1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * INET		An implementation of the TCP/IP protocol suite for the LINUX
4 *		operating system.  INET is implemented using the  BSD Socket
5 *		interface as the means of communication with the user level.
6 *
7 *		IPv4 Forwarding Information Base: FIB frontend.
8 *
9 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 */
11
12#include <linux/module.h>
13#include <linux/uaccess.h>
14#include <linux/bitops.h>
15#include <linux/capability.h>
16#include <linux/types.h>
17#include <linux/kernel.h>
18#include <linux/mm.h>
19#include <linux/string.h>
20#include <linux/socket.h>
21#include <linux/sockios.h>
22#include <linux/errno.h>
23#include <linux/in.h>
24#include <linux/inet.h>
25#include <linux/inetdevice.h>
26#include <linux/netdevice.h>
27#include <linux/if_addr.h>
28#include <linux/if_arp.h>
29#include <linux/skbuff.h>
30#include <linux/cache.h>
31#include <linux/init.h>
32#include <linux/list.h>
33#include <linux/slab.h>
34
35#include <net/ip.h>
36#include <net/protocol.h>
37#include <net/route.h>
38#include <net/tcp.h>
39#include <net/sock.h>
40#include <net/arp.h>
41#include <net/ip_fib.h>
42#include <net/nexthop.h>
43#include <net/rtnetlink.h>
44#include <net/xfrm.h>
45#include <net/l3mdev.h>
46#include <net/lwtunnel.h>
47#include <trace/events/fib.h>
48
49#ifndef CONFIG_IP_MULTIPLE_TABLES
50
51static int __net_init fib4_rules_init(struct net *net)
52{
53	struct fib_table *local_table, *main_table;
54
55	main_table  = fib_trie_table(RT_TABLE_MAIN, NULL);
56	if (!main_table)
57		return -ENOMEM;
58
59	local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
60	if (!local_table)
61		goto fail;
62
63	hlist_add_head_rcu(&local_table->tb_hlist,
64				&net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
65	hlist_add_head_rcu(&main_table->tb_hlist,
66				&net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
67	return 0;
68
69fail:
70	fib_free_table(main_table);
71	return -ENOMEM;
72}
73#else
74
75struct fib_table *fib_new_table(struct net *net, u32 id)
76{
77	struct fib_table *tb, *alias = NULL;
78	unsigned int h;
79
80	if (id == 0)
81		id = RT_TABLE_MAIN;
82	tb = fib_get_table(net, id);
83	if (tb)
84		return tb;
85
86	if (id == RT_TABLE_LOCAL && !net->ipv4.fib_has_custom_rules)
87		alias = fib_new_table(net, RT_TABLE_MAIN);
88
89	tb = fib_trie_table(id, alias);
90	if (!tb)
91		return NULL;
92
93	switch (id) {
94	case RT_TABLE_MAIN:
95		rcu_assign_pointer(net->ipv4.fib_main, tb);
96		break;
97	case RT_TABLE_DEFAULT:
98		rcu_assign_pointer(net->ipv4.fib_default, tb);
99		break;
100	default:
101		break;
102	}
103
104	h = id & (FIB_TABLE_HASHSZ - 1);
105	hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
106	return tb;
107}
108EXPORT_SYMBOL_GPL(fib_new_table);
109
110/* caller must hold either rtnl or rcu read lock */
111struct fib_table *fib_get_table(struct net *net, u32 id)
112{
113	struct fib_table *tb;
114	struct hlist_head *head;
115	unsigned int h;
116
117	if (id == 0)
118		id = RT_TABLE_MAIN;
119	h = id & (FIB_TABLE_HASHSZ - 1);
120
121	head = &net->ipv4.fib_table_hash[h];
122	hlist_for_each_entry_rcu(tb, head, tb_hlist,
123				 lockdep_rtnl_is_held()) {
124		if (tb->tb_id == id)
125			return tb;
126	}
127	return NULL;
128}
129#endif /* CONFIG_IP_MULTIPLE_TABLES */
130
131static void fib_replace_table(struct net *net, struct fib_table *old,
132			      struct fib_table *new)
133{
134#ifdef CONFIG_IP_MULTIPLE_TABLES
135	switch (new->tb_id) {
136	case RT_TABLE_MAIN:
137		rcu_assign_pointer(net->ipv4.fib_main, new);
138		break;
139	case RT_TABLE_DEFAULT:
140		rcu_assign_pointer(net->ipv4.fib_default, new);
141		break;
142	default:
143		break;
144	}
145
146#endif
147	/* replace the old table in the hlist */
148	hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
149}
150
151int fib_unmerge(struct net *net)
152{
153	struct fib_table *old, *new, *main_table;
154
155	/* attempt to fetch local table if it has been allocated */
156	old = fib_get_table(net, RT_TABLE_LOCAL);
157	if (!old)
158		return 0;
159
160	new = fib_trie_unmerge(old);
161	if (!new)
162		return -ENOMEM;
163
164	/* table is already unmerged */
165	if (new == old)
166		return 0;
167
168	/* replace merged table with clean table */
169	fib_replace_table(net, old, new);
170	fib_free_table(old);
171
172	/* attempt to fetch main table if it has been allocated */
173	main_table = fib_get_table(net, RT_TABLE_MAIN);
174	if (!main_table)
175		return 0;
176
177	/* flush local entries from main table */
178	fib_table_flush_external(main_table);
179
180	return 0;
181}
182
183void fib_flush(struct net *net)
184{
185	int flushed = 0;
186	unsigned int h;
187
188	for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
189		struct hlist_head *head = &net->ipv4.fib_table_hash[h];
190		struct hlist_node *tmp;
191		struct fib_table *tb;
192
193		hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
194			flushed += fib_table_flush(net, tb, false);
195	}
196
197	if (flushed)
198		rt_cache_flush(net);
199}
200
201/*
202 * Find address type as if only "dev" was present in the system. If
203 * on_dev is NULL then all interfaces are taken into consideration.
204 */
205static inline unsigned int __inet_dev_addr_type(struct net *net,
206						const struct net_device *dev,
207						__be32 addr, u32 tb_id)
208{
209	struct flowi4		fl4 = { .daddr = addr };
210	struct fib_result	res;
211	unsigned int ret = RTN_BROADCAST;
212	struct fib_table *table;
213
214	if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
215		return RTN_BROADCAST;
216	if (ipv4_is_multicast(addr))
217		return RTN_MULTICAST;
218
219	rcu_read_lock();
220
221	table = fib_get_table(net, tb_id);
222	if (table) {
223		ret = RTN_UNICAST;
224		if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
225			struct fib_nh_common *nhc = fib_info_nhc(res.fi, 0);
226
227			if (!dev || dev == nhc->nhc_dev)
228				ret = res.type;
229		}
230	}
231
232	rcu_read_unlock();
233	return ret;
234}
235
236unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id)
237{
238	return __inet_dev_addr_type(net, NULL, addr, tb_id);
239}
240EXPORT_SYMBOL(inet_addr_type_table);
241
242unsigned int inet_addr_type(struct net *net, __be32 addr)
243{
244	return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
245}
246EXPORT_SYMBOL(inet_addr_type);
247
248unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
249				__be32 addr)
250{
251	u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
252
253	return __inet_dev_addr_type(net, dev, addr, rt_table);
254}
255EXPORT_SYMBOL(inet_dev_addr_type);
256
257/* inet_addr_type with dev == NULL but using the table from a dev
258 * if one is associated
259 */
260unsigned int inet_addr_type_dev_table(struct net *net,
261				      const struct net_device *dev,
262				      __be32 addr)
263{
264	u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
265
266	return __inet_dev_addr_type(net, NULL, addr, rt_table);
267}
268EXPORT_SYMBOL(inet_addr_type_dev_table);
269
270__be32 fib_compute_spec_dst(struct sk_buff *skb)
271{
272	struct net_device *dev = skb->dev;
273	struct in_device *in_dev;
274	struct fib_result res;
275	struct rtable *rt;
276	struct net *net;
277	int scope;
278
279	rt = skb_rtable(skb);
280	if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
281	    RTCF_LOCAL)
282		return ip_hdr(skb)->daddr;
283
284	in_dev = __in_dev_get_rcu(dev);
285
286	net = dev_net(dev);
287
288	scope = RT_SCOPE_UNIVERSE;
289	if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
290		bool vmark = in_dev && IN_DEV_SRC_VMARK(in_dev);
291		struct flowi4 fl4 = {
292			.flowi4_iif = LOOPBACK_IFINDEX,
293			.flowi4_oif = l3mdev_master_ifindex_rcu(dev),
294			.daddr = ip_hdr(skb)->saddr,
295			.flowi4_tos = ip_hdr(skb)->tos & IPTOS_RT_MASK,
296			.flowi4_scope = scope,
297			.flowi4_mark = vmark ? skb->mark : 0,
298		};
299		if (!fib_lookup(net, &fl4, &res, 0))
300			return fib_result_prefsrc(net, &res);
301	} else {
302		scope = RT_SCOPE_LINK;
303	}
304
305	return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
306}
307
308bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev)
309{
310	bool dev_match = false;
311#ifdef CONFIG_IP_ROUTE_MULTIPATH
312	if (unlikely(fi->nh)) {
313		dev_match = nexthop_uses_dev(fi->nh, dev);
314	} else {
315		int ret;
316
317		for (ret = 0; ret < fib_info_num_path(fi); ret++) {
318			const struct fib_nh_common *nhc = fib_info_nhc(fi, ret);
319
320			if (nhc_l3mdev_matches_dev(nhc, dev)) {
321				dev_match = true;
322				break;
323			}
324		}
325	}
326#else
327	if (fib_info_nhc(fi, 0)->nhc_dev == dev)
328		dev_match = true;
329#endif
330
331	return dev_match;
332}
333EXPORT_SYMBOL_GPL(fib_info_nh_uses_dev);
334
335/* Given (packet source, input interface) and optional (dst, oif, tos):
336 * - (main) check, that source is valid i.e. not broadcast or our local
337 *   address.
338 * - figure out what "logical" interface this packet arrived
339 *   and calculate "specific destination" address.
340 * - check, that packet arrived from expected physical interface.
341 * called with rcu_read_lock()
342 */
343static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
344				 u8 tos, int oif, struct net_device *dev,
345				 int rpf, struct in_device *idev, u32 *itag)
346{
347	struct net *net = dev_net(dev);
348	struct flow_keys flkeys;
349	int ret, no_addr;
350	struct fib_result res;
351	struct flowi4 fl4;
352	bool dev_match;
353
354	fl4.flowi4_oif = 0;
355	fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev);
356	if (!fl4.flowi4_iif)
357		fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
358	fl4.daddr = src;
359	fl4.saddr = dst;
360	fl4.flowi4_tos = tos;
361	fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
362	fl4.flowi4_tun_key.tun_id = 0;
363	fl4.flowi4_flags = 0;
364	fl4.flowi4_uid = sock_net_uid(net, NULL);
365	fl4.flowi4_multipath_hash = 0;
366
367	no_addr = idev->ifa_list == NULL;
368
369	fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
370	if (!fib4_rules_early_flow_dissect(net, skb, &fl4, &flkeys)) {
371		fl4.flowi4_proto = 0;
372		fl4.fl4_sport = 0;
373		fl4.fl4_dport = 0;
374	} else {
375		swap(fl4.fl4_sport, fl4.fl4_dport);
376	}
377
378	if (fib_lookup(net, &fl4, &res, 0))
379		goto last_resort;
380	if (res.type != RTN_UNICAST &&
381	    (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
382		goto e_inval;
383	fib_combine_itag(itag, &res);
384
385	dev_match = fib_info_nh_uses_dev(res.fi, dev);
386	/* This is not common, loopback packets retain skb_dst so normally they
387	 * would not even hit this slow path.
388	 */
389	dev_match = dev_match || (res.type == RTN_LOCAL &&
390				  dev == net->loopback_dev);
391	if (dev_match) {
392		ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_LINK;
393		return ret;
394	}
395	if (no_addr)
396		goto last_resort;
397	if (rpf == 1)
398		goto e_rpf;
399	fl4.flowi4_oif = dev->ifindex;
400
401	ret = 0;
402	if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
403		if (res.type == RTN_UNICAST)
404			ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_LINK;
405	}
406	return ret;
407
408last_resort:
409	if (rpf)
410		goto e_rpf;
411	*itag = 0;
412	return 0;
413
414e_inval:
415	return -EINVAL;
416e_rpf:
417	return -EXDEV;
418}
419
420/* Ignore rp_filter for packets protected by IPsec. */
421int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
422			u8 tos, int oif, struct net_device *dev,
423			struct in_device *idev, u32 *itag)
424{
425	int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
426	struct net *net = dev_net(dev);
427
428	if (!r && !fib_num_tclassid_users(net) &&
429	    (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
430		if (IN_DEV_ACCEPT_LOCAL(idev))
431			goto ok;
432		/* with custom local routes in place, checking local addresses
433		 * only will be too optimistic, with custom rules, checking
434		 * local addresses only can be too strict, e.g. due to vrf
435		 */
436		if (net->ipv4.fib_has_custom_local_routes ||
437		    fib4_has_custom_rules(net))
438			goto full_check;
439		if (inet_lookup_ifaddr_rcu(net, src))
440			return -EINVAL;
441
442ok:
443		*itag = 0;
444		return 0;
445	}
446
447full_check:
448	return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
449}
450
451static inline __be32 sk_extract_addr(struct sockaddr *addr)
452{
453	return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
454}
455
456static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
457{
458	struct nlattr *nla;
459
460	nla = (struct nlattr *) ((char *) mx + len);
461	nla->nla_type = type;
462	nla->nla_len = nla_attr_size(4);
463	*(u32 *) nla_data(nla) = value;
464
465	return len + nla_total_size(4);
466}
467
468static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
469				 struct fib_config *cfg)
470{
471	__be32 addr;
472	int plen;
473
474	memset(cfg, 0, sizeof(*cfg));
475	cfg->fc_nlinfo.nl_net = net;
476
477	if (rt->rt_dst.sa_family != AF_INET)
478		return -EAFNOSUPPORT;
479
480	/*
481	 * Check mask for validity:
482	 * a) it must be contiguous.
483	 * b) destination must have all host bits clear.
484	 * c) if application forgot to set correct family (AF_INET),
485	 *    reject request unless it is absolutely clear i.e.
486	 *    both family and mask are zero.
487	 */
488	plen = 32;
489	addr = sk_extract_addr(&rt->rt_dst);
490	if (!(rt->rt_flags & RTF_HOST)) {
491		__be32 mask = sk_extract_addr(&rt->rt_genmask);
492
493		if (rt->rt_genmask.sa_family != AF_INET) {
494			if (mask || rt->rt_genmask.sa_family)
495				return -EAFNOSUPPORT;
496		}
497
498		if (bad_mask(mask, addr))
499			return -EINVAL;
500
501		plen = inet_mask_len(mask);
502	}
503
504	cfg->fc_dst_len = plen;
505	cfg->fc_dst = addr;
506
507	if (cmd != SIOCDELRT) {
508		cfg->fc_nlflags = NLM_F_CREATE;
509		cfg->fc_protocol = RTPROT_BOOT;
510	}
511
512	if (rt->rt_metric)
513		cfg->fc_priority = rt->rt_metric - 1;
514
515	if (rt->rt_flags & RTF_REJECT) {
516		cfg->fc_scope = RT_SCOPE_HOST;
517		cfg->fc_type = RTN_UNREACHABLE;
518		return 0;
519	}
520
521	cfg->fc_scope = RT_SCOPE_NOWHERE;
522	cfg->fc_type = RTN_UNICAST;
523
524	if (rt->rt_dev) {
525		char *colon;
526		struct net_device *dev;
527		char devname[IFNAMSIZ];
528
529		if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
530			return -EFAULT;
531
532		devname[IFNAMSIZ-1] = 0;
533		colon = strchr(devname, ':');
534		if (colon)
535			*colon = 0;
536		dev = __dev_get_by_name(net, devname);
537		if (!dev)
538			return -ENODEV;
539		cfg->fc_oif = dev->ifindex;
540		cfg->fc_table = l3mdev_fib_table(dev);
541		if (colon) {
542			const struct in_ifaddr *ifa;
543			struct in_device *in_dev;
544
545			in_dev = __in_dev_get_rtnl(dev);
546			if (!in_dev)
547				return -ENODEV;
548
549			*colon = ':';
550
551			rcu_read_lock();
552			in_dev_for_each_ifa_rcu(ifa, in_dev) {
553				if (strcmp(ifa->ifa_label, devname) == 0)
554					break;
555			}
556			rcu_read_unlock();
557
558			if (!ifa)
559				return -ENODEV;
560			cfg->fc_prefsrc = ifa->ifa_local;
561		}
562	}
563
564	addr = sk_extract_addr(&rt->rt_gateway);
565	if (rt->rt_gateway.sa_family == AF_INET && addr) {
566		unsigned int addr_type;
567
568		cfg->fc_gw4 = addr;
569		cfg->fc_gw_family = AF_INET;
570		addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
571		if (rt->rt_flags & RTF_GATEWAY &&
572		    addr_type == RTN_UNICAST)
573			cfg->fc_scope = RT_SCOPE_UNIVERSE;
574	}
575
576	if (!cfg->fc_table)
577		cfg->fc_table = RT_TABLE_MAIN;
578
579	if (cmd == SIOCDELRT)
580		return 0;
581
582	if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw_family)
583		return -EINVAL;
584
585	if (cfg->fc_scope == RT_SCOPE_NOWHERE)
586		cfg->fc_scope = RT_SCOPE_LINK;
587
588	if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
589		struct nlattr *mx;
590		int len = 0;
591
592		mx = kcalloc(3, nla_total_size(4), GFP_KERNEL);
593		if (!mx)
594			return -ENOMEM;
595
596		if (rt->rt_flags & RTF_MTU)
597			len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
598
599		if (rt->rt_flags & RTF_WINDOW)
600			len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
601
602		if (rt->rt_flags & RTF_IRTT)
603			len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
604
605		cfg->fc_mx = mx;
606		cfg->fc_mx_len = len;
607	}
608
609	return 0;
610}
611
612/*
613 * Handle IP routing ioctl calls.
614 * These are used to manipulate the routing tables
615 */
616int ip_rt_ioctl(struct net *net, unsigned int cmd, struct rtentry *rt)
617{
618	struct fib_config cfg;
619	int err;
620
621	switch (cmd) {
622	case SIOCADDRT:		/* Add a route */
623	case SIOCDELRT:		/* Delete a route */
624		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
625			return -EPERM;
626
627		rtnl_lock();
628		err = rtentry_to_fib_config(net, cmd, rt, &cfg);
629		if (err == 0) {
630			struct fib_table *tb;
631
632			if (cmd == SIOCDELRT) {
633				tb = fib_get_table(net, cfg.fc_table);
634				if (tb)
635					err = fib_table_delete(net, tb, &cfg,
636							       NULL);
637				else
638					err = -ESRCH;
639			} else {
640				tb = fib_new_table(net, cfg.fc_table);
641				if (tb)
642					err = fib_table_insert(net, tb,
643							       &cfg, NULL);
644				else
645					err = -ENOBUFS;
646			}
647
648			/* allocated by rtentry_to_fib_config() */
649			kfree(cfg.fc_mx);
650		}
651		rtnl_unlock();
652		return err;
653	}
654	return -EINVAL;
655}
656
657const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
658	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
659	[RTA_DST]		= { .type = NLA_U32 },
660	[RTA_SRC]		= { .type = NLA_U32 },
661	[RTA_IIF]		= { .type = NLA_U32 },
662	[RTA_OIF]		= { .type = NLA_U32 },
663	[RTA_GATEWAY]		= { .type = NLA_U32 },
664	[RTA_PRIORITY]		= { .type = NLA_U32 },
665	[RTA_PREFSRC]		= { .type = NLA_U32 },
666	[RTA_METRICS]		= { .type = NLA_NESTED },
667	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
668	[RTA_FLOW]		= { .type = NLA_U32 },
669	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
670	[RTA_ENCAP]		= { .type = NLA_NESTED },
671	[RTA_UID]		= { .type = NLA_U32 },
672	[RTA_MARK]		= { .type = NLA_U32 },
673	[RTA_TABLE]		= { .type = NLA_U32 },
674	[RTA_IP_PROTO]		= { .type = NLA_U8 },
675	[RTA_SPORT]		= { .type = NLA_U16 },
676	[RTA_DPORT]		= { .type = NLA_U16 },
677	[RTA_NH_ID]		= { .type = NLA_U32 },
678};
679
680int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla,
681		    struct netlink_ext_ack *extack)
682{
683	struct rtvia *via;
684	int alen;
685
686	if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
687		NL_SET_ERR_MSG(extack, "Invalid attribute length for RTA_VIA");
688		return -EINVAL;
689	}
690
691	via = nla_data(nla);
692	alen = nla_len(nla) - offsetof(struct rtvia, rtvia_addr);
693
694	switch (via->rtvia_family) {
695	case AF_INET:
696		if (alen != sizeof(__be32)) {
697			NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_VIA");
698			return -EINVAL;
699		}
700		cfg->fc_gw_family = AF_INET;
701		cfg->fc_gw4 = *((__be32 *)via->rtvia_addr);
702		break;
703	case AF_INET6:
704#if IS_ENABLED(CONFIG_IPV6)
705		if (alen != sizeof(struct in6_addr)) {
706			NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_VIA");
707			return -EINVAL;
708		}
709		cfg->fc_gw_family = AF_INET6;
710		cfg->fc_gw6 = *((struct in6_addr *)via->rtvia_addr);
711#else
712		NL_SET_ERR_MSG(extack, "IPv6 support not enabled in kernel");
713		return -EINVAL;
714#endif
715		break;
716	default:
717		NL_SET_ERR_MSG(extack, "Unsupported address family in RTA_VIA");
718		return -EINVAL;
719	}
720
721	return 0;
722}
723
724static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
725			     struct nlmsghdr *nlh, struct fib_config *cfg,
726			     struct netlink_ext_ack *extack)
727{
728	bool has_gw = false, has_via = false;
729	struct nlattr *attr;
730	int err, remaining;
731	struct rtmsg *rtm;
732
733	err = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
734					rtm_ipv4_policy, extack);
735	if (err < 0)
736		goto errout;
737
738	memset(cfg, 0, sizeof(*cfg));
739
740	rtm = nlmsg_data(nlh);
741	cfg->fc_dst_len = rtm->rtm_dst_len;
742	cfg->fc_tos = rtm->rtm_tos;
743	cfg->fc_table = rtm->rtm_table;
744	cfg->fc_protocol = rtm->rtm_protocol;
745	cfg->fc_scope = rtm->rtm_scope;
746	cfg->fc_type = rtm->rtm_type;
747	cfg->fc_flags = rtm->rtm_flags;
748	cfg->fc_nlflags = nlh->nlmsg_flags;
749
750	cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
751	cfg->fc_nlinfo.nlh = nlh;
752	cfg->fc_nlinfo.nl_net = net;
753
754	if (cfg->fc_type > RTN_MAX) {
755		NL_SET_ERR_MSG(extack, "Invalid route type");
756		err = -EINVAL;
757		goto errout;
758	}
759
760	nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
761		switch (nla_type(attr)) {
762		case RTA_DST:
763			cfg->fc_dst = nla_get_be32(attr);
764			break;
765		case RTA_OIF:
766			cfg->fc_oif = nla_get_u32(attr);
767			break;
768		case RTA_GATEWAY:
769			has_gw = true;
770			cfg->fc_gw4 = nla_get_be32(attr);
771			if (cfg->fc_gw4)
772				cfg->fc_gw_family = AF_INET;
773			break;
774		case RTA_VIA:
775			has_via = true;
776			err = fib_gw_from_via(cfg, attr, extack);
777			if (err)
778				goto errout;
779			break;
780		case RTA_PRIORITY:
781			cfg->fc_priority = nla_get_u32(attr);
782			break;
783		case RTA_PREFSRC:
784			cfg->fc_prefsrc = nla_get_be32(attr);
785			break;
786		case RTA_METRICS:
787			cfg->fc_mx = nla_data(attr);
788			cfg->fc_mx_len = nla_len(attr);
789			break;
790		case RTA_MULTIPATH:
791			err = lwtunnel_valid_encap_type_attr(nla_data(attr),
792							     nla_len(attr),
793							     extack);
794			if (err < 0)
795				goto errout;
796			cfg->fc_mp = nla_data(attr);
797			cfg->fc_mp_len = nla_len(attr);
798			break;
799		case RTA_FLOW:
800			cfg->fc_flow = nla_get_u32(attr);
801			break;
802		case RTA_TABLE:
803			cfg->fc_table = nla_get_u32(attr);
804			break;
805		case RTA_ENCAP:
806			cfg->fc_encap = attr;
807			break;
808		case RTA_ENCAP_TYPE:
809			cfg->fc_encap_type = nla_get_u16(attr);
810			err = lwtunnel_valid_encap_type(cfg->fc_encap_type,
811							extack);
812			if (err < 0)
813				goto errout;
814			break;
815		case RTA_NH_ID:
816			cfg->fc_nh_id = nla_get_u32(attr);
817			break;
818		}
819	}
820
821	if (cfg->fc_nh_id) {
822		if (cfg->fc_oif || cfg->fc_gw_family ||
823		    cfg->fc_encap || cfg->fc_mp) {
824			NL_SET_ERR_MSG(extack,
825				       "Nexthop specification and nexthop id are mutually exclusive");
826			return -EINVAL;
827		}
828	}
829
830	if (has_gw && has_via) {
831		NL_SET_ERR_MSG(extack,
832			       "Nexthop configuration can not contain both GATEWAY and VIA");
833		return -EINVAL;
834	}
835
836	if (!cfg->fc_table)
837		cfg->fc_table = RT_TABLE_MAIN;
838
839	return 0;
840errout:
841	return err;
842}
843
844static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
845			     struct netlink_ext_ack *extack)
846{
847	struct net *net = sock_net(skb->sk);
848	struct fib_config cfg;
849	struct fib_table *tb;
850	int err;
851
852	err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
853	if (err < 0)
854		goto errout;
855
856	if (cfg.fc_nh_id && !nexthop_find_by_id(net, cfg.fc_nh_id)) {
857		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
858		err = -EINVAL;
859		goto errout;
860	}
861
862	tb = fib_get_table(net, cfg.fc_table);
863	if (!tb) {
864		NL_SET_ERR_MSG(extack, "FIB table does not exist");
865		err = -ESRCH;
866		goto errout;
867	}
868
869	err = fib_table_delete(net, tb, &cfg, extack);
870errout:
871	return err;
872}
873
874static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
875			     struct netlink_ext_ack *extack)
876{
877	struct net *net = sock_net(skb->sk);
878	struct fib_config cfg;
879	struct fib_table *tb;
880	int err;
881
882	err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
883	if (err < 0)
884		goto errout;
885
886	tb = fib_new_table(net, cfg.fc_table);
887	if (!tb) {
888		err = -ENOBUFS;
889		goto errout;
890	}
891
892	err = fib_table_insert(net, tb, &cfg, extack);
893	if (!err && cfg.fc_type == RTN_LOCAL)
894		net->ipv4.fib_has_custom_local_routes = true;
895errout:
896	return err;
897}
898
899int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
900			  struct fib_dump_filter *filter,
901			  struct netlink_callback *cb)
902{
903	struct netlink_ext_ack *extack = cb->extack;
904	struct nlattr *tb[RTA_MAX + 1];
905	struct rtmsg *rtm;
906	int err, i;
907
908	ASSERT_RTNL();
909
910	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
911		NL_SET_ERR_MSG(extack, "Invalid header for FIB dump request");
912		return -EINVAL;
913	}
914
915	rtm = nlmsg_data(nlh);
916	if (rtm->rtm_dst_len || rtm->rtm_src_len  || rtm->rtm_tos   ||
917	    rtm->rtm_scope) {
918		NL_SET_ERR_MSG(extack, "Invalid values in header for FIB dump request");
919		return -EINVAL;
920	}
921
922	if (rtm->rtm_flags & ~(RTM_F_CLONED | RTM_F_PREFIX)) {
923		NL_SET_ERR_MSG(extack, "Invalid flags for FIB dump request");
924		return -EINVAL;
925	}
926	if (rtm->rtm_flags & RTM_F_CLONED)
927		filter->dump_routes = false;
928	else
929		filter->dump_exceptions = false;
930
931	filter->flags    = rtm->rtm_flags;
932	filter->protocol = rtm->rtm_protocol;
933	filter->rt_type  = rtm->rtm_type;
934	filter->table_id = rtm->rtm_table;
935
936	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
937					    rtm_ipv4_policy, extack);
938	if (err < 0)
939		return err;
940
941	for (i = 0; i <= RTA_MAX; ++i) {
942		int ifindex;
943
944		if (!tb[i])
945			continue;
946
947		switch (i) {
948		case RTA_TABLE:
949			filter->table_id = nla_get_u32(tb[i]);
950			break;
951		case RTA_OIF:
952			ifindex = nla_get_u32(tb[i]);
953			filter->dev = __dev_get_by_index(net, ifindex);
954			if (!filter->dev)
955				return -ENODEV;
956			break;
957		default:
958			NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
959			return -EINVAL;
960		}
961	}
962
963	if (filter->flags || filter->protocol || filter->rt_type ||
964	    filter->table_id || filter->dev) {
965		filter->filter_set = 1;
966		cb->answer_flags = NLM_F_DUMP_FILTERED;
967	}
968
969	return 0;
970}
971EXPORT_SYMBOL_GPL(ip_valid_fib_dump_req);
972
973static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
974{
975	struct fib_dump_filter filter = { .dump_routes = true,
976					  .dump_exceptions = true };
977	const struct nlmsghdr *nlh = cb->nlh;
978	struct net *net = sock_net(skb->sk);
979	unsigned int h, s_h;
980	unsigned int e = 0, s_e;
981	struct fib_table *tb;
982	struct hlist_head *head;
983	int dumped = 0, err;
984
985	if (cb->strict_check) {
986		err = ip_valid_fib_dump_req(net, nlh, &filter, cb);
987		if (err < 0)
988			return err;
989	} else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
990		struct rtmsg *rtm = nlmsg_data(nlh);
991
992		filter.flags = rtm->rtm_flags & (RTM_F_PREFIX | RTM_F_CLONED);
993	}
994
995	/* ipv4 does not use prefix flag */
996	if (filter.flags & RTM_F_PREFIX)
997		return skb->len;
998
999	if (filter.table_id) {
1000		tb = fib_get_table(net, filter.table_id);
1001		if (!tb) {
1002			if (rtnl_msg_family(cb->nlh) != PF_INET)
1003				return skb->len;
1004
1005			NL_SET_ERR_MSG(cb->extack, "ipv4: FIB table does not exist");
1006			return -ENOENT;
1007		}
1008
1009		rcu_read_lock();
1010		err = fib_table_dump(tb, skb, cb, &filter);
1011		rcu_read_unlock();
1012		return skb->len ? : err;
1013	}
1014
1015	s_h = cb->args[0];
1016	s_e = cb->args[1];
1017
1018	rcu_read_lock();
1019
1020	for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
1021		e = 0;
1022		head = &net->ipv4.fib_table_hash[h];
1023		hlist_for_each_entry_rcu(tb, head, tb_hlist) {
1024			if (e < s_e)
1025				goto next;
1026			if (dumped)
1027				memset(&cb->args[2], 0, sizeof(cb->args) -
1028						 2 * sizeof(cb->args[0]));
1029			err = fib_table_dump(tb, skb, cb, &filter);
1030			if (err < 0) {
1031				if (likely(skb->len))
1032					goto out;
1033
1034				goto out_err;
1035			}
1036			dumped = 1;
1037next:
1038			e++;
1039		}
1040	}
1041out:
1042	err = skb->len;
1043out_err:
1044	rcu_read_unlock();
1045
1046	cb->args[1] = e;
1047	cb->args[0] = h;
1048
1049	return err;
1050}
1051
1052/* Prepare and feed intra-kernel routing request.
1053 * Really, it should be netlink message, but :-( netlink
1054 * can be not configured, so that we feed it directly
1055 * to fib engine. It is legal, because all events occur
1056 * only when netlink is already locked.
1057 */
1058static void fib_magic(int cmd, int type, __be32 dst, int dst_len,
1059		      struct in_ifaddr *ifa, u32 rt_priority)
1060{
1061	struct net *net = dev_net(ifa->ifa_dev->dev);
1062	u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev);
1063	struct fib_table *tb;
1064	struct fib_config cfg = {
1065		.fc_protocol = RTPROT_KERNEL,
1066		.fc_type = type,
1067		.fc_dst = dst,
1068		.fc_dst_len = dst_len,
1069		.fc_priority = rt_priority,
1070		.fc_prefsrc = ifa->ifa_local,
1071		.fc_oif = ifa->ifa_dev->dev->ifindex,
1072		.fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
1073		.fc_nlinfo = {
1074			.nl_net = net,
1075		},
1076	};
1077
1078	if (!tb_id)
1079		tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
1080
1081	tb = fib_new_table(net, tb_id);
1082	if (!tb)
1083		return;
1084
1085	cfg.fc_table = tb->tb_id;
1086
1087	if (type != RTN_LOCAL)
1088		cfg.fc_scope = RT_SCOPE_LINK;
1089	else
1090		cfg.fc_scope = RT_SCOPE_HOST;
1091
1092	if (cmd == RTM_NEWROUTE)
1093		fib_table_insert(net, tb, &cfg, NULL);
1094	else
1095		fib_table_delete(net, tb, &cfg, NULL);
1096}
1097
1098void fib_add_ifaddr(struct in_ifaddr *ifa)
1099{
1100	struct in_device *in_dev = ifa->ifa_dev;
1101	struct net_device *dev = in_dev->dev;
1102	struct in_ifaddr *prim = ifa;
1103	__be32 mask = ifa->ifa_mask;
1104	__be32 addr = ifa->ifa_local;
1105	__be32 prefix = ifa->ifa_address & mask;
1106
1107	if (ifa->ifa_flags & IFA_F_SECONDARY) {
1108		prim = inet_ifa_byprefix(in_dev, prefix, mask);
1109		if (!prim) {
1110			pr_warn("%s: bug: prim == NULL\n", __func__);
1111			return;
1112		}
1113	}
1114
1115	fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim, 0);
1116
1117	if (!(dev->flags & IFF_UP))
1118		return;
1119
1120	/* Add broadcast address, if it is explicitly assigned. */
1121	if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) {
1122		fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
1123			  prim, 0);
1124		arp_invalidate(dev, ifa->ifa_broadcast, false);
1125	}
1126
1127	if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
1128	    (prefix != addr || ifa->ifa_prefixlen < 32)) {
1129		if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
1130			fib_magic(RTM_NEWROUTE,
1131				  dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1132				  prefix, ifa->ifa_prefixlen, prim,
1133				  ifa->ifa_rt_priority);
1134
1135		/* Add network specific broadcasts, when it takes a sense */
1136		if (ifa->ifa_prefixlen < 31) {
1137			fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32,
1138				  prim, 0);
1139			fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
1140				  32, prim, 0);
1141			arp_invalidate(dev, prefix | ~mask, false);
1142		}
1143	}
1144}
1145
1146void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric)
1147{
1148	__be32 prefix = ifa->ifa_address & ifa->ifa_mask;
1149	struct in_device *in_dev = ifa->ifa_dev;
1150	struct net_device *dev = in_dev->dev;
1151
1152	if (!(dev->flags & IFF_UP) ||
1153	    ifa->ifa_flags & (IFA_F_SECONDARY | IFA_F_NOPREFIXROUTE) ||
1154	    ipv4_is_zeronet(prefix) ||
1155	    (prefix == ifa->ifa_local && ifa->ifa_prefixlen == 32))
1156		return;
1157
1158	/* add the new */
1159	fib_magic(RTM_NEWROUTE,
1160		  dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1161		  prefix, ifa->ifa_prefixlen, ifa, new_metric);
1162
1163	/* delete the old */
1164	fib_magic(RTM_DELROUTE,
1165		  dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1166		  prefix, ifa->ifa_prefixlen, ifa, ifa->ifa_rt_priority);
1167}
1168
1169/* Delete primary or secondary address.
1170 * Optionally, on secondary address promotion consider the addresses
1171 * from subnet iprim as deleted, even if they are in device list.
1172 * In this case the secondary ifa can be in device list.
1173 */
1174void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
1175{
1176	struct in_device *in_dev = ifa->ifa_dev;
1177	struct net_device *dev = in_dev->dev;
1178	struct in_ifaddr *ifa1;
1179	struct in_ifaddr *prim = ifa, *prim1 = NULL;
1180	__be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
1181	__be32 any = ifa->ifa_address & ifa->ifa_mask;
1182#define LOCAL_OK	1
1183#define BRD_OK		2
1184#define BRD0_OK		4
1185#define BRD1_OK		8
1186	unsigned int ok = 0;
1187	int subnet = 0;		/* Primary network */
1188	int gone = 1;		/* Address is missing */
1189	int same_prefsrc = 0;	/* Another primary with same IP */
1190
1191	if (ifa->ifa_flags & IFA_F_SECONDARY) {
1192		prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
1193		if (!prim) {
1194			/* if the device has been deleted, we don't perform
1195			 * address promotion
1196			 */
1197			if (!in_dev->dead)
1198				pr_warn("%s: bug: prim == NULL\n", __func__);
1199			return;
1200		}
1201		if (iprim && iprim != prim) {
1202			pr_warn("%s: bug: iprim != prim\n", __func__);
1203			return;
1204		}
1205	} else if (!ipv4_is_zeronet(any) &&
1206		   (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
1207		if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
1208			fib_magic(RTM_DELROUTE,
1209				  dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1210				  any, ifa->ifa_prefixlen, prim, 0);
1211		subnet = 1;
1212	}
1213
1214	if (in_dev->dead)
1215		goto no_promotions;
1216
1217	/* Deletion is more complicated than add.
1218	 * We should take care of not to delete too much :-)
1219	 *
1220	 * Scan address list to be sure that addresses are really gone.
1221	 */
1222	rcu_read_lock();
1223	in_dev_for_each_ifa_rcu(ifa1, in_dev) {
1224		if (ifa1 == ifa) {
1225			/* promotion, keep the IP */
1226			gone = 0;
1227			continue;
1228		}
1229		/* Ignore IFAs from our subnet */
1230		if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
1231		    inet_ifa_match(ifa1->ifa_address, iprim))
1232			continue;
1233
1234		/* Ignore ifa1 if it uses different primary IP (prefsrc) */
1235		if (ifa1->ifa_flags & IFA_F_SECONDARY) {
1236			/* Another address from our subnet? */
1237			if (ifa1->ifa_mask == prim->ifa_mask &&
1238			    inet_ifa_match(ifa1->ifa_address, prim))
1239				prim1 = prim;
1240			else {
1241				/* We reached the secondaries, so
1242				 * same_prefsrc should be determined.
1243				 */
1244				if (!same_prefsrc)
1245					continue;
1246				/* Search new prim1 if ifa1 is not
1247				 * using the current prim1
1248				 */
1249				if (!prim1 ||
1250				    ifa1->ifa_mask != prim1->ifa_mask ||
1251				    !inet_ifa_match(ifa1->ifa_address, prim1))
1252					prim1 = inet_ifa_byprefix(in_dev,
1253							ifa1->ifa_address,
1254							ifa1->ifa_mask);
1255				if (!prim1)
1256					continue;
1257				if (prim1->ifa_local != prim->ifa_local)
1258					continue;
1259			}
1260		} else {
1261			if (prim->ifa_local != ifa1->ifa_local)
1262				continue;
1263			prim1 = ifa1;
1264			if (prim != prim1)
1265				same_prefsrc = 1;
1266		}
1267		if (ifa->ifa_local == ifa1->ifa_local)
1268			ok |= LOCAL_OK;
1269		if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
1270			ok |= BRD_OK;
1271		if (brd == ifa1->ifa_broadcast)
1272			ok |= BRD1_OK;
1273		if (any == ifa1->ifa_broadcast)
1274			ok |= BRD0_OK;
1275		/* primary has network specific broadcasts */
1276		if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
1277			__be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
1278			__be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
1279
1280			if (!ipv4_is_zeronet(any1)) {
1281				if (ifa->ifa_broadcast == brd1 ||
1282				    ifa->ifa_broadcast == any1)
1283					ok |= BRD_OK;
1284				if (brd == brd1 || brd == any1)
1285					ok |= BRD1_OK;
1286				if (any == brd1 || any == any1)
1287					ok |= BRD0_OK;
1288			}
1289		}
1290	}
1291	rcu_read_unlock();
1292
1293no_promotions:
1294	if (!(ok & BRD_OK))
1295		fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
1296			  prim, 0);
1297	if (subnet && ifa->ifa_prefixlen < 31) {
1298		if (!(ok & BRD1_OK))
1299			fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32,
1300				  prim, 0);
1301		if (!(ok & BRD0_OK))
1302			fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32,
1303				  prim, 0);
1304	}
1305	if (!(ok & LOCAL_OK)) {
1306		unsigned int addr_type;
1307
1308		fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim, 0);
1309
1310		/* Check, that this local address finally disappeared. */
1311		addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1312						     ifa->ifa_local);
1313		if (gone && addr_type != RTN_LOCAL) {
1314			/* And the last, but not the least thing.
1315			 * We must flush stray FIB entries.
1316			 *
1317			 * First of all, we scan fib_info list searching
1318			 * for stray nexthop entries, then ignite fib_flush.
1319			 */
1320			if (fib_sync_down_addr(dev, ifa->ifa_local))
1321				fib_flush(dev_net(dev));
1322		}
1323	}
1324#undef LOCAL_OK
1325#undef BRD_OK
1326#undef BRD0_OK
1327#undef BRD1_OK
1328}
1329
1330static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1331{
1332
1333	struct fib_result       res;
1334	struct flowi4           fl4 = {
1335		.flowi4_mark = frn->fl_mark,
1336		.daddr = frn->fl_addr,
1337		.flowi4_tos = frn->fl_tos,
1338		.flowi4_scope = frn->fl_scope,
1339	};
1340	struct fib_table *tb;
1341
1342	rcu_read_lock();
1343
1344	tb = fib_get_table(net, frn->tb_id_in);
1345
1346	frn->err = -ENOENT;
1347	if (tb) {
1348		local_bh_disable();
1349
1350		frn->tb_id = tb->tb_id;
1351		frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1352
1353		if (!frn->err) {
1354			frn->prefixlen = res.prefixlen;
1355			frn->nh_sel = res.nh_sel;
1356			frn->type = res.type;
1357			frn->scope = res.scope;
1358		}
1359		local_bh_enable();
1360	}
1361
1362	rcu_read_unlock();
1363}
1364
1365static void nl_fib_input(struct sk_buff *skb)
1366{
1367	struct net *net;
1368	struct fib_result_nl *frn;
1369	struct nlmsghdr *nlh;
1370	u32 portid;
1371
1372	net = sock_net(skb->sk);
1373	nlh = nlmsg_hdr(skb);
1374	if (skb->len < nlmsg_total_size(sizeof(*frn)) ||
1375	    skb->len < nlh->nlmsg_len ||
1376	    nlmsg_len(nlh) < sizeof(*frn))
1377		return;
1378
1379	skb = netlink_skb_clone(skb, GFP_KERNEL);
1380	if (!skb)
1381		return;
1382	nlh = nlmsg_hdr(skb);
1383
1384	frn = (struct fib_result_nl *) nlmsg_data(nlh);
1385	nl_fib_lookup(net, frn);
1386
1387	portid = NETLINK_CB(skb).portid;      /* netlink portid */
1388	NETLINK_CB(skb).portid = 0;        /* from kernel */
1389	NETLINK_CB(skb).dst_group = 0;  /* unicast */
1390	netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1391}
1392
1393static int __net_init nl_fib_lookup_init(struct net *net)
1394{
1395	struct sock *sk;
1396	struct netlink_kernel_cfg cfg = {
1397		.input	= nl_fib_input,
1398	};
1399
1400	sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1401	if (!sk)
1402		return -EAFNOSUPPORT;
1403	net->ipv4.fibnl = sk;
1404	return 0;
1405}
1406
1407static void nl_fib_lookup_exit(struct net *net)
1408{
1409	netlink_kernel_release(net->ipv4.fibnl);
1410	net->ipv4.fibnl = NULL;
1411}
1412
1413static void fib_disable_ip(struct net_device *dev, unsigned long event,
1414			   bool force)
1415{
1416	if (fib_sync_down_dev(dev, event, force))
1417		fib_flush(dev_net(dev));
1418	else
1419		rt_cache_flush(dev_net(dev));
1420	arp_ifdown(dev);
1421}
1422
1423static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1424{
1425	struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1426	struct net_device *dev = ifa->ifa_dev->dev;
1427	struct net *net = dev_net(dev);
1428
1429	switch (event) {
1430	case NETDEV_UP:
1431		fib_add_ifaddr(ifa);
1432#ifdef CONFIG_IP_ROUTE_MULTIPATH
1433		fib_sync_up(dev, RTNH_F_DEAD);
1434#endif
1435		atomic_inc(&net->ipv4.dev_addr_genid);
1436		rt_cache_flush(dev_net(dev));
1437		break;
1438	case NETDEV_DOWN:
1439		fib_del_ifaddr(ifa, NULL);
1440		atomic_inc(&net->ipv4.dev_addr_genid);
1441		if (!ifa->ifa_dev->ifa_list) {
1442			/* Last address was deleted from this interface.
1443			 * Disable IP.
1444			 */
1445			fib_disable_ip(dev, event, true);
1446		} else {
1447			rt_cache_flush(dev_net(dev));
1448		}
1449		break;
1450	}
1451	return NOTIFY_DONE;
1452}
1453
1454static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1455{
1456	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1457	struct netdev_notifier_changeupper_info *upper_info = ptr;
1458	struct netdev_notifier_info_ext *info_ext = ptr;
1459	struct in_device *in_dev;
1460	struct net *net = dev_net(dev);
1461	struct in_ifaddr *ifa;
1462	unsigned int flags;
1463
1464	if (event == NETDEV_UNREGISTER) {
1465		fib_disable_ip(dev, event, true);
1466		rt_flush_dev(dev);
1467		return NOTIFY_DONE;
1468	}
1469
1470	in_dev = __in_dev_get_rtnl(dev);
1471	if (!in_dev)
1472		return NOTIFY_DONE;
1473
1474	switch (event) {
1475	case NETDEV_UP:
1476		in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1477			fib_add_ifaddr(ifa);
1478		}
1479#ifdef CONFIG_IP_ROUTE_MULTIPATH
1480		fib_sync_up(dev, RTNH_F_DEAD);
1481#endif
1482		atomic_inc(&net->ipv4.dev_addr_genid);
1483		rt_cache_flush(net);
1484		break;
1485	case NETDEV_DOWN:
1486		fib_disable_ip(dev, event, false);
1487		break;
1488	case NETDEV_CHANGE:
1489		flags = dev_get_flags(dev);
1490		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1491			fib_sync_up(dev, RTNH_F_LINKDOWN);
1492		else
1493			fib_sync_down_dev(dev, event, false);
1494		rt_cache_flush(net);
1495		break;
1496	case NETDEV_CHANGEMTU:
1497		fib_sync_mtu(dev, info_ext->ext.mtu);
1498		rt_cache_flush(net);
1499		break;
1500	case NETDEV_CHANGEUPPER:
1501		upper_info = ptr;
1502		/* flush all routes if dev is linked to or unlinked from
1503		 * an L3 master device (e.g., VRF)
1504		 */
1505		if (upper_info->upper_dev &&
1506		    netif_is_l3_master(upper_info->upper_dev))
1507			fib_disable_ip(dev, NETDEV_DOWN, true);
1508		break;
1509	}
1510	return NOTIFY_DONE;
1511}
1512
1513static struct notifier_block fib_inetaddr_notifier = {
1514	.notifier_call = fib_inetaddr_event,
1515};
1516
1517static struct notifier_block fib_netdev_notifier = {
1518	.notifier_call = fib_netdev_event,
1519};
1520
1521static int __net_init ip_fib_net_init(struct net *net)
1522{
1523	int err;
1524	size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1525
1526	err = fib4_notifier_init(net);
1527	if (err)
1528		return err;
1529
1530	/* Avoid false sharing : Use at least a full cache line */
1531	size = max_t(size_t, size, L1_CACHE_BYTES);
1532
1533	net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1534	if (!net->ipv4.fib_table_hash) {
1535		err = -ENOMEM;
1536		goto err_table_hash_alloc;
1537	}
1538
1539	err = fib4_rules_init(net);
1540	if (err < 0)
1541		goto err_rules_init;
1542	return 0;
1543
1544err_rules_init:
1545	kfree(net->ipv4.fib_table_hash);
1546err_table_hash_alloc:
1547	fib4_notifier_exit(net);
1548	return err;
1549}
1550
1551static void ip_fib_net_exit(struct net *net)
1552{
1553	int i;
1554
1555	rtnl_lock();
1556#ifdef CONFIG_IP_MULTIPLE_TABLES
1557	RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1558	RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1559#endif
1560	/* Destroy the tables in reverse order to guarantee that the
1561	 * local table, ID 255, is destroyed before the main table, ID
1562	 * 254. This is necessary as the local table may contain
1563	 * references to data contained in the main table.
1564	 */
1565	for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) {
1566		struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1567		struct hlist_node *tmp;
1568		struct fib_table *tb;
1569
1570		hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1571			hlist_del(&tb->tb_hlist);
1572			fib_table_flush(net, tb, true);
1573			fib_free_table(tb);
1574		}
1575	}
1576
1577#ifdef CONFIG_IP_MULTIPLE_TABLES
1578	fib4_rules_exit(net);
1579#endif
1580	rtnl_unlock();
1581	kfree(net->ipv4.fib_table_hash);
1582	fib4_notifier_exit(net);
1583}
1584
1585static int __net_init fib_net_init(struct net *net)
1586{
1587	int error;
1588
1589#ifdef CONFIG_IP_ROUTE_CLASSID
1590	atomic_set(&net->ipv4.fib_num_tclassid_users, 0);
1591#endif
1592	error = ip_fib_net_init(net);
1593	if (error < 0)
1594		goto out;
1595	error = nl_fib_lookup_init(net);
1596	if (error < 0)
1597		goto out_nlfl;
1598	error = fib_proc_init(net);
1599	if (error < 0)
1600		goto out_proc;
1601out:
1602	return error;
1603
1604out_proc:
1605	nl_fib_lookup_exit(net);
1606out_nlfl:
1607	ip_fib_net_exit(net);
1608	goto out;
1609}
1610
1611static void __net_exit fib_net_exit(struct net *net)
1612{
1613	fib_proc_exit(net);
1614	nl_fib_lookup_exit(net);
1615	ip_fib_net_exit(net);
1616}
1617
1618static struct pernet_operations fib_net_ops = {
1619	.init = fib_net_init,
1620	.exit = fib_net_exit,
1621};
1622
1623void __init ip_fib_init(void)
1624{
1625	fib_trie_init();
1626
1627	register_pernet_subsys(&fib_net_ops);
1628
1629	register_netdevice_notifier(&fib_netdev_notifier);
1630	register_inetaddr_notifier(&fib_inetaddr_notifier);
1631
1632	rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, 0);
1633	rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, 0);
1634	rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, 0);
1635}
1636