Commit | Line | Data |
---|---|---|
0189197f EB |
1 | #include <linux/types.h> |
2 | #include <linux/skbuff.h> | |
3 | #include <linux/socket.h> | |
7720c01f | 4 | #include <linux/sysctl.h> |
0189197f EB |
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> | |
4b5edb2f | 10 | #include <linux/vmalloc.h> |
0189197f EB |
11 | #include <net/ip.h> |
12 | #include <net/dst.h> | |
13 | #include <net/sock.h> | |
14 | #include <net/arp.h> | |
15 | #include <net/ip_fib.h> | |
16 | #include <net/netevent.h> | |
17 | #include <net/netns/generic.h> | |
bf21563a RP |
18 | #if IS_ENABLED(CONFIG_IPV6) |
19 | #include <net/ipv6.h> | |
20 | #include <net/addrconf.h> | |
21 | #endif | |
f8efb73c | 22 | #include <net/nexthop.h> |
0189197f EB |
23 | #include "internal.h" |
24 | ||
1c78efa8 RS |
25 | /* Maximum number of labels to look ahead at when selecting a path of |
26 | * a multipath route | |
27 | */ | |
28 | #define MAX_MP_SELECT_LABELS 4 | |
29 | ||
7720c01f EB |
30 | static int zero = 0; |
31 | static int label_limit = (1 << 20) - 1; | |
32 | ||
8de147dc EB |
33 | static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, |
34 | struct nlmsghdr *nlh, struct net *net, u32 portid, | |
35 | unsigned int nlm_flags); | |
36 | ||
0189197f EB |
37 | static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index) |
38 | { | |
39 | struct mpls_route *rt = NULL; | |
40 | ||
41 | if (index < net->mpls.platform_labels) { | |
42 | struct mpls_route __rcu **platform_label = | |
43 | rcu_dereference(net->mpls.platform_label); | |
44 | rt = rcu_dereference(platform_label[index]); | |
45 | } | |
46 | return rt; | |
47 | } | |
48 | ||
03c57747 RS |
49 | static inline struct mpls_dev *mpls_dev_get(const struct net_device *dev) |
50 | { | |
51 | return rcu_dereference_rtnl(dev->mpls_ptr); | |
52 | } | |
53 | ||
face0188 | 54 | bool mpls_output_possible(const struct net_device *dev) |
0189197f EB |
55 | { |
56 | return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev); | |
57 | } | |
face0188 | 58 | EXPORT_SYMBOL_GPL(mpls_output_possible); |
0189197f | 59 | |
cf4b24f0 RS |
60 | static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh) |
61 | { | |
62 | u8 *nh0_via = PTR_ALIGN((u8 *)&rt->rt_nh[rt->rt_nhn], VIA_ALEN_ALIGN); | |
63 | int nh_index = nh - rt->rt_nh; | |
64 | ||
65 | return nh0_via + rt->rt_max_alen * nh_index; | |
66 | } | |
67 | ||
68 | static const u8 *mpls_nh_via(const struct mpls_route *rt, | |
69 | const struct mpls_nh *nh) | |
70 | { | |
71 | return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh); | |
72 | } | |
73 | ||
f8efb73c | 74 | static unsigned int mpls_nh_header_size(const struct mpls_nh *nh) |
0189197f EB |
75 | { |
76 | /* The size of the layer 2.5 labels to be added for this route */ | |
f8efb73c | 77 | return nh->nh_labels * sizeof(struct mpls_shim_hdr); |
0189197f EB |
78 | } |
79 | ||
face0188 | 80 | unsigned int mpls_dev_mtu(const struct net_device *dev) |
0189197f EB |
81 | { |
82 | /* The amount of data the layer 2 frame can hold */ | |
83 | return dev->mtu; | |
84 | } | |
face0188 | 85 | EXPORT_SYMBOL_GPL(mpls_dev_mtu); |
0189197f | 86 | |
face0188 | 87 | bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) |
0189197f EB |
88 | { |
89 | if (skb->len <= mtu) | |
90 | return false; | |
91 | ||
92 | if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu) | |
93 | return false; | |
94 | ||
95 | return true; | |
96 | } | |
face0188 | 97 | EXPORT_SYMBOL_GPL(mpls_pkt_too_big); |
0189197f | 98 | |
1c78efa8 RS |
99 | static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt, |
100 | struct sk_buff *skb, bool bos) | |
f8efb73c | 101 | { |
1c78efa8 RS |
102 | struct mpls_entry_decoded dec; |
103 | struct mpls_shim_hdr *hdr; | |
104 | bool eli_seen = false; | |
105 | int label_index; | |
106 | int nh_index = 0; | |
107 | u32 hash = 0; | |
108 | ||
109 | /* No need to look further into packet if there's only | |
110 | * one path | |
111 | */ | |
112 | if (rt->rt_nhn == 1) | |
113 | goto out; | |
114 | ||
115 | for (label_index = 0; label_index < MAX_MP_SELECT_LABELS && !bos; | |
116 | label_index++) { | |
117 | if (!pskb_may_pull(skb, sizeof(*hdr) * label_index)) | |
118 | break; | |
119 | ||
120 | /* Read and decode the current label */ | |
121 | hdr = mpls_hdr(skb) + label_index; | |
122 | dec = mpls_entry_decode(hdr); | |
123 | ||
124 | /* RFC6790 - reserved labels MUST NOT be used as keys | |
125 | * for the load-balancing function | |
126 | */ | |
127 | if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) { | |
128 | hash = jhash_1word(dec.label, hash); | |
129 | ||
130 | /* The entropy label follows the entropy label | |
131 | * indicator, so this means that the entropy | |
132 | * label was just added to the hash - no need to | |
133 | * go any deeper either in the label stack or in the | |
134 | * payload | |
135 | */ | |
136 | if (eli_seen) | |
137 | break; | |
138 | } else if (dec.label == MPLS_LABEL_ENTROPY) { | |
139 | eli_seen = true; | |
140 | } | |
141 | ||
142 | bos = dec.bos; | |
143 | if (bos && pskb_may_pull(skb, sizeof(*hdr) * label_index + | |
144 | sizeof(struct iphdr))) { | |
145 | const struct iphdr *v4hdr; | |
146 | ||
147 | v4hdr = (const struct iphdr *)(mpls_hdr(skb) + | |
148 | label_index); | |
149 | if (v4hdr->version == 4) { | |
150 | hash = jhash_3words(ntohl(v4hdr->saddr), | |
151 | ntohl(v4hdr->daddr), | |
152 | v4hdr->protocol, hash); | |
153 | } else if (v4hdr->version == 6 && | |
154 | pskb_may_pull(skb, sizeof(*hdr) * label_index + | |
155 | sizeof(struct ipv6hdr))) { | |
156 | const struct ipv6hdr *v6hdr; | |
157 | ||
158 | v6hdr = (const struct ipv6hdr *)(mpls_hdr(skb) + | |
159 | label_index); | |
160 | ||
161 | hash = __ipv6_addr_jhash(&v6hdr->saddr, hash); | |
162 | hash = __ipv6_addr_jhash(&v6hdr->daddr, hash); | |
163 | hash = jhash_1word(v6hdr->nexthdr, hash); | |
164 | } | |
165 | } | |
166 | } | |
167 | ||
168 | nh_index = hash % rt->rt_nhn; | |
169 | out: | |
170 | return &rt->rt_nh[nh_index]; | |
f8efb73c RP |
171 | } |
172 | ||
0189197f EB |
173 | static bool mpls_egress(struct mpls_route *rt, struct sk_buff *skb, |
174 | struct mpls_entry_decoded dec) | |
175 | { | |
118d5234 RS |
176 | enum mpls_payload_type payload_type; |
177 | bool success = false; | |
0189197f | 178 | |
76fecd82 EB |
179 | /* The IPv4 code below accesses through the IPv4 header |
180 | * checksum, which is 12 bytes into the packet. | |
181 | * The IPv6 code below accesses through the IPv6 hop limit | |
182 | * which is 8 bytes into the packet. | |
183 | * | |
184 | * For all supported cases there should always be at least 12 | |
185 | * bytes of packet data present. The IPv4 header is 20 bytes | |
186 | * without options and the IPv6 header is always 40 bytes | |
187 | * long. | |
188 | */ | |
189 | if (!pskb_may_pull(skb, 12)) | |
190 | return false; | |
191 | ||
118d5234 RS |
192 | payload_type = rt->rt_payload_type; |
193 | if (payload_type == MPT_UNSPEC) | |
194 | payload_type = ip_hdr(skb)->version; | |
195 | ||
196 | switch (payload_type) { | |
197 | case MPT_IPV4: { | |
198 | struct iphdr *hdr4 = ip_hdr(skb); | |
0189197f EB |
199 | skb->protocol = htons(ETH_P_IP); |
200 | csum_replace2(&hdr4->check, | |
201 | htons(hdr4->ttl << 8), | |
202 | htons(dec.ttl << 8)); | |
203 | hdr4->ttl = dec.ttl; | |
118d5234 RS |
204 | success = true; |
205 | break; | |
0189197f | 206 | } |
118d5234 | 207 | case MPT_IPV6: { |
0189197f EB |
208 | struct ipv6hdr *hdr6 = ipv6_hdr(skb); |
209 | skb->protocol = htons(ETH_P_IPV6); | |
210 | hdr6->hop_limit = dec.ttl; | |
118d5234 RS |
211 | success = true; |
212 | break; | |
0189197f | 213 | } |
118d5234 RS |
214 | case MPT_UNSPEC: |
215 | break; | |
216 | } | |
217 | ||
0189197f EB |
218 | return success; |
219 | } | |
220 | ||
221 | static int mpls_forward(struct sk_buff *skb, struct net_device *dev, | |
222 | struct packet_type *pt, struct net_device *orig_dev) | |
223 | { | |
224 | struct net *net = dev_net(dev); | |
225 | struct mpls_shim_hdr *hdr; | |
226 | struct mpls_route *rt; | |
f8efb73c | 227 | struct mpls_nh *nh; |
0189197f EB |
228 | struct mpls_entry_decoded dec; |
229 | struct net_device *out_dev; | |
03c57747 | 230 | struct mpls_dev *mdev; |
0189197f EB |
231 | unsigned int hh_len; |
232 | unsigned int new_header_size; | |
233 | unsigned int mtu; | |
234 | int err; | |
235 | ||
236 | /* Careful this entire function runs inside of an rcu critical section */ | |
237 | ||
03c57747 | 238 | mdev = mpls_dev_get(dev); |
37bde799 | 239 | if (!mdev || !mdev->input_enabled) |
03c57747 RS |
240 | goto drop; |
241 | ||
0189197f EB |
242 | if (skb->pkt_type != PACKET_HOST) |
243 | goto drop; | |
244 | ||
245 | if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) | |
246 | goto drop; | |
247 | ||
248 | if (!pskb_may_pull(skb, sizeof(*hdr))) | |
249 | goto drop; | |
250 | ||
251 | /* Read and decode the label */ | |
252 | hdr = mpls_hdr(skb); | |
253 | dec = mpls_entry_decode(hdr); | |
254 | ||
255 | /* Pop the label */ | |
256 | skb_pull(skb, sizeof(*hdr)); | |
257 | skb_reset_network_header(skb); | |
258 | ||
259 | skb_orphan(skb); | |
260 | ||
261 | rt = mpls_route_input_rcu(net, dec.label); | |
262 | if (!rt) | |
263 | goto drop; | |
264 | ||
1c78efa8 | 265 | nh = mpls_select_multipath(rt, skb, dec.bos); |
f8efb73c RP |
266 | if (!nh) |
267 | goto drop; | |
268 | ||
0189197f | 269 | /* Find the output device */ |
f8efb73c | 270 | out_dev = rcu_dereference(nh->nh_dev); |
0189197f EB |
271 | if (!mpls_output_possible(out_dev)) |
272 | goto drop; | |
273 | ||
274 | if (skb_warn_if_lro(skb)) | |
275 | goto drop; | |
276 | ||
277 | skb_forward_csum(skb); | |
278 | ||
279 | /* Verify ttl is valid */ | |
aa7da937 | 280 | if (dec.ttl <= 1) |
0189197f EB |
281 | goto drop; |
282 | dec.ttl -= 1; | |
283 | ||
284 | /* Verify the destination can hold the packet */ | |
f8efb73c | 285 | new_header_size = mpls_nh_header_size(nh); |
0189197f EB |
286 | mtu = mpls_dev_mtu(out_dev); |
287 | if (mpls_pkt_too_big(skb, mtu - new_header_size)) | |
288 | goto drop; | |
289 | ||
290 | hh_len = LL_RESERVED_SPACE(out_dev); | |
291 | if (!out_dev->header_ops) | |
292 | hh_len = 0; | |
293 | ||
294 | /* Ensure there is enough space for the headers in the skb */ | |
295 | if (skb_cow(skb, hh_len + new_header_size)) | |
296 | goto drop; | |
297 | ||
298 | skb->dev = out_dev; | |
299 | skb->protocol = htons(ETH_P_MPLS_UC); | |
300 | ||
301 | if (unlikely(!new_header_size && dec.bos)) { | |
302 | /* Penultimate hop popping */ | |
303 | if (!mpls_egress(rt, skb, dec)) | |
304 | goto drop; | |
305 | } else { | |
306 | bool bos; | |
307 | int i; | |
308 | skb_push(skb, new_header_size); | |
309 | skb_reset_network_header(skb); | |
310 | /* Push the new labels */ | |
311 | hdr = mpls_hdr(skb); | |
312 | bos = dec.bos; | |
f8efb73c RP |
313 | for (i = nh->nh_labels - 1; i >= 0; i--) { |
314 | hdr[i] = mpls_entry_encode(nh->nh_label[i], | |
315 | dec.ttl, 0, bos); | |
0189197f EB |
316 | bos = false; |
317 | } | |
318 | } | |
319 | ||
cf4b24f0 | 320 | err = neigh_xmit(nh->nh_via_table, out_dev, mpls_nh_via(rt, nh), skb); |
0189197f EB |
321 | if (err) |
322 | net_dbg_ratelimited("%s: packet transmission failed: %d\n", | |
323 | __func__, err); | |
324 | return 0; | |
325 | ||
326 | drop: | |
327 | kfree_skb(skb); | |
328 | return NET_RX_DROP; | |
329 | } | |
330 | ||
331 | static struct packet_type mpls_packet_type __read_mostly = { | |
332 | .type = cpu_to_be16(ETH_P_MPLS_UC), | |
333 | .func = mpls_forward, | |
334 | }; | |
335 | ||
f0126539 | 336 | static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = { |
03c05665 EB |
337 | [RTA_DST] = { .type = NLA_U32 }, |
338 | [RTA_OIF] = { .type = NLA_U32 }, | |
339 | }; | |
340 | ||
a2519929 | 341 | struct mpls_route_config { |
118d5234 RS |
342 | u32 rc_protocol; |
343 | u32 rc_ifindex; | |
f8efb73c RP |
344 | u8 rc_via_table; |
345 | u8 rc_via_alen; | |
118d5234 RS |
346 | u8 rc_via[MAX_VIA_ALEN]; |
347 | u32 rc_label; | |
f8efb73c | 348 | u8 rc_output_labels; |
118d5234 RS |
349 | u32 rc_output_label[MAX_NEW_LABELS]; |
350 | u32 rc_nlflags; | |
351 | enum mpls_payload_type rc_payload_type; | |
352 | struct nl_info rc_nlinfo; | |
f8efb73c RP |
353 | struct rtnexthop *rc_mp; |
354 | int rc_mp_len; | |
a2519929 EB |
355 | }; |
356 | ||
cf4b24f0 | 357 | static struct mpls_route *mpls_rt_alloc(int num_nh, u8 max_alen) |
0189197f | 358 | { |
cf4b24f0 | 359 | u8 max_alen_aligned = ALIGN(max_alen, VIA_ALEN_ALIGN); |
0189197f EB |
360 | struct mpls_route *rt; |
361 | ||
cf4b24f0 RS |
362 | rt = kzalloc(ALIGN(sizeof(*rt) + num_nh * sizeof(*rt->rt_nh), |
363 | VIA_ALEN_ALIGN) + | |
364 | num_nh * max_alen_aligned, | |
f8efb73c | 365 | GFP_KERNEL); |
cf4b24f0 | 366 | if (rt) { |
f8efb73c | 367 | rt->rt_nhn = num_nh; |
cf4b24f0 RS |
368 | rt->rt_max_alen = max_alen_aligned; |
369 | } | |
f8efb73c | 370 | |
0189197f EB |
371 | return rt; |
372 | } | |
373 | ||
374 | static void mpls_rt_free(struct mpls_route *rt) | |
375 | { | |
376 | if (rt) | |
377 | kfree_rcu(rt, rt_rcu); | |
378 | } | |
379 | ||
8de147dc EB |
380 | static void mpls_notify_route(struct net *net, unsigned index, |
381 | struct mpls_route *old, struct mpls_route *new, | |
382 | const struct nl_info *info) | |
383 | { | |
384 | struct nlmsghdr *nlh = info ? info->nlh : NULL; | |
385 | unsigned portid = info ? info->portid : 0; | |
386 | int event = new ? RTM_NEWROUTE : RTM_DELROUTE; | |
387 | struct mpls_route *rt = new ? new : old; | |
388 | unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0; | |
389 | /* Ignore reserved labels for now */ | |
a6affd24 | 390 | if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED)) |
8de147dc EB |
391 | rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags); |
392 | } | |
393 | ||
0189197f | 394 | static void mpls_route_update(struct net *net, unsigned index, |
f8efb73c | 395 | struct mpls_route *new, |
0189197f EB |
396 | const struct nl_info *info) |
397 | { | |
19d0c341 | 398 | struct mpls_route __rcu **platform_label; |
f8efb73c | 399 | struct mpls_route *rt; |
0189197f EB |
400 | |
401 | ASSERT_RTNL(); | |
402 | ||
19d0c341 EB |
403 | platform_label = rtnl_dereference(net->mpls.platform_label); |
404 | rt = rtnl_dereference(platform_label[index]); | |
f8efb73c | 405 | rcu_assign_pointer(platform_label[index], new); |
0189197f | 406 | |
f8efb73c | 407 | mpls_notify_route(net, index, rt, new, info); |
8de147dc | 408 | |
0189197f | 409 | /* If we removed a route free it now */ |
f8efb73c | 410 | mpls_rt_free(rt); |
0189197f EB |
411 | } |
412 | ||
a2519929 EB |
413 | static unsigned find_free_label(struct net *net) |
414 | { | |
19d0c341 EB |
415 | struct mpls_route __rcu **platform_label; |
416 | size_t platform_labels; | |
a2519929 | 417 | unsigned index; |
19d0c341 EB |
418 | |
419 | platform_label = rtnl_dereference(net->mpls.platform_label); | |
420 | platform_labels = net->mpls.platform_labels; | |
a6affd24 RS |
421 | for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels; |
422 | index++) { | |
19d0c341 | 423 | if (!rtnl_dereference(platform_label[index])) |
a2519929 EB |
424 | return index; |
425 | } | |
426 | return LABEL_NOT_SPECIFIED; | |
427 | } | |
428 | ||
bf21563a | 429 | #if IS_ENABLED(CONFIG_INET) |
cf4b24f0 RS |
430 | static struct net_device *inet_fib_lookup_dev(struct net *net, |
431 | const void *addr) | |
01faef2c | 432 | { |
5a9348b5 | 433 | struct net_device *dev; |
01faef2c RP |
434 | struct rtable *rt; |
435 | struct in_addr daddr; | |
436 | ||
437 | memcpy(&daddr, addr, sizeof(struct in_addr)); | |
438 | rt = ip_route_output(net, daddr.s_addr, 0, 0, 0); | |
439 | if (IS_ERR(rt)) | |
5a9348b5 | 440 | return ERR_CAST(rt); |
01faef2c RP |
441 | |
442 | dev = rt->dst.dev; | |
443 | dev_hold(dev); | |
444 | ||
445 | ip_rt_put(rt); | |
446 | ||
01faef2c | 447 | return dev; |
bf21563a RP |
448 | } |
449 | #else | |
cf4b24f0 RS |
450 | static struct net_device *inet_fib_lookup_dev(struct net *net, |
451 | const void *addr) | |
bf21563a RP |
452 | { |
453 | return ERR_PTR(-EAFNOSUPPORT); | |
01faef2c | 454 | } |
bf21563a | 455 | #endif |
01faef2c | 456 | |
bf21563a | 457 | #if IS_ENABLED(CONFIG_IPV6) |
cf4b24f0 RS |
458 | static struct net_device *inet6_fib_lookup_dev(struct net *net, |
459 | const void *addr) | |
01faef2c | 460 | { |
5a9348b5 | 461 | struct net_device *dev; |
01faef2c RP |
462 | struct dst_entry *dst; |
463 | struct flowi6 fl6; | |
bf21563a RP |
464 | int err; |
465 | ||
466 | if (!ipv6_stub) | |
467 | return ERR_PTR(-EAFNOSUPPORT); | |
01faef2c RP |
468 | |
469 | memset(&fl6, 0, sizeof(fl6)); | |
470 | memcpy(&fl6.daddr, addr, sizeof(struct in6_addr)); | |
bf21563a RP |
471 | err = ipv6_stub->ipv6_dst_lookup(net, NULL, &dst, &fl6); |
472 | if (err) | |
5a9348b5 | 473 | return ERR_PTR(err); |
01faef2c RP |
474 | |
475 | dev = dst->dev; | |
476 | dev_hold(dev); | |
01faef2c RP |
477 | dst_release(dst); |
478 | ||
479 | return dev; | |
480 | } | |
bf21563a | 481 | #else |
cf4b24f0 RS |
482 | static struct net_device *inet6_fib_lookup_dev(struct net *net, |
483 | const void *addr) | |
bf21563a RP |
484 | { |
485 | return ERR_PTR(-EAFNOSUPPORT); | |
486 | } | |
487 | #endif | |
01faef2c RP |
488 | |
489 | static struct net_device *find_outdev(struct net *net, | |
cf4b24f0 | 490 | struct mpls_route *rt, |
f8efb73c | 491 | struct mpls_nh *nh, int oif) |
01faef2c RP |
492 | { |
493 | struct net_device *dev = NULL; | |
494 | ||
f8efb73c RP |
495 | if (!oif) { |
496 | switch (nh->nh_via_table) { | |
01faef2c | 497 | case NEIGH_ARP_TABLE: |
cf4b24f0 | 498 | dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh)); |
01faef2c RP |
499 | break; |
500 | case NEIGH_ND_TABLE: | |
cf4b24f0 | 501 | dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh)); |
01faef2c RP |
502 | break; |
503 | case NEIGH_LINK_TABLE: | |
504 | break; | |
505 | } | |
506 | } else { | |
f8efb73c | 507 | dev = dev_get_by_index(net, oif); |
01faef2c RP |
508 | } |
509 | ||
3dcb615e RP |
510 | if (!dev) |
511 | return ERR_PTR(-ENODEV); | |
512 | ||
f8efb73c RP |
513 | /* The caller is holding rtnl anyways, so release the dev reference */ |
514 | dev_put(dev); | |
515 | ||
01faef2c RP |
516 | return dev; |
517 | } | |
518 | ||
cf4b24f0 RS |
519 | static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt, |
520 | struct mpls_nh *nh, int oif) | |
f8efb73c RP |
521 | { |
522 | struct net_device *dev = NULL; | |
523 | int err = -ENODEV; | |
524 | ||
cf4b24f0 | 525 | dev = find_outdev(net, rt, nh, oif); |
f8efb73c RP |
526 | if (IS_ERR(dev)) { |
527 | err = PTR_ERR(dev); | |
528 | dev = NULL; | |
529 | goto errout; | |
530 | } | |
531 | ||
532 | /* Ensure this is a supported device */ | |
533 | err = -EINVAL; | |
534 | if (!mpls_dev_get(dev)) | |
535 | goto errout; | |
536 | ||
537 | RCU_INIT_POINTER(nh->nh_dev, dev); | |
538 | ||
539 | return 0; | |
540 | ||
541 | errout: | |
542 | return err; | |
543 | } | |
544 | ||
545 | static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg, | |
546 | struct mpls_route *rt) | |
547 | { | |
548 | struct net *net = cfg->rc_nlinfo.nl_net; | |
549 | struct mpls_nh *nh = rt->rt_nh; | |
550 | int err; | |
551 | int i; | |
552 | ||
553 | if (!nh) | |
554 | return -ENOMEM; | |
555 | ||
556 | err = -EINVAL; | |
557 | /* Ensure only a supported number of labels are present */ | |
558 | if (cfg->rc_output_labels > MAX_NEW_LABELS) | |
559 | goto errout; | |
560 | ||
561 | nh->nh_labels = cfg->rc_output_labels; | |
562 | for (i = 0; i < nh->nh_labels; i++) | |
563 | nh->nh_label[i] = cfg->rc_output_label[i]; | |
564 | ||
565 | nh->nh_via_table = cfg->rc_via_table; | |
cf4b24f0 | 566 | memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen); |
f8efb73c RP |
567 | nh->nh_via_alen = cfg->rc_via_alen; |
568 | ||
cf4b24f0 | 569 | err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex); |
f8efb73c RP |
570 | if (err) |
571 | goto errout; | |
572 | ||
573 | return 0; | |
574 | ||
575 | errout: | |
576 | return err; | |
577 | } | |
578 | ||
cf4b24f0 RS |
579 | static int mpls_nh_build(struct net *net, struct mpls_route *rt, |
580 | struct mpls_nh *nh, int oif, | |
581 | struct nlattr *via, struct nlattr *newdst) | |
f8efb73c RP |
582 | { |
583 | int err = -ENOMEM; | |
584 | ||
585 | if (!nh) | |
586 | goto errout; | |
587 | ||
588 | if (newdst) { | |
589 | err = nla_get_labels(newdst, MAX_NEW_LABELS, | |
590 | &nh->nh_labels, nh->nh_label); | |
591 | if (err) | |
592 | goto errout; | |
593 | } | |
594 | ||
595 | err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table, | |
cf4b24f0 | 596 | __mpls_nh_via(rt, nh)); |
f8efb73c RP |
597 | if (err) |
598 | goto errout; | |
599 | ||
cf4b24f0 | 600 | err = mpls_nh_assign_dev(net, rt, nh, oif); |
f8efb73c RP |
601 | if (err) |
602 | goto errout; | |
603 | ||
604 | return 0; | |
605 | ||
606 | errout: | |
607 | return err; | |
608 | } | |
609 | ||
cf4b24f0 RS |
610 | static int mpls_count_nexthops(struct rtnexthop *rtnh, int len, |
611 | u8 cfg_via_alen, u8 *max_via_alen) | |
f8efb73c RP |
612 | { |
613 | int nhs = 0; | |
614 | int remaining = len; | |
615 | ||
cf4b24f0 RS |
616 | if (!rtnh) { |
617 | *max_via_alen = cfg_via_alen; | |
618 | return 1; | |
619 | } | |
620 | ||
621 | *max_via_alen = 0; | |
622 | ||
f8efb73c | 623 | while (rtnh_ok(rtnh, remaining)) { |
cf4b24f0 RS |
624 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); |
625 | int attrlen; | |
626 | ||
627 | attrlen = rtnh_attrlen(rtnh); | |
628 | nla = nla_find(attrs, attrlen, RTA_VIA); | |
629 | if (nla && nla_len(nla) >= | |
630 | offsetof(struct rtvia, rtvia_addr)) { | |
631 | int via_alen = nla_len(nla) - | |
632 | offsetof(struct rtvia, rtvia_addr); | |
633 | ||
634 | if (via_alen <= MAX_VIA_ALEN) | |
635 | *max_via_alen = max_t(u16, *max_via_alen, | |
636 | via_alen); | |
637 | } | |
638 | ||
f8efb73c RP |
639 | nhs++; |
640 | rtnh = rtnh_next(rtnh, &remaining); | |
641 | } | |
642 | ||
643 | /* leftover implies invalid nexthop configuration, discard it */ | |
644 | return remaining > 0 ? 0 : nhs; | |
645 | } | |
646 | ||
647 | static int mpls_nh_build_multi(struct mpls_route_config *cfg, | |
648 | struct mpls_route *rt) | |
649 | { | |
650 | struct rtnexthop *rtnh = cfg->rc_mp; | |
651 | struct nlattr *nla_via, *nla_newdst; | |
652 | int remaining = cfg->rc_mp_len; | |
653 | int nhs = 0; | |
654 | int err = 0; | |
655 | ||
656 | change_nexthops(rt) { | |
657 | int attrlen; | |
658 | ||
659 | nla_via = NULL; | |
660 | nla_newdst = NULL; | |
661 | ||
662 | err = -EINVAL; | |
663 | if (!rtnh_ok(rtnh, remaining)) | |
664 | goto errout; | |
665 | ||
1c78efa8 RS |
666 | /* neither weighted multipath nor any flags |
667 | * are supported | |
668 | */ | |
669 | if (rtnh->rtnh_hops || rtnh->rtnh_flags) | |
670 | goto errout; | |
671 | ||
f8efb73c RP |
672 | attrlen = rtnh_attrlen(rtnh); |
673 | if (attrlen > 0) { | |
674 | struct nlattr *attrs = rtnh_attrs(rtnh); | |
675 | ||
676 | nla_via = nla_find(attrs, attrlen, RTA_VIA); | |
677 | nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST); | |
678 | } | |
679 | ||
680 | if (!nla_via) | |
681 | goto errout; | |
682 | ||
cf4b24f0 | 683 | err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh, |
f8efb73c RP |
684 | rtnh->rtnh_ifindex, nla_via, |
685 | nla_newdst); | |
686 | if (err) | |
687 | goto errout; | |
688 | ||
689 | rtnh = rtnh_next(rtnh, &remaining); | |
690 | nhs++; | |
691 | } endfor_nexthops(rt); | |
692 | ||
693 | rt->rt_nhn = nhs; | |
694 | ||
695 | return 0; | |
696 | ||
697 | errout: | |
698 | return err; | |
699 | } | |
700 | ||
a2519929 EB |
701 | static int mpls_route_add(struct mpls_route_config *cfg) |
702 | { | |
19d0c341 | 703 | struct mpls_route __rcu **platform_label; |
a2519929 | 704 | struct net *net = cfg->rc_nlinfo.nl_net; |
a2519929 | 705 | struct mpls_route *rt, *old; |
a2519929 | 706 | int err = -EINVAL; |
cf4b24f0 | 707 | u8 max_via_alen; |
f8efb73c | 708 | unsigned index; |
cf4b24f0 | 709 | int nhs; |
a2519929 EB |
710 | |
711 | index = cfg->rc_label; | |
712 | ||
713 | /* If a label was not specified during insert pick one */ | |
714 | if ((index == LABEL_NOT_SPECIFIED) && | |
715 | (cfg->rc_nlflags & NLM_F_CREATE)) { | |
716 | index = find_free_label(net); | |
717 | } | |
718 | ||
a6affd24 RS |
719 | /* Reserved labels may not be set */ |
720 | if (index < MPLS_LABEL_FIRST_UNRESERVED) | |
a2519929 EB |
721 | goto errout; |
722 | ||
723 | /* The full 20 bit range may not be supported. */ | |
724 | if (index >= net->mpls.platform_labels) | |
725 | goto errout; | |
726 | ||
a2519929 | 727 | /* Append makes no sense with mpls */ |
0f7bbd58 | 728 | err = -EOPNOTSUPP; |
a2519929 EB |
729 | if (cfg->rc_nlflags & NLM_F_APPEND) |
730 | goto errout; | |
731 | ||
732 | err = -EEXIST; | |
19d0c341 EB |
733 | platform_label = rtnl_dereference(net->mpls.platform_label); |
734 | old = rtnl_dereference(platform_label[index]); | |
a2519929 EB |
735 | if ((cfg->rc_nlflags & NLM_F_EXCL) && old) |
736 | goto errout; | |
737 | ||
738 | err = -EEXIST; | |
739 | if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old) | |
740 | goto errout; | |
741 | ||
742 | err = -ENOENT; | |
743 | if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old) | |
744 | goto errout; | |
745 | ||
cf4b24f0 RS |
746 | err = -EINVAL; |
747 | nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len, | |
748 | cfg->rc_via_alen, &max_via_alen); | |
749 | if (nhs == 0) | |
750 | goto errout; | |
f8efb73c | 751 | |
a2519929 | 752 | err = -ENOMEM; |
cf4b24f0 | 753 | rt = mpls_rt_alloc(nhs, max_via_alen); |
a2519929 EB |
754 | if (!rt) |
755 | goto errout; | |
756 | ||
a2519929 | 757 | rt->rt_protocol = cfg->rc_protocol; |
118d5234 | 758 | rt->rt_payload_type = cfg->rc_payload_type; |
a2519929 | 759 | |
f8efb73c RP |
760 | if (cfg->rc_mp) |
761 | err = mpls_nh_build_multi(cfg, rt); | |
762 | else | |
763 | err = mpls_nh_build_from_cfg(cfg, rt); | |
764 | if (err) | |
765 | goto freert; | |
766 | ||
767 | mpls_route_update(net, index, rt, &cfg->rc_nlinfo); | |
a2519929 | 768 | |
a2519929 EB |
769 | return 0; |
770 | ||
f8efb73c RP |
771 | freert: |
772 | mpls_rt_free(rt); | |
a2519929 | 773 | errout: |
a2519929 EB |
774 | return err; |
775 | } | |
776 | ||
777 | static int mpls_route_del(struct mpls_route_config *cfg) | |
778 | { | |
779 | struct net *net = cfg->rc_nlinfo.nl_net; | |
780 | unsigned index; | |
781 | int err = -EINVAL; | |
782 | ||
783 | index = cfg->rc_label; | |
784 | ||
a6affd24 RS |
785 | /* Reserved labels may not be removed */ |
786 | if (index < MPLS_LABEL_FIRST_UNRESERVED) | |
a2519929 EB |
787 | goto errout; |
788 | ||
789 | /* The full 20 bit range may not be supported */ | |
790 | if (index >= net->mpls.platform_labels) | |
791 | goto errout; | |
792 | ||
f8efb73c | 793 | mpls_route_update(net, index, NULL, &cfg->rc_nlinfo); |
a2519929 EB |
794 | |
795 | err = 0; | |
796 | errout: | |
797 | return err; | |
798 | } | |
799 | ||
37bde799 RS |
800 | #define MPLS_PERDEV_SYSCTL_OFFSET(field) \ |
801 | (&((struct mpls_dev *)0)->field) | |
802 | ||
803 | static const struct ctl_table mpls_dev_table[] = { | |
804 | { | |
805 | .procname = "input", | |
806 | .maxlen = sizeof(int), | |
807 | .mode = 0644, | |
808 | .proc_handler = proc_dointvec, | |
809 | .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled), | |
810 | }, | |
811 | { } | |
812 | }; | |
813 | ||
814 | static int mpls_dev_sysctl_register(struct net_device *dev, | |
815 | struct mpls_dev *mdev) | |
816 | { | |
817 | char path[sizeof("net/mpls/conf/") + IFNAMSIZ]; | |
818 | struct ctl_table *table; | |
819 | int i; | |
820 | ||
821 | table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL); | |
822 | if (!table) | |
823 | goto out; | |
824 | ||
825 | /* Table data contains only offsets relative to the base of | |
826 | * the mdev at this point, so make them absolute. | |
827 | */ | |
828 | for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) | |
829 | table[i].data = (char *)mdev + (uintptr_t)table[i].data; | |
830 | ||
831 | snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name); | |
832 | ||
833 | mdev->sysctl = register_net_sysctl(dev_net(dev), path, table); | |
834 | if (!mdev->sysctl) | |
835 | goto free; | |
836 | ||
837 | return 0; | |
838 | ||
839 | free: | |
840 | kfree(table); | |
841 | out: | |
842 | return -ENOBUFS; | |
843 | } | |
844 | ||
845 | static void mpls_dev_sysctl_unregister(struct mpls_dev *mdev) | |
846 | { | |
847 | struct ctl_table *table; | |
848 | ||
849 | table = mdev->sysctl->ctl_table_arg; | |
850 | unregister_net_sysctl_table(mdev->sysctl); | |
851 | kfree(table); | |
852 | } | |
853 | ||
03c57747 RS |
854 | static struct mpls_dev *mpls_add_dev(struct net_device *dev) |
855 | { | |
856 | struct mpls_dev *mdev; | |
857 | int err = -ENOMEM; | |
858 | ||
859 | ASSERT_RTNL(); | |
860 | ||
861 | mdev = kzalloc(sizeof(*mdev), GFP_KERNEL); | |
862 | if (!mdev) | |
863 | return ERR_PTR(err); | |
864 | ||
37bde799 RS |
865 | err = mpls_dev_sysctl_register(dev, mdev); |
866 | if (err) | |
867 | goto free; | |
868 | ||
03c57747 RS |
869 | rcu_assign_pointer(dev->mpls_ptr, mdev); |
870 | ||
871 | return mdev; | |
37bde799 RS |
872 | |
873 | free: | |
874 | kfree(mdev); | |
875 | return ERR_PTR(err); | |
03c57747 RS |
876 | } |
877 | ||
0189197f EB |
878 | static void mpls_ifdown(struct net_device *dev) |
879 | { | |
19d0c341 | 880 | struct mpls_route __rcu **platform_label; |
0189197f | 881 | struct net *net = dev_net(dev); |
03c57747 | 882 | struct mpls_dev *mdev; |
0189197f EB |
883 | unsigned index; |
884 | ||
19d0c341 | 885 | platform_label = rtnl_dereference(net->mpls.platform_label); |
0189197f | 886 | for (index = 0; index < net->mpls.platform_labels; index++) { |
19d0c341 | 887 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); |
0189197f EB |
888 | if (!rt) |
889 | continue; | |
f8efb73c RP |
890 | for_nexthops(rt) { |
891 | if (rtnl_dereference(nh->nh_dev) != dev) | |
892 | continue; | |
893 | nh->nh_dev = NULL; | |
894 | } endfor_nexthops(rt); | |
0189197f | 895 | } |
03c57747 RS |
896 | |
897 | mdev = mpls_dev_get(dev); | |
898 | if (!mdev) | |
899 | return; | |
900 | ||
37bde799 RS |
901 | mpls_dev_sysctl_unregister(mdev); |
902 | ||
03c57747 RS |
903 | RCU_INIT_POINTER(dev->mpls_ptr, NULL); |
904 | ||
25cc8f07 | 905 | kfree_rcu(mdev, rcu); |
0189197f EB |
906 | } |
907 | ||
908 | static int mpls_dev_notify(struct notifier_block *this, unsigned long event, | |
909 | void *ptr) | |
910 | { | |
911 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
03c57747 | 912 | struct mpls_dev *mdev; |
0189197f EB |
913 | |
914 | switch(event) { | |
03c57747 RS |
915 | case NETDEV_REGISTER: |
916 | /* For now just support ethernet devices */ | |
917 | if ((dev->type == ARPHRD_ETHER) || | |
918 | (dev->type == ARPHRD_LOOPBACK)) { | |
919 | mdev = mpls_add_dev(dev); | |
920 | if (IS_ERR(mdev)) | |
921 | return notifier_from_errno(PTR_ERR(mdev)); | |
922 | } | |
923 | break; | |
924 | ||
0189197f EB |
925 | case NETDEV_UNREGISTER: |
926 | mpls_ifdown(dev); | |
927 | break; | |
0fae3bf0 RS |
928 | case NETDEV_CHANGENAME: |
929 | mdev = mpls_dev_get(dev); | |
930 | if (mdev) { | |
931 | int err; | |
932 | ||
933 | mpls_dev_sysctl_unregister(mdev); | |
934 | err = mpls_dev_sysctl_register(dev, mdev); | |
935 | if (err) | |
936 | return notifier_from_errno(err); | |
937 | } | |
938 | break; | |
0189197f EB |
939 | } |
940 | return NOTIFY_OK; | |
941 | } | |
942 | ||
943 | static struct notifier_block mpls_dev_notifier = { | |
944 | .notifier_call = mpls_dev_notify, | |
945 | }; | |
946 | ||
03c05665 | 947 | static int nla_put_via(struct sk_buff *skb, |
b79bda3d | 948 | u8 table, const void *addr, int alen) |
03c05665 | 949 | { |
b79bda3d EB |
950 | static const int table_to_family[NEIGH_NR_TABLES + 1] = { |
951 | AF_INET, AF_INET6, AF_DECnet, AF_PACKET, | |
952 | }; | |
03c05665 EB |
953 | struct nlattr *nla; |
954 | struct rtvia *via; | |
b79bda3d | 955 | int family = AF_UNSPEC; |
03c05665 EB |
956 | |
957 | nla = nla_reserve(skb, RTA_VIA, alen + 2); | |
958 | if (!nla) | |
959 | return -EMSGSIZE; | |
960 | ||
b79bda3d EB |
961 | if (table <= NEIGH_NR_TABLES) |
962 | family = table_to_family[table]; | |
963 | ||
03c05665 EB |
964 | via = nla_data(nla); |
965 | via->rtvia_family = family; | |
966 | memcpy(via->rtvia_addr, addr, alen); | |
967 | return 0; | |
968 | } | |
969 | ||
966bae33 EB |
970 | int nla_put_labels(struct sk_buff *skb, int attrtype, |
971 | u8 labels, const u32 label[]) | |
972 | { | |
973 | struct nlattr *nla; | |
974 | struct mpls_shim_hdr *nla_label; | |
975 | bool bos; | |
976 | int i; | |
977 | nla = nla_reserve(skb, attrtype, labels*4); | |
978 | if (!nla) | |
979 | return -EMSGSIZE; | |
980 | ||
981 | nla_label = nla_data(nla); | |
982 | bos = true; | |
983 | for (i = labels - 1; i >= 0; i--) { | |
984 | nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos); | |
985 | bos = false; | |
986 | } | |
987 | ||
988 | return 0; | |
989 | } | |
face0188 | 990 | EXPORT_SYMBOL_GPL(nla_put_labels); |
966bae33 EB |
991 | |
992 | int nla_get_labels(const struct nlattr *nla, | |
f8efb73c | 993 | u32 max_labels, u8 *labels, u32 label[]) |
966bae33 EB |
994 | { |
995 | unsigned len = nla_len(nla); | |
996 | unsigned nla_labels; | |
997 | struct mpls_shim_hdr *nla_label; | |
998 | bool bos; | |
999 | int i; | |
1000 | ||
1001 | /* len needs to be an even multiple of 4 (the label size) */ | |
1002 | if (len & 3) | |
1003 | return -EINVAL; | |
1004 | ||
1005 | /* Limit the number of new labels allowed */ | |
1006 | nla_labels = len/4; | |
1007 | if (nla_labels > max_labels) | |
1008 | return -EINVAL; | |
1009 | ||
1010 | nla_label = nla_data(nla); | |
1011 | bos = true; | |
1012 | for (i = nla_labels - 1; i >= 0; i--, bos = false) { | |
1013 | struct mpls_entry_decoded dec; | |
1014 | dec = mpls_entry_decode(nla_label + i); | |
1015 | ||
1016 | /* Ensure the bottom of stack flag is properly set | |
1017 | * and ttl and tc are both clear. | |
1018 | */ | |
1019 | if ((dec.bos != bos) || dec.ttl || dec.tc) | |
1020 | return -EINVAL; | |
1021 | ||
5a9ab017 | 1022 | switch (dec.label) { |
78f5b899 | 1023 | case MPLS_LABEL_IMPLNULL: |
5a9ab017 RS |
1024 | /* RFC3032: This is a label that an LSR may |
1025 | * assign and distribute, but which never | |
1026 | * actually appears in the encapsulation. | |
1027 | */ | |
1028 | return -EINVAL; | |
1029 | } | |
1030 | ||
966bae33 EB |
1031 | label[i] = dec.label; |
1032 | } | |
1033 | *labels = nla_labels; | |
1034 | return 0; | |
1035 | } | |
face0188 | 1036 | EXPORT_SYMBOL_GPL(nla_get_labels); |
966bae33 | 1037 | |
f8efb73c RP |
1038 | int nla_get_via(const struct nlattr *nla, u8 *via_alen, |
1039 | u8 *via_table, u8 via_addr[]) | |
1040 | { | |
1041 | struct rtvia *via = nla_data(nla); | |
1042 | int err = -EINVAL; | |
1043 | int alen; | |
1044 | ||
1045 | if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) | |
1046 | goto errout; | |
1047 | alen = nla_len(nla) - | |
1048 | offsetof(struct rtvia, rtvia_addr); | |
1049 | if (alen > MAX_VIA_ALEN) | |
1050 | goto errout; | |
1051 | ||
1052 | /* Validate the address family */ | |
1053 | switch (via->rtvia_family) { | |
1054 | case AF_PACKET: | |
1055 | *via_table = NEIGH_LINK_TABLE; | |
1056 | break; | |
1057 | case AF_INET: | |
1058 | *via_table = NEIGH_ARP_TABLE; | |
1059 | if (alen != 4) | |
1060 | goto errout; | |
1061 | break; | |
1062 | case AF_INET6: | |
1063 | *via_table = NEIGH_ND_TABLE; | |
1064 | if (alen != 16) | |
1065 | goto errout; | |
1066 | break; | |
1067 | default: | |
1068 | /* Unsupported address family */ | |
1069 | goto errout; | |
1070 | } | |
1071 | ||
1072 | memcpy(via_addr, via->rtvia_addr, alen); | |
1073 | *via_alen = alen; | |
1074 | err = 0; | |
1075 | ||
1076 | errout: | |
1077 | return err; | |
1078 | } | |
1079 | ||
03c05665 EB |
1080 | static int rtm_to_route_config(struct sk_buff *skb, struct nlmsghdr *nlh, |
1081 | struct mpls_route_config *cfg) | |
1082 | { | |
1083 | struct rtmsg *rtm; | |
1084 | struct nlattr *tb[RTA_MAX+1]; | |
1085 | int index; | |
1086 | int err; | |
1087 | ||
1088 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_mpls_policy); | |
1089 | if (err < 0) | |
1090 | goto errout; | |
1091 | ||
1092 | err = -EINVAL; | |
1093 | rtm = nlmsg_data(nlh); | |
1094 | memset(cfg, 0, sizeof(*cfg)); | |
1095 | ||
1096 | if (rtm->rtm_family != AF_MPLS) | |
1097 | goto errout; | |
1098 | if (rtm->rtm_dst_len != 20) | |
1099 | goto errout; | |
1100 | if (rtm->rtm_src_len != 0) | |
1101 | goto errout; | |
1102 | if (rtm->rtm_tos != 0) | |
1103 | goto errout; | |
1104 | if (rtm->rtm_table != RT_TABLE_MAIN) | |
1105 | goto errout; | |
1106 | /* Any value is acceptable for rtm_protocol */ | |
1107 | ||
1108 | /* As mpls uses destination specific addresses | |
1109 | * (or source specific address in the case of multicast) | |
1110 | * all addresses have universal scope. | |
1111 | */ | |
1112 | if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) | |
1113 | goto errout; | |
1114 | if (rtm->rtm_type != RTN_UNICAST) | |
1115 | goto errout; | |
1116 | if (rtm->rtm_flags != 0) | |
1117 | goto errout; | |
1118 | ||
1119 | cfg->rc_label = LABEL_NOT_SPECIFIED; | |
1120 | cfg->rc_protocol = rtm->rtm_protocol; | |
1121 | cfg->rc_nlflags = nlh->nlmsg_flags; | |
1122 | cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid; | |
1123 | cfg->rc_nlinfo.nlh = nlh; | |
1124 | cfg->rc_nlinfo.nl_net = sock_net(skb->sk); | |
1125 | ||
1126 | for (index = 0; index <= RTA_MAX; index++) { | |
1127 | struct nlattr *nla = tb[index]; | |
1128 | if (!nla) | |
1129 | continue; | |
1130 | ||
1131 | switch(index) { | |
1132 | case RTA_OIF: | |
1133 | cfg->rc_ifindex = nla_get_u32(nla); | |
1134 | break; | |
1135 | case RTA_NEWDST: | |
1136 | if (nla_get_labels(nla, MAX_NEW_LABELS, | |
1137 | &cfg->rc_output_labels, | |
1138 | cfg->rc_output_label)) | |
1139 | goto errout; | |
1140 | break; | |
1141 | case RTA_DST: | |
1142 | { | |
f8efb73c | 1143 | u8 label_count; |
03c05665 EB |
1144 | if (nla_get_labels(nla, 1, &label_count, |
1145 | &cfg->rc_label)) | |
1146 | goto errout; | |
1147 | ||
a6affd24 RS |
1148 | /* Reserved labels may not be set */ |
1149 | if (cfg->rc_label < MPLS_LABEL_FIRST_UNRESERVED) | |
03c05665 EB |
1150 | goto errout; |
1151 | ||
1152 | break; | |
1153 | } | |
1154 | case RTA_VIA: | |
1155 | { | |
f8efb73c RP |
1156 | if (nla_get_via(nla, &cfg->rc_via_alen, |
1157 | &cfg->rc_via_table, cfg->rc_via)) | |
03c05665 | 1158 | goto errout; |
f8efb73c RP |
1159 | break; |
1160 | } | |
1161 | case RTA_MULTIPATH: | |
1162 | { | |
1163 | cfg->rc_mp = nla_data(nla); | |
1164 | cfg->rc_mp_len = nla_len(nla); | |
03c05665 EB |
1165 | break; |
1166 | } | |
1167 | default: | |
1168 | /* Unsupported attribute */ | |
1169 | goto errout; | |
1170 | } | |
1171 | } | |
1172 | ||
1173 | err = 0; | |
1174 | errout: | |
1175 | return err; | |
1176 | } | |
1177 | ||
1178 | static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh) | |
1179 | { | |
1180 | struct mpls_route_config cfg; | |
1181 | int err; | |
1182 | ||
1183 | err = rtm_to_route_config(skb, nlh, &cfg); | |
1184 | if (err < 0) | |
1185 | return err; | |
1186 | ||
1187 | return mpls_route_del(&cfg); | |
1188 | } | |
1189 | ||
1190 | ||
1191 | static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh) | |
1192 | { | |
1193 | struct mpls_route_config cfg; | |
1194 | int err; | |
1195 | ||
1196 | err = rtm_to_route_config(skb, nlh, &cfg); | |
1197 | if (err < 0) | |
1198 | return err; | |
1199 | ||
1200 | return mpls_route_add(&cfg); | |
1201 | } | |
1202 | ||
1203 | static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event, | |
1204 | u32 label, struct mpls_route *rt, int flags) | |
1205 | { | |
19d0c341 | 1206 | struct net_device *dev; |
03c05665 EB |
1207 | struct nlmsghdr *nlh; |
1208 | struct rtmsg *rtm; | |
1209 | ||
1210 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags); | |
1211 | if (nlh == NULL) | |
1212 | return -EMSGSIZE; | |
1213 | ||
1214 | rtm = nlmsg_data(nlh); | |
1215 | rtm->rtm_family = AF_MPLS; | |
1216 | rtm->rtm_dst_len = 20; | |
1217 | rtm->rtm_src_len = 0; | |
1218 | rtm->rtm_tos = 0; | |
1219 | rtm->rtm_table = RT_TABLE_MAIN; | |
1220 | rtm->rtm_protocol = rt->rt_protocol; | |
1221 | rtm->rtm_scope = RT_SCOPE_UNIVERSE; | |
1222 | rtm->rtm_type = RTN_UNICAST; | |
1223 | rtm->rtm_flags = 0; | |
1224 | ||
03c05665 EB |
1225 | if (nla_put_labels(skb, RTA_DST, 1, &label)) |
1226 | goto nla_put_failure; | |
f8efb73c | 1227 | if (rt->rt_nhn == 1) { |
cf4b24f0 | 1228 | const struct mpls_nh *nh = rt->rt_nh; |
f8efb73c RP |
1229 | |
1230 | if (nh->nh_labels && | |
1231 | nla_put_labels(skb, RTA_NEWDST, nh->nh_labels, | |
1232 | nh->nh_label)) | |
1233 | goto nla_put_failure; | |
cf4b24f0 | 1234 | if (nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh), |
f8efb73c RP |
1235 | nh->nh_via_alen)) |
1236 | goto nla_put_failure; | |
1237 | dev = rtnl_dereference(nh->nh_dev); | |
1238 | if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) | |
1239 | goto nla_put_failure; | |
1240 | } else { | |
1241 | struct rtnexthop *rtnh; | |
1242 | struct nlattr *mp; | |
1243 | ||
1244 | mp = nla_nest_start(skb, RTA_MULTIPATH); | |
1245 | if (!mp) | |
1246 | goto nla_put_failure; | |
1247 | ||
1248 | for_nexthops(rt) { | |
1249 | rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); | |
1250 | if (!rtnh) | |
1251 | goto nla_put_failure; | |
1252 | ||
1253 | dev = rtnl_dereference(nh->nh_dev); | |
1254 | if (dev) | |
1255 | rtnh->rtnh_ifindex = dev->ifindex; | |
1256 | if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST, | |
1257 | nh->nh_labels, | |
1258 | nh->nh_label)) | |
1259 | goto nla_put_failure; | |
1260 | if (nla_put_via(skb, nh->nh_via_table, | |
cf4b24f0 | 1261 | mpls_nh_via(rt, nh), |
f8efb73c RP |
1262 | nh->nh_via_alen)) |
1263 | goto nla_put_failure; | |
1264 | ||
1265 | /* length of rtnetlink header + attributes */ | |
1266 | rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh; | |
1267 | } endfor_nexthops(rt); | |
1268 | ||
1269 | nla_nest_end(skb, mp); | |
1270 | } | |
03c05665 EB |
1271 | |
1272 | nlmsg_end(skb, nlh); | |
1273 | return 0; | |
1274 | ||
1275 | nla_put_failure: | |
1276 | nlmsg_cancel(skb, nlh); | |
1277 | return -EMSGSIZE; | |
1278 | } | |
1279 | ||
1280 | static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb) | |
1281 | { | |
1282 | struct net *net = sock_net(skb->sk); | |
19d0c341 EB |
1283 | struct mpls_route __rcu **platform_label; |
1284 | size_t platform_labels; | |
03c05665 EB |
1285 | unsigned int index; |
1286 | ||
1287 | ASSERT_RTNL(); | |
1288 | ||
1289 | index = cb->args[0]; | |
a6affd24 RS |
1290 | if (index < MPLS_LABEL_FIRST_UNRESERVED) |
1291 | index = MPLS_LABEL_FIRST_UNRESERVED; | |
03c05665 | 1292 | |
19d0c341 EB |
1293 | platform_label = rtnl_dereference(net->mpls.platform_label); |
1294 | platform_labels = net->mpls.platform_labels; | |
1295 | for (; index < platform_labels; index++) { | |
03c05665 | 1296 | struct mpls_route *rt; |
19d0c341 | 1297 | rt = rtnl_dereference(platform_label[index]); |
03c05665 EB |
1298 | if (!rt) |
1299 | continue; | |
1300 | ||
1301 | if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid, | |
1302 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, | |
1303 | index, rt, NLM_F_MULTI) < 0) | |
1304 | break; | |
1305 | } | |
1306 | cb->args[0] = index; | |
1307 | ||
1308 | return skb->len; | |
1309 | } | |
1310 | ||
8de147dc EB |
1311 | static inline size_t lfib_nlmsg_size(struct mpls_route *rt) |
1312 | { | |
1313 | size_t payload = | |
1314 | NLMSG_ALIGN(sizeof(struct rtmsg)) | |
8de147dc | 1315 | + nla_total_size(4); /* RTA_DST */ |
f8efb73c RP |
1316 | |
1317 | if (rt->rt_nhn == 1) { | |
1318 | struct mpls_nh *nh = rt->rt_nh; | |
1319 | ||
1320 | if (nh->nh_dev) | |
1321 | payload += nla_total_size(4); /* RTA_OIF */ | |
1322 | payload += nla_total_size(2 + nh->nh_via_alen); /* RTA_VIA */ | |
1323 | if (nh->nh_labels) /* RTA_NEWDST */ | |
1324 | payload += nla_total_size(nh->nh_labels * 4); | |
1325 | } else { | |
1326 | /* each nexthop is packed in an attribute */ | |
1327 | size_t nhsize = 0; | |
1328 | ||
1329 | for_nexthops(rt) { | |
1330 | nhsize += nla_total_size(sizeof(struct rtnexthop)); | |
1331 | nhsize += nla_total_size(2 + nh->nh_via_alen); | |
1332 | if (nh->nh_labels) | |
1333 | nhsize += nla_total_size(nh->nh_labels * 4); | |
1334 | } endfor_nexthops(rt); | |
1335 | /* nested attribute */ | |
1336 | payload += nla_total_size(nhsize); | |
1337 | } | |
1338 | ||
8de147dc EB |
1339 | return payload; |
1340 | } | |
1341 | ||
1342 | static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, | |
1343 | struct nlmsghdr *nlh, struct net *net, u32 portid, | |
1344 | unsigned int nlm_flags) | |
1345 | { | |
1346 | struct sk_buff *skb; | |
1347 | u32 seq = nlh ? nlh->nlmsg_seq : 0; | |
1348 | int err = -ENOBUFS; | |
1349 | ||
1350 | skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL); | |
1351 | if (skb == NULL) | |
1352 | goto errout; | |
1353 | ||
1354 | err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags); | |
1355 | if (err < 0) { | |
1356 | /* -EMSGSIZE implies BUG in lfib_nlmsg_size */ | |
1357 | WARN_ON(err == -EMSGSIZE); | |
1358 | kfree_skb(skb); | |
1359 | goto errout; | |
1360 | } | |
1361 | rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL); | |
1362 | ||
1363 | return; | |
1364 | errout: | |
1365 | if (err < 0) | |
1366 | rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err); | |
1367 | } | |
1368 | ||
7720c01f EB |
1369 | static int resize_platform_label_table(struct net *net, size_t limit) |
1370 | { | |
1371 | size_t size = sizeof(struct mpls_route *) * limit; | |
1372 | size_t old_limit; | |
1373 | size_t cp_size; | |
1374 | struct mpls_route __rcu **labels = NULL, **old; | |
1375 | struct mpls_route *rt0 = NULL, *rt2 = NULL; | |
1376 | unsigned index; | |
1377 | ||
1378 | if (size) { | |
1379 | labels = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY); | |
1380 | if (!labels) | |
1381 | labels = vzalloc(size); | |
1382 | ||
1383 | if (!labels) | |
1384 | goto nolabels; | |
1385 | } | |
1386 | ||
1387 | /* In case the predefined labels need to be populated */ | |
78f5b899 | 1388 | if (limit > MPLS_LABEL_IPV4NULL) { |
7720c01f | 1389 | struct net_device *lo = net->loopback_dev; |
cf4b24f0 | 1390 | rt0 = mpls_rt_alloc(1, lo->addr_len); |
7720c01f EB |
1391 | if (!rt0) |
1392 | goto nort0; | |
f8efb73c | 1393 | RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo); |
7720c01f | 1394 | rt0->rt_protocol = RTPROT_KERNEL; |
118d5234 | 1395 | rt0->rt_payload_type = MPT_IPV4; |
f8efb73c | 1396 | rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE; |
b4e04fc7 | 1397 | rt0->rt_nh->nh_via_alen = lo->addr_len; |
cf4b24f0 RS |
1398 | memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr, |
1399 | lo->addr_len); | |
7720c01f | 1400 | } |
78f5b899 | 1401 | if (limit > MPLS_LABEL_IPV6NULL) { |
7720c01f | 1402 | struct net_device *lo = net->loopback_dev; |
cf4b24f0 | 1403 | rt2 = mpls_rt_alloc(1, lo->addr_len); |
7720c01f EB |
1404 | if (!rt2) |
1405 | goto nort2; | |
f8efb73c | 1406 | RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo); |
7720c01f | 1407 | rt2->rt_protocol = RTPROT_KERNEL; |
118d5234 | 1408 | rt2->rt_payload_type = MPT_IPV6; |
f8efb73c | 1409 | rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE; |
b4e04fc7 | 1410 | rt2->rt_nh->nh_via_alen = lo->addr_len; |
cf4b24f0 RS |
1411 | memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr, |
1412 | lo->addr_len); | |
7720c01f EB |
1413 | } |
1414 | ||
1415 | rtnl_lock(); | |
1416 | /* Remember the original table */ | |
19d0c341 | 1417 | old = rtnl_dereference(net->mpls.platform_label); |
7720c01f EB |
1418 | old_limit = net->mpls.platform_labels; |
1419 | ||
1420 | /* Free any labels beyond the new table */ | |
1421 | for (index = limit; index < old_limit; index++) | |
f8efb73c | 1422 | mpls_route_update(net, index, NULL, NULL); |
7720c01f EB |
1423 | |
1424 | /* Copy over the old labels */ | |
1425 | cp_size = size; | |
1426 | if (old_limit < limit) | |
1427 | cp_size = old_limit * sizeof(struct mpls_route *); | |
1428 | ||
1429 | memcpy(labels, old, cp_size); | |
1430 | ||
1431 | /* If needed set the predefined labels */ | |
78f5b899 TH |
1432 | if ((old_limit <= MPLS_LABEL_IPV6NULL) && |
1433 | (limit > MPLS_LABEL_IPV6NULL)) { | |
1434 | RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2); | |
7720c01f EB |
1435 | rt2 = NULL; |
1436 | } | |
1437 | ||
78f5b899 TH |
1438 | if ((old_limit <= MPLS_LABEL_IPV4NULL) && |
1439 | (limit > MPLS_LABEL_IPV4NULL)) { | |
1440 | RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0); | |
7720c01f EB |
1441 | rt0 = NULL; |
1442 | } | |
1443 | ||
1444 | /* Update the global pointers */ | |
1445 | net->mpls.platform_labels = limit; | |
19d0c341 | 1446 | rcu_assign_pointer(net->mpls.platform_label, labels); |
7720c01f EB |
1447 | |
1448 | rtnl_unlock(); | |
1449 | ||
1450 | mpls_rt_free(rt2); | |
1451 | mpls_rt_free(rt0); | |
1452 | ||
1453 | if (old) { | |
1454 | synchronize_rcu(); | |
1455 | kvfree(old); | |
1456 | } | |
1457 | return 0; | |
1458 | ||
1459 | nort2: | |
1460 | mpls_rt_free(rt0); | |
1461 | nort0: | |
1462 | kvfree(labels); | |
1463 | nolabels: | |
1464 | return -ENOMEM; | |
1465 | } | |
1466 | ||
1467 | static int mpls_platform_labels(struct ctl_table *table, int write, | |
1468 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
1469 | { | |
1470 | struct net *net = table->data; | |
1471 | int platform_labels = net->mpls.platform_labels; | |
1472 | int ret; | |
1473 | struct ctl_table tmp = { | |
1474 | .procname = table->procname, | |
1475 | .data = &platform_labels, | |
1476 | .maxlen = sizeof(int), | |
1477 | .mode = table->mode, | |
1478 | .extra1 = &zero, | |
1479 | .extra2 = &label_limit, | |
1480 | }; | |
1481 | ||
1482 | ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); | |
1483 | ||
1484 | if (write && ret == 0) | |
1485 | ret = resize_platform_label_table(net, platform_labels); | |
1486 | ||
1487 | return ret; | |
1488 | } | |
1489 | ||
37bde799 | 1490 | static const struct ctl_table mpls_table[] = { |
7720c01f EB |
1491 | { |
1492 | .procname = "platform_labels", | |
1493 | .data = NULL, | |
1494 | .maxlen = sizeof(int), | |
1495 | .mode = 0644, | |
1496 | .proc_handler = mpls_platform_labels, | |
1497 | }, | |
1498 | { } | |
1499 | }; | |
1500 | ||
0189197f EB |
1501 | static int mpls_net_init(struct net *net) |
1502 | { | |
7720c01f EB |
1503 | struct ctl_table *table; |
1504 | ||
0189197f EB |
1505 | net->mpls.platform_labels = 0; |
1506 | net->mpls.platform_label = NULL; | |
1507 | ||
7720c01f EB |
1508 | table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL); |
1509 | if (table == NULL) | |
1510 | return -ENOMEM; | |
1511 | ||
1512 | table[0].data = net; | |
1513 | net->mpls.ctl = register_net_sysctl(net, "net/mpls", table); | |
6ea3c9d5 NA |
1514 | if (net->mpls.ctl == NULL) { |
1515 | kfree(table); | |
7720c01f | 1516 | return -ENOMEM; |
6ea3c9d5 | 1517 | } |
7720c01f | 1518 | |
0189197f EB |
1519 | return 0; |
1520 | } | |
1521 | ||
1522 | static void mpls_net_exit(struct net *net) | |
1523 | { | |
19d0c341 EB |
1524 | struct mpls_route __rcu **platform_label; |
1525 | size_t platform_labels; | |
7720c01f | 1526 | struct ctl_table *table; |
0189197f EB |
1527 | unsigned int index; |
1528 | ||
7720c01f EB |
1529 | table = net->mpls.ctl->ctl_table_arg; |
1530 | unregister_net_sysctl_table(net->mpls.ctl); | |
1531 | kfree(table); | |
1532 | ||
19d0c341 EB |
1533 | /* An rcu grace period has passed since there was a device in |
1534 | * the network namespace (and thus the last in flight packet) | |
0189197f EB |
1535 | * left this network namespace. This is because |
1536 | * unregister_netdevice_many and netdev_run_todo has completed | |
1537 | * for each network device that was in this network namespace. | |
1538 | * | |
1539 | * As such no additional rcu synchronization is necessary when | |
1540 | * freeing the platform_label table. | |
1541 | */ | |
1542 | rtnl_lock(); | |
19d0c341 EB |
1543 | platform_label = rtnl_dereference(net->mpls.platform_label); |
1544 | platform_labels = net->mpls.platform_labels; | |
1545 | for (index = 0; index < platform_labels; index++) { | |
1546 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); | |
1547 | RCU_INIT_POINTER(platform_label[index], NULL); | |
0189197f EB |
1548 | mpls_rt_free(rt); |
1549 | } | |
1550 | rtnl_unlock(); | |
1551 | ||
19d0c341 | 1552 | kvfree(platform_label); |
0189197f EB |
1553 | } |
1554 | ||
1555 | static struct pernet_operations mpls_net_ops = { | |
1556 | .init = mpls_net_init, | |
1557 | .exit = mpls_net_exit, | |
1558 | }; | |
1559 | ||
1560 | static int __init mpls_init(void) | |
1561 | { | |
1562 | int err; | |
1563 | ||
1564 | BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4); | |
1565 | ||
1566 | err = register_pernet_subsys(&mpls_net_ops); | |
1567 | if (err) | |
1568 | goto out; | |
1569 | ||
1570 | err = register_netdevice_notifier(&mpls_dev_notifier); | |
1571 | if (err) | |
1572 | goto out_unregister_pernet; | |
1573 | ||
1574 | dev_add_pack(&mpls_packet_type); | |
1575 | ||
03c05665 EB |
1576 | rtnl_register(PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, NULL); |
1577 | rtnl_register(PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, NULL); | |
1578 | rtnl_register(PF_MPLS, RTM_GETROUTE, NULL, mpls_dump_routes, NULL); | |
0189197f EB |
1579 | err = 0; |
1580 | out: | |
1581 | return err; | |
1582 | ||
1583 | out_unregister_pernet: | |
1584 | unregister_pernet_subsys(&mpls_net_ops); | |
1585 | goto out; | |
1586 | } | |
1587 | module_init(mpls_init); | |
1588 | ||
1589 | static void __exit mpls_exit(void) | |
1590 | { | |
03c05665 | 1591 | rtnl_unregister_all(PF_MPLS); |
0189197f EB |
1592 | dev_remove_pack(&mpls_packet_type); |
1593 | unregister_netdevice_notifier(&mpls_dev_notifier); | |
1594 | unregister_pernet_subsys(&mpls_net_ops); | |
1595 | } | |
1596 | module_exit(mpls_exit); | |
1597 | ||
1598 | MODULE_DESCRIPTION("MultiProtocol Label Switching"); | |
1599 | MODULE_LICENSE("GPL v2"); | |
1600 | MODULE_ALIAS_NETPROTO(PF_MPLS); |