xref: /openbmc/linux/net/ipv4/devinet.c (revision e8e0929d)
1 /*
2  *	NET3	IP device support routines.
3  *
4  *		This program is free software; you can redistribute it and/or
5  *		modify it under the terms of the GNU General Public License
6  *		as published by the Free Software Foundation; either version
7  *		2 of the License, or (at your option) any later version.
8  *
9  *	Derived from the IP parts of dev.c 1.0.19
10  * 		Authors:	Ross Biro
11  *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12  *				Mark Evans, <evansmp@uhura.aston.ac.uk>
13  *
14  *	Additional Authors:
15  *		Alan Cox, <gw4pts@gw4pts.ampr.org>
16  *		Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
17  *
18  *	Changes:
19  *		Alexey Kuznetsov:	pa_* fields are replaced with ifaddr
20  *					lists.
21  *		Cyrus Durgin:		updated for kmod
22  *		Matthias Andree:	in devinet_ioctl, compare label and
23  *					address (4.4BSD alias style support),
24  *					fall back to comparing just the label
25  *					if no match found.
26  */
27 
28 
29 #include <asm/uaccess.h>
30 #include <asm/system.h>
31 #include <linux/bitops.h>
32 #include <linux/capability.h>
33 #include <linux/module.h>
34 #include <linux/types.h>
35 #include <linux/kernel.h>
36 #include <linux/string.h>
37 #include <linux/mm.h>
38 #include <linux/socket.h>
39 #include <linux/sockios.h>
40 #include <linux/in.h>
41 #include <linux/errno.h>
42 #include <linux/interrupt.h>
43 #include <linux/if_addr.h>
44 #include <linux/if_ether.h>
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <linux/etherdevice.h>
48 #include <linux/skbuff.h>
49 #include <linux/init.h>
50 #include <linux/notifier.h>
51 #include <linux/inetdevice.h>
52 #include <linux/igmp.h>
53 #ifdef CONFIG_SYSCTL
54 #include <linux/sysctl.h>
55 #endif
56 #include <linux/kmod.h>
57 
58 #include <net/arp.h>
59 #include <net/ip.h>
60 #include <net/route.h>
61 #include <net/ip_fib.h>
62 #include <net/rtnetlink.h>
63 #include <net/net_namespace.h>
64 
65 static struct ipv4_devconf ipv4_devconf = {
66 	.data = {
67 		[NET_IPV4_CONF_ACCEPT_REDIRECTS - 1] = 1,
68 		[NET_IPV4_CONF_SEND_REDIRECTS - 1] = 1,
69 		[NET_IPV4_CONF_SECURE_REDIRECTS - 1] = 1,
70 		[NET_IPV4_CONF_SHARED_MEDIA - 1] = 1,
71 	},
72 };
73 
74 static struct ipv4_devconf ipv4_devconf_dflt = {
75 	.data = {
76 		[NET_IPV4_CONF_ACCEPT_REDIRECTS - 1] = 1,
77 		[NET_IPV4_CONF_SEND_REDIRECTS - 1] = 1,
78 		[NET_IPV4_CONF_SECURE_REDIRECTS - 1] = 1,
79 		[NET_IPV4_CONF_SHARED_MEDIA - 1] = 1,
80 		[NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
81 	},
82 };
83 
84 #define IPV4_DEVCONF_DFLT(net, attr) \
85 	IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
86 
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_LABEL]     	= { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
92 };
93 
94 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
95 
96 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
97 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
98 			 int destroy);
99 #ifdef CONFIG_SYSCTL
100 static void devinet_sysctl_register(struct in_device *idev);
101 static void devinet_sysctl_unregister(struct in_device *idev);
102 #else
103 static inline void devinet_sysctl_register(struct in_device *idev)
104 {
105 }
106 static inline void devinet_sysctl_unregister(struct in_device *idev)
107 {
108 }
109 #endif
110 
111 /* Locks all the inet devices. */
112 
113 static struct in_ifaddr *inet_alloc_ifa(void)
114 {
115 	return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
116 }
117 
118 static void inet_rcu_free_ifa(struct rcu_head *head)
119 {
120 	struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
121 	if (ifa->ifa_dev)
122 		in_dev_put(ifa->ifa_dev);
123 	kfree(ifa);
124 }
125 
126 static inline void inet_free_ifa(struct in_ifaddr *ifa)
127 {
128 	call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
129 }
130 
131 void in_dev_finish_destroy(struct in_device *idev)
132 {
133 	struct net_device *dev = idev->dev;
134 
135 	WARN_ON(idev->ifa_list);
136 	WARN_ON(idev->mc_list);
137 #ifdef NET_REFCNT_DEBUG
138 	printk(KERN_DEBUG "in_dev_finish_destroy: %p=%s\n",
139 	       idev, dev ? dev->name : "NIL");
140 #endif
141 	dev_put(dev);
142 	if (!idev->dead)
143 		printk("Freeing alive in_device %p\n", idev);
144 	else {
145 		kfree(idev);
146 	}
147 }
148 
149 static struct in_device *inetdev_init(struct net_device *dev)
150 {
151 	struct in_device *in_dev;
152 
153 	ASSERT_RTNL();
154 
155 	in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
156 	if (!in_dev)
157 		goto out;
158 	memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
159 			sizeof(in_dev->cnf));
160 	in_dev->cnf.sysctl = NULL;
161 	in_dev->dev = dev;
162 	if ((in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl)) == NULL)
163 		goto out_kfree;
164 	if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
165 		dev_disable_lro(dev);
166 	/* Reference in_dev->dev */
167 	dev_hold(dev);
168 	/* Account for reference dev->ip_ptr (below) */
169 	in_dev_hold(in_dev);
170 
171 	devinet_sysctl_register(in_dev);
172 	ip_mc_init_dev(in_dev);
173 	if (dev->flags & IFF_UP)
174 		ip_mc_up(in_dev);
175 
176 	/* we can receive as soon as ip_ptr is set -- do this last */
177 	rcu_assign_pointer(dev->ip_ptr, in_dev);
178 out:
179 	return in_dev;
180 out_kfree:
181 	kfree(in_dev);
182 	in_dev = NULL;
183 	goto out;
184 }
185 
186 static void in_dev_rcu_put(struct rcu_head *head)
187 {
188 	struct in_device *idev = container_of(head, struct in_device, rcu_head);
189 	in_dev_put(idev);
190 }
191 
192 static void inetdev_destroy(struct in_device *in_dev)
193 {
194 	struct in_ifaddr *ifa;
195 	struct net_device *dev;
196 
197 	ASSERT_RTNL();
198 
199 	dev = in_dev->dev;
200 
201 	in_dev->dead = 1;
202 
203 	ip_mc_destroy_dev(in_dev);
204 
205 	while ((ifa = in_dev->ifa_list) != NULL) {
206 		inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
207 		inet_free_ifa(ifa);
208 	}
209 
210 	dev->ip_ptr = NULL;
211 
212 	devinet_sysctl_unregister(in_dev);
213 	neigh_parms_release(&arp_tbl, in_dev->arp_parms);
214 	arp_ifdown(dev);
215 
216 	call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
217 }
218 
219 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
220 {
221 	rcu_read_lock();
222 	for_primary_ifa(in_dev) {
223 		if (inet_ifa_match(a, ifa)) {
224 			if (!b || inet_ifa_match(b, ifa)) {
225 				rcu_read_unlock();
226 				return 1;
227 			}
228 		}
229 	} endfor_ifa(in_dev);
230 	rcu_read_unlock();
231 	return 0;
232 }
233 
234 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
235 			 int destroy, struct nlmsghdr *nlh, u32 pid)
236 {
237 	struct in_ifaddr *promote = NULL;
238 	struct in_ifaddr *ifa, *ifa1 = *ifap;
239 	struct in_ifaddr *last_prim = in_dev->ifa_list;
240 	struct in_ifaddr *prev_prom = NULL;
241 	int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
242 
243 	ASSERT_RTNL();
244 
245 	/* 1. Deleting primary ifaddr forces deletion all secondaries
246 	 * unless alias promotion is set
247 	 **/
248 
249 	if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
250 		struct in_ifaddr **ifap1 = &ifa1->ifa_next;
251 
252 		while ((ifa = *ifap1) != NULL) {
253 			if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
254 			    ifa1->ifa_scope <= ifa->ifa_scope)
255 				last_prim = ifa;
256 
257 			if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
258 			    ifa1->ifa_mask != ifa->ifa_mask ||
259 			    !inet_ifa_match(ifa1->ifa_address, ifa)) {
260 				ifap1 = &ifa->ifa_next;
261 				prev_prom = ifa;
262 				continue;
263 			}
264 
265 			if (!do_promote) {
266 				*ifap1 = ifa->ifa_next;
267 
268 				rtmsg_ifa(RTM_DELADDR, ifa, nlh, pid);
269 				blocking_notifier_call_chain(&inetaddr_chain,
270 						NETDEV_DOWN, ifa);
271 				inet_free_ifa(ifa);
272 			} else {
273 				promote = ifa;
274 				break;
275 			}
276 		}
277 	}
278 
279 	/* 2. Unlink it */
280 
281 	*ifap = ifa1->ifa_next;
282 
283 	/* 3. Announce address deletion */
284 
285 	/* Send message first, then call notifier.
286 	   At first sight, FIB update triggered by notifier
287 	   will refer to already deleted ifaddr, that could confuse
288 	   netlink listeners. It is not true: look, gated sees
289 	   that route deleted and if it still thinks that ifaddr
290 	   is valid, it will try to restore deleted routes... Grr.
291 	   So that, this order is correct.
292 	 */
293 	rtmsg_ifa(RTM_DELADDR, ifa1, nlh, pid);
294 	blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
295 
296 	if (promote) {
297 
298 		if (prev_prom) {
299 			prev_prom->ifa_next = promote->ifa_next;
300 			promote->ifa_next = last_prim->ifa_next;
301 			last_prim->ifa_next = promote;
302 		}
303 
304 		promote->ifa_flags &= ~IFA_F_SECONDARY;
305 		rtmsg_ifa(RTM_NEWADDR, promote, nlh, pid);
306 		blocking_notifier_call_chain(&inetaddr_chain,
307 				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))
311 					continue;
312 			fib_add_ifaddr(ifa);
313 		}
314 
315 	}
316 	if (destroy)
317 		inet_free_ifa(ifa1);
318 }
319 
320 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
321 			 int destroy)
322 {
323 	__inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
324 }
325 
326 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
327 			     u32 pid)
328 {
329 	struct in_device *in_dev = ifa->ifa_dev;
330 	struct in_ifaddr *ifa1, **ifap, **last_primary;
331 
332 	ASSERT_RTNL();
333 
334 	if (!ifa->ifa_local) {
335 		inet_free_ifa(ifa);
336 		return 0;
337 	}
338 
339 	ifa->ifa_flags &= ~IFA_F_SECONDARY;
340 	last_primary = &in_dev->ifa_list;
341 
342 	for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
343 	     ifap = &ifa1->ifa_next) {
344 		if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
345 		    ifa->ifa_scope <= ifa1->ifa_scope)
346 			last_primary = &ifa1->ifa_next;
347 		if (ifa1->ifa_mask == ifa->ifa_mask &&
348 		    inet_ifa_match(ifa1->ifa_address, ifa)) {
349 			if (ifa1->ifa_local == ifa->ifa_local) {
350 				inet_free_ifa(ifa);
351 				return -EEXIST;
352 			}
353 			if (ifa1->ifa_scope != ifa->ifa_scope) {
354 				inet_free_ifa(ifa);
355 				return -EINVAL;
356 			}
357 			ifa->ifa_flags |= IFA_F_SECONDARY;
358 		}
359 	}
360 
361 	if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
362 		net_srandom(ifa->ifa_local);
363 		ifap = last_primary;
364 	}
365 
366 	ifa->ifa_next = *ifap;
367 	*ifap = ifa;
368 
369 	/* Send message first, then call notifier.
370 	   Notifier will trigger FIB update, so that
371 	   listeners of netlink will know about new ifaddr */
372 	rtmsg_ifa(RTM_NEWADDR, ifa, nlh, pid);
373 	blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
374 
375 	return 0;
376 }
377 
378 static int inet_insert_ifa(struct in_ifaddr *ifa)
379 {
380 	return __inet_insert_ifa(ifa, NULL, 0);
381 }
382 
383 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
384 {
385 	struct in_device *in_dev = __in_dev_get_rtnl(dev);
386 
387 	ASSERT_RTNL();
388 
389 	if (!in_dev) {
390 		inet_free_ifa(ifa);
391 		return -ENOBUFS;
392 	}
393 	ipv4_devconf_setall(in_dev);
394 	if (ifa->ifa_dev != in_dev) {
395 		WARN_ON(ifa->ifa_dev);
396 		in_dev_hold(in_dev);
397 		ifa->ifa_dev = in_dev;
398 	}
399 	if (ipv4_is_loopback(ifa->ifa_local))
400 		ifa->ifa_scope = RT_SCOPE_HOST;
401 	return inet_insert_ifa(ifa);
402 }
403 
404 struct in_device *inetdev_by_index(struct net *net, int ifindex)
405 {
406 	struct net_device *dev;
407 	struct in_device *in_dev = NULL;
408 	read_lock(&dev_base_lock);
409 	dev = __dev_get_by_index(net, ifindex);
410 	if (dev)
411 		in_dev = in_dev_get(dev);
412 	read_unlock(&dev_base_lock);
413 	return in_dev;
414 }
415 
416 /* Called only from RTNL semaphored context. No locks. */
417 
418 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
419 				    __be32 mask)
420 {
421 	ASSERT_RTNL();
422 
423 	for_primary_ifa(in_dev) {
424 		if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
425 			return ifa;
426 	} endfor_ifa(in_dev);
427 	return NULL;
428 }
429 
430 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
431 {
432 	struct net *net = sock_net(skb->sk);
433 	struct nlattr *tb[IFA_MAX+1];
434 	struct in_device *in_dev;
435 	struct ifaddrmsg *ifm;
436 	struct in_ifaddr *ifa, **ifap;
437 	int err = -EINVAL;
438 
439 	ASSERT_RTNL();
440 
441 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
442 	if (err < 0)
443 		goto errout;
444 
445 	ifm = nlmsg_data(nlh);
446 	in_dev = inetdev_by_index(net, ifm->ifa_index);
447 	if (in_dev == NULL) {
448 		err = -ENODEV;
449 		goto errout;
450 	}
451 
452 	__in_dev_put(in_dev);
453 
454 	for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
455 	     ifap = &ifa->ifa_next) {
456 		if (tb[IFA_LOCAL] &&
457 		    ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
458 			continue;
459 
460 		if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
461 			continue;
462 
463 		if (tb[IFA_ADDRESS] &&
464 		    (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
465 		    !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
466 			continue;
467 
468 		__inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).pid);
469 		return 0;
470 	}
471 
472 	err = -EADDRNOTAVAIL;
473 errout:
474 	return err;
475 }
476 
477 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh)
478 {
479 	struct nlattr *tb[IFA_MAX+1];
480 	struct in_ifaddr *ifa;
481 	struct ifaddrmsg *ifm;
482 	struct net_device *dev;
483 	struct in_device *in_dev;
484 	int err;
485 
486 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
487 	if (err < 0)
488 		goto errout;
489 
490 	ifm = nlmsg_data(nlh);
491 	err = -EINVAL;
492 	if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
493 		goto errout;
494 
495 	dev = __dev_get_by_index(net, ifm->ifa_index);
496 	err = -ENODEV;
497 	if (dev == NULL)
498 		goto errout;
499 
500 	in_dev = __in_dev_get_rtnl(dev);
501 	err = -ENOBUFS;
502 	if (in_dev == NULL)
503 		goto errout;
504 
505 	ifa = inet_alloc_ifa();
506 	if (ifa == NULL)
507 		/*
508 		 * A potential indev allocation can be left alive, it stays
509 		 * assigned to its device and is destroy with it.
510 		 */
511 		goto errout;
512 
513 	ipv4_devconf_setall(in_dev);
514 	in_dev_hold(in_dev);
515 
516 	if (tb[IFA_ADDRESS] == NULL)
517 		tb[IFA_ADDRESS] = tb[IFA_LOCAL];
518 
519 	ifa->ifa_prefixlen = ifm->ifa_prefixlen;
520 	ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
521 	ifa->ifa_flags = ifm->ifa_flags;
522 	ifa->ifa_scope = ifm->ifa_scope;
523 	ifa->ifa_dev = in_dev;
524 
525 	ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
526 	ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
527 
528 	if (tb[IFA_BROADCAST])
529 		ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
530 
531 	if (tb[IFA_LABEL])
532 		nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
533 	else
534 		memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
535 
536 	return ifa;
537 
538 errout:
539 	return ERR_PTR(err);
540 }
541 
542 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
543 {
544 	struct net *net = sock_net(skb->sk);
545 	struct in_ifaddr *ifa;
546 
547 	ASSERT_RTNL();
548 
549 	ifa = rtm_to_ifaddr(net, nlh);
550 	if (IS_ERR(ifa))
551 		return PTR_ERR(ifa);
552 
553 	return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).pid);
554 }
555 
556 /*
557  *	Determine a default network mask, based on the IP address.
558  */
559 
560 static __inline__ int inet_abc_len(__be32 addr)
561 {
562 	int rc = -1;	/* Something else, probably a multicast. */
563 
564 	if (ipv4_is_zeronet(addr))
565 		rc = 0;
566 	else {
567 		__u32 haddr = ntohl(addr);
568 
569 		if (IN_CLASSA(haddr))
570 			rc = 8;
571 		else if (IN_CLASSB(haddr))
572 			rc = 16;
573 		else if (IN_CLASSC(haddr))
574 			rc = 24;
575 	}
576 
577 	return rc;
578 }
579 
580 
581 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
582 {
583 	struct ifreq ifr;
584 	struct sockaddr_in sin_orig;
585 	struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
586 	struct in_device *in_dev;
587 	struct in_ifaddr **ifap = NULL;
588 	struct in_ifaddr *ifa = NULL;
589 	struct net_device *dev;
590 	char *colon;
591 	int ret = -EFAULT;
592 	int tryaddrmatch = 0;
593 
594 	/*
595 	 *	Fetch the caller's info block into kernel space
596 	 */
597 
598 	if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
599 		goto out;
600 	ifr.ifr_name[IFNAMSIZ - 1] = 0;
601 
602 	/* save original address for comparison */
603 	memcpy(&sin_orig, sin, sizeof(*sin));
604 
605 	colon = strchr(ifr.ifr_name, ':');
606 	if (colon)
607 		*colon = 0;
608 
609 	dev_load(net, ifr.ifr_name);
610 
611 	switch (cmd) {
612 	case SIOCGIFADDR:	/* Get interface address */
613 	case SIOCGIFBRDADDR:	/* Get the broadcast address */
614 	case SIOCGIFDSTADDR:	/* Get the destination address */
615 	case SIOCGIFNETMASK:	/* Get the netmask for the interface */
616 		/* Note that these ioctls will not sleep,
617 		   so that we do not impose a lock.
618 		   One day we will be forced to put shlock here (I mean SMP)
619 		 */
620 		tryaddrmatch = (sin_orig.sin_family == AF_INET);
621 		memset(sin, 0, sizeof(*sin));
622 		sin->sin_family = AF_INET;
623 		break;
624 
625 	case SIOCSIFFLAGS:
626 		ret = -EACCES;
627 		if (!capable(CAP_NET_ADMIN))
628 			goto out;
629 		break;
630 	case SIOCSIFADDR:	/* Set interface address (and family) */
631 	case SIOCSIFBRDADDR:	/* Set the broadcast address */
632 	case SIOCSIFDSTADDR:	/* Set the destination address */
633 	case SIOCSIFNETMASK: 	/* Set the netmask for the interface */
634 		ret = -EACCES;
635 		if (!capable(CAP_NET_ADMIN))
636 			goto out;
637 		ret = -EINVAL;
638 		if (sin->sin_family != AF_INET)
639 			goto out;
640 		break;
641 	default:
642 		ret = -EINVAL;
643 		goto out;
644 	}
645 
646 	rtnl_lock();
647 
648 	ret = -ENODEV;
649 	if ((dev = __dev_get_by_name(net, ifr.ifr_name)) == NULL)
650 		goto done;
651 
652 	if (colon)
653 		*colon = ':';
654 
655 	if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
656 		if (tryaddrmatch) {
657 			/* Matthias Andree */
658 			/* compare label and address (4.4BSD style) */
659 			/* note: we only do this for a limited set of ioctls
660 			   and only if the original address family was AF_INET.
661 			   This is checked above. */
662 			for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
663 			     ifap = &ifa->ifa_next) {
664 				if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
665 				    sin_orig.sin_addr.s_addr ==
666 							ifa->ifa_address) {
667 					break; /* found */
668 				}
669 			}
670 		}
671 		/* we didn't get a match, maybe the application is
672 		   4.3BSD-style and passed in junk so we fall back to
673 		   comparing just the label */
674 		if (!ifa) {
675 			for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
676 			     ifap = &ifa->ifa_next)
677 				if (!strcmp(ifr.ifr_name, ifa->ifa_label))
678 					break;
679 		}
680 	}
681 
682 	ret = -EADDRNOTAVAIL;
683 	if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
684 		goto done;
685 
686 	switch (cmd) {
687 	case SIOCGIFADDR:	/* Get interface address */
688 		sin->sin_addr.s_addr = ifa->ifa_local;
689 		goto rarok;
690 
691 	case SIOCGIFBRDADDR:	/* Get the broadcast address */
692 		sin->sin_addr.s_addr = ifa->ifa_broadcast;
693 		goto rarok;
694 
695 	case SIOCGIFDSTADDR:	/* Get the destination address */
696 		sin->sin_addr.s_addr = ifa->ifa_address;
697 		goto rarok;
698 
699 	case SIOCGIFNETMASK:	/* Get the netmask for the interface */
700 		sin->sin_addr.s_addr = ifa->ifa_mask;
701 		goto rarok;
702 
703 	case SIOCSIFFLAGS:
704 		if (colon) {
705 			ret = -EADDRNOTAVAIL;
706 			if (!ifa)
707 				break;
708 			ret = 0;
709 			if (!(ifr.ifr_flags & IFF_UP))
710 				inet_del_ifa(in_dev, ifap, 1);
711 			break;
712 		}
713 		ret = dev_change_flags(dev, ifr.ifr_flags);
714 		break;
715 
716 	case SIOCSIFADDR:	/* Set interface address (and family) */
717 		ret = -EINVAL;
718 		if (inet_abc_len(sin->sin_addr.s_addr) < 0)
719 			break;
720 
721 		if (!ifa) {
722 			ret = -ENOBUFS;
723 			if ((ifa = inet_alloc_ifa()) == NULL)
724 				break;
725 			if (colon)
726 				memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
727 			else
728 				memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
729 		} else {
730 			ret = 0;
731 			if (ifa->ifa_local == sin->sin_addr.s_addr)
732 				break;
733 			inet_del_ifa(in_dev, ifap, 0);
734 			ifa->ifa_broadcast = 0;
735 			ifa->ifa_scope = 0;
736 		}
737 
738 		ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
739 
740 		if (!(dev->flags & IFF_POINTOPOINT)) {
741 			ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
742 			ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
743 			if ((dev->flags & IFF_BROADCAST) &&
744 			    ifa->ifa_prefixlen < 31)
745 				ifa->ifa_broadcast = ifa->ifa_address |
746 						     ~ifa->ifa_mask;
747 		} else {
748 			ifa->ifa_prefixlen = 32;
749 			ifa->ifa_mask = inet_make_mask(32);
750 		}
751 		ret = inet_set_ifa(dev, ifa);
752 		break;
753 
754 	case SIOCSIFBRDADDR:	/* Set the broadcast address */
755 		ret = 0;
756 		if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
757 			inet_del_ifa(in_dev, ifap, 0);
758 			ifa->ifa_broadcast = sin->sin_addr.s_addr;
759 			inet_insert_ifa(ifa);
760 		}
761 		break;
762 
763 	case SIOCSIFDSTADDR:	/* Set the destination address */
764 		ret = 0;
765 		if (ifa->ifa_address == sin->sin_addr.s_addr)
766 			break;
767 		ret = -EINVAL;
768 		if (inet_abc_len(sin->sin_addr.s_addr) < 0)
769 			break;
770 		ret = 0;
771 		inet_del_ifa(in_dev, ifap, 0);
772 		ifa->ifa_address = sin->sin_addr.s_addr;
773 		inet_insert_ifa(ifa);
774 		break;
775 
776 	case SIOCSIFNETMASK: 	/* Set the netmask for the interface */
777 
778 		/*
779 		 *	The mask we set must be legal.
780 		 */
781 		ret = -EINVAL;
782 		if (bad_mask(sin->sin_addr.s_addr, 0))
783 			break;
784 		ret = 0;
785 		if (ifa->ifa_mask != sin->sin_addr.s_addr) {
786 			__be32 old_mask = ifa->ifa_mask;
787 			inet_del_ifa(in_dev, ifap, 0);
788 			ifa->ifa_mask = sin->sin_addr.s_addr;
789 			ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
790 
791 			/* See if current broadcast address matches
792 			 * with current netmask, then recalculate
793 			 * the broadcast address. Otherwise it's a
794 			 * funny address, so don't touch it since
795 			 * the user seems to know what (s)he's doing...
796 			 */
797 			if ((dev->flags & IFF_BROADCAST) &&
798 			    (ifa->ifa_prefixlen < 31) &&
799 			    (ifa->ifa_broadcast ==
800 			     (ifa->ifa_local|~old_mask))) {
801 				ifa->ifa_broadcast = (ifa->ifa_local |
802 						      ~sin->sin_addr.s_addr);
803 			}
804 			inet_insert_ifa(ifa);
805 		}
806 		break;
807 	}
808 done:
809 	rtnl_unlock();
810 out:
811 	return ret;
812 rarok:
813 	rtnl_unlock();
814 	ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
815 	goto out;
816 }
817 
818 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
819 {
820 	struct in_device *in_dev = __in_dev_get_rtnl(dev);
821 	struct in_ifaddr *ifa;
822 	struct ifreq ifr;
823 	int done = 0;
824 
825 	if (!in_dev || (ifa = in_dev->ifa_list) == NULL)
826 		goto out;
827 
828 	for (; ifa; ifa = ifa->ifa_next) {
829 		if (!buf) {
830 			done += sizeof(ifr);
831 			continue;
832 		}
833 		if (len < (int) sizeof(ifr))
834 			break;
835 		memset(&ifr, 0, sizeof(struct ifreq));
836 		if (ifa->ifa_label)
837 			strcpy(ifr.ifr_name, ifa->ifa_label);
838 		else
839 			strcpy(ifr.ifr_name, dev->name);
840 
841 		(*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
842 		(*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
843 								ifa->ifa_local;
844 
845 		if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
846 			done = -EFAULT;
847 			break;
848 		}
849 		buf  += sizeof(struct ifreq);
850 		len  -= sizeof(struct ifreq);
851 		done += sizeof(struct ifreq);
852 	}
853 out:
854 	return done;
855 }
856 
857 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
858 {
859 	__be32 addr = 0;
860 	struct in_device *in_dev;
861 	struct net *net = dev_net(dev);
862 
863 	rcu_read_lock();
864 	in_dev = __in_dev_get_rcu(dev);
865 	if (!in_dev)
866 		goto no_in_dev;
867 
868 	for_primary_ifa(in_dev) {
869 		if (ifa->ifa_scope > scope)
870 			continue;
871 		if (!dst || inet_ifa_match(dst, ifa)) {
872 			addr = ifa->ifa_local;
873 			break;
874 		}
875 		if (!addr)
876 			addr = ifa->ifa_local;
877 	} endfor_ifa(in_dev);
878 no_in_dev:
879 	rcu_read_unlock();
880 
881 	if (addr)
882 		goto out;
883 
884 	/* Not loopback addresses on loopback should be preferred
885 	   in this case. It is importnat that lo is the first interface
886 	   in dev_base list.
887 	 */
888 	read_lock(&dev_base_lock);
889 	rcu_read_lock();
890 	for_each_netdev(net, dev) {
891 		if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
892 			continue;
893 
894 		for_primary_ifa(in_dev) {
895 			if (ifa->ifa_scope != RT_SCOPE_LINK &&
896 			    ifa->ifa_scope <= scope) {
897 				addr = ifa->ifa_local;
898 				goto out_unlock_both;
899 			}
900 		} endfor_ifa(in_dev);
901 	}
902 out_unlock_both:
903 	read_unlock(&dev_base_lock);
904 	rcu_read_unlock();
905 out:
906 	return addr;
907 }
908 
909 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
910 			      __be32 local, int scope)
911 {
912 	int same = 0;
913 	__be32 addr = 0;
914 
915 	for_ifa(in_dev) {
916 		if (!addr &&
917 		    (local == ifa->ifa_local || !local) &&
918 		    ifa->ifa_scope <= scope) {
919 			addr = ifa->ifa_local;
920 			if (same)
921 				break;
922 		}
923 		if (!same) {
924 			same = (!local || inet_ifa_match(local, ifa)) &&
925 				(!dst || inet_ifa_match(dst, ifa));
926 			if (same && addr) {
927 				if (local || !dst)
928 					break;
929 				/* Is the selected addr into dst subnet? */
930 				if (inet_ifa_match(addr, ifa))
931 					break;
932 				/* No, then can we use new local src? */
933 				if (ifa->ifa_scope <= scope) {
934 					addr = ifa->ifa_local;
935 					break;
936 				}
937 				/* search for large dst subnet for addr */
938 				same = 0;
939 			}
940 		}
941 	} endfor_ifa(in_dev);
942 
943 	return same? addr : 0;
944 }
945 
946 /*
947  * Confirm that local IP address exists using wildcards:
948  * - in_dev: only on this interface, 0=any interface
949  * - dst: only in the same subnet as dst, 0=any dst
950  * - local: address, 0=autoselect the local address
951  * - scope: maximum allowed scope value for the local address
952  */
953 __be32 inet_confirm_addr(struct in_device *in_dev,
954 			 __be32 dst, __be32 local, int scope)
955 {
956 	__be32 addr = 0;
957 	struct net_device *dev;
958 	struct net *net;
959 
960 	if (scope != RT_SCOPE_LINK)
961 		return confirm_addr_indev(in_dev, dst, local, scope);
962 
963 	net = dev_net(in_dev->dev);
964 	read_lock(&dev_base_lock);
965 	rcu_read_lock();
966 	for_each_netdev(net, dev) {
967 		if ((in_dev = __in_dev_get_rcu(dev))) {
968 			addr = confirm_addr_indev(in_dev, dst, local, scope);
969 			if (addr)
970 				break;
971 		}
972 	}
973 	rcu_read_unlock();
974 	read_unlock(&dev_base_lock);
975 
976 	return addr;
977 }
978 
979 /*
980  *	Device notifier
981  */
982 
983 int register_inetaddr_notifier(struct notifier_block *nb)
984 {
985 	return blocking_notifier_chain_register(&inetaddr_chain, nb);
986 }
987 
988 int unregister_inetaddr_notifier(struct notifier_block *nb)
989 {
990 	return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
991 }
992 
993 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
994  * alias numbering and to create unique labels if possible.
995 */
996 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
997 {
998 	struct in_ifaddr *ifa;
999 	int named = 0;
1000 
1001 	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1002 		char old[IFNAMSIZ], *dot;
1003 
1004 		memcpy(old, ifa->ifa_label, IFNAMSIZ);
1005 		memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1006 		if (named++ == 0)
1007 			goto skip;
1008 		dot = strchr(old, ':');
1009 		if (dot == NULL) {
1010 			sprintf(old, ":%d", named);
1011 			dot = old;
1012 		}
1013 		if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) {
1014 			strcat(ifa->ifa_label, dot);
1015 		} else {
1016 			strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1017 		}
1018 skip:
1019 		rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1020 	}
1021 }
1022 
1023 static inline bool inetdev_valid_mtu(unsigned mtu)
1024 {
1025 	return mtu >= 68;
1026 }
1027 
1028 /* Called only under RTNL semaphore */
1029 
1030 static int inetdev_event(struct notifier_block *this, unsigned long event,
1031 			 void *ptr)
1032 {
1033 	struct net_device *dev = ptr;
1034 	struct in_device *in_dev = __in_dev_get_rtnl(dev);
1035 
1036 	ASSERT_RTNL();
1037 
1038 	if (!in_dev) {
1039 		if (event == NETDEV_REGISTER) {
1040 			in_dev = inetdev_init(dev);
1041 			if (!in_dev)
1042 				return notifier_from_errno(-ENOMEM);
1043 			if (dev->flags & IFF_LOOPBACK) {
1044 				IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1045 				IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1046 			}
1047 		} else if (event == NETDEV_CHANGEMTU) {
1048 			/* Re-enabling IP */
1049 			if (inetdev_valid_mtu(dev->mtu))
1050 				in_dev = inetdev_init(dev);
1051 		}
1052 		goto out;
1053 	}
1054 
1055 	switch (event) {
1056 	case NETDEV_REGISTER:
1057 		printk(KERN_DEBUG "inetdev_event: bug\n");
1058 		dev->ip_ptr = NULL;
1059 		break;
1060 	case NETDEV_UP:
1061 		if (!inetdev_valid_mtu(dev->mtu))
1062 			break;
1063 		if (dev->flags & IFF_LOOPBACK) {
1064 			struct in_ifaddr *ifa;
1065 			if ((ifa = inet_alloc_ifa()) != NULL) {
1066 				ifa->ifa_local =
1067 				  ifa->ifa_address = htonl(INADDR_LOOPBACK);
1068 				ifa->ifa_prefixlen = 8;
1069 				ifa->ifa_mask = inet_make_mask(8);
1070 				in_dev_hold(in_dev);
1071 				ifa->ifa_dev = in_dev;
1072 				ifa->ifa_scope = RT_SCOPE_HOST;
1073 				memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1074 				inet_insert_ifa(ifa);
1075 			}
1076 		}
1077 		ip_mc_up(in_dev);
1078 		/* fall through */
1079 	case NETDEV_CHANGEADDR:
1080 		if (IN_DEV_ARP_NOTIFY(in_dev))
1081 			arp_send(ARPOP_REQUEST, ETH_P_ARP,
1082 				 in_dev->ifa_list->ifa_address,
1083 				 dev,
1084 				 in_dev->ifa_list->ifa_address,
1085 				 NULL, dev->dev_addr, NULL);
1086 		break;
1087 	case NETDEV_DOWN:
1088 		ip_mc_down(in_dev);
1089 		break;
1090 	case NETDEV_BONDING_OLDTYPE:
1091 		ip_mc_unmap(in_dev);
1092 		break;
1093 	case NETDEV_BONDING_NEWTYPE:
1094 		ip_mc_remap(in_dev);
1095 		break;
1096 	case NETDEV_CHANGEMTU:
1097 		if (inetdev_valid_mtu(dev->mtu))
1098 			break;
1099 		/* disable IP when MTU is not enough */
1100 	case NETDEV_UNREGISTER:
1101 		inetdev_destroy(in_dev);
1102 		break;
1103 	case NETDEV_CHANGENAME:
1104 		/* Do not notify about label change, this event is
1105 		 * not interesting to applications using netlink.
1106 		 */
1107 		inetdev_changename(dev, in_dev);
1108 
1109 		devinet_sysctl_unregister(in_dev);
1110 		devinet_sysctl_register(in_dev);
1111 		break;
1112 	}
1113 out:
1114 	return NOTIFY_DONE;
1115 }
1116 
1117 static struct notifier_block ip_netdev_notifier = {
1118 	.notifier_call = inetdev_event,
1119 };
1120 
1121 static inline size_t inet_nlmsg_size(void)
1122 {
1123 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1124 	       + nla_total_size(4) /* IFA_ADDRESS */
1125 	       + nla_total_size(4) /* IFA_LOCAL */
1126 	       + nla_total_size(4) /* IFA_BROADCAST */
1127 	       + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
1128 }
1129 
1130 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1131 			    u32 pid, u32 seq, int event, unsigned int flags)
1132 {
1133 	struct ifaddrmsg *ifm;
1134 	struct nlmsghdr  *nlh;
1135 
1136 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*ifm), flags);
1137 	if (nlh == NULL)
1138 		return -EMSGSIZE;
1139 
1140 	ifm = nlmsg_data(nlh);
1141 	ifm->ifa_family = AF_INET;
1142 	ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1143 	ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
1144 	ifm->ifa_scope = ifa->ifa_scope;
1145 	ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1146 
1147 	if (ifa->ifa_address)
1148 		NLA_PUT_BE32(skb, IFA_ADDRESS, ifa->ifa_address);
1149 
1150 	if (ifa->ifa_local)
1151 		NLA_PUT_BE32(skb, IFA_LOCAL, ifa->ifa_local);
1152 
1153 	if (ifa->ifa_broadcast)
1154 		NLA_PUT_BE32(skb, IFA_BROADCAST, ifa->ifa_broadcast);
1155 
1156 	if (ifa->ifa_label[0])
1157 		NLA_PUT_STRING(skb, IFA_LABEL, ifa->ifa_label);
1158 
1159 	return nlmsg_end(skb, nlh);
1160 
1161 nla_put_failure:
1162 	nlmsg_cancel(skb, nlh);
1163 	return -EMSGSIZE;
1164 }
1165 
1166 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1167 {
1168 	struct net *net = sock_net(skb->sk);
1169 	int idx, ip_idx;
1170 	struct net_device *dev;
1171 	struct in_device *in_dev;
1172 	struct in_ifaddr *ifa;
1173 	int s_ip_idx, s_idx = cb->args[0];
1174 
1175 	s_ip_idx = ip_idx = cb->args[1];
1176 	idx = 0;
1177 	for_each_netdev(net, dev) {
1178 		if (idx < s_idx)
1179 			goto cont;
1180 		if (idx > s_idx)
1181 			s_ip_idx = 0;
1182 		if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
1183 			goto cont;
1184 
1185 		for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1186 		     ifa = ifa->ifa_next, ip_idx++) {
1187 			if (ip_idx < s_ip_idx)
1188 				continue;
1189 			if (inet_fill_ifaddr(skb, ifa, NETLINK_CB(cb->skb).pid,
1190 					     cb->nlh->nlmsg_seq,
1191 					     RTM_NEWADDR, NLM_F_MULTI) <= 0)
1192 				goto done;
1193 		}
1194 cont:
1195 		idx++;
1196 	}
1197 
1198 done:
1199 	cb->args[0] = idx;
1200 	cb->args[1] = ip_idx;
1201 
1202 	return skb->len;
1203 }
1204 
1205 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1206 		      u32 pid)
1207 {
1208 	struct sk_buff *skb;
1209 	u32 seq = nlh ? nlh->nlmsg_seq : 0;
1210 	int err = -ENOBUFS;
1211 	struct net *net;
1212 
1213 	net = dev_net(ifa->ifa_dev->dev);
1214 	skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1215 	if (skb == NULL)
1216 		goto errout;
1217 
1218 	err = inet_fill_ifaddr(skb, ifa, pid, seq, event, 0);
1219 	if (err < 0) {
1220 		/* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1221 		WARN_ON(err == -EMSGSIZE);
1222 		kfree_skb(skb);
1223 		goto errout;
1224 	}
1225 	rtnl_notify(skb, net, pid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1226 	return;
1227 errout:
1228 	if (err < 0)
1229 		rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1230 }
1231 
1232 #ifdef CONFIG_SYSCTL
1233 
1234 static void devinet_copy_dflt_conf(struct net *net, int i)
1235 {
1236 	struct net_device *dev;
1237 
1238 	read_lock(&dev_base_lock);
1239 	for_each_netdev(net, dev) {
1240 		struct in_device *in_dev;
1241 		rcu_read_lock();
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];
1245 		rcu_read_unlock();
1246 	}
1247 	read_unlock(&dev_base_lock);
1248 }
1249 
1250 static void inet_forward_change(struct net *net)
1251 {
1252 	struct net_device *dev;
1253 	int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1254 
1255 	IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1256 	IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
1257 
1258 	read_lock(&dev_base_lock);
1259 	for_each_netdev(net, dev) {
1260 		struct in_device *in_dev;
1261 		if (on)
1262 			dev_disable_lro(dev);
1263 		rcu_read_lock();
1264 		in_dev = __in_dev_get_rcu(dev);
1265 		if (in_dev)
1266 			IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1267 		rcu_read_unlock();
1268 	}
1269 	read_unlock(&dev_base_lock);
1270 }
1271 
1272 static int devinet_conf_proc(ctl_table *ctl, int write,
1273 			     void __user *buffer,
1274 			     size_t *lenp, loff_t *ppos)
1275 {
1276 	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1277 
1278 	if (write) {
1279 		struct ipv4_devconf *cnf = ctl->extra1;
1280 		struct net *net = ctl->extra2;
1281 		int i = (int *)ctl->data - cnf->data;
1282 
1283 		set_bit(i, cnf->state);
1284 
1285 		if (cnf == net->ipv4.devconf_dflt)
1286 			devinet_copy_dflt_conf(net, i);
1287 	}
1288 
1289 	return ret;
1290 }
1291 
1292 static int devinet_conf_sysctl(ctl_table *table,
1293 			       void __user *oldval, size_t __user *oldlenp,
1294 			       void __user *newval, size_t newlen)
1295 {
1296 	struct ipv4_devconf *cnf;
1297 	struct net *net;
1298 	int *valp = table->data;
1299 	int new;
1300 	int i;
1301 
1302 	if (!newval || !newlen)
1303 		return 0;
1304 
1305 	if (newlen != sizeof(int))
1306 		return -EINVAL;
1307 
1308 	if (get_user(new, (int __user *)newval))
1309 		return -EFAULT;
1310 
1311 	if (new == *valp)
1312 		return 0;
1313 
1314 	if (oldval && oldlenp) {
1315 		size_t len;
1316 
1317 		if (get_user(len, oldlenp))
1318 			return -EFAULT;
1319 
1320 		if (len) {
1321 			if (len > table->maxlen)
1322 				len = table->maxlen;
1323 			if (copy_to_user(oldval, valp, len))
1324 				return -EFAULT;
1325 			if (put_user(len, oldlenp))
1326 				return -EFAULT;
1327 		}
1328 	}
1329 
1330 	*valp = new;
1331 
1332 	cnf = table->extra1;
1333 	net = table->extra2;
1334 	i = (int *)table->data - cnf->data;
1335 
1336 	set_bit(i, cnf->state);
1337 
1338 	if (cnf == net->ipv4.devconf_dflt)
1339 		devinet_copy_dflt_conf(net, i);
1340 
1341 	return 1;
1342 }
1343 
1344 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1345 				  void __user *buffer,
1346 				  size_t *lenp, loff_t *ppos)
1347 {
1348 	int *valp = ctl->data;
1349 	int val = *valp;
1350 	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1351 
1352 	if (write && *valp != val) {
1353 		struct net *net = ctl->extra2;
1354 
1355 		if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
1356 			if (!rtnl_trylock())
1357 				return restart_syscall();
1358 			if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
1359 				inet_forward_change(net);
1360 			} else if (*valp) {
1361 				struct ipv4_devconf *cnf = ctl->extra1;
1362 				struct in_device *idev =
1363 					container_of(cnf, struct in_device, cnf);
1364 				dev_disable_lro(idev->dev);
1365 			}
1366 			rtnl_unlock();
1367 			rt_cache_flush(net, 0);
1368 		}
1369 	}
1370 
1371 	return ret;
1372 }
1373 
1374 int ipv4_doint_and_flush(ctl_table *ctl, int write,
1375 			 void __user *buffer,
1376 			 size_t *lenp, loff_t *ppos)
1377 {
1378 	int *valp = ctl->data;
1379 	int val = *valp;
1380 	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1381 	struct net *net = ctl->extra2;
1382 
1383 	if (write && *valp != val)
1384 		rt_cache_flush(net, 0);
1385 
1386 	return ret;
1387 }
1388 
1389 int ipv4_doint_and_flush_strategy(ctl_table *table,
1390 				  void __user *oldval, size_t __user *oldlenp,
1391 				  void __user *newval, size_t newlen)
1392 {
1393 	int ret = devinet_conf_sysctl(table, oldval, oldlenp, newval, newlen);
1394 	struct net *net = table->extra2;
1395 
1396 	if (ret == 1)
1397 		rt_cache_flush(net, 0);
1398 
1399 	return ret;
1400 }
1401 
1402 
1403 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc, sysctl) \
1404 	{ \
1405 		.ctl_name	= NET_IPV4_CONF_ ## attr, \
1406 		.procname	= name, \
1407 		.data		= ipv4_devconf.data + \
1408 				  NET_IPV4_CONF_ ## attr - 1, \
1409 		.maxlen		= sizeof(int), \
1410 		.mode		= mval, \
1411 		.proc_handler	= proc, \
1412 		.strategy	= sysctl, \
1413 		.extra1		= &ipv4_devconf, \
1414 	}
1415 
1416 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
1417 	DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc, \
1418 			     devinet_conf_sysctl)
1419 
1420 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
1421 	DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc, \
1422 			     devinet_conf_sysctl)
1423 
1424 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc, sysctl) \
1425 	DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc, sysctl)
1426 
1427 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
1428 	DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush, \
1429 				     ipv4_doint_and_flush_strategy)
1430 
1431 static struct devinet_sysctl_table {
1432 	struct ctl_table_header *sysctl_header;
1433 	struct ctl_table devinet_vars[__NET_IPV4_CONF_MAX];
1434 	char *dev_name;
1435 } devinet_sysctl = {
1436 	.devinet_vars = {
1437 		DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
1438 					     devinet_sysctl_forward,
1439 					     devinet_conf_sysctl),
1440 		DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
1441 
1442 		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
1443 		DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
1444 		DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
1445 		DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
1446 		DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
1447 		DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
1448 					"accept_source_route"),
1449 		DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
1450 		DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
1451 		DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
1452 		DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
1453 		DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
1454 		DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
1455 		DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
1456 		DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
1457 		DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
1458 		DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
1459 
1460 		DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
1461 		DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
1462 		DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
1463 					      "force_igmp_version"),
1464 		DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
1465 					      "promote_secondaries"),
1466 	},
1467 };
1468 
1469 static int __devinet_sysctl_register(struct net *net, char *dev_name,
1470 		int ctl_name, struct ipv4_devconf *p)
1471 {
1472 	int i;
1473 	struct devinet_sysctl_table *t;
1474 
1475 #define DEVINET_CTL_PATH_DEV	3
1476 
1477 	struct ctl_path devinet_ctl_path[] = {
1478 		{ .procname = "net", .ctl_name = CTL_NET, },
1479 		{ .procname = "ipv4", .ctl_name = NET_IPV4, },
1480 		{ .procname = "conf", .ctl_name = NET_IPV4_CONF, },
1481 		{ /* to be set */ },
1482 		{ },
1483 	};
1484 
1485 	t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1486 	if (!t)
1487 		goto out;
1488 
1489 	for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
1490 		t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
1491 		t->devinet_vars[i].extra1 = p;
1492 		t->devinet_vars[i].extra2 = net;
1493 	}
1494 
1495 	/*
1496 	 * Make a copy of dev_name, because '.procname' is regarded as const
1497 	 * by sysctl and we wouldn't want anyone to change it under our feet
1498 	 * (see SIOCSIFNAME).
1499 	 */
1500 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
1501 	if (!t->dev_name)
1502 		goto free;
1503 
1504 	devinet_ctl_path[DEVINET_CTL_PATH_DEV].procname = t->dev_name;
1505 	devinet_ctl_path[DEVINET_CTL_PATH_DEV].ctl_name = ctl_name;
1506 
1507 	t->sysctl_header = register_net_sysctl_table(net, devinet_ctl_path,
1508 			t->devinet_vars);
1509 	if (!t->sysctl_header)
1510 		goto free_procname;
1511 
1512 	p->sysctl = t;
1513 	return 0;
1514 
1515 free_procname:
1516 	kfree(t->dev_name);
1517 free:
1518 	kfree(t);
1519 out:
1520 	return -ENOBUFS;
1521 }
1522 
1523 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
1524 {
1525 	struct devinet_sysctl_table *t = cnf->sysctl;
1526 
1527 	if (t == NULL)
1528 		return;
1529 
1530 	cnf->sysctl = NULL;
1531 	unregister_sysctl_table(t->sysctl_header);
1532 	kfree(t->dev_name);
1533 	kfree(t);
1534 }
1535 
1536 static void devinet_sysctl_register(struct in_device *idev)
1537 {
1538 	neigh_sysctl_register(idev->dev, idev->arp_parms, NET_IPV4,
1539 			NET_IPV4_NEIGH, "ipv4", NULL, NULL);
1540 	__devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
1541 			idev->dev->ifindex, &idev->cnf);
1542 }
1543 
1544 static void devinet_sysctl_unregister(struct in_device *idev)
1545 {
1546 	__devinet_sysctl_unregister(&idev->cnf);
1547 	neigh_sysctl_unregister(idev->arp_parms);
1548 }
1549 
1550 static struct ctl_table ctl_forward_entry[] = {
1551 	{
1552 		.ctl_name	= NET_IPV4_FORWARD,
1553 		.procname	= "ip_forward",
1554 		.data		= &ipv4_devconf.data[
1555 					NET_IPV4_CONF_FORWARDING - 1],
1556 		.maxlen		= sizeof(int),
1557 		.mode		= 0644,
1558 		.proc_handler	= devinet_sysctl_forward,
1559 		.strategy	= devinet_conf_sysctl,
1560 		.extra1		= &ipv4_devconf,
1561 		.extra2		= &init_net,
1562 	},
1563 	{ },
1564 };
1565 
1566 static __net_initdata struct ctl_path net_ipv4_path[] = {
1567 	{ .procname = "net", .ctl_name = CTL_NET, },
1568 	{ .procname = "ipv4", .ctl_name = NET_IPV4, },
1569 	{ },
1570 };
1571 #endif
1572 
1573 static __net_init int devinet_init_net(struct net *net)
1574 {
1575 	int err;
1576 	struct ipv4_devconf *all, *dflt;
1577 #ifdef CONFIG_SYSCTL
1578 	struct ctl_table *tbl = ctl_forward_entry;
1579 	struct ctl_table_header *forw_hdr;
1580 #endif
1581 
1582 	err = -ENOMEM;
1583 	all = &ipv4_devconf;
1584 	dflt = &ipv4_devconf_dflt;
1585 
1586 	if (net != &init_net) {
1587 		all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
1588 		if (all == NULL)
1589 			goto err_alloc_all;
1590 
1591 		dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
1592 		if (dflt == NULL)
1593 			goto err_alloc_dflt;
1594 
1595 #ifdef CONFIG_SYSCTL
1596 		tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
1597 		if (tbl == NULL)
1598 			goto err_alloc_ctl;
1599 
1600 		tbl[0].data = &all->data[NET_IPV4_CONF_FORWARDING - 1];
1601 		tbl[0].extra1 = all;
1602 		tbl[0].extra2 = net;
1603 #endif
1604 	}
1605 
1606 #ifdef CONFIG_SYSCTL
1607 	err = __devinet_sysctl_register(net, "all",
1608 			NET_PROTO_CONF_ALL, all);
1609 	if (err < 0)
1610 		goto err_reg_all;
1611 
1612 	err = __devinet_sysctl_register(net, "default",
1613 			NET_PROTO_CONF_DEFAULT, dflt);
1614 	if (err < 0)
1615 		goto err_reg_dflt;
1616 
1617 	err = -ENOMEM;
1618 	forw_hdr = register_net_sysctl_table(net, net_ipv4_path, tbl);
1619 	if (forw_hdr == NULL)
1620 		goto err_reg_ctl;
1621 	net->ipv4.forw_hdr = forw_hdr;
1622 #endif
1623 
1624 	net->ipv4.devconf_all = all;
1625 	net->ipv4.devconf_dflt = dflt;
1626 	return 0;
1627 
1628 #ifdef CONFIG_SYSCTL
1629 err_reg_ctl:
1630 	__devinet_sysctl_unregister(dflt);
1631 err_reg_dflt:
1632 	__devinet_sysctl_unregister(all);
1633 err_reg_all:
1634 	if (tbl != ctl_forward_entry)
1635 		kfree(tbl);
1636 err_alloc_ctl:
1637 #endif
1638 	if (dflt != &ipv4_devconf_dflt)
1639 		kfree(dflt);
1640 err_alloc_dflt:
1641 	if (all != &ipv4_devconf)
1642 		kfree(all);
1643 err_alloc_all:
1644 	return err;
1645 }
1646 
1647 static __net_exit void devinet_exit_net(struct net *net)
1648 {
1649 #ifdef CONFIG_SYSCTL
1650 	struct ctl_table *tbl;
1651 
1652 	tbl = net->ipv4.forw_hdr->ctl_table_arg;
1653 	unregister_net_sysctl_table(net->ipv4.forw_hdr);
1654 	__devinet_sysctl_unregister(net->ipv4.devconf_dflt);
1655 	__devinet_sysctl_unregister(net->ipv4.devconf_all);
1656 	kfree(tbl);
1657 #endif
1658 	kfree(net->ipv4.devconf_dflt);
1659 	kfree(net->ipv4.devconf_all);
1660 }
1661 
1662 static __net_initdata struct pernet_operations devinet_ops = {
1663 	.init = devinet_init_net,
1664 	.exit = devinet_exit_net,
1665 };
1666 
1667 void __init devinet_init(void)
1668 {
1669 	register_pernet_subsys(&devinet_ops);
1670 
1671 	register_gifconf(PF_INET, inet_gifconf);
1672 	register_netdevice_notifier(&ip_netdev_notifier);
1673 
1674 	rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL);
1675 	rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL);
1676 	rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr);
1677 }
1678 
1679 EXPORT_SYMBOL(in_dev_finish_destroy);
1680 EXPORT_SYMBOL(inet_select_addr);
1681 EXPORT_SYMBOL(inetdev_by_index);
1682 EXPORT_SYMBOL(register_inetaddr_notifier);
1683 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1684