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