[RTNETLINK]: Unexport rtnl socket
[safe/jmp/linux-2.6] / net / core / rtnetlink.c
1 /*
2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
3  *              operating system.  INET is implemented using the  BSD Socket
4  *              interface as the means of communication with the user level.
5  *
6  *              Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *              This program is free software; you can redistribute it and/or
11  *              modify it under the terms of the GNU General Public License
12  *              as published by the Free Software Foundation; either version
13  *              2 of the License, or (at your option) any later version.
14  *
15  *      Fixes:
16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
17  */
18
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/sched.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
42 #include <asm/string.h>
43
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <net/arp.h>
49 #include <net/route.h>
50 #include <net/udp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/netlink.h>
55 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
56 #include <linux/wireless.h>
57 #include <net/iw_handler.h>
58 #endif  /* CONFIG_NET_WIRELESS_RTNETLINK */
59
60 static DEFINE_MUTEX(rtnl_mutex);
61 static struct sock *rtnl;
62
63 void rtnl_lock(void)
64 {
65         mutex_lock(&rtnl_mutex);
66 }
67
68 void __rtnl_unlock(void)
69 {
70         mutex_unlock(&rtnl_mutex);
71 }
72
73 void rtnl_unlock(void)
74 {
75         mutex_unlock(&rtnl_mutex);
76         if (rtnl && rtnl->sk_receive_queue.qlen)
77                 rtnl->sk_data_ready(rtnl, 0);
78         netdev_run_todo();
79 }
80
81 int rtnl_trylock(void)
82 {
83         return mutex_trylock(&rtnl_mutex);
84 }
85
86 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
87 {
88         memset(tb, 0, sizeof(struct rtattr*)*maxattr);
89
90         while (RTA_OK(rta, len)) {
91                 unsigned flavor = rta->rta_type;
92                 if (flavor && flavor <= maxattr)
93                         tb[flavor-1] = rta;
94                 rta = RTA_NEXT(rta, len);
95         }
96         return 0;
97 }
98
99 struct rtnetlink_link * rtnetlink_links[NPROTO];
100
101 static const int rtm_min[RTM_NR_FAMILIES] =
102 {
103         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
104         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
105         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
106         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
107         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
108         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
109         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
110         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
111         [RTM_FAM(RTM_NEWPREFIX)]    = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
112         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
113         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
114 };
115
116 static const int rta_max[RTM_NR_FAMILIES] =
117 {
118         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
119         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
120         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
121         [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
122         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
123         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
124         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
125         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
126 };
127
128 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
129 {
130         struct rtattr *rta;
131         int size = RTA_LENGTH(attrlen);
132
133         rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
134         rta->rta_type = attrtype;
135         rta->rta_len = size;
136         memcpy(RTA_DATA(rta), data, attrlen);
137         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
138 }
139
140 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
141 {
142         size_t ret = RTA_PAYLOAD(rta);
143         char *src = RTA_DATA(rta);
144
145         if (ret > 0 && src[ret - 1] == '\0')
146                 ret--;
147         if (size > 0) {
148                 size_t len = (ret >= size) ? size - 1 : ret;
149                 memset(dest, 0, size);
150                 memcpy(dest, src, len);
151         }
152         return ret;
153 }
154
155 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
156 {
157         int err = 0;
158
159         NETLINK_CB(skb).dst_group = group;
160         if (echo)
161                 atomic_inc(&skb->users);
162         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
163         if (echo)
164                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
165         return err;
166 }
167
168 int rtnl_unicast(struct sk_buff *skb, u32 pid)
169 {
170         return nlmsg_unicast(rtnl, skb, pid);
171 }
172
173 int rtnl_notify(struct sk_buff *skb, u32 pid, u32 group,
174                 struct nlmsghdr *nlh, gfp_t flags)
175 {
176         int report = 0;
177
178         if (nlh)
179                 report = nlmsg_report(nlh);
180
181         return nlmsg_notify(rtnl, skb, pid, group, report, flags);
182 }
183
184 void rtnl_set_sk_err(u32 group, int error)
185 {
186         netlink_set_err(rtnl, 0, group, error);
187 }
188
189 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
190 {
191         struct rtattr *mx = (struct rtattr*)skb->tail;
192         int i;
193
194         RTA_PUT(skb, RTA_METRICS, 0, NULL);
195         for (i=0; i<RTAX_MAX; i++) {
196                 if (metrics[i])
197                         RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
198         }
199         mx->rta_len = skb->tail - (u8*)mx;
200         if (mx->rta_len == RTA_LENGTH(0))
201                 skb_trim(skb, (u8*)mx - skb->data);
202         return 0;
203
204 rtattr_failure:
205         skb_trim(skb, (u8*)mx - skb->data);
206         return -1;
207 }
208
209
210 static void set_operstate(struct net_device *dev, unsigned char transition)
211 {
212         unsigned char operstate = dev->operstate;
213
214         switch(transition) {
215         case IF_OPER_UP:
216                 if ((operstate == IF_OPER_DORMANT ||
217                      operstate == IF_OPER_UNKNOWN) &&
218                     !netif_dormant(dev))
219                         operstate = IF_OPER_UP;
220                 break;
221
222         case IF_OPER_DORMANT:
223                 if (operstate == IF_OPER_UP ||
224                     operstate == IF_OPER_UNKNOWN)
225                         operstate = IF_OPER_DORMANT;
226                 break;
227         };
228
229         if (dev->operstate != operstate) {
230                 write_lock_bh(&dev_base_lock);
231                 dev->operstate = operstate;
232                 write_unlock_bh(&dev_base_lock);
233                 netdev_state_change(dev);
234         }
235 }
236
237 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
238                                  struct net_device_stats *b)
239 {
240         a->rx_packets = b->rx_packets;
241         a->tx_packets = b->tx_packets;
242         a->rx_bytes = b->rx_bytes;
243         a->tx_bytes = b->tx_bytes;
244         a->rx_errors = b->rx_errors;
245         a->tx_errors = b->tx_errors;
246         a->rx_dropped = b->rx_dropped;
247         a->tx_dropped = b->tx_dropped;
248
249         a->multicast = b->multicast;
250         a->collisions = b->collisions;
251
252         a->rx_length_errors = b->rx_length_errors;
253         a->rx_over_errors = b->rx_over_errors;
254         a->rx_crc_errors = b->rx_crc_errors;
255         a->rx_frame_errors = b->rx_frame_errors;
256         a->rx_fifo_errors = b->rx_fifo_errors;
257         a->rx_missed_errors = b->rx_missed_errors;
258
259         a->tx_aborted_errors = b->tx_aborted_errors;
260         a->tx_carrier_errors = b->tx_carrier_errors;
261         a->tx_fifo_errors = b->tx_fifo_errors;
262         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
263         a->tx_window_errors = b->tx_window_errors;
264
265         a->rx_compressed = b->rx_compressed;
266         a->tx_compressed = b->tx_compressed;
267 };
268
269 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
270                             void *iwbuf, int iwbuflen, int type, u32 pid,
271                             u32 seq, u32 change, unsigned int flags)
272 {
273         struct ifinfomsg *ifm;
274         struct nlmsghdr *nlh;
275
276         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
277         if (nlh == NULL)
278                 return -ENOBUFS;
279
280         ifm = nlmsg_data(nlh);
281         ifm->ifi_family = AF_UNSPEC;
282         ifm->__ifi_pad = 0;
283         ifm->ifi_type = dev->type;
284         ifm->ifi_index = dev->ifindex;
285         ifm->ifi_flags = dev_get_flags(dev);
286         ifm->ifi_change = change;
287
288         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
289         NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
290         NLA_PUT_U32(skb, IFLA_WEIGHT, dev->weight);
291         NLA_PUT_U8(skb, IFLA_OPERSTATE,
292                    netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
293         NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
294         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
295
296         if (dev->ifindex != dev->iflink)
297                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
298
299         if (dev->master)
300                 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
301
302         if (dev->qdisc_sleeping)
303                 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
304
305         if (1) {
306                 struct rtnl_link_ifmap map = {
307                         .mem_start   = dev->mem_start,
308                         .mem_end     = dev->mem_end,
309                         .base_addr   = dev->base_addr,
310                         .irq         = dev->irq,
311                         .dma         = dev->dma,
312                         .port        = dev->if_port,
313                 };
314                 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
315         }
316
317         if (dev->addr_len) {
318                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
319                 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
320         }
321
322         if (dev->get_stats) {
323                 struct net_device_stats *stats = dev->get_stats(dev);
324                 if (stats) {
325                         struct nlattr *attr;
326
327                         attr = nla_reserve(skb, IFLA_STATS,
328                                            sizeof(struct rtnl_link_stats));
329                         if (attr == NULL)
330                                 goto nla_put_failure;
331
332                         copy_rtnl_link_stats(nla_data(attr), stats);
333                 }
334         }
335
336         if (iwbuf)
337                 NLA_PUT(skb, IFLA_WIRELESS, iwbuflen, iwbuf);
338
339         return nlmsg_end(skb, nlh);
340
341 nla_put_failure:
342         return nlmsg_cancel(skb, nlh);
343 }
344
345 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
346 {
347         int idx;
348         int s_idx = cb->args[0];
349         struct net_device *dev;
350
351         read_lock(&dev_base_lock);
352         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
353                 if (idx < s_idx)
354                         continue;
355                 if (rtnl_fill_ifinfo(skb, dev, NULL, 0, RTM_NEWLINK,
356                                      NETLINK_CB(cb->skb).pid,
357                                      cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
358                         break;
359         }
360         read_unlock(&dev_base_lock);
361         cb->args[0] = idx;
362
363         return skb->len;
364 }
365
366 static struct nla_policy ifla_policy[IFLA_MAX+1] __read_mostly = {
367         [IFLA_IFNAME]           = { .type = NLA_STRING },
368         [IFLA_MAP]              = { .minlen = sizeof(struct rtnl_link_ifmap) },
369         [IFLA_MTU]              = { .type = NLA_U32 },
370         [IFLA_TXQLEN]           = { .type = NLA_U32 },
371         [IFLA_WEIGHT]           = { .type = NLA_U32 },
372         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
373         [IFLA_LINKMODE]         = { .type = NLA_U8 },
374 };
375
376 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
377 {
378         struct ifinfomsg *ifm;
379         struct net_device *dev;
380         int err, send_addr_notify = 0, modified = 0;
381         struct nlattr *tb[IFLA_MAX+1];
382         char ifname[IFNAMSIZ];
383
384         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
385         if (err < 0)
386                 goto errout;
387
388         if (tb[IFLA_IFNAME] &&
389             nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ) >= IFNAMSIZ)
390                 return -EINVAL;
391
392         err = -EINVAL;
393         ifm = nlmsg_data(nlh);
394         if (ifm->ifi_index >= 0)
395                 dev = dev_get_by_index(ifm->ifi_index);
396         else if (tb[IFLA_IFNAME])
397                 dev = dev_get_by_name(ifname);
398         else
399                 goto errout;
400
401         if (dev == NULL) {
402                 err = -ENODEV;
403                 goto errout;
404         }
405
406         if (tb[IFLA_ADDRESS] &&
407             nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
408                 goto errout_dev;
409
410         if (tb[IFLA_BROADCAST] &&
411             nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
412                 goto errout_dev;
413
414         if (tb[IFLA_MAP]) {
415                 struct rtnl_link_ifmap *u_map;
416                 struct ifmap k_map;
417
418                 if (!dev->set_config) {
419                         err = -EOPNOTSUPP;
420                         goto errout_dev;
421                 }
422
423                 if (!netif_device_present(dev)) {
424                         err = -ENODEV;
425                         goto errout_dev;
426                 }
427
428                 u_map = nla_data(tb[IFLA_MAP]);
429                 k_map.mem_start = (unsigned long) u_map->mem_start;
430                 k_map.mem_end = (unsigned long) u_map->mem_end;
431                 k_map.base_addr = (unsigned short) u_map->base_addr;
432                 k_map.irq = (unsigned char) u_map->irq;
433                 k_map.dma = (unsigned char) u_map->dma;
434                 k_map.port = (unsigned char) u_map->port;
435
436                 err = dev->set_config(dev, &k_map);
437                 if (err < 0)
438                         goto errout_dev;
439
440                 modified = 1;
441         }
442
443         if (tb[IFLA_ADDRESS]) {
444                 struct sockaddr *sa;
445                 int len;
446
447                 if (!dev->set_mac_address) {
448                         err = -EOPNOTSUPP;
449                         goto errout_dev;
450                 }
451
452                 if (!netif_device_present(dev)) {
453                         err = -ENODEV;
454                         goto errout_dev;
455                 }
456
457                 len = sizeof(sa_family_t) + dev->addr_len;
458                 sa = kmalloc(len, GFP_KERNEL);
459                 if (!sa) {
460                         err = -ENOMEM;
461                         goto errout_dev;
462                 }
463                 sa->sa_family = dev->type;
464                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
465                        dev->addr_len);
466                 err = dev->set_mac_address(dev, sa);
467                 kfree(sa);
468                 if (err)
469                         goto errout_dev;
470                 send_addr_notify = 1;
471                 modified = 1;
472         }
473
474         if (tb[IFLA_MTU]) {
475                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
476                 if (err < 0)
477                         goto errout_dev;
478                 modified = 1;
479         }
480
481         /*
482          * Interface selected by interface index but interface
483          * name provided implies that a name change has been
484          * requested.
485          */
486         if (ifm->ifi_index >= 0 && ifname[0]) {
487                 err = dev_change_name(dev, ifname);
488                 if (err < 0)
489                         goto errout_dev;
490                 modified = 1;
491         }
492
493 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
494         if (tb[IFLA_WIRELESS]) {
495                 /* Call Wireless Extensions.
496                  * Various stuff checked in there... */
497                 err = wireless_rtnetlink_set(dev, nla_data(tb[IFLA_WIRELESS]),
498                                              nla_len(tb[IFLA_WIRELESS]));
499                 if (err < 0)
500                         goto errout_dev;
501         }
502 #endif  /* CONFIG_NET_WIRELESS_RTNETLINK */
503
504         if (tb[IFLA_BROADCAST]) {
505                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
506                 send_addr_notify = 1;
507         }
508
509
510         if (ifm->ifi_flags)
511                 dev_change_flags(dev, ifm->ifi_flags);
512
513         if (tb[IFLA_TXQLEN])
514                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
515
516         if (tb[IFLA_WEIGHT])
517                 dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
518
519         if (tb[IFLA_OPERSTATE])
520                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
521
522         if (tb[IFLA_LINKMODE]) {
523                 write_lock_bh(&dev_base_lock);
524                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
525                 write_unlock_bh(&dev_base_lock);
526         }
527
528         err = 0;
529
530 errout_dev:
531         if (err < 0 && modified && net_ratelimit())
532                 printk(KERN_WARNING "A link change request failed with "
533                        "some changes comitted already. Interface %s may "
534                        "have been left with an inconsistent configuration, "
535                        "please check.\n", dev->name);
536
537         if (send_addr_notify)
538                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
539
540         dev_put(dev);
541 errout:
542         return err;
543 }
544
545 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
546 {
547         struct ifinfomsg *ifm;
548         struct nlattr *tb[IFLA_MAX+1];
549         struct net_device *dev = NULL;
550         struct sk_buff *nskb;
551         char *iw_buf = NULL, *iw = NULL;
552         int iw_buf_len = 0;
553         int err, payload;
554
555         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
556         if (err < 0)
557                 goto errout;
558
559         ifm = nlmsg_data(nlh);
560         if (ifm->ifi_index >= 0) {
561                 dev = dev_get_by_index(ifm->ifi_index);
562                 if (dev == NULL)
563                         return -ENODEV;
564         } else
565                 return -EINVAL;
566
567
568 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
569         if (tb[IFLA_WIRELESS]) {
570                 /* Call Wireless Extensions. We need to know the size before
571                  * we can alloc. Various stuff checked in there... */
572                 err = wireless_rtnetlink_get(dev, nla_data(tb[IFLA_WIRELESS]),
573                                              nla_len(tb[IFLA_WIRELESS]),
574                                              &iw_buf, &iw_buf_len);
575                 if (err < 0)
576                         goto errout;
577
578                 iw += IW_EV_POINT_OFF;
579         }
580 #endif  /* CONFIG_NET_WIRELESS_RTNETLINK */
581
582         payload = NLMSG_ALIGN(sizeof(struct ifinfomsg) +
583                               nla_total_size(iw_buf_len));
584         nskb = nlmsg_new(nlmsg_total_size(payload), GFP_KERNEL);
585         if (nskb == NULL) {
586                 err = -ENOBUFS;
587                 goto errout;
588         }
589
590         err = rtnl_fill_ifinfo(nskb, dev, iw, iw_buf_len, RTM_NEWLINK,
591                                NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 0);
592         if (err <= 0) {
593                 kfree_skb(skb);
594                 goto errout;
595         }
596
597         err = rtnl_unicast(skb, NETLINK_CB(skb).pid);
598 errout:
599         kfree(iw_buf);
600         dev_put(dev);
601
602         return err;
603 }
604
605 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
606 {
607         int idx;
608         int s_idx = cb->family;
609
610         if (s_idx == 0)
611                 s_idx = 1;
612         for (idx=1; idx<NPROTO; idx++) {
613                 int type = cb->nlh->nlmsg_type-RTM_BASE;
614                 if (idx < s_idx || idx == PF_PACKET)
615                         continue;
616                 if (rtnetlink_links[idx] == NULL ||
617                     rtnetlink_links[idx][type].dumpit == NULL)
618                         continue;
619                 if (idx > s_idx)
620                         memset(&cb->args[0], 0, sizeof(cb->args));
621                 if (rtnetlink_links[idx][type].dumpit(skb, cb))
622                         break;
623         }
624         cb->family = idx;
625
626         return skb->len;
627 }
628
629 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
630 {
631         struct sk_buff *skb;
632         int err = -ENOBUFS;
633
634         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
635         if (skb == NULL)
636                 goto errout;
637
638         err = rtnl_fill_ifinfo(skb, dev, NULL, 0, type, 0, 0, change, 0);
639         if (err < 0) {
640                 kfree_skb(skb);
641                 goto errout;
642         }
643
644         err = rtnl_notify(skb, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
645 errout:
646         if (err < 0)
647                 rtnl_set_sk_err(RTNLGRP_LINK, err);
648 }
649
650 /* Protected by RTNL sempahore.  */
651 static struct rtattr **rta_buf;
652 static int rtattr_max;
653
654 /* Process one rtnetlink message. */
655
656 static __inline__ int
657 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
658 {
659         struct rtnetlink_link *link;
660         struct rtnetlink_link *link_tab;
661         int sz_idx, kind;
662         int min_len;
663         int family;
664         int type;
665         int err;
666
667         /* Only requests are handled by kernel now */
668         if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
669                 return 0;
670
671         type = nlh->nlmsg_type;
672
673         /* A control message: ignore them */
674         if (type < RTM_BASE)
675                 return 0;
676
677         /* Unknown message: reply with EINVAL */
678         if (type > RTM_MAX)
679                 goto err_inval;
680
681         type -= RTM_BASE;
682
683         /* All the messages must have at least 1 byte length */
684         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
685                 return 0;
686
687         family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
688         if (family >= NPROTO) {
689                 *errp = -EAFNOSUPPORT;
690                 return -1;
691         }
692
693         link_tab = rtnetlink_links[family];
694         if (link_tab == NULL)
695                 link_tab = rtnetlink_links[PF_UNSPEC];
696         link = &link_tab[type];
697
698         sz_idx = type>>2;
699         kind = type&3;
700
701         if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN)) {
702                 *errp = -EPERM;
703                 return -1;
704         }
705
706         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
707                 if (link->dumpit == NULL)
708                         link = &(rtnetlink_links[PF_UNSPEC][type]);
709
710                 if (link->dumpit == NULL)
711                         goto err_inval;
712
713                 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
714                                                 link->dumpit, NULL)) != 0) {
715                         return -1;
716                 }
717
718                 netlink_queue_skip(nlh, skb);
719                 return -1;
720         }
721
722         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
723
724         min_len = rtm_min[sz_idx];
725         if (nlh->nlmsg_len < min_len)
726                 goto err_inval;
727
728         if (nlh->nlmsg_len > min_len) {
729                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
730                 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
731
732                 while (RTA_OK(attr, attrlen)) {
733                         unsigned flavor = attr->rta_type;
734                         if (flavor) {
735                                 if (flavor > rta_max[sz_idx])
736                                         goto err_inval;
737                                 rta_buf[flavor-1] = attr;
738                         }
739                         attr = RTA_NEXT(attr, attrlen);
740                 }
741         }
742
743         if (link->doit == NULL)
744                 link = &(rtnetlink_links[PF_UNSPEC][type]);
745         if (link->doit == NULL)
746                 goto err_inval;
747         err = link->doit(skb, nlh, (void *)&rta_buf[0]);
748
749         *errp = err;
750         return err;
751
752 err_inval:
753         *errp = -EINVAL;
754         return -1;
755 }
756
757 static void rtnetlink_rcv(struct sock *sk, int len)
758 {
759         unsigned int qlen = 0;
760
761         do {
762                 mutex_lock(&rtnl_mutex);
763                 netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
764                 mutex_unlock(&rtnl_mutex);
765
766                 netdev_run_todo();
767         } while (qlen);
768 }
769
770 static struct rtnetlink_link link_rtnetlink_table[RTM_NR_MSGTYPES] =
771 {
772         [RTM_GETLINK     - RTM_BASE] = { .doit   = rtnl_getlink,
773                                          .dumpit = rtnl_dump_ifinfo      },
774         [RTM_SETLINK     - RTM_BASE] = { .doit   = rtnl_setlink          },
775         [RTM_GETADDR     - RTM_BASE] = { .dumpit = rtnl_dump_all         },
776         [RTM_GETROUTE    - RTM_BASE] = { .dumpit = rtnl_dump_all         },
777         [RTM_NEWNEIGH    - RTM_BASE] = { .doit   = neigh_add             },
778         [RTM_DELNEIGH    - RTM_BASE] = { .doit   = neigh_delete          },
779         [RTM_GETNEIGH    - RTM_BASE] = { .dumpit = neigh_dump_info       },
780 #ifdef CONFIG_FIB_RULES
781         [RTM_NEWRULE     - RTM_BASE] = { .doit   = fib_nl_newrule        },
782         [RTM_DELRULE     - RTM_BASE] = { .doit   = fib_nl_delrule        },
783 #endif
784         [RTM_GETRULE     - RTM_BASE] = { .dumpit = rtnl_dump_all         },
785         [RTM_GETNEIGHTBL - RTM_BASE] = { .dumpit = neightbl_dump_info    },
786         [RTM_SETNEIGHTBL - RTM_BASE] = { .doit   = neightbl_set          },
787 };
788
789 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
790 {
791         struct net_device *dev = ptr;
792         switch (event) {
793         case NETDEV_UNREGISTER:
794                 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
795                 break;
796         case NETDEV_REGISTER:
797                 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
798                 break;
799         case NETDEV_UP:
800         case NETDEV_DOWN:
801                 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
802                 break;
803         case NETDEV_CHANGE:
804         case NETDEV_GOING_DOWN:
805                 break;
806         default:
807                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
808                 break;
809         }
810         return NOTIFY_DONE;
811 }
812
813 static struct notifier_block rtnetlink_dev_notifier = {
814         .notifier_call  = rtnetlink_event,
815 };
816
817 void __init rtnetlink_init(void)
818 {
819         int i;
820
821         rtattr_max = 0;
822         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
823                 if (rta_max[i] > rtattr_max)
824                         rtattr_max = rta_max[i];
825         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
826         if (!rta_buf)
827                 panic("rtnetlink_init: cannot allocate rta_buf\n");
828
829         rtnl = netlink_kernel_create(NETLINK_ROUTE, RTNLGRP_MAX, rtnetlink_rcv,
830                                      THIS_MODULE);
831         if (rtnl == NULL)
832                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
833         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
834         register_netdevice_notifier(&rtnetlink_dev_notifier);
835         rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
836         rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
837 }
838
839 EXPORT_SYMBOL(__rta_fill);
840 EXPORT_SYMBOL(rtattr_strlcpy);
841 EXPORT_SYMBOL(rtattr_parse);
842 EXPORT_SYMBOL(rtnetlink_links);
843 EXPORT_SYMBOL(rtnetlink_put_metrics);
844 EXPORT_SYMBOL(rtnl_lock);
845 EXPORT_SYMBOL(rtnl_trylock);
846 EXPORT_SYMBOL(rtnl_unlock);
847 EXPORT_SYMBOL(rtnl_unicast);
848 EXPORT_SYMBOL(rtnl_notify);
849 EXPORT_SYMBOL(rtnl_set_sk_err);