File: /usr/src/linux/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/config.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/major.h>
25 #include <linux/sched.h>
26 #include <linux/timer.h>
27 #include <linux/string.h>
28 #include <linux/sockios.h>
29 #include <linux/net.h>
30 #include <linux/fcntl.h>
31 #include <linux/mm.h>
32 #include <linux/slab.h>
33 #include <linux/interrupt.h>
34 #include <linux/capability.h>
35 #include <linux/skbuff.h>
36 #include <linux/init.h>
37
38 #include <asm/uaccess.h>
39 #include <asm/system.h>
40 #include <asm/string.h>
41
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <net/ip.h>
45 #include <net/protocol.h>
46 #include <net/arp.h>
47 #include <net/route.h>
48 #include <net/tcp.h>
49 #include <net/udp.h>
50 #include <net/sock.h>
51 #include <net/pkt_sched.h>
52
53 DECLARE_MUTEX(rtnl_sem);
54
55 void rtnl_lock(void)
56 {
57 rtnl_shlock();
58 rtnl_exlock();
59 }
60
61 void rtnl_unlock(void)
62 {
63 rtnl_exunlock();
64 rtnl_shunlock();
65 }
66
67 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
68 {
69 memset(tb, 0, sizeof(struct rtattr*)*maxattr);
70
71 while (RTA_OK(rta, len)) {
72 unsigned flavor = rta->rta_type;
73 if (flavor && flavor <= maxattr)
74 tb[flavor-1] = rta;
75 rta = RTA_NEXT(rta, len);
76 }
77 return 0;
78 }
79
80 #ifdef CONFIG_RTNETLINK
81 struct sock *rtnl;
82
83 struct rtnetlink_link * rtnetlink_links[NPROTO];
84
85 #define _S 1 /* superuser privileges required */
86 #define _X 2 /* exclusive access to tables required */
87 #define _G 4 /* GET request */
88
89 static const int rtm_min[(RTM_MAX+1-RTM_BASE)/4] =
90 {
91 NLMSG_LENGTH(sizeof(struct ifinfomsg)),
92 NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
93 NLMSG_LENGTH(sizeof(struct rtmsg)),
94 NLMSG_LENGTH(sizeof(struct ndmsg)),
95 NLMSG_LENGTH(sizeof(struct rtmsg)),
96 NLMSG_LENGTH(sizeof(struct tcmsg)),
97 NLMSG_LENGTH(sizeof(struct tcmsg)),
98 NLMSG_LENGTH(sizeof(struct tcmsg))
99 };
100
101 static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
102 {
103 IFLA_MAX,
104 IFA_MAX,
105 RTA_MAX,
106 NDA_MAX,
107 RTA_MAX,
108 TCA_MAX,
109 TCA_MAX,
110 TCA_MAX
111 };
112
113 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
114 {
115 struct rtattr *rta;
116 int size = RTA_LENGTH(attrlen);
117
118 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
119 rta->rta_type = attrtype;
120 rta->rta_len = size;
121 memcpy(RTA_DATA(rta), data, attrlen);
122 }
123
124 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
125 {
126 int err = 0;
127
128 NETLINK_CB(skb).dst_groups = group;
129 if (echo)
130 atomic_inc(&skb->users);
131 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
132 if (echo)
133 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
134 return err;
135 }
136
137 int rtnetlink_put_metrics(struct sk_buff *skb, unsigned *metrics)
138 {
139 struct rtattr *mx = (struct rtattr*)skb->tail;
140 int i;
141
142 RTA_PUT(skb, RTA_METRICS, 0, NULL);
143 for (i=0; i<RTAX_MAX; i++) {
144 if (metrics[i])
145 RTA_PUT(skb, i+1, sizeof(unsigned), metrics+i);
146 }
147 mx->rta_len = skb->tail - (u8*)mx;
148 if (mx->rta_len == RTA_LENGTH(0))
149 skb_trim(skb, (u8*)mx - skb->data);
150 return 0;
151
152 rtattr_failure:
153 skb_trim(skb, (u8*)mx - skb->data);
154 return -1;
155 }
156
157
158 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
159 int type, u32 pid, u32 seq, u32 change)
160 {
161 struct ifinfomsg *r;
162 struct nlmsghdr *nlh;
163 unsigned char *b = skb->tail;
164
165 nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
166 if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
167 r = NLMSG_DATA(nlh);
168 r->ifi_family = AF_UNSPEC;
169 r->ifi_type = dev->type;
170 r->ifi_index = dev->ifindex;
171 r->ifi_flags = dev->flags;
172 r->ifi_change = change;
173
174 if (!netif_running(dev) || !netif_carrier_ok(dev))
175 r->ifi_flags &= ~IFF_RUNNING;
176 else
177 r->ifi_flags |= IFF_RUNNING;
178
179 RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
180 if (dev->addr_len) {
181 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
182 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
183 }
184 if (1) {
185 unsigned mtu = dev->mtu;
186 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
187 }
188 if (dev->ifindex != dev->iflink)
189 RTA_PUT(skb, IFLA_LINK, sizeof(int), &dev->iflink);
190 if (dev->qdisc_sleeping)
191 RTA_PUT(skb, IFLA_QDISC,
192 strlen(dev->qdisc_sleeping->ops->id) + 1,
193 dev->qdisc_sleeping->ops->id);
194 if (dev->master)
195 RTA_PUT(skb, IFLA_MASTER, sizeof(int), &dev->master->ifindex);
196 if (dev->get_stats) {
197 struct net_device_stats *stats = dev->get_stats(dev);
198 if (stats)
199 RTA_PUT(skb, IFLA_STATS, sizeof(*stats), stats);
200 }
201 nlh->nlmsg_len = skb->tail - b;
202 return skb->len;
203
204 nlmsg_failure:
205 rtattr_failure:
206 skb_trim(skb, b - skb->data);
207 return -1;
208 }
209
210 int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
211 {
212 int idx;
213 int s_idx = cb->args[0];
214 struct net_device *dev;
215
216 read_lock(&dev_base_lock);
217 for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
218 if (idx < s_idx)
219 continue;
220 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
221 break;
222 }
223 read_unlock(&dev_base_lock);
224 cb->args[0] = idx;
225
226 return skb->len;
227 }
228
229 int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
230 {
231 int idx;
232 int s_idx = cb->family;
233
234 if (s_idx == 0)
235 s_idx = 1;
236 for (idx=1; idx<NPROTO; idx++) {
237 int type = cb->nlh->nlmsg_type-RTM_BASE;
238 if (idx < s_idx || idx == PF_PACKET)
239 continue;
240 if (rtnetlink_links[idx] == NULL ||
241 rtnetlink_links[idx][type].dumpit == NULL)
242 continue;
243 if (idx > s_idx)
244 memset(&cb->args[0], 0, sizeof(cb->args));
245 if (rtnetlink_links[idx][type].dumpit(skb, cb))
246 break;
247 }
248 cb->family = idx;
249
250 return skb->len;
251 }
252
253 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
254 {
255 struct sk_buff *skb;
256 int size = NLMSG_GOODSIZE;
257
258 skb = alloc_skb(size, GFP_KERNEL);
259 if (!skb)
260 return;
261
262 if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change) < 0) {
263 kfree_skb(skb);
264 return;
265 }
266 NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
267 netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
268 }
269
270 static int rtnetlink_done(struct netlink_callback *cb)
271 {
272 return 0;
273 }
274
275 /* Process one rtnetlink message. */
276
277 static __inline__ int
278 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
279 {
280 struct rtnetlink_link *link;
281 struct rtnetlink_link *link_tab;
282 struct rtattr *rta[RTATTR_MAX];
283
284 int exclusive = 0;
285 int sz_idx, kind;
286 int min_len;
287 int family;
288 int type;
289 int err;
290
291 /* Only requests are handled by kernel now */
292 if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
293 return 0;
294
295 type = nlh->nlmsg_type;
296
297 /* A control message: ignore them */
298 if (type < RTM_BASE)
299 return 0;
300
301 /* Unknown message: reply with EINVAL */
302 if (type > RTM_MAX)
303 goto err_inval;
304
305 type -= RTM_BASE;
306
307 /* All the messages must have at least 1 byte length */
308 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
309 return 0;
310
311 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
312 if (family > NPROTO) {
313 *errp = -EAFNOSUPPORT;
314 return -1;
315 }
316
317 link_tab = rtnetlink_links[family];
318 if (link_tab == NULL)
319 link_tab = rtnetlink_links[PF_UNSPEC];
320 link = &link_tab[type];
321
322 sz_idx = type>>2;
323 kind = type&3;
324
325 if (kind != 2 && !cap_raised(NETLINK_CB(skb).eff_cap, CAP_NET_ADMIN)) {
326 *errp = -EPERM;
327 return -1;
328 }
329
330 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
331 u32 rlen;
332
333 if (link->dumpit == NULL)
334 link = &(rtnetlink_links[PF_UNSPEC][type]);
335
336 if (link->dumpit == NULL)
337 goto err_inval;
338
339 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
340 link->dumpit,
341 rtnetlink_done)) != 0) {
342 return -1;
343 }
344 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
345 if (rlen > skb->len)
346 rlen = skb->len;
347 skb_pull(skb, rlen);
348 return -1;
349 }
350
351 if (kind != 2) {
352 if (rtnl_exlock_nowait()) {
353 *errp = 0;
354 return -1;
355 }
356 exclusive = 1;
357 }
358
359 memset(&rta, 0, sizeof(rta));
360
361 min_len = rtm_min[sz_idx];
362 if (nlh->nlmsg_len < min_len)
363 goto err_inval;
364
365 if (nlh->nlmsg_len > min_len) {
366 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
367 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
368
369 while (RTA_OK(attr, attrlen)) {
370 unsigned flavor = attr->rta_type;
371 if (flavor) {
372 if (flavor > rta_max[sz_idx])
373 goto err_inval;
374 rta[flavor-1] = attr;
375 }
376 attr = RTA_NEXT(attr, attrlen);
377 }
378 }
379
380 if (link->doit == NULL)
381 link = &(rtnetlink_links[PF_UNSPEC][type]);
382 if (link->doit == NULL)
383 goto err_inval;
384 err = link->doit(skb, nlh, (void *)&rta);
385
386 if (exclusive)
387 rtnl_exunlock();
388 *errp = err;
389 return err;
390
391 err_inval:
392 if (exclusive)
393 rtnl_exunlock();
394 *errp = -EINVAL;
395 return -1;
396 }
397
398 /*
399 * Process one packet of messages.
400 * Malformed skbs with wrong lengths of messages are discarded silently.
401 */
402
403 extern __inline__ int rtnetlink_rcv_skb(struct sk_buff *skb)
404 {
405 int err;
406 struct nlmsghdr * nlh;
407
408 while (skb->len >= NLMSG_SPACE(0)) {
409 u32 rlen;
410
411 nlh = (struct nlmsghdr *)skb->data;
412 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
413 return 0;
414 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
415 if (rlen > skb->len)
416 rlen = skb->len;
417 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
418 /* Not error, but we must interrupt processing here:
419 * Note, that in this case we do not pull message
420 * from skb, it will be processed later.
421 */
422 if (err == 0)
423 return -1;
424 netlink_ack(skb, nlh, err);
425 } else if (nlh->nlmsg_flags&NLM_F_ACK)
426 netlink_ack(skb, nlh, 0);
427 skb_pull(skb, rlen);
428 }
429
430 return 0;
431 }
432
433 /*
434 * rtnetlink input queue processing routine:
435 * - try to acquire shared lock. If it is failed, defer processing.
436 * - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
437 * that will occur, when a dump started and/or acquisition of
438 * exclusive lock failed.
439 */
440
441 static void rtnetlink_rcv(struct sock *sk, int len)
442 {
443 do {
444 struct sk_buff *skb;
445
446 if (rtnl_shlock_nowait())
447 return;
448
449 while ((skb = skb_dequeue(&sk->receive_queue)) != NULL) {
450 if (rtnetlink_rcv_skb(skb)) {
451 if (skb->len)
452 skb_queue_head(&sk->receive_queue, skb);
453 else
454 kfree_skb(skb);
455 break;
456 }
457 kfree_skb(skb);
458 }
459
460 up(&rtnl_sem);
461 } while (rtnl && rtnl->receive_queue.qlen);
462 }
463
464 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
465 {
466 { NULL, NULL, },
467 { NULL, NULL, },
468 { NULL, rtnetlink_dump_ifinfo, },
469 { NULL, NULL, },
470
471 { NULL, NULL, },
472 { NULL, NULL, },
473 { NULL, rtnetlink_dump_all, },
474 { NULL, NULL, },
475
476 { NULL, NULL, },
477 { NULL, NULL, },
478 { NULL, rtnetlink_dump_all, },
479 { NULL, NULL, },
480
481 { neigh_add, NULL, },
482 { neigh_delete, NULL, },
483 { NULL, neigh_dump_info, },
484 { NULL, NULL, },
485
486 { NULL, NULL, },
487 { NULL, NULL, },
488 { NULL, NULL, },
489 { NULL, NULL, },
490 };
491
492
493 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
494 {
495 struct net_device *dev = ptr;
496 switch (event) {
497 case NETDEV_UNREGISTER:
498 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
499 break;
500 case NETDEV_REGISTER:
501 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
502 break;
503 case NETDEV_UP:
504 case NETDEV_DOWN:
505 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
506 break;
507 case NETDEV_CHANGE:
508 case NETDEV_GOING_DOWN:
509 break;
510 default:
511 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
512 break;
513 }
514 return NOTIFY_DONE;
515 }
516
517 struct notifier_block rtnetlink_dev_notifier = {
518 rtnetlink_event,
519 NULL,
520 0
521 };
522
523
524 void __init rtnetlink_init(void)
525 {
526 #ifdef RTNL_DEBUG
527 printk("Initializing RT netlink socket\n");
528 #endif
529 rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
530 if (rtnl == NULL)
531 panic("rtnetlink_init: cannot initialize rtnetlink\n");
532 register_netdevice_notifier(&rtnetlink_dev_notifier);
533 rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
534 rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
535 }
536
537
538
539 #endif
540