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 
33 #include <net/arp.h>
34 #include <net/ip.h>
35 #include <net/protocol.h>
36 #include <net/route.h>
37 #include <net/tcp.h>
38 #include <net/sock.h>
39 #include <net/ip_fib.h>
40 #include <net/ip6_fib.h>
41 #include <net/nexthop.h>
42 #include <net/netlink.h>
43 #include <net/rtnh.h>
44 #include <net/lwtunnel.h>
45 #include <net/fib_notifier.h>
46 #include <net/addrconf.h>
47 
48 #include "fib_lookup.h"
49 
50 static DEFINE_SPINLOCK(fib_info_lock);
51 static struct hlist_head *fib_info_hash;
52 static struct hlist_head *fib_info_laddrhash;
53 static unsigned int fib_info_hash_size;
54 static unsigned int fib_info_cnt;
55 
56 #define DEVINDEX_HASHBITS 8
57 #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS)
58 static struct hlist_head fib_info_devhash[DEVINDEX_HASHSIZE];
59 
60 /* for_nexthops and change_nexthops only used when nexthop object
61  * is not set in a fib_info. The logic within can reference fib_nh.
62  */
63 #ifdef CONFIG_IP_ROUTE_MULTIPATH
64 
65 #define for_nexthops(fi) {						\
66 	int nhsel; const struct fib_nh *nh;				\
67 	for (nhsel = 0, nh = (fi)->fib_nh;				\
68 	     nhsel < fib_info_num_path((fi));				\
69 	     nh++, nhsel++)
70 
71 #define change_nexthops(fi) {						\
72 	int nhsel; struct fib_nh *nexthop_nh;				\
73 	for (nhsel = 0,	nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
74 	     nhsel < fib_info_num_path((fi));				\
75 	     nexthop_nh++, nhsel++)
76 
77 #else /* CONFIG_IP_ROUTE_MULTIPATH */
78 
79 /* Hope, that gcc will optimize it to get rid of dummy loop */
80 
81 #define for_nexthops(fi) {						\
82 	int nhsel; const struct fib_nh *nh = (fi)->fib_nh;		\
83 	for (nhsel = 0; nhsel < 1; nhsel++)
84 
85 #define change_nexthops(fi) {						\
86 	int nhsel;							\
87 	struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
88 	for (nhsel = 0; nhsel < 1; nhsel++)
89 
90 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
91 
92 #define endfor_nexthops(fi) }
93 
94 
95 const struct fib_prop fib_props[RTN_MAX + 1] = {
96 	[RTN_UNSPEC] = {
97 		.error	= 0,
98 		.scope	= RT_SCOPE_NOWHERE,
99 	},
100 	[RTN_UNICAST] = {
101 		.error	= 0,
102 		.scope	= RT_SCOPE_UNIVERSE,
103 	},
104 	[RTN_LOCAL] = {
105 		.error	= 0,
106 		.scope	= RT_SCOPE_HOST,
107 	},
108 	[RTN_BROADCAST] = {
109 		.error	= 0,
110 		.scope	= RT_SCOPE_LINK,
111 	},
112 	[RTN_ANYCAST] = {
113 		.error	= 0,
114 		.scope	= RT_SCOPE_LINK,
115 	},
116 	[RTN_MULTICAST] = {
117 		.error	= 0,
118 		.scope	= RT_SCOPE_UNIVERSE,
119 	},
120 	[RTN_BLACKHOLE] = {
121 		.error	= -EINVAL,
122 		.scope	= RT_SCOPE_UNIVERSE,
123 	},
124 	[RTN_UNREACHABLE] = {
125 		.error	= -EHOSTUNREACH,
126 		.scope	= RT_SCOPE_UNIVERSE,
127 	},
128 	[RTN_PROHIBIT] = {
129 		.error	= -EACCES,
130 		.scope	= RT_SCOPE_UNIVERSE,
131 	},
132 	[RTN_THROW] = {
133 		.error	= -EAGAIN,
134 		.scope	= RT_SCOPE_UNIVERSE,
135 	},
136 	[RTN_NAT] = {
137 		.error	= -EINVAL,
138 		.scope	= RT_SCOPE_NOWHERE,
139 	},
140 	[RTN_XRESOLVE] = {
141 		.error	= -EINVAL,
142 		.scope	= RT_SCOPE_NOWHERE,
143 	},
144 };
145 
rt_fibinfo_free(struct rtable __rcu ** rtp)146 static void rt_fibinfo_free(struct rtable __rcu **rtp)
147 {
148 	struct rtable *rt = rcu_dereference_protected(*rtp, 1);
149 
150 	if (!rt)
151 		return;
152 
153 	/* Not even needed : RCU_INIT_POINTER(*rtp, NULL);
154 	 * because we waited an RCU grace period before calling
155 	 * free_fib_info_rcu()
156 	 */
157 
158 	dst_dev_put(&rt->dst);
159 	dst_release_immediate(&rt->dst);
160 }
161 
free_nh_exceptions(struct fib_nh_common * nhc)162 static void free_nh_exceptions(struct fib_nh_common *nhc)
163 {
164 	struct fnhe_hash_bucket *hash;
165 	int i;
166 
167 	hash = rcu_dereference_protected(nhc->nhc_exceptions, 1);
168 	if (!hash)
169 		return;
170 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
171 		struct fib_nh_exception *fnhe;
172 
173 		fnhe = rcu_dereference_protected(hash[i].chain, 1);
174 		while (fnhe) {
175 			struct fib_nh_exception *next;
176 
177 			next = rcu_dereference_protected(fnhe->fnhe_next, 1);
178 
179 			rt_fibinfo_free(&fnhe->fnhe_rth_input);
180 			rt_fibinfo_free(&fnhe->fnhe_rth_output);
181 
182 			kfree(fnhe);
183 
184 			fnhe = next;
185 		}
186 	}
187 	kfree(hash);
188 }
189 
rt_fibinfo_free_cpus(struct rtable __rcu * __percpu * rtp)190 static void rt_fibinfo_free_cpus(struct rtable __rcu * __percpu *rtp)
191 {
192 	int cpu;
193 
194 	if (!rtp)
195 		return;
196 
197 	for_each_possible_cpu(cpu) {
198 		struct rtable *rt;
199 
200 		rt = rcu_dereference_protected(*per_cpu_ptr(rtp, cpu), 1);
201 		if (rt) {
202 			dst_dev_put(&rt->dst);
203 			dst_release_immediate(&rt->dst);
204 		}
205 	}
206 	free_percpu(rtp);
207 }
208 
fib_nh_common_release(struct fib_nh_common * nhc)209 void fib_nh_common_release(struct fib_nh_common *nhc)
210 {
211 	dev_put(nhc->nhc_dev);
212 	lwtstate_put(nhc->nhc_lwtstate);
213 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
214 	rt_fibinfo_free(&nhc->nhc_rth_input);
215 	free_nh_exceptions(nhc);
216 }
217 EXPORT_SYMBOL_GPL(fib_nh_common_release);
218 
fib_nh_release(struct net * net,struct fib_nh * fib_nh)219 void fib_nh_release(struct net *net, struct fib_nh *fib_nh)
220 {
221 #ifdef CONFIG_IP_ROUTE_CLASSID
222 	if (fib_nh->nh_tclassid)
223 		atomic_dec(&net->ipv4.fib_num_tclassid_users);
224 #endif
225 	fib_nh_common_release(&fib_nh->nh_common);
226 }
227 
228 /* Release a nexthop info record */
free_fib_info_rcu(struct rcu_head * head)229 static void free_fib_info_rcu(struct rcu_head *head)
230 {
231 	struct fib_info *fi = container_of(head, struct fib_info, rcu);
232 
233 	if (fi->nh) {
234 		nexthop_put(fi->nh);
235 	} else {
236 		change_nexthops(fi) {
237 			fib_nh_release(fi->fib_net, nexthop_nh);
238 		} endfor_nexthops(fi);
239 	}
240 
241 	ip_fib_metrics_put(fi->fib_metrics);
242 
243 	kfree(fi);
244 }
245 
free_fib_info(struct fib_info * fi)246 void free_fib_info(struct fib_info *fi)
247 {
248 	if (fi->fib_dead == 0) {
249 		pr_warn("Freeing alive fib_info %p\n", fi);
250 		return;
251 	}
252 	fib_info_cnt--;
253 
254 	call_rcu(&fi->rcu, free_fib_info_rcu);
255 }
256 EXPORT_SYMBOL_GPL(free_fib_info);
257 
fib_release_info(struct fib_info * fi)258 void fib_release_info(struct fib_info *fi)
259 {
260 	spin_lock_bh(&fib_info_lock);
261 	if (fi && refcount_dec_and_test(&fi->fib_treeref)) {
262 		hlist_del(&fi->fib_hash);
263 		if (fi->fib_prefsrc)
264 			hlist_del(&fi->fib_lhash);
265 		if (fi->nh) {
266 			list_del(&fi->nh_list);
267 		} else {
268 			change_nexthops(fi) {
269 				if (!nexthop_nh->fib_nh_dev)
270 					continue;
271 				hlist_del(&nexthop_nh->nh_hash);
272 			} endfor_nexthops(fi)
273 		}
274 		fi->fib_dead = 1;
275 		fib_info_put(fi);
276 	}
277 	spin_unlock_bh(&fib_info_lock);
278 }
279 
nh_comp(struct fib_info * fi,struct fib_info * ofi)280 static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi)
281 {
282 	const struct fib_nh *onh;
283 
284 	if (fi->nh || ofi->nh)
285 		return nexthop_cmp(fi->nh, ofi->nh) ? 0 : -1;
286 
287 	if (ofi->fib_nhs == 0)
288 		return 0;
289 
290 	for_nexthops(fi) {
291 		onh = fib_info_nh(ofi, nhsel);
292 
293 		if (nh->fib_nh_oif != onh->fib_nh_oif ||
294 		    nh->fib_nh_gw_family != onh->fib_nh_gw_family ||
295 		    nh->fib_nh_scope != onh->fib_nh_scope ||
296 #ifdef CONFIG_IP_ROUTE_MULTIPATH
297 		    nh->fib_nh_weight != onh->fib_nh_weight ||
298 #endif
299 #ifdef CONFIG_IP_ROUTE_CLASSID
300 		    nh->nh_tclassid != onh->nh_tclassid ||
301 #endif
302 		    lwtunnel_cmp_encap(nh->fib_nh_lws, onh->fib_nh_lws) ||
303 		    ((nh->fib_nh_flags ^ onh->fib_nh_flags) & ~RTNH_COMPARE_MASK))
304 			return -1;
305 
306 		if (nh->fib_nh_gw_family == AF_INET &&
307 		    nh->fib_nh_gw4 != onh->fib_nh_gw4)
308 			return -1;
309 
310 		if (nh->fib_nh_gw_family == AF_INET6 &&
311 		    ipv6_addr_cmp(&nh->fib_nh_gw6, &onh->fib_nh_gw6))
312 			return -1;
313 	} endfor_nexthops(fi);
314 	return 0;
315 }
316 
fib_devindex_hashfn(unsigned int val)317 static inline unsigned int fib_devindex_hashfn(unsigned int val)
318 {
319 	unsigned int mask = DEVINDEX_HASHSIZE - 1;
320 
321 	return (val ^
322 		(val >> DEVINDEX_HASHBITS) ^
323 		(val >> (DEVINDEX_HASHBITS * 2))) & mask;
324 }
325 
fib_info_hashfn_1(int init_val,u8 protocol,u8 scope,u32 prefsrc,u32 priority)326 static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope,
327 				      u32 prefsrc, u32 priority)
328 {
329 	unsigned int val = init_val;
330 
331 	val ^= (protocol << 8) | scope;
332 	val ^= prefsrc;
333 	val ^= priority;
334 
335 	return val;
336 }
337 
fib_info_hashfn_result(unsigned int val)338 static unsigned int fib_info_hashfn_result(unsigned int val)
339 {
340 	unsigned int mask = (fib_info_hash_size - 1);
341 
342 	return (val ^ (val >> 7) ^ (val >> 12)) & mask;
343 }
344 
fib_info_hashfn(struct fib_info * fi)345 static inline unsigned int fib_info_hashfn(struct fib_info *fi)
346 {
347 	unsigned int val;
348 
349 	val = fib_info_hashfn_1(fi->fib_nhs, fi->fib_protocol,
350 				fi->fib_scope, (__force u32)fi->fib_prefsrc,
351 				fi->fib_priority);
352 
353 	if (fi->nh) {
354 		val ^= fib_devindex_hashfn(fi->nh->id);
355 	} else {
356 		for_nexthops(fi) {
357 			val ^= fib_devindex_hashfn(nh->fib_nh_oif);
358 		} endfor_nexthops(fi)
359 	}
360 
361 	return fib_info_hashfn_result(val);
362 }
363 
364 /* no metrics, only nexthop id */
fib_find_info_nh(struct net * net,const struct fib_config * cfg)365 static struct fib_info *fib_find_info_nh(struct net *net,
366 					 const struct fib_config *cfg)
367 {
368 	struct hlist_head *head;
369 	struct fib_info *fi;
370 	unsigned int hash;
371 
372 	hash = fib_info_hashfn_1(fib_devindex_hashfn(cfg->fc_nh_id),
373 				 cfg->fc_protocol, cfg->fc_scope,
374 				 (__force u32)cfg->fc_prefsrc,
375 				 cfg->fc_priority);
376 	hash = fib_info_hashfn_result(hash);
377 	head = &fib_info_hash[hash];
378 
379 	hlist_for_each_entry(fi, head, fib_hash) {
380 		if (!net_eq(fi->fib_net, net))
381 			continue;
382 		if (!fi->nh || fi->nh->id != cfg->fc_nh_id)
383 			continue;
384 		if (cfg->fc_protocol == fi->fib_protocol &&
385 		    cfg->fc_scope == fi->fib_scope &&
386 		    cfg->fc_prefsrc == fi->fib_prefsrc &&
387 		    cfg->fc_priority == fi->fib_priority &&
388 		    cfg->fc_type == fi->fib_type &&
389 		    cfg->fc_table == fi->fib_tb_id &&
390 		    !((cfg->fc_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK))
391 			return fi;
392 	}
393 
394 	return NULL;
395 }
396 
fib_find_info(struct fib_info * nfi)397 static struct fib_info *fib_find_info(struct fib_info *nfi)
398 {
399 	struct hlist_head *head;
400 	struct fib_info *fi;
401 	unsigned int hash;
402 
403 	hash = fib_info_hashfn(nfi);
404 	head = &fib_info_hash[hash];
405 
406 	hlist_for_each_entry(fi, head, fib_hash) {
407 		if (!net_eq(fi->fib_net, nfi->fib_net))
408 			continue;
409 		if (fi->fib_nhs != nfi->fib_nhs)
410 			continue;
411 		if (nfi->fib_protocol == fi->fib_protocol &&
412 		    nfi->fib_scope == fi->fib_scope &&
413 		    nfi->fib_prefsrc == fi->fib_prefsrc &&
414 		    nfi->fib_priority == fi->fib_priority &&
415 		    nfi->fib_type == fi->fib_type &&
416 		    memcmp(nfi->fib_metrics, fi->fib_metrics,
417 			   sizeof(u32) * RTAX_MAX) == 0 &&
418 		    !((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK) &&
419 		    nh_comp(fi, nfi) == 0)
420 			return fi;
421 	}
422 
423 	return NULL;
424 }
425 
426 /* Check, that the gateway is already configured.
427  * Used only by redirect accept routine.
428  */
ip_fib_check_default(__be32 gw,struct net_device * dev)429 int ip_fib_check_default(__be32 gw, struct net_device *dev)
430 {
431 	struct hlist_head *head;
432 	struct fib_nh *nh;
433 	unsigned int hash;
434 
435 	spin_lock(&fib_info_lock);
436 
437 	hash = fib_devindex_hashfn(dev->ifindex);
438 	head = &fib_info_devhash[hash];
439 	hlist_for_each_entry(nh, head, nh_hash) {
440 		if (nh->fib_nh_dev == dev &&
441 		    nh->fib_nh_gw4 == gw &&
442 		    !(nh->fib_nh_flags & RTNH_F_DEAD)) {
443 			spin_unlock(&fib_info_lock);
444 			return 0;
445 		}
446 	}
447 
448 	spin_unlock(&fib_info_lock);
449 
450 	return -1;
451 }
452 
fib_nlmsg_size(struct fib_info * fi)453 size_t fib_nlmsg_size(struct fib_info *fi)
454 {
455 	size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
456 			 + nla_total_size(4) /* RTA_TABLE */
457 			 + nla_total_size(4) /* RTA_DST */
458 			 + nla_total_size(4) /* RTA_PRIORITY */
459 			 + nla_total_size(4) /* RTA_PREFSRC */
460 			 + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
461 	unsigned int nhs = fib_info_num_path(fi);
462 
463 	/* space for nested metrics */
464 	payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
465 
466 	if (fi->nh)
467 		payload += nla_total_size(4); /* RTA_NH_ID */
468 
469 	if (nhs) {
470 		size_t nh_encapsize = 0;
471 		/* Also handles the special case nhs == 1 */
472 
473 		/* each nexthop is packed in an attribute */
474 		size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
475 		unsigned int i;
476 
477 		/* may contain flow and gateway attribute */
478 		nhsize += 2 * nla_total_size(4);
479 
480 		/* grab encap info */
481 		for (i = 0; i < fib_info_num_path(fi); i++) {
482 			struct fib_nh_common *nhc = fib_info_nhc(fi, i);
483 
484 			if (nhc->nhc_lwtstate) {
485 				/* RTA_ENCAP_TYPE */
486 				nh_encapsize += lwtunnel_get_encap_size(
487 						nhc->nhc_lwtstate);
488 				/* RTA_ENCAP */
489 				nh_encapsize +=  nla_total_size(2);
490 			}
491 		}
492 
493 		/* all nexthops are packed in a nested attribute */
494 		payload += nla_total_size((nhs * nhsize) + nh_encapsize);
495 
496 	}
497 
498 	return payload;
499 }
500 
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)501 void rtmsg_fib(int event, __be32 key, struct fib_alias *fa,
502 	       int dst_len, u32 tb_id, const struct nl_info *info,
503 	       unsigned int nlm_flags)
504 {
505 	struct fib_rt_info fri;
506 	struct sk_buff *skb;
507 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
508 	int err = -ENOBUFS;
509 
510 	skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL);
511 	if (!skb)
512 		goto errout;
513 
514 	fri.fi = fa->fa_info;
515 	fri.tb_id = tb_id;
516 	fri.dst = key;
517 	fri.dst_len = dst_len;
518 	fri.tos = fa->fa_tos;
519 	fri.type = fa->fa_type;
520 	fri.offload = fa->offload;
521 	fri.trap = fa->trap;
522 	fri.offload_failed = fa->offload_failed;
523 	err = fib_dump_info(skb, info->portid, seq, event, &fri, nlm_flags);
524 	if (err < 0) {
525 		/* -EMSGSIZE implies BUG in fib_nlmsg_size() */
526 		WARN_ON(err == -EMSGSIZE);
527 		kfree_skb(skb);
528 		goto errout;
529 	}
530 	rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_IPV4_ROUTE,
531 		    info->nlh, GFP_KERNEL);
532 	return;
533 errout:
534 	if (err < 0)
535 		rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err);
536 }
537 
fib_detect_death(struct fib_info * fi,int order,struct fib_info ** last_resort,int * last_idx,int dflt)538 static int fib_detect_death(struct fib_info *fi, int order,
539 			    struct fib_info **last_resort, int *last_idx,
540 			    int dflt)
541 {
542 	const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
543 	struct neighbour *n;
544 	int state = NUD_NONE;
545 
546 	if (likely(nhc->nhc_gw_family == AF_INET))
547 		n = neigh_lookup(&arp_tbl, &nhc->nhc_gw.ipv4, nhc->nhc_dev);
548 	else if (nhc->nhc_gw_family == AF_INET6)
549 		n = neigh_lookup(ipv6_stub->nd_tbl, &nhc->nhc_gw.ipv6,
550 				 nhc->nhc_dev);
551 	else
552 		n = NULL;
553 
554 	if (n) {
555 		state = n->nud_state;
556 		neigh_release(n);
557 	} else {
558 		return 0;
559 	}
560 	if (state == NUD_REACHABLE)
561 		return 0;
562 	if ((state & NUD_VALID) && order != dflt)
563 		return 0;
564 	if ((state & NUD_VALID) ||
565 	    (*last_idx < 0 && order > dflt && state != NUD_INCOMPLETE)) {
566 		*last_resort = fi;
567 		*last_idx = order;
568 	}
569 	return 1;
570 }
571 
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)572 int fib_nh_common_init(struct net *net, struct fib_nh_common *nhc,
573 		       struct nlattr *encap, u16 encap_type,
574 		       void *cfg, gfp_t gfp_flags,
575 		       struct netlink_ext_ack *extack)
576 {
577 	int err;
578 
579 	nhc->nhc_pcpu_rth_output = alloc_percpu_gfp(struct rtable __rcu *,
580 						    gfp_flags);
581 	if (!nhc->nhc_pcpu_rth_output)
582 		return -ENOMEM;
583 
584 	if (encap) {
585 		struct lwtunnel_state *lwtstate;
586 
587 		if (encap_type == LWTUNNEL_ENCAP_NONE) {
588 			NL_SET_ERR_MSG(extack, "LWT encap type not specified");
589 			err = -EINVAL;
590 			goto lwt_failure;
591 		}
592 		err = lwtunnel_build_state(net, encap_type, encap,
593 					   nhc->nhc_family, cfg, &lwtstate,
594 					   extack);
595 		if (err)
596 			goto lwt_failure;
597 
598 		nhc->nhc_lwtstate = lwtstate_get(lwtstate);
599 	}
600 
601 	return 0;
602 
603 lwt_failure:
604 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
605 	nhc->nhc_pcpu_rth_output = NULL;
606 	return err;
607 }
608 EXPORT_SYMBOL_GPL(fib_nh_common_init);
609 
fib_nh_init(struct net * net,struct fib_nh * nh,struct fib_config * cfg,int nh_weight,struct netlink_ext_ack * extack)610 int fib_nh_init(struct net *net, struct fib_nh *nh,
611 		struct fib_config *cfg, int nh_weight,
612 		struct netlink_ext_ack *extack)
613 {
614 	int err;
615 
616 	nh->fib_nh_family = AF_INET;
617 
618 	err = fib_nh_common_init(net, &nh->nh_common, cfg->fc_encap,
619 				 cfg->fc_encap_type, cfg, GFP_KERNEL, extack);
620 	if (err)
621 		return err;
622 
623 	nh->fib_nh_oif = cfg->fc_oif;
624 	nh->fib_nh_gw_family = cfg->fc_gw_family;
625 	if (cfg->fc_gw_family == AF_INET)
626 		nh->fib_nh_gw4 = cfg->fc_gw4;
627 	else if (cfg->fc_gw_family == AF_INET6)
628 		nh->fib_nh_gw6 = cfg->fc_gw6;
629 
630 	nh->fib_nh_flags = cfg->fc_flags;
631 
632 #ifdef CONFIG_IP_ROUTE_CLASSID
633 	nh->nh_tclassid = cfg->fc_flow;
634 	if (nh->nh_tclassid)
635 		atomic_inc(&net->ipv4.fib_num_tclassid_users);
636 #endif
637 #ifdef CONFIG_IP_ROUTE_MULTIPATH
638 	nh->fib_nh_weight = nh_weight;
639 #endif
640 	return 0;
641 }
642 
643 #ifdef CONFIG_IP_ROUTE_MULTIPATH
644 
fib_count_nexthops(struct rtnexthop * rtnh,int remaining,struct netlink_ext_ack * extack)645 static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining,
646 			      struct netlink_ext_ack *extack)
647 {
648 	int nhs = 0;
649 
650 	while (rtnh_ok(rtnh, remaining)) {
651 		nhs++;
652 		rtnh = rtnh_next(rtnh, &remaining);
653 	}
654 
655 	/* leftover implies invalid nexthop configuration, discard it */
656 	if (remaining > 0) {
657 		NL_SET_ERR_MSG(extack,
658 			       "Invalid nexthop configuration - extra data after nexthops");
659 		nhs = 0;
660 	}
661 
662 	return nhs;
663 }
664 
fib_gw_from_attr(__be32 * gw,struct nlattr * nla,struct netlink_ext_ack * extack)665 static int fib_gw_from_attr(__be32 *gw, struct nlattr *nla,
666 			    struct netlink_ext_ack *extack)
667 {
668 	if (nla_len(nla) < sizeof(*gw)) {
669 		NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_GATEWAY");
670 		return -EINVAL;
671 	}
672 
673 	*gw = nla_get_in_addr(nla);
674 
675 	return 0;
676 }
677 
678 /* 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)679 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
680 		       int remaining, struct fib_config *cfg,
681 		       struct netlink_ext_ack *extack)
682 {
683 	struct net *net = fi->fib_net;
684 	struct fib_config fib_cfg;
685 	struct fib_nh *nh;
686 	int ret;
687 
688 	change_nexthops(fi) {
689 		int attrlen;
690 
691 		memset(&fib_cfg, 0, sizeof(fib_cfg));
692 
693 		if (!rtnh_ok(rtnh, remaining)) {
694 			NL_SET_ERR_MSG(extack,
695 				       "Invalid nexthop configuration - extra data after nexthop");
696 			return -EINVAL;
697 		}
698 
699 		if (rtnh->rtnh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
700 			NL_SET_ERR_MSG(extack,
701 				       "Invalid flags for nexthop - can not contain DEAD or LINKDOWN");
702 			return -EINVAL;
703 		}
704 
705 		fib_cfg.fc_flags = (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags;
706 		fib_cfg.fc_oif = rtnh->rtnh_ifindex;
707 
708 		attrlen = rtnh_attrlen(rtnh);
709 		if (attrlen > 0) {
710 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
711 
712 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
713 			nlav = nla_find(attrs, attrlen, RTA_VIA);
714 			if (nla && nlav) {
715 				NL_SET_ERR_MSG(extack,
716 					       "Nexthop configuration can not contain both GATEWAY and VIA");
717 				return -EINVAL;
718 			}
719 			if (nla) {
720 				ret = fib_gw_from_attr(&fib_cfg.fc_gw4, nla,
721 						       extack);
722 				if (ret)
723 					goto errout;
724 
725 				if (fib_cfg.fc_gw4)
726 					fib_cfg.fc_gw_family = AF_INET;
727 			} else if (nlav) {
728 				ret = fib_gw_from_via(&fib_cfg, nlav, extack);
729 				if (ret)
730 					goto errout;
731 			}
732 
733 			nla = nla_find(attrs, attrlen, RTA_FLOW);
734 			if (nla) {
735 				if (nla_len(nla) < sizeof(u32)) {
736 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
737 					return -EINVAL;
738 				}
739 				fib_cfg.fc_flow = nla_get_u32(nla);
740 			}
741 
742 			fib_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
743 			/* RTA_ENCAP_TYPE length checked in
744 			 * lwtunnel_valid_encap_type_attr
745 			 */
746 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
747 			if (nla)
748 				fib_cfg.fc_encap_type = nla_get_u16(nla);
749 		}
750 
751 		ret = fib_nh_init(net, nexthop_nh, &fib_cfg,
752 				  rtnh->rtnh_hops + 1, extack);
753 		if (ret)
754 			goto errout;
755 
756 		rtnh = rtnh_next(rtnh, &remaining);
757 	} endfor_nexthops(fi);
758 
759 	ret = -EINVAL;
760 	nh = fib_info_nh(fi, 0);
761 	if (cfg->fc_oif && nh->fib_nh_oif != cfg->fc_oif) {
762 		NL_SET_ERR_MSG(extack,
763 			       "Nexthop device index does not match RTA_OIF");
764 		goto errout;
765 	}
766 	if (cfg->fc_gw_family) {
767 		if (cfg->fc_gw_family != nh->fib_nh_gw_family ||
768 		    (cfg->fc_gw_family == AF_INET &&
769 		     nh->fib_nh_gw4 != cfg->fc_gw4) ||
770 		    (cfg->fc_gw_family == AF_INET6 &&
771 		     ipv6_addr_cmp(&nh->fib_nh_gw6, &cfg->fc_gw6))) {
772 			NL_SET_ERR_MSG(extack,
773 				       "Nexthop gateway does not match RTA_GATEWAY or RTA_VIA");
774 			goto errout;
775 		}
776 	}
777 #ifdef CONFIG_IP_ROUTE_CLASSID
778 	if (cfg->fc_flow && nh->nh_tclassid != cfg->fc_flow) {
779 		NL_SET_ERR_MSG(extack,
780 			       "Nexthop class id does not match RTA_FLOW");
781 		goto errout;
782 	}
783 #endif
784 	ret = 0;
785 errout:
786 	return ret;
787 }
788 
789 /* only called when fib_nh is integrated into fib_info */
fib_rebalance(struct fib_info * fi)790 static void fib_rebalance(struct fib_info *fi)
791 {
792 	int total;
793 	int w;
794 
795 	if (fib_info_num_path(fi) < 2)
796 		return;
797 
798 	total = 0;
799 	for_nexthops(fi) {
800 		if (nh->fib_nh_flags & RTNH_F_DEAD)
801 			continue;
802 
803 		if (ip_ignore_linkdown(nh->fib_nh_dev) &&
804 		    nh->fib_nh_flags & RTNH_F_LINKDOWN)
805 			continue;
806 
807 		total += nh->fib_nh_weight;
808 	} endfor_nexthops(fi);
809 
810 	w = 0;
811 	change_nexthops(fi) {
812 		int upper_bound;
813 
814 		if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD) {
815 			upper_bound = -1;
816 		} else if (ip_ignore_linkdown(nexthop_nh->fib_nh_dev) &&
817 			   nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN) {
818 			upper_bound = -1;
819 		} else {
820 			w += nexthop_nh->fib_nh_weight;
821 			upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31,
822 							    total) - 1;
823 		}
824 
825 		atomic_set(&nexthop_nh->fib_nh_upper_bound, upper_bound);
826 	} endfor_nexthops(fi);
827 }
828 #else /* CONFIG_IP_ROUTE_MULTIPATH */
829 
fib_get_nhs(struct fib_info * fi,struct rtnexthop * rtnh,int remaining,struct fib_config * cfg,struct netlink_ext_ack * extack)830 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
831 		       int remaining, struct fib_config *cfg,
832 		       struct netlink_ext_ack *extack)
833 {
834 	NL_SET_ERR_MSG(extack, "Multipath support not enabled in kernel");
835 
836 	return -EINVAL;
837 }
838 
839 #define fib_rebalance(fi) do { } while (0)
840 
841 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
842 
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)843 static int fib_encap_match(struct net *net, u16 encap_type,
844 			   struct nlattr *encap,
845 			   const struct fib_nh *nh,
846 			   const struct fib_config *cfg,
847 			   struct netlink_ext_ack *extack)
848 {
849 	struct lwtunnel_state *lwtstate;
850 	int ret, result = 0;
851 
852 	if (encap_type == LWTUNNEL_ENCAP_NONE)
853 		return 0;
854 
855 	ret = lwtunnel_build_state(net, encap_type, encap, AF_INET,
856 				   cfg, &lwtstate, extack);
857 	if (!ret) {
858 		result = lwtunnel_cmp_encap(lwtstate, nh->fib_nh_lws);
859 		lwtstate_free(lwtstate);
860 	}
861 
862 	return result;
863 }
864 
fib_nh_match(struct net * net,struct fib_config * cfg,struct fib_info * fi,struct netlink_ext_ack * extack)865 int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
866 		 struct netlink_ext_ack *extack)
867 {
868 #ifdef CONFIG_IP_ROUTE_MULTIPATH
869 	struct rtnexthop *rtnh;
870 	int remaining;
871 #endif
872 
873 	if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority)
874 		return 1;
875 
876 	if (cfg->fc_nh_id) {
877 		if (fi->nh && cfg->fc_nh_id == fi->nh->id)
878 			return 0;
879 		return 1;
880 	}
881 
882 	if (cfg->fc_oif || cfg->fc_gw_family) {
883 		struct fib_nh *nh = fib_info_nh(fi, 0);
884 
885 		if (cfg->fc_encap) {
886 			if (fib_encap_match(net, cfg->fc_encap_type,
887 					    cfg->fc_encap, nh, cfg, extack))
888 				return 1;
889 		}
890 #ifdef CONFIG_IP_ROUTE_CLASSID
891 		if (cfg->fc_flow &&
892 		    cfg->fc_flow != nh->nh_tclassid)
893 			return 1;
894 #endif
895 		if ((cfg->fc_oif && cfg->fc_oif != nh->fib_nh_oif) ||
896 		    (cfg->fc_gw_family &&
897 		     cfg->fc_gw_family != nh->fib_nh_gw_family))
898 			return 1;
899 
900 		if (cfg->fc_gw_family == AF_INET &&
901 		    cfg->fc_gw4 != nh->fib_nh_gw4)
902 			return 1;
903 
904 		if (cfg->fc_gw_family == AF_INET6 &&
905 		    ipv6_addr_cmp(&cfg->fc_gw6, &nh->fib_nh_gw6))
906 			return 1;
907 
908 		return 0;
909 	}
910 
911 #ifdef CONFIG_IP_ROUTE_MULTIPATH
912 	if (!cfg->fc_mp)
913 		return 0;
914 
915 	rtnh = cfg->fc_mp;
916 	remaining = cfg->fc_mp_len;
917 
918 	for_nexthops(fi) {
919 		int attrlen;
920 
921 		if (!rtnh_ok(rtnh, remaining))
922 			return -EINVAL;
923 
924 		if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->fib_nh_oif)
925 			return 1;
926 
927 		attrlen = rtnh_attrlen(rtnh);
928 		if (attrlen > 0) {
929 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
930 			int err;
931 
932 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
933 			nlav = nla_find(attrs, attrlen, RTA_VIA);
934 			if (nla && nlav) {
935 				NL_SET_ERR_MSG(extack,
936 					       "Nexthop configuration can not contain both GATEWAY and VIA");
937 				return -EINVAL;
938 			}
939 
940 			if (nla) {
941 				__be32 gw;
942 
943 				err = fib_gw_from_attr(&gw, nla, extack);
944 				if (err)
945 					return err;
946 
947 				if (nh->fib_nh_gw_family != AF_INET ||
948 				    gw != nh->fib_nh_gw4)
949 					return 1;
950 			} else if (nlav) {
951 				struct fib_config cfg2;
952 
953 				err = fib_gw_from_via(&cfg2, nlav, extack);
954 				if (err)
955 					return err;
956 
957 				switch (nh->fib_nh_gw_family) {
958 				case AF_INET:
959 					if (cfg2.fc_gw_family != AF_INET ||
960 					    cfg2.fc_gw4 != nh->fib_nh_gw4)
961 						return 1;
962 					break;
963 				case AF_INET6:
964 					if (cfg2.fc_gw_family != AF_INET6 ||
965 					    ipv6_addr_cmp(&cfg2.fc_gw6,
966 							  &nh->fib_nh_gw6))
967 						return 1;
968 					break;
969 				}
970 			}
971 
972 #ifdef CONFIG_IP_ROUTE_CLASSID
973 			nla = nla_find(attrs, attrlen, RTA_FLOW);
974 			if (nla) {
975 				if (nla_len(nla) < sizeof(u32)) {
976 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
977 					return -EINVAL;
978 				}
979 				if (nla_get_u32(nla) != nh->nh_tclassid)
980 					return 1;
981 			}
982 #endif
983 		}
984 
985 		rtnh = rtnh_next(rtnh, &remaining);
986 	} endfor_nexthops(fi);
987 #endif
988 	return 0;
989 }
990 
fib_metrics_match(struct fib_config * cfg,struct fib_info * fi)991 bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
992 {
993 	struct nlattr *nla;
994 	int remaining;
995 
996 	if (!cfg->fc_mx)
997 		return true;
998 
999 	nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1000 		int type = nla_type(nla);
1001 		u32 fi_val, val;
1002 
1003 		if (!type)
1004 			continue;
1005 		if (type > RTAX_MAX)
1006 			return false;
1007 
1008 		if (type == RTAX_CC_ALGO) {
1009 			char tmp[TCP_CA_NAME_MAX];
1010 			bool ecn_ca = false;
1011 
1012 			nla_strscpy(tmp, nla, sizeof(tmp));
1013 			val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
1014 		} else {
1015 			if (nla_len(nla) != sizeof(u32))
1016 				return false;
1017 			val = nla_get_u32(nla);
1018 		}
1019 
1020 		fi_val = fi->fib_metrics->metrics[type - 1];
1021 		if (type == RTAX_FEATURES)
1022 			fi_val &= ~DST_FEATURE_ECN_CA;
1023 
1024 		if (fi_val != val)
1025 			return false;
1026 	}
1027 
1028 	return true;
1029 }
1030 
fib_check_nh_v6_gw(struct net * net,struct fib_nh * nh,u32 table,struct netlink_ext_ack * extack)1031 static int fib_check_nh_v6_gw(struct net *net, struct fib_nh *nh,
1032 			      u32 table, struct netlink_ext_ack *extack)
1033 {
1034 	struct fib6_config cfg = {
1035 		.fc_table = table,
1036 		.fc_flags = nh->fib_nh_flags | RTF_GATEWAY,
1037 		.fc_ifindex = nh->fib_nh_oif,
1038 		.fc_gateway = nh->fib_nh_gw6,
1039 	};
1040 	struct fib6_nh fib6_nh = {};
1041 	int err;
1042 
1043 	err = ipv6_stub->fib6_nh_init(net, &fib6_nh, &cfg, GFP_KERNEL, extack);
1044 	if (!err) {
1045 		nh->fib_nh_dev = fib6_nh.fib_nh_dev;
1046 		dev_hold(nh->fib_nh_dev);
1047 		nh->fib_nh_oif = nh->fib_nh_dev->ifindex;
1048 		nh->fib_nh_scope = RT_SCOPE_LINK;
1049 
1050 		ipv6_stub->fib6_nh_release(&fib6_nh);
1051 	}
1052 
1053 	return err;
1054 }
1055 
1056 /*
1057  * Picture
1058  * -------
1059  *
1060  * Semantics of nexthop is very messy by historical reasons.
1061  * We have to take into account, that:
1062  * a) gateway can be actually local interface address,
1063  *    so that gatewayed route is direct.
1064  * b) gateway must be on-link address, possibly
1065  *    described not by an ifaddr, but also by a direct route.
1066  * c) If both gateway and interface are specified, they should not
1067  *    contradict.
1068  * d) If we use tunnel routes, gateway could be not on-link.
1069  *
1070  * Attempt to reconcile all of these (alas, self-contradictory) conditions
1071  * results in pretty ugly and hairy code with obscure logic.
1072  *
1073  * I chose to generalized it instead, so that the size
1074  * of code does not increase practically, but it becomes
1075  * much more general.
1076  * Every prefix is assigned a "scope" value: "host" is local address,
1077  * "link" is direct route,
1078  * [ ... "site" ... "interior" ... ]
1079  * and "universe" is true gateway route with global meaning.
1080  *
1081  * Every prefix refers to a set of "nexthop"s (gw, oif),
1082  * where gw must have narrower scope. This recursion stops
1083  * when gw has LOCAL scope or if "nexthop" is declared ONLINK,
1084  * which means that gw is forced to be on link.
1085  *
1086  * Code is still hairy, but now it is apparently logically
1087  * consistent and very flexible. F.e. as by-product it allows
1088  * to co-exists in peace independent exterior and interior
1089  * routing processes.
1090  *
1091  * Normally it looks as following.
1092  *
1093  * {universe prefix}  -> (gw, oif) [scope link]
1094  *		  |
1095  *		  |-> {link prefix} -> (gw, oif) [scope local]
1096  *					|
1097  *					|-> {local prefix} (terminal node)
1098  */
fib_check_nh_v4_gw(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1099 static int fib_check_nh_v4_gw(struct net *net, struct fib_nh *nh, u32 table,
1100 			      u8 scope, struct netlink_ext_ack *extack)
1101 {
1102 	struct net_device *dev;
1103 	struct fib_result res;
1104 	int err = 0;
1105 
1106 	if (nh->fib_nh_flags & RTNH_F_ONLINK) {
1107 		unsigned int addr_type;
1108 
1109 		if (scope >= RT_SCOPE_LINK) {
1110 			NL_SET_ERR_MSG(extack, "Nexthop has invalid scope");
1111 			return -EINVAL;
1112 		}
1113 		dev = __dev_get_by_index(net, nh->fib_nh_oif);
1114 		if (!dev) {
1115 			NL_SET_ERR_MSG(extack, "Nexthop device required for onlink");
1116 			return -ENODEV;
1117 		}
1118 		if (!(dev->flags & IFF_UP)) {
1119 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1120 			return -ENETDOWN;
1121 		}
1122 		addr_type = inet_addr_type_dev_table(net, dev, nh->fib_nh_gw4);
1123 		if (addr_type != RTN_UNICAST) {
1124 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1125 			return -EINVAL;
1126 		}
1127 		if (!netif_carrier_ok(dev))
1128 			nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1129 		nh->fib_nh_dev = dev;
1130 		dev_hold(dev);
1131 		nh->fib_nh_scope = RT_SCOPE_LINK;
1132 		return 0;
1133 	}
1134 	rcu_read_lock();
1135 	{
1136 		struct fib_table *tbl = NULL;
1137 		struct flowi4 fl4 = {
1138 			.daddr = nh->fib_nh_gw4,
1139 			.flowi4_scope = scope + 1,
1140 			.flowi4_oif = nh->fib_nh_oif,
1141 			.flowi4_iif = LOOPBACK_IFINDEX,
1142 		};
1143 
1144 		/* It is not necessary, but requires a bit of thinking */
1145 		if (fl4.flowi4_scope < RT_SCOPE_LINK)
1146 			fl4.flowi4_scope = RT_SCOPE_LINK;
1147 
1148 		if (table && table != RT_TABLE_MAIN)
1149 			tbl = fib_get_table(net, table);
1150 
1151 		if (tbl)
1152 			err = fib_table_lookup(tbl, &fl4, &res,
1153 					       FIB_LOOKUP_IGNORE_LINKSTATE |
1154 					       FIB_LOOKUP_NOREF);
1155 
1156 		/* on error or if no table given do full lookup. This
1157 		 * is needed for example when nexthops are in the local
1158 		 * table rather than the given table
1159 		 */
1160 		if (!tbl || err) {
1161 			err = fib_lookup(net, &fl4, &res,
1162 					 FIB_LOOKUP_IGNORE_LINKSTATE);
1163 		}
1164 
1165 		if (err) {
1166 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1167 			goto out;
1168 		}
1169 	}
1170 
1171 	err = -EINVAL;
1172 	if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) {
1173 		NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1174 		goto out;
1175 	}
1176 	nh->fib_nh_scope = res.scope;
1177 	nh->fib_nh_oif = FIB_RES_OIF(res);
1178 	nh->fib_nh_dev = dev = FIB_RES_DEV(res);
1179 	if (!dev) {
1180 		NL_SET_ERR_MSG(extack,
1181 			       "No egress device for nexthop gateway");
1182 		goto out;
1183 	}
1184 	dev_hold(dev);
1185 	if (!netif_carrier_ok(dev))
1186 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1187 	err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN;
1188 out:
1189 	rcu_read_unlock();
1190 	return err;
1191 }
1192 
fib_check_nh_nongw(struct net * net,struct fib_nh * nh,struct netlink_ext_ack * extack)1193 static int fib_check_nh_nongw(struct net *net, struct fib_nh *nh,
1194 			      struct netlink_ext_ack *extack)
1195 {
1196 	struct in_device *in_dev;
1197 	int err;
1198 
1199 	if (nh->fib_nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) {
1200 		NL_SET_ERR_MSG(extack,
1201 			       "Invalid flags for nexthop - PERVASIVE and ONLINK can not be set");
1202 		return -EINVAL;
1203 	}
1204 
1205 	rcu_read_lock();
1206 
1207 	err = -ENODEV;
1208 	in_dev = inetdev_by_index(net, nh->fib_nh_oif);
1209 	if (!in_dev)
1210 		goto out;
1211 	err = -ENETDOWN;
1212 	if (!(in_dev->dev->flags & IFF_UP)) {
1213 		NL_SET_ERR_MSG(extack, "Device for nexthop is not up");
1214 		goto out;
1215 	}
1216 
1217 	nh->fib_nh_dev = in_dev->dev;
1218 	dev_hold(nh->fib_nh_dev);
1219 	nh->fib_nh_scope = RT_SCOPE_HOST;
1220 	if (!netif_carrier_ok(nh->fib_nh_dev))
1221 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1222 	err = 0;
1223 out:
1224 	rcu_read_unlock();
1225 	return err;
1226 }
1227 
fib_check_nh(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1228 int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
1229 		 struct netlink_ext_ack *extack)
1230 {
1231 	int err;
1232 
1233 	if (nh->fib_nh_gw_family == AF_INET)
1234 		err = fib_check_nh_v4_gw(net, nh, table, scope, extack);
1235 	else if (nh->fib_nh_gw_family == AF_INET6)
1236 		err = fib_check_nh_v6_gw(net, nh, table, extack);
1237 	else
1238 		err = fib_check_nh_nongw(net, nh, extack);
1239 
1240 	return err;
1241 }
1242 
fib_laddr_hashfn(__be32 val)1243 static inline unsigned int fib_laddr_hashfn(__be32 val)
1244 {
1245 	unsigned int mask = (fib_info_hash_size - 1);
1246 
1247 	return ((__force u32)val ^
1248 		((__force u32)val >> 7) ^
1249 		((__force u32)val >> 14)) & mask;
1250 }
1251 
fib_info_hash_alloc(int bytes)1252 static struct hlist_head *fib_info_hash_alloc(int bytes)
1253 {
1254 	if (bytes <= PAGE_SIZE)
1255 		return kzalloc(bytes, GFP_KERNEL);
1256 	else
1257 		return (struct hlist_head *)
1258 			__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1259 					 get_order(bytes));
1260 }
1261 
fib_info_hash_free(struct hlist_head * hash,int bytes)1262 static void fib_info_hash_free(struct hlist_head *hash, int bytes)
1263 {
1264 	if (!hash)
1265 		return;
1266 
1267 	if (bytes <= PAGE_SIZE)
1268 		kfree(hash);
1269 	else
1270 		free_pages((unsigned long) hash, get_order(bytes));
1271 }
1272 
fib_info_hash_move(struct hlist_head * new_info_hash,struct hlist_head * new_laddrhash,unsigned int new_size)1273 static void fib_info_hash_move(struct hlist_head *new_info_hash,
1274 			       struct hlist_head *new_laddrhash,
1275 			       unsigned int new_size)
1276 {
1277 	struct hlist_head *old_info_hash, *old_laddrhash;
1278 	unsigned int old_size = fib_info_hash_size;
1279 	unsigned int i, bytes;
1280 
1281 	spin_lock_bh(&fib_info_lock);
1282 	old_info_hash = fib_info_hash;
1283 	old_laddrhash = fib_info_laddrhash;
1284 	fib_info_hash_size = new_size;
1285 
1286 	for (i = 0; i < old_size; i++) {
1287 		struct hlist_head *head = &fib_info_hash[i];
1288 		struct hlist_node *n;
1289 		struct fib_info *fi;
1290 
1291 		hlist_for_each_entry_safe(fi, n, head, fib_hash) {
1292 			struct hlist_head *dest;
1293 			unsigned int new_hash;
1294 
1295 			new_hash = fib_info_hashfn(fi);
1296 			dest = &new_info_hash[new_hash];
1297 			hlist_add_head(&fi->fib_hash, dest);
1298 		}
1299 	}
1300 	fib_info_hash = new_info_hash;
1301 
1302 	for (i = 0; i < old_size; i++) {
1303 		struct hlist_head *lhead = &fib_info_laddrhash[i];
1304 		struct hlist_node *n;
1305 		struct fib_info *fi;
1306 
1307 		hlist_for_each_entry_safe(fi, n, lhead, fib_lhash) {
1308 			struct hlist_head *ldest;
1309 			unsigned int new_hash;
1310 
1311 			new_hash = fib_laddr_hashfn(fi->fib_prefsrc);
1312 			ldest = &new_laddrhash[new_hash];
1313 			hlist_add_head(&fi->fib_lhash, ldest);
1314 		}
1315 	}
1316 	fib_info_laddrhash = new_laddrhash;
1317 
1318 	spin_unlock_bh(&fib_info_lock);
1319 
1320 	bytes = old_size * sizeof(struct hlist_head *);
1321 	fib_info_hash_free(old_info_hash, bytes);
1322 	fib_info_hash_free(old_laddrhash, bytes);
1323 }
1324 
fib_info_update_nhc_saddr(struct net * net,struct fib_nh_common * nhc,unsigned char scope)1325 __be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
1326 				 unsigned char scope)
1327 {
1328 	struct fib_nh *nh;
1329 
1330 	if (nhc->nhc_family != AF_INET)
1331 		return inet_select_addr(nhc->nhc_dev, 0, scope);
1332 
1333 	nh = container_of(nhc, struct fib_nh, nh_common);
1334 	nh->nh_saddr = inet_select_addr(nh->fib_nh_dev, nh->fib_nh_gw4, scope);
1335 	nh->nh_saddr_genid = atomic_read(&net->ipv4.dev_addr_genid);
1336 
1337 	return nh->nh_saddr;
1338 }
1339 
fib_result_prefsrc(struct net * net,struct fib_result * res)1340 __be32 fib_result_prefsrc(struct net *net, struct fib_result *res)
1341 {
1342 	struct fib_nh_common *nhc = res->nhc;
1343 
1344 	if (res->fi->fib_prefsrc)
1345 		return res->fi->fib_prefsrc;
1346 
1347 	if (nhc->nhc_family == AF_INET) {
1348 		struct fib_nh *nh;
1349 
1350 		nh = container_of(nhc, struct fib_nh, nh_common);
1351 		if (nh->nh_saddr_genid == atomic_read(&net->ipv4.dev_addr_genid))
1352 			return nh->nh_saddr;
1353 	}
1354 
1355 	return fib_info_update_nhc_saddr(net, nhc, res->fi->fib_scope);
1356 }
1357 
fib_valid_prefsrc(struct fib_config * cfg,__be32 fib_prefsrc)1358 static bool fib_valid_prefsrc(struct fib_config *cfg, __be32 fib_prefsrc)
1359 {
1360 	if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst ||
1361 	    fib_prefsrc != cfg->fc_dst) {
1362 		u32 tb_id = cfg->fc_table;
1363 		int rc;
1364 
1365 		if (tb_id == RT_TABLE_MAIN)
1366 			tb_id = RT_TABLE_LOCAL;
1367 
1368 		rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1369 					  fib_prefsrc, tb_id);
1370 
1371 		if (rc != RTN_LOCAL && tb_id != RT_TABLE_LOCAL) {
1372 			rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1373 						  fib_prefsrc, RT_TABLE_LOCAL);
1374 		}
1375 
1376 		if (rc != RTN_LOCAL)
1377 			return false;
1378 	}
1379 	return true;
1380 }
1381 
fib_create_info(struct fib_config * cfg,struct netlink_ext_ack * extack)1382 struct fib_info *fib_create_info(struct fib_config *cfg,
1383 				 struct netlink_ext_ack *extack)
1384 {
1385 	int err;
1386 	struct fib_info *fi = NULL;
1387 	struct nexthop *nh = NULL;
1388 	struct fib_info *ofi;
1389 	int nhs = 1;
1390 	struct net *net = cfg->fc_nlinfo.nl_net;
1391 
1392 	if (cfg->fc_type > RTN_MAX)
1393 		goto err_inval;
1394 
1395 	/* Fast check to catch the most weird cases */
1396 	if (fib_props[cfg->fc_type].scope > cfg->fc_scope) {
1397 		NL_SET_ERR_MSG(extack, "Invalid scope");
1398 		goto err_inval;
1399 	}
1400 
1401 	if (cfg->fc_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
1402 		NL_SET_ERR_MSG(extack,
1403 			       "Invalid rtm_flags - can not contain DEAD or LINKDOWN");
1404 		goto err_inval;
1405 	}
1406 
1407 	if (cfg->fc_nh_id) {
1408 		if (!cfg->fc_mx) {
1409 			fi = fib_find_info_nh(net, cfg);
1410 			if (fi) {
1411 				refcount_inc(&fi->fib_treeref);
1412 				return fi;
1413 			}
1414 		}
1415 
1416 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
1417 		if (!nh) {
1418 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
1419 			goto err_inval;
1420 		}
1421 		nhs = 0;
1422 	}
1423 
1424 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1425 	if (cfg->fc_mp) {
1426 		nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len, extack);
1427 		if (nhs == 0)
1428 			goto err_inval;
1429 	}
1430 #endif
1431 
1432 	err = -ENOBUFS;
1433 	if (fib_info_cnt >= fib_info_hash_size) {
1434 		unsigned int new_size = fib_info_hash_size << 1;
1435 		struct hlist_head *new_info_hash;
1436 		struct hlist_head *new_laddrhash;
1437 		unsigned int bytes;
1438 
1439 		if (!new_size)
1440 			new_size = 16;
1441 		bytes = new_size * sizeof(struct hlist_head *);
1442 		new_info_hash = fib_info_hash_alloc(bytes);
1443 		new_laddrhash = fib_info_hash_alloc(bytes);
1444 		if (!new_info_hash || !new_laddrhash) {
1445 			fib_info_hash_free(new_info_hash, bytes);
1446 			fib_info_hash_free(new_laddrhash, bytes);
1447 		} else
1448 			fib_info_hash_move(new_info_hash, new_laddrhash, new_size);
1449 
1450 		if (!fib_info_hash_size)
1451 			goto failure;
1452 	}
1453 
1454 	fi = kzalloc(struct_size(fi, fib_nh, nhs), GFP_KERNEL);
1455 	if (!fi)
1456 		goto failure;
1457 	fi->fib_metrics = ip_fib_metrics_init(fi->fib_net, cfg->fc_mx,
1458 					      cfg->fc_mx_len, extack);
1459 	if (IS_ERR(fi->fib_metrics)) {
1460 		err = PTR_ERR(fi->fib_metrics);
1461 		kfree(fi);
1462 		return ERR_PTR(err);
1463 	}
1464 
1465 	fib_info_cnt++;
1466 	fi->fib_net = net;
1467 	fi->fib_protocol = cfg->fc_protocol;
1468 	fi->fib_scope = cfg->fc_scope;
1469 	fi->fib_flags = cfg->fc_flags;
1470 	fi->fib_priority = cfg->fc_priority;
1471 	fi->fib_prefsrc = cfg->fc_prefsrc;
1472 	fi->fib_type = cfg->fc_type;
1473 	fi->fib_tb_id = cfg->fc_table;
1474 
1475 	fi->fib_nhs = nhs;
1476 	if (nh) {
1477 		if (!nexthop_get(nh)) {
1478 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1479 			err = -EINVAL;
1480 		} else {
1481 			err = 0;
1482 			fi->nh = nh;
1483 		}
1484 	} else {
1485 		change_nexthops(fi) {
1486 			nexthop_nh->nh_parent = fi;
1487 		} endfor_nexthops(fi)
1488 
1489 		if (cfg->fc_mp)
1490 			err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg,
1491 					  extack);
1492 		else
1493 			err = fib_nh_init(net, fi->fib_nh, cfg, 1, extack);
1494 	}
1495 
1496 	if (err != 0)
1497 		goto failure;
1498 
1499 	if (fib_props[cfg->fc_type].error) {
1500 		if (cfg->fc_gw_family || cfg->fc_oif || cfg->fc_mp) {
1501 			NL_SET_ERR_MSG(extack,
1502 				       "Gateway, device and multipath can not be specified for this route type");
1503 			goto err_inval;
1504 		}
1505 		goto link_it;
1506 	} else {
1507 		switch (cfg->fc_type) {
1508 		case RTN_UNICAST:
1509 		case RTN_LOCAL:
1510 		case RTN_BROADCAST:
1511 		case RTN_ANYCAST:
1512 		case RTN_MULTICAST:
1513 			break;
1514 		default:
1515 			NL_SET_ERR_MSG(extack, "Invalid route type");
1516 			goto err_inval;
1517 		}
1518 	}
1519 
1520 	if (cfg->fc_scope > RT_SCOPE_HOST) {
1521 		NL_SET_ERR_MSG(extack, "Invalid scope");
1522 		goto err_inval;
1523 	}
1524 
1525 	if (fi->nh) {
1526 		err = fib_check_nexthop(fi->nh, cfg->fc_scope, extack);
1527 		if (err)
1528 			goto failure;
1529 	} else if (cfg->fc_scope == RT_SCOPE_HOST) {
1530 		struct fib_nh *nh = fi->fib_nh;
1531 
1532 		/* Local address is added. */
1533 		if (nhs != 1) {
1534 			NL_SET_ERR_MSG(extack,
1535 				       "Route with host scope can not have multiple nexthops");
1536 			goto err_inval;
1537 		}
1538 		if (nh->fib_nh_gw_family) {
1539 			NL_SET_ERR_MSG(extack,
1540 				       "Route with host scope can not have a gateway");
1541 			goto err_inval;
1542 		}
1543 		nh->fib_nh_scope = RT_SCOPE_NOWHERE;
1544 		nh->fib_nh_dev = dev_get_by_index(net, nh->fib_nh_oif);
1545 		err = -ENODEV;
1546 		if (!nh->fib_nh_dev)
1547 			goto failure;
1548 	} else {
1549 		int linkdown = 0;
1550 
1551 		change_nexthops(fi) {
1552 			err = fib_check_nh(cfg->fc_nlinfo.nl_net, nexthop_nh,
1553 					   cfg->fc_table, cfg->fc_scope,
1554 					   extack);
1555 			if (err != 0)
1556 				goto failure;
1557 			if (nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN)
1558 				linkdown++;
1559 		} endfor_nexthops(fi)
1560 		if (linkdown == fi->fib_nhs)
1561 			fi->fib_flags |= RTNH_F_LINKDOWN;
1562 	}
1563 
1564 	if (fi->fib_prefsrc && !fib_valid_prefsrc(cfg, fi->fib_prefsrc)) {
1565 		NL_SET_ERR_MSG(extack, "Invalid prefsrc address");
1566 		goto err_inval;
1567 	}
1568 
1569 	if (!fi->nh) {
1570 		change_nexthops(fi) {
1571 			fib_info_update_nhc_saddr(net, &nexthop_nh->nh_common,
1572 						  fi->fib_scope);
1573 			if (nexthop_nh->fib_nh_gw_family == AF_INET6)
1574 				fi->fib_nh_is_v6 = true;
1575 		} endfor_nexthops(fi)
1576 
1577 		fib_rebalance(fi);
1578 	}
1579 
1580 link_it:
1581 	ofi = fib_find_info(fi);
1582 	if (ofi) {
1583 		fi->fib_dead = 1;
1584 		free_fib_info(fi);
1585 		refcount_inc(&ofi->fib_treeref);
1586 		return ofi;
1587 	}
1588 
1589 	refcount_set(&fi->fib_treeref, 1);
1590 	refcount_set(&fi->fib_clntref, 1);
1591 	spin_lock_bh(&fib_info_lock);
1592 	hlist_add_head(&fi->fib_hash,
1593 		       &fib_info_hash[fib_info_hashfn(fi)]);
1594 	if (fi->fib_prefsrc) {
1595 		struct hlist_head *head;
1596 
1597 		head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)];
1598 		hlist_add_head(&fi->fib_lhash, head);
1599 	}
1600 	if (fi->nh) {
1601 		list_add(&fi->nh_list, &nh->fi_list);
1602 	} else {
1603 		change_nexthops(fi) {
1604 			struct hlist_head *head;
1605 			unsigned int hash;
1606 
1607 			if (!nexthop_nh->fib_nh_dev)
1608 				continue;
1609 			hash = fib_devindex_hashfn(nexthop_nh->fib_nh_dev->ifindex);
1610 			head = &fib_info_devhash[hash];
1611 			hlist_add_head(&nexthop_nh->nh_hash, head);
1612 		} endfor_nexthops(fi)
1613 	}
1614 	spin_unlock_bh(&fib_info_lock);
1615 	return fi;
1616 
1617 err_inval:
1618 	err = -EINVAL;
1619 
1620 failure:
1621 	if (fi) {
1622 		fi->fib_dead = 1;
1623 		free_fib_info(fi);
1624 	}
1625 
1626 	return ERR_PTR(err);
1627 }
1628 
fib_nexthop_info(struct sk_buff * skb,const struct fib_nh_common * nhc,u8 rt_family,unsigned char * flags,bool skip_oif)1629 int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nhc,
1630 		     u8 rt_family, unsigned char *flags, bool skip_oif)
1631 {
1632 	if (nhc->nhc_flags & RTNH_F_DEAD)
1633 		*flags |= RTNH_F_DEAD;
1634 
1635 	if (nhc->nhc_flags & RTNH_F_LINKDOWN) {
1636 		*flags |= RTNH_F_LINKDOWN;
1637 
1638 		rcu_read_lock();
1639 		switch (nhc->nhc_family) {
1640 		case AF_INET:
1641 			if (ip_ignore_linkdown(nhc->nhc_dev))
1642 				*flags |= RTNH_F_DEAD;
1643 			break;
1644 		case AF_INET6:
1645 			if (ip6_ignore_linkdown(nhc->nhc_dev))
1646 				*flags |= RTNH_F_DEAD;
1647 			break;
1648 		}
1649 		rcu_read_unlock();
1650 	}
1651 
1652 	switch (nhc->nhc_gw_family) {
1653 	case AF_INET:
1654 		if (nla_put_in_addr(skb, RTA_GATEWAY, nhc->nhc_gw.ipv4))
1655 			goto nla_put_failure;
1656 		break;
1657 	case AF_INET6:
1658 		/* if gateway family does not match nexthop family
1659 		 * gateway is encoded as RTA_VIA
1660 		 */
1661 		if (rt_family != nhc->nhc_gw_family) {
1662 			int alen = sizeof(struct in6_addr);
1663 			struct nlattr *nla;
1664 			struct rtvia *via;
1665 
1666 			nla = nla_reserve(skb, RTA_VIA, alen + 2);
1667 			if (!nla)
1668 				goto nla_put_failure;
1669 
1670 			via = nla_data(nla);
1671 			via->rtvia_family = AF_INET6;
1672 			memcpy(via->rtvia_addr, &nhc->nhc_gw.ipv6, alen);
1673 		} else if (nla_put_in6_addr(skb, RTA_GATEWAY,
1674 					    &nhc->nhc_gw.ipv6) < 0) {
1675 			goto nla_put_failure;
1676 		}
1677 		break;
1678 	}
1679 
1680 	*flags |= (nhc->nhc_flags &
1681 		   (RTNH_F_ONLINK | RTNH_F_OFFLOAD | RTNH_F_TRAP));
1682 
1683 	if (!skip_oif && nhc->nhc_dev &&
1684 	    nla_put_u32(skb, RTA_OIF, nhc->nhc_dev->ifindex))
1685 		goto nla_put_failure;
1686 
1687 	if (nhc->nhc_lwtstate &&
1688 	    lwtunnel_fill_encap(skb, nhc->nhc_lwtstate,
1689 				RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
1690 		goto nla_put_failure;
1691 
1692 	return 0;
1693 
1694 nla_put_failure:
1695 	return -EMSGSIZE;
1696 }
1697 EXPORT_SYMBOL_GPL(fib_nexthop_info);
1698 
1699 #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)1700 int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nhc,
1701 		    int nh_weight, u8 rt_family, u32 nh_tclassid)
1702 {
1703 	const struct net_device *dev = nhc->nhc_dev;
1704 	struct rtnexthop *rtnh;
1705 	unsigned char flags = 0;
1706 
1707 	rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1708 	if (!rtnh)
1709 		goto nla_put_failure;
1710 
1711 	rtnh->rtnh_hops = nh_weight - 1;
1712 	rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
1713 
1714 	if (fib_nexthop_info(skb, nhc, rt_family, &flags, true) < 0)
1715 		goto nla_put_failure;
1716 
1717 	rtnh->rtnh_flags = flags;
1718 
1719 	if (nh_tclassid && nla_put_u32(skb, RTA_FLOW, nh_tclassid))
1720 		goto nla_put_failure;
1721 
1722 	/* length of rtnetlink header + attributes */
1723 	rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
1724 
1725 	return 0;
1726 
1727 nla_put_failure:
1728 	return -EMSGSIZE;
1729 }
1730 EXPORT_SYMBOL_GPL(fib_add_nexthop);
1731 #endif
1732 
1733 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1734 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1735 {
1736 	struct nlattr *mp;
1737 
1738 	mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
1739 	if (!mp)
1740 		goto nla_put_failure;
1741 
1742 	if (unlikely(fi->nh)) {
1743 		if (nexthop_mpath_fill_node(skb, fi->nh, AF_INET) < 0)
1744 			goto nla_put_failure;
1745 		goto mp_end;
1746 	}
1747 
1748 	for_nexthops(fi) {
1749 		u32 nh_tclassid = 0;
1750 #ifdef CONFIG_IP_ROUTE_CLASSID
1751 		nh_tclassid = nh->nh_tclassid;
1752 #endif
1753 		if (fib_add_nexthop(skb, &nh->nh_common, nh->fib_nh_weight,
1754 				    AF_INET, nh_tclassid) < 0)
1755 			goto nla_put_failure;
1756 	} endfor_nexthops(fi);
1757 
1758 mp_end:
1759 	nla_nest_end(skb, mp);
1760 
1761 	return 0;
1762 
1763 nla_put_failure:
1764 	return -EMSGSIZE;
1765 }
1766 #else
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1767 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1768 {
1769 	return 0;
1770 }
1771 #endif
1772 
fib_dump_info(struct sk_buff * skb,u32 portid,u32 seq,int event,const struct fib_rt_info * fri,unsigned int flags)1773 int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
1774 		  const struct fib_rt_info *fri, unsigned int flags)
1775 {
1776 	unsigned int nhs = fib_info_num_path(fri->fi);
1777 	struct fib_info *fi = fri->fi;
1778 	u32 tb_id = fri->tb_id;
1779 	struct nlmsghdr *nlh;
1780 	struct rtmsg *rtm;
1781 
1782 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1783 	if (!nlh)
1784 		return -EMSGSIZE;
1785 
1786 	rtm = nlmsg_data(nlh);
1787 	rtm->rtm_family = AF_INET;
1788 	rtm->rtm_dst_len = fri->dst_len;
1789 	rtm->rtm_src_len = 0;
1790 	rtm->rtm_tos = fri->tos;
1791 	if (tb_id < 256)
1792 		rtm->rtm_table = tb_id;
1793 	else
1794 		rtm->rtm_table = RT_TABLE_COMPAT;
1795 	if (nla_put_u32(skb, RTA_TABLE, tb_id))
1796 		goto nla_put_failure;
1797 	rtm->rtm_type = fri->type;
1798 	rtm->rtm_flags = fi->fib_flags;
1799 	rtm->rtm_scope = fi->fib_scope;
1800 	rtm->rtm_protocol = fi->fib_protocol;
1801 
1802 	if (rtm->rtm_dst_len &&
1803 	    nla_put_in_addr(skb, RTA_DST, fri->dst))
1804 		goto nla_put_failure;
1805 	if (fi->fib_priority &&
1806 	    nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority))
1807 		goto nla_put_failure;
1808 	if (rtnetlink_put_metrics(skb, fi->fib_metrics->metrics) < 0)
1809 		goto nla_put_failure;
1810 
1811 	if (fi->fib_prefsrc &&
1812 	    nla_put_in_addr(skb, RTA_PREFSRC, fi->fib_prefsrc))
1813 		goto nla_put_failure;
1814 
1815 	if (fi->nh) {
1816 		if (nla_put_u32(skb, RTA_NH_ID, fi->nh->id))
1817 			goto nla_put_failure;
1818 		if (nexthop_is_blackhole(fi->nh))
1819 			rtm->rtm_type = RTN_BLACKHOLE;
1820 		if (!fi->fib_net->ipv4.sysctl_nexthop_compat_mode)
1821 			goto offload;
1822 	}
1823 
1824 	if (nhs == 1) {
1825 		const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
1826 		unsigned char flags = 0;
1827 
1828 		if (fib_nexthop_info(skb, nhc, AF_INET, &flags, false) < 0)
1829 			goto nla_put_failure;
1830 
1831 		rtm->rtm_flags = flags;
1832 #ifdef CONFIG_IP_ROUTE_CLASSID
1833 		if (nhc->nhc_family == AF_INET) {
1834 			struct fib_nh *nh;
1835 
1836 			nh = container_of(nhc, struct fib_nh, nh_common);
1837 			if (nh->nh_tclassid &&
1838 			    nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid))
1839 				goto nla_put_failure;
1840 		}
1841 #endif
1842 	} else {
1843 		if (fib_add_multipath(skb, fi) < 0)
1844 			goto nla_put_failure;
1845 	}
1846 
1847 offload:
1848 	if (fri->offload)
1849 		rtm->rtm_flags |= RTM_F_OFFLOAD;
1850 	if (fri->trap)
1851 		rtm->rtm_flags |= RTM_F_TRAP;
1852 	if (fri->offload_failed)
1853 		rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
1854 
1855 	nlmsg_end(skb, nlh);
1856 	return 0;
1857 
1858 nla_put_failure:
1859 	nlmsg_cancel(skb, nlh);
1860 	return -EMSGSIZE;
1861 }
1862 
1863 /*
1864  * Update FIB if:
1865  * - local address disappeared -> we must delete all the entries
1866  *   referring to it.
1867  * - device went down -> we must shutdown all nexthops going via it.
1868  */
fib_sync_down_addr(struct net_device * dev,__be32 local)1869 int fib_sync_down_addr(struct net_device *dev, __be32 local)
1870 {
1871 	int ret = 0;
1872 	unsigned int hash = fib_laddr_hashfn(local);
1873 	struct hlist_head *head = &fib_info_laddrhash[hash];
1874 	int tb_id = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
1875 	struct net *net = dev_net(dev);
1876 	struct fib_info *fi;
1877 
1878 	if (!fib_info_laddrhash || local == 0)
1879 		return 0;
1880 
1881 	hlist_for_each_entry(fi, head, fib_lhash) {
1882 		if (!net_eq(fi->fib_net, net) ||
1883 		    fi->fib_tb_id != tb_id)
1884 			continue;
1885 		if (fi->fib_prefsrc == local) {
1886 			fi->fib_flags |= RTNH_F_DEAD;
1887 			ret++;
1888 		}
1889 	}
1890 	return ret;
1891 }
1892 
call_fib_nh_notifiers(struct fib_nh * nh,enum fib_event_type event_type)1893 static int call_fib_nh_notifiers(struct fib_nh *nh,
1894 				 enum fib_event_type event_type)
1895 {
1896 	bool ignore_link_down = ip_ignore_linkdown(nh->fib_nh_dev);
1897 	struct fib_nh_notifier_info info = {
1898 		.fib_nh = nh,
1899 	};
1900 
1901 	switch (event_type) {
1902 	case FIB_EVENT_NH_ADD:
1903 		if (nh->fib_nh_flags & RTNH_F_DEAD)
1904 			break;
1905 		if (ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN)
1906 			break;
1907 		return call_fib4_notifiers(dev_net(nh->fib_nh_dev), event_type,
1908 					   &info.info);
1909 	case FIB_EVENT_NH_DEL:
1910 		if ((ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN) ||
1911 		    (nh->fib_nh_flags & RTNH_F_DEAD))
1912 			return call_fib4_notifiers(dev_net(nh->fib_nh_dev),
1913 						   event_type, &info.info);
1914 		break;
1915 	default:
1916 		break;
1917 	}
1918 
1919 	return NOTIFY_DONE;
1920 }
1921 
1922 /* Update the PMTU of exceptions when:
1923  * - the new MTU of the first hop becomes smaller than the PMTU
1924  * - the old MTU was the same as the PMTU, and it limited discovery of
1925  *   larger MTUs on the path. With that limit raised, we can now
1926  *   discover larger MTUs
1927  * A special case is locked exceptions, for which the PMTU is smaller
1928  * than the minimal accepted PMTU:
1929  * - if the new MTU is greater than the PMTU, don't make any change
1930  * - otherwise, unlock and set PMTU
1931  */
fib_nhc_update_mtu(struct fib_nh_common * nhc,u32 new,u32 orig)1932 void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig)
1933 {
1934 	struct fnhe_hash_bucket *bucket;
1935 	int i;
1936 
1937 	bucket = rcu_dereference_protected(nhc->nhc_exceptions, 1);
1938 	if (!bucket)
1939 		return;
1940 
1941 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
1942 		struct fib_nh_exception *fnhe;
1943 
1944 		for (fnhe = rcu_dereference_protected(bucket[i].chain, 1);
1945 		     fnhe;
1946 		     fnhe = rcu_dereference_protected(fnhe->fnhe_next, 1)) {
1947 			if (fnhe->fnhe_mtu_locked) {
1948 				if (new <= fnhe->fnhe_pmtu) {
1949 					fnhe->fnhe_pmtu = new;
1950 					fnhe->fnhe_mtu_locked = false;
1951 				}
1952 			} else if (new < fnhe->fnhe_pmtu ||
1953 				   orig == fnhe->fnhe_pmtu) {
1954 				fnhe->fnhe_pmtu = new;
1955 			}
1956 		}
1957 	}
1958 }
1959 
fib_sync_mtu(struct net_device * dev,u32 orig_mtu)1960 void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
1961 {
1962 	unsigned int hash = fib_devindex_hashfn(dev->ifindex);
1963 	struct hlist_head *head = &fib_info_devhash[hash];
1964 	struct fib_nh *nh;
1965 
1966 	hlist_for_each_entry(nh, head, nh_hash) {
1967 		if (nh->fib_nh_dev == dev)
1968 			fib_nhc_update_mtu(&nh->nh_common, dev->mtu, orig_mtu);
1969 	}
1970 }
1971 
1972 /* Event              force Flags           Description
1973  * NETDEV_CHANGE      0     LINKDOWN        Carrier OFF, not for scope host
1974  * NETDEV_DOWN        0     LINKDOWN|DEAD   Link down, not for scope host
1975  * NETDEV_DOWN        1     LINKDOWN|DEAD   Last address removed
1976  * NETDEV_UNREGISTER  1     LINKDOWN|DEAD   Device removed
1977  *
1978  * only used when fib_nh is built into fib_info
1979  */
fib_sync_down_dev(struct net_device * dev,unsigned long event,bool force)1980 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force)
1981 {
1982 	int ret = 0;
1983 	int scope = RT_SCOPE_NOWHERE;
1984 	struct fib_info *prev_fi = NULL;
1985 	unsigned int hash = fib_devindex_hashfn(dev->ifindex);
1986 	struct hlist_head *head = &fib_info_devhash[hash];
1987 	struct fib_nh *nh;
1988 
1989 	if (force)
1990 		scope = -1;
1991 
1992 	hlist_for_each_entry(nh, head, nh_hash) {
1993 		struct fib_info *fi = nh->nh_parent;
1994 		int dead;
1995 
1996 		BUG_ON(!fi->fib_nhs);
1997 		if (nh->fib_nh_dev != dev || fi == prev_fi)
1998 			continue;
1999 		prev_fi = fi;
2000 		dead = 0;
2001 		change_nexthops(fi) {
2002 			if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD)
2003 				dead++;
2004 			else if (nexthop_nh->fib_nh_dev == dev &&
2005 				 nexthop_nh->fib_nh_scope != scope) {
2006 				switch (event) {
2007 				case NETDEV_DOWN:
2008 				case NETDEV_UNREGISTER:
2009 					nexthop_nh->fib_nh_flags |= RTNH_F_DEAD;
2010 					fallthrough;
2011 				case NETDEV_CHANGE:
2012 					nexthop_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
2013 					break;
2014 				}
2015 				call_fib_nh_notifiers(nexthop_nh,
2016 						      FIB_EVENT_NH_DEL);
2017 				dead++;
2018 			}
2019 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2020 			if (event == NETDEV_UNREGISTER &&
2021 			    nexthop_nh->fib_nh_dev == dev) {
2022 				dead = fi->fib_nhs;
2023 				break;
2024 			}
2025 #endif
2026 		} endfor_nexthops(fi)
2027 		if (dead == fi->fib_nhs) {
2028 			switch (event) {
2029 			case NETDEV_DOWN:
2030 			case NETDEV_UNREGISTER:
2031 				fi->fib_flags |= RTNH_F_DEAD;
2032 				fallthrough;
2033 			case NETDEV_CHANGE:
2034 				fi->fib_flags |= RTNH_F_LINKDOWN;
2035 				break;
2036 			}
2037 			ret++;
2038 		}
2039 
2040 		fib_rebalance(fi);
2041 	}
2042 
2043 	return ret;
2044 }
2045 
2046 /* Must be invoked inside of an RCU protected region.  */
fib_select_default(const struct flowi4 * flp,struct fib_result * res)2047 static void fib_select_default(const struct flowi4 *flp, struct fib_result *res)
2048 {
2049 	struct fib_info *fi = NULL, *last_resort = NULL;
2050 	struct hlist_head *fa_head = res->fa_head;
2051 	struct fib_table *tb = res->table;
2052 	u8 slen = 32 - res->prefixlen;
2053 	int order = -1, last_idx = -1;
2054 	struct fib_alias *fa, *fa1 = NULL;
2055 	u32 last_prio = res->fi->fib_priority;
2056 	u8 last_tos = 0;
2057 
2058 	hlist_for_each_entry_rcu(fa, fa_head, fa_list) {
2059 		struct fib_info *next_fi = fa->fa_info;
2060 		struct fib_nh_common *nhc;
2061 
2062 		if (fa->fa_slen != slen)
2063 			continue;
2064 		if (fa->fa_tos && fa->fa_tos != flp->flowi4_tos)
2065 			continue;
2066 		if (fa->tb_id != tb->tb_id)
2067 			continue;
2068 		if (next_fi->fib_priority > last_prio &&
2069 		    fa->fa_tos == last_tos) {
2070 			if (last_tos)
2071 				continue;
2072 			break;
2073 		}
2074 		if (next_fi->fib_flags & RTNH_F_DEAD)
2075 			continue;
2076 		last_tos = fa->fa_tos;
2077 		last_prio = next_fi->fib_priority;
2078 
2079 		if (next_fi->fib_scope != res->scope ||
2080 		    fa->fa_type != RTN_UNICAST)
2081 			continue;
2082 
2083 		nhc = fib_info_nhc(next_fi, 0);
2084 		if (!nhc->nhc_gw_family || nhc->nhc_scope != RT_SCOPE_LINK)
2085 			continue;
2086 
2087 		fib_alias_accessed(fa);
2088 
2089 		if (!fi) {
2090 			if (next_fi != res->fi)
2091 				break;
2092 			fa1 = fa;
2093 		} else if (!fib_detect_death(fi, order, &last_resort,
2094 					     &last_idx, fa1->fa_default)) {
2095 			fib_result_assign(res, fi);
2096 			fa1->fa_default = order;
2097 			goto out;
2098 		}
2099 		fi = next_fi;
2100 		order++;
2101 	}
2102 
2103 	if (order <= 0 || !fi) {
2104 		if (fa1)
2105 			fa1->fa_default = -1;
2106 		goto out;
2107 	}
2108 
2109 	if (!fib_detect_death(fi, order, &last_resort, &last_idx,
2110 			      fa1->fa_default)) {
2111 		fib_result_assign(res, fi);
2112 		fa1->fa_default = order;
2113 		goto out;
2114 	}
2115 
2116 	if (last_idx >= 0)
2117 		fib_result_assign(res, last_resort);
2118 	fa1->fa_default = last_idx;
2119 out:
2120 	return;
2121 }
2122 
2123 /*
2124  * Dead device goes up. We wake up dead nexthops.
2125  * It takes sense only on multipath routes.
2126  *
2127  * only used when fib_nh is built into fib_info
2128  */
fib_sync_up(struct net_device * dev,unsigned char nh_flags)2129 int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
2130 {
2131 	struct fib_info *prev_fi;
2132 	unsigned int hash;
2133 	struct hlist_head *head;
2134 	struct fib_nh *nh;
2135 	int ret;
2136 
2137 	if (!(dev->flags & IFF_UP))
2138 		return 0;
2139 
2140 	if (nh_flags & RTNH_F_DEAD) {
2141 		unsigned int flags = dev_get_flags(dev);
2142 
2143 		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
2144 			nh_flags |= RTNH_F_LINKDOWN;
2145 	}
2146 
2147 	prev_fi = NULL;
2148 	hash = fib_devindex_hashfn(dev->ifindex);
2149 	head = &fib_info_devhash[hash];
2150 	ret = 0;
2151 
2152 	hlist_for_each_entry(nh, head, nh_hash) {
2153 		struct fib_info *fi = nh->nh_parent;
2154 		int alive;
2155 
2156 		BUG_ON(!fi->fib_nhs);
2157 		if (nh->fib_nh_dev != dev || fi == prev_fi)
2158 			continue;
2159 
2160 		prev_fi = fi;
2161 		alive = 0;
2162 		change_nexthops(fi) {
2163 			if (!(nexthop_nh->fib_nh_flags & nh_flags)) {
2164 				alive++;
2165 				continue;
2166 			}
2167 			if (!nexthop_nh->fib_nh_dev ||
2168 			    !(nexthop_nh->fib_nh_dev->flags & IFF_UP))
2169 				continue;
2170 			if (nexthop_nh->fib_nh_dev != dev ||
2171 			    !__in_dev_get_rtnl(dev))
2172 				continue;
2173 			alive++;
2174 			nexthop_nh->fib_nh_flags &= ~nh_flags;
2175 			call_fib_nh_notifiers(nexthop_nh, FIB_EVENT_NH_ADD);
2176 		} endfor_nexthops(fi)
2177 
2178 		if (alive > 0) {
2179 			fi->fib_flags &= ~nh_flags;
2180 			ret++;
2181 		}
2182 
2183 		fib_rebalance(fi);
2184 	}
2185 
2186 	return ret;
2187 }
2188 
2189 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_good_nh(const struct fib_nh * nh)2190 static bool fib_good_nh(const struct fib_nh *nh)
2191 {
2192 	int state = NUD_REACHABLE;
2193 
2194 	if (nh->fib_nh_scope == RT_SCOPE_LINK) {
2195 		struct neighbour *n;
2196 
2197 		rcu_read_lock_bh();
2198 
2199 		if (likely(nh->fib_nh_gw_family == AF_INET))
2200 			n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
2201 						   (__force u32)nh->fib_nh_gw4);
2202 		else if (nh->fib_nh_gw_family == AF_INET6)
2203 			n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev,
2204 							   &nh->fib_nh_gw6);
2205 		else
2206 			n = NULL;
2207 		if (n)
2208 			state = n->nud_state;
2209 
2210 		rcu_read_unlock_bh();
2211 	}
2212 
2213 	return !!(state & NUD_VALID);
2214 }
2215 
fib_select_multipath(struct fib_result * res,int hash)2216 void fib_select_multipath(struct fib_result *res, int hash)
2217 {
2218 	struct fib_info *fi = res->fi;
2219 	struct net *net = fi->fib_net;
2220 	bool first = false;
2221 
2222 	if (unlikely(res->fi->nh)) {
2223 		nexthop_path_fib_result(res, hash);
2224 		return;
2225 	}
2226 
2227 	change_nexthops(fi) {
2228 		if (net->ipv4.sysctl_fib_multipath_use_neigh) {
2229 			if (!fib_good_nh(nexthop_nh))
2230 				continue;
2231 			if (!first) {
2232 				res->nh_sel = nhsel;
2233 				res->nhc = &nexthop_nh->nh_common;
2234 				first = true;
2235 			}
2236 		}
2237 
2238 		if (hash > atomic_read(&nexthop_nh->fib_nh_upper_bound))
2239 			continue;
2240 
2241 		res->nh_sel = nhsel;
2242 		res->nhc = &nexthop_nh->nh_common;
2243 		return;
2244 	} endfor_nexthops(fi);
2245 }
2246 #endif
2247 
fib_select_path(struct net * net,struct fib_result * res,struct flowi4 * fl4,const struct sk_buff * skb)2248 void fib_select_path(struct net *net, struct fib_result *res,
2249 		     struct flowi4 *fl4, const struct sk_buff *skb)
2250 {
2251 	if (fl4->flowi4_oif && !(fl4->flowi4_flags & FLOWI_FLAG_SKIP_NH_OIF))
2252 		goto check_saddr;
2253 
2254 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2255 	if (fib_info_num_path(res->fi) > 1) {
2256 		int h = fib_multipath_hash(net, fl4, skb, NULL);
2257 
2258 		fib_select_multipath(res, h);
2259 	}
2260 	else
2261 #endif
2262 	if (!res->prefixlen &&
2263 	    res->table->tb_num_default > 1 &&
2264 	    res->type == RTN_UNICAST)
2265 		fib_select_default(fl4, res);
2266 
2267 check_saddr:
2268 	if (!fl4->saddr)
2269 		fl4->saddr = fib_result_prefsrc(net, res);
2270 }
2271