1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/types.h>
3 #include <linux/skbuff.h>
4 #include <linux/socket.h>
5 #include <linux/sysctl.h>
6 #include <linux/net.h>
7 #include <linux/module.h>
8 #include <linux/if_arp.h>
9 #include <linux/ipv6.h>
10 #include <linux/mpls.h>
11 #include <linux/netconf.h>
12 #include <linux/nospec.h>
13 #include <linux/vmalloc.h>
14 #include <linux/percpu.h>
15 #include <net/ip.h>
16 #include <net/dst.h>
17 #include <net/sock.h>
18 #include <net/arp.h>
19 #include <net/ip_fib.h>
20 #include <net/netevent.h>
21 #include <net/ip_tunnels.h>
22 #include <net/netns/generic.h>
23 #if IS_ENABLED(CONFIG_IPV6)
24 #include <net/ipv6.h>
25 #endif
26 #include <net/ipv6_stubs.h>
27 #include <net/rtnh.h>
28 #include "internal.h"
29
30 /* max memory we will use for mpls_route */
31 #define MAX_MPLS_ROUTE_MEM 4096
32
33 /* Maximum number of labels to look ahead at when selecting a path of
34 * a multipath route
35 */
36 #define MAX_MP_SELECT_LABELS 4
37
38 #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1)
39
40 static int label_limit = (1 << 20) - 1;
41 static int ttl_max = 255;
42
43 #if IS_ENABLED(CONFIG_NET_IP_TUNNEL)
ipgre_mpls_encap_hlen(struct ip_tunnel_encap * e)44 static size_t ipgre_mpls_encap_hlen(struct ip_tunnel_encap *e)
45 {
46 return sizeof(struct mpls_shim_hdr);
47 }
48
49 static const struct ip_tunnel_encap_ops mpls_iptun_ops = {
50 .encap_hlen = ipgre_mpls_encap_hlen,
51 };
52
ipgre_tunnel_encap_add_mpls_ops(void)53 static int ipgre_tunnel_encap_add_mpls_ops(void)
54 {
55 return ip_tunnel_encap_add_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
56 }
57
ipgre_tunnel_encap_del_mpls_ops(void)58 static void ipgre_tunnel_encap_del_mpls_ops(void)
59 {
60 ip_tunnel_encap_del_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
61 }
62 #else
ipgre_tunnel_encap_add_mpls_ops(void)63 static int ipgre_tunnel_encap_add_mpls_ops(void)
64 {
65 return 0;
66 }
67
ipgre_tunnel_encap_del_mpls_ops(void)68 static void ipgre_tunnel_encap_del_mpls_ops(void)
69 {
70 }
71 #endif
72
73 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
74 struct nlmsghdr *nlh, struct net *net, u32 portid,
75 unsigned int nlm_flags);
76
mpls_route_input_rcu(struct net * net,unsigned index)77 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
78 {
79 struct mpls_route *rt = NULL;
80
81 if (index < net->mpls.platform_labels) {
82 struct mpls_route __rcu **platform_label =
83 rcu_dereference(net->mpls.platform_label);
84 rt = rcu_dereference(platform_label[index]);
85 }
86 return rt;
87 }
88
mpls_output_possible(const struct net_device * dev)89 bool mpls_output_possible(const struct net_device *dev)
90 {
91 return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);
92 }
93 EXPORT_SYMBOL_GPL(mpls_output_possible);
94
__mpls_nh_via(struct mpls_route * rt,struct mpls_nh * nh)95 static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh)
96 {
97 return (u8 *)nh + rt->rt_via_offset;
98 }
99
mpls_nh_via(const struct mpls_route * rt,const struct mpls_nh * nh)100 static const u8 *mpls_nh_via(const struct mpls_route *rt,
101 const struct mpls_nh *nh)
102 {
103 return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh);
104 }
105
mpls_nh_header_size(const struct mpls_nh * nh)106 static unsigned int mpls_nh_header_size(const struct mpls_nh *nh)
107 {
108 /* The size of the layer 2.5 labels to be added for this route */
109 return nh->nh_labels * sizeof(struct mpls_shim_hdr);
110 }
111
mpls_dev_mtu(const struct net_device * dev)112 unsigned int mpls_dev_mtu(const struct net_device *dev)
113 {
114 /* The amount of data the layer 2 frame can hold */
115 return dev->mtu;
116 }
117 EXPORT_SYMBOL_GPL(mpls_dev_mtu);
118
mpls_pkt_too_big(const struct sk_buff * skb,unsigned int mtu)119 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
120 {
121 if (skb->len <= mtu)
122 return false;
123
124 if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
125 return false;
126
127 return true;
128 }
129 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
130
mpls_stats_inc_outucastpkts(struct net_device * dev,const struct sk_buff * skb)131 void mpls_stats_inc_outucastpkts(struct net_device *dev,
132 const struct sk_buff *skb)
133 {
134 struct mpls_dev *mdev;
135
136 if (skb->protocol == htons(ETH_P_MPLS_UC)) {
137 mdev = mpls_dev_get(dev);
138 if (mdev)
139 MPLS_INC_STATS_LEN(mdev, skb->len,
140 tx_packets,
141 tx_bytes);
142 } else if (skb->protocol == htons(ETH_P_IP)) {
143 IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
144 #if IS_ENABLED(CONFIG_IPV6)
145 } else if (skb->protocol == htons(ETH_P_IPV6)) {
146 struct inet6_dev *in6dev = __in6_dev_get(dev);
147
148 if (in6dev)
149 IP6_UPD_PO_STATS(dev_net(dev), in6dev,
150 IPSTATS_MIB_OUT, skb->len);
151 #endif
152 }
153 }
154 EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
155
mpls_multipath_hash(struct mpls_route * rt,struct sk_buff * skb)156 static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb)
157 {
158 struct mpls_entry_decoded dec;
159 unsigned int mpls_hdr_len = 0;
160 struct mpls_shim_hdr *hdr;
161 bool eli_seen = false;
162 int label_index;
163 u32 hash = 0;
164
165 for (label_index = 0; label_index < MAX_MP_SELECT_LABELS;
166 label_index++) {
167 mpls_hdr_len += sizeof(*hdr);
168 if (!pskb_may_pull(skb, mpls_hdr_len))
169 break;
170
171 /* Read and decode the current label */
172 hdr = mpls_hdr(skb) + label_index;
173 dec = mpls_entry_decode(hdr);
174
175 /* RFC6790 - reserved labels MUST NOT be used as keys
176 * for the load-balancing function
177 */
178 if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) {
179 hash = jhash_1word(dec.label, hash);
180
181 /* The entropy label follows the entropy label
182 * indicator, so this means that the entropy
183 * label was just added to the hash - no need to
184 * go any deeper either in the label stack or in the
185 * payload
186 */
187 if (eli_seen)
188 break;
189 } else if (dec.label == MPLS_LABEL_ENTROPY) {
190 eli_seen = true;
191 }
192
193 if (!dec.bos)
194 continue;
195
196 /* found bottom label; does skb have room for a header? */
197 if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) {
198 const struct iphdr *v4hdr;
199
200 v4hdr = (const struct iphdr *)(hdr + 1);
201 if (v4hdr->version == 4) {
202 hash = jhash_3words(ntohl(v4hdr->saddr),
203 ntohl(v4hdr->daddr),
204 v4hdr->protocol, hash);
205 } else if (v4hdr->version == 6 &&
206 pskb_may_pull(skb, mpls_hdr_len +
207 sizeof(struct ipv6hdr))) {
208 const struct ipv6hdr *v6hdr;
209
210 v6hdr = (const struct ipv6hdr *)(hdr + 1);
211 hash = __ipv6_addr_jhash(&v6hdr->saddr, hash);
212 hash = __ipv6_addr_jhash(&v6hdr->daddr, hash);
213 hash = jhash_1word(v6hdr->nexthdr, hash);
214 }
215 }
216
217 break;
218 }
219
220 return hash;
221 }
222
mpls_get_nexthop(struct mpls_route * rt,u8 index)223 static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index)
224 {
225 return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size);
226 }
227
228 /* number of alive nexthops (rt->rt_nhn_alive) and the flags for
229 * a next hop (nh->nh_flags) are modified by netdev event handlers.
230 * Since those fields can change at any moment, use READ_ONCE to
231 * access both.
232 */
mpls_select_multipath(struct mpls_route * rt,struct sk_buff * skb)233 static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt,
234 struct sk_buff *skb)
235 {
236 u32 hash = 0;
237 int nh_index = 0;
238 int n = 0;
239 u8 alive;
240
241 /* No need to look further into packet if there's only
242 * one path
243 */
244 if (rt->rt_nhn == 1)
245 return rt->rt_nh;
246
247 alive = READ_ONCE(rt->rt_nhn_alive);
248 if (alive == 0)
249 return NULL;
250
251 hash = mpls_multipath_hash(rt, skb);
252 nh_index = hash % alive;
253 if (alive == rt->rt_nhn)
254 goto out;
255 for_nexthops(rt) {
256 unsigned int nh_flags = READ_ONCE(nh->nh_flags);
257
258 if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
259 continue;
260 if (n == nh_index)
261 return nh;
262 n++;
263 } endfor_nexthops(rt);
264
265 out:
266 return mpls_get_nexthop(rt, nh_index);
267 }
268
mpls_egress(struct net * net,struct mpls_route * rt,struct sk_buff * skb,struct mpls_entry_decoded dec)269 static bool mpls_egress(struct net *net, struct mpls_route *rt,
270 struct sk_buff *skb, struct mpls_entry_decoded dec)
271 {
272 enum mpls_payload_type payload_type;
273 bool success = false;
274
275 /* The IPv4 code below accesses through the IPv4 header
276 * checksum, which is 12 bytes into the packet.
277 * The IPv6 code below accesses through the IPv6 hop limit
278 * which is 8 bytes into the packet.
279 *
280 * For all supported cases there should always be at least 12
281 * bytes of packet data present. The IPv4 header is 20 bytes
282 * without options and the IPv6 header is always 40 bytes
283 * long.
284 */
285 if (!pskb_may_pull(skb, 12))
286 return false;
287
288 payload_type = rt->rt_payload_type;
289 if (payload_type == MPT_UNSPEC)
290 payload_type = ip_hdr(skb)->version;
291
292 switch (payload_type) {
293 case MPT_IPV4: {
294 struct iphdr *hdr4 = ip_hdr(skb);
295 u8 new_ttl;
296 skb->protocol = htons(ETH_P_IP);
297
298 /* If propagating TTL, take the decremented TTL from
299 * the incoming MPLS header, otherwise decrement the
300 * TTL, but only if not 0 to avoid underflow.
301 */
302 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
303 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
304 net->mpls.ip_ttl_propagate))
305 new_ttl = dec.ttl;
306 else
307 new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0;
308
309 csum_replace2(&hdr4->check,
310 htons(hdr4->ttl << 8),
311 htons(new_ttl << 8));
312 hdr4->ttl = new_ttl;
313 success = true;
314 break;
315 }
316 case MPT_IPV6: {
317 struct ipv6hdr *hdr6 = ipv6_hdr(skb);
318 skb->protocol = htons(ETH_P_IPV6);
319
320 /* If propagating TTL, take the decremented TTL from
321 * the incoming MPLS header, otherwise decrement the
322 * hop limit, but only if not 0 to avoid underflow.
323 */
324 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
325 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
326 net->mpls.ip_ttl_propagate))
327 hdr6->hop_limit = dec.ttl;
328 else if (hdr6->hop_limit)
329 hdr6->hop_limit = hdr6->hop_limit - 1;
330 success = true;
331 break;
332 }
333 case MPT_UNSPEC:
334 /* Should have decided which protocol it is by now */
335 break;
336 }
337
338 return success;
339 }
340
mpls_forward(struct sk_buff * skb,struct net_device * dev,struct packet_type * pt,struct net_device * orig_dev)341 static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
342 struct packet_type *pt, struct net_device *orig_dev)
343 {
344 struct net *net = dev_net(dev);
345 struct mpls_shim_hdr *hdr;
346 struct mpls_route *rt;
347 struct mpls_nh *nh;
348 struct mpls_entry_decoded dec;
349 struct net_device *out_dev;
350 struct mpls_dev *out_mdev;
351 struct mpls_dev *mdev;
352 unsigned int hh_len;
353 unsigned int new_header_size;
354 unsigned int mtu;
355 int err;
356
357 /* Careful this entire function runs inside of an rcu critical section */
358
359 mdev = mpls_dev_get(dev);
360 if (!mdev)
361 goto drop;
362
363 MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
364 rx_bytes);
365
366 if (!mdev->input_enabled) {
367 MPLS_INC_STATS(mdev, rx_dropped);
368 goto drop;
369 }
370
371 if (skb->pkt_type != PACKET_HOST)
372 goto err;
373
374 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
375 goto err;
376
377 if (!pskb_may_pull(skb, sizeof(*hdr)))
378 goto err;
379
380 skb_dst_drop(skb);
381
382 /* Read and decode the label */
383 hdr = mpls_hdr(skb);
384 dec = mpls_entry_decode(hdr);
385
386 rt = mpls_route_input_rcu(net, dec.label);
387 if (!rt) {
388 MPLS_INC_STATS(mdev, rx_noroute);
389 goto drop;
390 }
391
392 nh = mpls_select_multipath(rt, skb);
393 if (!nh)
394 goto err;
395
396 /* Pop the label */
397 skb_pull(skb, sizeof(*hdr));
398 skb_reset_network_header(skb);
399
400 skb_orphan(skb);
401
402 if (skb_warn_if_lro(skb))
403 goto err;
404
405 skb_forward_csum(skb);
406
407 /* Verify ttl is valid */
408 if (dec.ttl <= 1)
409 goto err;
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 dec.ttl -= 1;
434 if (unlikely(!new_header_size && dec.bos)) {
435 /* Penultimate hop popping */
436 if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
437 goto err;
438 } else {
439 bool bos;
440 int i;
441 skb_push(skb, new_header_size);
442 skb_reset_network_header(skb);
443 /* Push the new labels */
444 hdr = mpls_hdr(skb);
445 bos = dec.bos;
446 for (i = nh->nh_labels - 1; i >= 0; i--) {
447 hdr[i] = mpls_entry_encode(nh->nh_label[i],
448 dec.ttl, 0, bos);
449 bos = false;
450 }
451 }
452
453 mpls_stats_inc_outucastpkts(out_dev, skb);
454
455 /* If via wasn't specified then send out using device address */
456 if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
457 err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,
458 out_dev->dev_addr, skb);
459 else
460 err = neigh_xmit(nh->nh_via_table, out_dev,
461 mpls_nh_via(rt, nh), skb);
462 if (err)
463 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
464 __func__, err);
465 return 0;
466
467 tx_err:
468 out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
469 if (out_mdev)
470 MPLS_INC_STATS(out_mdev, tx_errors);
471 goto drop;
472 err:
473 MPLS_INC_STATS(mdev, rx_errors);
474 drop:
475 kfree_skb(skb);
476 return NET_RX_DROP;
477 }
478
479 static struct packet_type mpls_packet_type __read_mostly = {
480 .type = cpu_to_be16(ETH_P_MPLS_UC),
481 .func = mpls_forward,
482 };
483
484 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = {
485 [RTA_DST] = { .type = NLA_U32 },
486 [RTA_OIF] = { .type = NLA_U32 },
487 [RTA_TTL_PROPAGATE] = { .type = NLA_U8 },
488 };
489
490 struct mpls_route_config {
491 u32 rc_protocol;
492 u32 rc_ifindex;
493 u8 rc_via_table;
494 u8 rc_via_alen;
495 u8 rc_via[MAX_VIA_ALEN];
496 u32 rc_label;
497 u8 rc_ttl_propagate;
498 u8 rc_output_labels;
499 u32 rc_output_label[MAX_NEW_LABELS];
500 u32 rc_nlflags;
501 enum mpls_payload_type rc_payload_type;
502 struct nl_info rc_nlinfo;
503 struct rtnexthop *rc_mp;
504 int rc_mp_len;
505 };
506
507 /* all nexthops within a route have the same size based on max
508 * number of labels and max via length for a hop
509 */
mpls_rt_alloc(u8 num_nh,u8 max_alen,u8 max_labels)510 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
511 {
512 u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen);
513 struct mpls_route *rt;
514 size_t size;
515
516 size = sizeof(*rt) + num_nh * nh_size;
517 if (size > MAX_MPLS_ROUTE_MEM)
518 return ERR_PTR(-EINVAL);
519
520 rt = kzalloc(size, GFP_KERNEL);
521 if (!rt)
522 return ERR_PTR(-ENOMEM);
523
524 rt->rt_nhn = num_nh;
525 rt->rt_nhn_alive = num_nh;
526 rt->rt_nh_size = nh_size;
527 rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels);
528
529 return rt;
530 }
531
mpls_rt_free(struct mpls_route * rt)532 static void mpls_rt_free(struct mpls_route *rt)
533 {
534 if (rt)
535 kfree_rcu(rt, rt_rcu);
536 }
537
mpls_notify_route(struct net * net,unsigned index,struct mpls_route * old,struct mpls_route * new,const struct nl_info * info)538 static void mpls_notify_route(struct net *net, unsigned index,
539 struct mpls_route *old, struct mpls_route *new,
540 const struct nl_info *info)
541 {
542 struct nlmsghdr *nlh = info ? info->nlh : NULL;
543 unsigned portid = info ? info->portid : 0;
544 int event = new ? RTM_NEWROUTE : RTM_DELROUTE;
545 struct mpls_route *rt = new ? new : old;
546 unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0;
547 /* Ignore reserved labels for now */
548 if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED))
549 rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags);
550 }
551
mpls_route_update(struct net * net,unsigned index,struct mpls_route * new,const struct nl_info * info)552 static void mpls_route_update(struct net *net, unsigned index,
553 struct mpls_route *new,
554 const struct nl_info *info)
555 {
556 struct mpls_route __rcu **platform_label;
557 struct mpls_route *rt;
558
559 ASSERT_RTNL();
560
561 platform_label = rtnl_dereference(net->mpls.platform_label);
562 rt = rtnl_dereference(platform_label[index]);
563 rcu_assign_pointer(platform_label[index], new);
564
565 mpls_notify_route(net, index, rt, new, info);
566
567 /* If we removed a route free it now */
568 mpls_rt_free(rt);
569 }
570
find_free_label(struct net * net)571 static unsigned find_free_label(struct net *net)
572 {
573 struct mpls_route __rcu **platform_label;
574 size_t platform_labels;
575 unsigned index;
576
577 platform_label = rtnl_dereference(net->mpls.platform_label);
578 platform_labels = net->mpls.platform_labels;
579 for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
580 index++) {
581 if (!rtnl_dereference(platform_label[index]))
582 return index;
583 }
584 return LABEL_NOT_SPECIFIED;
585 }
586
587 #if IS_ENABLED(CONFIG_INET)
inet_fib_lookup_dev(struct net * net,const void * addr)588 static struct net_device *inet_fib_lookup_dev(struct net *net,
589 const void *addr)
590 {
591 struct net_device *dev;
592 struct rtable *rt;
593 struct in_addr daddr;
594
595 memcpy(&daddr, addr, sizeof(struct in_addr));
596 rt = ip_route_output(net, daddr.s_addr, 0, 0, 0);
597 if (IS_ERR(rt))
598 return ERR_CAST(rt);
599
600 dev = rt->dst.dev;
601 dev_hold(dev);
602
603 ip_rt_put(rt);
604
605 return dev;
606 }
607 #else
inet_fib_lookup_dev(struct net * net,const void * addr)608 static struct net_device *inet_fib_lookup_dev(struct net *net,
609 const void *addr)
610 {
611 return ERR_PTR(-EAFNOSUPPORT);
612 }
613 #endif
614
615 #if IS_ENABLED(CONFIG_IPV6)
inet6_fib_lookup_dev(struct net * net,const void * addr)616 static struct net_device *inet6_fib_lookup_dev(struct net *net,
617 const void *addr)
618 {
619 struct net_device *dev;
620 struct dst_entry *dst;
621 struct flowi6 fl6;
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 dst = ipv6_stub->ipv6_dst_lookup_flow(net, NULL, &fl6, NULL);
629 if (IS_ERR(dst))
630 return ERR_CAST(dst);
631
632 dev = dst->dev;
633 dev_hold(dev);
634 dst_release(dst);
635
636 return dev;
637 }
638 #else
inet6_fib_lookup_dev(struct net * net,const void * addr)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
find_outdev(struct net * net,struct mpls_route * rt,struct mpls_nh * nh,int oif)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
mpls_nh_assign_dev(struct net * net,struct mpls_route * rt,struct mpls_nh * nh,int oif)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
nla_get_via(const struct nlattr * nla,u8 * via_alen,u8 * via_table,u8 via_addr[],struct netlink_ext_ack * extack)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
mpls_nh_build_from_cfg(struct mpls_route_config * cfg,struct mpls_route * rt)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
mpls_nh_build(struct net * net,struct mpls_route * rt,struct mpls_nh * nh,int oif,struct nlattr * via,struct nlattr * newdst,u8 max_labels,struct netlink_ext_ack * extack)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
mpls_count_nexthops(struct rtnexthop * rtnh,int len,u8 cfg_via_alen,u8 * max_via_alen,u8 * max_labels)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
mpls_nh_build_multi(struct mpls_route_config * cfg,struct mpls_route * rt,u8 max_labels,struct netlink_ext_ack * extack)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
mpls_label_ok(struct net * net,unsigned int * index,struct netlink_ext_ack * extack)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
mpls_route_add(struct mpls_route_config * cfg,struct netlink_ext_ack * extack)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 rt = mpls_rt_alloc(nhs, max_via_alen, max_labels);
1023 if (IS_ERR(rt)) {
1024 err = PTR_ERR(rt);
1025 goto errout;
1026 }
1027
1028 rt->rt_protocol = cfg->rc_protocol;
1029 rt->rt_payload_type = cfg->rc_payload_type;
1030 rt->rt_ttl_propagate = cfg->rc_ttl_propagate;
1031
1032 if (cfg->rc_mp)
1033 err = mpls_nh_build_multi(cfg, rt, max_labels, extack);
1034 else
1035 err = mpls_nh_build_from_cfg(cfg, rt);
1036 if (err)
1037 goto freert;
1038
1039 mpls_route_update(net, index, rt, &cfg->rc_nlinfo);
1040
1041 return 0;
1042
1043 freert:
1044 mpls_rt_free(rt);
1045 errout:
1046 return err;
1047 }
1048
mpls_route_del(struct mpls_route_config * cfg,struct netlink_ext_ack * extack)1049 static int mpls_route_del(struct mpls_route_config *cfg,
1050 struct netlink_ext_ack *extack)
1051 {
1052 struct net *net = cfg->rc_nlinfo.nl_net;
1053 unsigned index;
1054 int err = -EINVAL;
1055
1056 index = cfg->rc_label;
1057
1058 if (!mpls_label_ok(net, &index, extack))
1059 goto errout;
1060
1061 mpls_route_update(net, index, NULL, &cfg->rc_nlinfo);
1062
1063 err = 0;
1064 errout:
1065 return err;
1066 }
1067
mpls_get_stats(struct mpls_dev * mdev,struct mpls_link_stats * stats)1068 static void mpls_get_stats(struct mpls_dev *mdev,
1069 struct mpls_link_stats *stats)
1070 {
1071 struct mpls_pcpu_stats *p;
1072 int i;
1073
1074 memset(stats, 0, sizeof(*stats));
1075
1076 for_each_possible_cpu(i) {
1077 struct mpls_link_stats local;
1078 unsigned int start;
1079
1080 p = per_cpu_ptr(mdev->stats, i);
1081 do {
1082 start = u64_stats_fetch_begin(&p->syncp);
1083 local = p->stats;
1084 } while (u64_stats_fetch_retry(&p->syncp, start));
1085
1086 stats->rx_packets += local.rx_packets;
1087 stats->rx_bytes += local.rx_bytes;
1088 stats->tx_packets += local.tx_packets;
1089 stats->tx_bytes += local.tx_bytes;
1090 stats->rx_errors += local.rx_errors;
1091 stats->tx_errors += local.tx_errors;
1092 stats->rx_dropped += local.rx_dropped;
1093 stats->tx_dropped += local.tx_dropped;
1094 stats->rx_noroute += local.rx_noroute;
1095 }
1096 }
1097
mpls_fill_stats_af(struct sk_buff * skb,const struct net_device * dev)1098 static int mpls_fill_stats_af(struct sk_buff *skb,
1099 const struct net_device *dev)
1100 {
1101 struct mpls_link_stats *stats;
1102 struct mpls_dev *mdev;
1103 struct nlattr *nla;
1104
1105 mdev = mpls_dev_get(dev);
1106 if (!mdev)
1107 return -ENODATA;
1108
1109 nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
1110 sizeof(struct mpls_link_stats),
1111 MPLS_STATS_UNSPEC);
1112 if (!nla)
1113 return -EMSGSIZE;
1114
1115 stats = nla_data(nla);
1116 mpls_get_stats(mdev, stats);
1117
1118 return 0;
1119 }
1120
mpls_get_stats_af_size(const struct net_device * dev)1121 static size_t mpls_get_stats_af_size(const struct net_device *dev)
1122 {
1123 struct mpls_dev *mdev;
1124
1125 mdev = mpls_dev_get(dev);
1126 if (!mdev)
1127 return 0;
1128
1129 return nla_total_size_64bit(sizeof(struct mpls_link_stats));
1130 }
1131
mpls_netconf_fill_devconf(struct sk_buff * skb,struct mpls_dev * mdev,u32 portid,u32 seq,int event,unsigned int flags,int type)1132 static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev,
1133 u32 portid, u32 seq, int event,
1134 unsigned int flags, int type)
1135 {
1136 struct nlmsghdr *nlh;
1137 struct netconfmsg *ncm;
1138 bool all = false;
1139
1140 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1141 flags);
1142 if (!nlh)
1143 return -EMSGSIZE;
1144
1145 if (type == NETCONFA_ALL)
1146 all = true;
1147
1148 ncm = nlmsg_data(nlh);
1149 ncm->ncm_family = AF_MPLS;
1150
1151 if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0)
1152 goto nla_put_failure;
1153
1154 if ((all || type == NETCONFA_INPUT) &&
1155 nla_put_s32(skb, NETCONFA_INPUT,
1156 mdev->input_enabled) < 0)
1157 goto nla_put_failure;
1158
1159 nlmsg_end(skb, nlh);
1160 return 0;
1161
1162 nla_put_failure:
1163 nlmsg_cancel(skb, nlh);
1164 return -EMSGSIZE;
1165 }
1166
mpls_netconf_msgsize_devconf(int type)1167 static int mpls_netconf_msgsize_devconf(int type)
1168 {
1169 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1170 + nla_total_size(4); /* NETCONFA_IFINDEX */
1171 bool all = false;
1172
1173 if (type == NETCONFA_ALL)
1174 all = true;
1175
1176 if (all || type == NETCONFA_INPUT)
1177 size += nla_total_size(4);
1178
1179 return size;
1180 }
1181
mpls_netconf_notify_devconf(struct net * net,int event,int type,struct mpls_dev * mdev)1182 static void mpls_netconf_notify_devconf(struct net *net, int event,
1183 int type, struct mpls_dev *mdev)
1184 {
1185 struct sk_buff *skb;
1186 int err = -ENOBUFS;
1187
1188 skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL);
1189 if (!skb)
1190 goto errout;
1191
1192 err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type);
1193 if (err < 0) {
1194 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1195 WARN_ON(err == -EMSGSIZE);
1196 kfree_skb(skb);
1197 goto errout;
1198 }
1199
1200 rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL);
1201 return;
1202 errout:
1203 if (err < 0)
1204 rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err);
1205 }
1206
1207 static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = {
1208 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1209 };
1210
mpls_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)1211 static int mpls_netconf_valid_get_req(struct sk_buff *skb,
1212 const struct nlmsghdr *nlh,
1213 struct nlattr **tb,
1214 struct netlink_ext_ack *extack)
1215 {
1216 int i, err;
1217
1218 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
1219 NL_SET_ERR_MSG_MOD(extack,
1220 "Invalid header for netconf get request");
1221 return -EINVAL;
1222 }
1223
1224 if (!netlink_strict_get_check(skb))
1225 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
1226 tb, NETCONFA_MAX,
1227 devconf_mpls_policy, extack);
1228
1229 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
1230 tb, NETCONFA_MAX,
1231 devconf_mpls_policy, extack);
1232 if (err)
1233 return err;
1234
1235 for (i = 0; i <= NETCONFA_MAX; i++) {
1236 if (!tb[i])
1237 continue;
1238
1239 switch (i) {
1240 case NETCONFA_IFINDEX:
1241 break;
1242 default:
1243 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
1244 return -EINVAL;
1245 }
1246 }
1247
1248 return 0;
1249 }
1250
mpls_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)1251 static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
1252 struct nlmsghdr *nlh,
1253 struct netlink_ext_ack *extack)
1254 {
1255 struct net *net = sock_net(in_skb->sk);
1256 struct nlattr *tb[NETCONFA_MAX + 1];
1257 struct net_device *dev;
1258 struct mpls_dev *mdev;
1259 struct sk_buff *skb;
1260 int ifindex;
1261 int err;
1262
1263 err = mpls_netconf_valid_get_req(in_skb, nlh, tb, extack);
1264 if (err < 0)
1265 goto errout;
1266
1267 err = -EINVAL;
1268 if (!tb[NETCONFA_IFINDEX])
1269 goto errout;
1270
1271 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1272 dev = __dev_get_by_index(net, ifindex);
1273 if (!dev)
1274 goto errout;
1275
1276 mdev = mpls_dev_get(dev);
1277 if (!mdev)
1278 goto errout;
1279
1280 err = -ENOBUFS;
1281 skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1282 if (!skb)
1283 goto errout;
1284
1285 err = mpls_netconf_fill_devconf(skb, mdev,
1286 NETLINK_CB(in_skb).portid,
1287 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1288 NETCONFA_ALL);
1289 if (err < 0) {
1290 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1291 WARN_ON(err == -EMSGSIZE);
1292 kfree_skb(skb);
1293 goto errout;
1294 }
1295 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1296 errout:
1297 return err;
1298 }
1299
mpls_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)1300 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
1301 struct netlink_callback *cb)
1302 {
1303 const struct nlmsghdr *nlh = cb->nlh;
1304 struct net *net = sock_net(skb->sk);
1305 struct hlist_head *head;
1306 struct net_device *dev;
1307 struct mpls_dev *mdev;
1308 int idx, s_idx;
1309 int h, s_h;
1310
1311 if (cb->strict_check) {
1312 struct netlink_ext_ack *extack = cb->extack;
1313 struct netconfmsg *ncm;
1314
1315 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
1316 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
1317 return -EINVAL;
1318 }
1319
1320 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
1321 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
1322 return -EINVAL;
1323 }
1324 }
1325
1326 s_h = cb->args[0];
1327 s_idx = idx = cb->args[1];
1328
1329 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1330 idx = 0;
1331 head = &net->dev_index_head[h];
1332 rcu_read_lock();
1333 cb->seq = net->dev_base_seq;
1334 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1335 if (idx < s_idx)
1336 goto cont;
1337 mdev = mpls_dev_get(dev);
1338 if (!mdev)
1339 goto cont;
1340 if (mpls_netconf_fill_devconf(skb, mdev,
1341 NETLINK_CB(cb->skb).portid,
1342 nlh->nlmsg_seq,
1343 RTM_NEWNETCONF,
1344 NLM_F_MULTI,
1345 NETCONFA_ALL) < 0) {
1346 rcu_read_unlock();
1347 goto done;
1348 }
1349 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1350 cont:
1351 idx++;
1352 }
1353 rcu_read_unlock();
1354 }
1355 done:
1356 cb->args[0] = h;
1357 cb->args[1] = idx;
1358
1359 return skb->len;
1360 }
1361
1362 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \
1363 (&((struct mpls_dev *)0)->field)
1364
mpls_conf_proc(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)1365 static int mpls_conf_proc(struct ctl_table *ctl, int write,
1366 void *buffer, size_t *lenp, loff_t *ppos)
1367 {
1368 int oval = *(int *)ctl->data;
1369 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1370
1371 if (write) {
1372 struct mpls_dev *mdev = ctl->extra1;
1373 int i = (int *)ctl->data - (int *)mdev;
1374 struct net *net = ctl->extra2;
1375 int val = *(int *)ctl->data;
1376
1377 if (i == offsetof(struct mpls_dev, input_enabled) &&
1378 val != oval) {
1379 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1380 NETCONFA_INPUT, mdev);
1381 }
1382 }
1383
1384 return ret;
1385 }
1386
1387 static const struct ctl_table mpls_dev_table[] = {
1388 {
1389 .procname = "input",
1390 .maxlen = sizeof(int),
1391 .mode = 0644,
1392 .proc_handler = mpls_conf_proc,
1393 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1394 },
1395 { }
1396 };
1397
mpls_dev_sysctl_register(struct net_device * dev,struct mpls_dev * mdev)1398 static int mpls_dev_sysctl_register(struct net_device *dev,
1399 struct mpls_dev *mdev)
1400 {
1401 char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1402 struct net *net = dev_net(dev);
1403 struct ctl_table *table;
1404 int i;
1405
1406 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1407 if (!table)
1408 goto out;
1409
1410 /* Table data contains only offsets relative to the base of
1411 * the mdev at this point, so make them absolute.
1412 */
1413 for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) {
1414 table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1415 table[i].extra1 = mdev;
1416 table[i].extra2 = net;
1417 }
1418
1419 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1420
1421 mdev->sysctl = register_net_sysctl(net, path, table);
1422 if (!mdev->sysctl)
1423 goto free;
1424
1425 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1426 return 0;
1427
1428 free:
1429 kfree(table);
1430 out:
1431 return -ENOBUFS;
1432 }
1433
mpls_dev_sysctl_unregister(struct net_device * dev,struct mpls_dev * mdev)1434 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1435 struct mpls_dev *mdev)
1436 {
1437 struct net *net = dev_net(dev);
1438 struct ctl_table *table;
1439
1440 table = mdev->sysctl->ctl_table_arg;
1441 unregister_net_sysctl_table(mdev->sysctl);
1442 kfree(table);
1443
1444 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1445 }
1446
mpls_add_dev(struct net_device * dev)1447 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1448 {
1449 struct mpls_dev *mdev;
1450 int err = -ENOMEM;
1451 int i;
1452
1453 ASSERT_RTNL();
1454
1455 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1456 if (!mdev)
1457 return ERR_PTR(err);
1458
1459 mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1460 if (!mdev->stats)
1461 goto free;
1462
1463 for_each_possible_cpu(i) {
1464 struct mpls_pcpu_stats *mpls_stats;
1465
1466 mpls_stats = per_cpu_ptr(mdev->stats, i);
1467 u64_stats_init(&mpls_stats->syncp);
1468 }
1469
1470 mdev->dev = dev;
1471
1472 err = mpls_dev_sysctl_register(dev, mdev);
1473 if (err)
1474 goto free;
1475
1476 rcu_assign_pointer(dev->mpls_ptr, mdev);
1477
1478 return mdev;
1479
1480 free:
1481 free_percpu(mdev->stats);
1482 kfree(mdev);
1483 return ERR_PTR(err);
1484 }
1485
mpls_dev_destroy_rcu(struct rcu_head * head)1486 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1487 {
1488 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1489
1490 free_percpu(mdev->stats);
1491 kfree(mdev);
1492 }
1493
mpls_ifdown(struct net_device * dev,int event)1494 static void mpls_ifdown(struct net_device *dev, int event)
1495 {
1496 struct mpls_route __rcu **platform_label;
1497 struct net *net = dev_net(dev);
1498 u8 alive, deleted;
1499 unsigned index;
1500
1501 platform_label = rtnl_dereference(net->mpls.platform_label);
1502 for (index = 0; index < net->mpls.platform_labels; index++) {
1503 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1504
1505 if (!rt)
1506 continue;
1507
1508 alive = 0;
1509 deleted = 0;
1510 change_nexthops(rt) {
1511 unsigned int nh_flags = nh->nh_flags;
1512
1513 if (rtnl_dereference(nh->nh_dev) != dev)
1514 goto next;
1515
1516 switch (event) {
1517 case NETDEV_DOWN:
1518 case NETDEV_UNREGISTER:
1519 nh_flags |= RTNH_F_DEAD;
1520 fallthrough;
1521 case NETDEV_CHANGE:
1522 nh_flags |= RTNH_F_LINKDOWN;
1523 break;
1524 }
1525 if (event == NETDEV_UNREGISTER)
1526 RCU_INIT_POINTER(nh->nh_dev, NULL);
1527
1528 if (nh->nh_flags != nh_flags)
1529 WRITE_ONCE(nh->nh_flags, nh_flags);
1530 next:
1531 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1532 alive++;
1533 if (!rtnl_dereference(nh->nh_dev))
1534 deleted++;
1535 } endfor_nexthops(rt);
1536
1537 WRITE_ONCE(rt->rt_nhn_alive, alive);
1538
1539 /* if there are no more nexthops, delete the route */
1540 if (event == NETDEV_UNREGISTER && deleted == rt->rt_nhn)
1541 mpls_route_update(net, index, NULL, NULL);
1542 }
1543 }
1544
mpls_ifup(struct net_device * dev,unsigned int flags)1545 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1546 {
1547 struct mpls_route __rcu **platform_label;
1548 struct net *net = dev_net(dev);
1549 unsigned index;
1550 u8 alive;
1551
1552 platform_label = rtnl_dereference(net->mpls.platform_label);
1553 for (index = 0; index < net->mpls.platform_labels; index++) {
1554 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1555
1556 if (!rt)
1557 continue;
1558
1559 alive = 0;
1560 change_nexthops(rt) {
1561 unsigned int nh_flags = nh->nh_flags;
1562 struct net_device *nh_dev =
1563 rtnl_dereference(nh->nh_dev);
1564
1565 if (!(nh_flags & flags)) {
1566 alive++;
1567 continue;
1568 }
1569 if (nh_dev != dev)
1570 continue;
1571 alive++;
1572 nh_flags &= ~flags;
1573 WRITE_ONCE(nh->nh_flags, nh_flags);
1574 } endfor_nexthops(rt);
1575
1576 WRITE_ONCE(rt->rt_nhn_alive, alive);
1577 }
1578 }
1579
mpls_dev_notify(struct notifier_block * this,unsigned long event,void * ptr)1580 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1581 void *ptr)
1582 {
1583 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1584 struct mpls_dev *mdev;
1585 unsigned int flags;
1586
1587 if (event == NETDEV_REGISTER) {
1588 mdev = mpls_add_dev(dev);
1589 if (IS_ERR(mdev))
1590 return notifier_from_errno(PTR_ERR(mdev));
1591
1592 return NOTIFY_OK;
1593 }
1594
1595 mdev = mpls_dev_get(dev);
1596 if (!mdev)
1597 return NOTIFY_OK;
1598
1599 switch (event) {
1600 case NETDEV_DOWN:
1601 mpls_ifdown(dev, event);
1602 break;
1603 case NETDEV_UP:
1604 flags = dev_get_flags(dev);
1605 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1606 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1607 else
1608 mpls_ifup(dev, RTNH_F_DEAD);
1609 break;
1610 case NETDEV_CHANGE:
1611 flags = dev_get_flags(dev);
1612 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1613 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1614 else
1615 mpls_ifdown(dev, event);
1616 break;
1617 case NETDEV_UNREGISTER:
1618 mpls_ifdown(dev, event);
1619 mdev = mpls_dev_get(dev);
1620 if (mdev) {
1621 mpls_dev_sysctl_unregister(dev, mdev);
1622 RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1623 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1624 }
1625 break;
1626 case NETDEV_CHANGENAME:
1627 mdev = mpls_dev_get(dev);
1628 if (mdev) {
1629 int err;
1630
1631 mpls_dev_sysctl_unregister(dev, mdev);
1632 err = mpls_dev_sysctl_register(dev, mdev);
1633 if (err)
1634 return notifier_from_errno(err);
1635 }
1636 break;
1637 }
1638 return NOTIFY_OK;
1639 }
1640
1641 static struct notifier_block mpls_dev_notifier = {
1642 .notifier_call = mpls_dev_notify,
1643 };
1644
nla_put_via(struct sk_buff * skb,u8 table,const void * addr,int alen)1645 static int nla_put_via(struct sk_buff *skb,
1646 u8 table, const void *addr, int alen)
1647 {
1648 static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1649 AF_INET, AF_INET6, AF_DECnet, AF_PACKET,
1650 };
1651 struct nlattr *nla;
1652 struct rtvia *via;
1653 int family = AF_UNSPEC;
1654
1655 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1656 if (!nla)
1657 return -EMSGSIZE;
1658
1659 if (table <= NEIGH_NR_TABLES)
1660 family = table_to_family[table];
1661
1662 via = nla_data(nla);
1663 via->rtvia_family = family;
1664 memcpy(via->rtvia_addr, addr, alen);
1665 return 0;
1666 }
1667
nla_put_labels(struct sk_buff * skb,int attrtype,u8 labels,const u32 label[])1668 int nla_put_labels(struct sk_buff *skb, int attrtype,
1669 u8 labels, const u32 label[])
1670 {
1671 struct nlattr *nla;
1672 struct mpls_shim_hdr *nla_label;
1673 bool bos;
1674 int i;
1675 nla = nla_reserve(skb, attrtype, labels*4);
1676 if (!nla)
1677 return -EMSGSIZE;
1678
1679 nla_label = nla_data(nla);
1680 bos = true;
1681 for (i = labels - 1; i >= 0; i--) {
1682 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1683 bos = false;
1684 }
1685
1686 return 0;
1687 }
1688 EXPORT_SYMBOL_GPL(nla_put_labels);
1689
nla_get_labels(const struct nlattr * nla,u8 max_labels,u8 * labels,u32 label[],struct netlink_ext_ack * extack)1690 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1691 u32 label[], struct netlink_ext_ack *extack)
1692 {
1693 unsigned len = nla_len(nla);
1694 struct mpls_shim_hdr *nla_label;
1695 u8 nla_labels;
1696 bool bos;
1697 int i;
1698
1699 /* len needs to be an even multiple of 4 (the label size). Number
1700 * of labels is a u8 so check for overflow.
1701 */
1702 if (len & 3 || len / 4 > 255) {
1703 NL_SET_ERR_MSG_ATTR(extack, nla,
1704 "Invalid length for labels attribute");
1705 return -EINVAL;
1706 }
1707
1708 /* Limit the number of new labels allowed */
1709 nla_labels = len/4;
1710 if (nla_labels > max_labels) {
1711 NL_SET_ERR_MSG(extack, "Too many labels");
1712 return -EINVAL;
1713 }
1714
1715 /* when label == NULL, caller wants number of labels */
1716 if (!label)
1717 goto out;
1718
1719 nla_label = nla_data(nla);
1720 bos = true;
1721 for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1722 struct mpls_entry_decoded dec;
1723 dec = mpls_entry_decode(nla_label + i);
1724
1725 /* Ensure the bottom of stack flag is properly set
1726 * and ttl and tc are both clear.
1727 */
1728 if (dec.ttl) {
1729 NL_SET_ERR_MSG_ATTR(extack, nla,
1730 "TTL in label must be 0");
1731 return -EINVAL;
1732 }
1733
1734 if (dec.tc) {
1735 NL_SET_ERR_MSG_ATTR(extack, nla,
1736 "Traffic class in label must be 0");
1737 return -EINVAL;
1738 }
1739
1740 if (dec.bos != bos) {
1741 NL_SET_BAD_ATTR(extack, nla);
1742 if (bos) {
1743 NL_SET_ERR_MSG(extack,
1744 "BOS bit must be set in first label");
1745 } else {
1746 NL_SET_ERR_MSG(extack,
1747 "BOS bit can only be set in first label");
1748 }
1749 return -EINVAL;
1750 }
1751
1752 switch (dec.label) {
1753 case MPLS_LABEL_IMPLNULL:
1754 /* RFC3032: This is a label that an LSR may
1755 * assign and distribute, but which never
1756 * actually appears in the encapsulation.
1757 */
1758 NL_SET_ERR_MSG_ATTR(extack, nla,
1759 "Implicit NULL Label (3) can not be used in encapsulation");
1760 return -EINVAL;
1761 }
1762
1763 label[i] = dec.label;
1764 }
1765 out:
1766 *labels = nla_labels;
1767 return 0;
1768 }
1769 EXPORT_SYMBOL_GPL(nla_get_labels);
1770
rtm_to_route_config(struct sk_buff * skb,struct nlmsghdr * nlh,struct mpls_route_config * cfg,struct netlink_ext_ack * extack)1771 static int rtm_to_route_config(struct sk_buff *skb,
1772 struct nlmsghdr *nlh,
1773 struct mpls_route_config *cfg,
1774 struct netlink_ext_ack *extack)
1775 {
1776 struct rtmsg *rtm;
1777 struct nlattr *tb[RTA_MAX+1];
1778 int index;
1779 int err;
1780
1781 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
1782 rtm_mpls_policy, extack);
1783 if (err < 0)
1784 goto errout;
1785
1786 err = -EINVAL;
1787 rtm = nlmsg_data(nlh);
1788
1789 if (rtm->rtm_family != AF_MPLS) {
1790 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1791 goto errout;
1792 }
1793 if (rtm->rtm_dst_len != 20) {
1794 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1795 goto errout;
1796 }
1797 if (rtm->rtm_src_len != 0) {
1798 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1799 goto errout;
1800 }
1801 if (rtm->rtm_tos != 0) {
1802 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1803 goto errout;
1804 }
1805 if (rtm->rtm_table != RT_TABLE_MAIN) {
1806 NL_SET_ERR_MSG(extack,
1807 "MPLS only supports the main route table");
1808 goto errout;
1809 }
1810 /* Any value is acceptable for rtm_protocol */
1811
1812 /* As mpls uses destination specific addresses
1813 * (or source specific address in the case of multicast)
1814 * all addresses have universal scope.
1815 */
1816 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1817 NL_SET_ERR_MSG(extack,
1818 "Invalid route scope - MPLS only supports UNIVERSE");
1819 goto errout;
1820 }
1821 if (rtm->rtm_type != RTN_UNICAST) {
1822 NL_SET_ERR_MSG(extack,
1823 "Invalid route type - MPLS only supports UNICAST");
1824 goto errout;
1825 }
1826 if (rtm->rtm_flags != 0) {
1827 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1828 goto errout;
1829 }
1830
1831 cfg->rc_label = LABEL_NOT_SPECIFIED;
1832 cfg->rc_protocol = rtm->rtm_protocol;
1833 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC;
1834 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
1835 cfg->rc_nlflags = nlh->nlmsg_flags;
1836 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid;
1837 cfg->rc_nlinfo.nlh = nlh;
1838 cfg->rc_nlinfo.nl_net = sock_net(skb->sk);
1839
1840 for (index = 0; index <= RTA_MAX; index++) {
1841 struct nlattr *nla = tb[index];
1842 if (!nla)
1843 continue;
1844
1845 switch (index) {
1846 case RTA_OIF:
1847 cfg->rc_ifindex = nla_get_u32(nla);
1848 break;
1849 case RTA_NEWDST:
1850 if (nla_get_labels(nla, MAX_NEW_LABELS,
1851 &cfg->rc_output_labels,
1852 cfg->rc_output_label, extack))
1853 goto errout;
1854 break;
1855 case RTA_DST:
1856 {
1857 u8 label_count;
1858 if (nla_get_labels(nla, 1, &label_count,
1859 &cfg->rc_label, extack))
1860 goto errout;
1861
1862 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1863 &cfg->rc_label, extack))
1864 goto errout;
1865 break;
1866 }
1867 case RTA_GATEWAY:
1868 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1869 goto errout;
1870 case RTA_VIA:
1871 {
1872 if (nla_get_via(nla, &cfg->rc_via_alen,
1873 &cfg->rc_via_table, cfg->rc_via,
1874 extack))
1875 goto errout;
1876 break;
1877 }
1878 case RTA_MULTIPATH:
1879 {
1880 cfg->rc_mp = nla_data(nla);
1881 cfg->rc_mp_len = nla_len(nla);
1882 break;
1883 }
1884 case RTA_TTL_PROPAGATE:
1885 {
1886 u8 ttl_propagate = nla_get_u8(nla);
1887
1888 if (ttl_propagate > 1) {
1889 NL_SET_ERR_MSG_ATTR(extack, nla,
1890 "RTA_TTL_PROPAGATE can only be 0 or 1");
1891 goto errout;
1892 }
1893 cfg->rc_ttl_propagate = ttl_propagate ?
1894 MPLS_TTL_PROP_ENABLED :
1895 MPLS_TTL_PROP_DISABLED;
1896 break;
1897 }
1898 default:
1899 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1900 /* Unsupported attribute */
1901 goto errout;
1902 }
1903 }
1904
1905 err = 0;
1906 errout:
1907 return err;
1908 }
1909
mpls_rtm_delroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)1910 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1911 struct netlink_ext_ack *extack)
1912 {
1913 struct mpls_route_config *cfg;
1914 int err;
1915
1916 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1917 if (!cfg)
1918 return -ENOMEM;
1919
1920 err = rtm_to_route_config(skb, nlh, cfg, extack);
1921 if (err < 0)
1922 goto out;
1923
1924 err = mpls_route_del(cfg, extack);
1925 out:
1926 kfree(cfg);
1927
1928 return err;
1929 }
1930
1931
mpls_rtm_newroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)1932 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1933 struct netlink_ext_ack *extack)
1934 {
1935 struct mpls_route_config *cfg;
1936 int err;
1937
1938 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1939 if (!cfg)
1940 return -ENOMEM;
1941
1942 err = rtm_to_route_config(skb, nlh, cfg, extack);
1943 if (err < 0)
1944 goto out;
1945
1946 err = mpls_route_add(cfg, extack);
1947 out:
1948 kfree(cfg);
1949
1950 return err;
1951 }
1952
mpls_dump_route(struct sk_buff * skb,u32 portid,u32 seq,int event,u32 label,struct mpls_route * rt,int flags)1953 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1954 u32 label, struct mpls_route *rt, int flags)
1955 {
1956 struct net_device *dev;
1957 struct nlmsghdr *nlh;
1958 struct rtmsg *rtm;
1959
1960 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1961 if (nlh == NULL)
1962 return -EMSGSIZE;
1963
1964 rtm = nlmsg_data(nlh);
1965 rtm->rtm_family = AF_MPLS;
1966 rtm->rtm_dst_len = 20;
1967 rtm->rtm_src_len = 0;
1968 rtm->rtm_tos = 0;
1969 rtm->rtm_table = RT_TABLE_MAIN;
1970 rtm->rtm_protocol = rt->rt_protocol;
1971 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
1972 rtm->rtm_type = RTN_UNICAST;
1973 rtm->rtm_flags = 0;
1974
1975 if (nla_put_labels(skb, RTA_DST, 1, &label))
1976 goto nla_put_failure;
1977
1978 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
1979 bool ttl_propagate =
1980 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
1981
1982 if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
1983 ttl_propagate))
1984 goto nla_put_failure;
1985 }
1986 if (rt->rt_nhn == 1) {
1987 const struct mpls_nh *nh = rt->rt_nh;
1988
1989 if (nh->nh_labels &&
1990 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
1991 nh->nh_label))
1992 goto nla_put_failure;
1993 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
1994 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
1995 nh->nh_via_alen))
1996 goto nla_put_failure;
1997 dev = rtnl_dereference(nh->nh_dev);
1998 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
1999 goto nla_put_failure;
2000 if (nh->nh_flags & RTNH_F_LINKDOWN)
2001 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2002 if (nh->nh_flags & RTNH_F_DEAD)
2003 rtm->rtm_flags |= RTNH_F_DEAD;
2004 } else {
2005 struct rtnexthop *rtnh;
2006 struct nlattr *mp;
2007 u8 linkdown = 0;
2008 u8 dead = 0;
2009
2010 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
2011 if (!mp)
2012 goto nla_put_failure;
2013
2014 for_nexthops(rt) {
2015 dev = rtnl_dereference(nh->nh_dev);
2016 if (!dev)
2017 continue;
2018
2019 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
2020 if (!rtnh)
2021 goto nla_put_failure;
2022
2023 rtnh->rtnh_ifindex = dev->ifindex;
2024 if (nh->nh_flags & RTNH_F_LINKDOWN) {
2025 rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
2026 linkdown++;
2027 }
2028 if (nh->nh_flags & RTNH_F_DEAD) {
2029 rtnh->rtnh_flags |= RTNH_F_DEAD;
2030 dead++;
2031 }
2032
2033 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
2034 nh->nh_labels,
2035 nh->nh_label))
2036 goto nla_put_failure;
2037 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2038 nla_put_via(skb, nh->nh_via_table,
2039 mpls_nh_via(rt, nh),
2040 nh->nh_via_alen))
2041 goto nla_put_failure;
2042
2043 /* length of rtnetlink header + attributes */
2044 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2045 } endfor_nexthops(rt);
2046
2047 if (linkdown == rt->rt_nhn)
2048 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2049 if (dead == rt->rt_nhn)
2050 rtm->rtm_flags |= RTNH_F_DEAD;
2051
2052 nla_nest_end(skb, mp);
2053 }
2054
2055 nlmsg_end(skb, nlh);
2056 return 0;
2057
2058 nla_put_failure:
2059 nlmsg_cancel(skb, nlh);
2060 return -EMSGSIZE;
2061 }
2062
2063 #if IS_ENABLED(CONFIG_INET)
mpls_valid_fib_dump_req(struct net * net,const struct nlmsghdr * nlh,struct fib_dump_filter * filter,struct netlink_callback * cb)2064 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2065 struct fib_dump_filter *filter,
2066 struct netlink_callback *cb)
2067 {
2068 return ip_valid_fib_dump_req(net, nlh, filter, cb);
2069 }
2070 #else
mpls_valid_fib_dump_req(struct net * net,const struct nlmsghdr * nlh,struct fib_dump_filter * filter,struct netlink_callback * cb)2071 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2072 struct fib_dump_filter *filter,
2073 struct netlink_callback *cb)
2074 {
2075 struct netlink_ext_ack *extack = cb->extack;
2076 struct nlattr *tb[RTA_MAX + 1];
2077 struct rtmsg *rtm;
2078 int err, i;
2079
2080 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2081 NL_SET_ERR_MSG_MOD(extack, "Invalid header for FIB dump request");
2082 return -EINVAL;
2083 }
2084
2085 rtm = nlmsg_data(nlh);
2086 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos ||
2087 rtm->rtm_table || rtm->rtm_scope || rtm->rtm_type ||
2088 rtm->rtm_flags) {
2089 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for FIB dump request");
2090 return -EINVAL;
2091 }
2092
2093 if (rtm->rtm_protocol) {
2094 filter->protocol = rtm->rtm_protocol;
2095 filter->filter_set = 1;
2096 cb->answer_flags = NLM_F_DUMP_FILTERED;
2097 }
2098
2099 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2100 rtm_mpls_policy, extack);
2101 if (err < 0)
2102 return err;
2103
2104 for (i = 0; i <= RTA_MAX; ++i) {
2105 int ifindex;
2106
2107 if (i == RTA_OIF) {
2108 ifindex = nla_get_u32(tb[i]);
2109 filter->dev = __dev_get_by_index(net, ifindex);
2110 if (!filter->dev)
2111 return -ENODEV;
2112 filter->filter_set = 1;
2113 } else if (tb[i]) {
2114 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
2115 return -EINVAL;
2116 }
2117 }
2118
2119 return 0;
2120 }
2121 #endif
2122
mpls_rt_uses_dev(struct mpls_route * rt,const struct net_device * dev)2123 static bool mpls_rt_uses_dev(struct mpls_route *rt,
2124 const struct net_device *dev)
2125 {
2126 struct net_device *nh_dev;
2127
2128 if (rt->rt_nhn == 1) {
2129 struct mpls_nh *nh = rt->rt_nh;
2130
2131 nh_dev = rtnl_dereference(nh->nh_dev);
2132 if (dev == nh_dev)
2133 return true;
2134 } else {
2135 for_nexthops(rt) {
2136 nh_dev = rtnl_dereference(nh->nh_dev);
2137 if (nh_dev == dev)
2138 return true;
2139 } endfor_nexthops(rt);
2140 }
2141
2142 return false;
2143 }
2144
mpls_dump_routes(struct sk_buff * skb,struct netlink_callback * cb)2145 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2146 {
2147 const struct nlmsghdr *nlh = cb->nlh;
2148 struct net *net = sock_net(skb->sk);
2149 struct mpls_route __rcu **platform_label;
2150 struct fib_dump_filter filter = {};
2151 unsigned int flags = NLM_F_MULTI;
2152 size_t platform_labels;
2153 unsigned int index;
2154
2155 ASSERT_RTNL();
2156
2157 if (cb->strict_check) {
2158 int err;
2159
2160 err = mpls_valid_fib_dump_req(net, nlh, &filter, cb);
2161 if (err < 0)
2162 return err;
2163
2164 /* for MPLS, there is only 1 table with fixed type and flags.
2165 * If either are set in the filter then return nothing.
2166 */
2167 if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) ||
2168 (filter.rt_type && filter.rt_type != RTN_UNICAST) ||
2169 filter.flags)
2170 return skb->len;
2171 }
2172
2173 index = cb->args[0];
2174 if (index < MPLS_LABEL_FIRST_UNRESERVED)
2175 index = MPLS_LABEL_FIRST_UNRESERVED;
2176
2177 platform_label = rtnl_dereference(net->mpls.platform_label);
2178 platform_labels = net->mpls.platform_labels;
2179
2180 if (filter.filter_set)
2181 flags |= NLM_F_DUMP_FILTERED;
2182
2183 for (; index < platform_labels; index++) {
2184 struct mpls_route *rt;
2185
2186 rt = rtnl_dereference(platform_label[index]);
2187 if (!rt)
2188 continue;
2189
2190 if ((filter.dev && !mpls_rt_uses_dev(rt, filter.dev)) ||
2191 (filter.protocol && rt->rt_protocol != filter.protocol))
2192 continue;
2193
2194 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2195 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2196 index, rt, flags) < 0)
2197 break;
2198 }
2199 cb->args[0] = index;
2200
2201 return skb->len;
2202 }
2203
lfib_nlmsg_size(struct mpls_route * rt)2204 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2205 {
2206 size_t payload =
2207 NLMSG_ALIGN(sizeof(struct rtmsg))
2208 + nla_total_size(4) /* RTA_DST */
2209 + nla_total_size(1); /* RTA_TTL_PROPAGATE */
2210
2211 if (rt->rt_nhn == 1) {
2212 struct mpls_nh *nh = rt->rt_nh;
2213
2214 if (nh->nh_dev)
2215 payload += nla_total_size(4); /* RTA_OIF */
2216 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2217 payload += nla_total_size(2 + nh->nh_via_alen);
2218 if (nh->nh_labels) /* RTA_NEWDST */
2219 payload += nla_total_size(nh->nh_labels * 4);
2220 } else {
2221 /* each nexthop is packed in an attribute */
2222 size_t nhsize = 0;
2223
2224 for_nexthops(rt) {
2225 if (!rtnl_dereference(nh->nh_dev))
2226 continue;
2227 nhsize += nla_total_size(sizeof(struct rtnexthop));
2228 /* RTA_VIA */
2229 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2230 nhsize += nla_total_size(2 + nh->nh_via_alen);
2231 if (nh->nh_labels)
2232 nhsize += nla_total_size(nh->nh_labels * 4);
2233 } endfor_nexthops(rt);
2234 /* nested attribute */
2235 payload += nla_total_size(nhsize);
2236 }
2237
2238 return payload;
2239 }
2240
rtmsg_lfib(int event,u32 label,struct mpls_route * rt,struct nlmsghdr * nlh,struct net * net,u32 portid,unsigned int nlm_flags)2241 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2242 struct nlmsghdr *nlh, struct net *net, u32 portid,
2243 unsigned int nlm_flags)
2244 {
2245 struct sk_buff *skb;
2246 u32 seq = nlh ? nlh->nlmsg_seq : 0;
2247 int err = -ENOBUFS;
2248
2249 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2250 if (skb == NULL)
2251 goto errout;
2252
2253 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2254 if (err < 0) {
2255 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2256 WARN_ON(err == -EMSGSIZE);
2257 kfree_skb(skb);
2258 goto errout;
2259 }
2260 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2261
2262 return;
2263 errout:
2264 if (err < 0)
2265 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2266 }
2267
mpls_valid_getroute_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)2268 static int mpls_valid_getroute_req(struct sk_buff *skb,
2269 const struct nlmsghdr *nlh,
2270 struct nlattr **tb,
2271 struct netlink_ext_ack *extack)
2272 {
2273 struct rtmsg *rtm;
2274 int i, err;
2275
2276 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2277 NL_SET_ERR_MSG_MOD(extack,
2278 "Invalid header for get route request");
2279 return -EINVAL;
2280 }
2281
2282 if (!netlink_strict_get_check(skb))
2283 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2284 rtm_mpls_policy, extack);
2285
2286 rtm = nlmsg_data(nlh);
2287 if ((rtm->rtm_dst_len && rtm->rtm_dst_len != 20) ||
2288 rtm->rtm_src_len || rtm->rtm_tos || rtm->rtm_table ||
2289 rtm->rtm_protocol || rtm->rtm_scope || rtm->rtm_type) {
2290 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
2291 return -EINVAL;
2292 }
2293 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
2294 NL_SET_ERR_MSG_MOD(extack,
2295 "Invalid flags for get route request");
2296 return -EINVAL;
2297 }
2298
2299 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2300 rtm_mpls_policy, extack);
2301 if (err)
2302 return err;
2303
2304 if ((tb[RTA_DST] || tb[RTA_NEWDST]) && !rtm->rtm_dst_len) {
2305 NL_SET_ERR_MSG_MOD(extack, "rtm_dst_len must be 20 for MPLS");
2306 return -EINVAL;
2307 }
2308
2309 for (i = 0; i <= RTA_MAX; i++) {
2310 if (!tb[i])
2311 continue;
2312
2313 switch (i) {
2314 case RTA_DST:
2315 case RTA_NEWDST:
2316 break;
2317 default:
2318 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
2319 return -EINVAL;
2320 }
2321 }
2322
2323 return 0;
2324 }
2325
mpls_getroute(struct sk_buff * in_skb,struct nlmsghdr * in_nlh,struct netlink_ext_ack * extack)2326 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2327 struct netlink_ext_ack *extack)
2328 {
2329 struct net *net = sock_net(in_skb->sk);
2330 u32 portid = NETLINK_CB(in_skb).portid;
2331 u32 in_label = LABEL_NOT_SPECIFIED;
2332 struct nlattr *tb[RTA_MAX + 1];
2333 u32 labels[MAX_NEW_LABELS];
2334 struct mpls_shim_hdr *hdr;
2335 unsigned int hdr_size = 0;
2336 struct net_device *dev;
2337 struct mpls_route *rt;
2338 struct rtmsg *rtm, *r;
2339 struct nlmsghdr *nlh;
2340 struct sk_buff *skb;
2341 struct mpls_nh *nh;
2342 u8 n_labels;
2343 int err;
2344
2345 err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack);
2346 if (err < 0)
2347 goto errout;
2348
2349 rtm = nlmsg_data(in_nlh);
2350
2351 if (tb[RTA_DST]) {
2352 u8 label_count;
2353
2354 if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2355 &in_label, extack)) {
2356 err = -EINVAL;
2357 goto errout;
2358 }
2359
2360 if (!mpls_label_ok(net, &in_label, extack)) {
2361 err = -EINVAL;
2362 goto errout;
2363 }
2364 }
2365
2366 rt = mpls_route_input_rcu(net, in_label);
2367 if (!rt) {
2368 err = -ENETUNREACH;
2369 goto errout;
2370 }
2371
2372 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2373 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2374 if (!skb) {
2375 err = -ENOBUFS;
2376 goto errout;
2377 }
2378
2379 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2380 RTM_NEWROUTE, in_label, rt, 0);
2381 if (err < 0) {
2382 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2383 WARN_ON(err == -EMSGSIZE);
2384 goto errout_free;
2385 }
2386
2387 return rtnl_unicast(skb, net, portid);
2388 }
2389
2390 if (tb[RTA_NEWDST]) {
2391 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2392 labels, extack) != 0) {
2393 err = -EINVAL;
2394 goto errout;
2395 }
2396
2397 hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2398 }
2399
2400 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2401 if (!skb) {
2402 err = -ENOBUFS;
2403 goto errout;
2404 }
2405
2406 skb->protocol = htons(ETH_P_MPLS_UC);
2407
2408 if (hdr_size) {
2409 bool bos;
2410 int i;
2411
2412 if (skb_cow(skb, hdr_size)) {
2413 err = -ENOBUFS;
2414 goto errout_free;
2415 }
2416
2417 skb_reserve(skb, hdr_size);
2418 skb_push(skb, hdr_size);
2419 skb_reset_network_header(skb);
2420
2421 /* Push new labels */
2422 hdr = mpls_hdr(skb);
2423 bos = true;
2424 for (i = n_labels - 1; i >= 0; i--) {
2425 hdr[i] = mpls_entry_encode(labels[i],
2426 1, 0, bos);
2427 bos = false;
2428 }
2429 }
2430
2431 nh = mpls_select_multipath(rt, skb);
2432 if (!nh) {
2433 err = -ENETUNREACH;
2434 goto errout_free;
2435 }
2436
2437 if (hdr_size) {
2438 skb_pull(skb, hdr_size);
2439 skb_reset_network_header(skb);
2440 }
2441
2442 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2443 RTM_NEWROUTE, sizeof(*r), 0);
2444 if (!nlh) {
2445 err = -EMSGSIZE;
2446 goto errout_free;
2447 }
2448
2449 r = nlmsg_data(nlh);
2450 r->rtm_family = AF_MPLS;
2451 r->rtm_dst_len = 20;
2452 r->rtm_src_len = 0;
2453 r->rtm_table = RT_TABLE_MAIN;
2454 r->rtm_type = RTN_UNICAST;
2455 r->rtm_scope = RT_SCOPE_UNIVERSE;
2456 r->rtm_protocol = rt->rt_protocol;
2457 r->rtm_flags = 0;
2458
2459 if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2460 goto nla_put_failure;
2461
2462 if (nh->nh_labels &&
2463 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2464 nh->nh_label))
2465 goto nla_put_failure;
2466
2467 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2468 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2469 nh->nh_via_alen))
2470 goto nla_put_failure;
2471 dev = rtnl_dereference(nh->nh_dev);
2472 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2473 goto nla_put_failure;
2474
2475 nlmsg_end(skb, nlh);
2476
2477 err = rtnl_unicast(skb, net, portid);
2478 errout:
2479 return err;
2480
2481 nla_put_failure:
2482 nlmsg_cancel(skb, nlh);
2483 err = -EMSGSIZE;
2484 errout_free:
2485 kfree_skb(skb);
2486 return err;
2487 }
2488
resize_platform_label_table(struct net * net,size_t limit)2489 static int resize_platform_label_table(struct net *net, size_t limit)
2490 {
2491 size_t size = sizeof(struct mpls_route *) * limit;
2492 size_t old_limit;
2493 size_t cp_size;
2494 struct mpls_route __rcu **labels = NULL, **old;
2495 struct mpls_route *rt0 = NULL, *rt2 = NULL;
2496 unsigned index;
2497
2498 if (size) {
2499 labels = kvzalloc(size, GFP_KERNEL);
2500 if (!labels)
2501 goto nolabels;
2502 }
2503
2504 /* In case the predefined labels need to be populated */
2505 if (limit > MPLS_LABEL_IPV4NULL) {
2506 struct net_device *lo = net->loopback_dev;
2507 rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2508 if (IS_ERR(rt0))
2509 goto nort0;
2510 RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo);
2511 rt0->rt_protocol = RTPROT_KERNEL;
2512 rt0->rt_payload_type = MPT_IPV4;
2513 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2514 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2515 rt0->rt_nh->nh_via_alen = lo->addr_len;
2516 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2517 lo->addr_len);
2518 }
2519 if (limit > MPLS_LABEL_IPV6NULL) {
2520 struct net_device *lo = net->loopback_dev;
2521 rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2522 if (IS_ERR(rt2))
2523 goto nort2;
2524 RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo);
2525 rt2->rt_protocol = RTPROT_KERNEL;
2526 rt2->rt_payload_type = MPT_IPV6;
2527 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2528 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2529 rt2->rt_nh->nh_via_alen = lo->addr_len;
2530 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2531 lo->addr_len);
2532 }
2533
2534 rtnl_lock();
2535 /* Remember the original table */
2536 old = rtnl_dereference(net->mpls.platform_label);
2537 old_limit = net->mpls.platform_labels;
2538
2539 /* Free any labels beyond the new table */
2540 for (index = limit; index < old_limit; index++)
2541 mpls_route_update(net, index, NULL, NULL);
2542
2543 /* Copy over the old labels */
2544 cp_size = size;
2545 if (old_limit < limit)
2546 cp_size = old_limit * sizeof(struct mpls_route *);
2547
2548 memcpy(labels, old, cp_size);
2549
2550 /* If needed set the predefined labels */
2551 if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2552 (limit > MPLS_LABEL_IPV6NULL)) {
2553 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2554 rt2 = NULL;
2555 }
2556
2557 if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2558 (limit > MPLS_LABEL_IPV4NULL)) {
2559 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2560 rt0 = NULL;
2561 }
2562
2563 /* Update the global pointers */
2564 net->mpls.platform_labels = limit;
2565 rcu_assign_pointer(net->mpls.platform_label, labels);
2566
2567 rtnl_unlock();
2568
2569 mpls_rt_free(rt2);
2570 mpls_rt_free(rt0);
2571
2572 if (old) {
2573 synchronize_rcu();
2574 kvfree(old);
2575 }
2576 return 0;
2577
2578 nort2:
2579 mpls_rt_free(rt0);
2580 nort0:
2581 kvfree(labels);
2582 nolabels:
2583 return -ENOMEM;
2584 }
2585
mpls_platform_labels(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)2586 static int mpls_platform_labels(struct ctl_table *table, int write,
2587 void *buffer, size_t *lenp, loff_t *ppos)
2588 {
2589 struct net *net = table->data;
2590 int platform_labels = net->mpls.platform_labels;
2591 int ret;
2592 struct ctl_table tmp = {
2593 .procname = table->procname,
2594 .data = &platform_labels,
2595 .maxlen = sizeof(int),
2596 .mode = table->mode,
2597 .extra1 = SYSCTL_ZERO,
2598 .extra2 = &label_limit,
2599 };
2600
2601 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2602
2603 if (write && ret == 0)
2604 ret = resize_platform_label_table(net, platform_labels);
2605
2606 return ret;
2607 }
2608
2609 #define MPLS_NS_SYSCTL_OFFSET(field) \
2610 (&((struct net *)0)->field)
2611
2612 static const struct ctl_table mpls_table[] = {
2613 {
2614 .procname = "platform_labels",
2615 .data = NULL,
2616 .maxlen = sizeof(int),
2617 .mode = 0644,
2618 .proc_handler = mpls_platform_labels,
2619 },
2620 {
2621 .procname = "ip_ttl_propagate",
2622 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2623 .maxlen = sizeof(int),
2624 .mode = 0644,
2625 .proc_handler = proc_dointvec_minmax,
2626 .extra1 = SYSCTL_ZERO,
2627 .extra2 = SYSCTL_ONE,
2628 },
2629 {
2630 .procname = "default_ttl",
2631 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2632 .maxlen = sizeof(int),
2633 .mode = 0644,
2634 .proc_handler = proc_dointvec_minmax,
2635 .extra1 = SYSCTL_ONE,
2636 .extra2 = &ttl_max,
2637 },
2638 { }
2639 };
2640
mpls_net_init(struct net * net)2641 static int mpls_net_init(struct net *net)
2642 {
2643 struct ctl_table *table;
2644 int i;
2645
2646 net->mpls.platform_labels = 0;
2647 net->mpls.platform_label = NULL;
2648 net->mpls.ip_ttl_propagate = 1;
2649 net->mpls.default_ttl = 255;
2650
2651 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2652 if (table == NULL)
2653 return -ENOMEM;
2654
2655 /* Table data contains only offsets relative to the base of
2656 * the mdev at this point, so make them absolute.
2657 */
2658 for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++)
2659 table[i].data = (char *)net + (uintptr_t)table[i].data;
2660
2661 net->mpls.ctl = register_net_sysctl(net, "net/mpls", table);
2662 if (net->mpls.ctl == NULL) {
2663 kfree(table);
2664 return -ENOMEM;
2665 }
2666
2667 return 0;
2668 }
2669
mpls_net_exit(struct net * net)2670 static void mpls_net_exit(struct net *net)
2671 {
2672 struct mpls_route __rcu **platform_label;
2673 size_t platform_labels;
2674 struct ctl_table *table;
2675 unsigned int index;
2676
2677 table = net->mpls.ctl->ctl_table_arg;
2678 unregister_net_sysctl_table(net->mpls.ctl);
2679 kfree(table);
2680
2681 /* An rcu grace period has passed since there was a device in
2682 * the network namespace (and thus the last in flight packet)
2683 * left this network namespace. This is because
2684 * unregister_netdevice_many and netdev_run_todo has completed
2685 * for each network device that was in this network namespace.
2686 *
2687 * As such no additional rcu synchronization is necessary when
2688 * freeing the platform_label table.
2689 */
2690 rtnl_lock();
2691 platform_label = rtnl_dereference(net->mpls.platform_label);
2692 platform_labels = net->mpls.platform_labels;
2693 for (index = 0; index < platform_labels; index++) {
2694 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2695 RCU_INIT_POINTER(platform_label[index], NULL);
2696 mpls_notify_route(net, index, rt, NULL, NULL);
2697 mpls_rt_free(rt);
2698 }
2699 rtnl_unlock();
2700
2701 kvfree(platform_label);
2702 }
2703
2704 static struct pernet_operations mpls_net_ops = {
2705 .init = mpls_net_init,
2706 .exit = mpls_net_exit,
2707 };
2708
2709 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2710 .family = AF_MPLS,
2711 .fill_stats_af = mpls_fill_stats_af,
2712 .get_stats_af_size = mpls_get_stats_af_size,
2713 };
2714
mpls_init(void)2715 static int __init mpls_init(void)
2716 {
2717 int err;
2718
2719 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2720
2721 err = register_pernet_subsys(&mpls_net_ops);
2722 if (err)
2723 goto out;
2724
2725 err = register_netdevice_notifier(&mpls_dev_notifier);
2726 if (err)
2727 goto out_unregister_pernet;
2728
2729 dev_add_pack(&mpls_packet_type);
2730
2731 rtnl_af_register(&mpls_af_ops);
2732
2733 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_NEWROUTE,
2734 mpls_rtm_newroute, NULL, 0);
2735 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_DELROUTE,
2736 mpls_rtm_delroute, NULL, 0);
2737 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETROUTE,
2738 mpls_getroute, mpls_dump_routes, 0);
2739 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
2740 mpls_netconf_get_devconf,
2741 mpls_netconf_dump_devconf, 0);
2742 err = ipgre_tunnel_encap_add_mpls_ops();
2743 if (err)
2744 pr_err("Can't add mpls over gre tunnel ops\n");
2745
2746 err = 0;
2747 out:
2748 return err;
2749
2750 out_unregister_pernet:
2751 unregister_pernet_subsys(&mpls_net_ops);
2752 goto out;
2753 }
2754 module_init(mpls_init);
2755
mpls_exit(void)2756 static void __exit mpls_exit(void)
2757 {
2758 rtnl_unregister_all(PF_MPLS);
2759 rtnl_af_unregister(&mpls_af_ops);
2760 dev_remove_pack(&mpls_packet_type);
2761 unregister_netdevice_notifier(&mpls_dev_notifier);
2762 unregister_pernet_subsys(&mpls_net_ops);
2763 ipgre_tunnel_encap_del_mpls_ops();
2764 }
2765 module_exit(mpls_exit);
2766
2767 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2768 MODULE_LICENSE("GPL v2");
2769 MODULE_ALIAS_NETPROTO(PF_MPLS);
2770
