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_NR_FAMILIES] =
91 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
92 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
93 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
94 [RTM_FAM(RTM_NEWNEIGH)] = NLMSG_LENGTH(sizeof(struct ndmsg)),
95 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
96 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
97 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
98 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
99 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
100 [RTM_FAM(RTM_NEWPREFIX)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
101 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
102 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
103 [RTM_FAM(RTM_NEWNEIGHTBL)] = NLMSG_LENGTH(sizeof(struct ndtmsg)),
106 static const int rta_max[RTM_NR_FAMILIES] =
108 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
109 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
110 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
111 [RTM_FAM(RTM_NEWNEIGH)] = NDA_MAX,
112 [RTM_FAM(RTM_NEWRULE)] = RTA_MAX,
113 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
114 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
115 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
116 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
117 [RTM_FAM(RTM_NEWNEIGHTBL)] = NDTA_MAX,
120 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
123 int size = RTA_LENGTH(attrlen);
125 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
126 rta->rta_type = attrtype;
128 memcpy(RTA_DATA(rta), data, attrlen);
129 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
132 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
134 size_t ret = RTA_PAYLOAD(rta);
135 char *src = RTA_DATA(rta);
137 if (ret > 0 && src[ret - 1] == '\0')
140 size_t len = (ret >= size) ? size - 1 : ret;
141 memset(dest, 0, size);
142 memcpy(dest, src, len);
147 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
151 NETLINK_CB(skb).dst_group = group;
153 atomic_inc(&skb->users);
154 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
156 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
160 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
162 struct rtattr *mx = (struct rtattr*)skb->tail;
165 RTA_PUT(skb, RTA_METRICS, 0, NULL);
166 for (i=0; i<RTAX_MAX; i++) {
168 RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
170 mx->rta_len = skb->tail - (u8*)mx;
171 if (mx->rta_len == RTA_LENGTH(0))
172 skb_trim(skb, (u8*)mx - skb->data);
176 skb_trim(skb, (u8*)mx - skb->data);
181 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
182 int type, u32 pid, u32 seq, u32 change,
186 struct nlmsghdr *nlh;
187 unsigned char *b = skb->tail;
189 nlh = NLMSG_NEW(skb, pid, seq, type, sizeof(*r), flags);
191 r->ifi_family = AF_UNSPEC;
193 r->ifi_type = dev->type;
194 r->ifi_index = dev->ifindex;
195 r->ifi_flags = dev_get_flags(dev);
196 r->ifi_change = change;
198 RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
201 u32 txqlen = dev->tx_queue_len;
202 RTA_PUT(skb, IFLA_TXQLEN, sizeof(txqlen), &txqlen);
206 u32 weight = dev->weight;
207 RTA_PUT(skb, IFLA_WEIGHT, sizeof(weight), &weight);
211 struct rtnl_link_ifmap map = {
212 .mem_start = dev->mem_start,
213 .mem_end = dev->mem_end,
214 .base_addr = dev->base_addr,
217 .port = dev->if_port,
219 RTA_PUT(skb, IFLA_MAP, sizeof(map), &map);
223 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
224 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
229 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
232 if (dev->ifindex != dev->iflink) {
233 u32 iflink = dev->iflink;
234 RTA_PUT(skb, IFLA_LINK, sizeof(iflink), &iflink);
237 if (dev->qdisc_sleeping)
238 RTA_PUT(skb, IFLA_QDISC,
239 strlen(dev->qdisc_sleeping->ops->id) + 1,
240 dev->qdisc_sleeping->ops->id);
243 u32 master = dev->master->ifindex;
244 RTA_PUT(skb, IFLA_MASTER, sizeof(master), &master);
247 if (dev->get_stats) {
248 unsigned long *stats = (unsigned long*)dev->get_stats(dev);
253 int n = sizeof(struct rtnl_link_stats)/4;
255 a = __RTA_PUT(skb, IFLA_STATS, n*4);
261 nlh->nlmsg_len = skb->tail - b;
266 skb_trim(skb, b - skb->data);
270 static int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
273 int s_idx = cb->args[0];
274 struct net_device *dev;
276 read_lock(&dev_base_lock);
277 for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
280 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK,
281 NETLINK_CB(cb->skb).pid,
282 cb->nlh->nlmsg_seq, 0,
286 read_unlock(&dev_base_lock);
292 static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
294 struct ifinfomsg *ifm = NLMSG_DATA(nlh);
295 struct rtattr **ida = arg;
296 struct net_device *dev;
297 int err, send_addr_notify = 0;
299 if (ifm->ifi_index >= 0)
300 dev = dev_get_by_index(ifm->ifi_index);
301 else if (ida[IFLA_IFNAME - 1]) {
302 char ifname[IFNAMSIZ];
304 if (rtattr_strlcpy(ifname, ida[IFLA_IFNAME - 1],
305 IFNAMSIZ) >= IFNAMSIZ)
307 dev = dev_get_by_name(ifname);
317 dev_change_flags(dev, ifm->ifi_flags);
319 if (ida[IFLA_MAP - 1]) {
320 struct rtnl_link_ifmap *u_map;
323 if (!dev->set_config) {
328 if (!netif_device_present(dev)) {
333 if (ida[IFLA_MAP - 1]->rta_len != RTA_LENGTH(sizeof(*u_map)))
336 u_map = RTA_DATA(ida[IFLA_MAP - 1]);
338 k_map.mem_start = (unsigned long) u_map->mem_start;
339 k_map.mem_end = (unsigned long) u_map->mem_end;
340 k_map.base_addr = (unsigned short) u_map->base_addr;
341 k_map.irq = (unsigned char) u_map->irq;
342 k_map.dma = (unsigned char) u_map->dma;
343 k_map.port = (unsigned char) u_map->port;
345 err = dev->set_config(dev, &k_map);
351 if (ida[IFLA_ADDRESS - 1]) {
352 if (!dev->set_mac_address) {
356 if (!netif_device_present(dev)) {
360 if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
363 err = dev->set_mac_address(dev, RTA_DATA(ida[IFLA_ADDRESS - 1]));
366 send_addr_notify = 1;
369 if (ida[IFLA_BROADCAST - 1]) {
370 if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
372 memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
374 send_addr_notify = 1;
377 if (ida[IFLA_MTU - 1]) {
378 if (ida[IFLA_MTU - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
380 err = dev_set_mtu(dev, *((u32 *) RTA_DATA(ida[IFLA_MTU - 1])));
387 if (ida[IFLA_TXQLEN - 1]) {
388 if (ida[IFLA_TXQLEN - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
391 dev->tx_queue_len = *((u32 *) RTA_DATA(ida[IFLA_TXQLEN - 1]));
394 if (ida[IFLA_WEIGHT - 1]) {
395 if (ida[IFLA_WEIGHT - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
398 dev->weight = *((u32 *) RTA_DATA(ida[IFLA_WEIGHT - 1]));
401 if (ifm->ifi_index >= 0 && ida[IFLA_IFNAME - 1]) {
402 char ifname[IFNAMSIZ];
404 if (rtattr_strlcpy(ifname, ida[IFLA_IFNAME - 1],
405 IFNAMSIZ) >= IFNAMSIZ)
407 err = dev_change_name(dev, ifname);
415 if (send_addr_notify)
416 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
422 static int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
425 int s_idx = cb->family;
429 for (idx=1; idx<NPROTO; idx++) {
430 int type = cb->nlh->nlmsg_type-RTM_BASE;
431 if (idx < s_idx || idx == PF_PACKET)
433 if (rtnetlink_links[idx] == NULL ||
434 rtnetlink_links[idx][type].dumpit == NULL)
437 memset(&cb->args[0], 0, sizeof(cb->args));
438 if (rtnetlink_links[idx][type].dumpit(skb, cb))
446 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
449 int size = NLMSG_SPACE(sizeof(struct ifinfomsg) +
450 sizeof(struct rtnl_link_ifmap) +
451 sizeof(struct rtnl_link_stats) + 128);
453 skb = alloc_skb(size, GFP_KERNEL);
457 if (rtnetlink_fill_ifinfo(skb, dev, type, current->pid, 0, change, 0) < 0) {
461 NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
462 netlink_broadcast(rtnl, skb, 0, RTNLGRP_LINK, GFP_KERNEL);
465 static int rtnetlink_done(struct netlink_callback *cb)
470 /* Protected by RTNL sempahore. */
471 static struct rtattr **rta_buf;
472 static int rtattr_max;
474 /* Process one rtnetlink message. */
476 static __inline__ int
477 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
479 struct rtnetlink_link *link;
480 struct rtnetlink_link *link_tab;
487 /* Only requests are handled by kernel now */
488 if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
491 type = nlh->nlmsg_type;
493 /* A control message: ignore them */
497 /* Unknown message: reply with EINVAL */
503 /* All the messages must have at least 1 byte length */
504 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
507 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
508 if (family >= NPROTO) {
509 *errp = -EAFNOSUPPORT;
513 link_tab = rtnetlink_links[family];
514 if (link_tab == NULL)
515 link_tab = rtnetlink_links[PF_UNSPEC];
516 link = &link_tab[type];
521 if (kind != 2 && security_netlink_recv(skb)) {
526 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
529 if (link->dumpit == NULL)
530 link = &(rtnetlink_links[PF_UNSPEC][type]);
532 if (link->dumpit == NULL)
535 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
537 rtnetlink_done)) != 0) {
540 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
547 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
549 min_len = rtm_min[sz_idx];
550 if (nlh->nlmsg_len < min_len)
553 if (nlh->nlmsg_len > min_len) {
554 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
555 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
557 while (RTA_OK(attr, attrlen)) {
558 unsigned flavor = attr->rta_type;
560 if (flavor > rta_max[sz_idx])
562 rta_buf[flavor-1] = attr;
564 attr = RTA_NEXT(attr, attrlen);
568 if (link->doit == NULL)
569 link = &(rtnetlink_links[PF_UNSPEC][type]);
570 if (link->doit == NULL)
572 err = link->doit(skb, nlh, (void *)&rta_buf[0]);
583 * Process one packet of messages.
584 * Malformed skbs with wrong lengths of messages are discarded silently.
587 static inline int rtnetlink_rcv_skb(struct sk_buff *skb)
590 struct nlmsghdr * nlh;
592 while (skb->len >= NLMSG_SPACE(0)) {
595 nlh = (struct nlmsghdr *)skb->data;
596 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
598 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
601 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
602 /* Not error, but we must interrupt processing here:
603 * Note, that in this case we do not pull message
604 * from skb, it will be processed later.
608 netlink_ack(skb, nlh, err);
609 } else if (nlh->nlmsg_flags&NLM_F_ACK)
610 netlink_ack(skb, nlh, 0);
618 * rtnetlink input queue processing routine:
619 * - process as much as there was in the queue upon entry.
620 * - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
621 * that will occur, when a dump started.
624 static void rtnetlink_rcv(struct sock *sk, int len)
626 unsigned int qlen = skb_queue_len(&sk->sk_receive_queue);
633 if (qlen > skb_queue_len(&sk->sk_receive_queue))
634 qlen = skb_queue_len(&sk->sk_receive_queue);
636 for (; qlen; qlen--) {
637 skb = skb_dequeue(&sk->sk_receive_queue);
638 if (rtnetlink_rcv_skb(skb)) {
640 skb_queue_head(&sk->sk_receive_queue,
657 static struct rtnetlink_link link_rtnetlink_table[RTM_NR_MSGTYPES] =
659 [RTM_GETLINK - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
660 [RTM_SETLINK - RTM_BASE] = { .doit = do_setlink },
661 [RTM_GETADDR - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
662 [RTM_GETROUTE - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
663 [RTM_NEWNEIGH - RTM_BASE] = { .doit = neigh_add },
664 [RTM_DELNEIGH - RTM_BASE] = { .doit = neigh_delete },
665 [RTM_GETNEIGH - RTM_BASE] = { .dumpit = neigh_dump_info },
666 [RTM_GETRULE - RTM_BASE] = { .dumpit = rtnetlink_dump_all },
667 [RTM_GETNEIGHTBL - RTM_BASE] = { .dumpit = neightbl_dump_info },
668 [RTM_SETNEIGHTBL - RTM_BASE] = { .doit = neightbl_set },
671 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
673 struct net_device *dev = ptr;
675 case NETDEV_UNREGISTER:
676 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
678 case NETDEV_REGISTER:
679 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
683 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
686 case NETDEV_GOING_DOWN:
689 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
695 static struct notifier_block rtnetlink_dev_notifier = {
696 .notifier_call = rtnetlink_event,
699 void __init rtnetlink_init(void)
704 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
705 if (rta_max[i] > rtattr_max)
706 rtattr_max = rta_max[i];
707 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
709 panic("rtnetlink_init: cannot allocate rta_buf\n");
711 rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv, THIS_MODULE);
713 panic("rtnetlink_init: cannot initialize rtnetlink\n");
714 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
715 register_netdevice_notifier(&rtnetlink_dev_notifier);
716 rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
717 rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
720 EXPORT_SYMBOL(__rta_fill);
721 EXPORT_SYMBOL(rtattr_strlcpy);
722 EXPORT_SYMBOL(rtattr_parse);
723 EXPORT_SYMBOL(rtnetlink_links);
724 EXPORT_SYMBOL(rtnetlink_put_metrics);
726 EXPORT_SYMBOL(rtnl_lock);
727 EXPORT_SYMBOL(rtnl_lock_interruptible);
728 EXPORT_SYMBOL(rtnl_sem);
729 EXPORT_SYMBOL(rtnl_unlock);