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/addrconf.h>
26 #include <net/nexthop.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)
ipgre_mpls_encap_hlen(struct ip_tunnel_encap * e)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
ipgre_tunnel_encap_add_mpls_ops(void)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
ipgre_tunnel_encap_del_mpls_ops(void)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
ipgre_tunnel_encap_add_mpls_ops(void)64 static int ipgre_tunnel_encap_add_mpls_ops(void)
65 {
66 return 0;
67 }
68
ipgre_tunnel_encap_del_mpls_ops(void)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
mpls_route_input_rcu(struct net * net,unsigned index)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
mpls_output_possible(const struct net_device * dev)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
__mpls_nh_via(struct mpls_route * rt,struct mpls_nh * nh)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
mpls_nh_via(const struct mpls_route * rt,const struct mpls_nh * nh)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
mpls_nh_header_size(const struct mpls_nh * nh)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
mpls_dev_mtu(const struct net_device * dev)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
mpls_pkt_too_big(const struct sk_buff * skb,unsigned int mtu)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
mpls_stats_inc_outucastpkts(struct net_device * dev,const struct sk_buff * skb)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
mpls_multipath_hash(struct mpls_route * rt,struct sk_buff * skb)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
mpls_get_nexthop(struct mpls_route * rt,u8 index)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 */
mpls_select_multipath(struct mpls_route * rt,struct sk_buff * skb)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
mpls_egress(struct net * net,struct mpls_route * rt,struct sk_buff * skb,struct mpls_entry_decoded dec)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
mpls_forward(struct sk_buff * skb,struct net_device * dev,struct packet_type * pt,struct net_device * orig_dev)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 */
mpls_rt_alloc(u8 num_nh,u8 max_alen,u8 max_labels)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
mpls_rt_free(struct mpls_route * rt)531 static void mpls_rt_free(struct mpls_route *rt)
532 {
533 if (rt)
534 kfree_rcu(rt, rt_rcu);
535 }
536
mpls_notify_route(struct net * net,unsigned index,struct mpls_route * old,struct mpls_route * new,const struct nl_info * info)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
mpls_route_update(struct net * net,unsigned index,struct mpls_route * new,const struct nl_info * info)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
find_free_label(struct net * net)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)
inet_fib_lookup_dev(struct net * net,const void * addr)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
inet_fib_lookup_dev(struct net * net,const void * addr)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)
inet6_fib_lookup_dev(struct net * net,const void * addr)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
622 if (!ipv6_stub)
623 return ERR_PTR(-EAFNOSUPPORT);
624
625 memset(&fl6, 0, sizeof(fl6));
626 memcpy(&fl6.daddr, addr, sizeof(struct in6_addr));
627 dst = ipv6_stub->ipv6_dst_lookup_flow(net, NULL, &fl6, NULL);
628 if (IS_ERR(dst))
629 return ERR_CAST(dst);
630
631 dev = dst->dev;
632 dev_hold(dev);
633 dst_release(dst);
634
635 return dev;
636 }
637 #else
inet6_fib_lookup_dev(struct net * net,const void * addr)638 static struct net_device *inet6_fib_lookup_dev(struct net *net,
639 const void *addr)
640 {
641 return ERR_PTR(-EAFNOSUPPORT);
642 }
643 #endif
644
find_outdev(struct net * net,struct mpls_route * rt,struct mpls_nh * nh,int oif)645 static struct net_device *find_outdev(struct net *net,
646 struct mpls_route *rt,
647 struct mpls_nh *nh, int oif)
648 {
649 struct net_device *dev = NULL;
650
651 if (!oif) {
652 switch (nh->nh_via_table) {
653 case NEIGH_ARP_TABLE:
654 dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh));
655 break;
656 case NEIGH_ND_TABLE:
657 dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh));
658 break;
659 case NEIGH_LINK_TABLE:
660 break;
661 }
662 } else {
663 dev = dev_get_by_index(net, oif);
664 }
665
666 if (!dev)
667 return ERR_PTR(-ENODEV);
668
669 if (IS_ERR(dev))
670 return dev;
671
672 /* The caller is holding rtnl anyways, so release the dev reference */
673 dev_put(dev);
674
675 return dev;
676 }
677
mpls_nh_assign_dev(struct net * net,struct mpls_route * rt,struct mpls_nh * nh,int oif)678 static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
679 struct mpls_nh *nh, int oif)
680 {
681 struct net_device *dev = NULL;
682 int err = -ENODEV;
683
684 dev = find_outdev(net, rt, nh, oif);
685 if (IS_ERR(dev)) {
686 err = PTR_ERR(dev);
687 dev = NULL;
688 goto errout;
689 }
690
691 /* Ensure this is a supported device */
692 err = -EINVAL;
693 if (!mpls_dev_get(dev))
694 goto errout;
695
696 if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
697 (dev->addr_len != nh->nh_via_alen))
698 goto errout;
699
700 RCU_INIT_POINTER(nh->nh_dev, dev);
701
702 if (!(dev->flags & IFF_UP)) {
703 nh->nh_flags |= RTNH_F_DEAD;
704 } else {
705 unsigned int flags;
706
707 flags = dev_get_flags(dev);
708 if (!(flags & (IFF_RUNNING | IFF_LOWER_UP)))
709 nh->nh_flags |= RTNH_F_LINKDOWN;
710 }
711
712 return 0;
713
714 errout:
715 return err;
716 }
717
nla_get_via(const struct nlattr * nla,u8 * via_alen,u8 * via_table,u8 via_addr[],struct netlink_ext_ack * extack)718 static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table,
719 u8 via_addr[], struct netlink_ext_ack *extack)
720 {
721 struct rtvia *via = nla_data(nla);
722 int err = -EINVAL;
723 int alen;
724
725 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
726 NL_SET_ERR_MSG_ATTR(extack, nla,
727 "Invalid attribute length for RTA_VIA");
728 goto errout;
729 }
730 alen = nla_len(nla) -
731 offsetof(struct rtvia, rtvia_addr);
732 if (alen > MAX_VIA_ALEN) {
733 NL_SET_ERR_MSG_ATTR(extack, nla,
734 "Invalid address length for RTA_VIA");
735 goto errout;
736 }
737
738 /* Validate the address family */
739 switch (via->rtvia_family) {
740 case AF_PACKET:
741 *via_table = NEIGH_LINK_TABLE;
742 break;
743 case AF_INET:
744 *via_table = NEIGH_ARP_TABLE;
745 if (alen != 4)
746 goto errout;
747 break;
748 case AF_INET6:
749 *via_table = NEIGH_ND_TABLE;
750 if (alen != 16)
751 goto errout;
752 break;
753 default:
754 /* Unsupported address family */
755 goto errout;
756 }
757
758 memcpy(via_addr, via->rtvia_addr, alen);
759 *via_alen = alen;
760 err = 0;
761
762 errout:
763 return err;
764 }
765
mpls_nh_build_from_cfg(struct mpls_route_config * cfg,struct mpls_route * rt)766 static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg,
767 struct mpls_route *rt)
768 {
769 struct net *net = cfg->rc_nlinfo.nl_net;
770 struct mpls_nh *nh = rt->rt_nh;
771 int err;
772 int i;
773
774 if (!nh)
775 return -ENOMEM;
776
777 nh->nh_labels = cfg->rc_output_labels;
778 for (i = 0; i < nh->nh_labels; i++)
779 nh->nh_label[i] = cfg->rc_output_label[i];
780
781 nh->nh_via_table = cfg->rc_via_table;
782 memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen);
783 nh->nh_via_alen = cfg->rc_via_alen;
784
785 err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex);
786 if (err)
787 goto errout;
788
789 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
790 rt->rt_nhn_alive--;
791
792 return 0;
793
794 errout:
795 return err;
796 }
797
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)798 static int mpls_nh_build(struct net *net, struct mpls_route *rt,
799 struct mpls_nh *nh, int oif, struct nlattr *via,
800 struct nlattr *newdst, u8 max_labels,
801 struct netlink_ext_ack *extack)
802 {
803 int err = -ENOMEM;
804
805 if (!nh)
806 goto errout;
807
808 if (newdst) {
809 err = nla_get_labels(newdst, max_labels, &nh->nh_labels,
810 nh->nh_label, extack);
811 if (err)
812 goto errout;
813 }
814
815 if (via) {
816 err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table,
817 __mpls_nh_via(rt, nh), extack);
818 if (err)
819 goto errout;
820 } else {
821 nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC;
822 }
823
824 err = mpls_nh_assign_dev(net, rt, nh, oif);
825 if (err)
826 goto errout;
827
828 return 0;
829
830 errout:
831 return err;
832 }
833
mpls_count_nexthops(struct rtnexthop * rtnh,int len,u8 cfg_via_alen,u8 * max_via_alen,u8 * max_labels)834 static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len,
835 u8 cfg_via_alen, u8 *max_via_alen,
836 u8 *max_labels)
837 {
838 int remaining = len;
839 u8 nhs = 0;
840
841 *max_via_alen = 0;
842 *max_labels = 0;
843
844 while (rtnh_ok(rtnh, remaining)) {
845 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
846 int attrlen;
847 u8 n_labels = 0;
848
849 attrlen = rtnh_attrlen(rtnh);
850 nla = nla_find(attrs, attrlen, RTA_VIA);
851 if (nla && nla_len(nla) >=
852 offsetof(struct rtvia, rtvia_addr)) {
853 int via_alen = nla_len(nla) -
854 offsetof(struct rtvia, rtvia_addr);
855
856 if (via_alen <= MAX_VIA_ALEN)
857 *max_via_alen = max_t(u16, *max_via_alen,
858 via_alen);
859 }
860
861 nla = nla_find(attrs, attrlen, RTA_NEWDST);
862 if (nla &&
863 nla_get_labels(nla, MAX_NEW_LABELS, &n_labels,
864 NULL, NULL) != 0)
865 return 0;
866
867 *max_labels = max_t(u8, *max_labels, n_labels);
868
869 /* number of nexthops is tracked by a u8.
870 * Check for overflow.
871 */
872 if (nhs == 255)
873 return 0;
874 nhs++;
875
876 rtnh = rtnh_next(rtnh, &remaining);
877 }
878
879 /* leftover implies invalid nexthop configuration, discard it */
880 return remaining > 0 ? 0 : nhs;
881 }
882
mpls_nh_build_multi(struct mpls_route_config * cfg,struct mpls_route * rt,u8 max_labels,struct netlink_ext_ack * extack)883 static int mpls_nh_build_multi(struct mpls_route_config *cfg,
884 struct mpls_route *rt, u8 max_labels,
885 struct netlink_ext_ack *extack)
886 {
887 struct rtnexthop *rtnh = cfg->rc_mp;
888 struct nlattr *nla_via, *nla_newdst;
889 int remaining = cfg->rc_mp_len;
890 int err = 0;
891 u8 nhs = 0;
892
893 change_nexthops(rt) {
894 int attrlen;
895
896 nla_via = NULL;
897 nla_newdst = NULL;
898
899 err = -EINVAL;
900 if (!rtnh_ok(rtnh, remaining))
901 goto errout;
902
903 /* neither weighted multipath nor any flags
904 * are supported
905 */
906 if (rtnh->rtnh_hops || rtnh->rtnh_flags)
907 goto errout;
908
909 attrlen = rtnh_attrlen(rtnh);
910 if (attrlen > 0) {
911 struct nlattr *attrs = rtnh_attrs(rtnh);
912
913 nla_via = nla_find(attrs, attrlen, RTA_VIA);
914 nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST);
915 }
916
917 err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh,
918 rtnh->rtnh_ifindex, nla_via, nla_newdst,
919 max_labels, extack);
920 if (err)
921 goto errout;
922
923 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
924 rt->rt_nhn_alive--;
925
926 rtnh = rtnh_next(rtnh, &remaining);
927 nhs++;
928 } endfor_nexthops(rt);
929
930 rt->rt_nhn = nhs;
931
932 return 0;
933
934 errout:
935 return err;
936 }
937
mpls_label_ok(struct net * net,unsigned int * index,struct netlink_ext_ack * extack)938 static bool mpls_label_ok(struct net *net, unsigned int *index,
939 struct netlink_ext_ack *extack)
940 {
941 bool is_ok = true;
942
943 /* Reserved labels may not be set */
944 if (*index < MPLS_LABEL_FIRST_UNRESERVED) {
945 NL_SET_ERR_MSG(extack,
946 "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher");
947 is_ok = false;
948 }
949
950 /* The full 20 bit range may not be supported. */
951 if (is_ok && *index >= net->mpls.platform_labels) {
952 NL_SET_ERR_MSG(extack,
953 "Label >= configured maximum in platform_labels");
954 is_ok = false;
955 }
956
957 *index = array_index_nospec(*index, net->mpls.platform_labels);
958 return is_ok;
959 }
960
mpls_route_add(struct mpls_route_config * cfg,struct netlink_ext_ack * extack)961 static int mpls_route_add(struct mpls_route_config *cfg,
962 struct netlink_ext_ack *extack)
963 {
964 struct mpls_route __rcu **platform_label;
965 struct net *net = cfg->rc_nlinfo.nl_net;
966 struct mpls_route *rt, *old;
967 int err = -EINVAL;
968 u8 max_via_alen;
969 unsigned index;
970 u8 max_labels;
971 u8 nhs;
972
973 index = cfg->rc_label;
974
975 /* If a label was not specified during insert pick one */
976 if ((index == LABEL_NOT_SPECIFIED) &&
977 (cfg->rc_nlflags & NLM_F_CREATE)) {
978 index = find_free_label(net);
979 }
980
981 if (!mpls_label_ok(net, &index, extack))
982 goto errout;
983
984 /* Append makes no sense with mpls */
985 err = -EOPNOTSUPP;
986 if (cfg->rc_nlflags & NLM_F_APPEND) {
987 NL_SET_ERR_MSG(extack, "MPLS does not support route append");
988 goto errout;
989 }
990
991 err = -EEXIST;
992 platform_label = rtnl_dereference(net->mpls.platform_label);
993 old = rtnl_dereference(platform_label[index]);
994 if ((cfg->rc_nlflags & NLM_F_EXCL) && old)
995 goto errout;
996
997 err = -EEXIST;
998 if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old)
999 goto errout;
1000
1001 err = -ENOENT;
1002 if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old)
1003 goto errout;
1004
1005 err = -EINVAL;
1006 if (cfg->rc_mp) {
1007 nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len,
1008 cfg->rc_via_alen, &max_via_alen,
1009 &max_labels);
1010 } else {
1011 max_via_alen = cfg->rc_via_alen;
1012 max_labels = cfg->rc_output_labels;
1013 nhs = 1;
1014 }
1015
1016 if (nhs == 0) {
1017 NL_SET_ERR_MSG(extack, "Route does not contain a nexthop");
1018 goto errout;
1019 }
1020
1021 err = -ENOMEM;
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_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)1211 static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
1212 struct nlmsghdr *nlh,
1213 struct netlink_ext_ack *extack)
1214 {
1215 struct net *net = sock_net(in_skb->sk);
1216 struct nlattr *tb[NETCONFA_MAX + 1];
1217 struct netconfmsg *ncm;
1218 struct net_device *dev;
1219 struct mpls_dev *mdev;
1220 struct sk_buff *skb;
1221 int ifindex;
1222 int err;
1223
1224 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1225 devconf_mpls_policy, NULL);
1226 if (err < 0)
1227 goto errout;
1228
1229 err = -EINVAL;
1230 if (!tb[NETCONFA_IFINDEX])
1231 goto errout;
1232
1233 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1234 dev = __dev_get_by_index(net, ifindex);
1235 if (!dev)
1236 goto errout;
1237
1238 mdev = mpls_dev_get(dev);
1239 if (!mdev)
1240 goto errout;
1241
1242 err = -ENOBUFS;
1243 skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1244 if (!skb)
1245 goto errout;
1246
1247 err = mpls_netconf_fill_devconf(skb, mdev,
1248 NETLINK_CB(in_skb).portid,
1249 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1250 NETCONFA_ALL);
1251 if (err < 0) {
1252 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1253 WARN_ON(err == -EMSGSIZE);
1254 kfree_skb(skb);
1255 goto errout;
1256 }
1257 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1258 errout:
1259 return err;
1260 }
1261
mpls_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)1262 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
1263 struct netlink_callback *cb)
1264 {
1265 struct net *net = sock_net(skb->sk);
1266 struct hlist_head *head;
1267 struct net_device *dev;
1268 struct mpls_dev *mdev;
1269 int idx, s_idx;
1270 int h, s_h;
1271
1272 s_h = cb->args[0];
1273 s_idx = idx = cb->args[1];
1274
1275 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1276 idx = 0;
1277 head = &net->dev_index_head[h];
1278 rcu_read_lock();
1279 cb->seq = net->dev_base_seq;
1280 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1281 if (idx < s_idx)
1282 goto cont;
1283 mdev = mpls_dev_get(dev);
1284 if (!mdev)
1285 goto cont;
1286 if (mpls_netconf_fill_devconf(skb, mdev,
1287 NETLINK_CB(cb->skb).portid,
1288 cb->nlh->nlmsg_seq,
1289 RTM_NEWNETCONF,
1290 NLM_F_MULTI,
1291 NETCONFA_ALL) < 0) {
1292 rcu_read_unlock();
1293 goto done;
1294 }
1295 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1296 cont:
1297 idx++;
1298 }
1299 rcu_read_unlock();
1300 }
1301 done:
1302 cb->args[0] = h;
1303 cb->args[1] = idx;
1304
1305 return skb->len;
1306 }
1307
1308 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \
1309 (&((struct mpls_dev *)0)->field)
1310
mpls_conf_proc(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)1311 static int mpls_conf_proc(struct ctl_table *ctl, int write,
1312 void __user *buffer,
1313 size_t *lenp, loff_t *ppos)
1314 {
1315 int oval = *(int *)ctl->data;
1316 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1317
1318 if (write) {
1319 struct mpls_dev *mdev = ctl->extra1;
1320 int i = (int *)ctl->data - (int *)mdev;
1321 struct net *net = ctl->extra2;
1322 int val = *(int *)ctl->data;
1323
1324 if (i == offsetof(struct mpls_dev, input_enabled) &&
1325 val != oval) {
1326 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1327 NETCONFA_INPUT, mdev);
1328 }
1329 }
1330
1331 return ret;
1332 }
1333
1334 static const struct ctl_table mpls_dev_table[] = {
1335 {
1336 .procname = "input",
1337 .maxlen = sizeof(int),
1338 .mode = 0644,
1339 .proc_handler = mpls_conf_proc,
1340 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1341 },
1342 { }
1343 };
1344
mpls_dev_sysctl_register(struct net_device * dev,struct mpls_dev * mdev)1345 static int mpls_dev_sysctl_register(struct net_device *dev,
1346 struct mpls_dev *mdev)
1347 {
1348 char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1349 struct net *net = dev_net(dev);
1350 struct ctl_table *table;
1351 int i;
1352
1353 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1354 if (!table)
1355 goto out;
1356
1357 /* Table data contains only offsets relative to the base of
1358 * the mdev at this point, so make them absolute.
1359 */
1360 for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) {
1361 table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1362 table[i].extra1 = mdev;
1363 table[i].extra2 = net;
1364 }
1365
1366 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1367
1368 mdev->sysctl = register_net_sysctl(net, path, table);
1369 if (!mdev->sysctl)
1370 goto free;
1371
1372 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1373 return 0;
1374
1375 free:
1376 kfree(table);
1377 out:
1378 mdev->sysctl = NULL;
1379 return -ENOBUFS;
1380 }
1381
mpls_dev_sysctl_unregister(struct net_device * dev,struct mpls_dev * mdev)1382 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1383 struct mpls_dev *mdev)
1384 {
1385 struct net *net = dev_net(dev);
1386 struct ctl_table *table;
1387
1388 if (!mdev->sysctl)
1389 return;
1390
1391 table = mdev->sysctl->ctl_table_arg;
1392 unregister_net_sysctl_table(mdev->sysctl);
1393 kfree(table);
1394
1395 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1396 }
1397
mpls_add_dev(struct net_device * dev)1398 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1399 {
1400 struct mpls_dev *mdev;
1401 int err = -ENOMEM;
1402 int i;
1403
1404 ASSERT_RTNL();
1405
1406 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1407 if (!mdev)
1408 return ERR_PTR(err);
1409
1410 mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1411 if (!mdev->stats)
1412 goto free;
1413
1414 for_each_possible_cpu(i) {
1415 struct mpls_pcpu_stats *mpls_stats;
1416
1417 mpls_stats = per_cpu_ptr(mdev->stats, i);
1418 u64_stats_init(&mpls_stats->syncp);
1419 }
1420
1421 mdev->dev = dev;
1422
1423 err = mpls_dev_sysctl_register(dev, mdev);
1424 if (err)
1425 goto free;
1426
1427 rcu_assign_pointer(dev->mpls_ptr, mdev);
1428
1429 return mdev;
1430
1431 free:
1432 free_percpu(mdev->stats);
1433 kfree(mdev);
1434 return ERR_PTR(err);
1435 }
1436
mpls_dev_destroy_rcu(struct rcu_head * head)1437 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1438 {
1439 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1440
1441 free_percpu(mdev->stats);
1442 kfree(mdev);
1443 }
1444
mpls_ifdown(struct net_device * dev,int event)1445 static void mpls_ifdown(struct net_device *dev, int event)
1446 {
1447 struct mpls_route __rcu **platform_label;
1448 struct net *net = dev_net(dev);
1449 u8 alive, deleted;
1450 unsigned index;
1451
1452 platform_label = rtnl_dereference(net->mpls.platform_label);
1453 for (index = 0; index < net->mpls.platform_labels; index++) {
1454 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1455
1456 if (!rt)
1457 continue;
1458
1459 alive = 0;
1460 deleted = 0;
1461 change_nexthops(rt) {
1462 unsigned int nh_flags = nh->nh_flags;
1463
1464 if (rtnl_dereference(nh->nh_dev) != dev)
1465 goto next;
1466
1467 switch (event) {
1468 case NETDEV_DOWN:
1469 case NETDEV_UNREGISTER:
1470 nh_flags |= RTNH_F_DEAD;
1471 /* fall through */
1472 case NETDEV_CHANGE:
1473 nh_flags |= RTNH_F_LINKDOWN;
1474 break;
1475 }
1476 if (event == NETDEV_UNREGISTER)
1477 RCU_INIT_POINTER(nh->nh_dev, NULL);
1478
1479 if (nh->nh_flags != nh_flags)
1480 WRITE_ONCE(nh->nh_flags, nh_flags);
1481 next:
1482 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1483 alive++;
1484 if (!rtnl_dereference(nh->nh_dev))
1485 deleted++;
1486 } endfor_nexthops(rt);
1487
1488 WRITE_ONCE(rt->rt_nhn_alive, alive);
1489
1490 /* if there are no more nexthops, delete the route */
1491 if (event == NETDEV_UNREGISTER && deleted == rt->rt_nhn)
1492 mpls_route_update(net, index, NULL, NULL);
1493 }
1494 }
1495
mpls_ifup(struct net_device * dev,unsigned int flags)1496 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1497 {
1498 struct mpls_route __rcu **platform_label;
1499 struct net *net = dev_net(dev);
1500 unsigned index;
1501 u8 alive;
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 change_nexthops(rt) {
1512 unsigned int nh_flags = nh->nh_flags;
1513 struct net_device *nh_dev =
1514 rtnl_dereference(nh->nh_dev);
1515
1516 if (!(nh_flags & flags)) {
1517 alive++;
1518 continue;
1519 }
1520 if (nh_dev != dev)
1521 continue;
1522 alive++;
1523 nh_flags &= ~flags;
1524 WRITE_ONCE(nh->nh_flags, nh_flags);
1525 } endfor_nexthops(rt);
1526
1527 WRITE_ONCE(rt->rt_nhn_alive, alive);
1528 }
1529 }
1530
mpls_dev_notify(struct notifier_block * this,unsigned long event,void * ptr)1531 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1532 void *ptr)
1533 {
1534 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1535 struct mpls_dev *mdev;
1536 unsigned int flags;
1537
1538 if (event == NETDEV_REGISTER) {
1539
1540 /* For now just support Ethernet, IPGRE, IP6GRE, SIT and
1541 * IPIP devices
1542 */
1543 if (dev->type == ARPHRD_ETHER ||
1544 dev->type == ARPHRD_LOOPBACK ||
1545 dev->type == ARPHRD_IPGRE ||
1546 dev->type == ARPHRD_IP6GRE ||
1547 dev->type == ARPHRD_SIT ||
1548 dev->type == ARPHRD_TUNNEL) {
1549 mdev = mpls_add_dev(dev);
1550 if (IS_ERR(mdev))
1551 return notifier_from_errno(PTR_ERR(mdev));
1552 }
1553 return NOTIFY_OK;
1554 }
1555
1556 mdev = mpls_dev_get(dev);
1557 if (!mdev)
1558 return NOTIFY_OK;
1559
1560 switch (event) {
1561 case NETDEV_DOWN:
1562 mpls_ifdown(dev, event);
1563 break;
1564 case NETDEV_UP:
1565 flags = dev_get_flags(dev);
1566 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1567 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1568 else
1569 mpls_ifup(dev, RTNH_F_DEAD);
1570 break;
1571 case NETDEV_CHANGE:
1572 flags = dev_get_flags(dev);
1573 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1574 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1575 else
1576 mpls_ifdown(dev, event);
1577 break;
1578 case NETDEV_UNREGISTER:
1579 mpls_ifdown(dev, event);
1580 mdev = mpls_dev_get(dev);
1581 if (mdev) {
1582 mpls_dev_sysctl_unregister(dev, mdev);
1583 RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1584 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1585 }
1586 break;
1587 case NETDEV_CHANGENAME:
1588 mdev = mpls_dev_get(dev);
1589 if (mdev) {
1590 int err;
1591
1592 mpls_dev_sysctl_unregister(dev, mdev);
1593 err = mpls_dev_sysctl_register(dev, mdev);
1594 if (err)
1595 return notifier_from_errno(err);
1596 }
1597 break;
1598 }
1599 return NOTIFY_OK;
1600 }
1601
1602 static struct notifier_block mpls_dev_notifier = {
1603 .notifier_call = mpls_dev_notify,
1604 };
1605
nla_put_via(struct sk_buff * skb,u8 table,const void * addr,int alen)1606 static int nla_put_via(struct sk_buff *skb,
1607 u8 table, const void *addr, int alen)
1608 {
1609 static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1610 AF_INET, AF_INET6, AF_DECnet, AF_PACKET,
1611 };
1612 struct nlattr *nla;
1613 struct rtvia *via;
1614 int family = AF_UNSPEC;
1615
1616 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1617 if (!nla)
1618 return -EMSGSIZE;
1619
1620 if (table <= NEIGH_NR_TABLES)
1621 family = table_to_family[table];
1622
1623 via = nla_data(nla);
1624 via->rtvia_family = family;
1625 memcpy(via->rtvia_addr, addr, alen);
1626 return 0;
1627 }
1628
nla_put_labels(struct sk_buff * skb,int attrtype,u8 labels,const u32 label[])1629 int nla_put_labels(struct sk_buff *skb, int attrtype,
1630 u8 labels, const u32 label[])
1631 {
1632 struct nlattr *nla;
1633 struct mpls_shim_hdr *nla_label;
1634 bool bos;
1635 int i;
1636 nla = nla_reserve(skb, attrtype, labels*4);
1637 if (!nla)
1638 return -EMSGSIZE;
1639
1640 nla_label = nla_data(nla);
1641 bos = true;
1642 for (i = labels - 1; i >= 0; i--) {
1643 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1644 bos = false;
1645 }
1646
1647 return 0;
1648 }
1649 EXPORT_SYMBOL_GPL(nla_put_labels);
1650
nla_get_labels(const struct nlattr * nla,u8 max_labels,u8 * labels,u32 label[],struct netlink_ext_ack * extack)1651 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1652 u32 label[], struct netlink_ext_ack *extack)
1653 {
1654 unsigned len = nla_len(nla);
1655 struct mpls_shim_hdr *nla_label;
1656 u8 nla_labels;
1657 bool bos;
1658 int i;
1659
1660 /* len needs to be an even multiple of 4 (the label size). Number
1661 * of labels is a u8 so check for overflow.
1662 */
1663 if (len & 3 || len / 4 > 255) {
1664 NL_SET_ERR_MSG_ATTR(extack, nla,
1665 "Invalid length for labels attribute");
1666 return -EINVAL;
1667 }
1668
1669 /* Limit the number of new labels allowed */
1670 nla_labels = len/4;
1671 if (nla_labels > max_labels) {
1672 NL_SET_ERR_MSG(extack, "Too many labels");
1673 return -EINVAL;
1674 }
1675
1676 /* when label == NULL, caller wants number of labels */
1677 if (!label)
1678 goto out;
1679
1680 nla_label = nla_data(nla);
1681 bos = true;
1682 for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1683 struct mpls_entry_decoded dec;
1684 dec = mpls_entry_decode(nla_label + i);
1685
1686 /* Ensure the bottom of stack flag is properly set
1687 * and ttl and tc are both clear.
1688 */
1689 if (dec.ttl) {
1690 NL_SET_ERR_MSG_ATTR(extack, nla,
1691 "TTL in label must be 0");
1692 return -EINVAL;
1693 }
1694
1695 if (dec.tc) {
1696 NL_SET_ERR_MSG_ATTR(extack, nla,
1697 "Traffic class in label must be 0");
1698 return -EINVAL;
1699 }
1700
1701 if (dec.bos != bos) {
1702 NL_SET_BAD_ATTR(extack, nla);
1703 if (bos) {
1704 NL_SET_ERR_MSG(extack,
1705 "BOS bit must be set in first label");
1706 } else {
1707 NL_SET_ERR_MSG(extack,
1708 "BOS bit can only be set in first label");
1709 }
1710 return -EINVAL;
1711 }
1712
1713 switch (dec.label) {
1714 case MPLS_LABEL_IMPLNULL:
1715 /* RFC3032: This is a label that an LSR may
1716 * assign and distribute, but which never
1717 * actually appears in the encapsulation.
1718 */
1719 NL_SET_ERR_MSG_ATTR(extack, nla,
1720 "Implicit NULL Label (3) can not be used in encapsulation");
1721 return -EINVAL;
1722 }
1723
1724 label[i] = dec.label;
1725 }
1726 out:
1727 *labels = nla_labels;
1728 return 0;
1729 }
1730 EXPORT_SYMBOL_GPL(nla_get_labels);
1731
rtm_to_route_config(struct sk_buff * skb,struct nlmsghdr * nlh,struct mpls_route_config * cfg,struct netlink_ext_ack * extack)1732 static int rtm_to_route_config(struct sk_buff *skb,
1733 struct nlmsghdr *nlh,
1734 struct mpls_route_config *cfg,
1735 struct netlink_ext_ack *extack)
1736 {
1737 struct rtmsg *rtm;
1738 struct nlattr *tb[RTA_MAX+1];
1739 int index;
1740 int err;
1741
1742 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_mpls_policy,
1743 extack);
1744 if (err < 0)
1745 goto errout;
1746
1747 err = -EINVAL;
1748 rtm = nlmsg_data(nlh);
1749
1750 if (rtm->rtm_family != AF_MPLS) {
1751 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1752 goto errout;
1753 }
1754 if (rtm->rtm_dst_len != 20) {
1755 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1756 goto errout;
1757 }
1758 if (rtm->rtm_src_len != 0) {
1759 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1760 goto errout;
1761 }
1762 if (rtm->rtm_tos != 0) {
1763 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1764 goto errout;
1765 }
1766 if (rtm->rtm_table != RT_TABLE_MAIN) {
1767 NL_SET_ERR_MSG(extack,
1768 "MPLS only supports the main route table");
1769 goto errout;
1770 }
1771 /* Any value is acceptable for rtm_protocol */
1772
1773 /* As mpls uses destination specific addresses
1774 * (or source specific address in the case of multicast)
1775 * all addresses have universal scope.
1776 */
1777 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1778 NL_SET_ERR_MSG(extack,
1779 "Invalid route scope - MPLS only supports UNIVERSE");
1780 goto errout;
1781 }
1782 if (rtm->rtm_type != RTN_UNICAST) {
1783 NL_SET_ERR_MSG(extack,
1784 "Invalid route type - MPLS only supports UNICAST");
1785 goto errout;
1786 }
1787 if (rtm->rtm_flags != 0) {
1788 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1789 goto errout;
1790 }
1791
1792 cfg->rc_label = LABEL_NOT_SPECIFIED;
1793 cfg->rc_protocol = rtm->rtm_protocol;
1794 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC;
1795 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
1796 cfg->rc_nlflags = nlh->nlmsg_flags;
1797 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid;
1798 cfg->rc_nlinfo.nlh = nlh;
1799 cfg->rc_nlinfo.nl_net = sock_net(skb->sk);
1800
1801 for (index = 0; index <= RTA_MAX; index++) {
1802 struct nlattr *nla = tb[index];
1803 if (!nla)
1804 continue;
1805
1806 switch (index) {
1807 case RTA_OIF:
1808 cfg->rc_ifindex = nla_get_u32(nla);
1809 break;
1810 case RTA_NEWDST:
1811 if (nla_get_labels(nla, MAX_NEW_LABELS,
1812 &cfg->rc_output_labels,
1813 cfg->rc_output_label, extack))
1814 goto errout;
1815 break;
1816 case RTA_DST:
1817 {
1818 u8 label_count;
1819 if (nla_get_labels(nla, 1, &label_count,
1820 &cfg->rc_label, extack))
1821 goto errout;
1822
1823 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1824 &cfg->rc_label, extack))
1825 goto errout;
1826 break;
1827 }
1828 case RTA_GATEWAY:
1829 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1830 goto errout;
1831 case RTA_VIA:
1832 {
1833 if (nla_get_via(nla, &cfg->rc_via_alen,
1834 &cfg->rc_via_table, cfg->rc_via,
1835 extack))
1836 goto errout;
1837 break;
1838 }
1839 case RTA_MULTIPATH:
1840 {
1841 cfg->rc_mp = nla_data(nla);
1842 cfg->rc_mp_len = nla_len(nla);
1843 break;
1844 }
1845 case RTA_TTL_PROPAGATE:
1846 {
1847 u8 ttl_propagate = nla_get_u8(nla);
1848
1849 if (ttl_propagate > 1) {
1850 NL_SET_ERR_MSG_ATTR(extack, nla,
1851 "RTA_TTL_PROPAGATE can only be 0 or 1");
1852 goto errout;
1853 }
1854 cfg->rc_ttl_propagate = ttl_propagate ?
1855 MPLS_TTL_PROP_ENABLED :
1856 MPLS_TTL_PROP_DISABLED;
1857 break;
1858 }
1859 default:
1860 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1861 /* Unsupported attribute */
1862 goto errout;
1863 }
1864 }
1865
1866 err = 0;
1867 errout:
1868 return err;
1869 }
1870
mpls_rtm_delroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)1871 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1872 struct netlink_ext_ack *extack)
1873 {
1874 struct mpls_route_config *cfg;
1875 int err;
1876
1877 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1878 if (!cfg)
1879 return -ENOMEM;
1880
1881 err = rtm_to_route_config(skb, nlh, cfg, extack);
1882 if (err < 0)
1883 goto out;
1884
1885 err = mpls_route_del(cfg, extack);
1886 out:
1887 kfree(cfg);
1888
1889 return err;
1890 }
1891
1892
mpls_rtm_newroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)1893 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1894 struct netlink_ext_ack *extack)
1895 {
1896 struct mpls_route_config *cfg;
1897 int err;
1898
1899 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1900 if (!cfg)
1901 return -ENOMEM;
1902
1903 err = rtm_to_route_config(skb, nlh, cfg, extack);
1904 if (err < 0)
1905 goto out;
1906
1907 err = mpls_route_add(cfg, extack);
1908 out:
1909 kfree(cfg);
1910
1911 return err;
1912 }
1913
mpls_dump_route(struct sk_buff * skb,u32 portid,u32 seq,int event,u32 label,struct mpls_route * rt,int flags)1914 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1915 u32 label, struct mpls_route *rt, int flags)
1916 {
1917 struct net_device *dev;
1918 struct nlmsghdr *nlh;
1919 struct rtmsg *rtm;
1920
1921 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1922 if (nlh == NULL)
1923 return -EMSGSIZE;
1924
1925 rtm = nlmsg_data(nlh);
1926 rtm->rtm_family = AF_MPLS;
1927 rtm->rtm_dst_len = 20;
1928 rtm->rtm_src_len = 0;
1929 rtm->rtm_tos = 0;
1930 rtm->rtm_table = RT_TABLE_MAIN;
1931 rtm->rtm_protocol = rt->rt_protocol;
1932 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
1933 rtm->rtm_type = RTN_UNICAST;
1934 rtm->rtm_flags = 0;
1935
1936 if (nla_put_labels(skb, RTA_DST, 1, &label))
1937 goto nla_put_failure;
1938
1939 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
1940 bool ttl_propagate =
1941 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
1942
1943 if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
1944 ttl_propagate))
1945 goto nla_put_failure;
1946 }
1947 if (rt->rt_nhn == 1) {
1948 const struct mpls_nh *nh = rt->rt_nh;
1949
1950 if (nh->nh_labels &&
1951 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
1952 nh->nh_label))
1953 goto nla_put_failure;
1954 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
1955 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
1956 nh->nh_via_alen))
1957 goto nla_put_failure;
1958 dev = rtnl_dereference(nh->nh_dev);
1959 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
1960 goto nla_put_failure;
1961 if (nh->nh_flags & RTNH_F_LINKDOWN)
1962 rtm->rtm_flags |= RTNH_F_LINKDOWN;
1963 if (nh->nh_flags & RTNH_F_DEAD)
1964 rtm->rtm_flags |= RTNH_F_DEAD;
1965 } else {
1966 struct rtnexthop *rtnh;
1967 struct nlattr *mp;
1968 u8 linkdown = 0;
1969 u8 dead = 0;
1970
1971 mp = nla_nest_start(skb, RTA_MULTIPATH);
1972 if (!mp)
1973 goto nla_put_failure;
1974
1975 for_nexthops(rt) {
1976 dev = rtnl_dereference(nh->nh_dev);
1977 if (!dev)
1978 continue;
1979
1980 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1981 if (!rtnh)
1982 goto nla_put_failure;
1983
1984 rtnh->rtnh_ifindex = dev->ifindex;
1985 if (nh->nh_flags & RTNH_F_LINKDOWN) {
1986 rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
1987 linkdown++;
1988 }
1989 if (nh->nh_flags & RTNH_F_DEAD) {
1990 rtnh->rtnh_flags |= RTNH_F_DEAD;
1991 dead++;
1992 }
1993
1994 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
1995 nh->nh_labels,
1996 nh->nh_label))
1997 goto nla_put_failure;
1998 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
1999 nla_put_via(skb, nh->nh_via_table,
2000 mpls_nh_via(rt, nh),
2001 nh->nh_via_alen))
2002 goto nla_put_failure;
2003
2004 /* length of rtnetlink header + attributes */
2005 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2006 } endfor_nexthops(rt);
2007
2008 if (linkdown == rt->rt_nhn)
2009 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2010 if (dead == rt->rt_nhn)
2011 rtm->rtm_flags |= RTNH_F_DEAD;
2012
2013 nla_nest_end(skb, mp);
2014 }
2015
2016 nlmsg_end(skb, nlh);
2017 return 0;
2018
2019 nla_put_failure:
2020 nlmsg_cancel(skb, nlh);
2021 return -EMSGSIZE;
2022 }
2023
mpls_dump_routes(struct sk_buff * skb,struct netlink_callback * cb)2024 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2025 {
2026 struct net *net = sock_net(skb->sk);
2027 struct mpls_route __rcu **platform_label;
2028 size_t platform_labels;
2029 unsigned int index;
2030
2031 ASSERT_RTNL();
2032
2033 index = cb->args[0];
2034 if (index < MPLS_LABEL_FIRST_UNRESERVED)
2035 index = MPLS_LABEL_FIRST_UNRESERVED;
2036
2037 platform_label = rtnl_dereference(net->mpls.platform_label);
2038 platform_labels = net->mpls.platform_labels;
2039 for (; index < platform_labels; index++) {
2040 struct mpls_route *rt;
2041 rt = rtnl_dereference(platform_label[index]);
2042 if (!rt)
2043 continue;
2044
2045 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2046 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2047 index, rt, NLM_F_MULTI) < 0)
2048 break;
2049 }
2050 cb->args[0] = index;
2051
2052 return skb->len;
2053 }
2054
lfib_nlmsg_size(struct mpls_route * rt)2055 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2056 {
2057 size_t payload =
2058 NLMSG_ALIGN(sizeof(struct rtmsg))
2059 + nla_total_size(4) /* RTA_DST */
2060 + nla_total_size(1); /* RTA_TTL_PROPAGATE */
2061
2062 if (rt->rt_nhn == 1) {
2063 struct mpls_nh *nh = rt->rt_nh;
2064
2065 if (nh->nh_dev)
2066 payload += nla_total_size(4); /* RTA_OIF */
2067 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2068 payload += nla_total_size(2 + nh->nh_via_alen);
2069 if (nh->nh_labels) /* RTA_NEWDST */
2070 payload += nla_total_size(nh->nh_labels * 4);
2071 } else {
2072 /* each nexthop is packed in an attribute */
2073 size_t nhsize = 0;
2074
2075 for_nexthops(rt) {
2076 if (!rtnl_dereference(nh->nh_dev))
2077 continue;
2078 nhsize += nla_total_size(sizeof(struct rtnexthop));
2079 /* RTA_VIA */
2080 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2081 nhsize += nla_total_size(2 + nh->nh_via_alen);
2082 if (nh->nh_labels)
2083 nhsize += nla_total_size(nh->nh_labels * 4);
2084 } endfor_nexthops(rt);
2085 /* nested attribute */
2086 payload += nla_total_size(nhsize);
2087 }
2088
2089 return payload;
2090 }
2091
rtmsg_lfib(int event,u32 label,struct mpls_route * rt,struct nlmsghdr * nlh,struct net * net,u32 portid,unsigned int nlm_flags)2092 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2093 struct nlmsghdr *nlh, struct net *net, u32 portid,
2094 unsigned int nlm_flags)
2095 {
2096 struct sk_buff *skb;
2097 u32 seq = nlh ? nlh->nlmsg_seq : 0;
2098 int err = -ENOBUFS;
2099
2100 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2101 if (skb == NULL)
2102 goto errout;
2103
2104 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2105 if (err < 0) {
2106 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2107 WARN_ON(err == -EMSGSIZE);
2108 kfree_skb(skb);
2109 goto errout;
2110 }
2111 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2112
2113 return;
2114 errout:
2115 if (err < 0)
2116 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2117 }
2118
mpls_getroute(struct sk_buff * in_skb,struct nlmsghdr * in_nlh,struct netlink_ext_ack * extack)2119 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2120 struct netlink_ext_ack *extack)
2121 {
2122 struct net *net = sock_net(in_skb->sk);
2123 u32 portid = NETLINK_CB(in_skb).portid;
2124 u32 in_label = LABEL_NOT_SPECIFIED;
2125 struct nlattr *tb[RTA_MAX + 1];
2126 u32 labels[MAX_NEW_LABELS];
2127 struct mpls_shim_hdr *hdr;
2128 unsigned int hdr_size = 0;
2129 struct net_device *dev;
2130 struct mpls_route *rt;
2131 struct rtmsg *rtm, *r;
2132 struct nlmsghdr *nlh;
2133 struct sk_buff *skb;
2134 struct mpls_nh *nh;
2135 u8 n_labels;
2136 int err;
2137
2138 err = nlmsg_parse(in_nlh, sizeof(*rtm), tb, RTA_MAX,
2139 rtm_mpls_policy, extack);
2140 if (err < 0)
2141 goto errout;
2142
2143 rtm = nlmsg_data(in_nlh);
2144
2145 if (tb[RTA_DST]) {
2146 u8 label_count;
2147
2148 if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2149 &in_label, extack)) {
2150 err = -EINVAL;
2151 goto errout;
2152 }
2153
2154 if (!mpls_label_ok(net, &in_label, extack)) {
2155 err = -EINVAL;
2156 goto errout;
2157 }
2158 }
2159
2160 rt = mpls_route_input_rcu(net, in_label);
2161 if (!rt) {
2162 err = -ENETUNREACH;
2163 goto errout;
2164 }
2165
2166 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2167 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2168 if (!skb) {
2169 err = -ENOBUFS;
2170 goto errout;
2171 }
2172
2173 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2174 RTM_NEWROUTE, in_label, rt, 0);
2175 if (err < 0) {
2176 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2177 WARN_ON(err == -EMSGSIZE);
2178 goto errout_free;
2179 }
2180
2181 return rtnl_unicast(skb, net, portid);
2182 }
2183
2184 if (tb[RTA_NEWDST]) {
2185 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2186 labels, extack) != 0) {
2187 err = -EINVAL;
2188 goto errout;
2189 }
2190
2191 hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2192 }
2193
2194 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2195 if (!skb) {
2196 err = -ENOBUFS;
2197 goto errout;
2198 }
2199
2200 skb->protocol = htons(ETH_P_MPLS_UC);
2201
2202 if (hdr_size) {
2203 bool bos;
2204 int i;
2205
2206 if (skb_cow(skb, hdr_size)) {
2207 err = -ENOBUFS;
2208 goto errout_free;
2209 }
2210
2211 skb_reserve(skb, hdr_size);
2212 skb_push(skb, hdr_size);
2213 skb_reset_network_header(skb);
2214
2215 /* Push new labels */
2216 hdr = mpls_hdr(skb);
2217 bos = true;
2218 for (i = n_labels - 1; i >= 0; i--) {
2219 hdr[i] = mpls_entry_encode(labels[i],
2220 1, 0, bos);
2221 bos = false;
2222 }
2223 }
2224
2225 nh = mpls_select_multipath(rt, skb);
2226 if (!nh) {
2227 err = -ENETUNREACH;
2228 goto errout_free;
2229 }
2230
2231 if (hdr_size) {
2232 skb_pull(skb, hdr_size);
2233 skb_reset_network_header(skb);
2234 }
2235
2236 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2237 RTM_NEWROUTE, sizeof(*r), 0);
2238 if (!nlh) {
2239 err = -EMSGSIZE;
2240 goto errout_free;
2241 }
2242
2243 r = nlmsg_data(nlh);
2244 r->rtm_family = AF_MPLS;
2245 r->rtm_dst_len = 20;
2246 r->rtm_src_len = 0;
2247 r->rtm_table = RT_TABLE_MAIN;
2248 r->rtm_type = RTN_UNICAST;
2249 r->rtm_scope = RT_SCOPE_UNIVERSE;
2250 r->rtm_protocol = rt->rt_protocol;
2251 r->rtm_flags = 0;
2252
2253 if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2254 goto nla_put_failure;
2255
2256 if (nh->nh_labels &&
2257 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2258 nh->nh_label))
2259 goto nla_put_failure;
2260
2261 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2262 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2263 nh->nh_via_alen))
2264 goto nla_put_failure;
2265 dev = rtnl_dereference(nh->nh_dev);
2266 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2267 goto nla_put_failure;
2268
2269 nlmsg_end(skb, nlh);
2270
2271 err = rtnl_unicast(skb, net, portid);
2272 errout:
2273 return err;
2274
2275 nla_put_failure:
2276 nlmsg_cancel(skb, nlh);
2277 err = -EMSGSIZE;
2278 errout_free:
2279 kfree_skb(skb);
2280 return err;
2281 }
2282
resize_platform_label_table(struct net * net,size_t limit)2283 static int resize_platform_label_table(struct net *net, size_t limit)
2284 {
2285 size_t size = sizeof(struct mpls_route *) * limit;
2286 size_t old_limit;
2287 size_t cp_size;
2288 struct mpls_route __rcu **labels = NULL, **old;
2289 struct mpls_route *rt0 = NULL, *rt2 = NULL;
2290 unsigned index;
2291
2292 if (size) {
2293 labels = kvzalloc(size, GFP_KERNEL);
2294 if (!labels)
2295 goto nolabels;
2296 }
2297
2298 /* In case the predefined labels need to be populated */
2299 if (limit > MPLS_LABEL_IPV4NULL) {
2300 struct net_device *lo = net->loopback_dev;
2301 rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2302 if (IS_ERR(rt0))
2303 goto nort0;
2304 RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo);
2305 rt0->rt_protocol = RTPROT_KERNEL;
2306 rt0->rt_payload_type = MPT_IPV4;
2307 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2308 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2309 rt0->rt_nh->nh_via_alen = lo->addr_len;
2310 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2311 lo->addr_len);
2312 }
2313 if (limit > MPLS_LABEL_IPV6NULL) {
2314 struct net_device *lo = net->loopback_dev;
2315 rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2316 if (IS_ERR(rt2))
2317 goto nort2;
2318 RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo);
2319 rt2->rt_protocol = RTPROT_KERNEL;
2320 rt2->rt_payload_type = MPT_IPV6;
2321 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2322 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2323 rt2->rt_nh->nh_via_alen = lo->addr_len;
2324 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2325 lo->addr_len);
2326 }
2327
2328 rtnl_lock();
2329 /* Remember the original table */
2330 old = rtnl_dereference(net->mpls.platform_label);
2331 old_limit = net->mpls.platform_labels;
2332
2333 /* Free any labels beyond the new table */
2334 for (index = limit; index < old_limit; index++)
2335 mpls_route_update(net, index, NULL, NULL);
2336
2337 /* Copy over the old labels */
2338 cp_size = size;
2339 if (old_limit < limit)
2340 cp_size = old_limit * sizeof(struct mpls_route *);
2341
2342 memcpy(labels, old, cp_size);
2343
2344 /* If needed set the predefined labels */
2345 if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2346 (limit > MPLS_LABEL_IPV6NULL)) {
2347 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2348 rt2 = NULL;
2349 }
2350
2351 if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2352 (limit > MPLS_LABEL_IPV4NULL)) {
2353 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2354 rt0 = NULL;
2355 }
2356
2357 /* Update the global pointers */
2358 net->mpls.platform_labels = limit;
2359 rcu_assign_pointer(net->mpls.platform_label, labels);
2360
2361 rtnl_unlock();
2362
2363 mpls_rt_free(rt2);
2364 mpls_rt_free(rt0);
2365
2366 if (old) {
2367 synchronize_rcu();
2368 kvfree(old);
2369 }
2370 return 0;
2371
2372 nort2:
2373 mpls_rt_free(rt0);
2374 nort0:
2375 kvfree(labels);
2376 nolabels:
2377 return -ENOMEM;
2378 }
2379
mpls_platform_labels(struct ctl_table * table,int write,void __user * buffer,size_t * lenp,loff_t * ppos)2380 static int mpls_platform_labels(struct ctl_table *table, int write,
2381 void __user *buffer, size_t *lenp, loff_t *ppos)
2382 {
2383 struct net *net = table->data;
2384 int platform_labels = net->mpls.platform_labels;
2385 int ret;
2386 struct ctl_table tmp = {
2387 .procname = table->procname,
2388 .data = &platform_labels,
2389 .maxlen = sizeof(int),
2390 .mode = table->mode,
2391 .extra1 = &zero,
2392 .extra2 = &label_limit,
2393 };
2394
2395 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2396
2397 if (write && ret == 0)
2398 ret = resize_platform_label_table(net, platform_labels);
2399
2400 return ret;
2401 }
2402
2403 #define MPLS_NS_SYSCTL_OFFSET(field) \
2404 (&((struct net *)0)->field)
2405
2406 static const struct ctl_table mpls_table[] = {
2407 {
2408 .procname = "platform_labels",
2409 .data = NULL,
2410 .maxlen = sizeof(int),
2411 .mode = 0644,
2412 .proc_handler = mpls_platform_labels,
2413 },
2414 {
2415 .procname = "ip_ttl_propagate",
2416 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2417 .maxlen = sizeof(int),
2418 .mode = 0644,
2419 .proc_handler = proc_dointvec_minmax,
2420 .extra1 = &zero,
2421 .extra2 = &one,
2422 },
2423 {
2424 .procname = "default_ttl",
2425 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2426 .maxlen = sizeof(int),
2427 .mode = 0644,
2428 .proc_handler = proc_dointvec_minmax,
2429 .extra1 = &one,
2430 .extra2 = &ttl_max,
2431 },
2432 { }
2433 };
2434
mpls_net_init(struct net * net)2435 static int mpls_net_init(struct net *net)
2436 {
2437 struct ctl_table *table;
2438 int i;
2439
2440 net->mpls.platform_labels = 0;
2441 net->mpls.platform_label = NULL;
2442 net->mpls.ip_ttl_propagate = 1;
2443 net->mpls.default_ttl = 255;
2444
2445 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2446 if (table == NULL)
2447 return -ENOMEM;
2448
2449 /* Table data contains only offsets relative to the base of
2450 * the mdev at this point, so make them absolute.
2451 */
2452 for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++)
2453 table[i].data = (char *)net + (uintptr_t)table[i].data;
2454
2455 net->mpls.ctl = register_net_sysctl(net, "net/mpls", table);
2456 if (net->mpls.ctl == NULL) {
2457 kfree(table);
2458 return -ENOMEM;
2459 }
2460
2461 return 0;
2462 }
2463
mpls_net_exit(struct net * net)2464 static void mpls_net_exit(struct net *net)
2465 {
2466 struct mpls_route __rcu **platform_label;
2467 size_t platform_labels;
2468 struct ctl_table *table;
2469 unsigned int index;
2470
2471 table = net->mpls.ctl->ctl_table_arg;
2472 unregister_net_sysctl_table(net->mpls.ctl);
2473 kfree(table);
2474
2475 /* An rcu grace period has passed since there was a device in
2476 * the network namespace (and thus the last in flight packet)
2477 * left this network namespace. This is because
2478 * unregister_netdevice_many and netdev_run_todo has completed
2479 * for each network device that was in this network namespace.
2480 *
2481 * As such no additional rcu synchronization is necessary when
2482 * freeing the platform_label table.
2483 */
2484 rtnl_lock();
2485 platform_label = rtnl_dereference(net->mpls.platform_label);
2486 platform_labels = net->mpls.platform_labels;
2487 for (index = 0; index < platform_labels; index++) {
2488 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2489 RCU_INIT_POINTER(platform_label[index], NULL);
2490 mpls_notify_route(net, index, rt, NULL, NULL);
2491 mpls_rt_free(rt);
2492 }
2493 rtnl_unlock();
2494
2495 kvfree(platform_label);
2496 }
2497
2498 static struct pernet_operations mpls_net_ops = {
2499 .init = mpls_net_init,
2500 .exit = mpls_net_exit,
2501 };
2502
2503 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2504 .family = AF_MPLS,
2505 .fill_stats_af = mpls_fill_stats_af,
2506 .get_stats_af_size = mpls_get_stats_af_size,
2507 };
2508
mpls_init(void)2509 static int __init mpls_init(void)
2510 {
2511 int err;
2512
2513 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2514
2515 err = register_pernet_subsys(&mpls_net_ops);
2516 if (err)
2517 goto out;
2518
2519 err = register_netdevice_notifier(&mpls_dev_notifier);
2520 if (err)
2521 goto out_unregister_pernet;
2522
2523 dev_add_pack(&mpls_packet_type);
2524
2525 rtnl_af_register(&mpls_af_ops);
2526
2527 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_NEWROUTE,
2528 mpls_rtm_newroute, NULL, 0);
2529 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_DELROUTE,
2530 mpls_rtm_delroute, NULL, 0);
2531 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETROUTE,
2532 mpls_getroute, mpls_dump_routes, 0);
2533 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
2534 mpls_netconf_get_devconf,
2535 mpls_netconf_dump_devconf, 0);
2536 err = ipgre_tunnel_encap_add_mpls_ops();
2537 if (err)
2538 pr_err("Can't add mpls over gre tunnel ops\n");
2539
2540 err = 0;
2541 out:
2542 return err;
2543
2544 out_unregister_pernet:
2545 unregister_pernet_subsys(&mpls_net_ops);
2546 goto out;
2547 }
2548 module_init(mpls_init);
2549
mpls_exit(void)2550 static void __exit mpls_exit(void)
2551 {
2552 rtnl_unregister_all(PF_MPLS);
2553 rtnl_af_unregister(&mpls_af_ops);
2554 dev_remove_pack(&mpls_packet_type);
2555 unregister_netdevice_notifier(&mpls_dev_notifier);
2556 unregister_pernet_subsys(&mpls_net_ops);
2557 ipgre_tunnel_encap_del_mpls_ops();
2558 }
2559 module_exit(mpls_exit);
2560
2561 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2562 MODULE_LICENSE("GPL v2");
2563 MODULE_ALIAS_NETPROTO(PF_MPLS);
2564