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/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>
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 #include <linux/security.h>
39 #include <asm/uaccess.h>
40 #include <asm/system.h>
41 #include <asm/string.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
46 #include <net/protocol.h>
48 #include <net/route.h>
51 #include <net/pkt_sched.h>
53 DECLARE_MUTEX(rtnl_sem);
60 int rtnl_lock_interruptible(void)
62 return down_interruptible(&rtnl_sem);
65 void rtnl_unlock(void)
72 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
74 memset(tb, 0, sizeof(struct rtattr*)*maxattr);
76 while (RTA_OK(rta, len)) {
77 unsigned flavor = rta->rta_type;
78 if (flavor && flavor <= maxattr)
80 rta = RTA_NEXT(rta, len);
87 struct rtnetlink_link * rtnetlink_links[NPROTO];
89 static const int rtm_min[(RTM_MAX+1-RTM_BASE)/4] =
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 NLMSG_LENGTH(sizeof(struct tcamsg))
102 static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
115 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
118 int size = RTA_LENGTH(attrlen);
120 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
121 rta->rta_type = attrtype;
123 memcpy(RTA_DATA(rta), data, attrlen);
126 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
128 size_t ret = RTA_PAYLOAD(rta);
129 char *src = RTA_DATA(rta);
131 if (ret > 0 && src[ret - 1] == '\0')
134 size_t len = (ret >= size) ? size - 1 : ret;
135 memset(dest, 0, size);
136 memcpy(dest, src, len);
141 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
145 NETLINK_CB(skb).dst_groups = group;
147 atomic_inc(&skb->users);
148 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
150 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
154 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
156 struct rtattr *mx = (struct rtattr*)skb->tail;
159 RTA_PUT(skb, RTA_METRICS, 0, NULL);
160 for (i=0; i<RTAX_MAX; i++) {
162 RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
164 mx->rta_len = skb->tail - (u8*)mx;
165 if (mx->rta_len == RTA_LENGTH(0))
166 skb_trim(skb, (u8*)mx - skb->data);
170 skb_trim(skb, (u8*)mx - skb->data);
175 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
176 int type, u32 pid, u32 seq, u32 change)
179 struct nlmsghdr *nlh;
180 unsigned char *b = skb->tail;
182 nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
183 if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
185 r->ifi_family = AF_UNSPEC;
186 r->ifi_type = dev->type;
187 r->ifi_index = dev->ifindex;
188 r->ifi_flags = dev_get_flags(dev);
189 r->ifi_change = change;
191 RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
194 u32 txqlen = dev->tx_queue_len;
195 RTA_PUT(skb, IFLA_TXQLEN, sizeof(txqlen), &txqlen);
199 u32 weight = dev->weight;
200 RTA_PUT(skb, IFLA_WEIGHT, sizeof(weight), &weight);
204 struct rtnl_link_ifmap map = {
205 .mem_start = dev->mem_start,
206 .mem_end = dev->mem_end,
207 .base_addr = dev->base_addr,
210 .port = dev->if_port,
212 RTA_PUT(skb, IFLA_MAP, sizeof(map), &map);
216 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
217 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
222 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
225 if (dev->ifindex != dev->iflink) {
226 u32 iflink = dev->iflink;
227 RTA_PUT(skb, IFLA_LINK, sizeof(iflink), &iflink);
230 if (dev->qdisc_sleeping)
231 RTA_PUT(skb, IFLA_QDISC,
232 strlen(dev->qdisc_sleeping->ops->id) + 1,
233 dev->qdisc_sleeping->ops->id);
236 u32 master = dev->master->ifindex;
237 RTA_PUT(skb, IFLA_MASTER, sizeof(master), &master);
240 if (dev->get_stats) {
241 unsigned long *stats = (unsigned long*)dev->get_stats(dev);
246 int n = sizeof(struct rtnl_link_stats)/4;
248 a = __RTA_PUT(skb, IFLA_STATS, n*4);
254 nlh->nlmsg_len = skb->tail - b;
259 skb_trim(skb, b - skb->data);
263 static int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
266 int s_idx = cb->args[0];
267 struct net_device *dev;
269 read_lock(&dev_base_lock);
270 for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
273 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
276 read_unlock(&dev_base_lock);
282 static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
284 struct ifinfomsg *ifm = NLMSG_DATA(nlh);
285 struct rtattr **ida = arg;
286 struct net_device *dev;
287 int err, send_addr_notify = 0;
289 if (ifm->ifi_index >= 0)
290 dev = dev_get_by_index(ifm->ifi_index);
291 else if (ida[IFLA_IFNAME - 1]) {
292 char ifname[IFNAMSIZ];
294 if (rtattr_strlcpy(ifname, ida[IFLA_IFNAME - 1],
295 IFNAMSIZ) >= IFNAMSIZ)
297 dev = dev_get_by_name(ifname);
307 dev_change_flags(dev, ifm->ifi_flags);
309 if (ida[IFLA_MAP - 1]) {
310 struct rtnl_link_ifmap *u_map;
313 if (!dev->set_config) {
318 if (!netif_device_present(dev)) {
323 if (ida[IFLA_MAP - 1]->rta_len != RTA_LENGTH(sizeof(*u_map)))
326 u_map = RTA_DATA(ida[IFLA_MAP - 1]);
328 k_map.mem_start = (unsigned long) u_map->mem_start;
329 k_map.mem_end = (unsigned long) u_map->mem_end;
330 k_map.base_addr = (unsigned short) u_map->base_addr;
331 k_map.irq = (unsigned char) u_map->irq;
332 k_map.dma = (unsigned char) u_map->dma;
333 k_map.port = (unsigned char) u_map->port;
335 err = dev->set_config(dev, &k_map);
341 if (ida[IFLA_ADDRESS - 1]) {
342 if (!dev->set_mac_address) {
346 if (!netif_device_present(dev)) {
350 if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
353 err = dev->set_mac_address(dev, RTA_DATA(ida[IFLA_ADDRESS - 1]));
356 send_addr_notify = 1;
359 if (ida[IFLA_BROADCAST - 1]) {
360 if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
362 memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
364 send_addr_notify = 1;
367 if (ida[IFLA_MTU - 1]) {
368 if (ida[IFLA_MTU - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
370 err = dev_set_mtu(dev, *((u32 *) RTA_DATA(ida[IFLA_MTU - 1])));
377 if (ida[IFLA_TXQLEN - 1]) {
378 if (ida[IFLA_TXQLEN - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
381 dev->tx_queue_len = *((u32 *) RTA_DATA(ida[IFLA_TXQLEN - 1]));
384 if (ida[IFLA_WEIGHT - 1]) {
385 if (ida[IFLA_WEIGHT - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
388 dev->weight = *((u32 *) RTA_DATA(ida[IFLA_WEIGHT - 1]));
391 if (ifm->ifi_index >= 0 && ida[IFLA_IFNAME - 1]) {
392 char ifname[IFNAMSIZ];
394 if (rtattr_strlcpy(ifname, ida[IFLA_IFNAME - 1],
395 IFNAMSIZ) >= IFNAMSIZ)
397 err = dev_change_name(dev, ifname);
405 if (send_addr_notify)
406 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
412 static int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
415 int s_idx = cb->family;
419 for (idx=1; idx<NPROTO; idx++) {
420 int type = cb->nlh->nlmsg_type-RTM_BASE;
421 if (idx < s_idx || idx == PF_PACKET)
423 if (rtnetlink_links[idx] == NULL ||
424 rtnetlink_links[idx][type].dumpit == NULL)
427 memset(&cb->args[0], 0, sizeof(cb->args));
428 if (rtnetlink_links[idx][type].dumpit(skb, cb))
436 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
439 int size = NLMSG_SPACE(sizeof(struct ifinfomsg) +
440 sizeof(struct rtnl_link_ifmap) +
441 sizeof(struct rtnl_link_stats) + 128);
443 skb = alloc_skb(size, GFP_KERNEL);
447 if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change) < 0) {
451 NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
452 netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
455 static int rtnetlink_done(struct netlink_callback *cb)
460 /* Protected by RTNL sempahore. */
461 static struct rtattr **rta_buf;
462 static int rtattr_max;
464 /* Process one rtnetlink message. */
466 static __inline__ int
467 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
469 struct rtnetlink_link *link;
470 struct rtnetlink_link *link_tab;
477 /* Only requests are handled by kernel now */
478 if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
481 type = nlh->nlmsg_type;
483 /* A control message: ignore them */
487 /* Unknown message: reply with EINVAL */
493 /* All the messages must have at least 1 byte length */
494 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
497 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
498 if (family >= NPROTO) {
499 *errp = -EAFNOSUPPORT;
503 link_tab = rtnetlink_links[family];
504 if (link_tab == NULL)
505 link_tab = rtnetlink_links[PF_UNSPEC];
506 link = &link_tab[type];
511 if (kind != 2 && security_netlink_recv(skb)) {
516 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
519 if (link->dumpit == NULL)
520 link = &(rtnetlink_links[PF_UNSPEC][type]);
522 if (link->dumpit == NULL)
525 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
527 rtnetlink_done)) != 0) {
530 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
537 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
539 min_len = rtm_min[sz_idx];
540 if (nlh->nlmsg_len < min_len)
543 if (nlh->nlmsg_len > min_len) {
544 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
545 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
547 while (RTA_OK(attr, attrlen)) {
548 unsigned flavor = attr->rta_type;
550 if (flavor > rta_max[sz_idx])
552 rta_buf[flavor-1] = attr;
554 attr = RTA_NEXT(attr, attrlen);
558 if (link->doit == NULL)
559 link = &(rtnetlink_links[PF_UNSPEC][type]);
560 if (link->doit == NULL)
562 err = link->doit(skb, nlh, (void *)&rta_buf[0]);
573 * Process one packet of messages.
574 * Malformed skbs with wrong lengths of messages are discarded silently.
577 static inline int rtnetlink_rcv_skb(struct sk_buff *skb)
580 struct nlmsghdr * nlh;
582 while (skb->len >= NLMSG_SPACE(0)) {
585 nlh = (struct nlmsghdr *)skb->data;
586 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
588 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
591 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
592 /* Not error, but we must interrupt processing here:
593 * Note, that in this case we do not pull message
594 * from skb, it will be processed later.
598 netlink_ack(skb, nlh, err);
599 } else if (nlh->nlmsg_flags&NLM_F_ACK)
600 netlink_ack(skb, nlh, 0);
608 * rtnetlink input queue processing routine:
609 * - try to acquire shared lock. If it is failed, defer processing.
610 * - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
611 * that will occur, when a dump started and/or acquisition of
612 * exclusive lock failed.
615 static void rtnetlink_rcv(struct sock *sk, int len)
620 if (rtnl_shlock_nowait())
623 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
624 if (rtnetlink_rcv_skb(skb)) {
626 skb_queue_head(&sk->sk_receive_queue,
638 } while (rtnl && rtnl->sk_receive_queue.qlen);
641 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
643 [RTM_GETLINK - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
644 [RTM_SETLINK - RTM_BASE] = { .doit = do_setlink },
645 [RTM_GETADDR - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
646 [RTM_GETROUTE - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
647 [RTM_NEWNEIGH - RTM_BASE] = { .doit = neigh_add },
648 [RTM_DELNEIGH - RTM_BASE] = { .doit = neigh_delete },
649 [RTM_GETNEIGH - RTM_BASE] = { .dumpit = neigh_dump_info },
650 [RTM_GETRULE - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
653 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
655 struct net_device *dev = ptr;
657 case NETDEV_UNREGISTER:
658 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
660 case NETDEV_REGISTER:
661 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
665 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
668 case NETDEV_GOING_DOWN:
671 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
677 static struct notifier_block rtnetlink_dev_notifier = {
678 .notifier_call = rtnetlink_event,
681 void __init rtnetlink_init(void)
686 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
687 if (rta_max[i] > rtattr_max)
688 rtattr_max = rta_max[i];
689 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
691 panic("rtnetlink_init: cannot allocate rta_buf\n");
693 rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
695 panic("rtnetlink_init: cannot initialize rtnetlink\n");
696 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
697 register_netdevice_notifier(&rtnetlink_dev_notifier);
698 rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
699 rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
702 EXPORT_SYMBOL(__rta_fill);
703 EXPORT_SYMBOL(rtattr_strlcpy);
704 EXPORT_SYMBOL(rtattr_parse);
705 EXPORT_SYMBOL(rtnetlink_links);
706 EXPORT_SYMBOL(rtnetlink_put_metrics);
708 EXPORT_SYMBOL(rtnl_lock);
709 EXPORT_SYMBOL(rtnl_lock_interruptible);
710 EXPORT_SYMBOL(rtnl_sem);
711 EXPORT_SYMBOL(rtnl_unlock);