VServer 1.9.2 (patch-2.6.8.1-vs1.9.2.diff)
[linux-2.6.git] / 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/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/major.h>
26 #include <linux/sched.h>
27 #include <linux/timer.h>
28 #include <linux/string.h>
29 #include <linux/sockios.h>
30 #include <linux/net.h>
31 #include <linux/fcntl.h>
32 #include <linux/mm.h>
33 #include <linux/slab.h>
34 #include <linux/interrupt.h>
35 #include <linux/capability.h>
36 #include <linux/skbuff.h>
37 #include <linux/init.h>
38 #include <linux/security.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
54 DECLARE_MUTEX(rtnl_sem);
55
56 void rtnl_lock(void)
57 {
58         rtnl_shlock();
59 }
60  
61 void rtnl_unlock(void)
62 {
63         rtnl_shunlock();
64
65         netdev_run_todo();
66 }
67
68 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
69 {
70         memset(tb, 0, sizeof(struct rtattr*)*maxattr);
71
72         while (RTA_OK(rta, len)) {
73                 unsigned flavor = rta->rta_type;
74                 if (flavor && flavor <= maxattr)
75                         tb[flavor-1] = rta;
76                 rta = RTA_NEXT(rta, len);
77         }
78         return 0;
79 }
80
81 struct sock *rtnl;
82
83 struct rtnetlink_link * rtnetlink_links[NPROTO];
84
85 static const int rtm_min[(RTM_MAX+1-RTM_BASE)/4] =
86 {
87         NLMSG_LENGTH(sizeof(struct ifinfomsg)),
88         NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
89         NLMSG_LENGTH(sizeof(struct rtmsg)),
90         NLMSG_LENGTH(sizeof(struct ndmsg)),
91         NLMSG_LENGTH(sizeof(struct rtmsg)),
92         NLMSG_LENGTH(sizeof(struct tcmsg)),
93         NLMSG_LENGTH(sizeof(struct tcmsg)),
94         NLMSG_LENGTH(sizeof(struct tcmsg)),
95         NLMSG_LENGTH(sizeof(struct tcamsg))
96 };
97
98 static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
99 {
100         IFLA_MAX,
101         IFA_MAX,
102         RTA_MAX,
103         NDA_MAX,
104         RTA_MAX,
105         TCA_MAX,
106         TCA_MAX,
107         TCA_MAX,
108         TCAA_MAX
109 };
110
111 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
112 {
113         struct rtattr *rta;
114         int size = RTA_LENGTH(attrlen);
115
116         rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
117         rta->rta_type = attrtype;
118         rta->rta_len = size;
119         memcpy(RTA_DATA(rta), data, attrlen);
120 }
121
122 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
123 {
124         int err = 0;
125
126         NETLINK_CB(skb).dst_groups = group;
127         if (echo)
128                 atomic_inc(&skb->users);
129         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
130         if (echo)
131                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
132         return err;
133 }
134
135 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
136 {
137         struct rtattr *mx = (struct rtattr*)skb->tail;
138         int i;
139
140         RTA_PUT(skb, RTA_METRICS, 0, NULL);
141         for (i=0; i<RTAX_MAX; i++) {
142                 if (metrics[i])
143                         RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
144         }
145         mx->rta_len = skb->tail - (u8*)mx;
146         if (mx->rta_len == RTA_LENGTH(0))
147                 skb_trim(skb, (u8*)mx - skb->data);
148         return 0;
149
150 rtattr_failure:
151         skb_trim(skb, (u8*)mx - skb->data);
152         return -1;
153 }
154
155
156 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
157                                  int type, u32 pid, u32 seq, u32 change)
158 {
159         struct ifinfomsg *r;
160         struct nlmsghdr  *nlh;
161         unsigned char    *b = skb->tail;
162
163         nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
164         if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
165         r = NLMSG_DATA(nlh);
166         r->ifi_family = AF_UNSPEC;
167         r->ifi_type = dev->type;
168         r->ifi_index = dev->ifindex;
169         r->ifi_flags = dev->flags;
170         r->ifi_change = change;
171
172         if (!netif_running(dev) || !netif_carrier_ok(dev))
173                 r->ifi_flags &= ~IFF_RUNNING;
174         else
175                 r->ifi_flags |= IFF_RUNNING;
176
177         RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
178         if (dev->addr_len) {
179                 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
180                 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
181         }
182         if (1) {
183                 unsigned mtu = dev->mtu;
184                 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
185         }
186         if (dev->ifindex != dev->iflink)
187                 RTA_PUT(skb, IFLA_LINK, sizeof(int), &dev->iflink);
188         if (dev->qdisc_sleeping)
189                 RTA_PUT(skb, IFLA_QDISC,
190                         strlen(dev->qdisc_sleeping->ops->id) + 1,
191                         dev->qdisc_sleeping->ops->id);
192         if (dev->master)
193                 RTA_PUT(skb, IFLA_MASTER, sizeof(int), &dev->master->ifindex);
194         if (dev->get_stats) {
195                 unsigned long *stats = (unsigned long*)dev->get_stats(dev);
196                 if (stats) {
197                         struct rtattr  *a;
198                         __u32          *s;
199                         int             i;
200                         int             n = sizeof(struct rtnl_link_stats)/4;
201
202                         a = __RTA_PUT(skb, IFLA_STATS, n*4);
203                         s = RTA_DATA(a);
204                         for (i=0; i<n; i++)
205                                 s[i] = stats[i];
206                 }
207         }
208         nlh->nlmsg_len = skb->tail - b;
209         return skb->len;
210
211 nlmsg_failure:
212 rtattr_failure:
213         skb_trim(skb, b - skb->data);
214         return -1;
215 }
216
217 int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
218 {
219         int idx;
220         int s_idx = cb->args[0];
221         struct net_device *dev;
222
223         read_lock(&dev_base_lock);
224         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
225                 if (idx < s_idx)
226                         continue;
227                 if (!dev_in_nx_info(dev, current->nx_info))
228                         continue;
229                 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
230                         break;
231         }
232         read_unlock(&dev_base_lock);
233         cb->args[0] = idx;
234
235         return skb->len;
236 }
237
238 static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
239 {
240         struct ifinfomsg  *ifm = NLMSG_DATA(nlh);
241         struct rtattr    **ida = arg;
242         struct net_device *dev;
243         int err;
244
245         dev = dev_get_by_index(ifm->ifi_index);
246         if (!dev)
247                 return -ENODEV;
248
249         err = -EINVAL;
250
251         if (ida[IFLA_ADDRESS - 1]) {
252                 if (!dev->set_mac_address) {
253                         err = -EOPNOTSUPP;
254                         goto out;
255                 }
256                 if (!netif_device_present(dev)) {
257                         err = -ENODEV;
258                         goto out;
259                 }
260                 if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
261                         goto out;
262
263                 err = dev->set_mac_address(dev, RTA_DATA(ida[IFLA_ADDRESS - 1]));
264                 if (err)
265                         goto out;
266         }
267
268         if (ida[IFLA_BROADCAST - 1]) {
269                 if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
270                         goto out;
271                 memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
272                        dev->addr_len);
273         }
274
275         err = 0;
276
277 out:
278         if (!err)
279                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
280
281         dev_put(dev);
282         return err;
283 }
284
285 static int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
286 {
287         int idx;
288         int s_idx = cb->family;
289
290         if (s_idx == 0)
291                 s_idx = 1;
292         for (idx=1; idx<NPROTO; idx++) {
293                 int type = cb->nlh->nlmsg_type-RTM_BASE;
294                 if (idx < s_idx || idx == PF_PACKET)
295                         continue;
296                 if (rtnetlink_links[idx] == NULL ||
297                     rtnetlink_links[idx][type].dumpit == NULL)
298                         continue;
299                 if (idx > s_idx)
300                         memset(&cb->args[0], 0, sizeof(cb->args));
301                 if (rtnetlink_links[idx][type].dumpit(skb, cb))
302                         break;
303         }
304         cb->family = idx;
305
306         return skb->len;
307 }
308
309 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
310 {
311         struct sk_buff *skb;
312         int size = NLMSG_GOODSIZE;
313
314         if (!dev_in_nx_info(dev, current->nx_info))
315                 return;
316         skb = alloc_skb(size, GFP_KERNEL);
317         if (!skb)
318                 return;
319
320         if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change) < 0) {
321                 kfree_skb(skb);
322                 return;
323         }
324         NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
325         netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
326 }
327
328 static int rtnetlink_done(struct netlink_callback *cb)
329 {
330         return 0;
331 }
332
333 /* Process one rtnetlink message. */
334
335 static __inline__ int
336 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
337 {
338         struct rtnetlink_link *link;
339         struct rtnetlink_link *link_tab;
340         struct rtattr   *rta[RTATTR_MAX];
341
342         int sz_idx, kind;
343         int min_len;
344         int family;
345         int type;
346         int err;
347
348         /* Only requests are handled by kernel now */
349         if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
350                 return 0;
351
352         type = nlh->nlmsg_type;
353
354         /* A control message: ignore them */
355         if (type < RTM_BASE)
356                 return 0;
357
358         /* Unknown message: reply with EINVAL */
359         if (type > RTM_MAX)
360                 goto err_inval;
361
362         type -= RTM_BASE;
363
364         /* All the messages must have at least 1 byte length */
365         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
366                 return 0;
367
368         family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
369         if (family >= NPROTO) {
370                 *errp = -EAFNOSUPPORT;
371                 return -1;
372         }
373
374         link_tab = rtnetlink_links[family];
375         if (link_tab == NULL)
376                 link_tab = rtnetlink_links[PF_UNSPEC];
377         link = &link_tab[type];
378
379         sz_idx = type>>2;
380         kind = type&3;
381
382         if (kind != 2 && security_netlink_recv(skb)) {
383                 *errp = -EPERM;
384                 return -1;
385         }
386
387         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
388                 u32 rlen;
389
390                 if (link->dumpit == NULL)
391                         link = &(rtnetlink_links[PF_UNSPEC][type]);
392
393                 if (link->dumpit == NULL)
394                         goto err_inval;
395
396                 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
397                                                 link->dumpit,
398                                                 rtnetlink_done)) != 0) {
399                         return -1;
400                 }
401                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
402                 if (rlen > skb->len)
403                         rlen = skb->len;
404                 skb_pull(skb, rlen);
405                 return -1;
406         }
407
408         memset(&rta, 0, sizeof(rta));
409
410         min_len = rtm_min[sz_idx];
411         if (nlh->nlmsg_len < min_len)
412                 goto err_inval;
413
414         if (nlh->nlmsg_len > min_len) {
415                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
416                 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
417
418                 while (RTA_OK(attr, attrlen)) {
419                         unsigned flavor = attr->rta_type;
420                         if (flavor) {
421                                 if (flavor > rta_max[sz_idx])
422                                         goto err_inval;
423                                 rta[flavor-1] = attr;
424                         }
425                         attr = RTA_NEXT(attr, attrlen);
426                 }
427         }
428
429         if (link->doit == NULL)
430                 link = &(rtnetlink_links[PF_UNSPEC][type]);
431         if (link->doit == NULL)
432                 goto err_inval;
433         err = link->doit(skb, nlh, (void *)&rta);
434
435         *errp = err;
436         return err;
437
438 err_inval:
439         *errp = -EINVAL;
440         return -1;
441 }
442
443 /* 
444  * Process one packet of messages.
445  * Malformed skbs with wrong lengths of messages are discarded silently.
446  */
447
448 static inline int rtnetlink_rcv_skb(struct sk_buff *skb)
449 {
450         int err;
451         struct nlmsghdr * nlh;
452
453         while (skb->len >= NLMSG_SPACE(0)) {
454                 u32 rlen;
455
456                 nlh = (struct nlmsghdr *)skb->data;
457                 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
458                         return 0;
459                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
460                 if (rlen > skb->len)
461                         rlen = skb->len;
462                 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
463                         /* Not error, but we must interrupt processing here:
464                          *   Note, that in this case we do not pull message
465                          *   from skb, it will be processed later.
466                          */
467                         if (err == 0)
468                                 return -1;
469                         netlink_ack(skb, nlh, err);
470                 } else if (nlh->nlmsg_flags&NLM_F_ACK)
471                         netlink_ack(skb, nlh, 0);
472                 skb_pull(skb, rlen);
473         }
474
475         return 0;
476 }
477
478 /*
479  *  rtnetlink input queue processing routine:
480  *      - try to acquire shared lock. If it is failed, defer processing.
481  *      - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
482  *        that will occur, when a dump started and/or acquisition of
483  *        exclusive lock failed.
484  */
485
486 static void rtnetlink_rcv(struct sock *sk, int len)
487 {
488         do {
489                 struct sk_buff *skb;
490
491                 if (rtnl_shlock_nowait())
492                         return;
493
494                 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
495                         if (rtnetlink_rcv_skb(skb)) {
496                                 if (skb->len)
497                                         skb_queue_head(&sk->sk_receive_queue,
498                                                        skb);
499                                 else
500                                         kfree_skb(skb);
501                                 break;
502                         }
503                         kfree_skb(skb);
504                 }
505
506                 up(&rtnl_sem);
507
508                 netdev_run_todo();
509         } while (rtnl && rtnl->sk_receive_queue.qlen);
510 }
511
512 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
513 {
514         [RTM_GETLINK  - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
515         [RTM_SETLINK  - RTM_BASE] = { .doit   = do_setlink            },
516         [RTM_GETADDR  - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
517         [RTM_GETROUTE - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
518         [RTM_NEWNEIGH - RTM_BASE] = { .doit   = neigh_add             },
519         [RTM_DELNEIGH - RTM_BASE] = { .doit   = neigh_delete          },
520         [RTM_GETNEIGH - RTM_BASE] = { .dumpit = neigh_dump_info       }
521 };
522
523 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
524 {
525         struct net_device *dev = ptr;
526         switch (event) {
527         case NETDEV_UNREGISTER:
528                 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
529                 break;
530         case NETDEV_REGISTER:
531                 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
532                 break;
533         case NETDEV_UP:
534         case NETDEV_DOWN:
535                 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
536                 break;
537         case NETDEV_CHANGE:
538         case NETDEV_GOING_DOWN:
539                 break;
540         default:
541                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
542                 break;
543         }
544         return NOTIFY_DONE;
545 }
546
547 static struct notifier_block rtnetlink_dev_notifier = {
548         .notifier_call  = rtnetlink_event,
549 };
550
551 void __init rtnetlink_init(void)
552 {
553         rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
554         if (rtnl == NULL)
555                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
556         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
557         register_netdevice_notifier(&rtnetlink_dev_notifier);
558         rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
559         rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
560 }
561
562 EXPORT_SYMBOL(__rta_fill);
563 EXPORT_SYMBOL(rtattr_parse);
564 EXPORT_SYMBOL(rtnetlink_dump_ifinfo);
565 EXPORT_SYMBOL(rtnetlink_links);
566 EXPORT_SYMBOL(rtnetlink_put_metrics);
567 EXPORT_SYMBOL(rtnl);
568 EXPORT_SYMBOL(rtnl_lock);
569 EXPORT_SYMBOL(rtnl_sem);
570 EXPORT_SYMBOL(rtnl_unlock);