int rc
= -1; /* Something else, probably a multicast. */
588 __u32 haddr
= ntohl(addr
);
590 if (IN_CLASSA(haddr
))
592 else if (IN_CLASSB(haddr
))
594 else if (IN_CLASSC(haddr
))
602 int devinet_ioctl(unsigned int cmd
, void __user
*arg
)
605 struct sockaddr_in sin_orig
;
606 struct sockaddr_in
*sin
= (struct sockaddr_in
*)&ifr
.ifr_addr
;
607 struct in_device
*in_dev
;
608 struct in_ifaddr
**ifap
= NULL
;
609 struct in_ifaddr
*ifa
= NULL
;
610 struct net_device
*dev
;
613 int tryaddrmatch
= 0;
616 * Fetch the caller's info block into kernel space
619 if (copy_from_user(&ifr
, arg
, sizeof(struct ifreq
)))
621 ifr
.ifr_name
[IFNAMSIZ
- 1] = 0;
623 /* save original address for comparison */
624 memcpy(&sin_orig
, sin
, sizeof(*sin
));
626 colon
= strchr(ifr
.ifr_name
, ':');
631 dev_load(ifr
.ifr_name
);
635 case SIOCGIFADDR
: /* Get interface address */
636 case SIOCGIFBRDADDR
: /* Get the broadcast address */
637 case SIOCGIFDSTADDR
: /* Get the destination address */
638 case SIOCGIFNETMASK
: /* Get the netmask for the interface */
639 /* Note that these ioctls will not sleep,
640 so that we do not impose a lock.
641 One day we will be forced to put shlock here (I mean SMP)
643 tryaddrmatch
= (sin_orig
.sin_family
== AF_INET
);
644 memset(sin
, 0, sizeof(*sin
));
645 sin
->sin_family
= AF_INET
;
650 if (!capable(CAP_NET_ADMIN
))
653 case SIOCSIFADDR
: /* Set interface address (and family) */
654 case SIOCSIFBRDADDR
: /* Set the broadcast address */
655 case SIOCSIFDSTADDR
: /* Set the destination address */
656 case SIOCSIFNETMASK
: /* Set the netmask for the interface */
658 if (!capable(CAP_NET_ADMIN
))
661 if (sin
->sin_family
!= AF_INET
)
672 if ((dev
= __dev_get_by_name(ifr
.ifr_name
)) == NULL
)
678 if ((in_dev
= __in_dev_get_rtnl(dev
)) != NULL
) {
680 /* Matthias Andree */
681 /* compare label and address (4.4BSD style) */
682 /* note: we only do this for a limited set of ioctls
683 and only if the original address family was AF_INET.
684 This is checked above. */
685 for (ifap
= &in_dev
->ifa_list
; (ifa
= *ifap
) != NULL
;
686 ifap
= &ifa
->ifa_next
) {
687 if (!strcmp(ifr
.ifr_name
, ifa
->ifa_label
) &&
688 sin_orig
.sin_addr
.s_addr
==
694 /* we didn't get a match, maybe the application is
695 4.3BSD-style and passed in junk so we fall back to
696 comparing just the label */
698 for (ifap
= &in_dev
->ifa_list
; (ifa
= *ifap
) != NULL
;
699 ifap
= &ifa
->ifa_next
)
700 if (!strcmp(ifr
.ifr_name
, ifa
->ifa_label
))
705 ret
= -EADDRNOTAVAIL
;
706 if (!ifa
&& cmd
!= SIOCSIFADDR
&& cmd
!= SIOCSIFFLAGS
)
710 case SIOCGIFADDR
: /* Get interface address */
711 sin
->sin_addr
.s_addr
= ifa
->ifa_local
;
714 case SIOCGIFBRDADDR
: /* Get the broadcast address */
715 sin
->sin_addr
.s_addr
= ifa
->ifa_broadcast
;
718 case SIOCGIFDSTADDR
: /* Get the destination address */
719 sin
->sin_addr
.s_addr
= ifa
->ifa_address
;
722 case SIOCGIFNETMASK
: /* Get the netmask for the interface */
723 sin
->sin_addr
.s_addr
= ifa
->ifa_mask
;
728 ret
= -EADDRNOTAVAIL
;
732 if (!(ifr
.ifr_flags
& IFF_UP
))
733 inet_del_ifa(in_dev
, ifap
, 1);
736 ret
= dev_change_flags(dev
, ifr
.ifr_flags
);
739 case SIOCSIFADDR
: /* Set interface address (and family) */
741 if (inet_abc_len(sin
->sin_addr
.s_addr
) < 0)
746 if ((ifa
= inet_alloc_ifa()) == NULL
)
749 memcpy(ifa
->ifa_label
, ifr
.ifr_name
, IFNAMSIZ
);
751 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
754 if (ifa
->ifa_local
== sin
->sin_addr
.s_addr
)
756 inet_del_ifa(in_dev
, ifap
, 0);
757 ifa
->ifa_broadcast
= 0;
758 ifa
->ifa_anycast
= 0;
761 ifa
->ifa_address
= ifa
->ifa_local
= sin
->sin_addr
.s_addr
;
763 if (!(dev
->flags
& IFF_POINTOPOINT
)) {
764 ifa
->ifa_prefixlen
= inet_abc_len(ifa
->ifa_address
);
765 ifa
->ifa_mask
= inet_make_mask(ifa
->ifa_prefixlen
);
766 if ((dev
->flags
& IFF_BROADCAST
) &&
767 ifa
->ifa_prefixlen
< 31)
768 ifa
->ifa_broadcast
= ifa
->ifa_address
|
771 ifa
->ifa_prefixlen
= 32;
772 ifa
->ifa_mask
= inet_make_mask(32);
774 ret
= inet_set_ifa(dev
, ifa
);
777 case SIOCSIFBRDADDR
: /* Set the broadcast address */
779 if (ifa
->ifa_broadcast
!= sin
->sin_addr
.s_addr
) {
780 inet_del_ifa(in_dev
, ifap
, 0);
781 ifa
->ifa_broadcast
= sin
->sin_addr
.s_addr
;
782 inet_insert_ifa(ifa
);
786 case SIOCSIFDSTADDR
: /* Set the destination address */
788 if (ifa
->ifa_address
== sin
->sin_addr
.s_addr
)
791 if (inet_abc_len(sin
->sin_addr
.s_addr
) < 0)
794 inet_del_ifa(in_dev
, ifap
, 0);
795 ifa
->ifa_address
= sin
->sin_addr
.s_addr
;
796 inet_insert_ifa(ifa
);
799 case SIOCSIFNETMASK
: /* Set the netmask for the interface */
802 * The mask we set must be legal.
805 if (bad_mask(sin
->sin_addr
.s_addr
, 0))
808 if (ifa
->ifa_mask
!= sin
->sin_addr
.s_addr
) {
809 __be32 old_mask
= ifa
->ifa_mask
;
810 inet_del_ifa(in_dev
, ifap
, 0);
811 ifa
->ifa_mask
= sin
->sin_addr
.s_addr
;
812 ifa
->ifa_prefixlen
= inet_mask_len(ifa
->ifa_mask
);
814 /* See if current broadcast address matches
815 * with current netmask, then recalculate
816 * the broadcast address. Otherwise it's a
817 * funny address, so don't touch it since
818 * the user seems to know what (s)he's doing...
820 if ((dev
->flags
& IFF_BROADCAST
) &&
821 (ifa
->ifa_prefixlen
< 31) &&
822 (ifa
->ifa_broadcast
==
823 (ifa
->ifa_local
|~old_mask
))) {
824 ifa
->ifa_broadcast
= (ifa
->ifa_local
|
825 ~sin
->sin_addr
.s_addr
);
827 inet_insert_ifa(ifa
);
837 ret
= copy_to_user(arg
, &ifr
, sizeof(struct ifreq
)) ? -EFAULT
: 0;
841 static int inet_gifconf(struct net_device
*dev
, char __user
*buf
, int len
)
843 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
844 struct in_ifaddr
*ifa
;
848 if (!in_dev
|| (ifa
= in_dev
->ifa_list
) == NULL
)
851 for (; ifa
; ifa
= ifa
->ifa_next
) {
856 if (len
< (int) sizeof(ifr
))
858 memset(&ifr
, 0, sizeof(struct ifreq
));
860 strcpy(ifr
.ifr_name
, ifa
->ifa_label
);
862 strcpy(ifr
.ifr_name
, dev
->name
);
864 (*(struct sockaddr_in
*)&ifr
.ifr_addr
).sin_family
= AF_INET
;
865 (*(struct sockaddr_in
*)&ifr
.ifr_addr
).sin_addr
.s_addr
=
868 if (copy_to_user(buf
, &ifr
, sizeof(struct ifreq
))) {
872 buf
+= sizeof(struct ifreq
);
873 len
-= sizeof(struct ifreq
);
874 done
+= sizeof(struct ifreq
);
880 __be32
inet_select_addr(const struct net_device
*dev
, __be32 dst
, int scope
)
883 struct in_device
*in_dev
;
886 in_dev
= __in_dev_get_rcu(dev
);
890 for_primary_ifa(in_dev
) {
891 if (ifa
->ifa_scope
> scope
)
893 if (!dst
|| inet_ifa_match(dst
, ifa
)) {
894 addr
= ifa
->ifa_local
;
898 addr
= ifa
->ifa_local
;
899 } endfor_ifa(in_dev
);
906 /* Not loopback addresses on loopback should be preferred
907 in this case. It is importnat that lo is the first interface
910 read_lock(&dev_base_lock
);
912 for_each_netdev(dev
) {
913 if ((in_dev
= __in_dev_get_rcu(dev
)) == NULL
)
916 for_primary_ifa(in_dev
) {
917 if (ifa
->ifa_scope
!= RT_SCOPE_LINK
&&
918 ifa
->ifa_scope
<= scope
) {
919 addr
= ifa
->ifa_local
;
920 goto out_unlock_both
;
922 } endfor_ifa(in_dev
);
925 read_unlock(&dev_base_lock
);
931 static __be32
confirm_addr_indev(struct in_device
*in_dev
, __be32 dst
,
932 __be32 local
, int scope
)
939 (local
== ifa
->ifa_local
|| !local
) &&
940 ifa
->ifa_scope
<= scope
) {
941 addr
= ifa
->ifa_local
;
946 same
= (!local
|| inet_ifa_match(local
, ifa
)) &&
947 (!dst
|| inet_ifa_match(dst
, ifa
));
951 /* Is the selected addr into dst subnet? */
952 if (inet_ifa_match(addr
, ifa
))
954 /* No, then can we use new local src? */
955 if (ifa
->ifa_scope
<= scope
) {
956 addr
= ifa
->ifa_local
;
959 /* search for large dst subnet for addr */
963 } endfor_ifa(in_dev
);
965 return same
? addr
: 0;
969 * Confirm that local IP address exists using wildcards:
970 * - dev: only on this interface, 0=any interface
971 * - dst: only in the same subnet as dst, 0=any dst
972 * - local: address, 0=autoselect the local address
973 * - scope: maximum allowed scope value for the local address
975 __be32
inet_confirm_addr(const struct net_device
*dev
, __be32 dst
, __be32 local
, int scope
)
978 struct in_device
*in_dev
;
982 if ((in_dev
= __in_dev_get_rcu(dev
)))
983 addr
= confirm_addr_indev(in_dev
, dst
, local
, scope
);
989 read_lock(&dev_base_lock
);
991 for_each_netdev(dev
) {
992 if ((in_dev
= __in_dev_get_rcu(dev
))) {
993 addr
= confirm_addr_indev(in_dev
, dst
, local
, scope
);
999 read_unlock(&dev_base_lock
);
1008 int register_inetaddr_notifier(struct notifier_block
*nb
)
1010 return blocking_notifier_chain_register(&inetaddr_chain
, nb
);
1013 int unregister_inetaddr_notifier(struct notifier_block
*nb
)
1015 return blocking_notifier_chain_unregister(&inetaddr_chain
, nb
);
1018 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
1019 * alias numbering and to create unique labels if possible.
1021 static void inetdev_changename(struct net_device
*dev
, struct in_device
*in_dev
)
1023 struct in_ifaddr
*ifa
;
1026 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
1027 char old
[IFNAMSIZ
], *dot
;
1029 memcpy(old
, ifa
->ifa_label
, IFNAMSIZ
);
1030 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
1033 dot
= strchr(ifa
->ifa_label
, ':');
1035 sprintf(old
, ":%d", named
);
1038 if (strlen(dot
) + strlen(dev
->name
) < IFNAMSIZ
) {
1039 strcat(ifa
->ifa_label
, dot
);
1041 strcpy(ifa
->ifa_label
+ (IFNAMSIZ
- strlen(dot
) - 1), dot
);
1046 /* Called only under RTNL semaphore */
1048 static int inetdev_event(struct notifier_block
*this, unsigned long event
,
1051 struct net_device
*dev
= ptr
;
1052 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
1057 if (event
== NETDEV_REGISTER
) {
1058 in_dev
= inetdev_init(dev
);
1059 if (dev
== &loopback_dev
) {
1062 "Failed to create loopback\n");
1063 IN_DEV_CONF_SET(in_dev
, NOXFRM
, 1);
1064 IN_DEV_CONF_SET(in_dev
, NOPOLICY
, 1);
1071 case NETDEV_REGISTER
:
1072 printk(KERN_DEBUG
"inetdev_event: bug\n");
1078 if (dev
== &loopback_dev
) {
1079 struct in_ifaddr
*ifa
;
1080 if ((ifa
= inet_alloc_ifa()) != NULL
) {
1082 ifa
->ifa_address
= htonl(INADDR_LOOPBACK
);
1083 ifa
->ifa_prefixlen
= 8;
1084 ifa
->ifa_mask
= inet_make_mask(8);
1085 in_dev_hold(in_dev
);
1086 ifa
->ifa_dev
= in_dev
;
1087 ifa
->ifa_scope
= RT_SCOPE_HOST
;
1088 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
1089 inet_insert_ifa(ifa
);
1097 case NETDEV_CHANGEMTU
:
1100 /* MTU falled under 68, disable IP */
1101 case NETDEV_UNREGISTER
:
1102 inetdev_destroy(in_dev
);
1104 case NETDEV_CHANGENAME
:
1105 /* Do not notify about label change, this event is
1106 * not interesting to applications using netlink.
1108 inetdev_changename(dev
, in_dev
);
1110 #ifdef CONFIG_SYSCTL
1111 devinet_sysctl_unregister(&in_dev
->cnf
);
1112 neigh_sysctl_unregister(in_dev
->arp_parms
);
1113 neigh_sysctl_register(dev
, in_dev
->arp_parms
, NET_IPV4
,
1114 NET_IPV4_NEIGH
, "ipv4", NULL
, NULL
);
1115 devinet_sysctl_register(in_dev
, &in_dev
->cnf
);
1123 static struct notifier_block ip_netdev_notifier
= {
1124 .notifier_call
=inetdev_event
,
1127 static inline size_t inet_nlmsg_size(void)
1129 return NLMSG_ALIGN(sizeof(struct ifaddrmsg
))
1130 + nla_total_size(4) /* IFA_ADDRESS */
1131 + nla_total_size(4) /* IFA_LOCAL */
1132 + nla_total_size(4) /* IFA_BROADCAST */
1133 + nla_total_size(4) /* IFA_ANYCAST */
1134 + nla_total_size(IFNAMSIZ
); /* IFA_LABEL */
1137 static int inet_fill_ifaddr(struct sk_buff
*skb
, struct in_ifaddr
*ifa
,
1138 u32 pid
, u32 seq
, int event
, unsigned int flags
)
1140 struct ifaddrmsg
*ifm
;
1141 struct nlmsghdr
*nlh
;
1143 nlh
= nlmsg_put(skb
, pid
, seq
, event
, sizeof(*ifm
), flags
);
1147 ifm
= nlmsg_data(nlh
);
1148 ifm
->ifa_family
= AF_INET
;
1149 ifm
->ifa_prefixlen
= ifa
->ifa_prefixlen
;
1150 ifm
->ifa_flags
= ifa
->ifa_flags
|IFA_F_PERMANENT
;
1151 ifm
->ifa_scope
= ifa
->ifa_scope
;
1152 ifm
->ifa_index
= ifa
->ifa_dev
->dev
->ifindex
;
1154 if (ifa
->ifa_address
)
1155 NLA_PUT_BE32(skb
, IFA_ADDRESS
, ifa
->ifa_address
);
1158 NLA_PUT_BE32(skb
, IFA_LOCAL
, ifa
->ifa_local
);
1160 if (ifa
->ifa_broadcast
)
1161 NLA_PUT_BE32(skb
, IFA_BROADCAST
, ifa
->ifa_broadcast
);
1163 if (ifa
->ifa_anycast
)
1164 NLA_PUT_BE32(skb
, IFA_ANYCAST
, ifa
->ifa_anycast
);
1166 if (ifa
->ifa_label
[0])
1167 NLA_PUT_STRING(skb
, IFA_LABEL
, ifa
->ifa_label
);
1169 return nlmsg_end(skb
, nlh
);
1172 nlmsg_cancel(skb
, nlh
);
1176 static int inet_dump_ifaddr(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1179 struct net_device
*dev
;
1180 struct in_device
*in_dev
;
1181 struct in_ifaddr
*ifa
;
1182 int s_ip_idx
, s_idx
= cb
->args
[0];
1184 s_ip_idx
= ip_idx
= cb
->args
[1];
1186 for_each_netdev(dev
) {
1191 if ((in_dev
= __in_dev_get_rtnl(dev
)) == NULL
)
1194 for (ifa
= in_dev
->ifa_list
, ip_idx
= 0; ifa
;
1195 ifa
= ifa
->ifa_next
, ip_idx
++) {
1196 if (ip_idx
< s_ip_idx
)
1198 if (inet_fill_ifaddr(skb
, ifa
, NETLINK_CB(cb
->skb
).pid
,
1200 RTM_NEWADDR
, NLM_F_MULTI
) <= 0)
1209 cb
->args
[1] = ip_idx
;
1214 static void rtmsg_ifa(int event
, struct in_ifaddr
* ifa
, struct nlmsghdr
*nlh
,
1217 struct sk_buff
*skb
;
1218 u32 seq
= nlh
? nlh
->nlmsg_seq
: 0;
1221 skb
= nlmsg_new(inet_nlmsg_size(), GFP_KERNEL
);
1225 err
= inet_fill_ifaddr(skb
, ifa
, pid
, seq
, event
, 0);
1227 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1228 WARN_ON(err
== -EMSGSIZE
);
1232 err
= rtnl_notify(skb
, pid
, RTNLGRP_IPV4_IFADDR
, nlh
, GFP_KERNEL
);
1235 rtnl_set_sk_err(RTNLGRP_IPV4_IFADDR
, err
);
1238 #ifdef CONFIG_SYSCTL
1240 static void devinet_copy_dflt_conf(int i
)
1242 struct net_device
*dev
;
1244 read_lock(&dev_base_lock
);
1245 for_each_netdev(dev
) {
1246 struct in_device
*in_dev
;
1248 in_dev
= __in_dev_get_rcu(dev
);
1249 if (in_dev
&& !test_bit(i
, in_dev
->cnf
.state
))
1250 in_dev
->cnf
.data
[i
] = ipv4_devconf_dflt
.data
[i
];
1253 read_unlock(&dev_base_lock
);
1256 static int devinet_conf_proc(ctl_table
*ctl
, int write
,
1257 struct file
* filp
, void __user
*buffer
,
1258 size_t *lenp
, loff_t
*ppos
)
1260 int ret
= proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
1263 struct ipv4_devconf
*cnf
= ctl
->extra1
;
1264 int i
= (int *)ctl
->data
- cnf
->data
;
1266 set_bit(i
, cnf
->state
);
1268 if (cnf
== &ipv4_devconf_dflt
)
1269 devinet_copy_dflt_conf(i
);
1275 static int devinet_conf_sysctl(ctl_table
*table
, int __user
*name
, int nlen
,
1276 void __user
*oldval
, size_t __user
*oldlenp
,
1277 void __user
*newval
, size_t newlen
)
1279 struct ipv4_devconf
*cnf
;
1280 int *valp
= table
->data
;
1284 if (!newval
|| !newlen
)
1287 if (newlen
!= sizeof(int))
1290 if (get_user(new, (int __user
*)newval
))
1296 if (oldval
&& oldlenp
) {
1299 if (get_user(len
, oldlenp
))
1303 if (len
> table
->maxlen
)
1304 len
= table
->maxlen
;
1305 if (copy_to_user(oldval
, valp
, len
))
1307 if (put_user(len
, oldlenp
))
1314 cnf
= table
->extra1
;
1315 i
= (int *)table
->data
- cnf
->data
;
1317 set_bit(i
, cnf
->state
);
1319 if (cnf
== &ipv4_devconf_dflt
)
1320 devinet_copy_dflt_conf(i
);
1325 void inet_forward_change(void)
1327 struct net_device
*dev
;
1328 int on
= IPV4_DEVCONF_ALL(FORWARDING
);
1330 IPV4_DEVCONF_ALL(ACCEPT_REDIRECTS
) = !on
;
1331 IPV4_DEVCONF_DFLT(FORWARDING
) = on
;
1333 read_lock(&dev_base_lock
);
1334 for_each_netdev(dev
) {
1335 struct in_device
*in_dev
;
1337 in_dev
= __in_dev_get_rcu(dev
);
1339 IN_DEV_CONF_SET(in_dev
, FORWARDING
, on
);
1342 read_unlock(&dev_base_lock
);
1347 static int devinet_sysctl_forward(ctl_table
*ctl
, int write
,
1348 struct file
* filp
, void __user
*buffer
,
1349 size_t *lenp
, loff_t
*ppos
)
1351 int *valp
= ctl
->data
;
1353 int ret
= proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
1355 if (write
&& *valp
!= val
) {
1356 if (valp
== &IPV4_DEVCONF_ALL(FORWARDING
))
1357 inet_forward_change();
1358 else if (valp
!= &IPV4_DEVCONF_DFLT(FORWARDING
))
1365 int ipv4_doint_and_flush(ctl_table
*ctl
, int write
,
1366 struct file
* filp
, void __user
*buffer
,
1367 size_t *lenp
, loff_t
*ppos
)
1369 int *valp
= ctl
->data
;
1371 int ret
= proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
1373 if (write
&& *valp
!= val
)
1379 int ipv4_doint_and_flush_strategy(ctl_table
*table
, int __user
*name
, int nlen
,
1380 void __user
*oldval
, size_t __user
*oldlenp
,
1381 void __user
*newval
, size_t newlen
)
1383 int ret
= devinet_conf_sysctl(table
, name
, nlen
, oldval
, oldlenp
,
1393 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc, sysctl) \
1395 .ctl_name = NET_IPV4_CONF_ ## attr, \
1397 .data = ipv4_devconf.data + \
1398 NET_IPV4_CONF_ ## attr - 1, \
1399 .maxlen = sizeof(int), \
1401 .proc_handler = proc, \
1402 .strategy = sysctl, \
1403 .extra1 = &ipv4_devconf, \
1406 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
1407 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc, \
1408 devinet_conf_sysctl)
1410 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
1411 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc, \
1412 devinet_conf_sysctl)
1414 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc, sysctl) \
1415 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc, sysctl)
1417 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
1418 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush, \
1419 ipv4_doint_and_flush_strategy)
1421 static struct devinet_sysctl_table
{
1422 struct ctl_table_header
*sysctl_header
;
1423 ctl_table devinet_vars
[__NET_IPV4_CONF_MAX
];
1424 ctl_table devinet_dev
[2];
1425 ctl_table devinet_conf_dir
[2];
1426 ctl_table devinet_proto_dir
[2];
1427 ctl_table devinet_root_dir
[2];
1428 } devinet_sysctl
= {
1430 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING
, "forwarding",
1431 devinet_sysctl_forward
,
1432 devinet_conf_sysctl
),
1433 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING
, "mc_forwarding"),
1435 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS
, "accept_redirects"),
1436 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS
, "secure_redirects"),
1437 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA
, "shared_media"),
1438 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER
, "rp_filter"),
1439 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS
, "send_redirects"),
1440 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE
,
1441 "accept_source_route"),
1442 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP
, "proxy_arp"),
1443 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID
, "medium_id"),
1444 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY
, "bootp_relay"),
1445 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS
, "log_martians"),
1446 DEVINET_SYSCTL_RW_ENTRY(TAG
, "tag"),
1447 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER
, "arp_filter"),
1448 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE
, "arp_announce"),
1449 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE
, "arp_ignore"),
1450 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT
, "arp_accept"),
1452 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM
, "disable_xfrm"),
1453 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY
, "disable_policy"),
1454 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION
,
1455 "force_igmp_version"),
1456 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES
,
1457 "promote_secondaries"),
1461 .ctl_name
= NET_PROTO_CONF_ALL
,
1464 .child
= devinet_sysctl
.devinet_vars
,
1467 .devinet_conf_dir
= {
1469 .ctl_name
= NET_IPV4_CONF
,
1472 .child
= devinet_sysctl
.devinet_dev
,
1475 .devinet_proto_dir
= {
1477 .ctl_name
= NET_IPV4
,
1480 .child
= devinet_sysctl
.devinet_conf_dir
,
1483 .devinet_root_dir
= {
1485 .ctl_name
= CTL_NET
,
1488 .child
= devinet_sysctl
.devinet_proto_dir
,
1493 static void devinet_sysctl_register(struct in_device
*in_dev
,
1494 struct ipv4_devconf
*p
)
1497 struct net_device
*dev
= in_dev
? in_dev
->dev
: NULL
;
1498 struct devinet_sysctl_table
*t
= kmemdup(&devinet_sysctl
, sizeof(*t
),
1500 char *dev_name
= NULL
;
1504 for (i
= 0; i
< ARRAY_SIZE(t
->devinet_vars
) - 1; i
++) {
1505 t
->devinet_vars
[i
].data
+= (char *)p
- (char *)&ipv4_devconf
;
1506 t
->devinet_vars
[i
].extra1
= p
;
1510 dev_name
= dev
->name
;
1511 t
->devinet_dev
[0].ctl_name
= dev
->ifindex
;
1513 dev_name
= "default";
1514 t
->devinet_dev
[0].ctl_name
= NET_PROTO_CONF_DEFAULT
;
1518 * Make a copy of dev_name, because '.procname' is regarded as const
1519 * by sysctl and we wouldn't want anyone to change it under our feet
1520 * (see SIOCSIFNAME).
1522 dev_name
= kstrdup(dev_name
, GFP_KERNEL
);
1526 t
->devinet_dev
[0].procname
= dev_name
;
1527 t
->devinet_dev
[0].child
= t
->devinet_vars
;
1528 t
->devinet_conf_dir
[0].child
= t
->devinet_dev
;
1529 t
->devinet_proto_dir
[0].child
= t
->devinet_conf_dir
;
1530 t
->devinet_root_dir
[0].child
= t
->devinet_proto_dir
;
1532 t
->sysctl_header
= register_sysctl_table(t
->devinet_root_dir
);
1533 if (!t
->sysctl_header
)
1547 static void devinet_sysctl_unregister(struct ipv4_devconf
*p
)
1550 struct devinet_sysctl_table
*t
= p
->sysctl
;
1552 unregister_sysctl_table(t
->sysctl_header
);
1553 kfree(t
->devinet_dev
[0].procname
);
1559 void __init
devinet_init(void)
1561 register_gifconf(PF_INET
, inet_gifconf
);
1562 register_netdevice_notifier(&ip_netdev_notifier
);
1564 rtnl_register(PF_INET
, RTM_NEWADDR
, inet_rtm_newaddr
, NULL
);
1565 rtnl_register(PF_INET
, RTM_DELADDR
, inet_rtm_deladdr
, NULL
);
1566 rtnl_register(PF_INET
, RTM_GETADDR
, NULL
, inet_dump_ifaddr
);
1567 #ifdef CONFIG_SYSCTL
1568 devinet_sysctl
.sysctl_header
=
1569 register_sysctl_table(devinet_sysctl
.devinet_root_dir
);
1570 devinet_sysctl_register(NULL
, &ipv4_devconf_dflt
);
1574 EXPORT_SYMBOL(in_dev_finish_destroy
);
1575 EXPORT_SYMBOL(inet_select_addr
);
1576 EXPORT_SYMBOL(inetdev_by_index
);
1577 EXPORT_SYMBOL(register_inetaddr_notifier
);
1578 EXPORT_SYMBOL(unregister_inetaddr_notifier
);