?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
writeback: split writeback_inodes_wb
[linux-2.6/next.git] / net / xfrm / xfrm_policy.c
blobaf1c173be4ad88906d55d7efa24fb47fcd0cf19d
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
13 *
14 */
15
16 #include <linux/err.h>
17 #include <linux/slab.h>
18 #include <linux/kmod.h>
19 #include <linux/list.h>
20 #include <linux/spinlock.h>
21 #include <linux/workqueue.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/module.h>
26 #include <linux/cache.h>
27 #include <linux/audit.h>
28 #include <net/dst.h>
29 #include <net/xfrm.h>
30 #include <net/ip.h>
31 #ifdef CONFIG_XFRM_STATISTICS
32 #include <net/snmp.h>
33 #endif
34
35 #include "xfrm_hash.h"
36
37 DEFINE_MUTEX(xfrm_cfg_mutex);
38 EXPORT_SYMBOL(xfrm_cfg_mutex);
39
40 static DEFINE_SPINLOCK(xfrm_policy_sk_bundle_lock);
41 static struct dst_entry *xfrm_policy_sk_bundles;
42 static DEFINE_RWLOCK(xfrm_policy_lock);
43
44 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
45 static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
46
47 static struct kmem_cache *xfrm_dst_cache __read_mostly;
48
49 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
50 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
51 static void xfrm_init_pmtu(struct dst_entry *dst);
52 static int stale_bundle(struct dst_entry *dst);
53
54 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
55 int dir);
56
57 static inline int
58 __xfrm4_selector_match(struct xfrm_selector *sel, struct flowi *fl)
59 {
60 return addr_match(&fl->fl4_dst, &sel->daddr, sel->prefixlen_d) &&
61 addr_match(&fl->fl4_src, &sel->saddr, sel->prefixlen_s) &&
62 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
63 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
64 (fl->proto == sel->proto || !sel->proto) &&
65 (fl->oif == sel->ifindex || !sel->ifindex);
66 }
67
68 static inline int
69 __xfrm6_selector_match(struct xfrm_selector *sel, struct flowi *fl)
70 {
71 return addr_match(&fl->fl6_dst, &sel->daddr, sel->prefixlen_d) &&
72 addr_match(&fl->fl6_src, &sel->saddr, sel->prefixlen_s) &&
73 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
74 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
75 (fl->proto == sel->proto || !sel->proto) &&
76 (fl->oif == sel->ifindex || !sel->ifindex);
77 }
78
79 int xfrm_selector_match(struct xfrm_selector *sel, struct flowi *fl,
80 unsigned short family)
81 {
82 switch (family) {
83 case AF_INET:
84 return __xfrm4_selector_match(sel, fl);
85 case AF_INET6:
86 return __xfrm6_selector_match(sel, fl);
87 }
88 return 0;
89 }
90
91 static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
92 xfrm_address_t *saddr,
93 xfrm_address_t *daddr,
94 int family)
95 {
96 struct xfrm_policy_afinfo *afinfo;
97 struct dst_entry *dst;
98
99 afinfo = xfrm_policy_get_afinfo(family);
100 if (unlikely(afinfo == NULL))
101 return ERR_PTR(-EAFNOSUPPORT);
102
103 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
104
105 xfrm_policy_put_afinfo(afinfo);
106
107 return dst;
108 }
109
110 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
111 xfrm_address_t *prev_saddr,
112 xfrm_address_t *prev_daddr,
113 int family)
114 {
115 struct net *net = xs_net(x);
116 xfrm_address_t *saddr = &x->props.saddr;
117 xfrm_address_t *daddr = &x->id.daddr;
118 struct dst_entry *dst;
119
120 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
121 saddr = x->coaddr;
122 daddr = prev_daddr;
123 }
124 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
125 saddr = prev_saddr;
126 daddr = x->coaddr;
127 }
128
129 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
130
131 if (!IS_ERR(dst)) {
132 if (prev_saddr != saddr)
133 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
134 if (prev_daddr != daddr)
135 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
136 }
137
138 return dst;
139 }
140
141 static inline unsigned long make_jiffies(long secs)
142 {
143 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
144 return MAX_SCHEDULE_TIMEOUT-1;
145 else
146 return secs*HZ;
147 }
148
149 static void xfrm_policy_timer(unsigned long data)
150 {
151 struct xfrm_policy *xp = (struct xfrm_policy*)data;
152 unsigned long now = get_seconds();
153 long next = LONG_MAX;
154 int warn = 0;
155 int dir;
156
157 read_lock(&xp->lock);
158
159 if (unlikely(xp->walk.dead))
160 goto out;
161
162 dir = xfrm_policy_id2dir(xp->index);
163
164 if (xp->lft.hard_add_expires_seconds) {
165 long tmo = xp->lft.hard_add_expires_seconds +
166 xp->curlft.add_time - now;
167 if (tmo <= 0)
168 goto expired;
169 if (tmo < next)
170 next = tmo;
171 }
172 if (xp->lft.hard_use_expires_seconds) {
173 long tmo = xp->lft.hard_use_expires_seconds +
174 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
175 if (tmo <= 0)
176 goto expired;
177 if (tmo < next)
178 next = tmo;
179 }
180 if (xp->lft.soft_add_expires_seconds) {
181 long tmo = xp->lft.soft_add_expires_seconds +
182 xp->curlft.add_time - now;
183 if (tmo <= 0) {
184 warn = 1;
185 tmo = XFRM_KM_TIMEOUT;
186 }
187 if (tmo < next)
188 next = tmo;
189 }
190 if (xp->lft.soft_use_expires_seconds) {
191 long tmo = xp->lft.soft_use_expires_seconds +
192 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
193 if (tmo <= 0) {
194 warn = 1;
195 tmo = XFRM_KM_TIMEOUT;
196 }
197 if (tmo < next)
198 next = tmo;
199 }
200
201 if (warn)
202 km_policy_expired(xp, dir, 0, 0);
203 if (next != LONG_MAX &&
204 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
205 xfrm_pol_hold(xp);
206
207 out:
208 read_unlock(&xp->lock);
209 xfrm_pol_put(xp);
210 return;
211
212 expired:
213 read_unlock(&xp->lock);
214 if (!xfrm_policy_delete(xp, dir))
215 km_policy_expired(xp, dir, 1, 0);
216 xfrm_pol_put(xp);
217 }
218
219 static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
220 {
221 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
222
223 if (unlikely(pol->walk.dead))
224 flo = NULL;
225 else
226 xfrm_pol_hold(pol);
227
228 return flo;
229 }
230
231 static int xfrm_policy_flo_check(struct flow_cache_object *flo)
232 {
233 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
234
235 return !pol->walk.dead;
236 }
237
238 static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
239 {
240 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
241 }
242
243 static const struct flow_cache_ops xfrm_policy_fc_ops = {
244 .get = xfrm_policy_flo_get,
245 .check = xfrm_policy_flo_check,
246 .delete = xfrm_policy_flo_delete,
247 };
248
249 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
250 * SPD calls.
251 */
252
253 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
254 {
255 struct xfrm_policy *policy;
256
257 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
258
259 if (policy) {
260 write_pnet(&policy->xp_net, net);
261 INIT_LIST_HEAD(&policy->walk.all);
262 INIT_HLIST_NODE(&policy->bydst);
263 INIT_HLIST_NODE(&policy->byidx);
264 rwlock_init(&policy->lock);
265 atomic_set(&policy->refcnt, 1);
266 setup_timer(&policy->timer, xfrm_policy_timer,
267 (unsigned long)policy);
268 policy->flo.ops = &xfrm_policy_fc_ops;
269 }
270 return policy;
271 }
272 EXPORT_SYMBOL(xfrm_policy_alloc);
273
274 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
275
276 void xfrm_policy_destroy(struct xfrm_policy *policy)
277 {
278 BUG_ON(!policy->walk.dead);
279
280 if (del_timer(&policy->timer))
281 BUG();
282
283 security_xfrm_policy_free(policy->security);
284 kfree(policy);
285 }
286 EXPORT_SYMBOL(xfrm_policy_destroy);
287
288 /* Rule must be locked. Release descentant resources, announce
289 * entry dead. The rule must be unlinked from lists to the moment.
290 */
291
292 static void xfrm_policy_kill(struct xfrm_policy *policy)
293 {
294 policy->walk.dead = 1;
295
296 atomic_inc(&policy->genid);
297
298 if (del_timer(&policy->timer))
299 xfrm_pol_put(policy);
300
301 xfrm_pol_put(policy);
302 }
303
304 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
305
306 static inline unsigned int idx_hash(struct net *net, u32 index)
307 {
308 return __idx_hash(index, net->xfrm.policy_idx_hmask);
309 }
310
311 static struct hlist_head *policy_hash_bysel(struct net *net, struct xfrm_selector *sel, unsigned short family, int dir)
312 {
313 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
314 unsigned int hash = __sel_hash(sel, family, hmask);
315
316 return (hash == hmask + 1 ?
317 &net->xfrm.policy_inexact[dir] :
318 net->xfrm.policy_bydst[dir].table + hash);
319 }
320
321 static struct hlist_head *policy_hash_direct(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
322 {
323 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
324 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
325
326 return net->xfrm.policy_bydst[dir].table + hash;
327 }
328
329 static void xfrm_dst_hash_transfer(struct hlist_head *list,
330 struct hlist_head *ndsttable,
331 unsigned int nhashmask)
332 {
333 struct hlist_node *entry, *tmp, *entry0 = NULL;
334 struct xfrm_policy *pol;
335 unsigned int h0 = 0;
336
337 redo:
338 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
339 unsigned int h;
340
341 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
342 pol->family, nhashmask);
343 if (!entry0) {
344 hlist_del(entry);
345 hlist_add_head(&pol->bydst, ndsttable+h);
346 h0 = h;
347 } else {
348 if (h != h0)
349 continue;
350 hlist_del(entry);
351 hlist_add_after(entry0, &pol->bydst);
352 }
353 entry0 = entry;
354 }
355 if (!hlist_empty(list)) {
356 entry0 = NULL;
357 goto redo;
358 }
359 }
360
361 static void xfrm_idx_hash_transfer(struct hlist_head *list,
362 struct hlist_head *nidxtable,
363 unsigned int nhashmask)
364 {
365 struct hlist_node *entry, *tmp;
366 struct xfrm_policy *pol;
367
368 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
369 unsigned int h;
370
371 h = __idx_hash(pol->index, nhashmask);
372 hlist_add_head(&pol->byidx, nidxtable+h);
373 }
374 }
375
376 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
377 {
378 return ((old_hmask + 1) << 1) - 1;
379 }
380
381 static void xfrm_bydst_resize(struct net *net, int dir)
382 {
383 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
384 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
385 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
386 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
387 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
388 int i;
389
390 if (!ndst)
391 return;
392
393 write_lock_bh(&xfrm_policy_lock);
394
395 for (i = hmask; i >= 0; i--)
396 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
397
398 net->xfrm.policy_bydst[dir].table = ndst;
399 net->xfrm.policy_bydst[dir].hmask = nhashmask;
400
401 write_unlock_bh(&xfrm_policy_lock);
402
403 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
404 }
405
406 static void xfrm_byidx_resize(struct net *net, int total)
407 {
408 unsigned int hmask = net->xfrm.policy_idx_hmask;
409 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
410 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
411 struct hlist_head *oidx = net->xfrm.policy_byidx;
412 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
413 int i;
414
415 if (!nidx)
416 return;
417
418 write_lock_bh(&xfrm_policy_lock);
419
420 for (i = hmask; i >= 0; i--)
421 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
422
423 net->xfrm.policy_byidx = nidx;
424 net->xfrm.policy_idx_hmask = nhashmask;
425
426 write_unlock_bh(&xfrm_policy_lock);
427
428 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
429 }
430
431 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
432 {
433 unsigned int cnt = net->xfrm.policy_count[dir];
434 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
435
436 if (total)
437 *total += cnt;
438
439 if ((hmask + 1) < xfrm_policy_hashmax &&
440 cnt > hmask)
441 return 1;
442
443 return 0;
444 }
445
446 static inline int xfrm_byidx_should_resize(struct net *net, int total)
447 {
448 unsigned int hmask = net->xfrm.policy_idx_hmask;
449
450 if ((hmask + 1) < xfrm_policy_hashmax &&
451 total > hmask)
452 return 1;
453
454 return 0;
455 }
456
457 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
458 {
459 read_lock_bh(&xfrm_policy_lock);
460 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
461 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
462 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
463 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
464 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
465 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
466 si->spdhcnt = net->xfrm.policy_idx_hmask;
467 si->spdhmcnt = xfrm_policy_hashmax;
468 read_unlock_bh(&xfrm_policy_lock);
469 }
470 EXPORT_SYMBOL(xfrm_spd_getinfo);
471
472 static DEFINE_MUTEX(hash_resize_mutex);
473 static void xfrm_hash_resize(struct work_struct *work)
474 {
475 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
476 int dir, total;
477
478 mutex_lock(&hash_resize_mutex);
479
480 total = 0;
481 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
482 if (xfrm_bydst_should_resize(net, dir, &total))
483 xfrm_bydst_resize(net, dir);
484 }
485 if (xfrm_byidx_should_resize(net, total))
486 xfrm_byidx_resize(net, total);
487
488 mutex_unlock(&hash_resize_mutex);
489 }
490
491 /* Generate new index... KAME seems to generate them ordered by cost
492 * of an absolute inpredictability of ordering of rules. This will not pass. */
493 static u32 xfrm_gen_index(struct net *net, int dir)
494 {
495 static u32 idx_generator;
496
497 for (;;) {
498 struct hlist_node *entry;
499 struct hlist_head *list;
500 struct xfrm_policy *p;
501 u32 idx;
502 int found;
503
504 idx = (idx_generator | dir);
505 idx_generator += 8;
506 if (idx == 0)
507 idx = 8;
508 list = net->xfrm.policy_byidx + idx_hash(net, idx);
509 found = 0;
510 hlist_for_each_entry(p, entry, list, byidx) {
511 if (p->index == idx) {
512 found = 1;
513 break;
514 }
515 }
516 if (!found)
517 return idx;
518 }
519 }
520
521 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
522 {
523 u32 *p1 = (u32 *) s1;
524 u32 *p2 = (u32 *) s2;
525 int len = sizeof(struct xfrm_selector) / sizeof(u32);
526 int i;
527
528 for (i = 0; i < len; i++) {
529 if (p1[i] != p2[i])
530 return 1;
531 }
532
533 return 0;
534 }
535
536 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
537 {
538 struct net *net = xp_net(policy);
539 struct xfrm_policy *pol;
540 struct xfrm_policy *delpol;
541 struct hlist_head *chain;
542 struct hlist_node *entry, *newpos;
543 u32 mark = policy->mark.v & policy->mark.m;
544
545 write_lock_bh(&xfrm_policy_lock);
546 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
547 delpol = NULL;
548 newpos = NULL;
549 hlist_for_each_entry(pol, entry, chain, bydst) {
550 if (pol->type == policy->type &&
551 !selector_cmp(&pol->selector, &policy->selector) &&
552 (mark & pol->mark.m) == pol->mark.v &&
553 xfrm_sec_ctx_match(pol->security, policy->security) &&
554 !WARN_ON(delpol)) {
555 if (excl) {
556 write_unlock_bh(&xfrm_policy_lock);
557 return -EEXIST;
558 }
559 delpol = pol;
560 if (policy->priority > pol->priority)
561 continue;
562 } else if (policy->priority >= pol->priority) {
563 newpos = &pol->bydst;
564 continue;
565 }
566 if (delpol)
567 break;
568 }
569 if (newpos)
570 hlist_add_after(newpos, &policy->bydst);
571 else
572 hlist_add_head(&policy->bydst, chain);
573 xfrm_pol_hold(policy);
574 net->xfrm.policy_count[dir]++;
575 atomic_inc(&flow_cache_genid);
576 if (delpol)
577 __xfrm_policy_unlink(delpol, dir);
578 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
579 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
580 policy->curlft.add_time = get_seconds();
581 policy->curlft.use_time = 0;
582 if (!mod_timer(&policy->timer, jiffies + HZ))
583 xfrm_pol_hold(policy);
584 list_add(&policy->walk.all, &net->xfrm.policy_all);
585 write_unlock_bh(&xfrm_policy_lock);
586
587 if (delpol)
588 xfrm_policy_kill(delpol);
589 else if (xfrm_bydst_should_resize(net, dir, NULL))
590 schedule_work(&net->xfrm.policy_hash_work);
591
592 return 0;
593 }
594 EXPORT_SYMBOL(xfrm_policy_insert);
595
596 struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
597 int dir, struct xfrm_selector *sel,
598 struct xfrm_sec_ctx *ctx, int delete,
599 int *err)
600 {
601 struct xfrm_policy *pol, *ret;
602 struct hlist_head *chain;
603 struct hlist_node *entry;
604
605 *err = 0;
606 write_lock_bh(&xfrm_policy_lock);
607 chain = policy_hash_bysel(net, sel, sel->family, dir);
608 ret = NULL;
609 hlist_for_each_entry(pol, entry, chain, bydst) {
610 if (pol->type == type &&
611 (mark & pol->mark.m) == pol->mark.v &&
612 !selector_cmp(sel, &pol->selector) &&
613 xfrm_sec_ctx_match(ctx, pol->security)) {
614 xfrm_pol_hold(pol);
615 if (delete) {
616 *err = security_xfrm_policy_delete(
617 pol->security);
618 if (*err) {
619 write_unlock_bh(&xfrm_policy_lock);
620 return pol;
621 }
622 __xfrm_policy_unlink(pol, dir);
623 }
624 ret = pol;
625 break;
626 }
627 }
628 write_unlock_bh(&xfrm_policy_lock);
629
630 if (ret && delete)
631 xfrm_policy_kill(ret);
632 return ret;
633 }
634 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
635
636 struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
637 int dir, u32 id, int delete, int *err)
638 {
639 struct xfrm_policy *pol, *ret;
640 struct hlist_head *chain;
641 struct hlist_node *entry;
642
643 *err = -ENOENT;
644 if (xfrm_policy_id2dir(id) != dir)
645 return NULL;
646
647 *err = 0;
648 write_lock_bh(&xfrm_policy_lock);
649 chain = net->xfrm.policy_byidx + idx_hash(net, id);
650 ret = NULL;
651 hlist_for_each_entry(pol, entry, chain, byidx) {
652 if (pol->type == type && pol->index == id &&
653 (mark & pol->mark.m) == pol->mark.v) {
654 xfrm_pol_hold(pol);
655 if (delete) {
656 *err = security_xfrm_policy_delete(
657 pol->security);
658 if (*err) {
659 write_unlock_bh(&xfrm_policy_lock);
660 return pol;
661 }
662 __xfrm_policy_unlink(pol, dir);
663 }
664 ret = pol;
665 break;
666 }
667 }
668 write_unlock_bh(&xfrm_policy_lock);
669
670 if (ret && delete)
671 xfrm_policy_kill(ret);
672 return ret;
673 }
674 EXPORT_SYMBOL(xfrm_policy_byid);
675
676 #ifdef CONFIG_SECURITY_NETWORK_XFRM
677 static inline int
678 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
679 {
680 int dir, err = 0;
681
682 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
683 struct xfrm_policy *pol;
684 struct hlist_node *entry;
685 int i;
686
687 hlist_for_each_entry(pol, entry,
688 &net->xfrm.policy_inexact[dir], bydst) {
689 if (pol->type != type)
690 continue;
691 err = security_xfrm_policy_delete(pol->security);
692 if (err) {
693 xfrm_audit_policy_delete(pol, 0,
694 audit_info->loginuid,
695 audit_info->sessionid,
696 audit_info->secid);
697 return err;
698 }
699 }
700 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
701 hlist_for_each_entry(pol, entry,
702 net->xfrm.policy_bydst[dir].table + i,
703 bydst) {
704 if (pol->type != type)
705 continue;
706 err = security_xfrm_policy_delete(
707 pol->security);
708 if (err) {
709 xfrm_audit_policy_delete(pol, 0,
710 audit_info->loginuid,
711 audit_info->sessionid,
712 audit_info->secid);
713 return err;
714 }
715 }
716 }
717 }
718 return err;
719 }
720 #else
721 static inline int
722 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
723 {
724 return 0;
725 }
726 #endif
727
728 int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
729 {
730 int dir, err = 0, cnt = 0;
731
732 write_lock_bh(&xfrm_policy_lock);
733
734 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
735 if (err)
736 goto out;
737
738 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
739 struct xfrm_policy *pol;
740 struct hlist_node *entry;
741 int i;
742
743 again1:
744 hlist_for_each_entry(pol, entry,
745 &net->xfrm.policy_inexact[dir], bydst) {
746 if (pol->type != type)
747 continue;
748 __xfrm_policy_unlink(pol, dir);
749 write_unlock_bh(&xfrm_policy_lock);
750 cnt++;
751
752 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
753 audit_info->sessionid,
754 audit_info->secid);
755
756 xfrm_policy_kill(pol);
757
758 write_lock_bh(&xfrm_policy_lock);
759 goto again1;
760 }
761
762 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
763 again2:
764 hlist_for_each_entry(pol, entry,
765 net->xfrm.policy_bydst[dir].table + i,
766 bydst) {
767 if (pol->type != type)
768 continue;
769 __xfrm_policy_unlink(pol, dir);
770 write_unlock_bh(&xfrm_policy_lock);
771 cnt++;
772
773 xfrm_audit_policy_delete(pol, 1,
774 audit_info->loginuid,
775 audit_info->sessionid,
776 audit_info->secid);
777 xfrm_policy_kill(pol);
778
779 write_lock_bh(&xfrm_policy_lock);
780 goto again2;
781 }
782 }
783
784 }
785 if (!cnt)
786 err = -ESRCH;
787 out:
788 write_unlock_bh(&xfrm_policy_lock);
789 return err;
790 }
791 EXPORT_SYMBOL(xfrm_policy_flush);
792
793 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
794 int (*func)(struct xfrm_policy *, int, int, void*),
795 void *data)
796 {
797 struct xfrm_policy *pol;
798 struct xfrm_policy_walk_entry *x;
799 int error = 0;
800
801 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
802 walk->type != XFRM_POLICY_TYPE_ANY)
803 return -EINVAL;
804
805 if (list_empty(&walk->walk.all) && walk->seq != 0)
806 return 0;
807
808 write_lock_bh(&xfrm_policy_lock);
809 if (list_empty(&walk->walk.all))
810 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
811 else
812 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
813 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
814 if (x->dead)
815 continue;
816 pol = container_of(x, struct xfrm_policy, walk);
817 if (walk->type != XFRM_POLICY_TYPE_ANY &&
818 walk->type != pol->type)
819 continue;
820 error = func(pol, xfrm_policy_id2dir(pol->index),
821 walk->seq, data);
822 if (error) {
823 list_move_tail(&walk->walk.all, &x->all);
824 goto out;
825 }
826 walk->seq++;
827 }
828 if (walk->seq == 0) {
829 error = -ENOENT;
830 goto out;
831 }
832 list_del_init(&walk->walk.all);
833 out:
834 write_unlock_bh(&xfrm_policy_lock);
835 return error;
836 }
837 EXPORT_SYMBOL(xfrm_policy_walk);
838
839 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
840 {
841 INIT_LIST_HEAD(&walk->walk.all);
842 walk->walk.dead = 1;
843 walk->type = type;
844 walk->seq = 0;
845 }
846 EXPORT_SYMBOL(xfrm_policy_walk_init);
847
848 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
849 {
850 if (list_empty(&walk->walk.all))
851 return;
852
853 write_lock_bh(&xfrm_policy_lock);
854 list_del(&walk->walk.all);
855 write_unlock_bh(&xfrm_policy_lock);
856 }
857 EXPORT_SYMBOL(xfrm_policy_walk_done);
858
859 /*
860 * Find policy to apply to this flow.
861 *
862 * Returns 0 if policy found, else an -errno.
863 */
864 static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
865 u8 type, u16 family, int dir)
866 {
867 struct xfrm_selector *sel = &pol->selector;
868 int match, ret = -ESRCH;
869
870 if (pol->family != family ||
871 (fl->mark & pol->mark.m) != pol->mark.v ||
872 pol->type != type)
873 return ret;
874
875 match = xfrm_selector_match(sel, fl, family);
876 if (match)
877 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
878 dir);
879
880 return ret;
881 }
882
883 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
884 struct flowi *fl,
885 u16 family, u8 dir)
886 {
887 int err;
888 struct xfrm_policy *pol, *ret;
889 xfrm_address_t *daddr, *saddr;
890 struct hlist_node *entry;
891 struct hlist_head *chain;
892 u32 priority = ~0U;
893
894 daddr = xfrm_flowi_daddr(fl, family);
895 saddr = xfrm_flowi_saddr(fl, family);
896 if (unlikely(!daddr || !saddr))
897 return NULL;
898
899 read_lock_bh(&xfrm_policy_lock);
900 chain = policy_hash_direct(net, daddr, saddr, family, dir);
901 ret = NULL;
902 hlist_for_each_entry(pol, entry, chain, bydst) {
903 err = xfrm_policy_match(pol, fl, type, family, dir);
904 if (err) {
905 if (err == -ESRCH)
906 continue;
907 else {
908 ret = ERR_PTR(err);
909 goto fail;
910 }
911 } else {
912 ret = pol;
913 priority = ret->priority;
914 break;
915 }
916 }
917 chain = &net->xfrm.policy_inexact[dir];
918 hlist_for_each_entry(pol, entry, chain, bydst) {
919 err = xfrm_policy_match(pol, fl, type, family, dir);
920 if (err) {
921 if (err == -ESRCH)
922 continue;
923 else {
924 ret = ERR_PTR(err);
925 goto fail;
926 }
927 } else if (pol->priority < priority) {
928 ret = pol;
929 break;
930 }
931 }
932 if (ret)
933 xfrm_pol_hold(ret);
934 fail:
935 read_unlock_bh(&xfrm_policy_lock);
936
937 return ret;
938 }
939
940 static struct xfrm_policy *
941 __xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family, u8 dir)
942 {
943 #ifdef CONFIG_XFRM_SUB_POLICY
944 struct xfrm_policy *pol;
945
946 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
947 if (pol != NULL)
948 return pol;
949 #endif
950 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
951 }
952
953 static struct flow_cache_object *
954 xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family,
955 u8 dir, struct flow_cache_object *old_obj, void *ctx)
956 {
957 struct xfrm_policy *pol;
958
959 if (old_obj)
960 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
961
962 pol = __xfrm_policy_lookup(net, fl, family, dir);
963 if (IS_ERR_OR_NULL(pol))
964 return ERR_CAST(pol);
965
966 /* Resolver returns two references:
967 * one for cache and one for caller of flow_cache_lookup() */
968 xfrm_pol_hold(pol);
969
970 return &pol->flo;
971 }
972
973 static inline int policy_to_flow_dir(int dir)
974 {
975 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
976 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
977 XFRM_POLICY_FWD == FLOW_DIR_FWD)
978 return dir;
979 switch (dir) {
980 default:
981 case XFRM_POLICY_IN:
982 return FLOW_DIR_IN;
983 case XFRM_POLICY_OUT:
984 return FLOW_DIR_OUT;
985 case XFRM_POLICY_FWD:
986 return FLOW_DIR_FWD;
987 }
988 }
989
990 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
991 {
992 struct xfrm_policy *pol;
993
994 read_lock_bh(&xfrm_policy_lock);
995 if ((pol = sk->sk_policy[dir]) != NULL) {
996 int match = xfrm_selector_match(&pol->selector, fl,
997 sk->sk_family);
998 int err = 0;
999
1000 if (match) {
1001 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1002 pol = NULL;
1003 goto out;
1004 }
1005 err = security_xfrm_policy_lookup(pol->security,
1006 fl->secid,
1007 policy_to_flow_dir(dir));
1008 if (!err)
1009 xfrm_pol_hold(pol);
1010 else if (err == -ESRCH)
1011 pol = NULL;
1012 else
1013 pol = ERR_PTR(err);
1014 } else
1015 pol = NULL;
1016 }
1017 out:
1018 read_unlock_bh(&xfrm_policy_lock);
1019 return pol;
1020 }
1021
1022 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1023 {
1024 struct net *net = xp_net(pol);
1025 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1026 pol->family, dir);
1027
1028 list_add(&pol->walk.all, &net->xfrm.policy_all);
1029 hlist_add_head(&pol->bydst, chain);
1030 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1031 net->xfrm.policy_count[dir]++;
1032 xfrm_pol_hold(pol);
1033
1034 if (xfrm_bydst_should_resize(net, dir, NULL))
1035 schedule_work(&net->xfrm.policy_hash_work);
1036 }
1037
1038 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1039 int dir)
1040 {
1041 struct net *net = xp_net(pol);
1042
1043 if (hlist_unhashed(&pol->bydst))
1044 return NULL;
1045
1046 hlist_del(&pol->bydst);
1047 hlist_del(&pol->byidx);
1048 list_del(&pol->walk.all);
1049 net->xfrm.policy_count[dir]--;
1050
1051 return pol;
1052 }
1053
1054 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1055 {
1056 write_lock_bh(&xfrm_policy_lock);
1057 pol = __xfrm_policy_unlink(pol, dir);
1058 write_unlock_bh(&xfrm_policy_lock);
1059 if (pol) {
1060 xfrm_policy_kill(pol);
1061 return 0;
1062 }
1063 return -ENOENT;
1064 }
1065 EXPORT_SYMBOL(xfrm_policy_delete);
1066
1067 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1068 {
1069 struct net *net = xp_net(pol);
1070 struct xfrm_policy *old_pol;
1071
1072 #ifdef CONFIG_XFRM_SUB_POLICY
1073 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1074 return -EINVAL;
1075 #endif
1076
1077 write_lock_bh(&xfrm_policy_lock);
1078 old_pol = sk->sk_policy[dir];
1079 sk->sk_policy[dir] = pol;
1080 if (pol) {
1081 pol->curlft.add_time = get_seconds();
1082 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1083 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1084 }
1085 if (old_pol)
1086 /* Unlinking succeeds always. This is the only function
1087 * allowed to delete or replace socket policy.
1088 */
1089 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1090 write_unlock_bh(&xfrm_policy_lock);
1091
1092 if (old_pol) {
1093 xfrm_policy_kill(old_pol);
1094 }
1095 return 0;
1096 }
1097
1098 static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1099 {
1100 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1101
1102 if (newp) {
1103 newp->selector = old->selector;
1104 if (security_xfrm_policy_clone(old->security,
1105 &newp->security)) {
1106 kfree(newp);
1107 return NULL; /* ENOMEM */
1108 }
1109 newp->lft = old->lft;
1110 newp->curlft = old->curlft;
1111 newp->mark = old->mark;
1112 newp->action = old->action;
1113 newp->flags = old->flags;
1114 newp->xfrm_nr = old->xfrm_nr;
1115 newp->index = old->index;
1116 newp->type = old->type;
1117 memcpy(newp->xfrm_vec, old->xfrm_vec,
1118 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1119 write_lock_bh(&xfrm_policy_lock);
1120 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1121 write_unlock_bh(&xfrm_policy_lock);
1122 xfrm_pol_put(newp);
1123 }
1124 return newp;
1125 }
1126
1127 int __xfrm_sk_clone_policy(struct sock *sk)
1128 {
1129 struct xfrm_policy *p0 = sk->sk_policy[0],
1130 *p1 = sk->sk_policy[1];
1131
1132 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1133 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1134 return -ENOMEM;
1135 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1136 return -ENOMEM;
1137 return 0;
1138 }
1139
1140 static int
1141 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1142 unsigned short family)
1143 {
1144 int err;
1145 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1146
1147 if (unlikely(afinfo == NULL))
1148 return -EINVAL;
1149 err = afinfo->get_saddr(net, local, remote);
1150 xfrm_policy_put_afinfo(afinfo);
1151 return err;
1152 }
1153
1154 /* Resolve list of templates for the flow, given policy. */
1155
1156 static int
1157 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
1158 struct xfrm_state **xfrm,
1159 unsigned short family)
1160 {
1161 struct net *net = xp_net(policy);
1162 int nx;
1163 int i, error;
1164 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1165 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1166 xfrm_address_t tmp;
1167
1168 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1169 struct xfrm_state *x;
1170 xfrm_address_t *remote = daddr;
1171 xfrm_address_t *local = saddr;
1172 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1173
1174 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1175 tmpl->mode == XFRM_MODE_BEET) {
1176 remote = &tmpl->id.daddr;
1177 local = &tmpl->saddr;
1178 family = tmpl->encap_family;
1179 if (xfrm_addr_any(local, family)) {
1180 error = xfrm_get_saddr(net, &tmp, remote, family);
1181 if (error)
1182 goto fail;
1183 local = &tmp;
1184 }
1185 }
1186
1187 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1188
1189 if (x && x->km.state == XFRM_STATE_VALID) {
1190 xfrm[nx++] = x;
1191 daddr = remote;
1192 saddr = local;
1193 continue;
1194 }
1195 if (x) {
1196 error = (x->km.state == XFRM_STATE_ERROR ?
1197 -EINVAL : -EAGAIN);
1198 xfrm_state_put(x);
1199 }
1200 else if (error == -ESRCH)
1201 error = -EAGAIN;
1202
1203 if (!tmpl->optional)
1204 goto fail;
1205 }
1206 return nx;
1207
1208 fail:
1209 for (nx--; nx>=0; nx--)
1210 xfrm_state_put(xfrm[nx]);
1211 return error;
1212 }
1213
1214 static int
1215 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1216 struct xfrm_state **xfrm,
1217 unsigned short family)
1218 {
1219 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1220 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1221 int cnx = 0;
1222 int error;
1223 int ret;
1224 int i;
1225
1226 for (i = 0; i < npols; i++) {
1227 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1228 error = -ENOBUFS;
1229 goto fail;
1230 }
1231
1232 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1233 if (ret < 0) {
1234 error = ret;
1235 goto fail;
1236 } else
1237 cnx += ret;
1238 }
1239
1240 /* found states are sorted for outbound processing */
1241 if (npols > 1)
1242 xfrm_state_sort(xfrm, tpp, cnx, family);
1243
1244 return cnx;
1245
1246 fail:
1247 for (cnx--; cnx>=0; cnx--)
1248 xfrm_state_put(tpp[cnx]);
1249 return error;
1250
1251 }
1252
1253 /* Check that the bundle accepts the flow and its components are
1254 * still valid.
1255 */
1256
1257 static inline int xfrm_get_tos(struct flowi *fl, int family)
1258 {
1259 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1260 int tos;
1261
1262 if (!afinfo)
1263 return -EINVAL;
1264
1265 tos = afinfo->get_tos(fl);
1266
1267 xfrm_policy_put_afinfo(afinfo);
1268
1269 return tos;
1270 }
1271
1272 static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1273 {
1274 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1275 struct dst_entry *dst = &xdst->u.dst;
1276
1277 if (xdst->route == NULL) {
1278 /* Dummy bundle - if it has xfrms we were not
1279 * able to build bundle as template resolution failed.
1280 * It means we need to try again resolving. */
1281 if (xdst->num_xfrms > 0)
1282 return NULL;
1283 } else {
1284 /* Real bundle */
1285 if (stale_bundle(dst))
1286 return NULL;
1287 }
1288
1289 dst_hold(dst);
1290 return flo;
1291 }
1292
1293 static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1294 {
1295 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1296 struct dst_entry *dst = &xdst->u.dst;
1297
1298 if (!xdst->route)
1299 return 0;
1300 if (stale_bundle(dst))
1301 return 0;
1302
1303 return 1;
1304 }
1305
1306 static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1307 {
1308 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1309 struct dst_entry *dst = &xdst->u.dst;
1310
1311 dst_free(dst);
1312 }
1313
1314 static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1315 .get = xfrm_bundle_flo_get,
1316 .check = xfrm_bundle_flo_check,
1317 .delete = xfrm_bundle_flo_delete,
1318 };
1319
1320 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1321 {
1322 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1323 struct dst_ops *dst_ops;
1324 struct xfrm_dst *xdst;
1325
1326 if (!afinfo)
1327 return ERR_PTR(-EINVAL);
1328
1329 switch (family) {
1330 case AF_INET:
1331 dst_ops = &net->xfrm.xfrm4_dst_ops;
1332 break;
1333 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1334 case AF_INET6:
1335 dst_ops = &net->xfrm.xfrm6_dst_ops;
1336 break;
1337 #endif
1338 default:
1339 BUG();
1340 }
1341 xdst = dst_alloc(dst_ops) ?: ERR_PTR(-ENOBUFS);
1342 xfrm_policy_put_afinfo(afinfo);
1343
1344 xdst->flo.ops = &xfrm_bundle_fc_ops;
1345
1346 return xdst;
1347 }
1348
1349 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1350 int nfheader_len)
1351 {
1352 struct xfrm_policy_afinfo *afinfo =
1353 xfrm_policy_get_afinfo(dst->ops->family);
1354 int err;
1355
1356 if (!afinfo)
1357 return -EINVAL;
1358
1359 err = afinfo->init_path(path, dst, nfheader_len);
1360
1361 xfrm_policy_put_afinfo(afinfo);
1362
1363 return err;
1364 }
1365
1366 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
1367 struct flowi *fl)
1368 {
1369 struct xfrm_policy_afinfo *afinfo =
1370 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1371 int err;
1372
1373 if (!afinfo)
1374 return -EINVAL;
1375
1376 err = afinfo->fill_dst(xdst, dev, fl);
1377
1378 xfrm_policy_put_afinfo(afinfo);
1379
1380 return err;
1381 }
1382
1383
1384 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1385 * all the metrics... Shortly, bundle a bundle.
1386 */
1387
1388 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1389 struct xfrm_state **xfrm, int nx,
1390 struct flowi *fl,
1391 struct dst_entry *dst)
1392 {
1393 struct net *net = xp_net(policy);
1394 unsigned long now = jiffies;
1395 struct net_device *dev;
1396 struct dst_entry *dst_prev = NULL;
1397 struct dst_entry *dst0 = NULL;
1398 int i = 0;
1399 int err;
1400 int header_len = 0;
1401 int nfheader_len = 0;
1402 int trailer_len = 0;
1403 int tos;
1404 int family = policy->selector.family;
1405 xfrm_address_t saddr, daddr;
1406
1407 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1408
1409 tos = xfrm_get_tos(fl, family);
1410 err = tos;
1411 if (tos < 0)
1412 goto put_states;
1413
1414 dst_hold(dst);
1415
1416 for (; i < nx; i++) {
1417 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1418 struct dst_entry *dst1 = &xdst->u.dst;
1419
1420 err = PTR_ERR(xdst);
1421 if (IS_ERR(xdst)) {
1422 dst_release(dst);
1423 goto put_states;
1424 }
1425
1426 if (!dst_prev)
1427 dst0 = dst1;
1428 else {
1429 dst_prev->child = dst_clone(dst1);
1430 dst1->flags |= DST_NOHASH;
1431 }
1432
1433 xdst->route = dst;
1434 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1435
1436 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1437 family = xfrm[i]->props.family;
1438 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1439 family);
1440 err = PTR_ERR(dst);
1441 if (IS_ERR(dst))
1442 goto put_states;
1443 } else
1444 dst_hold(dst);
1445
1446 dst1->xfrm = xfrm[i];
1447 xdst->xfrm_genid = xfrm[i]->genid;
1448
1449 dst1->obsolete = -1;
1450 dst1->flags |= DST_HOST;
1451 dst1->lastuse = now;
1452
1453 dst1->input = dst_discard;
1454 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1455
1456 dst1->next = dst_prev;
1457 dst_prev = dst1;
1458
1459 header_len += xfrm[i]->props.header_len;
1460 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1461 nfheader_len += xfrm[i]->props.header_len;
1462 trailer_len += xfrm[i]->props.trailer_len;
1463 }
1464
1465 dst_prev->child = dst;
1466 dst0->path = dst;
1467
1468 err = -ENODEV;
1469 dev = dst->dev;
1470 if (!dev)
1471 goto free_dst;
1472
1473 /* Copy neighbour for reachability confirmation */
1474 dst0->neighbour = neigh_clone(dst->neighbour);
1475
1476 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1477 xfrm_init_pmtu(dst_prev);
1478
1479 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1480 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1481
1482 err = xfrm_fill_dst(xdst, dev, fl);
1483 if (err)
1484 goto free_dst;
1485
1486 dst_prev->header_len = header_len;
1487 dst_prev->trailer_len = trailer_len;
1488 header_len -= xdst->u.dst.xfrm->props.header_len;
1489 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1490 }
1491
1492 out:
1493 return dst0;
1494
1495 put_states:
1496 for (; i < nx; i++)
1497 xfrm_state_put(xfrm[i]);
1498 free_dst:
1499 if (dst0)
1500 dst_free(dst0);
1501 dst0 = ERR_PTR(err);
1502 goto out;
1503 }
1504
1505 static int inline
1506 xfrm_dst_alloc_copy(void **target, void *src, int size)
1507 {
1508 if (!*target) {
1509 *target = kmalloc(size, GFP_ATOMIC);
1510 if (!*target)
1511 return -ENOMEM;
1512 }
1513 memcpy(*target, src, size);
1514 return 0;
1515 }
1516
1517 static int inline
1518 xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1519 {
1520 #ifdef CONFIG_XFRM_SUB_POLICY
1521 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1522 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1523 sel, sizeof(*sel));
1524 #else
1525 return 0;
1526 #endif
1527 }
1528
1529 static int inline
1530 xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1531 {
1532 #ifdef CONFIG_XFRM_SUB_POLICY
1533 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1534 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1535 #else
1536 return 0;
1537 #endif
1538 }
1539
1540 static int xfrm_expand_policies(struct flowi *fl, u16 family,
1541 struct xfrm_policy **pols,
1542 int *num_pols, int *num_xfrms)
1543 {
1544 int i;
1545
1546 if (*num_pols == 0 || !pols[0]) {
1547 *num_pols = 0;
1548 *num_xfrms = 0;
1549 return 0;
1550 }
1551 if (IS_ERR(pols[0]))
1552 return PTR_ERR(pols[0]);
1553
1554 *num_xfrms = pols[0]->xfrm_nr;
1555
1556 #ifdef CONFIG_XFRM_SUB_POLICY
1557 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1558 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1559 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1560 XFRM_POLICY_TYPE_MAIN,
1561 fl, family,
1562 XFRM_POLICY_OUT);
1563 if (pols[1]) {
1564 if (IS_ERR(pols[1])) {
1565 xfrm_pols_put(pols, *num_pols);
1566 return PTR_ERR(pols[1]);
1567 }
1568 (*num_pols) ++;
1569 (*num_xfrms) += pols[1]->xfrm_nr;
1570 }
1571 }
1572 #endif
1573 for (i = 0; i < *num_pols; i++) {
1574 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1575 *num_xfrms = -1;
1576 break;
1577 }
1578 }
1579
1580 return 0;
1581
1582 }
1583
1584 static struct xfrm_dst *
1585 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
1586 struct flowi *fl, u16 family,
1587 struct dst_entry *dst_orig)
1588 {
1589 struct net *net = xp_net(pols[0]);
1590 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1591 struct dst_entry *dst;
1592 struct xfrm_dst *xdst;
1593 int err;
1594
1595 /* Try to instantiate a bundle */
1596 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1597 if (err < 0) {
1598 if (err != -EAGAIN)
1599 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1600 return ERR_PTR(err);
1601 }
1602
1603 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1604 if (IS_ERR(dst)) {
1605 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1606 return ERR_CAST(dst);
1607 }
1608
1609 xdst = (struct xfrm_dst *)dst;
1610 xdst->num_xfrms = err;
1611 if (num_pols > 1)
1612 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1613 else
1614 err = xfrm_dst_update_origin(dst, fl);
1615 if (unlikely(err)) {
1616 dst_free(dst);
1617 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1618 return ERR_PTR(err);
1619 }
1620
1621 xdst->num_pols = num_pols;
1622 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1623 xdst->policy_genid = atomic_read(&pols[0]->genid);
1624
1625 return xdst;
1626 }
1627
1628 static struct flow_cache_object *
1629 xfrm_bundle_lookup(struct net *net, struct flowi *fl, u16 family, u8 dir,
1630 struct flow_cache_object *oldflo, void *ctx)
1631 {
1632 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1633 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1634 struct xfrm_dst *xdst, *new_xdst;
1635 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1636
1637 /* Check if the policies from old bundle are usable */
1638 xdst = NULL;
1639 if (oldflo) {
1640 xdst = container_of(oldflo, struct xfrm_dst, flo);
1641 num_pols = xdst->num_pols;
1642 num_xfrms = xdst->num_xfrms;
1643 pol_dead = 0;
1644 for (i = 0; i < num_pols; i++) {
1645 pols[i] = xdst->pols[i];
1646 pol_dead |= pols[i]->walk.dead;
1647 }
1648 if (pol_dead) {
1649 dst_free(&xdst->u.dst);
1650 xdst = NULL;
1651 num_pols = 0;
1652 num_xfrms = 0;
1653 oldflo = NULL;
1654 }
1655 }
1656
1657 /* Resolve policies to use if we couldn't get them from
1658 * previous cache entry */
1659 if (xdst == NULL) {
1660 num_pols = 1;
1661 pols[0] = __xfrm_policy_lookup(net, fl, family, dir);
1662 err = xfrm_expand_policies(fl, family, pols,
1663 &num_pols, &num_xfrms);
1664 if (err < 0)
1665 goto inc_error;
1666 if (num_pols == 0)
1667 return NULL;
1668 if (num_xfrms <= 0)
1669 goto make_dummy_bundle;
1670 }
1671
1672 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1673 if (IS_ERR(new_xdst)) {
1674 err = PTR_ERR(new_xdst);
1675 if (err != -EAGAIN)
1676 goto error;
1677 if (oldflo == NULL)
1678 goto make_dummy_bundle;
1679 dst_hold(&xdst->u.dst);
1680 return oldflo;
1681 }
1682
1683 /* Kill the previous bundle */
1684 if (xdst) {
1685 /* The policies were stolen for newly generated bundle */
1686 xdst->num_pols = 0;
1687 dst_free(&xdst->u.dst);
1688 }
1689
1690 /* Flow cache does not have reference, it dst_free()'s,
1691 * but we do need to return one reference for original caller */
1692 dst_hold(&new_xdst->u.dst);
1693 return &new_xdst->flo;
1694
1695 make_dummy_bundle:
1696 /* We found policies, but there's no bundles to instantiate:
1697 * either because the policy blocks, has no transformations or
1698 * we could not build template (no xfrm_states).*/
1699 xdst = xfrm_alloc_dst(net, family);
1700 if (IS_ERR(xdst)) {
1701 xfrm_pols_put(pols, num_pols);
1702 return ERR_CAST(xdst);
1703 }
1704 xdst->num_pols = num_pols;
1705 xdst->num_xfrms = num_xfrms;
1706 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1707
1708 dst_hold(&xdst->u.dst);
1709 return &xdst->flo;
1710
1711 inc_error:
1712 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1713 error:
1714 if (xdst != NULL)
1715 dst_free(&xdst->u.dst);
1716 else
1717 xfrm_pols_put(pols, num_pols);
1718 return ERR_PTR(err);
1719 }
1720
1721 /* Main function: finds/creates a bundle for given flow.
1722 *
1723 * At the moment we eat a raw IP route. Mostly to speed up lookups
1724 * on interfaces with disabled IPsec.
1725 */
1726 int __xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1727 struct sock *sk, int flags)
1728 {
1729 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1730 struct flow_cache_object *flo;
1731 struct xfrm_dst *xdst;
1732 struct dst_entry *dst, *dst_orig = *dst_p, *route;
1733 u16 family = dst_orig->ops->family;
1734 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1735 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
1736
1737 restart:
1738 dst = NULL;
1739 xdst = NULL;
1740 route = NULL;
1741
1742 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1743 num_pols = 1;
1744 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1745 err = xfrm_expand_policies(fl, family, pols,
1746 &num_pols, &num_xfrms);
1747 if (err < 0)
1748 goto dropdst;
1749
1750 if (num_pols) {
1751 if (num_xfrms <= 0) {
1752 drop_pols = num_pols;
1753 goto no_transform;
1754 }
1755
1756 xdst = xfrm_resolve_and_create_bundle(
1757 pols, num_pols, fl,
1758 family, dst_orig);
1759 if (IS_ERR(xdst)) {
1760 xfrm_pols_put(pols, num_pols);
1761 err = PTR_ERR(xdst);
1762 goto dropdst;
1763 }
1764
1765 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
1766 xdst->u.dst.next = xfrm_policy_sk_bundles;
1767 xfrm_policy_sk_bundles = &xdst->u.dst;
1768 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
1769
1770 route = xdst->route;
1771 }
1772 }
1773
1774 if (xdst == NULL) {
1775 /* To accelerate a bit... */
1776 if ((dst_orig->flags & DST_NOXFRM) ||
1777 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1778 goto nopol;
1779
1780 flo = flow_cache_lookup(net, fl, family, dir,
1781 xfrm_bundle_lookup, dst_orig);
1782 if (flo == NULL)
1783 goto nopol;
1784 if (IS_ERR(flo)) {
1785 err = PTR_ERR(flo);
1786 goto dropdst;
1787 }
1788 xdst = container_of(flo, struct xfrm_dst, flo);
1789
1790 num_pols = xdst->num_pols;
1791 num_xfrms = xdst->num_xfrms;
1792 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
1793 route = xdst->route;
1794 }
1795
1796 dst = &xdst->u.dst;
1797 if (route == NULL && num_xfrms > 0) {
1798 /* The only case when xfrm_bundle_lookup() returns a
1799 * bundle with null route, is when the template could
1800 * not be resolved. It means policies are there, but
1801 * bundle could not be created, since we don't yet
1802 * have the xfrm_state's. We need to wait for KM to
1803 * negotiate new SA's or bail out with error.*/
1804 if (net->xfrm.sysctl_larval_drop) {
1805 /* EREMOTE tells the caller to generate
1806 * a one-shot blackhole route. */
1807 dst_release(dst);
1808 xfrm_pols_put(pols, drop_pols);
1809 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1810 return -EREMOTE;
1811 }
1812 if (flags & XFRM_LOOKUP_WAIT) {
1813 DECLARE_WAITQUEUE(wait, current);
1814
1815 add_wait_queue(&net->xfrm.km_waitq, &wait);
1816 set_current_state(TASK_INTERRUPTIBLE);
1817 schedule();
1818 set_current_state(TASK_RUNNING);
1819 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1820
1821 if (!signal_pending(current)) {
1822 dst_release(dst);
1823 goto restart;
1824 }
1825
1826 err = -ERESTART;
1827 } else
1828 err = -EAGAIN;
1829
1830 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1831 goto error;
1832 }
1833
1834 no_transform:
1835 if (num_pols == 0)
1836 goto nopol;
1837
1838 if ((flags & XFRM_LOOKUP_ICMP) &&
1839 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
1840 err = -ENOENT;
1841 goto error;
1842 }
1843
1844 for (i = 0; i < num_pols; i++)
1845 pols[i]->curlft.use_time = get_seconds();
1846
1847 if (num_xfrms < 0) {
1848 /* Prohibit the flow */
1849 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1850 err = -EPERM;
1851 goto error;
1852 } else if (num_xfrms > 0) {
1853 /* Flow transformed */
1854 *dst_p = dst;
1855 dst_release(dst_orig);
1856 } else {
1857 /* Flow passes untransformed */
1858 dst_release(dst);
1859 }
1860 ok:
1861 xfrm_pols_put(pols, drop_pols);
1862 return 0;
1863
1864 nopol:
1865 if (!(flags & XFRM_LOOKUP_ICMP))
1866 goto ok;
1867 err = -ENOENT;
1868 error:
1869 dst_release(dst);
1870 dropdst:
1871 dst_release(dst_orig);
1872 *dst_p = NULL;
1873 xfrm_pols_put(pols, drop_pols);
1874 return err;
1875 }
1876 EXPORT_SYMBOL(__xfrm_lookup);
1877
1878 int xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1879 struct sock *sk, int flags)
1880 {
1881 int err = __xfrm_lookup(net, dst_p, fl, sk, flags);
1882
1883 if (err == -EREMOTE) {
1884 dst_release(*dst_p);
1885 *dst_p = NULL;
1886 err = -EAGAIN;
1887 }
1888
1889 return err;
1890 }
1891 EXPORT_SYMBOL(xfrm_lookup);
1892
1893 static inline int
1894 xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1895 {
1896 struct xfrm_state *x;
1897
1898 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1899 return 0;
1900 x = skb->sp->xvec[idx];
1901 if (!x->type->reject)
1902 return 0;
1903 return x->type->reject(x, skb, fl);
1904 }
1905
1906 /* When skb is transformed back to its "native" form, we have to
1907 * check policy restrictions. At the moment we make this in maximally
1908 * stupid way. Shame on me. :-) Of course, connected sockets must
1909 * have policy cached at them.
1910 */
1911
1912 static inline int
1913 xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1914 unsigned short family)
1915 {
1916 if (xfrm_state_kern(x))
1917 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1918 return x->id.proto == tmpl->id.proto &&
1919 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1920 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1921 x->props.mode == tmpl->mode &&
1922 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1923 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1924 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1925 xfrm_state_addr_cmp(tmpl, x, family));
1926 }
1927
1928 /*
1929 * 0 or more than 0 is returned when validation is succeeded (either bypass
1930 * because of optional transport mode, or next index of the mathced secpath
1931 * state with the template.
1932 * -1 is returned when no matching template is found.
1933 * Otherwise "-2 - errored_index" is returned.
1934 */
1935 static inline int
1936 xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1937 unsigned short family)
1938 {
1939 int idx = start;
1940
1941 if (tmpl->optional) {
1942 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1943 return start;
1944 } else
1945 start = -1;
1946 for (; idx < sp->len; idx++) {
1947 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1948 return ++idx;
1949 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1950 if (start == -1)
1951 start = -2-idx;
1952 break;
1953 }
1954 }
1955 return start;
1956 }
1957
1958 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1959 unsigned int family, int reverse)
1960 {
1961 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1962 int err;
1963
1964 if (unlikely(afinfo == NULL))
1965 return -EAFNOSUPPORT;
1966
1967 afinfo->decode_session(skb, fl, reverse);
1968 err = security_xfrm_decode_session(skb, &fl->secid);
1969 xfrm_policy_put_afinfo(afinfo);
1970 return err;
1971 }
1972 EXPORT_SYMBOL(__xfrm_decode_session);
1973
1974 static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1975 {
1976 for (; k < sp->len; k++) {
1977 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
1978 *idxp = k;
1979 return 1;
1980 }
1981 }
1982
1983 return 0;
1984 }
1985
1986 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1987 unsigned short family)
1988 {
1989 struct net *net = dev_net(skb->dev);
1990 struct xfrm_policy *pol;
1991 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1992 int npols = 0;
1993 int xfrm_nr;
1994 int pi;
1995 int reverse;
1996 struct flowi fl;
1997 u8 fl_dir;
1998 int xerr_idx = -1;
1999
2000 reverse = dir & ~XFRM_POLICY_MASK;
2001 dir &= XFRM_POLICY_MASK;
2002 fl_dir = policy_to_flow_dir(dir);
2003
2004 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
2005 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
2006 return 0;
2007 }
2008
2009 nf_nat_decode_session(skb, &fl, family);
2010
2011 /* First, check used SA against their selectors. */
2012 if (skb->sp) {
2013 int i;
2014
2015 for (i=skb->sp->len-1; i>=0; i--) {
2016 struct xfrm_state *x = skb->sp->xvec[i];
2017 if (!xfrm_selector_match(&x->sel, &fl, family)) {
2018 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
2019 return 0;
2020 }
2021 }
2022 }
2023
2024 pol = NULL;
2025 if (sk && sk->sk_policy[dir]) {
2026 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
2027 if (IS_ERR(pol)) {
2028 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2029 return 0;
2030 }
2031 }
2032
2033 if (!pol) {
2034 struct flow_cache_object *flo;
2035
2036 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2037 xfrm_policy_lookup, NULL);
2038 if (IS_ERR_OR_NULL(flo))
2039 pol = ERR_CAST(flo);
2040 else
2041 pol = container_of(flo, struct xfrm_policy, flo);
2042 }
2043
2044 if (IS_ERR(pol)) {
2045 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2046 return 0;
2047 }
2048
2049 if (!pol) {
2050 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
2051 xfrm_secpath_reject(xerr_idx, skb, &fl);
2052 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
2053 return 0;
2054 }
2055 return 1;
2056 }
2057
2058 pol->curlft.use_time = get_seconds();
2059
2060 pols[0] = pol;
2061 npols ++;
2062 #ifdef CONFIG_XFRM_SUB_POLICY
2063 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
2064 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
2065 &fl, family,
2066 XFRM_POLICY_IN);
2067 if (pols[1]) {
2068 if (IS_ERR(pols[1])) {
2069 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2070 return 0;
2071 }
2072 pols[1]->curlft.use_time = get_seconds();
2073 npols ++;
2074 }
2075 }
2076 #endif
2077
2078 if (pol->action == XFRM_POLICY_ALLOW) {
2079 struct sec_path *sp;
2080 static struct sec_path dummy;
2081 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
2082 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
2083 struct xfrm_tmpl **tpp = tp;
2084 int ti = 0;
2085 int i, k;
2086
2087 if ((sp = skb->sp) == NULL)
2088 sp = &dummy;
2089
2090 for (pi = 0; pi < npols; pi++) {
2091 if (pols[pi] != pol &&
2092 pols[pi]->action != XFRM_POLICY_ALLOW) {
2093 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2094 goto reject;
2095 }
2096 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2097 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2098 goto reject_error;
2099 }
2100 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2101 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2102 }
2103 xfrm_nr = ti;
2104 if (npols > 1) {
2105 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2106 tpp = stp;
2107 }
2108
2109 /* For each tunnel xfrm, find the first matching tmpl.
2110 * For each tmpl before that, find corresponding xfrm.
2111 * Order is _important_. Later we will implement
2112 * some barriers, but at the moment barriers
2113 * are implied between each two transformations.
2114 */
2115 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2116 k = xfrm_policy_ok(tpp[i], sp, k, family);
2117 if (k < 0) {
2118 if (k < -1)
2119 /* "-2 - errored_index" returned */
2120 xerr_idx = -(2+k);
2121 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2122 goto reject;
2123 }
2124 }
2125
2126 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2127 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2128 goto reject;
2129 }
2130
2131 xfrm_pols_put(pols, npols);
2132 return 1;
2133 }
2134 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2135
2136 reject:
2137 xfrm_secpath_reject(xerr_idx, skb, &fl);
2138 reject_error:
2139 xfrm_pols_put(pols, npols);
2140 return 0;
2141 }
2142 EXPORT_SYMBOL(__xfrm_policy_check);
2143
2144 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2145 {
2146 struct net *net = dev_net(skb->dev);
2147 struct flowi fl;
2148 struct dst_entry *dst;
2149 int res;
2150
2151 if (xfrm_decode_session(skb, &fl, family) < 0) {
2152 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2153 return 0;
2154 }
2155
2156 skb_dst_force(skb);
2157 dst = skb_dst(skb);
2158
2159 res = xfrm_lookup(net, &dst, &fl, NULL, 0) == 0;
2160 skb_dst_set(skb, dst);
2161 return res;
2162 }
2163 EXPORT_SYMBOL(__xfrm_route_forward);
2164
2165 /* Optimize later using cookies and generation ids. */
2166
2167 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2168 {
2169 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2170 * to "-1" to force all XFRM destinations to get validated by
2171 * dst_ops->check on every use. We do this because when a
2172 * normal route referenced by an XFRM dst is obsoleted we do
2173 * not go looking around for all parent referencing XFRM dsts
2174 * so that we can invalidate them. It is just too much work.
2175 * Instead we make the checks here on every use. For example:
2176 *
2177 * XFRM dst A --> IPv4 dst X
2178 *
2179 * X is the "xdst->route" of A (X is also the "dst->path" of A
2180 * in this example). If X is marked obsolete, "A" will not
2181 * notice. That's what we are validating here via the
2182 * stale_bundle() check.
2183 *
2184 * When a policy's bundle is pruned, we dst_free() the XFRM
2185 * dst which causes it's ->obsolete field to be set to a
2186 * positive non-zero integer. If an XFRM dst has been pruned
2187 * like this, we want to force a new route lookup.
2188 */
2189 if (dst->obsolete < 0 && !stale_bundle(dst))
2190 return dst;
2191
2192 return NULL;
2193 }
2194
2195 static int stale_bundle(struct dst_entry *dst)
2196 {
2197 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
2198 }
2199
2200 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2201 {
2202 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2203 dst->dev = dev_net(dev)->loopback_dev;
2204 dev_hold(dst->dev);
2205 dev_put(dev);
2206 }
2207 }
2208 EXPORT_SYMBOL(xfrm_dst_ifdown);
2209
2210 static void xfrm_link_failure(struct sk_buff *skb)
2211 {
2212 /* Impossible. Such dst must be popped before reaches point of failure. */
2213 }
2214
2215 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2216 {
2217 if (dst) {
2218 if (dst->obsolete) {
2219 dst_release(dst);
2220 dst = NULL;
2221 }
2222 }
2223 return dst;
2224 }
2225
2226 static void __xfrm_garbage_collect(struct net *net)
2227 {
2228 struct dst_entry *head, *next;
2229
2230 flow_cache_flush();
2231
2232 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2233 head = xfrm_policy_sk_bundles;
2234 xfrm_policy_sk_bundles = NULL;
2235 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2236
2237 while (head) {
2238 next = head->next;
2239 dst_free(head);
2240 head = next;
2241 }
2242 }
2243
2244 static void xfrm_init_pmtu(struct dst_entry *dst)
2245 {
2246 do {
2247 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2248 u32 pmtu, route_mtu_cached;
2249
2250 pmtu = dst_mtu(dst->child);
2251 xdst->child_mtu_cached = pmtu;
2252
2253 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2254
2255 route_mtu_cached = dst_mtu(xdst->route);
2256 xdst->route_mtu_cached = route_mtu_cached;
2257
2258 if (pmtu > route_mtu_cached)
2259 pmtu = route_mtu_cached;
2260
2261 dst->metrics[RTAX_MTU-1] = pmtu;
2262 } while ((dst = dst->next));
2263 }
2264
2265 /* Check that the bundle accepts the flow and its components are
2266 * still valid.
2267 */
2268
2269 int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2270 struct flowi *fl, int family, int strict)
2271 {
2272 struct dst_entry *dst = &first->u.dst;
2273 struct xfrm_dst *last;
2274 u32 mtu;
2275
2276 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2277 (dst->dev && !netif_running(dst->dev)))
2278 return 0;
2279 #ifdef CONFIG_XFRM_SUB_POLICY
2280 if (fl) {
2281 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2282 return 0;
2283 if (first->partner &&
2284 !xfrm_selector_match(first->partner, fl, family))
2285 return 0;
2286 }
2287 #endif
2288
2289 last = NULL;
2290
2291 do {
2292 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2293
2294 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2295 return 0;
2296 if (fl && pol &&
2297 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
2298 return 0;
2299 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2300 return 0;
2301 if (xdst->xfrm_genid != dst->xfrm->genid)
2302 return 0;
2303 if (xdst->num_pols > 0 &&
2304 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
2305 return 0;
2306
2307 if (strict && fl &&
2308 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2309 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2310 return 0;
2311
2312 mtu = dst_mtu(dst->child);
2313 if (xdst->child_mtu_cached != mtu) {
2314 last = xdst;
2315 xdst->child_mtu_cached = mtu;
2316 }
2317
2318 if (!dst_check(xdst->route, xdst->route_cookie))
2319 return 0;
2320 mtu = dst_mtu(xdst->route);
2321 if (xdst->route_mtu_cached != mtu) {
2322 last = xdst;
2323 xdst->route_mtu_cached = mtu;
2324 }
2325
2326 dst = dst->child;
2327 } while (dst->xfrm);
2328
2329 if (likely(!last))
2330 return 1;
2331
2332 mtu = last->child_mtu_cached;
2333 for (;;) {
2334 dst = &last->u.dst;
2335
2336 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2337 if (mtu > last->route_mtu_cached)
2338 mtu = last->route_mtu_cached;
2339 dst->metrics[RTAX_MTU-1] = mtu;
2340
2341 if (last == first)
2342 break;
2343
2344 last = (struct xfrm_dst *)last->u.dst.next;
2345 last->child_mtu_cached = mtu;
2346 }
2347
2348 return 1;
2349 }
2350
2351 EXPORT_SYMBOL(xfrm_bundle_ok);
2352
2353 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2354 {
2355 struct net *net;
2356 int err = 0;
2357 if (unlikely(afinfo == NULL))
2358 return -EINVAL;
2359 if (unlikely(afinfo->family >= NPROTO))
2360 return -EAFNOSUPPORT;
2361 write_lock_bh(&xfrm_policy_afinfo_lock);
2362 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2363 err = -ENOBUFS;
2364 else {
2365 struct dst_ops *dst_ops = afinfo->dst_ops;
2366 if (likely(dst_ops->kmem_cachep == NULL))
2367 dst_ops->kmem_cachep = xfrm_dst_cache;
2368 if (likely(dst_ops->check == NULL))
2369 dst_ops->check = xfrm_dst_check;
2370 if (likely(dst_ops->negative_advice == NULL))
2371 dst_ops->negative_advice = xfrm_negative_advice;
2372 if (likely(dst_ops->link_failure == NULL))
2373 dst_ops->link_failure = xfrm_link_failure;
2374 if (likely(afinfo->garbage_collect == NULL))
2375 afinfo->garbage_collect = __xfrm_garbage_collect;
2376 xfrm_policy_afinfo[afinfo->family] = afinfo;
2377 }
2378 write_unlock_bh(&xfrm_policy_afinfo_lock);
2379
2380 rtnl_lock();
2381 for_each_net(net) {
2382 struct dst_ops *xfrm_dst_ops;
2383
2384 switch (afinfo->family) {
2385 case AF_INET:
2386 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2387 break;
2388 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2389 case AF_INET6:
2390 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2391 break;
2392 #endif
2393 default:
2394 BUG();
2395 }
2396 *xfrm_dst_ops = *afinfo->dst_ops;
2397 }
2398 rtnl_unlock();
2399
2400 return err;
2401 }
2402 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2403
2404 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2405 {
2406 int err = 0;
2407 if (unlikely(afinfo == NULL))
2408 return -EINVAL;
2409 if (unlikely(afinfo->family >= NPROTO))
2410 return -EAFNOSUPPORT;
2411 write_lock_bh(&xfrm_policy_afinfo_lock);
2412 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2413 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2414 err = -EINVAL;
2415 else {
2416 struct dst_ops *dst_ops = afinfo->dst_ops;
2417 xfrm_policy_afinfo[afinfo->family] = NULL;
2418 dst_ops->kmem_cachep = NULL;
2419 dst_ops->check = NULL;
2420 dst_ops->negative_advice = NULL;
2421 dst_ops->link_failure = NULL;
2422 afinfo->garbage_collect = NULL;
2423 }
2424 }
2425 write_unlock_bh(&xfrm_policy_afinfo_lock);
2426 return err;
2427 }
2428 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2429
2430 static void __net_init xfrm_dst_ops_init(struct net *net)
2431 {
2432 struct xfrm_policy_afinfo *afinfo;
2433
2434 read_lock_bh(&xfrm_policy_afinfo_lock);
2435 afinfo = xfrm_policy_afinfo[AF_INET];
2436 if (afinfo)
2437 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2438 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2439 afinfo = xfrm_policy_afinfo[AF_INET6];
2440 if (afinfo)
2441 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2442 #endif
2443 read_unlock_bh(&xfrm_policy_afinfo_lock);
2444 }
2445
2446 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2447 {
2448 struct xfrm_policy_afinfo *afinfo;
2449 if (unlikely(family >= NPROTO))
2450 return NULL;
2451 read_lock(&xfrm_policy_afinfo_lock);
2452 afinfo = xfrm_policy_afinfo[family];
2453 if (unlikely(!afinfo))
2454 read_unlock(&xfrm_policy_afinfo_lock);
2455 return afinfo;
2456 }
2457
2458 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2459 {
2460 read_unlock(&xfrm_policy_afinfo_lock);
2461 }
2462
2463 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2464 {
2465 struct net_device *dev = ptr;
2466
2467 switch (event) {
2468 case NETDEV_DOWN:
2469 __xfrm_garbage_collect(dev_net(dev));
2470 }
2471 return NOTIFY_DONE;
2472 }
2473
2474 static struct notifier_block xfrm_dev_notifier = {
2475 .notifier_call = xfrm_dev_event,
2476 };
2477
2478 #ifdef CONFIG_XFRM_STATISTICS
2479 static int __net_init xfrm_statistics_init(struct net *net)
2480 {
2481 int rv;
2482
2483 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2484 sizeof(struct linux_xfrm_mib)) < 0)
2485 return -ENOMEM;
2486 rv = xfrm_proc_init(net);
2487 if (rv < 0)
2488 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2489 return rv;
2490 }
2491
2492 static void xfrm_statistics_fini(struct net *net)
2493 {
2494 xfrm_proc_fini(net);
2495 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2496 }
2497 #else
2498 static int __net_init xfrm_statistics_init(struct net *net)
2499 {
2500 return 0;
2501 }
2502
2503 static void xfrm_statistics_fini(struct net *net)
2504 {
2505 }
2506 #endif
2507
2508 static int __net_init xfrm_policy_init(struct net *net)
2509 {
2510 unsigned int hmask, sz;
2511 int dir;
2512
2513 if (net_eq(net, &init_net))
2514 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2515 sizeof(struct xfrm_dst),
2516 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2517 NULL);
2518
2519 hmask = 8 - 1;
2520 sz = (hmask+1) * sizeof(struct hlist_head);
2521
2522 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2523 if (!net->xfrm.policy_byidx)
2524 goto out_byidx;
2525 net->xfrm.policy_idx_hmask = hmask;
2526
2527 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2528 struct xfrm_policy_hash *htab;
2529
2530 net->xfrm.policy_count[dir] = 0;
2531 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2532
2533 htab = &net->xfrm.policy_bydst[dir];
2534 htab->table = xfrm_hash_alloc(sz);
2535 if (!htab->table)
2536 goto out_bydst;
2537 htab->hmask = hmask;
2538 }
2539
2540 INIT_LIST_HEAD(&net->xfrm.policy_all);
2541 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2542 if (net_eq(net, &init_net))
2543 register_netdevice_notifier(&xfrm_dev_notifier);
2544 return 0;
2545
2546 out_bydst:
2547 for (dir--; dir >= 0; dir--) {
2548 struct xfrm_policy_hash *htab;
2549
2550 htab = &net->xfrm.policy_bydst[dir];
2551 xfrm_hash_free(htab->table, sz);
2552 }
2553 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2554 out_byidx:
2555 return -ENOMEM;
2556 }
2557
2558 static void xfrm_policy_fini(struct net *net)
2559 {
2560 struct xfrm_audit audit_info;
2561 unsigned int sz;
2562 int dir;
2563
2564 flush_work(&net->xfrm.policy_hash_work);
2565 #ifdef CONFIG_XFRM_SUB_POLICY
2566 audit_info.loginuid = -1;
2567 audit_info.sessionid = -1;
2568 audit_info.secid = 0;
2569 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2570 #endif
2571 audit_info.loginuid = -1;
2572 audit_info.sessionid = -1;
2573 audit_info.secid = 0;
2574 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2575
2576 WARN_ON(!list_empty(&net->xfrm.policy_all));
2577
2578 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2579 struct xfrm_policy_hash *htab;
2580
2581 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2582
2583 htab = &net->xfrm.policy_bydst[dir];
2584 sz = (htab->hmask + 1);
2585 WARN_ON(!hlist_empty(htab->table));
2586 xfrm_hash_free(htab->table, sz);
2587 }
2588
2589 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2590 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2591 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2592 }
2593
2594 static int __net_init xfrm_net_init(struct net *net)
2595 {
2596 int rv;
2597
2598 rv = xfrm_statistics_init(net);
2599 if (rv < 0)
2600 goto out_statistics;
2601 rv = xfrm_state_init(net);
2602 if (rv < 0)
2603 goto out_state;
2604 rv = xfrm_policy_init(net);
2605 if (rv < 0)
2606 goto out_policy;
2607 xfrm_dst_ops_init(net);
2608 rv = xfrm_sysctl_init(net);
2609 if (rv < 0)
2610 goto out_sysctl;
2611 return 0;
2612
2613 out_sysctl:
2614 xfrm_policy_fini(net);
2615 out_policy:
2616 xfrm_state_fini(net);
2617 out_state:
2618 xfrm_statistics_fini(net);
2619 out_statistics:
2620 return rv;
2621 }
2622
2623 static void __net_exit xfrm_net_exit(struct net *net)
2624 {
2625 xfrm_sysctl_fini(net);
2626 xfrm_policy_fini(net);
2627 xfrm_state_fini(net);
2628 xfrm_statistics_fini(net);
2629 }
2630
2631 static struct pernet_operations __net_initdata xfrm_net_ops = {
2632 .init = xfrm_net_init,
2633 .exit = xfrm_net_exit,
2634 };
2635
2636 void __init xfrm_init(void)
2637 {
2638 register_pernet_subsys(&xfrm_net_ops);
2639 xfrm_input_init();
2640 }
2641
2642 #ifdef CONFIG_AUDITSYSCALL
2643 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2644 struct audit_buffer *audit_buf)
2645 {
2646 struct xfrm_sec_ctx *ctx = xp->security;
2647 struct xfrm_selector *sel = &xp->selector;
2648
2649 if (ctx)
2650 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2651 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2652
2653 switch(sel->family) {
2654 case AF_INET:
2655 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2656 if (sel->prefixlen_s != 32)
2657 audit_log_format(audit_buf, " src_prefixlen=%d",
2658 sel->prefixlen_s);
2659 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2660 if (sel->prefixlen_d != 32)
2661 audit_log_format(audit_buf, " dst_prefixlen=%d",
2662 sel->prefixlen_d);
2663 break;
2664 case AF_INET6:
2665 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2666 if (sel->prefixlen_s != 128)
2667 audit_log_format(audit_buf, " src_prefixlen=%d",
2668 sel->prefixlen_s);
2669 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2670 if (sel->prefixlen_d != 128)
2671 audit_log_format(audit_buf, " dst_prefixlen=%d",
2672 sel->prefixlen_d);
2673 break;
2674 }
2675 }
2676
2677 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2678 uid_t auid, u32 sessionid, u32 secid)
2679 {
2680 struct audit_buffer *audit_buf;
2681
2682 audit_buf = xfrm_audit_start("SPD-add");
2683 if (audit_buf == NULL)
2684 return;
2685 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2686 audit_log_format(audit_buf, " res=%u", result);
2687 xfrm_audit_common_policyinfo(xp, audit_buf);
2688 audit_log_end(audit_buf);
2689 }
2690 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2691
2692 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2693 uid_t auid, u32 sessionid, u32 secid)
2694 {
2695 struct audit_buffer *audit_buf;
2696
2697 audit_buf = xfrm_audit_start("SPD-delete");
2698 if (audit_buf == NULL)
2699 return;
2700 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2701 audit_log_format(audit_buf, " res=%u", result);
2702 xfrm_audit_common_policyinfo(xp, audit_buf);
2703 audit_log_end(audit_buf);
2704 }
2705 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2706 #endif
2707
2708 #ifdef CONFIG_XFRM_MIGRATE
2709 static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2710 struct xfrm_selector *sel_tgt)
2711 {
2712 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2713 if (sel_tgt->family == sel_cmp->family &&
2714 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2715 sel_cmp->family) == 0 &&
2716 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2717 sel_cmp->family) == 0 &&
2718 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2719 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2720 return 1;
2721 }
2722 } else {
2723 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2724 return 1;
2725 }
2726 }
2727 return 0;
2728 }
2729
2730 static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2731 u8 dir, u8 type)
2732 {
2733 struct xfrm_policy *pol, *ret = NULL;
2734 struct hlist_node *entry;
2735 struct hlist_head *chain;
2736 u32 priority = ~0U;
2737
2738 read_lock_bh(&xfrm_policy_lock);
2739 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2740 hlist_for_each_entry(pol, entry, chain, bydst) {
2741 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2742 pol->type == type) {
2743 ret = pol;
2744 priority = ret->priority;
2745 break;
2746 }
2747 }
2748 chain = &init_net.xfrm.policy_inexact[dir];
2749 hlist_for_each_entry(pol, entry, chain, bydst) {
2750 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2751 pol->type == type &&
2752 pol->priority < priority) {
2753 ret = pol;
2754 break;
2755 }
2756 }
2757
2758 if (ret)
2759 xfrm_pol_hold(ret);
2760
2761 read_unlock_bh(&xfrm_policy_lock);
2762
2763 return ret;
2764 }
2765
2766 static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2767 {
2768 int match = 0;
2769
2770 if (t->mode == m->mode && t->id.proto == m->proto &&
2771 (m->reqid == 0 || t->reqid == m->reqid)) {
2772 switch (t->mode) {
2773 case XFRM_MODE_TUNNEL:
2774 case XFRM_MODE_BEET:
2775 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2776 m->old_family) == 0 &&
2777 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2778 m->old_family) == 0) {
2779 match = 1;
2780 }
2781 break;
2782 case XFRM_MODE_TRANSPORT:
2783 /* in case of transport mode, template does not store
2784 any IP addresses, hence we just compare mode and
2785 protocol */
2786 match = 1;
2787 break;
2788 default:
2789 break;
2790 }
2791 }
2792 return match;
2793 }
2794
2795 /* update endpoint address(es) of template(s) */
2796 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2797 struct xfrm_migrate *m, int num_migrate)
2798 {
2799 struct xfrm_migrate *mp;
2800 int i, j, n = 0;
2801
2802 write_lock_bh(&pol->lock);
2803 if (unlikely(pol->walk.dead)) {
2804 /* target policy has been deleted */
2805 write_unlock_bh(&pol->lock);
2806 return -ENOENT;
2807 }
2808
2809 for (i = 0; i < pol->xfrm_nr; i++) {
2810 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2811 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2812 continue;
2813 n++;
2814 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2815 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2816 continue;
2817 /* update endpoints */
2818 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2819 sizeof(pol->xfrm_vec[i].id.daddr));
2820 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2821 sizeof(pol->xfrm_vec[i].saddr));
2822 pol->xfrm_vec[i].encap_family = mp->new_family;
2823 /* flush bundles */
2824 atomic_inc(&pol->genid);
2825 }
2826 }
2827
2828 write_unlock_bh(&pol->lock);
2829
2830 if (!n)
2831 return -ENODATA;
2832
2833 return 0;
2834 }
2835
2836 static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2837 {
2838 int i, j;
2839
2840 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2841 return -EINVAL;
2842
2843 for (i = 0; i < num_migrate; i++) {
2844 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2845 m[i].old_family) == 0) &&
2846 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2847 m[i].old_family) == 0))
2848 return -EINVAL;
2849 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2850 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2851 return -EINVAL;
2852
2853 /* check if there is any duplicated entry */
2854 for (j = i + 1; j < num_migrate; j++) {
2855 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2856 sizeof(m[i].old_daddr)) &&
2857 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2858 sizeof(m[i].old_saddr)) &&
2859 m[i].proto == m[j].proto &&
2860 m[i].mode == m[j].mode &&
2861 m[i].reqid == m[j].reqid &&
2862 m[i].old_family == m[j].old_family)
2863 return -EINVAL;
2864 }
2865 }
2866
2867 return 0;
2868 }
2869
2870 int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
2871 struct xfrm_migrate *m, int num_migrate,
2872 struct xfrm_kmaddress *k)
2873 {
2874 int i, err, nx_cur = 0, nx_new = 0;
2875 struct xfrm_policy *pol = NULL;
2876 struct xfrm_state *x, *xc;
2877 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2878 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2879 struct xfrm_migrate *mp;
2880
2881 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2882 goto out;
2883
2884 /* Stage 1 - find policy */
2885 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2886 err = -ENOENT;
2887 goto out;
2888 }
2889
2890 /* Stage 2 - find and update state(s) */
2891 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2892 if ((x = xfrm_migrate_state_find(mp))) {
2893 x_cur[nx_cur] = x;
2894 nx_cur++;
2895 if ((xc = xfrm_state_migrate(x, mp))) {
2896 x_new[nx_new] = xc;
2897 nx_new++;
2898 } else {
2899 err = -ENODATA;
2900 goto restore_state;
2901 }
2902 }
2903 }
2904
2905 /* Stage 3 - update policy */
2906 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2907 goto restore_state;
2908
2909 /* Stage 4 - delete old state(s) */
2910 if (nx_cur) {
2911 xfrm_states_put(x_cur, nx_cur);
2912 xfrm_states_delete(x_cur, nx_cur);
2913 }
2914
2915 /* Stage 5 - announce */
2916 km_migrate(sel, dir, type, m, num_migrate, k);
2917
2918 xfrm_pol_put(pol);
2919
2920 return 0;
2921 out:
2922 return err;
2923
2924 restore_state:
2925 if (pol)
2926 xfrm_pol_put(pol);
2927 if (nx_cur)
2928 xfrm_states_put(x_cur, nx_cur);
2929 if (nx_new)
2930 xfrm_states_delete(x_new, nx_new);
2931
2932 return err;
2933 }
2934 EXPORT_SYMBOL(xfrm_migrate);
2935 #endif
linux-next