?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
perf kmem: Remove die() calls
[linux-2.6.git] / net / xfrm / xfrm_policy.c
blob5a2aa17e4d3c4a3667783266ddbb8992a3eff6bd
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/flow.h>
30 #include <net/xfrm.h>
31 #include <net/ip.h>
32 #ifdef CONFIG_XFRM_STATISTICS
33 #include <net/snmp.h>
34 #endif
35
36 #include "xfrm_hash.h"
37
38 DEFINE_MUTEX(xfrm_cfg_mutex);
39 EXPORT_SYMBOL(xfrm_cfg_mutex);
40
41 static DEFINE_SPINLOCK(xfrm_policy_sk_bundle_lock);
42 static struct dst_entry *xfrm_policy_sk_bundles;
43 static DEFINE_RWLOCK(xfrm_policy_lock);
44
45 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
46 static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
47
48 static struct kmem_cache *xfrm_dst_cache __read_mostly;
49
50 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
51 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
52 static void xfrm_init_pmtu(struct dst_entry *dst);
53 static int stale_bundle(struct dst_entry *dst);
54 static int xfrm_bundle_ok(struct xfrm_dst *xdst);
55
56
57 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
58 int dir);
59
60 static inline bool
61 __xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
62 {
63 const struct flowi4 *fl4 = &fl->u.ip4;
64
65 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
66 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
67 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
68 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
69 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
70 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
71 }
72
73 static inline bool
74 __xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
75 {
76 const struct flowi6 *fl6 = &fl->u.ip6;
77
78 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
79 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
80 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
81 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
82 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
83 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
84 }
85
86 bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
87 unsigned short family)
88 {
89 switch (family) {
90 case AF_INET:
91 return __xfrm4_selector_match(sel, fl);
92 case AF_INET6:
93 return __xfrm6_selector_match(sel, fl);
94 }
95 return false;
96 }
97
98 static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
99 const xfrm_address_t *saddr,
100 const xfrm_address_t *daddr,
101 int family)
102 {
103 struct xfrm_policy_afinfo *afinfo;
104 struct dst_entry *dst;
105
106 afinfo = xfrm_policy_get_afinfo(family);
107 if (unlikely(afinfo == NULL))
108 return ERR_PTR(-EAFNOSUPPORT);
109
110 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
111
112 xfrm_policy_put_afinfo(afinfo);
113
114 return dst;
115 }
116
117 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
118 xfrm_address_t *prev_saddr,
119 xfrm_address_t *prev_daddr,
120 int family)
121 {
122 struct net *net = xs_net(x);
123 xfrm_address_t *saddr = &x->props.saddr;
124 xfrm_address_t *daddr = &x->id.daddr;
125 struct dst_entry *dst;
126
127 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
128 saddr = x->coaddr;
129 daddr = prev_daddr;
130 }
131 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
132 saddr = prev_saddr;
133 daddr = x->coaddr;
134 }
135
136 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
137
138 if (!IS_ERR(dst)) {
139 if (prev_saddr != saddr)
140 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
141 if (prev_daddr != daddr)
142 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
143 }
144
145 return dst;
146 }
147
148 static inline unsigned long make_jiffies(long secs)
149 {
150 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
151 return MAX_SCHEDULE_TIMEOUT-1;
152 else
153 return secs*HZ;
154 }
155
156 static void xfrm_policy_timer(unsigned long data)
157 {
158 struct xfrm_policy *xp = (struct xfrm_policy*)data;
159 unsigned long now = get_seconds();
160 long next = LONG_MAX;
161 int warn = 0;
162 int dir;
163
164 read_lock(&xp->lock);
165
166 if (unlikely(xp->walk.dead))
167 goto out;
168
169 dir = xfrm_policy_id2dir(xp->index);
170
171 if (xp->lft.hard_add_expires_seconds) {
172 long tmo = xp->lft.hard_add_expires_seconds +
173 xp->curlft.add_time - now;
174 if (tmo <= 0)
175 goto expired;
176 if (tmo < next)
177 next = tmo;
178 }
179 if (xp->lft.hard_use_expires_seconds) {
180 long tmo = xp->lft.hard_use_expires_seconds +
181 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
182 if (tmo <= 0)
183 goto expired;
184 if (tmo < next)
185 next = tmo;
186 }
187 if (xp->lft.soft_add_expires_seconds) {
188 long tmo = xp->lft.soft_add_expires_seconds +
189 xp->curlft.add_time - now;
190 if (tmo <= 0) {
191 warn = 1;
192 tmo = XFRM_KM_TIMEOUT;
193 }
194 if (tmo < next)
195 next = tmo;
196 }
197 if (xp->lft.soft_use_expires_seconds) {
198 long tmo = xp->lft.soft_use_expires_seconds +
199 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
200 if (tmo <= 0) {
201 warn = 1;
202 tmo = XFRM_KM_TIMEOUT;
203 }
204 if (tmo < next)
205 next = tmo;
206 }
207
208 if (warn)
209 km_policy_expired(xp, dir, 0, 0);
210 if (next != LONG_MAX &&
211 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
212 xfrm_pol_hold(xp);
213
214 out:
215 read_unlock(&xp->lock);
216 xfrm_pol_put(xp);
217 return;
218
219 expired:
220 read_unlock(&xp->lock);
221 if (!xfrm_policy_delete(xp, dir))
222 km_policy_expired(xp, dir, 1, 0);
223 xfrm_pol_put(xp);
224 }
225
226 static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
227 {
228 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
229
230 if (unlikely(pol->walk.dead))
231 flo = NULL;
232 else
233 xfrm_pol_hold(pol);
234
235 return flo;
236 }
237
238 static int xfrm_policy_flo_check(struct flow_cache_object *flo)
239 {
240 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
241
242 return !pol->walk.dead;
243 }
244
245 static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
246 {
247 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
248 }
249
250 static const struct flow_cache_ops xfrm_policy_fc_ops = {
251 .get = xfrm_policy_flo_get,
252 .check = xfrm_policy_flo_check,
253 .delete = xfrm_policy_flo_delete,
254 };
255
256 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
257 * SPD calls.
258 */
259
260 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
261 {
262 struct xfrm_policy *policy;
263
264 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
265
266 if (policy) {
267 write_pnet(&policy->xp_net, net);
268 INIT_LIST_HEAD(&policy->walk.all);
269 INIT_HLIST_NODE(&policy->bydst);
270 INIT_HLIST_NODE(&policy->byidx);
271 rwlock_init(&policy->lock);
272 atomic_set(&policy->refcnt, 1);
273 setup_timer(&policy->timer, xfrm_policy_timer,
274 (unsigned long)policy);
275 policy->flo.ops = &xfrm_policy_fc_ops;
276 }
277 return policy;
278 }
279 EXPORT_SYMBOL(xfrm_policy_alloc);
280
281 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
282
283 void xfrm_policy_destroy(struct xfrm_policy *policy)
284 {
285 BUG_ON(!policy->walk.dead);
286
287 if (del_timer(&policy->timer))
288 BUG();
289
290 security_xfrm_policy_free(policy->security);
291 kfree(policy);
292 }
293 EXPORT_SYMBOL(xfrm_policy_destroy);
294
295 /* Rule must be locked. Release descentant resources, announce
296 * entry dead. The rule must be unlinked from lists to the moment.
297 */
298
299 static void xfrm_policy_kill(struct xfrm_policy *policy)
300 {
301 policy->walk.dead = 1;
302
303 atomic_inc(&policy->genid);
304
305 if (del_timer(&policy->timer))
306 xfrm_pol_put(policy);
307
308 xfrm_pol_put(policy);
309 }
310
311 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
312
313 static inline unsigned int idx_hash(struct net *net, u32 index)
314 {
315 return __idx_hash(index, net->xfrm.policy_idx_hmask);
316 }
317
318 static struct hlist_head *policy_hash_bysel(struct net *net,
319 const struct xfrm_selector *sel,
320 unsigned short family, int dir)
321 {
322 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
323 unsigned int hash = __sel_hash(sel, family, hmask);
324
325 return (hash == hmask + 1 ?
326 &net->xfrm.policy_inexact[dir] :
327 net->xfrm.policy_bydst[dir].table + hash);
328 }
329
330 static struct hlist_head *policy_hash_direct(struct net *net,
331 const xfrm_address_t *daddr,
332 const xfrm_address_t *saddr,
333 unsigned short family, int dir)
334 {
335 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
336 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
337
338 return net->xfrm.policy_bydst[dir].table + hash;
339 }
340
341 static void xfrm_dst_hash_transfer(struct hlist_head *list,
342 struct hlist_head *ndsttable,
343 unsigned int nhashmask)
344 {
345 struct hlist_node *entry, *tmp, *entry0 = NULL;
346 struct xfrm_policy *pol;
347 unsigned int h0 = 0;
348
349 redo:
350 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
351 unsigned int h;
352
353 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
354 pol->family, nhashmask);
355 if (!entry0) {
356 hlist_del(entry);
357 hlist_add_head(&pol->bydst, ndsttable+h);
358 h0 = h;
359 } else {
360 if (h != h0)
361 continue;
362 hlist_del(entry);
363 hlist_add_after(entry0, &pol->bydst);
364 }
365 entry0 = entry;
366 }
367 if (!hlist_empty(list)) {
368 entry0 = NULL;
369 goto redo;
370 }
371 }
372
373 static void xfrm_idx_hash_transfer(struct hlist_head *list,
374 struct hlist_head *nidxtable,
375 unsigned int nhashmask)
376 {
377 struct hlist_node *entry, *tmp;
378 struct xfrm_policy *pol;
379
380 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
381 unsigned int h;
382
383 h = __idx_hash(pol->index, nhashmask);
384 hlist_add_head(&pol->byidx, nidxtable+h);
385 }
386 }
387
388 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
389 {
390 return ((old_hmask + 1) << 1) - 1;
391 }
392
393 static void xfrm_bydst_resize(struct net *net, int dir)
394 {
395 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
396 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
397 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
398 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
399 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
400 int i;
401
402 if (!ndst)
403 return;
404
405 write_lock_bh(&xfrm_policy_lock);
406
407 for (i = hmask; i >= 0; i--)
408 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
409
410 net->xfrm.policy_bydst[dir].table = ndst;
411 net->xfrm.policy_bydst[dir].hmask = nhashmask;
412
413 write_unlock_bh(&xfrm_policy_lock);
414
415 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
416 }
417
418 static void xfrm_byidx_resize(struct net *net, int total)
419 {
420 unsigned int hmask = net->xfrm.policy_idx_hmask;
421 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
422 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
423 struct hlist_head *oidx = net->xfrm.policy_byidx;
424 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
425 int i;
426
427 if (!nidx)
428 return;
429
430 write_lock_bh(&xfrm_policy_lock);
431
432 for (i = hmask; i >= 0; i--)
433 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
434
435 net->xfrm.policy_byidx = nidx;
436 net->xfrm.policy_idx_hmask = nhashmask;
437
438 write_unlock_bh(&xfrm_policy_lock);
439
440 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
441 }
442
443 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
444 {
445 unsigned int cnt = net->xfrm.policy_count[dir];
446 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
447
448 if (total)
449 *total += cnt;
450
451 if ((hmask + 1) < xfrm_policy_hashmax &&
452 cnt > hmask)
453 return 1;
454
455 return 0;
456 }
457
458 static inline int xfrm_byidx_should_resize(struct net *net, int total)
459 {
460 unsigned int hmask = net->xfrm.policy_idx_hmask;
461
462 if ((hmask + 1) < xfrm_policy_hashmax &&
463 total > hmask)
464 return 1;
465
466 return 0;
467 }
468
469 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
470 {
471 read_lock_bh(&xfrm_policy_lock);
472 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
473 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
474 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
475 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
476 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
477 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
478 si->spdhcnt = net->xfrm.policy_idx_hmask;
479 si->spdhmcnt = xfrm_policy_hashmax;
480 read_unlock_bh(&xfrm_policy_lock);
481 }
482 EXPORT_SYMBOL(xfrm_spd_getinfo);
483
484 static DEFINE_MUTEX(hash_resize_mutex);
485 static void xfrm_hash_resize(struct work_struct *work)
486 {
487 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
488 int dir, total;
489
490 mutex_lock(&hash_resize_mutex);
491
492 total = 0;
493 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
494 if (xfrm_bydst_should_resize(net, dir, &total))
495 xfrm_bydst_resize(net, dir);
496 }
497 if (xfrm_byidx_should_resize(net, total))
498 xfrm_byidx_resize(net, total);
499
500 mutex_unlock(&hash_resize_mutex);
501 }
502
503 /* Generate new index... KAME seems to generate them ordered by cost
504 * of an absolute inpredictability of ordering of rules. This will not pass. */
505 static u32 xfrm_gen_index(struct net *net, int dir)
506 {
507 static u32 idx_generator;
508
509 for (;;) {
510 struct hlist_node *entry;
511 struct hlist_head *list;
512 struct xfrm_policy *p;
513 u32 idx;
514 int found;
515
516 idx = (idx_generator | dir);
517 idx_generator += 8;
518 if (idx == 0)
519 idx = 8;
520 list = net->xfrm.policy_byidx + idx_hash(net, idx);
521 found = 0;
522 hlist_for_each_entry(p, entry, list, byidx) {
523 if (p->index == idx) {
524 found = 1;
525 break;
526 }
527 }
528 if (!found)
529 return idx;
530 }
531 }
532
533 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
534 {
535 u32 *p1 = (u32 *) s1;
536 u32 *p2 = (u32 *) s2;
537 int len = sizeof(struct xfrm_selector) / sizeof(u32);
538 int i;
539
540 for (i = 0; i < len; i++) {
541 if (p1[i] != p2[i])
542 return 1;
543 }
544
545 return 0;
546 }
547
548 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
549 {
550 struct net *net = xp_net(policy);
551 struct xfrm_policy *pol;
552 struct xfrm_policy *delpol;
553 struct hlist_head *chain;
554 struct hlist_node *entry, *newpos;
555 u32 mark = policy->mark.v & policy->mark.m;
556
557 write_lock_bh(&xfrm_policy_lock);
558 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
559 delpol = NULL;
560 newpos = NULL;
561 hlist_for_each_entry(pol, entry, chain, bydst) {
562 if (pol->type == policy->type &&
563 !selector_cmp(&pol->selector, &policy->selector) &&
564 (mark & pol->mark.m) == pol->mark.v &&
565 xfrm_sec_ctx_match(pol->security, policy->security) &&
566 !WARN_ON(delpol)) {
567 if (excl) {
568 write_unlock_bh(&xfrm_policy_lock);
569 return -EEXIST;
570 }
571 delpol = pol;
572 if (policy->priority > pol->priority)
573 continue;
574 } else if (policy->priority >= pol->priority) {
575 newpos = &pol->bydst;
576 continue;
577 }
578 if (delpol)
579 break;
580 }
581 if (newpos)
582 hlist_add_after(newpos, &policy->bydst);
583 else
584 hlist_add_head(&policy->bydst, chain);
585 xfrm_pol_hold(policy);
586 net->xfrm.policy_count[dir]++;
587 atomic_inc(&flow_cache_genid);
588 if (delpol)
589 __xfrm_policy_unlink(delpol, dir);
590 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
591 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
592 policy->curlft.add_time = get_seconds();
593 policy->curlft.use_time = 0;
594 if (!mod_timer(&policy->timer, jiffies + HZ))
595 xfrm_pol_hold(policy);
596 list_add(&policy->walk.all, &net->xfrm.policy_all);
597 write_unlock_bh(&xfrm_policy_lock);
598
599 if (delpol)
600 xfrm_policy_kill(delpol);
601 else if (xfrm_bydst_should_resize(net, dir, NULL))
602 schedule_work(&net->xfrm.policy_hash_work);
603
604 return 0;
605 }
606 EXPORT_SYMBOL(xfrm_policy_insert);
607
608 struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
609 int dir, struct xfrm_selector *sel,
610 struct xfrm_sec_ctx *ctx, int delete,
611 int *err)
612 {
613 struct xfrm_policy *pol, *ret;
614 struct hlist_head *chain;
615 struct hlist_node *entry;
616
617 *err = 0;
618 write_lock_bh(&xfrm_policy_lock);
619 chain = policy_hash_bysel(net, sel, sel->family, dir);
620 ret = NULL;
621 hlist_for_each_entry(pol, entry, chain, bydst) {
622 if (pol->type == type &&
623 (mark & pol->mark.m) == pol->mark.v &&
624 !selector_cmp(sel, &pol->selector) &&
625 xfrm_sec_ctx_match(ctx, pol->security)) {
626 xfrm_pol_hold(pol);
627 if (delete) {
628 *err = security_xfrm_policy_delete(
629 pol->security);
630 if (*err) {
631 write_unlock_bh(&xfrm_policy_lock);
632 return pol;
633 }
634 __xfrm_policy_unlink(pol, dir);
635 }
636 ret = pol;
637 break;
638 }
639 }
640 write_unlock_bh(&xfrm_policy_lock);
641
642 if (ret && delete)
643 xfrm_policy_kill(ret);
644 return ret;
645 }
646 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
647
648 struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
649 int dir, u32 id, int delete, int *err)
650 {
651 struct xfrm_policy *pol, *ret;
652 struct hlist_head *chain;
653 struct hlist_node *entry;
654
655 *err = -ENOENT;
656 if (xfrm_policy_id2dir(id) != dir)
657 return NULL;
658
659 *err = 0;
660 write_lock_bh(&xfrm_policy_lock);
661 chain = net->xfrm.policy_byidx + idx_hash(net, id);
662 ret = NULL;
663 hlist_for_each_entry(pol, entry, chain, byidx) {
664 if (pol->type == type && pol->index == id &&
665 (mark & pol->mark.m) == pol->mark.v) {
666 xfrm_pol_hold(pol);
667 if (delete) {
668 *err = security_xfrm_policy_delete(
669 pol->security);
670 if (*err) {
671 write_unlock_bh(&xfrm_policy_lock);
672 return pol;
673 }
674 __xfrm_policy_unlink(pol, dir);
675 }
676 ret = pol;
677 break;
678 }
679 }
680 write_unlock_bh(&xfrm_policy_lock);
681
682 if (ret && delete)
683 xfrm_policy_kill(ret);
684 return ret;
685 }
686 EXPORT_SYMBOL(xfrm_policy_byid);
687
688 #ifdef CONFIG_SECURITY_NETWORK_XFRM
689 static inline int
690 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
691 {
692 int dir, err = 0;
693
694 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
695 struct xfrm_policy *pol;
696 struct hlist_node *entry;
697 int i;
698
699 hlist_for_each_entry(pol, entry,
700 &net->xfrm.policy_inexact[dir], bydst) {
701 if (pol->type != type)
702 continue;
703 err = security_xfrm_policy_delete(pol->security);
704 if (err) {
705 xfrm_audit_policy_delete(pol, 0,
706 audit_info->loginuid,
707 audit_info->sessionid,
708 audit_info->secid);
709 return err;
710 }
711 }
712 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
713 hlist_for_each_entry(pol, entry,
714 net->xfrm.policy_bydst[dir].table + i,
715 bydst) {
716 if (pol->type != type)
717 continue;
718 err = security_xfrm_policy_delete(
719 pol->security);
720 if (err) {
721 xfrm_audit_policy_delete(pol, 0,
722 audit_info->loginuid,
723 audit_info->sessionid,
724 audit_info->secid);
725 return err;
726 }
727 }
728 }
729 }
730 return err;
731 }
732 #else
733 static inline int
734 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
735 {
736 return 0;
737 }
738 #endif
739
740 int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
741 {
742 int dir, err = 0, cnt = 0;
743
744 write_lock_bh(&xfrm_policy_lock);
745
746 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
747 if (err)
748 goto out;
749
750 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
751 struct xfrm_policy *pol;
752 struct hlist_node *entry;
753 int i;
754
755 again1:
756 hlist_for_each_entry(pol, entry,
757 &net->xfrm.policy_inexact[dir], bydst) {
758 if (pol->type != type)
759 continue;
760 __xfrm_policy_unlink(pol, dir);
761 write_unlock_bh(&xfrm_policy_lock);
762 cnt++;
763
764 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
765 audit_info->sessionid,
766 audit_info->secid);
767
768 xfrm_policy_kill(pol);
769
770 write_lock_bh(&xfrm_policy_lock);
771 goto again1;
772 }
773
774 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
775 again2:
776 hlist_for_each_entry(pol, entry,
777 net->xfrm.policy_bydst[dir].table + i,
778 bydst) {
779 if (pol->type != type)
780 continue;
781 __xfrm_policy_unlink(pol, dir);
782 write_unlock_bh(&xfrm_policy_lock);
783 cnt++;
784
785 xfrm_audit_policy_delete(pol, 1,
786 audit_info->loginuid,
787 audit_info->sessionid,
788 audit_info->secid);
789 xfrm_policy_kill(pol);
790
791 write_lock_bh(&xfrm_policy_lock);
792 goto again2;
793 }
794 }
795
796 }
797 if (!cnt)
798 err = -ESRCH;
799 out:
800 write_unlock_bh(&xfrm_policy_lock);
801 return err;
802 }
803 EXPORT_SYMBOL(xfrm_policy_flush);
804
805 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
806 int (*func)(struct xfrm_policy *, int, int, void*),
807 void *data)
808 {
809 struct xfrm_policy *pol;
810 struct xfrm_policy_walk_entry *x;
811 int error = 0;
812
813 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
814 walk->type != XFRM_POLICY_TYPE_ANY)
815 return -EINVAL;
816
817 if (list_empty(&walk->walk.all) && walk->seq != 0)
818 return 0;
819
820 write_lock_bh(&xfrm_policy_lock);
821 if (list_empty(&walk->walk.all))
822 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
823 else
824 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
825 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
826 if (x->dead)
827 continue;
828 pol = container_of(x, struct xfrm_policy, walk);
829 if (walk->type != XFRM_POLICY_TYPE_ANY &&
830 walk->type != pol->type)
831 continue;
832 error = func(pol, xfrm_policy_id2dir(pol->index),
833 walk->seq, data);
834 if (error) {
835 list_move_tail(&walk->walk.all, &x->all);
836 goto out;
837 }
838 walk->seq++;
839 }
840 if (walk->seq == 0) {
841 error = -ENOENT;
842 goto out;
843 }
844 list_del_init(&walk->walk.all);
845 out:
846 write_unlock_bh(&xfrm_policy_lock);
847 return error;
848 }
849 EXPORT_SYMBOL(xfrm_policy_walk);
850
851 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
852 {
853 INIT_LIST_HEAD(&walk->walk.all);
854 walk->walk.dead = 1;
855 walk->type = type;
856 walk->seq = 0;
857 }
858 EXPORT_SYMBOL(xfrm_policy_walk_init);
859
860 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
861 {
862 if (list_empty(&walk->walk.all))
863 return;
864
865 write_lock_bh(&xfrm_policy_lock);
866 list_del(&walk->walk.all);
867 write_unlock_bh(&xfrm_policy_lock);
868 }
869 EXPORT_SYMBOL(xfrm_policy_walk_done);
870
871 /*
872 * Find policy to apply to this flow.
873 *
874 * Returns 0 if policy found, else an -errno.
875 */
876 static int xfrm_policy_match(const struct xfrm_policy *pol,
877 const struct flowi *fl,
878 u8 type, u16 family, int dir)
879 {
880 const struct xfrm_selector *sel = &pol->selector;
881 int ret = -ESRCH;
882 bool match;
883
884 if (pol->family != family ||
885 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
886 pol->type != type)
887 return ret;
888
889 match = xfrm_selector_match(sel, fl, family);
890 if (match)
891 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
892 dir);
893
894 return ret;
895 }
896
897 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
898 const struct flowi *fl,
899 u16 family, u8 dir)
900 {
901 int err;
902 struct xfrm_policy *pol, *ret;
903 const xfrm_address_t *daddr, *saddr;
904 struct hlist_node *entry;
905 struct hlist_head *chain;
906 u32 priority = ~0U;
907
908 daddr = xfrm_flowi_daddr(fl, family);
909 saddr = xfrm_flowi_saddr(fl, family);
910 if (unlikely(!daddr || !saddr))
911 return NULL;
912
913 read_lock_bh(&xfrm_policy_lock);
914 chain = policy_hash_direct(net, daddr, saddr, family, dir);
915 ret = NULL;
916 hlist_for_each_entry(pol, entry, chain, bydst) {
917 err = xfrm_policy_match(pol, fl, type, family, dir);
918 if (err) {
919 if (err == -ESRCH)
920 continue;
921 else {
922 ret = ERR_PTR(err);
923 goto fail;
924 }
925 } else {
926 ret = pol;
927 priority = ret->priority;
928 break;
929 }
930 }
931 chain = &net->xfrm.policy_inexact[dir];
932 hlist_for_each_entry(pol, entry, chain, bydst) {
933 err = xfrm_policy_match(pol, fl, type, family, dir);
934 if (err) {
935 if (err == -ESRCH)
936 continue;
937 else {
938 ret = ERR_PTR(err);
939 goto fail;
940 }
941 } else if (pol->priority < priority) {
942 ret = pol;
943 break;
944 }
945 }
946 if (ret)
947 xfrm_pol_hold(ret);
948 fail:
949 read_unlock_bh(&xfrm_policy_lock);
950
951 return ret;
952 }
953
954 static struct xfrm_policy *
955 __xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
956 {
957 #ifdef CONFIG_XFRM_SUB_POLICY
958 struct xfrm_policy *pol;
959
960 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
961 if (pol != NULL)
962 return pol;
963 #endif
964 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
965 }
966
967 static struct flow_cache_object *
968 xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
969 u8 dir, struct flow_cache_object *old_obj, void *ctx)
970 {
971 struct xfrm_policy *pol;
972
973 if (old_obj)
974 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
975
976 pol = __xfrm_policy_lookup(net, fl, family, dir);
977 if (IS_ERR_OR_NULL(pol))
978 return ERR_CAST(pol);
979
980 /* Resolver returns two references:
981 * one for cache and one for caller of flow_cache_lookup() */
982 xfrm_pol_hold(pol);
983
984 return &pol->flo;
985 }
986
987 static inline int policy_to_flow_dir(int dir)
988 {
989 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
990 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
991 XFRM_POLICY_FWD == FLOW_DIR_FWD)
992 return dir;
993 switch (dir) {
994 default:
995 case XFRM_POLICY_IN:
996 return FLOW_DIR_IN;
997 case XFRM_POLICY_OUT:
998 return FLOW_DIR_OUT;
999 case XFRM_POLICY_FWD:
1000 return FLOW_DIR_FWD;
1001 }
1002 }
1003
1004 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1005 const struct flowi *fl)
1006 {
1007 struct xfrm_policy *pol;
1008
1009 read_lock_bh(&xfrm_policy_lock);
1010 if ((pol = sk->sk_policy[dir]) != NULL) {
1011 bool match = xfrm_selector_match(&pol->selector, fl,
1012 sk->sk_family);
1013 int err = 0;
1014
1015 if (match) {
1016 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1017 pol = NULL;
1018 goto out;
1019 }
1020 err = security_xfrm_policy_lookup(pol->security,
1021 fl->flowi_secid,
1022 policy_to_flow_dir(dir));
1023 if (!err)
1024 xfrm_pol_hold(pol);
1025 else if (err == -ESRCH)
1026 pol = NULL;
1027 else
1028 pol = ERR_PTR(err);
1029 } else
1030 pol = NULL;
1031 }
1032 out:
1033 read_unlock_bh(&xfrm_policy_lock);
1034 return pol;
1035 }
1036
1037 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1038 {
1039 struct net *net = xp_net(pol);
1040 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1041 pol->family, dir);
1042
1043 list_add(&pol->walk.all, &net->xfrm.policy_all);
1044 hlist_add_head(&pol->bydst, chain);
1045 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1046 net->xfrm.policy_count[dir]++;
1047 xfrm_pol_hold(pol);
1048
1049 if (xfrm_bydst_should_resize(net, dir, NULL))
1050 schedule_work(&net->xfrm.policy_hash_work);
1051 }
1052
1053 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1054 int dir)
1055 {
1056 struct net *net = xp_net(pol);
1057
1058 if (hlist_unhashed(&pol->bydst))
1059 return NULL;
1060
1061 hlist_del(&pol->bydst);
1062 hlist_del(&pol->byidx);
1063 list_del(&pol->walk.all);
1064 net->xfrm.policy_count[dir]--;
1065
1066 return pol;
1067 }
1068
1069 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1070 {
1071 write_lock_bh(&xfrm_policy_lock);
1072 pol = __xfrm_policy_unlink(pol, dir);
1073 write_unlock_bh(&xfrm_policy_lock);
1074 if (pol) {
1075 xfrm_policy_kill(pol);
1076 return 0;
1077 }
1078 return -ENOENT;
1079 }
1080 EXPORT_SYMBOL(xfrm_policy_delete);
1081
1082 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1083 {
1084 struct net *net = xp_net(pol);
1085 struct xfrm_policy *old_pol;
1086
1087 #ifdef CONFIG_XFRM_SUB_POLICY
1088 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1089 return -EINVAL;
1090 #endif
1091
1092 write_lock_bh(&xfrm_policy_lock);
1093 old_pol = sk->sk_policy[dir];
1094 sk->sk_policy[dir] = pol;
1095 if (pol) {
1096 pol->curlft.add_time = get_seconds();
1097 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1098 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1099 }
1100 if (old_pol)
1101 /* Unlinking succeeds always. This is the only function
1102 * allowed to delete or replace socket policy.
1103 */
1104 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1105 write_unlock_bh(&xfrm_policy_lock);
1106
1107 if (old_pol) {
1108 xfrm_policy_kill(old_pol);
1109 }
1110 return 0;
1111 }
1112
1113 static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1114 {
1115 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1116
1117 if (newp) {
1118 newp->selector = old->selector;
1119 if (security_xfrm_policy_clone(old->security,
1120 &newp->security)) {
1121 kfree(newp);
1122 return NULL; /* ENOMEM */
1123 }
1124 newp->lft = old->lft;
1125 newp->curlft = old->curlft;
1126 newp->mark = old->mark;
1127 newp->action = old->action;
1128 newp->flags = old->flags;
1129 newp->xfrm_nr = old->xfrm_nr;
1130 newp->index = old->index;
1131 newp->type = old->type;
1132 memcpy(newp->xfrm_vec, old->xfrm_vec,
1133 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1134 write_lock_bh(&xfrm_policy_lock);
1135 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1136 write_unlock_bh(&xfrm_policy_lock);
1137 xfrm_pol_put(newp);
1138 }
1139 return newp;
1140 }
1141
1142 int __xfrm_sk_clone_policy(struct sock *sk)
1143 {
1144 struct xfrm_policy *p0 = sk->sk_policy[0],
1145 *p1 = sk->sk_policy[1];
1146
1147 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1148 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1149 return -ENOMEM;
1150 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1151 return -ENOMEM;
1152 return 0;
1153 }
1154
1155 static int
1156 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1157 unsigned short family)
1158 {
1159 int err;
1160 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1161
1162 if (unlikely(afinfo == NULL))
1163 return -EINVAL;
1164 err = afinfo->get_saddr(net, local, remote);
1165 xfrm_policy_put_afinfo(afinfo);
1166 return err;
1167 }
1168
1169 /* Resolve list of templates for the flow, given policy. */
1170
1171 static int
1172 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1173 struct xfrm_state **xfrm, unsigned short family)
1174 {
1175 struct net *net = xp_net(policy);
1176 int nx;
1177 int i, error;
1178 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1179 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1180 xfrm_address_t tmp;
1181
1182 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1183 struct xfrm_state *x;
1184 xfrm_address_t *remote = daddr;
1185 xfrm_address_t *local = saddr;
1186 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1187
1188 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1189 tmpl->mode == XFRM_MODE_BEET) {
1190 remote = &tmpl->id.daddr;
1191 local = &tmpl->saddr;
1192 if (xfrm_addr_any(local, tmpl->encap_family)) {
1193 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
1194 if (error)
1195 goto fail;
1196 local = &tmp;
1197 }
1198 }
1199
1200 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1201
1202 if (x && x->km.state == XFRM_STATE_VALID) {
1203 xfrm[nx++] = x;
1204 daddr = remote;
1205 saddr = local;
1206 continue;
1207 }
1208 if (x) {
1209 error = (x->km.state == XFRM_STATE_ERROR ?
1210 -EINVAL : -EAGAIN);
1211 xfrm_state_put(x);
1212 }
1213 else if (error == -ESRCH)
1214 error = -EAGAIN;
1215
1216 if (!tmpl->optional)
1217 goto fail;
1218 }
1219 return nx;
1220
1221 fail:
1222 for (nx--; nx>=0; nx--)
1223 xfrm_state_put(xfrm[nx]);
1224 return error;
1225 }
1226
1227 static int
1228 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1229 struct xfrm_state **xfrm, unsigned short family)
1230 {
1231 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1232 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1233 int cnx = 0;
1234 int error;
1235 int ret;
1236 int i;
1237
1238 for (i = 0; i < npols; i++) {
1239 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1240 error = -ENOBUFS;
1241 goto fail;
1242 }
1243
1244 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1245 if (ret < 0) {
1246 error = ret;
1247 goto fail;
1248 } else
1249 cnx += ret;
1250 }
1251
1252 /* found states are sorted for outbound processing */
1253 if (npols > 1)
1254 xfrm_state_sort(xfrm, tpp, cnx, family);
1255
1256 return cnx;
1257
1258 fail:
1259 for (cnx--; cnx>=0; cnx--)
1260 xfrm_state_put(tpp[cnx]);
1261 return error;
1262
1263 }
1264
1265 /* Check that the bundle accepts the flow and its components are
1266 * still valid.
1267 */
1268
1269 static inline int xfrm_get_tos(const struct flowi *fl, int family)
1270 {
1271 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1272 int tos;
1273
1274 if (!afinfo)
1275 return -EINVAL;
1276
1277 tos = afinfo->get_tos(fl);
1278
1279 xfrm_policy_put_afinfo(afinfo);
1280
1281 return tos;
1282 }
1283
1284 static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1285 {
1286 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1287 struct dst_entry *dst = &xdst->u.dst;
1288
1289 if (xdst->route == NULL) {
1290 /* Dummy bundle - if it has xfrms we were not
1291 * able to build bundle as template resolution failed.
1292 * It means we need to try again resolving. */
1293 if (xdst->num_xfrms > 0)
1294 return NULL;
1295 } else {
1296 /* Real bundle */
1297 if (stale_bundle(dst))
1298 return NULL;
1299 }
1300
1301 dst_hold(dst);
1302 return flo;
1303 }
1304
1305 static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1306 {
1307 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1308 struct dst_entry *dst = &xdst->u.dst;
1309
1310 if (!xdst->route)
1311 return 0;
1312 if (stale_bundle(dst))
1313 return 0;
1314
1315 return 1;
1316 }
1317
1318 static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1319 {
1320 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1321 struct dst_entry *dst = &xdst->u.dst;
1322
1323 dst_free(dst);
1324 }
1325
1326 static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1327 .get = xfrm_bundle_flo_get,
1328 .check = xfrm_bundle_flo_check,
1329 .delete = xfrm_bundle_flo_delete,
1330 };
1331
1332 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1333 {
1334 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1335 struct dst_ops *dst_ops;
1336 struct xfrm_dst *xdst;
1337
1338 if (!afinfo)
1339 return ERR_PTR(-EINVAL);
1340
1341 switch (family) {
1342 case AF_INET:
1343 dst_ops = &net->xfrm.xfrm4_dst_ops;
1344 break;
1345 #if IS_ENABLED(CONFIG_IPV6)
1346 case AF_INET6:
1347 dst_ops = &net->xfrm.xfrm6_dst_ops;
1348 break;
1349 #endif
1350 default:
1351 BUG();
1352 }
1353 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
1354
1355 if (likely(xdst)) {
1356 struct dst_entry *dst = &xdst->u.dst;
1357
1358 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
1359 xdst->flo.ops = &xfrm_bundle_fc_ops;
1360 if (afinfo->init_dst)
1361 afinfo->init_dst(net, xdst);
1362 } else
1363 xdst = ERR_PTR(-ENOBUFS);
1364
1365 xfrm_policy_put_afinfo(afinfo);
1366
1367 return xdst;
1368 }
1369
1370 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1371 int nfheader_len)
1372 {
1373 struct xfrm_policy_afinfo *afinfo =
1374 xfrm_policy_get_afinfo(dst->ops->family);
1375 int err;
1376
1377 if (!afinfo)
1378 return -EINVAL;
1379
1380 err = afinfo->init_path(path, dst, nfheader_len);
1381
1382 xfrm_policy_put_afinfo(afinfo);
1383
1384 return err;
1385 }
1386
1387 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
1388 const struct flowi *fl)
1389 {
1390 struct xfrm_policy_afinfo *afinfo =
1391 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1392 int err;
1393
1394 if (!afinfo)
1395 return -EINVAL;
1396
1397 err = afinfo->fill_dst(xdst, dev, fl);
1398
1399 xfrm_policy_put_afinfo(afinfo);
1400
1401 return err;
1402 }
1403
1404
1405 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1406 * all the metrics... Shortly, bundle a bundle.
1407 */
1408
1409 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1410 struct xfrm_state **xfrm, int nx,
1411 const struct flowi *fl,
1412 struct dst_entry *dst)
1413 {
1414 struct net *net = xp_net(policy);
1415 unsigned long now = jiffies;
1416 struct net_device *dev;
1417 struct xfrm_mode *inner_mode;
1418 struct dst_entry *dst_prev = NULL;
1419 struct dst_entry *dst0 = NULL;
1420 int i = 0;
1421 int err;
1422 int header_len = 0;
1423 int nfheader_len = 0;
1424 int trailer_len = 0;
1425 int tos;
1426 int family = policy->selector.family;
1427 xfrm_address_t saddr, daddr;
1428
1429 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1430
1431 tos = xfrm_get_tos(fl, family);
1432 err = tos;
1433 if (tos < 0)
1434 goto put_states;
1435
1436 dst_hold(dst);
1437
1438 for (; i < nx; i++) {
1439 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1440 struct dst_entry *dst1 = &xdst->u.dst;
1441
1442 err = PTR_ERR(xdst);
1443 if (IS_ERR(xdst)) {
1444 dst_release(dst);
1445 goto put_states;
1446 }
1447
1448 if (xfrm[i]->sel.family == AF_UNSPEC) {
1449 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1450 xfrm_af2proto(family));
1451 if (!inner_mode) {
1452 err = -EAFNOSUPPORT;
1453 dst_release(dst);
1454 goto put_states;
1455 }
1456 } else
1457 inner_mode = xfrm[i]->inner_mode;
1458
1459 if (!dst_prev)
1460 dst0 = dst1;
1461 else {
1462 dst_prev->child = dst_clone(dst1);
1463 dst1->flags |= DST_NOHASH;
1464 }
1465
1466 xdst->route = dst;
1467 dst_copy_metrics(dst1, dst);
1468
1469 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1470 family = xfrm[i]->props.family;
1471 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1472 family);
1473 err = PTR_ERR(dst);
1474 if (IS_ERR(dst))
1475 goto put_states;
1476 } else
1477 dst_hold(dst);
1478
1479 dst1->xfrm = xfrm[i];
1480 xdst->xfrm_genid = xfrm[i]->genid;
1481
1482 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
1483 dst1->flags |= DST_HOST;
1484 dst1->lastuse = now;
1485
1486 dst1->input = dst_discard;
1487 dst1->output = inner_mode->afinfo->output;
1488
1489 dst1->next = dst_prev;
1490 dst_prev = dst1;
1491
1492 header_len += xfrm[i]->props.header_len;
1493 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1494 nfheader_len += xfrm[i]->props.header_len;
1495 trailer_len += xfrm[i]->props.trailer_len;
1496 }
1497
1498 dst_prev->child = dst;
1499 dst0->path = dst;
1500
1501 err = -ENODEV;
1502 dev = dst->dev;
1503 if (!dev)
1504 goto free_dst;
1505
1506 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1507 xfrm_init_pmtu(dst_prev);
1508
1509 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1510 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1511
1512 err = xfrm_fill_dst(xdst, dev, fl);
1513 if (err)
1514 goto free_dst;
1515
1516 dst_prev->header_len = header_len;
1517 dst_prev->trailer_len = trailer_len;
1518 header_len -= xdst->u.dst.xfrm->props.header_len;
1519 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1520 }
1521
1522 out:
1523 return dst0;
1524
1525 put_states:
1526 for (; i < nx; i++)
1527 xfrm_state_put(xfrm[i]);
1528 free_dst:
1529 if (dst0)
1530 dst_free(dst0);
1531 dst0 = ERR_PTR(err);
1532 goto out;
1533 }
1534
1535 static int inline
1536 xfrm_dst_alloc_copy(void **target, const void *src, int size)
1537 {
1538 if (!*target) {
1539 *target = kmalloc(size, GFP_ATOMIC);
1540 if (!*target)
1541 return -ENOMEM;
1542 }
1543 memcpy(*target, src, size);
1544 return 0;
1545 }
1546
1547 static int inline
1548 xfrm_dst_update_parent(struct dst_entry *dst, const struct xfrm_selector *sel)
1549 {
1550 #ifdef CONFIG_XFRM_SUB_POLICY
1551 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1552 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1553 sel, sizeof(*sel));
1554 #else
1555 return 0;
1556 #endif
1557 }
1558
1559 static int inline
1560 xfrm_dst_update_origin(struct dst_entry *dst, const struct flowi *fl)
1561 {
1562 #ifdef CONFIG_XFRM_SUB_POLICY
1563 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1564 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1565 #else
1566 return 0;
1567 #endif
1568 }
1569
1570 static int xfrm_expand_policies(const struct flowi *fl, u16 family,
1571 struct xfrm_policy **pols,
1572 int *num_pols, int *num_xfrms)
1573 {
1574 int i;
1575
1576 if (*num_pols == 0 || !pols[0]) {
1577 *num_pols = 0;
1578 *num_xfrms = 0;
1579 return 0;
1580 }
1581 if (IS_ERR(pols[0]))
1582 return PTR_ERR(pols[0]);
1583
1584 *num_xfrms = pols[0]->xfrm_nr;
1585
1586 #ifdef CONFIG_XFRM_SUB_POLICY
1587 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1588 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1589 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1590 XFRM_POLICY_TYPE_MAIN,
1591 fl, family,
1592 XFRM_POLICY_OUT);
1593 if (pols[1]) {
1594 if (IS_ERR(pols[1])) {
1595 xfrm_pols_put(pols, *num_pols);
1596 return PTR_ERR(pols[1]);
1597 }
1598 (*num_pols) ++;
1599 (*num_xfrms) += pols[1]->xfrm_nr;
1600 }
1601 }
1602 #endif
1603 for (i = 0; i < *num_pols; i++) {
1604 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1605 *num_xfrms = -1;
1606 break;
1607 }
1608 }
1609
1610 return 0;
1611
1612 }
1613
1614 static struct xfrm_dst *
1615 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
1616 const struct flowi *fl, u16 family,
1617 struct dst_entry *dst_orig)
1618 {
1619 struct net *net = xp_net(pols[0]);
1620 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1621 struct dst_entry *dst;
1622 struct xfrm_dst *xdst;
1623 int err;
1624
1625 /* Try to instantiate a bundle */
1626 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1627 if (err <= 0) {
1628 if (err != 0 && err != -EAGAIN)
1629 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1630 return ERR_PTR(err);
1631 }
1632
1633 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1634 if (IS_ERR(dst)) {
1635 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1636 return ERR_CAST(dst);
1637 }
1638
1639 xdst = (struct xfrm_dst *)dst;
1640 xdst->num_xfrms = err;
1641 if (num_pols > 1)
1642 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1643 else
1644 err = xfrm_dst_update_origin(dst, fl);
1645 if (unlikely(err)) {
1646 dst_free(dst);
1647 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1648 return ERR_PTR(err);
1649 }
1650
1651 xdst->num_pols = num_pols;
1652 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1653 xdst->policy_genid = atomic_read(&pols[0]->genid);
1654
1655 return xdst;
1656 }
1657
1658 static struct flow_cache_object *
1659 xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
1660 struct flow_cache_object *oldflo, void *ctx)
1661 {
1662 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1663 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1664 struct xfrm_dst *xdst, *new_xdst;
1665 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1666
1667 /* Check if the policies from old bundle are usable */
1668 xdst = NULL;
1669 if (oldflo) {
1670 xdst = container_of(oldflo, struct xfrm_dst, flo);
1671 num_pols = xdst->num_pols;
1672 num_xfrms = xdst->num_xfrms;
1673 pol_dead = 0;
1674 for (i = 0; i < num_pols; i++) {
1675 pols[i] = xdst->pols[i];
1676 pol_dead |= pols[i]->walk.dead;
1677 }
1678 if (pol_dead) {
1679 dst_free(&xdst->u.dst);
1680 xdst = NULL;
1681 num_pols = 0;
1682 num_xfrms = 0;
1683 oldflo = NULL;
1684 }
1685 }
1686
1687 /* Resolve policies to use if we couldn't get them from
1688 * previous cache entry */
1689 if (xdst == NULL) {
1690 num_pols = 1;
1691 pols[0] = __xfrm_policy_lookup(net, fl, family, dir);
1692 err = xfrm_expand_policies(fl, family, pols,
1693 &num_pols, &num_xfrms);
1694 if (err < 0)
1695 goto inc_error;
1696 if (num_pols == 0)
1697 return NULL;
1698 if (num_xfrms <= 0)
1699 goto make_dummy_bundle;
1700 }
1701
1702 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1703 if (IS_ERR(new_xdst)) {
1704 err = PTR_ERR(new_xdst);
1705 if (err != -EAGAIN)
1706 goto error;
1707 if (oldflo == NULL)
1708 goto make_dummy_bundle;
1709 dst_hold(&xdst->u.dst);
1710 return oldflo;
1711 } else if (new_xdst == NULL) {
1712 num_xfrms = 0;
1713 if (oldflo == NULL)
1714 goto make_dummy_bundle;
1715 xdst->num_xfrms = 0;
1716 dst_hold(&xdst->u.dst);
1717 return oldflo;
1718 }
1719
1720 /* Kill the previous bundle */
1721 if (xdst) {
1722 /* The policies were stolen for newly generated bundle */
1723 xdst->num_pols = 0;
1724 dst_free(&xdst->u.dst);
1725 }
1726
1727 /* Flow cache does not have reference, it dst_free()'s,
1728 * but we do need to return one reference for original caller */
1729 dst_hold(&new_xdst->u.dst);
1730 return &new_xdst->flo;
1731
1732 make_dummy_bundle:
1733 /* We found policies, but there's no bundles to instantiate:
1734 * either because the policy blocks, has no transformations or
1735 * we could not build template (no xfrm_states).*/
1736 xdst = xfrm_alloc_dst(net, family);
1737 if (IS_ERR(xdst)) {
1738 xfrm_pols_put(pols, num_pols);
1739 return ERR_CAST(xdst);
1740 }
1741 xdst->num_pols = num_pols;
1742 xdst->num_xfrms = num_xfrms;
1743 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1744
1745 dst_hold(&xdst->u.dst);
1746 return &xdst->flo;
1747
1748 inc_error:
1749 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1750 error:
1751 if (xdst != NULL)
1752 dst_free(&xdst->u.dst);
1753 else
1754 xfrm_pols_put(pols, num_pols);
1755 return ERR_PTR(err);
1756 }
1757
1758 static struct dst_entry *make_blackhole(struct net *net, u16 family,
1759 struct dst_entry *dst_orig)
1760 {
1761 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1762 struct dst_entry *ret;
1763
1764 if (!afinfo) {
1765 dst_release(dst_orig);
1766 ret = ERR_PTR(-EINVAL);
1767 } else {
1768 ret = afinfo->blackhole_route(net, dst_orig);
1769 }
1770 xfrm_policy_put_afinfo(afinfo);
1771
1772 return ret;
1773 }
1774
1775 /* Main function: finds/creates a bundle for given flow.
1776 *
1777 * At the moment we eat a raw IP route. Mostly to speed up lookups
1778 * on interfaces with disabled IPsec.
1779 */
1780 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
1781 const struct flowi *fl,
1782 struct sock *sk, int flags)
1783 {
1784 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1785 struct flow_cache_object *flo;
1786 struct xfrm_dst *xdst;
1787 struct dst_entry *dst, *route;
1788 u16 family = dst_orig->ops->family;
1789 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1790 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
1791
1792 restart:
1793 dst = NULL;
1794 xdst = NULL;
1795 route = NULL;
1796
1797 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1798 num_pols = 1;
1799 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1800 err = xfrm_expand_policies(fl, family, pols,
1801 &num_pols, &num_xfrms);
1802 if (err < 0)
1803 goto dropdst;
1804
1805 if (num_pols) {
1806 if (num_xfrms <= 0) {
1807 drop_pols = num_pols;
1808 goto no_transform;
1809 }
1810
1811 xdst = xfrm_resolve_and_create_bundle(
1812 pols, num_pols, fl,
1813 family, dst_orig);
1814 if (IS_ERR(xdst)) {
1815 xfrm_pols_put(pols, num_pols);
1816 err = PTR_ERR(xdst);
1817 goto dropdst;
1818 } else if (xdst == NULL) {
1819 num_xfrms = 0;
1820 drop_pols = num_pols;
1821 goto no_transform;
1822 }
1823
1824 dst_hold(&xdst->u.dst);
1825
1826 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
1827 xdst->u.dst.next = xfrm_policy_sk_bundles;
1828 xfrm_policy_sk_bundles = &xdst->u.dst;
1829 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
1830
1831 route = xdst->route;
1832 }
1833 }
1834
1835 if (xdst == NULL) {
1836 /* To accelerate a bit... */
1837 if ((dst_orig->flags & DST_NOXFRM) ||
1838 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1839 goto nopol;
1840
1841 flo = flow_cache_lookup(net, fl, family, dir,
1842 xfrm_bundle_lookup, dst_orig);
1843 if (flo == NULL)
1844 goto nopol;
1845 if (IS_ERR(flo)) {
1846 err = PTR_ERR(flo);
1847 goto dropdst;
1848 }
1849 xdst = container_of(flo, struct xfrm_dst, flo);
1850
1851 num_pols = xdst->num_pols;
1852 num_xfrms = xdst->num_xfrms;
1853 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
1854 route = xdst->route;
1855 }
1856
1857 dst = &xdst->u.dst;
1858 if (route == NULL && num_xfrms > 0) {
1859 /* The only case when xfrm_bundle_lookup() returns a
1860 * bundle with null route, is when the template could
1861 * not be resolved. It means policies are there, but
1862 * bundle could not be created, since we don't yet
1863 * have the xfrm_state's. We need to wait for KM to
1864 * negotiate new SA's or bail out with error.*/
1865 if (net->xfrm.sysctl_larval_drop) {
1866 /* EREMOTE tells the caller to generate
1867 * a one-shot blackhole route. */
1868 dst_release(dst);
1869 xfrm_pols_put(pols, drop_pols);
1870 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1871
1872 return make_blackhole(net, family, dst_orig);
1873 }
1874 if (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP) {
1875 DECLARE_WAITQUEUE(wait, current);
1876
1877 add_wait_queue(&net->xfrm.km_waitq, &wait);
1878 set_current_state(TASK_INTERRUPTIBLE);
1879 schedule();
1880 set_current_state(TASK_RUNNING);
1881 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1882
1883 if (!signal_pending(current)) {
1884 dst_release(dst);
1885 goto restart;
1886 }
1887
1888 err = -ERESTART;
1889 } else
1890 err = -EAGAIN;
1891
1892 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1893 goto error;
1894 }
1895
1896 no_transform:
1897 if (num_pols == 0)
1898 goto nopol;
1899
1900 if ((flags & XFRM_LOOKUP_ICMP) &&
1901 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
1902 err = -ENOENT;
1903 goto error;
1904 }
1905
1906 for (i = 0; i < num_pols; i++)
1907 pols[i]->curlft.use_time = get_seconds();
1908
1909 if (num_xfrms < 0) {
1910 /* Prohibit the flow */
1911 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1912 err = -EPERM;
1913 goto error;
1914 } else if (num_xfrms > 0) {
1915 /* Flow transformed */
1916 dst_release(dst_orig);
1917 } else {
1918 /* Flow passes untransformed */
1919 dst_release(dst);
1920 dst = dst_orig;
1921 }
1922 ok:
1923 xfrm_pols_put(pols, drop_pols);
1924 if (dst && dst->xfrm &&
1925 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
1926 dst->flags |= DST_XFRM_TUNNEL;
1927 return dst;
1928
1929 nopol:
1930 if (!(flags & XFRM_LOOKUP_ICMP)) {
1931 dst = dst_orig;
1932 goto ok;
1933 }
1934 err = -ENOENT;
1935 error:
1936 dst_release(dst);
1937 dropdst:
1938 dst_release(dst_orig);
1939 xfrm_pols_put(pols, drop_pols);
1940 return ERR_PTR(err);
1941 }
1942 EXPORT_SYMBOL(xfrm_lookup);
1943
1944 static inline int
1945 xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
1946 {
1947 struct xfrm_state *x;
1948
1949 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1950 return 0;
1951 x = skb->sp->xvec[idx];
1952 if (!x->type->reject)
1953 return 0;
1954 return x->type->reject(x, skb, fl);
1955 }
1956
1957 /* When skb is transformed back to its "native" form, we have to
1958 * check policy restrictions. At the moment we make this in maximally
1959 * stupid way. Shame on me. :-) Of course, connected sockets must
1960 * have policy cached at them.
1961 */
1962
1963 static inline int
1964 xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1965 unsigned short family)
1966 {
1967 if (xfrm_state_kern(x))
1968 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1969 return x->id.proto == tmpl->id.proto &&
1970 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1971 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1972 x->props.mode == tmpl->mode &&
1973 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1974 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1975 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1976 xfrm_state_addr_cmp(tmpl, x, family));
1977 }
1978
1979 /*
1980 * 0 or more than 0 is returned when validation is succeeded (either bypass
1981 * because of optional transport mode, or next index of the mathced secpath
1982 * state with the template.
1983 * -1 is returned when no matching template is found.
1984 * Otherwise "-2 - errored_index" is returned.
1985 */
1986 static inline int
1987 xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1988 unsigned short family)
1989 {
1990 int idx = start;
1991
1992 if (tmpl->optional) {
1993 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1994 return start;
1995 } else
1996 start = -1;
1997 for (; idx < sp->len; idx++) {
1998 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1999 return ++idx;
2000 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2001 if (start == -1)
2002 start = -2-idx;
2003 break;
2004 }
2005 }
2006 return start;
2007 }
2008
2009 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2010 unsigned int family, int reverse)
2011 {
2012 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2013 int err;
2014
2015 if (unlikely(afinfo == NULL))
2016 return -EAFNOSUPPORT;
2017
2018 afinfo->decode_session(skb, fl, reverse);
2019 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
2020 xfrm_policy_put_afinfo(afinfo);
2021 return err;
2022 }
2023 EXPORT_SYMBOL(__xfrm_decode_session);
2024
2025 static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
2026 {
2027 for (; k < sp->len; k++) {
2028 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
2029 *idxp = k;
2030 return 1;
2031 }
2032 }
2033
2034 return 0;
2035 }
2036
2037 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
2038 unsigned short family)
2039 {
2040 struct net *net = dev_net(skb->dev);
2041 struct xfrm_policy *pol;
2042 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2043 int npols = 0;
2044 int xfrm_nr;
2045 int pi;
2046 int reverse;
2047 struct flowi fl;
2048 u8 fl_dir;
2049 int xerr_idx = -1;
2050
2051 reverse = dir & ~XFRM_POLICY_MASK;
2052 dir &= XFRM_POLICY_MASK;
2053 fl_dir = policy_to_flow_dir(dir);
2054
2055 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
2056 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
2057 return 0;
2058 }
2059
2060 nf_nat_decode_session(skb, &fl, family);
2061
2062 /* First, check used SA against their selectors. */
2063 if (skb->sp) {
2064 int i;
2065
2066 for (i=skb->sp->len-1; i>=0; i--) {
2067 struct xfrm_state *x = skb->sp->xvec[i];
2068 if (!xfrm_selector_match(&x->sel, &fl, family)) {
2069 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
2070 return 0;
2071 }
2072 }
2073 }
2074
2075 pol = NULL;
2076 if (sk && sk->sk_policy[dir]) {
2077 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
2078 if (IS_ERR(pol)) {
2079 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2080 return 0;
2081 }
2082 }
2083
2084 if (!pol) {
2085 struct flow_cache_object *flo;
2086
2087 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2088 xfrm_policy_lookup, NULL);
2089 if (IS_ERR_OR_NULL(flo))
2090 pol = ERR_CAST(flo);
2091 else
2092 pol = container_of(flo, struct xfrm_policy, flo);
2093 }
2094
2095 if (IS_ERR(pol)) {
2096 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2097 return 0;
2098 }
2099
2100 if (!pol) {
2101 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
2102 xfrm_secpath_reject(xerr_idx, skb, &fl);
2103 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
2104 return 0;
2105 }
2106 return 1;
2107 }
2108
2109 pol->curlft.use_time = get_seconds();
2110
2111 pols[0] = pol;
2112 npols ++;
2113 #ifdef CONFIG_XFRM_SUB_POLICY
2114 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
2115 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
2116 &fl, family,
2117 XFRM_POLICY_IN);
2118 if (pols[1]) {
2119 if (IS_ERR(pols[1])) {
2120 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2121 return 0;
2122 }
2123 pols[1]->curlft.use_time = get_seconds();
2124 npols ++;
2125 }
2126 }
2127 #endif
2128
2129 if (pol->action == XFRM_POLICY_ALLOW) {
2130 struct sec_path *sp;
2131 static struct sec_path dummy;
2132 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
2133 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
2134 struct xfrm_tmpl **tpp = tp;
2135 int ti = 0;
2136 int i, k;
2137
2138 if ((sp = skb->sp) == NULL)
2139 sp = &dummy;
2140
2141 for (pi = 0; pi < npols; pi++) {
2142 if (pols[pi] != pol &&
2143 pols[pi]->action != XFRM_POLICY_ALLOW) {
2144 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2145 goto reject;
2146 }
2147 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2148 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2149 goto reject_error;
2150 }
2151 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2152 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2153 }
2154 xfrm_nr = ti;
2155 if (npols > 1) {
2156 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2157 tpp = stp;
2158 }
2159
2160 /* For each tunnel xfrm, find the first matching tmpl.
2161 * For each tmpl before that, find corresponding xfrm.
2162 * Order is _important_. Later we will implement
2163 * some barriers, but at the moment barriers
2164 * are implied between each two transformations.
2165 */
2166 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2167 k = xfrm_policy_ok(tpp[i], sp, k, family);
2168 if (k < 0) {
2169 if (k < -1)
2170 /* "-2 - errored_index" returned */
2171 xerr_idx = -(2+k);
2172 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2173 goto reject;
2174 }
2175 }
2176
2177 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2178 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2179 goto reject;
2180 }
2181
2182 xfrm_pols_put(pols, npols);
2183 return 1;
2184 }
2185 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2186
2187 reject:
2188 xfrm_secpath_reject(xerr_idx, skb, &fl);
2189 reject_error:
2190 xfrm_pols_put(pols, npols);
2191 return 0;
2192 }
2193 EXPORT_SYMBOL(__xfrm_policy_check);
2194
2195 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2196 {
2197 struct net *net = dev_net(skb->dev);
2198 struct flowi fl;
2199 struct dst_entry *dst;
2200 int res = 1;
2201
2202 if (xfrm_decode_session(skb, &fl, family) < 0) {
2203 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2204 return 0;
2205 }
2206
2207 skb_dst_force(skb);
2208
2209 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2210 if (IS_ERR(dst)) {
2211 res = 0;
2212 dst = NULL;
2213 }
2214 skb_dst_set(skb, dst);
2215 return res;
2216 }
2217 EXPORT_SYMBOL(__xfrm_route_forward);
2218
2219 /* Optimize later using cookies and generation ids. */
2220
2221 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2222 {
2223 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2224 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2225 * get validated by dst_ops->check on every use. We do this
2226 * because when a normal route referenced by an XFRM dst is
2227 * obsoleted we do not go looking around for all parent
2228 * referencing XFRM dsts so that we can invalidate them. It
2229 * is just too much work. Instead we make the checks here on
2230 * every use. For example:
2231 *
2232 * XFRM dst A --> IPv4 dst X
2233 *
2234 * X is the "xdst->route" of A (X is also the "dst->path" of A
2235 * in this example). If X is marked obsolete, "A" will not
2236 * notice. That's what we are validating here via the
2237 * stale_bundle() check.
2238 *
2239 * When a policy's bundle is pruned, we dst_free() the XFRM
2240 * dst which causes it's ->obsolete field to be set to
2241 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2242 * this, we want to force a new route lookup.
2243 */
2244 if (dst->obsolete < 0 && !stale_bundle(dst))
2245 return dst;
2246
2247 return NULL;
2248 }
2249
2250 static int stale_bundle(struct dst_entry *dst)
2251 {
2252 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
2253 }
2254
2255 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2256 {
2257 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2258 dst->dev = dev_net(dev)->loopback_dev;
2259 dev_hold(dst->dev);
2260 dev_put(dev);
2261 }
2262 }
2263 EXPORT_SYMBOL(xfrm_dst_ifdown);
2264
2265 static void xfrm_link_failure(struct sk_buff *skb)
2266 {
2267 /* Impossible. Such dst must be popped before reaches point of failure. */
2268 }
2269
2270 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2271 {
2272 if (dst) {
2273 if (dst->obsolete) {
2274 dst_release(dst);
2275 dst = NULL;
2276 }
2277 }
2278 return dst;
2279 }
2280
2281 static void __xfrm_garbage_collect(struct net *net)
2282 {
2283 struct dst_entry *head, *next;
2284
2285 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2286 head = xfrm_policy_sk_bundles;
2287 xfrm_policy_sk_bundles = NULL;
2288 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2289
2290 while (head) {
2291 next = head->next;
2292 dst_free(head);
2293 head = next;
2294 }
2295 }
2296
2297 static void xfrm_garbage_collect(struct net *net)
2298 {
2299 flow_cache_flush();
2300 __xfrm_garbage_collect(net);
2301 }
2302
2303 static void xfrm_garbage_collect_deferred(struct net *net)
2304 {
2305 flow_cache_flush_deferred();
2306 __xfrm_garbage_collect(net);
2307 }
2308
2309 static void xfrm_init_pmtu(struct dst_entry *dst)
2310 {
2311 do {
2312 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2313 u32 pmtu, route_mtu_cached;
2314
2315 pmtu = dst_mtu(dst->child);
2316 xdst->child_mtu_cached = pmtu;
2317
2318 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2319
2320 route_mtu_cached = dst_mtu(xdst->route);
2321 xdst->route_mtu_cached = route_mtu_cached;
2322
2323 if (pmtu > route_mtu_cached)
2324 pmtu = route_mtu_cached;
2325
2326 dst_metric_set(dst, RTAX_MTU, pmtu);
2327 } while ((dst = dst->next));
2328 }
2329
2330 /* Check that the bundle accepts the flow and its components are
2331 * still valid.
2332 */
2333
2334 static int xfrm_bundle_ok(struct xfrm_dst *first)
2335 {
2336 struct dst_entry *dst = &first->u.dst;
2337 struct xfrm_dst *last;
2338 u32 mtu;
2339
2340 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2341 (dst->dev && !netif_running(dst->dev)))
2342 return 0;
2343
2344 last = NULL;
2345
2346 do {
2347 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2348
2349 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2350 return 0;
2351 if (xdst->xfrm_genid != dst->xfrm->genid)
2352 return 0;
2353 if (xdst->num_pols > 0 &&
2354 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
2355 return 0;
2356
2357 mtu = dst_mtu(dst->child);
2358 if (xdst->child_mtu_cached != mtu) {
2359 last = xdst;
2360 xdst->child_mtu_cached = mtu;
2361 }
2362
2363 if (!dst_check(xdst->route, xdst->route_cookie))
2364 return 0;
2365 mtu = dst_mtu(xdst->route);
2366 if (xdst->route_mtu_cached != mtu) {
2367 last = xdst;
2368 xdst->route_mtu_cached = mtu;
2369 }
2370
2371 dst = dst->child;
2372 } while (dst->xfrm);
2373
2374 if (likely(!last))
2375 return 1;
2376
2377 mtu = last->child_mtu_cached;
2378 for (;;) {
2379 dst = &last->u.dst;
2380
2381 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2382 if (mtu > last->route_mtu_cached)
2383 mtu = last->route_mtu_cached;
2384 dst_metric_set(dst, RTAX_MTU, mtu);
2385
2386 if (last == first)
2387 break;
2388
2389 last = (struct xfrm_dst *)last->u.dst.next;
2390 last->child_mtu_cached = mtu;
2391 }
2392
2393 return 1;
2394 }
2395
2396 static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2397 {
2398 return dst_metric_advmss(dst->path);
2399 }
2400
2401 static unsigned int xfrm_mtu(const struct dst_entry *dst)
2402 {
2403 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2404
2405 return mtu ? : dst_mtu(dst->path);
2406 }
2407
2408 static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2409 struct sk_buff *skb,
2410 const void *daddr)
2411 {
2412 return dst->path->ops->neigh_lookup(dst, skb, daddr);
2413 }
2414
2415 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2416 {
2417 struct net *net;
2418 int err = 0;
2419 if (unlikely(afinfo == NULL))
2420 return -EINVAL;
2421 if (unlikely(afinfo->family >= NPROTO))
2422 return -EAFNOSUPPORT;
2423 write_lock_bh(&xfrm_policy_afinfo_lock);
2424 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2425 err = -ENOBUFS;
2426 else {
2427 struct dst_ops *dst_ops = afinfo->dst_ops;
2428 if (likely(dst_ops->kmem_cachep == NULL))
2429 dst_ops->kmem_cachep = xfrm_dst_cache;
2430 if (likely(dst_ops->check == NULL))
2431 dst_ops->check = xfrm_dst_check;
2432 if (likely(dst_ops->default_advmss == NULL))
2433 dst_ops->default_advmss = xfrm_default_advmss;
2434 if (likely(dst_ops->mtu == NULL))
2435 dst_ops->mtu = xfrm_mtu;
2436 if (likely(dst_ops->negative_advice == NULL))
2437 dst_ops->negative_advice = xfrm_negative_advice;
2438 if (likely(dst_ops->link_failure == NULL))
2439 dst_ops->link_failure = xfrm_link_failure;
2440 if (likely(dst_ops->neigh_lookup == NULL))
2441 dst_ops->neigh_lookup = xfrm_neigh_lookup;
2442 if (likely(afinfo->garbage_collect == NULL))
2443 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
2444 xfrm_policy_afinfo[afinfo->family] = afinfo;
2445 }
2446 write_unlock_bh(&xfrm_policy_afinfo_lock);
2447
2448 rtnl_lock();
2449 for_each_net(net) {
2450 struct dst_ops *xfrm_dst_ops;
2451
2452 switch (afinfo->family) {
2453 case AF_INET:
2454 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2455 break;
2456 #if IS_ENABLED(CONFIG_IPV6)
2457 case AF_INET6:
2458 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2459 break;
2460 #endif
2461 default:
2462 BUG();
2463 }
2464 *xfrm_dst_ops = *afinfo->dst_ops;
2465 }
2466 rtnl_unlock();
2467
2468 return err;
2469 }
2470 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2471
2472 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2473 {
2474 int err = 0;
2475 if (unlikely(afinfo == NULL))
2476 return -EINVAL;
2477 if (unlikely(afinfo->family >= NPROTO))
2478 return -EAFNOSUPPORT;
2479 write_lock_bh(&xfrm_policy_afinfo_lock);
2480 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2481 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2482 err = -EINVAL;
2483 else {
2484 struct dst_ops *dst_ops = afinfo->dst_ops;
2485 xfrm_policy_afinfo[afinfo->family] = NULL;
2486 dst_ops->kmem_cachep = NULL;
2487 dst_ops->check = NULL;
2488 dst_ops->negative_advice = NULL;
2489 dst_ops->link_failure = NULL;
2490 afinfo->garbage_collect = NULL;
2491 }
2492 }
2493 write_unlock_bh(&xfrm_policy_afinfo_lock);
2494 return err;
2495 }
2496 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2497
2498 static void __net_init xfrm_dst_ops_init(struct net *net)
2499 {
2500 struct xfrm_policy_afinfo *afinfo;
2501
2502 read_lock_bh(&xfrm_policy_afinfo_lock);
2503 afinfo = xfrm_policy_afinfo[AF_INET];
2504 if (afinfo)
2505 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2506 #if IS_ENABLED(CONFIG_IPV6)
2507 afinfo = xfrm_policy_afinfo[AF_INET6];
2508 if (afinfo)
2509 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2510 #endif
2511 read_unlock_bh(&xfrm_policy_afinfo_lock);
2512 }
2513
2514 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2515 {
2516 struct xfrm_policy_afinfo *afinfo;
2517 if (unlikely(family >= NPROTO))
2518 return NULL;
2519 read_lock(&xfrm_policy_afinfo_lock);
2520 afinfo = xfrm_policy_afinfo[family];
2521 if (unlikely(!afinfo))
2522 read_unlock(&xfrm_policy_afinfo_lock);
2523 return afinfo;
2524 }
2525
2526 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2527 {
2528 read_unlock(&xfrm_policy_afinfo_lock);
2529 }
2530
2531 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2532 {
2533 struct net_device *dev = ptr;
2534
2535 switch (event) {
2536 case NETDEV_DOWN:
2537 xfrm_garbage_collect(dev_net(dev));
2538 }
2539 return NOTIFY_DONE;
2540 }
2541
2542 static struct notifier_block xfrm_dev_notifier = {
2543 .notifier_call = xfrm_dev_event,
2544 };
2545
2546 #ifdef CONFIG_XFRM_STATISTICS
2547 static int __net_init xfrm_statistics_init(struct net *net)
2548 {
2549 int rv;
2550
2551 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2552 sizeof(struct linux_xfrm_mib),
2553 __alignof__(struct linux_xfrm_mib)) < 0)
2554 return -ENOMEM;
2555 rv = xfrm_proc_init(net);
2556 if (rv < 0)
2557 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2558 return rv;
2559 }
2560
2561 static void xfrm_statistics_fini(struct net *net)
2562 {
2563 xfrm_proc_fini(net);
2564 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2565 }
2566 #else
2567 static int __net_init xfrm_statistics_init(struct net *net)
2568 {
2569 return 0;
2570 }
2571
2572 static void xfrm_statistics_fini(struct net *net)
2573 {
2574 }
2575 #endif
2576
2577 static int __net_init xfrm_policy_init(struct net *net)
2578 {
2579 unsigned int hmask, sz;
2580 int dir;
2581
2582 if (net_eq(net, &init_net))
2583 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2584 sizeof(struct xfrm_dst),
2585 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2586 NULL);
2587
2588 hmask = 8 - 1;
2589 sz = (hmask+1) * sizeof(struct hlist_head);
2590
2591 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2592 if (!net->xfrm.policy_byidx)
2593 goto out_byidx;
2594 net->xfrm.policy_idx_hmask = hmask;
2595
2596 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2597 struct xfrm_policy_hash *htab;
2598
2599 net->xfrm.policy_count[dir] = 0;
2600 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2601
2602 htab = &net->xfrm.policy_bydst[dir];
2603 htab->table = xfrm_hash_alloc(sz);
2604 if (!htab->table)
2605 goto out_bydst;
2606 htab->hmask = hmask;
2607 }
2608
2609 INIT_LIST_HEAD(&net->xfrm.policy_all);
2610 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2611 if (net_eq(net, &init_net))
2612 register_netdevice_notifier(&xfrm_dev_notifier);
2613 return 0;
2614
2615 out_bydst:
2616 for (dir--; dir >= 0; dir--) {
2617 struct xfrm_policy_hash *htab;
2618
2619 htab = &net->xfrm.policy_bydst[dir];
2620 xfrm_hash_free(htab->table, sz);
2621 }
2622 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2623 out_byidx:
2624 return -ENOMEM;
2625 }
2626
2627 static void xfrm_policy_fini(struct net *net)
2628 {
2629 struct xfrm_audit audit_info;
2630 unsigned int sz;
2631 int dir;
2632
2633 flush_work(&net->xfrm.policy_hash_work);
2634 #ifdef CONFIG_XFRM_SUB_POLICY
2635 audit_info.loginuid = -1;
2636 audit_info.sessionid = -1;
2637 audit_info.secid = 0;
2638 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2639 #endif
2640 audit_info.loginuid = -1;
2641 audit_info.sessionid = -1;
2642 audit_info.secid = 0;
2643 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2644
2645 WARN_ON(!list_empty(&net->xfrm.policy_all));
2646
2647 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2648 struct xfrm_policy_hash *htab;
2649
2650 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2651
2652 htab = &net->xfrm.policy_bydst[dir];
2653 sz = (htab->hmask + 1);
2654 WARN_ON(!hlist_empty(htab->table));
2655 xfrm_hash_free(htab->table, sz);
2656 }
2657
2658 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2659 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2660 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2661 }
2662
2663 static int __net_init xfrm_net_init(struct net *net)
2664 {
2665 int rv;
2666
2667 rv = xfrm_statistics_init(net);
2668 if (rv < 0)
2669 goto out_statistics;
2670 rv = xfrm_state_init(net);
2671 if (rv < 0)
2672 goto out_state;
2673 rv = xfrm_policy_init(net);
2674 if (rv < 0)
2675 goto out_policy;
2676 xfrm_dst_ops_init(net);
2677 rv = xfrm_sysctl_init(net);
2678 if (rv < 0)
2679 goto out_sysctl;
2680 return 0;
2681
2682 out_sysctl:
2683 xfrm_policy_fini(net);
2684 out_policy:
2685 xfrm_state_fini(net);
2686 out_state:
2687 xfrm_statistics_fini(net);
2688 out_statistics:
2689 return rv;
2690 }
2691
2692 static void __net_exit xfrm_net_exit(struct net *net)
2693 {
2694 xfrm_sysctl_fini(net);
2695 xfrm_policy_fini(net);
2696 xfrm_state_fini(net);
2697 xfrm_statistics_fini(net);
2698 }
2699
2700 static struct pernet_operations __net_initdata xfrm_net_ops = {
2701 .init = xfrm_net_init,
2702 .exit = xfrm_net_exit,
2703 };
2704
2705 void __init xfrm_init(void)
2706 {
2707 register_pernet_subsys(&xfrm_net_ops);
2708 xfrm_input_init();
2709 }
2710
2711 #ifdef CONFIG_AUDITSYSCALL
2712 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2713 struct audit_buffer *audit_buf)
2714 {
2715 struct xfrm_sec_ctx *ctx = xp->security;
2716 struct xfrm_selector *sel = &xp->selector;
2717
2718 if (ctx)
2719 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2720 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2721
2722 switch(sel->family) {
2723 case AF_INET:
2724 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2725 if (sel->prefixlen_s != 32)
2726 audit_log_format(audit_buf, " src_prefixlen=%d",
2727 sel->prefixlen_s);
2728 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2729 if (sel->prefixlen_d != 32)
2730 audit_log_format(audit_buf, " dst_prefixlen=%d",
2731 sel->prefixlen_d);
2732 break;
2733 case AF_INET6:
2734 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2735 if (sel->prefixlen_s != 128)
2736 audit_log_format(audit_buf, " src_prefixlen=%d",
2737 sel->prefixlen_s);
2738 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2739 if (sel->prefixlen_d != 128)
2740 audit_log_format(audit_buf, " dst_prefixlen=%d",
2741 sel->prefixlen_d);
2742 break;
2743 }
2744 }
2745
2746 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2747 uid_t auid, u32 sessionid, u32 secid)
2748 {
2749 struct audit_buffer *audit_buf;
2750
2751 audit_buf = xfrm_audit_start("SPD-add");
2752 if (audit_buf == NULL)
2753 return;
2754 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2755 audit_log_format(audit_buf, " res=%u", result);
2756 xfrm_audit_common_policyinfo(xp, audit_buf);
2757 audit_log_end(audit_buf);
2758 }
2759 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2760
2761 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2762 uid_t auid, u32 sessionid, u32 secid)
2763 {
2764 struct audit_buffer *audit_buf;
2765
2766 audit_buf = xfrm_audit_start("SPD-delete");
2767 if (audit_buf == NULL)
2768 return;
2769 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2770 audit_log_format(audit_buf, " res=%u", result);
2771 xfrm_audit_common_policyinfo(xp, audit_buf);
2772 audit_log_end(audit_buf);
2773 }
2774 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2775 #endif
2776
2777 #ifdef CONFIG_XFRM_MIGRATE
2778 static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
2779 const struct xfrm_selector *sel_tgt)
2780 {
2781 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2782 if (sel_tgt->family == sel_cmp->family &&
2783 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2784 sel_cmp->family) == 0 &&
2785 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2786 sel_cmp->family) == 0 &&
2787 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2788 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2789 return true;
2790 }
2791 } else {
2792 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2793 return true;
2794 }
2795 }
2796 return false;
2797 }
2798
2799 static struct xfrm_policy * xfrm_migrate_policy_find(const struct xfrm_selector *sel,
2800 u8 dir, u8 type)
2801 {
2802 struct xfrm_policy *pol, *ret = NULL;
2803 struct hlist_node *entry;
2804 struct hlist_head *chain;
2805 u32 priority = ~0U;
2806
2807 read_lock_bh(&xfrm_policy_lock);
2808 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2809 hlist_for_each_entry(pol, entry, chain, bydst) {
2810 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2811 pol->type == type) {
2812 ret = pol;
2813 priority = ret->priority;
2814 break;
2815 }
2816 }
2817 chain = &init_net.xfrm.policy_inexact[dir];
2818 hlist_for_each_entry(pol, entry, chain, bydst) {
2819 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2820 pol->type == type &&
2821 pol->priority < priority) {
2822 ret = pol;
2823 break;
2824 }
2825 }
2826
2827 if (ret)
2828 xfrm_pol_hold(ret);
2829
2830 read_unlock_bh(&xfrm_policy_lock);
2831
2832 return ret;
2833 }
2834
2835 static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
2836 {
2837 int match = 0;
2838
2839 if (t->mode == m->mode && t->id.proto == m->proto &&
2840 (m->reqid == 0 || t->reqid == m->reqid)) {
2841 switch (t->mode) {
2842 case XFRM_MODE_TUNNEL:
2843 case XFRM_MODE_BEET:
2844 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2845 m->old_family) == 0 &&
2846 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2847 m->old_family) == 0) {
2848 match = 1;
2849 }
2850 break;
2851 case XFRM_MODE_TRANSPORT:
2852 /* in case of transport mode, template does not store
2853 any IP addresses, hence we just compare mode and
2854 protocol */
2855 match = 1;
2856 break;
2857 default:
2858 break;
2859 }
2860 }
2861 return match;
2862 }
2863
2864 /* update endpoint address(es) of template(s) */
2865 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2866 struct xfrm_migrate *m, int num_migrate)
2867 {
2868 struct xfrm_migrate *mp;
2869 int i, j, n = 0;
2870
2871 write_lock_bh(&pol->lock);
2872 if (unlikely(pol->walk.dead)) {
2873 /* target policy has been deleted */
2874 write_unlock_bh(&pol->lock);
2875 return -ENOENT;
2876 }
2877
2878 for (i = 0; i < pol->xfrm_nr; i++) {
2879 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2880 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2881 continue;
2882 n++;
2883 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2884 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2885 continue;
2886 /* update endpoints */
2887 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2888 sizeof(pol->xfrm_vec[i].id.daddr));
2889 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2890 sizeof(pol->xfrm_vec[i].saddr));
2891 pol->xfrm_vec[i].encap_family = mp->new_family;
2892 /* flush bundles */
2893 atomic_inc(&pol->genid);
2894 }
2895 }
2896
2897 write_unlock_bh(&pol->lock);
2898
2899 if (!n)
2900 return -ENODATA;
2901
2902 return 0;
2903 }
2904
2905 static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
2906 {
2907 int i, j;
2908
2909 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2910 return -EINVAL;
2911
2912 for (i = 0; i < num_migrate; i++) {
2913 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2914 m[i].old_family) == 0) &&
2915 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2916 m[i].old_family) == 0))
2917 return -EINVAL;
2918 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2919 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2920 return -EINVAL;
2921
2922 /* check if there is any duplicated entry */
2923 for (j = i + 1; j < num_migrate; j++) {
2924 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2925 sizeof(m[i].old_daddr)) &&
2926 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2927 sizeof(m[i].old_saddr)) &&
2928 m[i].proto == m[j].proto &&
2929 m[i].mode == m[j].mode &&
2930 m[i].reqid == m[j].reqid &&
2931 m[i].old_family == m[j].old_family)
2932 return -EINVAL;
2933 }
2934 }
2935
2936 return 0;
2937 }
2938
2939 int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2940 struct xfrm_migrate *m, int num_migrate,
2941 struct xfrm_kmaddress *k)
2942 {
2943 int i, err, nx_cur = 0, nx_new = 0;
2944 struct xfrm_policy *pol = NULL;
2945 struct xfrm_state *x, *xc;
2946 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2947 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2948 struct xfrm_migrate *mp;
2949
2950 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2951 goto out;
2952
2953 /* Stage 1 - find policy */
2954 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2955 err = -ENOENT;
2956 goto out;
2957 }
2958
2959 /* Stage 2 - find and update state(s) */
2960 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2961 if ((x = xfrm_migrate_state_find(mp))) {
2962 x_cur[nx_cur] = x;
2963 nx_cur++;
2964 if ((xc = xfrm_state_migrate(x, mp))) {
2965 x_new[nx_new] = xc;
2966 nx_new++;
2967 } else {
2968 err = -ENODATA;
2969 goto restore_state;
2970 }
2971 }
2972 }
2973
2974 /* Stage 3 - update policy */
2975 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2976 goto restore_state;
2977
2978 /* Stage 4 - delete old state(s) */
2979 if (nx_cur) {
2980 xfrm_states_put(x_cur, nx_cur);
2981 xfrm_states_delete(x_cur, nx_cur);
2982 }
2983
2984 /* Stage 5 - announce */
2985 km_migrate(sel, dir, type, m, num_migrate, k);
2986
2987 xfrm_pol_put(pol);
2988
2989 return 0;
2990 out:
2991 return err;
2992
2993 restore_state:
2994 if (pol)
2995 xfrm_pol_put(pol);
2996 if (nx_cur)
2997 xfrm_states_put(x_cur, nx_cur);
2998 if (nx_new)
2999 xfrm_states_delete(x_new, nx_new);
3000
3001 return err;
3002 }
3003 EXPORT_SYMBOL(xfrm_migrate);
3004 #endif
Linus' kernel tree