]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (c) 2011-2012 Apple Inc. All rights reserved. | |
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 | ||
120 | #include <dev/random/randomdev.h> | |
121 | ||
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> | |
127 | #include <net/kpi_interface.h> | |
128 | #include <net/route.h> | |
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 | ||
139 | static unsigned int iflr_size; /* size of if_llreach */ | |
140 | static struct zone *iflr_zone; /* zone for if_llreach */ | |
141 | ||
142 | #define IFLR_ZONE_MAX 128 /* maximum elements in zone */ | |
143 | #define IFLR_ZONE_NAME "if_llreach" /* zone name */ | |
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 | |
165 | #define LL_COMPUTE_RTIME(x) ND_COMPUTE_RTIME(x) | |
166 | #else | |
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() & \ | |
171 | ((LL_MAX_RANDOM_FACTOR - LL_MIN_RANDOM_FACTOR) * (x >> 10)))) / 1000) | |
172 | #endif /* !INET6 */ | |
173 | ||
174 | void | |
175 | ifnet_llreach_init(void) | |
176 | { | |
177 | iflr_size = sizeof (struct if_llreach); | |
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) */ | |
193 | if (!reuse) | |
194 | RB_INIT(&ifp->if_ll_srcs); | |
195 | lck_rw_done(&ifp->if_llreach_lock); | |
196 | } | |
197 | ||
198 | void | |
199 | ifnet_llreach_ifdetach(struct ifnet *ifp) | |
200 | { | |
201 | #pragma unused(ifp) | |
202 | /* | |
203 | * Nothing to do for now; the link-layer source tree might | |
204 | * contain entries at this point, that are still referred | |
205 | * to by route entries pointing to this ifp. | |
206 | */ | |
207 | } | |
208 | ||
209 | /* | |
210 | * Link-layer source tree comparison function. | |
211 | * | |
212 | * An ordered predicate is necessary; bcmp() is not documented to return | |
213 | * an indication of order, memcmp() is, and is an ISO C99 requirement. | |
214 | */ | |
215 | static __inline int | |
216 | iflr_cmp(const struct if_llreach *a, const struct if_llreach *b) | |
217 | { | |
218 | return (memcmp(&a->lr_key, &b->lr_key, sizeof (a->lr_key))); | |
219 | } | |
220 | ||
221 | static __inline int | |
222 | iflr_reachable(struct if_llreach *lr, int cmp_delta, u_int64_t tval) | |
223 | { | |
224 | u_int64_t now; | |
225 | u_int64_t expire; | |
226 | ||
227 | now = net_uptime(); /* current approx. uptime */ | |
228 | /* | |
229 | * No need for lr_lock; atomically read the last rcvd uptime. | |
230 | */ | |
231 | expire = lr->lr_lastrcvd + lr->lr_reachable; | |
232 | /* | |
233 | * If we haven't heard back from the local host for over | |
234 | * lr_reachable seconds, consider that the host is no | |
235 | * longer reachable. | |
236 | */ | |
237 | if (!cmp_delta) | |
238 | return (expire >= now); | |
239 | /* | |
240 | * If the caller supplied a reference time, consider the | |
241 | * host is reachable if the record hasn't expired (see above) | |
242 | * and if the reference time is within the past lr_reachable | |
243 | * seconds. | |
244 | */ | |
245 | return ((expire >= now) && (now - tval) < lr->lr_reachable); | |
246 | } | |
247 | ||
248 | int | |
249 | ifnet_llreach_reachable(struct if_llreach *lr) | |
250 | { | |
251 | /* | |
252 | * Check whether the cache is too old to be trusted. | |
253 | */ | |
254 | return (iflr_reachable(lr, 0, 0)); | |
255 | } | |
256 | ||
257 | int | |
258 | ifnet_llreach_reachable_delta(struct if_llreach *lr, u_int64_t tval) | |
259 | { | |
260 | /* | |
261 | * Check whether the cache is too old to be trusted. | |
262 | */ | |
263 | return (iflr_reachable(lr, 1, tval)); | |
264 | } | |
265 | ||
266 | void | |
267 | ifnet_llreach_set_reachable(struct ifnet *ifp, u_int16_t llproto, void *addr, | |
268 | unsigned int alen) | |
269 | { | |
270 | struct if_llreach find, *lr; | |
271 | ||
272 | VERIFY(alen == IF_LLREACH_MAXLEN); /* for now */ | |
273 | ||
274 | find.lr_key.proto = llproto; | |
275 | bcopy(addr, &find.lr_key.addr, IF_LLREACH_MAXLEN); | |
276 | ||
277 | lck_rw_lock_shared(&ifp->if_llreach_lock); | |
278 | lr = RB_FIND(ll_reach_tree, &ifp->if_ll_srcs, &find); | |
279 | if (lr == NULL) { | |
280 | lck_rw_done(&ifp->if_llreach_lock); | |
281 | return; | |
282 | } | |
283 | /* | |
284 | * No need for lr_lock; atomically update the last rcvd uptime. | |
285 | */ | |
286 | lr->lr_lastrcvd = net_uptime(); | |
287 | lck_rw_done(&ifp->if_llreach_lock); | |
288 | } | |
289 | ||
290 | struct if_llreach * | |
291 | ifnet_llreach_alloc(struct ifnet *ifp, u_int16_t llproto, void *addr, | |
292 | unsigned int alen, u_int64_t llreach_base) | |
293 | { | |
294 | struct if_llreach find, *lr; | |
295 | struct timeval cnow; | |
296 | ||
297 | if (llreach_base == 0) | |
298 | return (NULL); | |
299 | ||
300 | VERIFY(alen == IF_LLREACH_MAXLEN); /* for now */ | |
301 | ||
302 | find.lr_key.proto = llproto; | |
303 | bcopy(addr, &find.lr_key.addr, IF_LLREACH_MAXLEN); | |
304 | ||
305 | lck_rw_lock_shared(&ifp->if_llreach_lock); | |
306 | lr = RB_FIND(ll_reach_tree, &ifp->if_ll_srcs, &find); | |
307 | if (lr != NULL) { | |
308 | found: | |
309 | IFLR_LOCK(lr); | |
310 | VERIFY(lr->lr_reqcnt >= 1); | |
311 | lr->lr_reqcnt++; | |
312 | VERIFY(lr->lr_reqcnt != 0); | |
313 | IFLR_ADDREF_LOCKED(lr); /* for caller */ | |
314 | lr->lr_lastrcvd = net_uptime(); /* current approx. uptime */ | |
315 | IFLR_UNLOCK(lr); | |
316 | lck_rw_done(&ifp->if_llreach_lock); | |
317 | return (lr); | |
318 | } | |
319 | ||
320 | if (!lck_rw_lock_shared_to_exclusive(&ifp->if_llreach_lock)) | |
321 | lck_rw_lock_exclusive(&ifp->if_llreach_lock); | |
322 | ||
323 | lck_rw_assert(&ifp->if_llreach_lock, LCK_RW_ASSERT_EXCLUSIVE); | |
324 | ||
325 | /* in case things have changed while becoming writer */ | |
326 | lr = RB_FIND(ll_reach_tree, &ifp->if_ll_srcs, &find); | |
327 | if (lr != NULL) | |
328 | goto found; | |
329 | ||
330 | lr = iflr_alloc(M_WAITOK); | |
331 | if (lr == NULL) { | |
332 | lck_rw_done(&ifp->if_llreach_lock); | |
333 | return (NULL); | |
334 | } | |
335 | IFLR_LOCK(lr); | |
336 | lr->lr_reqcnt++; | |
337 | VERIFY(lr->lr_reqcnt == 1); | |
338 | IFLR_ADDREF_LOCKED(lr); /* for RB tree */ | |
339 | IFLR_ADDREF_LOCKED(lr); /* for caller */ | |
340 | lr->lr_lastrcvd = net_uptime(); /* current approx. uptime */ | |
341 | lr->lr_baseup = lr->lr_lastrcvd; /* base uptime */ | |
342 | getmicrotime(&cnow); | |
343 | lr->lr_basecal = cnow.tv_sec; /* base calendar time */ | |
344 | lr->lr_basereachable = llreach_base; | |
345 | lr->lr_reachable = LL_COMPUTE_RTIME(lr->lr_basereachable * 1000); | |
346 | lr->lr_debug |= IFD_ATTACHED; | |
347 | lr->lr_ifp = ifp; | |
348 | lr->lr_key.proto = llproto; | |
349 | bcopy(addr, &lr->lr_key.addr, IF_LLREACH_MAXLEN); | |
350 | lr->lr_rssi = IFNET_RSSI_UNKNOWN; | |
351 | lr->lr_lqm = IFNET_LQM_THRESH_UNKNOWN; | |
352 | lr->lr_npm = IFNET_NPM_THRESH_UNKNOWN; | |
353 | RB_INSERT(ll_reach_tree, &ifp->if_ll_srcs, lr); | |
354 | IFLR_UNLOCK(lr); | |
355 | lck_rw_done(&ifp->if_llreach_lock); | |
356 | ||
357 | return (lr); | |
358 | } | |
359 | ||
360 | void | |
361 | ifnet_llreach_free(struct if_llreach *lr) | |
362 | { | |
363 | struct ifnet *ifp; | |
364 | ||
365 | /* no need to lock here; lr_ifp never changes */ | |
366 | ifp = lr->lr_ifp; | |
367 | ||
368 | lck_rw_lock_exclusive(&ifp->if_llreach_lock); | |
369 | IFLR_LOCK(lr); | |
370 | if (lr->lr_reqcnt == 0) { | |
371 | panic("%s: lr=%p negative reqcnt", __func__, lr); | |
372 | /* NOTREACHED */ | |
373 | } | |
374 | --lr->lr_reqcnt; | |
375 | if (lr->lr_reqcnt > 0) { | |
376 | IFLR_UNLOCK(lr); | |
377 | lck_rw_done(&ifp->if_llreach_lock); | |
378 | IFLR_REMREF(lr); /* for caller */ | |
379 | return; | |
380 | } | |
381 | if (!(lr->lr_debug & IFD_ATTACHED)) { | |
382 | panic("%s: Attempt to detach an unattached llreach lr=%p", | |
383 | __func__, lr); | |
384 | /* NOTREACHED */ | |
385 | } | |
386 | lr->lr_debug &= ~IFD_ATTACHED; | |
387 | RB_REMOVE(ll_reach_tree, &ifp->if_ll_srcs, lr); | |
388 | IFLR_UNLOCK(lr); | |
389 | lck_rw_done(&ifp->if_llreach_lock); | |
390 | ||
391 | IFLR_REMREF(lr); /* for RB tree */ | |
392 | IFLR_REMREF(lr); /* for caller */ | |
393 | } | |
394 | ||
395 | u_int64_t | |
396 | ifnet_llreach_up2calexp(struct if_llreach *lr, u_int64_t uptime) | |
397 | { | |
398 | u_int64_t calendar = 0; | |
399 | ||
400 | if (uptime != 0) { | |
401 | struct timeval cnow; | |
402 | u_int64_t unow; | |
403 | ||
404 | getmicrotime(&cnow); /* current calendar time */ | |
405 | unow = net_uptime(); /* current approx. uptime */ | |
406 | /* | |
407 | * Take into account possible calendar time changes; | |
408 | * adjust base calendar value if necessary, i.e. | |
409 | * the calendar skew should equate to the uptime skew. | |
410 | */ | |
411 | lr->lr_basecal += (cnow.tv_sec - lr->lr_basecal) - | |
412 | (unow - lr->lr_baseup); | |
413 | ||
414 | calendar = lr->lr_basecal + lr->lr_reachable + | |
415 | (uptime - lr->lr_baseup); | |
416 | } | |
417 | ||
418 | return (calendar); | |
419 | } | |
420 | ||
421 | u_int64_t | |
422 | ifnet_llreach_up2upexp(struct if_llreach *lr, u_int64_t uptime) | |
423 | { | |
424 | return (lr->lr_reachable + uptime); | |
425 | } | |
426 | ||
427 | int | |
428 | ifnet_llreach_get_defrouter(struct ifnet *ifp, int af, | |
429 | struct ifnet_llreach_info *iflri) | |
430 | { | |
431 | struct radix_node_head *rnh; | |
432 | struct sockaddr_storage dst_ss, mask_ss; | |
433 | struct rtentry *rt; | |
434 | int error = ESRCH; | |
435 | ||
436 | VERIFY(ifp != NULL && iflri != NULL && | |
437 | (af == AF_INET || af == AF_INET6)); | |
438 | ||
439 | bzero(iflri, sizeof (*iflri)); | |
440 | ||
441 | if ((rnh = rt_tables[af]) == NULL) | |
442 | return (error); | |
443 | ||
444 | bzero(&dst_ss, sizeof (dst_ss)); | |
445 | bzero(&mask_ss, sizeof (mask_ss)); | |
446 | dst_ss.ss_family = af; | |
447 | dst_ss.ss_len = (af == AF_INET) ? sizeof (struct sockaddr_in) : | |
448 | sizeof (struct sockaddr_in6); | |
449 | ||
450 | lck_mtx_lock(rnh_lock); | |
451 | rt = rt_lookup(TRUE, SA(&dst_ss), SA(&mask_ss), rnh, ifp->if_index); | |
452 | if (rt != NULL) { | |
453 | struct rtentry *gwrt; | |
454 | ||
455 | RT_LOCK(rt); | |
456 | if ((rt->rt_flags & RTF_GATEWAY) && | |
457 | (gwrt = rt->rt_gwroute) != NULL && | |
458 | rt_key(rt)->sa_family == rt_key(gwrt)->sa_family && | |
459 | (gwrt->rt_flags & RTF_UP)) { | |
460 | RT_UNLOCK(rt); | |
461 | RT_LOCK(gwrt); | |
462 | if (gwrt->rt_llinfo_get_iflri != NULL) { | |
463 | (*gwrt->rt_llinfo_get_iflri)(gwrt, iflri); | |
464 | error = 0; | |
465 | } | |
466 | RT_UNLOCK(gwrt); | |
467 | } else { | |
468 | RT_UNLOCK(rt); | |
469 | } | |
470 | rtfree_locked(rt); | |
471 | } | |
472 | lck_mtx_unlock(rnh_lock); | |
473 | ||
474 | return (error); | |
475 | } | |
476 | ||
477 | static struct if_llreach * | |
478 | iflr_alloc(int how) | |
479 | { | |
480 | struct if_llreach *lr; | |
481 | ||
482 | lr = (how == M_WAITOK) ? zalloc(iflr_zone) : zalloc_noblock(iflr_zone); | |
483 | if (lr != NULL) { | |
484 | bzero(lr, iflr_size); | |
485 | lck_mtx_init(&lr->lr_lock, ifnet_lock_group, ifnet_lock_attr); | |
486 | lr->lr_debug |= IFD_ALLOC; | |
487 | } | |
488 | return (lr); | |
489 | } | |
490 | ||
491 | static void | |
492 | iflr_free(struct if_llreach *lr) | |
493 | { | |
494 | IFLR_LOCK(lr); | |
495 | if (lr->lr_debug & IFD_ATTACHED) { | |
496 | panic("%s: attached lr=%p is being freed", __func__, lr); | |
497 | /* NOTREACHED */ | |
498 | } else if (!(lr->lr_debug & IFD_ALLOC)) { | |
499 | panic("%s: lr %p cannot be freed", __func__, lr); | |
500 | /* NOTREACHED */ | |
501 | } else if (lr->lr_refcnt != 0) { | |
502 | panic("%s: non-zero refcount lr=%p", __func__, lr); | |
503 | /* NOTREACHED */ | |
504 | } else if (lr->lr_reqcnt != 0) { | |
505 | panic("%s: non-zero reqcnt lr=%p", __func__, lr); | |
506 | /* NOTREACHED */ | |
507 | } | |
508 | lr->lr_debug &= ~IFD_ALLOC; | |
509 | IFLR_UNLOCK(lr); | |
510 | ||
511 | lck_mtx_destroy(&lr->lr_lock, ifnet_lock_group); | |
512 | zfree(iflr_zone, lr); | |
513 | } | |
514 | ||
515 | void | |
516 | iflr_addref(struct if_llreach *lr, int locked) | |
517 | { | |
518 | if (!locked) | |
519 | IFLR_LOCK(lr); | |
520 | else | |
521 | IFLR_LOCK_ASSERT_HELD(lr); | |
522 | ||
523 | if (++lr->lr_refcnt == 0) { | |
524 | panic("%s: lr=%p wraparound refcnt", __func__, lr); | |
525 | /* NOTREACHED */ | |
526 | } | |
527 | if (!locked) | |
528 | IFLR_UNLOCK(lr); | |
529 | } | |
530 | ||
531 | void | |
532 | iflr_remref(struct if_llreach *lr) | |
533 | { | |
534 | IFLR_LOCK(lr); | |
535 | if (lr->lr_refcnt == 0) { | |
536 | panic("%s: lr=%p negative refcnt", __func__, lr); | |
537 | /* NOTREACHED */ | |
538 | } | |
539 | --lr->lr_refcnt; | |
540 | if (lr->lr_refcnt > 0) { | |
541 | IFLR_UNLOCK(lr); | |
542 | return; | |
543 | } | |
544 | IFLR_UNLOCK(lr); | |
545 | ||
546 | iflr_free(lr); /* deallocate it */ | |
547 | } | |
548 | ||
549 | void | |
550 | ifnet_lr2ri(struct if_llreach *lr, struct rt_reach_info *ri) | |
551 | { | |
552 | struct if_llreach_info lri; | |
553 | ||
554 | IFLR_LOCK_ASSERT_HELD(lr); | |
555 | ||
556 | bzero(ri, sizeof (*ri)); | |
557 | ifnet_lr2lri(lr, &lri); | |
558 | ri->ri_refcnt = lri.lri_refcnt; | |
559 | ri->ri_probes = lri.lri_probes; | |
560 | ri->ri_rcv_expire = lri.lri_expire; | |
561 | ri->ri_rssi = lri.lri_rssi; | |
562 | ri->ri_lqm = lri.lri_lqm; | |
563 | ri->ri_npm = lri.lri_npm; | |
564 | } | |
565 | ||
566 | void | |
567 | ifnet_lr2iflri(struct if_llreach *lr, struct ifnet_llreach_info *iflri) | |
568 | { | |
569 | IFLR_LOCK_ASSERT_HELD(lr); | |
570 | ||
571 | bzero(iflri, sizeof (*iflri)); | |
572 | /* | |
573 | * Note here we return request count, not actual memory refcnt. | |
574 | */ | |
575 | iflri->iflri_refcnt = lr->lr_reqcnt; | |
576 | iflri->iflri_probes = lr->lr_probes; | |
577 | iflri->iflri_rcv_expire = ifnet_llreach_up2upexp(lr, lr->lr_lastrcvd); | |
578 | iflri->iflri_curtime = net_uptime(); | |
579 | switch (lr->lr_key.proto) { | |
580 | case ETHERTYPE_IP: | |
581 | iflri->iflri_netproto = PF_INET; | |
582 | break; | |
583 | case ETHERTYPE_IPV6: | |
584 | iflri->iflri_netproto = PF_INET6; | |
585 | break; | |
586 | default: | |
587 | /* | |
588 | * This shouldn't be possible for the time being, | |
589 | * since link-layer reachability records are only | |
590 | * kept for ARP and ND6. | |
591 | */ | |
592 | iflri->iflri_netproto = PF_UNSPEC; | |
593 | break; | |
594 | } | |
595 | bcopy(&lr->lr_key.addr, &iflri->iflri_addr, IF_LLREACH_MAXLEN); | |
596 | iflri->iflri_rssi = lr->lr_rssi; | |
597 | iflri->iflri_lqm = lr->lr_lqm; | |
598 | iflri->iflri_npm = lr->lr_npm; | |
599 | } | |
600 | ||
601 | void | |
602 | ifnet_lr2lri(struct if_llreach *lr, struct if_llreach_info *lri) | |
603 | { | |
604 | IFLR_LOCK_ASSERT_HELD(lr); | |
605 | ||
606 | bzero(lri, sizeof (*lri)); | |
607 | /* | |
608 | * Note here we return request count, not actual memory refcnt. | |
609 | */ | |
610 | lri->lri_refcnt = lr->lr_reqcnt; | |
611 | lri->lri_ifindex = lr->lr_ifp->if_index; | |
612 | lri->lri_probes = lr->lr_probes; | |
613 | lri->lri_expire = ifnet_llreach_up2calexp(lr, lr->lr_lastrcvd); | |
614 | lri->lri_proto = lr->lr_key.proto; | |
615 | bcopy(&lr->lr_key.addr, &lri->lri_addr, IF_LLREACH_MAXLEN); | |
616 | lri->lri_rssi = lr->lr_rssi; | |
617 | lri->lri_lqm = lr->lr_lqm; | |
618 | lri->lri_npm = lr->lr_npm; | |
619 | } | |
620 | ||
621 | static int | |
622 | sysctl_llreach_ifinfo SYSCTL_HANDLER_ARGS | |
623 | { | |
624 | #pragma unused(oidp) | |
625 | int *name, retval = 0; | |
626 | unsigned int namelen; | |
627 | uint32_t ifindex; | |
628 | struct if_llreach *lr; | |
629 | struct if_llreach_info lri; | |
630 | struct ifnet *ifp; | |
631 | ||
632 | name = (int *)arg1; | |
633 | namelen = (unsigned int)arg2; | |
634 | ||
635 | if (req->newptr != USER_ADDR_NULL) | |
636 | return (EPERM); | |
637 | ||
638 | if (namelen != 1) | |
639 | return (EINVAL); | |
640 | ||
641 | ifindex = name[0]; | |
642 | ifnet_head_lock_shared(); | |
643 | if (ifindex <= 0 || ifindex > (u_int)if_index) { | |
644 | printf("%s: ifindex %u out of range\n", __func__, ifindex); | |
645 | ifnet_head_done(); | |
646 | return (ENOENT); | |
647 | } | |
648 | ||
649 | ifp = ifindex2ifnet[ifindex]; | |
650 | ifnet_head_done(); | |
651 | if (ifp == NULL) { | |
652 | printf("%s: no ifp for ifindex %u\n", __func__, ifindex); | |
653 | return (ENOENT); | |
654 | } | |
655 | ||
656 | lck_rw_lock_shared(&ifp->if_llreach_lock); | |
657 | RB_FOREACH(lr, ll_reach_tree, &ifp->if_ll_srcs) { | |
658 | /* Export to if_llreach_info structure */ | |
659 | IFLR_LOCK(lr); | |
660 | ifnet_lr2lri(lr, &lri); | |
661 | IFLR_UNLOCK(lr); | |
662 | ||
663 | if ((retval = SYSCTL_OUT(req, &lri, sizeof (lri))) != 0) | |
664 | break; | |
665 | } | |
666 | lck_rw_done(&ifp->if_llreach_lock); | |
667 | ||
668 | return (retval); | |
669 | } |