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
30 #include <linux/config.h>
32 #include <asm/uaccess.h>
33 #include <asm/system.h>
34 #include <linux/bitops.h>
35 #include <linux/capability.h>
36 #include <linux/module.h>
37 #include <linux/types.h>
38 #include <linux/kernel.h>
39 #include <linux/sched.h>
40 #include <linux/string.h>
42 #include <linux/socket.h>
43 #include <linux/sockios.h>
45 #include <linux/errno.h>
46 #include <linux/interrupt.h>
47 #include <linux/if_ether.h>
48 #include <linux/inet.h>
49 #include <linux/netdevice.h>
50 #include <linux/etherdevice.h>
51 #include <linux/skbuff.h>
52 #include <linux/rtnetlink.h>
53 #include <linux/init.h>
54 #include <linux/notifier.h>
55 #include <linux/inetdevice.h>
56 #include <linux/igmp.h>
58 #include <linux/sysctl.h>
60 #include <linux/kmod.h>
64 #include <net/route.h>
65 #include <net/ip_fib.h>
67 struct ipv4_devconf ipv4_devconf
= {
68 .accept_redirects
= 1,
70 .secure_redirects
= 1,
74 static struct ipv4_devconf ipv4_devconf_dflt
= {
75 .accept_redirects
= 1,
77 .secure_redirects
= 1,
79 .accept_source_route
= 1,
82 static void rtmsg_ifa(int event
, struct in_ifaddr
*);
84 static struct notifier_block
*inetaddr_chain
;
85 static void inet_del_ifa(struct in_device
*in_dev
, struct in_ifaddr
**ifap
,
88 static void devinet_sysctl_register(struct in_device
*in_dev
,
89 struct ipv4_devconf
*p
);
90 static void devinet_sysctl_unregister(struct ipv4_devconf
*p
);
93 /* Locks all the inet devices. */
95 static struct in_ifaddr
*inet_alloc_ifa(void)
97 struct in_ifaddr
*ifa
= kmalloc(sizeof(*ifa
), GFP_KERNEL
);
100 memset(ifa
, 0, sizeof(*ifa
));
101 INIT_RCU_HEAD(&ifa
->rcu_head
);
107 static void inet_rcu_free_ifa(struct rcu_head
*head
)
109 struct in_ifaddr
*ifa
= container_of(head
, struct in_ifaddr
, rcu_head
);
111 in_dev_put(ifa
->ifa_dev
);
115 static inline void inet_free_ifa(struct in_ifaddr
*ifa
)
117 call_rcu(&ifa
->rcu_head
, inet_rcu_free_ifa
);
120 void in_dev_finish_destroy(struct in_device
*idev
)
122 struct net_device
*dev
= idev
->dev
;
124 BUG_TRAP(!idev
->ifa_list
);
125 BUG_TRAP(!idev
->mc_list
);
126 #ifdef NET_REFCNT_DEBUG
127 printk(KERN_DEBUG
"in_dev_finish_destroy: %p=%s\n",
128 idev
, dev
? dev
->name
: "NIL");
132 printk("Freeing alive in_device %p\n", idev
);
138 struct in_device
*inetdev_init(struct net_device
*dev
)
140 struct in_device
*in_dev
;
144 in_dev
= kmalloc(sizeof(*in_dev
), GFP_KERNEL
);
147 memset(in_dev
, 0, sizeof(*in_dev
));
148 INIT_RCU_HEAD(&in_dev
->rcu_head
);
149 memcpy(&in_dev
->cnf
, &ipv4_devconf_dflt
, sizeof(in_dev
->cnf
));
150 in_dev
->cnf
.sysctl
= NULL
;
152 if ((in_dev
->arp_parms
= neigh_parms_alloc(dev
, &arp_tbl
)) == NULL
)
154 /* Reference in_dev->dev */
157 neigh_sysctl_register(dev
, in_dev
->arp_parms
, NET_IPV4
,
158 NET_IPV4_NEIGH
, "ipv4", NULL
, NULL
);
161 /* Account for reference dev->ip_ptr (below) */
165 devinet_sysctl_register(in_dev
, &in_dev
->cnf
);
167 ip_mc_init_dev(in_dev
);
168 if (dev
->flags
& IFF_UP
)
171 /* we can receive as soon as ip_ptr is set -- do this last */
172 rcu_assign_pointer(dev
->ip_ptr
, in_dev
);
180 static void in_dev_rcu_put(struct rcu_head
*head
)
182 struct in_device
*idev
= container_of(head
, struct in_device
, rcu_head
);
186 static void inetdev_destroy(struct in_device
*in_dev
)
188 struct in_ifaddr
*ifa
;
189 struct net_device
*dev
;
194 if (dev
== &loopback_dev
)
199 ip_mc_destroy_dev(in_dev
);
201 while ((ifa
= in_dev
->ifa_list
) != NULL
) {
202 inet_del_ifa(in_dev
, &in_dev
->ifa_list
, 0);
207 devinet_sysctl_unregister(&in_dev
->cnf
);
213 neigh_sysctl_unregister(in_dev
->arp_parms
);
215 neigh_parms_release(&arp_tbl
, in_dev
->arp_parms
);
218 call_rcu(&in_dev
->rcu_head
, in_dev_rcu_put
);
221 int inet_addr_onlink(struct in_device
*in_dev
, u32 a
, u32 b
)
224 for_primary_ifa(in_dev
) {
225 if (inet_ifa_match(a
, ifa
)) {
226 if (!b
|| inet_ifa_match(b
, ifa
)) {
231 } endfor_ifa(in_dev
);
236 static void inet_del_ifa(struct in_device
*in_dev
, struct in_ifaddr
**ifap
,
239 struct in_ifaddr
*promote
= NULL
;
240 struct in_ifaddr
*ifa
, *ifa1
= *ifap
;
241 struct in_ifaddr
*last_prim
= in_dev
->ifa_list
;
242 struct in_ifaddr
*prev_prom
= NULL
;
243 int do_promote
= IN_DEV_PROMOTE_SECONDARIES(in_dev
);
247 /* 1. Deleting primary ifaddr forces deletion all secondaries
248 * unless alias promotion is set
251 if (!(ifa1
->ifa_flags
& IFA_F_SECONDARY
)) {
252 struct in_ifaddr
**ifap1
= &ifa1
->ifa_next
;
254 while ((ifa
= *ifap1
) != NULL
) {
255 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
) &&
256 ifa1
->ifa_scope
<= ifa
->ifa_scope
)
259 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
) ||
260 ifa1
->ifa_mask
!= ifa
->ifa_mask
||
261 !inet_ifa_match(ifa1
->ifa_address
, ifa
)) {
262 ifap1
= &ifa
->ifa_next
;
268 *ifap1
= ifa
->ifa_next
;
270 rtmsg_ifa(RTM_DELADDR
, ifa
);
271 notifier_call_chain(&inetaddr_chain
, NETDEV_DOWN
, ifa
);
282 *ifap
= ifa1
->ifa_next
;
284 /* 3. Announce address deletion */
286 /* Send message first, then call notifier.
287 At first sight, FIB update triggered by notifier
288 will refer to already deleted ifaddr, that could confuse
289 netlink listeners. It is not true: look, gated sees
290 that route deleted and if it still thinks that ifaddr
291 is valid, it will try to restore deleted routes... Grr.
292 So that, this order is correct.
294 rtmsg_ifa(RTM_DELADDR
, ifa1
);
295 notifier_call_chain(&inetaddr_chain
, NETDEV_DOWN
, ifa1
);
300 prev_prom
->ifa_next
= promote
->ifa_next
;
301 promote
->ifa_next
= last_prim
->ifa_next
;
302 last_prim
->ifa_next
= promote
;
305 promote
->ifa_flags
&= ~IFA_F_SECONDARY
;
306 rtmsg_ifa(RTM_NEWADDR
, promote
);
307 notifier_call_chain(&inetaddr_chain
, NETDEV_UP
, promote
);
308 for (ifa
= promote
->ifa_next
; ifa
; ifa
= ifa
->ifa_next
) {
309 if (ifa1
->ifa_mask
!= ifa
->ifa_mask
||
310 !inet_ifa_match(ifa1
->ifa_address
, ifa
))
319 if (!in_dev
->ifa_list
)
320 inetdev_destroy(in_dev
);
324 static int inet_insert_ifa(struct in_ifaddr
*ifa
)
326 struct in_device
*in_dev
= ifa
->ifa_dev
;
327 struct in_ifaddr
*ifa1
, **ifap
, **last_primary
;
331 if (!ifa
->ifa_local
) {
336 ifa
->ifa_flags
&= ~IFA_F_SECONDARY
;
337 last_primary
= &in_dev
->ifa_list
;
339 for (ifap
= &in_dev
->ifa_list
; (ifa1
= *ifap
) != NULL
;
340 ifap
= &ifa1
->ifa_next
) {
341 if (!(ifa1
->ifa_flags
& IFA_F_SECONDARY
) &&
342 ifa
->ifa_scope
<= ifa1
->ifa_scope
)
343 last_primary
= &ifa1
->ifa_next
;
344 if (ifa1
->ifa_mask
== ifa
->ifa_mask
&&
345 inet_ifa_match(ifa1
->ifa_address
, ifa
)) {
346 if (ifa1
->ifa_local
== ifa
->ifa_local
) {
350 if (ifa1
->ifa_scope
!= ifa
->ifa_scope
) {
354 ifa
->ifa_flags
|= IFA_F_SECONDARY
;
358 if (!(ifa
->ifa_flags
& IFA_F_SECONDARY
)) {
359 net_srandom(ifa
->ifa_local
);
363 ifa
->ifa_next
= *ifap
;
366 /* Send message first, then call notifier.
367 Notifier will trigger FIB update, so that
368 listeners of netlink will know about new ifaddr */
369 rtmsg_ifa(RTM_NEWADDR
, ifa
);
370 notifier_call_chain(&inetaddr_chain
, NETDEV_UP
, ifa
);
375 static int inet_set_ifa(struct net_device
*dev
, struct in_ifaddr
*ifa
)
377 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
382 in_dev
= inetdev_init(dev
);
388 if (ifa
->ifa_dev
!= in_dev
) {
389 BUG_TRAP(!ifa
->ifa_dev
);
391 ifa
->ifa_dev
= in_dev
;
393 if (LOOPBACK(ifa
->ifa_local
))
394 ifa
->ifa_scope
= RT_SCOPE_HOST
;
395 return inet_insert_ifa(ifa
);
398 struct in_device
*inetdev_by_index(int ifindex
)
400 struct net_device
*dev
;
401 struct in_device
*in_dev
= NULL
;
402 read_lock(&dev_base_lock
);
403 dev
= __dev_get_by_index(ifindex
);
405 in_dev
= in_dev_get(dev
);
406 read_unlock(&dev_base_lock
);
410 /* Called only from RTNL semaphored context. No locks. */
412 struct in_ifaddr
*inet_ifa_byprefix(struct in_device
*in_dev
, u32 prefix
,
417 for_primary_ifa(in_dev
) {
418 if (ifa
->ifa_mask
== mask
&& inet_ifa_match(prefix
, ifa
))
420 } endfor_ifa(in_dev
);
424 static int inet_rtm_deladdr(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
426 struct rtattr
**rta
= arg
;
427 struct in_device
*in_dev
;
428 struct ifaddrmsg
*ifm
= NLMSG_DATA(nlh
);
429 struct in_ifaddr
*ifa
, **ifap
;
433 if ((in_dev
= inetdev_by_index(ifm
->ifa_index
)) == NULL
)
435 __in_dev_put(in_dev
);
437 for (ifap
= &in_dev
->ifa_list
; (ifa
= *ifap
) != NULL
;
438 ifap
= &ifa
->ifa_next
) {
439 if ((rta
[IFA_LOCAL
- 1] &&
440 memcmp(RTA_DATA(rta
[IFA_LOCAL
- 1]),
441 &ifa
->ifa_local
, 4)) ||
442 (rta
[IFA_LABEL
- 1] &&
443 rtattr_strcmp(rta
[IFA_LABEL
- 1], ifa
->ifa_label
)) ||
444 (rta
[IFA_ADDRESS
- 1] &&
445 (ifm
->ifa_prefixlen
!= ifa
->ifa_prefixlen
||
446 !inet_ifa_match(*(u32
*)RTA_DATA(rta
[IFA_ADDRESS
- 1]),
449 inet_del_ifa(in_dev
, ifap
, 1);
453 return -EADDRNOTAVAIL
;
456 static int inet_rtm_newaddr(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
458 struct rtattr
**rta
= arg
;
459 struct net_device
*dev
;
460 struct in_device
*in_dev
;
461 struct ifaddrmsg
*ifm
= NLMSG_DATA(nlh
);
462 struct in_ifaddr
*ifa
;
467 if (ifm
->ifa_prefixlen
> 32 || !rta
[IFA_LOCAL
- 1])
471 if ((dev
= __dev_get_by_index(ifm
->ifa_index
)) == NULL
)
475 if ((in_dev
= __in_dev_get_rtnl(dev
)) == NULL
) {
476 in_dev
= inetdev_init(dev
);
481 if ((ifa
= inet_alloc_ifa()) == NULL
)
484 if (!rta
[IFA_ADDRESS
- 1])
485 rta
[IFA_ADDRESS
- 1] = rta
[IFA_LOCAL
- 1];
486 memcpy(&ifa
->ifa_local
, RTA_DATA(rta
[IFA_LOCAL
- 1]), 4);
487 memcpy(&ifa
->ifa_address
, RTA_DATA(rta
[IFA_ADDRESS
- 1]), 4);
488 ifa
->ifa_prefixlen
= ifm
->ifa_prefixlen
;
489 ifa
->ifa_mask
= inet_make_mask(ifm
->ifa_prefixlen
);
490 if (rta
[IFA_BROADCAST
- 1])
491 memcpy(&ifa
->ifa_broadcast
,
492 RTA_DATA(rta
[IFA_BROADCAST
- 1]), 4);
493 if (rta
[IFA_ANYCAST
- 1])
494 memcpy(&ifa
->ifa_anycast
, RTA_DATA(rta
[IFA_ANYCAST
- 1]), 4);
495 ifa
->ifa_flags
= ifm
->ifa_flags
;
496 ifa
->ifa_scope
= ifm
->ifa_scope
;
498 ifa
->ifa_dev
= in_dev
;
499 if (rta
[IFA_LABEL
- 1])
500 rtattr_strlcpy(ifa
->ifa_label
, rta
[IFA_LABEL
- 1], IFNAMSIZ
);
502 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
504 rc
= inet_insert_ifa(ifa
);
510 * Determine a default network mask, based on the IP address.
513 static __inline__
int inet_abc_len(u32 addr
)
515 int rc
= -1; /* Something else, probably a multicast. */
524 else if (IN_CLASSB(addr
))
526 else if (IN_CLASSC(addr
))
534 int devinet_ioctl(unsigned int cmd
, void __user
*arg
)
537 struct sockaddr_in sin_orig
;
538 struct sockaddr_in
*sin
= (struct sockaddr_in
*)&ifr
.ifr_addr
;
539 struct in_device
*in_dev
;
540 struct in_ifaddr
**ifap
= NULL
;
541 struct in_ifaddr
*ifa
= NULL
;
542 struct net_device
*dev
;
545 int tryaddrmatch
= 0;
548 * Fetch the caller's info block into kernel space
551 if (copy_from_user(&ifr
, arg
, sizeof(struct ifreq
)))
553 ifr
.ifr_name
[IFNAMSIZ
- 1] = 0;
555 /* save original address for comparison */
556 memcpy(&sin_orig
, sin
, sizeof(*sin
));
558 colon
= strchr(ifr
.ifr_name
, ':');
563 dev_load(ifr
.ifr_name
);
567 case SIOCGIFADDR
: /* Get interface address */
568 case SIOCGIFBRDADDR
: /* Get the broadcast address */
569 case SIOCGIFDSTADDR
: /* Get the destination address */
570 case SIOCGIFNETMASK
: /* Get the netmask for the interface */
571 /* Note that these ioctls will not sleep,
572 so that we do not impose a lock.
573 One day we will be forced to put shlock here (I mean SMP)
575 tryaddrmatch
= (sin_orig
.sin_family
== AF_INET
);
576 memset(sin
, 0, sizeof(*sin
));
577 sin
->sin_family
= AF_INET
;
582 if (!capable(CAP_NET_ADMIN
))
585 case SIOCSIFADDR
: /* Set interface address (and family) */
586 case SIOCSIFBRDADDR
: /* Set the broadcast address */
587 case SIOCSIFDSTADDR
: /* Set the destination address */
588 case SIOCSIFNETMASK
: /* Set the netmask for the interface */
590 if (!capable(CAP_NET_ADMIN
))
593 if (sin
->sin_family
!= AF_INET
)
604 if ((dev
= __dev_get_by_name(ifr
.ifr_name
)) == NULL
)
610 if ((in_dev
= __in_dev_get_rtnl(dev
)) != NULL
) {
612 /* Matthias Andree */
613 /* compare label and address (4.4BSD style) */
614 /* note: we only do this for a limited set of ioctls
615 and only if the original address family was AF_INET.
616 This is checked above. */
617 for (ifap
= &in_dev
->ifa_list
; (ifa
= *ifap
) != NULL
;
618 ifap
= &ifa
->ifa_next
) {
619 if (!strcmp(ifr
.ifr_name
, ifa
->ifa_label
) &&
620 sin_orig
.sin_addr
.s_addr
==
626 /* we didn't get a match, maybe the application is
627 4.3BSD-style and passed in junk so we fall back to
628 comparing just the label */
630 for (ifap
= &in_dev
->ifa_list
; (ifa
= *ifap
) != NULL
;
631 ifap
= &ifa
->ifa_next
)
632 if (!strcmp(ifr
.ifr_name
, ifa
->ifa_label
))
637 ret
= -EADDRNOTAVAIL
;
638 if (!ifa
&& cmd
!= SIOCSIFADDR
&& cmd
!= SIOCSIFFLAGS
)
642 case SIOCGIFADDR
: /* Get interface address */
643 sin
->sin_addr
.s_addr
= ifa
->ifa_local
;
646 case SIOCGIFBRDADDR
: /* Get the broadcast address */
647 sin
->sin_addr
.s_addr
= ifa
->ifa_broadcast
;
650 case SIOCGIFDSTADDR
: /* Get the destination address */
651 sin
->sin_addr
.s_addr
= ifa
->ifa_address
;
654 case SIOCGIFNETMASK
: /* Get the netmask for the interface */
655 sin
->sin_addr
.s_addr
= ifa
->ifa_mask
;
660 ret
= -EADDRNOTAVAIL
;
664 if (!(ifr
.ifr_flags
& IFF_UP
))
665 inet_del_ifa(in_dev
, ifap
, 1);
668 ret
= dev_change_flags(dev
, ifr
.ifr_flags
);
671 case SIOCSIFADDR
: /* Set interface address (and family) */
673 if (inet_abc_len(sin
->sin_addr
.s_addr
) < 0)
678 if ((ifa
= inet_alloc_ifa()) == NULL
)
681 memcpy(ifa
->ifa_label
, ifr
.ifr_name
, IFNAMSIZ
);
683 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
686 if (ifa
->ifa_local
== sin
->sin_addr
.s_addr
)
688 inet_del_ifa(in_dev
, ifap
, 0);
689 ifa
->ifa_broadcast
= 0;
690 ifa
->ifa_anycast
= 0;
693 ifa
->ifa_address
= ifa
->ifa_local
= sin
->sin_addr
.s_addr
;
695 if (!(dev
->flags
& IFF_POINTOPOINT
)) {
696 ifa
->ifa_prefixlen
= inet_abc_len(ifa
->ifa_address
);
697 ifa
->ifa_mask
= inet_make_mask(ifa
->ifa_prefixlen
);
698 if ((dev
->flags
& IFF_BROADCAST
) &&
699 ifa
->ifa_prefixlen
< 31)
700 ifa
->ifa_broadcast
= ifa
->ifa_address
|
703 ifa
->ifa_prefixlen
= 32;
704 ifa
->ifa_mask
= inet_make_mask(32);
706 ret
= inet_set_ifa(dev
, ifa
);
709 case SIOCSIFBRDADDR
: /* Set the broadcast address */
711 if (ifa
->ifa_broadcast
!= sin
->sin_addr
.s_addr
) {
712 inet_del_ifa(in_dev
, ifap
, 0);
713 ifa
->ifa_broadcast
= sin
->sin_addr
.s_addr
;
714 inet_insert_ifa(ifa
);
718 case SIOCSIFDSTADDR
: /* Set the destination address */
720 if (ifa
->ifa_address
== sin
->sin_addr
.s_addr
)
723 if (inet_abc_len(sin
->sin_addr
.s_addr
) < 0)
726 inet_del_ifa(in_dev
, ifap
, 0);
727 ifa
->ifa_address
= sin
->sin_addr
.s_addr
;
728 inet_insert_ifa(ifa
);
731 case SIOCSIFNETMASK
: /* Set the netmask for the interface */
734 * The mask we set must be legal.
737 if (bad_mask(sin
->sin_addr
.s_addr
, 0))
740 if (ifa
->ifa_mask
!= sin
->sin_addr
.s_addr
) {
741 u32 old_mask
= ifa
->ifa_mask
;
742 inet_del_ifa(in_dev
, ifap
, 0);
743 ifa
->ifa_mask
= sin
->sin_addr
.s_addr
;
744 ifa
->ifa_prefixlen
= inet_mask_len(ifa
->ifa_mask
);
746 /* See if current broadcast address matches
747 * with current netmask, then recalculate
748 * the broadcast address. Otherwise it's a
749 * funny address, so don't touch it since
750 * the user seems to know what (s)he's doing...
752 if ((dev
->flags
& IFF_BROADCAST
) &&
753 (ifa
->ifa_prefixlen
< 31) &&
754 (ifa
->ifa_broadcast
==
755 (ifa
->ifa_local
|~old_mask
))) {
756 ifa
->ifa_broadcast
= (ifa
->ifa_local
|
757 ~sin
->sin_addr
.s_addr
);
759 inet_insert_ifa(ifa
);
769 ret
= copy_to_user(arg
, &ifr
, sizeof(struct ifreq
)) ? -EFAULT
: 0;
773 static int inet_gifconf(struct net_device
*dev
, char __user
*buf
, int len
)
775 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
776 struct in_ifaddr
*ifa
;
780 if (!in_dev
|| (ifa
= in_dev
->ifa_list
) == NULL
)
783 for (; ifa
; ifa
= ifa
->ifa_next
) {
788 if (len
< (int) sizeof(ifr
))
790 memset(&ifr
, 0, sizeof(struct ifreq
));
792 strcpy(ifr
.ifr_name
, ifa
->ifa_label
);
794 strcpy(ifr
.ifr_name
, dev
->name
);
796 (*(struct sockaddr_in
*)&ifr
.ifr_addr
).sin_family
= AF_INET
;
797 (*(struct sockaddr_in
*)&ifr
.ifr_addr
).sin_addr
.s_addr
=
800 if (copy_to_user(buf
, &ifr
, sizeof(struct ifreq
))) {
804 buf
+= sizeof(struct ifreq
);
805 len
-= sizeof(struct ifreq
);
806 done
+= sizeof(struct ifreq
);
812 u32
inet_select_addr(const struct net_device
*dev
, u32 dst
, int scope
)
815 struct in_device
*in_dev
;
818 in_dev
= __in_dev_get_rcu(dev
);
822 for_primary_ifa(in_dev
) {
823 if (ifa
->ifa_scope
> scope
)
825 if (!dst
|| inet_ifa_match(dst
, ifa
)) {
826 addr
= ifa
->ifa_local
;
830 addr
= ifa
->ifa_local
;
831 } endfor_ifa(in_dev
);
838 /* Not loopback addresses on loopback should be preferred
839 in this case. It is importnat that lo is the first interface
842 read_lock(&dev_base_lock
);
844 for (dev
= dev_base
; dev
; dev
= dev
->next
) {
845 if ((in_dev
= __in_dev_get_rcu(dev
)) == NULL
)
848 for_primary_ifa(in_dev
) {
849 if (ifa
->ifa_scope
!= RT_SCOPE_LINK
&&
850 ifa
->ifa_scope
<= scope
) {
851 addr
= ifa
->ifa_local
;
852 goto out_unlock_both
;
854 } endfor_ifa(in_dev
);
857 read_unlock(&dev_base_lock
);
863 static u32
confirm_addr_indev(struct in_device
*in_dev
, u32 dst
,
864 u32 local
, int scope
)
871 (local
== ifa
->ifa_local
|| !local
) &&
872 ifa
->ifa_scope
<= scope
) {
873 addr
= ifa
->ifa_local
;
878 same
= (!local
|| inet_ifa_match(local
, ifa
)) &&
879 (!dst
|| inet_ifa_match(dst
, ifa
));
883 /* Is the selected addr into dst subnet? */
884 if (inet_ifa_match(addr
, ifa
))
886 /* No, then can we use new local src? */
887 if (ifa
->ifa_scope
<= scope
) {
888 addr
= ifa
->ifa_local
;
891 /* search for large dst subnet for addr */
895 } endfor_ifa(in_dev
);
897 return same
? addr
: 0;
901 * Confirm that local IP address exists using wildcards:
902 * - dev: only on this interface, 0=any interface
903 * - dst: only in the same subnet as dst, 0=any dst
904 * - local: address, 0=autoselect the local address
905 * - scope: maximum allowed scope value for the local address
907 u32
inet_confirm_addr(const struct net_device
*dev
, u32 dst
, u32 local
, int scope
)
910 struct in_device
*in_dev
;
914 if ((in_dev
= __in_dev_get_rcu(dev
)))
915 addr
= confirm_addr_indev(in_dev
, dst
, local
, scope
);
921 read_lock(&dev_base_lock
);
923 for (dev
= dev_base
; dev
; dev
= dev
->next
) {
924 if ((in_dev
= __in_dev_get_rcu(dev
))) {
925 addr
= confirm_addr_indev(in_dev
, dst
, local
, scope
);
931 read_unlock(&dev_base_lock
);
940 int register_inetaddr_notifier(struct notifier_block
*nb
)
942 return notifier_chain_register(&inetaddr_chain
, nb
);
945 int unregister_inetaddr_notifier(struct notifier_block
*nb
)
947 return notifier_chain_unregister(&inetaddr_chain
, nb
);
950 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
951 * alias numbering and to create unique labels if possible.
953 static void inetdev_changename(struct net_device
*dev
, struct in_device
*in_dev
)
955 struct in_ifaddr
*ifa
;
958 for (ifa
= in_dev
->ifa_list
; ifa
; ifa
= ifa
->ifa_next
) {
959 char old
[IFNAMSIZ
], *dot
;
961 memcpy(old
, ifa
->ifa_label
, IFNAMSIZ
);
962 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
965 dot
= strchr(ifa
->ifa_label
, ':');
967 sprintf(old
, ":%d", named
);
970 if (strlen(dot
) + strlen(dev
->name
) < IFNAMSIZ
) {
971 strcat(ifa
->ifa_label
, dot
);
973 strcpy(ifa
->ifa_label
+ (IFNAMSIZ
- strlen(dot
) - 1), dot
);
978 /* Called only under RTNL semaphore */
980 static int inetdev_event(struct notifier_block
*this, unsigned long event
,
983 struct net_device
*dev
= ptr
;
984 struct in_device
*in_dev
= __in_dev_get_rtnl(dev
);
989 if (event
== NETDEV_REGISTER
&& dev
== &loopback_dev
) {
990 in_dev
= inetdev_init(dev
);
992 panic("devinet: Failed to create loopback\n");
993 in_dev
->cnf
.no_xfrm
= 1;
994 in_dev
->cnf
.no_policy
= 1;
1000 case NETDEV_REGISTER
:
1001 printk(KERN_DEBUG
"inetdev_event: bug\n");
1007 if (dev
== &loopback_dev
) {
1008 struct in_ifaddr
*ifa
;
1009 if ((ifa
= inet_alloc_ifa()) != NULL
) {
1011 ifa
->ifa_address
= htonl(INADDR_LOOPBACK
);
1012 ifa
->ifa_prefixlen
= 8;
1013 ifa
->ifa_mask
= inet_make_mask(8);
1014 in_dev_hold(in_dev
);
1015 ifa
->ifa_dev
= in_dev
;
1016 ifa
->ifa_scope
= RT_SCOPE_HOST
;
1017 memcpy(ifa
->ifa_label
, dev
->name
, IFNAMSIZ
);
1018 inet_insert_ifa(ifa
);
1026 case NETDEV_CHANGEMTU
:
1029 /* MTU falled under 68, disable IP */
1030 case NETDEV_UNREGISTER
:
1031 inetdev_destroy(in_dev
);
1033 case NETDEV_CHANGENAME
:
1034 /* Do not notify about label change, this event is
1035 * not interesting to applications using netlink.
1037 inetdev_changename(dev
, in_dev
);
1039 #ifdef CONFIG_SYSCTL
1040 devinet_sysctl_unregister(&in_dev
->cnf
);
1041 neigh_sysctl_unregister(in_dev
->arp_parms
);
1042 neigh_sysctl_register(dev
, in_dev
->arp_parms
, NET_IPV4
,
1043 NET_IPV4_NEIGH
, "ipv4", NULL
, NULL
);
1044 devinet_sysctl_register(in_dev
, &in_dev
->cnf
);
1052 static struct notifier_block ip_netdev_notifier
= {
1053 .notifier_call
=inetdev_event
,
1056 static int inet_fill_ifaddr(struct sk_buff
*skb
, struct in_ifaddr
*ifa
,
1057 u32 pid
, u32 seq
, int event
, unsigned int flags
)
1059 struct ifaddrmsg
*ifm
;
1060 struct nlmsghdr
*nlh
;
1061 unsigned char *b
= skb
->tail
;
1063 nlh
= NLMSG_NEW(skb
, pid
, seq
, event
, sizeof(*ifm
), flags
);
1064 ifm
= NLMSG_DATA(nlh
);
1065 ifm
->ifa_family
= AF_INET
;
1066 ifm
->ifa_prefixlen
= ifa
->ifa_prefixlen
;
1067 ifm
->ifa_flags
= ifa
->ifa_flags
|IFA_F_PERMANENT
;
1068 ifm
->ifa_scope
= ifa
->ifa_scope
;
1069 ifm
->ifa_index
= ifa
->ifa_dev
->dev
->ifindex
;
1070 if (ifa
->ifa_address
)
1071 RTA_PUT(skb
, IFA_ADDRESS
, 4, &ifa
->ifa_address
);
1073 RTA_PUT(skb
, IFA_LOCAL
, 4, &ifa
->ifa_local
);
1074 if (ifa
->ifa_broadcast
)
1075 RTA_PUT(skb
, IFA_BROADCAST
, 4, &ifa
->ifa_broadcast
);
1076 if (ifa
->ifa_anycast
)
1077 RTA_PUT(skb
, IFA_ANYCAST
, 4, &ifa
->ifa_anycast
);
1078 if (ifa
->ifa_label
[0])
1079 RTA_PUT(skb
, IFA_LABEL
, IFNAMSIZ
, &ifa
->ifa_label
);
1080 nlh
->nlmsg_len
= skb
->tail
- b
;
1085 skb_trim(skb
, b
- skb
->data
);
1089 static int inet_dump_ifaddr(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1092 struct net_device
*dev
;
1093 struct in_device
*in_dev
;
1094 struct in_ifaddr
*ifa
;
1095 int s_ip_idx
, s_idx
= cb
->args
[0];
1097 s_ip_idx
= ip_idx
= cb
->args
[1];
1098 read_lock(&dev_base_lock
);
1099 for (dev
= dev_base
, idx
= 0; dev
; dev
= dev
->next
, idx
++) {
1105 if ((in_dev
= __in_dev_get_rcu(dev
)) == NULL
) {
1110 for (ifa
= in_dev
->ifa_list
, ip_idx
= 0; ifa
;
1111 ifa
= ifa
->ifa_next
, ip_idx
++) {
1112 if (ip_idx
< s_ip_idx
)
1114 if (inet_fill_ifaddr(skb
, ifa
, NETLINK_CB(cb
->skb
).pid
,
1116 RTM_NEWADDR
, NLM_F_MULTI
) <= 0) {
1125 read_unlock(&dev_base_lock
);
1127 cb
->args
[1] = ip_idx
;
1132 static void rtmsg_ifa(int event
, struct in_ifaddr
* ifa
)
1134 int size
= NLMSG_SPACE(sizeof(struct ifaddrmsg
) + 128);
1135 struct sk_buff
*skb
= alloc_skb(size
, GFP_KERNEL
);
1138 netlink_set_err(rtnl
, 0, RTNLGRP_IPV4_IFADDR
, ENOBUFS
);
1139 else if (inet_fill_ifaddr(skb
, ifa
, 0, 0, event
, 0) < 0) {
1141 netlink_set_err(rtnl
, 0, RTNLGRP_IPV4_IFADDR
, EINVAL
);
1143 netlink_broadcast(rtnl
, skb
, 0, RTNLGRP_IPV4_IFADDR
, GFP_KERNEL
);
1147 static struct rtnetlink_link inet_rtnetlink_table
[RTM_NR_MSGTYPES
] = {
1148 [RTM_NEWADDR
- RTM_BASE
] = { .doit
= inet_rtm_newaddr
, },
1149 [RTM_DELADDR
- RTM_BASE
] = { .doit
= inet_rtm_deladdr
, },
1150 [RTM_GETADDR
- RTM_BASE
] = { .dumpit
= inet_dump_ifaddr
, },
1151 [RTM_NEWROUTE
- RTM_BASE
] = { .doit
= inet_rtm_newroute
, },
1152 [RTM_DELROUTE
- RTM_BASE
] = { .doit
= inet_rtm_delroute
, },
1153 [RTM_GETROUTE
- RTM_BASE
] = { .doit
= inet_rtm_getroute
,
1154 .dumpit
= inet_dump_fib
, },
1155 #ifdef CONFIG_IP_MULTIPLE_TABLES
1156 [RTM_NEWRULE
- RTM_BASE
] = { .doit
= inet_rtm_newrule
, },
1157 [RTM_DELRULE
- RTM_BASE
] = { .doit
= inet_rtm_delrule
, },
1158 [RTM_GETRULE
- RTM_BASE
] = { .dumpit
= inet_dump_rules
, },
1162 #ifdef CONFIG_SYSCTL
1164 void inet_forward_change(void)
1166 struct net_device
*dev
;
1167 int on
= ipv4_devconf
.forwarding
;
1169 ipv4_devconf
.accept_redirects
= !on
;
1170 ipv4_devconf_dflt
.forwarding
= on
;
1172 read_lock(&dev_base_lock
);
1173 for (dev
= dev_base
; dev
; dev
= dev
->next
) {
1174 struct in_device
*in_dev
;
1176 in_dev
= __in_dev_get_rcu(dev
);
1178 in_dev
->cnf
.forwarding
= on
;
1181 read_unlock(&dev_base_lock
);
1186 static int devinet_sysctl_forward(ctl_table
*ctl
, int write
,
1187 struct file
* filp
, void __user
*buffer
,
1188 size_t *lenp
, loff_t
*ppos
)
1190 int *valp
= ctl
->data
;
1192 int ret
= proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
1194 if (write
&& *valp
!= val
) {
1195 if (valp
== &ipv4_devconf
.forwarding
)
1196 inet_forward_change();
1197 else if (valp
!= &ipv4_devconf_dflt
.forwarding
)
1204 int ipv4_doint_and_flush(ctl_table
*ctl
, int write
,
1205 struct file
* filp
, void __user
*buffer
,
1206 size_t *lenp
, loff_t
*ppos
)
1208 int *valp
= ctl
->data
;
1210 int ret
= proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
1212 if (write
&& *valp
!= val
)
1218 int ipv4_doint_and_flush_strategy(ctl_table
*table
, int __user
*name
, int nlen
,
1219 void __user
*oldval
, size_t __user
*oldlenp
,
1220 void __user
*newval
, size_t newlen
,
1223 int *valp
= table
->data
;
1226 if (!newval
|| !newlen
)
1229 if (newlen
!= sizeof(int))
1232 if (get_user(new, (int __user
*)newval
))
1238 if (oldval
&& oldlenp
) {
1241 if (get_user(len
, oldlenp
))
1245 if (len
> table
->maxlen
)
1246 len
= table
->maxlen
;
1247 if (copy_to_user(oldval
, valp
, len
))
1249 if (put_user(len
, oldlenp
))
1260 static struct devinet_sysctl_table
{
1261 struct ctl_table_header
*sysctl_header
;
1262 ctl_table devinet_vars
[__NET_IPV4_CONF_MAX
];
1263 ctl_table devinet_dev
[2];
1264 ctl_table devinet_conf_dir
[2];
1265 ctl_table devinet_proto_dir
[2];
1266 ctl_table devinet_root_dir
[2];
1267 } devinet_sysctl
= {
1270 .ctl_name
= NET_IPV4_CONF_FORWARDING
,
1271 .procname
= "forwarding",
1272 .data
= &ipv4_devconf
.forwarding
,
1273 .maxlen
= sizeof(int),
1275 .proc_handler
= &devinet_sysctl_forward
,
1278 .ctl_name
= NET_IPV4_CONF_MC_FORWARDING
,
1279 .procname
= "mc_forwarding",
1280 .data
= &ipv4_devconf
.mc_forwarding
,
1281 .maxlen
= sizeof(int),
1283 .proc_handler
= &proc_dointvec
,
1286 .ctl_name
= NET_IPV4_CONF_ACCEPT_REDIRECTS
,
1287 .procname
= "accept_redirects",
1288 .data
= &ipv4_devconf
.accept_redirects
,
1289 .maxlen
= sizeof(int),
1291 .proc_handler
= &proc_dointvec
,
1294 .ctl_name
= NET_IPV4_CONF_SECURE_REDIRECTS
,
1295 .procname
= "secure_redirects",
1296 .data
= &ipv4_devconf
.secure_redirects
,
1297 .maxlen
= sizeof(int),
1299 .proc_handler
= &proc_dointvec
,
1302 .ctl_name
= NET_IPV4_CONF_SHARED_MEDIA
,
1303 .procname
= "shared_media",
1304 .data
= &ipv4_devconf
.shared_media
,
1305 .maxlen
= sizeof(int),
1307 .proc_handler
= &proc_dointvec
,
1310 .ctl_name
= NET_IPV4_CONF_RP_FILTER
,
1311 .procname
= "rp_filter",
1312 .data
= &ipv4_devconf
.rp_filter
,
1313 .maxlen
= sizeof(int),
1315 .proc_handler
= &proc_dointvec
,
1318 .ctl_name
= NET_IPV4_CONF_SEND_REDIRECTS
,
1319 .procname
= "send_redirects",
1320 .data
= &ipv4_devconf
.send_redirects
,
1321 .maxlen
= sizeof(int),
1323 .proc_handler
= &proc_dointvec
,
1326 .ctl_name
= NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE
,
1327 .procname
= "accept_source_route",
1328 .data
= &ipv4_devconf
.accept_source_route
,
1329 .maxlen
= sizeof(int),
1331 .proc_handler
= &proc_dointvec
,
1334 .ctl_name
= NET_IPV4_CONF_PROXY_ARP
,
1335 .procname
= "proxy_arp",
1336 .data
= &ipv4_devconf
.proxy_arp
,
1337 .maxlen
= sizeof(int),
1339 .proc_handler
= &proc_dointvec
,
1342 .ctl_name
= NET_IPV4_CONF_MEDIUM_ID
,
1343 .procname
= "medium_id",
1344 .data
= &ipv4_devconf
.medium_id
,
1345 .maxlen
= sizeof(int),
1347 .proc_handler
= &proc_dointvec
,
1350 .ctl_name
= NET_IPV4_CONF_BOOTP_RELAY
,
1351 .procname
= "bootp_relay",
1352 .data
= &ipv4_devconf
.bootp_relay
,
1353 .maxlen
= sizeof(int),
1355 .proc_handler
= &proc_dointvec
,
1358 .ctl_name
= NET_IPV4_CONF_LOG_MARTIANS
,
1359 .procname
= "log_martians",
1360 .data
= &ipv4_devconf
.log_martians
,
1361 .maxlen
= sizeof(int),
1363 .proc_handler
= &proc_dointvec
,
1366 .ctl_name
= NET_IPV4_CONF_TAG
,
1368 .data
= &ipv4_devconf
.tag
,
1369 .maxlen
= sizeof(int),
1371 .proc_handler
= &proc_dointvec
,
1374 .ctl_name
= NET_IPV4_CONF_ARPFILTER
,
1375 .procname
= "arp_filter",
1376 .data
= &ipv4_devconf
.arp_filter
,
1377 .maxlen
= sizeof(int),
1379 .proc_handler
= &proc_dointvec
,
1382 .ctl_name
= NET_IPV4_CONF_ARP_ANNOUNCE
,
1383 .procname
= "arp_announce",
1384 .data
= &ipv4_devconf
.arp_announce
,
1385 .maxlen
= sizeof(int),
1387 .proc_handler
= &proc_dointvec
,
1390 .ctl_name
= NET_IPV4_CONF_ARP_IGNORE
,
1391 .procname
= "arp_ignore",
1392 .data
= &ipv4_devconf
.arp_ignore
,
1393 .maxlen
= sizeof(int),
1395 .proc_handler
= &proc_dointvec
,
1398 .ctl_name
= NET_IPV4_CONF_NOXFRM
,
1399 .procname
= "disable_xfrm",
1400 .data
= &ipv4_devconf
.no_xfrm
,
1401 .maxlen
= sizeof(int),
1403 .proc_handler
= &ipv4_doint_and_flush
,
1404 .strategy
= &ipv4_doint_and_flush_strategy
,
1407 .ctl_name
= NET_IPV4_CONF_NOPOLICY
,
1408 .procname
= "disable_policy",
1409 .data
= &ipv4_devconf
.no_policy
,
1410 .maxlen
= sizeof(int),
1412 .proc_handler
= &ipv4_doint_and_flush
,
1413 .strategy
= &ipv4_doint_and_flush_strategy
,
1416 .ctl_name
= NET_IPV4_CONF_FORCE_IGMP_VERSION
,
1417 .procname
= "force_igmp_version",
1418 .data
= &ipv4_devconf
.force_igmp_version
,
1419 .maxlen
= sizeof(int),
1421 .proc_handler
= &ipv4_doint_and_flush
,
1422 .strategy
= &ipv4_doint_and_flush_strategy
,
1425 .ctl_name
= NET_IPV4_CONF_PROMOTE_SECONDARIES
,
1426 .procname
= "promote_secondaries",
1427 .data
= &ipv4_devconf
.promote_secondaries
,
1428 .maxlen
= sizeof(int),
1430 .proc_handler
= &ipv4_doint_and_flush
,
1431 .strategy
= &ipv4_doint_and_flush_strategy
,
1436 .ctl_name
= NET_PROTO_CONF_ALL
,
1439 .child
= devinet_sysctl
.devinet_vars
,
1442 .devinet_conf_dir
= {
1444 .ctl_name
= NET_IPV4_CONF
,
1447 .child
= devinet_sysctl
.devinet_dev
,
1450 .devinet_proto_dir
= {
1452 .ctl_name
= NET_IPV4
,
1455 .child
= devinet_sysctl
.devinet_conf_dir
,
1458 .devinet_root_dir
= {
1460 .ctl_name
= CTL_NET
,
1463 .child
= devinet_sysctl
.devinet_proto_dir
,
1468 static void devinet_sysctl_register(struct in_device
*in_dev
,
1469 struct ipv4_devconf
*p
)
1472 struct net_device
*dev
= in_dev
? in_dev
->dev
: NULL
;
1473 struct devinet_sysctl_table
*t
= kmalloc(sizeof(*t
), GFP_KERNEL
);
1474 char *dev_name
= NULL
;
1478 memcpy(t
, &devinet_sysctl
, sizeof(*t
));
1479 for (i
= 0; i
< ARRAY_SIZE(t
->devinet_vars
) - 1; i
++) {
1480 t
->devinet_vars
[i
].data
+= (char *)p
- (char *)&ipv4_devconf
;
1481 t
->devinet_vars
[i
].de
= NULL
;
1485 dev_name
= dev
->name
;
1486 t
->devinet_dev
[0].ctl_name
= dev
->ifindex
;
1488 dev_name
= "default";
1489 t
->devinet_dev
[0].ctl_name
= NET_PROTO_CONF_DEFAULT
;
1493 * Make a copy of dev_name, because '.procname' is regarded as const
1494 * by sysctl and we wouldn't want anyone to change it under our feet
1495 * (see SIOCSIFNAME).
1497 dev_name
= kstrdup(dev_name
, GFP_KERNEL
);
1501 t
->devinet_dev
[0].procname
= dev_name
;
1502 t
->devinet_dev
[0].child
= t
->devinet_vars
;
1503 t
->devinet_dev
[0].de
= NULL
;
1504 t
->devinet_conf_dir
[0].child
= t
->devinet_dev
;
1505 t
->devinet_conf_dir
[0].de
= NULL
;
1506 t
->devinet_proto_dir
[0].child
= t
->devinet_conf_dir
;
1507 t
->devinet_proto_dir
[0].de
= NULL
;
1508 t
->devinet_root_dir
[0].child
= t
->devinet_proto_dir
;
1509 t
->devinet_root_dir
[0].de
= NULL
;
1511 t
->sysctl_header
= register_sysctl_table(t
->devinet_root_dir
, 0);
1512 if (!t
->sysctl_header
)
1526 static void devinet_sysctl_unregister(struct ipv4_devconf
*p
)
1529 struct devinet_sysctl_table
*t
= p
->sysctl
;
1531 unregister_sysctl_table(t
->sysctl_header
);
1532 kfree(t
->devinet_dev
[0].procname
);
1538 void __init
devinet_init(void)
1540 register_gifconf(PF_INET
, inet_gifconf
);
1541 register_netdevice_notifier(&ip_netdev_notifier
);
1542 rtnetlink_links
[PF_INET
] = inet_rtnetlink_table
;
1543 #ifdef CONFIG_SYSCTL
1544 devinet_sysctl
.sysctl_header
=
1545 register_sysctl_table(devinet_sysctl
.devinet_root_dir
, 0);
1546 devinet_sysctl_register(NULL
, &ipv4_devconf_dflt
);
1550 EXPORT_SYMBOL(devinet_ioctl
);
1551 EXPORT_SYMBOL(in_dev_finish_destroy
);
1552 EXPORT_SYMBOL(inet_select_addr
);
1553 EXPORT_SYMBOL(inetdev_by_index
);
1554 EXPORT_SYMBOL(register_inetaddr_notifier
);
1555 EXPORT_SYMBOL(unregister_inetaddr_notifier
);