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))
99 static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
111 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
114 int size = RTA_LENGTH(attrlen);
116 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
117 rta->rta_type = attrtype;
119 memcpy(RTA_DATA(rta), data, attrlen);
122 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
126 NETLINK_CB(skb).dst_groups = group;
128 atomic_inc(&skb->users);
129 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
131 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
135 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
137 struct rtattr *mx = (struct rtattr*)skb->tail;
140 RTA_PUT(skb, RTA_METRICS, 0, NULL);
141 for (i=0; i<RTAX_MAX; i++) {
143 RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
145 mx->rta_len = skb->tail - (u8*)mx;
146 if (mx->rta_len == RTA_LENGTH(0))
147 skb_trim(skb, (u8*)mx - skb->data);
151 skb_trim(skb, (u8*)mx - skb->data);
156 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
157 int type, u32 pid, u32 seq, u32 change)
160 struct nlmsghdr *nlh;
161 unsigned char *b = skb->tail;
163 nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
164 if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
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;
172 if (!netif_running(dev) || !netif_carrier_ok(dev))
173 r->ifi_flags &= ~IFF_RUNNING;
175 r->ifi_flags |= IFF_RUNNING;
177 RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
179 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
180 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
183 unsigned mtu = dev->mtu;
184 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
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);
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);
200 int n = sizeof(struct rtnl_link_stats)/4;
202 a = __RTA_PUT(skb, IFLA_STATS, n*4);
208 nlh->nlmsg_len = skb->tail - b;
213 skb_trim(skb, b - skb->data);
217 int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
220 int s_idx = cb->args[0];
221 struct net_device *dev;
223 read_lock(&dev_base_lock);
224 for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
227 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
230 read_unlock(&dev_base_lock);
236 static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
238 struct ifinfomsg *ifm = NLMSG_DATA(nlh);
239 struct rtattr **ida = arg;
240 struct net_device *dev;
243 dev = dev_get_by_index(ifm->ifi_index);
249 if (ida[IFLA_ADDRESS - 1]) {
250 if (!dev->set_mac_address) {
254 if (!netif_device_present(dev)) {
258 if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
261 err = dev->set_mac_address(dev, RTA_DATA(ida[IFLA_ADDRESS - 1]));
266 if (ida[IFLA_BROADCAST - 1]) {
267 if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
269 memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
277 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
283 static int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
286 int s_idx = cb->family;
290 for (idx=1; idx<NPROTO; idx++) {
291 int type = cb->nlh->nlmsg_type-RTM_BASE;
292 if (idx < s_idx || idx == PF_PACKET)
294 if (rtnetlink_links[idx] == NULL ||
295 rtnetlink_links[idx][type].dumpit == NULL)
298 memset(&cb->args[0], 0, sizeof(cb->args));
299 if (rtnetlink_links[idx][type].dumpit(skb, cb))
307 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
310 int size = NLMSG_GOODSIZE;
312 skb = alloc_skb(size, GFP_KERNEL);
316 if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change) < 0) {
320 NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
321 netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
324 static int rtnetlink_done(struct netlink_callback *cb)
329 /* Process one rtnetlink message. */
331 static __inline__ int
332 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
334 struct rtnetlink_link *link;
335 struct rtnetlink_link *link_tab;
336 struct rtattr *rta[RTATTR_MAX];
345 /* Only requests are handled by kernel now */
346 if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
349 type = nlh->nlmsg_type;
351 /* A control message: ignore them */
355 /* Unknown message: reply with EINVAL */
361 /* All the messages must have at least 1 byte length */
362 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
365 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
366 if (family >= NPROTO) {
367 *errp = -EAFNOSUPPORT;
371 link_tab = rtnetlink_links[family];
372 if (link_tab == NULL)
373 link_tab = rtnetlink_links[PF_UNSPEC];
374 link = &link_tab[type];
379 if (kind != 2 && security_netlink_recv(skb)) {
384 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
387 if (link->dumpit == NULL)
388 link = &(rtnetlink_links[PF_UNSPEC][type]);
390 if (link->dumpit == NULL)
393 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
395 rtnetlink_done)) != 0) {
398 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
406 if (rtnl_exlock_nowait()) {
413 memset(&rta, 0, sizeof(rta));
415 min_len = rtm_min[sz_idx];
416 if (nlh->nlmsg_len < min_len)
419 if (nlh->nlmsg_len > min_len) {
420 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
421 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
423 while (RTA_OK(attr, attrlen)) {
424 unsigned flavor = attr->rta_type;
426 if (flavor > rta_max[sz_idx])
428 rta[flavor-1] = attr;
430 attr = RTA_NEXT(attr, attrlen);
434 if (link->doit == NULL)
435 link = &(rtnetlink_links[PF_UNSPEC][type]);
436 if (link->doit == NULL)
438 err = link->doit(skb, nlh, (void *)&rta);
453 * Process one packet of messages.
454 * Malformed skbs with wrong lengths of messages are discarded silently.
457 static inline int rtnetlink_rcv_skb(struct sk_buff *skb)
460 struct nlmsghdr * nlh;
462 while (skb->len >= NLMSG_SPACE(0)) {
465 nlh = (struct nlmsghdr *)skb->data;
466 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
468 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
471 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
472 /* Not error, but we must interrupt processing here:
473 * Note, that in this case we do not pull message
474 * from skb, it will be processed later.
478 netlink_ack(skb, nlh, err);
479 } else if (nlh->nlmsg_flags&NLM_F_ACK)
480 netlink_ack(skb, nlh, 0);
488 * rtnetlink input queue processing routine:
489 * - try to acquire shared lock. If it is failed, defer processing.
490 * - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
491 * that will occur, when a dump started and/or acquisition of
492 * exclusive lock failed.
495 static void rtnetlink_rcv(struct sock *sk, int len)
500 if (rtnl_shlock_nowait())
503 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
504 if (rtnetlink_rcv_skb(skb)) {
506 skb_queue_head(&sk->sk_receive_queue,
518 } while (rtnl && rtnl->sk_receive_queue.qlen);
521 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
523 [RTM_GETLINK - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
524 [RTM_SETLINK - RTM_BASE] = { .doit = do_setlink },
525 [RTM_GETADDR - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
526 [RTM_GETROUTE - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
527 [RTM_NEWNEIGH - RTM_BASE] = { .doit = neigh_add },
528 [RTM_DELNEIGH - RTM_BASE] = { .doit = neigh_delete },
529 [RTM_GETNEIGH - RTM_BASE] = { .dumpit = neigh_dump_info }
532 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
534 struct net_device *dev = ptr;
536 case NETDEV_UNREGISTER:
537 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
539 case NETDEV_REGISTER:
540 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
544 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
547 case NETDEV_GOING_DOWN:
550 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
556 static struct notifier_block rtnetlink_dev_notifier = {
557 .notifier_call = rtnetlink_event,
560 void __init rtnetlink_init(void)
562 rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
564 panic("rtnetlink_init: cannot initialize rtnetlink\n");
565 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
566 register_netdevice_notifier(&rtnetlink_dev_notifier);
567 rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
568 rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
571 EXPORT_SYMBOL(__rta_fill);
572 EXPORT_SYMBOL(rtattr_parse);
573 EXPORT_SYMBOL(rtnetlink_dump_ifinfo);
574 EXPORT_SYMBOL(rtnetlink_links);
575 EXPORT_SYMBOL(rtnetlink_put_metrics);
577 EXPORT_SYMBOL(rtnl_lock);
578 EXPORT_SYMBOL(rtnl_sem);
579 EXPORT_SYMBOL(rtnl_unlock);