xfrm: Mark flowi arg to xfrm_dst_{alloc_copy,update_origin}() const.
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / xfrm / xfrm_policy.c
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 static int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *xdst,
54 const struct flowi *fl, int family);
55
56
57 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
58 int dir);
59
60 static inline int
61 __xfrm4_selector_match(struct xfrm_selector *sel, const struct flowi *fl)
62 {
63 return addr_match(&fl->fl4_dst, &sel->daddr, sel->prefixlen_d) &&
64 addr_match(&fl->fl4_src, &sel->saddr, sel->prefixlen_s) &&
65 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
66 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
67 (fl->proto == sel->proto || !sel->proto) &&
68 (fl->oif == sel->ifindex || !sel->ifindex);
69 }
70
71 static inline int
72 __xfrm6_selector_match(struct xfrm_selector *sel, const struct flowi *fl)
73 {
74 return addr_match(&fl->fl6_dst, &sel->daddr, sel->prefixlen_d) &&
75 addr_match(&fl->fl6_src, &sel->saddr, sel->prefixlen_s) &&
76 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
77 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
78 (fl->proto == sel->proto || !sel->proto) &&
79 (fl->oif == sel->ifindex || !sel->ifindex);
80 }
81
82 int xfrm_selector_match(struct xfrm_selector *sel, const struct flowi *fl,
83 unsigned short family)
84 {
85 switch (family) {
86 case AF_INET:
87 return __xfrm4_selector_match(sel, fl);
88 case AF_INET6:
89 return __xfrm6_selector_match(sel, fl);
90 }
91 return 0;
92 }
93
94 static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
95 xfrm_address_t *saddr,
96 xfrm_address_t *daddr,
97 int family)
98 {
99 struct xfrm_policy_afinfo *afinfo;
100 struct dst_entry *dst;
101
102 afinfo = xfrm_policy_get_afinfo(family);
103 if (unlikely(afinfo == NULL))
104 return ERR_PTR(-EAFNOSUPPORT);
105
106 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
107
108 xfrm_policy_put_afinfo(afinfo);
109
110 return dst;
111 }
112
113 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
114 xfrm_address_t *prev_saddr,
115 xfrm_address_t *prev_daddr,
116 int family)
117 {
118 struct net *net = xs_net(x);
119 xfrm_address_t *saddr = &x->props.saddr;
120 xfrm_address_t *daddr = &x->id.daddr;
121 struct dst_entry *dst;
122
123 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
124 saddr = x->coaddr;
125 daddr = prev_daddr;
126 }
127 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
128 saddr = prev_saddr;
129 daddr = x->coaddr;
130 }
131
132 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
133
134 if (!IS_ERR(dst)) {
135 if (prev_saddr != saddr)
136 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
137 if (prev_daddr != daddr)
138 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
139 }
140
141 return dst;
142 }
143
144 static inline unsigned long make_jiffies(long secs)
145 {
146 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
147 return MAX_SCHEDULE_TIMEOUT-1;
148 else
149 return secs*HZ;
150 }
151
152 static void xfrm_policy_timer(unsigned long data)
153 {
154 struct xfrm_policy *xp = (struct xfrm_policy*)data;
155 unsigned long now = get_seconds();
156 long next = LONG_MAX;
157 int warn = 0;
158 int dir;
159
160 read_lock(&xp->lock);
161
162 if (unlikely(xp->walk.dead))
163 goto out;
164
165 dir = xfrm_policy_id2dir(xp->index);
166
167 if (xp->lft.hard_add_expires_seconds) {
168 long tmo = xp->lft.hard_add_expires_seconds +
169 xp->curlft.add_time - now;
170 if (tmo <= 0)
171 goto expired;
172 if (tmo < next)
173 next = tmo;
174 }
175 if (xp->lft.hard_use_expires_seconds) {
176 long tmo = xp->lft.hard_use_expires_seconds +
177 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
178 if (tmo <= 0)
179 goto expired;
180 if (tmo < next)
181 next = tmo;
182 }
183 if (xp->lft.soft_add_expires_seconds) {
184 long tmo = xp->lft.soft_add_expires_seconds +
185 xp->curlft.add_time - now;
186 if (tmo <= 0) {
187 warn = 1;
188 tmo = XFRM_KM_TIMEOUT;
189 }
190 if (tmo < next)
191 next = tmo;
192 }
193 if (xp->lft.soft_use_expires_seconds) {
194 long tmo = xp->lft.soft_use_expires_seconds +
195 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
196 if (tmo <= 0) {
197 warn = 1;
198 tmo = XFRM_KM_TIMEOUT;
199 }
200 if (tmo < next)
201 next = tmo;
202 }
203
204 if (warn)
205 km_policy_expired(xp, dir, 0, 0);
206 if (next != LONG_MAX &&
207 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
208 xfrm_pol_hold(xp);
209
210 out:
211 read_unlock(&xp->lock);
212 xfrm_pol_put(xp);
213 return;
214
215 expired:
216 read_unlock(&xp->lock);
217 if (!xfrm_policy_delete(xp, dir))
218 km_policy_expired(xp, dir, 1, 0);
219 xfrm_pol_put(xp);
220 }
221
222 static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
223 {
224 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
225
226 if (unlikely(pol->walk.dead))
227 flo = NULL;
228 else
229 xfrm_pol_hold(pol);
230
231 return flo;
232 }
233
234 static int xfrm_policy_flo_check(struct flow_cache_object *flo)
235 {
236 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
237
238 return !pol->walk.dead;
239 }
240
241 static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
242 {
243 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
244 }
245
246 static const struct flow_cache_ops xfrm_policy_fc_ops = {
247 .get = xfrm_policy_flo_get,
248 .check = xfrm_policy_flo_check,
249 .delete = xfrm_policy_flo_delete,
250 };
251
252 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
253 * SPD calls.
254 */
255
256 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
257 {
258 struct xfrm_policy *policy;
259
260 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
261
262 if (policy) {
263 write_pnet(&policy->xp_net, net);
264 INIT_LIST_HEAD(&policy->walk.all);
265 INIT_HLIST_NODE(&policy->bydst);
266 INIT_HLIST_NODE(&policy->byidx);
267 rwlock_init(&policy->lock);
268 atomic_set(&policy->refcnt, 1);
269 setup_timer(&policy->timer, xfrm_policy_timer,
270 (unsigned long)policy);
271 policy->flo.ops = &xfrm_policy_fc_ops;
272 }
273 return policy;
274 }
275 EXPORT_SYMBOL(xfrm_policy_alloc);
276
277 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
278
279 void xfrm_policy_destroy(struct xfrm_policy *policy)
280 {
281 BUG_ON(!policy->walk.dead);
282
283 if (del_timer(&policy->timer))
284 BUG();
285
286 security_xfrm_policy_free(policy->security);
287 kfree(policy);
288 }
289 EXPORT_SYMBOL(xfrm_policy_destroy);
290
291 /* Rule must be locked. Release descentant resources, announce
292 * entry dead. The rule must be unlinked from lists to the moment.
293 */
294
295 static void xfrm_policy_kill(struct xfrm_policy *policy)
296 {
297 policy->walk.dead = 1;
298
299 atomic_inc(&policy->genid);
300
301 if (del_timer(&policy->timer))
302 xfrm_pol_put(policy);
303
304 xfrm_pol_put(policy);
305 }
306
307 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
308
309 static inline unsigned int idx_hash(struct net *net, u32 index)
310 {
311 return __idx_hash(index, net->xfrm.policy_idx_hmask);
312 }
313
314 static struct hlist_head *policy_hash_bysel(struct net *net, struct xfrm_selector *sel, unsigned short family, int dir)
315 {
316 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
317 unsigned int hash = __sel_hash(sel, family, hmask);
318
319 return (hash == hmask + 1 ?
320 &net->xfrm.policy_inexact[dir] :
321 net->xfrm.policy_bydst[dir].table + hash);
322 }
323
324 static struct hlist_head *policy_hash_direct(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
325 {
326 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
327 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
328
329 return net->xfrm.policy_bydst[dir].table + hash;
330 }
331
332 static void xfrm_dst_hash_transfer(struct hlist_head *list,
333 struct hlist_head *ndsttable,
334 unsigned int nhashmask)
335 {
336 struct hlist_node *entry, *tmp, *entry0 = NULL;
337 struct xfrm_policy *pol;
338 unsigned int h0 = 0;
339
340 redo:
341 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
342 unsigned int h;
343
344 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
345 pol->family, nhashmask);
346 if (!entry0) {
347 hlist_del(entry);
348 hlist_add_head(&pol->bydst, ndsttable+h);
349 h0 = h;
350 } else {
351 if (h != h0)
352 continue;
353 hlist_del(entry);
354 hlist_add_after(entry0, &pol->bydst);
355 }
356 entry0 = entry;
357 }
358 if (!hlist_empty(list)) {
359 entry0 = NULL;
360 goto redo;
361 }
362 }
363
364 static void xfrm_idx_hash_transfer(struct hlist_head *list,
365 struct hlist_head *nidxtable,
366 unsigned int nhashmask)
367 {
368 struct hlist_node *entry, *tmp;
369 struct xfrm_policy *pol;
370
371 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
372 unsigned int h;
373
374 h = __idx_hash(pol->index, nhashmask);
375 hlist_add_head(&pol->byidx, nidxtable+h);
376 }
377 }
378
379 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
380 {
381 return ((old_hmask + 1) << 1) - 1;
382 }
383
384 static void xfrm_bydst_resize(struct net *net, int dir)
385 {
386 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
387 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
388 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
389 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
390 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
391 int i;
392
393 if (!ndst)
394 return;
395
396 write_lock_bh(&xfrm_policy_lock);
397
398 for (i = hmask; i >= 0; i--)
399 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
400
401 net->xfrm.policy_bydst[dir].table = ndst;
402 net->xfrm.policy_bydst[dir].hmask = nhashmask;
403
404 write_unlock_bh(&xfrm_policy_lock);
405
406 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
407 }
408
409 static void xfrm_byidx_resize(struct net *net, int total)
410 {
411 unsigned int hmask = net->xfrm.policy_idx_hmask;
412 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
413 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
414 struct hlist_head *oidx = net->xfrm.policy_byidx;
415 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
416 int i;
417
418 if (!nidx)
419 return;
420
421 write_lock_bh(&xfrm_policy_lock);
422
423 for (i = hmask; i >= 0; i--)
424 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
425
426 net->xfrm.policy_byidx = nidx;
427 net->xfrm.policy_idx_hmask = nhashmask;
428
429 write_unlock_bh(&xfrm_policy_lock);
430
431 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
432 }
433
434 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
435 {
436 unsigned int cnt = net->xfrm.policy_count[dir];
437 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
438
439 if (total)
440 *total += cnt;
441
442 if ((hmask + 1) < xfrm_policy_hashmax &&
443 cnt > hmask)
444 return 1;
445
446 return 0;
447 }
448
449 static inline int xfrm_byidx_should_resize(struct net *net, int total)
450 {
451 unsigned int hmask = net->xfrm.policy_idx_hmask;
452
453 if ((hmask + 1) < xfrm_policy_hashmax &&
454 total > hmask)
455 return 1;
456
457 return 0;
458 }
459
460 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
461 {
462 read_lock_bh(&xfrm_policy_lock);
463 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
464 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
465 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
466 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
467 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
468 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
469 si->spdhcnt = net->xfrm.policy_idx_hmask;
470 si->spdhmcnt = xfrm_policy_hashmax;
471 read_unlock_bh(&xfrm_policy_lock);
472 }
473 EXPORT_SYMBOL(xfrm_spd_getinfo);
474
475 static DEFINE_MUTEX(hash_resize_mutex);
476 static void xfrm_hash_resize(struct work_struct *work)
477 {
478 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
479 int dir, total;
480
481 mutex_lock(&hash_resize_mutex);
482
483 total = 0;
484 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
485 if (xfrm_bydst_should_resize(net, dir, &total))
486 xfrm_bydst_resize(net, dir);
487 }
488 if (xfrm_byidx_should_resize(net, total))
489 xfrm_byidx_resize(net, total);
490
491 mutex_unlock(&hash_resize_mutex);
492 }
493
494 /* Generate new index... KAME seems to generate them ordered by cost
495 * of an absolute inpredictability of ordering of rules. This will not pass. */
496 static u32 xfrm_gen_index(struct net *net, int dir)
497 {
498 static u32 idx_generator;
499
500 for (;;) {
501 struct hlist_node *entry;
502 struct hlist_head *list;
503 struct xfrm_policy *p;
504 u32 idx;
505 int found;
506
507 idx = (idx_generator | dir);
508 idx_generator += 8;
509 if (idx == 0)
510 idx = 8;
511 list = net->xfrm.policy_byidx + idx_hash(net, idx);
512 found = 0;
513 hlist_for_each_entry(p, entry, list, byidx) {
514 if (p->index == idx) {
515 found = 1;
516 break;
517 }
518 }
519 if (!found)
520 return idx;
521 }
522 }
523
524 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
525 {
526 u32 *p1 = (u32 *) s1;
527 u32 *p2 = (u32 *) s2;
528 int len = sizeof(struct xfrm_selector) / sizeof(u32);
529 int i;
530
531 for (i = 0; i < len; i++) {
532 if (p1[i] != p2[i])
533 return 1;
534 }
535
536 return 0;
537 }
538
539 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
540 {
541 struct net *net = xp_net(policy);
542 struct xfrm_policy *pol;
543 struct xfrm_policy *delpol;
544 struct hlist_head *chain;
545 struct hlist_node *entry, *newpos;
546 u32 mark = policy->mark.v & policy->mark.m;
547
548 write_lock_bh(&xfrm_policy_lock);
549 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
550 delpol = NULL;
551 newpos = NULL;
552 hlist_for_each_entry(pol, entry, chain, bydst) {
553 if (pol->type == policy->type &&
554 !selector_cmp(&pol->selector, &policy->selector) &&
555 (mark & pol->mark.m) == pol->mark.v &&
556 xfrm_sec_ctx_match(pol->security, policy->security) &&
557 !WARN_ON(delpol)) {
558 if (excl) {
559 write_unlock_bh(&xfrm_policy_lock);
560 return -EEXIST;
561 }
562 delpol = pol;
563 if (policy->priority > pol->priority)
564 continue;
565 } else if (policy->priority >= pol->priority) {
566 newpos = &pol->bydst;
567 continue;
568 }
569 if (delpol)
570 break;
571 }
572 if (newpos)
573 hlist_add_after(newpos, &policy->bydst);
574 else
575 hlist_add_head(&policy->bydst, chain);
576 xfrm_pol_hold(policy);
577 net->xfrm.policy_count[dir]++;
578 atomic_inc(&flow_cache_genid);
579 if (delpol)
580 __xfrm_policy_unlink(delpol, dir);
581 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
582 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
583 policy->curlft.add_time = get_seconds();
584 policy->curlft.use_time = 0;
585 if (!mod_timer(&policy->timer, jiffies + HZ))
586 xfrm_pol_hold(policy);
587 list_add(&policy->walk.all, &net->xfrm.policy_all);
588 write_unlock_bh(&xfrm_policy_lock);
589
590 if (delpol)
591 xfrm_policy_kill(delpol);
592 else if (xfrm_bydst_should_resize(net, dir, NULL))
593 schedule_work(&net->xfrm.policy_hash_work);
594
595 return 0;
596 }
597 EXPORT_SYMBOL(xfrm_policy_insert);
598
599 struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
600 int dir, struct xfrm_selector *sel,
601 struct xfrm_sec_ctx *ctx, int delete,
602 int *err)
603 {
604 struct xfrm_policy *pol, *ret;
605 struct hlist_head *chain;
606 struct hlist_node *entry;
607
608 *err = 0;
609 write_lock_bh(&xfrm_policy_lock);
610 chain = policy_hash_bysel(net, sel, sel->family, dir);
611 ret = NULL;
612 hlist_for_each_entry(pol, entry, chain, bydst) {
613 if (pol->type == type &&
614 (mark & pol->mark.m) == pol->mark.v &&
615 !selector_cmp(sel, &pol->selector) &&
616 xfrm_sec_ctx_match(ctx, pol->security)) {
617 xfrm_pol_hold(pol);
618 if (delete) {
619 *err = security_xfrm_policy_delete(
620 pol->security);
621 if (*err) {
622 write_unlock_bh(&xfrm_policy_lock);
623 return pol;
624 }
625 __xfrm_policy_unlink(pol, dir);
626 }
627 ret = pol;
628 break;
629 }
630 }
631 write_unlock_bh(&xfrm_policy_lock);
632
633 if (ret && delete)
634 xfrm_policy_kill(ret);
635 return ret;
636 }
637 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
638
639 struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
640 int dir, u32 id, int delete, int *err)
641 {
642 struct xfrm_policy *pol, *ret;
643 struct hlist_head *chain;
644 struct hlist_node *entry;
645
646 *err = -ENOENT;
647 if (xfrm_policy_id2dir(id) != dir)
648 return NULL;
649
650 *err = 0;
651 write_lock_bh(&xfrm_policy_lock);
652 chain = net->xfrm.policy_byidx + idx_hash(net, id);
653 ret = NULL;
654 hlist_for_each_entry(pol, entry, chain, byidx) {
655 if (pol->type == type && pol->index == id &&
656 (mark & pol->mark.m) == pol->mark.v) {
657 xfrm_pol_hold(pol);
658 if (delete) {
659 *err = security_xfrm_policy_delete(
660 pol->security);
661 if (*err) {
662 write_unlock_bh(&xfrm_policy_lock);
663 return pol;
664 }
665 __xfrm_policy_unlink(pol, dir);
666 }
667 ret = pol;
668 break;
669 }
670 }
671 write_unlock_bh(&xfrm_policy_lock);
672
673 if (ret && delete)
674 xfrm_policy_kill(ret);
675 return ret;
676 }
677 EXPORT_SYMBOL(xfrm_policy_byid);
678
679 #ifdef CONFIG_SECURITY_NETWORK_XFRM
680 static inline int
681 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
682 {
683 int dir, err = 0;
684
685 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
686 struct xfrm_policy *pol;
687 struct hlist_node *entry;
688 int i;
689
690 hlist_for_each_entry(pol, entry,
691 &net->xfrm.policy_inexact[dir], bydst) {
692 if (pol->type != type)
693 continue;
694 err = security_xfrm_policy_delete(pol->security);
695 if (err) {
696 xfrm_audit_policy_delete(pol, 0,
697 audit_info->loginuid,
698 audit_info->sessionid,
699 audit_info->secid);
700 return err;
701 }
702 }
703 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
704 hlist_for_each_entry(pol, entry,
705 net->xfrm.policy_bydst[dir].table + i,
706 bydst) {
707 if (pol->type != type)
708 continue;
709 err = security_xfrm_policy_delete(
710 pol->security);
711 if (err) {
712 xfrm_audit_policy_delete(pol, 0,
713 audit_info->loginuid,
714 audit_info->sessionid,
715 audit_info->secid);
716 return err;
717 }
718 }
719 }
720 }
721 return err;
722 }
723 #else
724 static inline int
725 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
726 {
727 return 0;
728 }
729 #endif
730
731 int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
732 {
733 int dir, err = 0, cnt = 0;
734
735 write_lock_bh(&xfrm_policy_lock);
736
737 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
738 if (err)
739 goto out;
740
741 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
742 struct xfrm_policy *pol;
743 struct hlist_node *entry;
744 int i;
745
746 again1:
747 hlist_for_each_entry(pol, entry,
748 &net->xfrm.policy_inexact[dir], bydst) {
749 if (pol->type != type)
750 continue;
751 __xfrm_policy_unlink(pol, dir);
752 write_unlock_bh(&xfrm_policy_lock);
753 cnt++;
754
755 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
756 audit_info->sessionid,
757 audit_info->secid);
758
759 xfrm_policy_kill(pol);
760
761 write_lock_bh(&xfrm_policy_lock);
762 goto again1;
763 }
764
765 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
766 again2:
767 hlist_for_each_entry(pol, entry,
768 net->xfrm.policy_bydst[dir].table + i,
769 bydst) {
770 if (pol->type != type)
771 continue;
772 __xfrm_policy_unlink(pol, dir);
773 write_unlock_bh(&xfrm_policy_lock);
774 cnt++;
775
776 xfrm_audit_policy_delete(pol, 1,
777 audit_info->loginuid,
778 audit_info->sessionid,
779 audit_info->secid);
780 xfrm_policy_kill(pol);
781
782 write_lock_bh(&xfrm_policy_lock);
783 goto again2;
784 }
785 }
786
787 }
788 if (!cnt)
789 err = -ESRCH;
790 out:
791 write_unlock_bh(&xfrm_policy_lock);
792 return err;
793 }
794 EXPORT_SYMBOL(xfrm_policy_flush);
795
796 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
797 int (*func)(struct xfrm_policy *, int, int, void*),
798 void *data)
799 {
800 struct xfrm_policy *pol;
801 struct xfrm_policy_walk_entry *x;
802 int error = 0;
803
804 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
805 walk->type != XFRM_POLICY_TYPE_ANY)
806 return -EINVAL;
807
808 if (list_empty(&walk->walk.all) && walk->seq != 0)
809 return 0;
810
811 write_lock_bh(&xfrm_policy_lock);
812 if (list_empty(&walk->walk.all))
813 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
814 else
815 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
816 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
817 if (x->dead)
818 continue;
819 pol = container_of(x, struct xfrm_policy, walk);
820 if (walk->type != XFRM_POLICY_TYPE_ANY &&
821 walk->type != pol->type)
822 continue;
823 error = func(pol, xfrm_policy_id2dir(pol->index),
824 walk->seq, data);
825 if (error) {
826 list_move_tail(&walk->walk.all, &x->all);
827 goto out;
828 }
829 walk->seq++;
830 }
831 if (walk->seq == 0) {
832 error = -ENOENT;
833 goto out;
834 }
835 list_del_init(&walk->walk.all);
836 out:
837 write_unlock_bh(&xfrm_policy_lock);
838 return error;
839 }
840 EXPORT_SYMBOL(xfrm_policy_walk);
841
842 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
843 {
844 INIT_LIST_HEAD(&walk->walk.all);
845 walk->walk.dead = 1;
846 walk->type = type;
847 walk->seq = 0;
848 }
849 EXPORT_SYMBOL(xfrm_policy_walk_init);
850
851 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
852 {
853 if (list_empty(&walk->walk.all))
854 return;
855
856 write_lock_bh(&xfrm_policy_lock);
857 list_del(&walk->walk.all);
858 write_unlock_bh(&xfrm_policy_lock);
859 }
860 EXPORT_SYMBOL(xfrm_policy_walk_done);
861
862 /*
863 * Find policy to apply to this flow.
864 *
865 * Returns 0 if policy found, else an -errno.
866 */
867 static int xfrm_policy_match(struct xfrm_policy *pol, const struct flowi *fl,
868 u8 type, u16 family, int dir)
869 {
870 struct xfrm_selector *sel = &pol->selector;
871 int match, ret = -ESRCH;
872
873 if (pol->family != family ||
874 (fl->mark & pol->mark.m) != pol->mark.v ||
875 pol->type != type)
876 return ret;
877
878 match = xfrm_selector_match(sel, fl, family);
879 if (match)
880 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
881 dir);
882
883 return ret;
884 }
885
886 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
887 const struct flowi *fl,
888 u16 family, u8 dir)
889 {
890 int err;
891 struct xfrm_policy *pol, *ret;
892 xfrm_address_t *daddr, *saddr;
893 struct hlist_node *entry;
894 struct hlist_head *chain;
895 u32 priority = ~0U;
896
897 daddr = xfrm_flowi_daddr(fl, family);
898 saddr = xfrm_flowi_saddr(fl, family);
899 if (unlikely(!daddr || !saddr))
900 return NULL;
901
902 read_lock_bh(&xfrm_policy_lock);
903 chain = policy_hash_direct(net, daddr, saddr, family, dir);
904 ret = NULL;
905 hlist_for_each_entry(pol, entry, chain, bydst) {
906 err = xfrm_policy_match(pol, fl, type, family, dir);
907 if (err) {
908 if (err == -ESRCH)
909 continue;
910 else {
911 ret = ERR_PTR(err);
912 goto fail;
913 }
914 } else {
915 ret = pol;
916 priority = ret->priority;
917 break;
918 }
919 }
920 chain = &net->xfrm.policy_inexact[dir];
921 hlist_for_each_entry(pol, entry, chain, bydst) {
922 err = xfrm_policy_match(pol, fl, type, family, dir);
923 if (err) {
924 if (err == -ESRCH)
925 continue;
926 else {
927 ret = ERR_PTR(err);
928 goto fail;
929 }
930 } else if (pol->priority < priority) {
931 ret = pol;
932 break;
933 }
934 }
935 if (ret)
936 xfrm_pol_hold(ret);
937 fail:
938 read_unlock_bh(&xfrm_policy_lock);
939
940 return ret;
941 }
942
943 static struct xfrm_policy *
944 __xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
945 {
946 #ifdef CONFIG_XFRM_SUB_POLICY
947 struct xfrm_policy *pol;
948
949 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
950 if (pol != NULL)
951 return pol;
952 #endif
953 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
954 }
955
956 static struct flow_cache_object *
957 xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family,
958 u8 dir, struct flow_cache_object *old_obj, void *ctx)
959 {
960 struct xfrm_policy *pol;
961
962 if (old_obj)
963 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
964
965 pol = __xfrm_policy_lookup(net, fl, family, dir);
966 if (IS_ERR_OR_NULL(pol))
967 return ERR_CAST(pol);
968
969 /* Resolver returns two references:
970 * one for cache and one for caller of flow_cache_lookup() */
971 xfrm_pol_hold(pol);
972
973 return &pol->flo;
974 }
975
976 static inline int policy_to_flow_dir(int dir)
977 {
978 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
979 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
980 XFRM_POLICY_FWD == FLOW_DIR_FWD)
981 return dir;
982 switch (dir) {
983 default:
984 case XFRM_POLICY_IN:
985 return FLOW_DIR_IN;
986 case XFRM_POLICY_OUT:
987 return FLOW_DIR_OUT;
988 case XFRM_POLICY_FWD:
989 return FLOW_DIR_FWD;
990 }
991 }
992
993 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
994 {
995 struct xfrm_policy *pol;
996
997 read_lock_bh(&xfrm_policy_lock);
998 if ((pol = sk->sk_policy[dir]) != NULL) {
999 int match = xfrm_selector_match(&pol->selector, fl,
1000 sk->sk_family);
1001 int err = 0;
1002
1003 if (match) {
1004 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1005 pol = NULL;
1006 goto out;
1007 }
1008 err = security_xfrm_policy_lookup(pol->security,
1009 fl->secid,
1010 policy_to_flow_dir(dir));
1011 if (!err)
1012 xfrm_pol_hold(pol);
1013 else if (err == -ESRCH)
1014 pol = NULL;
1015 else
1016 pol = ERR_PTR(err);
1017 } else
1018 pol = NULL;
1019 }
1020 out:
1021 read_unlock_bh(&xfrm_policy_lock);
1022 return pol;
1023 }
1024
1025 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1026 {
1027 struct net *net = xp_net(pol);
1028 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1029 pol->family, dir);
1030
1031 list_add(&pol->walk.all, &net->xfrm.policy_all);
1032 hlist_add_head(&pol->bydst, chain);
1033 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1034 net->xfrm.policy_count[dir]++;
1035 xfrm_pol_hold(pol);
1036
1037 if (xfrm_bydst_should_resize(net, dir, NULL))
1038 schedule_work(&net->xfrm.policy_hash_work);
1039 }
1040
1041 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1042 int dir)
1043 {
1044 struct net *net = xp_net(pol);
1045
1046 if (hlist_unhashed(&pol->bydst))
1047 return NULL;
1048
1049 hlist_del(&pol->bydst);
1050 hlist_del(&pol->byidx);
1051 list_del(&pol->walk.all);
1052 net->xfrm.policy_count[dir]--;
1053
1054 return pol;
1055 }
1056
1057 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1058 {
1059 write_lock_bh(&xfrm_policy_lock);
1060 pol = __xfrm_policy_unlink(pol, dir);
1061 write_unlock_bh(&xfrm_policy_lock);
1062 if (pol) {
1063 xfrm_policy_kill(pol);
1064 return 0;
1065 }
1066 return -ENOENT;
1067 }
1068 EXPORT_SYMBOL(xfrm_policy_delete);
1069
1070 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1071 {
1072 struct net *net = xp_net(pol);
1073 struct xfrm_policy *old_pol;
1074
1075 #ifdef CONFIG_XFRM_SUB_POLICY
1076 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1077 return -EINVAL;
1078 #endif
1079
1080 write_lock_bh(&xfrm_policy_lock);
1081 old_pol = sk->sk_policy[dir];
1082 sk->sk_policy[dir] = pol;
1083 if (pol) {
1084 pol->curlft.add_time = get_seconds();
1085 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1086 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1087 }
1088 if (old_pol)
1089 /* Unlinking succeeds always. This is the only function
1090 * allowed to delete or replace socket policy.
1091 */
1092 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1093 write_unlock_bh(&xfrm_policy_lock);
1094
1095 if (old_pol) {
1096 xfrm_policy_kill(old_pol);
1097 }
1098 return 0;
1099 }
1100
1101 static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1102 {
1103 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1104
1105 if (newp) {
1106 newp->selector = old->selector;
1107 if (security_xfrm_policy_clone(old->security,
1108 &newp->security)) {
1109 kfree(newp);
1110 return NULL; /* ENOMEM */
1111 }
1112 newp->lft = old->lft;
1113 newp->curlft = old->curlft;
1114 newp->mark = old->mark;
1115 newp->action = old->action;
1116 newp->flags = old->flags;
1117 newp->xfrm_nr = old->xfrm_nr;
1118 newp->index = old->index;
1119 newp->type = old->type;
1120 memcpy(newp->xfrm_vec, old->xfrm_vec,
1121 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1122 write_lock_bh(&xfrm_policy_lock);
1123 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1124 write_unlock_bh(&xfrm_policy_lock);
1125 xfrm_pol_put(newp);
1126 }
1127 return newp;
1128 }
1129
1130 int __xfrm_sk_clone_policy(struct sock *sk)
1131 {
1132 struct xfrm_policy *p0 = sk->sk_policy[0],
1133 *p1 = sk->sk_policy[1];
1134
1135 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1136 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1137 return -ENOMEM;
1138 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1139 return -ENOMEM;
1140 return 0;
1141 }
1142
1143 static int
1144 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1145 unsigned short family)
1146 {
1147 int err;
1148 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1149
1150 if (unlikely(afinfo == NULL))
1151 return -EINVAL;
1152 err = afinfo->get_saddr(net, local, remote);
1153 xfrm_policy_put_afinfo(afinfo);
1154 return err;
1155 }
1156
1157 /* Resolve list of templates for the flow, given policy. */
1158
1159 static int
1160 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1161 struct xfrm_state **xfrm, unsigned short family)
1162 {
1163 struct net *net = xp_net(policy);
1164 int nx;
1165 int i, error;
1166 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1167 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1168 xfrm_address_t tmp;
1169
1170 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1171 struct xfrm_state *x;
1172 xfrm_address_t *remote = daddr;
1173 xfrm_address_t *local = saddr;
1174 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1175
1176 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1177 tmpl->mode == XFRM_MODE_BEET) {
1178 remote = &tmpl->id.daddr;
1179 local = &tmpl->saddr;
1180 if (xfrm_addr_any(local, tmpl->encap_family)) {
1181 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
1182 if (error)
1183 goto fail;
1184 local = &tmp;
1185 }
1186 }
1187
1188 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1189
1190 if (x && x->km.state == XFRM_STATE_VALID) {
1191 xfrm[nx++] = x;
1192 daddr = remote;
1193 saddr = local;
1194 continue;
1195 }
1196 if (x) {
1197 error = (x->km.state == XFRM_STATE_ERROR ?
1198 -EINVAL : -EAGAIN);
1199 xfrm_state_put(x);
1200 }
1201 else if (error == -ESRCH)
1202 error = -EAGAIN;
1203
1204 if (!tmpl->optional)
1205 goto fail;
1206 }
1207 return nx;
1208
1209 fail:
1210 for (nx--; nx>=0; nx--)
1211 xfrm_state_put(xfrm[nx]);
1212 return error;
1213 }
1214
1215 static int
1216 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1217 struct xfrm_state **xfrm, 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(const 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, 0);
1342 xfrm_policy_put_afinfo(afinfo);
1343
1344 if (likely(xdst))
1345 xdst->flo.ops = &xfrm_bundle_fc_ops;
1346 else
1347 xdst = ERR_PTR(-ENOBUFS);
1348
1349 return xdst;
1350 }
1351
1352 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1353 int nfheader_len)
1354 {
1355 struct xfrm_policy_afinfo *afinfo =
1356 xfrm_policy_get_afinfo(dst->ops->family);
1357 int err;
1358
1359 if (!afinfo)
1360 return -EINVAL;
1361
1362 err = afinfo->init_path(path, dst, nfheader_len);
1363
1364 xfrm_policy_put_afinfo(afinfo);
1365
1366 return err;
1367 }
1368
1369 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
1370 const struct flowi *fl)
1371 {
1372 struct xfrm_policy_afinfo *afinfo =
1373 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1374 int err;
1375
1376 if (!afinfo)
1377 return -EINVAL;
1378
1379 err = afinfo->fill_dst(xdst, dev, fl);
1380
1381 xfrm_policy_put_afinfo(afinfo);
1382
1383 return err;
1384 }
1385
1386
1387 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1388 * all the metrics... Shortly, bundle a bundle.
1389 */
1390
1391 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1392 struct xfrm_state **xfrm, int nx,
1393 const struct flowi *fl,
1394 struct dst_entry *dst)
1395 {
1396 struct net *net = xp_net(policy);
1397 unsigned long now = jiffies;
1398 struct net_device *dev;
1399 struct dst_entry *dst_prev = NULL;
1400 struct dst_entry *dst0 = NULL;
1401 int i = 0;
1402 int err;
1403 int header_len = 0;
1404 int nfheader_len = 0;
1405 int trailer_len = 0;
1406 int tos;
1407 int family = policy->selector.family;
1408 xfrm_address_t saddr, daddr;
1409
1410 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1411
1412 tos = xfrm_get_tos(fl, family);
1413 err = tos;
1414 if (tos < 0)
1415 goto put_states;
1416
1417 dst_hold(dst);
1418
1419 for (; i < nx; i++) {
1420 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1421 struct dst_entry *dst1 = &xdst->u.dst;
1422
1423 err = PTR_ERR(xdst);
1424 if (IS_ERR(xdst)) {
1425 dst_release(dst);
1426 goto put_states;
1427 }
1428
1429 if (!dst_prev)
1430 dst0 = dst1;
1431 else {
1432 dst_prev->child = dst_clone(dst1);
1433 dst1->flags |= DST_NOHASH;
1434 }
1435
1436 xdst->route = dst;
1437 dst_copy_metrics(dst1, dst);
1438
1439 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1440 family = xfrm[i]->props.family;
1441 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1442 family);
1443 err = PTR_ERR(dst);
1444 if (IS_ERR(dst))
1445 goto put_states;
1446 } else
1447 dst_hold(dst);
1448
1449 dst1->xfrm = xfrm[i];
1450 xdst->xfrm_genid = xfrm[i]->genid;
1451
1452 dst1->obsolete = -1;
1453 dst1->flags |= DST_HOST;
1454 dst1->lastuse = now;
1455
1456 dst1->input = dst_discard;
1457 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1458
1459 dst1->next = dst_prev;
1460 dst_prev = dst1;
1461
1462 header_len += xfrm[i]->props.header_len;
1463 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1464 nfheader_len += xfrm[i]->props.header_len;
1465 trailer_len += xfrm[i]->props.trailer_len;
1466 }
1467
1468 dst_prev->child = dst;
1469 dst0->path = dst;
1470
1471 err = -ENODEV;
1472 dev = dst->dev;
1473 if (!dev)
1474 goto free_dst;
1475
1476 /* Copy neighbour for reachability confirmation */
1477 dst0->neighbour = neigh_clone(dst->neighbour);
1478
1479 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1480 xfrm_init_pmtu(dst_prev);
1481
1482 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1483 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1484
1485 err = xfrm_fill_dst(xdst, dev, fl);
1486 if (err)
1487 goto free_dst;
1488
1489 dst_prev->header_len = header_len;
1490 dst_prev->trailer_len = trailer_len;
1491 header_len -= xdst->u.dst.xfrm->props.header_len;
1492 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1493 }
1494
1495 out:
1496 return dst0;
1497
1498 put_states:
1499 for (; i < nx; i++)
1500 xfrm_state_put(xfrm[i]);
1501 free_dst:
1502 if (dst0)
1503 dst_free(dst0);
1504 dst0 = ERR_PTR(err);
1505 goto out;
1506 }
1507
1508 static int inline
1509 xfrm_dst_alloc_copy(void **target, const void *src, int size)
1510 {
1511 if (!*target) {
1512 *target = kmalloc(size, GFP_ATOMIC);
1513 if (!*target)
1514 return -ENOMEM;
1515 }
1516 memcpy(*target, src, size);
1517 return 0;
1518 }
1519
1520 static int inline
1521 xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1522 {
1523 #ifdef CONFIG_XFRM_SUB_POLICY
1524 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1525 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1526 sel, sizeof(*sel));
1527 #else
1528 return 0;
1529 #endif
1530 }
1531
1532 static int inline
1533 xfrm_dst_update_origin(struct dst_entry *dst, const struct flowi *fl)
1534 {
1535 #ifdef CONFIG_XFRM_SUB_POLICY
1536 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1537 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1538 #else
1539 return 0;
1540 #endif
1541 }
1542
1543 static int xfrm_expand_policies(const struct flowi *fl, u16 family,
1544 struct xfrm_policy **pols,
1545 int *num_pols, int *num_xfrms)
1546 {
1547 int i;
1548
1549 if (*num_pols == 0 || !pols[0]) {
1550 *num_pols = 0;
1551 *num_xfrms = 0;
1552 return 0;
1553 }
1554 if (IS_ERR(pols[0]))
1555 return PTR_ERR(pols[0]);
1556
1557 *num_xfrms = pols[0]->xfrm_nr;
1558
1559 #ifdef CONFIG_XFRM_SUB_POLICY
1560 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1561 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1562 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1563 XFRM_POLICY_TYPE_MAIN,
1564 fl, family,
1565 XFRM_POLICY_OUT);
1566 if (pols[1]) {
1567 if (IS_ERR(pols[1])) {
1568 xfrm_pols_put(pols, *num_pols);
1569 return PTR_ERR(pols[1]);
1570 }
1571 (*num_pols) ++;
1572 (*num_xfrms) += pols[1]->xfrm_nr;
1573 }
1574 }
1575 #endif
1576 for (i = 0; i < *num_pols; i++) {
1577 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1578 *num_xfrms = -1;
1579 break;
1580 }
1581 }
1582
1583 return 0;
1584
1585 }
1586
1587 static struct xfrm_dst *
1588 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
1589 struct flowi *fl, u16 family,
1590 struct dst_entry *dst_orig)
1591 {
1592 struct net *net = xp_net(pols[0]);
1593 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1594 struct dst_entry *dst;
1595 struct xfrm_dst *xdst;
1596 int err;
1597
1598 /* Try to instantiate a bundle */
1599 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1600 if (err <= 0) {
1601 if (err != 0 && err != -EAGAIN)
1602 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1603 return ERR_PTR(err);
1604 }
1605
1606 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1607 if (IS_ERR(dst)) {
1608 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1609 return ERR_CAST(dst);
1610 }
1611
1612 xdst = (struct xfrm_dst *)dst;
1613 xdst->num_xfrms = err;
1614 if (num_pols > 1)
1615 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1616 else
1617 err = xfrm_dst_update_origin(dst, fl);
1618 if (unlikely(err)) {
1619 dst_free(dst);
1620 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1621 return ERR_PTR(err);
1622 }
1623
1624 xdst->num_pols = num_pols;
1625 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1626 xdst->policy_genid = atomic_read(&pols[0]->genid);
1627
1628 return xdst;
1629 }
1630
1631 static struct flow_cache_object *
1632 xfrm_bundle_lookup(struct net *net, struct flowi *fl, u16 family, u8 dir,
1633 struct flow_cache_object *oldflo, void *ctx)
1634 {
1635 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1636 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1637 struct xfrm_dst *xdst, *new_xdst;
1638 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1639
1640 /* Check if the policies from old bundle are usable */
1641 xdst = NULL;
1642 if (oldflo) {
1643 xdst = container_of(oldflo, struct xfrm_dst, flo);
1644 num_pols = xdst->num_pols;
1645 num_xfrms = xdst->num_xfrms;
1646 pol_dead = 0;
1647 for (i = 0; i < num_pols; i++) {
1648 pols[i] = xdst->pols[i];
1649 pol_dead |= pols[i]->walk.dead;
1650 }
1651 if (pol_dead) {
1652 dst_free(&xdst->u.dst);
1653 xdst = NULL;
1654 num_pols = 0;
1655 num_xfrms = 0;
1656 oldflo = NULL;
1657 }
1658 }
1659
1660 /* Resolve policies to use if we couldn't get them from
1661 * previous cache entry */
1662 if (xdst == NULL) {
1663 num_pols = 1;
1664 pols[0] = __xfrm_policy_lookup(net, fl, family, dir);
1665 err = xfrm_expand_policies(fl, family, pols,
1666 &num_pols, &num_xfrms);
1667 if (err < 0)
1668 goto inc_error;
1669 if (num_pols == 0)
1670 return NULL;
1671 if (num_xfrms <= 0)
1672 goto make_dummy_bundle;
1673 }
1674
1675 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1676 if (IS_ERR(new_xdst)) {
1677 err = PTR_ERR(new_xdst);
1678 if (err != -EAGAIN)
1679 goto error;
1680 if (oldflo == NULL)
1681 goto make_dummy_bundle;
1682 dst_hold(&xdst->u.dst);
1683 return oldflo;
1684 } else if (new_xdst == NULL) {
1685 num_xfrms = 0;
1686 if (oldflo == NULL)
1687 goto make_dummy_bundle;
1688 xdst->num_xfrms = 0;
1689 dst_hold(&xdst->u.dst);
1690 return oldflo;
1691 }
1692
1693 /* Kill the previous bundle */
1694 if (xdst) {
1695 /* The policies were stolen for newly generated bundle */
1696 xdst->num_pols = 0;
1697 dst_free(&xdst->u.dst);
1698 }
1699
1700 /* Flow cache does not have reference, it dst_free()'s,
1701 * but we do need to return one reference for original caller */
1702 dst_hold(&new_xdst->u.dst);
1703 return &new_xdst->flo;
1704
1705 make_dummy_bundle:
1706 /* We found policies, but there's no bundles to instantiate:
1707 * either because the policy blocks, has no transformations or
1708 * we could not build template (no xfrm_states).*/
1709 xdst = xfrm_alloc_dst(net, family);
1710 if (IS_ERR(xdst)) {
1711 xfrm_pols_put(pols, num_pols);
1712 return ERR_CAST(xdst);
1713 }
1714 xdst->num_pols = num_pols;
1715 xdst->num_xfrms = num_xfrms;
1716 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1717
1718 dst_hold(&xdst->u.dst);
1719 return &xdst->flo;
1720
1721 inc_error:
1722 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1723 error:
1724 if (xdst != NULL)
1725 dst_free(&xdst->u.dst);
1726 else
1727 xfrm_pols_put(pols, num_pols);
1728 return ERR_PTR(err);
1729 }
1730
1731 /* Main function: finds/creates a bundle for given flow.
1732 *
1733 * At the moment we eat a raw IP route. Mostly to speed up lookups
1734 * on interfaces with disabled IPsec.
1735 */
1736 int __xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1737 struct sock *sk, int flags)
1738 {
1739 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1740 struct flow_cache_object *flo;
1741 struct xfrm_dst *xdst;
1742 struct dst_entry *dst, *dst_orig = *dst_p, *route;
1743 u16 family = dst_orig->ops->family;
1744 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1745 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
1746
1747 restart:
1748 dst = NULL;
1749 xdst = NULL;
1750 route = NULL;
1751
1752 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1753 num_pols = 1;
1754 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1755 err = xfrm_expand_policies(fl, family, pols,
1756 &num_pols, &num_xfrms);
1757 if (err < 0)
1758 goto dropdst;
1759
1760 if (num_pols) {
1761 if (num_xfrms <= 0) {
1762 drop_pols = num_pols;
1763 goto no_transform;
1764 }
1765
1766 xdst = xfrm_resolve_and_create_bundle(
1767 pols, num_pols, fl,
1768 family, dst_orig);
1769 if (IS_ERR(xdst)) {
1770 xfrm_pols_put(pols, num_pols);
1771 err = PTR_ERR(xdst);
1772 goto dropdst;
1773 } else if (xdst == NULL) {
1774 num_xfrms = 0;
1775 drop_pols = num_pols;
1776 goto no_transform;
1777 }
1778
1779 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
1780 xdst->u.dst.next = xfrm_policy_sk_bundles;
1781 xfrm_policy_sk_bundles = &xdst->u.dst;
1782 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
1783
1784 route = xdst->route;
1785 }
1786 }
1787
1788 if (xdst == NULL) {
1789 /* To accelerate a bit... */
1790 if ((dst_orig->flags & DST_NOXFRM) ||
1791 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1792 goto nopol;
1793
1794 flo = flow_cache_lookup(net, fl, family, dir,
1795 xfrm_bundle_lookup, dst_orig);
1796 if (flo == NULL)
1797 goto nopol;
1798 if (IS_ERR(flo)) {
1799 err = PTR_ERR(flo);
1800 goto dropdst;
1801 }
1802 xdst = container_of(flo, struct xfrm_dst, flo);
1803
1804 num_pols = xdst->num_pols;
1805 num_xfrms = xdst->num_xfrms;
1806 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
1807 route = xdst->route;
1808 }
1809
1810 dst = &xdst->u.dst;
1811 if (route == NULL && num_xfrms > 0) {
1812 /* The only case when xfrm_bundle_lookup() returns a
1813 * bundle with null route, is when the template could
1814 * not be resolved. It means policies are there, but
1815 * bundle could not be created, since we don't yet
1816 * have the xfrm_state's. We need to wait for KM to
1817 * negotiate new SA's or bail out with error.*/
1818 if (net->xfrm.sysctl_larval_drop) {
1819 /* EREMOTE tells the caller to generate
1820 * a one-shot blackhole route. */
1821 dst_release(dst);
1822 xfrm_pols_put(pols, drop_pols);
1823 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1824 return -EREMOTE;
1825 }
1826 if (flags & XFRM_LOOKUP_WAIT) {
1827 DECLARE_WAITQUEUE(wait, current);
1828
1829 add_wait_queue(&net->xfrm.km_waitq, &wait);
1830 set_current_state(TASK_INTERRUPTIBLE);
1831 schedule();
1832 set_current_state(TASK_RUNNING);
1833 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1834
1835 if (!signal_pending(current)) {
1836 dst_release(dst);
1837 goto restart;
1838 }
1839
1840 err = -ERESTART;
1841 } else
1842 err = -EAGAIN;
1843
1844 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1845 goto error;
1846 }
1847
1848 no_transform:
1849 if (num_pols == 0)
1850 goto nopol;
1851
1852 if ((flags & XFRM_LOOKUP_ICMP) &&
1853 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
1854 err = -ENOENT;
1855 goto error;
1856 }
1857
1858 for (i = 0; i < num_pols; i++)
1859 pols[i]->curlft.use_time = get_seconds();
1860
1861 if (num_xfrms < 0) {
1862 /* Prohibit the flow */
1863 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1864 err = -EPERM;
1865 goto error;
1866 } else if (num_xfrms > 0) {
1867 /* Flow transformed */
1868 *dst_p = dst;
1869 dst_release(dst_orig);
1870 } else {
1871 /* Flow passes untransformed */
1872 dst_release(dst);
1873 }
1874 ok:
1875 xfrm_pols_put(pols, drop_pols);
1876 return 0;
1877
1878 nopol:
1879 if (!(flags & XFRM_LOOKUP_ICMP))
1880 goto ok;
1881 err = -ENOENT;
1882 error:
1883 dst_release(dst);
1884 dropdst:
1885 dst_release(dst_orig);
1886 *dst_p = NULL;
1887 xfrm_pols_put(pols, drop_pols);
1888 return err;
1889 }
1890 EXPORT_SYMBOL(__xfrm_lookup);
1891
1892 int xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1893 struct sock *sk, int flags)
1894 {
1895 int err = __xfrm_lookup(net, dst_p, fl, sk, flags);
1896
1897 if (err == -EREMOTE) {
1898 dst_release(*dst_p);
1899 *dst_p = NULL;
1900 err = -EAGAIN;
1901 }
1902
1903 return err;
1904 }
1905 EXPORT_SYMBOL(xfrm_lookup);
1906
1907 static inline int
1908 xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
1909 {
1910 struct xfrm_state *x;
1911
1912 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1913 return 0;
1914 x = skb->sp->xvec[idx];
1915 if (!x->type->reject)
1916 return 0;
1917 return x->type->reject(x, skb, fl);
1918 }
1919
1920 /* When skb is transformed back to its "native" form, we have to
1921 * check policy restrictions. At the moment we make this in maximally
1922 * stupid way. Shame on me. :-) Of course, connected sockets must
1923 * have policy cached at them.
1924 */
1925
1926 static inline int
1927 xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1928 unsigned short family)
1929 {
1930 if (xfrm_state_kern(x))
1931 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1932 return x->id.proto == tmpl->id.proto &&
1933 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1934 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1935 x->props.mode == tmpl->mode &&
1936 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1937 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1938 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1939 xfrm_state_addr_cmp(tmpl, x, family));
1940 }
1941
1942 /*
1943 * 0 or more than 0 is returned when validation is succeeded (either bypass
1944 * because of optional transport mode, or next index of the mathced secpath
1945 * state with the template.
1946 * -1 is returned when no matching template is found.
1947 * Otherwise "-2 - errored_index" is returned.
1948 */
1949 static inline int
1950 xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1951 unsigned short family)
1952 {
1953 int idx = start;
1954
1955 if (tmpl->optional) {
1956 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1957 return start;
1958 } else
1959 start = -1;
1960 for (; idx < sp->len; idx++) {
1961 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1962 return ++idx;
1963 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1964 if (start == -1)
1965 start = -2-idx;
1966 break;
1967 }
1968 }
1969 return start;
1970 }
1971
1972 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1973 unsigned int family, int reverse)
1974 {
1975 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1976 int err;
1977
1978 if (unlikely(afinfo == NULL))
1979 return -EAFNOSUPPORT;
1980
1981 afinfo->decode_session(skb, fl, reverse);
1982 err = security_xfrm_decode_session(skb, &fl->secid);
1983 xfrm_policy_put_afinfo(afinfo);
1984 return err;
1985 }
1986 EXPORT_SYMBOL(__xfrm_decode_session);
1987
1988 static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1989 {
1990 for (; k < sp->len; k++) {
1991 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
1992 *idxp = k;
1993 return 1;
1994 }
1995 }
1996
1997 return 0;
1998 }
1999
2000 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
2001 unsigned short family)
2002 {
2003 struct net *net = dev_net(skb->dev);
2004 struct xfrm_policy *pol;
2005 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2006 int npols = 0;
2007 int xfrm_nr;
2008 int pi;
2009 int reverse;
2010 struct flowi fl;
2011 u8 fl_dir;
2012 int xerr_idx = -1;
2013
2014 reverse = dir & ~XFRM_POLICY_MASK;
2015 dir &= XFRM_POLICY_MASK;
2016 fl_dir = policy_to_flow_dir(dir);
2017
2018 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
2019 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
2020 return 0;
2021 }
2022
2023 nf_nat_decode_session(skb, &fl, family);
2024
2025 /* First, check used SA against their selectors. */
2026 if (skb->sp) {
2027 int i;
2028
2029 for (i=skb->sp->len-1; i>=0; i--) {
2030 struct xfrm_state *x = skb->sp->xvec[i];
2031 if (!xfrm_selector_match(&x->sel, &fl, family)) {
2032 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
2033 return 0;
2034 }
2035 }
2036 }
2037
2038 pol = NULL;
2039 if (sk && sk->sk_policy[dir]) {
2040 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
2041 if (IS_ERR(pol)) {
2042 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2043 return 0;
2044 }
2045 }
2046
2047 if (!pol) {
2048 struct flow_cache_object *flo;
2049
2050 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2051 xfrm_policy_lookup, NULL);
2052 if (IS_ERR_OR_NULL(flo))
2053 pol = ERR_CAST(flo);
2054 else
2055 pol = container_of(flo, struct xfrm_policy, flo);
2056 }
2057
2058 if (IS_ERR(pol)) {
2059 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2060 return 0;
2061 }
2062
2063 if (!pol) {
2064 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
2065 xfrm_secpath_reject(xerr_idx, skb, &fl);
2066 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
2067 return 0;
2068 }
2069 return 1;
2070 }
2071
2072 pol->curlft.use_time = get_seconds();
2073
2074 pols[0] = pol;
2075 npols ++;
2076 #ifdef CONFIG_XFRM_SUB_POLICY
2077 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
2078 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
2079 &fl, family,
2080 XFRM_POLICY_IN);
2081 if (pols[1]) {
2082 if (IS_ERR(pols[1])) {
2083 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2084 return 0;
2085 }
2086 pols[1]->curlft.use_time = get_seconds();
2087 npols ++;
2088 }
2089 }
2090 #endif
2091
2092 if (pol->action == XFRM_POLICY_ALLOW) {
2093 struct sec_path *sp;
2094 static struct sec_path dummy;
2095 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
2096 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
2097 struct xfrm_tmpl **tpp = tp;
2098 int ti = 0;
2099 int i, k;
2100
2101 if ((sp = skb->sp) == NULL)
2102 sp = &dummy;
2103
2104 for (pi = 0; pi < npols; pi++) {
2105 if (pols[pi] != pol &&
2106 pols[pi]->action != XFRM_POLICY_ALLOW) {
2107 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2108 goto reject;
2109 }
2110 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2111 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2112 goto reject_error;
2113 }
2114 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2115 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2116 }
2117 xfrm_nr = ti;
2118 if (npols > 1) {
2119 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2120 tpp = stp;
2121 }
2122
2123 /* For each tunnel xfrm, find the first matching tmpl.
2124 * For each tmpl before that, find corresponding xfrm.
2125 * Order is _important_. Later we will implement
2126 * some barriers, but at the moment barriers
2127 * are implied between each two transformations.
2128 */
2129 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2130 k = xfrm_policy_ok(tpp[i], sp, k, family);
2131 if (k < 0) {
2132 if (k < -1)
2133 /* "-2 - errored_index" returned */
2134 xerr_idx = -(2+k);
2135 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2136 goto reject;
2137 }
2138 }
2139
2140 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2141 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2142 goto reject;
2143 }
2144
2145 xfrm_pols_put(pols, npols);
2146 return 1;
2147 }
2148 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2149
2150 reject:
2151 xfrm_secpath_reject(xerr_idx, skb, &fl);
2152 reject_error:
2153 xfrm_pols_put(pols, npols);
2154 return 0;
2155 }
2156 EXPORT_SYMBOL(__xfrm_policy_check);
2157
2158 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2159 {
2160 struct net *net = dev_net(skb->dev);
2161 struct flowi fl;
2162 struct dst_entry *dst;
2163 int res;
2164
2165 if (xfrm_decode_session(skb, &fl, family) < 0) {
2166 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2167 return 0;
2168 }
2169
2170 skb_dst_force(skb);
2171 dst = skb_dst(skb);
2172
2173 res = xfrm_lookup(net, &dst, &fl, NULL, 0) == 0;
2174 skb_dst_set(skb, dst);
2175 return res;
2176 }
2177 EXPORT_SYMBOL(__xfrm_route_forward);
2178
2179 /* Optimize later using cookies and generation ids. */
2180
2181 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2182 {
2183 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2184 * to "-1" to force all XFRM destinations to get validated by
2185 * dst_ops->check on every use. We do this because when a
2186 * normal route referenced by an XFRM dst is obsoleted we do
2187 * not go looking around for all parent referencing XFRM dsts
2188 * so that we can invalidate them. It is just too much work.
2189 * Instead we make the checks here on every use. For example:
2190 *
2191 * XFRM dst A --> IPv4 dst X
2192 *
2193 * X is the "xdst->route" of A (X is also the "dst->path" of A
2194 * in this example). If X is marked obsolete, "A" will not
2195 * notice. That's what we are validating here via the
2196 * stale_bundle() check.
2197 *
2198 * When a policy's bundle is pruned, we dst_free() the XFRM
2199 * dst which causes it's ->obsolete field to be set to a
2200 * positive non-zero integer. If an XFRM dst has been pruned
2201 * like this, we want to force a new route lookup.
2202 */
2203 if (dst->obsolete < 0 && !stale_bundle(dst))
2204 return dst;
2205
2206 return NULL;
2207 }
2208
2209 static int stale_bundle(struct dst_entry *dst)
2210 {
2211 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC);
2212 }
2213
2214 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2215 {
2216 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2217 dst->dev = dev_net(dev)->loopback_dev;
2218 dev_hold(dst->dev);
2219 dev_put(dev);
2220 }
2221 }
2222 EXPORT_SYMBOL(xfrm_dst_ifdown);
2223
2224 static void xfrm_link_failure(struct sk_buff *skb)
2225 {
2226 /* Impossible. Such dst must be popped before reaches point of failure. */
2227 }
2228
2229 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2230 {
2231 if (dst) {
2232 if (dst->obsolete) {
2233 dst_release(dst);
2234 dst = NULL;
2235 }
2236 }
2237 return dst;
2238 }
2239
2240 static void __xfrm_garbage_collect(struct net *net)
2241 {
2242 struct dst_entry *head, *next;
2243
2244 flow_cache_flush();
2245
2246 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2247 head = xfrm_policy_sk_bundles;
2248 xfrm_policy_sk_bundles = NULL;
2249 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2250
2251 while (head) {
2252 next = head->next;
2253 dst_free(head);
2254 head = next;
2255 }
2256 }
2257
2258 static void xfrm_init_pmtu(struct dst_entry *dst)
2259 {
2260 do {
2261 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2262 u32 pmtu, route_mtu_cached;
2263
2264 pmtu = dst_mtu(dst->child);
2265 xdst->child_mtu_cached = pmtu;
2266
2267 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2268
2269 route_mtu_cached = dst_mtu(xdst->route);
2270 xdst->route_mtu_cached = route_mtu_cached;
2271
2272 if (pmtu > route_mtu_cached)
2273 pmtu = route_mtu_cached;
2274
2275 dst_metric_set(dst, RTAX_MTU, pmtu);
2276 } while ((dst = dst->next));
2277 }
2278
2279 /* Check that the bundle accepts the flow and its components are
2280 * still valid.
2281 */
2282
2283 static int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2284 const struct flowi *fl, int family)
2285 {
2286 struct dst_entry *dst = &first->u.dst;
2287 struct xfrm_dst *last;
2288 u32 mtu;
2289
2290 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2291 (dst->dev && !netif_running(dst->dev)))
2292 return 0;
2293 #ifdef CONFIG_XFRM_SUB_POLICY
2294 if (fl) {
2295 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2296 return 0;
2297 if (first->partner &&
2298 !xfrm_selector_match(first->partner, fl, family))
2299 return 0;
2300 }
2301 #endif
2302
2303 last = NULL;
2304
2305 do {
2306 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2307
2308 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2309 return 0;
2310 if (fl && pol &&
2311 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
2312 return 0;
2313 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2314 return 0;
2315 if (xdst->xfrm_genid != dst->xfrm->genid)
2316 return 0;
2317 if (xdst->num_pols > 0 &&
2318 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
2319 return 0;
2320
2321 mtu = dst_mtu(dst->child);
2322 if (xdst->child_mtu_cached != mtu) {
2323 last = xdst;
2324 xdst->child_mtu_cached = mtu;
2325 }
2326
2327 if (!dst_check(xdst->route, xdst->route_cookie))
2328 return 0;
2329 mtu = dst_mtu(xdst->route);
2330 if (xdst->route_mtu_cached != mtu) {
2331 last = xdst;
2332 xdst->route_mtu_cached = mtu;
2333 }
2334
2335 dst = dst->child;
2336 } while (dst->xfrm);
2337
2338 if (likely(!last))
2339 return 1;
2340
2341 mtu = last->child_mtu_cached;
2342 for (;;) {
2343 dst = &last->u.dst;
2344
2345 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2346 if (mtu > last->route_mtu_cached)
2347 mtu = last->route_mtu_cached;
2348 dst_metric_set(dst, RTAX_MTU, mtu);
2349
2350 if (last == first)
2351 break;
2352
2353 last = (struct xfrm_dst *)last->u.dst.next;
2354 last->child_mtu_cached = mtu;
2355 }
2356
2357 return 1;
2358 }
2359
2360 static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2361 {
2362 return dst_metric_advmss(dst->path);
2363 }
2364
2365 static unsigned int xfrm_default_mtu(const struct dst_entry *dst)
2366 {
2367 return dst_mtu(dst->path);
2368 }
2369
2370 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2371 {
2372 struct net *net;
2373 int err = 0;
2374 if (unlikely(afinfo == NULL))
2375 return -EINVAL;
2376 if (unlikely(afinfo->family >= NPROTO))
2377 return -EAFNOSUPPORT;
2378 write_lock_bh(&xfrm_policy_afinfo_lock);
2379 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2380 err = -ENOBUFS;
2381 else {
2382 struct dst_ops *dst_ops = afinfo->dst_ops;
2383 if (likely(dst_ops->kmem_cachep == NULL))
2384 dst_ops->kmem_cachep = xfrm_dst_cache;
2385 if (likely(dst_ops->check == NULL))
2386 dst_ops->check = xfrm_dst_check;
2387 if (likely(dst_ops->default_advmss == NULL))
2388 dst_ops->default_advmss = xfrm_default_advmss;
2389 if (likely(dst_ops->default_mtu == NULL))
2390 dst_ops->default_mtu = xfrm_default_mtu;
2391 if (likely(dst_ops->negative_advice == NULL))
2392 dst_ops->negative_advice = xfrm_negative_advice;
2393 if (likely(dst_ops->link_failure == NULL))
2394 dst_ops->link_failure = xfrm_link_failure;
2395 if (likely(afinfo->garbage_collect == NULL))
2396 afinfo->garbage_collect = __xfrm_garbage_collect;
2397 xfrm_policy_afinfo[afinfo->family] = afinfo;
2398 }
2399 write_unlock_bh(&xfrm_policy_afinfo_lock);
2400
2401 rtnl_lock();
2402 for_each_net(net) {
2403 struct dst_ops *xfrm_dst_ops;
2404
2405 switch (afinfo->family) {
2406 case AF_INET:
2407 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2408 break;
2409 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2410 case AF_INET6:
2411 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2412 break;
2413 #endif
2414 default:
2415 BUG();
2416 }
2417 *xfrm_dst_ops = *afinfo->dst_ops;
2418 }
2419 rtnl_unlock();
2420
2421 return err;
2422 }
2423 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2424
2425 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2426 {
2427 int err = 0;
2428 if (unlikely(afinfo == NULL))
2429 return -EINVAL;
2430 if (unlikely(afinfo->family >= NPROTO))
2431 return -EAFNOSUPPORT;
2432 write_lock_bh(&xfrm_policy_afinfo_lock);
2433 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2434 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2435 err = -EINVAL;
2436 else {
2437 struct dst_ops *dst_ops = afinfo->dst_ops;
2438 xfrm_policy_afinfo[afinfo->family] = NULL;
2439 dst_ops->kmem_cachep = NULL;
2440 dst_ops->check = NULL;
2441 dst_ops->negative_advice = NULL;
2442 dst_ops->link_failure = NULL;
2443 afinfo->garbage_collect = NULL;
2444 }
2445 }
2446 write_unlock_bh(&xfrm_policy_afinfo_lock);
2447 return err;
2448 }
2449 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2450
2451 static void __net_init xfrm_dst_ops_init(struct net *net)
2452 {
2453 struct xfrm_policy_afinfo *afinfo;
2454
2455 read_lock_bh(&xfrm_policy_afinfo_lock);
2456 afinfo = xfrm_policy_afinfo[AF_INET];
2457 if (afinfo)
2458 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2459 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2460 afinfo = xfrm_policy_afinfo[AF_INET6];
2461 if (afinfo)
2462 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2463 #endif
2464 read_unlock_bh(&xfrm_policy_afinfo_lock);
2465 }
2466
2467 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2468 {
2469 struct xfrm_policy_afinfo *afinfo;
2470 if (unlikely(family >= NPROTO))
2471 return NULL;
2472 read_lock(&xfrm_policy_afinfo_lock);
2473 afinfo = xfrm_policy_afinfo[family];
2474 if (unlikely(!afinfo))
2475 read_unlock(&xfrm_policy_afinfo_lock);
2476 return afinfo;
2477 }
2478
2479 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2480 {
2481 read_unlock(&xfrm_policy_afinfo_lock);
2482 }
2483
2484 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2485 {
2486 struct net_device *dev = ptr;
2487
2488 switch (event) {
2489 case NETDEV_DOWN:
2490 __xfrm_garbage_collect(dev_net(dev));
2491 }
2492 return NOTIFY_DONE;
2493 }
2494
2495 static struct notifier_block xfrm_dev_notifier = {
2496 .notifier_call = xfrm_dev_event,
2497 };
2498
2499 #ifdef CONFIG_XFRM_STATISTICS
2500 static int __net_init xfrm_statistics_init(struct net *net)
2501 {
2502 int rv;
2503
2504 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2505 sizeof(struct linux_xfrm_mib),
2506 __alignof__(struct linux_xfrm_mib)) < 0)
2507 return -ENOMEM;
2508 rv = xfrm_proc_init(net);
2509 if (rv < 0)
2510 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2511 return rv;
2512 }
2513
2514 static void xfrm_statistics_fini(struct net *net)
2515 {
2516 xfrm_proc_fini(net);
2517 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2518 }
2519 #else
2520 static int __net_init xfrm_statistics_init(struct net *net)
2521 {
2522 return 0;
2523 }
2524
2525 static void xfrm_statistics_fini(struct net *net)
2526 {
2527 }
2528 #endif
2529
2530 static int __net_init xfrm_policy_init(struct net *net)
2531 {
2532 unsigned int hmask, sz;
2533 int dir;
2534
2535 if (net_eq(net, &init_net))
2536 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2537 sizeof(struct xfrm_dst),
2538 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2539 NULL);
2540
2541 hmask = 8 - 1;
2542 sz = (hmask+1) * sizeof(struct hlist_head);
2543
2544 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2545 if (!net->xfrm.policy_byidx)
2546 goto out_byidx;
2547 net->xfrm.policy_idx_hmask = hmask;
2548
2549 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2550 struct xfrm_policy_hash *htab;
2551
2552 net->xfrm.policy_count[dir] = 0;
2553 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2554
2555 htab = &net->xfrm.policy_bydst[dir];
2556 htab->table = xfrm_hash_alloc(sz);
2557 if (!htab->table)
2558 goto out_bydst;
2559 htab->hmask = hmask;
2560 }
2561
2562 INIT_LIST_HEAD(&net->xfrm.policy_all);
2563 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2564 if (net_eq(net, &init_net))
2565 register_netdevice_notifier(&xfrm_dev_notifier);
2566 return 0;
2567
2568 out_bydst:
2569 for (dir--; dir >= 0; dir--) {
2570 struct xfrm_policy_hash *htab;
2571
2572 htab = &net->xfrm.policy_bydst[dir];
2573 xfrm_hash_free(htab->table, sz);
2574 }
2575 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2576 out_byidx:
2577 return -ENOMEM;
2578 }
2579
2580 static void xfrm_policy_fini(struct net *net)
2581 {
2582 struct xfrm_audit audit_info;
2583 unsigned int sz;
2584 int dir;
2585
2586 flush_work(&net->xfrm.policy_hash_work);
2587 #ifdef CONFIG_XFRM_SUB_POLICY
2588 audit_info.loginuid = -1;
2589 audit_info.sessionid = -1;
2590 audit_info.secid = 0;
2591 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2592 #endif
2593 audit_info.loginuid = -1;
2594 audit_info.sessionid = -1;
2595 audit_info.secid = 0;
2596 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2597
2598 WARN_ON(!list_empty(&net->xfrm.policy_all));
2599
2600 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2601 struct xfrm_policy_hash *htab;
2602
2603 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2604
2605 htab = &net->xfrm.policy_bydst[dir];
2606 sz = (htab->hmask + 1);
2607 WARN_ON(!hlist_empty(htab->table));
2608 xfrm_hash_free(htab->table, sz);
2609 }
2610
2611 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2612 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2613 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2614 }
2615
2616 static int __net_init xfrm_net_init(struct net *net)
2617 {
2618 int rv;
2619
2620 rv = xfrm_statistics_init(net);
2621 if (rv < 0)
2622 goto out_statistics;
2623 rv = xfrm_state_init(net);
2624 if (rv < 0)
2625 goto out_state;
2626 rv = xfrm_policy_init(net);
2627 if (rv < 0)
2628 goto out_policy;
2629 xfrm_dst_ops_init(net);
2630 rv = xfrm_sysctl_init(net);
2631 if (rv < 0)
2632 goto out_sysctl;
2633 return 0;
2634
2635 out_sysctl:
2636 xfrm_policy_fini(net);
2637 out_policy:
2638 xfrm_state_fini(net);
2639 out_state:
2640 xfrm_statistics_fini(net);
2641 out_statistics:
2642 return rv;
2643 }
2644
2645 static void __net_exit xfrm_net_exit(struct net *net)
2646 {
2647 xfrm_sysctl_fini(net);
2648 xfrm_policy_fini(net);
2649 xfrm_state_fini(net);
2650 xfrm_statistics_fini(net);
2651 }
2652
2653 static struct pernet_operations __net_initdata xfrm_net_ops = {
2654 .init = xfrm_net_init,
2655 .exit = xfrm_net_exit,
2656 };
2657
2658 void __init xfrm_init(void)
2659 {
2660 register_pernet_subsys(&xfrm_net_ops);
2661 xfrm_input_init();
2662 }
2663
2664 #ifdef CONFIG_AUDITSYSCALL
2665 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2666 struct audit_buffer *audit_buf)
2667 {
2668 struct xfrm_sec_ctx *ctx = xp->security;
2669 struct xfrm_selector *sel = &xp->selector;
2670
2671 if (ctx)
2672 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2673 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2674
2675 switch(sel->family) {
2676 case AF_INET:
2677 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2678 if (sel->prefixlen_s != 32)
2679 audit_log_format(audit_buf, " src_prefixlen=%d",
2680 sel->prefixlen_s);
2681 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2682 if (sel->prefixlen_d != 32)
2683 audit_log_format(audit_buf, " dst_prefixlen=%d",
2684 sel->prefixlen_d);
2685 break;
2686 case AF_INET6:
2687 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2688 if (sel->prefixlen_s != 128)
2689 audit_log_format(audit_buf, " src_prefixlen=%d",
2690 sel->prefixlen_s);
2691 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2692 if (sel->prefixlen_d != 128)
2693 audit_log_format(audit_buf, " dst_prefixlen=%d",
2694 sel->prefixlen_d);
2695 break;
2696 }
2697 }
2698
2699 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2700 uid_t auid, u32 sessionid, u32 secid)
2701 {
2702 struct audit_buffer *audit_buf;
2703
2704 audit_buf = xfrm_audit_start("SPD-add");
2705 if (audit_buf == NULL)
2706 return;
2707 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2708 audit_log_format(audit_buf, " res=%u", result);
2709 xfrm_audit_common_policyinfo(xp, audit_buf);
2710 audit_log_end(audit_buf);
2711 }
2712 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2713
2714 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2715 uid_t auid, u32 sessionid, u32 secid)
2716 {
2717 struct audit_buffer *audit_buf;
2718
2719 audit_buf = xfrm_audit_start("SPD-delete");
2720 if (audit_buf == NULL)
2721 return;
2722 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2723 audit_log_format(audit_buf, " res=%u", result);
2724 xfrm_audit_common_policyinfo(xp, audit_buf);
2725 audit_log_end(audit_buf);
2726 }
2727 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2728 #endif
2729
2730 #ifdef CONFIG_XFRM_MIGRATE
2731 static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2732 struct xfrm_selector *sel_tgt)
2733 {
2734 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2735 if (sel_tgt->family == sel_cmp->family &&
2736 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2737 sel_cmp->family) == 0 &&
2738 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2739 sel_cmp->family) == 0 &&
2740 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2741 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2742 return 1;
2743 }
2744 } else {
2745 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2746 return 1;
2747 }
2748 }
2749 return 0;
2750 }
2751
2752 static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2753 u8 dir, u8 type)
2754 {
2755 struct xfrm_policy *pol, *ret = NULL;
2756 struct hlist_node *entry;
2757 struct hlist_head *chain;
2758 u32 priority = ~0U;
2759
2760 read_lock_bh(&xfrm_policy_lock);
2761 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2762 hlist_for_each_entry(pol, entry, chain, bydst) {
2763 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2764 pol->type == type) {
2765 ret = pol;
2766 priority = ret->priority;
2767 break;
2768 }
2769 }
2770 chain = &init_net.xfrm.policy_inexact[dir];
2771 hlist_for_each_entry(pol, entry, chain, bydst) {
2772 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2773 pol->type == type &&
2774 pol->priority < priority) {
2775 ret = pol;
2776 break;
2777 }
2778 }
2779
2780 if (ret)
2781 xfrm_pol_hold(ret);
2782
2783 read_unlock_bh(&xfrm_policy_lock);
2784
2785 return ret;
2786 }
2787
2788 static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2789 {
2790 int match = 0;
2791
2792 if (t->mode == m->mode && t->id.proto == m->proto &&
2793 (m->reqid == 0 || t->reqid == m->reqid)) {
2794 switch (t->mode) {
2795 case XFRM_MODE_TUNNEL:
2796 case XFRM_MODE_BEET:
2797 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2798 m->old_family) == 0 &&
2799 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2800 m->old_family) == 0) {
2801 match = 1;
2802 }
2803 break;
2804 case XFRM_MODE_TRANSPORT:
2805 /* in case of transport mode, template does not store
2806 any IP addresses, hence we just compare mode and
2807 protocol */
2808 match = 1;
2809 break;
2810 default:
2811 break;
2812 }
2813 }
2814 return match;
2815 }
2816
2817 /* update endpoint address(es) of template(s) */
2818 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2819 struct xfrm_migrate *m, int num_migrate)
2820 {
2821 struct xfrm_migrate *mp;
2822 int i, j, n = 0;
2823
2824 write_lock_bh(&pol->lock);
2825 if (unlikely(pol->walk.dead)) {
2826 /* target policy has been deleted */
2827 write_unlock_bh(&pol->lock);
2828 return -ENOENT;
2829 }
2830
2831 for (i = 0; i < pol->xfrm_nr; i++) {
2832 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2833 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2834 continue;
2835 n++;
2836 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2837 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2838 continue;
2839 /* update endpoints */
2840 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2841 sizeof(pol->xfrm_vec[i].id.daddr));
2842 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2843 sizeof(pol->xfrm_vec[i].saddr));
2844 pol->xfrm_vec[i].encap_family = mp->new_family;
2845 /* flush bundles */
2846 atomic_inc(&pol->genid);
2847 }
2848 }
2849
2850 write_unlock_bh(&pol->lock);
2851
2852 if (!n)
2853 return -ENODATA;
2854
2855 return 0;
2856 }
2857
2858 static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2859 {
2860 int i, j;
2861
2862 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2863 return -EINVAL;
2864
2865 for (i = 0; i < num_migrate; i++) {
2866 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2867 m[i].old_family) == 0) &&
2868 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2869 m[i].old_family) == 0))
2870 return -EINVAL;
2871 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2872 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2873 return -EINVAL;
2874
2875 /* check if there is any duplicated entry */
2876 for (j = i + 1; j < num_migrate; j++) {
2877 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2878 sizeof(m[i].old_daddr)) &&
2879 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2880 sizeof(m[i].old_saddr)) &&
2881 m[i].proto == m[j].proto &&
2882 m[i].mode == m[j].mode &&
2883 m[i].reqid == m[j].reqid &&
2884 m[i].old_family == m[j].old_family)
2885 return -EINVAL;
2886 }
2887 }
2888
2889 return 0;
2890 }
2891
2892 int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
2893 struct xfrm_migrate *m, int num_migrate,
2894 struct xfrm_kmaddress *k)
2895 {
2896 int i, err, nx_cur = 0, nx_new = 0;
2897 struct xfrm_policy *pol = NULL;
2898 struct xfrm_state *x, *xc;
2899 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2900 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2901 struct xfrm_migrate *mp;
2902
2903 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2904 goto out;
2905
2906 /* Stage 1 - find policy */
2907 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2908 err = -ENOENT;
2909 goto out;
2910 }
2911
2912 /* Stage 2 - find and update state(s) */
2913 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2914 if ((x = xfrm_migrate_state_find(mp))) {
2915 x_cur[nx_cur] = x;
2916 nx_cur++;
2917 if ((xc = xfrm_state_migrate(x, mp))) {
2918 x_new[nx_new] = xc;
2919 nx_new++;
2920 } else {
2921 err = -ENODATA;
2922 goto restore_state;
2923 }
2924 }
2925 }
2926
2927 /* Stage 3 - update policy */
2928 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2929 goto restore_state;
2930
2931 /* Stage 4 - delete old state(s) */
2932 if (nx_cur) {
2933 xfrm_states_put(x_cur, nx_cur);
2934 xfrm_states_delete(x_cur, nx_cur);
2935 }
2936
2937 /* Stage 5 - announce */
2938 km_migrate(sel, dir, type, m, num_migrate, k);
2939
2940 xfrm_pol_put(pol);
2941
2942 return 0;
2943 out:
2944 return err;
2945
2946 restore_state:
2947 if (pol)
2948 xfrm_pol_put(pol);
2949 if (nx_cur)
2950 xfrm_states_put(x_cur, nx_cur);
2951 if (nx_new)
2952 xfrm_states_delete(x_new, nx_new);
2953
2954 return err;
2955 }
2956 EXPORT_SYMBOL(xfrm_migrate);
2957 #endif