]>
Commit | Line | Data |
---|---|---|
6d2010ae | 1 | /* |
316670eb | 2 | * Copyright (c) 2011-2012 Apple Inc. All rights reserved. |
6d2010ae A |
3 | * |
4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ | |
5 | * | |
6 | * This file contains Original Code and/or Modifications of Original Code | |
7 | * as defined in and that are subject to the Apple Public Source License | |
8 | * Version 2.0 (the 'License'). You may not use this file except in | |
9 | * compliance with the License. The rights granted to you under the License | |
10 | * may not be used to create, or enable the creation or redistribution of, | |
11 | * unlawful or unlicensed copies of an Apple operating system, or to | |
12 | * circumvent, violate, or enable the circumvention or violation of, any | |
13 | * terms of an Apple operating system software license agreement. | |
14 | * | |
15 | * Please obtain a copy of the License at | |
16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. | |
17 | * | |
18 | * The Original Code and all software distributed under the License are | |
19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER | |
20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, | |
21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, | |
22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. | |
23 | * Please see the License for the specific language governing rights and | |
24 | * limitations under the License. | |
25 | * | |
26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ | |
27 | */ | |
28 | ||
29 | /* | |
30 | * Link-layer Reachability Record | |
31 | * | |
32 | * Each interface maintains a red-black tree which contains records related | |
33 | * to the on-link nodes which we are interested in communicating with. Each | |
34 | * record gets allocated and inserted into the tree in the following manner: | |
35 | * upon processing an ARP announcement or reply from a known node (i.e. there | |
36 | * exists a ARP route entry for the node), and if a link-layer reachability | |
37 | * record for the node doesn't yet exist; and, upon processing a ND6 RS/RA/ | |
38 | * NS/NA/redirect from a node, and if a link-layer reachability record for the | |
39 | * node doesn't yet exist. | |
40 | * | |
41 | * Each newly created record is then referred to by the resolver route entry; | |
42 | * if a record already exists, its reference count gets increased for the new | |
43 | * resolver entry which now refers to it. A record gets removed from the tree | |
44 | * and freed once its reference counts drops to zero, i.e. when there is no | |
45 | * more resolver entry referring to it. | |
46 | * | |
47 | * A record contains the link-layer protocol (e.g. Ethertype IP/IPv6), the | |
48 | * HW address of the sender, the "last heard from" timestamp (lr_lastrcvd) and | |
49 | * the number of references made to it (lr_reqcnt). Because the key for each | |
50 | * record in the red-black tree consists of the link-layer protocol, therefore | |
51 | * the namespace for the records is partitioned based on the type of link-layer | |
52 | * protocol, i.e. an Ethertype IP link-layer record is only referred to by one | |
53 | * or more ARP entries; an Ethernet IPv6 link-layer record is only referred to | |
54 | * by one or more ND6 entries. Therefore, lr_reqcnt represents the number of | |
55 | * resolver entry references to the record for the same protocol family. | |
56 | * | |
57 | * Upon receiving packets from the network, the protocol's input callback | |
58 | * (e.g. ether_inet{6}_input) informs the corresponding resolver (ARP/ND6) | |
59 | * about the (link-layer) origin of the packet. This results in searching | |
60 | * for a matching record in the red-black tree for the interface where the | |
61 | * packet arrived on. If there's no match, no further processing takes place. | |
62 | * Otherwise, the lr_lastrcvd timestamp of the record is updated. | |
63 | * | |
64 | * When an IP/IPv6 packet is transmitted to the resolver (i.e. the destination | |
65 | * is on-link), ARP/ND6 records the "last spoken to" timestamp in the route | |
66 | * entry ({la,ln}_lastused). | |
67 | * | |
68 | * The reachability of the on-link node is determined by the following logic, | |
69 | * upon sending a packet thru the resolver: | |
70 | * | |
71 | * a) If the record is only used by exactly one resolver entry (lr_reqcnt | |
72 | * is 1), i.e. the target host does not have IP/IPv6 aliases that we know | |
73 | * of, check if lr_lastrcvd is "recent." If so, simply send the packet; | |
74 | * otherwise, re-resolve the target node. | |
75 | * | |
76 | * b) If the record is shared by multiple resolver entries (lr_reqcnt is | |
77 | * greater than 1), i.e. the target host has more than one IP/IPv6 aliases | |
78 | * on the same network interface, we can't rely on lr_lastrcvd alone, as | |
79 | * one of the IP/IPv6 aliases could have been silently moved to another | |
80 | * node for which we don't have a link-layer record. If lr_lastrcvd is | |
81 | * not "recent", we re-resolve the target node. Otherwise, we perform | |
82 | * an additional check against {la,ln}_lastused to see whether it is also | |
83 | * "recent", relative to lr_lastrcvd. If so, simply send the packet; | |
84 | * otherwise, re-resolve the target node. | |
85 | * | |
86 | * The value for "recent" is configurable by adjusting the basetime value for | |
87 | * net.link.ether.inet.arp_llreach_base or net.inet6.icmp6.nd6_llreach_base. | |
88 | * The default basetime value is 30 seconds, and the actual expiration time | |
89 | * is calculated by multiplying the basetime value with some random factor, | |
90 | * which results in a number between 15 to 45 seconds. Setting the basetime | |
91 | * value to 0 effectively disables this feature for the corresponding resolver. | |
92 | * | |
93 | * Assumptions: | |
94 | * | |
95 | * The above logic is based upon the following assumptions: | |
96 | * | |
97 | * i) Network traffics are mostly bi-directional, i.e. the act of sending | |
98 | * packets to an on-link node would most likely cause us to receive | |
99 | * packets from that node. | |
100 | * | |
101 | * ii) If the on-link node's IP/IPv6 address silently moves to another | |
102 | * on-link node for which we are not aware of, non-unicast packets | |
103 | * from the old node would trigger the record's lr_lastrcvd to be | |
104 | * kept recent. | |
105 | * | |
106 | * We can mitigate the above by having the resolver check its {la,ln}_lastused | |
107 | * timestamp at all times, i.e. not only when lr_reqcnt is greater than 1; but | |
108 | * we currently optimize for the common cases. | |
109 | */ | |
110 | ||
111 | #include <sys/param.h> | |
112 | #include <sys/systm.h> | |
113 | #include <sys/kernel.h> | |
114 | #include <sys/malloc.h> | |
115 | #include <sys/tree.h> | |
116 | #include <sys/sysctl.h> | |
117 | #include <sys/mcache.h> | |
118 | #include <sys/protosw.h> | |
119 | ||
39236c6e A |
120 | #include <dev/random/randomdev.h> |
121 | ||
6d2010ae A |
122 | #include <net/if_dl.h> |
123 | #include <net/if.h> | |
124 | #include <net/if_var.h> | |
125 | #include <net/if_llreach.h> | |
126 | #include <net/dlil.h> | |
316670eb A |
127 | #include <net/kpi_interface.h> |
128 | #include <net/route.h> | |
6d2010ae A |
129 | |
130 | #include <kern/assert.h> | |
131 | #include <kern/locks.h> | |
132 | #include <kern/zalloc.h> | |
133 | ||
134 | #if INET6 | |
135 | #include <netinet6/in6_var.h> | |
136 | #include <netinet6/nd6.h> | |
137 | #endif /* INET6 */ | |
138 | ||
0a7de745 A |
139 | static unsigned int iflr_size; /* size of if_llreach */ |
140 | static struct zone *iflr_zone; /* zone for if_llreach */ | |
6d2010ae | 141 | |
0a7de745 A |
142 | #define IFLR_ZONE_MAX 128 /* maximum elements in zone */ |
143 | #define IFLR_ZONE_NAME "if_llreach" /* zone name */ | |
6d2010ae A |
144 | |
145 | static struct if_llreach *iflr_alloc(int); | |
146 | static void iflr_free(struct if_llreach *); | |
147 | static __inline int iflr_cmp(const struct if_llreach *, | |
148 | const struct if_llreach *); | |
149 | static __inline int iflr_reachable(struct if_llreach *, int, u_int64_t); | |
150 | static int sysctl_llreach_ifinfo SYSCTL_HANDLER_ARGS; | |
151 | ||
152 | /* The following is protected by if_llreach_lock */ | |
153 | RB_GENERATE_PREV(ll_reach_tree, if_llreach, lr_link, iflr_cmp); | |
154 | ||
155 | SYSCTL_DECL(_net_link_generic_system); | |
156 | ||
157 | SYSCTL_NODE(_net_link_generic_system, OID_AUTO, llreach_info, | |
158 | CTLFLAG_RD | CTLFLAG_LOCKED, sysctl_llreach_ifinfo, | |
159 | "Per-interface tree of source link-layer reachability records"); | |
160 | ||
161 | /* | |
162 | * Link-layer reachability is based off node constants in RFC4861. | |
163 | */ | |
164 | #if INET6 | |
0a7de745 | 165 | #define LL_COMPUTE_RTIME(x) ND_COMPUTE_RTIME(x) |
6d2010ae | 166 | #else |
0a7de745 A |
167 | #define LL_MIN_RANDOM_FACTOR 512 /* 1024 * 0.5 */ |
168 | #define LL_MAX_RANDOM_FACTOR 1536 /* 1024 * 1.5 */ | |
169 | #define LL_COMPUTE_RTIME(x) \ | |
170 | (((LL_MIN_RANDOM_FACTOR * (x >> 10)) + (RandomULong() & \ | |
6d2010ae A |
171 | ((LL_MAX_RANDOM_FACTOR - LL_MIN_RANDOM_FACTOR) * (x >> 10)))) / 1000) |
172 | #endif /* !INET6 */ | |
173 | ||
174 | void | |
175 | ifnet_llreach_init(void) | |
176 | { | |
0a7de745 | 177 | iflr_size = sizeof(struct if_llreach); |
6d2010ae A |
178 | iflr_zone = zinit(iflr_size, |
179 | IFLR_ZONE_MAX * iflr_size, 0, IFLR_ZONE_NAME); | |
180 | if (iflr_zone == NULL) { | |
181 | panic("%s: failed allocating %s", __func__, IFLR_ZONE_NAME); | |
182 | /* NOTREACHED */ | |
183 | } | |
184 | zone_change(iflr_zone, Z_EXPAND, TRUE); | |
185 | zone_change(iflr_zone, Z_CALLERACCT, FALSE); | |
186 | } | |
187 | ||
188 | void | |
189 | ifnet_llreach_ifattach(struct ifnet *ifp, boolean_t reuse) | |
190 | { | |
191 | lck_rw_lock_exclusive(&ifp->if_llreach_lock); | |
192 | /* Initialize link-layer source tree (if not already) */ | |
0a7de745 | 193 | if (!reuse) { |
6d2010ae | 194 | RB_INIT(&ifp->if_ll_srcs); |
0a7de745 | 195 | } |
6d2010ae A |
196 | lck_rw_done(&ifp->if_llreach_lock); |
197 | } | |
198 | ||
199 | void | |
200 | ifnet_llreach_ifdetach(struct ifnet *ifp) | |
201 | { | |
202 | #pragma unused(ifp) | |
203 | /* | |
204 | * Nothing to do for now; the link-layer source tree might | |
205 | * contain entries at this point, that are still referred | |
206 | * to by route entries pointing to this ifp. | |
207 | */ | |
208 | } | |
209 | ||
210 | /* | |
211 | * Link-layer source tree comparison function. | |
212 | * | |
213 | * An ordered predicate is necessary; bcmp() is not documented to return | |
214 | * an indication of order, memcmp() is, and is an ISO C99 requirement. | |
215 | */ | |
216 | static __inline int | |
217 | iflr_cmp(const struct if_llreach *a, const struct if_llreach *b) | |
218 | { | |
0a7de745 | 219 | return memcmp(&a->lr_key, &b->lr_key, sizeof(a->lr_key)); |
6d2010ae A |
220 | } |
221 | ||
222 | static __inline int | |
223 | iflr_reachable(struct if_llreach *lr, int cmp_delta, u_int64_t tval) | |
224 | { | |
225 | u_int64_t now; | |
226 | u_int64_t expire; | |
227 | ||
0a7de745 | 228 | now = net_uptime(); /* current approx. uptime */ |
6d2010ae A |
229 | /* |
230 | * No need for lr_lock; atomically read the last rcvd uptime. | |
231 | */ | |
232 | expire = lr->lr_lastrcvd + lr->lr_reachable; | |
233 | /* | |
234 | * If we haven't heard back from the local host for over | |
235 | * lr_reachable seconds, consider that the host is no | |
236 | * longer reachable. | |
237 | */ | |
0a7de745 A |
238 | if (!cmp_delta) { |
239 | return expire >= now; | |
240 | } | |
6d2010ae A |
241 | /* |
242 | * If the caller supplied a reference time, consider the | |
243 | * host is reachable if the record hasn't expired (see above) | |
244 | * and if the reference time is within the past lr_reachable | |
245 | * seconds. | |
246 | */ | |
0a7de745 | 247 | return (expire >= now) && (now - tval) < lr->lr_reachable; |
6d2010ae A |
248 | } |
249 | ||
250 | int | |
251 | ifnet_llreach_reachable(struct if_llreach *lr) | |
252 | { | |
253 | /* | |
254 | * Check whether the cache is too old to be trusted. | |
255 | */ | |
0a7de745 | 256 | return iflr_reachable(lr, 0, 0); |
6d2010ae A |
257 | } |
258 | ||
259 | int | |
260 | ifnet_llreach_reachable_delta(struct if_llreach *lr, u_int64_t tval) | |
261 | { | |
262 | /* | |
263 | * Check whether the cache is too old to be trusted. | |
264 | */ | |
0a7de745 | 265 | return iflr_reachable(lr, 1, tval); |
6d2010ae A |
266 | } |
267 | ||
268 | void | |
269 | ifnet_llreach_set_reachable(struct ifnet *ifp, u_int16_t llproto, void *addr, | |
270 | unsigned int alen) | |
271 | { | |
272 | struct if_llreach find, *lr; | |
273 | ||
0a7de745 | 274 | VERIFY(alen == IF_LLREACH_MAXLEN); /* for now */ |
6d2010ae A |
275 | |
276 | find.lr_key.proto = llproto; | |
277 | bcopy(addr, &find.lr_key.addr, IF_LLREACH_MAXLEN); | |
278 | ||
279 | lck_rw_lock_shared(&ifp->if_llreach_lock); | |
280 | lr = RB_FIND(ll_reach_tree, &ifp->if_ll_srcs, &find); | |
281 | if (lr == NULL) { | |
282 | lck_rw_done(&ifp->if_llreach_lock); | |
283 | return; | |
284 | } | |
285 | /* | |
286 | * No need for lr_lock; atomically update the last rcvd uptime. | |
287 | */ | |
288 | lr->lr_lastrcvd = net_uptime(); | |
289 | lck_rw_done(&ifp->if_llreach_lock); | |
290 | } | |
291 | ||
292 | struct if_llreach * | |
293 | ifnet_llreach_alloc(struct ifnet *ifp, u_int16_t llproto, void *addr, | |
294 | unsigned int alen, u_int64_t llreach_base) | |
295 | { | |
296 | struct if_llreach find, *lr; | |
39236c6e | 297 | struct timeval cnow; |
6d2010ae | 298 | |
0a7de745 A |
299 | if (llreach_base == 0) { |
300 | return NULL; | |
301 | } | |
6d2010ae | 302 | |
0a7de745 | 303 | VERIFY(alen == IF_LLREACH_MAXLEN); /* for now */ |
6d2010ae A |
304 | |
305 | find.lr_key.proto = llproto; | |
306 | bcopy(addr, &find.lr_key.addr, IF_LLREACH_MAXLEN); | |
307 | ||
308 | lck_rw_lock_shared(&ifp->if_llreach_lock); | |
309 | lr = RB_FIND(ll_reach_tree, &ifp->if_ll_srcs, &find); | |
310 | if (lr != NULL) { | |
311 | found: | |
312 | IFLR_LOCK(lr); | |
313 | VERIFY(lr->lr_reqcnt >= 1); | |
314 | lr->lr_reqcnt++; | |
315 | VERIFY(lr->lr_reqcnt != 0); | |
0a7de745 A |
316 | IFLR_ADDREF_LOCKED(lr); /* for caller */ |
317 | lr->lr_lastrcvd = net_uptime(); /* current approx. uptime */ | |
6d2010ae A |
318 | IFLR_UNLOCK(lr); |
319 | lck_rw_done(&ifp->if_llreach_lock); | |
0a7de745 | 320 | return lr; |
6d2010ae A |
321 | } |
322 | ||
0a7de745 | 323 | if (!lck_rw_lock_shared_to_exclusive(&ifp->if_llreach_lock)) { |
6d2010ae | 324 | lck_rw_lock_exclusive(&ifp->if_llreach_lock); |
0a7de745 | 325 | } |
6d2010ae | 326 | |
5ba3f43e | 327 | LCK_RW_ASSERT(&ifp->if_llreach_lock, LCK_RW_ASSERT_EXCLUSIVE); |
6d2010ae A |
328 | |
329 | /* in case things have changed while becoming writer */ | |
330 | lr = RB_FIND(ll_reach_tree, &ifp->if_ll_srcs, &find); | |
0a7de745 | 331 | if (lr != NULL) { |
6d2010ae | 332 | goto found; |
0a7de745 | 333 | } |
6d2010ae A |
334 | |
335 | lr = iflr_alloc(M_WAITOK); | |
336 | if (lr == NULL) { | |
337 | lck_rw_done(&ifp->if_llreach_lock); | |
0a7de745 | 338 | return NULL; |
6d2010ae A |
339 | } |
340 | IFLR_LOCK(lr); | |
341 | lr->lr_reqcnt++; | |
342 | VERIFY(lr->lr_reqcnt == 1); | |
0a7de745 A |
343 | IFLR_ADDREF_LOCKED(lr); /* for RB tree */ |
344 | IFLR_ADDREF_LOCKED(lr); /* for caller */ | |
345 | lr->lr_lastrcvd = net_uptime(); /* current approx. uptime */ | |
346 | lr->lr_baseup = lr->lr_lastrcvd; /* base uptime */ | |
39236c6e | 347 | getmicrotime(&cnow); |
0a7de745 | 348 | lr->lr_basecal = cnow.tv_sec; /* base calendar time */ |
6d2010ae A |
349 | lr->lr_basereachable = llreach_base; |
350 | lr->lr_reachable = LL_COMPUTE_RTIME(lr->lr_basereachable * 1000); | |
351 | lr->lr_debug |= IFD_ATTACHED; | |
352 | lr->lr_ifp = ifp; | |
353 | lr->lr_key.proto = llproto; | |
354 | bcopy(addr, &lr->lr_key.addr, IF_LLREACH_MAXLEN); | |
316670eb A |
355 | lr->lr_rssi = IFNET_RSSI_UNKNOWN; |
356 | lr->lr_lqm = IFNET_LQM_THRESH_UNKNOWN; | |
357 | lr->lr_npm = IFNET_NPM_THRESH_UNKNOWN; | |
6d2010ae A |
358 | RB_INSERT(ll_reach_tree, &ifp->if_ll_srcs, lr); |
359 | IFLR_UNLOCK(lr); | |
360 | lck_rw_done(&ifp->if_llreach_lock); | |
361 | ||
0a7de745 | 362 | return lr; |
6d2010ae A |
363 | } |
364 | ||
365 | void | |
366 | ifnet_llreach_free(struct if_llreach *lr) | |
367 | { | |
368 | struct ifnet *ifp; | |
369 | ||
370 | /* no need to lock here; lr_ifp never changes */ | |
371 | ifp = lr->lr_ifp; | |
372 | ||
373 | lck_rw_lock_exclusive(&ifp->if_llreach_lock); | |
374 | IFLR_LOCK(lr); | |
375 | if (lr->lr_reqcnt == 0) { | |
376 | panic("%s: lr=%p negative reqcnt", __func__, lr); | |
377 | /* NOTREACHED */ | |
378 | } | |
379 | --lr->lr_reqcnt; | |
380 | if (lr->lr_reqcnt > 0) { | |
381 | IFLR_UNLOCK(lr); | |
382 | lck_rw_done(&ifp->if_llreach_lock); | |
0a7de745 | 383 | IFLR_REMREF(lr); /* for caller */ |
6d2010ae A |
384 | return; |
385 | } | |
386 | if (!(lr->lr_debug & IFD_ATTACHED)) { | |
387 | panic("%s: Attempt to detach an unattached llreach lr=%p", | |
388 | __func__, lr); | |
389 | /* NOTREACHED */ | |
390 | } | |
391 | lr->lr_debug &= ~IFD_ATTACHED; | |
392 | RB_REMOVE(ll_reach_tree, &ifp->if_ll_srcs, lr); | |
393 | IFLR_UNLOCK(lr); | |
394 | lck_rw_done(&ifp->if_llreach_lock); | |
395 | ||
0a7de745 A |
396 | IFLR_REMREF(lr); /* for RB tree */ |
397 | IFLR_REMREF(lr); /* for caller */ | |
6d2010ae A |
398 | } |
399 | ||
400 | u_int64_t | |
316670eb | 401 | ifnet_llreach_up2calexp(struct if_llreach *lr, u_int64_t uptime) |
6d2010ae A |
402 | { |
403 | u_int64_t calendar = 0; | |
404 | ||
405 | if (uptime != 0) { | |
406 | struct timeval cnow; | |
407 | u_int64_t unow; | |
408 | ||
0a7de745 A |
409 | getmicrotime(&cnow); /* current calendar time */ |
410 | unow = net_uptime(); /* current approx. uptime */ | |
6d2010ae A |
411 | /* |
412 | * Take into account possible calendar time changes; | |
413 | * adjust base calendar value if necessary, i.e. | |
414 | * the calendar skew should equate to the uptime skew. | |
415 | */ | |
416 | lr->lr_basecal += (cnow.tv_sec - lr->lr_basecal) - | |
417 | (unow - lr->lr_baseup); | |
418 | ||
419 | calendar = lr->lr_basecal + lr->lr_reachable + | |
420 | (uptime - lr->lr_baseup); | |
421 | } | |
422 | ||
0a7de745 | 423 | return calendar; |
6d2010ae A |
424 | } |
425 | ||
316670eb A |
426 | u_int64_t |
427 | ifnet_llreach_up2upexp(struct if_llreach *lr, u_int64_t uptime) | |
428 | { | |
0a7de745 | 429 | return lr->lr_reachable + uptime; |
316670eb A |
430 | } |
431 | ||
432 | int | |
433 | ifnet_llreach_get_defrouter(struct ifnet *ifp, int af, | |
434 | struct ifnet_llreach_info *iflri) | |
435 | { | |
436 | struct radix_node_head *rnh; | |
437 | struct sockaddr_storage dst_ss, mask_ss; | |
438 | struct rtentry *rt; | |
439 | int error = ESRCH; | |
440 | ||
441 | VERIFY(ifp != NULL && iflri != NULL && | |
442 | (af == AF_INET || af == AF_INET6)); | |
443 | ||
0a7de745 | 444 | bzero(iflri, sizeof(*iflri)); |
316670eb | 445 | |
0a7de745 A |
446 | if ((rnh = rt_tables[af]) == NULL) { |
447 | return error; | |
448 | } | |
316670eb | 449 | |
0a7de745 A |
450 | bzero(&dst_ss, sizeof(dst_ss)); |
451 | bzero(&mask_ss, sizeof(mask_ss)); | |
316670eb | 452 | dst_ss.ss_family = af; |
0a7de745 A |
453 | dst_ss.ss_len = (af == AF_INET) ? sizeof(struct sockaddr_in) : |
454 | sizeof(struct sockaddr_in6); | |
316670eb A |
455 | |
456 | lck_mtx_lock(rnh_lock); | |
457 | rt = rt_lookup(TRUE, SA(&dst_ss), SA(&mask_ss), rnh, ifp->if_index); | |
458 | if (rt != NULL) { | |
459 | struct rtentry *gwrt; | |
460 | ||
461 | RT_LOCK(rt); | |
462 | if ((rt->rt_flags & RTF_GATEWAY) && | |
463 | (gwrt = rt->rt_gwroute) != NULL && | |
464 | rt_key(rt)->sa_family == rt_key(gwrt)->sa_family && | |
465 | (gwrt->rt_flags & RTF_UP)) { | |
466 | RT_UNLOCK(rt); | |
467 | RT_LOCK(gwrt); | |
468 | if (gwrt->rt_llinfo_get_iflri != NULL) { | |
469 | (*gwrt->rt_llinfo_get_iflri)(gwrt, iflri); | |
470 | error = 0; | |
471 | } | |
472 | RT_UNLOCK(gwrt); | |
473 | } else { | |
474 | RT_UNLOCK(rt); | |
475 | } | |
476 | rtfree_locked(rt); | |
477 | } | |
478 | lck_mtx_unlock(rnh_lock); | |
479 | ||
0a7de745 | 480 | return error; |
316670eb A |
481 | } |
482 | ||
6d2010ae A |
483 | static struct if_llreach * |
484 | iflr_alloc(int how) | |
485 | { | |
486 | struct if_llreach *lr; | |
487 | ||
488 | lr = (how == M_WAITOK) ? zalloc(iflr_zone) : zalloc_noblock(iflr_zone); | |
489 | if (lr != NULL) { | |
490 | bzero(lr, iflr_size); | |
491 | lck_mtx_init(&lr->lr_lock, ifnet_lock_group, ifnet_lock_attr); | |
492 | lr->lr_debug |= IFD_ALLOC; | |
493 | } | |
0a7de745 | 494 | return lr; |
6d2010ae A |
495 | } |
496 | ||
497 | static void | |
498 | iflr_free(struct if_llreach *lr) | |
499 | { | |
500 | IFLR_LOCK(lr); | |
501 | if (lr->lr_debug & IFD_ATTACHED) { | |
502 | panic("%s: attached lr=%p is being freed", __func__, lr); | |
503 | /* NOTREACHED */ | |
504 | } else if (!(lr->lr_debug & IFD_ALLOC)) { | |
505 | panic("%s: lr %p cannot be freed", __func__, lr); | |
506 | /* NOTREACHED */ | |
507 | } else if (lr->lr_refcnt != 0) { | |
508 | panic("%s: non-zero refcount lr=%p", __func__, lr); | |
509 | /* NOTREACHED */ | |
510 | } else if (lr->lr_reqcnt != 0) { | |
511 | panic("%s: non-zero reqcnt lr=%p", __func__, lr); | |
512 | /* NOTREACHED */ | |
513 | } | |
514 | lr->lr_debug &= ~IFD_ALLOC; | |
515 | IFLR_UNLOCK(lr); | |
516 | ||
517 | lck_mtx_destroy(&lr->lr_lock, ifnet_lock_group); | |
518 | zfree(iflr_zone, lr); | |
519 | } | |
520 | ||
521 | void | |
522 | iflr_addref(struct if_llreach *lr, int locked) | |
523 | { | |
0a7de745 | 524 | if (!locked) { |
6d2010ae | 525 | IFLR_LOCK(lr); |
0a7de745 | 526 | } else { |
6d2010ae | 527 | IFLR_LOCK_ASSERT_HELD(lr); |
0a7de745 | 528 | } |
6d2010ae A |
529 | |
530 | if (++lr->lr_refcnt == 0) { | |
531 | panic("%s: lr=%p wraparound refcnt", __func__, lr); | |
532 | /* NOTREACHED */ | |
533 | } | |
0a7de745 | 534 | if (!locked) { |
6d2010ae | 535 | IFLR_UNLOCK(lr); |
0a7de745 | 536 | } |
6d2010ae A |
537 | } |
538 | ||
539 | void | |
540 | iflr_remref(struct if_llreach *lr) | |
541 | { | |
542 | IFLR_LOCK(lr); | |
543 | if (lr->lr_refcnt == 0) { | |
544 | panic("%s: lr=%p negative refcnt", __func__, lr); | |
545 | /* NOTREACHED */ | |
546 | } | |
547 | --lr->lr_refcnt; | |
548 | if (lr->lr_refcnt > 0) { | |
549 | IFLR_UNLOCK(lr); | |
550 | return; | |
551 | } | |
552 | IFLR_UNLOCK(lr); | |
553 | ||
0a7de745 | 554 | iflr_free(lr); /* deallocate it */ |
6d2010ae A |
555 | } |
556 | ||
557 | void | |
558 | ifnet_lr2ri(struct if_llreach *lr, struct rt_reach_info *ri) | |
559 | { | |
560 | struct if_llreach_info lri; | |
561 | ||
562 | IFLR_LOCK_ASSERT_HELD(lr); | |
563 | ||
0a7de745 | 564 | bzero(ri, sizeof(*ri)); |
6d2010ae A |
565 | ifnet_lr2lri(lr, &lri); |
566 | ri->ri_refcnt = lri.lri_refcnt; | |
567 | ri->ri_probes = lri.lri_probes; | |
568 | ri->ri_rcv_expire = lri.lri_expire; | |
316670eb A |
569 | ri->ri_rssi = lri.lri_rssi; |
570 | ri->ri_lqm = lri.lri_lqm; | |
571 | ri->ri_npm = lri.lri_npm; | |
572 | } | |
573 | ||
574 | void | |
575 | ifnet_lr2iflri(struct if_llreach *lr, struct ifnet_llreach_info *iflri) | |
576 | { | |
577 | IFLR_LOCK_ASSERT_HELD(lr); | |
578 | ||
0a7de745 | 579 | bzero(iflri, sizeof(*iflri)); |
316670eb A |
580 | /* |
581 | * Note here we return request count, not actual memory refcnt. | |
582 | */ | |
583 | iflri->iflri_refcnt = lr->lr_reqcnt; | |
584 | iflri->iflri_probes = lr->lr_probes; | |
585 | iflri->iflri_rcv_expire = ifnet_llreach_up2upexp(lr, lr->lr_lastrcvd); | |
586 | iflri->iflri_curtime = net_uptime(); | |
587 | switch (lr->lr_key.proto) { | |
588 | case ETHERTYPE_IP: | |
589 | iflri->iflri_netproto = PF_INET; | |
590 | break; | |
591 | case ETHERTYPE_IPV6: | |
592 | iflri->iflri_netproto = PF_INET6; | |
593 | break; | |
594 | default: | |
595 | /* | |
596 | * This shouldn't be possible for the time being, | |
597 | * since link-layer reachability records are only | |
598 | * kept for ARP and ND6. | |
599 | */ | |
600 | iflri->iflri_netproto = PF_UNSPEC; | |
601 | break; | |
602 | } | |
603 | bcopy(&lr->lr_key.addr, &iflri->iflri_addr, IF_LLREACH_MAXLEN); | |
604 | iflri->iflri_rssi = lr->lr_rssi; | |
605 | iflri->iflri_lqm = lr->lr_lqm; | |
606 | iflri->iflri_npm = lr->lr_npm; | |
6d2010ae A |
607 | } |
608 | ||
609 | void | |
610 | ifnet_lr2lri(struct if_llreach *lr, struct if_llreach_info *lri) | |
611 | { | |
612 | IFLR_LOCK_ASSERT_HELD(lr); | |
613 | ||
0a7de745 | 614 | bzero(lri, sizeof(*lri)); |
6d2010ae A |
615 | /* |
616 | * Note here we return request count, not actual memory refcnt. | |
617 | */ | |
0a7de745 | 618 | lri->lri_refcnt = lr->lr_reqcnt; |
6d2010ae | 619 | lri->lri_ifindex = lr->lr_ifp->if_index; |
0a7de745 | 620 | lri->lri_probes = lr->lr_probes; |
316670eb | 621 | lri->lri_expire = ifnet_llreach_up2calexp(lr, lr->lr_lastrcvd); |
6d2010ae A |
622 | lri->lri_proto = lr->lr_key.proto; |
623 | bcopy(&lr->lr_key.addr, &lri->lri_addr, IF_LLREACH_MAXLEN); | |
316670eb A |
624 | lri->lri_rssi = lr->lr_rssi; |
625 | lri->lri_lqm = lr->lr_lqm; | |
626 | lri->lri_npm = lr->lr_npm; | |
6d2010ae A |
627 | } |
628 | ||
629 | static int | |
630 | sysctl_llreach_ifinfo SYSCTL_HANDLER_ARGS | |
631 | { | |
632 | #pragma unused(oidp) | |
0a7de745 A |
633 | int *name, retval = 0; |
634 | unsigned int namelen; | |
635 | uint32_t ifindex; | |
6d2010ae | 636 | struct if_llreach *lr; |
527f9951 | 637 | struct if_llreach_info lri = {}; |
0a7de745 | 638 | struct ifnet *ifp; |
6d2010ae A |
639 | |
640 | name = (int *)arg1; | |
641 | namelen = (unsigned int)arg2; | |
642 | ||
0a7de745 A |
643 | if (req->newptr != USER_ADDR_NULL) { |
644 | return EPERM; | |
645 | } | |
6d2010ae | 646 | |
0a7de745 A |
647 | if (namelen != 1) { |
648 | return EINVAL; | |
649 | } | |
6d2010ae A |
650 | |
651 | ifindex = name[0]; | |
652 | ifnet_head_lock_shared(); | |
653 | if (ifindex <= 0 || ifindex > (u_int)if_index) { | |
654 | printf("%s: ifindex %u out of range\n", __func__, ifindex); | |
655 | ifnet_head_done(); | |
0a7de745 | 656 | return ENOENT; |
6d2010ae A |
657 | } |
658 | ||
659 | ifp = ifindex2ifnet[ifindex]; | |
660 | ifnet_head_done(); | |
661 | if (ifp == NULL) { | |
662 | printf("%s: no ifp for ifindex %u\n", __func__, ifindex); | |
0a7de745 | 663 | return ENOENT; |
6d2010ae A |
664 | } |
665 | ||
666 | lck_rw_lock_shared(&ifp->if_llreach_lock); | |
667 | RB_FOREACH(lr, ll_reach_tree, &ifp->if_ll_srcs) { | |
668 | /* Export to if_llreach_info structure */ | |
669 | IFLR_LOCK(lr); | |
670 | ifnet_lr2lri(lr, &lri); | |
671 | IFLR_UNLOCK(lr); | |
672 | ||
0a7de745 | 673 | if ((retval = SYSCTL_OUT(req, &lri, sizeof(lri))) != 0) { |
6d2010ae | 674 | break; |
0a7de745 | 675 | } |
6d2010ae A |
676 | } |
677 | lck_rw_done(&ifp->if_llreach_lock); | |
678 | ||
0a7de745 | 679 | return retval; |
6d2010ae | 680 | } |