• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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/ip.h>
37 #include <net/protocol.h>
38 #include <net/route.h>
39 #include <net/tcp.h>
40 #include <net/sock.h>
41 #include <net/ip_fib.h>
42 #include <net/ip6_fib.h>
43 #include <net/nexthop.h>
44 #include <net/netlink.h>
45 #include <net/rtnh.h>
46 #include <net/lwtunnel.h>
47 #include <net/fib_notifier.h>
48 #include <net/addrconf.h>
49 
50 #include "fib_lookup.h"
51 
52 static DEFINE_SPINLOCK(fib_info_lock);
53 static struct hlist_head *fib_info_hash;
54 static struct hlist_head *fib_info_laddrhash;
55 static unsigned int fib_info_hash_size;
56 static unsigned int fib_info_cnt;
57 
58 #define DEVINDEX_HASHBITS 8
59 #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS)
60 static struct hlist_head fib_info_devhash[DEVINDEX_HASHSIZE];
61 
62 /* for_nexthops and change_nexthops only used when nexthop object
63  * is not set in a fib_info. The logic within can reference fib_nh.
64  */
65 #ifdef CONFIG_IP_ROUTE_MULTIPATH
66 
67 #define for_nexthops(fi) {						\
68 	int nhsel; const struct fib_nh *nh;				\
69 	for (nhsel = 0, nh = (fi)->fib_nh;				\
70 	     nhsel < fib_info_num_path((fi));				\
71 	     nh++, nhsel++)
72 
73 #define change_nexthops(fi) {						\
74 	int nhsel; struct fib_nh *nexthop_nh;				\
75 	for (nhsel = 0,	nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
76 	     nhsel < fib_info_num_path((fi));				\
77 	     nexthop_nh++, nhsel++)
78 
79 #else /* CONFIG_IP_ROUTE_MULTIPATH */
80 
81 /* Hope, that gcc will optimize it to get rid of dummy loop */
82 
83 #define for_nexthops(fi) {						\
84 	int nhsel; const struct fib_nh *nh = (fi)->fib_nh;		\
85 	for (nhsel = 0; nhsel < 1; nhsel++)
86 
87 #define change_nexthops(fi) {						\
88 	int nhsel;							\
89 	struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
90 	for (nhsel = 0; nhsel < 1; nhsel++)
91 
92 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
93 
94 #define endfor_nexthops(fi) }
95 
96 
97 const struct fib_prop fib_props[RTN_MAX + 1] = {
98 	[RTN_UNSPEC] = {
99 		.error	= 0,
100 		.scope	= RT_SCOPE_NOWHERE,
101 	},
102 	[RTN_UNICAST] = {
103 		.error	= 0,
104 		.scope	= RT_SCOPE_UNIVERSE,
105 	},
106 	[RTN_LOCAL] = {
107 		.error	= 0,
108 		.scope	= RT_SCOPE_HOST,
109 	},
110 	[RTN_BROADCAST] = {
111 		.error	= 0,
112 		.scope	= RT_SCOPE_LINK,
113 	},
114 	[RTN_ANYCAST] = {
115 		.error	= 0,
116 		.scope	= RT_SCOPE_LINK,
117 	},
118 	[RTN_MULTICAST] = {
119 		.error	= 0,
120 		.scope	= RT_SCOPE_UNIVERSE,
121 	},
122 	[RTN_BLACKHOLE] = {
123 		.error	= -EINVAL,
124 		.scope	= RT_SCOPE_UNIVERSE,
125 	},
126 	[RTN_UNREACHABLE] = {
127 		.error	= -EHOSTUNREACH,
128 		.scope	= RT_SCOPE_UNIVERSE,
129 	},
130 	[RTN_PROHIBIT] = {
131 		.error	= -EACCES,
132 		.scope	= RT_SCOPE_UNIVERSE,
133 	},
134 	[RTN_THROW] = {
135 		.error	= -EAGAIN,
136 		.scope	= RT_SCOPE_UNIVERSE,
137 	},
138 	[RTN_NAT] = {
139 		.error	= -EINVAL,
140 		.scope	= RT_SCOPE_NOWHERE,
141 	},
142 	[RTN_XRESOLVE] = {
143 		.error	= -EINVAL,
144 		.scope	= RT_SCOPE_NOWHERE,
145 	},
146 };
147 
rt_fibinfo_free(struct rtable __rcu ** rtp)148 static void rt_fibinfo_free(struct rtable __rcu **rtp)
149 {
150 	struct rtable *rt = rcu_dereference_protected(*rtp, 1);
151 
152 	if (!rt)
153 		return;
154 
155 	/* Not even needed : RCU_INIT_POINTER(*rtp, NULL);
156 	 * because we waited an RCU grace period before calling
157 	 * free_fib_info_rcu()
158 	 */
159 
160 	dst_dev_put(&rt->dst);
161 	dst_release_immediate(&rt->dst);
162 }
163 
free_nh_exceptions(struct fib_nh_common * nhc)164 static void free_nh_exceptions(struct fib_nh_common *nhc)
165 {
166 	struct fnhe_hash_bucket *hash;
167 	int i;
168 
169 	hash = rcu_dereference_protected(nhc->nhc_exceptions, 1);
170 	if (!hash)
171 		return;
172 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
173 		struct fib_nh_exception *fnhe;
174 
175 		fnhe = rcu_dereference_protected(hash[i].chain, 1);
176 		while (fnhe) {
177 			struct fib_nh_exception *next;
178 
179 			next = rcu_dereference_protected(fnhe->fnhe_next, 1);
180 
181 			rt_fibinfo_free(&fnhe->fnhe_rth_input);
182 			rt_fibinfo_free(&fnhe->fnhe_rth_output);
183 
184 			kfree(fnhe);
185 
186 			fnhe = next;
187 		}
188 	}
189 	kfree(hash);
190 }
191 
rt_fibinfo_free_cpus(struct rtable __rcu * __percpu * rtp)192 static void rt_fibinfo_free_cpus(struct rtable __rcu * __percpu *rtp)
193 {
194 	int cpu;
195 
196 	if (!rtp)
197 		return;
198 
199 	for_each_possible_cpu(cpu) {
200 		struct rtable *rt;
201 
202 		rt = rcu_dereference_protected(*per_cpu_ptr(rtp, cpu), 1);
203 		if (rt) {
204 			dst_dev_put(&rt->dst);
205 			dst_release_immediate(&rt->dst);
206 		}
207 	}
208 	free_percpu(rtp);
209 }
210 
fib_nh_common_release(struct fib_nh_common * nhc)211 void fib_nh_common_release(struct fib_nh_common *nhc)
212 {
213 	if (nhc->nhc_dev)
214 		dev_put(nhc->nhc_dev);
215 
216 	lwtstate_put(nhc->nhc_lwtstate);
217 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
218 	rt_fibinfo_free(&nhc->nhc_rth_input);
219 	free_nh_exceptions(nhc);
220 }
221 EXPORT_SYMBOL_GPL(fib_nh_common_release);
222 
fib_nh_release(struct net * net,struct fib_nh * fib_nh)223 void fib_nh_release(struct net *net, struct fib_nh *fib_nh)
224 {
225 #ifdef CONFIG_IP_ROUTE_CLASSID
226 	if (fib_nh->nh_tclassid)
227 		atomic_dec(&net->ipv4.fib_num_tclassid_users);
228 #endif
229 	fib_nh_common_release(&fib_nh->nh_common);
230 }
231 
232 /* Release a nexthop info record */
free_fib_info_rcu(struct rcu_head * head)233 static void free_fib_info_rcu(struct rcu_head *head)
234 {
235 	struct fib_info *fi = container_of(head, struct fib_info, rcu);
236 
237 	if (fi->nh) {
238 		nexthop_put(fi->nh);
239 	} else {
240 		change_nexthops(fi) {
241 			fib_nh_release(fi->fib_net, nexthop_nh);
242 		} endfor_nexthops(fi);
243 	}
244 
245 	ip_fib_metrics_put(fi->fib_metrics);
246 
247 	kfree(fi);
248 }
249 
free_fib_info(struct fib_info * fi)250 void free_fib_info(struct fib_info *fi)
251 {
252 	if (fi->fib_dead == 0) {
253 		pr_warn("Freeing alive fib_info %p\n", fi);
254 		return;
255 	}
256 	fib_info_cnt--;
257 
258 	call_rcu(&fi->rcu, free_fib_info_rcu);
259 }
260 EXPORT_SYMBOL_GPL(free_fib_info);
261 
fib_release_info(struct fib_info * fi)262 void fib_release_info(struct fib_info *fi)
263 {
264 	spin_lock_bh(&fib_info_lock);
265 	if (fi && --fi->fib_treeref == 0) {
266 		hlist_del(&fi->fib_hash);
267 		if (fi->fib_prefsrc)
268 			hlist_del(&fi->fib_lhash);
269 		if (fi->nh) {
270 			list_del(&fi->nh_list);
271 		} else {
272 			change_nexthops(fi) {
273 				if (!nexthop_nh->fib_nh_dev)
274 					continue;
275 				hlist_del(&nexthop_nh->nh_hash);
276 			} endfor_nexthops(fi)
277 		}
278 		/* Paired with READ_ONCE() from fib_table_lookup() */
279 		WRITE_ONCE(fi->fib_dead, 1);
280 		fib_info_put(fi);
281 	}
282 	spin_unlock_bh(&fib_info_lock);
283 }
284 
nh_comp(struct fib_info * fi,struct fib_info * ofi)285 static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi)
286 {
287 	const struct fib_nh *onh;
288 
289 	if (fi->nh || ofi->nh)
290 		return nexthop_cmp(fi->nh, ofi->nh) ? 0 : -1;
291 
292 	if (ofi->fib_nhs == 0)
293 		return 0;
294 
295 	for_nexthops(fi) {
296 		onh = fib_info_nh(ofi, nhsel);
297 
298 		if (nh->fib_nh_oif != onh->fib_nh_oif ||
299 		    nh->fib_nh_gw_family != onh->fib_nh_gw_family ||
300 		    nh->fib_nh_scope != onh->fib_nh_scope ||
301 #ifdef CONFIG_IP_ROUTE_MULTIPATH
302 		    nh->fib_nh_weight != onh->fib_nh_weight ||
303 #endif
304 #ifdef CONFIG_IP_ROUTE_CLASSID
305 		    nh->nh_tclassid != onh->nh_tclassid ||
306 #endif
307 		    lwtunnel_cmp_encap(nh->fib_nh_lws, onh->fib_nh_lws) ||
308 		    ((nh->fib_nh_flags ^ onh->fib_nh_flags) & ~RTNH_COMPARE_MASK))
309 			return -1;
310 
311 		if (nh->fib_nh_gw_family == AF_INET &&
312 		    nh->fib_nh_gw4 != onh->fib_nh_gw4)
313 			return -1;
314 
315 		if (nh->fib_nh_gw_family == AF_INET6 &&
316 		    ipv6_addr_cmp(&nh->fib_nh_gw6, &onh->fib_nh_gw6))
317 			return -1;
318 	} endfor_nexthops(fi);
319 	return 0;
320 }
321 
fib_devindex_hashfn(unsigned int val)322 static inline unsigned int fib_devindex_hashfn(unsigned int val)
323 {
324 	return hash_32(val, DEVINDEX_HASHBITS);
325 }
326 
327 static struct hlist_head *
fib_info_devhash_bucket(const struct net_device * dev)328 fib_info_devhash_bucket(const struct net_device *dev)
329 {
330 	u32 val = net_hash_mix(dev_net(dev)) ^ dev->ifindex;
331 
332 	return &fib_info_devhash[fib_devindex_hashfn(val)];
333 }
334 
fib_info_hashfn_1(int init_val,u8 protocol,u8 scope,u32 prefsrc,u32 priority)335 static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope,
336 				      u32 prefsrc, u32 priority)
337 {
338 	unsigned int val = init_val;
339 
340 	val ^= (protocol << 8) | scope;
341 	val ^= prefsrc;
342 	val ^= priority;
343 
344 	return val;
345 }
346 
fib_info_hashfn_result(unsigned int val)347 static unsigned int fib_info_hashfn_result(unsigned int val)
348 {
349 	unsigned int mask = (fib_info_hash_size - 1);
350 
351 	return (val ^ (val >> 7) ^ (val >> 12)) & mask;
352 }
353 
fib_info_hashfn(struct fib_info * fi)354 static inline unsigned int fib_info_hashfn(struct fib_info *fi)
355 {
356 	unsigned int val;
357 
358 	val = fib_info_hashfn_1(fi->fib_nhs, fi->fib_protocol,
359 				fi->fib_scope, (__force u32)fi->fib_prefsrc,
360 				fi->fib_priority);
361 
362 	if (fi->nh) {
363 		val ^= fib_devindex_hashfn(fi->nh->id);
364 	} else {
365 		for_nexthops(fi) {
366 			val ^= fib_devindex_hashfn(nh->fib_nh_oif);
367 		} endfor_nexthops(fi)
368 	}
369 
370 	return fib_info_hashfn_result(val);
371 }
372 
373 /* no metrics, only nexthop id */
fib_find_info_nh(struct net * net,const struct fib_config * cfg)374 static struct fib_info *fib_find_info_nh(struct net *net,
375 					 const struct fib_config *cfg)
376 {
377 	struct hlist_head *head;
378 	struct fib_info *fi;
379 	unsigned int hash;
380 
381 	hash = fib_info_hashfn_1(fib_devindex_hashfn(cfg->fc_nh_id),
382 				 cfg->fc_protocol, cfg->fc_scope,
383 				 (__force u32)cfg->fc_prefsrc,
384 				 cfg->fc_priority);
385 	hash = fib_info_hashfn_result(hash);
386 	head = &fib_info_hash[hash];
387 
388 	hlist_for_each_entry(fi, head, fib_hash) {
389 		if (!net_eq(fi->fib_net, net))
390 			continue;
391 		if (!fi->nh || fi->nh->id != cfg->fc_nh_id)
392 			continue;
393 		if (cfg->fc_protocol == fi->fib_protocol &&
394 		    cfg->fc_scope == fi->fib_scope &&
395 		    cfg->fc_prefsrc == fi->fib_prefsrc &&
396 		    cfg->fc_priority == fi->fib_priority &&
397 		    cfg->fc_type == fi->fib_type &&
398 		    cfg->fc_table == fi->fib_tb_id &&
399 		    !((cfg->fc_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK))
400 			return fi;
401 	}
402 
403 	return NULL;
404 }
405 
fib_find_info(struct fib_info * nfi)406 static struct fib_info *fib_find_info(struct fib_info *nfi)
407 {
408 	struct hlist_head *head;
409 	struct fib_info *fi;
410 	unsigned int hash;
411 
412 	hash = fib_info_hashfn(nfi);
413 	head = &fib_info_hash[hash];
414 
415 	hlist_for_each_entry(fi, head, fib_hash) {
416 		if (!net_eq(fi->fib_net, nfi->fib_net))
417 			continue;
418 		if (fi->fib_nhs != nfi->fib_nhs)
419 			continue;
420 		if (nfi->fib_protocol == fi->fib_protocol &&
421 		    nfi->fib_scope == fi->fib_scope &&
422 		    nfi->fib_prefsrc == fi->fib_prefsrc &&
423 		    nfi->fib_priority == fi->fib_priority &&
424 		    nfi->fib_type == fi->fib_type &&
425 		    nfi->fib_tb_id == fi->fib_tb_id &&
426 		    memcmp(nfi->fib_metrics, fi->fib_metrics,
427 			   sizeof(u32) * RTAX_MAX) == 0 &&
428 		    !((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK) &&
429 		    nh_comp(fi, nfi) == 0)
430 			return fi;
431 	}
432 
433 	return NULL;
434 }
435 
436 /* Check, that the gateway is already configured.
437  * Used only by redirect accept routine.
438  */
ip_fib_check_default(__be32 gw,struct net_device * dev)439 int ip_fib_check_default(__be32 gw, struct net_device *dev)
440 {
441 	struct hlist_head *head;
442 	struct fib_nh *nh;
443 
444 	spin_lock(&fib_info_lock);
445 
446 	head = fib_info_devhash_bucket(dev);
447 
448 	hlist_for_each_entry(nh, head, nh_hash) {
449 		if (nh->fib_nh_dev == dev &&
450 		    nh->fib_nh_gw4 == gw &&
451 		    !(nh->fib_nh_flags & RTNH_F_DEAD)) {
452 			spin_unlock(&fib_info_lock);
453 			return 0;
454 		}
455 	}
456 
457 	spin_unlock(&fib_info_lock);
458 
459 	return -1;
460 }
461 
fib_nlmsg_size(struct fib_info * fi)462 static inline size_t fib_nlmsg_size(struct fib_info *fi)
463 {
464 	size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
465 			 + nla_total_size(4) /* RTA_TABLE */
466 			 + nla_total_size(4) /* RTA_DST */
467 			 + nla_total_size(4) /* RTA_PRIORITY */
468 			 + nla_total_size(4) /* RTA_PREFSRC */
469 			 + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
470 	unsigned int nhs = fib_info_num_path(fi);
471 
472 	/* space for nested metrics */
473 	payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
474 
475 	if (fi->nh)
476 		payload += nla_total_size(4); /* RTA_NH_ID */
477 
478 	if (nhs) {
479 		size_t nh_encapsize = 0;
480 		/* Also handles the special case nhs == 1 */
481 
482 		/* each nexthop is packed in an attribute */
483 		size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
484 		unsigned int i;
485 
486 		/* may contain flow and gateway attribute */
487 		nhsize += 2 * nla_total_size(4);
488 
489 		/* grab encap info */
490 		for (i = 0; i < fib_info_num_path(fi); i++) {
491 			struct fib_nh_common *nhc = fib_info_nhc(fi, i);
492 
493 			if (nhc->nhc_lwtstate) {
494 				/* RTA_ENCAP_TYPE */
495 				nh_encapsize += lwtunnel_get_encap_size(
496 						nhc->nhc_lwtstate);
497 				/* RTA_ENCAP */
498 				nh_encapsize +=  nla_total_size(2);
499 			}
500 		}
501 
502 		/* all nexthops are packed in a nested attribute */
503 		payload += nla_total_size((nhs * nhsize) + nh_encapsize);
504 
505 	}
506 
507 	return payload;
508 }
509 
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)510 void rtmsg_fib(int event, __be32 key, struct fib_alias *fa,
511 	       int dst_len, u32 tb_id, const struct nl_info *info,
512 	       unsigned int nlm_flags)
513 {
514 	struct sk_buff *skb;
515 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
516 	int err = -ENOBUFS;
517 
518 	skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL);
519 	if (!skb)
520 		goto errout;
521 
522 	err = fib_dump_info(skb, info->portid, seq, event, tb_id,
523 			    fa->fa_type, key, dst_len,
524 			    fa->fa_tos, fa->fa_info, nlm_flags);
525 	if (err < 0) {
526 		/* -EMSGSIZE implies BUG in fib_nlmsg_size() */
527 		WARN_ON(err == -EMSGSIZE);
528 		kfree_skb(skb);
529 		goto errout;
530 	}
531 	rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_IPV4_ROUTE,
532 		    info->nlh, GFP_KERNEL);
533 	return;
534 errout:
535 	if (err < 0)
536 		rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err);
537 }
538 
fib_detect_death(struct fib_info * fi,int order,struct fib_info ** last_resort,int * last_idx,int dflt)539 static int fib_detect_death(struct fib_info *fi, int order,
540 			    struct fib_info **last_resort, int *last_idx,
541 			    int dflt)
542 {
543 	const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
544 	struct neighbour *n;
545 	int state = NUD_NONE;
546 
547 	if (likely(nhc->nhc_gw_family == AF_INET))
548 		n = neigh_lookup(&arp_tbl, &nhc->nhc_gw.ipv4, nhc->nhc_dev);
549 	else if (nhc->nhc_gw_family == AF_INET6)
550 		n = neigh_lookup(ipv6_stub->nd_tbl, &nhc->nhc_gw.ipv6,
551 				 nhc->nhc_dev);
552 	else
553 		n = NULL;
554 
555 	if (n) {
556 		state = n->nud_state;
557 		neigh_release(n);
558 	} else {
559 		return 0;
560 	}
561 	if (state == NUD_REACHABLE)
562 		return 0;
563 	if ((state & NUD_VALID) && order != dflt)
564 		return 0;
565 	if ((state & NUD_VALID) ||
566 	    (*last_idx < 0 && order > dflt && state != NUD_INCOMPLETE)) {
567 		*last_resort = fi;
568 		*last_idx = order;
569 	}
570 	return 1;
571 }
572 
fib_nh_common_init(struct fib_nh_common * nhc,struct nlattr * encap,u16 encap_type,void * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)573 int fib_nh_common_init(struct fib_nh_common *nhc, struct nlattr *encap,
574 		       u16 encap_type, 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(encap_type, encap, nhc->nhc_family,
593 					   cfg, &lwtstate, extack);
594 		if (err)
595 			goto lwt_failure;
596 
597 		nhc->nhc_lwtstate = lwtstate_get(lwtstate);
598 	}
599 
600 	return 0;
601 
602 lwt_failure:
603 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
604 	nhc->nhc_pcpu_rth_output = NULL;
605 	return err;
606 }
607 EXPORT_SYMBOL_GPL(fib_nh_common_init);
608 
fib_nh_init(struct net * net,struct fib_nh * nh,struct fib_config * cfg,int nh_weight,struct netlink_ext_ack * extack)609 int fib_nh_init(struct net *net, struct fib_nh *nh,
610 		struct fib_config *cfg, int nh_weight,
611 		struct netlink_ext_ack *extack)
612 {
613 	int err;
614 
615 	nh->fib_nh_family = AF_INET;
616 
617 	err = fib_nh_common_init(&nh->nh_common, cfg->fc_encap,
618 				 cfg->fc_encap_type, cfg, GFP_KERNEL, extack);
619 	if (err)
620 		return err;
621 
622 	nh->fib_nh_oif = cfg->fc_oif;
623 	nh->fib_nh_gw_family = cfg->fc_gw_family;
624 	if (cfg->fc_gw_family == AF_INET)
625 		nh->fib_nh_gw4 = cfg->fc_gw4;
626 	else if (cfg->fc_gw_family == AF_INET6)
627 		nh->fib_nh_gw6 = cfg->fc_gw6;
628 
629 	nh->fib_nh_flags = cfg->fc_flags;
630 
631 #ifdef CONFIG_IP_ROUTE_CLASSID
632 	nh->nh_tclassid = cfg->fc_flow;
633 	if (nh->nh_tclassid)
634 		atomic_inc(&net->ipv4.fib_num_tclassid_users);
635 #endif
636 #ifdef CONFIG_IP_ROUTE_MULTIPATH
637 	nh->fib_nh_weight = nh_weight;
638 #endif
639 	return 0;
640 }
641 
642 #ifdef CONFIG_IP_ROUTE_MULTIPATH
643 
fib_count_nexthops(struct rtnexthop * rtnh,int remaining,struct netlink_ext_ack * extack)644 static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining,
645 			      struct netlink_ext_ack *extack)
646 {
647 	int nhs = 0;
648 
649 	while (rtnh_ok(rtnh, remaining)) {
650 		nhs++;
651 		rtnh = rtnh_next(rtnh, &remaining);
652 	}
653 
654 	/* leftover implies invalid nexthop configuration, discard it */
655 	if (remaining > 0) {
656 		NL_SET_ERR_MSG(extack,
657 			       "Invalid nexthop configuration - extra data after nexthops");
658 		nhs = 0;
659 	}
660 
661 	return nhs;
662 }
663 
fib_gw_from_attr(__be32 * gw,struct nlattr * nla,struct netlink_ext_ack * extack)664 static int fib_gw_from_attr(__be32 *gw, struct nlattr *nla,
665 			    struct netlink_ext_ack *extack)
666 {
667 	if (nla_len(nla) < sizeof(*gw)) {
668 		NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_GATEWAY");
669 		return -EINVAL;
670 	}
671 
672 	*gw = nla_get_in_addr(nla);
673 
674 	return 0;
675 }
676 
677 /* 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)678 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
679 		       int remaining, struct fib_config *cfg,
680 		       struct netlink_ext_ack *extack)
681 {
682 	struct net *net = fi->fib_net;
683 	struct fib_config fib_cfg;
684 	struct fib_nh *nh;
685 	int ret;
686 
687 	change_nexthops(fi) {
688 		int attrlen;
689 
690 		memset(&fib_cfg, 0, sizeof(fib_cfg));
691 
692 		if (!rtnh_ok(rtnh, remaining)) {
693 			NL_SET_ERR_MSG(extack,
694 				       "Invalid nexthop configuration - extra data after nexthop");
695 			return -EINVAL;
696 		}
697 
698 		if (rtnh->rtnh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
699 			NL_SET_ERR_MSG(extack,
700 				       "Invalid flags for nexthop - can not contain DEAD or LINKDOWN");
701 			return -EINVAL;
702 		}
703 
704 		fib_cfg.fc_flags = (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags;
705 		fib_cfg.fc_oif = rtnh->rtnh_ifindex;
706 
707 		attrlen = rtnh_attrlen(rtnh);
708 		if (attrlen > 0) {
709 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
710 
711 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
712 			nlav = nla_find(attrs, attrlen, RTA_VIA);
713 			if (nla && nlav) {
714 				NL_SET_ERR_MSG(extack,
715 					       "Nexthop configuration can not contain both GATEWAY and VIA");
716 				return -EINVAL;
717 			}
718 			if (nla) {
719 				ret = fib_gw_from_attr(&fib_cfg.fc_gw4, nla,
720 						       extack);
721 				if (ret)
722 					goto errout;
723 
724 				if (fib_cfg.fc_gw4)
725 					fib_cfg.fc_gw_family = AF_INET;
726 			} else if (nlav) {
727 				ret = fib_gw_from_via(&fib_cfg, nlav, extack);
728 				if (ret)
729 					goto errout;
730 			}
731 
732 			nla = nla_find(attrs, attrlen, RTA_FLOW);
733 			if (nla) {
734 				if (nla_len(nla) < sizeof(u32)) {
735 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
736 					return -EINVAL;
737 				}
738 				fib_cfg.fc_flow = nla_get_u32(nla);
739 			}
740 
741 			fib_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
742 			/* RTA_ENCAP_TYPE length checked in
743 			 * lwtunnel_valid_encap_type_attr
744 			 */
745 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
746 			if (nla)
747 				fib_cfg.fc_encap_type = nla_get_u16(nla);
748 		}
749 
750 		ret = fib_nh_init(net, nexthop_nh, &fib_cfg,
751 				  rtnh->rtnh_hops + 1, extack);
752 		if (ret)
753 			goto errout;
754 
755 		rtnh = rtnh_next(rtnh, &remaining);
756 	} endfor_nexthops(fi);
757 
758 	ret = -EINVAL;
759 	nh = fib_info_nh(fi, 0);
760 	if (cfg->fc_oif && nh->fib_nh_oif != cfg->fc_oif) {
761 		NL_SET_ERR_MSG(extack,
762 			       "Nexthop device index does not match RTA_OIF");
763 		goto errout;
764 	}
765 	if (cfg->fc_gw_family) {
766 		if (cfg->fc_gw_family != nh->fib_nh_gw_family ||
767 		    (cfg->fc_gw_family == AF_INET &&
768 		     nh->fib_nh_gw4 != cfg->fc_gw4) ||
769 		    (cfg->fc_gw_family == AF_INET6 &&
770 		     ipv6_addr_cmp(&nh->fib_nh_gw6, &cfg->fc_gw6))) {
771 			NL_SET_ERR_MSG(extack,
772 				       "Nexthop gateway does not match RTA_GATEWAY or RTA_VIA");
773 			goto errout;
774 		}
775 	}
776 #ifdef CONFIG_IP_ROUTE_CLASSID
777 	if (cfg->fc_flow && nh->nh_tclassid != cfg->fc_flow) {
778 		NL_SET_ERR_MSG(extack,
779 			       "Nexthop class id does not match RTA_FLOW");
780 		goto errout;
781 	}
782 #endif
783 	ret = 0;
784 errout:
785 	return ret;
786 }
787 
788 /* only called when fib_nh is integrated into fib_info */
fib_rebalance(struct fib_info * fi)789 static void fib_rebalance(struct fib_info *fi)
790 {
791 	int total;
792 	int w;
793 
794 	if (fib_info_num_path(fi) < 2)
795 		return;
796 
797 	total = 0;
798 	for_nexthops(fi) {
799 		if (nh->fib_nh_flags & RTNH_F_DEAD)
800 			continue;
801 
802 		if (ip_ignore_linkdown(nh->fib_nh_dev) &&
803 		    nh->fib_nh_flags & RTNH_F_LINKDOWN)
804 			continue;
805 
806 		total += nh->fib_nh_weight;
807 	} endfor_nexthops(fi);
808 
809 	w = 0;
810 	change_nexthops(fi) {
811 		int upper_bound;
812 
813 		if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD) {
814 			upper_bound = -1;
815 		} else if (ip_ignore_linkdown(nexthop_nh->fib_nh_dev) &&
816 			   nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN) {
817 			upper_bound = -1;
818 		} else {
819 			w += nexthop_nh->fib_nh_weight;
820 			upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31,
821 							    total) - 1;
822 		}
823 
824 		atomic_set(&nexthop_nh->fib_nh_upper_bound, upper_bound);
825 	} endfor_nexthops(fi);
826 }
827 #else /* CONFIG_IP_ROUTE_MULTIPATH */
828 
fib_get_nhs(struct fib_info * fi,struct rtnexthop * rtnh,int remaining,struct fib_config * cfg,struct netlink_ext_ack * extack)829 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
830 		       int remaining, struct fib_config *cfg,
831 		       struct netlink_ext_ack *extack)
832 {
833 	NL_SET_ERR_MSG(extack, "Multipath support not enabled in kernel");
834 
835 	return -EINVAL;
836 }
837 
838 #define fib_rebalance(fi) do { } while (0)
839 
840 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
841 
fib_encap_match(u16 encap_type,struct nlattr * encap,const struct fib_nh * nh,const struct fib_config * cfg,struct netlink_ext_ack * extack)842 static int fib_encap_match(u16 encap_type,
843 			   struct nlattr *encap,
844 			   const struct fib_nh *nh,
845 			   const struct fib_config *cfg,
846 			   struct netlink_ext_ack *extack)
847 {
848 	struct lwtunnel_state *lwtstate;
849 	int ret, result = 0;
850 
851 	if (encap_type == LWTUNNEL_ENCAP_NONE)
852 		return 0;
853 
854 	ret = lwtunnel_build_state(encap_type, encap, AF_INET,
855 				   cfg, &lwtstate, extack);
856 	if (!ret) {
857 		result = lwtunnel_cmp_encap(lwtstate, nh->fib_nh_lws);
858 		lwtstate_free(lwtstate);
859 	}
860 
861 	return result;
862 }
863 
fib_nh_match(struct fib_config * cfg,struct fib_info * fi,struct netlink_ext_ack * extack)864 int fib_nh_match(struct fib_config *cfg, struct fib_info *fi,
865 		 struct netlink_ext_ack *extack)
866 {
867 #ifdef CONFIG_IP_ROUTE_MULTIPATH
868 	struct rtnexthop *rtnh;
869 	int remaining;
870 #endif
871 
872 	if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority)
873 		return 1;
874 
875 	if (cfg->fc_nh_id) {
876 		if (fi->nh && cfg->fc_nh_id == fi->nh->id)
877 			return 0;
878 		return 1;
879 	}
880 
881 	if (fi->nh) {
882 		if (cfg->fc_oif || cfg->fc_gw_family || cfg->fc_mp)
883 			return 1;
884 		return 0;
885 	}
886 
887 	if (cfg->fc_oif || cfg->fc_gw_family) {
888 		struct fib_nh *nh;
889 
890 		nh = fib_info_nh(fi, 0);
891 		if (cfg->fc_encap) {
892 			if (fib_encap_match(cfg->fc_encap_type, cfg->fc_encap,
893 					    nh, cfg, extack))
894 				return 1;
895 		}
896 #ifdef CONFIG_IP_ROUTE_CLASSID
897 		if (cfg->fc_flow &&
898 		    cfg->fc_flow != nh->nh_tclassid)
899 			return 1;
900 #endif
901 		if ((cfg->fc_oif && cfg->fc_oif != nh->fib_nh_oif) ||
902 		    (cfg->fc_gw_family &&
903 		     cfg->fc_gw_family != nh->fib_nh_gw_family))
904 			return 1;
905 
906 		if (cfg->fc_gw_family == AF_INET &&
907 		    cfg->fc_gw4 != nh->fib_nh_gw4)
908 			return 1;
909 
910 		if (cfg->fc_gw_family == AF_INET6 &&
911 		    ipv6_addr_cmp(&cfg->fc_gw6, &nh->fib_nh_gw6))
912 			return 1;
913 
914 		return 0;
915 	}
916 
917 #ifdef CONFIG_IP_ROUTE_MULTIPATH
918 	if (!cfg->fc_mp)
919 		return 0;
920 
921 	rtnh = cfg->fc_mp;
922 	remaining = cfg->fc_mp_len;
923 
924 	for_nexthops(fi) {
925 		int attrlen;
926 
927 		if (!rtnh_ok(rtnh, remaining))
928 			return -EINVAL;
929 
930 		if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->fib_nh_oif)
931 			return 1;
932 
933 		attrlen = rtnh_attrlen(rtnh);
934 		if (attrlen > 0) {
935 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
936 			int err;
937 
938 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
939 			nlav = nla_find(attrs, attrlen, RTA_VIA);
940 			if (nla && nlav) {
941 				NL_SET_ERR_MSG(extack,
942 					       "Nexthop configuration can not contain both GATEWAY and VIA");
943 				return -EINVAL;
944 			}
945 
946 			if (nla) {
947 				__be32 gw;
948 
949 				err = fib_gw_from_attr(&gw, nla, extack);
950 				if (err)
951 					return err;
952 
953 				if (nh->fib_nh_gw_family != AF_INET ||
954 				    gw != nh->fib_nh_gw4)
955 					return 1;
956 			} else if (nlav) {
957 				struct fib_config cfg2;
958 
959 				err = fib_gw_from_via(&cfg2, nlav, extack);
960 				if (err)
961 					return err;
962 
963 				switch (nh->fib_nh_gw_family) {
964 				case AF_INET:
965 					if (cfg2.fc_gw_family != AF_INET ||
966 					    cfg2.fc_gw4 != nh->fib_nh_gw4)
967 						return 1;
968 					break;
969 				case AF_INET6:
970 					if (cfg2.fc_gw_family != AF_INET6 ||
971 					    ipv6_addr_cmp(&cfg2.fc_gw6,
972 							  &nh->fib_nh_gw6))
973 						return 1;
974 					break;
975 				}
976 			}
977 
978 #ifdef CONFIG_IP_ROUTE_CLASSID
979 			nla = nla_find(attrs, attrlen, RTA_FLOW);
980 			if (nla) {
981 				if (nla_len(nla) < sizeof(u32)) {
982 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
983 					return -EINVAL;
984 				}
985 				if (nla_get_u32(nla) != nh->nh_tclassid)
986 					return 1;
987 			}
988 #endif
989 		}
990 
991 		rtnh = rtnh_next(rtnh, &remaining);
992 	} endfor_nexthops(fi);
993 #endif
994 	return 0;
995 }
996 
fib_metrics_match(struct fib_config * cfg,struct fib_info * fi)997 bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
998 {
999 	struct nlattr *nla;
1000 	int remaining;
1001 
1002 	if (!cfg->fc_mx)
1003 		return true;
1004 
1005 	nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1006 		int type = nla_type(nla);
1007 		u32 fi_val, val;
1008 
1009 		if (!type)
1010 			continue;
1011 		if (type > RTAX_MAX)
1012 			return false;
1013 
1014 		type = array_index_nospec(type, RTAX_MAX + 1);
1015 		if (type == RTAX_CC_ALGO) {
1016 			char tmp[TCP_CA_NAME_MAX];
1017 			bool ecn_ca = false;
1018 
1019 			nla_strlcpy(tmp, nla, sizeof(tmp));
1020 			val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
1021 		} else {
1022 			if (nla_len(nla) != sizeof(u32))
1023 				return false;
1024 			val = nla_get_u32(nla);
1025 		}
1026 
1027 		fi_val = fi->fib_metrics->metrics[type - 1];
1028 		if (type == RTAX_FEATURES)
1029 			fi_val &= ~DST_FEATURE_ECN_CA;
1030 
1031 		if (fi_val != val)
1032 			return false;
1033 	}
1034 
1035 	return true;
1036 }
1037 
fib_check_nh_v6_gw(struct net * net,struct fib_nh * nh,u32 table,struct netlink_ext_ack * extack)1038 static int fib_check_nh_v6_gw(struct net *net, struct fib_nh *nh,
1039 			      u32 table, struct netlink_ext_ack *extack)
1040 {
1041 	struct fib6_config cfg = {
1042 		.fc_table = table,
1043 		.fc_flags = nh->fib_nh_flags | RTF_GATEWAY,
1044 		.fc_ifindex = nh->fib_nh_oif,
1045 		.fc_gateway = nh->fib_nh_gw6,
1046 	};
1047 	struct fib6_nh fib6_nh = {};
1048 	int err;
1049 
1050 	err = ipv6_stub->fib6_nh_init(net, &fib6_nh, &cfg, GFP_KERNEL, extack);
1051 	if (!err) {
1052 		nh->fib_nh_dev = fib6_nh.fib_nh_dev;
1053 		dev_hold(nh->fib_nh_dev);
1054 		nh->fib_nh_oif = nh->fib_nh_dev->ifindex;
1055 		nh->fib_nh_scope = RT_SCOPE_LINK;
1056 
1057 		ipv6_stub->fib6_nh_release(&fib6_nh);
1058 	}
1059 
1060 	return err;
1061 }
1062 
1063 /*
1064  * Picture
1065  * -------
1066  *
1067  * Semantics of nexthop is very messy by historical reasons.
1068  * We have to take into account, that:
1069  * a) gateway can be actually local interface address,
1070  *    so that gatewayed route is direct.
1071  * b) gateway must be on-link address, possibly
1072  *    described not by an ifaddr, but also by a direct route.
1073  * c) If both gateway and interface are specified, they should not
1074  *    contradict.
1075  * d) If we use tunnel routes, gateway could be not on-link.
1076  *
1077  * Attempt to reconcile all of these (alas, self-contradictory) conditions
1078  * results in pretty ugly and hairy code with obscure logic.
1079  *
1080  * I chose to generalized it instead, so that the size
1081  * of code does not increase practically, but it becomes
1082  * much more general.
1083  * Every prefix is assigned a "scope" value: "host" is local address,
1084  * "link" is direct route,
1085  * [ ... "site" ... "interior" ... ]
1086  * and "universe" is true gateway route with global meaning.
1087  *
1088  * Every prefix refers to a set of "nexthop"s (gw, oif),
1089  * where gw must have narrower scope. This recursion stops
1090  * when gw has LOCAL scope or if "nexthop" is declared ONLINK,
1091  * which means that gw is forced to be on link.
1092  *
1093  * Code is still hairy, but now it is apparently logically
1094  * consistent and very flexible. F.e. as by-product it allows
1095  * to co-exists in peace independent exterior and interior
1096  * routing processes.
1097  *
1098  * Normally it looks as following.
1099  *
1100  * {universe prefix}  -> (gw, oif) [scope link]
1101  *		  |
1102  *		  |-> {link prefix} -> (gw, oif) [scope local]
1103  *					|
1104  *					|-> {local prefix} (terminal node)
1105  */
fib_check_nh_v4_gw(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1106 static int fib_check_nh_v4_gw(struct net *net, struct fib_nh *nh, u32 table,
1107 			      u8 scope, struct netlink_ext_ack *extack)
1108 {
1109 	struct net_device *dev;
1110 	struct fib_result res;
1111 	int err = 0;
1112 
1113 	if (nh->fib_nh_flags & RTNH_F_ONLINK) {
1114 		unsigned int addr_type;
1115 
1116 		if (scope >= RT_SCOPE_LINK) {
1117 			NL_SET_ERR_MSG(extack, "Nexthop has invalid scope");
1118 			return -EINVAL;
1119 		}
1120 		dev = __dev_get_by_index(net, nh->fib_nh_oif);
1121 		if (!dev) {
1122 			NL_SET_ERR_MSG(extack, "Nexthop device required for onlink");
1123 			return -ENODEV;
1124 		}
1125 		if (!(dev->flags & IFF_UP)) {
1126 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1127 			return -ENETDOWN;
1128 		}
1129 		addr_type = inet_addr_type_dev_table(net, dev, nh->fib_nh_gw4);
1130 		if (addr_type != RTN_UNICAST) {
1131 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1132 			return -EINVAL;
1133 		}
1134 		if (!netif_carrier_ok(dev))
1135 			nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1136 		nh->fib_nh_dev = dev;
1137 		dev_hold(dev);
1138 		nh->fib_nh_scope = RT_SCOPE_LINK;
1139 		return 0;
1140 	}
1141 	rcu_read_lock();
1142 	{
1143 		struct fib_table *tbl = NULL;
1144 		struct flowi4 fl4 = {
1145 			.daddr = nh->fib_nh_gw4,
1146 			.flowi4_scope = scope + 1,
1147 			.flowi4_oif = nh->fib_nh_oif,
1148 			.flowi4_iif = LOOPBACK_IFINDEX,
1149 		};
1150 
1151 		/* It is not necessary, but requires a bit of thinking */
1152 		if (fl4.flowi4_scope < RT_SCOPE_LINK)
1153 			fl4.flowi4_scope = RT_SCOPE_LINK;
1154 
1155 		if (table && table != RT_TABLE_MAIN)
1156 			tbl = fib_get_table(net, table);
1157 
1158 		if (tbl)
1159 			err = fib_table_lookup(tbl, &fl4, &res,
1160 					       FIB_LOOKUP_IGNORE_LINKSTATE |
1161 					       FIB_LOOKUP_NOREF);
1162 
1163 		/* on error or if no table given do full lookup. This
1164 		 * is needed for example when nexthops are in the local
1165 		 * table rather than the given table
1166 		 */
1167 		if (!tbl || err) {
1168 			err = fib_lookup(net, &fl4, &res,
1169 					 FIB_LOOKUP_IGNORE_LINKSTATE);
1170 		}
1171 
1172 		if (err) {
1173 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1174 			goto out;
1175 		}
1176 	}
1177 
1178 	err = -EINVAL;
1179 	if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) {
1180 		NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1181 		goto out;
1182 	}
1183 	nh->fib_nh_scope = res.scope;
1184 	nh->fib_nh_oif = FIB_RES_OIF(res);
1185 	nh->fib_nh_dev = dev = FIB_RES_DEV(res);
1186 	if (!dev) {
1187 		NL_SET_ERR_MSG(extack,
1188 			       "No egress device for nexthop gateway");
1189 		goto out;
1190 	}
1191 	dev_hold(dev);
1192 	if (!netif_carrier_ok(dev))
1193 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1194 	err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN;
1195 out:
1196 	rcu_read_unlock();
1197 	return err;
1198 }
1199 
fib_check_nh_nongw(struct net * net,struct fib_nh * nh,struct netlink_ext_ack * extack)1200 static int fib_check_nh_nongw(struct net *net, struct fib_nh *nh,
1201 			      struct netlink_ext_ack *extack)
1202 {
1203 	struct in_device *in_dev;
1204 	int err;
1205 
1206 	if (nh->fib_nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) {
1207 		NL_SET_ERR_MSG(extack,
1208 			       "Invalid flags for nexthop - PERVASIVE and ONLINK can not be set");
1209 		return -EINVAL;
1210 	}
1211 
1212 	rcu_read_lock();
1213 
1214 	err = -ENODEV;
1215 	in_dev = inetdev_by_index(net, nh->fib_nh_oif);
1216 	if (!in_dev)
1217 		goto out;
1218 	err = -ENETDOWN;
1219 	if (!(in_dev->dev->flags & IFF_UP)) {
1220 		NL_SET_ERR_MSG(extack, "Device for nexthop is not up");
1221 		goto out;
1222 	}
1223 
1224 	nh->fib_nh_dev = in_dev->dev;
1225 	dev_hold(nh->fib_nh_dev);
1226 	nh->fib_nh_scope = RT_SCOPE_LINK;
1227 	if (!netif_carrier_ok(nh->fib_nh_dev))
1228 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1229 	err = 0;
1230 out:
1231 	rcu_read_unlock();
1232 	return err;
1233 }
1234 
fib_check_nh(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1235 int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
1236 		 struct netlink_ext_ack *extack)
1237 {
1238 	int err;
1239 
1240 	if (nh->fib_nh_gw_family == AF_INET)
1241 		err = fib_check_nh_v4_gw(net, nh, table, scope, extack);
1242 	else if (nh->fib_nh_gw_family == AF_INET6)
1243 		err = fib_check_nh_v6_gw(net, nh, table, extack);
1244 	else
1245 		err = fib_check_nh_nongw(net, nh, extack);
1246 
1247 	return err;
1248 }
1249 
fib_laddr_hashfn(__be32 val)1250 static inline unsigned int fib_laddr_hashfn(__be32 val)
1251 {
1252 	unsigned int mask = (fib_info_hash_size - 1);
1253 
1254 	return ((__force u32)val ^
1255 		((__force u32)val >> 7) ^
1256 		((__force u32)val >> 14)) & mask;
1257 }
1258 
fib_info_hash_alloc(int bytes)1259 static struct hlist_head *fib_info_hash_alloc(int bytes)
1260 {
1261 	if (bytes <= PAGE_SIZE)
1262 		return kzalloc(bytes, GFP_KERNEL);
1263 	else
1264 		return (struct hlist_head *)
1265 			__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1266 					 get_order(bytes));
1267 }
1268 
fib_info_hash_free(struct hlist_head * hash,int bytes)1269 static void fib_info_hash_free(struct hlist_head *hash, int bytes)
1270 {
1271 	if (!hash)
1272 		return;
1273 
1274 	if (bytes <= PAGE_SIZE)
1275 		kfree(hash);
1276 	else
1277 		free_pages((unsigned long) hash, get_order(bytes));
1278 }
1279 
fib_info_hash_move(struct hlist_head * new_info_hash,struct hlist_head * new_laddrhash,unsigned int new_size)1280 static void fib_info_hash_move(struct hlist_head *new_info_hash,
1281 			       struct hlist_head *new_laddrhash,
1282 			       unsigned int new_size)
1283 {
1284 	struct hlist_head *old_info_hash, *old_laddrhash;
1285 	unsigned int old_size = fib_info_hash_size;
1286 	unsigned int i, bytes;
1287 
1288 	spin_lock_bh(&fib_info_lock);
1289 	old_info_hash = fib_info_hash;
1290 	old_laddrhash = fib_info_laddrhash;
1291 	fib_info_hash_size = new_size;
1292 
1293 	for (i = 0; i < old_size; i++) {
1294 		struct hlist_head *head = &fib_info_hash[i];
1295 		struct hlist_node *n;
1296 		struct fib_info *fi;
1297 
1298 		hlist_for_each_entry_safe(fi, n, head, fib_hash) {
1299 			struct hlist_head *dest;
1300 			unsigned int new_hash;
1301 
1302 			new_hash = fib_info_hashfn(fi);
1303 			dest = &new_info_hash[new_hash];
1304 			hlist_add_head(&fi->fib_hash, dest);
1305 		}
1306 	}
1307 	fib_info_hash = new_info_hash;
1308 
1309 	for (i = 0; i < old_size; i++) {
1310 		struct hlist_head *lhead = &fib_info_laddrhash[i];
1311 		struct hlist_node *n;
1312 		struct fib_info *fi;
1313 
1314 		hlist_for_each_entry_safe(fi, n, lhead, fib_lhash) {
1315 			struct hlist_head *ldest;
1316 			unsigned int new_hash;
1317 
1318 			new_hash = fib_laddr_hashfn(fi->fib_prefsrc);
1319 			ldest = &new_laddrhash[new_hash];
1320 			hlist_add_head(&fi->fib_lhash, ldest);
1321 		}
1322 	}
1323 	fib_info_laddrhash = new_laddrhash;
1324 
1325 	spin_unlock_bh(&fib_info_lock);
1326 
1327 	bytes = old_size * sizeof(struct hlist_head *);
1328 	fib_info_hash_free(old_info_hash, bytes);
1329 	fib_info_hash_free(old_laddrhash, bytes);
1330 }
1331 
fib_info_update_nhc_saddr(struct net * net,struct fib_nh_common * nhc,unsigned char scope)1332 __be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
1333 				 unsigned char scope)
1334 {
1335 	struct fib_nh *nh;
1336 	__be32 saddr;
1337 
1338 	if (nhc->nhc_family != AF_INET)
1339 		return inet_select_addr(nhc->nhc_dev, 0, scope);
1340 
1341 	nh = container_of(nhc, struct fib_nh, nh_common);
1342 	saddr = inet_select_addr(nh->fib_nh_dev, nh->fib_nh_gw4, scope);
1343 
1344 	WRITE_ONCE(nh->nh_saddr, saddr);
1345 	WRITE_ONCE(nh->nh_saddr_genid, atomic_read(&net->ipv4.dev_addr_genid));
1346 
1347 	return saddr;
1348 }
1349 
fib_result_prefsrc(struct net * net,struct fib_result * res)1350 __be32 fib_result_prefsrc(struct net *net, struct fib_result *res)
1351 {
1352 	struct fib_nh_common *nhc = res->nhc;
1353 
1354 	if (res->fi->fib_prefsrc)
1355 		return res->fi->fib_prefsrc;
1356 
1357 	if (nhc->nhc_family == AF_INET) {
1358 		struct fib_nh *nh;
1359 
1360 		nh = container_of(nhc, struct fib_nh, nh_common);
1361 		if (READ_ONCE(nh->nh_saddr_genid) ==
1362 		    atomic_read(&net->ipv4.dev_addr_genid))
1363 			return READ_ONCE(nh->nh_saddr);
1364 	}
1365 
1366 	return fib_info_update_nhc_saddr(net, nhc, res->fi->fib_scope);
1367 }
1368 
fib_valid_prefsrc(struct fib_config * cfg,__be32 fib_prefsrc)1369 static bool fib_valid_prefsrc(struct fib_config *cfg, __be32 fib_prefsrc)
1370 {
1371 	if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst ||
1372 	    fib_prefsrc != cfg->fc_dst) {
1373 		u32 tb_id = cfg->fc_table;
1374 		int rc;
1375 
1376 		if (tb_id == RT_TABLE_MAIN)
1377 			tb_id = RT_TABLE_LOCAL;
1378 
1379 		rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1380 					  fib_prefsrc, tb_id);
1381 
1382 		if (rc != RTN_LOCAL && tb_id != RT_TABLE_LOCAL) {
1383 			rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1384 						  fib_prefsrc, RT_TABLE_LOCAL);
1385 		}
1386 
1387 		if (rc != RTN_LOCAL)
1388 			return false;
1389 	}
1390 	return true;
1391 }
1392 
fib_create_info(struct fib_config * cfg,struct netlink_ext_ack * extack)1393 struct fib_info *fib_create_info(struct fib_config *cfg,
1394 				 struct netlink_ext_ack *extack)
1395 {
1396 	int err;
1397 	struct fib_info *fi = NULL;
1398 	struct nexthop *nh = NULL;
1399 	struct fib_info *ofi;
1400 	int nhs = 1;
1401 	struct net *net = cfg->fc_nlinfo.nl_net;
1402 
1403 	if (cfg->fc_type > RTN_MAX)
1404 		goto err_inval;
1405 
1406 	/* Fast check to catch the most weird cases */
1407 	if (fib_props[cfg->fc_type].scope > cfg->fc_scope) {
1408 		NL_SET_ERR_MSG(extack, "Invalid scope");
1409 		goto err_inval;
1410 	}
1411 
1412 	if (cfg->fc_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
1413 		NL_SET_ERR_MSG(extack,
1414 			       "Invalid rtm_flags - can not contain DEAD or LINKDOWN");
1415 		goto err_inval;
1416 	}
1417 
1418 	if (cfg->fc_nh_id) {
1419 		if (!cfg->fc_mx) {
1420 			fi = fib_find_info_nh(net, cfg);
1421 			if (fi) {
1422 				fi->fib_treeref++;
1423 				return fi;
1424 			}
1425 		}
1426 
1427 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
1428 		if (!nh) {
1429 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
1430 			goto err_inval;
1431 		}
1432 		nhs = 0;
1433 	}
1434 
1435 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1436 	if (cfg->fc_mp) {
1437 		nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len, extack);
1438 		if (nhs == 0)
1439 			goto err_inval;
1440 	}
1441 #endif
1442 
1443 	err = -ENOBUFS;
1444 	if (fib_info_cnt >= fib_info_hash_size) {
1445 		unsigned int new_size = fib_info_hash_size << 1;
1446 		struct hlist_head *new_info_hash;
1447 		struct hlist_head *new_laddrhash;
1448 		unsigned int bytes;
1449 
1450 		if (!new_size)
1451 			new_size = 16;
1452 		bytes = new_size * sizeof(struct hlist_head *);
1453 		new_info_hash = fib_info_hash_alloc(bytes);
1454 		new_laddrhash = fib_info_hash_alloc(bytes);
1455 		if (!new_info_hash || !new_laddrhash) {
1456 			fib_info_hash_free(new_info_hash, bytes);
1457 			fib_info_hash_free(new_laddrhash, bytes);
1458 		} else
1459 			fib_info_hash_move(new_info_hash, new_laddrhash, new_size);
1460 
1461 		if (!fib_info_hash_size)
1462 			goto failure;
1463 	}
1464 
1465 	fi = kzalloc(struct_size(fi, fib_nh, nhs), GFP_KERNEL);
1466 	if (!fi)
1467 		goto failure;
1468 	fi->fib_metrics = ip_fib_metrics_init(fi->fib_net, cfg->fc_mx,
1469 					      cfg->fc_mx_len, extack);
1470 	if (IS_ERR(fi->fib_metrics)) {
1471 		err = PTR_ERR(fi->fib_metrics);
1472 		kfree(fi);
1473 		return ERR_PTR(err);
1474 	}
1475 
1476 	fib_info_cnt++;
1477 	fi->fib_net = net;
1478 	fi->fib_protocol = cfg->fc_protocol;
1479 	fi->fib_scope = cfg->fc_scope;
1480 	fi->fib_flags = cfg->fc_flags;
1481 	fi->fib_priority = cfg->fc_priority;
1482 	fi->fib_prefsrc = cfg->fc_prefsrc;
1483 	fi->fib_type = cfg->fc_type;
1484 	fi->fib_tb_id = cfg->fc_table;
1485 
1486 	fi->fib_nhs = nhs;
1487 	if (nh) {
1488 		if (!nexthop_get(nh)) {
1489 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1490 			err = -EINVAL;
1491 		} else {
1492 			err = 0;
1493 			fi->nh = nh;
1494 		}
1495 	} else {
1496 		change_nexthops(fi) {
1497 			nexthop_nh->nh_parent = fi;
1498 		} endfor_nexthops(fi)
1499 
1500 		if (cfg->fc_mp)
1501 			err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg,
1502 					  extack);
1503 		else
1504 			err = fib_nh_init(net, fi->fib_nh, cfg, 1, extack);
1505 	}
1506 
1507 	if (err != 0)
1508 		goto failure;
1509 
1510 	if (fib_props[cfg->fc_type].error) {
1511 		if (cfg->fc_gw_family || cfg->fc_oif || cfg->fc_mp) {
1512 			NL_SET_ERR_MSG(extack,
1513 				       "Gateway, device and multipath can not be specified for this route type");
1514 			goto err_inval;
1515 		}
1516 		goto link_it;
1517 	} else {
1518 		switch (cfg->fc_type) {
1519 		case RTN_UNICAST:
1520 		case RTN_LOCAL:
1521 		case RTN_BROADCAST:
1522 		case RTN_ANYCAST:
1523 		case RTN_MULTICAST:
1524 			break;
1525 		default:
1526 			NL_SET_ERR_MSG(extack, "Invalid route type");
1527 			goto err_inval;
1528 		}
1529 	}
1530 
1531 	if (cfg->fc_scope > RT_SCOPE_HOST) {
1532 		NL_SET_ERR_MSG(extack, "Invalid scope");
1533 		goto err_inval;
1534 	}
1535 
1536 	if (fi->nh) {
1537 		err = fib_check_nexthop(fi->nh, cfg->fc_scope, extack);
1538 		if (err)
1539 			goto failure;
1540 	} else if (cfg->fc_scope == RT_SCOPE_HOST) {
1541 		struct fib_nh *nh = fi->fib_nh;
1542 
1543 		/* Local address is added. */
1544 		if (nhs != 1) {
1545 			NL_SET_ERR_MSG(extack,
1546 				       "Route with host scope can not have multiple nexthops");
1547 			goto err_inval;
1548 		}
1549 		if (nh->fib_nh_gw_family) {
1550 			NL_SET_ERR_MSG(extack,
1551 				       "Route with host scope can not have a gateway");
1552 			goto err_inval;
1553 		}
1554 		nh->fib_nh_scope = RT_SCOPE_NOWHERE;
1555 		nh->fib_nh_dev = dev_get_by_index(net, nh->fib_nh_oif);
1556 		err = -ENODEV;
1557 		if (!nh->fib_nh_dev)
1558 			goto failure;
1559 	} else {
1560 		int linkdown = 0;
1561 
1562 		change_nexthops(fi) {
1563 			err = fib_check_nh(cfg->fc_nlinfo.nl_net, nexthop_nh,
1564 					   cfg->fc_table, cfg->fc_scope,
1565 					   extack);
1566 			if (err != 0)
1567 				goto failure;
1568 			if (nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN)
1569 				linkdown++;
1570 		} endfor_nexthops(fi)
1571 		if (linkdown == fi->fib_nhs)
1572 			fi->fib_flags |= RTNH_F_LINKDOWN;
1573 	}
1574 
1575 	if (fi->fib_prefsrc && !fib_valid_prefsrc(cfg, fi->fib_prefsrc)) {
1576 		NL_SET_ERR_MSG(extack, "Invalid prefsrc address");
1577 		goto err_inval;
1578 	}
1579 
1580 	if (!fi->nh) {
1581 		change_nexthops(fi) {
1582 			fib_info_update_nhc_saddr(net, &nexthop_nh->nh_common,
1583 						  fi->fib_scope);
1584 			if (nexthop_nh->fib_nh_gw_family == AF_INET6)
1585 				fi->fib_nh_is_v6 = true;
1586 		} endfor_nexthops(fi)
1587 
1588 		fib_rebalance(fi);
1589 	}
1590 
1591 link_it:
1592 	ofi = fib_find_info(fi);
1593 	if (ofi) {
1594 		/* fib_table_lookup() should not see @fi yet. */
1595 		fi->fib_dead = 1;
1596 		free_fib_info(fi);
1597 		ofi->fib_treeref++;
1598 		return ofi;
1599 	}
1600 
1601 	fi->fib_treeref++;
1602 	refcount_set(&fi->fib_clntref, 1);
1603 	spin_lock_bh(&fib_info_lock);
1604 	hlist_add_head(&fi->fib_hash,
1605 		       &fib_info_hash[fib_info_hashfn(fi)]);
1606 	if (fi->fib_prefsrc) {
1607 		struct hlist_head *head;
1608 
1609 		head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)];
1610 		hlist_add_head(&fi->fib_lhash, head);
1611 	}
1612 	if (fi->nh) {
1613 		list_add(&fi->nh_list, &nh->fi_list);
1614 	} else {
1615 		change_nexthops(fi) {
1616 			struct hlist_head *head;
1617 
1618 			if (!nexthop_nh->fib_nh_dev)
1619 				continue;
1620 			head = fib_info_devhash_bucket(nexthop_nh->fib_nh_dev);
1621 			hlist_add_head(&nexthop_nh->nh_hash, head);
1622 		} endfor_nexthops(fi)
1623 	}
1624 	spin_unlock_bh(&fib_info_lock);
1625 	return fi;
1626 
1627 err_inval:
1628 	err = -EINVAL;
1629 
1630 failure:
1631 	if (fi) {
1632 		/* fib_table_lookup() should not see @fi yet. */
1633 		fi->fib_dead = 1;
1634 		free_fib_info(fi);
1635 	}
1636 
1637 	return ERR_PTR(err);
1638 }
1639 
fib_nexthop_info(struct sk_buff * skb,const struct fib_nh_common * nhc,u8 rt_family,unsigned char * flags,bool skip_oif)1640 int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nhc,
1641 		     u8 rt_family, unsigned char *flags, bool skip_oif)
1642 {
1643 	if (nhc->nhc_flags & RTNH_F_DEAD)
1644 		*flags |= RTNH_F_DEAD;
1645 
1646 	if (nhc->nhc_flags & RTNH_F_LINKDOWN) {
1647 		*flags |= RTNH_F_LINKDOWN;
1648 
1649 		rcu_read_lock();
1650 		switch (nhc->nhc_family) {
1651 		case AF_INET:
1652 			if (ip_ignore_linkdown(nhc->nhc_dev))
1653 				*flags |= RTNH_F_DEAD;
1654 			break;
1655 		case AF_INET6:
1656 			if (ip6_ignore_linkdown(nhc->nhc_dev))
1657 				*flags |= RTNH_F_DEAD;
1658 			break;
1659 		}
1660 		rcu_read_unlock();
1661 	}
1662 
1663 	switch (nhc->nhc_gw_family) {
1664 	case AF_INET:
1665 		if (nla_put_in_addr(skb, RTA_GATEWAY, nhc->nhc_gw.ipv4))
1666 			goto nla_put_failure;
1667 		break;
1668 	case AF_INET6:
1669 		/* if gateway family does not match nexthop family
1670 		 * gateway is encoded as RTA_VIA
1671 		 */
1672 		if (rt_family != nhc->nhc_gw_family) {
1673 			int alen = sizeof(struct in6_addr);
1674 			struct nlattr *nla;
1675 			struct rtvia *via;
1676 
1677 			nla = nla_reserve(skb, RTA_VIA, alen + 2);
1678 			if (!nla)
1679 				goto nla_put_failure;
1680 
1681 			via = nla_data(nla);
1682 			via->rtvia_family = AF_INET6;
1683 			memcpy(via->rtvia_addr, &nhc->nhc_gw.ipv6, alen);
1684 		} else if (nla_put_in6_addr(skb, RTA_GATEWAY,
1685 					    &nhc->nhc_gw.ipv6) < 0) {
1686 			goto nla_put_failure;
1687 		}
1688 		break;
1689 	}
1690 
1691 	*flags |= (nhc->nhc_flags & RTNH_F_ONLINK);
1692 	if (nhc->nhc_flags & RTNH_F_OFFLOAD)
1693 		*flags |= RTNH_F_OFFLOAD;
1694 
1695 	if (!skip_oif && nhc->nhc_dev &&
1696 	    nla_put_u32(skb, RTA_OIF, nhc->nhc_dev->ifindex))
1697 		goto nla_put_failure;
1698 
1699 	if (nhc->nhc_lwtstate &&
1700 	    lwtunnel_fill_encap(skb, nhc->nhc_lwtstate,
1701 				RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
1702 		goto nla_put_failure;
1703 
1704 	return 0;
1705 
1706 nla_put_failure:
1707 	return -EMSGSIZE;
1708 }
1709 EXPORT_SYMBOL_GPL(fib_nexthop_info);
1710 
1711 #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)1712 int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nhc,
1713 		    int nh_weight, u8 rt_family, u32 nh_tclassid)
1714 {
1715 	const struct net_device *dev = nhc->nhc_dev;
1716 	struct rtnexthop *rtnh;
1717 	unsigned char flags = 0;
1718 
1719 	rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1720 	if (!rtnh)
1721 		goto nla_put_failure;
1722 
1723 	rtnh->rtnh_hops = nh_weight - 1;
1724 	rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
1725 
1726 	if (fib_nexthop_info(skb, nhc, rt_family, &flags, true) < 0)
1727 		goto nla_put_failure;
1728 
1729 	rtnh->rtnh_flags = flags;
1730 
1731 	if (nh_tclassid && nla_put_u32(skb, RTA_FLOW, nh_tclassid))
1732 		goto nla_put_failure;
1733 
1734 	/* length of rtnetlink header + attributes */
1735 	rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
1736 
1737 	return 0;
1738 
1739 nla_put_failure:
1740 	return -EMSGSIZE;
1741 }
1742 EXPORT_SYMBOL_GPL(fib_add_nexthop);
1743 #endif
1744 
1745 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1746 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1747 {
1748 	struct nlattr *mp;
1749 
1750 	mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
1751 	if (!mp)
1752 		goto nla_put_failure;
1753 
1754 	if (unlikely(fi->nh)) {
1755 		if (nexthop_mpath_fill_node(skb, fi->nh, AF_INET) < 0)
1756 			goto nla_put_failure;
1757 		goto mp_end;
1758 	}
1759 
1760 	for_nexthops(fi) {
1761 		u32 nh_tclassid = 0;
1762 #ifdef CONFIG_IP_ROUTE_CLASSID
1763 		nh_tclassid = nh->nh_tclassid;
1764 #endif
1765 		if (fib_add_nexthop(skb, &nh->nh_common, nh->fib_nh_weight,
1766 				    AF_INET, nh_tclassid) < 0)
1767 			goto nla_put_failure;
1768 	} endfor_nexthops(fi);
1769 
1770 mp_end:
1771 	nla_nest_end(skb, mp);
1772 
1773 	return 0;
1774 
1775 nla_put_failure:
1776 	return -EMSGSIZE;
1777 }
1778 #else
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1779 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1780 {
1781 	return 0;
1782 }
1783 #endif
1784 
fib_dump_info(struct sk_buff * skb,u32 portid,u32 seq,int event,u32 tb_id,u8 type,__be32 dst,int dst_len,u8 tos,struct fib_info * fi,unsigned int flags)1785 int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
1786 		  u32 tb_id, u8 type, __be32 dst, int dst_len, u8 tos,
1787 		  struct fib_info *fi, unsigned int flags)
1788 {
1789 	unsigned int nhs = fib_info_num_path(fi);
1790 	struct nlmsghdr *nlh;
1791 	struct rtmsg *rtm;
1792 
1793 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1794 	if (!nlh)
1795 		return -EMSGSIZE;
1796 
1797 	rtm = nlmsg_data(nlh);
1798 	rtm->rtm_family = AF_INET;
1799 	rtm->rtm_dst_len = dst_len;
1800 	rtm->rtm_src_len = 0;
1801 	rtm->rtm_tos = tos;
1802 	if (tb_id < 256)
1803 		rtm->rtm_table = tb_id;
1804 	else
1805 		rtm->rtm_table = RT_TABLE_COMPAT;
1806 	if (nla_put_u32(skb, RTA_TABLE, tb_id))
1807 		goto nla_put_failure;
1808 	rtm->rtm_type = type;
1809 	rtm->rtm_flags = fi->fib_flags;
1810 	rtm->rtm_scope = fi->fib_scope;
1811 	rtm->rtm_protocol = fi->fib_protocol;
1812 
1813 	if (rtm->rtm_dst_len &&
1814 	    nla_put_in_addr(skb, RTA_DST, dst))
1815 		goto nla_put_failure;
1816 	if (fi->fib_priority &&
1817 	    nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority))
1818 		goto nla_put_failure;
1819 	if (rtnetlink_put_metrics(skb, fi->fib_metrics->metrics) < 0)
1820 		goto nla_put_failure;
1821 
1822 	if (fi->fib_prefsrc &&
1823 	    nla_put_in_addr(skb, RTA_PREFSRC, fi->fib_prefsrc))
1824 		goto nla_put_failure;
1825 
1826 	if (fi->nh) {
1827 		if (nla_put_u32(skb, RTA_NH_ID, fi->nh->id))
1828 			goto nla_put_failure;
1829 		if (nexthop_is_blackhole(fi->nh))
1830 			rtm->rtm_type = RTN_BLACKHOLE;
1831 	}
1832 
1833 	if (nhs == 1) {
1834 		const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
1835 		unsigned char flags = 0;
1836 
1837 		if (fib_nexthop_info(skb, nhc, AF_INET, &flags, false) < 0)
1838 			goto nla_put_failure;
1839 
1840 		rtm->rtm_flags = flags;
1841 #ifdef CONFIG_IP_ROUTE_CLASSID
1842 		if (nhc->nhc_family == AF_INET) {
1843 			struct fib_nh *nh;
1844 
1845 			nh = container_of(nhc, struct fib_nh, nh_common);
1846 			if (nh->nh_tclassid &&
1847 			    nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid))
1848 				goto nla_put_failure;
1849 		}
1850 #endif
1851 	} else {
1852 		if (fib_add_multipath(skb, fi) < 0)
1853 			goto nla_put_failure;
1854 	}
1855 
1856 	nlmsg_end(skb, nlh);
1857 	return 0;
1858 
1859 nla_put_failure:
1860 	nlmsg_cancel(skb, nlh);
1861 	return -EMSGSIZE;
1862 }
1863 
1864 /*
1865  * Update FIB if:
1866  * - local address disappeared -> we must delete all the entries
1867  *   referring to it.
1868  * - device went down -> we must shutdown all nexthops going via it.
1869  */
fib_sync_down_addr(struct net_device * dev,__be32 local)1870 int fib_sync_down_addr(struct net_device *dev, __be32 local)
1871 {
1872 	int ret = 0;
1873 	unsigned int hash = fib_laddr_hashfn(local);
1874 	struct hlist_head *head = &fib_info_laddrhash[hash];
1875 	int tb_id = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
1876 	struct net *net = dev_net(dev);
1877 	struct fib_info *fi;
1878 
1879 	if (!fib_info_laddrhash || local == 0)
1880 		return 0;
1881 
1882 	hlist_for_each_entry(fi, head, fib_lhash) {
1883 		if (!net_eq(fi->fib_net, net) ||
1884 		    fi->fib_tb_id != tb_id)
1885 			continue;
1886 		if (fi->fib_prefsrc == local) {
1887 			fi->fib_flags |= RTNH_F_DEAD;
1888 			ret++;
1889 		}
1890 	}
1891 	return ret;
1892 }
1893 
call_fib_nh_notifiers(struct fib_nh * nh,enum fib_event_type event_type)1894 static int call_fib_nh_notifiers(struct fib_nh *nh,
1895 				 enum fib_event_type event_type)
1896 {
1897 	bool ignore_link_down = ip_ignore_linkdown(nh->fib_nh_dev);
1898 	struct fib_nh_notifier_info info = {
1899 		.fib_nh = nh,
1900 	};
1901 
1902 	switch (event_type) {
1903 	case FIB_EVENT_NH_ADD:
1904 		if (nh->fib_nh_flags & RTNH_F_DEAD)
1905 			break;
1906 		if (ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN)
1907 			break;
1908 		return call_fib4_notifiers(dev_net(nh->fib_nh_dev), event_type,
1909 					   &info.info);
1910 	case FIB_EVENT_NH_DEL:
1911 		if ((ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN) ||
1912 		    (nh->fib_nh_flags & RTNH_F_DEAD))
1913 			return call_fib4_notifiers(dev_net(nh->fib_nh_dev),
1914 						   event_type, &info.info);
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 	struct hlist_head *head = fib_info_devhash_bucket(dev);
1963 	struct fib_nh *nh;
1964 
1965 	hlist_for_each_entry(nh, head, nh_hash) {
1966 		if (nh->fib_nh_dev == dev)
1967 			fib_nhc_update_mtu(&nh->nh_common, dev->mtu, orig_mtu);
1968 	}
1969 }
1970 
1971 /* Event              force Flags           Description
1972  * NETDEV_CHANGE      0     LINKDOWN        Carrier OFF, not for scope host
1973  * NETDEV_DOWN        0     LINKDOWN|DEAD   Link down, not for scope host
1974  * NETDEV_DOWN        1     LINKDOWN|DEAD   Last address removed
1975  * NETDEV_UNREGISTER  1     LINKDOWN|DEAD   Device removed
1976  *
1977  * only used when fib_nh is built into fib_info
1978  */
fib_sync_down_dev(struct net_device * dev,unsigned long event,bool force)1979 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force)
1980 {
1981 	struct hlist_head *head = fib_info_devhash_bucket(dev);
1982 	struct fib_info *prev_fi = NULL;
1983 	int scope = RT_SCOPE_NOWHERE;
1984 	struct fib_nh *nh;
1985 	int ret = 0;
1986 
1987 	if (force)
1988 		scope = -1;
1989 
1990 	hlist_for_each_entry(nh, head, nh_hash) {
1991 		struct fib_info *fi = nh->nh_parent;
1992 		int dead;
1993 
1994 		BUG_ON(!fi->fib_nhs);
1995 		if (nh->fib_nh_dev != dev || fi == prev_fi)
1996 			continue;
1997 		prev_fi = fi;
1998 		dead = 0;
1999 		change_nexthops(fi) {
2000 			if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD)
2001 				dead++;
2002 			else if (nexthop_nh->fib_nh_dev == dev &&
2003 				 nexthop_nh->fib_nh_scope != scope) {
2004 				switch (event) {
2005 				case NETDEV_DOWN:
2006 				case NETDEV_UNREGISTER:
2007 					nexthop_nh->fib_nh_flags |= RTNH_F_DEAD;
2008 					/* fall through */
2009 				case NETDEV_CHANGE:
2010 					nexthop_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
2011 					break;
2012 				}
2013 				call_fib_nh_notifiers(nexthop_nh,
2014 						      FIB_EVENT_NH_DEL);
2015 				dead++;
2016 			}
2017 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2018 			if (event == NETDEV_UNREGISTER &&
2019 			    nexthop_nh->fib_nh_dev == dev) {
2020 				dead = fi->fib_nhs;
2021 				break;
2022 			}
2023 #endif
2024 		} endfor_nexthops(fi)
2025 		if (dead == fi->fib_nhs) {
2026 			switch (event) {
2027 			case NETDEV_DOWN:
2028 			case NETDEV_UNREGISTER:
2029 				fi->fib_flags |= RTNH_F_DEAD;
2030 				/* fall through */
2031 			case NETDEV_CHANGE:
2032 				fi->fib_flags |= RTNH_F_LINKDOWN;
2033 				break;
2034 			}
2035 			ret++;
2036 		}
2037 
2038 		fib_rebalance(fi);
2039 	}
2040 
2041 	return ret;
2042 }
2043 
2044 /* Must be invoked inside of an RCU protected region.  */
fib_select_default(const struct flowi4 * flp,struct fib_result * res)2045 static void fib_select_default(const struct flowi4 *flp, struct fib_result *res)
2046 {
2047 	struct fib_info *fi = NULL, *last_resort = NULL;
2048 	struct hlist_head *fa_head = res->fa_head;
2049 	struct fib_table *tb = res->table;
2050 	u8 slen = 32 - res->prefixlen;
2051 	int order = -1, last_idx = -1;
2052 	struct fib_alias *fa, *fa1 = NULL;
2053 	u32 last_prio = res->fi->fib_priority;
2054 	u8 last_tos = 0;
2055 
2056 	hlist_for_each_entry_rcu(fa, fa_head, fa_list) {
2057 		struct fib_info *next_fi = fa->fa_info;
2058 		struct fib_nh_common *nhc;
2059 
2060 		if (fa->fa_slen != slen)
2061 			continue;
2062 		if (fa->fa_tos && fa->fa_tos != flp->flowi4_tos)
2063 			continue;
2064 		if (fa->tb_id != tb->tb_id)
2065 			continue;
2066 		if (next_fi->fib_priority > last_prio &&
2067 		    fa->fa_tos == last_tos) {
2068 			if (last_tos)
2069 				continue;
2070 			break;
2071 		}
2072 		if (next_fi->fib_flags & RTNH_F_DEAD)
2073 			continue;
2074 		last_tos = fa->fa_tos;
2075 		last_prio = next_fi->fib_priority;
2076 
2077 		if (next_fi->fib_scope != res->scope ||
2078 		    fa->fa_type != RTN_UNICAST)
2079 			continue;
2080 
2081 		nhc = fib_info_nhc(next_fi, 0);
2082 		if (!nhc->nhc_gw_family || nhc->nhc_scope != RT_SCOPE_LINK)
2083 			continue;
2084 
2085 		fib_alias_accessed(fa);
2086 
2087 		if (!fi) {
2088 			if (next_fi != res->fi)
2089 				break;
2090 			fa1 = fa;
2091 		} else if (!fib_detect_death(fi, order, &last_resort,
2092 					     &last_idx, fa1->fa_default)) {
2093 			fib_result_assign(res, fi);
2094 			fa1->fa_default = order;
2095 			goto out;
2096 		}
2097 		fi = next_fi;
2098 		order++;
2099 	}
2100 
2101 	if (order <= 0 || !fi) {
2102 		if (fa1)
2103 			fa1->fa_default = -1;
2104 		goto out;
2105 	}
2106 
2107 	if (!fib_detect_death(fi, order, &last_resort, &last_idx,
2108 			      fa1->fa_default)) {
2109 		fib_result_assign(res, fi);
2110 		fa1->fa_default = order;
2111 		goto out;
2112 	}
2113 
2114 	if (last_idx >= 0)
2115 		fib_result_assign(res, last_resort);
2116 	fa1->fa_default = last_idx;
2117 out:
2118 	return;
2119 }
2120 
2121 /*
2122  * Dead device goes up. We wake up dead nexthops.
2123  * It takes sense only on multipath routes.
2124  *
2125  * only used when fib_nh is built into fib_info
2126  */
fib_sync_up(struct net_device * dev,unsigned char nh_flags)2127 int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
2128 {
2129 	struct fib_info *prev_fi;
2130 	struct hlist_head *head;
2131 	struct fib_nh *nh;
2132 	int ret;
2133 
2134 	if (!(dev->flags & IFF_UP))
2135 		return 0;
2136 
2137 	if (nh_flags & RTNH_F_DEAD) {
2138 		unsigned int flags = dev_get_flags(dev);
2139 
2140 		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
2141 			nh_flags |= RTNH_F_LINKDOWN;
2142 	}
2143 
2144 	prev_fi = NULL;
2145 	head = fib_info_devhash_bucket(dev);
2146 	ret = 0;
2147 
2148 	hlist_for_each_entry(nh, head, nh_hash) {
2149 		struct fib_info *fi = nh->nh_parent;
2150 		int alive;
2151 
2152 		BUG_ON(!fi->fib_nhs);
2153 		if (nh->fib_nh_dev != dev || fi == prev_fi)
2154 			continue;
2155 
2156 		prev_fi = fi;
2157 		alive = 0;
2158 		change_nexthops(fi) {
2159 			if (!(nexthop_nh->fib_nh_flags & nh_flags)) {
2160 				alive++;
2161 				continue;
2162 			}
2163 			if (!nexthop_nh->fib_nh_dev ||
2164 			    !(nexthop_nh->fib_nh_dev->flags & IFF_UP))
2165 				continue;
2166 			if (nexthop_nh->fib_nh_dev != dev ||
2167 			    !__in_dev_get_rtnl(dev))
2168 				continue;
2169 			alive++;
2170 			nexthop_nh->fib_nh_flags &= ~nh_flags;
2171 			call_fib_nh_notifiers(nexthop_nh, FIB_EVENT_NH_ADD);
2172 		} endfor_nexthops(fi)
2173 
2174 		if (alive > 0) {
2175 			fi->fib_flags &= ~nh_flags;
2176 			ret++;
2177 		}
2178 
2179 		fib_rebalance(fi);
2180 	}
2181 
2182 	return ret;
2183 }
2184 
2185 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_good_nh(const struct fib_nh * nh)2186 static bool fib_good_nh(const struct fib_nh *nh)
2187 {
2188 	int state = NUD_REACHABLE;
2189 
2190 	if (nh->fib_nh_scope == RT_SCOPE_LINK) {
2191 		struct neighbour *n;
2192 
2193 		rcu_read_lock_bh();
2194 
2195 		if (likely(nh->fib_nh_gw_family == AF_INET))
2196 			n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
2197 						   (__force u32)nh->fib_nh_gw4);
2198 		else if (nh->fib_nh_gw_family == AF_INET6)
2199 			n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev,
2200 							   &nh->fib_nh_gw6);
2201 		else
2202 			n = NULL;
2203 		if (n)
2204 			state = n->nud_state;
2205 
2206 		rcu_read_unlock_bh();
2207 	}
2208 
2209 	return !!(state & NUD_VALID);
2210 }
2211 
fib_select_multipath(struct fib_result * res,int hash)2212 void fib_select_multipath(struct fib_result *res, int hash)
2213 {
2214 	struct fib_info *fi = res->fi;
2215 	struct net *net = fi->fib_net;
2216 	bool first = false;
2217 
2218 	if (unlikely(res->fi->nh)) {
2219 		nexthop_path_fib_result(res, hash);
2220 		return;
2221 	}
2222 
2223 	change_nexthops(fi) {
2224 		if (READ_ONCE(net->ipv4.sysctl_fib_multipath_use_neigh)) {
2225 			if (!fib_good_nh(nexthop_nh))
2226 				continue;
2227 			if (!first) {
2228 				res->nh_sel = nhsel;
2229 				res->nhc = &nexthop_nh->nh_common;
2230 				first = true;
2231 			}
2232 		}
2233 
2234 		if (hash > atomic_read(&nexthop_nh->fib_nh_upper_bound))
2235 			continue;
2236 
2237 		res->nh_sel = nhsel;
2238 		res->nhc = &nexthop_nh->nh_common;
2239 		return;
2240 	} endfor_nexthops(fi);
2241 }
2242 #endif
2243 
fib_select_path(struct net * net,struct fib_result * res,struct flowi4 * fl4,const struct sk_buff * skb)2244 void fib_select_path(struct net *net, struct fib_result *res,
2245 		     struct flowi4 *fl4, const struct sk_buff *skb)
2246 {
2247 	if (fl4->flowi4_oif && !(fl4->flowi4_flags & FLOWI_FLAG_SKIP_NH_OIF))
2248 		goto check_saddr;
2249 
2250 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2251 	if (fib_info_num_path(res->fi) > 1) {
2252 		int h = fib_multipath_hash(net, fl4, skb, NULL);
2253 
2254 		fib_select_multipath(res, h);
2255 	}
2256 	else
2257 #endif
2258 	if (!res->prefixlen &&
2259 	    res->table->tb_num_default > 1 &&
2260 	    res->type == RTN_UNICAST)
2261 		fib_select_default(fl4, res);
2262 
2263 check_saddr:
2264 	if (!fl4->saddr)
2265 		fl4->saddr = fib_result_prefsrc(net, res);
2266 }
2267