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.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
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.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
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>
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>
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
42 #include <asm/string.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
54 DECLARE_MUTEX(rtnl_sem);
62 void rtnl_unlock(void)
70 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
72 memset(tb, 0, sizeof(struct rtattr*)*maxattr);
74 while (RTA_OK(rta, len)) {
75 unsigned flavor = rta->rta_type;
76 if (flavor && flavor <= maxattr)
78 rta = RTA_NEXT(rta, len);
85 struct rtnetlink_link * rtnetlink_links[NPROTO];
87 static const int rtm_min[(RTM_MAX+1-RTM_BASE)/4] =
89 NLMSG_LENGTH(sizeof(struct ifinfomsg)),
90 NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
91 NLMSG_LENGTH(sizeof(struct rtmsg)),
92 NLMSG_LENGTH(sizeof(struct ndmsg)),
93 NLMSG_LENGTH(sizeof(struct rtmsg)),
94 NLMSG_LENGTH(sizeof(struct tcmsg)),
95 NLMSG_LENGTH(sizeof(struct tcmsg)),
96 NLMSG_LENGTH(sizeof(struct tcmsg)),
97 NLMSG_LENGTH(sizeof(struct tcamsg))
100 static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
113 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
116 int size = RTA_LENGTH(attrlen);
118 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
119 rta->rta_type = attrtype;
121 memcpy(RTA_DATA(rta), data, attrlen);
124 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
128 NETLINK_CB(skb).dst_groups = group;
130 atomic_inc(&skb->users);
131 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
133 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
137 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
139 struct rtattr *mx = (struct rtattr*)skb->tail;
142 RTA_PUT(skb, RTA_METRICS, 0, NULL);
143 for (i=0; i<RTAX_MAX; i++) {
145 RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
147 mx->rta_len = skb->tail - (u8*)mx;
148 if (mx->rta_len == RTA_LENGTH(0))
149 skb_trim(skb, (u8*)mx - skb->data);
153 skb_trim(skb, (u8*)mx - skb->data);
158 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
159 int type, u32 pid, u32 seq, u32 change)
162 struct nlmsghdr *nlh;
163 unsigned char *b = skb->tail;
165 nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
166 if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
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;
174 if (!netif_running(dev) || !netif_carrier_ok(dev))
175 r->ifi_flags &= ~IFF_RUNNING;
177 r->ifi_flags |= IFF_RUNNING;
179 RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
181 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
182 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
185 unsigned mtu = dev->mtu;
186 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
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);
195 RTA_PUT(skb, IFLA_MASTER, sizeof(int), &dev->master->ifindex);
196 if (dev->get_stats) {
197 unsigned long *stats = (unsigned long*)dev->get_stats(dev);
202 int n = sizeof(struct rtnl_link_stats)/4;
204 a = __RTA_PUT(skb, IFLA_STATS, n*4);
210 nlh->nlmsg_len = skb->tail - b;
215 skb_trim(skb, b - skb->data);
219 int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
222 int s_idx = cb->args[0];
223 struct net_device *dev;
225 read_lock(&dev_base_lock);
226 for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
229 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
232 read_unlock(&dev_base_lock);
238 static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
240 struct ifinfomsg *ifm = NLMSG_DATA(nlh);
241 struct rtattr **ida = arg;
242 struct net_device *dev;
245 dev = dev_get_by_index(ifm->ifi_index);
251 if (ida[IFLA_ADDRESS - 1]) {
252 if (!dev->set_mac_address) {
256 if (!netif_device_present(dev)) {
260 if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
263 err = dev->set_mac_address(dev, RTA_DATA(ida[IFLA_ADDRESS - 1]));
268 if (ida[IFLA_BROADCAST - 1]) {
269 if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
271 memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
279 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
285 static int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
288 int s_idx = cb->family;
292 for (idx=1; idx<NPROTO; idx++) {
293 int type = cb->nlh->nlmsg_type-RTM_BASE;
294 if (idx < s_idx || idx == PF_PACKET)
296 if (rtnetlink_links[idx] == NULL ||
297 rtnetlink_links[idx][type].dumpit == NULL)
300 memset(&cb->args[0], 0, sizeof(cb->args));
301 if (rtnetlink_links[idx][type].dumpit(skb, cb))
309 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
312 int size = NLMSG_GOODSIZE;
314 skb = alloc_skb(size, GFP_KERNEL);
318 if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change) < 0) {
322 NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
323 netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
326 static int rtnetlink_done(struct netlink_callback *cb)
331 /* Process one rtnetlink message. */
333 static __inline__ int
334 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
336 struct rtnetlink_link *link;
337 struct rtnetlink_link *link_tab;
338 struct rtattr *rta[RTATTR_MAX];
347 /* Only requests are handled by kernel now */
348 if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
351 type = nlh->nlmsg_type;
353 /* A control message: ignore them */
357 /* Unknown message: reply with EINVAL */
363 /* All the messages must have at least 1 byte length */
364 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
367 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
368 if (family >= NPROTO) {
369 *errp = -EAFNOSUPPORT;
373 link_tab = rtnetlink_links[family];
374 if (link_tab == NULL)
375 link_tab = rtnetlink_links[PF_UNSPEC];
376 link = &link_tab[type];
381 if (kind != 2 && security_netlink_recv(skb)) {
386 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
389 if (link->dumpit == NULL)
390 link = &(rtnetlink_links[PF_UNSPEC][type]);
392 if (link->dumpit == NULL)
395 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
397 rtnetlink_done)) != 0) {
400 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
408 if (rtnl_exlock_nowait()) {
415 memset(&rta, 0, sizeof(rta));
417 min_len = rtm_min[sz_idx];
418 if (nlh->nlmsg_len < min_len)
421 if (nlh->nlmsg_len > min_len) {
422 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
423 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
425 while (RTA_OK(attr, attrlen)) {
426 unsigned flavor = attr->rta_type;
428 if (flavor > rta_max[sz_idx])
430 rta[flavor-1] = attr;
432 attr = RTA_NEXT(attr, attrlen);
436 if (link->doit == NULL)
437 link = &(rtnetlink_links[PF_UNSPEC][type]);
438 if (link->doit == NULL)
440 err = link->doit(skb, nlh, (void *)&rta);
455 * Process one packet of messages.
456 * Malformed skbs with wrong lengths of messages are discarded silently.
459 static inline int rtnetlink_rcv_skb(struct sk_buff *skb)
462 struct nlmsghdr * nlh;
464 while (skb->len >= NLMSG_SPACE(0)) {
467 nlh = (struct nlmsghdr *)skb->data;
468 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
470 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
473 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
474 /* Not error, but we must interrupt processing here:
475 * Note, that in this case we do not pull message
476 * from skb, it will be processed later.
480 netlink_ack(skb, nlh, err);
481 } else if (nlh->nlmsg_flags&NLM_F_ACK)
482 netlink_ack(skb, nlh, 0);
490 * rtnetlink input queue processing routine:
491 * - try to acquire shared lock. If it is failed, defer processing.
492 * - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
493 * that will occur, when a dump started and/or acquisition of
494 * exclusive lock failed.
497 static void rtnetlink_rcv(struct sock *sk, int len)
502 if (rtnl_shlock_nowait())
505 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
506 if (rtnetlink_rcv_skb(skb)) {
508 skb_queue_head(&sk->sk_receive_queue,
520 } while (rtnl && rtnl->sk_receive_queue.qlen);
523 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
525 [RTM_GETLINK - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
526 [RTM_SETLINK - RTM_BASE] = { .doit = do_setlink },
527 [RTM_GETADDR - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
528 [RTM_GETROUTE - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
529 [RTM_NEWNEIGH - RTM_BASE] = { .doit = neigh_add },
530 [RTM_DELNEIGH - RTM_BASE] = { .doit = neigh_delete },
531 [RTM_GETNEIGH - RTM_BASE] = { .dumpit = neigh_dump_info }
534 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
536 struct net_device *dev = ptr;
538 case NETDEV_UNREGISTER:
539 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
541 case NETDEV_REGISTER:
542 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
546 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
549 case NETDEV_GOING_DOWN:
552 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
558 static struct notifier_block rtnetlink_dev_notifier = {
559 .notifier_call = rtnetlink_event,
562 void __init rtnetlink_init(void)
564 rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
566 panic("rtnetlink_init: cannot initialize rtnetlink\n");
567 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
568 register_netdevice_notifier(&rtnetlink_dev_notifier);
569 rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
570 rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
573 EXPORT_SYMBOL(__rta_fill);
574 EXPORT_SYMBOL(rtattr_parse);
575 EXPORT_SYMBOL(rtnetlink_dump_ifinfo);
576 EXPORT_SYMBOL(rtnetlink_links);
577 EXPORT_SYMBOL(rtnetlink_put_metrics);
579 EXPORT_SYMBOL(rtnl_lock);
580 EXPORT_SYMBOL(rtnl_sem);
581 EXPORT_SYMBOL(rtnl_unlock);