]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * xfrm_policy.c | |
3 | * | |
4 | * Changes: | |
5 | * Mitsuru KANDA @USAGI | |
6 | * Kazunori MIYAZAWA @USAGI | |
7 | * Kunihiro Ishiguro <kunihiro@ipinfusion.com> | |
8 | * IPv6 support | |
9 | * Kazunori MIYAZAWA @USAGI | |
10 | * YOSHIFUJI Hideaki | |
11 | * Split up af-specific portion | |
12 | * Derek Atkins <derek@ihtfp.com> Add the post_input processor | |
df71837d | 13 | * |
1da177e4 LT |
14 | */ |
15 | ||
1da177e4 LT |
16 | #include <linux/slab.h> |
17 | #include <linux/kmod.h> | |
18 | #include <linux/list.h> | |
19 | #include <linux/spinlock.h> | |
20 | #include <linux/workqueue.h> | |
21 | #include <linux/notifier.h> | |
22 | #include <linux/netdevice.h> | |
eb9c7ebe | 23 | #include <linux/netfilter.h> |
1da177e4 | 24 | #include <linux/module.h> |
2518c7c2 | 25 | #include <linux/cache.h> |
1da177e4 LT |
26 | #include <net/xfrm.h> |
27 | #include <net/ip.h> | |
28 | ||
44e36b42 DM |
29 | #include "xfrm_hash.h" |
30 | ||
4a3e2f71 AV |
31 | DEFINE_MUTEX(xfrm_cfg_mutex); |
32 | EXPORT_SYMBOL(xfrm_cfg_mutex); | |
1da177e4 LT |
33 | |
34 | static DEFINE_RWLOCK(xfrm_policy_lock); | |
35 | ||
2518c7c2 DM |
36 | unsigned int xfrm_policy_count[XFRM_POLICY_MAX*2]; |
37 | EXPORT_SYMBOL(xfrm_policy_count); | |
1da177e4 LT |
38 | |
39 | static DEFINE_RWLOCK(xfrm_policy_afinfo_lock); | |
40 | static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO]; | |
41 | ||
ba89966c | 42 | static kmem_cache_t *xfrm_dst_cache __read_mostly; |
1da177e4 LT |
43 | |
44 | static struct work_struct xfrm_policy_gc_work; | |
2518c7c2 | 45 | static HLIST_HEAD(xfrm_policy_gc_list); |
1da177e4 LT |
46 | static DEFINE_SPINLOCK(xfrm_policy_gc_lock); |
47 | ||
48 | static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family); | |
49 | static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo); | |
546be240 HX |
50 | static struct xfrm_policy_afinfo *xfrm_policy_lock_afinfo(unsigned int family); |
51 | static void xfrm_policy_unlock_afinfo(struct xfrm_policy_afinfo *afinfo); | |
1da177e4 LT |
52 | |
53 | int xfrm_register_type(struct xfrm_type *type, unsigned short family) | |
54 | { | |
546be240 HX |
55 | struct xfrm_policy_afinfo *afinfo = xfrm_policy_lock_afinfo(family); |
56 | struct xfrm_type **typemap; | |
1da177e4 LT |
57 | int err = 0; |
58 | ||
59 | if (unlikely(afinfo == NULL)) | |
60 | return -EAFNOSUPPORT; | |
61 | typemap = afinfo->type_map; | |
62 | ||
546be240 HX |
63 | if (likely(typemap[type->proto] == NULL)) |
64 | typemap[type->proto] = type; | |
1da177e4 LT |
65 | else |
66 | err = -EEXIST; | |
546be240 | 67 | xfrm_policy_unlock_afinfo(afinfo); |
1da177e4 LT |
68 | return err; |
69 | } | |
70 | EXPORT_SYMBOL(xfrm_register_type); | |
71 | ||
72 | int xfrm_unregister_type(struct xfrm_type *type, unsigned short family) | |
73 | { | |
546be240 HX |
74 | struct xfrm_policy_afinfo *afinfo = xfrm_policy_lock_afinfo(family); |
75 | struct xfrm_type **typemap; | |
1da177e4 LT |
76 | int err = 0; |
77 | ||
78 | if (unlikely(afinfo == NULL)) | |
79 | return -EAFNOSUPPORT; | |
80 | typemap = afinfo->type_map; | |
81 | ||
546be240 | 82 | if (unlikely(typemap[type->proto] != type)) |
1da177e4 LT |
83 | err = -ENOENT; |
84 | else | |
546be240 HX |
85 | typemap[type->proto] = NULL; |
86 | xfrm_policy_unlock_afinfo(afinfo); | |
1da177e4 LT |
87 | return err; |
88 | } | |
89 | EXPORT_SYMBOL(xfrm_unregister_type); | |
90 | ||
91 | struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family) | |
92 | { | |
93 | struct xfrm_policy_afinfo *afinfo; | |
546be240 | 94 | struct xfrm_type **typemap; |
1da177e4 LT |
95 | struct xfrm_type *type; |
96 | int modload_attempted = 0; | |
97 | ||
98 | retry: | |
99 | afinfo = xfrm_policy_get_afinfo(family); | |
100 | if (unlikely(afinfo == NULL)) | |
101 | return NULL; | |
102 | typemap = afinfo->type_map; | |
103 | ||
546be240 | 104 | type = typemap[proto]; |
1da177e4 LT |
105 | if (unlikely(type && !try_module_get(type->owner))) |
106 | type = NULL; | |
1da177e4 LT |
107 | if (!type && !modload_attempted) { |
108 | xfrm_policy_put_afinfo(afinfo); | |
109 | request_module("xfrm-type-%d-%d", | |
110 | (int) family, (int) proto); | |
111 | modload_attempted = 1; | |
112 | goto retry; | |
113 | } | |
114 | ||
115 | xfrm_policy_put_afinfo(afinfo); | |
116 | return type; | |
117 | } | |
1da177e4 LT |
118 | |
119 | int xfrm_dst_lookup(struct xfrm_dst **dst, struct flowi *fl, | |
120 | unsigned short family) | |
121 | { | |
122 | struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); | |
123 | int err = 0; | |
124 | ||
125 | if (unlikely(afinfo == NULL)) | |
126 | return -EAFNOSUPPORT; | |
127 | ||
128 | if (likely(afinfo->dst_lookup != NULL)) | |
129 | err = afinfo->dst_lookup(dst, fl); | |
130 | else | |
131 | err = -EINVAL; | |
132 | xfrm_policy_put_afinfo(afinfo); | |
133 | return err; | |
134 | } | |
135 | EXPORT_SYMBOL(xfrm_dst_lookup); | |
136 | ||
137 | void xfrm_put_type(struct xfrm_type *type) | |
138 | { | |
139 | module_put(type->owner); | |
140 | } | |
141 | ||
b59f45d0 HX |
142 | int xfrm_register_mode(struct xfrm_mode *mode, int family) |
143 | { | |
144 | struct xfrm_policy_afinfo *afinfo; | |
145 | struct xfrm_mode **modemap; | |
146 | int err; | |
147 | ||
148 | if (unlikely(mode->encap >= XFRM_MODE_MAX)) | |
149 | return -EINVAL; | |
150 | ||
151 | afinfo = xfrm_policy_lock_afinfo(family); | |
152 | if (unlikely(afinfo == NULL)) | |
153 | return -EAFNOSUPPORT; | |
154 | ||
155 | err = -EEXIST; | |
156 | modemap = afinfo->mode_map; | |
157 | if (likely(modemap[mode->encap] == NULL)) { | |
158 | modemap[mode->encap] = mode; | |
159 | err = 0; | |
160 | } | |
161 | ||
162 | xfrm_policy_unlock_afinfo(afinfo); | |
163 | return err; | |
164 | } | |
165 | EXPORT_SYMBOL(xfrm_register_mode); | |
166 | ||
167 | int xfrm_unregister_mode(struct xfrm_mode *mode, int family) | |
168 | { | |
169 | struct xfrm_policy_afinfo *afinfo; | |
170 | struct xfrm_mode **modemap; | |
171 | int err; | |
172 | ||
173 | if (unlikely(mode->encap >= XFRM_MODE_MAX)) | |
174 | return -EINVAL; | |
175 | ||
176 | afinfo = xfrm_policy_lock_afinfo(family); | |
177 | if (unlikely(afinfo == NULL)) | |
178 | return -EAFNOSUPPORT; | |
179 | ||
180 | err = -ENOENT; | |
181 | modemap = afinfo->mode_map; | |
182 | if (likely(modemap[mode->encap] == mode)) { | |
183 | modemap[mode->encap] = NULL; | |
184 | err = 0; | |
185 | } | |
186 | ||
187 | xfrm_policy_unlock_afinfo(afinfo); | |
188 | return err; | |
189 | } | |
190 | EXPORT_SYMBOL(xfrm_unregister_mode); | |
191 | ||
192 | struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family) | |
193 | { | |
194 | struct xfrm_policy_afinfo *afinfo; | |
195 | struct xfrm_mode *mode; | |
196 | int modload_attempted = 0; | |
197 | ||
198 | if (unlikely(encap >= XFRM_MODE_MAX)) | |
199 | return NULL; | |
200 | ||
201 | retry: | |
202 | afinfo = xfrm_policy_get_afinfo(family); | |
203 | if (unlikely(afinfo == NULL)) | |
204 | return NULL; | |
205 | ||
206 | mode = afinfo->mode_map[encap]; | |
207 | if (unlikely(mode && !try_module_get(mode->owner))) | |
208 | mode = NULL; | |
209 | if (!mode && !modload_attempted) { | |
210 | xfrm_policy_put_afinfo(afinfo); | |
211 | request_module("xfrm-mode-%d-%d", family, encap); | |
212 | modload_attempted = 1; | |
213 | goto retry; | |
214 | } | |
215 | ||
216 | xfrm_policy_put_afinfo(afinfo); | |
217 | return mode; | |
218 | } | |
219 | ||
220 | void xfrm_put_mode(struct xfrm_mode *mode) | |
221 | { | |
222 | module_put(mode->owner); | |
223 | } | |
224 | ||
1da177e4 LT |
225 | static inline unsigned long make_jiffies(long secs) |
226 | { | |
227 | if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ) | |
228 | return MAX_SCHEDULE_TIMEOUT-1; | |
229 | else | |
230 | return secs*HZ; | |
231 | } | |
232 | ||
233 | static void xfrm_policy_timer(unsigned long data) | |
234 | { | |
235 | struct xfrm_policy *xp = (struct xfrm_policy*)data; | |
236 | unsigned long now = (unsigned long)xtime.tv_sec; | |
237 | long next = LONG_MAX; | |
238 | int warn = 0; | |
239 | int dir; | |
240 | ||
241 | read_lock(&xp->lock); | |
242 | ||
243 | if (xp->dead) | |
244 | goto out; | |
245 | ||
77d8d7a6 | 246 | dir = xfrm_policy_id2dir(xp->index); |
1da177e4 LT |
247 | |
248 | if (xp->lft.hard_add_expires_seconds) { | |
249 | long tmo = xp->lft.hard_add_expires_seconds + | |
250 | xp->curlft.add_time - now; | |
251 | if (tmo <= 0) | |
252 | goto expired; | |
253 | if (tmo < next) | |
254 | next = tmo; | |
255 | } | |
256 | if (xp->lft.hard_use_expires_seconds) { | |
257 | long tmo = xp->lft.hard_use_expires_seconds + | |
258 | (xp->curlft.use_time ? : xp->curlft.add_time) - now; | |
259 | if (tmo <= 0) | |
260 | goto expired; | |
261 | if (tmo < next) | |
262 | next = tmo; | |
263 | } | |
264 | if (xp->lft.soft_add_expires_seconds) { | |
265 | long tmo = xp->lft.soft_add_expires_seconds + | |
266 | xp->curlft.add_time - now; | |
267 | if (tmo <= 0) { | |
268 | warn = 1; | |
269 | tmo = XFRM_KM_TIMEOUT; | |
270 | } | |
271 | if (tmo < next) | |
272 | next = tmo; | |
273 | } | |
274 | if (xp->lft.soft_use_expires_seconds) { | |
275 | long tmo = xp->lft.soft_use_expires_seconds + | |
276 | (xp->curlft.use_time ? : xp->curlft.add_time) - now; | |
277 | if (tmo <= 0) { | |
278 | warn = 1; | |
279 | tmo = XFRM_KM_TIMEOUT; | |
280 | } | |
281 | if (tmo < next) | |
282 | next = tmo; | |
283 | } | |
284 | ||
285 | if (warn) | |
6c5c8ca7 | 286 | km_policy_expired(xp, dir, 0, 0); |
1da177e4 LT |
287 | if (next != LONG_MAX && |
288 | !mod_timer(&xp->timer, jiffies + make_jiffies(next))) | |
289 | xfrm_pol_hold(xp); | |
290 | ||
291 | out: | |
292 | read_unlock(&xp->lock); | |
293 | xfrm_pol_put(xp); | |
294 | return; | |
295 | ||
296 | expired: | |
297 | read_unlock(&xp->lock); | |
4666faab | 298 | if (!xfrm_policy_delete(xp, dir)) |
6c5c8ca7 | 299 | km_policy_expired(xp, dir, 1, 0); |
1da177e4 LT |
300 | xfrm_pol_put(xp); |
301 | } | |
302 | ||
303 | ||
304 | /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2 | |
305 | * SPD calls. | |
306 | */ | |
307 | ||
dd0fc66f | 308 | struct xfrm_policy *xfrm_policy_alloc(gfp_t gfp) |
1da177e4 LT |
309 | { |
310 | struct xfrm_policy *policy; | |
311 | ||
0da974f4 | 312 | policy = kzalloc(sizeof(struct xfrm_policy), gfp); |
1da177e4 LT |
313 | |
314 | if (policy) { | |
2518c7c2 DM |
315 | INIT_HLIST_NODE(&policy->bydst); |
316 | INIT_HLIST_NODE(&policy->byidx); | |
1da177e4 | 317 | rwlock_init(&policy->lock); |
2518c7c2 | 318 | atomic_set(&policy->refcnt, 1); |
1da177e4 LT |
319 | init_timer(&policy->timer); |
320 | policy->timer.data = (unsigned long)policy; | |
321 | policy->timer.function = xfrm_policy_timer; | |
322 | } | |
323 | return policy; | |
324 | } | |
325 | EXPORT_SYMBOL(xfrm_policy_alloc); | |
326 | ||
327 | /* Destroy xfrm_policy: descendant resources must be released to this moment. */ | |
328 | ||
329 | void __xfrm_policy_destroy(struct xfrm_policy *policy) | |
330 | { | |
09a62660 | 331 | BUG_ON(!policy->dead); |
1da177e4 | 332 | |
09a62660 | 333 | BUG_ON(policy->bundles); |
1da177e4 LT |
334 | |
335 | if (del_timer(&policy->timer)) | |
336 | BUG(); | |
337 | ||
df71837d | 338 | security_xfrm_policy_free(policy); |
1da177e4 LT |
339 | kfree(policy); |
340 | } | |
341 | EXPORT_SYMBOL(__xfrm_policy_destroy); | |
342 | ||
343 | static void xfrm_policy_gc_kill(struct xfrm_policy *policy) | |
344 | { | |
345 | struct dst_entry *dst; | |
346 | ||
347 | while ((dst = policy->bundles) != NULL) { | |
348 | policy->bundles = dst->next; | |
349 | dst_free(dst); | |
350 | } | |
351 | ||
352 | if (del_timer(&policy->timer)) | |
353 | atomic_dec(&policy->refcnt); | |
354 | ||
355 | if (atomic_read(&policy->refcnt) > 1) | |
356 | flow_cache_flush(); | |
357 | ||
358 | xfrm_pol_put(policy); | |
359 | } | |
360 | ||
361 | static void xfrm_policy_gc_task(void *data) | |
362 | { | |
363 | struct xfrm_policy *policy; | |
2518c7c2 DM |
364 | struct hlist_node *entry, *tmp; |
365 | struct hlist_head gc_list; | |
1da177e4 LT |
366 | |
367 | spin_lock_bh(&xfrm_policy_gc_lock); | |
2518c7c2 DM |
368 | gc_list.first = xfrm_policy_gc_list.first; |
369 | INIT_HLIST_HEAD(&xfrm_policy_gc_list); | |
1da177e4 LT |
370 | spin_unlock_bh(&xfrm_policy_gc_lock); |
371 | ||
2518c7c2 | 372 | hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst) |
1da177e4 | 373 | xfrm_policy_gc_kill(policy); |
1da177e4 LT |
374 | } |
375 | ||
376 | /* Rule must be locked. Release descentant resources, announce | |
377 | * entry dead. The rule must be unlinked from lists to the moment. | |
378 | */ | |
379 | ||
380 | static void xfrm_policy_kill(struct xfrm_policy *policy) | |
381 | { | |
382 | int dead; | |
383 | ||
384 | write_lock_bh(&policy->lock); | |
385 | dead = policy->dead; | |
386 | policy->dead = 1; | |
387 | write_unlock_bh(&policy->lock); | |
388 | ||
389 | if (unlikely(dead)) { | |
390 | WARN_ON(1); | |
391 | return; | |
392 | } | |
393 | ||
394 | spin_lock(&xfrm_policy_gc_lock); | |
2518c7c2 | 395 | hlist_add_head(&policy->bydst, &xfrm_policy_gc_list); |
1da177e4 LT |
396 | spin_unlock(&xfrm_policy_gc_lock); |
397 | ||
398 | schedule_work(&xfrm_policy_gc_work); | |
399 | } | |
400 | ||
2518c7c2 DM |
401 | struct xfrm_policy_hash { |
402 | struct hlist_head *table; | |
403 | unsigned int hmask; | |
404 | }; | |
405 | ||
406 | static struct hlist_head xfrm_policy_inexact[XFRM_POLICY_MAX*2]; | |
407 | static struct xfrm_policy_hash xfrm_policy_bydst[XFRM_POLICY_MAX*2] __read_mostly; | |
408 | static struct hlist_head *xfrm_policy_byidx __read_mostly; | |
409 | static unsigned int xfrm_idx_hmask __read_mostly; | |
410 | static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024; | |
411 | ||
2518c7c2 DM |
412 | static inline unsigned int idx_hash(u32 index) |
413 | { | |
414 | return __idx_hash(index, xfrm_idx_hmask); | |
415 | } | |
416 | ||
2518c7c2 DM |
417 | static struct hlist_head *policy_hash_bysel(struct xfrm_selector *sel, unsigned short family, int dir) |
418 | { | |
419 | unsigned int hmask = xfrm_policy_bydst[dir].hmask; | |
420 | unsigned int hash = __sel_hash(sel, family, hmask); | |
421 | ||
422 | return (hash == hmask + 1 ? | |
423 | &xfrm_policy_inexact[dir] : | |
424 | xfrm_policy_bydst[dir].table + hash); | |
425 | } | |
426 | ||
427 | static struct hlist_head *policy_hash_direct(xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir) | |
428 | { | |
429 | unsigned int hmask = xfrm_policy_bydst[dir].hmask; | |
430 | unsigned int hash = __addr_hash(daddr, saddr, family, hmask); | |
431 | ||
432 | return xfrm_policy_bydst[dir].table + hash; | |
433 | } | |
434 | ||
2518c7c2 DM |
435 | static void xfrm_dst_hash_transfer(struct hlist_head *list, |
436 | struct hlist_head *ndsttable, | |
437 | unsigned int nhashmask) | |
438 | { | |
439 | struct hlist_node *entry, *tmp; | |
440 | struct xfrm_policy *pol; | |
441 | ||
442 | hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) { | |
443 | unsigned int h; | |
444 | ||
445 | h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr, | |
446 | pol->family, nhashmask); | |
447 | hlist_add_head(&pol->bydst, ndsttable+h); | |
448 | } | |
449 | } | |
450 | ||
451 | static void xfrm_idx_hash_transfer(struct hlist_head *list, | |
452 | struct hlist_head *nidxtable, | |
453 | unsigned int nhashmask) | |
454 | { | |
455 | struct hlist_node *entry, *tmp; | |
456 | struct xfrm_policy *pol; | |
457 | ||
458 | hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) { | |
459 | unsigned int h; | |
460 | ||
461 | h = __idx_hash(pol->index, nhashmask); | |
462 | hlist_add_head(&pol->byidx, nidxtable+h); | |
463 | } | |
464 | } | |
465 | ||
466 | static unsigned long xfrm_new_hash_mask(unsigned int old_hmask) | |
467 | { | |
468 | return ((old_hmask + 1) << 1) - 1; | |
469 | } | |
470 | ||
471 | static void xfrm_bydst_resize(int dir) | |
472 | { | |
473 | unsigned int hmask = xfrm_policy_bydst[dir].hmask; | |
474 | unsigned int nhashmask = xfrm_new_hash_mask(hmask); | |
475 | unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head); | |
476 | struct hlist_head *odst = xfrm_policy_bydst[dir].table; | |
44e36b42 | 477 | struct hlist_head *ndst = xfrm_hash_alloc(nsize); |
2518c7c2 DM |
478 | int i; |
479 | ||
480 | if (!ndst) | |
481 | return; | |
482 | ||
483 | write_lock_bh(&xfrm_policy_lock); | |
484 | ||
485 | for (i = hmask; i >= 0; i--) | |
486 | xfrm_dst_hash_transfer(odst + i, ndst, nhashmask); | |
487 | ||
488 | xfrm_policy_bydst[dir].table = ndst; | |
489 | xfrm_policy_bydst[dir].hmask = nhashmask; | |
490 | ||
491 | write_unlock_bh(&xfrm_policy_lock); | |
492 | ||
44e36b42 | 493 | xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head)); |
2518c7c2 DM |
494 | } |
495 | ||
496 | static void xfrm_byidx_resize(int total) | |
497 | { | |
498 | unsigned int hmask = xfrm_idx_hmask; | |
499 | unsigned int nhashmask = xfrm_new_hash_mask(hmask); | |
500 | unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head); | |
501 | struct hlist_head *oidx = xfrm_policy_byidx; | |
44e36b42 | 502 | struct hlist_head *nidx = xfrm_hash_alloc(nsize); |
2518c7c2 DM |
503 | int i; |
504 | ||
505 | if (!nidx) | |
506 | return; | |
507 | ||
508 | write_lock_bh(&xfrm_policy_lock); | |
509 | ||
510 | for (i = hmask; i >= 0; i--) | |
511 | xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask); | |
512 | ||
513 | xfrm_policy_byidx = nidx; | |
514 | xfrm_idx_hmask = nhashmask; | |
515 | ||
516 | write_unlock_bh(&xfrm_policy_lock); | |
517 | ||
44e36b42 | 518 | xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head)); |
2518c7c2 DM |
519 | } |
520 | ||
521 | static inline int xfrm_bydst_should_resize(int dir, int *total) | |
522 | { | |
523 | unsigned int cnt = xfrm_policy_count[dir]; | |
524 | unsigned int hmask = xfrm_policy_bydst[dir].hmask; | |
525 | ||
526 | if (total) | |
527 | *total += cnt; | |
528 | ||
529 | if ((hmask + 1) < xfrm_policy_hashmax && | |
530 | cnt > hmask) | |
531 | return 1; | |
532 | ||
533 | return 0; | |
534 | } | |
535 | ||
536 | static inline int xfrm_byidx_should_resize(int total) | |
537 | { | |
538 | unsigned int hmask = xfrm_idx_hmask; | |
539 | ||
540 | if ((hmask + 1) < xfrm_policy_hashmax && | |
541 | total > hmask) | |
542 | return 1; | |
543 | ||
544 | return 0; | |
545 | } | |
546 | ||
547 | static DEFINE_MUTEX(hash_resize_mutex); | |
548 | ||
549 | static void xfrm_hash_resize(void *__unused) | |
550 | { | |
551 | int dir, total; | |
552 | ||
553 | mutex_lock(&hash_resize_mutex); | |
554 | ||
555 | total = 0; | |
556 | for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) { | |
557 | if (xfrm_bydst_should_resize(dir, &total)) | |
558 | xfrm_bydst_resize(dir); | |
559 | } | |
560 | if (xfrm_byidx_should_resize(total)) | |
561 | xfrm_byidx_resize(total); | |
562 | ||
563 | mutex_unlock(&hash_resize_mutex); | |
564 | } | |
565 | ||
566 | static DECLARE_WORK(xfrm_hash_work, xfrm_hash_resize, NULL); | |
567 | ||
1da177e4 LT |
568 | /* Generate new index... KAME seems to generate them ordered by cost |
569 | * of an absolute inpredictability of ordering of rules. This will not pass. */ | |
4e81bb83 | 570 | static u32 xfrm_gen_index(u8 type, int dir) |
1da177e4 | 571 | { |
1da177e4 LT |
572 | static u32 idx_generator; |
573 | ||
574 | for (;;) { | |
2518c7c2 DM |
575 | struct hlist_node *entry; |
576 | struct hlist_head *list; | |
577 | struct xfrm_policy *p; | |
578 | u32 idx; | |
579 | int found; | |
580 | ||
1da177e4 LT |
581 | idx = (idx_generator | dir); |
582 | idx_generator += 8; | |
583 | if (idx == 0) | |
584 | idx = 8; | |
2518c7c2 DM |
585 | list = xfrm_policy_byidx + idx_hash(idx); |
586 | found = 0; | |
587 | hlist_for_each_entry(p, entry, list, byidx) { | |
588 | if (p->index == idx) { | |
589 | found = 1; | |
1da177e4 | 590 | break; |
2518c7c2 | 591 | } |
1da177e4 | 592 | } |
2518c7c2 | 593 | if (!found) |
1da177e4 LT |
594 | return idx; |
595 | } | |
596 | } | |
597 | ||
2518c7c2 DM |
598 | static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2) |
599 | { | |
600 | u32 *p1 = (u32 *) s1; | |
601 | u32 *p2 = (u32 *) s2; | |
602 | int len = sizeof(struct xfrm_selector) / sizeof(u32); | |
603 | int i; | |
604 | ||
605 | for (i = 0; i < len; i++) { | |
606 | if (p1[i] != p2[i]) | |
607 | return 1; | |
608 | } | |
609 | ||
610 | return 0; | |
611 | } | |
612 | ||
1da177e4 LT |
613 | int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl) |
614 | { | |
2518c7c2 DM |
615 | struct xfrm_policy *pol; |
616 | struct xfrm_policy *delpol; | |
617 | struct hlist_head *chain; | |
618 | struct hlist_node *entry, *newpos, *last; | |
9b78a82c | 619 | struct dst_entry *gc_list; |
1da177e4 LT |
620 | |
621 | write_lock_bh(&xfrm_policy_lock); | |
2518c7c2 DM |
622 | chain = policy_hash_bysel(&policy->selector, policy->family, dir); |
623 | delpol = NULL; | |
624 | newpos = NULL; | |
625 | last = NULL; | |
626 | hlist_for_each_entry(pol, entry, chain, bydst) { | |
627 | if (!delpol && | |
628 | pol->type == policy->type && | |
629 | !selector_cmp(&pol->selector, &policy->selector) && | |
df71837d | 630 | xfrm_sec_ctx_match(pol->security, policy->security)) { |
1da177e4 LT |
631 | if (excl) { |
632 | write_unlock_bh(&xfrm_policy_lock); | |
633 | return -EEXIST; | |
634 | } | |
1da177e4 LT |
635 | delpol = pol; |
636 | if (policy->priority > pol->priority) | |
637 | continue; | |
638 | } else if (policy->priority >= pol->priority) { | |
2518c7c2 | 639 | last = &pol->bydst; |
1da177e4 LT |
640 | continue; |
641 | } | |
642 | if (!newpos) | |
2518c7c2 | 643 | newpos = &pol->bydst; |
1da177e4 LT |
644 | if (delpol) |
645 | break; | |
2518c7c2 | 646 | last = &pol->bydst; |
1da177e4 | 647 | } |
2518c7c2 DM |
648 | if (!newpos) |
649 | newpos = last; | |
1da177e4 | 650 | if (newpos) |
2518c7c2 DM |
651 | hlist_add_after(newpos, &policy->bydst); |
652 | else | |
653 | hlist_add_head(&policy->bydst, chain); | |
1da177e4 | 654 | xfrm_pol_hold(policy); |
2518c7c2 | 655 | xfrm_policy_count[dir]++; |
1da177e4 | 656 | atomic_inc(&flow_cache_genid); |
2518c7c2 DM |
657 | if (delpol) { |
658 | hlist_del(&delpol->bydst); | |
659 | hlist_del(&delpol->byidx); | |
660 | xfrm_policy_count[dir]--; | |
661 | } | |
4e81bb83 | 662 | policy->index = delpol ? delpol->index : xfrm_gen_index(policy->type, dir); |
2518c7c2 | 663 | hlist_add_head(&policy->byidx, xfrm_policy_byidx+idx_hash(policy->index)); |
1da177e4 LT |
664 | policy->curlft.add_time = (unsigned long)xtime.tv_sec; |
665 | policy->curlft.use_time = 0; | |
666 | if (!mod_timer(&policy->timer, jiffies + HZ)) | |
667 | xfrm_pol_hold(policy); | |
668 | write_unlock_bh(&xfrm_policy_lock); | |
669 | ||
9b78a82c | 670 | if (delpol) |
1da177e4 | 671 | xfrm_policy_kill(delpol); |
2518c7c2 DM |
672 | else if (xfrm_bydst_should_resize(dir, NULL)) |
673 | schedule_work(&xfrm_hash_work); | |
9b78a82c DM |
674 | |
675 | read_lock_bh(&xfrm_policy_lock); | |
676 | gc_list = NULL; | |
2518c7c2 DM |
677 | entry = &policy->bydst; |
678 | hlist_for_each_entry_continue(policy, entry, bydst) { | |
9b78a82c DM |
679 | struct dst_entry *dst; |
680 | ||
681 | write_lock(&policy->lock); | |
682 | dst = policy->bundles; | |
683 | if (dst) { | |
684 | struct dst_entry *tail = dst; | |
685 | while (tail->next) | |
686 | tail = tail->next; | |
687 | tail->next = gc_list; | |
688 | gc_list = dst; | |
689 | ||
690 | policy->bundles = NULL; | |
691 | } | |
692 | write_unlock(&policy->lock); | |
1da177e4 | 693 | } |
9b78a82c DM |
694 | read_unlock_bh(&xfrm_policy_lock); |
695 | ||
696 | while (gc_list) { | |
697 | struct dst_entry *dst = gc_list; | |
698 | ||
699 | gc_list = dst->next; | |
700 | dst_free(dst); | |
701 | } | |
702 | ||
1da177e4 LT |
703 | return 0; |
704 | } | |
705 | EXPORT_SYMBOL(xfrm_policy_insert); | |
706 | ||
4e81bb83 MN |
707 | struct xfrm_policy *xfrm_policy_bysel_ctx(u8 type, int dir, |
708 | struct xfrm_selector *sel, | |
df71837d | 709 | struct xfrm_sec_ctx *ctx, int delete) |
1da177e4 | 710 | { |
2518c7c2 DM |
711 | struct xfrm_policy *pol, *ret; |
712 | struct hlist_head *chain; | |
713 | struct hlist_node *entry; | |
1da177e4 LT |
714 | |
715 | write_lock_bh(&xfrm_policy_lock); | |
2518c7c2 DM |
716 | chain = policy_hash_bysel(sel, sel->family, dir); |
717 | ret = NULL; | |
718 | hlist_for_each_entry(pol, entry, chain, bydst) { | |
719 | if (pol->type == type && | |
720 | !selector_cmp(sel, &pol->selector) && | |
721 | xfrm_sec_ctx_match(ctx, pol->security)) { | |
1da177e4 | 722 | xfrm_pol_hold(pol); |
2518c7c2 DM |
723 | if (delete) { |
724 | hlist_del(&pol->bydst); | |
725 | hlist_del(&pol->byidx); | |
726 | xfrm_policy_count[dir]--; | |
727 | } | |
728 | ret = pol; | |
1da177e4 LT |
729 | break; |
730 | } | |
731 | } | |
732 | write_unlock_bh(&xfrm_policy_lock); | |
733 | ||
2518c7c2 | 734 | if (ret && delete) { |
1da177e4 | 735 | atomic_inc(&flow_cache_genid); |
2518c7c2 | 736 | xfrm_policy_kill(ret); |
1da177e4 | 737 | } |
2518c7c2 | 738 | return ret; |
1da177e4 | 739 | } |
df71837d | 740 | EXPORT_SYMBOL(xfrm_policy_bysel_ctx); |
1da177e4 | 741 | |
4e81bb83 | 742 | struct xfrm_policy *xfrm_policy_byid(u8 type, int dir, u32 id, int delete) |
1da177e4 | 743 | { |
2518c7c2 DM |
744 | struct xfrm_policy *pol, *ret; |
745 | struct hlist_head *chain; | |
746 | struct hlist_node *entry; | |
1da177e4 LT |
747 | |
748 | write_lock_bh(&xfrm_policy_lock); | |
2518c7c2 DM |
749 | chain = xfrm_policy_byidx + idx_hash(id); |
750 | ret = NULL; | |
751 | hlist_for_each_entry(pol, entry, chain, byidx) { | |
752 | if (pol->type == type && pol->index == id) { | |
1da177e4 | 753 | xfrm_pol_hold(pol); |
2518c7c2 DM |
754 | if (delete) { |
755 | hlist_del(&pol->bydst); | |
756 | hlist_del(&pol->byidx); | |
757 | xfrm_policy_count[dir]--; | |
758 | } | |
759 | ret = pol; | |
1da177e4 LT |
760 | break; |
761 | } | |
762 | } | |
763 | write_unlock_bh(&xfrm_policy_lock); | |
764 | ||
2518c7c2 | 765 | if (ret && delete) { |
1da177e4 | 766 | atomic_inc(&flow_cache_genid); |
2518c7c2 | 767 | xfrm_policy_kill(ret); |
1da177e4 | 768 | } |
2518c7c2 | 769 | return ret; |
1da177e4 LT |
770 | } |
771 | EXPORT_SYMBOL(xfrm_policy_byid); | |
772 | ||
4e81bb83 | 773 | void xfrm_policy_flush(u8 type) |
1da177e4 | 774 | { |
1da177e4 LT |
775 | int dir; |
776 | ||
777 | write_lock_bh(&xfrm_policy_lock); | |
778 | for (dir = 0; dir < XFRM_POLICY_MAX; dir++) { | |
2518c7c2 DM |
779 | struct xfrm_policy *pol; |
780 | struct hlist_node *entry; | |
781 | int i; | |
782 | ||
783 | again1: | |
784 | hlist_for_each_entry(pol, entry, | |
785 | &xfrm_policy_inexact[dir], bydst) { | |
786 | if (pol->type != type) | |
787 | continue; | |
788 | hlist_del(&pol->bydst); | |
789 | hlist_del(&pol->byidx); | |
1da177e4 LT |
790 | write_unlock_bh(&xfrm_policy_lock); |
791 | ||
2518c7c2 | 792 | xfrm_policy_kill(pol); |
1da177e4 LT |
793 | |
794 | write_lock_bh(&xfrm_policy_lock); | |
2518c7c2 DM |
795 | goto again1; |
796 | } | |
797 | ||
798 | for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) { | |
799 | again2: | |
800 | hlist_for_each_entry(pol, entry, | |
801 | xfrm_policy_bydst[dir].table + i, | |
802 | bydst) { | |
803 | if (pol->type != type) | |
804 | continue; | |
805 | hlist_del(&pol->bydst); | |
806 | hlist_del(&pol->byidx); | |
807 | write_unlock_bh(&xfrm_policy_lock); | |
808 | ||
809 | xfrm_policy_kill(pol); | |
810 | ||
811 | write_lock_bh(&xfrm_policy_lock); | |
812 | goto again2; | |
813 | } | |
1da177e4 | 814 | } |
2518c7c2 DM |
815 | |
816 | xfrm_policy_count[dir] = 0; | |
1da177e4 LT |
817 | } |
818 | atomic_inc(&flow_cache_genid); | |
819 | write_unlock_bh(&xfrm_policy_lock); | |
820 | } | |
821 | EXPORT_SYMBOL(xfrm_policy_flush); | |
822 | ||
4e81bb83 | 823 | int xfrm_policy_walk(u8 type, int (*func)(struct xfrm_policy *, int, int, void*), |
1da177e4 LT |
824 | void *data) |
825 | { | |
2518c7c2 DM |
826 | struct xfrm_policy *pol; |
827 | struct hlist_node *entry; | |
828 | int dir, count, error; | |
1da177e4 LT |
829 | |
830 | read_lock_bh(&xfrm_policy_lock); | |
2518c7c2 | 831 | count = 0; |
1da177e4 | 832 | for (dir = 0; dir < 2*XFRM_POLICY_MAX; dir++) { |
2518c7c2 DM |
833 | struct hlist_head *table = xfrm_policy_bydst[dir].table; |
834 | int i; | |
835 | ||
836 | hlist_for_each_entry(pol, entry, | |
837 | &xfrm_policy_inexact[dir], bydst) { | |
838 | if (pol->type == type) | |
839 | count++; | |
840 | } | |
841 | for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) { | |
842 | hlist_for_each_entry(pol, entry, table + i, bydst) { | |
843 | if (pol->type == type) | |
844 | count++; | |
845 | } | |
846 | } | |
1da177e4 LT |
847 | } |
848 | ||
849 | if (count == 0) { | |
850 | error = -ENOENT; | |
851 | goto out; | |
852 | } | |
853 | ||
854 | for (dir = 0; dir < 2*XFRM_POLICY_MAX; dir++) { | |
2518c7c2 DM |
855 | struct hlist_head *table = xfrm_policy_bydst[dir].table; |
856 | int i; | |
857 | ||
858 | hlist_for_each_entry(pol, entry, | |
859 | &xfrm_policy_inexact[dir], bydst) { | |
860 | if (pol->type != type) | |
861 | continue; | |
862 | error = func(pol, dir % XFRM_POLICY_MAX, --count, data); | |
1da177e4 LT |
863 | if (error) |
864 | goto out; | |
865 | } | |
2518c7c2 DM |
866 | for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) { |
867 | hlist_for_each_entry(pol, entry, table + i, bydst) { | |
868 | if (pol->type != type) | |
869 | continue; | |
870 | error = func(pol, dir % XFRM_POLICY_MAX, --count, data); | |
871 | if (error) | |
872 | goto out; | |
873 | } | |
874 | } | |
1da177e4 | 875 | } |
2518c7c2 | 876 | error = 0; |
1da177e4 LT |
877 | out: |
878 | read_unlock_bh(&xfrm_policy_lock); | |
879 | return error; | |
880 | } | |
881 | EXPORT_SYMBOL(xfrm_policy_walk); | |
882 | ||
883 | /* Find policy to apply to this flow. */ | |
884 | ||
2518c7c2 DM |
885 | static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl, |
886 | u8 type, u16 family, int dir) | |
1da177e4 | 887 | { |
2518c7c2 DM |
888 | struct xfrm_selector *sel = &pol->selector; |
889 | int match; | |
1da177e4 | 890 | |
2518c7c2 DM |
891 | if (pol->family != family || |
892 | pol->type != type) | |
893 | return 0; | |
1da177e4 | 894 | |
2518c7c2 DM |
895 | match = xfrm_selector_match(sel, fl, family); |
896 | if (match) { | |
897 | if (!security_xfrm_policy_lookup(pol, fl->secid, dir)) | |
898 | return 1; | |
899 | } | |
900 | ||
901 | return 0; | |
902 | } | |
1da177e4 | 903 | |
2518c7c2 DM |
904 | static struct xfrm_policy *xfrm_policy_lookup_bytype(u8 type, struct flowi *fl, |
905 | u16 family, u8 dir) | |
906 | { | |
907 | struct xfrm_policy *pol, *ret; | |
908 | xfrm_address_t *daddr, *saddr; | |
909 | struct hlist_node *entry; | |
910 | struct hlist_head *chain; | |
df71837d | 911 | |
2518c7c2 DM |
912 | daddr = xfrm_flowi_daddr(fl, family); |
913 | saddr = xfrm_flowi_saddr(fl, family); | |
914 | if (unlikely(!daddr || !saddr)) | |
915 | return NULL; | |
916 | ||
917 | read_lock_bh(&xfrm_policy_lock); | |
918 | chain = policy_hash_direct(daddr, saddr, family, dir); | |
919 | ret = NULL; | |
920 | hlist_for_each_entry(pol, entry, chain, bydst) { | |
921 | if (xfrm_policy_match(pol, fl, type, family, dir)) { | |
922 | xfrm_pol_hold(pol); | |
923 | ret = pol; | |
924 | break; | |
925 | } | |
926 | } | |
927 | if (!ret) { | |
928 | chain = &xfrm_policy_inexact[dir]; | |
929 | hlist_for_each_entry(pol, entry, chain, bydst) { | |
930 | if (xfrm_policy_match(pol, fl, type, family, dir)) { | |
df71837d | 931 | xfrm_pol_hold(pol); |
2518c7c2 | 932 | ret = pol; |
df71837d TJ |
933 | break; |
934 | } | |
1da177e4 LT |
935 | } |
936 | } | |
937 | read_unlock_bh(&xfrm_policy_lock); | |
4e81bb83 | 938 | |
2518c7c2 | 939 | return ret; |
4e81bb83 MN |
940 | } |
941 | ||
942 | static void xfrm_policy_lookup(struct flowi *fl, u16 family, u8 dir, | |
943 | void **objp, atomic_t **obj_refp) | |
944 | { | |
945 | struct xfrm_policy *pol; | |
946 | ||
947 | #ifdef CONFIG_XFRM_SUB_POLICY | |
948 | pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB, fl, family, dir); | |
949 | if (pol) | |
950 | goto end; | |
951 | #endif | |
952 | pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN, fl, family, dir); | |
953 | ||
954 | #ifdef CONFIG_XFRM_SUB_POLICY | |
2518c7c2 | 955 | end: |
4e81bb83 | 956 | #endif |
1da177e4 LT |
957 | if ((*objp = (void *) pol) != NULL) |
958 | *obj_refp = &pol->refcnt; | |
959 | } | |
960 | ||
df71837d TJ |
961 | static inline int policy_to_flow_dir(int dir) |
962 | { | |
963 | if (XFRM_POLICY_IN == FLOW_DIR_IN && | |
964 | XFRM_POLICY_OUT == FLOW_DIR_OUT && | |
965 | XFRM_POLICY_FWD == FLOW_DIR_FWD) | |
966 | return dir; | |
967 | switch (dir) { | |
968 | default: | |
969 | case XFRM_POLICY_IN: | |
970 | return FLOW_DIR_IN; | |
971 | case XFRM_POLICY_OUT: | |
972 | return FLOW_DIR_OUT; | |
973 | case XFRM_POLICY_FWD: | |
974 | return FLOW_DIR_FWD; | |
975 | }; | |
976 | } | |
977 | ||
e0d1caa7 | 978 | static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl) |
1da177e4 LT |
979 | { |
980 | struct xfrm_policy *pol; | |
981 | ||
982 | read_lock_bh(&xfrm_policy_lock); | |
983 | if ((pol = sk->sk_policy[dir]) != NULL) { | |
df71837d | 984 | int match = xfrm_selector_match(&pol->selector, fl, |
1da177e4 | 985 | sk->sk_family); |
df71837d TJ |
986 | int err = 0; |
987 | ||
1da177e4 | 988 | if (match) |
e0d1caa7 | 989 | err = security_xfrm_policy_lookup(pol, fl->secid, policy_to_flow_dir(dir)); |
df71837d TJ |
990 | |
991 | if (match && !err) | |
1da177e4 LT |
992 | xfrm_pol_hold(pol); |
993 | else | |
994 | pol = NULL; | |
995 | } | |
996 | read_unlock_bh(&xfrm_policy_lock); | |
997 | return pol; | |
998 | } | |
999 | ||
1000 | static void __xfrm_policy_link(struct xfrm_policy *pol, int dir) | |
1001 | { | |
2518c7c2 DM |
1002 | struct hlist_head *chain = policy_hash_bysel(&pol->selector, |
1003 | pol->family, dir); | |
4e81bb83 | 1004 | |
2518c7c2 DM |
1005 | hlist_add_head(&pol->bydst, chain); |
1006 | hlist_add_head(&pol->byidx, xfrm_policy_byidx+idx_hash(pol->index)); | |
1007 | xfrm_policy_count[dir]++; | |
1da177e4 | 1008 | xfrm_pol_hold(pol); |
2518c7c2 DM |
1009 | |
1010 | if (xfrm_bydst_should_resize(dir, NULL)) | |
1011 | schedule_work(&xfrm_hash_work); | |
1da177e4 LT |
1012 | } |
1013 | ||
1014 | static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol, | |
1015 | int dir) | |
1016 | { | |
2518c7c2 DM |
1017 | if (hlist_unhashed(&pol->bydst)) |
1018 | return NULL; | |
1da177e4 | 1019 | |
2518c7c2 DM |
1020 | hlist_del(&pol->bydst); |
1021 | hlist_del(&pol->byidx); | |
1022 | xfrm_policy_count[dir]--; | |
1023 | ||
1024 | return pol; | |
1da177e4 LT |
1025 | } |
1026 | ||
4666faab | 1027 | int xfrm_policy_delete(struct xfrm_policy *pol, int dir) |
1da177e4 LT |
1028 | { |
1029 | write_lock_bh(&xfrm_policy_lock); | |
1030 | pol = __xfrm_policy_unlink(pol, dir); | |
1031 | write_unlock_bh(&xfrm_policy_lock); | |
1032 | if (pol) { | |
1033 | if (dir < XFRM_POLICY_MAX) | |
1034 | atomic_inc(&flow_cache_genid); | |
1035 | xfrm_policy_kill(pol); | |
4666faab | 1036 | return 0; |
1da177e4 | 1037 | } |
4666faab | 1038 | return -ENOENT; |
1da177e4 | 1039 | } |
a70fcb0b | 1040 | EXPORT_SYMBOL(xfrm_policy_delete); |
1da177e4 LT |
1041 | |
1042 | int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol) | |
1043 | { | |
1044 | struct xfrm_policy *old_pol; | |
1045 | ||
4e81bb83 MN |
1046 | #ifdef CONFIG_XFRM_SUB_POLICY |
1047 | if (pol && pol->type != XFRM_POLICY_TYPE_MAIN) | |
1048 | return -EINVAL; | |
1049 | #endif | |
1050 | ||
1da177e4 LT |
1051 | write_lock_bh(&xfrm_policy_lock); |
1052 | old_pol = sk->sk_policy[dir]; | |
1053 | sk->sk_policy[dir] = pol; | |
1054 | if (pol) { | |
1055 | pol->curlft.add_time = (unsigned long)xtime.tv_sec; | |
4e81bb83 | 1056 | pol->index = xfrm_gen_index(pol->type, XFRM_POLICY_MAX+dir); |
1da177e4 LT |
1057 | __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir); |
1058 | } | |
1059 | if (old_pol) | |
1060 | __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir); | |
1061 | write_unlock_bh(&xfrm_policy_lock); | |
1062 | ||
1063 | if (old_pol) { | |
1064 | xfrm_policy_kill(old_pol); | |
1065 | } | |
1066 | return 0; | |
1067 | } | |
1068 | ||
1069 | static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir) | |
1070 | { | |
1071 | struct xfrm_policy *newp = xfrm_policy_alloc(GFP_ATOMIC); | |
1072 | ||
1073 | if (newp) { | |
1074 | newp->selector = old->selector; | |
df71837d TJ |
1075 | if (security_xfrm_policy_clone(old, newp)) { |
1076 | kfree(newp); | |
1077 | return NULL; /* ENOMEM */ | |
1078 | } | |
1da177e4 LT |
1079 | newp->lft = old->lft; |
1080 | newp->curlft = old->curlft; | |
1081 | newp->action = old->action; | |
1082 | newp->flags = old->flags; | |
1083 | newp->xfrm_nr = old->xfrm_nr; | |
1084 | newp->index = old->index; | |
4e81bb83 | 1085 | newp->type = old->type; |
1da177e4 LT |
1086 | memcpy(newp->xfrm_vec, old->xfrm_vec, |
1087 | newp->xfrm_nr*sizeof(struct xfrm_tmpl)); | |
1088 | write_lock_bh(&xfrm_policy_lock); | |
1089 | __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir); | |
1090 | write_unlock_bh(&xfrm_policy_lock); | |
1091 | xfrm_pol_put(newp); | |
1092 | } | |
1093 | return newp; | |
1094 | } | |
1095 | ||
1096 | int __xfrm_sk_clone_policy(struct sock *sk) | |
1097 | { | |
1098 | struct xfrm_policy *p0 = sk->sk_policy[0], | |
1099 | *p1 = sk->sk_policy[1]; | |
1100 | ||
1101 | sk->sk_policy[0] = sk->sk_policy[1] = NULL; | |
1102 | if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL) | |
1103 | return -ENOMEM; | |
1104 | if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL) | |
1105 | return -ENOMEM; | |
1106 | return 0; | |
1107 | } | |
1108 | ||
1109 | /* Resolve list of templates for the flow, given policy. */ | |
1110 | ||
1111 | static int | |
4e81bb83 MN |
1112 | xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl, |
1113 | struct xfrm_state **xfrm, | |
1114 | unsigned short family) | |
1da177e4 LT |
1115 | { |
1116 | int nx; | |
1117 | int i, error; | |
1118 | xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family); | |
1119 | xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family); | |
1120 | ||
1121 | for (nx=0, i = 0; i < policy->xfrm_nr; i++) { | |
1122 | struct xfrm_state *x; | |
1123 | xfrm_address_t *remote = daddr; | |
1124 | xfrm_address_t *local = saddr; | |
1125 | struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i]; | |
1126 | ||
7e49e6de | 1127 | if (tmpl->mode == XFRM_MODE_TUNNEL) { |
1da177e4 LT |
1128 | remote = &tmpl->id.daddr; |
1129 | local = &tmpl->saddr; | |
1130 | } | |
1131 | ||
1132 | x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family); | |
1133 | ||
1134 | if (x && x->km.state == XFRM_STATE_VALID) { | |
1135 | xfrm[nx++] = x; | |
1136 | daddr = remote; | |
1137 | saddr = local; | |
1138 | continue; | |
1139 | } | |
1140 | if (x) { | |
1141 | error = (x->km.state == XFRM_STATE_ERROR ? | |
1142 | -EINVAL : -EAGAIN); | |
1143 | xfrm_state_put(x); | |
1144 | } | |
1145 | ||
1146 | if (!tmpl->optional) | |
1147 | goto fail; | |
1148 | } | |
1149 | return nx; | |
1150 | ||
1151 | fail: | |
1152 | for (nx--; nx>=0; nx--) | |
1153 | xfrm_state_put(xfrm[nx]); | |
1154 | return error; | |
1155 | } | |
1156 | ||
4e81bb83 MN |
1157 | static int |
1158 | xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl, | |
1159 | struct xfrm_state **xfrm, | |
1160 | unsigned short family) | |
1161 | { | |
41a49cc3 MN |
1162 | struct xfrm_state *tp[XFRM_MAX_DEPTH]; |
1163 | struct xfrm_state **tpp = (npols > 1) ? tp : xfrm; | |
4e81bb83 MN |
1164 | int cnx = 0; |
1165 | int error; | |
1166 | int ret; | |
1167 | int i; | |
1168 | ||
1169 | for (i = 0; i < npols; i++) { | |
1170 | if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) { | |
1171 | error = -ENOBUFS; | |
1172 | goto fail; | |
1173 | } | |
41a49cc3 MN |
1174 | |
1175 | ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family); | |
4e81bb83 MN |
1176 | if (ret < 0) { |
1177 | error = ret; | |
1178 | goto fail; | |
1179 | } else | |
1180 | cnx += ret; | |
1181 | } | |
1182 | ||
41a49cc3 MN |
1183 | /* found states are sorted for outbound processing */ |
1184 | if (npols > 1) | |
1185 | xfrm_state_sort(xfrm, tpp, cnx, family); | |
1186 | ||
4e81bb83 MN |
1187 | return cnx; |
1188 | ||
1189 | fail: | |
1190 | for (cnx--; cnx>=0; cnx--) | |
41a49cc3 | 1191 | xfrm_state_put(tpp[cnx]); |
4e81bb83 MN |
1192 | return error; |
1193 | ||
1194 | } | |
1195 | ||
1da177e4 LT |
1196 | /* Check that the bundle accepts the flow and its components are |
1197 | * still valid. | |
1198 | */ | |
1199 | ||
1200 | static struct dst_entry * | |
1201 | xfrm_find_bundle(struct flowi *fl, struct xfrm_policy *policy, unsigned short family) | |
1202 | { | |
1203 | struct dst_entry *x; | |
1204 | struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); | |
1205 | if (unlikely(afinfo == NULL)) | |
1206 | return ERR_PTR(-EINVAL); | |
1207 | x = afinfo->find_bundle(fl, policy); | |
1208 | xfrm_policy_put_afinfo(afinfo); | |
1209 | return x; | |
1210 | } | |
1211 | ||
1212 | /* Allocate chain of dst_entry's, attach known xfrm's, calculate | |
1213 | * all the metrics... Shortly, bundle a bundle. | |
1214 | */ | |
1215 | ||
1216 | static int | |
1217 | xfrm_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx, | |
1218 | struct flowi *fl, struct dst_entry **dst_p, | |
1219 | unsigned short family) | |
1220 | { | |
1221 | int err; | |
1222 | struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); | |
1223 | if (unlikely(afinfo == NULL)) | |
1224 | return -EINVAL; | |
1225 | err = afinfo->bundle_create(policy, xfrm, nx, fl, dst_p); | |
1226 | xfrm_policy_put_afinfo(afinfo); | |
1227 | return err; | |
1228 | } | |
1229 | ||
1da177e4 LT |
1230 | |
1231 | static int stale_bundle(struct dst_entry *dst); | |
1232 | ||
1233 | /* Main function: finds/creates a bundle for given flow. | |
1234 | * | |
1235 | * At the moment we eat a raw IP route. Mostly to speed up lookups | |
1236 | * on interfaces with disabled IPsec. | |
1237 | */ | |
1238 | int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl, | |
1239 | struct sock *sk, int flags) | |
1240 | { | |
1241 | struct xfrm_policy *policy; | |
4e81bb83 MN |
1242 | struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX]; |
1243 | int npols; | |
1244 | int pol_dead; | |
1245 | int xfrm_nr; | |
1246 | int pi; | |
1da177e4 LT |
1247 | struct xfrm_state *xfrm[XFRM_MAX_DEPTH]; |
1248 | struct dst_entry *dst, *dst_orig = *dst_p; | |
1249 | int nx = 0; | |
1250 | int err; | |
1251 | u32 genid; | |
42cf93cd | 1252 | u16 family; |
df71837d | 1253 | u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT); |
e0d1caa7 | 1254 | |
1da177e4 LT |
1255 | restart: |
1256 | genid = atomic_read(&flow_cache_genid); | |
1257 | policy = NULL; | |
4e81bb83 MN |
1258 | for (pi = 0; pi < ARRAY_SIZE(pols); pi++) |
1259 | pols[pi] = NULL; | |
1260 | npols = 0; | |
1261 | pol_dead = 0; | |
1262 | xfrm_nr = 0; | |
1263 | ||
1da177e4 | 1264 | if (sk && sk->sk_policy[1]) |
e0d1caa7 | 1265 | policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl); |
1da177e4 LT |
1266 | |
1267 | if (!policy) { | |
1268 | /* To accelerate a bit... */ | |
2518c7c2 DM |
1269 | if ((dst_orig->flags & DST_NOXFRM) || |
1270 | !xfrm_policy_count[XFRM_POLICY_OUT]) | |
1da177e4 LT |
1271 | return 0; |
1272 | ||
e0d1caa7 | 1273 | policy = flow_cache_lookup(fl, dst_orig->ops->family, |
42cf93cd | 1274 | dir, xfrm_policy_lookup); |
1da177e4 LT |
1275 | } |
1276 | ||
1277 | if (!policy) | |
1278 | return 0; | |
1279 | ||
42cf93cd | 1280 | family = dst_orig->ops->family; |
1da177e4 | 1281 | policy->curlft.use_time = (unsigned long)xtime.tv_sec; |
4e81bb83 MN |
1282 | pols[0] = policy; |
1283 | npols ++; | |
1284 | xfrm_nr += pols[0]->xfrm_nr; | |
1da177e4 LT |
1285 | |
1286 | switch (policy->action) { | |
1287 | case XFRM_POLICY_BLOCK: | |
1288 | /* Prohibit the flow */ | |
e104411b PM |
1289 | err = -EPERM; |
1290 | goto error; | |
1da177e4 LT |
1291 | |
1292 | case XFRM_POLICY_ALLOW: | |
4e81bb83 | 1293 | #ifndef CONFIG_XFRM_SUB_POLICY |
1da177e4 LT |
1294 | if (policy->xfrm_nr == 0) { |
1295 | /* Flow passes not transformed. */ | |
1296 | xfrm_pol_put(policy); | |
1297 | return 0; | |
1298 | } | |
4e81bb83 | 1299 | #endif |
1da177e4 LT |
1300 | |
1301 | /* Try to find matching bundle. | |
1302 | * | |
1303 | * LATER: help from flow cache. It is optional, this | |
1304 | * is required only for output policy. | |
1305 | */ | |
1306 | dst = xfrm_find_bundle(fl, policy, family); | |
1307 | if (IS_ERR(dst)) { | |
e104411b PM |
1308 | err = PTR_ERR(dst); |
1309 | goto error; | |
1da177e4 LT |
1310 | } |
1311 | ||
1312 | if (dst) | |
1313 | break; | |
1314 | ||
4e81bb83 MN |
1315 | #ifdef CONFIG_XFRM_SUB_POLICY |
1316 | if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) { | |
1317 | pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN, | |
1318 | fl, family, | |
1319 | XFRM_POLICY_OUT); | |
1320 | if (pols[1]) { | |
1321 | if (pols[1]->action == XFRM_POLICY_BLOCK) { | |
1322 | err = -EPERM; | |
1323 | goto error; | |
1324 | } | |
1325 | npols ++; | |
1326 | xfrm_nr += pols[1]->xfrm_nr; | |
1327 | } | |
1328 | } | |
1329 | ||
1330 | /* | |
1331 | * Because neither flowi nor bundle information knows about | |
1332 | * transformation template size. On more than one policy usage | |
1333 | * we can realize whether all of them is bypass or not after | |
1334 | * they are searched. See above not-transformed bypass | |
1335 | * is surrounded by non-sub policy configuration, too. | |
1336 | */ | |
1337 | if (xfrm_nr == 0) { | |
1338 | /* Flow passes not transformed. */ | |
1339 | xfrm_pols_put(pols, npols); | |
1340 | return 0; | |
1341 | } | |
1342 | ||
1343 | #endif | |
1344 | nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family); | |
1da177e4 LT |
1345 | |
1346 | if (unlikely(nx<0)) { | |
1347 | err = nx; | |
1348 | if (err == -EAGAIN && flags) { | |
1349 | DECLARE_WAITQUEUE(wait, current); | |
1350 | ||
1351 | add_wait_queue(&km_waitq, &wait); | |
1352 | set_current_state(TASK_INTERRUPTIBLE); | |
1353 | schedule(); | |
1354 | set_current_state(TASK_RUNNING); | |
1355 | remove_wait_queue(&km_waitq, &wait); | |
1356 | ||
4e81bb83 | 1357 | nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family); |
1da177e4 LT |
1358 | |
1359 | if (nx == -EAGAIN && signal_pending(current)) { | |
1360 | err = -ERESTART; | |
1361 | goto error; | |
1362 | } | |
1363 | if (nx == -EAGAIN || | |
1364 | genid != atomic_read(&flow_cache_genid)) { | |
4e81bb83 | 1365 | xfrm_pols_put(pols, npols); |
1da177e4 LT |
1366 | goto restart; |
1367 | } | |
1368 | err = nx; | |
1369 | } | |
1370 | if (err < 0) | |
1371 | goto error; | |
1372 | } | |
1373 | if (nx == 0) { | |
1374 | /* Flow passes not transformed. */ | |
4e81bb83 | 1375 | xfrm_pols_put(pols, npols); |
1da177e4 LT |
1376 | return 0; |
1377 | } | |
1378 | ||
1379 | dst = dst_orig; | |
1380 | err = xfrm_bundle_create(policy, xfrm, nx, fl, &dst, family); | |
1381 | ||
1382 | if (unlikely(err)) { | |
1383 | int i; | |
1384 | for (i=0; i<nx; i++) | |
1385 | xfrm_state_put(xfrm[i]); | |
1386 | goto error; | |
1387 | } | |
1388 | ||
4e81bb83 MN |
1389 | for (pi = 0; pi < npols; pi++) { |
1390 | read_lock_bh(&pols[pi]->lock); | |
1391 | pol_dead |= pols[pi]->dead; | |
1392 | read_unlock_bh(&pols[pi]->lock); | |
1393 | } | |
1394 | ||
1da177e4 | 1395 | write_lock_bh(&policy->lock); |
4e81bb83 | 1396 | if (unlikely(pol_dead || stale_bundle(dst))) { |
1da177e4 LT |
1397 | /* Wow! While we worked on resolving, this |
1398 | * policy has gone. Retry. It is not paranoia, | |
1399 | * we just cannot enlist new bundle to dead object. | |
1400 | * We can't enlist stable bundles either. | |
1401 | */ | |
1402 | write_unlock_bh(&policy->lock); | |
1da177e4 LT |
1403 | if (dst) |
1404 | dst_free(dst); | |
00de651d HX |
1405 | |
1406 | err = -EHOSTUNREACH; | |
1407 | goto error; | |
1da177e4 LT |
1408 | } |
1409 | dst->next = policy->bundles; | |
1410 | policy->bundles = dst; | |
1411 | dst_hold(dst); | |
1412 | write_unlock_bh(&policy->lock); | |
1413 | } | |
1414 | *dst_p = dst; | |
1415 | dst_release(dst_orig); | |
4e81bb83 | 1416 | xfrm_pols_put(pols, npols); |
1da177e4 LT |
1417 | return 0; |
1418 | ||
1419 | error: | |
1420 | dst_release(dst_orig); | |
4e81bb83 | 1421 | xfrm_pols_put(pols, npols); |
1da177e4 LT |
1422 | *dst_p = NULL; |
1423 | return err; | |
1424 | } | |
1425 | EXPORT_SYMBOL(xfrm_lookup); | |
1426 | ||
df0ba92a MN |
1427 | static inline int |
1428 | xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl) | |
1429 | { | |
1430 | struct xfrm_state *x; | |
1431 | int err; | |
1432 | ||
1433 | if (!skb->sp || idx < 0 || idx >= skb->sp->len) | |
1434 | return 0; | |
1435 | x = skb->sp->xvec[idx]; | |
1436 | if (!x->type->reject) | |
1437 | return 0; | |
1438 | xfrm_state_hold(x); | |
1439 | err = x->type->reject(x, skb, fl); | |
1440 | xfrm_state_put(x); | |
1441 | return err; | |
1442 | } | |
1443 | ||
1da177e4 LT |
1444 | /* When skb is transformed back to its "native" form, we have to |
1445 | * check policy restrictions. At the moment we make this in maximally | |
1446 | * stupid way. Shame on me. :-) Of course, connected sockets must | |
1447 | * have policy cached at them. | |
1448 | */ | |
1449 | ||
1450 | static inline int | |
1451 | xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x, | |
1452 | unsigned short family) | |
1453 | { | |
1454 | if (xfrm_state_kern(x)) | |
1455 | return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, family); | |
1456 | return x->id.proto == tmpl->id.proto && | |
1457 | (x->id.spi == tmpl->id.spi || !tmpl->id.spi) && | |
1458 | (x->props.reqid == tmpl->reqid || !tmpl->reqid) && | |
1459 | x->props.mode == tmpl->mode && | |
f3bd4840 MN |
1460 | ((tmpl->aalgos & (1<<x->props.aalgo)) || |
1461 | !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) && | |
7e49e6de MN |
1462 | !(x->props.mode != XFRM_MODE_TRANSPORT && |
1463 | xfrm_state_addr_cmp(tmpl, x, family)); | |
1da177e4 LT |
1464 | } |
1465 | ||
df0ba92a MN |
1466 | /* |
1467 | * 0 or more than 0 is returned when validation is succeeded (either bypass | |
1468 | * because of optional transport mode, or next index of the mathced secpath | |
1469 | * state with the template. | |
1470 | * -1 is returned when no matching template is found. | |
1471 | * Otherwise "-2 - errored_index" is returned. | |
1472 | */ | |
1da177e4 LT |
1473 | static inline int |
1474 | xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start, | |
1475 | unsigned short family) | |
1476 | { | |
1477 | int idx = start; | |
1478 | ||
1479 | if (tmpl->optional) { | |
7e49e6de | 1480 | if (tmpl->mode == XFRM_MODE_TRANSPORT) |
1da177e4 LT |
1481 | return start; |
1482 | } else | |
1483 | start = -1; | |
1484 | for (; idx < sp->len; idx++) { | |
dbe5b4aa | 1485 | if (xfrm_state_ok(tmpl, sp->xvec[idx], family)) |
1da177e4 | 1486 | return ++idx; |
df0ba92a MN |
1487 | if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) { |
1488 | if (start == -1) | |
1489 | start = -2-idx; | |
1da177e4 | 1490 | break; |
df0ba92a | 1491 | } |
1da177e4 LT |
1492 | } |
1493 | return start; | |
1494 | } | |
1495 | ||
3e3850e9 PM |
1496 | int |
1497 | xfrm_decode_session(struct sk_buff *skb, struct flowi *fl, unsigned short family) | |
1da177e4 LT |
1498 | { |
1499 | struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); | |
e0d1caa7 | 1500 | int err; |
1da177e4 LT |
1501 | |
1502 | if (unlikely(afinfo == NULL)) | |
1503 | return -EAFNOSUPPORT; | |
1504 | ||
1505 | afinfo->decode_session(skb, fl); | |
beb8d13b | 1506 | err = security_xfrm_decode_session(skb, &fl->secid); |
1da177e4 | 1507 | xfrm_policy_put_afinfo(afinfo); |
e0d1caa7 | 1508 | return err; |
1da177e4 | 1509 | } |
3e3850e9 | 1510 | EXPORT_SYMBOL(xfrm_decode_session); |
1da177e4 | 1511 | |
df0ba92a | 1512 | static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp) |
1da177e4 LT |
1513 | { |
1514 | for (; k < sp->len; k++) { | |
df0ba92a MN |
1515 | if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) { |
1516 | if (idxp) | |
1517 | *idxp = k; | |
1da177e4 | 1518 | return 1; |
df0ba92a | 1519 | } |
1da177e4 LT |
1520 | } |
1521 | ||
1522 | return 0; | |
1523 | } | |
1524 | ||
1525 | int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb, | |
1526 | unsigned short family) | |
1527 | { | |
1528 | struct xfrm_policy *pol; | |
4e81bb83 MN |
1529 | struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX]; |
1530 | int npols = 0; | |
1531 | int xfrm_nr; | |
1532 | int pi; | |
1da177e4 | 1533 | struct flowi fl; |
df71837d | 1534 | u8 fl_dir = policy_to_flow_dir(dir); |
df0ba92a MN |
1535 | int xerr_idx = -1; |
1536 | int *xerr_idxp = &xerr_idx; | |
1da177e4 | 1537 | |
3e3850e9 | 1538 | if (xfrm_decode_session(skb, &fl, family) < 0) |
1da177e4 | 1539 | return 0; |
eb9c7ebe | 1540 | nf_nat_decode_session(skb, &fl, family); |
1da177e4 LT |
1541 | |
1542 | /* First, check used SA against their selectors. */ | |
1543 | if (skb->sp) { | |
1544 | int i; | |
1545 | ||
1546 | for (i=skb->sp->len-1; i>=0; i--) { | |
dbe5b4aa HX |
1547 | struct xfrm_state *x = skb->sp->xvec[i]; |
1548 | if (!xfrm_selector_match(&x->sel, &fl, family)) | |
1da177e4 | 1549 | return 0; |
1da177e4 LT |
1550 | } |
1551 | } | |
1552 | ||
1553 | pol = NULL; | |
1554 | if (sk && sk->sk_policy[dir]) | |
e0d1caa7 | 1555 | pol = xfrm_sk_policy_lookup(sk, dir, &fl); |
1da177e4 LT |
1556 | |
1557 | if (!pol) | |
e0d1caa7 | 1558 | pol = flow_cache_lookup(&fl, family, fl_dir, |
1da177e4 LT |
1559 | xfrm_policy_lookup); |
1560 | ||
df0ba92a MN |
1561 | if (!pol) { |
1562 | if (skb->sp && secpath_has_nontransport(skb->sp, 0, xerr_idxp)) { | |
1563 | xfrm_secpath_reject(xerr_idx, skb, &fl); | |
1564 | return 0; | |
1565 | } | |
1566 | return 1; | |
1567 | } | |
1da177e4 LT |
1568 | |
1569 | pol->curlft.use_time = (unsigned long)xtime.tv_sec; | |
1570 | ||
4e81bb83 MN |
1571 | pols[0] = pol; |
1572 | npols ++; | |
1573 | #ifdef CONFIG_XFRM_SUB_POLICY | |
1574 | if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) { | |
1575 | pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN, | |
1576 | &fl, family, | |
1577 | XFRM_POLICY_IN); | |
1578 | if (pols[1]) { | |
1579 | pols[1]->curlft.use_time = (unsigned long)xtime.tv_sec; | |
1580 | npols ++; | |
1581 | } | |
1582 | } | |
1583 | #endif | |
1584 | ||
1da177e4 LT |
1585 | if (pol->action == XFRM_POLICY_ALLOW) { |
1586 | struct sec_path *sp; | |
1587 | static struct sec_path dummy; | |
4e81bb83 | 1588 | struct xfrm_tmpl *tp[XFRM_MAX_DEPTH]; |
41a49cc3 | 1589 | struct xfrm_tmpl *stp[XFRM_MAX_DEPTH]; |
4e81bb83 MN |
1590 | struct xfrm_tmpl **tpp = tp; |
1591 | int ti = 0; | |
1da177e4 LT |
1592 | int i, k; |
1593 | ||
1594 | if ((sp = skb->sp) == NULL) | |
1595 | sp = &dummy; | |
1596 | ||
4e81bb83 MN |
1597 | for (pi = 0; pi < npols; pi++) { |
1598 | if (pols[pi] != pol && | |
1599 | pols[pi]->action != XFRM_POLICY_ALLOW) | |
1600 | goto reject; | |
1601 | if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) | |
1602 | goto reject_error; | |
1603 | for (i = 0; i < pols[pi]->xfrm_nr; i++) | |
1604 | tpp[ti++] = &pols[pi]->xfrm_vec[i]; | |
1605 | } | |
1606 | xfrm_nr = ti; | |
41a49cc3 MN |
1607 | if (npols > 1) { |
1608 | xfrm_tmpl_sort(stp, tpp, xfrm_nr, family); | |
1609 | tpp = stp; | |
1610 | } | |
4e81bb83 | 1611 | |
1da177e4 LT |
1612 | /* For each tunnel xfrm, find the first matching tmpl. |
1613 | * For each tmpl before that, find corresponding xfrm. | |
1614 | * Order is _important_. Later we will implement | |
1615 | * some barriers, but at the moment barriers | |
1616 | * are implied between each two transformations. | |
1617 | */ | |
4e81bb83 MN |
1618 | for (i = xfrm_nr-1, k = 0; i >= 0; i--) { |
1619 | k = xfrm_policy_ok(tpp[i], sp, k, family); | |
df0ba92a MN |
1620 | if (k < 0) { |
1621 | if (k < -1 && xerr_idxp) | |
1622 | *xerr_idxp = -(2+k); | |
1da177e4 | 1623 | goto reject; |
df0ba92a | 1624 | } |
1da177e4 LT |
1625 | } |
1626 | ||
df0ba92a | 1627 | if (secpath_has_nontransport(sp, k, xerr_idxp)) |
1da177e4 LT |
1628 | goto reject; |
1629 | ||
4e81bb83 | 1630 | xfrm_pols_put(pols, npols); |
1da177e4 LT |
1631 | return 1; |
1632 | } | |
1633 | ||
1634 | reject: | |
df0ba92a | 1635 | xfrm_secpath_reject(xerr_idx, skb, &fl); |
4e81bb83 MN |
1636 | reject_error: |
1637 | xfrm_pols_put(pols, npols); | |
1da177e4 LT |
1638 | return 0; |
1639 | } | |
1640 | EXPORT_SYMBOL(__xfrm_policy_check); | |
1641 | ||
1642 | int __xfrm_route_forward(struct sk_buff *skb, unsigned short family) | |
1643 | { | |
1644 | struct flowi fl; | |
1645 | ||
3e3850e9 | 1646 | if (xfrm_decode_session(skb, &fl, family) < 0) |
1da177e4 LT |
1647 | return 0; |
1648 | ||
1649 | return xfrm_lookup(&skb->dst, &fl, NULL, 0) == 0; | |
1650 | } | |
1651 | EXPORT_SYMBOL(__xfrm_route_forward); | |
1652 | ||
d49c73c7 DM |
1653 | /* Optimize later using cookies and generation ids. */ |
1654 | ||
1da177e4 LT |
1655 | static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie) |
1656 | { | |
d49c73c7 DM |
1657 | /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete |
1658 | * to "-1" to force all XFRM destinations to get validated by | |
1659 | * dst_ops->check on every use. We do this because when a | |
1660 | * normal route referenced by an XFRM dst is obsoleted we do | |
1661 | * not go looking around for all parent referencing XFRM dsts | |
1662 | * so that we can invalidate them. It is just too much work. | |
1663 | * Instead we make the checks here on every use. For example: | |
1664 | * | |
1665 | * XFRM dst A --> IPv4 dst X | |
1666 | * | |
1667 | * X is the "xdst->route" of A (X is also the "dst->path" of A | |
1668 | * in this example). If X is marked obsolete, "A" will not | |
1669 | * notice. That's what we are validating here via the | |
1670 | * stale_bundle() check. | |
1671 | * | |
1672 | * When a policy's bundle is pruned, we dst_free() the XFRM | |
1673 | * dst which causes it's ->obsolete field to be set to a | |
1674 | * positive non-zero integer. If an XFRM dst has been pruned | |
1675 | * like this, we want to force a new route lookup. | |
399c180a | 1676 | */ |
d49c73c7 DM |
1677 | if (dst->obsolete < 0 && !stale_bundle(dst)) |
1678 | return dst; | |
1679 | ||
1da177e4 LT |
1680 | return NULL; |
1681 | } | |
1682 | ||
1683 | static int stale_bundle(struct dst_entry *dst) | |
1684 | { | |
e53820de | 1685 | return !xfrm_bundle_ok((struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0); |
1da177e4 LT |
1686 | } |
1687 | ||
aabc9761 | 1688 | void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev) |
1da177e4 | 1689 | { |
1da177e4 LT |
1690 | while ((dst = dst->child) && dst->xfrm && dst->dev == dev) { |
1691 | dst->dev = &loopback_dev; | |
1692 | dev_hold(&loopback_dev); | |
1693 | dev_put(dev); | |
1694 | } | |
1695 | } | |
aabc9761 | 1696 | EXPORT_SYMBOL(xfrm_dst_ifdown); |
1da177e4 LT |
1697 | |
1698 | static void xfrm_link_failure(struct sk_buff *skb) | |
1699 | { | |
1700 | /* Impossible. Such dst must be popped before reaches point of failure. */ | |
1701 | return; | |
1702 | } | |
1703 | ||
1704 | static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst) | |
1705 | { | |
1706 | if (dst) { | |
1707 | if (dst->obsolete) { | |
1708 | dst_release(dst); | |
1709 | dst = NULL; | |
1710 | } | |
1711 | } | |
1712 | return dst; | |
1713 | } | |
1714 | ||
2518c7c2 DM |
1715 | static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p) |
1716 | { | |
1717 | struct dst_entry *dst, **dstp; | |
1718 | ||
1719 | write_lock(&pol->lock); | |
1720 | dstp = &pol->bundles; | |
1721 | while ((dst=*dstp) != NULL) { | |
1722 | if (func(dst)) { | |
1723 | *dstp = dst->next; | |
1724 | dst->next = *gc_list_p; | |
1725 | *gc_list_p = dst; | |
1726 | } else { | |
1727 | dstp = &dst->next; | |
1728 | } | |
1729 | } | |
1730 | write_unlock(&pol->lock); | |
1731 | } | |
1732 | ||
1da177e4 LT |
1733 | static void xfrm_prune_bundles(int (*func)(struct dst_entry *)) |
1734 | { | |
2518c7c2 DM |
1735 | struct dst_entry *gc_list = NULL; |
1736 | int dir; | |
1da177e4 LT |
1737 | |
1738 | read_lock_bh(&xfrm_policy_lock); | |
2518c7c2 DM |
1739 | for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) { |
1740 | struct xfrm_policy *pol; | |
1741 | struct hlist_node *entry; | |
1742 | struct hlist_head *table; | |
1743 | int i; | |
4e81bb83 | 1744 | |
2518c7c2 DM |
1745 | hlist_for_each_entry(pol, entry, |
1746 | &xfrm_policy_inexact[dir], bydst) | |
1747 | prune_one_bundle(pol, func, &gc_list); | |
1748 | ||
1749 | table = xfrm_policy_bydst[dir].table; | |
1750 | for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) { | |
1751 | hlist_for_each_entry(pol, entry, table + i, bydst) | |
1752 | prune_one_bundle(pol, func, &gc_list); | |
1da177e4 LT |
1753 | } |
1754 | } | |
1755 | read_unlock_bh(&xfrm_policy_lock); | |
1756 | ||
1757 | while (gc_list) { | |
2518c7c2 | 1758 | struct dst_entry *dst = gc_list; |
1da177e4 LT |
1759 | gc_list = dst->next; |
1760 | dst_free(dst); | |
1761 | } | |
1762 | } | |
1763 | ||
1764 | static int unused_bundle(struct dst_entry *dst) | |
1765 | { | |
1766 | return !atomic_read(&dst->__refcnt); | |
1767 | } | |
1768 | ||
1769 | static void __xfrm_garbage_collect(void) | |
1770 | { | |
1771 | xfrm_prune_bundles(unused_bundle); | |
1772 | } | |
1773 | ||
1c095399 | 1774 | static int xfrm_flush_bundles(void) |
1da177e4 LT |
1775 | { |
1776 | xfrm_prune_bundles(stale_bundle); | |
1777 | return 0; | |
1778 | } | |
1779 | ||
1780 | void xfrm_init_pmtu(struct dst_entry *dst) | |
1781 | { | |
1782 | do { | |
1783 | struct xfrm_dst *xdst = (struct xfrm_dst *)dst; | |
1784 | u32 pmtu, route_mtu_cached; | |
1785 | ||
1786 | pmtu = dst_mtu(dst->child); | |
1787 | xdst->child_mtu_cached = pmtu; | |
1788 | ||
1789 | pmtu = xfrm_state_mtu(dst->xfrm, pmtu); | |
1790 | ||
1791 | route_mtu_cached = dst_mtu(xdst->route); | |
1792 | xdst->route_mtu_cached = route_mtu_cached; | |
1793 | ||
1794 | if (pmtu > route_mtu_cached) | |
1795 | pmtu = route_mtu_cached; | |
1796 | ||
1797 | dst->metrics[RTAX_MTU-1] = pmtu; | |
1798 | } while ((dst = dst->next)); | |
1799 | } | |
1800 | ||
1801 | EXPORT_SYMBOL(xfrm_init_pmtu); | |
1802 | ||
1803 | /* Check that the bundle accepts the flow and its components are | |
1804 | * still valid. | |
1805 | */ | |
1806 | ||
e53820de | 1807 | int xfrm_bundle_ok(struct xfrm_dst *first, struct flowi *fl, int family, int strict) |
1da177e4 LT |
1808 | { |
1809 | struct dst_entry *dst = &first->u.dst; | |
1810 | struct xfrm_dst *last; | |
1811 | u32 mtu; | |
1812 | ||
92d63dec | 1813 | if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) || |
1da177e4 LT |
1814 | (dst->dev && !netif_running(dst->dev))) |
1815 | return 0; | |
1816 | ||
1817 | last = NULL; | |
1818 | ||
1819 | do { | |
1820 | struct xfrm_dst *xdst = (struct xfrm_dst *)dst; | |
1821 | ||
1822 | if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family)) | |
1823 | return 0; | |
e0d1caa7 VY |
1824 | if (fl && !security_xfrm_flow_state_match(fl, dst->xfrm)) |
1825 | return 0; | |
1da177e4 LT |
1826 | if (dst->xfrm->km.state != XFRM_STATE_VALID) |
1827 | return 0; | |
9d4a706d DM |
1828 | if (xdst->genid != dst->xfrm->genid) |
1829 | return 0; | |
e53820de MN |
1830 | |
1831 | if (strict && fl && dst->xfrm->props.mode != XFRM_MODE_TUNNEL && | |
1832 | !xfrm_state_addr_flow_check(dst->xfrm, fl, family)) | |
1833 | return 0; | |
1da177e4 LT |
1834 | |
1835 | mtu = dst_mtu(dst->child); | |
1836 | if (xdst->child_mtu_cached != mtu) { | |
1837 | last = xdst; | |
1838 | xdst->child_mtu_cached = mtu; | |
1839 | } | |
1840 | ||
92d63dec | 1841 | if (!dst_check(xdst->route, xdst->route_cookie)) |
1da177e4 LT |
1842 | return 0; |
1843 | mtu = dst_mtu(xdst->route); | |
1844 | if (xdst->route_mtu_cached != mtu) { | |
1845 | last = xdst; | |
1846 | xdst->route_mtu_cached = mtu; | |
1847 | } | |
1848 | ||
1849 | dst = dst->child; | |
1850 | } while (dst->xfrm); | |
1851 | ||
1852 | if (likely(!last)) | |
1853 | return 1; | |
1854 | ||
1855 | mtu = last->child_mtu_cached; | |
1856 | for (;;) { | |
1857 | dst = &last->u.dst; | |
1858 | ||
1859 | mtu = xfrm_state_mtu(dst->xfrm, mtu); | |
1860 | if (mtu > last->route_mtu_cached) | |
1861 | mtu = last->route_mtu_cached; | |
1862 | dst->metrics[RTAX_MTU-1] = mtu; | |
1863 | ||
1864 | if (last == first) | |
1865 | break; | |
1866 | ||
1867 | last = last->u.next; | |
1868 | last->child_mtu_cached = mtu; | |
1869 | } | |
1870 | ||
1871 | return 1; | |
1872 | } | |
1873 | ||
1874 | EXPORT_SYMBOL(xfrm_bundle_ok); | |
1875 | ||
1da177e4 LT |
1876 | int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo) |
1877 | { | |
1878 | int err = 0; | |
1879 | if (unlikely(afinfo == NULL)) | |
1880 | return -EINVAL; | |
1881 | if (unlikely(afinfo->family >= NPROTO)) | |
1882 | return -EAFNOSUPPORT; | |
e959d812 | 1883 | write_lock_bh(&xfrm_policy_afinfo_lock); |
1da177e4 LT |
1884 | if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL)) |
1885 | err = -ENOBUFS; | |
1886 | else { | |
1887 | struct dst_ops *dst_ops = afinfo->dst_ops; | |
1888 | if (likely(dst_ops->kmem_cachep == NULL)) | |
1889 | dst_ops->kmem_cachep = xfrm_dst_cache; | |
1890 | if (likely(dst_ops->check == NULL)) | |
1891 | dst_ops->check = xfrm_dst_check; | |
1da177e4 LT |
1892 | if (likely(dst_ops->negative_advice == NULL)) |
1893 | dst_ops->negative_advice = xfrm_negative_advice; | |
1894 | if (likely(dst_ops->link_failure == NULL)) | |
1895 | dst_ops->link_failure = xfrm_link_failure; | |
1da177e4 LT |
1896 | if (likely(afinfo->garbage_collect == NULL)) |
1897 | afinfo->garbage_collect = __xfrm_garbage_collect; | |
1898 | xfrm_policy_afinfo[afinfo->family] = afinfo; | |
1899 | } | |
e959d812 | 1900 | write_unlock_bh(&xfrm_policy_afinfo_lock); |
1da177e4 LT |
1901 | return err; |
1902 | } | |
1903 | EXPORT_SYMBOL(xfrm_policy_register_afinfo); | |
1904 | ||
1905 | int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo) | |
1906 | { | |
1907 | int err = 0; | |
1908 | if (unlikely(afinfo == NULL)) | |
1909 | return -EINVAL; | |
1910 | if (unlikely(afinfo->family >= NPROTO)) | |
1911 | return -EAFNOSUPPORT; | |
e959d812 | 1912 | write_lock_bh(&xfrm_policy_afinfo_lock); |
1da177e4 LT |
1913 | if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) { |
1914 | if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo)) | |
1915 | err = -EINVAL; | |
1916 | else { | |
1917 | struct dst_ops *dst_ops = afinfo->dst_ops; | |
1918 | xfrm_policy_afinfo[afinfo->family] = NULL; | |
1919 | dst_ops->kmem_cachep = NULL; | |
1920 | dst_ops->check = NULL; | |
1da177e4 LT |
1921 | dst_ops->negative_advice = NULL; |
1922 | dst_ops->link_failure = NULL; | |
1da177e4 LT |
1923 | afinfo->garbage_collect = NULL; |
1924 | } | |
1925 | } | |
e959d812 | 1926 | write_unlock_bh(&xfrm_policy_afinfo_lock); |
1da177e4 LT |
1927 | return err; |
1928 | } | |
1929 | EXPORT_SYMBOL(xfrm_policy_unregister_afinfo); | |
1930 | ||
1931 | static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family) | |
1932 | { | |
1933 | struct xfrm_policy_afinfo *afinfo; | |
1934 | if (unlikely(family >= NPROTO)) | |
1935 | return NULL; | |
1936 | read_lock(&xfrm_policy_afinfo_lock); | |
1937 | afinfo = xfrm_policy_afinfo[family]; | |
546be240 HX |
1938 | if (unlikely(!afinfo)) |
1939 | read_unlock(&xfrm_policy_afinfo_lock); | |
1da177e4 LT |
1940 | return afinfo; |
1941 | } | |
1942 | ||
1943 | static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo) | |
1944 | { | |
546be240 HX |
1945 | read_unlock(&xfrm_policy_afinfo_lock); |
1946 | } | |
1947 | ||
1948 | static struct xfrm_policy_afinfo *xfrm_policy_lock_afinfo(unsigned int family) | |
1949 | { | |
1950 | struct xfrm_policy_afinfo *afinfo; | |
1951 | if (unlikely(family >= NPROTO)) | |
1952 | return NULL; | |
1953 | write_lock_bh(&xfrm_policy_afinfo_lock); | |
1954 | afinfo = xfrm_policy_afinfo[family]; | |
1955 | if (unlikely(!afinfo)) | |
1956 | write_unlock_bh(&xfrm_policy_afinfo_lock); | |
1957 | return afinfo; | |
1958 | } | |
1959 | ||
1960 | static void xfrm_policy_unlock_afinfo(struct xfrm_policy_afinfo *afinfo) | |
1961 | { | |
1962 | write_unlock_bh(&xfrm_policy_afinfo_lock); | |
1da177e4 LT |
1963 | } |
1964 | ||
1965 | static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1966 | { | |
1967 | switch (event) { | |
1968 | case NETDEV_DOWN: | |
1969 | xfrm_flush_bundles(); | |
1970 | } | |
1971 | return NOTIFY_DONE; | |
1972 | } | |
1973 | ||
1974 | static struct notifier_block xfrm_dev_notifier = { | |
1975 | xfrm_dev_event, | |
1976 | NULL, | |
1977 | 0 | |
1978 | }; | |
1979 | ||
1980 | static void __init xfrm_policy_init(void) | |
1981 | { | |
2518c7c2 DM |
1982 | unsigned int hmask, sz; |
1983 | int dir; | |
1984 | ||
1da177e4 LT |
1985 | xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache", |
1986 | sizeof(struct xfrm_dst), | |
1987 | 0, SLAB_HWCACHE_ALIGN, | |
1988 | NULL, NULL); | |
1989 | if (!xfrm_dst_cache) | |
1990 | panic("XFRM: failed to allocate xfrm_dst_cache\n"); | |
1991 | ||
2518c7c2 DM |
1992 | hmask = 8 - 1; |
1993 | sz = (hmask+1) * sizeof(struct hlist_head); | |
1994 | ||
44e36b42 | 1995 | xfrm_policy_byidx = xfrm_hash_alloc(sz); |
2518c7c2 DM |
1996 | xfrm_idx_hmask = hmask; |
1997 | if (!xfrm_policy_byidx) | |
1998 | panic("XFRM: failed to allocate byidx hash\n"); | |
1999 | ||
2000 | for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) { | |
2001 | struct xfrm_policy_hash *htab; | |
2002 | ||
2003 | INIT_HLIST_HEAD(&xfrm_policy_inexact[dir]); | |
2004 | ||
2005 | htab = &xfrm_policy_bydst[dir]; | |
44e36b42 | 2006 | htab->table = xfrm_hash_alloc(sz); |
2518c7c2 DM |
2007 | htab->hmask = hmask; |
2008 | if (!htab->table) | |
2009 | panic("XFRM: failed to allocate bydst hash\n"); | |
2010 | } | |
2011 | ||
1da177e4 LT |
2012 | INIT_WORK(&xfrm_policy_gc_work, xfrm_policy_gc_task, NULL); |
2013 | register_netdevice_notifier(&xfrm_dev_notifier); | |
2014 | } | |
2015 | ||
2016 | void __init xfrm_init(void) | |
2017 | { | |
2018 | xfrm_state_init(); | |
2019 | xfrm_policy_init(); | |
2020 | xfrm_input_init(); | |
2021 | } | |
2022 |