xref: /linux/net/ipv4/fib_semantics.c (revision 32a92f8c89326985e05dce8b22d3f0aa07a3e1bd)
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: semantics.
8  *
9  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10  */
11 
12 #include <linux/uaccess.h>
13 #include <linux/bitops.h>
14 #include <linux/types.h>
15 #include <linux/kernel.h>
16 #include <linux/jiffies.h>
17 #include <linux/mm.h>
18 #include <linux/string.h>
19 #include <linux/socket.h>
20 #include <linux/sockios.h>
21 #include <linux/errno.h>
22 #include <linux/in.h>
23 #include <linux/inet.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <linux/if_arp.h>
27 #include <linux/proc_fs.h>
28 #include <linux/skbuff.h>
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/netlink.h>
32 #include <linux/hash.h>
33 #include <linux/nospec.h>
34 
35 #include <net/arp.h>
36 #include <net/inet_dscp.h>
37 #include <net/ip.h>
38 #include <net/protocol.h>
39 #include <net/route.h>
40 #include <net/tcp.h>
41 #include <net/sock.h>
42 #include <net/ip_fib.h>
43 #include <net/ip6_fib.h>
44 #include <net/nexthop.h>
45 #include <net/netlink.h>
46 #include <net/rtnh.h>
47 #include <net/lwtunnel.h>
48 #include <net/fib_notifier.h>
49 #include <net/addrconf.h>
50 
51 #include "fib_lookup.h"
52 
53 /* for_nexthops and change_nexthops only used when nexthop object
54  * is not set in a fib_info. The logic within can reference fib_nh.
55  */
56 #ifdef CONFIG_IP_ROUTE_MULTIPATH
57 
58 #define for_nexthops(fi) {						\
59 	int nhsel; const struct fib_nh *nh;				\
60 	for (nhsel = 0, nh = (fi)->fib_nh;				\
61 	     nhsel < fib_info_num_path((fi));				\
62 	     nh++, nhsel++)
63 
64 #define change_nexthops(fi) {						\
65 	int nhsel; struct fib_nh *nexthop_nh;				\
66 	for (nhsel = 0,	nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
67 	     nhsel < fib_info_num_path((fi));				\
68 	     nexthop_nh++, nhsel++)
69 
70 #else /* CONFIG_IP_ROUTE_MULTIPATH */
71 
72 /* Hope, that gcc will optimize it to get rid of dummy loop */
73 
74 #define for_nexthops(fi) {						\
75 	int nhsel; const struct fib_nh *nh = (fi)->fib_nh;		\
76 	for (nhsel = 0; nhsel < 1; nhsel++)
77 
78 #define change_nexthops(fi) {						\
79 	int nhsel;							\
80 	struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
81 	for (nhsel = 0; nhsel < 1; nhsel++)
82 
83 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
84 
85 #define endfor_nexthops(fi) }
86 
87 
88 const struct fib_prop fib_props[RTN_MAX + 1] = {
89 	[RTN_UNSPEC] = {
90 		.error	= 0,
91 		.scope	= RT_SCOPE_NOWHERE,
92 	},
93 	[RTN_UNICAST] = {
94 		.error	= 0,
95 		.scope	= RT_SCOPE_UNIVERSE,
96 	},
97 	[RTN_LOCAL] = {
98 		.error	= 0,
99 		.scope	= RT_SCOPE_HOST,
100 	},
101 	[RTN_BROADCAST] = {
102 		.error	= 0,
103 		.scope	= RT_SCOPE_LINK,
104 	},
105 	[RTN_ANYCAST] = {
106 		.error	= 0,
107 		.scope	= RT_SCOPE_LINK,
108 	},
109 	[RTN_MULTICAST] = {
110 		.error	= 0,
111 		.scope	= RT_SCOPE_UNIVERSE,
112 	},
113 	[RTN_BLACKHOLE] = {
114 		.error	= -EINVAL,
115 		.scope	= RT_SCOPE_UNIVERSE,
116 	},
117 	[RTN_UNREACHABLE] = {
118 		.error	= -EHOSTUNREACH,
119 		.scope	= RT_SCOPE_UNIVERSE,
120 	},
121 	[RTN_PROHIBIT] = {
122 		.error	= -EACCES,
123 		.scope	= RT_SCOPE_UNIVERSE,
124 	},
125 	[RTN_THROW] = {
126 		.error	= -EAGAIN,
127 		.scope	= RT_SCOPE_UNIVERSE,
128 	},
129 	[RTN_NAT] = {
130 		.error	= -EINVAL,
131 		.scope	= RT_SCOPE_NOWHERE,
132 	},
133 	[RTN_XRESOLVE] = {
134 		.error	= -EINVAL,
135 		.scope	= RT_SCOPE_NOWHERE,
136 	},
137 };
138 
rt_fibinfo_free(struct rtable __rcu ** rtp)139 static void rt_fibinfo_free(struct rtable __rcu **rtp)
140 {
141 	struct rtable *rt = rcu_dereference_protected(*rtp, 1);
142 
143 	if (!rt)
144 		return;
145 
146 	/* Not even needed : RCU_INIT_POINTER(*rtp, NULL);
147 	 * because we waited an RCU grace period before calling
148 	 * free_fib_info_rcu()
149 	 */
150 
151 	dst_dev_put(&rt->dst);
152 	dst_release_immediate(&rt->dst);
153 }
154 
free_nh_exceptions(struct fib_nh_common * nhc)155 static void free_nh_exceptions(struct fib_nh_common *nhc)
156 {
157 	struct fnhe_hash_bucket *hash;
158 	int i;
159 
160 	hash = rcu_dereference_protected(nhc->nhc_exceptions, 1);
161 	if (!hash)
162 		return;
163 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
164 		struct fib_nh_exception *fnhe;
165 
166 		fnhe = rcu_dereference_protected(hash[i].chain, 1);
167 		while (fnhe) {
168 			struct fib_nh_exception *next;
169 
170 			next = rcu_dereference_protected(fnhe->fnhe_next, 1);
171 
172 			rt_fibinfo_free(&fnhe->fnhe_rth_input);
173 			rt_fibinfo_free(&fnhe->fnhe_rth_output);
174 
175 			kfree(fnhe);
176 
177 			fnhe = next;
178 		}
179 	}
180 	kfree(hash);
181 }
182 
rt_fibinfo_free_cpus(struct rtable __rcu * __percpu * rtp)183 static void rt_fibinfo_free_cpus(struct rtable __rcu * __percpu *rtp)
184 {
185 	int cpu;
186 
187 	if (!rtp)
188 		return;
189 
190 	for_each_possible_cpu(cpu) {
191 		struct rtable *rt;
192 
193 		rt = rcu_dereference_protected(*per_cpu_ptr(rtp, cpu), 1);
194 		if (rt) {
195 			dst_dev_put(&rt->dst);
196 			dst_release_immediate(&rt->dst);
197 		}
198 	}
199 	free_percpu(rtp);
200 }
201 
fib_nh_common_release(struct fib_nh_common * nhc)202 void fib_nh_common_release(struct fib_nh_common *nhc)
203 {
204 	netdev_put(nhc->nhc_dev, &nhc->nhc_dev_tracker);
205 	lwtstate_put(nhc->nhc_lwtstate);
206 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
207 	rt_fibinfo_free(&nhc->nhc_rth_input);
208 	free_nh_exceptions(nhc);
209 }
210 EXPORT_SYMBOL_GPL(fib_nh_common_release);
211 
fib_nh_release(struct net * net,struct fib_nh * fib_nh)212 void fib_nh_release(struct net *net, struct fib_nh *fib_nh)
213 {
214 #ifdef CONFIG_IP_ROUTE_CLASSID
215 	if (fib_nh->nh_tclassid)
216 		atomic_dec(&net->ipv4.fib_num_tclassid_users);
217 #endif
218 	fib_nh_common_release(&fib_nh->nh_common);
219 }
220 
221 /* Release a nexthop info record */
free_fib_info_rcu(struct rcu_head * head)222 static void free_fib_info_rcu(struct rcu_head *head)
223 {
224 	struct fib_info *fi = container_of(head, struct fib_info, rcu);
225 
226 	if (fi->nh) {
227 		nexthop_put(fi->nh);
228 	} else {
229 		change_nexthops(fi) {
230 			fib_nh_release(fi->fib_net, nexthop_nh);
231 		} endfor_nexthops(fi);
232 	}
233 
234 	ip_fib_metrics_put(fi->fib_metrics);
235 
236 	kfree(fi);
237 }
238 
free_fib_info(struct fib_info * fi)239 void free_fib_info(struct fib_info *fi)
240 {
241 	if (fi->fib_dead == 0) {
242 		pr_warn("Freeing alive fib_info %p\n", fi);
243 		return;
244 	}
245 
246 	call_rcu_hurry(&fi->rcu, free_fib_info_rcu);
247 }
248 EXPORT_SYMBOL_GPL(free_fib_info);
249 
fib_release_info(struct fib_info * fi)250 void fib_release_info(struct fib_info *fi)
251 {
252 	ASSERT_RTNL();
253 	if (fi && refcount_dec_and_test(&fi->fib_treeref)) {
254 		hlist_del(&fi->fib_hash);
255 		fi->fib_net->ipv4.fib_info_cnt--;
256 
257 		if (fi->fib_prefsrc)
258 			hlist_del(&fi->fib_lhash);
259 		if (fi->nh) {
260 			list_del(&fi->nh_list);
261 		} else {
262 			change_nexthops(fi) {
263 				if (!nexthop_nh->fib_nh_dev)
264 					continue;
265 				hlist_del_rcu(&nexthop_nh->nh_hash);
266 			} endfor_nexthops(fi)
267 		}
268 		/* Paired with READ_ONCE() from fib_table_lookup() */
269 		WRITE_ONCE(fi->fib_dead, 1);
270 		fib_info_put(fi);
271 	}
272 }
273 
nh_comp(struct fib_info * fi,struct fib_info * ofi)274 static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi)
275 {
276 	const struct fib_nh *onh;
277 
278 	if (fi->nh || ofi->nh)
279 		return nexthop_cmp(fi->nh, ofi->nh) ? 0 : -1;
280 
281 	if (ofi->fib_nhs == 0)
282 		return 0;
283 
284 	for_nexthops(fi) {
285 		onh = fib_info_nh(ofi, nhsel);
286 
287 		if (nh->fib_nh_oif != onh->fib_nh_oif ||
288 		    nh->fib_nh_gw_family != onh->fib_nh_gw_family ||
289 		    nh->fib_nh_scope != onh->fib_nh_scope ||
290 #ifdef CONFIG_IP_ROUTE_MULTIPATH
291 		    nh->fib_nh_weight != onh->fib_nh_weight ||
292 #endif
293 #ifdef CONFIG_IP_ROUTE_CLASSID
294 		    nh->nh_tclassid != onh->nh_tclassid ||
295 #endif
296 		    lwtunnel_cmp_encap(nh->fib_nh_lws, onh->fib_nh_lws) ||
297 		    ((nh->fib_nh_flags ^ onh->fib_nh_flags) & ~RTNH_COMPARE_MASK))
298 			return -1;
299 
300 		if (nh->fib_nh_gw_family == AF_INET &&
301 		    nh->fib_nh_gw4 != onh->fib_nh_gw4)
302 			return -1;
303 
304 		if (nh->fib_nh_gw_family == AF_INET6 &&
305 		    ipv6_addr_cmp(&nh->fib_nh_gw6, &onh->fib_nh_gw6))
306 			return -1;
307 	} endfor_nexthops(fi);
308 	return 0;
309 }
310 
fib_nh_head(struct net_device * dev)311 static struct hlist_head *fib_nh_head(struct net_device *dev)
312 {
313 	return &dev->fib_nh_head;
314 }
315 
fib_info_hashfn_1(int init_val,u8 protocol,u8 scope,u32 prefsrc,u32 priority)316 static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope,
317 				      u32 prefsrc, u32 priority)
318 {
319 	unsigned int val = init_val;
320 
321 	val ^= (protocol << 8) | scope;
322 	val ^= prefsrc;
323 	val ^= priority;
324 
325 	return val;
326 }
327 
fib_info_hashfn_result(const struct net * net,unsigned int val)328 static unsigned int fib_info_hashfn_result(const struct net *net,
329 					   unsigned int val)
330 {
331 	return hash_32(val ^ net_hash_mix(net), net->ipv4.fib_info_hash_bits);
332 }
333 
fib_info_hash_bucket(struct fib_info * fi)334 static struct hlist_head *fib_info_hash_bucket(struct fib_info *fi)
335 {
336 	struct net *net = fi->fib_net;
337 	unsigned int val;
338 
339 	val = fib_info_hashfn_1(fi->fib_nhs, fi->fib_protocol,
340 				fi->fib_scope, (__force u32)fi->fib_prefsrc,
341 				fi->fib_priority);
342 
343 	if (fi->nh) {
344 		val ^= fi->nh->id;
345 	} else {
346 		for_nexthops(fi) {
347 			val ^= nh->fib_nh_oif;
348 		} endfor_nexthops(fi)
349 	}
350 
351 	return &net->ipv4.fib_info_hash[fib_info_hashfn_result(net, val)];
352 }
353 
fib_info_laddrhash_bucket(const struct net * net,__be32 val)354 static struct hlist_head *fib_info_laddrhash_bucket(const struct net *net,
355 						    __be32 val)
356 {
357 	unsigned int hash_bits = net->ipv4.fib_info_hash_bits;
358 	u32 slot;
359 
360 	slot = hash_32(net_hash_mix(net) ^ (__force u32)val, hash_bits);
361 
362 	return &net->ipv4.fib_info_hash[(1 << hash_bits) + slot];
363 }
364 
fib_info_hash_alloc(unsigned int hash_bits)365 static struct hlist_head *fib_info_hash_alloc(unsigned int hash_bits)
366 {
367 	/* The second half is used for prefsrc */
368 	return kvzalloc_objs(struct hlist_head, (1 << hash_bits) * 2);
369 }
370 
fib_info_hash_free(struct hlist_head * head)371 static void fib_info_hash_free(struct hlist_head *head)
372 {
373 	kvfree(head);
374 }
375 
fib_info_hash_grow(struct net * net)376 static void fib_info_hash_grow(struct net *net)
377 {
378 	unsigned int old_size = 1 << net->ipv4.fib_info_hash_bits;
379 	struct hlist_head *new_info_hash, *old_info_hash;
380 	unsigned int i;
381 
382 	if (net->ipv4.fib_info_cnt < old_size)
383 		return;
384 
385 	new_info_hash = fib_info_hash_alloc(net->ipv4.fib_info_hash_bits + 1);
386 	if (!new_info_hash)
387 		return;
388 
389 	old_info_hash = net->ipv4.fib_info_hash;
390 	net->ipv4.fib_info_hash = new_info_hash;
391 	net->ipv4.fib_info_hash_bits += 1;
392 
393 	for (i = 0; i < old_size; i++) {
394 		struct hlist_head *head = &old_info_hash[i];
395 		struct hlist_node *n;
396 		struct fib_info *fi;
397 
398 		hlist_for_each_entry_safe(fi, n, head, fib_hash)
399 			hlist_add_head(&fi->fib_hash, fib_info_hash_bucket(fi));
400 	}
401 
402 	for (i = 0; i < old_size; i++) {
403 		struct hlist_head *lhead = &old_info_hash[old_size + i];
404 		struct hlist_node *n;
405 		struct fib_info *fi;
406 
407 		hlist_for_each_entry_safe(fi, n, lhead, fib_lhash)
408 			hlist_add_head(&fi->fib_lhash,
409 				       fib_info_laddrhash_bucket(fi->fib_net,
410 								 fi->fib_prefsrc));
411 	}
412 
413 	fib_info_hash_free(old_info_hash);
414 }
415 
416 /* no metrics, only nexthop id */
fib_find_info_nh(struct net * net,const struct fib_config * cfg)417 static struct fib_info *fib_find_info_nh(struct net *net,
418 					 const struct fib_config *cfg)
419 {
420 	struct hlist_head *head;
421 	struct fib_info *fi;
422 	unsigned int hash;
423 
424 	hash = fib_info_hashfn_1(cfg->fc_nh_id,
425 				 cfg->fc_protocol, cfg->fc_scope,
426 				 (__force u32)cfg->fc_prefsrc,
427 				 cfg->fc_priority);
428 	hash = fib_info_hashfn_result(net, hash);
429 	head = &net->ipv4.fib_info_hash[hash];
430 
431 	hlist_for_each_entry(fi, head, fib_hash) {
432 		if (!fi->nh || fi->nh->id != cfg->fc_nh_id)
433 			continue;
434 
435 		if (cfg->fc_protocol == fi->fib_protocol &&
436 		    cfg->fc_scope == fi->fib_scope &&
437 		    cfg->fc_prefsrc == fi->fib_prefsrc &&
438 		    cfg->fc_priority == fi->fib_priority &&
439 		    cfg->fc_type == fi->fib_type &&
440 		    cfg->fc_table == fi->fib_tb_id &&
441 		    !((cfg->fc_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK))
442 			return fi;
443 	}
444 
445 	return NULL;
446 }
447 
fib_find_info(struct fib_info * nfi)448 static struct fib_info *fib_find_info(struct fib_info *nfi)
449 {
450 	struct hlist_head *head = fib_info_hash_bucket(nfi);
451 	struct fib_info *fi;
452 
453 	hlist_for_each_entry(fi, head, fib_hash) {
454 		if (fi->fib_nhs != nfi->fib_nhs)
455 			continue;
456 
457 		if (nfi->fib_protocol == fi->fib_protocol &&
458 		    nfi->fib_scope == fi->fib_scope &&
459 		    nfi->fib_prefsrc == fi->fib_prefsrc &&
460 		    nfi->fib_priority == fi->fib_priority &&
461 		    nfi->fib_type == fi->fib_type &&
462 		    nfi->fib_tb_id == fi->fib_tb_id &&
463 		    memcmp(nfi->fib_metrics, fi->fib_metrics,
464 			   sizeof(u32) * RTAX_MAX) == 0 &&
465 		    !((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK) &&
466 		    nh_comp(fi, nfi) == 0)
467 			return fi;
468 	}
469 
470 	return NULL;
471 }
472 
473 /* Check, that the gateway is already configured.
474  * Used only by redirect accept routine, under rcu_read_lock();
475  */
ip_fib_check_default(__be32 gw,struct net_device * dev)476 int ip_fib_check_default(__be32 gw, struct net_device *dev)
477 {
478 	struct hlist_head *head;
479 	struct fib_nh *nh;
480 
481 	head = fib_nh_head(dev);
482 
483 	hlist_for_each_entry_rcu(nh, head, nh_hash) {
484 		DEBUG_NET_WARN_ON_ONCE(nh->fib_nh_dev != dev);
485 		if (nh->fib_nh_gw4 == gw &&
486 		    !(nh->fib_nh_flags & RTNH_F_DEAD)) {
487 			return 0;
488 		}
489 	}
490 
491 	return -1;
492 }
493 
fib_nlmsg_size(struct fib_info * fi)494 size_t fib_nlmsg_size(struct fib_info *fi)
495 {
496 	size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
497 			 + nla_total_size(4) /* RTA_TABLE */
498 			 + nla_total_size(4) /* RTA_DST */
499 			 + nla_total_size(4) /* RTA_PRIORITY */
500 			 + nla_total_size(4) /* RTA_PREFSRC */
501 			 + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
502 	unsigned int nhs = fib_info_num_path(fi);
503 
504 	/* space for nested metrics */
505 	payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
506 
507 	if (fi->nh)
508 		payload += nla_total_size(4); /* RTA_NH_ID */
509 
510 	if (nhs) {
511 		size_t nh_encapsize = 0;
512 		/* Also handles the special case nhs == 1 */
513 
514 		/* each nexthop is packed in an attribute */
515 		size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
516 		unsigned int i;
517 
518 		/* may contain flow and gateway attribute */
519 		nhsize += 2 * nla_total_size(4);
520 
521 		/* grab encap info */
522 		for (i = 0; i < fib_info_num_path(fi); i++) {
523 			struct fib_nh_common *nhc = fib_info_nhc(fi, i);
524 
525 			if (nhc->nhc_lwtstate) {
526 				/* RTA_ENCAP_TYPE */
527 				nh_encapsize += lwtunnel_get_encap_size(
528 						nhc->nhc_lwtstate);
529 				/* RTA_ENCAP */
530 				nh_encapsize +=  nla_total_size(2);
531 			}
532 		}
533 
534 		/* all nexthops are packed in a nested attribute */
535 		payload += nla_total_size((nhs * nhsize) + nh_encapsize);
536 
537 	}
538 
539 	return payload;
540 }
541 
rtmsg_fib(int event,__be32 key,struct fib_alias * fa,int dst_len,u32 tb_id,const struct nl_info * info,unsigned int nlm_flags)542 void rtmsg_fib(int event, __be32 key, struct fib_alias *fa,
543 	       int dst_len, u32 tb_id, const struct nl_info *info,
544 	       unsigned int nlm_flags)
545 {
546 	struct fib_rt_info fri;
547 	struct sk_buff *skb;
548 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
549 	int err = -ENOBUFS;
550 
551 	skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL);
552 	if (!skb)
553 		goto errout;
554 
555 	fri.fi = fa->fa_info;
556 	fri.tb_id = tb_id;
557 	fri.dst = key;
558 	fri.dst_len = dst_len;
559 	fri.dscp = fa->fa_dscp;
560 	fri.type = fa->fa_type;
561 	fri.offload = READ_ONCE(fa->offload);
562 	fri.trap = READ_ONCE(fa->trap);
563 	fri.offload_failed = READ_ONCE(fa->offload_failed);
564 	err = fib_dump_info(skb, info->portid, seq, event, &fri, nlm_flags);
565 	if (err < 0) {
566 		/* -EMSGSIZE implies BUG in fib_nlmsg_size() */
567 		WARN_ON(err == -EMSGSIZE);
568 		kfree_skb(skb);
569 		goto errout;
570 	}
571 	rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_IPV4_ROUTE,
572 		    info->nlh, GFP_KERNEL);
573 	return;
574 errout:
575 	rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err);
576 }
577 
fib_detect_death(struct fib_info * fi,int order,struct fib_info ** last_resort,int * last_idx,int dflt)578 static int fib_detect_death(struct fib_info *fi, int order,
579 			    struct fib_info **last_resort, int *last_idx,
580 			    int dflt)
581 {
582 	const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
583 	struct neighbour *n;
584 	int state = NUD_NONE;
585 
586 	if (likely(nhc->nhc_gw_family == AF_INET))
587 		n = neigh_lookup(&arp_tbl, &nhc->nhc_gw.ipv4, nhc->nhc_dev);
588 	else if (nhc->nhc_gw_family == AF_INET6)
589 		n = neigh_lookup(ipv6_stub->nd_tbl, &nhc->nhc_gw.ipv6,
590 				 nhc->nhc_dev);
591 	else
592 		n = NULL;
593 
594 	if (n) {
595 		state = READ_ONCE(n->nud_state);
596 		neigh_release(n);
597 	} else {
598 		return 0;
599 	}
600 	if (state == NUD_REACHABLE)
601 		return 0;
602 	if ((state & NUD_VALID) && order != dflt)
603 		return 0;
604 	if ((state & NUD_VALID) ||
605 	    (*last_idx < 0 && order > dflt && state != NUD_INCOMPLETE)) {
606 		*last_resort = fi;
607 		*last_idx = order;
608 	}
609 	return 1;
610 }
611 
fib_nh_common_init(struct net * net,struct fib_nh_common * nhc,struct nlattr * encap,u16 encap_type,void * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)612 int fib_nh_common_init(struct net *net, struct fib_nh_common *nhc,
613 		       struct nlattr *encap, u16 encap_type,
614 		       void *cfg, gfp_t gfp_flags,
615 		       struct netlink_ext_ack *extack)
616 {
617 	int err;
618 
619 	nhc->nhc_pcpu_rth_output = alloc_percpu_gfp(struct rtable __rcu *,
620 						    gfp_flags);
621 	if (!nhc->nhc_pcpu_rth_output)
622 		return -ENOMEM;
623 
624 	if (encap) {
625 		struct lwtunnel_state *lwtstate;
626 
627 		err = lwtunnel_build_state(net, encap_type, encap,
628 					   nhc->nhc_family, cfg, &lwtstate,
629 					   extack);
630 		if (err)
631 			goto lwt_failure;
632 
633 		nhc->nhc_lwtstate = lwtstate_get(lwtstate);
634 	}
635 
636 	return 0;
637 
638 lwt_failure:
639 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
640 	nhc->nhc_pcpu_rth_output = NULL;
641 	return err;
642 }
643 EXPORT_SYMBOL_GPL(fib_nh_common_init);
644 
fib_nh_init(struct net * net,struct fib_nh * nh,struct fib_config * cfg,int nh_weight,struct netlink_ext_ack * extack)645 int fib_nh_init(struct net *net, struct fib_nh *nh,
646 		struct fib_config *cfg, int nh_weight,
647 		struct netlink_ext_ack *extack)
648 {
649 	int err;
650 
651 	nh->fib_nh_family = AF_INET;
652 
653 	err = fib_nh_common_init(net, &nh->nh_common, cfg->fc_encap,
654 				 cfg->fc_encap_type, cfg, GFP_KERNEL, extack);
655 	if (err)
656 		return err;
657 
658 	nh->fib_nh_oif = cfg->fc_oif;
659 	nh->fib_nh_gw_family = cfg->fc_gw_family;
660 	if (cfg->fc_gw_family == AF_INET)
661 		nh->fib_nh_gw4 = cfg->fc_gw4;
662 	else if (cfg->fc_gw_family == AF_INET6)
663 		nh->fib_nh_gw6 = cfg->fc_gw6;
664 
665 	nh->fib_nh_flags = cfg->fc_flags;
666 
667 #ifdef CONFIG_IP_ROUTE_CLASSID
668 	nh->nh_tclassid = cfg->fc_flow;
669 	if (nh->nh_tclassid)
670 		atomic_inc(&net->ipv4.fib_num_tclassid_users);
671 #endif
672 #ifdef CONFIG_IP_ROUTE_MULTIPATH
673 	nh->fib_nh_weight = nh_weight;
674 #endif
675 	return 0;
676 }
677 
678 #ifdef CONFIG_IP_ROUTE_MULTIPATH
679 
fib_count_nexthops(struct rtnexthop * rtnh,int remaining,struct netlink_ext_ack * extack)680 static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining,
681 			      struct netlink_ext_ack *extack)
682 {
683 	int nhs = 0;
684 
685 	while (rtnh_ok(rtnh, remaining)) {
686 		nhs++;
687 		rtnh = rtnh_next(rtnh, &remaining);
688 	}
689 
690 	/* leftover implies invalid nexthop configuration, discard it */
691 	if (remaining > 0) {
692 		NL_SET_ERR_MSG(extack,
693 			       "Invalid nexthop configuration - extra data after nexthops");
694 		nhs = 0;
695 	}
696 
697 	return nhs;
698 }
699 
fib_gw_from_attr(__be32 * gw,struct nlattr * nla,struct netlink_ext_ack * extack)700 static int fib_gw_from_attr(__be32 *gw, struct nlattr *nla,
701 			    struct netlink_ext_ack *extack)
702 {
703 	if (nla_len(nla) < sizeof(*gw)) {
704 		NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_GATEWAY");
705 		return -EINVAL;
706 	}
707 
708 	*gw = nla_get_in_addr(nla);
709 
710 	return 0;
711 }
712 
713 /* only called when fib_nh is integrated into fib_info */
fib_get_nhs(struct fib_info * fi,struct rtnexthop * rtnh,int remaining,struct fib_config * cfg,struct netlink_ext_ack * extack)714 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
715 		       int remaining, struct fib_config *cfg,
716 		       struct netlink_ext_ack *extack)
717 {
718 	struct net *net = fi->fib_net;
719 	struct fib_config fib_cfg;
720 	struct fib_nh *nh;
721 	int ret;
722 
723 	change_nexthops(fi) {
724 		int attrlen;
725 
726 		memset(&fib_cfg, 0, sizeof(fib_cfg));
727 
728 		if (!rtnh_ok(rtnh, remaining)) {
729 			NL_SET_ERR_MSG(extack,
730 				       "Invalid nexthop configuration - extra data after nexthop");
731 			return -EINVAL;
732 		}
733 
734 		if (rtnh->rtnh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
735 			NL_SET_ERR_MSG(extack,
736 				       "Invalid flags for nexthop - can not contain DEAD or LINKDOWN");
737 			return -EINVAL;
738 		}
739 
740 		fib_cfg.fc_flags = (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags;
741 		fib_cfg.fc_oif = rtnh->rtnh_ifindex;
742 
743 		attrlen = rtnh_attrlen(rtnh);
744 		if (attrlen > 0) {
745 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
746 
747 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
748 			nlav = nla_find(attrs, attrlen, RTA_VIA);
749 			if (nla && nlav) {
750 				NL_SET_ERR_MSG(extack,
751 					       "Nexthop configuration can not contain both GATEWAY and VIA");
752 				return -EINVAL;
753 			}
754 			if (nla) {
755 				ret = fib_gw_from_attr(&fib_cfg.fc_gw4, nla,
756 						       extack);
757 				if (ret)
758 					goto errout;
759 
760 				if (fib_cfg.fc_gw4)
761 					fib_cfg.fc_gw_family = AF_INET;
762 			} else if (nlav) {
763 				ret = fib_gw_from_via(&fib_cfg, nlav, extack);
764 				if (ret)
765 					goto errout;
766 			}
767 
768 			nla = nla_find(attrs, attrlen, RTA_FLOW);
769 			if (nla) {
770 				if (nla_len(nla) < sizeof(u32)) {
771 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
772 					return -EINVAL;
773 				}
774 				fib_cfg.fc_flow = nla_get_u32(nla);
775 			}
776 
777 			fib_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
778 			/* RTA_ENCAP_TYPE length checked in
779 			 * lwtunnel_valid_encap_type_attr
780 			 */
781 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
782 			if (nla)
783 				fib_cfg.fc_encap_type = nla_get_u16(nla);
784 		}
785 
786 		ret = fib_nh_init(net, nexthop_nh, &fib_cfg,
787 				  rtnh->rtnh_hops + 1, extack);
788 		if (ret)
789 			goto errout;
790 
791 		rtnh = rtnh_next(rtnh, &remaining);
792 	} endfor_nexthops(fi);
793 
794 	ret = -EINVAL;
795 	nh = fib_info_nh(fi, 0);
796 	if (cfg->fc_oif && nh->fib_nh_oif != cfg->fc_oif) {
797 		NL_SET_ERR_MSG(extack,
798 			       "Nexthop device index does not match RTA_OIF");
799 		goto errout;
800 	}
801 	if (cfg->fc_gw_family) {
802 		if (cfg->fc_gw_family != nh->fib_nh_gw_family ||
803 		    (cfg->fc_gw_family == AF_INET &&
804 		     nh->fib_nh_gw4 != cfg->fc_gw4) ||
805 		    (cfg->fc_gw_family == AF_INET6 &&
806 		     ipv6_addr_cmp(&nh->fib_nh_gw6, &cfg->fc_gw6))) {
807 			NL_SET_ERR_MSG(extack,
808 				       "Nexthop gateway does not match RTA_GATEWAY or RTA_VIA");
809 			goto errout;
810 		}
811 	}
812 #ifdef CONFIG_IP_ROUTE_CLASSID
813 	if (cfg->fc_flow && nh->nh_tclassid != cfg->fc_flow) {
814 		NL_SET_ERR_MSG(extack,
815 			       "Nexthop class id does not match RTA_FLOW");
816 		goto errout;
817 	}
818 #endif
819 	ret = 0;
820 errout:
821 	return ret;
822 }
823 
824 /* only called when fib_nh is integrated into fib_info */
fib_rebalance(struct fib_info * fi)825 static void fib_rebalance(struct fib_info *fi)
826 {
827 	int total;
828 	int w;
829 
830 	if (fib_info_num_path(fi) < 2)
831 		return;
832 
833 	total = 0;
834 	for_nexthops(fi) {
835 		if (nh->fib_nh_flags & RTNH_F_DEAD)
836 			continue;
837 
838 		if (ip_ignore_linkdown(nh->fib_nh_dev) &&
839 		    nh->fib_nh_flags & RTNH_F_LINKDOWN)
840 			continue;
841 
842 		total += nh->fib_nh_weight;
843 	} endfor_nexthops(fi);
844 
845 	w = 0;
846 	change_nexthops(fi) {
847 		int upper_bound;
848 
849 		if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD) {
850 			upper_bound = -1;
851 		} else if (ip_ignore_linkdown(nexthop_nh->fib_nh_dev) &&
852 			   nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN) {
853 			upper_bound = -1;
854 		} else {
855 			w += nexthop_nh->fib_nh_weight;
856 			upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31,
857 							    total) - 1;
858 		}
859 
860 		atomic_set(&nexthop_nh->fib_nh_upper_bound, upper_bound);
861 	} endfor_nexthops(fi);
862 }
863 #else /* CONFIG_IP_ROUTE_MULTIPATH */
864 
fib_get_nhs(struct fib_info * fi,struct rtnexthop * rtnh,int remaining,struct fib_config * cfg,struct netlink_ext_ack * extack)865 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
866 		       int remaining, struct fib_config *cfg,
867 		       struct netlink_ext_ack *extack)
868 {
869 	NL_SET_ERR_MSG(extack, "Multipath support not enabled in kernel");
870 
871 	return -EINVAL;
872 }
873 
874 #define fib_rebalance(fi) do { } while (0)
875 
876 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
877 
fib_encap_match(struct net * net,u16 encap_type,struct nlattr * encap,const struct fib_nh * nh,const struct fib_config * cfg,struct netlink_ext_ack * extack)878 static int fib_encap_match(struct net *net, u16 encap_type,
879 			   struct nlattr *encap,
880 			   const struct fib_nh *nh,
881 			   const struct fib_config *cfg,
882 			   struct netlink_ext_ack *extack)
883 {
884 	struct lwtunnel_state *lwtstate;
885 	int ret, result = 0;
886 
887 	if (encap_type == LWTUNNEL_ENCAP_NONE)
888 		return 0;
889 
890 	ret = lwtunnel_build_state(net, encap_type, encap, AF_INET,
891 				   cfg, &lwtstate, extack);
892 	if (!ret) {
893 		result = lwtunnel_cmp_encap(lwtstate, nh->fib_nh_lws);
894 		lwtstate_free(lwtstate);
895 	}
896 
897 	return result;
898 }
899 
fib_nh_match(struct net * net,struct fib_config * cfg,struct fib_info * fi,struct netlink_ext_ack * extack)900 int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
901 		 struct netlink_ext_ack *extack)
902 {
903 #ifdef CONFIG_IP_ROUTE_MULTIPATH
904 	struct rtnexthop *rtnh;
905 	int remaining;
906 #endif
907 
908 	if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority)
909 		return 1;
910 
911 	if (cfg->fc_nh_id) {
912 		if (fi->nh && cfg->fc_nh_id == fi->nh->id)
913 			return 0;
914 		return 1;
915 	}
916 
917 	if (fi->nh) {
918 		if (cfg->fc_oif || cfg->fc_gw_family || cfg->fc_mp)
919 			return 1;
920 		return 0;
921 	}
922 
923 	if (cfg->fc_oif || cfg->fc_gw_family) {
924 		struct fib_nh *nh;
925 
926 		nh = fib_info_nh(fi, 0);
927 		if (cfg->fc_encap) {
928 			if (fib_encap_match(net, cfg->fc_encap_type,
929 					    cfg->fc_encap, nh, cfg, extack))
930 				return 1;
931 		}
932 #ifdef CONFIG_IP_ROUTE_CLASSID
933 		if (cfg->fc_flow &&
934 		    cfg->fc_flow != nh->nh_tclassid)
935 			return 1;
936 #endif
937 		if ((cfg->fc_oif && cfg->fc_oif != nh->fib_nh_oif) ||
938 		    (cfg->fc_gw_family &&
939 		     cfg->fc_gw_family != nh->fib_nh_gw_family))
940 			return 1;
941 
942 		if (cfg->fc_gw_family == AF_INET &&
943 		    cfg->fc_gw4 != nh->fib_nh_gw4)
944 			return 1;
945 
946 		if (cfg->fc_gw_family == AF_INET6 &&
947 		    ipv6_addr_cmp(&cfg->fc_gw6, &nh->fib_nh_gw6))
948 			return 1;
949 
950 		return 0;
951 	}
952 
953 #ifdef CONFIG_IP_ROUTE_MULTIPATH
954 	if (!cfg->fc_mp)
955 		return 0;
956 
957 	rtnh = cfg->fc_mp;
958 	remaining = cfg->fc_mp_len;
959 
960 	for_nexthops(fi) {
961 		int attrlen;
962 
963 		if (!rtnh_ok(rtnh, remaining))
964 			return -EINVAL;
965 
966 		if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->fib_nh_oif)
967 			return 1;
968 
969 		attrlen = rtnh_attrlen(rtnh);
970 		if (attrlen > 0) {
971 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
972 			int err;
973 
974 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
975 			nlav = nla_find(attrs, attrlen, RTA_VIA);
976 			if (nla && nlav) {
977 				NL_SET_ERR_MSG(extack,
978 					       "Nexthop configuration can not contain both GATEWAY and VIA");
979 				return -EINVAL;
980 			}
981 
982 			if (nla) {
983 				__be32 gw;
984 
985 				err = fib_gw_from_attr(&gw, nla, extack);
986 				if (err)
987 					return err;
988 
989 				if (nh->fib_nh_gw_family != AF_INET ||
990 				    gw != nh->fib_nh_gw4)
991 					return 1;
992 			} else if (nlav) {
993 				struct fib_config cfg2;
994 
995 				err = fib_gw_from_via(&cfg2, nlav, extack);
996 				if (err)
997 					return err;
998 
999 				switch (nh->fib_nh_gw_family) {
1000 				case AF_INET:
1001 					if (cfg2.fc_gw_family != AF_INET ||
1002 					    cfg2.fc_gw4 != nh->fib_nh_gw4)
1003 						return 1;
1004 					break;
1005 				case AF_INET6:
1006 					if (cfg2.fc_gw_family != AF_INET6 ||
1007 					    ipv6_addr_cmp(&cfg2.fc_gw6,
1008 							  &nh->fib_nh_gw6))
1009 						return 1;
1010 					break;
1011 				}
1012 			}
1013 
1014 #ifdef CONFIG_IP_ROUTE_CLASSID
1015 			nla = nla_find(attrs, attrlen, RTA_FLOW);
1016 			if (nla) {
1017 				if (nla_len(nla) < sizeof(u32)) {
1018 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
1019 					return -EINVAL;
1020 				}
1021 				if (nla_get_u32(nla) != nh->nh_tclassid)
1022 					return 1;
1023 			}
1024 #endif
1025 		}
1026 
1027 		rtnh = rtnh_next(rtnh, &remaining);
1028 	} endfor_nexthops(fi);
1029 #endif
1030 	return 0;
1031 }
1032 
fib_metrics_match(struct fib_config * cfg,struct fib_info * fi)1033 bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
1034 {
1035 	struct nlattr *nla;
1036 	int remaining;
1037 
1038 	if (!cfg->fc_mx)
1039 		return true;
1040 
1041 	nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1042 		int type = nla_type(nla);
1043 		u32 fi_val, val;
1044 
1045 		if (!type)
1046 			continue;
1047 		if (type > RTAX_MAX)
1048 			return false;
1049 
1050 		type = array_index_nospec(type, RTAX_MAX + 1);
1051 		if (type == RTAX_CC_ALGO) {
1052 			char tmp[TCP_CA_NAME_MAX];
1053 			bool ecn_ca = false;
1054 
1055 			nla_strscpy(tmp, nla, sizeof(tmp));
1056 			val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1057 		} else {
1058 			if (nla_len(nla) != sizeof(u32))
1059 				return false;
1060 			val = nla_get_u32(nla);
1061 		}
1062 
1063 		fi_val = fi->fib_metrics->metrics[type - 1];
1064 		if (type == RTAX_FEATURES)
1065 			fi_val &= ~DST_FEATURE_ECN_CA;
1066 
1067 		if (fi_val != val)
1068 			return false;
1069 	}
1070 
1071 	return true;
1072 }
1073 
fib_check_nh_v6_gw(struct net * net,struct fib_nh * nh,u32 table,struct netlink_ext_ack * extack)1074 static int fib_check_nh_v6_gw(struct net *net, struct fib_nh *nh,
1075 			      u32 table, struct netlink_ext_ack *extack)
1076 {
1077 	struct fib6_config cfg = {
1078 		.fc_table = table,
1079 		.fc_flags = nh->fib_nh_flags | RTF_GATEWAY,
1080 		.fc_ifindex = nh->fib_nh_oif,
1081 		.fc_gateway = nh->fib_nh_gw6,
1082 	};
1083 	struct fib6_nh fib6_nh = {};
1084 	int err;
1085 
1086 	err = ipv6_stub->fib6_nh_init(net, &fib6_nh, &cfg, GFP_KERNEL, extack);
1087 	if (!err) {
1088 		nh->fib_nh_dev = fib6_nh.fib_nh_dev;
1089 		netdev_hold(nh->fib_nh_dev, &nh->fib_nh_dev_tracker,
1090 			    GFP_KERNEL);
1091 		nh->fib_nh_oif = nh->fib_nh_dev->ifindex;
1092 		nh->fib_nh_scope = RT_SCOPE_LINK;
1093 
1094 		ipv6_stub->fib6_nh_release(&fib6_nh);
1095 	}
1096 
1097 	return err;
1098 }
1099 
1100 /*
1101  * Picture
1102  * -------
1103  *
1104  * Semantics of nexthop is very messy by historical reasons.
1105  * We have to take into account, that:
1106  * a) gateway can be actually local interface address,
1107  *    so that gatewayed route is direct.
1108  * b) gateway must be on-link address, possibly
1109  *    described not by an ifaddr, but also by a direct route.
1110  * c) If both gateway and interface are specified, they should not
1111  *    contradict.
1112  * d) If we use tunnel routes, gateway could be not on-link.
1113  *
1114  * Attempt to reconcile all of these (alas, self-contradictory) conditions
1115  * results in pretty ugly and hairy code with obscure logic.
1116  *
1117  * I chose to generalized it instead, so that the size
1118  * of code does not increase practically, but it becomes
1119  * much more general.
1120  * Every prefix is assigned a "scope" value: "host" is local address,
1121  * "link" is direct route,
1122  * [ ... "site" ... "interior" ... ]
1123  * and "universe" is true gateway route with global meaning.
1124  *
1125  * Every prefix refers to a set of "nexthop"s (gw, oif),
1126  * where gw must have narrower scope. This recursion stops
1127  * when gw has LOCAL scope or if "nexthop" is declared ONLINK,
1128  * which means that gw is forced to be on link.
1129  *
1130  * Code is still hairy, but now it is apparently logically
1131  * consistent and very flexible. F.e. as by-product it allows
1132  * to co-exists in peace independent exterior and interior
1133  * routing processes.
1134  *
1135  * Normally it looks as following.
1136  *
1137  * {universe prefix}  -> (gw, oif) [scope link]
1138  *		  |
1139  *		  |-> {link prefix} -> (gw, oif) [scope local]
1140  *					|
1141  *					|-> {local prefix} (terminal node)
1142  */
fib_check_nh_v4_gw(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1143 static int fib_check_nh_v4_gw(struct net *net, struct fib_nh *nh, u32 table,
1144 			      u8 scope, struct netlink_ext_ack *extack)
1145 {
1146 	struct net_device *dev;
1147 	struct fib_result res;
1148 	int err = 0;
1149 
1150 	if (nh->fib_nh_flags & RTNH_F_ONLINK) {
1151 		unsigned int addr_type;
1152 
1153 		if (scope >= RT_SCOPE_LINK) {
1154 			NL_SET_ERR_MSG(extack, "Nexthop has invalid scope");
1155 			return -EINVAL;
1156 		}
1157 		dev = __dev_get_by_index(net, nh->fib_nh_oif);
1158 		if (!dev) {
1159 			NL_SET_ERR_MSG(extack, "Nexthop device required for onlink");
1160 			return -ENODEV;
1161 		}
1162 		if (!(dev->flags & IFF_UP)) {
1163 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1164 			return -ENETDOWN;
1165 		}
1166 		addr_type = inet_addr_type_dev_table(net, dev, nh->fib_nh_gw4);
1167 		if (addr_type != RTN_UNICAST) {
1168 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1169 			return -EINVAL;
1170 		}
1171 		if (!netif_carrier_ok(dev))
1172 			nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1173 		nh->fib_nh_dev = dev;
1174 		netdev_hold(dev, &nh->fib_nh_dev_tracker, GFP_ATOMIC);
1175 		nh->fib_nh_scope = RT_SCOPE_LINK;
1176 		return 0;
1177 	}
1178 	rcu_read_lock();
1179 	{
1180 		struct fib_table *tbl = NULL;
1181 		struct flowi4 fl4 = {
1182 			.daddr = nh->fib_nh_gw4,
1183 			.flowi4_scope = scope + 1,
1184 			.flowi4_oif = nh->fib_nh_oif,
1185 			.flowi4_iif = LOOPBACK_IFINDEX,
1186 		};
1187 
1188 		/* It is not necessary, but requires a bit of thinking */
1189 		if (fl4.flowi4_scope < RT_SCOPE_LINK)
1190 			fl4.flowi4_scope = RT_SCOPE_LINK;
1191 
1192 		if (table && table != RT_TABLE_MAIN)
1193 			tbl = fib_get_table(net, table);
1194 
1195 		if (tbl)
1196 			err = fib_table_lookup(tbl, &fl4, &res,
1197 					       FIB_LOOKUP_IGNORE_LINKSTATE |
1198 					       FIB_LOOKUP_NOREF);
1199 
1200 		/* on error or if no table given do full lookup. This
1201 		 * is needed for example when nexthops are in the local
1202 		 * table rather than the given table
1203 		 */
1204 		if (!tbl || err) {
1205 			err = fib_lookup(net, &fl4, &res,
1206 					 FIB_LOOKUP_IGNORE_LINKSTATE);
1207 		}
1208 
1209 		if (err) {
1210 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1211 			goto out;
1212 		}
1213 	}
1214 
1215 	err = -EINVAL;
1216 	if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) {
1217 		NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1218 		goto out;
1219 	}
1220 	nh->fib_nh_scope = res.scope;
1221 	nh->fib_nh_oif = FIB_RES_OIF(res);
1222 	nh->fib_nh_dev = dev = FIB_RES_DEV(res);
1223 	if (!dev) {
1224 		NL_SET_ERR_MSG(extack,
1225 			       "No egress device for nexthop gateway");
1226 		goto out;
1227 	}
1228 	netdev_hold(dev, &nh->fib_nh_dev_tracker, GFP_ATOMIC);
1229 	if (!netif_carrier_ok(dev))
1230 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1231 	err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN;
1232 out:
1233 	rcu_read_unlock();
1234 	return err;
1235 }
1236 
fib_check_nh_nongw(struct net * net,struct fib_nh * nh,struct netlink_ext_ack * extack)1237 static int fib_check_nh_nongw(struct net *net, struct fib_nh *nh,
1238 			      struct netlink_ext_ack *extack)
1239 {
1240 	struct in_device *in_dev;
1241 	int err;
1242 
1243 	if (nh->fib_nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) {
1244 		NL_SET_ERR_MSG(extack,
1245 			       "Invalid flags for nexthop - PERVASIVE and ONLINK can not be set");
1246 		return -EINVAL;
1247 	}
1248 
1249 	rcu_read_lock();
1250 
1251 	err = -ENODEV;
1252 	in_dev = inetdev_by_index(net, nh->fib_nh_oif);
1253 	if (!in_dev)
1254 		goto out;
1255 	err = -ENETDOWN;
1256 	if (!(in_dev->dev->flags & IFF_UP)) {
1257 		NL_SET_ERR_MSG(extack, "Device for nexthop is not up");
1258 		goto out;
1259 	}
1260 
1261 	nh->fib_nh_dev = in_dev->dev;
1262 	netdev_hold(nh->fib_nh_dev, &nh->fib_nh_dev_tracker, GFP_ATOMIC);
1263 	nh->fib_nh_scope = RT_SCOPE_HOST;
1264 	if (!netif_carrier_ok(nh->fib_nh_dev))
1265 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1266 	err = 0;
1267 out:
1268 	rcu_read_unlock();
1269 	return err;
1270 }
1271 
fib_check_nh(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1272 int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
1273 		 struct netlink_ext_ack *extack)
1274 {
1275 	int err;
1276 
1277 	if (nh->fib_nh_gw_family == AF_INET)
1278 		err = fib_check_nh_v4_gw(net, nh, table, scope, extack);
1279 	else if (nh->fib_nh_gw_family == AF_INET6)
1280 		err = fib_check_nh_v6_gw(net, nh, table, extack);
1281 	else
1282 		err = fib_check_nh_nongw(net, nh, extack);
1283 
1284 	return err;
1285 }
1286 
fib_info_update_nhc_saddr(struct net * net,struct fib_nh_common * nhc,unsigned char scope)1287 __be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
1288 				 unsigned char scope)
1289 {
1290 	struct fib_nh *nh;
1291 	__be32 saddr;
1292 
1293 	if (nhc->nhc_family != AF_INET)
1294 		return inet_select_addr(nhc->nhc_dev, 0, scope);
1295 
1296 	nh = container_of(nhc, struct fib_nh, nh_common);
1297 	saddr = inet_select_addr(nh->fib_nh_dev, nh->fib_nh_gw4, scope);
1298 
1299 	WRITE_ONCE(nh->nh_saddr, saddr);
1300 	WRITE_ONCE(nh->nh_saddr_genid, atomic_read(&net->ipv4.dev_addr_genid));
1301 
1302 	return saddr;
1303 }
1304 
fib_result_prefsrc(struct net * net,struct fib_result * res)1305 __be32 fib_result_prefsrc(struct net *net, struct fib_result *res)
1306 {
1307 	struct fib_nh_common *nhc = res->nhc;
1308 
1309 	if (res->fi->fib_prefsrc)
1310 		return res->fi->fib_prefsrc;
1311 
1312 	if (nhc->nhc_family == AF_INET) {
1313 		struct fib_nh *nh;
1314 
1315 		nh = container_of(nhc, struct fib_nh, nh_common);
1316 		if (READ_ONCE(nh->nh_saddr_genid) ==
1317 		    atomic_read(&net->ipv4.dev_addr_genid))
1318 			return READ_ONCE(nh->nh_saddr);
1319 	}
1320 
1321 	return fib_info_update_nhc_saddr(net, nhc, res->fi->fib_scope);
1322 }
1323 
fib_valid_prefsrc(struct fib_config * cfg,__be32 fib_prefsrc)1324 static bool fib_valid_prefsrc(struct fib_config *cfg, __be32 fib_prefsrc)
1325 {
1326 	if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst ||
1327 	    fib_prefsrc != cfg->fc_dst) {
1328 		u32 tb_id = cfg->fc_table;
1329 		int rc;
1330 
1331 		if (tb_id == RT_TABLE_MAIN)
1332 			tb_id = RT_TABLE_LOCAL;
1333 
1334 		rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1335 					  fib_prefsrc, tb_id);
1336 
1337 		if (rc != RTN_LOCAL && tb_id != RT_TABLE_LOCAL) {
1338 			rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1339 						  fib_prefsrc, RT_TABLE_LOCAL);
1340 		}
1341 
1342 		if (rc != RTN_LOCAL)
1343 			return false;
1344 	}
1345 	return true;
1346 }
1347 
fib_create_info(struct fib_config * cfg,struct netlink_ext_ack * extack)1348 struct fib_info *fib_create_info(struct fib_config *cfg,
1349 				 struct netlink_ext_ack *extack)
1350 {
1351 	int err;
1352 	struct fib_info *fi = NULL;
1353 	struct nexthop *nh = NULL;
1354 	struct fib_info *ofi;
1355 	int nhs = 1;
1356 	struct net *net = cfg->fc_nlinfo.nl_net;
1357 
1358 	ASSERT_RTNL();
1359 	if (cfg->fc_type > RTN_MAX)
1360 		goto err_inval;
1361 
1362 	/* Fast check to catch the most weird cases */
1363 	if (fib_props[cfg->fc_type].scope > cfg->fc_scope) {
1364 		NL_SET_ERR_MSG(extack, "Invalid scope");
1365 		goto err_inval;
1366 	}
1367 
1368 	if (cfg->fc_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
1369 		NL_SET_ERR_MSG(extack,
1370 			       "Invalid rtm_flags - can not contain DEAD or LINKDOWN");
1371 		goto err_inval;
1372 	}
1373 
1374 	if (cfg->fc_nh_id) {
1375 		if (!cfg->fc_mx) {
1376 			fi = fib_find_info_nh(net, cfg);
1377 			if (fi) {
1378 				refcount_inc(&fi->fib_treeref);
1379 				return fi;
1380 			}
1381 		}
1382 
1383 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
1384 		if (!nh) {
1385 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
1386 			goto err_inval;
1387 		}
1388 		nhs = 0;
1389 	}
1390 
1391 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1392 	if (cfg->fc_mp) {
1393 		nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len, extack);
1394 		if (nhs == 0)
1395 			goto err_inval;
1396 	}
1397 #endif
1398 
1399 	fib_info_hash_grow(net);
1400 
1401 	fi = kzalloc_flex(*fi, fib_nh, nhs);
1402 	if (!fi) {
1403 		err = -ENOBUFS;
1404 		goto failure;
1405 	}
1406 
1407 	fi->fib_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len, extack);
1408 	if (IS_ERR(fi->fib_metrics)) {
1409 		err = PTR_ERR(fi->fib_metrics);
1410 		kfree(fi);
1411 		return ERR_PTR(err);
1412 	}
1413 
1414 	fi->fib_net = net;
1415 	fi->fib_protocol = cfg->fc_protocol;
1416 	fi->fib_scope = cfg->fc_scope;
1417 	fi->fib_flags = cfg->fc_flags;
1418 	fi->fib_priority = cfg->fc_priority;
1419 	fi->fib_prefsrc = cfg->fc_prefsrc;
1420 	fi->fib_type = cfg->fc_type;
1421 	fi->fib_tb_id = cfg->fc_table;
1422 
1423 	fi->fib_nhs = nhs;
1424 	if (nh) {
1425 		if (!nexthop_get(nh)) {
1426 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1427 			err = -EINVAL;
1428 		} else {
1429 			err = 0;
1430 			fi->nh = nh;
1431 		}
1432 	} else {
1433 		change_nexthops(fi) {
1434 			nexthop_nh->nh_parent = fi;
1435 		} endfor_nexthops(fi)
1436 
1437 		if (cfg->fc_mp)
1438 			err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg,
1439 					  extack);
1440 		else
1441 			err = fib_nh_init(net, fi->fib_nh, cfg, 1, extack);
1442 	}
1443 
1444 	if (err != 0)
1445 		goto failure;
1446 
1447 	if (fib_props[cfg->fc_type].error) {
1448 		if (cfg->fc_gw_family || cfg->fc_oif || cfg->fc_mp) {
1449 			NL_SET_ERR_MSG(extack,
1450 				       "Gateway, device and multipath can not be specified for this route type");
1451 			goto err_inval;
1452 		}
1453 		goto link_it;
1454 	} else {
1455 		switch (cfg->fc_type) {
1456 		case RTN_UNICAST:
1457 		case RTN_LOCAL:
1458 		case RTN_BROADCAST:
1459 		case RTN_ANYCAST:
1460 		case RTN_MULTICAST:
1461 			break;
1462 		default:
1463 			NL_SET_ERR_MSG(extack, "Invalid route type");
1464 			goto err_inval;
1465 		}
1466 	}
1467 
1468 	if (cfg->fc_scope > RT_SCOPE_HOST) {
1469 		NL_SET_ERR_MSG(extack, "Invalid scope");
1470 		goto err_inval;
1471 	}
1472 
1473 	if (fi->nh) {
1474 		err = fib_check_nexthop(fi->nh, cfg->fc_scope, extack);
1475 		if (err)
1476 			goto failure;
1477 	} else if (cfg->fc_scope == RT_SCOPE_HOST) {
1478 		struct fib_nh *nh = fi->fib_nh;
1479 
1480 		/* Local address is added. */
1481 		if (nhs != 1) {
1482 			NL_SET_ERR_MSG(extack,
1483 				       "Route with host scope can not have multiple nexthops");
1484 			goto err_inval;
1485 		}
1486 		if (nh->fib_nh_gw_family) {
1487 			NL_SET_ERR_MSG(extack,
1488 				       "Route with host scope can not have a gateway");
1489 			goto err_inval;
1490 		}
1491 		nh->fib_nh_scope = RT_SCOPE_NOWHERE;
1492 		nh->fib_nh_dev = dev_get_by_index(net, nh->fib_nh_oif);
1493 		err = -ENODEV;
1494 		if (!nh->fib_nh_dev)
1495 			goto failure;
1496 		netdev_tracker_alloc(nh->fib_nh_dev, &nh->fib_nh_dev_tracker,
1497 				     GFP_KERNEL);
1498 	} else {
1499 		int linkdown = 0;
1500 
1501 		change_nexthops(fi) {
1502 			err = fib_check_nh(cfg->fc_nlinfo.nl_net, nexthop_nh,
1503 					   cfg->fc_table, cfg->fc_scope,
1504 					   extack);
1505 			if (err != 0)
1506 				goto failure;
1507 			if (nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN)
1508 				linkdown++;
1509 		} endfor_nexthops(fi)
1510 		if (linkdown == fi->fib_nhs)
1511 			fi->fib_flags |= RTNH_F_LINKDOWN;
1512 	}
1513 
1514 	if (fi->fib_prefsrc && !fib_valid_prefsrc(cfg, fi->fib_prefsrc)) {
1515 		NL_SET_ERR_MSG(extack, "Invalid prefsrc address");
1516 		goto err_inval;
1517 	}
1518 
1519 	if (!fi->nh) {
1520 		change_nexthops(fi) {
1521 			fib_info_update_nhc_saddr(net, &nexthop_nh->nh_common,
1522 						  fi->fib_scope);
1523 			if (nexthop_nh->fib_nh_gw_family == AF_INET6)
1524 				fi->fib_nh_is_v6 = true;
1525 		} endfor_nexthops(fi)
1526 
1527 		fib_rebalance(fi);
1528 	}
1529 
1530 link_it:
1531 	ofi = fib_find_info(fi);
1532 	if (ofi) {
1533 		/* fib_table_lookup() should not see @fi yet. */
1534 		fi->fib_dead = 1;
1535 		free_fib_info(fi);
1536 		refcount_inc(&ofi->fib_treeref);
1537 		return ofi;
1538 	}
1539 
1540 	refcount_set(&fi->fib_treeref, 1);
1541 	refcount_set(&fi->fib_clntref, 1);
1542 
1543 	net->ipv4.fib_info_cnt++;
1544 	hlist_add_head(&fi->fib_hash, fib_info_hash_bucket(fi));
1545 
1546 	if (fi->fib_prefsrc) {
1547 		struct hlist_head *head;
1548 
1549 		head = fib_info_laddrhash_bucket(net, fi->fib_prefsrc);
1550 		hlist_add_head(&fi->fib_lhash, head);
1551 	}
1552 	if (fi->nh) {
1553 		list_add(&fi->nh_list, &nh->fi_list);
1554 	} else {
1555 		change_nexthops(fi) {
1556 			struct hlist_head *head;
1557 
1558 			if (!nexthop_nh->fib_nh_dev)
1559 				continue;
1560 			head = fib_nh_head(nexthop_nh->fib_nh_dev);
1561 			hlist_add_head_rcu(&nexthop_nh->nh_hash, head);
1562 		} endfor_nexthops(fi)
1563 	}
1564 	return fi;
1565 
1566 err_inval:
1567 	err = -EINVAL;
1568 
1569 failure:
1570 	if (fi) {
1571 		/* fib_table_lookup() should not see @fi yet. */
1572 		fi->fib_dead = 1;
1573 		free_fib_info(fi);
1574 	}
1575 
1576 	return ERR_PTR(err);
1577 }
1578 
fib_nexthop_info(struct sk_buff * skb,const struct fib_nh_common * nhc,u8 rt_family,unsigned char * flags,bool skip_oif)1579 int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nhc,
1580 		     u8 rt_family, unsigned char *flags, bool skip_oif)
1581 {
1582 	if (nhc->nhc_flags & RTNH_F_DEAD)
1583 		*flags |= RTNH_F_DEAD;
1584 
1585 	if (nhc->nhc_flags & RTNH_F_LINKDOWN) {
1586 		*flags |= RTNH_F_LINKDOWN;
1587 
1588 		rcu_read_lock();
1589 		switch (nhc->nhc_family) {
1590 		case AF_INET:
1591 			if (ip_ignore_linkdown(nhc->nhc_dev))
1592 				*flags |= RTNH_F_DEAD;
1593 			break;
1594 		case AF_INET6:
1595 			if (ip6_ignore_linkdown(nhc->nhc_dev))
1596 				*flags |= RTNH_F_DEAD;
1597 			break;
1598 		}
1599 		rcu_read_unlock();
1600 	}
1601 
1602 	switch (nhc->nhc_gw_family) {
1603 	case AF_INET:
1604 		if (nla_put_in_addr(skb, RTA_GATEWAY, nhc->nhc_gw.ipv4))
1605 			goto nla_put_failure;
1606 		break;
1607 	case AF_INET6:
1608 		/* if gateway family does not match nexthop family
1609 		 * gateway is encoded as RTA_VIA
1610 		 */
1611 		if (rt_family != nhc->nhc_gw_family) {
1612 			int alen = sizeof(struct in6_addr);
1613 			struct nlattr *nla;
1614 			struct rtvia *via;
1615 
1616 			nla = nla_reserve(skb, RTA_VIA, alen + 2);
1617 			if (!nla)
1618 				goto nla_put_failure;
1619 
1620 			via = nla_data(nla);
1621 			via->rtvia_family = AF_INET6;
1622 			memcpy(via->rtvia_addr, &nhc->nhc_gw.ipv6, alen);
1623 		} else if (nla_put_in6_addr(skb, RTA_GATEWAY,
1624 					    &nhc->nhc_gw.ipv6) < 0) {
1625 			goto nla_put_failure;
1626 		}
1627 		break;
1628 	}
1629 
1630 	*flags |= (nhc->nhc_flags &
1631 		   (RTNH_F_ONLINK | RTNH_F_OFFLOAD | RTNH_F_TRAP));
1632 
1633 	if (!skip_oif && nhc->nhc_dev &&
1634 	    nla_put_u32(skb, RTA_OIF, nhc->nhc_dev->ifindex))
1635 		goto nla_put_failure;
1636 
1637 	if (lwtunnel_fill_encap(skb, nhc->nhc_lwtstate,
1638 				RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
1639 		goto nla_put_failure;
1640 
1641 	return 0;
1642 
1643 nla_put_failure:
1644 	return -EMSGSIZE;
1645 }
1646 EXPORT_SYMBOL_GPL(fib_nexthop_info);
1647 
1648 #if IS_ENABLED(CONFIG_IP_ROUTE_MULTIPATH) || IS_ENABLED(CONFIG_IPV6)
fib_add_nexthop(struct sk_buff * skb,const struct fib_nh_common * nhc,int nh_weight,u8 rt_family,u32 nh_tclassid)1649 int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nhc,
1650 		    int nh_weight, u8 rt_family, u32 nh_tclassid)
1651 {
1652 	const struct net_device *dev = nhc->nhc_dev;
1653 	struct rtnexthop *rtnh;
1654 	unsigned char flags = 0;
1655 
1656 	rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1657 	if (!rtnh)
1658 		goto nla_put_failure;
1659 
1660 	rtnh->rtnh_hops = nh_weight - 1;
1661 	rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
1662 
1663 	if (fib_nexthop_info(skb, nhc, rt_family, &flags, true) < 0)
1664 		goto nla_put_failure;
1665 
1666 	rtnh->rtnh_flags = flags;
1667 
1668 	if (nh_tclassid && nla_put_u32(skb, RTA_FLOW, nh_tclassid))
1669 		goto nla_put_failure;
1670 
1671 	/* length of rtnetlink header + attributes */
1672 	rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
1673 
1674 	return 0;
1675 
1676 nla_put_failure:
1677 	return -EMSGSIZE;
1678 }
1679 EXPORT_SYMBOL_GPL(fib_add_nexthop);
1680 #endif
1681 
1682 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1683 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1684 {
1685 	struct nlattr *mp;
1686 
1687 	mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
1688 	if (!mp)
1689 		goto nla_put_failure;
1690 
1691 	if (unlikely(fi->nh)) {
1692 		if (nexthop_mpath_fill_node(skb, fi->nh, AF_INET) < 0)
1693 			goto nla_put_failure;
1694 		goto mp_end;
1695 	}
1696 
1697 	for_nexthops(fi) {
1698 		u32 nh_tclassid = 0;
1699 #ifdef CONFIG_IP_ROUTE_CLASSID
1700 		nh_tclassid = nh->nh_tclassid;
1701 #endif
1702 		if (fib_add_nexthop(skb, &nh->nh_common, nh->fib_nh_weight,
1703 				    AF_INET, nh_tclassid) < 0)
1704 			goto nla_put_failure;
1705 	} endfor_nexthops(fi);
1706 
1707 mp_end:
1708 	nla_nest_end(skb, mp);
1709 
1710 	return 0;
1711 
1712 nla_put_failure:
1713 	return -EMSGSIZE;
1714 }
1715 #else
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1716 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1717 {
1718 	return 0;
1719 }
1720 #endif
1721 
fib_dump_info(struct sk_buff * skb,u32 portid,u32 seq,int event,const struct fib_rt_info * fri,unsigned int flags)1722 int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
1723 		  const struct fib_rt_info *fri, unsigned int flags)
1724 {
1725 	unsigned int nhs = fib_info_num_path(fri->fi);
1726 	struct fib_info *fi = fri->fi;
1727 	u32 tb_id = fri->tb_id;
1728 	struct nlmsghdr *nlh;
1729 	struct rtmsg *rtm;
1730 
1731 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1732 	if (!nlh)
1733 		return -EMSGSIZE;
1734 
1735 	rtm = nlmsg_data(nlh);
1736 	rtm->rtm_family = AF_INET;
1737 	rtm->rtm_dst_len = fri->dst_len;
1738 	rtm->rtm_src_len = 0;
1739 	rtm->rtm_tos = inet_dscp_to_dsfield(fri->dscp);
1740 	if (tb_id < 256)
1741 		rtm->rtm_table = tb_id;
1742 	else
1743 		rtm->rtm_table = RT_TABLE_COMPAT;
1744 	if (nla_put_u32(skb, RTA_TABLE, tb_id))
1745 		goto nla_put_failure;
1746 	rtm->rtm_type = fri->type;
1747 	rtm->rtm_flags = fi->fib_flags;
1748 	rtm->rtm_scope = fi->fib_scope;
1749 	rtm->rtm_protocol = fi->fib_protocol;
1750 
1751 	if (rtm->rtm_dst_len &&
1752 	    nla_put_in_addr(skb, RTA_DST, fri->dst))
1753 		goto nla_put_failure;
1754 	if (fi->fib_priority &&
1755 	    nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority))
1756 		goto nla_put_failure;
1757 	if (rtnetlink_put_metrics(skb, fi->fib_metrics->metrics) < 0)
1758 		goto nla_put_failure;
1759 
1760 	if (fi->fib_prefsrc &&
1761 	    nla_put_in_addr(skb, RTA_PREFSRC, fi->fib_prefsrc))
1762 		goto nla_put_failure;
1763 
1764 	if (fi->nh) {
1765 		if (nla_put_u32(skb, RTA_NH_ID, fi->nh->id))
1766 			goto nla_put_failure;
1767 		if (nexthop_is_blackhole(fi->nh))
1768 			rtm->rtm_type = RTN_BLACKHOLE;
1769 		if (!READ_ONCE(fi->fib_net->ipv4.sysctl_nexthop_compat_mode))
1770 			goto offload;
1771 	}
1772 
1773 	if (nhs == 1) {
1774 		const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
1775 		unsigned char flags = 0;
1776 
1777 		if (fib_nexthop_info(skb, nhc, AF_INET, &flags, false) < 0)
1778 			goto nla_put_failure;
1779 
1780 		rtm->rtm_flags = flags;
1781 #ifdef CONFIG_IP_ROUTE_CLASSID
1782 		if (nhc->nhc_family == AF_INET) {
1783 			struct fib_nh *nh;
1784 
1785 			nh = container_of(nhc, struct fib_nh, nh_common);
1786 			if (nh->nh_tclassid &&
1787 			    nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid))
1788 				goto nla_put_failure;
1789 		}
1790 #endif
1791 	} else {
1792 		if (fib_add_multipath(skb, fi) < 0)
1793 			goto nla_put_failure;
1794 	}
1795 
1796 offload:
1797 	if (fri->offload)
1798 		rtm->rtm_flags |= RTM_F_OFFLOAD;
1799 	if (fri->trap)
1800 		rtm->rtm_flags |= RTM_F_TRAP;
1801 	if (fri->offload_failed)
1802 		rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
1803 
1804 	nlmsg_end(skb, nlh);
1805 	return 0;
1806 
1807 nla_put_failure:
1808 	nlmsg_cancel(skb, nlh);
1809 	return -EMSGSIZE;
1810 }
1811 
1812 /*
1813  * Update FIB if:
1814  * - local address disappeared -> we must delete all the entries
1815  *   referring to it.
1816  * - device went down -> we must shutdown all nexthops going via it.
1817  */
fib_sync_down_addr(struct net_device * dev,__be32 local)1818 int fib_sync_down_addr(struct net_device *dev, __be32 local)
1819 {
1820 	int tb_id = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
1821 	struct net *net = dev_net(dev);
1822 	struct hlist_head *head;
1823 	struct fib_info *fi;
1824 	int ret = 0;
1825 
1826 	if (!local)
1827 		return 0;
1828 
1829 	head = fib_info_laddrhash_bucket(net, local);
1830 	hlist_for_each_entry(fi, head, fib_lhash) {
1831 		if (!net_eq(fi->fib_net, net) ||
1832 		    fi->fib_tb_id != tb_id)
1833 			continue;
1834 		if (fi->fib_prefsrc == local) {
1835 			fi->fib_flags |= RTNH_F_DEAD;
1836 			fi->pfsrc_removed = true;
1837 			ret++;
1838 		}
1839 	}
1840 	return ret;
1841 }
1842 
call_fib_nh_notifiers(struct fib_nh * nh,enum fib_event_type event_type)1843 static int call_fib_nh_notifiers(struct fib_nh *nh,
1844 				 enum fib_event_type event_type)
1845 {
1846 	bool ignore_link_down = ip_ignore_linkdown(nh->fib_nh_dev);
1847 	struct fib_nh_notifier_info info = {
1848 		.fib_nh = nh,
1849 	};
1850 
1851 	switch (event_type) {
1852 	case FIB_EVENT_NH_ADD:
1853 		if (nh->fib_nh_flags & RTNH_F_DEAD)
1854 			break;
1855 		if (ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN)
1856 			break;
1857 		return call_fib4_notifiers(dev_net(nh->fib_nh_dev), event_type,
1858 					   &info.info);
1859 	case FIB_EVENT_NH_DEL:
1860 		if ((ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN) ||
1861 		    (nh->fib_nh_flags & RTNH_F_DEAD))
1862 			return call_fib4_notifiers(dev_net(nh->fib_nh_dev),
1863 						   event_type, &info.info);
1864 		break;
1865 	default:
1866 		break;
1867 	}
1868 
1869 	return NOTIFY_DONE;
1870 }
1871 
1872 /* Update the PMTU of exceptions when:
1873  * - the new MTU of the first hop becomes smaller than the PMTU
1874  * - the old MTU was the same as the PMTU, and it limited discovery of
1875  *   larger MTUs on the path. With that limit raised, we can now
1876  *   discover larger MTUs
1877  * A special case is locked exceptions, for which the PMTU is smaller
1878  * than the minimal accepted PMTU:
1879  * - if the new MTU is greater than the PMTU, don't make any change
1880  * - otherwise, unlock and set PMTU
1881  */
fib_nhc_update_mtu(struct fib_nh_common * nhc,u32 new,u32 orig)1882 void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig)
1883 {
1884 	struct fnhe_hash_bucket *bucket;
1885 	int i;
1886 
1887 	bucket = rcu_dereference_protected(nhc->nhc_exceptions, 1);
1888 	if (!bucket)
1889 		return;
1890 
1891 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
1892 		struct fib_nh_exception *fnhe;
1893 
1894 		for (fnhe = rcu_dereference_protected(bucket[i].chain, 1);
1895 		     fnhe;
1896 		     fnhe = rcu_dereference_protected(fnhe->fnhe_next, 1)) {
1897 			if (fnhe->fnhe_mtu_locked) {
1898 				if (new <= fnhe->fnhe_pmtu) {
1899 					fnhe->fnhe_pmtu = new;
1900 					fnhe->fnhe_mtu_locked = false;
1901 				}
1902 			} else if (new < fnhe->fnhe_pmtu ||
1903 				   orig == fnhe->fnhe_pmtu) {
1904 				fnhe->fnhe_pmtu = new;
1905 			}
1906 		}
1907 	}
1908 }
1909 
fib_sync_mtu(struct net_device * dev,u32 orig_mtu)1910 void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
1911 {
1912 	struct hlist_head *head = fib_nh_head(dev);
1913 	struct fib_nh *nh;
1914 
1915 	hlist_for_each_entry(nh, head, nh_hash) {
1916 		DEBUG_NET_WARN_ON_ONCE(nh->fib_nh_dev != dev);
1917 		fib_nhc_update_mtu(&nh->nh_common, dev->mtu, orig_mtu);
1918 	}
1919 }
1920 
1921 /* Event              force Flags           Description
1922  * NETDEV_CHANGE      0     LINKDOWN        Carrier OFF, not for scope host
1923  * NETDEV_DOWN        0     LINKDOWN|DEAD   Link down, not for scope host
1924  * NETDEV_DOWN        1     LINKDOWN|DEAD   Last address removed
1925  * NETDEV_UNREGISTER  1     LINKDOWN|DEAD   Device removed
1926  *
1927  * only used when fib_nh is built into fib_info
1928  */
fib_sync_down_dev(struct net_device * dev,unsigned long event,bool force)1929 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force)
1930 {
1931 	struct hlist_head *head = fib_nh_head(dev);
1932 	struct fib_info *prev_fi = NULL;
1933 	int scope = RT_SCOPE_NOWHERE;
1934 	struct fib_nh *nh;
1935 	int ret = 0;
1936 
1937 	if (force)
1938 		scope = -1;
1939 
1940 	hlist_for_each_entry(nh, head, nh_hash) {
1941 		struct fib_info *fi = nh->nh_parent;
1942 		int dead;
1943 
1944 		BUG_ON(!fi->fib_nhs);
1945 		DEBUG_NET_WARN_ON_ONCE(nh->fib_nh_dev != dev);
1946 		if (fi == prev_fi)
1947 			continue;
1948 		prev_fi = fi;
1949 		dead = 0;
1950 		change_nexthops(fi) {
1951 			if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD)
1952 				dead++;
1953 			else if (nexthop_nh->fib_nh_dev == dev &&
1954 				 nexthop_nh->fib_nh_scope != scope) {
1955 				switch (event) {
1956 				case NETDEV_DOWN:
1957 				case NETDEV_UNREGISTER:
1958 					nexthop_nh->fib_nh_flags |= RTNH_F_DEAD;
1959 					fallthrough;
1960 				case NETDEV_CHANGE:
1961 					nexthop_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1962 					break;
1963 				}
1964 				call_fib_nh_notifiers(nexthop_nh,
1965 						      FIB_EVENT_NH_DEL);
1966 				dead++;
1967 			}
1968 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1969 			if (event == NETDEV_UNREGISTER &&
1970 			    nexthop_nh->fib_nh_dev == dev) {
1971 				dead = fi->fib_nhs;
1972 				break;
1973 			}
1974 #endif
1975 		} endfor_nexthops(fi)
1976 		if (dead == fi->fib_nhs) {
1977 			switch (event) {
1978 			case NETDEV_DOWN:
1979 			case NETDEV_UNREGISTER:
1980 				fi->fib_flags |= RTNH_F_DEAD;
1981 				fallthrough;
1982 			case NETDEV_CHANGE:
1983 				fi->fib_flags |= RTNH_F_LINKDOWN;
1984 				break;
1985 			}
1986 			ret++;
1987 		}
1988 
1989 		fib_rebalance(fi);
1990 	}
1991 
1992 	return ret;
1993 }
1994 
1995 /* Must be invoked inside of an RCU protected region.  */
fib_select_default(const struct flowi4 * flp,struct fib_result * res)1996 static void fib_select_default(const struct flowi4 *flp, struct fib_result *res)
1997 {
1998 	struct fib_info *fi = NULL, *last_resort = NULL;
1999 	struct hlist_head *fa_head = res->fa_head;
2000 	struct fib_table *tb = res->table;
2001 	u8 slen = 32 - res->prefixlen;
2002 	int order = -1, last_idx = -1;
2003 	struct fib_alias *fa, *fa1 = NULL;
2004 	u32 last_prio = res->fi->fib_priority;
2005 	dscp_t last_dscp = 0;
2006 
2007 	hlist_for_each_entry_rcu(fa, fa_head, fa_list) {
2008 		struct fib_info *next_fi = fa->fa_info;
2009 		struct fib_nh_common *nhc;
2010 
2011 		if (fa->fa_slen != slen)
2012 			continue;
2013 		if (fa->fa_dscp && !fib_dscp_masked_match(fa->fa_dscp, flp))
2014 			continue;
2015 		if (fa->tb_id != tb->tb_id)
2016 			continue;
2017 		if (next_fi->fib_priority > last_prio &&
2018 		    fa->fa_dscp == last_dscp) {
2019 			if (last_dscp)
2020 				continue;
2021 			break;
2022 		}
2023 		if (next_fi->fib_flags & RTNH_F_DEAD)
2024 			continue;
2025 		last_dscp = fa->fa_dscp;
2026 		last_prio = next_fi->fib_priority;
2027 
2028 		if (next_fi->fib_scope != res->scope ||
2029 		    fa->fa_type != RTN_UNICAST)
2030 			continue;
2031 
2032 		nhc = fib_info_nhc(next_fi, 0);
2033 		if (!nhc->nhc_gw_family || nhc->nhc_scope != RT_SCOPE_LINK)
2034 			continue;
2035 
2036 		fib_alias_accessed(fa);
2037 
2038 		if (!fi) {
2039 			if (next_fi != res->fi)
2040 				break;
2041 			fa1 = fa;
2042 		} else if (!fib_detect_death(fi, order, &last_resort,
2043 					     &last_idx, fa1->fa_default)) {
2044 			fib_result_assign(res, fi);
2045 			fa1->fa_default = order;
2046 			goto out;
2047 		}
2048 		fi = next_fi;
2049 		order++;
2050 	}
2051 
2052 	if (order <= 0 || !fi) {
2053 		if (fa1)
2054 			fa1->fa_default = -1;
2055 		goto out;
2056 	}
2057 
2058 	if (!fib_detect_death(fi, order, &last_resort, &last_idx,
2059 			      fa1->fa_default)) {
2060 		fib_result_assign(res, fi);
2061 		fa1->fa_default = order;
2062 		goto out;
2063 	}
2064 
2065 	if (last_idx >= 0)
2066 		fib_result_assign(res, last_resort);
2067 	fa1->fa_default = last_idx;
2068 out:
2069 	return;
2070 }
2071 
2072 /*
2073  * Dead device goes up. We wake up dead nexthops.
2074  * It takes sense only on multipath routes.
2075  *
2076  * only used when fib_nh is built into fib_info
2077  */
fib_sync_up(struct net_device * dev,unsigned char nh_flags)2078 int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
2079 {
2080 	struct fib_info *prev_fi;
2081 	struct hlist_head *head;
2082 	struct fib_nh *nh;
2083 	int ret;
2084 
2085 	if (!(dev->flags & IFF_UP))
2086 		return 0;
2087 
2088 	if (nh_flags & RTNH_F_DEAD) {
2089 		unsigned int flags = netif_get_flags(dev);
2090 
2091 		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
2092 			nh_flags |= RTNH_F_LINKDOWN;
2093 	}
2094 
2095 	prev_fi = NULL;
2096 	head = fib_nh_head(dev);
2097 	ret = 0;
2098 
2099 	hlist_for_each_entry(nh, head, nh_hash) {
2100 		struct fib_info *fi = nh->nh_parent;
2101 		int alive;
2102 
2103 		BUG_ON(!fi->fib_nhs);
2104 		DEBUG_NET_WARN_ON_ONCE(nh->fib_nh_dev != dev);
2105 		if (fi == prev_fi)
2106 			continue;
2107 
2108 		prev_fi = fi;
2109 		alive = 0;
2110 		change_nexthops(fi) {
2111 			if (!(nexthop_nh->fib_nh_flags & nh_flags)) {
2112 				alive++;
2113 				continue;
2114 			}
2115 			if (!nexthop_nh->fib_nh_dev ||
2116 			    !(nexthop_nh->fib_nh_dev->flags & IFF_UP))
2117 				continue;
2118 			if (nexthop_nh->fib_nh_dev != dev ||
2119 			    !__in_dev_get_rtnl(dev))
2120 				continue;
2121 			alive++;
2122 			nexthop_nh->fib_nh_flags &= ~nh_flags;
2123 			call_fib_nh_notifiers(nexthop_nh, FIB_EVENT_NH_ADD);
2124 		} endfor_nexthops(fi)
2125 
2126 		if (alive > 0) {
2127 			fi->fib_flags &= ~nh_flags;
2128 			ret++;
2129 		}
2130 
2131 		fib_rebalance(fi);
2132 	}
2133 
2134 	return ret;
2135 }
2136 
2137 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_good_nh(const struct fib_nh * nh)2138 static bool fib_good_nh(const struct fib_nh *nh)
2139 {
2140 	int state = NUD_REACHABLE;
2141 
2142 	if (nh->fib_nh_scope == RT_SCOPE_LINK) {
2143 		struct neighbour *n;
2144 
2145 		rcu_read_lock();
2146 
2147 		if (likely(nh->fib_nh_gw_family == AF_INET))
2148 			n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
2149 						   (__force u32)nh->fib_nh_gw4);
2150 		else if (nh->fib_nh_gw_family == AF_INET6)
2151 			n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev,
2152 							   &nh->fib_nh_gw6);
2153 		else
2154 			n = NULL;
2155 		if (n)
2156 			state = READ_ONCE(n->nud_state);
2157 
2158 		rcu_read_unlock();
2159 	}
2160 
2161 	return !!(state & NUD_VALID);
2162 }
2163 
fib_select_multipath(struct fib_result * res,int hash,const struct flowi4 * fl4)2164 void fib_select_multipath(struct fib_result *res, int hash,
2165 			  const struct flowi4 *fl4)
2166 {
2167 	struct fib_info *fi = res->fi;
2168 	struct net *net = fi->fib_net;
2169 	bool use_neigh;
2170 	int score = -1;
2171 	__be32 saddr;
2172 
2173 	if (unlikely(res->fi->nh)) {
2174 		nexthop_path_fib_result(res, hash);
2175 		return;
2176 	}
2177 
2178 	use_neigh = READ_ONCE(net->ipv4.sysctl_fib_multipath_use_neigh);
2179 	saddr = fl4 ? fl4->saddr : 0;
2180 
2181 	change_nexthops(fi) {
2182 		int nh_upper_bound, nh_score = 0;
2183 
2184 		/* Nexthops without a carrier are assigned an upper bound of
2185 		 * minus one when "ignore_routes_with_linkdown" is set.
2186 		 */
2187 		nh_upper_bound = atomic_read(&nexthop_nh->fib_nh_upper_bound);
2188 		if (nh_upper_bound == -1 ||
2189 		    (use_neigh && !fib_good_nh(nexthop_nh)))
2190 			continue;
2191 
2192 		if (saddr && nexthop_nh->nh_saddr == saddr)
2193 			nh_score += 2;
2194 		if (hash <= nh_upper_bound)
2195 			nh_score++;
2196 		if (score < nh_score) {
2197 			res->nh_sel = nhsel;
2198 			res->nhc = &nexthop_nh->nh_common;
2199 			if (nh_score == 3 || (!saddr && nh_score == 1))
2200 				return;
2201 			score = nh_score;
2202 		}
2203 
2204 	} endfor_nexthops(fi);
2205 }
2206 #endif
2207 
fib_select_path(struct net * net,struct fib_result * res,struct flowi4 * fl4,const struct sk_buff * skb)2208 void fib_select_path(struct net *net, struct fib_result *res,
2209 		     struct flowi4 *fl4, const struct sk_buff *skb)
2210 {
2211 	if (fl4->flowi4_oif)
2212 		goto check_saddr;
2213 
2214 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2215 	if (fib_info_num_path(res->fi) > 1) {
2216 		int h = fib_multipath_hash(net, fl4, skb, NULL);
2217 
2218 		fib_select_multipath(res, h, fl4);
2219 	}
2220 	else
2221 #endif
2222 	if (!res->prefixlen &&
2223 	    res->table->tb_num_default > 1 &&
2224 	    res->type == RTN_UNICAST)
2225 		fib_select_default(fl4, res);
2226 
2227 check_saddr:
2228 	if (!fl4->saddr) {
2229 		struct net_device *l3mdev;
2230 
2231 		l3mdev = dev_get_by_index_rcu(net, fl4->flowi4_l3mdev);
2232 
2233 		if (!l3mdev ||
2234 		    l3mdev_master_dev_rcu(FIB_RES_DEV(*res)) == l3mdev)
2235 			fl4->saddr = fib_result_prefsrc(net, res);
2236 		else
2237 			fl4->saddr = inet_select_addr(l3mdev, 0, RT_SCOPE_LINK);
2238 	}
2239 }
2240 
fib4_semantics_init(struct net * net)2241 int __net_init fib4_semantics_init(struct net *net)
2242 {
2243 	unsigned int hash_bits = 4;
2244 
2245 	net->ipv4.fib_info_hash = fib_info_hash_alloc(hash_bits);
2246 	if (!net->ipv4.fib_info_hash)
2247 		return -ENOMEM;
2248 
2249 	net->ipv4.fib_info_hash_bits = hash_bits;
2250 	net->ipv4.fib_info_cnt = 0;
2251 
2252 	return 0;
2253 }
2254 
fib4_semantics_exit(struct net * net)2255 void __net_exit fib4_semantics_exit(struct net *net)
2256 {
2257 	fib_info_hash_free(net->ipv4.fib_info_hash);
2258 }
2259