]>
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> | |
24045a03 | 10 | #include <linux/netconf.h> |
4b5edb2f | 11 | #include <linux/vmalloc.h> |
27d69105 | 12 | #include <linux/percpu.h> |
0189197f EB |
13 | #include <net/ip.h> |
14 | #include <net/dst.h> | |
15 | #include <net/sock.h> | |
16 | #include <net/arp.h> | |
17 | #include <net/ip_fib.h> | |
18 | #include <net/netevent.h> | |
19 | #include <net/netns/generic.h> | |
bf21563a RP |
20 | #if IS_ENABLED(CONFIG_IPV6) |
21 | #include <net/ipv6.h> | |
bf21563a | 22 | #endif |
27d69105 | 23 | #include <net/addrconf.h> |
f8efb73c | 24 | #include <net/nexthop.h> |
0189197f EB |
25 | #include "internal.h" |
26 | ||
df1c6316 DA |
27 | /* max memory we will use for mpls_route */ |
28 | #define MAX_MPLS_ROUTE_MEM 4096 | |
29 | ||
1c78efa8 RS |
30 | /* Maximum number of labels to look ahead at when selecting a path of |
31 | * a multipath route | |
32 | */ | |
33 | #define MAX_MP_SELECT_LABELS 4 | |
34 | ||
eb7809f0 RS |
35 | #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1) |
36 | ||
7720c01f | 37 | static int zero = 0; |
5b441ac8 | 38 | static int one = 1; |
7720c01f | 39 | static int label_limit = (1 << 20) - 1; |
a59166e4 | 40 | static int ttl_max = 255; |
7720c01f | 41 | |
8de147dc EB |
42 | static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, |
43 | struct nlmsghdr *nlh, struct net *net, u32 portid, | |
44 | unsigned int nlm_flags); | |
45 | ||
0189197f EB |
46 | static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index) |
47 | { | |
48 | struct mpls_route *rt = NULL; | |
49 | ||
50 | if (index < net->mpls.platform_labels) { | |
51 | struct mpls_route __rcu **platform_label = | |
52 | rcu_dereference(net->mpls.platform_label); | |
3b2ec7eb ER |
53 | |
54 | osb(); | |
0189197f EB |
55 | rt = rcu_dereference(platform_label[index]); |
56 | } | |
57 | return rt; | |
58 | } | |
59 | ||
face0188 | 60 | bool mpls_output_possible(const struct net_device *dev) |
0189197f EB |
61 | { |
62 | return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev); | |
63 | } | |
face0188 | 64 | EXPORT_SYMBOL_GPL(mpls_output_possible); |
0189197f | 65 | |
cf4b24f0 RS |
66 | static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh) |
67 | { | |
59b20966 | 68 | return (u8 *)nh + rt->rt_via_offset; |
cf4b24f0 RS |
69 | } |
70 | ||
71 | static const u8 *mpls_nh_via(const struct mpls_route *rt, | |
72 | const struct mpls_nh *nh) | |
73 | { | |
74 | return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh); | |
75 | } | |
76 | ||
f8efb73c | 77 | static unsigned int mpls_nh_header_size(const struct mpls_nh *nh) |
0189197f EB |
78 | { |
79 | /* The size of the layer 2.5 labels to be added for this route */ | |
f8efb73c | 80 | return nh->nh_labels * sizeof(struct mpls_shim_hdr); |
0189197f EB |
81 | } |
82 | ||
face0188 | 83 | unsigned int mpls_dev_mtu(const struct net_device *dev) |
0189197f EB |
84 | { |
85 | /* The amount of data the layer 2 frame can hold */ | |
86 | return dev->mtu; | |
87 | } | |
face0188 | 88 | EXPORT_SYMBOL_GPL(mpls_dev_mtu); |
0189197f | 89 | |
face0188 | 90 | bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) |
0189197f EB |
91 | { |
92 | if (skb->len <= mtu) | |
93 | return false; | |
94 | ||
ae7ef81e | 95 | if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu)) |
0189197f EB |
96 | return false; |
97 | ||
98 | return true; | |
99 | } | |
face0188 | 100 | EXPORT_SYMBOL_GPL(mpls_pkt_too_big); |
0189197f | 101 | |
27d69105 RS |
102 | void mpls_stats_inc_outucastpkts(struct net_device *dev, |
103 | const struct sk_buff *skb) | |
104 | { | |
105 | struct mpls_dev *mdev; | |
106 | ||
107 | if (skb->protocol == htons(ETH_P_MPLS_UC)) { | |
108 | mdev = mpls_dev_get(dev); | |
109 | if (mdev) | |
110 | MPLS_INC_STATS_LEN(mdev, skb->len, | |
111 | tx_packets, | |
112 | tx_bytes); | |
113 | } else if (skb->protocol == htons(ETH_P_IP)) { | |
114 | IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len); | |
115 | #if IS_ENABLED(CONFIG_IPV6) | |
116 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
117 | struct inet6_dev *in6dev = __in6_dev_get(dev); | |
118 | ||
119 | if (in6dev) | |
120 | IP6_UPD_PO_STATS(dev_net(dev), in6dev, | |
121 | IPSTATS_MIB_OUT, skb->len); | |
122 | #endif | |
123 | } | |
124 | } | |
125 | EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts); | |
126 | ||
9f427a0e | 127 | static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb) |
f8efb73c | 128 | { |
1c78efa8 | 129 | struct mpls_entry_decoded dec; |
9f427a0e | 130 | unsigned int mpls_hdr_len = 0; |
1c78efa8 RS |
131 | struct mpls_shim_hdr *hdr; |
132 | bool eli_seen = false; | |
133 | int label_index; | |
1c78efa8 RS |
134 | u32 hash = 0; |
135 | ||
9f427a0e | 136 | for (label_index = 0; label_index < MAX_MP_SELECT_LABELS; |
1c78efa8 | 137 | label_index++) { |
9f427a0e DA |
138 | mpls_hdr_len += sizeof(*hdr); |
139 | if (!pskb_may_pull(skb, mpls_hdr_len)) | |
1c78efa8 RS |
140 | break; |
141 | ||
142 | /* Read and decode the current label */ | |
143 | hdr = mpls_hdr(skb) + label_index; | |
144 | dec = mpls_entry_decode(hdr); | |
145 | ||
146 | /* RFC6790 - reserved labels MUST NOT be used as keys | |
147 | * for the load-balancing function | |
148 | */ | |
149 | if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) { | |
150 | hash = jhash_1word(dec.label, hash); | |
151 | ||
152 | /* The entropy label follows the entropy label | |
153 | * indicator, so this means that the entropy | |
154 | * label was just added to the hash - no need to | |
155 | * go any deeper either in the label stack or in the | |
156 | * payload | |
157 | */ | |
158 | if (eli_seen) | |
159 | break; | |
160 | } else if (dec.label == MPLS_LABEL_ENTROPY) { | |
161 | eli_seen = true; | |
162 | } | |
163 | ||
9f427a0e DA |
164 | if (!dec.bos) |
165 | continue; | |
166 | ||
167 | /* found bottom label; does skb have room for a header? */ | |
168 | if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) { | |
1c78efa8 RS |
169 | const struct iphdr *v4hdr; |
170 | ||
9f427a0e | 171 | v4hdr = (const struct iphdr *)(hdr + 1); |
1c78efa8 RS |
172 | if (v4hdr->version == 4) { |
173 | hash = jhash_3words(ntohl(v4hdr->saddr), | |
174 | ntohl(v4hdr->daddr), | |
175 | v4hdr->protocol, hash); | |
176 | } else if (v4hdr->version == 6 && | |
9f427a0e DA |
177 | pskb_may_pull(skb, mpls_hdr_len + |
178 | sizeof(struct ipv6hdr))) { | |
1c78efa8 RS |
179 | const struct ipv6hdr *v6hdr; |
180 | ||
9f427a0e | 181 | v6hdr = (const struct ipv6hdr *)(hdr + 1); |
1c78efa8 RS |
182 | hash = __ipv6_addr_jhash(&v6hdr->saddr, hash); |
183 | hash = __ipv6_addr_jhash(&v6hdr->daddr, hash); | |
184 | hash = jhash_1word(v6hdr->nexthdr, hash); | |
185 | } | |
186 | } | |
9f427a0e DA |
187 | |
188 | break; | |
1c78efa8 RS |
189 | } |
190 | ||
c89359a4 RP |
191 | return hash; |
192 | } | |
193 | ||
59b20966 DA |
194 | static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index) |
195 | { | |
196 | return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size); | |
197 | } | |
198 | ||
39eb8cd1 DA |
199 | /* number of alive nexthops (rt->rt_nhn_alive) and the flags for |
200 | * a next hop (nh->nh_flags) are modified by netdev event handlers. | |
201 | * Since those fields can change at any moment, use READ_ONCE to | |
202 | * access both. | |
203 | */ | |
c89359a4 | 204 | static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt, |
9f427a0e | 205 | struct sk_buff *skb) |
c89359a4 | 206 | { |
c89359a4 RP |
207 | u32 hash = 0; |
208 | int nh_index = 0; | |
209 | int n = 0; | |
77ef013a | 210 | u8 alive; |
c89359a4 RP |
211 | |
212 | /* No need to look further into packet if there's only | |
213 | * one path | |
214 | */ | |
215 | if (rt->rt_nhn == 1) | |
59b20966 | 216 | return rt->rt_nh; |
c89359a4 | 217 | |
39eb8cd1 DA |
218 | alive = READ_ONCE(rt->rt_nhn_alive); |
219 | if (alive == 0) | |
c89359a4 RP |
220 | return NULL; |
221 | ||
9f427a0e | 222 | hash = mpls_multipath_hash(rt, skb); |
c89359a4 RP |
223 | nh_index = hash % alive; |
224 | if (alive == rt->rt_nhn) | |
225 | goto out; | |
226 | for_nexthops(rt) { | |
39eb8cd1 DA |
227 | unsigned int nh_flags = READ_ONCE(nh->nh_flags); |
228 | ||
229 | if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) | |
c89359a4 RP |
230 | continue; |
231 | if (n == nh_index) | |
232 | return nh; | |
233 | n++; | |
234 | } endfor_nexthops(rt); | |
235 | ||
1c78efa8 | 236 | out: |
59b20966 | 237 | return mpls_get_nexthop(rt, nh_index); |
f8efb73c RP |
238 | } |
239 | ||
5b441ac8 RS |
240 | static bool mpls_egress(struct net *net, struct mpls_route *rt, |
241 | struct sk_buff *skb, struct mpls_entry_decoded dec) | |
0189197f | 242 | { |
118d5234 RS |
243 | enum mpls_payload_type payload_type; |
244 | bool success = false; | |
0189197f | 245 | |
76fecd82 EB |
246 | /* The IPv4 code below accesses through the IPv4 header |
247 | * checksum, which is 12 bytes into the packet. | |
248 | * The IPv6 code below accesses through the IPv6 hop limit | |
249 | * which is 8 bytes into the packet. | |
250 | * | |
251 | * For all supported cases there should always be at least 12 | |
252 | * bytes of packet data present. The IPv4 header is 20 bytes | |
253 | * without options and the IPv6 header is always 40 bytes | |
254 | * long. | |
255 | */ | |
256 | if (!pskb_may_pull(skb, 12)) | |
257 | return false; | |
258 | ||
118d5234 RS |
259 | payload_type = rt->rt_payload_type; |
260 | if (payload_type == MPT_UNSPEC) | |
261 | payload_type = ip_hdr(skb)->version; | |
262 | ||
263 | switch (payload_type) { | |
264 | case MPT_IPV4: { | |
265 | struct iphdr *hdr4 = ip_hdr(skb); | |
5b441ac8 | 266 | u8 new_ttl; |
0189197f | 267 | skb->protocol = htons(ETH_P_IP); |
5b441ac8 RS |
268 | |
269 | /* If propagating TTL, take the decremented TTL from | |
270 | * the incoming MPLS header, otherwise decrement the | |
271 | * TTL, but only if not 0 to avoid underflow. | |
272 | */ | |
273 | if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED || | |
274 | (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT && | |
275 | net->mpls.ip_ttl_propagate)) | |
276 | new_ttl = dec.ttl; | |
277 | else | |
278 | new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0; | |
279 | ||
0189197f EB |
280 | csum_replace2(&hdr4->check, |
281 | htons(hdr4->ttl << 8), | |
5b441ac8 RS |
282 | htons(new_ttl << 8)); |
283 | hdr4->ttl = new_ttl; | |
118d5234 RS |
284 | success = true; |
285 | break; | |
0189197f | 286 | } |
118d5234 | 287 | case MPT_IPV6: { |
0189197f EB |
288 | struct ipv6hdr *hdr6 = ipv6_hdr(skb); |
289 | skb->protocol = htons(ETH_P_IPV6); | |
5b441ac8 RS |
290 | |
291 | /* If propagating TTL, take the decremented TTL from | |
292 | * the incoming MPLS header, otherwise decrement the | |
293 | * hop limit, but only if not 0 to avoid underflow. | |
294 | */ | |
295 | if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED || | |
296 | (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT && | |
297 | net->mpls.ip_ttl_propagate)) | |
298 | hdr6->hop_limit = dec.ttl; | |
299 | else if (hdr6->hop_limit) | |
300 | hdr6->hop_limit = hdr6->hop_limit - 1; | |
118d5234 RS |
301 | success = true; |
302 | break; | |
0189197f | 303 | } |
118d5234 | 304 | case MPT_UNSPEC: |
5b441ac8 | 305 | /* Should have decided which protocol it is by now */ |
118d5234 RS |
306 | break; |
307 | } | |
308 | ||
0189197f EB |
309 | return success; |
310 | } | |
311 | ||
312 | static int mpls_forward(struct sk_buff *skb, struct net_device *dev, | |
313 | struct packet_type *pt, struct net_device *orig_dev) | |
314 | { | |
315 | struct net *net = dev_net(dev); | |
316 | struct mpls_shim_hdr *hdr; | |
317 | struct mpls_route *rt; | |
f8efb73c | 318 | struct mpls_nh *nh; |
0189197f EB |
319 | struct mpls_entry_decoded dec; |
320 | struct net_device *out_dev; | |
27d69105 | 321 | struct mpls_dev *out_mdev; |
03c57747 | 322 | struct mpls_dev *mdev; |
0189197f EB |
323 | unsigned int hh_len; |
324 | unsigned int new_header_size; | |
325 | unsigned int mtu; | |
326 | int err; | |
327 | ||
328 | /* Careful this entire function runs inside of an rcu critical section */ | |
329 | ||
03c57747 | 330 | mdev = mpls_dev_get(dev); |
27d69105 | 331 | if (!mdev) |
03c57747 RS |
332 | goto drop; |
333 | ||
27d69105 RS |
334 | MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets, |
335 | rx_bytes); | |
336 | ||
337 | if (!mdev->input_enabled) { | |
338 | MPLS_INC_STATS(mdev, rx_dropped); | |
0189197f | 339 | goto drop; |
27d69105 RS |
340 | } |
341 | ||
342 | if (skb->pkt_type != PACKET_HOST) | |
343 | goto err; | |
0189197f EB |
344 | |
345 | if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) | |
27d69105 | 346 | goto err; |
0189197f EB |
347 | |
348 | if (!pskb_may_pull(skb, sizeof(*hdr))) | |
27d69105 | 349 | goto err; |
0189197f EB |
350 | |
351 | /* Read and decode the label */ | |
352 | hdr = mpls_hdr(skb); | |
353 | dec = mpls_entry_decode(hdr); | |
354 | ||
0189197f | 355 | rt = mpls_route_input_rcu(net, dec.label); |
27d69105 RS |
356 | if (!rt) { |
357 | MPLS_INC_STATS(mdev, rx_noroute); | |
0189197f | 358 | goto drop; |
27d69105 | 359 | } |
0189197f | 360 | |
9f427a0e | 361 | nh = mpls_select_multipath(rt, skb); |
f8efb73c | 362 | if (!nh) |
27d69105 | 363 | goto err; |
0189197f | 364 | |
9f427a0e DA |
365 | /* Pop the label */ |
366 | skb_pull(skb, sizeof(*hdr)); | |
367 | skb_reset_network_header(skb); | |
368 | ||
369 | skb_orphan(skb); | |
370 | ||
0189197f | 371 | if (skb_warn_if_lro(skb)) |
27d69105 | 372 | goto err; |
0189197f EB |
373 | |
374 | skb_forward_csum(skb); | |
375 | ||
376 | /* Verify ttl is valid */ | |
aa7da937 | 377 | if (dec.ttl <= 1) |
27d69105 | 378 | goto err; |
0189197f EB |
379 | dec.ttl -= 1; |
380 | ||
27d69105 RS |
381 | /* Find the output device */ |
382 | out_dev = rcu_dereference(nh->nh_dev); | |
383 | if (!mpls_output_possible(out_dev)) | |
384 | goto tx_err; | |
385 | ||
0189197f | 386 | /* Verify the destination can hold the packet */ |
f8efb73c | 387 | new_header_size = mpls_nh_header_size(nh); |
0189197f EB |
388 | mtu = mpls_dev_mtu(out_dev); |
389 | if (mpls_pkt_too_big(skb, mtu - new_header_size)) | |
27d69105 | 390 | goto tx_err; |
0189197f EB |
391 | |
392 | hh_len = LL_RESERVED_SPACE(out_dev); | |
393 | if (!out_dev->header_ops) | |
394 | hh_len = 0; | |
395 | ||
396 | /* Ensure there is enough space for the headers in the skb */ | |
397 | if (skb_cow(skb, hh_len + new_header_size)) | |
27d69105 | 398 | goto tx_err; |
0189197f EB |
399 | |
400 | skb->dev = out_dev; | |
401 | skb->protocol = htons(ETH_P_MPLS_UC); | |
402 | ||
403 | if (unlikely(!new_header_size && dec.bos)) { | |
404 | /* Penultimate hop popping */ | |
5b441ac8 | 405 | if (!mpls_egress(dev_net(out_dev), rt, skb, dec)) |
27d69105 | 406 | goto err; |
0189197f EB |
407 | } else { |
408 | bool bos; | |
409 | int i; | |
410 | skb_push(skb, new_header_size); | |
411 | skb_reset_network_header(skb); | |
412 | /* Push the new labels */ | |
413 | hdr = mpls_hdr(skb); | |
414 | bos = dec.bos; | |
f8efb73c RP |
415 | for (i = nh->nh_labels - 1; i >= 0; i--) { |
416 | hdr[i] = mpls_entry_encode(nh->nh_label[i], | |
417 | dec.ttl, 0, bos); | |
0189197f EB |
418 | bos = false; |
419 | } | |
420 | } | |
421 | ||
27d69105 RS |
422 | mpls_stats_inc_outucastpkts(out_dev, skb); |
423 | ||
eb7809f0 RS |
424 | /* If via wasn't specified then send out using device address */ |
425 | if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC) | |
426 | err = neigh_xmit(NEIGH_LINK_TABLE, out_dev, | |
427 | out_dev->dev_addr, skb); | |
428 | else | |
429 | err = neigh_xmit(nh->nh_via_table, out_dev, | |
430 | mpls_nh_via(rt, nh), skb); | |
0189197f EB |
431 | if (err) |
432 | net_dbg_ratelimited("%s: packet transmission failed: %d\n", | |
433 | __func__, err); | |
434 | return 0; | |
435 | ||
27d69105 RS |
436 | tx_err: |
437 | out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL; | |
438 | if (out_mdev) | |
439 | MPLS_INC_STATS(out_mdev, tx_errors); | |
440 | goto drop; | |
441 | err: | |
442 | MPLS_INC_STATS(mdev, rx_errors); | |
0189197f EB |
443 | drop: |
444 | kfree_skb(skb); | |
445 | return NET_RX_DROP; | |
446 | } | |
447 | ||
448 | static struct packet_type mpls_packet_type __read_mostly = { | |
449 | .type = cpu_to_be16(ETH_P_MPLS_UC), | |
450 | .func = mpls_forward, | |
451 | }; | |
452 | ||
f0126539 | 453 | static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = { |
03c05665 EB |
454 | [RTA_DST] = { .type = NLA_U32 }, |
455 | [RTA_OIF] = { .type = NLA_U32 }, | |
5b441ac8 | 456 | [RTA_TTL_PROPAGATE] = { .type = NLA_U8 }, |
03c05665 EB |
457 | }; |
458 | ||
a2519929 | 459 | struct mpls_route_config { |
118d5234 RS |
460 | u32 rc_protocol; |
461 | u32 rc_ifindex; | |
f8efb73c RP |
462 | u8 rc_via_table; |
463 | u8 rc_via_alen; | |
118d5234 RS |
464 | u8 rc_via[MAX_VIA_ALEN]; |
465 | u32 rc_label; | |
5b441ac8 | 466 | u8 rc_ttl_propagate; |
f8efb73c | 467 | u8 rc_output_labels; |
118d5234 RS |
468 | u32 rc_output_label[MAX_NEW_LABELS]; |
469 | u32 rc_nlflags; | |
470 | enum mpls_payload_type rc_payload_type; | |
471 | struct nl_info rc_nlinfo; | |
f8efb73c RP |
472 | struct rtnexthop *rc_mp; |
473 | int rc_mp_len; | |
a2519929 EB |
474 | }; |
475 | ||
59b20966 DA |
476 | /* all nexthops within a route have the same size based on max |
477 | * number of labels and max via length for a hop | |
478 | */ | |
479 | static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels) | |
0189197f | 480 | { |
59b20966 | 481 | u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen); |
0189197f | 482 | struct mpls_route *rt; |
df1c6316 | 483 | size_t size; |
0189197f | 484 | |
df1c6316 DA |
485 | size = sizeof(*rt) + num_nh * nh_size; |
486 | if (size > MAX_MPLS_ROUTE_MEM) | |
487 | return ERR_PTR(-EINVAL); | |
488 | ||
489 | rt = kzalloc(size, GFP_KERNEL); | |
490 | if (!rt) | |
491 | return ERR_PTR(-ENOMEM); | |
492 | ||
493 | rt->rt_nhn = num_nh; | |
494 | rt->rt_nhn_alive = num_nh; | |
495 | rt->rt_nh_size = nh_size; | |
496 | rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels); | |
f8efb73c | 497 | |
0189197f EB |
498 | return rt; |
499 | } | |
500 | ||
501 | static void mpls_rt_free(struct mpls_route *rt) | |
502 | { | |
503 | if (rt) | |
504 | kfree_rcu(rt, rt_rcu); | |
505 | } | |
506 | ||
8de147dc EB |
507 | static void mpls_notify_route(struct net *net, unsigned index, |
508 | struct mpls_route *old, struct mpls_route *new, | |
509 | const struct nl_info *info) | |
510 | { | |
511 | struct nlmsghdr *nlh = info ? info->nlh : NULL; | |
512 | unsigned portid = info ? info->portid : 0; | |
513 | int event = new ? RTM_NEWROUTE : RTM_DELROUTE; | |
514 | struct mpls_route *rt = new ? new : old; | |
515 | unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0; | |
516 | /* Ignore reserved labels for now */ | |
a6affd24 | 517 | if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED)) |
8de147dc EB |
518 | rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags); |
519 | } | |
520 | ||
0189197f | 521 | static void mpls_route_update(struct net *net, unsigned index, |
f8efb73c | 522 | struct mpls_route *new, |
0189197f EB |
523 | const struct nl_info *info) |
524 | { | |
19d0c341 | 525 | struct mpls_route __rcu **platform_label; |
f8efb73c | 526 | struct mpls_route *rt; |
0189197f EB |
527 | |
528 | ASSERT_RTNL(); | |
529 | ||
19d0c341 EB |
530 | platform_label = rtnl_dereference(net->mpls.platform_label); |
531 | rt = rtnl_dereference(platform_label[index]); | |
f8efb73c | 532 | rcu_assign_pointer(platform_label[index], new); |
0189197f | 533 | |
f8efb73c | 534 | mpls_notify_route(net, index, rt, new, info); |
8de147dc | 535 | |
0189197f | 536 | /* If we removed a route free it now */ |
f8efb73c | 537 | mpls_rt_free(rt); |
0189197f EB |
538 | } |
539 | ||
a2519929 EB |
540 | static unsigned find_free_label(struct net *net) |
541 | { | |
19d0c341 EB |
542 | struct mpls_route __rcu **platform_label; |
543 | size_t platform_labels; | |
a2519929 | 544 | unsigned index; |
19d0c341 EB |
545 | |
546 | platform_label = rtnl_dereference(net->mpls.platform_label); | |
547 | platform_labels = net->mpls.platform_labels; | |
a6affd24 RS |
548 | for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels; |
549 | index++) { | |
19d0c341 | 550 | if (!rtnl_dereference(platform_label[index])) |
a2519929 EB |
551 | return index; |
552 | } | |
553 | return LABEL_NOT_SPECIFIED; | |
554 | } | |
555 | ||
bf21563a | 556 | #if IS_ENABLED(CONFIG_INET) |
cf4b24f0 RS |
557 | static struct net_device *inet_fib_lookup_dev(struct net *net, |
558 | const void *addr) | |
01faef2c | 559 | { |
5a9348b5 | 560 | struct net_device *dev; |
01faef2c RP |
561 | struct rtable *rt; |
562 | struct in_addr daddr; | |
563 | ||
564 | memcpy(&daddr, addr, sizeof(struct in_addr)); | |
565 | rt = ip_route_output(net, daddr.s_addr, 0, 0, 0); | |
566 | if (IS_ERR(rt)) | |
5a9348b5 | 567 | return ERR_CAST(rt); |
01faef2c RP |
568 | |
569 | dev = rt->dst.dev; | |
570 | dev_hold(dev); | |
571 | ||
572 | ip_rt_put(rt); | |
573 | ||
01faef2c | 574 | return dev; |
bf21563a RP |
575 | } |
576 | #else | |
cf4b24f0 RS |
577 | static struct net_device *inet_fib_lookup_dev(struct net *net, |
578 | const void *addr) | |
bf21563a RP |
579 | { |
580 | return ERR_PTR(-EAFNOSUPPORT); | |
01faef2c | 581 | } |
bf21563a | 582 | #endif |
01faef2c | 583 | |
bf21563a | 584 | #if IS_ENABLED(CONFIG_IPV6) |
cf4b24f0 RS |
585 | static struct net_device *inet6_fib_lookup_dev(struct net *net, |
586 | const void *addr) | |
01faef2c | 587 | { |
5a9348b5 | 588 | struct net_device *dev; |
01faef2c RP |
589 | struct dst_entry *dst; |
590 | struct flowi6 fl6; | |
bf21563a RP |
591 | int err; |
592 | ||
593 | if (!ipv6_stub) | |
594 | return ERR_PTR(-EAFNOSUPPORT); | |
01faef2c RP |
595 | |
596 | memset(&fl6, 0, sizeof(fl6)); | |
597 | memcpy(&fl6.daddr, addr, sizeof(struct in6_addr)); | |
bf21563a RP |
598 | err = ipv6_stub->ipv6_dst_lookup(net, NULL, &dst, &fl6); |
599 | if (err) | |
5a9348b5 | 600 | return ERR_PTR(err); |
01faef2c RP |
601 | |
602 | dev = dst->dev; | |
603 | dev_hold(dev); | |
01faef2c RP |
604 | dst_release(dst); |
605 | ||
606 | return dev; | |
607 | } | |
bf21563a | 608 | #else |
cf4b24f0 RS |
609 | static struct net_device *inet6_fib_lookup_dev(struct net *net, |
610 | const void *addr) | |
bf21563a RP |
611 | { |
612 | return ERR_PTR(-EAFNOSUPPORT); | |
613 | } | |
614 | #endif | |
01faef2c RP |
615 | |
616 | static struct net_device *find_outdev(struct net *net, | |
cf4b24f0 | 617 | struct mpls_route *rt, |
f8efb73c | 618 | struct mpls_nh *nh, int oif) |
01faef2c RP |
619 | { |
620 | struct net_device *dev = NULL; | |
621 | ||
f8efb73c RP |
622 | if (!oif) { |
623 | switch (nh->nh_via_table) { | |
01faef2c | 624 | case NEIGH_ARP_TABLE: |
cf4b24f0 | 625 | dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh)); |
01faef2c RP |
626 | break; |
627 | case NEIGH_ND_TABLE: | |
cf4b24f0 | 628 | dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh)); |
01faef2c RP |
629 | break; |
630 | case NEIGH_LINK_TABLE: | |
631 | break; | |
632 | } | |
633 | } else { | |
f8efb73c | 634 | dev = dev_get_by_index(net, oif); |
01faef2c RP |
635 | } |
636 | ||
3dcb615e RP |
637 | if (!dev) |
638 | return ERR_PTR(-ENODEV); | |
639 | ||
94a57f1f RP |
640 | if (IS_ERR(dev)) |
641 | return dev; | |
642 | ||
f8efb73c RP |
643 | /* The caller is holding rtnl anyways, so release the dev reference */ |
644 | dev_put(dev); | |
645 | ||
01faef2c RP |
646 | return dev; |
647 | } | |
648 | ||
cf4b24f0 RS |
649 | static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt, |
650 | struct mpls_nh *nh, int oif) | |
f8efb73c RP |
651 | { |
652 | struct net_device *dev = NULL; | |
653 | int err = -ENODEV; | |
654 | ||
cf4b24f0 | 655 | dev = find_outdev(net, rt, nh, oif); |
f8efb73c RP |
656 | if (IS_ERR(dev)) { |
657 | err = PTR_ERR(dev); | |
658 | dev = NULL; | |
659 | goto errout; | |
660 | } | |
661 | ||
662 | /* Ensure this is a supported device */ | |
663 | err = -EINVAL; | |
664 | if (!mpls_dev_get(dev)) | |
665 | goto errout; | |
666 | ||
a3e948e8 RS |
667 | if ((nh->nh_via_table == NEIGH_LINK_TABLE) && |
668 | (dev->addr_len != nh->nh_via_alen)) | |
669 | goto errout; | |
670 | ||
f8efb73c RP |
671 | RCU_INIT_POINTER(nh->nh_dev, dev); |
672 | ||
c89359a4 RP |
673 | if (!(dev->flags & IFF_UP)) { |
674 | nh->nh_flags |= RTNH_F_DEAD; | |
675 | } else { | |
676 | unsigned int flags; | |
677 | ||
678 | flags = dev_get_flags(dev); | |
679 | if (!(flags & (IFF_RUNNING | IFF_LOWER_UP))) | |
680 | nh->nh_flags |= RTNH_F_LINKDOWN; | |
681 | } | |
682 | ||
f8efb73c RP |
683 | return 0; |
684 | ||
685 | errout: | |
686 | return err; | |
687 | } | |
688 | ||
d4e72560 DA |
689 | static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table, |
690 | u8 via_addr[], struct netlink_ext_ack *extack) | |
691 | { | |
692 | struct rtvia *via = nla_data(nla); | |
693 | int err = -EINVAL; | |
694 | int alen; | |
695 | ||
696 | if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) { | |
697 | NL_SET_ERR_MSG_ATTR(extack, nla, | |
698 | "Invalid attribute length for RTA_VIA"); | |
699 | goto errout; | |
700 | } | |
701 | alen = nla_len(nla) - | |
702 | offsetof(struct rtvia, rtvia_addr); | |
703 | if (alen > MAX_VIA_ALEN) { | |
704 | NL_SET_ERR_MSG_ATTR(extack, nla, | |
705 | "Invalid address length for RTA_VIA"); | |
706 | goto errout; | |
707 | } | |
708 | ||
709 | /* Validate the address family */ | |
710 | switch (via->rtvia_family) { | |
711 | case AF_PACKET: | |
712 | *via_table = NEIGH_LINK_TABLE; | |
713 | break; | |
714 | case AF_INET: | |
715 | *via_table = NEIGH_ARP_TABLE; | |
716 | if (alen != 4) | |
717 | goto errout; | |
718 | break; | |
719 | case AF_INET6: | |
720 | *via_table = NEIGH_ND_TABLE; | |
721 | if (alen != 16) | |
722 | goto errout; | |
723 | break; | |
724 | default: | |
725 | /* Unsupported address family */ | |
726 | goto errout; | |
727 | } | |
728 | ||
729 | memcpy(via_addr, via->rtvia_addr, alen); | |
730 | *via_alen = alen; | |
731 | err = 0; | |
732 | ||
733 | errout: | |
734 | return err; | |
735 | } | |
736 | ||
f8efb73c RP |
737 | static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg, |
738 | struct mpls_route *rt) | |
739 | { | |
740 | struct net *net = cfg->rc_nlinfo.nl_net; | |
741 | struct mpls_nh *nh = rt->rt_nh; | |
742 | int err; | |
743 | int i; | |
744 | ||
745 | if (!nh) | |
746 | return -ENOMEM; | |
747 | ||
f8efb73c RP |
748 | nh->nh_labels = cfg->rc_output_labels; |
749 | for (i = 0; i < nh->nh_labels; i++) | |
750 | nh->nh_label[i] = cfg->rc_output_label[i]; | |
751 | ||
752 | nh->nh_via_table = cfg->rc_via_table; | |
cf4b24f0 | 753 | memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen); |
f8efb73c RP |
754 | nh->nh_via_alen = cfg->rc_via_alen; |
755 | ||
cf4b24f0 | 756 | err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex); |
f8efb73c RP |
757 | if (err) |
758 | goto errout; | |
759 | ||
c89359a4 RP |
760 | if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) |
761 | rt->rt_nhn_alive--; | |
762 | ||
f8efb73c RP |
763 | return 0; |
764 | ||
765 | errout: | |
766 | return err; | |
767 | } | |
768 | ||
cf4b24f0 | 769 | static int mpls_nh_build(struct net *net, struct mpls_route *rt, |
c89359a4 | 770 | struct mpls_nh *nh, int oif, struct nlattr *via, |
074350e2 DA |
771 | struct nlattr *newdst, u8 max_labels, |
772 | struct netlink_ext_ack *extack) | |
f8efb73c RP |
773 | { |
774 | int err = -ENOMEM; | |
775 | ||
776 | if (!nh) | |
777 | goto errout; | |
778 | ||
779 | if (newdst) { | |
a1f10abe | 780 | err = nla_get_labels(newdst, max_labels, &nh->nh_labels, |
074350e2 | 781 | nh->nh_label, extack); |
f8efb73c RP |
782 | if (err) |
783 | goto errout; | |
784 | } | |
785 | ||
f20367df RS |
786 | if (via) { |
787 | err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table, | |
074350e2 | 788 | __mpls_nh_via(rt, nh), extack); |
f20367df RS |
789 | if (err) |
790 | goto errout; | |
791 | } else { | |
792 | nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC; | |
793 | } | |
f8efb73c | 794 | |
cf4b24f0 | 795 | err = mpls_nh_assign_dev(net, rt, nh, oif); |
f8efb73c RP |
796 | if (err) |
797 | goto errout; | |
798 | ||
799 | return 0; | |
800 | ||
801 | errout: | |
802 | return err; | |
803 | } | |
804 | ||
77ef013a | 805 | static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len, |
a4ac8c98 DA |
806 | u8 cfg_via_alen, u8 *max_via_alen, |
807 | u8 *max_labels) | |
f8efb73c | 808 | { |
f8efb73c | 809 | int remaining = len; |
77ef013a | 810 | u8 nhs = 0; |
f8efb73c | 811 | |
cf4b24f0 | 812 | *max_via_alen = 0; |
a4ac8c98 | 813 | *max_labels = 0; |
cf4b24f0 | 814 | |
f8efb73c | 815 | while (rtnh_ok(rtnh, remaining)) { |
cf4b24f0 RS |
816 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); |
817 | int attrlen; | |
a4ac8c98 | 818 | u8 n_labels = 0; |
cf4b24f0 RS |
819 | |
820 | attrlen = rtnh_attrlen(rtnh); | |
821 | nla = nla_find(attrs, attrlen, RTA_VIA); | |
822 | if (nla && nla_len(nla) >= | |
823 | offsetof(struct rtvia, rtvia_addr)) { | |
824 | int via_alen = nla_len(nla) - | |
825 | offsetof(struct rtvia, rtvia_addr); | |
826 | ||
827 | if (via_alen <= MAX_VIA_ALEN) | |
828 | *max_via_alen = max_t(u16, *max_via_alen, | |
829 | via_alen); | |
830 | } | |
831 | ||
a4ac8c98 DA |
832 | nla = nla_find(attrs, attrlen, RTA_NEWDST); |
833 | if (nla && | |
a1f10abe DA |
834 | nla_get_labels(nla, MAX_NEW_LABELS, &n_labels, |
835 | NULL, NULL) != 0) | |
a4ac8c98 DA |
836 | return 0; |
837 | ||
838 | *max_labels = max_t(u8, *max_labels, n_labels); | |
839 | ||
77ef013a DA |
840 | /* number of nexthops is tracked by a u8. |
841 | * Check for overflow. | |
842 | */ | |
843 | if (nhs == 255) | |
844 | return 0; | |
f8efb73c | 845 | nhs++; |
77ef013a | 846 | |
f8efb73c RP |
847 | rtnh = rtnh_next(rtnh, &remaining); |
848 | } | |
849 | ||
850 | /* leftover implies invalid nexthop configuration, discard it */ | |
851 | return remaining > 0 ? 0 : nhs; | |
852 | } | |
853 | ||
854 | static int mpls_nh_build_multi(struct mpls_route_config *cfg, | |
074350e2 DA |
855 | struct mpls_route *rt, u8 max_labels, |
856 | struct netlink_ext_ack *extack) | |
f8efb73c RP |
857 | { |
858 | struct rtnexthop *rtnh = cfg->rc_mp; | |
859 | struct nlattr *nla_via, *nla_newdst; | |
860 | int remaining = cfg->rc_mp_len; | |
f8efb73c | 861 | int err = 0; |
77ef013a | 862 | u8 nhs = 0; |
f8efb73c RP |
863 | |
864 | change_nexthops(rt) { | |
865 | int attrlen; | |
866 | ||
867 | nla_via = NULL; | |
868 | nla_newdst = NULL; | |
869 | ||
870 | err = -EINVAL; | |
871 | if (!rtnh_ok(rtnh, remaining)) | |
872 | goto errout; | |
873 | ||
1c78efa8 RS |
874 | /* neither weighted multipath nor any flags |
875 | * are supported | |
876 | */ | |
877 | if (rtnh->rtnh_hops || rtnh->rtnh_flags) | |
878 | goto errout; | |
879 | ||
f8efb73c RP |
880 | attrlen = rtnh_attrlen(rtnh); |
881 | if (attrlen > 0) { | |
882 | struct nlattr *attrs = rtnh_attrs(rtnh); | |
883 | ||
884 | nla_via = nla_find(attrs, attrlen, RTA_VIA); | |
885 | nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST); | |
886 | } | |
887 | ||
cf4b24f0 | 888 | err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh, |
a4ac8c98 | 889 | rtnh->rtnh_ifindex, nla_via, nla_newdst, |
074350e2 | 890 | max_labels, extack); |
f8efb73c RP |
891 | if (err) |
892 | goto errout; | |
893 | ||
c89359a4 RP |
894 | if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) |
895 | rt->rt_nhn_alive--; | |
896 | ||
f8efb73c RP |
897 | rtnh = rtnh_next(rtnh, &remaining); |
898 | nhs++; | |
899 | } endfor_nexthops(rt); | |
900 | ||
901 | rt->rt_nhn = nhs; | |
902 | ||
903 | return 0; | |
904 | ||
905 | errout: | |
906 | return err; | |
907 | } | |
908 | ||
074350e2 DA |
909 | static bool mpls_label_ok(struct net *net, unsigned int index, |
910 | struct netlink_ext_ack *extack) | |
b7b386f4 DA |
911 | { |
912 | /* Reserved labels may not be set */ | |
074350e2 DA |
913 | if (index < MPLS_LABEL_FIRST_UNRESERVED) { |
914 | NL_SET_ERR_MSG(extack, | |
915 | "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher"); | |
b7b386f4 | 916 | return false; |
074350e2 | 917 | } |
b7b386f4 DA |
918 | |
919 | /* The full 20 bit range may not be supported. */ | |
074350e2 DA |
920 | if (index >= net->mpls.platform_labels) { |
921 | NL_SET_ERR_MSG(extack, | |
922 | "Label >= configured maximum in platform_labels"); | |
b7b386f4 | 923 | return false; |
074350e2 | 924 | } |
b7b386f4 DA |
925 | |
926 | return true; | |
927 | } | |
928 | ||
074350e2 DA |
929 | static int mpls_route_add(struct mpls_route_config *cfg, |
930 | struct netlink_ext_ack *extack) | |
a2519929 | 931 | { |
19d0c341 | 932 | struct mpls_route __rcu **platform_label; |
a2519929 | 933 | struct net *net = cfg->rc_nlinfo.nl_net; |
a2519929 | 934 | struct mpls_route *rt, *old; |
a2519929 | 935 | int err = -EINVAL; |
cf4b24f0 | 936 | u8 max_via_alen; |
f8efb73c | 937 | unsigned index; |
a4ac8c98 | 938 | u8 max_labels; |
77ef013a | 939 | u8 nhs; |
a2519929 EB |
940 | |
941 | index = cfg->rc_label; | |
942 | ||
943 | /* If a label was not specified during insert pick one */ | |
944 | if ((index == LABEL_NOT_SPECIFIED) && | |
945 | (cfg->rc_nlflags & NLM_F_CREATE)) { | |
946 | index = find_free_label(net); | |
947 | } | |
948 | ||
074350e2 | 949 | if (!mpls_label_ok(net, index, extack)) |
a2519929 EB |
950 | goto errout; |
951 | ||
a2519929 | 952 | /* Append makes no sense with mpls */ |
0f7bbd58 | 953 | err = -EOPNOTSUPP; |
074350e2 DA |
954 | if (cfg->rc_nlflags & NLM_F_APPEND) { |
955 | NL_SET_ERR_MSG(extack, "MPLS does not support route append"); | |
a2519929 | 956 | goto errout; |
074350e2 | 957 | } |
a2519929 EB |
958 | |
959 | err = -EEXIST; | |
19d0c341 EB |
960 | platform_label = rtnl_dereference(net->mpls.platform_label); |
961 | old = rtnl_dereference(platform_label[index]); | |
a2519929 EB |
962 | if ((cfg->rc_nlflags & NLM_F_EXCL) && old) |
963 | goto errout; | |
964 | ||
965 | err = -EEXIST; | |
966 | if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old) | |
967 | goto errout; | |
968 | ||
969 | err = -ENOENT; | |
970 | if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old) | |
971 | goto errout; | |
972 | ||
cf4b24f0 | 973 | err = -EINVAL; |
a4ac8c98 DA |
974 | if (cfg->rc_mp) { |
975 | nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len, | |
976 | cfg->rc_via_alen, &max_via_alen, | |
977 | &max_labels); | |
978 | } else { | |
979 | max_via_alen = cfg->rc_via_alen; | |
980 | max_labels = cfg->rc_output_labels; | |
981 | nhs = 1; | |
982 | } | |
983 | ||
074350e2 DA |
984 | if (nhs == 0) { |
985 | NL_SET_ERR_MSG(extack, "Route does not contain a nexthop"); | |
cf4b24f0 | 986 | goto errout; |
074350e2 | 987 | } |
f8efb73c | 988 | |
a2519929 | 989 | err = -ENOMEM; |
a4ac8c98 | 990 | rt = mpls_rt_alloc(nhs, max_via_alen, max_labels); |
df1c6316 DA |
991 | if (IS_ERR(rt)) { |
992 | err = PTR_ERR(rt); | |
a2519929 | 993 | goto errout; |
df1c6316 | 994 | } |
a2519929 | 995 | |
a2519929 | 996 | rt->rt_protocol = cfg->rc_protocol; |
118d5234 | 997 | rt->rt_payload_type = cfg->rc_payload_type; |
5b441ac8 | 998 | rt->rt_ttl_propagate = cfg->rc_ttl_propagate; |
a2519929 | 999 | |
f8efb73c | 1000 | if (cfg->rc_mp) |
074350e2 | 1001 | err = mpls_nh_build_multi(cfg, rt, max_labels, extack); |
f8efb73c RP |
1002 | else |
1003 | err = mpls_nh_build_from_cfg(cfg, rt); | |
1004 | if (err) | |
1005 | goto freert; | |
1006 | ||
1007 | mpls_route_update(net, index, rt, &cfg->rc_nlinfo); | |
a2519929 | 1008 | |
a2519929 EB |
1009 | return 0; |
1010 | ||
f8efb73c RP |
1011 | freert: |
1012 | mpls_rt_free(rt); | |
a2519929 | 1013 | errout: |
a2519929 EB |
1014 | return err; |
1015 | } | |
1016 | ||
074350e2 DA |
1017 | static int mpls_route_del(struct mpls_route_config *cfg, |
1018 | struct netlink_ext_ack *extack) | |
a2519929 EB |
1019 | { |
1020 | struct net *net = cfg->rc_nlinfo.nl_net; | |
1021 | unsigned index; | |
1022 | int err = -EINVAL; | |
1023 | ||
1024 | index = cfg->rc_label; | |
1025 | ||
074350e2 | 1026 | if (!mpls_label_ok(net, index, extack)) |
a2519929 EB |
1027 | goto errout; |
1028 | ||
f8efb73c | 1029 | mpls_route_update(net, index, NULL, &cfg->rc_nlinfo); |
a2519929 EB |
1030 | |
1031 | err = 0; | |
1032 | errout: | |
1033 | return err; | |
1034 | } | |
1035 | ||
27d69105 RS |
1036 | static void mpls_get_stats(struct mpls_dev *mdev, |
1037 | struct mpls_link_stats *stats) | |
1038 | { | |
1039 | struct mpls_pcpu_stats *p; | |
1040 | int i; | |
1041 | ||
1042 | memset(stats, 0, sizeof(*stats)); | |
1043 | ||
1044 | for_each_possible_cpu(i) { | |
1045 | struct mpls_link_stats local; | |
1046 | unsigned int start; | |
1047 | ||
1048 | p = per_cpu_ptr(mdev->stats, i); | |
1049 | do { | |
1050 | start = u64_stats_fetch_begin(&p->syncp); | |
1051 | local = p->stats; | |
1052 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
1053 | ||
1054 | stats->rx_packets += local.rx_packets; | |
1055 | stats->rx_bytes += local.rx_bytes; | |
1056 | stats->tx_packets += local.tx_packets; | |
1057 | stats->tx_bytes += local.tx_bytes; | |
1058 | stats->rx_errors += local.rx_errors; | |
1059 | stats->tx_errors += local.tx_errors; | |
1060 | stats->rx_dropped += local.rx_dropped; | |
1061 | stats->tx_dropped += local.tx_dropped; | |
1062 | stats->rx_noroute += local.rx_noroute; | |
1063 | } | |
1064 | } | |
1065 | ||
1066 | static int mpls_fill_stats_af(struct sk_buff *skb, | |
1067 | const struct net_device *dev) | |
1068 | { | |
1069 | struct mpls_link_stats *stats; | |
1070 | struct mpls_dev *mdev; | |
1071 | struct nlattr *nla; | |
1072 | ||
1073 | mdev = mpls_dev_get(dev); | |
1074 | if (!mdev) | |
1075 | return -ENODATA; | |
1076 | ||
1077 | nla = nla_reserve_64bit(skb, MPLS_STATS_LINK, | |
1078 | sizeof(struct mpls_link_stats), | |
1079 | MPLS_STATS_UNSPEC); | |
1080 | if (!nla) | |
1081 | return -EMSGSIZE; | |
1082 | ||
1083 | stats = nla_data(nla); | |
1084 | mpls_get_stats(mdev, stats); | |
1085 | ||
1086 | return 0; | |
1087 | } | |
1088 | ||
1089 | static size_t mpls_get_stats_af_size(const struct net_device *dev) | |
1090 | { | |
1091 | struct mpls_dev *mdev; | |
1092 | ||
1093 | mdev = mpls_dev_get(dev); | |
1094 | if (!mdev) | |
1095 | return 0; | |
1096 | ||
1097 | return nla_total_size_64bit(sizeof(struct mpls_link_stats)); | |
1098 | } | |
1099 | ||
24045a03 DA |
1100 | static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev, |
1101 | u32 portid, u32 seq, int event, | |
1102 | unsigned int flags, int type) | |
1103 | { | |
1104 | struct nlmsghdr *nlh; | |
1105 | struct netconfmsg *ncm; | |
1106 | bool all = false; | |
1107 | ||
1108 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), | |
1109 | flags); | |
1110 | if (!nlh) | |
1111 | return -EMSGSIZE; | |
1112 | ||
1113 | if (type == NETCONFA_ALL) | |
1114 | all = true; | |
1115 | ||
1116 | ncm = nlmsg_data(nlh); | |
1117 | ncm->ncm_family = AF_MPLS; | |
1118 | ||
1119 | if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0) | |
1120 | goto nla_put_failure; | |
1121 | ||
1122 | if ((all || type == NETCONFA_INPUT) && | |
1123 | nla_put_s32(skb, NETCONFA_INPUT, | |
1124 | mdev->input_enabled) < 0) | |
1125 | goto nla_put_failure; | |
1126 | ||
1127 | nlmsg_end(skb, nlh); | |
1128 | return 0; | |
1129 | ||
1130 | nla_put_failure: | |
1131 | nlmsg_cancel(skb, nlh); | |
1132 | return -EMSGSIZE; | |
1133 | } | |
1134 | ||
1135 | static int mpls_netconf_msgsize_devconf(int type) | |
1136 | { | |
1137 | int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) | |
1138 | + nla_total_size(4); /* NETCONFA_IFINDEX */ | |
1139 | bool all = false; | |
1140 | ||
1141 | if (type == NETCONFA_ALL) | |
1142 | all = true; | |
1143 | ||
1144 | if (all || type == NETCONFA_INPUT) | |
1145 | size += nla_total_size(4); | |
1146 | ||
1147 | return size; | |
1148 | } | |
1149 | ||
823566ae DA |
1150 | static void mpls_netconf_notify_devconf(struct net *net, int event, |
1151 | int type, struct mpls_dev *mdev) | |
24045a03 DA |
1152 | { |
1153 | struct sk_buff *skb; | |
1154 | int err = -ENOBUFS; | |
1155 | ||
1156 | skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL); | |
1157 | if (!skb) | |
1158 | goto errout; | |
1159 | ||
823566ae | 1160 | err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type); |
24045a03 DA |
1161 | if (err < 0) { |
1162 | /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */ | |
1163 | WARN_ON(err == -EMSGSIZE); | |
1164 | kfree_skb(skb); | |
1165 | goto errout; | |
1166 | } | |
1167 | ||
1168 | rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL); | |
1169 | return; | |
1170 | errout: | |
1171 | if (err < 0) | |
1172 | rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err); | |
1173 | } | |
1174 | ||
1175 | static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = { | |
1176 | [NETCONFA_IFINDEX] = { .len = sizeof(int) }, | |
1177 | }; | |
1178 | ||
1179 | static int mpls_netconf_get_devconf(struct sk_buff *in_skb, | |
c21ef3e3 DA |
1180 | struct nlmsghdr *nlh, |
1181 | struct netlink_ext_ack *extack) | |
24045a03 DA |
1182 | { |
1183 | struct net *net = sock_net(in_skb->sk); | |
1184 | struct nlattr *tb[NETCONFA_MAX + 1]; | |
1185 | struct netconfmsg *ncm; | |
1186 | struct net_device *dev; | |
1187 | struct mpls_dev *mdev; | |
1188 | struct sk_buff *skb; | |
1189 | int ifindex; | |
1190 | int err; | |
1191 | ||
1192 | err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX, | |
fceb6435 | 1193 | devconf_mpls_policy, NULL); |
24045a03 DA |
1194 | if (err < 0) |
1195 | goto errout; | |
1196 | ||
1197 | err = -EINVAL; | |
1198 | if (!tb[NETCONFA_IFINDEX]) | |
1199 | goto errout; | |
1200 | ||
1201 | ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); | |
1202 | dev = __dev_get_by_index(net, ifindex); | |
1203 | if (!dev) | |
1204 | goto errout; | |
1205 | ||
1206 | mdev = mpls_dev_get(dev); | |
1207 | if (!mdev) | |
1208 | goto errout; | |
1209 | ||
1210 | err = -ENOBUFS; | |
1211 | skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); | |
1212 | if (!skb) | |
1213 | goto errout; | |
1214 | ||
1215 | err = mpls_netconf_fill_devconf(skb, mdev, | |
1216 | NETLINK_CB(in_skb).portid, | |
1217 | nlh->nlmsg_seq, RTM_NEWNETCONF, 0, | |
1218 | NETCONFA_ALL); | |
1219 | if (err < 0) { | |
1220 | /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */ | |
1221 | WARN_ON(err == -EMSGSIZE); | |
1222 | kfree_skb(skb); | |
1223 | goto errout; | |
1224 | } | |
1225 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); | |
1226 | errout: | |
1227 | return err; | |
1228 | } | |
1229 | ||
1230 | static int mpls_netconf_dump_devconf(struct sk_buff *skb, | |
1231 | struct netlink_callback *cb) | |
1232 | { | |
1233 | struct net *net = sock_net(skb->sk); | |
1234 | struct hlist_head *head; | |
1235 | struct net_device *dev; | |
1236 | struct mpls_dev *mdev; | |
1237 | int idx, s_idx; | |
1238 | int h, s_h; | |
1239 | ||
1240 | s_h = cb->args[0]; | |
1241 | s_idx = idx = cb->args[1]; | |
1242 | ||
1243 | for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { | |
1244 | idx = 0; | |
1245 | head = &net->dev_index_head[h]; | |
1246 | rcu_read_lock(); | |
1247 | cb->seq = net->dev_base_seq; | |
1248 | hlist_for_each_entry_rcu(dev, head, index_hlist) { | |
1249 | if (idx < s_idx) | |
1250 | goto cont; | |
1251 | mdev = mpls_dev_get(dev); | |
1252 | if (!mdev) | |
1253 | goto cont; | |
1254 | if (mpls_netconf_fill_devconf(skb, mdev, | |
1255 | NETLINK_CB(cb->skb).portid, | |
1256 | cb->nlh->nlmsg_seq, | |
1257 | RTM_NEWNETCONF, | |
1258 | NLM_F_MULTI, | |
1259 | NETCONFA_ALL) < 0) { | |
1260 | rcu_read_unlock(); | |
1261 | goto done; | |
1262 | } | |
1263 | nl_dump_check_consistent(cb, nlmsg_hdr(skb)); | |
1264 | cont: | |
1265 | idx++; | |
1266 | } | |
1267 | rcu_read_unlock(); | |
1268 | } | |
1269 | done: | |
1270 | cb->args[0] = h; | |
1271 | cb->args[1] = idx; | |
1272 | ||
1273 | return skb->len; | |
1274 | } | |
1275 | ||
37bde799 RS |
1276 | #define MPLS_PERDEV_SYSCTL_OFFSET(field) \ |
1277 | (&((struct mpls_dev *)0)->field) | |
1278 | ||
24045a03 DA |
1279 | static int mpls_conf_proc(struct ctl_table *ctl, int write, |
1280 | void __user *buffer, | |
1281 | size_t *lenp, loff_t *ppos) | |
1282 | { | |
1283 | int oval = *(int *)ctl->data; | |
1284 | int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); | |
1285 | ||
1286 | if (write) { | |
1287 | struct mpls_dev *mdev = ctl->extra1; | |
1288 | int i = (int *)ctl->data - (int *)mdev; | |
1289 | struct net *net = ctl->extra2; | |
1290 | int val = *(int *)ctl->data; | |
1291 | ||
1292 | if (i == offsetof(struct mpls_dev, input_enabled) && | |
1293 | val != oval) { | |
823566ae DA |
1294 | mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, |
1295 | NETCONFA_INPUT, mdev); | |
24045a03 DA |
1296 | } |
1297 | } | |
1298 | ||
1299 | return ret; | |
1300 | } | |
1301 | ||
37bde799 RS |
1302 | static const struct ctl_table mpls_dev_table[] = { |
1303 | { | |
1304 | .procname = "input", | |
1305 | .maxlen = sizeof(int), | |
1306 | .mode = 0644, | |
24045a03 | 1307 | .proc_handler = mpls_conf_proc, |
37bde799 RS |
1308 | .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled), |
1309 | }, | |
1310 | { } | |
1311 | }; | |
1312 | ||
1313 | static int mpls_dev_sysctl_register(struct net_device *dev, | |
1314 | struct mpls_dev *mdev) | |
1315 | { | |
1316 | char path[sizeof("net/mpls/conf/") + IFNAMSIZ]; | |
24045a03 | 1317 | struct net *net = dev_net(dev); |
37bde799 RS |
1318 | struct ctl_table *table; |
1319 | int i; | |
1320 | ||
1321 | table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL); | |
1322 | if (!table) | |
1323 | goto out; | |
1324 | ||
1325 | /* Table data contains only offsets relative to the base of | |
1326 | * the mdev at this point, so make them absolute. | |
1327 | */ | |
24045a03 | 1328 | for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) { |
37bde799 | 1329 | table[i].data = (char *)mdev + (uintptr_t)table[i].data; |
24045a03 DA |
1330 | table[i].extra1 = mdev; |
1331 | table[i].extra2 = net; | |
1332 | } | |
37bde799 RS |
1333 | |
1334 | snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name); | |
1335 | ||
1182e4d0 | 1336 | mdev->sysctl = register_net_sysctl(net, path, table); |
37bde799 RS |
1337 | if (!mdev->sysctl) |
1338 | goto free; | |
1339 | ||
1182e4d0 | 1340 | mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev); |
37bde799 RS |
1341 | return 0; |
1342 | ||
1343 | free: | |
1344 | kfree(table); | |
1345 | out: | |
1346 | return -ENOBUFS; | |
1347 | } | |
1348 | ||
1182e4d0 DA |
1349 | static void mpls_dev_sysctl_unregister(struct net_device *dev, |
1350 | struct mpls_dev *mdev) | |
37bde799 | 1351 | { |
1182e4d0 | 1352 | struct net *net = dev_net(dev); |
37bde799 RS |
1353 | struct ctl_table *table; |
1354 | ||
1355 | table = mdev->sysctl->ctl_table_arg; | |
1356 | unregister_net_sysctl_table(mdev->sysctl); | |
1357 | kfree(table); | |
1182e4d0 DA |
1358 | |
1359 | mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev); | |
37bde799 RS |
1360 | } |
1361 | ||
03c57747 RS |
1362 | static struct mpls_dev *mpls_add_dev(struct net_device *dev) |
1363 | { | |
1364 | struct mpls_dev *mdev; | |
1365 | int err = -ENOMEM; | |
27d69105 | 1366 | int i; |
03c57747 RS |
1367 | |
1368 | ASSERT_RTNL(); | |
1369 | ||
1370 | mdev = kzalloc(sizeof(*mdev), GFP_KERNEL); | |
1371 | if (!mdev) | |
1372 | return ERR_PTR(err); | |
1373 | ||
27d69105 RS |
1374 | mdev->stats = alloc_percpu(struct mpls_pcpu_stats); |
1375 | if (!mdev->stats) | |
1376 | goto free; | |
1377 | ||
1378 | for_each_possible_cpu(i) { | |
1379 | struct mpls_pcpu_stats *mpls_stats; | |
1380 | ||
1381 | mpls_stats = per_cpu_ptr(mdev->stats, i); | |
1382 | u64_stats_init(&mpls_stats->syncp); | |
1383 | } | |
1384 | ||
1182e4d0 DA |
1385 | mdev->dev = dev; |
1386 | ||
37bde799 RS |
1387 | err = mpls_dev_sysctl_register(dev, mdev); |
1388 | if (err) | |
1389 | goto free; | |
1390 | ||
03c57747 RS |
1391 | rcu_assign_pointer(dev->mpls_ptr, mdev); |
1392 | ||
1393 | return mdev; | |
37bde799 RS |
1394 | |
1395 | free: | |
27d69105 | 1396 | free_percpu(mdev->stats); |
37bde799 RS |
1397 | kfree(mdev); |
1398 | return ERR_PTR(err); | |
03c57747 RS |
1399 | } |
1400 | ||
27d69105 RS |
1401 | static void mpls_dev_destroy_rcu(struct rcu_head *head) |
1402 | { | |
1403 | struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu); | |
1404 | ||
1405 | free_percpu(mdev->stats); | |
1406 | kfree(mdev); | |
1407 | } | |
1408 | ||
c89359a4 | 1409 | static void mpls_ifdown(struct net_device *dev, int event) |
0189197f | 1410 | { |
19d0c341 | 1411 | struct mpls_route __rcu **platform_label; |
0189197f | 1412 | struct net *net = dev_net(dev); |
77ef013a | 1413 | u8 alive, deleted; |
0189197f EB |
1414 | unsigned index; |
1415 | ||
19d0c341 | 1416 | platform_label = rtnl_dereference(net->mpls.platform_label); |
0189197f | 1417 | for (index = 0; index < net->mpls.platform_labels; index++) { |
19d0c341 | 1418 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); |
c89359a4 | 1419 | |
0189197f EB |
1420 | if (!rt) |
1421 | continue; | |
c89359a4 | 1422 | |
61733c91 | 1423 | alive = 0; |
4ea8efad | 1424 | deleted = 0; |
c89359a4 | 1425 | change_nexthops(rt) { |
39eb8cd1 DA |
1426 | unsigned int nh_flags = nh->nh_flags; |
1427 | ||
f8efb73c | 1428 | if (rtnl_dereference(nh->nh_dev) != dev) |
61733c91 DA |
1429 | goto next; |
1430 | ||
c89359a4 RP |
1431 | switch (event) { |
1432 | case NETDEV_DOWN: | |
1433 | case NETDEV_UNREGISTER: | |
39eb8cd1 | 1434 | nh_flags |= RTNH_F_DEAD; |
c89359a4 RP |
1435 | /* fall through */ |
1436 | case NETDEV_CHANGE: | |
39eb8cd1 | 1437 | nh_flags |= RTNH_F_LINKDOWN; |
c89359a4 RP |
1438 | break; |
1439 | } | |
1440 | if (event == NETDEV_UNREGISTER) | |
1441 | RCU_INIT_POINTER(nh->nh_dev, NULL); | |
39eb8cd1 DA |
1442 | |
1443 | if (nh->nh_flags != nh_flags) | |
1444 | WRITE_ONCE(nh->nh_flags, nh_flags); | |
61733c91 | 1445 | next: |
39eb8cd1 | 1446 | if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))) |
61733c91 | 1447 | alive++; |
4ea8efad DA |
1448 | if (!rtnl_dereference(nh->nh_dev)) |
1449 | deleted++; | |
f8efb73c | 1450 | } endfor_nexthops(rt); |
61733c91 DA |
1451 | |
1452 | WRITE_ONCE(rt->rt_nhn_alive, alive); | |
4ea8efad DA |
1453 | |
1454 | /* if there are no more nexthops, delete the route */ | |
1455 | if (event == NETDEV_UNREGISTER && deleted == rt->rt_nhn) | |
1456 | mpls_route_update(net, index, NULL, NULL); | |
0189197f | 1457 | } |
c89359a4 RP |
1458 | } |
1459 | ||
39eb8cd1 | 1460 | static void mpls_ifup(struct net_device *dev, unsigned int flags) |
c89359a4 RP |
1461 | { |
1462 | struct mpls_route __rcu **platform_label; | |
1463 | struct net *net = dev_net(dev); | |
1464 | unsigned index; | |
77ef013a | 1465 | u8 alive; |
c89359a4 RP |
1466 | |
1467 | platform_label = rtnl_dereference(net->mpls.platform_label); | |
1468 | for (index = 0; index < net->mpls.platform_labels; index++) { | |
1469 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); | |
1470 | ||
1471 | if (!rt) | |
1472 | continue; | |
37bde799 | 1473 | |
c89359a4 RP |
1474 | alive = 0; |
1475 | change_nexthops(rt) { | |
39eb8cd1 | 1476 | unsigned int nh_flags = nh->nh_flags; |
c89359a4 RP |
1477 | struct net_device *nh_dev = |
1478 | rtnl_dereference(nh->nh_dev); | |
03c57747 | 1479 | |
39eb8cd1 | 1480 | if (!(nh_flags & flags)) { |
c89359a4 RP |
1481 | alive++; |
1482 | continue; | |
1483 | } | |
1484 | if (nh_dev != dev) | |
1485 | continue; | |
1486 | alive++; | |
39eb8cd1 | 1487 | nh_flags &= ~flags; |
c2e8471d | 1488 | WRITE_ONCE(nh->nh_flags, nh_flags); |
c89359a4 RP |
1489 | } endfor_nexthops(rt); |
1490 | ||
39eb8cd1 | 1491 | WRITE_ONCE(rt->rt_nhn_alive, alive); |
c89359a4 | 1492 | } |
0189197f EB |
1493 | } |
1494 | ||
1495 | static int mpls_dev_notify(struct notifier_block *this, unsigned long event, | |
1496 | void *ptr) | |
1497 | { | |
1498 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
03c57747 | 1499 | struct mpls_dev *mdev; |
c89359a4 | 1500 | unsigned int flags; |
0189197f | 1501 | |
c89359a4 | 1502 | if (event == NETDEV_REGISTER) { |
407f31be | 1503 | /* For now just support Ethernet, IPGRE, SIT and IPIP devices */ |
0d227a86 SH |
1504 | if (dev->type == ARPHRD_ETHER || |
1505 | dev->type == ARPHRD_LOOPBACK || | |
407f31be SH |
1506 | dev->type == ARPHRD_IPGRE || |
1507 | dev->type == ARPHRD_SIT || | |
1508 | dev->type == ARPHRD_TUNNEL) { | |
03c57747 RS |
1509 | mdev = mpls_add_dev(dev); |
1510 | if (IS_ERR(mdev)) | |
1511 | return notifier_from_errno(PTR_ERR(mdev)); | |
1512 | } | |
c89359a4 RP |
1513 | return NOTIFY_OK; |
1514 | } | |
03c57747 | 1515 | |
c89359a4 RP |
1516 | mdev = mpls_dev_get(dev); |
1517 | if (!mdev) | |
1518 | return NOTIFY_OK; | |
1519 | ||
1520 | switch (event) { | |
1521 | case NETDEV_DOWN: | |
1522 | mpls_ifdown(dev, event); | |
1523 | break; | |
1524 | case NETDEV_UP: | |
1525 | flags = dev_get_flags(dev); | |
1526 | if (flags & (IFF_RUNNING | IFF_LOWER_UP)) | |
1527 | mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN); | |
1528 | else | |
1529 | mpls_ifup(dev, RTNH_F_DEAD); | |
1530 | break; | |
1531 | case NETDEV_CHANGE: | |
1532 | flags = dev_get_flags(dev); | |
1533 | if (flags & (IFF_RUNNING | IFF_LOWER_UP)) | |
1534 | mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN); | |
1535 | else | |
1536 | mpls_ifdown(dev, event); | |
1537 | break; | |
0189197f | 1538 | case NETDEV_UNREGISTER: |
c89359a4 RP |
1539 | mpls_ifdown(dev, event); |
1540 | mdev = mpls_dev_get(dev); | |
1541 | if (mdev) { | |
1182e4d0 | 1542 | mpls_dev_sysctl_unregister(dev, mdev); |
c89359a4 | 1543 | RCU_INIT_POINTER(dev->mpls_ptr, NULL); |
27d69105 | 1544 | call_rcu(&mdev->rcu, mpls_dev_destroy_rcu); |
c89359a4 | 1545 | } |
0189197f | 1546 | break; |
0fae3bf0 RS |
1547 | case NETDEV_CHANGENAME: |
1548 | mdev = mpls_dev_get(dev); | |
1549 | if (mdev) { | |
1550 | int err; | |
1551 | ||
1182e4d0 | 1552 | mpls_dev_sysctl_unregister(dev, mdev); |
0fae3bf0 RS |
1553 | err = mpls_dev_sysctl_register(dev, mdev); |
1554 | if (err) | |
1555 | return notifier_from_errno(err); | |
1556 | } | |
1557 | break; | |
0189197f EB |
1558 | } |
1559 | return NOTIFY_OK; | |
1560 | } | |
1561 | ||
1562 | static struct notifier_block mpls_dev_notifier = { | |
1563 | .notifier_call = mpls_dev_notify, | |
1564 | }; | |
1565 | ||
03c05665 | 1566 | static int nla_put_via(struct sk_buff *skb, |
b79bda3d | 1567 | u8 table, const void *addr, int alen) |
03c05665 | 1568 | { |
b79bda3d EB |
1569 | static const int table_to_family[NEIGH_NR_TABLES + 1] = { |
1570 | AF_INET, AF_INET6, AF_DECnet, AF_PACKET, | |
1571 | }; | |
03c05665 EB |
1572 | struct nlattr *nla; |
1573 | struct rtvia *via; | |
b79bda3d | 1574 | int family = AF_UNSPEC; |
03c05665 EB |
1575 | |
1576 | nla = nla_reserve(skb, RTA_VIA, alen + 2); | |
1577 | if (!nla) | |
1578 | return -EMSGSIZE; | |
1579 | ||
b79bda3d EB |
1580 | if (table <= NEIGH_NR_TABLES) |
1581 | family = table_to_family[table]; | |
1582 | ||
03c05665 EB |
1583 | via = nla_data(nla); |
1584 | via->rtvia_family = family; | |
1585 | memcpy(via->rtvia_addr, addr, alen); | |
1586 | return 0; | |
1587 | } | |
1588 | ||
966bae33 EB |
1589 | int nla_put_labels(struct sk_buff *skb, int attrtype, |
1590 | u8 labels, const u32 label[]) | |
1591 | { | |
1592 | struct nlattr *nla; | |
1593 | struct mpls_shim_hdr *nla_label; | |
1594 | bool bos; | |
1595 | int i; | |
1596 | nla = nla_reserve(skb, attrtype, labels*4); | |
1597 | if (!nla) | |
1598 | return -EMSGSIZE; | |
1599 | ||
1600 | nla_label = nla_data(nla); | |
1601 | bos = true; | |
1602 | for (i = labels - 1; i >= 0; i--) { | |
1603 | nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos); | |
1604 | bos = false; | |
1605 | } | |
1606 | ||
1607 | return 0; | |
1608 | } | |
face0188 | 1609 | EXPORT_SYMBOL_GPL(nla_put_labels); |
966bae33 | 1610 | |
a1f10abe DA |
1611 | int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels, |
1612 | u32 label[], struct netlink_ext_ack *extack) | |
966bae33 EB |
1613 | { |
1614 | unsigned len = nla_len(nla); | |
966bae33 | 1615 | struct mpls_shim_hdr *nla_label; |
a4ac8c98 | 1616 | u8 nla_labels; |
966bae33 EB |
1617 | bool bos; |
1618 | int i; | |
1619 | ||
a4ac8c98 DA |
1620 | /* len needs to be an even multiple of 4 (the label size). Number |
1621 | * of labels is a u8 so check for overflow. | |
1622 | */ | |
a1f10abe DA |
1623 | if (len & 3 || len / 4 > 255) { |
1624 | NL_SET_ERR_MSG_ATTR(extack, nla, | |
1625 | "Invalid length for labels attribute"); | |
966bae33 | 1626 | return -EINVAL; |
a1f10abe | 1627 | } |
966bae33 EB |
1628 | |
1629 | /* Limit the number of new labels allowed */ | |
1630 | nla_labels = len/4; | |
a1f10abe DA |
1631 | if (nla_labels > max_labels) { |
1632 | NL_SET_ERR_MSG(extack, "Too many labels"); | |
966bae33 | 1633 | return -EINVAL; |
a1f10abe | 1634 | } |
966bae33 | 1635 | |
a4ac8c98 DA |
1636 | /* when label == NULL, caller wants number of labels */ |
1637 | if (!label) | |
1638 | goto out; | |
1639 | ||
966bae33 EB |
1640 | nla_label = nla_data(nla); |
1641 | bos = true; | |
1642 | for (i = nla_labels - 1; i >= 0; i--, bos = false) { | |
1643 | struct mpls_entry_decoded dec; | |
1644 | dec = mpls_entry_decode(nla_label + i); | |
1645 | ||
1646 | /* Ensure the bottom of stack flag is properly set | |
1647 | * and ttl and tc are both clear. | |
1648 | */ | |
a1f10abe DA |
1649 | if (dec.ttl) { |
1650 | NL_SET_ERR_MSG_ATTR(extack, nla, | |
1651 | "TTL in label must be 0"); | |
1652 | return -EINVAL; | |
1653 | } | |
1654 | ||
1655 | if (dec.tc) { | |
1656 | NL_SET_ERR_MSG_ATTR(extack, nla, | |
1657 | "Traffic class in label must be 0"); | |
966bae33 | 1658 | return -EINVAL; |
a1f10abe DA |
1659 | } |
1660 | ||
1661 | if (dec.bos != bos) { | |
1662 | NL_SET_BAD_ATTR(extack, nla); | |
1663 | if (bos) { | |
1664 | NL_SET_ERR_MSG(extack, | |
1665 | "BOS bit must be set in first label"); | |
1666 | } else { | |
1667 | NL_SET_ERR_MSG(extack, | |
1668 | "BOS bit can only be set in first label"); | |
1669 | } | |
1670 | return -EINVAL; | |
1671 | } | |
966bae33 | 1672 | |
5a9ab017 | 1673 | switch (dec.label) { |
78f5b899 | 1674 | case MPLS_LABEL_IMPLNULL: |
5a9ab017 RS |
1675 | /* RFC3032: This is a label that an LSR may |
1676 | * assign and distribute, but which never | |
1677 | * actually appears in the encapsulation. | |
1678 | */ | |
a1f10abe DA |
1679 | NL_SET_ERR_MSG_ATTR(extack, nla, |
1680 | "Implicit NULL Label (3) can not be used in encapsulation"); | |
5a9ab017 RS |
1681 | return -EINVAL; |
1682 | } | |
1683 | ||
966bae33 EB |
1684 | label[i] = dec.label; |
1685 | } | |
a4ac8c98 | 1686 | out: |
966bae33 EB |
1687 | *labels = nla_labels; |
1688 | return 0; | |
1689 | } | |
face0188 | 1690 | EXPORT_SYMBOL_GPL(nla_get_labels); |
966bae33 | 1691 | |
074350e2 DA |
1692 | static int rtm_to_route_config(struct sk_buff *skb, |
1693 | struct nlmsghdr *nlh, | |
1694 | struct mpls_route_config *cfg, | |
1695 | struct netlink_ext_ack *extack) | |
03c05665 EB |
1696 | { |
1697 | struct rtmsg *rtm; | |
1698 | struct nlattr *tb[RTA_MAX+1]; | |
1699 | int index; | |
1700 | int err; | |
1701 | ||
fceb6435 | 1702 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_mpls_policy, |
074350e2 | 1703 | extack); |
03c05665 EB |
1704 | if (err < 0) |
1705 | goto errout; | |
1706 | ||
1707 | err = -EINVAL; | |
1708 | rtm = nlmsg_data(nlh); | |
03c05665 | 1709 | |
074350e2 DA |
1710 | if (rtm->rtm_family != AF_MPLS) { |
1711 | NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg"); | |
03c05665 | 1712 | goto errout; |
074350e2 DA |
1713 | } |
1714 | if (rtm->rtm_dst_len != 20) { | |
1715 | NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS"); | |
03c05665 | 1716 | goto errout; |
074350e2 DA |
1717 | } |
1718 | if (rtm->rtm_src_len != 0) { | |
1719 | NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS"); | |
03c05665 | 1720 | goto errout; |
074350e2 DA |
1721 | } |
1722 | if (rtm->rtm_tos != 0) { | |
1723 | NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS"); | |
03c05665 | 1724 | goto errout; |
074350e2 DA |
1725 | } |
1726 | if (rtm->rtm_table != RT_TABLE_MAIN) { | |
1727 | NL_SET_ERR_MSG(extack, | |
1728 | "MPLS only supports the main route table"); | |
03c05665 | 1729 | goto errout; |
074350e2 | 1730 | } |
03c05665 EB |
1731 | /* Any value is acceptable for rtm_protocol */ |
1732 | ||
1733 | /* As mpls uses destination specific addresses | |
1734 | * (or source specific address in the case of multicast) | |
1735 | * all addresses have universal scope. | |
1736 | */ | |
074350e2 DA |
1737 | if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) { |
1738 | NL_SET_ERR_MSG(extack, | |
1739 | "Invalid route scope - MPLS only supports UNIVERSE"); | |
03c05665 | 1740 | goto errout; |
074350e2 DA |
1741 | } |
1742 | if (rtm->rtm_type != RTN_UNICAST) { | |
1743 | NL_SET_ERR_MSG(extack, | |
1744 | "Invalid route type - MPLS only supports UNICAST"); | |
03c05665 | 1745 | goto errout; |
074350e2 DA |
1746 | } |
1747 | if (rtm->rtm_flags != 0) { | |
1748 | NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS"); | |
03c05665 | 1749 | goto errout; |
074350e2 | 1750 | } |
03c05665 EB |
1751 | |
1752 | cfg->rc_label = LABEL_NOT_SPECIFIED; | |
1753 | cfg->rc_protocol = rtm->rtm_protocol; | |
eb7809f0 | 1754 | cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC; |
5b441ac8 | 1755 | cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT; |
03c05665 EB |
1756 | cfg->rc_nlflags = nlh->nlmsg_flags; |
1757 | cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid; | |
1758 | cfg->rc_nlinfo.nlh = nlh; | |
1759 | cfg->rc_nlinfo.nl_net = sock_net(skb->sk); | |
1760 | ||
1761 | for (index = 0; index <= RTA_MAX; index++) { | |
1762 | struct nlattr *nla = tb[index]; | |
1763 | if (!nla) | |
1764 | continue; | |
1765 | ||
14dd3e1b | 1766 | switch (index) { |
03c05665 EB |
1767 | case RTA_OIF: |
1768 | cfg->rc_ifindex = nla_get_u32(nla); | |
1769 | break; | |
1770 | case RTA_NEWDST: | |
1771 | if (nla_get_labels(nla, MAX_NEW_LABELS, | |
1772 | &cfg->rc_output_labels, | |
074350e2 | 1773 | cfg->rc_output_label, extack)) |
03c05665 EB |
1774 | goto errout; |
1775 | break; | |
1776 | case RTA_DST: | |
1777 | { | |
f8efb73c | 1778 | u8 label_count; |
03c05665 | 1779 | if (nla_get_labels(nla, 1, &label_count, |
074350e2 | 1780 | &cfg->rc_label, extack)) |
03c05665 EB |
1781 | goto errout; |
1782 | ||
b7b386f4 | 1783 | if (!mpls_label_ok(cfg->rc_nlinfo.nl_net, |
074350e2 | 1784 | cfg->rc_label, extack)) |
03c05665 | 1785 | goto errout; |
03c05665 EB |
1786 | break; |
1787 | } | |
1788 | case RTA_VIA: | |
1789 | { | |
f8efb73c | 1790 | if (nla_get_via(nla, &cfg->rc_via_alen, |
074350e2 DA |
1791 | &cfg->rc_via_table, cfg->rc_via, |
1792 | extack)) | |
03c05665 | 1793 | goto errout; |
f8efb73c RP |
1794 | break; |
1795 | } | |
1796 | case RTA_MULTIPATH: | |
1797 | { | |
1798 | cfg->rc_mp = nla_data(nla); | |
1799 | cfg->rc_mp_len = nla_len(nla); | |
03c05665 EB |
1800 | break; |
1801 | } | |
5b441ac8 RS |
1802 | case RTA_TTL_PROPAGATE: |
1803 | { | |
1804 | u8 ttl_propagate = nla_get_u8(nla); | |
1805 | ||
074350e2 DA |
1806 | if (ttl_propagate > 1) { |
1807 | NL_SET_ERR_MSG_ATTR(extack, nla, | |
1808 | "RTA_TTL_PROPAGATE can only be 0 or 1"); | |
5b441ac8 | 1809 | goto errout; |
074350e2 | 1810 | } |
5b441ac8 RS |
1811 | cfg->rc_ttl_propagate = ttl_propagate ? |
1812 | MPLS_TTL_PROP_ENABLED : | |
1813 | MPLS_TTL_PROP_DISABLED; | |
1814 | break; | |
1815 | } | |
03c05665 | 1816 | default: |
074350e2 | 1817 | NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute"); |
03c05665 EB |
1818 | /* Unsupported attribute */ |
1819 | goto errout; | |
1820 | } | |
1821 | } | |
1822 | ||
1823 | err = 0; | |
1824 | errout: | |
1825 | return err; | |
1826 | } | |
1827 | ||
c21ef3e3 DA |
1828 | static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, |
1829 | struct netlink_ext_ack *extack) | |
03c05665 | 1830 | { |
a4ac8c98 | 1831 | struct mpls_route_config *cfg; |
03c05665 EB |
1832 | int err; |
1833 | ||
a4ac8c98 DA |
1834 | cfg = kzalloc(sizeof(*cfg), GFP_KERNEL); |
1835 | if (!cfg) | |
1836 | return -ENOMEM; | |
1837 | ||
074350e2 | 1838 | err = rtm_to_route_config(skb, nlh, cfg, extack); |
03c05665 | 1839 | if (err < 0) |
a4ac8c98 | 1840 | goto out; |
03c05665 | 1841 | |
074350e2 | 1842 | err = mpls_route_del(cfg, extack); |
a4ac8c98 DA |
1843 | out: |
1844 | kfree(cfg); | |
1845 | ||
1846 | return err; | |
03c05665 EB |
1847 | } |
1848 | ||
1849 | ||
c21ef3e3 DA |
1850 | static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, |
1851 | struct netlink_ext_ack *extack) | |
03c05665 | 1852 | { |
a4ac8c98 | 1853 | struct mpls_route_config *cfg; |
03c05665 EB |
1854 | int err; |
1855 | ||
a4ac8c98 DA |
1856 | cfg = kzalloc(sizeof(*cfg), GFP_KERNEL); |
1857 | if (!cfg) | |
1858 | return -ENOMEM; | |
1859 | ||
074350e2 | 1860 | err = rtm_to_route_config(skb, nlh, cfg, extack); |
03c05665 | 1861 | if (err < 0) |
a4ac8c98 | 1862 | goto out; |
03c05665 | 1863 | |
074350e2 | 1864 | err = mpls_route_add(cfg, extack); |
a4ac8c98 DA |
1865 | out: |
1866 | kfree(cfg); | |
1867 | ||
1868 | return err; | |
03c05665 EB |
1869 | } |
1870 | ||
1871 | static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event, | |
1872 | u32 label, struct mpls_route *rt, int flags) | |
1873 | { | |
19d0c341 | 1874 | struct net_device *dev; |
03c05665 EB |
1875 | struct nlmsghdr *nlh; |
1876 | struct rtmsg *rtm; | |
1877 | ||
1878 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags); | |
1879 | if (nlh == NULL) | |
1880 | return -EMSGSIZE; | |
1881 | ||
1882 | rtm = nlmsg_data(nlh); | |
1883 | rtm->rtm_family = AF_MPLS; | |
1884 | rtm->rtm_dst_len = 20; | |
1885 | rtm->rtm_src_len = 0; | |
1886 | rtm->rtm_tos = 0; | |
1887 | rtm->rtm_table = RT_TABLE_MAIN; | |
1888 | rtm->rtm_protocol = rt->rt_protocol; | |
1889 | rtm->rtm_scope = RT_SCOPE_UNIVERSE; | |
1890 | rtm->rtm_type = RTN_UNICAST; | |
1891 | rtm->rtm_flags = 0; | |
1892 | ||
03c05665 EB |
1893 | if (nla_put_labels(skb, RTA_DST, 1, &label)) |
1894 | goto nla_put_failure; | |
5b441ac8 RS |
1895 | |
1896 | if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) { | |
1897 | bool ttl_propagate = | |
1898 | rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED; | |
1899 | ||
1900 | if (nla_put_u8(skb, RTA_TTL_PROPAGATE, | |
1901 | ttl_propagate)) | |
1902 | goto nla_put_failure; | |
1903 | } | |
f8efb73c | 1904 | if (rt->rt_nhn == 1) { |
cf4b24f0 | 1905 | const struct mpls_nh *nh = rt->rt_nh; |
f8efb73c RP |
1906 | |
1907 | if (nh->nh_labels && | |
1908 | nla_put_labels(skb, RTA_NEWDST, nh->nh_labels, | |
1909 | nh->nh_label)) | |
1910 | goto nla_put_failure; | |
eb7809f0 | 1911 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && |
72dcac96 | 1912 | nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh), |
f8efb73c RP |
1913 | nh->nh_via_alen)) |
1914 | goto nla_put_failure; | |
1915 | dev = rtnl_dereference(nh->nh_dev); | |
1916 | if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) | |
1917 | goto nla_put_failure; | |
c89359a4 RP |
1918 | if (nh->nh_flags & RTNH_F_LINKDOWN) |
1919 | rtm->rtm_flags |= RTNH_F_LINKDOWN; | |
1920 | if (nh->nh_flags & RTNH_F_DEAD) | |
1921 | rtm->rtm_flags |= RTNH_F_DEAD; | |
f8efb73c RP |
1922 | } else { |
1923 | struct rtnexthop *rtnh; | |
1924 | struct nlattr *mp; | |
77ef013a DA |
1925 | u8 linkdown = 0; |
1926 | u8 dead = 0; | |
f8efb73c RP |
1927 | |
1928 | mp = nla_nest_start(skb, RTA_MULTIPATH); | |
1929 | if (!mp) | |
1930 | goto nla_put_failure; | |
1931 | ||
1932 | for_nexthops(rt) { | |
c00e51dd DA |
1933 | dev = rtnl_dereference(nh->nh_dev); |
1934 | if (!dev) | |
1935 | continue; | |
1936 | ||
f8efb73c RP |
1937 | rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); |
1938 | if (!rtnh) | |
1939 | goto nla_put_failure; | |
1940 | ||
c00e51dd | 1941 | rtnh->rtnh_ifindex = dev->ifindex; |
c89359a4 RP |
1942 | if (nh->nh_flags & RTNH_F_LINKDOWN) { |
1943 | rtnh->rtnh_flags |= RTNH_F_LINKDOWN; | |
1944 | linkdown++; | |
1945 | } | |
1946 | if (nh->nh_flags & RTNH_F_DEAD) { | |
1947 | rtnh->rtnh_flags |= RTNH_F_DEAD; | |
1948 | dead++; | |
1949 | } | |
1950 | ||
f8efb73c RP |
1951 | if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST, |
1952 | nh->nh_labels, | |
1953 | nh->nh_label)) | |
1954 | goto nla_put_failure; | |
f20367df RS |
1955 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && |
1956 | nla_put_via(skb, nh->nh_via_table, | |
cf4b24f0 | 1957 | mpls_nh_via(rt, nh), |
f8efb73c RP |
1958 | nh->nh_via_alen)) |
1959 | goto nla_put_failure; | |
1960 | ||
1961 | /* length of rtnetlink header + attributes */ | |
1962 | rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh; | |
1963 | } endfor_nexthops(rt); | |
1964 | ||
c89359a4 RP |
1965 | if (linkdown == rt->rt_nhn) |
1966 | rtm->rtm_flags |= RTNH_F_LINKDOWN; | |
1967 | if (dead == rt->rt_nhn) | |
1968 | rtm->rtm_flags |= RTNH_F_DEAD; | |
1969 | ||
f8efb73c RP |
1970 | nla_nest_end(skb, mp); |
1971 | } | |
03c05665 EB |
1972 | |
1973 | nlmsg_end(skb, nlh); | |
1974 | return 0; | |
1975 | ||
1976 | nla_put_failure: | |
1977 | nlmsg_cancel(skb, nlh); | |
1978 | return -EMSGSIZE; | |
1979 | } | |
1980 | ||
1981 | static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb) | |
1982 | { | |
1983 | struct net *net = sock_net(skb->sk); | |
19d0c341 EB |
1984 | struct mpls_route __rcu **platform_label; |
1985 | size_t platform_labels; | |
03c05665 EB |
1986 | unsigned int index; |
1987 | ||
1988 | ASSERT_RTNL(); | |
1989 | ||
1990 | index = cb->args[0]; | |
a6affd24 RS |
1991 | if (index < MPLS_LABEL_FIRST_UNRESERVED) |
1992 | index = MPLS_LABEL_FIRST_UNRESERVED; | |
03c05665 | 1993 | |
19d0c341 EB |
1994 | platform_label = rtnl_dereference(net->mpls.platform_label); |
1995 | platform_labels = net->mpls.platform_labels; | |
1996 | for (; index < platform_labels; index++) { | |
03c05665 | 1997 | struct mpls_route *rt; |
19d0c341 | 1998 | rt = rtnl_dereference(platform_label[index]); |
03c05665 EB |
1999 | if (!rt) |
2000 | continue; | |
2001 | ||
2002 | if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid, | |
2003 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, | |
2004 | index, rt, NLM_F_MULTI) < 0) | |
2005 | break; | |
2006 | } | |
2007 | cb->args[0] = index; | |
2008 | ||
2009 | return skb->len; | |
2010 | } | |
2011 | ||
8de147dc EB |
2012 | static inline size_t lfib_nlmsg_size(struct mpls_route *rt) |
2013 | { | |
2014 | size_t payload = | |
2015 | NLMSG_ALIGN(sizeof(struct rtmsg)) | |
5b441ac8 RS |
2016 | + nla_total_size(4) /* RTA_DST */ |
2017 | + nla_total_size(1); /* RTA_TTL_PROPAGATE */ | |
f8efb73c RP |
2018 | |
2019 | if (rt->rt_nhn == 1) { | |
2020 | struct mpls_nh *nh = rt->rt_nh; | |
2021 | ||
2022 | if (nh->nh_dev) | |
2023 | payload += nla_total_size(4); /* RTA_OIF */ | |
eb7809f0 | 2024 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */ |
72dcac96 | 2025 | payload += nla_total_size(2 + nh->nh_via_alen); |
f8efb73c RP |
2026 | if (nh->nh_labels) /* RTA_NEWDST */ |
2027 | payload += nla_total_size(nh->nh_labels * 4); | |
2028 | } else { | |
2029 | /* each nexthop is packed in an attribute */ | |
2030 | size_t nhsize = 0; | |
2031 | ||
2032 | for_nexthops(rt) { | |
e944e97a DA |
2033 | if (!rtnl_dereference(nh->nh_dev)) |
2034 | continue; | |
f8efb73c | 2035 | nhsize += nla_total_size(sizeof(struct rtnexthop)); |
f20367df RS |
2036 | /* RTA_VIA */ |
2037 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) | |
2038 | nhsize += nla_total_size(2 + nh->nh_via_alen); | |
f8efb73c RP |
2039 | if (nh->nh_labels) |
2040 | nhsize += nla_total_size(nh->nh_labels * 4); | |
2041 | } endfor_nexthops(rt); | |
2042 | /* nested attribute */ | |
2043 | payload += nla_total_size(nhsize); | |
2044 | } | |
2045 | ||
8de147dc EB |
2046 | return payload; |
2047 | } | |
2048 | ||
2049 | static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, | |
2050 | struct nlmsghdr *nlh, struct net *net, u32 portid, | |
2051 | unsigned int nlm_flags) | |
2052 | { | |
2053 | struct sk_buff *skb; | |
2054 | u32 seq = nlh ? nlh->nlmsg_seq : 0; | |
2055 | int err = -ENOBUFS; | |
2056 | ||
2057 | skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL); | |
2058 | if (skb == NULL) | |
2059 | goto errout; | |
2060 | ||
2061 | err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags); | |
2062 | if (err < 0) { | |
2063 | /* -EMSGSIZE implies BUG in lfib_nlmsg_size */ | |
2064 | WARN_ON(err == -EMSGSIZE); | |
2065 | kfree_skb(skb); | |
2066 | goto errout; | |
2067 | } | |
2068 | rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL); | |
2069 | ||
2070 | return; | |
2071 | errout: | |
2072 | if (err < 0) | |
2073 | rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err); | |
2074 | } | |
2075 | ||
397fc9e5 RP |
2076 | static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh, |
2077 | struct netlink_ext_ack *extack) | |
2078 | { | |
2079 | struct net *net = sock_net(in_skb->sk); | |
2080 | u32 portid = NETLINK_CB(in_skb).portid; | |
a906c1aa | 2081 | u32 in_label = LABEL_NOT_SPECIFIED; |
397fc9e5 RP |
2082 | struct nlattr *tb[RTA_MAX + 1]; |
2083 | u32 labels[MAX_NEW_LABELS]; | |
2084 | struct mpls_shim_hdr *hdr; | |
2085 | unsigned int hdr_size = 0; | |
2086 | struct net_device *dev; | |
2087 | struct mpls_route *rt; | |
2088 | struct rtmsg *rtm, *r; | |
2089 | struct nlmsghdr *nlh; | |
2090 | struct sk_buff *skb; | |
2091 | struct mpls_nh *nh; | |
397fc9e5 | 2092 | u8 n_labels; |
a906c1aa | 2093 | int err; |
397fc9e5 RP |
2094 | |
2095 | err = nlmsg_parse(in_nlh, sizeof(*rtm), tb, RTA_MAX, | |
ca4a1cd9 | 2096 | rtm_mpls_policy, extack); |
397fc9e5 RP |
2097 | if (err < 0) |
2098 | goto errout; | |
2099 | ||
2100 | rtm = nlmsg_data(in_nlh); | |
2101 | ||
2102 | if (tb[RTA_DST]) { | |
2103 | u8 label_count; | |
2104 | ||
2105 | if (nla_get_labels(tb[RTA_DST], 1, &label_count, | |
a906c1aa RP |
2106 | &in_label, extack)) { |
2107 | err = -EINVAL; | |
397fc9e5 | 2108 | goto errout; |
a906c1aa | 2109 | } |
397fc9e5 | 2110 | |
a906c1aa RP |
2111 | if (!mpls_label_ok(net, in_label, extack)) { |
2112 | err = -EINVAL; | |
397fc9e5 | 2113 | goto errout; |
a906c1aa | 2114 | } |
397fc9e5 RP |
2115 | } |
2116 | ||
2117 | rt = mpls_route_input_rcu(net, in_label); | |
2118 | if (!rt) { | |
2119 | err = -ENETUNREACH; | |
2120 | goto errout; | |
2121 | } | |
2122 | ||
2123 | if (rtm->rtm_flags & RTM_F_FIB_MATCH) { | |
2124 | skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL); | |
2125 | if (!skb) { | |
2126 | err = -ENOBUFS; | |
2127 | goto errout; | |
2128 | } | |
2129 | ||
2130 | err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq, | |
2131 | RTM_NEWROUTE, in_label, rt, 0); | |
2132 | if (err < 0) { | |
2133 | /* -EMSGSIZE implies BUG in lfib_nlmsg_size */ | |
2134 | WARN_ON(err == -EMSGSIZE); | |
2135 | goto errout_free; | |
2136 | } | |
2137 | ||
2138 | return rtnl_unicast(skb, net, portid); | |
2139 | } | |
2140 | ||
2141 | if (tb[RTA_NEWDST]) { | |
2142 | if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels, | |
2143 | labels, extack) != 0) { | |
2144 | err = -EINVAL; | |
2145 | goto errout; | |
2146 | } | |
2147 | ||
2148 | hdr_size = n_labels * sizeof(struct mpls_shim_hdr); | |
2149 | } | |
2150 | ||
2151 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); | |
2152 | if (!skb) { | |
2153 | err = -ENOBUFS; | |
2154 | goto errout; | |
2155 | } | |
2156 | ||
2157 | skb->protocol = htons(ETH_P_MPLS_UC); | |
2158 | ||
2159 | if (hdr_size) { | |
2160 | bool bos; | |
2161 | int i; | |
2162 | ||
2163 | if (skb_cow(skb, hdr_size)) { | |
2164 | err = -ENOBUFS; | |
2165 | goto errout_free; | |
2166 | } | |
2167 | ||
2168 | skb_reserve(skb, hdr_size); | |
2169 | skb_push(skb, hdr_size); | |
2170 | skb_reset_network_header(skb); | |
2171 | ||
2172 | /* Push new labels */ | |
2173 | hdr = mpls_hdr(skb); | |
2174 | bos = true; | |
2175 | for (i = n_labels - 1; i >= 0; i--) { | |
2176 | hdr[i] = mpls_entry_encode(labels[i], | |
2177 | 1, 0, bos); | |
2178 | bos = false; | |
2179 | } | |
2180 | } | |
2181 | ||
2182 | nh = mpls_select_multipath(rt, skb); | |
2183 | if (!nh) { | |
2184 | err = -ENETUNREACH; | |
2185 | goto errout_free; | |
2186 | } | |
2187 | ||
2188 | if (hdr_size) { | |
2189 | skb_pull(skb, hdr_size); | |
2190 | skb_reset_network_header(skb); | |
2191 | } | |
2192 | ||
2193 | nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq, | |
2194 | RTM_NEWROUTE, sizeof(*r), 0); | |
2195 | if (!nlh) { | |
2196 | err = -EMSGSIZE; | |
2197 | goto errout_free; | |
2198 | } | |
2199 | ||
2200 | r = nlmsg_data(nlh); | |
2201 | r->rtm_family = AF_MPLS; | |
2202 | r->rtm_dst_len = 20; | |
2203 | r->rtm_src_len = 0; | |
2204 | r->rtm_table = RT_TABLE_MAIN; | |
2205 | r->rtm_type = RTN_UNICAST; | |
2206 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2207 | r->rtm_protocol = rt->rt_protocol; | |
2208 | r->rtm_flags = 0; | |
2209 | ||
2210 | if (nla_put_labels(skb, RTA_DST, 1, &in_label)) | |
2211 | goto nla_put_failure; | |
2212 | ||
2213 | if (nh->nh_labels && | |
2214 | nla_put_labels(skb, RTA_NEWDST, nh->nh_labels, | |
2215 | nh->nh_label)) | |
2216 | goto nla_put_failure; | |
2217 | ||
2218 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && | |
2219 | nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh), | |
2220 | nh->nh_via_alen)) | |
2221 | goto nla_put_failure; | |
2222 | dev = rtnl_dereference(nh->nh_dev); | |
2223 | if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) | |
2224 | goto nla_put_failure; | |
2225 | ||
2226 | nlmsg_end(skb, nlh); | |
2227 | ||
2228 | err = rtnl_unicast(skb, net, portid); | |
2229 | errout: | |
2230 | return err; | |
2231 | ||
2232 | nla_put_failure: | |
2233 | nlmsg_cancel(skb, nlh); | |
2234 | err = -EMSGSIZE; | |
2235 | errout_free: | |
2236 | kfree_skb(skb); | |
2237 | return err; | |
2238 | } | |
2239 | ||
7720c01f EB |
2240 | static int resize_platform_label_table(struct net *net, size_t limit) |
2241 | { | |
2242 | size_t size = sizeof(struct mpls_route *) * limit; | |
2243 | size_t old_limit; | |
2244 | size_t cp_size; | |
2245 | struct mpls_route __rcu **labels = NULL, **old; | |
2246 | struct mpls_route *rt0 = NULL, *rt2 = NULL; | |
2247 | unsigned index; | |
2248 | ||
2249 | if (size) { | |
752ade68 | 2250 | labels = kvzalloc(size, GFP_KERNEL); |
7720c01f EB |
2251 | if (!labels) |
2252 | goto nolabels; | |
2253 | } | |
2254 | ||
2255 | /* In case the predefined labels need to be populated */ | |
78f5b899 | 2256 | if (limit > MPLS_LABEL_IPV4NULL) { |
7720c01f | 2257 | struct net_device *lo = net->loopback_dev; |
a4ac8c98 | 2258 | rt0 = mpls_rt_alloc(1, lo->addr_len, 0); |
df1c6316 | 2259 | if (IS_ERR(rt0)) |
7720c01f | 2260 | goto nort0; |
f8efb73c | 2261 | RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo); |
7720c01f | 2262 | rt0->rt_protocol = RTPROT_KERNEL; |
118d5234 | 2263 | rt0->rt_payload_type = MPT_IPV4; |
5b441ac8 | 2264 | rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT; |
f8efb73c | 2265 | rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE; |
b4e04fc7 | 2266 | rt0->rt_nh->nh_via_alen = lo->addr_len; |
cf4b24f0 RS |
2267 | memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr, |
2268 | lo->addr_len); | |
7720c01f | 2269 | } |
78f5b899 | 2270 | if (limit > MPLS_LABEL_IPV6NULL) { |
7720c01f | 2271 | struct net_device *lo = net->loopback_dev; |
a4ac8c98 | 2272 | rt2 = mpls_rt_alloc(1, lo->addr_len, 0); |
df1c6316 | 2273 | if (IS_ERR(rt2)) |
7720c01f | 2274 | goto nort2; |
f8efb73c | 2275 | RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo); |
7720c01f | 2276 | rt2->rt_protocol = RTPROT_KERNEL; |
118d5234 | 2277 | rt2->rt_payload_type = MPT_IPV6; |
6a18c312 | 2278 | rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT; |
f8efb73c | 2279 | rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE; |
b4e04fc7 | 2280 | rt2->rt_nh->nh_via_alen = lo->addr_len; |
cf4b24f0 RS |
2281 | memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr, |
2282 | lo->addr_len); | |
7720c01f EB |
2283 | } |
2284 | ||
2285 | rtnl_lock(); | |
2286 | /* Remember the original table */ | |
19d0c341 | 2287 | old = rtnl_dereference(net->mpls.platform_label); |
7720c01f EB |
2288 | old_limit = net->mpls.platform_labels; |
2289 | ||
2290 | /* Free any labels beyond the new table */ | |
2291 | for (index = limit; index < old_limit; index++) | |
f8efb73c | 2292 | mpls_route_update(net, index, NULL, NULL); |
7720c01f EB |
2293 | |
2294 | /* Copy over the old labels */ | |
2295 | cp_size = size; | |
2296 | if (old_limit < limit) | |
2297 | cp_size = old_limit * sizeof(struct mpls_route *); | |
2298 | ||
2299 | memcpy(labels, old, cp_size); | |
2300 | ||
2301 | /* If needed set the predefined labels */ | |
78f5b899 TH |
2302 | if ((old_limit <= MPLS_LABEL_IPV6NULL) && |
2303 | (limit > MPLS_LABEL_IPV6NULL)) { | |
2304 | RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2); | |
7720c01f EB |
2305 | rt2 = NULL; |
2306 | } | |
2307 | ||
78f5b899 TH |
2308 | if ((old_limit <= MPLS_LABEL_IPV4NULL) && |
2309 | (limit > MPLS_LABEL_IPV4NULL)) { | |
2310 | RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0); | |
7720c01f EB |
2311 | rt0 = NULL; |
2312 | } | |
2313 | ||
2314 | /* Update the global pointers */ | |
2315 | net->mpls.platform_labels = limit; | |
19d0c341 | 2316 | rcu_assign_pointer(net->mpls.platform_label, labels); |
7720c01f EB |
2317 | |
2318 | rtnl_unlock(); | |
2319 | ||
2320 | mpls_rt_free(rt2); | |
2321 | mpls_rt_free(rt0); | |
2322 | ||
2323 | if (old) { | |
2324 | synchronize_rcu(); | |
2325 | kvfree(old); | |
2326 | } | |
2327 | return 0; | |
2328 | ||
2329 | nort2: | |
2330 | mpls_rt_free(rt0); | |
2331 | nort0: | |
2332 | kvfree(labels); | |
2333 | nolabels: | |
2334 | return -ENOMEM; | |
2335 | } | |
2336 | ||
2337 | static int mpls_platform_labels(struct ctl_table *table, int write, | |
2338 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2339 | { | |
2340 | struct net *net = table->data; | |
2341 | int platform_labels = net->mpls.platform_labels; | |
2342 | int ret; | |
2343 | struct ctl_table tmp = { | |
2344 | .procname = table->procname, | |
2345 | .data = &platform_labels, | |
2346 | .maxlen = sizeof(int), | |
2347 | .mode = table->mode, | |
2348 | .extra1 = &zero, | |
2349 | .extra2 = &label_limit, | |
2350 | }; | |
2351 | ||
2352 | ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); | |
2353 | ||
2354 | if (write && ret == 0) | |
2355 | ret = resize_platform_label_table(net, platform_labels); | |
2356 | ||
2357 | return ret; | |
2358 | } | |
2359 | ||
5b441ac8 RS |
2360 | #define MPLS_NS_SYSCTL_OFFSET(field) \ |
2361 | (&((struct net *)0)->field) | |
2362 | ||
37bde799 | 2363 | static const struct ctl_table mpls_table[] = { |
7720c01f EB |
2364 | { |
2365 | .procname = "platform_labels", | |
2366 | .data = NULL, | |
2367 | .maxlen = sizeof(int), | |
2368 | .mode = 0644, | |
2369 | .proc_handler = mpls_platform_labels, | |
2370 | }, | |
5b441ac8 RS |
2371 | { |
2372 | .procname = "ip_ttl_propagate", | |
2373 | .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate), | |
2374 | .maxlen = sizeof(int), | |
2375 | .mode = 0644, | |
2376 | .proc_handler = proc_dointvec_minmax, | |
2377 | .extra1 = &zero, | |
2378 | .extra2 = &one, | |
2379 | }, | |
a59166e4 RS |
2380 | { |
2381 | .procname = "default_ttl", | |
2382 | .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl), | |
2383 | .maxlen = sizeof(int), | |
2384 | .mode = 0644, | |
2385 | .proc_handler = proc_dointvec_minmax, | |
2386 | .extra1 = &one, | |
2387 | .extra2 = &ttl_max, | |
2388 | }, | |
7720c01f EB |
2389 | { } |
2390 | }; | |
2391 | ||
0189197f EB |
2392 | static int mpls_net_init(struct net *net) |
2393 | { | |
7720c01f | 2394 | struct ctl_table *table; |
5b441ac8 | 2395 | int i; |
7720c01f | 2396 | |
0189197f EB |
2397 | net->mpls.platform_labels = 0; |
2398 | net->mpls.platform_label = NULL; | |
5b441ac8 | 2399 | net->mpls.ip_ttl_propagate = 1; |
a59166e4 | 2400 | net->mpls.default_ttl = 255; |
0189197f | 2401 | |
7720c01f EB |
2402 | table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL); |
2403 | if (table == NULL) | |
2404 | return -ENOMEM; | |
2405 | ||
5b441ac8 RS |
2406 | /* Table data contains only offsets relative to the base of |
2407 | * the mdev at this point, so make them absolute. | |
2408 | */ | |
2409 | for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++) | |
2410 | table[i].data = (char *)net + (uintptr_t)table[i].data; | |
2411 | ||
7720c01f | 2412 | net->mpls.ctl = register_net_sysctl(net, "net/mpls", table); |
6ea3c9d5 NA |
2413 | if (net->mpls.ctl == NULL) { |
2414 | kfree(table); | |
7720c01f | 2415 | return -ENOMEM; |
6ea3c9d5 | 2416 | } |
7720c01f | 2417 | |
0189197f EB |
2418 | return 0; |
2419 | } | |
2420 | ||
2421 | static void mpls_net_exit(struct net *net) | |
2422 | { | |
19d0c341 EB |
2423 | struct mpls_route __rcu **platform_label; |
2424 | size_t platform_labels; | |
7720c01f | 2425 | struct ctl_table *table; |
0189197f EB |
2426 | unsigned int index; |
2427 | ||
7720c01f EB |
2428 | table = net->mpls.ctl->ctl_table_arg; |
2429 | unregister_net_sysctl_table(net->mpls.ctl); | |
2430 | kfree(table); | |
2431 | ||
19d0c341 EB |
2432 | /* An rcu grace period has passed since there was a device in |
2433 | * the network namespace (and thus the last in flight packet) | |
0189197f EB |
2434 | * left this network namespace. This is because |
2435 | * unregister_netdevice_many and netdev_run_todo has completed | |
2436 | * for each network device that was in this network namespace. | |
2437 | * | |
2438 | * As such no additional rcu synchronization is necessary when | |
2439 | * freeing the platform_label table. | |
2440 | */ | |
2441 | rtnl_lock(); | |
19d0c341 EB |
2442 | platform_label = rtnl_dereference(net->mpls.platform_label); |
2443 | platform_labels = net->mpls.platform_labels; | |
2444 | for (index = 0; index < platform_labels; index++) { | |
2445 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); | |
2446 | RCU_INIT_POINTER(platform_label[index], NULL); | |
e37791ec | 2447 | mpls_notify_route(net, index, rt, NULL, NULL); |
0189197f EB |
2448 | mpls_rt_free(rt); |
2449 | } | |
2450 | rtnl_unlock(); | |
2451 | ||
19d0c341 | 2452 | kvfree(platform_label); |
0189197f EB |
2453 | } |
2454 | ||
2455 | static struct pernet_operations mpls_net_ops = { | |
2456 | .init = mpls_net_init, | |
2457 | .exit = mpls_net_exit, | |
2458 | }; | |
2459 | ||
27d69105 RS |
2460 | static struct rtnl_af_ops mpls_af_ops __read_mostly = { |
2461 | .family = AF_MPLS, | |
2462 | .fill_stats_af = mpls_fill_stats_af, | |
2463 | .get_stats_af_size = mpls_get_stats_af_size, | |
2464 | }; | |
2465 | ||
0189197f EB |
2466 | static int __init mpls_init(void) |
2467 | { | |
2468 | int err; | |
2469 | ||
2470 | BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4); | |
2471 | ||
2472 | err = register_pernet_subsys(&mpls_net_ops); | |
2473 | if (err) | |
2474 | goto out; | |
2475 | ||
2476 | err = register_netdevice_notifier(&mpls_dev_notifier); | |
2477 | if (err) | |
2478 | goto out_unregister_pernet; | |
2479 | ||
2480 | dev_add_pack(&mpls_packet_type); | |
2481 | ||
27d69105 RS |
2482 | rtnl_af_register(&mpls_af_ops); |
2483 | ||
03c05665 EB |
2484 | rtnl_register(PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, NULL); |
2485 | rtnl_register(PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, NULL); | |
397fc9e5 RP |
2486 | rtnl_register(PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes, |
2487 | NULL); | |
24045a03 DA |
2488 | rtnl_register(PF_MPLS, RTM_GETNETCONF, mpls_netconf_get_devconf, |
2489 | mpls_netconf_dump_devconf, NULL); | |
0189197f EB |
2490 | err = 0; |
2491 | out: | |
2492 | return err; | |
2493 | ||
2494 | out_unregister_pernet: | |
2495 | unregister_pernet_subsys(&mpls_net_ops); | |
2496 | goto out; | |
2497 | } | |
2498 | module_init(mpls_init); | |
2499 | ||
2500 | static void __exit mpls_exit(void) | |
2501 | { | |
03c05665 | 2502 | rtnl_unregister_all(PF_MPLS); |
27d69105 | 2503 | rtnl_af_unregister(&mpls_af_ops); |
0189197f EB |
2504 | dev_remove_pack(&mpls_packet_type); |
2505 | unregister_netdevice_notifier(&mpls_dev_notifier); | |
2506 | unregister_pernet_subsys(&mpls_net_ops); | |
2507 | } | |
2508 | module_exit(mpls_exit); | |
2509 | ||
2510 | MODULE_DESCRIPTION("MultiProtocol Label Switching"); | |
2511 | MODULE_LICENSE("GPL v2"); | |
2512 | MODULE_ALIAS_NETPROTO(PF_MPLS); |