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 | ||
1c78efa8 RS |
27 | /* Maximum number of labels to look ahead at when selecting a path of |
28 | * a multipath route | |
29 | */ | |
30 | #define MAX_MP_SELECT_LABELS 4 | |
31 | ||
eb7809f0 RS |
32 | #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1) |
33 | ||
7720c01f | 34 | static int zero = 0; |
5b441ac8 | 35 | static int one = 1; |
7720c01f | 36 | static int label_limit = (1 << 20) - 1; |
a59166e4 | 37 | static int ttl_max = 255; |
7720c01f | 38 | |
8de147dc EB |
39 | static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, |
40 | struct nlmsghdr *nlh, struct net *net, u32 portid, | |
41 | unsigned int nlm_flags); | |
42 | ||
0189197f EB |
43 | static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index) |
44 | { | |
45 | struct mpls_route *rt = NULL; | |
46 | ||
47 | if (index < net->mpls.platform_labels) { | |
48 | struct mpls_route __rcu **platform_label = | |
49 | rcu_dereference(net->mpls.platform_label); | |
50 | rt = rcu_dereference(platform_label[index]); | |
51 | } | |
52 | return rt; | |
53 | } | |
54 | ||
face0188 | 55 | bool mpls_output_possible(const struct net_device *dev) |
0189197f EB |
56 | { |
57 | return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev); | |
58 | } | |
face0188 | 59 | EXPORT_SYMBOL_GPL(mpls_output_possible); |
0189197f | 60 | |
cf4b24f0 RS |
61 | static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh) |
62 | { | |
63 | u8 *nh0_via = PTR_ALIGN((u8 *)&rt->rt_nh[rt->rt_nhn], VIA_ALEN_ALIGN); | |
64 | int nh_index = nh - rt->rt_nh; | |
65 | ||
66 | return nh0_via + rt->rt_max_alen * nh_index; | |
67 | } | |
68 | ||
69 | static const u8 *mpls_nh_via(const struct mpls_route *rt, | |
70 | const struct mpls_nh *nh) | |
71 | { | |
72 | return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh); | |
73 | } | |
74 | ||
f8efb73c | 75 | static unsigned int mpls_nh_header_size(const struct mpls_nh *nh) |
0189197f EB |
76 | { |
77 | /* The size of the layer 2.5 labels to be added for this route */ | |
f8efb73c | 78 | return nh->nh_labels * sizeof(struct mpls_shim_hdr); |
0189197f EB |
79 | } |
80 | ||
face0188 | 81 | unsigned int mpls_dev_mtu(const struct net_device *dev) |
0189197f EB |
82 | { |
83 | /* The amount of data the layer 2 frame can hold */ | |
84 | return dev->mtu; | |
85 | } | |
face0188 | 86 | EXPORT_SYMBOL_GPL(mpls_dev_mtu); |
0189197f | 87 | |
face0188 | 88 | bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu) |
0189197f EB |
89 | { |
90 | if (skb->len <= mtu) | |
91 | return false; | |
92 | ||
ae7ef81e | 93 | if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu)) |
0189197f EB |
94 | return false; |
95 | ||
96 | return true; | |
97 | } | |
face0188 | 98 | EXPORT_SYMBOL_GPL(mpls_pkt_too_big); |
0189197f | 99 | |
27d69105 RS |
100 | void mpls_stats_inc_outucastpkts(struct net_device *dev, |
101 | const struct sk_buff *skb) | |
102 | { | |
103 | struct mpls_dev *mdev; | |
104 | ||
105 | if (skb->protocol == htons(ETH_P_MPLS_UC)) { | |
106 | mdev = mpls_dev_get(dev); | |
107 | if (mdev) | |
108 | MPLS_INC_STATS_LEN(mdev, skb->len, | |
109 | tx_packets, | |
110 | tx_bytes); | |
111 | } else if (skb->protocol == htons(ETH_P_IP)) { | |
112 | IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len); | |
113 | #if IS_ENABLED(CONFIG_IPV6) | |
114 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
115 | struct inet6_dev *in6dev = __in6_dev_get(dev); | |
116 | ||
117 | if (in6dev) | |
118 | IP6_UPD_PO_STATS(dev_net(dev), in6dev, | |
119 | IPSTATS_MIB_OUT, skb->len); | |
120 | #endif | |
121 | } | |
122 | } | |
123 | EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts); | |
124 | ||
9f427a0e | 125 | static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb) |
f8efb73c | 126 | { |
1c78efa8 | 127 | struct mpls_entry_decoded dec; |
9f427a0e | 128 | unsigned int mpls_hdr_len = 0; |
1c78efa8 RS |
129 | struct mpls_shim_hdr *hdr; |
130 | bool eli_seen = false; | |
131 | int label_index; | |
1c78efa8 RS |
132 | u32 hash = 0; |
133 | ||
9f427a0e | 134 | for (label_index = 0; label_index < MAX_MP_SELECT_LABELS; |
1c78efa8 | 135 | label_index++) { |
9f427a0e DA |
136 | mpls_hdr_len += sizeof(*hdr); |
137 | if (!pskb_may_pull(skb, mpls_hdr_len)) | |
1c78efa8 RS |
138 | break; |
139 | ||
140 | /* Read and decode the current label */ | |
141 | hdr = mpls_hdr(skb) + label_index; | |
142 | dec = mpls_entry_decode(hdr); | |
143 | ||
144 | /* RFC6790 - reserved labels MUST NOT be used as keys | |
145 | * for the load-balancing function | |
146 | */ | |
147 | if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) { | |
148 | hash = jhash_1word(dec.label, hash); | |
149 | ||
150 | /* The entropy label follows the entropy label | |
151 | * indicator, so this means that the entropy | |
152 | * label was just added to the hash - no need to | |
153 | * go any deeper either in the label stack or in the | |
154 | * payload | |
155 | */ | |
156 | if (eli_seen) | |
157 | break; | |
158 | } else if (dec.label == MPLS_LABEL_ENTROPY) { | |
159 | eli_seen = true; | |
160 | } | |
161 | ||
9f427a0e DA |
162 | if (!dec.bos) |
163 | continue; | |
164 | ||
165 | /* found bottom label; does skb have room for a header? */ | |
166 | if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) { | |
1c78efa8 RS |
167 | const struct iphdr *v4hdr; |
168 | ||
9f427a0e | 169 | v4hdr = (const struct iphdr *)(hdr + 1); |
1c78efa8 RS |
170 | if (v4hdr->version == 4) { |
171 | hash = jhash_3words(ntohl(v4hdr->saddr), | |
172 | ntohl(v4hdr->daddr), | |
173 | v4hdr->protocol, hash); | |
174 | } else if (v4hdr->version == 6 && | |
9f427a0e DA |
175 | pskb_may_pull(skb, mpls_hdr_len + |
176 | sizeof(struct ipv6hdr))) { | |
1c78efa8 RS |
177 | const struct ipv6hdr *v6hdr; |
178 | ||
9f427a0e | 179 | v6hdr = (const struct ipv6hdr *)(hdr + 1); |
1c78efa8 RS |
180 | hash = __ipv6_addr_jhash(&v6hdr->saddr, hash); |
181 | hash = __ipv6_addr_jhash(&v6hdr->daddr, hash); | |
182 | hash = jhash_1word(v6hdr->nexthdr, hash); | |
183 | } | |
184 | } | |
9f427a0e DA |
185 | |
186 | break; | |
1c78efa8 RS |
187 | } |
188 | ||
c89359a4 RP |
189 | return hash; |
190 | } | |
191 | ||
39eb8cd1 DA |
192 | /* number of alive nexthops (rt->rt_nhn_alive) and the flags for |
193 | * a next hop (nh->nh_flags) are modified by netdev event handlers. | |
194 | * Since those fields can change at any moment, use READ_ONCE to | |
195 | * access both. | |
196 | */ | |
c89359a4 | 197 | static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt, |
9f427a0e | 198 | struct sk_buff *skb) |
c89359a4 | 199 | { |
39eb8cd1 | 200 | unsigned int alive; |
c89359a4 RP |
201 | u32 hash = 0; |
202 | int nh_index = 0; | |
203 | int n = 0; | |
204 | ||
205 | /* No need to look further into packet if there's only | |
206 | * one path | |
207 | */ | |
208 | if (rt->rt_nhn == 1) | |
209 | goto out; | |
210 | ||
39eb8cd1 DA |
211 | alive = READ_ONCE(rt->rt_nhn_alive); |
212 | if (alive == 0) | |
c89359a4 RP |
213 | return NULL; |
214 | ||
9f427a0e | 215 | hash = mpls_multipath_hash(rt, skb); |
c89359a4 RP |
216 | nh_index = hash % alive; |
217 | if (alive == rt->rt_nhn) | |
218 | goto out; | |
219 | for_nexthops(rt) { | |
39eb8cd1 DA |
220 | unsigned int nh_flags = READ_ONCE(nh->nh_flags); |
221 | ||
222 | if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) | |
c89359a4 RP |
223 | continue; |
224 | if (n == nh_index) | |
225 | return nh; | |
226 | n++; | |
227 | } endfor_nexthops(rt); | |
228 | ||
1c78efa8 RS |
229 | out: |
230 | return &rt->rt_nh[nh_index]; | |
f8efb73c RP |
231 | } |
232 | ||
5b441ac8 RS |
233 | static bool mpls_egress(struct net *net, struct mpls_route *rt, |
234 | struct sk_buff *skb, struct mpls_entry_decoded dec) | |
0189197f | 235 | { |
118d5234 RS |
236 | enum mpls_payload_type payload_type; |
237 | bool success = false; | |
0189197f | 238 | |
76fecd82 EB |
239 | /* The IPv4 code below accesses through the IPv4 header |
240 | * checksum, which is 12 bytes into the packet. | |
241 | * The IPv6 code below accesses through the IPv6 hop limit | |
242 | * which is 8 bytes into the packet. | |
243 | * | |
244 | * For all supported cases there should always be at least 12 | |
245 | * bytes of packet data present. The IPv4 header is 20 bytes | |
246 | * without options and the IPv6 header is always 40 bytes | |
247 | * long. | |
248 | */ | |
249 | if (!pskb_may_pull(skb, 12)) | |
250 | return false; | |
251 | ||
118d5234 RS |
252 | payload_type = rt->rt_payload_type; |
253 | if (payload_type == MPT_UNSPEC) | |
254 | payload_type = ip_hdr(skb)->version; | |
255 | ||
256 | switch (payload_type) { | |
257 | case MPT_IPV4: { | |
258 | struct iphdr *hdr4 = ip_hdr(skb); | |
5b441ac8 | 259 | u8 new_ttl; |
0189197f | 260 | skb->protocol = htons(ETH_P_IP); |
5b441ac8 RS |
261 | |
262 | /* If propagating TTL, take the decremented TTL from | |
263 | * the incoming MPLS header, otherwise decrement the | |
264 | * TTL, but only if not 0 to avoid underflow. | |
265 | */ | |
266 | if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED || | |
267 | (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT && | |
268 | net->mpls.ip_ttl_propagate)) | |
269 | new_ttl = dec.ttl; | |
270 | else | |
271 | new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0; | |
272 | ||
0189197f EB |
273 | csum_replace2(&hdr4->check, |
274 | htons(hdr4->ttl << 8), | |
5b441ac8 RS |
275 | htons(new_ttl << 8)); |
276 | hdr4->ttl = new_ttl; | |
118d5234 RS |
277 | success = true; |
278 | break; | |
0189197f | 279 | } |
118d5234 | 280 | case MPT_IPV6: { |
0189197f EB |
281 | struct ipv6hdr *hdr6 = ipv6_hdr(skb); |
282 | skb->protocol = htons(ETH_P_IPV6); | |
5b441ac8 RS |
283 | |
284 | /* If propagating TTL, take the decremented TTL from | |
285 | * the incoming MPLS header, otherwise decrement the | |
286 | * hop limit, but only if not 0 to avoid underflow. | |
287 | */ | |
288 | if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED || | |
289 | (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT && | |
290 | net->mpls.ip_ttl_propagate)) | |
291 | hdr6->hop_limit = dec.ttl; | |
292 | else if (hdr6->hop_limit) | |
293 | hdr6->hop_limit = hdr6->hop_limit - 1; | |
118d5234 RS |
294 | success = true; |
295 | break; | |
0189197f | 296 | } |
118d5234 | 297 | case MPT_UNSPEC: |
5b441ac8 | 298 | /* Should have decided which protocol it is by now */ |
118d5234 RS |
299 | break; |
300 | } | |
301 | ||
0189197f EB |
302 | return success; |
303 | } | |
304 | ||
305 | static int mpls_forward(struct sk_buff *skb, struct net_device *dev, | |
306 | struct packet_type *pt, struct net_device *orig_dev) | |
307 | { | |
308 | struct net *net = dev_net(dev); | |
309 | struct mpls_shim_hdr *hdr; | |
310 | struct mpls_route *rt; | |
f8efb73c | 311 | struct mpls_nh *nh; |
0189197f EB |
312 | struct mpls_entry_decoded dec; |
313 | struct net_device *out_dev; | |
27d69105 | 314 | struct mpls_dev *out_mdev; |
03c57747 | 315 | struct mpls_dev *mdev; |
0189197f EB |
316 | unsigned int hh_len; |
317 | unsigned int new_header_size; | |
318 | unsigned int mtu; | |
319 | int err; | |
320 | ||
321 | /* Careful this entire function runs inside of an rcu critical section */ | |
322 | ||
03c57747 | 323 | mdev = mpls_dev_get(dev); |
27d69105 | 324 | if (!mdev) |
03c57747 RS |
325 | goto drop; |
326 | ||
27d69105 RS |
327 | MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets, |
328 | rx_bytes); | |
329 | ||
330 | if (!mdev->input_enabled) { | |
331 | MPLS_INC_STATS(mdev, rx_dropped); | |
0189197f | 332 | goto drop; |
27d69105 RS |
333 | } |
334 | ||
335 | if (skb->pkt_type != PACKET_HOST) | |
336 | goto err; | |
0189197f EB |
337 | |
338 | if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) | |
27d69105 | 339 | goto err; |
0189197f EB |
340 | |
341 | if (!pskb_may_pull(skb, sizeof(*hdr))) | |
27d69105 | 342 | goto err; |
0189197f EB |
343 | |
344 | /* Read and decode the label */ | |
345 | hdr = mpls_hdr(skb); | |
346 | dec = mpls_entry_decode(hdr); | |
347 | ||
0189197f | 348 | rt = mpls_route_input_rcu(net, dec.label); |
27d69105 RS |
349 | if (!rt) { |
350 | MPLS_INC_STATS(mdev, rx_noroute); | |
0189197f | 351 | goto drop; |
27d69105 | 352 | } |
0189197f | 353 | |
9f427a0e | 354 | nh = mpls_select_multipath(rt, skb); |
f8efb73c | 355 | if (!nh) |
27d69105 | 356 | goto err; |
0189197f | 357 | |
9f427a0e DA |
358 | /* Pop the label */ |
359 | skb_pull(skb, sizeof(*hdr)); | |
360 | skb_reset_network_header(skb); | |
361 | ||
362 | skb_orphan(skb); | |
363 | ||
0189197f | 364 | if (skb_warn_if_lro(skb)) |
27d69105 | 365 | goto err; |
0189197f EB |
366 | |
367 | skb_forward_csum(skb); | |
368 | ||
369 | /* Verify ttl is valid */ | |
aa7da937 | 370 | if (dec.ttl <= 1) |
27d69105 | 371 | goto err; |
0189197f EB |
372 | dec.ttl -= 1; |
373 | ||
27d69105 RS |
374 | /* Find the output device */ |
375 | out_dev = rcu_dereference(nh->nh_dev); | |
376 | if (!mpls_output_possible(out_dev)) | |
377 | goto tx_err; | |
378 | ||
0189197f | 379 | /* Verify the destination can hold the packet */ |
f8efb73c | 380 | new_header_size = mpls_nh_header_size(nh); |
0189197f EB |
381 | mtu = mpls_dev_mtu(out_dev); |
382 | if (mpls_pkt_too_big(skb, mtu - new_header_size)) | |
27d69105 | 383 | goto tx_err; |
0189197f EB |
384 | |
385 | hh_len = LL_RESERVED_SPACE(out_dev); | |
386 | if (!out_dev->header_ops) | |
387 | hh_len = 0; | |
388 | ||
389 | /* Ensure there is enough space for the headers in the skb */ | |
390 | if (skb_cow(skb, hh_len + new_header_size)) | |
27d69105 | 391 | goto tx_err; |
0189197f EB |
392 | |
393 | skb->dev = out_dev; | |
394 | skb->protocol = htons(ETH_P_MPLS_UC); | |
395 | ||
396 | if (unlikely(!new_header_size && dec.bos)) { | |
397 | /* Penultimate hop popping */ | |
5b441ac8 | 398 | if (!mpls_egress(dev_net(out_dev), rt, skb, dec)) |
27d69105 | 399 | goto err; |
0189197f EB |
400 | } else { |
401 | bool bos; | |
402 | int i; | |
403 | skb_push(skb, new_header_size); | |
404 | skb_reset_network_header(skb); | |
405 | /* Push the new labels */ | |
406 | hdr = mpls_hdr(skb); | |
407 | bos = dec.bos; | |
f8efb73c RP |
408 | for (i = nh->nh_labels - 1; i >= 0; i--) { |
409 | hdr[i] = mpls_entry_encode(nh->nh_label[i], | |
410 | dec.ttl, 0, bos); | |
0189197f EB |
411 | bos = false; |
412 | } | |
413 | } | |
414 | ||
27d69105 RS |
415 | mpls_stats_inc_outucastpkts(out_dev, skb); |
416 | ||
eb7809f0 RS |
417 | /* If via wasn't specified then send out using device address */ |
418 | if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC) | |
419 | err = neigh_xmit(NEIGH_LINK_TABLE, out_dev, | |
420 | out_dev->dev_addr, skb); | |
421 | else | |
422 | err = neigh_xmit(nh->nh_via_table, out_dev, | |
423 | mpls_nh_via(rt, nh), skb); | |
0189197f EB |
424 | if (err) |
425 | net_dbg_ratelimited("%s: packet transmission failed: %d\n", | |
426 | __func__, err); | |
427 | return 0; | |
428 | ||
27d69105 RS |
429 | tx_err: |
430 | out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL; | |
431 | if (out_mdev) | |
432 | MPLS_INC_STATS(out_mdev, tx_errors); | |
433 | goto drop; | |
434 | err: | |
435 | MPLS_INC_STATS(mdev, rx_errors); | |
0189197f EB |
436 | drop: |
437 | kfree_skb(skb); | |
438 | return NET_RX_DROP; | |
439 | } | |
440 | ||
441 | static struct packet_type mpls_packet_type __read_mostly = { | |
442 | .type = cpu_to_be16(ETH_P_MPLS_UC), | |
443 | .func = mpls_forward, | |
444 | }; | |
445 | ||
f0126539 | 446 | static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = { |
03c05665 EB |
447 | [RTA_DST] = { .type = NLA_U32 }, |
448 | [RTA_OIF] = { .type = NLA_U32 }, | |
5b441ac8 | 449 | [RTA_TTL_PROPAGATE] = { .type = NLA_U8 }, |
03c05665 EB |
450 | }; |
451 | ||
a2519929 | 452 | struct mpls_route_config { |
118d5234 RS |
453 | u32 rc_protocol; |
454 | u32 rc_ifindex; | |
f8efb73c RP |
455 | u8 rc_via_table; |
456 | u8 rc_via_alen; | |
118d5234 RS |
457 | u8 rc_via[MAX_VIA_ALEN]; |
458 | u32 rc_label; | |
5b441ac8 | 459 | u8 rc_ttl_propagate; |
f8efb73c | 460 | u8 rc_output_labels; |
118d5234 RS |
461 | u32 rc_output_label[MAX_NEW_LABELS]; |
462 | u32 rc_nlflags; | |
463 | enum mpls_payload_type rc_payload_type; | |
464 | struct nl_info rc_nlinfo; | |
f8efb73c RP |
465 | struct rtnexthop *rc_mp; |
466 | int rc_mp_len; | |
a2519929 EB |
467 | }; |
468 | ||
cf4b24f0 | 469 | static struct mpls_route *mpls_rt_alloc(int num_nh, u8 max_alen) |
0189197f | 470 | { |
cf4b24f0 | 471 | u8 max_alen_aligned = ALIGN(max_alen, VIA_ALEN_ALIGN); |
0189197f EB |
472 | struct mpls_route *rt; |
473 | ||
cf4b24f0 RS |
474 | rt = kzalloc(ALIGN(sizeof(*rt) + num_nh * sizeof(*rt->rt_nh), |
475 | VIA_ALEN_ALIGN) + | |
476 | num_nh * max_alen_aligned, | |
f8efb73c | 477 | GFP_KERNEL); |
cf4b24f0 | 478 | if (rt) { |
f8efb73c | 479 | rt->rt_nhn = num_nh; |
c89359a4 | 480 | rt->rt_nhn_alive = num_nh; |
cf4b24f0 RS |
481 | rt->rt_max_alen = max_alen_aligned; |
482 | } | |
f8efb73c | 483 | |
0189197f EB |
484 | return rt; |
485 | } | |
486 | ||
487 | static void mpls_rt_free(struct mpls_route *rt) | |
488 | { | |
489 | if (rt) | |
490 | kfree_rcu(rt, rt_rcu); | |
491 | } | |
492 | ||
8de147dc EB |
493 | static void mpls_notify_route(struct net *net, unsigned index, |
494 | struct mpls_route *old, struct mpls_route *new, | |
495 | const struct nl_info *info) | |
496 | { | |
497 | struct nlmsghdr *nlh = info ? info->nlh : NULL; | |
498 | unsigned portid = info ? info->portid : 0; | |
499 | int event = new ? RTM_NEWROUTE : RTM_DELROUTE; | |
500 | struct mpls_route *rt = new ? new : old; | |
501 | unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0; | |
502 | /* Ignore reserved labels for now */ | |
a6affd24 | 503 | if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED)) |
8de147dc EB |
504 | rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags); |
505 | } | |
506 | ||
0189197f | 507 | static void mpls_route_update(struct net *net, unsigned index, |
f8efb73c | 508 | struct mpls_route *new, |
0189197f EB |
509 | const struct nl_info *info) |
510 | { | |
19d0c341 | 511 | struct mpls_route __rcu **platform_label; |
f8efb73c | 512 | struct mpls_route *rt; |
0189197f EB |
513 | |
514 | ASSERT_RTNL(); | |
515 | ||
19d0c341 EB |
516 | platform_label = rtnl_dereference(net->mpls.platform_label); |
517 | rt = rtnl_dereference(platform_label[index]); | |
f8efb73c | 518 | rcu_assign_pointer(platform_label[index], new); |
0189197f | 519 | |
f8efb73c | 520 | mpls_notify_route(net, index, rt, new, info); |
8de147dc | 521 | |
0189197f | 522 | /* If we removed a route free it now */ |
f8efb73c | 523 | mpls_rt_free(rt); |
0189197f EB |
524 | } |
525 | ||
a2519929 EB |
526 | static unsigned find_free_label(struct net *net) |
527 | { | |
19d0c341 EB |
528 | struct mpls_route __rcu **platform_label; |
529 | size_t platform_labels; | |
a2519929 | 530 | unsigned index; |
19d0c341 EB |
531 | |
532 | platform_label = rtnl_dereference(net->mpls.platform_label); | |
533 | platform_labels = net->mpls.platform_labels; | |
a6affd24 RS |
534 | for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels; |
535 | index++) { | |
19d0c341 | 536 | if (!rtnl_dereference(platform_label[index])) |
a2519929 EB |
537 | return index; |
538 | } | |
539 | return LABEL_NOT_SPECIFIED; | |
540 | } | |
541 | ||
bf21563a | 542 | #if IS_ENABLED(CONFIG_INET) |
cf4b24f0 RS |
543 | static struct net_device *inet_fib_lookup_dev(struct net *net, |
544 | const void *addr) | |
01faef2c | 545 | { |
5a9348b5 | 546 | struct net_device *dev; |
01faef2c RP |
547 | struct rtable *rt; |
548 | struct in_addr daddr; | |
549 | ||
550 | memcpy(&daddr, addr, sizeof(struct in_addr)); | |
551 | rt = ip_route_output(net, daddr.s_addr, 0, 0, 0); | |
552 | if (IS_ERR(rt)) | |
5a9348b5 | 553 | return ERR_CAST(rt); |
01faef2c RP |
554 | |
555 | dev = rt->dst.dev; | |
556 | dev_hold(dev); | |
557 | ||
558 | ip_rt_put(rt); | |
559 | ||
01faef2c | 560 | return dev; |
bf21563a RP |
561 | } |
562 | #else | |
cf4b24f0 RS |
563 | static struct net_device *inet_fib_lookup_dev(struct net *net, |
564 | const void *addr) | |
bf21563a RP |
565 | { |
566 | return ERR_PTR(-EAFNOSUPPORT); | |
01faef2c | 567 | } |
bf21563a | 568 | #endif |
01faef2c | 569 | |
bf21563a | 570 | #if IS_ENABLED(CONFIG_IPV6) |
cf4b24f0 RS |
571 | static struct net_device *inet6_fib_lookup_dev(struct net *net, |
572 | const void *addr) | |
01faef2c | 573 | { |
5a9348b5 | 574 | struct net_device *dev; |
01faef2c RP |
575 | struct dst_entry *dst; |
576 | struct flowi6 fl6; | |
bf21563a RP |
577 | int err; |
578 | ||
579 | if (!ipv6_stub) | |
580 | return ERR_PTR(-EAFNOSUPPORT); | |
01faef2c RP |
581 | |
582 | memset(&fl6, 0, sizeof(fl6)); | |
583 | memcpy(&fl6.daddr, addr, sizeof(struct in6_addr)); | |
bf21563a RP |
584 | err = ipv6_stub->ipv6_dst_lookup(net, NULL, &dst, &fl6); |
585 | if (err) | |
5a9348b5 | 586 | return ERR_PTR(err); |
01faef2c RP |
587 | |
588 | dev = dst->dev; | |
589 | dev_hold(dev); | |
01faef2c RP |
590 | dst_release(dst); |
591 | ||
592 | return dev; | |
593 | } | |
bf21563a | 594 | #else |
cf4b24f0 RS |
595 | static struct net_device *inet6_fib_lookup_dev(struct net *net, |
596 | const void *addr) | |
bf21563a RP |
597 | { |
598 | return ERR_PTR(-EAFNOSUPPORT); | |
599 | } | |
600 | #endif | |
01faef2c RP |
601 | |
602 | static struct net_device *find_outdev(struct net *net, | |
cf4b24f0 | 603 | struct mpls_route *rt, |
f8efb73c | 604 | struct mpls_nh *nh, int oif) |
01faef2c RP |
605 | { |
606 | struct net_device *dev = NULL; | |
607 | ||
f8efb73c RP |
608 | if (!oif) { |
609 | switch (nh->nh_via_table) { | |
01faef2c | 610 | case NEIGH_ARP_TABLE: |
cf4b24f0 | 611 | dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh)); |
01faef2c RP |
612 | break; |
613 | case NEIGH_ND_TABLE: | |
cf4b24f0 | 614 | dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh)); |
01faef2c RP |
615 | break; |
616 | case NEIGH_LINK_TABLE: | |
617 | break; | |
618 | } | |
619 | } else { | |
f8efb73c | 620 | dev = dev_get_by_index(net, oif); |
01faef2c RP |
621 | } |
622 | ||
3dcb615e RP |
623 | if (!dev) |
624 | return ERR_PTR(-ENODEV); | |
625 | ||
94a57f1f RP |
626 | if (IS_ERR(dev)) |
627 | return dev; | |
628 | ||
f8efb73c RP |
629 | /* The caller is holding rtnl anyways, so release the dev reference */ |
630 | dev_put(dev); | |
631 | ||
01faef2c RP |
632 | return dev; |
633 | } | |
634 | ||
cf4b24f0 RS |
635 | static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt, |
636 | struct mpls_nh *nh, int oif) | |
f8efb73c RP |
637 | { |
638 | struct net_device *dev = NULL; | |
639 | int err = -ENODEV; | |
640 | ||
cf4b24f0 | 641 | dev = find_outdev(net, rt, nh, oif); |
f8efb73c RP |
642 | if (IS_ERR(dev)) { |
643 | err = PTR_ERR(dev); | |
644 | dev = NULL; | |
645 | goto errout; | |
646 | } | |
647 | ||
648 | /* Ensure this is a supported device */ | |
649 | err = -EINVAL; | |
650 | if (!mpls_dev_get(dev)) | |
651 | goto errout; | |
652 | ||
a3e948e8 RS |
653 | if ((nh->nh_via_table == NEIGH_LINK_TABLE) && |
654 | (dev->addr_len != nh->nh_via_alen)) | |
655 | goto errout; | |
656 | ||
f8efb73c RP |
657 | RCU_INIT_POINTER(nh->nh_dev, dev); |
658 | ||
c89359a4 RP |
659 | if (!(dev->flags & IFF_UP)) { |
660 | nh->nh_flags |= RTNH_F_DEAD; | |
661 | } else { | |
662 | unsigned int flags; | |
663 | ||
664 | flags = dev_get_flags(dev); | |
665 | if (!(flags & (IFF_RUNNING | IFF_LOWER_UP))) | |
666 | nh->nh_flags |= RTNH_F_LINKDOWN; | |
667 | } | |
668 | ||
f8efb73c RP |
669 | return 0; |
670 | ||
671 | errout: | |
672 | return err; | |
673 | } | |
674 | ||
675 | static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg, | |
676 | struct mpls_route *rt) | |
677 | { | |
678 | struct net *net = cfg->rc_nlinfo.nl_net; | |
679 | struct mpls_nh *nh = rt->rt_nh; | |
680 | int err; | |
681 | int i; | |
682 | ||
683 | if (!nh) | |
684 | return -ENOMEM; | |
685 | ||
686 | err = -EINVAL; | |
687 | /* Ensure only a supported number of labels are present */ | |
688 | if (cfg->rc_output_labels > MAX_NEW_LABELS) | |
689 | goto errout; | |
690 | ||
691 | nh->nh_labels = cfg->rc_output_labels; | |
692 | for (i = 0; i < nh->nh_labels; i++) | |
693 | nh->nh_label[i] = cfg->rc_output_label[i]; | |
694 | ||
695 | nh->nh_via_table = cfg->rc_via_table; | |
cf4b24f0 | 696 | memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen); |
f8efb73c RP |
697 | nh->nh_via_alen = cfg->rc_via_alen; |
698 | ||
cf4b24f0 | 699 | err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex); |
f8efb73c RP |
700 | if (err) |
701 | goto errout; | |
702 | ||
c89359a4 RP |
703 | if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) |
704 | rt->rt_nhn_alive--; | |
705 | ||
f8efb73c RP |
706 | return 0; |
707 | ||
708 | errout: | |
709 | return err; | |
710 | } | |
711 | ||
cf4b24f0 | 712 | static int mpls_nh_build(struct net *net, struct mpls_route *rt, |
c89359a4 RP |
713 | struct mpls_nh *nh, int oif, struct nlattr *via, |
714 | struct nlattr *newdst) | |
f8efb73c RP |
715 | { |
716 | int err = -ENOMEM; | |
717 | ||
718 | if (!nh) | |
719 | goto errout; | |
720 | ||
721 | if (newdst) { | |
722 | err = nla_get_labels(newdst, MAX_NEW_LABELS, | |
723 | &nh->nh_labels, nh->nh_label); | |
724 | if (err) | |
725 | goto errout; | |
726 | } | |
727 | ||
f20367df RS |
728 | if (via) { |
729 | err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table, | |
730 | __mpls_nh_via(rt, nh)); | |
731 | if (err) | |
732 | goto errout; | |
733 | } else { | |
734 | nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC; | |
735 | } | |
f8efb73c | 736 | |
cf4b24f0 | 737 | err = mpls_nh_assign_dev(net, rt, nh, oif); |
f8efb73c RP |
738 | if (err) |
739 | goto errout; | |
740 | ||
741 | return 0; | |
742 | ||
743 | errout: | |
744 | return err; | |
745 | } | |
746 | ||
cf4b24f0 RS |
747 | static int mpls_count_nexthops(struct rtnexthop *rtnh, int len, |
748 | u8 cfg_via_alen, u8 *max_via_alen) | |
f8efb73c RP |
749 | { |
750 | int nhs = 0; | |
751 | int remaining = len; | |
752 | ||
cf4b24f0 RS |
753 | if (!rtnh) { |
754 | *max_via_alen = cfg_via_alen; | |
755 | return 1; | |
756 | } | |
757 | ||
758 | *max_via_alen = 0; | |
759 | ||
f8efb73c | 760 | while (rtnh_ok(rtnh, remaining)) { |
cf4b24f0 RS |
761 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); |
762 | int attrlen; | |
763 | ||
764 | attrlen = rtnh_attrlen(rtnh); | |
765 | nla = nla_find(attrs, attrlen, RTA_VIA); | |
766 | if (nla && nla_len(nla) >= | |
767 | offsetof(struct rtvia, rtvia_addr)) { | |
768 | int via_alen = nla_len(nla) - | |
769 | offsetof(struct rtvia, rtvia_addr); | |
770 | ||
771 | if (via_alen <= MAX_VIA_ALEN) | |
772 | *max_via_alen = max_t(u16, *max_via_alen, | |
773 | via_alen); | |
774 | } | |
775 | ||
f8efb73c RP |
776 | nhs++; |
777 | rtnh = rtnh_next(rtnh, &remaining); | |
778 | } | |
779 | ||
780 | /* leftover implies invalid nexthop configuration, discard it */ | |
781 | return remaining > 0 ? 0 : nhs; | |
782 | } | |
783 | ||
784 | static int mpls_nh_build_multi(struct mpls_route_config *cfg, | |
785 | struct mpls_route *rt) | |
786 | { | |
787 | struct rtnexthop *rtnh = cfg->rc_mp; | |
788 | struct nlattr *nla_via, *nla_newdst; | |
789 | int remaining = cfg->rc_mp_len; | |
790 | int nhs = 0; | |
791 | int err = 0; | |
792 | ||
793 | change_nexthops(rt) { | |
794 | int attrlen; | |
795 | ||
796 | nla_via = NULL; | |
797 | nla_newdst = NULL; | |
798 | ||
799 | err = -EINVAL; | |
800 | if (!rtnh_ok(rtnh, remaining)) | |
801 | goto errout; | |
802 | ||
1c78efa8 RS |
803 | /* neither weighted multipath nor any flags |
804 | * are supported | |
805 | */ | |
806 | if (rtnh->rtnh_hops || rtnh->rtnh_flags) | |
807 | goto errout; | |
808 | ||
f8efb73c RP |
809 | attrlen = rtnh_attrlen(rtnh); |
810 | if (attrlen > 0) { | |
811 | struct nlattr *attrs = rtnh_attrs(rtnh); | |
812 | ||
813 | nla_via = nla_find(attrs, attrlen, RTA_VIA); | |
814 | nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST); | |
815 | } | |
816 | ||
cf4b24f0 | 817 | err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh, |
c89359a4 | 818 | rtnh->rtnh_ifindex, nla_via, nla_newdst); |
f8efb73c RP |
819 | if (err) |
820 | goto errout; | |
821 | ||
c89359a4 RP |
822 | if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) |
823 | rt->rt_nhn_alive--; | |
824 | ||
f8efb73c RP |
825 | rtnh = rtnh_next(rtnh, &remaining); |
826 | nhs++; | |
827 | } endfor_nexthops(rt); | |
828 | ||
829 | rt->rt_nhn = nhs; | |
830 | ||
831 | return 0; | |
832 | ||
833 | errout: | |
834 | return err; | |
835 | } | |
836 | ||
a2519929 EB |
837 | static int mpls_route_add(struct mpls_route_config *cfg) |
838 | { | |
19d0c341 | 839 | struct mpls_route __rcu **platform_label; |
a2519929 | 840 | struct net *net = cfg->rc_nlinfo.nl_net; |
a2519929 | 841 | struct mpls_route *rt, *old; |
a2519929 | 842 | int err = -EINVAL; |
cf4b24f0 | 843 | u8 max_via_alen; |
f8efb73c | 844 | unsigned index; |
cf4b24f0 | 845 | int nhs; |
a2519929 EB |
846 | |
847 | index = cfg->rc_label; | |
848 | ||
849 | /* If a label was not specified during insert pick one */ | |
850 | if ((index == LABEL_NOT_SPECIFIED) && | |
851 | (cfg->rc_nlflags & NLM_F_CREATE)) { | |
852 | index = find_free_label(net); | |
853 | } | |
854 | ||
a6affd24 RS |
855 | /* Reserved labels may not be set */ |
856 | if (index < MPLS_LABEL_FIRST_UNRESERVED) | |
a2519929 EB |
857 | goto errout; |
858 | ||
859 | /* The full 20 bit range may not be supported. */ | |
860 | if (index >= net->mpls.platform_labels) | |
861 | goto errout; | |
862 | ||
a2519929 | 863 | /* Append makes no sense with mpls */ |
0f7bbd58 | 864 | err = -EOPNOTSUPP; |
a2519929 EB |
865 | if (cfg->rc_nlflags & NLM_F_APPEND) |
866 | goto errout; | |
867 | ||
868 | err = -EEXIST; | |
19d0c341 EB |
869 | platform_label = rtnl_dereference(net->mpls.platform_label); |
870 | old = rtnl_dereference(platform_label[index]); | |
a2519929 EB |
871 | if ((cfg->rc_nlflags & NLM_F_EXCL) && old) |
872 | goto errout; | |
873 | ||
874 | err = -EEXIST; | |
875 | if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old) | |
876 | goto errout; | |
877 | ||
878 | err = -ENOENT; | |
879 | if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old) | |
880 | goto errout; | |
881 | ||
cf4b24f0 RS |
882 | err = -EINVAL; |
883 | nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len, | |
884 | cfg->rc_via_alen, &max_via_alen); | |
885 | if (nhs == 0) | |
886 | goto errout; | |
f8efb73c | 887 | |
a2519929 | 888 | err = -ENOMEM; |
cf4b24f0 | 889 | rt = mpls_rt_alloc(nhs, max_via_alen); |
a2519929 EB |
890 | if (!rt) |
891 | goto errout; | |
892 | ||
a2519929 | 893 | rt->rt_protocol = cfg->rc_protocol; |
118d5234 | 894 | rt->rt_payload_type = cfg->rc_payload_type; |
5b441ac8 | 895 | rt->rt_ttl_propagate = cfg->rc_ttl_propagate; |
a2519929 | 896 | |
f8efb73c RP |
897 | if (cfg->rc_mp) |
898 | err = mpls_nh_build_multi(cfg, rt); | |
899 | else | |
900 | err = mpls_nh_build_from_cfg(cfg, rt); | |
901 | if (err) | |
902 | goto freert; | |
903 | ||
904 | mpls_route_update(net, index, rt, &cfg->rc_nlinfo); | |
a2519929 | 905 | |
a2519929 EB |
906 | return 0; |
907 | ||
f8efb73c RP |
908 | freert: |
909 | mpls_rt_free(rt); | |
a2519929 | 910 | errout: |
a2519929 EB |
911 | return err; |
912 | } | |
913 | ||
914 | static int mpls_route_del(struct mpls_route_config *cfg) | |
915 | { | |
916 | struct net *net = cfg->rc_nlinfo.nl_net; | |
917 | unsigned index; | |
918 | int err = -EINVAL; | |
919 | ||
920 | index = cfg->rc_label; | |
921 | ||
a6affd24 RS |
922 | /* Reserved labels may not be removed */ |
923 | if (index < MPLS_LABEL_FIRST_UNRESERVED) | |
a2519929 EB |
924 | goto errout; |
925 | ||
926 | /* The full 20 bit range may not be supported */ | |
927 | if (index >= net->mpls.platform_labels) | |
928 | goto errout; | |
929 | ||
f8efb73c | 930 | mpls_route_update(net, index, NULL, &cfg->rc_nlinfo); |
a2519929 EB |
931 | |
932 | err = 0; | |
933 | errout: | |
934 | return err; | |
935 | } | |
936 | ||
27d69105 RS |
937 | static void mpls_get_stats(struct mpls_dev *mdev, |
938 | struct mpls_link_stats *stats) | |
939 | { | |
940 | struct mpls_pcpu_stats *p; | |
941 | int i; | |
942 | ||
943 | memset(stats, 0, sizeof(*stats)); | |
944 | ||
945 | for_each_possible_cpu(i) { | |
946 | struct mpls_link_stats local; | |
947 | unsigned int start; | |
948 | ||
949 | p = per_cpu_ptr(mdev->stats, i); | |
950 | do { | |
951 | start = u64_stats_fetch_begin(&p->syncp); | |
952 | local = p->stats; | |
953 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
954 | ||
955 | stats->rx_packets += local.rx_packets; | |
956 | stats->rx_bytes += local.rx_bytes; | |
957 | stats->tx_packets += local.tx_packets; | |
958 | stats->tx_bytes += local.tx_bytes; | |
959 | stats->rx_errors += local.rx_errors; | |
960 | stats->tx_errors += local.tx_errors; | |
961 | stats->rx_dropped += local.rx_dropped; | |
962 | stats->tx_dropped += local.tx_dropped; | |
963 | stats->rx_noroute += local.rx_noroute; | |
964 | } | |
965 | } | |
966 | ||
967 | static int mpls_fill_stats_af(struct sk_buff *skb, | |
968 | const struct net_device *dev) | |
969 | { | |
970 | struct mpls_link_stats *stats; | |
971 | struct mpls_dev *mdev; | |
972 | struct nlattr *nla; | |
973 | ||
974 | mdev = mpls_dev_get(dev); | |
975 | if (!mdev) | |
976 | return -ENODATA; | |
977 | ||
978 | nla = nla_reserve_64bit(skb, MPLS_STATS_LINK, | |
979 | sizeof(struct mpls_link_stats), | |
980 | MPLS_STATS_UNSPEC); | |
981 | if (!nla) | |
982 | return -EMSGSIZE; | |
983 | ||
984 | stats = nla_data(nla); | |
985 | mpls_get_stats(mdev, stats); | |
986 | ||
987 | return 0; | |
988 | } | |
989 | ||
990 | static size_t mpls_get_stats_af_size(const struct net_device *dev) | |
991 | { | |
992 | struct mpls_dev *mdev; | |
993 | ||
994 | mdev = mpls_dev_get(dev); | |
995 | if (!mdev) | |
996 | return 0; | |
997 | ||
998 | return nla_total_size_64bit(sizeof(struct mpls_link_stats)); | |
999 | } | |
1000 | ||
24045a03 DA |
1001 | static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev, |
1002 | u32 portid, u32 seq, int event, | |
1003 | unsigned int flags, int type) | |
1004 | { | |
1005 | struct nlmsghdr *nlh; | |
1006 | struct netconfmsg *ncm; | |
1007 | bool all = false; | |
1008 | ||
1009 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), | |
1010 | flags); | |
1011 | if (!nlh) | |
1012 | return -EMSGSIZE; | |
1013 | ||
1014 | if (type == NETCONFA_ALL) | |
1015 | all = true; | |
1016 | ||
1017 | ncm = nlmsg_data(nlh); | |
1018 | ncm->ncm_family = AF_MPLS; | |
1019 | ||
1020 | if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0) | |
1021 | goto nla_put_failure; | |
1022 | ||
1023 | if ((all || type == NETCONFA_INPUT) && | |
1024 | nla_put_s32(skb, NETCONFA_INPUT, | |
1025 | mdev->input_enabled) < 0) | |
1026 | goto nla_put_failure; | |
1027 | ||
1028 | nlmsg_end(skb, nlh); | |
1029 | return 0; | |
1030 | ||
1031 | nla_put_failure: | |
1032 | nlmsg_cancel(skb, nlh); | |
1033 | return -EMSGSIZE; | |
1034 | } | |
1035 | ||
1036 | static int mpls_netconf_msgsize_devconf(int type) | |
1037 | { | |
1038 | int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) | |
1039 | + nla_total_size(4); /* NETCONFA_IFINDEX */ | |
1040 | bool all = false; | |
1041 | ||
1042 | if (type == NETCONFA_ALL) | |
1043 | all = true; | |
1044 | ||
1045 | if (all || type == NETCONFA_INPUT) | |
1046 | size += nla_total_size(4); | |
1047 | ||
1048 | return size; | |
1049 | } | |
1050 | ||
823566ae DA |
1051 | static void mpls_netconf_notify_devconf(struct net *net, int event, |
1052 | int type, struct mpls_dev *mdev) | |
24045a03 DA |
1053 | { |
1054 | struct sk_buff *skb; | |
1055 | int err = -ENOBUFS; | |
1056 | ||
1057 | skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL); | |
1058 | if (!skb) | |
1059 | goto errout; | |
1060 | ||
823566ae | 1061 | err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type); |
24045a03 DA |
1062 | if (err < 0) { |
1063 | /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */ | |
1064 | WARN_ON(err == -EMSGSIZE); | |
1065 | kfree_skb(skb); | |
1066 | goto errout; | |
1067 | } | |
1068 | ||
1069 | rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL); | |
1070 | return; | |
1071 | errout: | |
1072 | if (err < 0) | |
1073 | rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err); | |
1074 | } | |
1075 | ||
1076 | static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = { | |
1077 | [NETCONFA_IFINDEX] = { .len = sizeof(int) }, | |
1078 | }; | |
1079 | ||
1080 | static int mpls_netconf_get_devconf(struct sk_buff *in_skb, | |
1081 | struct nlmsghdr *nlh) | |
1082 | { | |
1083 | struct net *net = sock_net(in_skb->sk); | |
1084 | struct nlattr *tb[NETCONFA_MAX + 1]; | |
1085 | struct netconfmsg *ncm; | |
1086 | struct net_device *dev; | |
1087 | struct mpls_dev *mdev; | |
1088 | struct sk_buff *skb; | |
1089 | int ifindex; | |
1090 | int err; | |
1091 | ||
1092 | err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX, | |
1093 | devconf_mpls_policy); | |
1094 | if (err < 0) | |
1095 | goto errout; | |
1096 | ||
1097 | err = -EINVAL; | |
1098 | if (!tb[NETCONFA_IFINDEX]) | |
1099 | goto errout; | |
1100 | ||
1101 | ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); | |
1102 | dev = __dev_get_by_index(net, ifindex); | |
1103 | if (!dev) | |
1104 | goto errout; | |
1105 | ||
1106 | mdev = mpls_dev_get(dev); | |
1107 | if (!mdev) | |
1108 | goto errout; | |
1109 | ||
1110 | err = -ENOBUFS; | |
1111 | skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); | |
1112 | if (!skb) | |
1113 | goto errout; | |
1114 | ||
1115 | err = mpls_netconf_fill_devconf(skb, mdev, | |
1116 | NETLINK_CB(in_skb).portid, | |
1117 | nlh->nlmsg_seq, RTM_NEWNETCONF, 0, | |
1118 | NETCONFA_ALL); | |
1119 | if (err < 0) { | |
1120 | /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */ | |
1121 | WARN_ON(err == -EMSGSIZE); | |
1122 | kfree_skb(skb); | |
1123 | goto errout; | |
1124 | } | |
1125 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); | |
1126 | errout: | |
1127 | return err; | |
1128 | } | |
1129 | ||
1130 | static int mpls_netconf_dump_devconf(struct sk_buff *skb, | |
1131 | struct netlink_callback *cb) | |
1132 | { | |
1133 | struct net *net = sock_net(skb->sk); | |
1134 | struct hlist_head *head; | |
1135 | struct net_device *dev; | |
1136 | struct mpls_dev *mdev; | |
1137 | int idx, s_idx; | |
1138 | int h, s_h; | |
1139 | ||
1140 | s_h = cb->args[0]; | |
1141 | s_idx = idx = cb->args[1]; | |
1142 | ||
1143 | for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { | |
1144 | idx = 0; | |
1145 | head = &net->dev_index_head[h]; | |
1146 | rcu_read_lock(); | |
1147 | cb->seq = net->dev_base_seq; | |
1148 | hlist_for_each_entry_rcu(dev, head, index_hlist) { | |
1149 | if (idx < s_idx) | |
1150 | goto cont; | |
1151 | mdev = mpls_dev_get(dev); | |
1152 | if (!mdev) | |
1153 | goto cont; | |
1154 | if (mpls_netconf_fill_devconf(skb, mdev, | |
1155 | NETLINK_CB(cb->skb).portid, | |
1156 | cb->nlh->nlmsg_seq, | |
1157 | RTM_NEWNETCONF, | |
1158 | NLM_F_MULTI, | |
1159 | NETCONFA_ALL) < 0) { | |
1160 | rcu_read_unlock(); | |
1161 | goto done; | |
1162 | } | |
1163 | nl_dump_check_consistent(cb, nlmsg_hdr(skb)); | |
1164 | cont: | |
1165 | idx++; | |
1166 | } | |
1167 | rcu_read_unlock(); | |
1168 | } | |
1169 | done: | |
1170 | cb->args[0] = h; | |
1171 | cb->args[1] = idx; | |
1172 | ||
1173 | return skb->len; | |
1174 | } | |
1175 | ||
37bde799 RS |
1176 | #define MPLS_PERDEV_SYSCTL_OFFSET(field) \ |
1177 | (&((struct mpls_dev *)0)->field) | |
1178 | ||
24045a03 DA |
1179 | static int mpls_conf_proc(struct ctl_table *ctl, int write, |
1180 | void __user *buffer, | |
1181 | size_t *lenp, loff_t *ppos) | |
1182 | { | |
1183 | int oval = *(int *)ctl->data; | |
1184 | int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); | |
1185 | ||
1186 | if (write) { | |
1187 | struct mpls_dev *mdev = ctl->extra1; | |
1188 | int i = (int *)ctl->data - (int *)mdev; | |
1189 | struct net *net = ctl->extra2; | |
1190 | int val = *(int *)ctl->data; | |
1191 | ||
1192 | if (i == offsetof(struct mpls_dev, input_enabled) && | |
1193 | val != oval) { | |
823566ae DA |
1194 | mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, |
1195 | NETCONFA_INPUT, mdev); | |
24045a03 DA |
1196 | } |
1197 | } | |
1198 | ||
1199 | return ret; | |
1200 | } | |
1201 | ||
37bde799 RS |
1202 | static const struct ctl_table mpls_dev_table[] = { |
1203 | { | |
1204 | .procname = "input", | |
1205 | .maxlen = sizeof(int), | |
1206 | .mode = 0644, | |
24045a03 | 1207 | .proc_handler = mpls_conf_proc, |
37bde799 RS |
1208 | .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled), |
1209 | }, | |
1210 | { } | |
1211 | }; | |
1212 | ||
1213 | static int mpls_dev_sysctl_register(struct net_device *dev, | |
1214 | struct mpls_dev *mdev) | |
1215 | { | |
1216 | char path[sizeof("net/mpls/conf/") + IFNAMSIZ]; | |
24045a03 | 1217 | struct net *net = dev_net(dev); |
37bde799 RS |
1218 | struct ctl_table *table; |
1219 | int i; | |
1220 | ||
1221 | table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL); | |
1222 | if (!table) | |
1223 | goto out; | |
1224 | ||
1225 | /* Table data contains only offsets relative to the base of | |
1226 | * the mdev at this point, so make them absolute. | |
1227 | */ | |
24045a03 | 1228 | for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) { |
37bde799 | 1229 | table[i].data = (char *)mdev + (uintptr_t)table[i].data; |
24045a03 DA |
1230 | table[i].extra1 = mdev; |
1231 | table[i].extra2 = net; | |
1232 | } | |
37bde799 RS |
1233 | |
1234 | snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name); | |
1235 | ||
1182e4d0 | 1236 | mdev->sysctl = register_net_sysctl(net, path, table); |
37bde799 RS |
1237 | if (!mdev->sysctl) |
1238 | goto free; | |
1239 | ||
1182e4d0 | 1240 | mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev); |
37bde799 RS |
1241 | return 0; |
1242 | ||
1243 | free: | |
1244 | kfree(table); | |
1245 | out: | |
1246 | return -ENOBUFS; | |
1247 | } | |
1248 | ||
1182e4d0 DA |
1249 | static void mpls_dev_sysctl_unregister(struct net_device *dev, |
1250 | struct mpls_dev *mdev) | |
37bde799 | 1251 | { |
1182e4d0 | 1252 | struct net *net = dev_net(dev); |
37bde799 RS |
1253 | struct ctl_table *table; |
1254 | ||
1255 | table = mdev->sysctl->ctl_table_arg; | |
1256 | unregister_net_sysctl_table(mdev->sysctl); | |
1257 | kfree(table); | |
1182e4d0 DA |
1258 | |
1259 | mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev); | |
37bde799 RS |
1260 | } |
1261 | ||
03c57747 RS |
1262 | static struct mpls_dev *mpls_add_dev(struct net_device *dev) |
1263 | { | |
1264 | struct mpls_dev *mdev; | |
1265 | int err = -ENOMEM; | |
27d69105 | 1266 | int i; |
03c57747 RS |
1267 | |
1268 | ASSERT_RTNL(); | |
1269 | ||
1270 | mdev = kzalloc(sizeof(*mdev), GFP_KERNEL); | |
1271 | if (!mdev) | |
1272 | return ERR_PTR(err); | |
1273 | ||
27d69105 RS |
1274 | mdev->stats = alloc_percpu(struct mpls_pcpu_stats); |
1275 | if (!mdev->stats) | |
1276 | goto free; | |
1277 | ||
1278 | for_each_possible_cpu(i) { | |
1279 | struct mpls_pcpu_stats *mpls_stats; | |
1280 | ||
1281 | mpls_stats = per_cpu_ptr(mdev->stats, i); | |
1282 | u64_stats_init(&mpls_stats->syncp); | |
1283 | } | |
1284 | ||
1182e4d0 DA |
1285 | mdev->dev = dev; |
1286 | ||
37bde799 RS |
1287 | err = mpls_dev_sysctl_register(dev, mdev); |
1288 | if (err) | |
1289 | goto free; | |
1290 | ||
03c57747 RS |
1291 | rcu_assign_pointer(dev->mpls_ptr, mdev); |
1292 | ||
1293 | return mdev; | |
37bde799 RS |
1294 | |
1295 | free: | |
27d69105 | 1296 | free_percpu(mdev->stats); |
37bde799 RS |
1297 | kfree(mdev); |
1298 | return ERR_PTR(err); | |
03c57747 RS |
1299 | } |
1300 | ||
27d69105 RS |
1301 | static void mpls_dev_destroy_rcu(struct rcu_head *head) |
1302 | { | |
1303 | struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu); | |
1304 | ||
1305 | free_percpu(mdev->stats); | |
1306 | kfree(mdev); | |
1307 | } | |
1308 | ||
c89359a4 | 1309 | static void mpls_ifdown(struct net_device *dev, int event) |
0189197f | 1310 | { |
19d0c341 | 1311 | struct mpls_route __rcu **platform_label; |
0189197f | 1312 | struct net *net = dev_net(dev); |
4ea8efad | 1313 | unsigned int alive, deleted; |
0189197f EB |
1314 | unsigned index; |
1315 | ||
19d0c341 | 1316 | platform_label = rtnl_dereference(net->mpls.platform_label); |
0189197f | 1317 | for (index = 0; index < net->mpls.platform_labels; index++) { |
19d0c341 | 1318 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); |
c89359a4 | 1319 | |
0189197f EB |
1320 | if (!rt) |
1321 | continue; | |
c89359a4 | 1322 | |
61733c91 | 1323 | alive = 0; |
4ea8efad | 1324 | deleted = 0; |
c89359a4 | 1325 | change_nexthops(rt) { |
39eb8cd1 DA |
1326 | unsigned int nh_flags = nh->nh_flags; |
1327 | ||
f8efb73c | 1328 | if (rtnl_dereference(nh->nh_dev) != dev) |
61733c91 DA |
1329 | goto next; |
1330 | ||
c89359a4 RP |
1331 | switch (event) { |
1332 | case NETDEV_DOWN: | |
1333 | case NETDEV_UNREGISTER: | |
39eb8cd1 | 1334 | nh_flags |= RTNH_F_DEAD; |
c89359a4 RP |
1335 | /* fall through */ |
1336 | case NETDEV_CHANGE: | |
39eb8cd1 | 1337 | nh_flags |= RTNH_F_LINKDOWN; |
c89359a4 RP |
1338 | break; |
1339 | } | |
1340 | if (event == NETDEV_UNREGISTER) | |
1341 | RCU_INIT_POINTER(nh->nh_dev, NULL); | |
39eb8cd1 DA |
1342 | |
1343 | if (nh->nh_flags != nh_flags) | |
1344 | WRITE_ONCE(nh->nh_flags, nh_flags); | |
61733c91 | 1345 | next: |
39eb8cd1 | 1346 | if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))) |
61733c91 | 1347 | alive++; |
4ea8efad DA |
1348 | if (!rtnl_dereference(nh->nh_dev)) |
1349 | deleted++; | |
f8efb73c | 1350 | } endfor_nexthops(rt); |
61733c91 DA |
1351 | |
1352 | WRITE_ONCE(rt->rt_nhn_alive, alive); | |
4ea8efad DA |
1353 | |
1354 | /* if there are no more nexthops, delete the route */ | |
1355 | if (event == NETDEV_UNREGISTER && deleted == rt->rt_nhn) | |
1356 | mpls_route_update(net, index, NULL, NULL); | |
0189197f | 1357 | } |
c89359a4 RP |
1358 | } |
1359 | ||
39eb8cd1 | 1360 | static void mpls_ifup(struct net_device *dev, unsigned int flags) |
c89359a4 RP |
1361 | { |
1362 | struct mpls_route __rcu **platform_label; | |
1363 | struct net *net = dev_net(dev); | |
1364 | unsigned index; | |
1365 | int alive; | |
1366 | ||
1367 | platform_label = rtnl_dereference(net->mpls.platform_label); | |
1368 | for (index = 0; index < net->mpls.platform_labels; index++) { | |
1369 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); | |
1370 | ||
1371 | if (!rt) | |
1372 | continue; | |
37bde799 | 1373 | |
c89359a4 RP |
1374 | alive = 0; |
1375 | change_nexthops(rt) { | |
39eb8cd1 | 1376 | unsigned int nh_flags = nh->nh_flags; |
c89359a4 RP |
1377 | struct net_device *nh_dev = |
1378 | rtnl_dereference(nh->nh_dev); | |
03c57747 | 1379 | |
39eb8cd1 | 1380 | if (!(nh_flags & flags)) { |
c89359a4 RP |
1381 | alive++; |
1382 | continue; | |
1383 | } | |
1384 | if (nh_dev != dev) | |
1385 | continue; | |
1386 | alive++; | |
39eb8cd1 DA |
1387 | nh_flags &= ~flags; |
1388 | WRITE_ONCE(nh->nh_flags, flags); | |
c89359a4 RP |
1389 | } endfor_nexthops(rt); |
1390 | ||
39eb8cd1 | 1391 | WRITE_ONCE(rt->rt_nhn_alive, alive); |
c89359a4 | 1392 | } |
0189197f EB |
1393 | } |
1394 | ||
1395 | static int mpls_dev_notify(struct notifier_block *this, unsigned long event, | |
1396 | void *ptr) | |
1397 | { | |
1398 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
03c57747 | 1399 | struct mpls_dev *mdev; |
c89359a4 | 1400 | unsigned int flags; |
0189197f | 1401 | |
c89359a4 | 1402 | if (event == NETDEV_REGISTER) { |
407f31be | 1403 | /* For now just support Ethernet, IPGRE, SIT and IPIP devices */ |
0d227a86 SH |
1404 | if (dev->type == ARPHRD_ETHER || |
1405 | dev->type == ARPHRD_LOOPBACK || | |
407f31be SH |
1406 | dev->type == ARPHRD_IPGRE || |
1407 | dev->type == ARPHRD_SIT || | |
1408 | dev->type == ARPHRD_TUNNEL) { | |
03c57747 RS |
1409 | mdev = mpls_add_dev(dev); |
1410 | if (IS_ERR(mdev)) | |
1411 | return notifier_from_errno(PTR_ERR(mdev)); | |
1412 | } | |
c89359a4 RP |
1413 | return NOTIFY_OK; |
1414 | } | |
03c57747 | 1415 | |
c89359a4 RP |
1416 | mdev = mpls_dev_get(dev); |
1417 | if (!mdev) | |
1418 | return NOTIFY_OK; | |
1419 | ||
1420 | switch (event) { | |
1421 | case NETDEV_DOWN: | |
1422 | mpls_ifdown(dev, event); | |
1423 | break; | |
1424 | case NETDEV_UP: | |
1425 | flags = dev_get_flags(dev); | |
1426 | if (flags & (IFF_RUNNING | IFF_LOWER_UP)) | |
1427 | mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN); | |
1428 | else | |
1429 | mpls_ifup(dev, RTNH_F_DEAD); | |
1430 | break; | |
1431 | case NETDEV_CHANGE: | |
1432 | flags = dev_get_flags(dev); | |
1433 | if (flags & (IFF_RUNNING | IFF_LOWER_UP)) | |
1434 | mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN); | |
1435 | else | |
1436 | mpls_ifdown(dev, event); | |
1437 | break; | |
0189197f | 1438 | case NETDEV_UNREGISTER: |
c89359a4 RP |
1439 | mpls_ifdown(dev, event); |
1440 | mdev = mpls_dev_get(dev); | |
1441 | if (mdev) { | |
1182e4d0 | 1442 | mpls_dev_sysctl_unregister(dev, mdev); |
c89359a4 | 1443 | RCU_INIT_POINTER(dev->mpls_ptr, NULL); |
27d69105 | 1444 | call_rcu(&mdev->rcu, mpls_dev_destroy_rcu); |
c89359a4 | 1445 | } |
0189197f | 1446 | break; |
0fae3bf0 RS |
1447 | case NETDEV_CHANGENAME: |
1448 | mdev = mpls_dev_get(dev); | |
1449 | if (mdev) { | |
1450 | int err; | |
1451 | ||
1182e4d0 | 1452 | mpls_dev_sysctl_unregister(dev, mdev); |
0fae3bf0 RS |
1453 | err = mpls_dev_sysctl_register(dev, mdev); |
1454 | if (err) | |
1455 | return notifier_from_errno(err); | |
1456 | } | |
1457 | break; | |
0189197f EB |
1458 | } |
1459 | return NOTIFY_OK; | |
1460 | } | |
1461 | ||
1462 | static struct notifier_block mpls_dev_notifier = { | |
1463 | .notifier_call = mpls_dev_notify, | |
1464 | }; | |
1465 | ||
03c05665 | 1466 | static int nla_put_via(struct sk_buff *skb, |
b79bda3d | 1467 | u8 table, const void *addr, int alen) |
03c05665 | 1468 | { |
b79bda3d EB |
1469 | static const int table_to_family[NEIGH_NR_TABLES + 1] = { |
1470 | AF_INET, AF_INET6, AF_DECnet, AF_PACKET, | |
1471 | }; | |
03c05665 EB |
1472 | struct nlattr *nla; |
1473 | struct rtvia *via; | |
b79bda3d | 1474 | int family = AF_UNSPEC; |
03c05665 EB |
1475 | |
1476 | nla = nla_reserve(skb, RTA_VIA, alen + 2); | |
1477 | if (!nla) | |
1478 | return -EMSGSIZE; | |
1479 | ||
b79bda3d EB |
1480 | if (table <= NEIGH_NR_TABLES) |
1481 | family = table_to_family[table]; | |
1482 | ||
03c05665 EB |
1483 | via = nla_data(nla); |
1484 | via->rtvia_family = family; | |
1485 | memcpy(via->rtvia_addr, addr, alen); | |
1486 | return 0; | |
1487 | } | |
1488 | ||
966bae33 EB |
1489 | int nla_put_labels(struct sk_buff *skb, int attrtype, |
1490 | u8 labels, const u32 label[]) | |
1491 | { | |
1492 | struct nlattr *nla; | |
1493 | struct mpls_shim_hdr *nla_label; | |
1494 | bool bos; | |
1495 | int i; | |
1496 | nla = nla_reserve(skb, attrtype, labels*4); | |
1497 | if (!nla) | |
1498 | return -EMSGSIZE; | |
1499 | ||
1500 | nla_label = nla_data(nla); | |
1501 | bos = true; | |
1502 | for (i = labels - 1; i >= 0; i--) { | |
1503 | nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos); | |
1504 | bos = false; | |
1505 | } | |
1506 | ||
1507 | return 0; | |
1508 | } | |
face0188 | 1509 | EXPORT_SYMBOL_GPL(nla_put_labels); |
966bae33 EB |
1510 | |
1511 | int nla_get_labels(const struct nlattr *nla, | |
f8efb73c | 1512 | u32 max_labels, u8 *labels, u32 label[]) |
966bae33 EB |
1513 | { |
1514 | unsigned len = nla_len(nla); | |
1515 | unsigned nla_labels; | |
1516 | struct mpls_shim_hdr *nla_label; | |
1517 | bool bos; | |
1518 | int i; | |
1519 | ||
1520 | /* len needs to be an even multiple of 4 (the label size) */ | |
1521 | if (len & 3) | |
1522 | return -EINVAL; | |
1523 | ||
1524 | /* Limit the number of new labels allowed */ | |
1525 | nla_labels = len/4; | |
1526 | if (nla_labels > max_labels) | |
1527 | return -EINVAL; | |
1528 | ||
1529 | nla_label = nla_data(nla); | |
1530 | bos = true; | |
1531 | for (i = nla_labels - 1; i >= 0; i--, bos = false) { | |
1532 | struct mpls_entry_decoded dec; | |
1533 | dec = mpls_entry_decode(nla_label + i); | |
1534 | ||
1535 | /* Ensure the bottom of stack flag is properly set | |
1536 | * and ttl and tc are both clear. | |
1537 | */ | |
1538 | if ((dec.bos != bos) || dec.ttl || dec.tc) | |
1539 | return -EINVAL; | |
1540 | ||
5a9ab017 | 1541 | switch (dec.label) { |
78f5b899 | 1542 | case MPLS_LABEL_IMPLNULL: |
5a9ab017 RS |
1543 | /* RFC3032: This is a label that an LSR may |
1544 | * assign and distribute, but which never | |
1545 | * actually appears in the encapsulation. | |
1546 | */ | |
1547 | return -EINVAL; | |
1548 | } | |
1549 | ||
966bae33 EB |
1550 | label[i] = dec.label; |
1551 | } | |
1552 | *labels = nla_labels; | |
1553 | return 0; | |
1554 | } | |
face0188 | 1555 | EXPORT_SYMBOL_GPL(nla_get_labels); |
966bae33 | 1556 | |
f8efb73c RP |
1557 | int nla_get_via(const struct nlattr *nla, u8 *via_alen, |
1558 | u8 *via_table, u8 via_addr[]) | |
1559 | { | |
1560 | struct rtvia *via = nla_data(nla); | |
1561 | int err = -EINVAL; | |
1562 | int alen; | |
1563 | ||
1564 | if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) | |
1565 | goto errout; | |
1566 | alen = nla_len(nla) - | |
1567 | offsetof(struct rtvia, rtvia_addr); | |
1568 | if (alen > MAX_VIA_ALEN) | |
1569 | goto errout; | |
1570 | ||
1571 | /* Validate the address family */ | |
1572 | switch (via->rtvia_family) { | |
1573 | case AF_PACKET: | |
1574 | *via_table = NEIGH_LINK_TABLE; | |
1575 | break; | |
1576 | case AF_INET: | |
1577 | *via_table = NEIGH_ARP_TABLE; | |
1578 | if (alen != 4) | |
1579 | goto errout; | |
1580 | break; | |
1581 | case AF_INET6: | |
1582 | *via_table = NEIGH_ND_TABLE; | |
1583 | if (alen != 16) | |
1584 | goto errout; | |
1585 | break; | |
1586 | default: | |
1587 | /* Unsupported address family */ | |
1588 | goto errout; | |
1589 | } | |
1590 | ||
1591 | memcpy(via_addr, via->rtvia_addr, alen); | |
1592 | *via_alen = alen; | |
1593 | err = 0; | |
1594 | ||
1595 | errout: | |
1596 | return err; | |
1597 | } | |
1598 | ||
03c05665 EB |
1599 | static int rtm_to_route_config(struct sk_buff *skb, struct nlmsghdr *nlh, |
1600 | struct mpls_route_config *cfg) | |
1601 | { | |
1602 | struct rtmsg *rtm; | |
1603 | struct nlattr *tb[RTA_MAX+1]; | |
1604 | int index; | |
1605 | int err; | |
1606 | ||
1607 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_mpls_policy); | |
1608 | if (err < 0) | |
1609 | goto errout; | |
1610 | ||
1611 | err = -EINVAL; | |
1612 | rtm = nlmsg_data(nlh); | |
1613 | memset(cfg, 0, sizeof(*cfg)); | |
1614 | ||
1615 | if (rtm->rtm_family != AF_MPLS) | |
1616 | goto errout; | |
1617 | if (rtm->rtm_dst_len != 20) | |
1618 | goto errout; | |
1619 | if (rtm->rtm_src_len != 0) | |
1620 | goto errout; | |
1621 | if (rtm->rtm_tos != 0) | |
1622 | goto errout; | |
1623 | if (rtm->rtm_table != RT_TABLE_MAIN) | |
1624 | goto errout; | |
1625 | /* Any value is acceptable for rtm_protocol */ | |
1626 | ||
1627 | /* As mpls uses destination specific addresses | |
1628 | * (or source specific address in the case of multicast) | |
1629 | * all addresses have universal scope. | |
1630 | */ | |
1631 | if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) | |
1632 | goto errout; | |
1633 | if (rtm->rtm_type != RTN_UNICAST) | |
1634 | goto errout; | |
1635 | if (rtm->rtm_flags != 0) | |
1636 | goto errout; | |
1637 | ||
1638 | cfg->rc_label = LABEL_NOT_SPECIFIED; | |
1639 | cfg->rc_protocol = rtm->rtm_protocol; | |
eb7809f0 | 1640 | cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC; |
5b441ac8 | 1641 | cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT; |
03c05665 EB |
1642 | cfg->rc_nlflags = nlh->nlmsg_flags; |
1643 | cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid; | |
1644 | cfg->rc_nlinfo.nlh = nlh; | |
1645 | cfg->rc_nlinfo.nl_net = sock_net(skb->sk); | |
1646 | ||
1647 | for (index = 0; index <= RTA_MAX; index++) { | |
1648 | struct nlattr *nla = tb[index]; | |
1649 | if (!nla) | |
1650 | continue; | |
1651 | ||
14dd3e1b | 1652 | switch (index) { |
03c05665 EB |
1653 | case RTA_OIF: |
1654 | cfg->rc_ifindex = nla_get_u32(nla); | |
1655 | break; | |
1656 | case RTA_NEWDST: | |
1657 | if (nla_get_labels(nla, MAX_NEW_LABELS, | |
1658 | &cfg->rc_output_labels, | |
1659 | cfg->rc_output_label)) | |
1660 | goto errout; | |
1661 | break; | |
1662 | case RTA_DST: | |
1663 | { | |
f8efb73c | 1664 | u8 label_count; |
03c05665 EB |
1665 | if (nla_get_labels(nla, 1, &label_count, |
1666 | &cfg->rc_label)) | |
1667 | goto errout; | |
1668 | ||
a6affd24 RS |
1669 | /* Reserved labels may not be set */ |
1670 | if (cfg->rc_label < MPLS_LABEL_FIRST_UNRESERVED) | |
03c05665 EB |
1671 | goto errout; |
1672 | ||
1673 | break; | |
1674 | } | |
1675 | case RTA_VIA: | |
1676 | { | |
f8efb73c RP |
1677 | if (nla_get_via(nla, &cfg->rc_via_alen, |
1678 | &cfg->rc_via_table, cfg->rc_via)) | |
03c05665 | 1679 | goto errout; |
f8efb73c RP |
1680 | break; |
1681 | } | |
1682 | case RTA_MULTIPATH: | |
1683 | { | |
1684 | cfg->rc_mp = nla_data(nla); | |
1685 | cfg->rc_mp_len = nla_len(nla); | |
03c05665 EB |
1686 | break; |
1687 | } | |
5b441ac8 RS |
1688 | case RTA_TTL_PROPAGATE: |
1689 | { | |
1690 | u8 ttl_propagate = nla_get_u8(nla); | |
1691 | ||
1692 | if (ttl_propagate > 1) | |
1693 | goto errout; | |
1694 | cfg->rc_ttl_propagate = ttl_propagate ? | |
1695 | MPLS_TTL_PROP_ENABLED : | |
1696 | MPLS_TTL_PROP_DISABLED; | |
1697 | break; | |
1698 | } | |
03c05665 EB |
1699 | default: |
1700 | /* Unsupported attribute */ | |
1701 | goto errout; | |
1702 | } | |
1703 | } | |
1704 | ||
1705 | err = 0; | |
1706 | errout: | |
1707 | return err; | |
1708 | } | |
1709 | ||
1710 | static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh) | |
1711 | { | |
1712 | struct mpls_route_config cfg; | |
1713 | int err; | |
1714 | ||
1715 | err = rtm_to_route_config(skb, nlh, &cfg); | |
1716 | if (err < 0) | |
1717 | return err; | |
1718 | ||
1719 | return mpls_route_del(&cfg); | |
1720 | } | |
1721 | ||
1722 | ||
1723 | static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh) | |
1724 | { | |
1725 | struct mpls_route_config cfg; | |
1726 | int err; | |
1727 | ||
1728 | err = rtm_to_route_config(skb, nlh, &cfg); | |
1729 | if (err < 0) | |
1730 | return err; | |
1731 | ||
1732 | return mpls_route_add(&cfg); | |
1733 | } | |
1734 | ||
1735 | static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event, | |
1736 | u32 label, struct mpls_route *rt, int flags) | |
1737 | { | |
19d0c341 | 1738 | struct net_device *dev; |
03c05665 EB |
1739 | struct nlmsghdr *nlh; |
1740 | struct rtmsg *rtm; | |
1741 | ||
1742 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags); | |
1743 | if (nlh == NULL) | |
1744 | return -EMSGSIZE; | |
1745 | ||
1746 | rtm = nlmsg_data(nlh); | |
1747 | rtm->rtm_family = AF_MPLS; | |
1748 | rtm->rtm_dst_len = 20; | |
1749 | rtm->rtm_src_len = 0; | |
1750 | rtm->rtm_tos = 0; | |
1751 | rtm->rtm_table = RT_TABLE_MAIN; | |
1752 | rtm->rtm_protocol = rt->rt_protocol; | |
1753 | rtm->rtm_scope = RT_SCOPE_UNIVERSE; | |
1754 | rtm->rtm_type = RTN_UNICAST; | |
1755 | rtm->rtm_flags = 0; | |
1756 | ||
03c05665 EB |
1757 | if (nla_put_labels(skb, RTA_DST, 1, &label)) |
1758 | goto nla_put_failure; | |
5b441ac8 RS |
1759 | |
1760 | if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) { | |
1761 | bool ttl_propagate = | |
1762 | rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED; | |
1763 | ||
1764 | if (nla_put_u8(skb, RTA_TTL_PROPAGATE, | |
1765 | ttl_propagate)) | |
1766 | goto nla_put_failure; | |
1767 | } | |
f8efb73c | 1768 | if (rt->rt_nhn == 1) { |
cf4b24f0 | 1769 | const struct mpls_nh *nh = rt->rt_nh; |
f8efb73c RP |
1770 | |
1771 | if (nh->nh_labels && | |
1772 | nla_put_labels(skb, RTA_NEWDST, nh->nh_labels, | |
1773 | nh->nh_label)) | |
1774 | goto nla_put_failure; | |
eb7809f0 | 1775 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && |
72dcac96 | 1776 | nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh), |
f8efb73c RP |
1777 | nh->nh_via_alen)) |
1778 | goto nla_put_failure; | |
1779 | dev = rtnl_dereference(nh->nh_dev); | |
1780 | if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) | |
1781 | goto nla_put_failure; | |
c89359a4 RP |
1782 | if (nh->nh_flags & RTNH_F_LINKDOWN) |
1783 | rtm->rtm_flags |= RTNH_F_LINKDOWN; | |
1784 | if (nh->nh_flags & RTNH_F_DEAD) | |
1785 | rtm->rtm_flags |= RTNH_F_DEAD; | |
f8efb73c RP |
1786 | } else { |
1787 | struct rtnexthop *rtnh; | |
1788 | struct nlattr *mp; | |
c89359a4 RP |
1789 | int dead = 0; |
1790 | int linkdown = 0; | |
f8efb73c RP |
1791 | |
1792 | mp = nla_nest_start(skb, RTA_MULTIPATH); | |
1793 | if (!mp) | |
1794 | goto nla_put_failure; | |
1795 | ||
1796 | for_nexthops(rt) { | |
c00e51dd DA |
1797 | dev = rtnl_dereference(nh->nh_dev); |
1798 | if (!dev) | |
1799 | continue; | |
1800 | ||
f8efb73c RP |
1801 | rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); |
1802 | if (!rtnh) | |
1803 | goto nla_put_failure; | |
1804 | ||
c00e51dd | 1805 | rtnh->rtnh_ifindex = dev->ifindex; |
c89359a4 RP |
1806 | if (nh->nh_flags & RTNH_F_LINKDOWN) { |
1807 | rtnh->rtnh_flags |= RTNH_F_LINKDOWN; | |
1808 | linkdown++; | |
1809 | } | |
1810 | if (nh->nh_flags & RTNH_F_DEAD) { | |
1811 | rtnh->rtnh_flags |= RTNH_F_DEAD; | |
1812 | dead++; | |
1813 | } | |
1814 | ||
f8efb73c RP |
1815 | if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST, |
1816 | nh->nh_labels, | |
1817 | nh->nh_label)) | |
1818 | goto nla_put_failure; | |
f20367df RS |
1819 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC && |
1820 | nla_put_via(skb, nh->nh_via_table, | |
cf4b24f0 | 1821 | mpls_nh_via(rt, nh), |
f8efb73c RP |
1822 | nh->nh_via_alen)) |
1823 | goto nla_put_failure; | |
1824 | ||
1825 | /* length of rtnetlink header + attributes */ | |
1826 | rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh; | |
1827 | } endfor_nexthops(rt); | |
1828 | ||
c89359a4 RP |
1829 | if (linkdown == rt->rt_nhn) |
1830 | rtm->rtm_flags |= RTNH_F_LINKDOWN; | |
1831 | if (dead == rt->rt_nhn) | |
1832 | rtm->rtm_flags |= RTNH_F_DEAD; | |
1833 | ||
f8efb73c RP |
1834 | nla_nest_end(skb, mp); |
1835 | } | |
03c05665 EB |
1836 | |
1837 | nlmsg_end(skb, nlh); | |
1838 | return 0; | |
1839 | ||
1840 | nla_put_failure: | |
1841 | nlmsg_cancel(skb, nlh); | |
1842 | return -EMSGSIZE; | |
1843 | } | |
1844 | ||
1845 | static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb) | |
1846 | { | |
1847 | struct net *net = sock_net(skb->sk); | |
19d0c341 EB |
1848 | struct mpls_route __rcu **platform_label; |
1849 | size_t platform_labels; | |
03c05665 EB |
1850 | unsigned int index; |
1851 | ||
1852 | ASSERT_RTNL(); | |
1853 | ||
1854 | index = cb->args[0]; | |
a6affd24 RS |
1855 | if (index < MPLS_LABEL_FIRST_UNRESERVED) |
1856 | index = MPLS_LABEL_FIRST_UNRESERVED; | |
03c05665 | 1857 | |
19d0c341 EB |
1858 | platform_label = rtnl_dereference(net->mpls.platform_label); |
1859 | platform_labels = net->mpls.platform_labels; | |
1860 | for (; index < platform_labels; index++) { | |
03c05665 | 1861 | struct mpls_route *rt; |
19d0c341 | 1862 | rt = rtnl_dereference(platform_label[index]); |
03c05665 EB |
1863 | if (!rt) |
1864 | continue; | |
1865 | ||
1866 | if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid, | |
1867 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, | |
1868 | index, rt, NLM_F_MULTI) < 0) | |
1869 | break; | |
1870 | } | |
1871 | cb->args[0] = index; | |
1872 | ||
1873 | return skb->len; | |
1874 | } | |
1875 | ||
8de147dc EB |
1876 | static inline size_t lfib_nlmsg_size(struct mpls_route *rt) |
1877 | { | |
1878 | size_t payload = | |
1879 | NLMSG_ALIGN(sizeof(struct rtmsg)) | |
5b441ac8 RS |
1880 | + nla_total_size(4) /* RTA_DST */ |
1881 | + nla_total_size(1); /* RTA_TTL_PROPAGATE */ | |
f8efb73c RP |
1882 | |
1883 | if (rt->rt_nhn == 1) { | |
1884 | struct mpls_nh *nh = rt->rt_nh; | |
1885 | ||
1886 | if (nh->nh_dev) | |
1887 | payload += nla_total_size(4); /* RTA_OIF */ | |
eb7809f0 | 1888 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */ |
72dcac96 | 1889 | payload += nla_total_size(2 + nh->nh_via_alen); |
f8efb73c RP |
1890 | if (nh->nh_labels) /* RTA_NEWDST */ |
1891 | payload += nla_total_size(nh->nh_labels * 4); | |
1892 | } else { | |
1893 | /* each nexthop is packed in an attribute */ | |
1894 | size_t nhsize = 0; | |
1895 | ||
1896 | for_nexthops(rt) { | |
e944e97a DA |
1897 | if (!rtnl_dereference(nh->nh_dev)) |
1898 | continue; | |
f8efb73c | 1899 | nhsize += nla_total_size(sizeof(struct rtnexthop)); |
f20367df RS |
1900 | /* RTA_VIA */ |
1901 | if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) | |
1902 | nhsize += nla_total_size(2 + nh->nh_via_alen); | |
f8efb73c RP |
1903 | if (nh->nh_labels) |
1904 | nhsize += nla_total_size(nh->nh_labels * 4); | |
1905 | } endfor_nexthops(rt); | |
1906 | /* nested attribute */ | |
1907 | payload += nla_total_size(nhsize); | |
1908 | } | |
1909 | ||
8de147dc EB |
1910 | return payload; |
1911 | } | |
1912 | ||
1913 | static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt, | |
1914 | struct nlmsghdr *nlh, struct net *net, u32 portid, | |
1915 | unsigned int nlm_flags) | |
1916 | { | |
1917 | struct sk_buff *skb; | |
1918 | u32 seq = nlh ? nlh->nlmsg_seq : 0; | |
1919 | int err = -ENOBUFS; | |
1920 | ||
1921 | skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL); | |
1922 | if (skb == NULL) | |
1923 | goto errout; | |
1924 | ||
1925 | err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags); | |
1926 | if (err < 0) { | |
1927 | /* -EMSGSIZE implies BUG in lfib_nlmsg_size */ | |
1928 | WARN_ON(err == -EMSGSIZE); | |
1929 | kfree_skb(skb); | |
1930 | goto errout; | |
1931 | } | |
1932 | rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL); | |
1933 | ||
1934 | return; | |
1935 | errout: | |
1936 | if (err < 0) | |
1937 | rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err); | |
1938 | } | |
1939 | ||
7720c01f EB |
1940 | static int resize_platform_label_table(struct net *net, size_t limit) |
1941 | { | |
1942 | size_t size = sizeof(struct mpls_route *) * limit; | |
1943 | size_t old_limit; | |
1944 | size_t cp_size; | |
1945 | struct mpls_route __rcu **labels = NULL, **old; | |
1946 | struct mpls_route *rt0 = NULL, *rt2 = NULL; | |
1947 | unsigned index; | |
1948 | ||
1949 | if (size) { | |
1950 | labels = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY); | |
1951 | if (!labels) | |
1952 | labels = vzalloc(size); | |
1953 | ||
1954 | if (!labels) | |
1955 | goto nolabels; | |
1956 | } | |
1957 | ||
1958 | /* In case the predefined labels need to be populated */ | |
78f5b899 | 1959 | if (limit > MPLS_LABEL_IPV4NULL) { |
7720c01f | 1960 | struct net_device *lo = net->loopback_dev; |
cf4b24f0 | 1961 | rt0 = mpls_rt_alloc(1, lo->addr_len); |
7720c01f EB |
1962 | if (!rt0) |
1963 | goto nort0; | |
f8efb73c | 1964 | RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo); |
7720c01f | 1965 | rt0->rt_protocol = RTPROT_KERNEL; |
118d5234 | 1966 | rt0->rt_payload_type = MPT_IPV4; |
5b441ac8 | 1967 | rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT; |
f8efb73c | 1968 | rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE; |
b4e04fc7 | 1969 | rt0->rt_nh->nh_via_alen = lo->addr_len; |
cf4b24f0 RS |
1970 | memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr, |
1971 | lo->addr_len); | |
7720c01f | 1972 | } |
78f5b899 | 1973 | if (limit > MPLS_LABEL_IPV6NULL) { |
7720c01f | 1974 | struct net_device *lo = net->loopback_dev; |
cf4b24f0 | 1975 | rt2 = mpls_rt_alloc(1, lo->addr_len); |
7720c01f EB |
1976 | if (!rt2) |
1977 | goto nort2; | |
f8efb73c | 1978 | RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo); |
7720c01f | 1979 | rt2->rt_protocol = RTPROT_KERNEL; |
118d5234 | 1980 | rt2->rt_payload_type = MPT_IPV6; |
6a18c312 | 1981 | rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT; |
f8efb73c | 1982 | rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE; |
b4e04fc7 | 1983 | rt2->rt_nh->nh_via_alen = lo->addr_len; |
cf4b24f0 RS |
1984 | memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr, |
1985 | lo->addr_len); | |
7720c01f EB |
1986 | } |
1987 | ||
1988 | rtnl_lock(); | |
1989 | /* Remember the original table */ | |
19d0c341 | 1990 | old = rtnl_dereference(net->mpls.platform_label); |
7720c01f EB |
1991 | old_limit = net->mpls.platform_labels; |
1992 | ||
1993 | /* Free any labels beyond the new table */ | |
1994 | for (index = limit; index < old_limit; index++) | |
f8efb73c | 1995 | mpls_route_update(net, index, NULL, NULL); |
7720c01f EB |
1996 | |
1997 | /* Copy over the old labels */ | |
1998 | cp_size = size; | |
1999 | if (old_limit < limit) | |
2000 | cp_size = old_limit * sizeof(struct mpls_route *); | |
2001 | ||
2002 | memcpy(labels, old, cp_size); | |
2003 | ||
2004 | /* If needed set the predefined labels */ | |
78f5b899 TH |
2005 | if ((old_limit <= MPLS_LABEL_IPV6NULL) && |
2006 | (limit > MPLS_LABEL_IPV6NULL)) { | |
2007 | RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2); | |
7720c01f EB |
2008 | rt2 = NULL; |
2009 | } | |
2010 | ||
78f5b899 TH |
2011 | if ((old_limit <= MPLS_LABEL_IPV4NULL) && |
2012 | (limit > MPLS_LABEL_IPV4NULL)) { | |
2013 | RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0); | |
7720c01f EB |
2014 | rt0 = NULL; |
2015 | } | |
2016 | ||
2017 | /* Update the global pointers */ | |
2018 | net->mpls.platform_labels = limit; | |
19d0c341 | 2019 | rcu_assign_pointer(net->mpls.platform_label, labels); |
7720c01f EB |
2020 | |
2021 | rtnl_unlock(); | |
2022 | ||
2023 | mpls_rt_free(rt2); | |
2024 | mpls_rt_free(rt0); | |
2025 | ||
2026 | if (old) { | |
2027 | synchronize_rcu(); | |
2028 | kvfree(old); | |
2029 | } | |
2030 | return 0; | |
2031 | ||
2032 | nort2: | |
2033 | mpls_rt_free(rt0); | |
2034 | nort0: | |
2035 | kvfree(labels); | |
2036 | nolabels: | |
2037 | return -ENOMEM; | |
2038 | } | |
2039 | ||
2040 | static int mpls_platform_labels(struct ctl_table *table, int write, | |
2041 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2042 | { | |
2043 | struct net *net = table->data; | |
2044 | int platform_labels = net->mpls.platform_labels; | |
2045 | int ret; | |
2046 | struct ctl_table tmp = { | |
2047 | .procname = table->procname, | |
2048 | .data = &platform_labels, | |
2049 | .maxlen = sizeof(int), | |
2050 | .mode = table->mode, | |
2051 | .extra1 = &zero, | |
2052 | .extra2 = &label_limit, | |
2053 | }; | |
2054 | ||
2055 | ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); | |
2056 | ||
2057 | if (write && ret == 0) | |
2058 | ret = resize_platform_label_table(net, platform_labels); | |
2059 | ||
2060 | return ret; | |
2061 | } | |
2062 | ||
5b441ac8 RS |
2063 | #define MPLS_NS_SYSCTL_OFFSET(field) \ |
2064 | (&((struct net *)0)->field) | |
2065 | ||
37bde799 | 2066 | static const struct ctl_table mpls_table[] = { |
7720c01f EB |
2067 | { |
2068 | .procname = "platform_labels", | |
2069 | .data = NULL, | |
2070 | .maxlen = sizeof(int), | |
2071 | .mode = 0644, | |
2072 | .proc_handler = mpls_platform_labels, | |
2073 | }, | |
5b441ac8 RS |
2074 | { |
2075 | .procname = "ip_ttl_propagate", | |
2076 | .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate), | |
2077 | .maxlen = sizeof(int), | |
2078 | .mode = 0644, | |
2079 | .proc_handler = proc_dointvec_minmax, | |
2080 | .extra1 = &zero, | |
2081 | .extra2 = &one, | |
2082 | }, | |
a59166e4 RS |
2083 | { |
2084 | .procname = "default_ttl", | |
2085 | .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl), | |
2086 | .maxlen = sizeof(int), | |
2087 | .mode = 0644, | |
2088 | .proc_handler = proc_dointvec_minmax, | |
2089 | .extra1 = &one, | |
2090 | .extra2 = &ttl_max, | |
2091 | }, | |
7720c01f EB |
2092 | { } |
2093 | }; | |
2094 | ||
0189197f EB |
2095 | static int mpls_net_init(struct net *net) |
2096 | { | |
7720c01f | 2097 | struct ctl_table *table; |
5b441ac8 | 2098 | int i; |
7720c01f | 2099 | |
0189197f EB |
2100 | net->mpls.platform_labels = 0; |
2101 | net->mpls.platform_label = NULL; | |
5b441ac8 | 2102 | net->mpls.ip_ttl_propagate = 1; |
a59166e4 | 2103 | net->mpls.default_ttl = 255; |
0189197f | 2104 | |
7720c01f EB |
2105 | table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL); |
2106 | if (table == NULL) | |
2107 | return -ENOMEM; | |
2108 | ||
5b441ac8 RS |
2109 | /* Table data contains only offsets relative to the base of |
2110 | * the mdev at this point, so make them absolute. | |
2111 | */ | |
2112 | for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++) | |
2113 | table[i].data = (char *)net + (uintptr_t)table[i].data; | |
2114 | ||
7720c01f | 2115 | net->mpls.ctl = register_net_sysctl(net, "net/mpls", table); |
6ea3c9d5 NA |
2116 | if (net->mpls.ctl == NULL) { |
2117 | kfree(table); | |
7720c01f | 2118 | return -ENOMEM; |
6ea3c9d5 | 2119 | } |
7720c01f | 2120 | |
0189197f EB |
2121 | return 0; |
2122 | } | |
2123 | ||
2124 | static void mpls_net_exit(struct net *net) | |
2125 | { | |
19d0c341 EB |
2126 | struct mpls_route __rcu **platform_label; |
2127 | size_t platform_labels; | |
7720c01f | 2128 | struct ctl_table *table; |
0189197f EB |
2129 | unsigned int index; |
2130 | ||
7720c01f EB |
2131 | table = net->mpls.ctl->ctl_table_arg; |
2132 | unregister_net_sysctl_table(net->mpls.ctl); | |
2133 | kfree(table); | |
2134 | ||
19d0c341 EB |
2135 | /* An rcu grace period has passed since there was a device in |
2136 | * the network namespace (and thus the last in flight packet) | |
0189197f EB |
2137 | * left this network namespace. This is because |
2138 | * unregister_netdevice_many and netdev_run_todo has completed | |
2139 | * for each network device that was in this network namespace. | |
2140 | * | |
2141 | * As such no additional rcu synchronization is necessary when | |
2142 | * freeing the platform_label table. | |
2143 | */ | |
2144 | rtnl_lock(); | |
19d0c341 EB |
2145 | platform_label = rtnl_dereference(net->mpls.platform_label); |
2146 | platform_labels = net->mpls.platform_labels; | |
2147 | for (index = 0; index < platform_labels; index++) { | |
2148 | struct mpls_route *rt = rtnl_dereference(platform_label[index]); | |
2149 | RCU_INIT_POINTER(platform_label[index], NULL); | |
e37791ec | 2150 | mpls_notify_route(net, index, rt, NULL, NULL); |
0189197f EB |
2151 | mpls_rt_free(rt); |
2152 | } | |
2153 | rtnl_unlock(); | |
2154 | ||
19d0c341 | 2155 | kvfree(platform_label); |
0189197f EB |
2156 | } |
2157 | ||
2158 | static struct pernet_operations mpls_net_ops = { | |
2159 | .init = mpls_net_init, | |
2160 | .exit = mpls_net_exit, | |
2161 | }; | |
2162 | ||
27d69105 RS |
2163 | static struct rtnl_af_ops mpls_af_ops __read_mostly = { |
2164 | .family = AF_MPLS, | |
2165 | .fill_stats_af = mpls_fill_stats_af, | |
2166 | .get_stats_af_size = mpls_get_stats_af_size, | |
2167 | }; | |
2168 | ||
0189197f EB |
2169 | static int __init mpls_init(void) |
2170 | { | |
2171 | int err; | |
2172 | ||
2173 | BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4); | |
2174 | ||
2175 | err = register_pernet_subsys(&mpls_net_ops); | |
2176 | if (err) | |
2177 | goto out; | |
2178 | ||
2179 | err = register_netdevice_notifier(&mpls_dev_notifier); | |
2180 | if (err) | |
2181 | goto out_unregister_pernet; | |
2182 | ||
2183 | dev_add_pack(&mpls_packet_type); | |
2184 | ||
27d69105 RS |
2185 | rtnl_af_register(&mpls_af_ops); |
2186 | ||
03c05665 EB |
2187 | rtnl_register(PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, NULL); |
2188 | rtnl_register(PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, NULL); | |
2189 | rtnl_register(PF_MPLS, RTM_GETROUTE, NULL, mpls_dump_routes, NULL); | |
24045a03 DA |
2190 | rtnl_register(PF_MPLS, RTM_GETNETCONF, mpls_netconf_get_devconf, |
2191 | mpls_netconf_dump_devconf, NULL); | |
0189197f EB |
2192 | err = 0; |
2193 | out: | |
2194 | return err; | |
2195 | ||
2196 | out_unregister_pernet: | |
2197 | unregister_pernet_subsys(&mpls_net_ops); | |
2198 | goto out; | |
2199 | } | |
2200 | module_init(mpls_init); | |
2201 | ||
2202 | static void __exit mpls_exit(void) | |
2203 | { | |
03c05665 | 2204 | rtnl_unregister_all(PF_MPLS); |
27d69105 | 2205 | rtnl_af_unregister(&mpls_af_ops); |
0189197f EB |
2206 | dev_remove_pack(&mpls_packet_type); |
2207 | unregister_netdevice_notifier(&mpls_dev_notifier); | |
2208 | unregister_pernet_subsys(&mpls_net_ops); | |
2209 | } | |
2210 | module_exit(mpls_exit); | |
2211 | ||
2212 | MODULE_DESCRIPTION("MultiProtocol Label Switching"); | |
2213 | MODULE_LICENSE("GPL v2"); | |
2214 | MODULE_ALIAS_NETPROTO(PF_MPLS); |