xref: /openbmc/linux/net/mpls/af_mpls.c (revision cb1aaebe)
1 #include <linux/types.h>
2 #include <linux/skbuff.h>
3 #include <linux/socket.h>
4 #include <linux/sysctl.h>
5 #include <linux/net.h>
6 #include <linux/module.h>
7 #include <linux/if_arp.h>
8 #include <linux/ipv6.h>
9 #include <linux/mpls.h>
10 #include <linux/netconf.h>
11 #include <linux/nospec.h>
12 #include <linux/vmalloc.h>
13 #include <linux/percpu.h>
14 #include <net/ip.h>
15 #include <net/dst.h>
16 #include <net/sock.h>
17 #include <net/arp.h>
18 #include <net/ip_fib.h>
19 #include <net/netevent.h>
20 #include <net/ip_tunnels.h>
21 #include <net/netns/generic.h>
22 #if IS_ENABLED(CONFIG_IPV6)
23 #include <net/ipv6.h>
24 #endif
25 #include <net/ipv6_stubs.h>
26 #include <net/rtnh.h>
27 #include "internal.h"
28 
29 /* max memory we will use for mpls_route */
30 #define MAX_MPLS_ROUTE_MEM	4096
31 
32 /* Maximum number of labels to look ahead at when selecting a path of
33  * a multipath route
34  */
35 #define MAX_MP_SELECT_LABELS 4
36 
37 #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1)
38 
39 static int zero = 0;
40 static int one = 1;
41 static int label_limit = (1 << 20) - 1;
42 static int ttl_max = 255;
43 
44 #if IS_ENABLED(CONFIG_NET_IP_TUNNEL)
45 static size_t ipgre_mpls_encap_hlen(struct ip_tunnel_encap *e)
46 {
47 	return sizeof(struct mpls_shim_hdr);
48 }
49 
50 static const struct ip_tunnel_encap_ops mpls_iptun_ops = {
51 	.encap_hlen	= ipgre_mpls_encap_hlen,
52 };
53 
54 static int ipgre_tunnel_encap_add_mpls_ops(void)
55 {
56 	return ip_tunnel_encap_add_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
57 }
58 
59 static void ipgre_tunnel_encap_del_mpls_ops(void)
60 {
61 	ip_tunnel_encap_del_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
62 }
63 #else
64 static int ipgre_tunnel_encap_add_mpls_ops(void)
65 {
66 	return 0;
67 }
68 
69 static void ipgre_tunnel_encap_del_mpls_ops(void)
70 {
71 }
72 #endif
73 
74 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
75 		       struct nlmsghdr *nlh, struct net *net, u32 portid,
76 		       unsigned int nlm_flags);
77 
78 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
79 {
80 	struct mpls_route *rt = NULL;
81 
82 	if (index < net->mpls.platform_labels) {
83 		struct mpls_route __rcu **platform_label =
84 			rcu_dereference(net->mpls.platform_label);
85 		rt = rcu_dereference(platform_label[index]);
86 	}
87 	return rt;
88 }
89 
90 bool mpls_output_possible(const struct net_device *dev)
91 {
92 	return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);
93 }
94 EXPORT_SYMBOL_GPL(mpls_output_possible);
95 
96 static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh)
97 {
98 	return (u8 *)nh + rt->rt_via_offset;
99 }
100 
101 static const u8 *mpls_nh_via(const struct mpls_route *rt,
102 			     const struct mpls_nh *nh)
103 {
104 	return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh);
105 }
106 
107 static unsigned int mpls_nh_header_size(const struct mpls_nh *nh)
108 {
109 	/* The size of the layer 2.5 labels to be added for this route */
110 	return nh->nh_labels * sizeof(struct mpls_shim_hdr);
111 }
112 
113 unsigned int mpls_dev_mtu(const struct net_device *dev)
114 {
115 	/* The amount of data the layer 2 frame can hold */
116 	return dev->mtu;
117 }
118 EXPORT_SYMBOL_GPL(mpls_dev_mtu);
119 
120 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
121 {
122 	if (skb->len <= mtu)
123 		return false;
124 
125 	if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
126 		return false;
127 
128 	return true;
129 }
130 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
131 
132 void mpls_stats_inc_outucastpkts(struct net_device *dev,
133 				 const struct sk_buff *skb)
134 {
135 	struct mpls_dev *mdev;
136 
137 	if (skb->protocol == htons(ETH_P_MPLS_UC)) {
138 		mdev = mpls_dev_get(dev);
139 		if (mdev)
140 			MPLS_INC_STATS_LEN(mdev, skb->len,
141 					   tx_packets,
142 					   tx_bytes);
143 	} else if (skb->protocol == htons(ETH_P_IP)) {
144 		IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
145 #if IS_ENABLED(CONFIG_IPV6)
146 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
147 		struct inet6_dev *in6dev = __in6_dev_get(dev);
148 
149 		if (in6dev)
150 			IP6_UPD_PO_STATS(dev_net(dev), in6dev,
151 					 IPSTATS_MIB_OUT, skb->len);
152 #endif
153 	}
154 }
155 EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
156 
157 static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb)
158 {
159 	struct mpls_entry_decoded dec;
160 	unsigned int mpls_hdr_len = 0;
161 	struct mpls_shim_hdr *hdr;
162 	bool eli_seen = false;
163 	int label_index;
164 	u32 hash = 0;
165 
166 	for (label_index = 0; label_index < MAX_MP_SELECT_LABELS;
167 	     label_index++) {
168 		mpls_hdr_len += sizeof(*hdr);
169 		if (!pskb_may_pull(skb, mpls_hdr_len))
170 			break;
171 
172 		/* Read and decode the current label */
173 		hdr = mpls_hdr(skb) + label_index;
174 		dec = mpls_entry_decode(hdr);
175 
176 		/* RFC6790 - reserved labels MUST NOT be used as keys
177 		 * for the load-balancing function
178 		 */
179 		if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) {
180 			hash = jhash_1word(dec.label, hash);
181 
182 			/* The entropy label follows the entropy label
183 			 * indicator, so this means that the entropy
184 			 * label was just added to the hash - no need to
185 			 * go any deeper either in the label stack or in the
186 			 * payload
187 			 */
188 			if (eli_seen)
189 				break;
190 		} else if (dec.label == MPLS_LABEL_ENTROPY) {
191 			eli_seen = true;
192 		}
193 
194 		if (!dec.bos)
195 			continue;
196 
197 		/* found bottom label; does skb have room for a header? */
198 		if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) {
199 			const struct iphdr *v4hdr;
200 
201 			v4hdr = (const struct iphdr *)(hdr + 1);
202 			if (v4hdr->version == 4) {
203 				hash = jhash_3words(ntohl(v4hdr->saddr),
204 						    ntohl(v4hdr->daddr),
205 						    v4hdr->protocol, hash);
206 			} else if (v4hdr->version == 6 &&
207 				   pskb_may_pull(skb, mpls_hdr_len +
208 						 sizeof(struct ipv6hdr))) {
209 				const struct ipv6hdr *v6hdr;
210 
211 				v6hdr = (const struct ipv6hdr *)(hdr + 1);
212 				hash = __ipv6_addr_jhash(&v6hdr->saddr, hash);
213 				hash = __ipv6_addr_jhash(&v6hdr->daddr, hash);
214 				hash = jhash_1word(v6hdr->nexthdr, hash);
215 			}
216 		}
217 
218 		break;
219 	}
220 
221 	return hash;
222 }
223 
224 static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index)
225 {
226 	return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size);
227 }
228 
229 /* number of alive nexthops (rt->rt_nhn_alive) and the flags for
230  * a next hop (nh->nh_flags) are modified by netdev event handlers.
231  * Since those fields can change at any moment, use READ_ONCE to
232  * access both.
233  */
234 static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt,
235 					     struct sk_buff *skb)
236 {
237 	u32 hash = 0;
238 	int nh_index = 0;
239 	int n = 0;
240 	u8 alive;
241 
242 	/* No need to look further into packet if there's only
243 	 * one path
244 	 */
245 	if (rt->rt_nhn == 1)
246 		return rt->rt_nh;
247 
248 	alive = READ_ONCE(rt->rt_nhn_alive);
249 	if (alive == 0)
250 		return NULL;
251 
252 	hash = mpls_multipath_hash(rt, skb);
253 	nh_index = hash % alive;
254 	if (alive == rt->rt_nhn)
255 		goto out;
256 	for_nexthops(rt) {
257 		unsigned int nh_flags = READ_ONCE(nh->nh_flags);
258 
259 		if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
260 			continue;
261 		if (n == nh_index)
262 			return nh;
263 		n++;
264 	} endfor_nexthops(rt);
265 
266 out:
267 	return mpls_get_nexthop(rt, nh_index);
268 }
269 
270 static bool mpls_egress(struct net *net, struct mpls_route *rt,
271 			struct sk_buff *skb, struct mpls_entry_decoded dec)
272 {
273 	enum mpls_payload_type payload_type;
274 	bool success = false;
275 
276 	/* The IPv4 code below accesses through the IPv4 header
277 	 * checksum, which is 12 bytes into the packet.
278 	 * The IPv6 code below accesses through the IPv6 hop limit
279 	 * which is 8 bytes into the packet.
280 	 *
281 	 * For all supported cases there should always be at least 12
282 	 * bytes of packet data present.  The IPv4 header is 20 bytes
283 	 * without options and the IPv6 header is always 40 bytes
284 	 * long.
285 	 */
286 	if (!pskb_may_pull(skb, 12))
287 		return false;
288 
289 	payload_type = rt->rt_payload_type;
290 	if (payload_type == MPT_UNSPEC)
291 		payload_type = ip_hdr(skb)->version;
292 
293 	switch (payload_type) {
294 	case MPT_IPV4: {
295 		struct iphdr *hdr4 = ip_hdr(skb);
296 		u8 new_ttl;
297 		skb->protocol = htons(ETH_P_IP);
298 
299 		/* If propagating TTL, take the decremented TTL from
300 		 * the incoming MPLS header, otherwise decrement the
301 		 * TTL, but only if not 0 to avoid underflow.
302 		 */
303 		if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
304 		    (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
305 		     net->mpls.ip_ttl_propagate))
306 			new_ttl = dec.ttl;
307 		else
308 			new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0;
309 
310 		csum_replace2(&hdr4->check,
311 			      htons(hdr4->ttl << 8),
312 			      htons(new_ttl << 8));
313 		hdr4->ttl = new_ttl;
314 		success = true;
315 		break;
316 	}
317 	case MPT_IPV6: {
318 		struct ipv6hdr *hdr6 = ipv6_hdr(skb);
319 		skb->protocol = htons(ETH_P_IPV6);
320 
321 		/* If propagating TTL, take the decremented TTL from
322 		 * the incoming MPLS header, otherwise decrement the
323 		 * hop limit, but only if not 0 to avoid underflow.
324 		 */
325 		if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
326 		    (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
327 		     net->mpls.ip_ttl_propagate))
328 			hdr6->hop_limit = dec.ttl;
329 		else if (hdr6->hop_limit)
330 			hdr6->hop_limit = hdr6->hop_limit - 1;
331 		success = true;
332 		break;
333 	}
334 	case MPT_UNSPEC:
335 		/* Should have decided which protocol it is by now */
336 		break;
337 	}
338 
339 	return success;
340 }
341 
342 static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
343 			struct packet_type *pt, struct net_device *orig_dev)
344 {
345 	struct net *net = dev_net(dev);
346 	struct mpls_shim_hdr *hdr;
347 	struct mpls_route *rt;
348 	struct mpls_nh *nh;
349 	struct mpls_entry_decoded dec;
350 	struct net_device *out_dev;
351 	struct mpls_dev *out_mdev;
352 	struct mpls_dev *mdev;
353 	unsigned int hh_len;
354 	unsigned int new_header_size;
355 	unsigned int mtu;
356 	int err;
357 
358 	/* Careful this entire function runs inside of an rcu critical section */
359 
360 	mdev = mpls_dev_get(dev);
361 	if (!mdev)
362 		goto drop;
363 
364 	MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
365 			   rx_bytes);
366 
367 	if (!mdev->input_enabled) {
368 		MPLS_INC_STATS(mdev, rx_dropped);
369 		goto drop;
370 	}
371 
372 	if (skb->pkt_type != PACKET_HOST)
373 		goto err;
374 
375 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
376 		goto err;
377 
378 	if (!pskb_may_pull(skb, sizeof(*hdr)))
379 		goto err;
380 
381 	/* Read and decode the label */
382 	hdr = mpls_hdr(skb);
383 	dec = mpls_entry_decode(hdr);
384 
385 	rt = mpls_route_input_rcu(net, dec.label);
386 	if (!rt) {
387 		MPLS_INC_STATS(mdev, rx_noroute);
388 		goto drop;
389 	}
390 
391 	nh = mpls_select_multipath(rt, skb);
392 	if (!nh)
393 		goto err;
394 
395 	/* Pop the label */
396 	skb_pull(skb, sizeof(*hdr));
397 	skb_reset_network_header(skb);
398 
399 	skb_orphan(skb);
400 
401 	if (skb_warn_if_lro(skb))
402 		goto err;
403 
404 	skb_forward_csum(skb);
405 
406 	/* Verify ttl is valid */
407 	if (dec.ttl <= 1)
408 		goto err;
409 	dec.ttl -= 1;
410 
411 	/* Find the output device */
412 	out_dev = rcu_dereference(nh->nh_dev);
413 	if (!mpls_output_possible(out_dev))
414 		goto tx_err;
415 
416 	/* Verify the destination can hold the packet */
417 	new_header_size = mpls_nh_header_size(nh);
418 	mtu = mpls_dev_mtu(out_dev);
419 	if (mpls_pkt_too_big(skb, mtu - new_header_size))
420 		goto tx_err;
421 
422 	hh_len = LL_RESERVED_SPACE(out_dev);
423 	if (!out_dev->header_ops)
424 		hh_len = 0;
425 
426 	/* Ensure there is enough space for the headers in the skb */
427 	if (skb_cow(skb, hh_len + new_header_size))
428 		goto tx_err;
429 
430 	skb->dev = out_dev;
431 	skb->protocol = htons(ETH_P_MPLS_UC);
432 
433 	if (unlikely(!new_header_size && dec.bos)) {
434 		/* Penultimate hop popping */
435 		if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
436 			goto err;
437 	} else {
438 		bool bos;
439 		int i;
440 		skb_push(skb, new_header_size);
441 		skb_reset_network_header(skb);
442 		/* Push the new labels */
443 		hdr = mpls_hdr(skb);
444 		bos = dec.bos;
445 		for (i = nh->nh_labels - 1; i >= 0; i--) {
446 			hdr[i] = mpls_entry_encode(nh->nh_label[i],
447 						   dec.ttl, 0, bos);
448 			bos = false;
449 		}
450 	}
451 
452 	mpls_stats_inc_outucastpkts(out_dev, skb);
453 
454 	/* If via wasn't specified then send out using device address */
455 	if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
456 		err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,
457 				 out_dev->dev_addr, skb);
458 	else
459 		err = neigh_xmit(nh->nh_via_table, out_dev,
460 				 mpls_nh_via(rt, nh), skb);
461 	if (err)
462 		net_dbg_ratelimited("%s: packet transmission failed: %d\n",
463 				    __func__, err);
464 	return 0;
465 
466 tx_err:
467 	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
468 	if (out_mdev)
469 		MPLS_INC_STATS(out_mdev, tx_errors);
470 	goto drop;
471 err:
472 	MPLS_INC_STATS(mdev, rx_errors);
473 drop:
474 	kfree_skb(skb);
475 	return NET_RX_DROP;
476 }
477 
478 static struct packet_type mpls_packet_type __read_mostly = {
479 	.type = cpu_to_be16(ETH_P_MPLS_UC),
480 	.func = mpls_forward,
481 };
482 
483 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = {
484 	[RTA_DST]		= { .type = NLA_U32 },
485 	[RTA_OIF]		= { .type = NLA_U32 },
486 	[RTA_TTL_PROPAGATE]	= { .type = NLA_U8 },
487 };
488 
489 struct mpls_route_config {
490 	u32			rc_protocol;
491 	u32			rc_ifindex;
492 	u8			rc_via_table;
493 	u8			rc_via_alen;
494 	u8			rc_via[MAX_VIA_ALEN];
495 	u32			rc_label;
496 	u8			rc_ttl_propagate;
497 	u8			rc_output_labels;
498 	u32			rc_output_label[MAX_NEW_LABELS];
499 	u32			rc_nlflags;
500 	enum mpls_payload_type	rc_payload_type;
501 	struct nl_info		rc_nlinfo;
502 	struct rtnexthop	*rc_mp;
503 	int			rc_mp_len;
504 };
505 
506 /* all nexthops within a route have the same size based on max
507  * number of labels and max via length for a hop
508  */
509 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
510 {
511 	u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen);
512 	struct mpls_route *rt;
513 	size_t size;
514 
515 	size = sizeof(*rt) + num_nh * nh_size;
516 	if (size > MAX_MPLS_ROUTE_MEM)
517 		return ERR_PTR(-EINVAL);
518 
519 	rt = kzalloc(size, GFP_KERNEL);
520 	if (!rt)
521 		return ERR_PTR(-ENOMEM);
522 
523 	rt->rt_nhn = num_nh;
524 	rt->rt_nhn_alive = num_nh;
525 	rt->rt_nh_size = nh_size;
526 	rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels);
527 
528 	return rt;
529 }
530 
531 static void mpls_rt_free(struct mpls_route *rt)
532 {
533 	if (rt)
534 		kfree_rcu(rt, rt_rcu);
535 }
536 
537 static void mpls_notify_route(struct net *net, unsigned index,
538 			      struct mpls_route *old, struct mpls_route *new,
539 			      const struct nl_info *info)
540 {
541 	struct nlmsghdr *nlh = info ? info->nlh : NULL;
542 	unsigned portid = info ? info->portid : 0;
543 	int event = new ? RTM_NEWROUTE : RTM_DELROUTE;
544 	struct mpls_route *rt = new ? new : old;
545 	unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0;
546 	/* Ignore reserved labels for now */
547 	if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED))
548 		rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags);
549 }
550 
551 static void mpls_route_update(struct net *net, unsigned index,
552 			      struct mpls_route *new,
553 			      const struct nl_info *info)
554 {
555 	struct mpls_route __rcu **platform_label;
556 	struct mpls_route *rt;
557 
558 	ASSERT_RTNL();
559 
560 	platform_label = rtnl_dereference(net->mpls.platform_label);
561 	rt = rtnl_dereference(platform_label[index]);
562 	rcu_assign_pointer(platform_label[index], new);
563 
564 	mpls_notify_route(net, index, rt, new, info);
565 
566 	/* If we removed a route free it now */
567 	mpls_rt_free(rt);
568 }
569 
570 static unsigned find_free_label(struct net *net)
571 {
572 	struct mpls_route __rcu **platform_label;
573 	size_t platform_labels;
574 	unsigned index;
575 
576 	platform_label = rtnl_dereference(net->mpls.platform_label);
577 	platform_labels = net->mpls.platform_labels;
578 	for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
579 	     index++) {
580 		if (!rtnl_dereference(platform_label[index]))
581 			return index;
582 	}
583 	return LABEL_NOT_SPECIFIED;
584 }
585 
586 #if IS_ENABLED(CONFIG_INET)
587 static struct net_device *inet_fib_lookup_dev(struct net *net,
588 					      const void *addr)
589 {
590 	struct net_device *dev;
591 	struct rtable *rt;
592 	struct in_addr daddr;
593 
594 	memcpy(&daddr, addr, sizeof(struct in_addr));
595 	rt = ip_route_output(net, daddr.s_addr, 0, 0, 0);
596 	if (IS_ERR(rt))
597 		return ERR_CAST(rt);
598 
599 	dev = rt->dst.dev;
600 	dev_hold(dev);
601 
602 	ip_rt_put(rt);
603 
604 	return dev;
605 }
606 #else
607 static struct net_device *inet_fib_lookup_dev(struct net *net,
608 					      const void *addr)
609 {
610 	return ERR_PTR(-EAFNOSUPPORT);
611 }
612 #endif
613 
614 #if IS_ENABLED(CONFIG_IPV6)
615 static struct net_device *inet6_fib_lookup_dev(struct net *net,
616 					       const void *addr)
617 {
618 	struct net_device *dev;
619 	struct dst_entry *dst;
620 	struct flowi6 fl6;
621 	int err;
622 
623 	if (!ipv6_stub)
624 		return ERR_PTR(-EAFNOSUPPORT);
625 
626 	memset(&fl6, 0, sizeof(fl6));
627 	memcpy(&fl6.daddr, addr, sizeof(struct in6_addr));
628 	err = ipv6_stub->ipv6_dst_lookup(net, NULL, &dst, &fl6);
629 	if (err)
630 		return ERR_PTR(err);
631 
632 	dev = dst->dev;
633 	dev_hold(dev);
634 	dst_release(dst);
635 
636 	return dev;
637 }
638 #else
639 static struct net_device *inet6_fib_lookup_dev(struct net *net,
640 					       const void *addr)
641 {
642 	return ERR_PTR(-EAFNOSUPPORT);
643 }
644 #endif
645 
646 static struct net_device *find_outdev(struct net *net,
647 				      struct mpls_route *rt,
648 				      struct mpls_nh *nh, int oif)
649 {
650 	struct net_device *dev = NULL;
651 
652 	if (!oif) {
653 		switch (nh->nh_via_table) {
654 		case NEIGH_ARP_TABLE:
655 			dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh));
656 			break;
657 		case NEIGH_ND_TABLE:
658 			dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh));
659 			break;
660 		case NEIGH_LINK_TABLE:
661 			break;
662 		}
663 	} else {
664 		dev = dev_get_by_index(net, oif);
665 	}
666 
667 	if (!dev)
668 		return ERR_PTR(-ENODEV);
669 
670 	if (IS_ERR(dev))
671 		return dev;
672 
673 	/* The caller is holding rtnl anyways, so release the dev reference */
674 	dev_put(dev);
675 
676 	return dev;
677 }
678 
679 static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
680 			      struct mpls_nh *nh, int oif)
681 {
682 	struct net_device *dev = NULL;
683 	int err = -ENODEV;
684 
685 	dev = find_outdev(net, rt, nh, oif);
686 	if (IS_ERR(dev)) {
687 		err = PTR_ERR(dev);
688 		dev = NULL;
689 		goto errout;
690 	}
691 
692 	/* Ensure this is a supported device */
693 	err = -EINVAL;
694 	if (!mpls_dev_get(dev))
695 		goto errout;
696 
697 	if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
698 	    (dev->addr_len != nh->nh_via_alen))
699 		goto errout;
700 
701 	RCU_INIT_POINTER(nh->nh_dev, dev);
702 
703 	if (!(dev->flags & IFF_UP)) {
704 		nh->nh_flags |= RTNH_F_DEAD;
705 	} else {
706 		unsigned int flags;
707 
708 		flags = dev_get_flags(dev);
709 		if (!(flags & (IFF_RUNNING | IFF_LOWER_UP)))
710 			nh->nh_flags |= RTNH_F_LINKDOWN;
711 	}
712 
713 	return 0;
714 
715 errout:
716 	return err;
717 }
718 
719 static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table,
720 		       u8 via_addr[], struct netlink_ext_ack *extack)
721 {
722 	struct rtvia *via = nla_data(nla);
723 	int err = -EINVAL;
724 	int alen;
725 
726 	if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
727 		NL_SET_ERR_MSG_ATTR(extack, nla,
728 				    "Invalid attribute length for RTA_VIA");
729 		goto errout;
730 	}
731 	alen = nla_len(nla) -
732 			offsetof(struct rtvia, rtvia_addr);
733 	if (alen > MAX_VIA_ALEN) {
734 		NL_SET_ERR_MSG_ATTR(extack, nla,
735 				    "Invalid address length for RTA_VIA");
736 		goto errout;
737 	}
738 
739 	/* Validate the address family */
740 	switch (via->rtvia_family) {
741 	case AF_PACKET:
742 		*via_table = NEIGH_LINK_TABLE;
743 		break;
744 	case AF_INET:
745 		*via_table = NEIGH_ARP_TABLE;
746 		if (alen != 4)
747 			goto errout;
748 		break;
749 	case AF_INET6:
750 		*via_table = NEIGH_ND_TABLE;
751 		if (alen != 16)
752 			goto errout;
753 		break;
754 	default:
755 		/* Unsupported address family */
756 		goto errout;
757 	}
758 
759 	memcpy(via_addr, via->rtvia_addr, alen);
760 	*via_alen = alen;
761 	err = 0;
762 
763 errout:
764 	return err;
765 }
766 
767 static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg,
768 				  struct mpls_route *rt)
769 {
770 	struct net *net = cfg->rc_nlinfo.nl_net;
771 	struct mpls_nh *nh = rt->rt_nh;
772 	int err;
773 	int i;
774 
775 	if (!nh)
776 		return -ENOMEM;
777 
778 	nh->nh_labels = cfg->rc_output_labels;
779 	for (i = 0; i < nh->nh_labels; i++)
780 		nh->nh_label[i] = cfg->rc_output_label[i];
781 
782 	nh->nh_via_table = cfg->rc_via_table;
783 	memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen);
784 	nh->nh_via_alen = cfg->rc_via_alen;
785 
786 	err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex);
787 	if (err)
788 		goto errout;
789 
790 	if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
791 		rt->rt_nhn_alive--;
792 
793 	return 0;
794 
795 errout:
796 	return err;
797 }
798 
799 static int mpls_nh_build(struct net *net, struct mpls_route *rt,
800 			 struct mpls_nh *nh, int oif, struct nlattr *via,
801 			 struct nlattr *newdst, u8 max_labels,
802 			 struct netlink_ext_ack *extack)
803 {
804 	int err = -ENOMEM;
805 
806 	if (!nh)
807 		goto errout;
808 
809 	if (newdst) {
810 		err = nla_get_labels(newdst, max_labels, &nh->nh_labels,
811 				     nh->nh_label, extack);
812 		if (err)
813 			goto errout;
814 	}
815 
816 	if (via) {
817 		err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table,
818 				  __mpls_nh_via(rt, nh), extack);
819 		if (err)
820 			goto errout;
821 	} else {
822 		nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC;
823 	}
824 
825 	err = mpls_nh_assign_dev(net, rt, nh, oif);
826 	if (err)
827 		goto errout;
828 
829 	return 0;
830 
831 errout:
832 	return err;
833 }
834 
835 static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len,
836 			      u8 cfg_via_alen, u8 *max_via_alen,
837 			      u8 *max_labels)
838 {
839 	int remaining = len;
840 	u8 nhs = 0;
841 
842 	*max_via_alen = 0;
843 	*max_labels = 0;
844 
845 	while (rtnh_ok(rtnh, remaining)) {
846 		struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
847 		int attrlen;
848 		u8 n_labels = 0;
849 
850 		attrlen = rtnh_attrlen(rtnh);
851 		nla = nla_find(attrs, attrlen, RTA_VIA);
852 		if (nla && nla_len(nla) >=
853 		    offsetof(struct rtvia, rtvia_addr)) {
854 			int via_alen = nla_len(nla) -
855 				offsetof(struct rtvia, rtvia_addr);
856 
857 			if (via_alen <= MAX_VIA_ALEN)
858 				*max_via_alen = max_t(u16, *max_via_alen,
859 						      via_alen);
860 		}
861 
862 		nla = nla_find(attrs, attrlen, RTA_NEWDST);
863 		if (nla &&
864 		    nla_get_labels(nla, MAX_NEW_LABELS, &n_labels,
865 				   NULL, NULL) != 0)
866 			return 0;
867 
868 		*max_labels = max_t(u8, *max_labels, n_labels);
869 
870 		/* number of nexthops is tracked by a u8.
871 		 * Check for overflow.
872 		 */
873 		if (nhs == 255)
874 			return 0;
875 		nhs++;
876 
877 		rtnh = rtnh_next(rtnh, &remaining);
878 	}
879 
880 	/* leftover implies invalid nexthop configuration, discard it */
881 	return remaining > 0 ? 0 : nhs;
882 }
883 
884 static int mpls_nh_build_multi(struct mpls_route_config *cfg,
885 			       struct mpls_route *rt, u8 max_labels,
886 			       struct netlink_ext_ack *extack)
887 {
888 	struct rtnexthop *rtnh = cfg->rc_mp;
889 	struct nlattr *nla_via, *nla_newdst;
890 	int remaining = cfg->rc_mp_len;
891 	int err = 0;
892 	u8 nhs = 0;
893 
894 	change_nexthops(rt) {
895 		int attrlen;
896 
897 		nla_via = NULL;
898 		nla_newdst = NULL;
899 
900 		err = -EINVAL;
901 		if (!rtnh_ok(rtnh, remaining))
902 			goto errout;
903 
904 		/* neither weighted multipath nor any flags
905 		 * are supported
906 		 */
907 		if (rtnh->rtnh_hops || rtnh->rtnh_flags)
908 			goto errout;
909 
910 		attrlen = rtnh_attrlen(rtnh);
911 		if (attrlen > 0) {
912 			struct nlattr *attrs = rtnh_attrs(rtnh);
913 
914 			nla_via = nla_find(attrs, attrlen, RTA_VIA);
915 			nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST);
916 		}
917 
918 		err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh,
919 				    rtnh->rtnh_ifindex, nla_via, nla_newdst,
920 				    max_labels, extack);
921 		if (err)
922 			goto errout;
923 
924 		if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
925 			rt->rt_nhn_alive--;
926 
927 		rtnh = rtnh_next(rtnh, &remaining);
928 		nhs++;
929 	} endfor_nexthops(rt);
930 
931 	rt->rt_nhn = nhs;
932 
933 	return 0;
934 
935 errout:
936 	return err;
937 }
938 
939 static bool mpls_label_ok(struct net *net, unsigned int *index,
940 			  struct netlink_ext_ack *extack)
941 {
942 	bool is_ok = true;
943 
944 	/* Reserved labels may not be set */
945 	if (*index < MPLS_LABEL_FIRST_UNRESERVED) {
946 		NL_SET_ERR_MSG(extack,
947 			       "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher");
948 		is_ok = false;
949 	}
950 
951 	/* The full 20 bit range may not be supported. */
952 	if (is_ok && *index >= net->mpls.platform_labels) {
953 		NL_SET_ERR_MSG(extack,
954 			       "Label >= configured maximum in platform_labels");
955 		is_ok = false;
956 	}
957 
958 	*index = array_index_nospec(*index, net->mpls.platform_labels);
959 	return is_ok;
960 }
961 
962 static int mpls_route_add(struct mpls_route_config *cfg,
963 			  struct netlink_ext_ack *extack)
964 {
965 	struct mpls_route __rcu **platform_label;
966 	struct net *net = cfg->rc_nlinfo.nl_net;
967 	struct mpls_route *rt, *old;
968 	int err = -EINVAL;
969 	u8 max_via_alen;
970 	unsigned index;
971 	u8 max_labels;
972 	u8 nhs;
973 
974 	index = cfg->rc_label;
975 
976 	/* If a label was not specified during insert pick one */
977 	if ((index == LABEL_NOT_SPECIFIED) &&
978 	    (cfg->rc_nlflags & NLM_F_CREATE)) {
979 		index = find_free_label(net);
980 	}
981 
982 	if (!mpls_label_ok(net, &index, extack))
983 		goto errout;
984 
985 	/* Append makes no sense with mpls */
986 	err = -EOPNOTSUPP;
987 	if (cfg->rc_nlflags & NLM_F_APPEND) {
988 		NL_SET_ERR_MSG(extack, "MPLS does not support route append");
989 		goto errout;
990 	}
991 
992 	err = -EEXIST;
993 	platform_label = rtnl_dereference(net->mpls.platform_label);
994 	old = rtnl_dereference(platform_label[index]);
995 	if ((cfg->rc_nlflags & NLM_F_EXCL) && old)
996 		goto errout;
997 
998 	err = -EEXIST;
999 	if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old)
1000 		goto errout;
1001 
1002 	err = -ENOENT;
1003 	if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old)
1004 		goto errout;
1005 
1006 	err = -EINVAL;
1007 	if (cfg->rc_mp) {
1008 		nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len,
1009 					  cfg->rc_via_alen, &max_via_alen,
1010 					  &max_labels);
1011 	} else {
1012 		max_via_alen = cfg->rc_via_alen;
1013 		max_labels = cfg->rc_output_labels;
1014 		nhs = 1;
1015 	}
1016 
1017 	if (nhs == 0) {
1018 		NL_SET_ERR_MSG(extack, "Route does not contain a nexthop");
1019 		goto errout;
1020 	}
1021 
1022 	err = -ENOMEM;
1023 	rt = mpls_rt_alloc(nhs, max_via_alen, max_labels);
1024 	if (IS_ERR(rt)) {
1025 		err = PTR_ERR(rt);
1026 		goto errout;
1027 	}
1028 
1029 	rt->rt_protocol = cfg->rc_protocol;
1030 	rt->rt_payload_type = cfg->rc_payload_type;
1031 	rt->rt_ttl_propagate = cfg->rc_ttl_propagate;
1032 
1033 	if (cfg->rc_mp)
1034 		err = mpls_nh_build_multi(cfg, rt, max_labels, extack);
1035 	else
1036 		err = mpls_nh_build_from_cfg(cfg, rt);
1037 	if (err)
1038 		goto freert;
1039 
1040 	mpls_route_update(net, index, rt, &cfg->rc_nlinfo);
1041 
1042 	return 0;
1043 
1044 freert:
1045 	mpls_rt_free(rt);
1046 errout:
1047 	return err;
1048 }
1049 
1050 static int mpls_route_del(struct mpls_route_config *cfg,
1051 			  struct netlink_ext_ack *extack)
1052 {
1053 	struct net *net = cfg->rc_nlinfo.nl_net;
1054 	unsigned index;
1055 	int err = -EINVAL;
1056 
1057 	index = cfg->rc_label;
1058 
1059 	if (!mpls_label_ok(net, &index, extack))
1060 		goto errout;
1061 
1062 	mpls_route_update(net, index, NULL, &cfg->rc_nlinfo);
1063 
1064 	err = 0;
1065 errout:
1066 	return err;
1067 }
1068 
1069 static void mpls_get_stats(struct mpls_dev *mdev,
1070 			   struct mpls_link_stats *stats)
1071 {
1072 	struct mpls_pcpu_stats *p;
1073 	int i;
1074 
1075 	memset(stats, 0, sizeof(*stats));
1076 
1077 	for_each_possible_cpu(i) {
1078 		struct mpls_link_stats local;
1079 		unsigned int start;
1080 
1081 		p = per_cpu_ptr(mdev->stats, i);
1082 		do {
1083 			start = u64_stats_fetch_begin(&p->syncp);
1084 			local = p->stats;
1085 		} while (u64_stats_fetch_retry(&p->syncp, start));
1086 
1087 		stats->rx_packets	+= local.rx_packets;
1088 		stats->rx_bytes		+= local.rx_bytes;
1089 		stats->tx_packets	+= local.tx_packets;
1090 		stats->tx_bytes		+= local.tx_bytes;
1091 		stats->rx_errors	+= local.rx_errors;
1092 		stats->tx_errors	+= local.tx_errors;
1093 		stats->rx_dropped	+= local.rx_dropped;
1094 		stats->tx_dropped	+= local.tx_dropped;
1095 		stats->rx_noroute	+= local.rx_noroute;
1096 	}
1097 }
1098 
1099 static int mpls_fill_stats_af(struct sk_buff *skb,
1100 			      const struct net_device *dev)
1101 {
1102 	struct mpls_link_stats *stats;
1103 	struct mpls_dev *mdev;
1104 	struct nlattr *nla;
1105 
1106 	mdev = mpls_dev_get(dev);
1107 	if (!mdev)
1108 		return -ENODATA;
1109 
1110 	nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
1111 				sizeof(struct mpls_link_stats),
1112 				MPLS_STATS_UNSPEC);
1113 	if (!nla)
1114 		return -EMSGSIZE;
1115 
1116 	stats = nla_data(nla);
1117 	mpls_get_stats(mdev, stats);
1118 
1119 	return 0;
1120 }
1121 
1122 static size_t mpls_get_stats_af_size(const struct net_device *dev)
1123 {
1124 	struct mpls_dev *mdev;
1125 
1126 	mdev = mpls_dev_get(dev);
1127 	if (!mdev)
1128 		return 0;
1129 
1130 	return nla_total_size_64bit(sizeof(struct mpls_link_stats));
1131 }
1132 
1133 static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev,
1134 				     u32 portid, u32 seq, int event,
1135 				     unsigned int flags, int type)
1136 {
1137 	struct nlmsghdr  *nlh;
1138 	struct netconfmsg *ncm;
1139 	bool all = false;
1140 
1141 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1142 			flags);
1143 	if (!nlh)
1144 		return -EMSGSIZE;
1145 
1146 	if (type == NETCONFA_ALL)
1147 		all = true;
1148 
1149 	ncm = nlmsg_data(nlh);
1150 	ncm->ncm_family = AF_MPLS;
1151 
1152 	if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0)
1153 		goto nla_put_failure;
1154 
1155 	if ((all || type == NETCONFA_INPUT) &&
1156 	    nla_put_s32(skb, NETCONFA_INPUT,
1157 			mdev->input_enabled) < 0)
1158 		goto nla_put_failure;
1159 
1160 	nlmsg_end(skb, nlh);
1161 	return 0;
1162 
1163 nla_put_failure:
1164 	nlmsg_cancel(skb, nlh);
1165 	return -EMSGSIZE;
1166 }
1167 
1168 static int mpls_netconf_msgsize_devconf(int type)
1169 {
1170 	int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1171 			+ nla_total_size(4); /* NETCONFA_IFINDEX */
1172 	bool all = false;
1173 
1174 	if (type == NETCONFA_ALL)
1175 		all = true;
1176 
1177 	if (all || type == NETCONFA_INPUT)
1178 		size += nla_total_size(4);
1179 
1180 	return size;
1181 }
1182 
1183 static void mpls_netconf_notify_devconf(struct net *net, int event,
1184 					int type, struct mpls_dev *mdev)
1185 {
1186 	struct sk_buff *skb;
1187 	int err = -ENOBUFS;
1188 
1189 	skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL);
1190 	if (!skb)
1191 		goto errout;
1192 
1193 	err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type);
1194 	if (err < 0) {
1195 		/* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1196 		WARN_ON(err == -EMSGSIZE);
1197 		kfree_skb(skb);
1198 		goto errout;
1199 	}
1200 
1201 	rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL);
1202 	return;
1203 errout:
1204 	if (err < 0)
1205 		rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err);
1206 }
1207 
1208 static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = {
1209 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
1210 };
1211 
1212 static int mpls_netconf_valid_get_req(struct sk_buff *skb,
1213 				      const struct nlmsghdr *nlh,
1214 				      struct nlattr **tb,
1215 				      struct netlink_ext_ack *extack)
1216 {
1217 	int i, err;
1218 
1219 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
1220 		NL_SET_ERR_MSG_MOD(extack,
1221 				   "Invalid header for netconf get request");
1222 		return -EINVAL;
1223 	}
1224 
1225 	if (!netlink_strict_get_check(skb))
1226 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
1227 					      tb, NETCONFA_MAX,
1228 					      devconf_mpls_policy, extack);
1229 
1230 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
1231 					    tb, NETCONFA_MAX,
1232 					    devconf_mpls_policy, extack);
1233 	if (err)
1234 		return err;
1235 
1236 	for (i = 0; i <= NETCONFA_MAX; i++) {
1237 		if (!tb[i])
1238 			continue;
1239 
1240 		switch (i) {
1241 		case NETCONFA_IFINDEX:
1242 			break;
1243 		default:
1244 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
1245 			return -EINVAL;
1246 		}
1247 	}
1248 
1249 	return 0;
1250 }
1251 
1252 static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
1253 				    struct nlmsghdr *nlh,
1254 				    struct netlink_ext_ack *extack)
1255 {
1256 	struct net *net = sock_net(in_skb->sk);
1257 	struct nlattr *tb[NETCONFA_MAX + 1];
1258 	struct net_device *dev;
1259 	struct mpls_dev *mdev;
1260 	struct sk_buff *skb;
1261 	int ifindex;
1262 	int err;
1263 
1264 	err = mpls_netconf_valid_get_req(in_skb, nlh, tb, extack);
1265 	if (err < 0)
1266 		goto errout;
1267 
1268 	err = -EINVAL;
1269 	if (!tb[NETCONFA_IFINDEX])
1270 		goto errout;
1271 
1272 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1273 	dev = __dev_get_by_index(net, ifindex);
1274 	if (!dev)
1275 		goto errout;
1276 
1277 	mdev = mpls_dev_get(dev);
1278 	if (!mdev)
1279 		goto errout;
1280 
1281 	err = -ENOBUFS;
1282 	skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1283 	if (!skb)
1284 		goto errout;
1285 
1286 	err = mpls_netconf_fill_devconf(skb, mdev,
1287 					NETLINK_CB(in_skb).portid,
1288 					nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1289 					NETCONFA_ALL);
1290 	if (err < 0) {
1291 		/* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1292 		WARN_ON(err == -EMSGSIZE);
1293 		kfree_skb(skb);
1294 		goto errout;
1295 	}
1296 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1297 errout:
1298 	return err;
1299 }
1300 
1301 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
1302 				     struct netlink_callback *cb)
1303 {
1304 	const struct nlmsghdr *nlh = cb->nlh;
1305 	struct net *net = sock_net(skb->sk);
1306 	struct hlist_head *head;
1307 	struct net_device *dev;
1308 	struct mpls_dev *mdev;
1309 	int idx, s_idx;
1310 	int h, s_h;
1311 
1312 	if (cb->strict_check) {
1313 		struct netlink_ext_ack *extack = cb->extack;
1314 		struct netconfmsg *ncm;
1315 
1316 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
1317 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
1318 			return -EINVAL;
1319 		}
1320 
1321 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
1322 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
1323 			return -EINVAL;
1324 		}
1325 	}
1326 
1327 	s_h = cb->args[0];
1328 	s_idx = idx = cb->args[1];
1329 
1330 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1331 		idx = 0;
1332 		head = &net->dev_index_head[h];
1333 		rcu_read_lock();
1334 		cb->seq = net->dev_base_seq;
1335 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1336 			if (idx < s_idx)
1337 				goto cont;
1338 			mdev = mpls_dev_get(dev);
1339 			if (!mdev)
1340 				goto cont;
1341 			if (mpls_netconf_fill_devconf(skb, mdev,
1342 						      NETLINK_CB(cb->skb).portid,
1343 						      nlh->nlmsg_seq,
1344 						      RTM_NEWNETCONF,
1345 						      NLM_F_MULTI,
1346 						      NETCONFA_ALL) < 0) {
1347 				rcu_read_unlock();
1348 				goto done;
1349 			}
1350 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1351 cont:
1352 			idx++;
1353 		}
1354 		rcu_read_unlock();
1355 	}
1356 done:
1357 	cb->args[0] = h;
1358 	cb->args[1] = idx;
1359 
1360 	return skb->len;
1361 }
1362 
1363 #define MPLS_PERDEV_SYSCTL_OFFSET(field)	\
1364 	(&((struct mpls_dev *)0)->field)
1365 
1366 static int mpls_conf_proc(struct ctl_table *ctl, int write,
1367 			  void __user *buffer,
1368 			  size_t *lenp, loff_t *ppos)
1369 {
1370 	int oval = *(int *)ctl->data;
1371 	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1372 
1373 	if (write) {
1374 		struct mpls_dev *mdev = ctl->extra1;
1375 		int i = (int *)ctl->data - (int *)mdev;
1376 		struct net *net = ctl->extra2;
1377 		int val = *(int *)ctl->data;
1378 
1379 		if (i == offsetof(struct mpls_dev, input_enabled) &&
1380 		    val != oval) {
1381 			mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1382 						    NETCONFA_INPUT, mdev);
1383 		}
1384 	}
1385 
1386 	return ret;
1387 }
1388 
1389 static const struct ctl_table mpls_dev_table[] = {
1390 	{
1391 		.procname	= "input",
1392 		.maxlen		= sizeof(int),
1393 		.mode		= 0644,
1394 		.proc_handler	= mpls_conf_proc,
1395 		.data		= MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1396 	},
1397 	{ }
1398 };
1399 
1400 static int mpls_dev_sysctl_register(struct net_device *dev,
1401 				    struct mpls_dev *mdev)
1402 {
1403 	char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1404 	struct net *net = dev_net(dev);
1405 	struct ctl_table *table;
1406 	int i;
1407 
1408 	table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1409 	if (!table)
1410 		goto out;
1411 
1412 	/* Table data contains only offsets relative to the base of
1413 	 * the mdev at this point, so make them absolute.
1414 	 */
1415 	for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) {
1416 		table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1417 		table[i].extra1 = mdev;
1418 		table[i].extra2 = net;
1419 	}
1420 
1421 	snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1422 
1423 	mdev->sysctl = register_net_sysctl(net, path, table);
1424 	if (!mdev->sysctl)
1425 		goto free;
1426 
1427 	mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1428 	return 0;
1429 
1430 free:
1431 	kfree(table);
1432 out:
1433 	return -ENOBUFS;
1434 }
1435 
1436 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1437 				       struct mpls_dev *mdev)
1438 {
1439 	struct net *net = dev_net(dev);
1440 	struct ctl_table *table;
1441 
1442 	table = mdev->sysctl->ctl_table_arg;
1443 	unregister_net_sysctl_table(mdev->sysctl);
1444 	kfree(table);
1445 
1446 	mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1447 }
1448 
1449 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1450 {
1451 	struct mpls_dev *mdev;
1452 	int err = -ENOMEM;
1453 	int i;
1454 
1455 	ASSERT_RTNL();
1456 
1457 	mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1458 	if (!mdev)
1459 		return ERR_PTR(err);
1460 
1461 	mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1462 	if (!mdev->stats)
1463 		goto free;
1464 
1465 	for_each_possible_cpu(i) {
1466 		struct mpls_pcpu_stats *mpls_stats;
1467 
1468 		mpls_stats = per_cpu_ptr(mdev->stats, i);
1469 		u64_stats_init(&mpls_stats->syncp);
1470 	}
1471 
1472 	mdev->dev = dev;
1473 
1474 	err = mpls_dev_sysctl_register(dev, mdev);
1475 	if (err)
1476 		goto free;
1477 
1478 	rcu_assign_pointer(dev->mpls_ptr, mdev);
1479 
1480 	return mdev;
1481 
1482 free:
1483 	free_percpu(mdev->stats);
1484 	kfree(mdev);
1485 	return ERR_PTR(err);
1486 }
1487 
1488 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1489 {
1490 	struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1491 
1492 	free_percpu(mdev->stats);
1493 	kfree(mdev);
1494 }
1495 
1496 static void mpls_ifdown(struct net_device *dev, int event)
1497 {
1498 	struct mpls_route __rcu **platform_label;
1499 	struct net *net = dev_net(dev);
1500 	u8 alive, deleted;
1501 	unsigned index;
1502 
1503 	platform_label = rtnl_dereference(net->mpls.platform_label);
1504 	for (index = 0; index < net->mpls.platform_labels; index++) {
1505 		struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1506 
1507 		if (!rt)
1508 			continue;
1509 
1510 		alive = 0;
1511 		deleted = 0;
1512 		change_nexthops(rt) {
1513 			unsigned int nh_flags = nh->nh_flags;
1514 
1515 			if (rtnl_dereference(nh->nh_dev) != dev)
1516 				goto next;
1517 
1518 			switch (event) {
1519 			case NETDEV_DOWN:
1520 			case NETDEV_UNREGISTER:
1521 				nh_flags |= RTNH_F_DEAD;
1522 				/* fall through */
1523 			case NETDEV_CHANGE:
1524 				nh_flags |= RTNH_F_LINKDOWN;
1525 				break;
1526 			}
1527 			if (event == NETDEV_UNREGISTER)
1528 				RCU_INIT_POINTER(nh->nh_dev, NULL);
1529 
1530 			if (nh->nh_flags != nh_flags)
1531 				WRITE_ONCE(nh->nh_flags, nh_flags);
1532 next:
1533 			if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1534 				alive++;
1535 			if (!rtnl_dereference(nh->nh_dev))
1536 				deleted++;
1537 		} endfor_nexthops(rt);
1538 
1539 		WRITE_ONCE(rt->rt_nhn_alive, alive);
1540 
1541 		/* if there are no more nexthops, delete the route */
1542 		if (event == NETDEV_UNREGISTER && deleted == rt->rt_nhn)
1543 			mpls_route_update(net, index, NULL, NULL);
1544 	}
1545 }
1546 
1547 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1548 {
1549 	struct mpls_route __rcu **platform_label;
1550 	struct net *net = dev_net(dev);
1551 	unsigned index;
1552 	u8 alive;
1553 
1554 	platform_label = rtnl_dereference(net->mpls.platform_label);
1555 	for (index = 0; index < net->mpls.platform_labels; index++) {
1556 		struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1557 
1558 		if (!rt)
1559 			continue;
1560 
1561 		alive = 0;
1562 		change_nexthops(rt) {
1563 			unsigned int nh_flags = nh->nh_flags;
1564 			struct net_device *nh_dev =
1565 				rtnl_dereference(nh->nh_dev);
1566 
1567 			if (!(nh_flags & flags)) {
1568 				alive++;
1569 				continue;
1570 			}
1571 			if (nh_dev != dev)
1572 				continue;
1573 			alive++;
1574 			nh_flags &= ~flags;
1575 			WRITE_ONCE(nh->nh_flags, nh_flags);
1576 		} endfor_nexthops(rt);
1577 
1578 		WRITE_ONCE(rt->rt_nhn_alive, alive);
1579 	}
1580 }
1581 
1582 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1583 			   void *ptr)
1584 {
1585 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1586 	struct mpls_dev *mdev;
1587 	unsigned int flags;
1588 
1589 	if (event == NETDEV_REGISTER) {
1590 
1591 		/* For now just support Ethernet, IPGRE, IP6GRE, SIT and
1592 		 * IPIP devices
1593 		 */
1594 		if (dev->type == ARPHRD_ETHER ||
1595 		    dev->type == ARPHRD_LOOPBACK ||
1596 		    dev->type == ARPHRD_IPGRE ||
1597 		    dev->type == ARPHRD_IP6GRE ||
1598 		    dev->type == ARPHRD_SIT ||
1599 		    dev->type == ARPHRD_TUNNEL) {
1600 			mdev = mpls_add_dev(dev);
1601 			if (IS_ERR(mdev))
1602 				return notifier_from_errno(PTR_ERR(mdev));
1603 		}
1604 		return NOTIFY_OK;
1605 	}
1606 
1607 	mdev = mpls_dev_get(dev);
1608 	if (!mdev)
1609 		return NOTIFY_OK;
1610 
1611 	switch (event) {
1612 	case NETDEV_DOWN:
1613 		mpls_ifdown(dev, event);
1614 		break;
1615 	case NETDEV_UP:
1616 		flags = dev_get_flags(dev);
1617 		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1618 			mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1619 		else
1620 			mpls_ifup(dev, RTNH_F_DEAD);
1621 		break;
1622 	case NETDEV_CHANGE:
1623 		flags = dev_get_flags(dev);
1624 		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1625 			mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1626 		else
1627 			mpls_ifdown(dev, event);
1628 		break;
1629 	case NETDEV_UNREGISTER:
1630 		mpls_ifdown(dev, event);
1631 		mdev = mpls_dev_get(dev);
1632 		if (mdev) {
1633 			mpls_dev_sysctl_unregister(dev, mdev);
1634 			RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1635 			call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1636 		}
1637 		break;
1638 	case NETDEV_CHANGENAME:
1639 		mdev = mpls_dev_get(dev);
1640 		if (mdev) {
1641 			int err;
1642 
1643 			mpls_dev_sysctl_unregister(dev, mdev);
1644 			err = mpls_dev_sysctl_register(dev, mdev);
1645 			if (err)
1646 				return notifier_from_errno(err);
1647 		}
1648 		break;
1649 	}
1650 	return NOTIFY_OK;
1651 }
1652 
1653 static struct notifier_block mpls_dev_notifier = {
1654 	.notifier_call = mpls_dev_notify,
1655 };
1656 
1657 static int nla_put_via(struct sk_buff *skb,
1658 		       u8 table, const void *addr, int alen)
1659 {
1660 	static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1661 		AF_INET, AF_INET6, AF_DECnet, AF_PACKET,
1662 	};
1663 	struct nlattr *nla;
1664 	struct rtvia *via;
1665 	int family = AF_UNSPEC;
1666 
1667 	nla = nla_reserve(skb, RTA_VIA, alen + 2);
1668 	if (!nla)
1669 		return -EMSGSIZE;
1670 
1671 	if (table <= NEIGH_NR_TABLES)
1672 		family = table_to_family[table];
1673 
1674 	via = nla_data(nla);
1675 	via->rtvia_family = family;
1676 	memcpy(via->rtvia_addr, addr, alen);
1677 	return 0;
1678 }
1679 
1680 int nla_put_labels(struct sk_buff *skb, int attrtype,
1681 		   u8 labels, const u32 label[])
1682 {
1683 	struct nlattr *nla;
1684 	struct mpls_shim_hdr *nla_label;
1685 	bool bos;
1686 	int i;
1687 	nla = nla_reserve(skb, attrtype, labels*4);
1688 	if (!nla)
1689 		return -EMSGSIZE;
1690 
1691 	nla_label = nla_data(nla);
1692 	bos = true;
1693 	for (i = labels - 1; i >= 0; i--) {
1694 		nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1695 		bos = false;
1696 	}
1697 
1698 	return 0;
1699 }
1700 EXPORT_SYMBOL_GPL(nla_put_labels);
1701 
1702 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1703 		   u32 label[], struct netlink_ext_ack *extack)
1704 {
1705 	unsigned len = nla_len(nla);
1706 	struct mpls_shim_hdr *nla_label;
1707 	u8 nla_labels;
1708 	bool bos;
1709 	int i;
1710 
1711 	/* len needs to be an even multiple of 4 (the label size). Number
1712 	 * of labels is a u8 so check for overflow.
1713 	 */
1714 	if (len & 3 || len / 4 > 255) {
1715 		NL_SET_ERR_MSG_ATTR(extack, nla,
1716 				    "Invalid length for labels attribute");
1717 		return -EINVAL;
1718 	}
1719 
1720 	/* Limit the number of new labels allowed */
1721 	nla_labels = len/4;
1722 	if (nla_labels > max_labels) {
1723 		NL_SET_ERR_MSG(extack, "Too many labels");
1724 		return -EINVAL;
1725 	}
1726 
1727 	/* when label == NULL, caller wants number of labels */
1728 	if (!label)
1729 		goto out;
1730 
1731 	nla_label = nla_data(nla);
1732 	bos = true;
1733 	for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1734 		struct mpls_entry_decoded dec;
1735 		dec = mpls_entry_decode(nla_label + i);
1736 
1737 		/* Ensure the bottom of stack flag is properly set
1738 		 * and ttl and tc are both clear.
1739 		 */
1740 		if (dec.ttl) {
1741 			NL_SET_ERR_MSG_ATTR(extack, nla,
1742 					    "TTL in label must be 0");
1743 			return -EINVAL;
1744 		}
1745 
1746 		if (dec.tc) {
1747 			NL_SET_ERR_MSG_ATTR(extack, nla,
1748 					    "Traffic class in label must be 0");
1749 			return -EINVAL;
1750 		}
1751 
1752 		if (dec.bos != bos) {
1753 			NL_SET_BAD_ATTR(extack, nla);
1754 			if (bos) {
1755 				NL_SET_ERR_MSG(extack,
1756 					       "BOS bit must be set in first label");
1757 			} else {
1758 				NL_SET_ERR_MSG(extack,
1759 					       "BOS bit can only be set in first label");
1760 			}
1761 			return -EINVAL;
1762 		}
1763 
1764 		switch (dec.label) {
1765 		case MPLS_LABEL_IMPLNULL:
1766 			/* RFC3032: This is a label that an LSR may
1767 			 * assign and distribute, but which never
1768 			 * actually appears in the encapsulation.
1769 			 */
1770 			NL_SET_ERR_MSG_ATTR(extack, nla,
1771 					    "Implicit NULL Label (3) can not be used in encapsulation");
1772 			return -EINVAL;
1773 		}
1774 
1775 		label[i] = dec.label;
1776 	}
1777 out:
1778 	*labels = nla_labels;
1779 	return 0;
1780 }
1781 EXPORT_SYMBOL_GPL(nla_get_labels);
1782 
1783 static int rtm_to_route_config(struct sk_buff *skb,
1784 			       struct nlmsghdr *nlh,
1785 			       struct mpls_route_config *cfg,
1786 			       struct netlink_ext_ack *extack)
1787 {
1788 	struct rtmsg *rtm;
1789 	struct nlattr *tb[RTA_MAX+1];
1790 	int index;
1791 	int err;
1792 
1793 	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
1794 				     rtm_mpls_policy, extack);
1795 	if (err < 0)
1796 		goto errout;
1797 
1798 	err = -EINVAL;
1799 	rtm = nlmsg_data(nlh);
1800 
1801 	if (rtm->rtm_family != AF_MPLS) {
1802 		NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1803 		goto errout;
1804 	}
1805 	if (rtm->rtm_dst_len != 20) {
1806 		NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1807 		goto errout;
1808 	}
1809 	if (rtm->rtm_src_len != 0) {
1810 		NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1811 		goto errout;
1812 	}
1813 	if (rtm->rtm_tos != 0) {
1814 		NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1815 		goto errout;
1816 	}
1817 	if (rtm->rtm_table != RT_TABLE_MAIN) {
1818 		NL_SET_ERR_MSG(extack,
1819 			       "MPLS only supports the main route table");
1820 		goto errout;
1821 	}
1822 	/* Any value is acceptable for rtm_protocol */
1823 
1824 	/* As mpls uses destination specific addresses
1825 	 * (or source specific address in the case of multicast)
1826 	 * all addresses have universal scope.
1827 	 */
1828 	if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1829 		NL_SET_ERR_MSG(extack,
1830 			       "Invalid route scope  - MPLS only supports UNIVERSE");
1831 		goto errout;
1832 	}
1833 	if (rtm->rtm_type != RTN_UNICAST) {
1834 		NL_SET_ERR_MSG(extack,
1835 			       "Invalid route type - MPLS only supports UNICAST");
1836 		goto errout;
1837 	}
1838 	if (rtm->rtm_flags != 0) {
1839 		NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1840 		goto errout;
1841 	}
1842 
1843 	cfg->rc_label		= LABEL_NOT_SPECIFIED;
1844 	cfg->rc_protocol	= rtm->rtm_protocol;
1845 	cfg->rc_via_table	= MPLS_NEIGH_TABLE_UNSPEC;
1846 	cfg->rc_ttl_propagate	= MPLS_TTL_PROP_DEFAULT;
1847 	cfg->rc_nlflags		= nlh->nlmsg_flags;
1848 	cfg->rc_nlinfo.portid	= NETLINK_CB(skb).portid;
1849 	cfg->rc_nlinfo.nlh	= nlh;
1850 	cfg->rc_nlinfo.nl_net	= sock_net(skb->sk);
1851 
1852 	for (index = 0; index <= RTA_MAX; index++) {
1853 		struct nlattr *nla = tb[index];
1854 		if (!nla)
1855 			continue;
1856 
1857 		switch (index) {
1858 		case RTA_OIF:
1859 			cfg->rc_ifindex = nla_get_u32(nla);
1860 			break;
1861 		case RTA_NEWDST:
1862 			if (nla_get_labels(nla, MAX_NEW_LABELS,
1863 					   &cfg->rc_output_labels,
1864 					   cfg->rc_output_label, extack))
1865 				goto errout;
1866 			break;
1867 		case RTA_DST:
1868 		{
1869 			u8 label_count;
1870 			if (nla_get_labels(nla, 1, &label_count,
1871 					   &cfg->rc_label, extack))
1872 				goto errout;
1873 
1874 			if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1875 					   &cfg->rc_label, extack))
1876 				goto errout;
1877 			break;
1878 		}
1879 		case RTA_GATEWAY:
1880 			NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1881 			goto errout;
1882 		case RTA_VIA:
1883 		{
1884 			if (nla_get_via(nla, &cfg->rc_via_alen,
1885 					&cfg->rc_via_table, cfg->rc_via,
1886 					extack))
1887 				goto errout;
1888 			break;
1889 		}
1890 		case RTA_MULTIPATH:
1891 		{
1892 			cfg->rc_mp = nla_data(nla);
1893 			cfg->rc_mp_len = nla_len(nla);
1894 			break;
1895 		}
1896 		case RTA_TTL_PROPAGATE:
1897 		{
1898 			u8 ttl_propagate = nla_get_u8(nla);
1899 
1900 			if (ttl_propagate > 1) {
1901 				NL_SET_ERR_MSG_ATTR(extack, nla,
1902 						    "RTA_TTL_PROPAGATE can only be 0 or 1");
1903 				goto errout;
1904 			}
1905 			cfg->rc_ttl_propagate = ttl_propagate ?
1906 				MPLS_TTL_PROP_ENABLED :
1907 				MPLS_TTL_PROP_DISABLED;
1908 			break;
1909 		}
1910 		default:
1911 			NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1912 			/* Unsupported attribute */
1913 			goto errout;
1914 		}
1915 	}
1916 
1917 	err = 0;
1918 errout:
1919 	return err;
1920 }
1921 
1922 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1923 			     struct netlink_ext_ack *extack)
1924 {
1925 	struct mpls_route_config *cfg;
1926 	int err;
1927 
1928 	cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1929 	if (!cfg)
1930 		return -ENOMEM;
1931 
1932 	err = rtm_to_route_config(skb, nlh, cfg, extack);
1933 	if (err < 0)
1934 		goto out;
1935 
1936 	err = mpls_route_del(cfg, extack);
1937 out:
1938 	kfree(cfg);
1939 
1940 	return err;
1941 }
1942 
1943 
1944 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1945 			     struct netlink_ext_ack *extack)
1946 {
1947 	struct mpls_route_config *cfg;
1948 	int err;
1949 
1950 	cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1951 	if (!cfg)
1952 		return -ENOMEM;
1953 
1954 	err = rtm_to_route_config(skb, nlh, cfg, extack);
1955 	if (err < 0)
1956 		goto out;
1957 
1958 	err = mpls_route_add(cfg, extack);
1959 out:
1960 	kfree(cfg);
1961 
1962 	return err;
1963 }
1964 
1965 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1966 			   u32 label, struct mpls_route *rt, int flags)
1967 {
1968 	struct net_device *dev;
1969 	struct nlmsghdr *nlh;
1970 	struct rtmsg *rtm;
1971 
1972 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1973 	if (nlh == NULL)
1974 		return -EMSGSIZE;
1975 
1976 	rtm = nlmsg_data(nlh);
1977 	rtm->rtm_family = AF_MPLS;
1978 	rtm->rtm_dst_len = 20;
1979 	rtm->rtm_src_len = 0;
1980 	rtm->rtm_tos = 0;
1981 	rtm->rtm_table = RT_TABLE_MAIN;
1982 	rtm->rtm_protocol = rt->rt_protocol;
1983 	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
1984 	rtm->rtm_type = RTN_UNICAST;
1985 	rtm->rtm_flags = 0;
1986 
1987 	if (nla_put_labels(skb, RTA_DST, 1, &label))
1988 		goto nla_put_failure;
1989 
1990 	if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
1991 		bool ttl_propagate =
1992 			rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
1993 
1994 		if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
1995 			       ttl_propagate))
1996 			goto nla_put_failure;
1997 	}
1998 	if (rt->rt_nhn == 1) {
1999 		const struct mpls_nh *nh = rt->rt_nh;
2000 
2001 		if (nh->nh_labels &&
2002 		    nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2003 				   nh->nh_label))
2004 			goto nla_put_failure;
2005 		if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2006 		    nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2007 				nh->nh_via_alen))
2008 			goto nla_put_failure;
2009 		dev = rtnl_dereference(nh->nh_dev);
2010 		if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2011 			goto nla_put_failure;
2012 		if (nh->nh_flags & RTNH_F_LINKDOWN)
2013 			rtm->rtm_flags |= RTNH_F_LINKDOWN;
2014 		if (nh->nh_flags & RTNH_F_DEAD)
2015 			rtm->rtm_flags |= RTNH_F_DEAD;
2016 	} else {
2017 		struct rtnexthop *rtnh;
2018 		struct nlattr *mp;
2019 		u8 linkdown = 0;
2020 		u8 dead = 0;
2021 
2022 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
2023 		if (!mp)
2024 			goto nla_put_failure;
2025 
2026 		for_nexthops(rt) {
2027 			dev = rtnl_dereference(nh->nh_dev);
2028 			if (!dev)
2029 				continue;
2030 
2031 			rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
2032 			if (!rtnh)
2033 				goto nla_put_failure;
2034 
2035 			rtnh->rtnh_ifindex = dev->ifindex;
2036 			if (nh->nh_flags & RTNH_F_LINKDOWN) {
2037 				rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
2038 				linkdown++;
2039 			}
2040 			if (nh->nh_flags & RTNH_F_DEAD) {
2041 				rtnh->rtnh_flags |= RTNH_F_DEAD;
2042 				dead++;
2043 			}
2044 
2045 			if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
2046 							    nh->nh_labels,
2047 							    nh->nh_label))
2048 				goto nla_put_failure;
2049 			if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2050 			    nla_put_via(skb, nh->nh_via_table,
2051 					mpls_nh_via(rt, nh),
2052 					nh->nh_via_alen))
2053 				goto nla_put_failure;
2054 
2055 			/* length of rtnetlink header + attributes */
2056 			rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2057 		} endfor_nexthops(rt);
2058 
2059 		if (linkdown == rt->rt_nhn)
2060 			rtm->rtm_flags |= RTNH_F_LINKDOWN;
2061 		if (dead == rt->rt_nhn)
2062 			rtm->rtm_flags |= RTNH_F_DEAD;
2063 
2064 		nla_nest_end(skb, mp);
2065 	}
2066 
2067 	nlmsg_end(skb, nlh);
2068 	return 0;
2069 
2070 nla_put_failure:
2071 	nlmsg_cancel(skb, nlh);
2072 	return -EMSGSIZE;
2073 }
2074 
2075 #if IS_ENABLED(CONFIG_INET)
2076 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2077 				   struct fib_dump_filter *filter,
2078 				   struct netlink_callback *cb)
2079 {
2080 	return ip_valid_fib_dump_req(net, nlh, filter, cb);
2081 }
2082 #else
2083 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2084 				   struct fib_dump_filter *filter,
2085 				   struct netlink_callback *cb)
2086 {
2087 	struct netlink_ext_ack *extack = cb->extack;
2088 	struct nlattr *tb[RTA_MAX + 1];
2089 	struct rtmsg *rtm;
2090 	int err, i;
2091 
2092 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2093 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for FIB dump request");
2094 		return -EINVAL;
2095 	}
2096 
2097 	rtm = nlmsg_data(nlh);
2098 	if (rtm->rtm_dst_len || rtm->rtm_src_len  || rtm->rtm_tos   ||
2099 	    rtm->rtm_table   || rtm->rtm_scope    || rtm->rtm_type  ||
2100 	    rtm->rtm_flags) {
2101 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for FIB dump request");
2102 		return -EINVAL;
2103 	}
2104 
2105 	if (rtm->rtm_protocol) {
2106 		filter->protocol = rtm->rtm_protocol;
2107 		filter->filter_set = 1;
2108 		cb->answer_flags = NLM_F_DUMP_FILTERED;
2109 	}
2110 
2111 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2112 					    rtm_mpls_policy, extack);
2113 	if (err < 0)
2114 		return err;
2115 
2116 	for (i = 0; i <= RTA_MAX; ++i) {
2117 		int ifindex;
2118 
2119 		if (i == RTA_OIF) {
2120 			ifindex = nla_get_u32(tb[i]);
2121 			filter->dev = __dev_get_by_index(net, ifindex);
2122 			if (!filter->dev)
2123 				return -ENODEV;
2124 			filter->filter_set = 1;
2125 		} else if (tb[i]) {
2126 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
2127 			return -EINVAL;
2128 		}
2129 	}
2130 
2131 	return 0;
2132 }
2133 #endif
2134 
2135 static bool mpls_rt_uses_dev(struct mpls_route *rt,
2136 			     const struct net_device *dev)
2137 {
2138 	struct net_device *nh_dev;
2139 
2140 	if (rt->rt_nhn == 1) {
2141 		struct mpls_nh *nh = rt->rt_nh;
2142 
2143 		nh_dev = rtnl_dereference(nh->nh_dev);
2144 		if (dev == nh_dev)
2145 			return true;
2146 	} else {
2147 		for_nexthops(rt) {
2148 			nh_dev = rtnl_dereference(nh->nh_dev);
2149 			if (nh_dev == dev)
2150 				return true;
2151 		} endfor_nexthops(rt);
2152 	}
2153 
2154 	return false;
2155 }
2156 
2157 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2158 {
2159 	const struct nlmsghdr *nlh = cb->nlh;
2160 	struct net *net = sock_net(skb->sk);
2161 	struct mpls_route __rcu **platform_label;
2162 	struct fib_dump_filter filter = {};
2163 	unsigned int flags = NLM_F_MULTI;
2164 	size_t platform_labels;
2165 	unsigned int index;
2166 
2167 	ASSERT_RTNL();
2168 
2169 	if (cb->strict_check) {
2170 		int err;
2171 
2172 		err = mpls_valid_fib_dump_req(net, nlh, &filter, cb);
2173 		if (err < 0)
2174 			return err;
2175 
2176 		/* for MPLS, there is only 1 table with fixed type and flags.
2177 		 * If either are set in the filter then return nothing.
2178 		 */
2179 		if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) ||
2180 		    (filter.rt_type && filter.rt_type != RTN_UNICAST) ||
2181 		     filter.flags)
2182 			return skb->len;
2183 	}
2184 
2185 	index = cb->args[0];
2186 	if (index < MPLS_LABEL_FIRST_UNRESERVED)
2187 		index = MPLS_LABEL_FIRST_UNRESERVED;
2188 
2189 	platform_label = rtnl_dereference(net->mpls.platform_label);
2190 	platform_labels = net->mpls.platform_labels;
2191 
2192 	if (filter.filter_set)
2193 		flags |= NLM_F_DUMP_FILTERED;
2194 
2195 	for (; index < platform_labels; index++) {
2196 		struct mpls_route *rt;
2197 
2198 		rt = rtnl_dereference(platform_label[index]);
2199 		if (!rt)
2200 			continue;
2201 
2202 		if ((filter.dev && !mpls_rt_uses_dev(rt, filter.dev)) ||
2203 		    (filter.protocol && rt->rt_protocol != filter.protocol))
2204 			continue;
2205 
2206 		if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2207 				    cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2208 				    index, rt, flags) < 0)
2209 			break;
2210 	}
2211 	cb->args[0] = index;
2212 
2213 	return skb->len;
2214 }
2215 
2216 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2217 {
2218 	size_t payload =
2219 		NLMSG_ALIGN(sizeof(struct rtmsg))
2220 		+ nla_total_size(4)			/* RTA_DST */
2221 		+ nla_total_size(1);			/* RTA_TTL_PROPAGATE */
2222 
2223 	if (rt->rt_nhn == 1) {
2224 		struct mpls_nh *nh = rt->rt_nh;
2225 
2226 		if (nh->nh_dev)
2227 			payload += nla_total_size(4); /* RTA_OIF */
2228 		if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2229 			payload += nla_total_size(2 + nh->nh_via_alen);
2230 		if (nh->nh_labels) /* RTA_NEWDST */
2231 			payload += nla_total_size(nh->nh_labels * 4);
2232 	} else {
2233 		/* each nexthop is packed in an attribute */
2234 		size_t nhsize = 0;
2235 
2236 		for_nexthops(rt) {
2237 			if (!rtnl_dereference(nh->nh_dev))
2238 				continue;
2239 			nhsize += nla_total_size(sizeof(struct rtnexthop));
2240 			/* RTA_VIA */
2241 			if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2242 				nhsize += nla_total_size(2 + nh->nh_via_alen);
2243 			if (nh->nh_labels)
2244 				nhsize += nla_total_size(nh->nh_labels * 4);
2245 		} endfor_nexthops(rt);
2246 		/* nested attribute */
2247 		payload += nla_total_size(nhsize);
2248 	}
2249 
2250 	return payload;
2251 }
2252 
2253 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2254 		       struct nlmsghdr *nlh, struct net *net, u32 portid,
2255 		       unsigned int nlm_flags)
2256 {
2257 	struct sk_buff *skb;
2258 	u32 seq = nlh ? nlh->nlmsg_seq : 0;
2259 	int err = -ENOBUFS;
2260 
2261 	skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2262 	if (skb == NULL)
2263 		goto errout;
2264 
2265 	err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2266 	if (err < 0) {
2267 		/* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2268 		WARN_ON(err == -EMSGSIZE);
2269 		kfree_skb(skb);
2270 		goto errout;
2271 	}
2272 	rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2273 
2274 	return;
2275 errout:
2276 	if (err < 0)
2277 		rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2278 }
2279 
2280 static int mpls_valid_getroute_req(struct sk_buff *skb,
2281 				   const struct nlmsghdr *nlh,
2282 				   struct nlattr **tb,
2283 				   struct netlink_ext_ack *extack)
2284 {
2285 	struct rtmsg *rtm;
2286 	int i, err;
2287 
2288 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2289 		NL_SET_ERR_MSG_MOD(extack,
2290 				   "Invalid header for get route request");
2291 		return -EINVAL;
2292 	}
2293 
2294 	if (!netlink_strict_get_check(skb))
2295 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2296 					      rtm_mpls_policy, extack);
2297 
2298 	rtm = nlmsg_data(nlh);
2299 	if ((rtm->rtm_dst_len && rtm->rtm_dst_len != 20) ||
2300 	    rtm->rtm_src_len || rtm->rtm_tos || rtm->rtm_table ||
2301 	    rtm->rtm_protocol || rtm->rtm_scope || rtm->rtm_type) {
2302 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
2303 		return -EINVAL;
2304 	}
2305 	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
2306 		NL_SET_ERR_MSG_MOD(extack,
2307 				   "Invalid flags for get route request");
2308 		return -EINVAL;
2309 	}
2310 
2311 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2312 					    rtm_mpls_policy, extack);
2313 	if (err)
2314 		return err;
2315 
2316 	if ((tb[RTA_DST] || tb[RTA_NEWDST]) && !rtm->rtm_dst_len) {
2317 		NL_SET_ERR_MSG_MOD(extack, "rtm_dst_len must be 20 for MPLS");
2318 		return -EINVAL;
2319 	}
2320 
2321 	for (i = 0; i <= RTA_MAX; i++) {
2322 		if (!tb[i])
2323 			continue;
2324 
2325 		switch (i) {
2326 		case RTA_DST:
2327 		case RTA_NEWDST:
2328 			break;
2329 		default:
2330 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
2331 			return -EINVAL;
2332 		}
2333 	}
2334 
2335 	return 0;
2336 }
2337 
2338 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2339 			 struct netlink_ext_ack *extack)
2340 {
2341 	struct net *net = sock_net(in_skb->sk);
2342 	u32 portid = NETLINK_CB(in_skb).portid;
2343 	u32 in_label = LABEL_NOT_SPECIFIED;
2344 	struct nlattr *tb[RTA_MAX + 1];
2345 	u32 labels[MAX_NEW_LABELS];
2346 	struct mpls_shim_hdr *hdr;
2347 	unsigned int hdr_size = 0;
2348 	struct net_device *dev;
2349 	struct mpls_route *rt;
2350 	struct rtmsg *rtm, *r;
2351 	struct nlmsghdr *nlh;
2352 	struct sk_buff *skb;
2353 	struct mpls_nh *nh;
2354 	u8 n_labels;
2355 	int err;
2356 
2357 	err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack);
2358 	if (err < 0)
2359 		goto errout;
2360 
2361 	rtm = nlmsg_data(in_nlh);
2362 
2363 	if (tb[RTA_DST]) {
2364 		u8 label_count;
2365 
2366 		if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2367 				   &in_label, extack)) {
2368 			err = -EINVAL;
2369 			goto errout;
2370 		}
2371 
2372 		if (!mpls_label_ok(net, &in_label, extack)) {
2373 			err = -EINVAL;
2374 			goto errout;
2375 		}
2376 	}
2377 
2378 	rt = mpls_route_input_rcu(net, in_label);
2379 	if (!rt) {
2380 		err = -ENETUNREACH;
2381 		goto errout;
2382 	}
2383 
2384 	if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2385 		skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2386 		if (!skb) {
2387 			err = -ENOBUFS;
2388 			goto errout;
2389 		}
2390 
2391 		err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2392 				      RTM_NEWROUTE, in_label, rt, 0);
2393 		if (err < 0) {
2394 			/* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2395 			WARN_ON(err == -EMSGSIZE);
2396 			goto errout_free;
2397 		}
2398 
2399 		return rtnl_unicast(skb, net, portid);
2400 	}
2401 
2402 	if (tb[RTA_NEWDST]) {
2403 		if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2404 				   labels, extack) != 0) {
2405 			err = -EINVAL;
2406 			goto errout;
2407 		}
2408 
2409 		hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2410 	}
2411 
2412 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2413 	if (!skb) {
2414 		err = -ENOBUFS;
2415 		goto errout;
2416 	}
2417 
2418 	skb->protocol = htons(ETH_P_MPLS_UC);
2419 
2420 	if (hdr_size) {
2421 		bool bos;
2422 		int i;
2423 
2424 		if (skb_cow(skb, hdr_size)) {
2425 			err = -ENOBUFS;
2426 			goto errout_free;
2427 		}
2428 
2429 		skb_reserve(skb, hdr_size);
2430 		skb_push(skb, hdr_size);
2431 		skb_reset_network_header(skb);
2432 
2433 		/* Push new labels */
2434 		hdr = mpls_hdr(skb);
2435 		bos = true;
2436 		for (i = n_labels - 1; i >= 0; i--) {
2437 			hdr[i] = mpls_entry_encode(labels[i],
2438 						   1, 0, bos);
2439 			bos = false;
2440 		}
2441 	}
2442 
2443 	nh = mpls_select_multipath(rt, skb);
2444 	if (!nh) {
2445 		err = -ENETUNREACH;
2446 		goto errout_free;
2447 	}
2448 
2449 	if (hdr_size) {
2450 		skb_pull(skb, hdr_size);
2451 		skb_reset_network_header(skb);
2452 	}
2453 
2454 	nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2455 			RTM_NEWROUTE, sizeof(*r), 0);
2456 	if (!nlh) {
2457 		err = -EMSGSIZE;
2458 		goto errout_free;
2459 	}
2460 
2461 	r = nlmsg_data(nlh);
2462 	r->rtm_family	 = AF_MPLS;
2463 	r->rtm_dst_len	= 20;
2464 	r->rtm_src_len	= 0;
2465 	r->rtm_table	= RT_TABLE_MAIN;
2466 	r->rtm_type	= RTN_UNICAST;
2467 	r->rtm_scope	= RT_SCOPE_UNIVERSE;
2468 	r->rtm_protocol = rt->rt_protocol;
2469 	r->rtm_flags	= 0;
2470 
2471 	if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2472 		goto nla_put_failure;
2473 
2474 	if (nh->nh_labels &&
2475 	    nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2476 			   nh->nh_label))
2477 		goto nla_put_failure;
2478 
2479 	if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2480 	    nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2481 			nh->nh_via_alen))
2482 		goto nla_put_failure;
2483 	dev = rtnl_dereference(nh->nh_dev);
2484 	if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2485 		goto nla_put_failure;
2486 
2487 	nlmsg_end(skb, nlh);
2488 
2489 	err = rtnl_unicast(skb, net, portid);
2490 errout:
2491 	return err;
2492 
2493 nla_put_failure:
2494 	nlmsg_cancel(skb, nlh);
2495 	err = -EMSGSIZE;
2496 errout_free:
2497 	kfree_skb(skb);
2498 	return err;
2499 }
2500 
2501 static int resize_platform_label_table(struct net *net, size_t limit)
2502 {
2503 	size_t size = sizeof(struct mpls_route *) * limit;
2504 	size_t old_limit;
2505 	size_t cp_size;
2506 	struct mpls_route __rcu **labels = NULL, **old;
2507 	struct mpls_route *rt0 = NULL, *rt2 = NULL;
2508 	unsigned index;
2509 
2510 	if (size) {
2511 		labels = kvzalloc(size, GFP_KERNEL);
2512 		if (!labels)
2513 			goto nolabels;
2514 	}
2515 
2516 	/* In case the predefined labels need to be populated */
2517 	if (limit > MPLS_LABEL_IPV4NULL) {
2518 		struct net_device *lo = net->loopback_dev;
2519 		rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2520 		if (IS_ERR(rt0))
2521 			goto nort0;
2522 		RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo);
2523 		rt0->rt_protocol = RTPROT_KERNEL;
2524 		rt0->rt_payload_type = MPT_IPV4;
2525 		rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2526 		rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2527 		rt0->rt_nh->nh_via_alen = lo->addr_len;
2528 		memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2529 		       lo->addr_len);
2530 	}
2531 	if (limit > MPLS_LABEL_IPV6NULL) {
2532 		struct net_device *lo = net->loopback_dev;
2533 		rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2534 		if (IS_ERR(rt2))
2535 			goto nort2;
2536 		RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo);
2537 		rt2->rt_protocol = RTPROT_KERNEL;
2538 		rt2->rt_payload_type = MPT_IPV6;
2539 		rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2540 		rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2541 		rt2->rt_nh->nh_via_alen = lo->addr_len;
2542 		memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2543 		       lo->addr_len);
2544 	}
2545 
2546 	rtnl_lock();
2547 	/* Remember the original table */
2548 	old = rtnl_dereference(net->mpls.platform_label);
2549 	old_limit = net->mpls.platform_labels;
2550 
2551 	/* Free any labels beyond the new table */
2552 	for (index = limit; index < old_limit; index++)
2553 		mpls_route_update(net, index, NULL, NULL);
2554 
2555 	/* Copy over the old labels */
2556 	cp_size = size;
2557 	if (old_limit < limit)
2558 		cp_size = old_limit * sizeof(struct mpls_route *);
2559 
2560 	memcpy(labels, old, cp_size);
2561 
2562 	/* If needed set the predefined labels */
2563 	if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2564 	    (limit > MPLS_LABEL_IPV6NULL)) {
2565 		RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2566 		rt2 = NULL;
2567 	}
2568 
2569 	if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2570 	    (limit > MPLS_LABEL_IPV4NULL)) {
2571 		RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2572 		rt0 = NULL;
2573 	}
2574 
2575 	/* Update the global pointers */
2576 	net->mpls.platform_labels = limit;
2577 	rcu_assign_pointer(net->mpls.platform_label, labels);
2578 
2579 	rtnl_unlock();
2580 
2581 	mpls_rt_free(rt2);
2582 	mpls_rt_free(rt0);
2583 
2584 	if (old) {
2585 		synchronize_rcu();
2586 		kvfree(old);
2587 	}
2588 	return 0;
2589 
2590 nort2:
2591 	mpls_rt_free(rt0);
2592 nort0:
2593 	kvfree(labels);
2594 nolabels:
2595 	return -ENOMEM;
2596 }
2597 
2598 static int mpls_platform_labels(struct ctl_table *table, int write,
2599 				void __user *buffer, size_t *lenp, loff_t *ppos)
2600 {
2601 	struct net *net = table->data;
2602 	int platform_labels = net->mpls.platform_labels;
2603 	int ret;
2604 	struct ctl_table tmp = {
2605 		.procname	= table->procname,
2606 		.data		= &platform_labels,
2607 		.maxlen		= sizeof(int),
2608 		.mode		= table->mode,
2609 		.extra1		= &zero,
2610 		.extra2		= &label_limit,
2611 	};
2612 
2613 	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2614 
2615 	if (write && ret == 0)
2616 		ret = resize_platform_label_table(net, platform_labels);
2617 
2618 	return ret;
2619 }
2620 
2621 #define MPLS_NS_SYSCTL_OFFSET(field)		\
2622 	(&((struct net *)0)->field)
2623 
2624 static const struct ctl_table mpls_table[] = {
2625 	{
2626 		.procname	= "platform_labels",
2627 		.data		= NULL,
2628 		.maxlen		= sizeof(int),
2629 		.mode		= 0644,
2630 		.proc_handler	= mpls_platform_labels,
2631 	},
2632 	{
2633 		.procname	= "ip_ttl_propagate",
2634 		.data		= MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2635 		.maxlen		= sizeof(int),
2636 		.mode		= 0644,
2637 		.proc_handler	= proc_dointvec_minmax,
2638 		.extra1		= &zero,
2639 		.extra2		= &one,
2640 	},
2641 	{
2642 		.procname	= "default_ttl",
2643 		.data		= MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2644 		.maxlen		= sizeof(int),
2645 		.mode		= 0644,
2646 		.proc_handler	= proc_dointvec_minmax,
2647 		.extra1		= &one,
2648 		.extra2		= &ttl_max,
2649 	},
2650 	{ }
2651 };
2652 
2653 static int mpls_net_init(struct net *net)
2654 {
2655 	struct ctl_table *table;
2656 	int i;
2657 
2658 	net->mpls.platform_labels = 0;
2659 	net->mpls.platform_label = NULL;
2660 	net->mpls.ip_ttl_propagate = 1;
2661 	net->mpls.default_ttl = 255;
2662 
2663 	table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2664 	if (table == NULL)
2665 		return -ENOMEM;
2666 
2667 	/* Table data contains only offsets relative to the base of
2668 	 * the mdev at this point, so make them absolute.
2669 	 */
2670 	for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++)
2671 		table[i].data = (char *)net + (uintptr_t)table[i].data;
2672 
2673 	net->mpls.ctl = register_net_sysctl(net, "net/mpls", table);
2674 	if (net->mpls.ctl == NULL) {
2675 		kfree(table);
2676 		return -ENOMEM;
2677 	}
2678 
2679 	return 0;
2680 }
2681 
2682 static void mpls_net_exit(struct net *net)
2683 {
2684 	struct mpls_route __rcu **platform_label;
2685 	size_t platform_labels;
2686 	struct ctl_table *table;
2687 	unsigned int index;
2688 
2689 	table = net->mpls.ctl->ctl_table_arg;
2690 	unregister_net_sysctl_table(net->mpls.ctl);
2691 	kfree(table);
2692 
2693 	/* An rcu grace period has passed since there was a device in
2694 	 * the network namespace (and thus the last in flight packet)
2695 	 * left this network namespace.  This is because
2696 	 * unregister_netdevice_many and netdev_run_todo has completed
2697 	 * for each network device that was in this network namespace.
2698 	 *
2699 	 * As such no additional rcu synchronization is necessary when
2700 	 * freeing the platform_label table.
2701 	 */
2702 	rtnl_lock();
2703 	platform_label = rtnl_dereference(net->mpls.platform_label);
2704 	platform_labels = net->mpls.platform_labels;
2705 	for (index = 0; index < platform_labels; index++) {
2706 		struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2707 		RCU_INIT_POINTER(platform_label[index], NULL);
2708 		mpls_notify_route(net, index, rt, NULL, NULL);
2709 		mpls_rt_free(rt);
2710 	}
2711 	rtnl_unlock();
2712 
2713 	kvfree(platform_label);
2714 }
2715 
2716 static struct pernet_operations mpls_net_ops = {
2717 	.init = mpls_net_init,
2718 	.exit = mpls_net_exit,
2719 };
2720 
2721 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2722 	.family		   = AF_MPLS,
2723 	.fill_stats_af	   = mpls_fill_stats_af,
2724 	.get_stats_af_size = mpls_get_stats_af_size,
2725 };
2726 
2727 static int __init mpls_init(void)
2728 {
2729 	int err;
2730 
2731 	BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2732 
2733 	err = register_pernet_subsys(&mpls_net_ops);
2734 	if (err)
2735 		goto out;
2736 
2737 	err = register_netdevice_notifier(&mpls_dev_notifier);
2738 	if (err)
2739 		goto out_unregister_pernet;
2740 
2741 	dev_add_pack(&mpls_packet_type);
2742 
2743 	rtnl_af_register(&mpls_af_ops);
2744 
2745 	rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_NEWROUTE,
2746 			     mpls_rtm_newroute, NULL, 0);
2747 	rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_DELROUTE,
2748 			     mpls_rtm_delroute, NULL, 0);
2749 	rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETROUTE,
2750 			     mpls_getroute, mpls_dump_routes, 0);
2751 	rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
2752 			     mpls_netconf_get_devconf,
2753 			     mpls_netconf_dump_devconf, 0);
2754 	err = ipgre_tunnel_encap_add_mpls_ops();
2755 	if (err)
2756 		pr_err("Can't add mpls over gre tunnel ops\n");
2757 
2758 	err = 0;
2759 out:
2760 	return err;
2761 
2762 out_unregister_pernet:
2763 	unregister_pernet_subsys(&mpls_net_ops);
2764 	goto out;
2765 }
2766 module_init(mpls_init);
2767 
2768 static void __exit mpls_exit(void)
2769 {
2770 	rtnl_unregister_all(PF_MPLS);
2771 	rtnl_af_unregister(&mpls_af_ops);
2772 	dev_remove_pack(&mpls_packet_type);
2773 	unregister_netdevice_notifier(&mpls_dev_notifier);
2774 	unregister_pernet_subsys(&mpls_net_ops);
2775 	ipgre_tunnel_encap_del_mpls_ops();
2776 }
2777 module_exit(mpls_exit);
2778 
2779 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2780 MODULE_LICENSE("GPL v2");
2781 MODULE_ALIAS_NETPROTO(PF_MPLS);
2782