projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: Rearrange dst_entry layout to avoid useless padding.
[linux-2.6-block.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT		 2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 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>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
ec30d78c 27#include <linux/cpu.h>
68277acc 28#include <linux/audit.h>
25ee3286 29#include <net/dst.h>
6ce74ec7 30#include <net/flow.h>
1da177e4
LT		 31#include <net/xfrm.h>
32#include <net/ip.h>
558f82ef
MN		 33#ifdef CONFIG_XFRM_STATISTICS
34#include <net/snmp.h>
35#endif
1da177e4 36
44e36b42
DM		 37#include "xfrm_hash.h"
38
a0073fe1
SK		 39#define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
40#define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
41#define XFRM_MAX_QUEUE_LEN 100
42
b8c203b2
SK		 43struct xfrm_flo {
44 struct dst_entry *dst_orig;
45 u8 flags;
46};
47
ec30d78c
FW		 48static DEFINE_PER_CPU(struct xfrm_dst *, xfrm_last_dst);
49static struct work_struct *xfrm_pcpu_work __read_mostly;
418a99ac 50static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
37b10383 51static struct xfrm_policy_afinfo const __rcu *xfrm_policy_afinfo[AF_INET6 + 1]
418a99ac 52 __read_mostly;
1da177e4 53
e18b890b 54static struct kmem_cache *xfrm_dst_cache __read_mostly;
30846090 55static __read_mostly seqcount_t xfrm_policy_hash_generation;
1da177e4 56
25ee3286 57static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 58static int stale_bundle(struct dst_entry *dst);
12fdb4d3 59static int xfrm_bundle_ok(struct xfrm_dst *xdst);
c3aed709 60static void xfrm_policy_queue_process(struct timer_list *t);
1da177e4 61
12bfa8bd 62static void __xfrm_policy_link(struct xfrm_policy *pol, int dir);
29fa0b30
WY		 63static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
64 int dir);
65
e37cc8ad
FW		 66static inline bool xfrm_pol_hold_rcu(struct xfrm_policy *policy)
67{
850a6212 68 return refcount_inc_not_zero(&policy->refcnt);
e37cc8ad
FW		 69}
70
bc9b35ad 71static inline bool
200ce96e 72__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 73{
7e1dc7b6
DM		 74 const struct flowi4 *fl4 = &fl->u.ip4;
75
26bff940
AD		 76 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
77 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 78 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
79 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
80 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
81 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 82}
83
bc9b35ad 84static inline bool
200ce96e 85__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 86{
7e1dc7b6
DM		 87 const struct flowi6 *fl6 = &fl->u.ip6;
88
89 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
90 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
91 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
92 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
93 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
94 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 95}
96
bc9b35ad
DM		 97bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
98 unsigned short family)
77681021
AM		 99{
100 switch (family) {
101 case AF_INET:
102 return __xfrm4_selector_match(sel, fl);
103 case AF_INET6:
104 return __xfrm6_selector_match(sel, fl);
105 }
bc9b35ad 106 return false;
77681021
AM		 107}
108
a2817d8b 109static const struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
ef8531b6 110{
a2817d8b 111 const struct xfrm_policy_afinfo *afinfo;
ef8531b6 112
a2817d8b 113 if (unlikely(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
ef8531b6
ED		 114 return NULL;
115 rcu_read_lock();
116 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
117 if (unlikely(!afinfo))
118 rcu_read_unlock();
119 return afinfo;
120}
121
d77e38e6
SK		 122struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos, int oif,
123 const xfrm_address_t *saddr,
124 const xfrm_address_t *daddr,
077fbac4 125 int family, u32 mark)
9bb182a7 126{
37b10383 127 const struct xfrm_policy_afinfo *afinfo;
9bb182a7
YH		 128 struct dst_entry *dst;
129
130 afinfo = xfrm_policy_get_afinfo(family);
131 if (unlikely(afinfo == NULL))
132 return ERR_PTR(-EAFNOSUPPORT);
133
077fbac4 134 dst = afinfo->dst_lookup(net, tos, oif, saddr, daddr, mark);
9bb182a7 135
bdba9fe0 136 rcu_read_unlock();
9bb182a7
YH		 137
138 return dst;
139}
d77e38e6 140EXPORT_SYMBOL(__xfrm_dst_lookup);
9bb182a7 141
42a7b32b
DA		 142static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
143 int tos, int oif,
9bb182a7
YH		 144 xfrm_address_t *prev_saddr,
145 xfrm_address_t *prev_daddr,
077fbac4 146 int family, u32 mark)
1da177e4 147{
c5b3cf46 148 struct net *net = xs_net(x);
66cdb3ca
HX		 149 xfrm_address_t *saddr = &x->props.saddr;
150 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 151 struct dst_entry *dst;
152
9bb182a7 153 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 154 saddr = x->coaddr;
9bb182a7
YH		 155 daddr = prev_daddr;
156 }
157 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
158 saddr = prev_saddr;
66cdb3ca 159 daddr = x->coaddr;
9bb182a7 160 }
1da177e4 161
077fbac4 162 dst = __xfrm_dst_lookup(net, tos, oif, saddr, daddr, family, mark);
9bb182a7
YH		 163
164 if (!IS_ERR(dst)) {
165 if (prev_saddr != saddr)
166 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
167 if (prev_daddr != daddr)
168 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
169 }
1da177e4 170
66cdb3ca 171 return dst;
1da177e4 172}
1da177e4 173
1da177e4
LT		 174static inline unsigned long make_jiffies(long secs)
175{
176 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
177 return MAX_SCHEDULE_TIMEOUT-1;
178 else
a716c119 179 return secs*HZ;
1da177e4
LT		 180}
181
c3aed709 182static void xfrm_policy_timer(struct timer_list *t)
1da177e4 183{
c3aed709 184 struct xfrm_policy *xp = from_timer(xp, t, timer);
9d729f72 185 unsigned long now = get_seconds();
1da177e4
LT		 186 long next = LONG_MAX;
187 int warn = 0;
188 int dir;
189
190 read_lock(&xp->lock);
191
ea2dea9d 192 if (unlikely(xp->walk.dead))
1da177e4
LT		 193 goto out;
194
77d8d7a6 195 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 196
197 if (xp->lft.hard_add_expires_seconds) {
198 long tmo = xp->lft.hard_add_expires_seconds +
199 xp->curlft.add_time - now;
200 if (tmo <= 0)
201 goto expired;
202 if (tmo < next)
203 next = tmo;
204 }
205 if (xp->lft.hard_use_expires_seconds) {
206 long tmo = xp->lft.hard_use_expires_seconds +
207 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
208 if (tmo <= 0)
209 goto expired;
210 if (tmo < next)
211 next = tmo;
212 }
213 if (xp->lft.soft_add_expires_seconds) {
214 long tmo = xp->lft.soft_add_expires_seconds +
215 xp->curlft.add_time - now;
216 if (tmo <= 0) {
217 warn = 1;
218 tmo = XFRM_KM_TIMEOUT;
219 }
220 if (tmo < next)
221 next = tmo;
222 }
223 if (xp->lft.soft_use_expires_seconds) {
224 long tmo = xp->lft.soft_use_expires_seconds +
225 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
226 if (tmo <= 0) {
227 warn = 1;
228 tmo = XFRM_KM_TIMEOUT;
229 }
230 if (tmo < next)
231 next = tmo;
232 }
233
234 if (warn)
6c5c8ca7 235 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 236 if (next != LONG_MAX &&
237 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
238 xfrm_pol_hold(xp);
239
240out:
241 read_unlock(&xp->lock);
242 xfrm_pol_put(xp);
243 return;
244
245expired:
246 read_unlock(&xp->lock);
4666faab 247 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 248 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 249 xfrm_pol_put(xp);
250}
251
1da177e4
LT		 252/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
253 * SPD calls.
254 */
255
0331b1f3 256struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 257{
258 struct xfrm_policy *policy;
259
0da974f4 260 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 261
262 if (policy) {
0331b1f3 263 write_pnet(&policy->xp_net, net);
12a169e7 264 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 265 INIT_HLIST_NODE(&policy->bydst);
266 INIT_HLIST_NODE(&policy->byidx);
1da177e4 267 rwlock_init(&policy->lock);
850a6212 268 refcount_set(&policy->refcnt, 1);
a0073fe1 269 skb_queue_head_init(&policy->polq.hold_queue);
c3aed709
KC		 270 timer_setup(&policy->timer, xfrm_policy_timer, 0);
271 timer_setup(&policy->polq.hold_timer,
272 xfrm_policy_queue_process, 0);
1da177e4
LT		 273 }
274 return policy;
275}
276EXPORT_SYMBOL(xfrm_policy_alloc);
277
56f04730
ED		 278static void xfrm_policy_destroy_rcu(struct rcu_head *head)
279{
280 struct xfrm_policy *policy = container_of(head, struct xfrm_policy, rcu);
281
282 security_xfrm_policy_free(policy->security);
283 kfree(policy);
284}
285
1da177e4
LT		 286/* Destroy xfrm_policy: descendant resources must be released to this moment. */
287
64c31b3f 288void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 289{
12a169e7 290 BUG_ON(!policy->walk.dead);
1da177e4 291
0659eea9 292 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 293 BUG();
294
56f04730 295 call_rcu(&policy->rcu, xfrm_policy_destroy_rcu);
1da177e4 296}
64c31b3f 297EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 298
1365e547 299/* Rule must be locked. Release descendant resources, announce
1da177e4
LT		 300 * entry dead. The rule must be unlinked from lists to the moment.
301 */
302
303static void xfrm_policy_kill(struct xfrm_policy *policy)
304{
12a169e7 305 policy->walk.dead = 1;
1da177e4 306
285ead17 307 atomic_inc(&policy->genid);
1da177e4 308
e7d8f6cb
SK		 309 if (del_timer(&policy->polq.hold_timer))
310 xfrm_pol_put(policy);
1ee5e667 311 skb_queue_purge(&policy->polq.hold_queue);
a0073fe1 312
285ead17
TT		 313 if (del_timer(&policy->timer))
314 xfrm_pol_put(policy);
315
316 xfrm_pol_put(policy);
1da177e4
LT		 317}
318
2518c7c2
DM		 319static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
320
e92303f8 321static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 322{
e92303f8 323 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 324}
325
b58555f1
CG		 326/* calculate policy hash thresholds */
327static void __get_hash_thresh(struct net *net,
328 unsigned short family, int dir,
329 u8 *dbits, u8 *sbits)
330{
331 switch (family) {
332 case AF_INET:
333 *dbits = net->xfrm.policy_bydst[dir].dbits4;
334 *sbits = net->xfrm.policy_bydst[dir].sbits4;
335 break;
336
337 case AF_INET6:
338 *dbits = net->xfrm.policy_bydst[dir].dbits6;
339 *sbits = net->xfrm.policy_bydst[dir].sbits6;
340 break;
341
342 default:
343 *dbits = 0;
344 *sbits = 0;
345 }
346}
347
5f803b58
DM		 348static struct hlist_head *policy_hash_bysel(struct net *net,
349 const struct xfrm_selector *sel,
350 unsigned short family, int dir)
2518c7c2 351{
1121994c 352 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 353 unsigned int hash;
354 u8 dbits;
355 u8 sbits;
356
357 __get_hash_thresh(net, family, dir, &dbits, &sbits);
358 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2 359
e1e551bc
FW		 360 if (hash == hmask + 1)
361 return &net->xfrm.policy_inexact[dir];
362
363 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
364 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM		 365}
366
5f803b58
DM		 367static struct hlist_head *policy_hash_direct(struct net *net,
368 const xfrm_address_t *daddr,
369 const xfrm_address_t *saddr,
370 unsigned short family, int dir)
2518c7c2 371{
1121994c 372 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 373 unsigned int hash;
374 u8 dbits;
375 u8 sbits;
376
377 __get_hash_thresh(net, family, dir, &dbits, &sbits);
378 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 379
e1e551bc
FW		 380 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
381 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM		 382}
383
b58555f1
CG		 384static void xfrm_dst_hash_transfer(struct net *net,
385 struct hlist_head *list,
2518c7c2 386 struct hlist_head *ndsttable,
b58555f1
CG		 387 unsigned int nhashmask,
388 int dir)
2518c7c2 389{
b67bfe0d 390 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 391 struct xfrm_policy *pol;
b791160b 392 unsigned int h0 = 0;
b58555f1
CG		 393 u8 dbits;
394 u8 sbits;
2518c7c2 395
b791160b 396redo:
b67bfe0d 397 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 398 unsigned int h;
399
b58555f1 400 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 401 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 402 pol->family, nhashmask, dbits, sbits);
b791160b 403 if (!entry0) {
a5eefc1d
FW		 404 hlist_del_rcu(&pol->bydst);
405 hlist_add_head_rcu(&pol->bydst, ndsttable + h);
b791160b
YH		 406 h0 = h;
407 } else {
408 if (h != h0)
409 continue;
a5eefc1d
FW		 410 hlist_del_rcu(&pol->bydst);
411 hlist_add_behind_rcu(&pol->bydst, entry0);
b791160b 412 }
b67bfe0d 413 entry0 = &pol->bydst;
b791160b
YH		 414 }
415 if (!hlist_empty(list)) {
416 entry0 = NULL;
417 goto redo;
2518c7c2
DM		 418 }
419}
420
421static void xfrm_idx_hash_transfer(struct hlist_head *list,
422 struct hlist_head *nidxtable,
423 unsigned int nhashmask)
424{
b67bfe0d 425 struct hlist_node *tmp;
2518c7c2
DM		 426 struct xfrm_policy *pol;
427
b67bfe0d 428 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 429 unsigned int h;
430
431 h = __idx_hash(pol->index, nhashmask);
432 hlist_add_head(&pol->byidx, nidxtable+h);
433 }
434}
435
436static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
437{
438 return ((old_hmask + 1) << 1) - 1;
439}
440
66caf628 441static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 442{
66caf628 443 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 444 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
445 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
44e36b42 446 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
e1e551bc 447 struct hlist_head *odst;
2518c7c2
DM		 448 int i;
449
450 if (!ndst)
451 return;
452
9d0380df 453 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
30846090
FW		 454 write_seqcount_begin(&xfrm_policy_hash_generation);
455
456 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
457 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
2518c7c2 458
e1e551bc
FW		 459 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
460 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
461
2518c7c2 462 for (i = hmask; i >= 0; i--)
b58555f1 463 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 464
e1e551bc 465 rcu_assign_pointer(net->xfrm.policy_bydst[dir].table, ndst);
66caf628 466 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 467
30846090 468 write_seqcount_end(&xfrm_policy_hash_generation);
9d0380df 469 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 470
e1e551bc
FW		 471 synchronize_rcu();
472
44e36b42 473 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 474}
475
66caf628 476static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 477{
66caf628 478 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 479 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
480 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 481 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 482 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 483 int i;
484
485 if (!nidx)
486 return;
487
9d0380df 488 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 489
490 for (i = hmask; i >= 0; i--)
491 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
492
66caf628
AD		 493 net->xfrm.policy_byidx = nidx;
494 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 495
9d0380df 496 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 497
44e36b42 498 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 499}
500
66caf628 501static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 502{
66caf628
AD		 503 unsigned int cnt = net->xfrm.policy_count[dir];
504 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 505
506 if (total)
507 *total += cnt;
508
509 if ((hmask + 1) < xfrm_policy_hashmax &&
510 cnt > hmask)
511 return 1;
512
513 return 0;
514}
515
66caf628 516static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 517{
66caf628 518 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 519
520 if ((hmask + 1) < xfrm_policy_hashmax &&
521 total > hmask)
522 return 1;
523
524 return 0;
525}
526
e071041b 527void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 528{
e071041b
AD		 529 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
530 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
531 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
532 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
533 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
534 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
535 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 536 si->spdhmcnt = xfrm_policy_hashmax;
ecfd6b18
JHS		 537}
538EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 539
ecfd6b18 540static DEFINE_MUTEX(hash_resize_mutex);
66caf628 541static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 542{
66caf628 543 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 544 int dir, total;
545
546 mutex_lock(&hash_resize_mutex);
547
548 total = 0;
53c2e285 549 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD		 550 if (xfrm_bydst_should_resize(net, dir, &total))
551 xfrm_bydst_resize(net, dir);
2518c7c2 552 }
66caf628
AD		 553 if (xfrm_byidx_should_resize(net, total))
554 xfrm_byidx_resize(net, total);
2518c7c2
DM		 555
556 mutex_unlock(&hash_resize_mutex);
557}
558
880a6fab
CG		 559static void xfrm_hash_rebuild(struct work_struct *work)
560{
561 struct net *net = container_of(work, struct net,
562 xfrm.policy_hthresh.work);
563 unsigned int hmask;
564 struct xfrm_policy *pol;
565 struct xfrm_policy *policy;
566 struct hlist_head *chain;
567 struct hlist_head *odst;
568 struct hlist_node *newpos;
569 int i;
570 int dir;
571 unsigned seq;
572 u8 lbits4, rbits4, lbits6, rbits6;
573
574 mutex_lock(&hash_resize_mutex);
575
576 /* read selector prefixlen thresholds */
577 do {
578 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
579
580 lbits4 = net->xfrm.policy_hthresh.lbits4;
581 rbits4 = net->xfrm.policy_hthresh.rbits4;
582 lbits6 = net->xfrm.policy_hthresh.lbits6;
583 rbits6 = net->xfrm.policy_hthresh.rbits6;
584 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
585
9d0380df 586 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 587
588 /* reset the bydst and inexact table in all directions */
53c2e285 589 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG		 590 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
591 hmask = net->xfrm.policy_bydst[dir].hmask;
592 odst = net->xfrm.policy_bydst[dir].table;
593 for (i = hmask; i >= 0; i--)
594 INIT_HLIST_HEAD(odst + i);
595 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
596 /* dir out => dst = remote, src = local */
597 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
598 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
599 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
600 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
601 } else {
602 /* dir in/fwd => dst = local, src = remote */
603 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
604 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
605 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
606 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
607 }
608 }
609
610 /* re-insert all policies by order of creation */
611 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
6916fb3b
TB		 612 if (xfrm_policy_id2dir(policy->index) >= XFRM_POLICY_MAX) {
613 /* skip socket policies */
614 continue;
615 }
880a6fab
CG		 616 newpos = NULL;
617 chain = policy_hash_bysel(net, &policy->selector,
618 policy->family,
619 xfrm_policy_id2dir(policy->index));
620 hlist_for_each_entry(pol, chain, bydst) {
621 if (policy->priority >= pol->priority)
622 newpos = &pol->bydst;
623 else
624 break;
625 }
626 if (newpos)
627 hlist_add_behind(&policy->bydst, newpos);
628 else
629 hlist_add_head(&policy->bydst, chain);
630 }
631
9d0380df 632 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 633
634 mutex_unlock(&hash_resize_mutex);
635}
636
637void xfrm_policy_hash_rebuild(struct net *net)
638{
639 schedule_work(&net->xfrm.policy_hthresh.work);
640}
641EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
642
1da177e4
LT		 643/* Generate new index... KAME seems to generate them ordered by cost
644 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 645static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 646{
1da177e4
LT		 647 static u32 idx_generator;
648
649 for (;;) {
2518c7c2
DM		 650 struct hlist_head *list;
651 struct xfrm_policy *p;
652 u32 idx;
653 int found;
654
e682adf0
FD		 655 if (!index) {
656 idx = (idx_generator | dir);
657 idx_generator += 8;
658 } else {
659 idx = index;
660 index = 0;
661 }
662
1da177e4
LT		 663 if (idx == 0)
664 idx = 8;
1121994c 665 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 666 found = 0;
b67bfe0d 667 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 668 if (p->index == idx) {
669 found = 1;
1da177e4 670 break;
2518c7c2 671 }
1da177e4 672 }
2518c7c2 673 if (!found)
1da177e4
LT		 674 return idx;
675 }
676}
677
2518c7c2
DM		 678static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
679{
680 u32 *p1 = (u32 *) s1;
681 u32 *p2 = (u32 *) s2;
682 int len = sizeof(struct xfrm_selector) / sizeof(u32);
683 int i;
684
685 for (i = 0; i < len; i++) {
686 if (p1[i] != p2[i])
687 return 1;
688 }
689
690 return 0;
691}
692
a0073fe1
SK		 693static void xfrm_policy_requeue(struct xfrm_policy *old,
694 struct xfrm_policy *new)
695{
696 struct xfrm_policy_queue *pq = &old->polq;
697 struct sk_buff_head list;
698
de2ad486
LR		 699 if (skb_queue_empty(&pq->hold_queue))
700 return;
701
a0073fe1
SK		 702 __skb_queue_head_init(&list);
703
704 spin_lock_bh(&pq->hold_queue.lock);
705 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 706 if (del_timer(&pq->hold_timer))
707 xfrm_pol_put(old);
a0073fe1
SK		 708 spin_unlock_bh(&pq->hold_queue.lock);
709
a0073fe1
SK		 710 pq = &new->polq;
711
712 spin_lock_bh(&pq->hold_queue.lock);
713 skb_queue_splice(&list, &pq->hold_queue);
714 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 715 if (!mod_timer(&pq->hold_timer, jiffies))
716 xfrm_pol_hold(new);
a0073fe1
SK		 717 spin_unlock_bh(&pq->hold_queue.lock);
718}
719
7cb8a939
SK		 720static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
721 struct xfrm_policy *pol)
722{
723 u32 mark = policy->mark.v & policy->mark.m;
724
725 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
726 return true;
727
728 if ((mark & pol->mark.m) == pol->mark.v &&
729 policy->priority == pol->priority)
730 return true;
731
732 return false;
733}
734
1da177e4
LT		 735int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
736{
1121994c 737 struct net *net = xp_net(policy);
2518c7c2
DM		 738 struct xfrm_policy *pol;
739 struct xfrm_policy *delpol;
740 struct hlist_head *chain;
b67bfe0d 741 struct hlist_node *newpos;
1da177e4 742
9d0380df 743 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 744 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 745 delpol = NULL;
746 newpos = NULL;
b67bfe0d 747 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 748 if (pol->type == policy->type &&
2518c7c2 749 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 750 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 751 xfrm_sec_ctx_match(pol->security, policy->security) &&
752 !WARN_ON(delpol)) {
1da177e4 753 if (excl) {
9d0380df 754 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 755 return -EEXIST;
756 }
1da177e4
LT		 757 delpol = pol;
758 if (policy->priority > pol->priority)
759 continue;
760 } else if (policy->priority >= pol->priority) {
a6c7ab55 761 newpos = &pol->bydst;
1da177e4
LT		 762 continue;
763 }
1da177e4
LT		 764 if (delpol)
765 break;
1da177e4
LT		 766 }
767 if (newpos)
1d023284 768 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 769 else
770 hlist_add_head(&policy->bydst, chain);
12bfa8bd 771 __xfrm_policy_link(policy, dir);
ca4c3fc2 772
773 /* After previous checking, family can either be AF_INET or AF_INET6 */
774 if (policy->family == AF_INET)
775 rt_genid_bump_ipv4(net);
776 else
777 rt_genid_bump_ipv6(net);
778
a0073fe1
SK		 779 if (delpol) {
780 xfrm_policy_requeue(delpol, policy);
29fa0b30 781 __xfrm_policy_unlink(delpol, dir);
a0073fe1 782 }
e682adf0 783 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 784 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 785 policy->curlft.add_time = get_seconds();
1da177e4
LT		 786 policy->curlft.use_time = 0;
787 if (!mod_timer(&policy->timer, jiffies + HZ))
788 xfrm_pol_hold(policy);
9d0380df 789 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 790
9b78a82c 791 if (delpol)
1da177e4 792 xfrm_policy_kill(delpol);
1121994c
AD		 793 else if (xfrm_bydst_should_resize(net, dir, NULL))
794 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 795
1da177e4
LT		 796 return 0;
797}
798EXPORT_SYMBOL(xfrm_policy_insert);
799
8ca2e93b
JHS		 800struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
801 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 802 struct xfrm_sec_ctx *ctx, int delete,
803 int *err)
1da177e4 804{
2518c7c2
DM		 805 struct xfrm_policy *pol, *ret;
806 struct hlist_head *chain;
1da177e4 807
ef41aaa0 808 *err = 0;
9d0380df 809 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 810 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 811 ret = NULL;
b67bfe0d 812 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 813 if (pol->type == type &&
34f8d884 814 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 815 !selector_cmp(sel, &pol->selector) &&
816 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 817 xfrm_pol_hold(pol);
2518c7c2 818 if (delete) {
03e1ad7b
PM		 819 *err = security_xfrm_policy_delete(
820 pol->security);
ef41aaa0 821 if (*err) {
9d0380df 822 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 823 return pol;
824 }
29fa0b30 825 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 826 }
827 ret = pol;
1da177e4
LT		 828 break;
829 }
830 }
9d0380df 831 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 832
fe1a5f03 833 if (ret && delete)
2518c7c2 834 xfrm_policy_kill(ret);
2518c7c2 835 return ret;
1da177e4 836}
df71837d 837EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 838
8ca2e93b
JHS		 839struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
840 int dir, u32 id, int delete, int *err)
1da177e4 841{
2518c7c2
DM		 842 struct xfrm_policy *pol, *ret;
843 struct hlist_head *chain;
1da177e4 844
b5505c6e
HX		 845 *err = -ENOENT;
846 if (xfrm_policy_id2dir(id) != dir)
847 return NULL;
848
ef41aaa0 849 *err = 0;
9d0380df 850 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 851 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 852 ret = NULL;
b67bfe0d 853 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 854 if (pol->type == type && pol->index == id &&
855 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 856 xfrm_pol_hold(pol);
2518c7c2 857 if (delete) {
03e1ad7b
PM		 858 *err = security_xfrm_policy_delete(
859 pol->security);
ef41aaa0 860 if (*err) {
9d0380df 861 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 862 return pol;
863 }
29fa0b30 864 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 865 }
866 ret = pol;
1da177e4
LT		 867 break;
868 }
869 }
9d0380df 870 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 871
fe1a5f03 872 if (ret && delete)
2518c7c2 873 xfrm_policy_kill(ret);
2518c7c2 874 return ret;
1da177e4
LT		 875}
876EXPORT_SYMBOL(xfrm_policy_byid);
877
4aa2e62c
JL		 878#ifdef CONFIG_SECURITY_NETWORK_XFRM
879static inline int
2e71029e 880xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 881{
4aa2e62c
JL		 882 int dir, err = 0;
883
884 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
885 struct xfrm_policy *pol;
4aa2e62c
JL		 886 int i;
887
b67bfe0d 888 hlist_for_each_entry(pol,
33ffbbd5 889 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 890 if (pol->type != type)
891 continue;
03e1ad7b 892 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 893 if (err) {
2e71029e 894 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 895 return err;
896 }
7dc12d6d 897 }
33ffbbd5 898 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 899 hlist_for_each_entry(pol,
33ffbbd5 900 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 901 bydst) {
902 if (pol->type != type)
903 continue;
03e1ad7b
PM		 904 err = security_xfrm_policy_delete(
905 pol->security);
4aa2e62c 906 if (err) {
ab5f5e8b 907 xfrm_audit_policy_delete(pol, 0,
2e71029e 908 task_valid);
4aa2e62c
JL		 909 return err;
910 }
911 }
912 }
913 }
914 return err;
915}
916#else
917static inline int
2e71029e 918xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 919{
920 return 0;
921}
922#endif
923
2e71029e 924int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 925{
2f1eb65f 926 int dir, err = 0, cnt = 0;
1da177e4 927
9d0380df 928 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 929
2e71029e 930 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 931 if (err)
932 goto out;
933
1da177e4 934 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 935 struct xfrm_policy *pol;
29fa0b30 936 int i;
2518c7c2
DM		 937
938 again1:
b67bfe0d 939 hlist_for_each_entry(pol,
33ffbbd5 940 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 941 if (pol->type != type)
942 continue;
ea2dea9d 943 __xfrm_policy_unlink(pol, dir);
9d0380df 944 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 945 cnt++;
1da177e4 946
2e71029e 947 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 948
2518c7c2 949 xfrm_policy_kill(pol);
1da177e4 950
4141b36a 951 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 952 goto again1;
953 }
954
33ffbbd5 955 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 956 again2:
b67bfe0d 957 hlist_for_each_entry(pol,
33ffbbd5 958 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 959 bydst) {
960 if (pol->type != type)
961 continue;
ea2dea9d 962 __xfrm_policy_unlink(pol, dir);
9d0380df 963 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 964 cnt++;
2518c7c2 965
2e71029e 966 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 967 xfrm_policy_kill(pol);
968
9d0380df 969 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 970 goto again2;
971 }
1da177e4 972 }
2518c7c2 973
1da177e4 974 }
2f1eb65f
JHS		 975 if (!cnt)
976 err = -ESRCH;
ec30d78c
FW		 977 else
978 xfrm_policy_cache_flush();
4aa2e62c 979out:
9d0380df 980 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 981 return err;
1da177e4
LT		 982}
983EXPORT_SYMBOL(xfrm_policy_flush);
984
cdcbca7c 985int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 986 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 987 void *data)
988{
12a169e7
HX		 989 struct xfrm_policy *pol;
990 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 991 int error = 0;
992
993 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
994 walk->type != XFRM_POLICY_TYPE_ANY)
995 return -EINVAL;
1da177e4 996
12a169e7 997 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 998 return 0;
999
9d0380df 1000 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1001 if (list_empty(&walk->walk.all))
cdcbca7c 1002 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1003 else
80077702
LR		 1004 x = list_first_entry(&walk->walk.all,
1005 struct xfrm_policy_walk_entry, all);
1006
cdcbca7c 1007 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1008 if (x->dead)
4c563f76 1009 continue;
12a169e7
HX		 1010 pol = container_of(x, struct xfrm_policy, walk);
1011 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1012 walk->type != pol->type)
1013 continue;
1014 error = func(pol, xfrm_policy_id2dir(pol->index),
1015 walk->seq, data);
1016 if (error) {
1017 list_move_tail(&walk->walk.all, &x->all);
1018 goto out;
2518c7c2 1019 }
12a169e7 1020 walk->seq++;
1da177e4 1021 }
12a169e7 1022 if (walk->seq == 0) {
baf5d743
JHS		 1023 error = -ENOENT;
1024 goto out;
1025 }
12a169e7 1026 list_del_init(&walk->walk.all);
1da177e4 1027out:
9d0380df 1028 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1029 return error;
1030}
1031EXPORT_SYMBOL(xfrm_policy_walk);
1032
12a169e7
HX		 1033void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1034{
1035 INIT_LIST_HEAD(&walk->walk.all);
1036 walk->walk.dead = 1;
1037 walk->type = type;
1038 walk->seq = 0;
1039}
1040EXPORT_SYMBOL(xfrm_policy_walk_init);
1041
283bc9f3 1042void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1043{
1044 if (list_empty(&walk->walk.all))
1045 return;
1046
9d0380df 1047 spin_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1048 list_del(&walk->walk.all);
9d0380df 1049 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1050}
1051EXPORT_SYMBOL(xfrm_policy_walk_done);
1052
134b0fc5
JM		 1053/*
1054 * Find policy to apply to this flow.
1055 *
1056 * Returns 0 if policy found, else an -errno.
1057 */
f299d557
DM		 1058static int xfrm_policy_match(const struct xfrm_policy *pol,
1059 const struct flowi *fl,
2518c7c2 1060 u8 type, u16 family, int dir)
1da177e4 1061{
f299d557 1062 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1063 int ret = -ESRCH;
1064 bool match;
1da177e4 1065
2518c7c2 1066 if (pol->family != family ||
1d28f42c 1067 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1068 pol->type != type)
134b0fc5 1069 return ret;
1da177e4 1070
2518c7c2 1071 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1072 if (match)
1d28f42c 1073 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1074 dir);
2518c7c2 1075
134b0fc5 1076 return ret;
2518c7c2 1077}
1da177e4 1078
52479b62 1079static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1080 const struct flowi *fl,
2518c7c2
DM		 1081 u16 family, u8 dir)
1082{
134b0fc5 1083 int err;
2518c7c2 1084 struct xfrm_policy *pol, *ret;
0b597e7e 1085 const xfrm_address_t *daddr, *saddr;
2518c7c2 1086 struct hlist_head *chain;
30846090
FW		 1087 unsigned int sequence;
1088 u32 priority;
df71837d 1089
2518c7c2
DM		 1090 daddr = xfrm_flowi_daddr(fl, family);
1091 saddr = xfrm_flowi_saddr(fl, family);
1092 if (unlikely(!daddr || !saddr))
1093 return NULL;
1094
a7c44247 1095 rcu_read_lock();
30846090
FW		 1096 retry:
1097 do {
1098 sequence = read_seqcount_begin(&xfrm_policy_hash_generation);
1099 chain = policy_hash_direct(net, daddr, saddr, family, dir);
1100 } while (read_seqcount_retry(&xfrm_policy_hash_generation, sequence));
1101
1102 priority = ~0U;
2518c7c2 1103 ret = NULL;
a5eefc1d 1104 hlist_for_each_entry_rcu(pol, chain, bydst) {
134b0fc5
JM		 1105 err = xfrm_policy_match(pol, fl, type, family, dir);
1106 if (err) {
1107 if (err == -ESRCH)
1108 continue;
1109 else {
1110 ret = ERR_PTR(err);
1111 goto fail;
1112 }
1113 } else {
2518c7c2 1114 ret = pol;
acba48e1 1115 priority = ret->priority;
2518c7c2
DM		 1116 break;
1117 }
1118 }
52479b62 1119 chain = &net->xfrm.policy_inexact[dir];
a5eefc1d 1120 hlist_for_each_entry_rcu(pol, chain, bydst) {
8faf491e
LR		 1121 if ((pol->priority >= priority) && ret)
1122 break;
1123
134b0fc5
JM		 1124 err = xfrm_policy_match(pol, fl, type, family, dir);
1125 if (err) {
1126 if (err == -ESRCH)
1127 continue;
1128 else {
1129 ret = ERR_PTR(err);
1130 goto fail;
1131 }
8faf491e 1132 } else {
acba48e1
DM		 1133 ret = pol;
1134 break;
1da177e4
LT		 1135 }
1136 }
586f2eb4 1137
30846090
FW		 1138 if (read_seqcount_retry(&xfrm_policy_hash_generation, sequence))
1139 goto retry;
1140
e37cc8ad
FW		 1141 if (ret && !xfrm_pol_hold_rcu(ret))
1142 goto retry;
134b0fc5 1143fail:
a7c44247 1144 rcu_read_unlock();
4e81bb83 1145
2518c7c2 1146 return ret;
4e81bb83
MN		 1147}
1148
80c802f3 1149static struct xfrm_policy *
86dc8ee0 1150xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1151{
1152#ifdef CONFIG_XFRM_SUB_POLICY
1153 struct xfrm_policy *pol;
1154
1155 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1156 if (pol != NULL)
1157 return pol;
1158#endif
1159 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1160}
1161
6f9c9615 1162static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir,
4c86d777 1163 const struct flowi *fl, u16 family)
1da177e4
LT		 1164{
1165 struct xfrm_policy *pol;
1166
d188ba86 1167 rcu_read_lock();
ae33786f 1168 again:
d188ba86
ED		 1169 pol = rcu_dereference(sk->sk_policy[dir]);
1170 if (pol != NULL) {
4c86d777 1171 bool match = xfrm_selector_match(&pol->selector, fl, family);
a716c119 1172 int err = 0;
df71837d 1173
3bccfbc7 1174 if (match) {
34f8d884
JHS		 1175 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1176 pol = NULL;
1177 goto out;
1178 }
03e1ad7b 1179 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1180 fl->flowi_secid,
aff669bc 1181 dir);
330e832a
FW		 1182 if (!err) {
1183 if (!xfrm_pol_hold_rcu(pol))
1184 goto again;
1185 } else if (err == -ESRCH) {
3bccfbc7 1186 pol = NULL;
330e832a 1187 } else {
3bccfbc7 1188 pol = ERR_PTR(err);
330e832a 1189 }
3bccfbc7 1190 } else
1da177e4
LT		 1191 pol = NULL;
1192 }
34f8d884 1193out:
d188ba86 1194 rcu_read_unlock();
1da177e4
LT		 1195 return pol;
1196}
1197
1198static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1199{
98806f75 1200 struct net *net = xp_net(pol);
4e81bb83 1201
98806f75 1202 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1203 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1204 xfrm_pol_hold(pol);
1205}
1206
1207static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1208 int dir)
1209{
98806f75
AD		 1210 struct net *net = xp_net(pol);
1211
53c2e285 1212 if (list_empty(&pol->walk.all))
2518c7c2 1213 return NULL;
1da177e4 1214
53c2e285
HX		 1215 /* Socket policies are not hashed. */
1216 if (!hlist_unhashed(&pol->bydst)) {
a5eefc1d 1217 hlist_del_rcu(&pol->bydst);
53c2e285
HX		 1218 hlist_del(&pol->byidx);
1219 }
1220
1221 list_del_init(&pol->walk.all);
98806f75 1222 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1223
1224 return pol;
1da177e4
LT		 1225}
1226
53c2e285
HX		 1227static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1228{
1229 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1230}
1231
1232static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1233{
1234 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1235}
1236
4666faab 1237int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1238{
283bc9f3
FD		 1239 struct net *net = xp_net(pol);
1240
9d0380df 1241 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1242 pol = __xfrm_policy_unlink(pol, dir);
9d0380df 1243 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1244 if (pol) {
1da177e4 1245 xfrm_policy_kill(pol);
4666faab 1246 return 0;
1da177e4 1247 }
4666faab 1248 return -ENOENT;
1da177e4 1249}
a70fcb0b 1250EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1251
1252int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1253{
1121994c 1254 struct net *net = xp_net(pol);
1da177e4
LT		 1255 struct xfrm_policy *old_pol;
1256
4e81bb83
MN		 1257#ifdef CONFIG_XFRM_SUB_POLICY
1258 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1259 return -EINVAL;
1260#endif
1261
9d0380df 1262 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
d188ba86
ED		 1263 old_pol = rcu_dereference_protected(sk->sk_policy[dir],
1264 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
1da177e4 1265 if (pol) {
9d729f72 1266 pol->curlft.add_time = get_seconds();
e682adf0 1267 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1268 xfrm_sk_policy_link(pol, dir);
1da177e4 1269 }
d188ba86 1270 rcu_assign_pointer(sk->sk_policy[dir], pol);
a0073fe1
SK		 1271 if (old_pol) {
1272 if (pol)
1273 xfrm_policy_requeue(old_pol, pol);
1274
ea2dea9d
TT		 1275 /* Unlinking succeeds always. This is the only function
1276 * allowed to delete or replace socket policy.
1277 */
53c2e285 1278 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1279 }
9d0380df 1280 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1281
1282 if (old_pol) {
1283 xfrm_policy_kill(old_pol);
1284 }
1285 return 0;
1286}
1287
d3e40a9f 1288static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1289{
0331b1f3 1290 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1291 struct net *net = xp_net(old);
1da177e4
LT		 1292
1293 if (newp) {
1294 newp->selector = old->selector;
03e1ad7b
PM		 1295 if (security_xfrm_policy_clone(old->security,
1296 &newp->security)) {
df71837d
TJ		 1297 kfree(newp);
1298 return NULL; /* ENOMEM */
1299 }
1da177e4
LT		 1300 newp->lft = old->lft;
1301 newp->curlft = old->curlft;
fb977e2c 1302 newp->mark = old->mark;
1da177e4
LT		 1303 newp->action = old->action;
1304 newp->flags = old->flags;
1305 newp->xfrm_nr = old->xfrm_nr;
1306 newp->index = old->index;
4e81bb83 1307 newp->type = old->type;
0e74aa1d 1308 newp->family = old->family;
1da177e4
LT		 1309 memcpy(newp->xfrm_vec, old->xfrm_vec,
1310 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
9d0380df 1311 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1312 xfrm_sk_policy_link(newp, dir);
9d0380df 1313 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1314 xfrm_pol_put(newp);
1315 }
1316 return newp;
1317}
1318
d188ba86 1319int __xfrm_sk_clone_policy(struct sock *sk, const struct sock *osk)
1da177e4 1320{
d188ba86
ED		 1321 const struct xfrm_policy *p;
1322 struct xfrm_policy *np;
1323 int i, ret = 0;
1324
1325 rcu_read_lock();
1326 for (i = 0; i < 2; i++) {
1327 p = rcu_dereference(osk->sk_policy[i]);
1328 if (p) {
1329 np = clone_policy(p, i);
1330 if (unlikely(!np)) {
1331 ret = -ENOMEM;
1332 break;
1333 }
1334 rcu_assign_pointer(sk->sk_policy[i], np);
1335 }
1336 }
1337 rcu_read_unlock();
1338 return ret;
1da177e4
LT		 1339}
1340
a1e59abf 1341static int
42a7b32b 1342xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local,
077fbac4 1343 xfrm_address_t *remote, unsigned short family, u32 mark)
a1e59abf
PM		 1344{
1345 int err;
37b10383 1346 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
a1e59abf
PM		 1347
1348 if (unlikely(afinfo == NULL))
1349 return -EINVAL;
077fbac4 1350 err = afinfo->get_saddr(net, oif, local, remote, mark);
bdba9fe0 1351 rcu_read_unlock();
a1e59abf
PM		 1352 return err;
1353}
1354
1da177e4
LT		 1355/* Resolve list of templates for the flow, given policy. */
1356
1357static int
a6c2e611
DM		 1358xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1359 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1360{
fbda33b2 1361 struct net *net = xp_net(policy);
1da177e4
LT		 1362 int nx;
1363 int i, error;
94802151
SK		 1364 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1365 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1366 xfrm_address_t tmp;
1da177e4 1367
9b7a787d 1368 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4 1369 struct xfrm_state *x;
94802151
SK		 1370 xfrm_address_t *remote = daddr;
1371 xfrm_address_t *local = saddr;
1da177e4
LT		 1372 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1373
94802151
SK		 1374 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1375 tmpl->mode == XFRM_MODE_BEET) {
1376 remote = &tmpl->id.daddr;
1377 local = &tmpl->saddr;
1378 if (xfrm_addr_any(local, tmpl->encap_family)) {
1379 error = xfrm_get_saddr(net, fl->flowi_oif,
1380 &tmp, remote,
1381 tmpl->encap_family, 0);
1382 if (error)
1383 goto fail;
1384 local = &tmp;
1385 }
1da177e4
LT		 1386 }
1387
1388 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1389
1390 if (x && x->km.state == XFRM_STATE_VALID) {
1391 xfrm[nx++] = x;
94802151
SK		 1392 daddr = remote;
1393 saddr = local;
1da177e4
LT		 1394 continue;
1395 }
1396 if (x) {
1397 error = (x->km.state == XFRM_STATE_ERROR ?
1398 -EINVAL : -EAGAIN);
1399 xfrm_state_put(x);
42054569 1400 } else if (error == -ESRCH) {
a4322266 1401 error = -EAGAIN;
42054569 1402 }
1da177e4
LT		 1403
1404 if (!tmpl->optional)
1405 goto fail;
1406 }
1407 return nx;
1408
1409fail:
9b7a787d 1410 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1411 xfrm_state_put(xfrm[nx]);
1412 return error;
1413}
1414
4e81bb83 1415static int
a6c2e611
DM		 1416xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1417 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1418{
41a49cc3
MN		 1419 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1420 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1421 int cnx = 0;
1422 int error;
1423 int ret;
1424 int i;
1425
1426 for (i = 0; i < npols; i++) {
1427 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1428 error = -ENOBUFS;
1429 goto fail;
1430 }
41a49cc3
MN		 1431
1432 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1433 if (ret < 0) {
1434 error = ret;
1435 goto fail;
1436 } else
1437 cnx += ret;
1438 }
1439
41a49cc3
MN		 1440 /* found states are sorted for outbound processing */
1441 if (npols > 1)
1442 xfrm_state_sort(xfrm, tpp, cnx, family);
1443
4e81bb83
MN		 1444 return cnx;
1445
1446 fail:
9b7a787d 1447 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1448 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1449 return error;
1450
1451}
1452
f5e2bb4f 1453static int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286 1454{
37b10383 1455 const struct xfrm_policy_afinfo *afinfo;
f5e2bb4f 1456 int tos = 0;
1da177e4 1457
f5e2bb4f
FW		 1458 afinfo = xfrm_policy_get_afinfo(family);
1459 tos = afinfo ? afinfo->get_tos(fl) : 0;
25ee3286 1460
bdba9fe0 1461 rcu_read_unlock();
25ee3286
HX		 1462
1463 return tos;
1464}
1465
d7c7544c 1466static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1467{
37b10383 1468 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1469 struct dst_ops *dst_ops;
25ee3286
HX		 1470 struct xfrm_dst *xdst;
1471
1472 if (!afinfo)
1473 return ERR_PTR(-EINVAL);
1474
d7c7544c
AD		 1475 switch (family) {
1476 case AF_INET:
1477 dst_ops = &net->xfrm.xfrm4_dst_ops;
1478 break;
dfd56b8b 1479#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1480 case AF_INET6:
1481 dst_ops = &net->xfrm.xfrm6_dst_ops;
1482 break;
1483#endif
1484 default:
1485 BUG();
1486 }
b2a9c0ed 1487 xdst = dst_alloc(dst_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
25ee3286 1488
d4cae562 1489 if (likely(xdst)) {
141e369d
SK		 1490 struct dst_entry *dst = &xdst->u.dst;
1491
1492 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
d4cae562 1493 } else
0b150932 1494 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1495
bdba9fe0 1496 rcu_read_unlock();
d4cae562 1497
25ee3286
HX		 1498 return xdst;
1499}
1500
a1b05140
MN		 1501static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1502 int nfheader_len)
1503{
37b10383 1504 const struct xfrm_policy_afinfo *afinfo =
a1b05140
MN		 1505 xfrm_policy_get_afinfo(dst->ops->family);
1506 int err;
1507
1508 if (!afinfo)
1509 return -EINVAL;
1510
1511 err = afinfo->init_path(path, dst, nfheader_len);
1512
bdba9fe0 1513 rcu_read_unlock();
a1b05140
MN		 1514
1515 return err;
1516}
1517
87c1e12b 1518static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1519 const struct flowi *fl)
25ee3286 1520{
37b10383 1521 const struct xfrm_policy_afinfo *afinfo =
25ee3286
HX		 1522 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1523 int err;
1524
1525 if (!afinfo)
1da177e4 1526 return -EINVAL;
25ee3286 1527
87c1e12b 1528 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1529
bdba9fe0 1530 rcu_read_unlock();
25ee3286 1531
1da177e4
LT		 1532 return err;
1533}
1534
80c802f3 1535
25ee3286
HX		 1536/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1537 * all the metrics... Shortly, bundle a bundle.
1538 */
1539
1540static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1541 struct xfrm_state **xfrm, int nx,
98313ada 1542 const struct flowi *fl,
25ee3286
HX		 1543 struct dst_entry *dst)
1544{
d7c7544c 1545 struct net *net = xp_net(policy);
25ee3286
HX		 1546 unsigned long now = jiffies;
1547 struct net_device *dev;
43a4dea4 1548 struct xfrm_mode *inner_mode;
45b018be
DM		 1549 struct xfrm_dst *xdst_prev = NULL;
1550 struct xfrm_dst *xdst0 = NULL;
25ee3286
HX		 1551 int i = 0;
1552 int err;
1553 int header_len = 0;
a1b05140 1554 int nfheader_len = 0;
25ee3286
HX		 1555 int trailer_len = 0;
1556 int tos;
1557 int family = policy->selector.family;
9bb182a7
YH		 1558 xfrm_address_t saddr, daddr;
1559
1560 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1561
1562 tos = xfrm_get_tos(fl, family);
25ee3286
HX		 1563
1564 dst_hold(dst);
1565
1566 for (; i < nx; i++) {
d7c7544c 1567 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1568 struct dst_entry *dst1 = &xdst->u.dst;
1569
1570 err = PTR_ERR(xdst);
1571 if (IS_ERR(xdst)) {
1572 dst_release(dst);
1573 goto put_states;
1574 }
1575
45b018be
DM		 1576 if (!xdst_prev)
1577 xdst0 = xdst;
10a7ef33
DM		 1578 else
1579 /* Ref count is taken during xfrm_alloc_dst()
1580 * No need to do dst_clone() on dst1
1581 */
45b018be 1582 xfrm_dst_set_child(xdst_prev, &xdst->u.dst);
10a7ef33 1583
43a4dea4
SK		 1584 if (xfrm[i]->sel.family == AF_UNSPEC) {
1585 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1586 xfrm_af2proto(family));
1587 if (!inner_mode) {
1588 err = -EAFNOSUPPORT;
1589 dst_release(dst);
1590 goto put_states;
1591 }
1592 } else
1593 inner_mode = xfrm[i]->inner_mode;
1594
25ee3286 1595 xdst->route = dst;
defb3519 1596 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1597
1598 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1599 family = xfrm[i]->props.family;
42a7b32b 1600 dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
077fbac4
LC		 1601 &saddr, &daddr, family,
1602 xfrm[i]->props.output_mark);
25ee3286
HX		 1603 err = PTR_ERR(dst);
1604 if (IS_ERR(dst))
1605 goto put_states;
1606 } else
1607 dst_hold(dst);
1608
1609 dst1->xfrm = xfrm[i];
80c802f3 1610 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1611
f5b0a874 1612 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1613 dst1->flags |= DST_HOST;
1614 dst1->lastuse = now;
1615
1616 dst1->input = dst_discard;
43a4dea4 1617 dst1->output = inner_mode->afinfo->output;
25ee3286 1618
45b018be
DM		 1619 dst1->next = &xdst_prev->u.dst;
1620 xdst_prev = xdst;
25ee3286
HX		 1621
1622 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1623 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1624 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1625 trailer_len += xfrm[i]->props.trailer_len;
1626 }
1627
45b018be 1628 xfrm_dst_set_child(xdst_prev, dst);
0f6c480f 1629 xdst0->path = dst;
25ee3286
HX		 1630
1631 err = -ENODEV;
1632 dev = dst->dev;
1633 if (!dev)
1634 goto free_dst;
1635
45b018be
DM		 1636 xfrm_init_path(xdst0, dst, nfheader_len);
1637 xfrm_init_pmtu(&xdst_prev->u.dst);
25ee3286 1638
45b018be
DM		 1639 for (xdst_prev = xdst0; xdst_prev != (struct xfrm_dst *)dst;
1640 xdst_prev = (struct xfrm_dst *) xfrm_dst_child(&xdst_prev->u.dst)) {
1641 err = xfrm_fill_dst(xdst_prev, dev, fl);
25ee3286
HX		 1642 if (err)
1643 goto free_dst;
1644
45b018be
DM		 1645 xdst_prev->u.dst.header_len = header_len;
1646 xdst_prev->u.dst.trailer_len = trailer_len;
1647 header_len -= xdst_prev->u.dst.xfrm->props.header_len;
1648 trailer_len -= xdst_prev->u.dst.xfrm->props.trailer_len;
25ee3286
HX		 1649 }
1650
1651out:
45b018be 1652 return &xdst0->u.dst;
25ee3286
HX		 1653
1654put_states:
1655 for (; i < nx; i++)
1656 xfrm_state_put(xfrm[i]);
1657free_dst:
45b018be
DM		 1658 if (xdst0)
1659 dst_release_immediate(&xdst0->u.dst);
1660 xdst0 = ERR_PTR(err);
25ee3286
HX		 1661 goto out;
1662}
1663
73ff93cd 1664static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1665 struct xfrm_policy **pols,
1666 int *num_pols, int *num_xfrms)
1667{
1668 int i;
1669
1670 if (*num_pols == 0 || !pols[0]) {
1671 *num_pols = 0;
1672 *num_xfrms = 0;
1673 return 0;
1674 }
1675 if (IS_ERR(pols[0]))
1676 return PTR_ERR(pols[0]);
1677
1678 *num_xfrms = pols[0]->xfrm_nr;
1679
1680#ifdef CONFIG_XFRM_SUB_POLICY
1681 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1682 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1683 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1684 XFRM_POLICY_TYPE_MAIN,
1685 fl, family,
1686 XFRM_POLICY_OUT);
1687 if (pols[1]) {
1688 if (IS_ERR(pols[1])) {
1689 xfrm_pols_put(pols, *num_pols);
1690 return PTR_ERR(pols[1]);
1691 }
02d0892f 1692 (*num_pols)++;
80c802f3
TT		 1693 (*num_xfrms) += pols[1]->xfrm_nr;
1694 }
1695 }
1696#endif
1697 for (i = 0; i < *num_pols; i++) {
1698 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1699 *num_xfrms = -1;
1700 break;
1701 }
1702 }
1703
1704 return 0;
1705
1706}
1707
ec30d78c
FW		 1708static void xfrm_last_dst_update(struct xfrm_dst *xdst, struct xfrm_dst *old)
1709{
1710 this_cpu_write(xfrm_last_dst, xdst);
1711 if (old)
1712 dst_release(&old->u.dst);
1713}
1714
1715static void __xfrm_pcpu_work_fn(void)
1716{
1717 struct xfrm_dst *old;
1718
1719 old = this_cpu_read(xfrm_last_dst);
1720 if (old && !xfrm_bundle_ok(old))
1721 xfrm_last_dst_update(NULL, old);
1722}
1723
1724static void xfrm_pcpu_work_fn(struct work_struct *work)
1725{
1726 local_bh_disable();
1727 rcu_read_lock();
1728 __xfrm_pcpu_work_fn();
1729 rcu_read_unlock();
1730 local_bh_enable();
1731}
1732
1733void xfrm_policy_cache_flush(void)
1734{
1735 struct xfrm_dst *old;
1736 bool found = 0;
1737 int cpu;
1738
1739 local_bh_disable();
1740 rcu_read_lock();
1741 for_each_possible_cpu(cpu) {
1742 old = per_cpu(xfrm_last_dst, cpu);
1743 if (old && !xfrm_bundle_ok(old)) {
1744 if (smp_processor_id() == cpu) {
1745 __xfrm_pcpu_work_fn();
1746 continue;
1747 }
1748 found = true;
1749 break;
1750 }
1751 }
1752
1753 rcu_read_unlock();
1754 local_bh_enable();
1755
1756 if (!found)
1757 return;
1758
1759 get_online_cpus();
1760
1761 for_each_possible_cpu(cpu) {
1762 bool bundle_release;
1763
1764 rcu_read_lock();
1765 old = per_cpu(xfrm_last_dst, cpu);
1766 bundle_release = old && !xfrm_bundle_ok(old);
1767 rcu_read_unlock();
1768
1769 if (!bundle_release)
1770 continue;
1771
1772 if (cpu_online(cpu)) {
1773 schedule_work_on(cpu, &xfrm_pcpu_work[cpu]);
1774 continue;
1775 }
1776
1777 rcu_read_lock();
1778 old = per_cpu(xfrm_last_dst, cpu);
1779 if (old && !xfrm_bundle_ok(old)) {
1780 per_cpu(xfrm_last_dst, cpu) = NULL;
1781 dst_release(&old->u.dst);
1782 }
1783 rcu_read_unlock();
1784 }
1785
1786 put_online_cpus();
1787}
1788
cf379667
FW		 1789static bool xfrm_xdst_can_reuse(struct xfrm_dst *xdst,
1790 struct xfrm_state * const xfrm[],
1791 int num)
ec30d78c 1792{
cf379667
FW		 1793 const struct dst_entry *dst = &xdst->u.dst;
1794 int i;
ec30d78c 1795
cf379667
FW		 1796 if (xdst->num_xfrms != num)
1797 return false;
ec30d78c 1798
cf379667
FW		 1799 for (i = 0; i < num; i++) {
1800 if (!dst || dst->xfrm != xfrm[i])
1801 return false;
b92cf4aa 1802 dst = xfrm_dst_child(dst);
cf379667 1803 }
ec30d78c 1804
cf379667 1805 return xfrm_bundle_ok(xdst);
ec30d78c
FW		 1806}
1807
80c802f3
TT		 1808static struct xfrm_dst *
1809xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1810 const struct flowi *fl, u16 family,
80c802f3
TT		 1811 struct dst_entry *dst_orig)
1812{
1813 struct net *net = xp_net(pols[0]);
1814 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
ec30d78c 1815 struct xfrm_dst *xdst, *old;
80c802f3 1816 struct dst_entry *dst;
80c802f3
TT		 1817 int err;
1818
cf379667
FW		 1819 /* Try to instantiate a bundle */
1820 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1821 if (err <= 0) {
1822 if (err != 0 && err != -EAGAIN)
1823 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1824 return ERR_PTR(err);
1825 }
1826
ec30d78c
FW		 1827 xdst = this_cpu_read(xfrm_last_dst);
1828 if (xdst &&
1829 xdst->u.dst.dev == dst_orig->dev &&
1830 xdst->num_pols == num_pols &&
ec30d78c 1831 memcmp(xdst->pols, pols,
13ead5c4 1832 sizeof(struct xfrm_policy *) * num_pols) == 0 &&
cf379667 1833 xfrm_xdst_can_reuse(xdst, xfrm, err)) {
ec30d78c 1834 dst_hold(&xdst->u.dst);
cf379667
FW		 1835 while (err > 0)
1836 xfrm_state_put(xfrm[--err]);
ec30d78c
FW		 1837 return xdst;
1838 }
1839
1840 old = xdst;
80c802f3
TT		 1841
1842 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1843 if (IS_ERR(dst)) {
1844 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1845 return ERR_CAST(dst);
1846 }
1847
1848 xdst = (struct xfrm_dst *)dst;
1849 xdst->num_xfrms = err;
80c802f3 1850 xdst->num_pols = num_pols;
3e94c2dc 1851 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1852 xdst->policy_genid = atomic_read(&pols[0]->genid);
1853
ec30d78c
FW		 1854 atomic_set(&xdst->u.dst.__refcnt, 2);
1855 xfrm_last_dst_update(xdst, old);
1856
80c802f3
TT		 1857 return xdst;
1858}
1859
c3aed709 1860static void xfrm_policy_queue_process(struct timer_list *t)
a0073fe1 1861{
a0073fe1
SK		 1862 struct sk_buff *skb;
1863 struct sock *sk;
1864 struct dst_entry *dst;
c3aed709 1865 struct xfrm_policy *pol = from_timer(pol, t, polq.hold_timer);
3f5312ae 1866 struct net *net = xp_net(pol);
a0073fe1
SK		 1867 struct xfrm_policy_queue *pq = &pol->polq;
1868 struct flowi fl;
1869 struct sk_buff_head list;
1870
1871 spin_lock(&pq->hold_queue.lock);
1872 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1873 if (!skb) {
1874 spin_unlock(&pq->hold_queue.lock);
1875 goto out;
1876 }
a0073fe1
SK		 1877 dst = skb_dst(skb);
1878 sk = skb->sk;
1879 xfrm_decode_session(skb, &fl, dst->ops->family);
1880 spin_unlock(&pq->hold_queue.lock);
1881
0f6c480f
DM		 1882 dst_hold(xfrm_dst_path(dst));
1883 dst = xfrm_lookup(net, xfrm_dst_path(dst), &fl, sk, 0);
a0073fe1
SK		 1884 if (IS_ERR(dst))
1885 goto purge_queue;
1886
1887 if (dst->flags & DST_XFRM_QUEUE) {
1888 dst_release(dst);
1889
1890 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1891 goto purge_queue;
1892
1893 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1894 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1895 xfrm_pol_hold(pol);
1896 goto out;
a0073fe1
SK		 1897 }
1898
1899 dst_release(dst);
1900
1901 __skb_queue_head_init(&list);
1902
1903 spin_lock(&pq->hold_queue.lock);
1904 pq->timeout = 0;
1905 skb_queue_splice_init(&pq->hold_queue, &list);
1906 spin_unlock(&pq->hold_queue.lock);
1907
1908 while (!skb_queue_empty(&list)) {
1909 skb = __skb_dequeue(&list);
1910
1911 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
0f6c480f
DM		 1912 dst_hold(xfrm_dst_path(skb_dst(skb)));
1913 dst = xfrm_lookup(net, xfrm_dst_path(skb_dst(skb)), &fl, skb->sk, 0);
a0073fe1 1914 if (IS_ERR(dst)) {
a0073fe1
SK		 1915 kfree_skb(skb);
1916 continue;
1917 }
1918
1919 nf_reset(skb);
1920 skb_dst_drop(skb);
1921 skb_dst_set(skb, dst);
1922
13206b6b 1923 dst_output(net, skb->sk, skb);
a0073fe1
SK		 1924 }
1925
e7d8f6cb
SK		 1926out:
1927 xfrm_pol_put(pol);
a0073fe1
SK		 1928 return;
1929
1930purge_queue:
1931 pq->timeout = 0;
1ee5e667 1932 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1933 xfrm_pol_put(pol);
a0073fe1
SK		 1934}
1935
ede2059d 1936static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1937{
1938 unsigned long sched_next;
1939 struct dst_entry *dst = skb_dst(skb);
1940 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1941 struct xfrm_policy *pol = xdst->pols[0];
1942 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1943
39bb5e62 1944 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1945 kfree_skb(skb);
1946 return 0;
1947 }
a0073fe1
SK		 1948
1949 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1950 kfree_skb(skb);
1951 return -EAGAIN;
1952 }
1953
1954 skb_dst_force(skb);
a0073fe1
SK		 1955
1956 spin_lock_bh(&pq->hold_queue.lock);
1957
1958 if (!pq->timeout)
1959 pq->timeout = XFRM_QUEUE_TMO_MIN;
1960
1961 sched_next = jiffies + pq->timeout;
1962
1963 if (del_timer(&pq->hold_timer)) {
1964 if (time_before(pq->hold_timer.expires, sched_next))
1965 sched_next = pq->hold_timer.expires;
e7d8f6cb 1966 xfrm_pol_put(pol);
a0073fe1
SK		 1967 }
1968
1969 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1970 if (!mod_timer(&pq->hold_timer, sched_next))
1971 xfrm_pol_hold(pol);
a0073fe1
SK		 1972
1973 spin_unlock_bh(&pq->hold_queue.lock);
1974
1975 return 0;
1976}
1977
1978static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 1979 struct xfrm_flo *xflo,
a0073fe1
SK		 1980 const struct flowi *fl,
1981 int num_xfrms,
1982 u16 family)
1983{
1984 int err;
1985 struct net_device *dev;
b8c203b2 1986 struct dst_entry *dst;
a0073fe1
SK		 1987 struct dst_entry *dst1;
1988 struct xfrm_dst *xdst;
1989
1990 xdst = xfrm_alloc_dst(net, family);
1991 if (IS_ERR(xdst))
1992 return xdst;
1993
b8c203b2
SK		 1994 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
1995 net->xfrm.sysctl_larval_drop ||
1996 num_xfrms <= 0)
a0073fe1
SK		 1997 return xdst;
1998
b8c203b2 1999 dst = xflo->dst_orig;
a0073fe1
SK		 2000 dst1 = &xdst->u.dst;
2001 dst_hold(dst);
2002 xdst->route = dst;
2003
2004 dst_copy_metrics(dst1, dst);
2005
2006 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2007 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2008 dst1->lastuse = jiffies;
2009
2010 dst1->input = dst_discard;
2011 dst1->output = xdst_queue_output;
2012
2013 dst_hold(dst);
45b018be 2014 xfrm_dst_set_child(xdst, dst);
0f6c480f 2015 xdst->path = dst;
a0073fe1
SK		 2016
2017 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2018
2019 err = -ENODEV;
2020 dev = dst->dev;
2021 if (!dev)
2022 goto free_dst;
2023
2024 err = xfrm_fill_dst(xdst, dev, fl);
2025 if (err)
2026 goto free_dst;
2027
2028out:
2029 return xdst;
2030
2031free_dst:
2032 dst_release(dst1);
2033 xdst = ERR_PTR(err);
2034 goto out;
2035}
2036
bd45c539 2037static struct xfrm_dst *
3ca28286 2038xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir, struct xfrm_flo *xflo)
80c802f3 2039{
80c802f3 2040 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
855dad99 2041 int num_pols = 0, num_xfrms = 0, err;
bd45c539 2042 struct xfrm_dst *xdst;
80c802f3 2043
80c802f3
TT		 2044 /* Resolve policies to use if we couldn't get them from
2045 * previous cache entry */
855dad99 2046 num_pols = 1;
86dc8ee0 2047 pols[0] = xfrm_policy_lookup(net, fl, family, dir);
855dad99 2048 err = xfrm_expand_policies(fl, family, pols,
80c802f3 2049 &num_pols, &num_xfrms);
855dad99
FW		 2050 if (err < 0)
2051 goto inc_error;
2052 if (num_pols == 0)
2053 return NULL;
2054 if (num_xfrms <= 0)
2055 goto make_dummy_bundle;
80c802f3 2056
bd45c539 2057 xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
b8c203b2 2058 xflo->dst_orig);
bd45c539
FW		 2059 if (IS_ERR(xdst)) {
2060 err = PTR_ERR(xdst);
80c802f3
TT		 2061 if (err != -EAGAIN)
2062 goto error;
855dad99 2063 goto make_dummy_bundle;
bd45c539 2064 } else if (xdst == NULL) {
d809ec89 2065 num_xfrms = 0;
855dad99 2066 goto make_dummy_bundle;
80c802f3
TT		 2067 }
2068
bd45c539 2069 return xdst;
80c802f3
TT		 2070
2071make_dummy_bundle:
2072 /* We found policies, but there's no bundles to instantiate:
2073 * either because the policy blocks, has no transformations or
2074 * we could not build template (no xfrm_states).*/
b8c203b2 2075 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2076 if (IS_ERR(xdst)) {
2077 xfrm_pols_put(pols, num_pols);
2078 return ERR_CAST(xdst);
2079 }
2080 xdst->num_pols = num_pols;
2081 xdst->num_xfrms = num_xfrms;
3e94c2dc 2082 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3 2083
bd45c539 2084 return xdst;
80c802f3
TT		 2085
2086inc_error:
2087 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2088error:
855dad99 2089 xfrm_pols_put(pols, num_pols);
80c802f3
TT		 2090 return ERR_PTR(err);
2091}
1da177e4 2092
2774c131
DM		 2093static struct dst_entry *make_blackhole(struct net *net, u16 family,
2094 struct dst_entry *dst_orig)
2095{
37b10383 2096 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2774c131
DM		 2097 struct dst_entry *ret;
2098
2099 if (!afinfo) {
2100 dst_release(dst_orig);
433a1954 2101 return ERR_PTR(-EINVAL);
2774c131
DM		 2102 } else {
2103 ret = afinfo->blackhole_route(net, dst_orig);
2104 }
bdba9fe0 2105 rcu_read_unlock();
2774c131
DM		 2106
2107 return ret;
2108}
2109
1da177e4
LT		 2110/* Main function: finds/creates a bundle for given flow.
2111 *
2112 * At the moment we eat a raw IP route. Mostly to speed up lookups
2113 * on interfaces with disabled IPsec.
2114 */
452edd59
DM		 2115struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2116 const struct flowi *fl,
6f9c9615 2117 const struct sock *sk, int flags)
1da177e4 2118{
4e81bb83 2119 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3 2120 struct xfrm_dst *xdst;
452edd59 2121 struct dst_entry *dst, *route;
80c802f3 2122 u16 family = dst_orig->ops->family;
aff669bc 2123 u8 dir = XFRM_POLICY_OUT;
4b021628 2124 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2125
80c802f3
TT		 2126 dst = NULL;
2127 xdst = NULL;
2128 route = NULL;
4e81bb83 2129
bd5eb35f 2130 sk = sk_const_to_full_sk(sk);
f7944fb1 2131 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3 2132 num_pols = 1;
4c86d777 2133 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, family);
80c802f3
TT		 2134 err = xfrm_expand_policies(fl, family, pols,
2135 &num_pols, &num_xfrms);
2136 if (err < 0)
75b8c133 2137 goto dropdst;
80c802f3
TT		 2138
2139 if (num_pols) {
2140 if (num_xfrms <= 0) {
2141 drop_pols = num_pols;
2142 goto no_transform;
2143 }
2144
2145 xdst = xfrm_resolve_and_create_bundle(
2146 pols, num_pols, fl,
2147 family, dst_orig);
2148 if (IS_ERR(xdst)) {
2149 xfrm_pols_put(pols, num_pols);
2150 err = PTR_ERR(xdst);
2151 goto dropdst;
d809ec89
TT		 2152 } else if (xdst == NULL) {
2153 num_xfrms = 0;
2154 drop_pols = num_pols;
2155 goto no_transform;
80c802f3
TT		 2156 }
2157
80c802f3 2158 route = xdst->route;
0aa64774 2159 }
3bccfbc7 2160 }
1da177e4 2161
80c802f3 2162 if (xdst == NULL) {
b8c203b2
SK		 2163 struct xfrm_flo xflo;
2164
2165 xflo.dst_orig = dst_orig;
2166 xflo.flags = flags;
2167
1da177e4 2168 /* To accelerate a bit... */
2518c7c2 2169 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2170 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2171 goto nopol;
1da177e4 2172
bd45c539
FW		 2173 xdst = xfrm_bundle_lookup(net, fl, family, dir, &xflo);
2174 if (xdst == NULL)
80c802f3 2175 goto nopol;
bd45c539
FW		 2176 if (IS_ERR(xdst)) {
2177 err = PTR_ERR(xdst);
75b8c133 2178 goto dropdst;
d66e37a9 2179 }
80c802f3
TT		 2180
2181 num_pols = xdst->num_pols;
2182 num_xfrms = xdst->num_xfrms;
3e94c2dc 2183 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2184 route = xdst->route;
2185 }
2186
2187 dst = &xdst->u.dst;
2188 if (route == NULL && num_xfrms > 0) {
2189 /* The only case when xfrm_bundle_lookup() returns a
2190 * bundle with null route, is when the template could
2191 * not be resolved. It means policies are there, but
2192 * bundle could not be created, since we don't yet
2193 * have the xfrm_state's. We need to wait for KM to
2194 * negotiate new SA's or bail out with error.*/
2195 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2196 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2197 err = -EREMOTE;
2198 goto error;
80c802f3 2199 }
80c802f3 2200
5b8ef341 2201 err = -EAGAIN;
80c802f3
TT		 2202
2203 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2204 goto error;
1da177e4
LT		 2205 }
2206
80c802f3
TT		 2207no_transform:
2208 if (num_pols == 0)
8b7817f3 2209 goto nopol;
1da177e4 2210
80c802f3
TT		 2211 if ((flags & XFRM_LOOKUP_ICMP) &&
2212 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2213 err = -ENOENT;
8b7817f3 2214 goto error;
80c802f3 2215 }
8b7817f3 2216
80c802f3
TT		 2217 for (i = 0; i < num_pols; i++)
2218 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2219
80c802f3 2220 if (num_xfrms < 0) {
1da177e4 2221 /* Prohibit the flow */
59c9940e 2222 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2223 err = -EPERM;
2224 goto error;
80c802f3
TT		 2225 } else if (num_xfrms > 0) {
2226 /* Flow transformed */
80c802f3
TT		 2227 dst_release(dst_orig);
2228 } else {
2229 /* Flow passes untransformed */
2230 dst_release(dst);
452edd59 2231 dst = dst_orig;
1da177e4 2232 }
80c802f3
TT		 2233ok:
2234 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2235 if (dst && dst->xfrm &&
2236 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2237 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2238 return dst;
1da177e4 2239
80c802f3 2240nopol:
452edd59
DM		 2241 if (!(flags & XFRM_LOOKUP_ICMP)) {
2242 dst = dst_orig;
80c802f3 2243 goto ok;
452edd59 2244 }
80c802f3 2245 err = -ENOENT;
1da177e4 2246error:
80c802f3 2247 dst_release(dst);
75b8c133 2248dropdst:
ac37e251
W		 2249 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2250 dst_release(dst_orig);
80c802f3 2251 xfrm_pols_put(pols, drop_pols);
452edd59 2252 return ERR_PTR(err);
1da177e4
LT		 2253}
2254EXPORT_SYMBOL(xfrm_lookup);
2255
f92ee619
SK		 2256/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2257 * Otherwise we may send out blackholed packets.
2258 */
2259struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2260 const struct flowi *fl,
6f9c9615 2261 const struct sock *sk, int flags)
f92ee619 2262{
b8c203b2 2263 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2264 flags | XFRM_LOOKUP_QUEUE |
2265 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2266
2267 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2268 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2269
2270 return dst;
2271}
2272EXPORT_SYMBOL(xfrm_lookup_route);
2273
df0ba92a 2274static inline int
8f029de2 2275xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2276{
2277 struct xfrm_state *x;
df0ba92a
MN		 2278
2279 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2280 return 0;
2281 x = skb->sp->xvec[idx];
2282 if (!x->type->reject)
2283 return 0;
1ecafede 2284 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2285}
2286
1da177e4
LT		 2287/* When skb is transformed back to its "native" form, we have to
2288 * check policy restrictions. At the moment we make this in maximally
2289 * stupid way. Shame on me. :-) Of course, connected sockets must
2290 * have policy cached at them.
2291 */
2292
2293static inline int
7db454b9 2294xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2295 unsigned short family)
2296{
2297 if (xfrm_state_kern(x))
928ba416 2298 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2299 return x->id.proto == tmpl->id.proto &&
2300 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2301 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2302 x->props.mode == tmpl->mode &&
c5d18e98 2303 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2304 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2305 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2306 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2307}
2308
df0ba92a
MN		 2309/*
2310 * 0 or more than 0 is returned when validation is succeeded (either bypass
2311 * because of optional transport mode, or next index of the mathced secpath
2312 * state with the template.
2313 * -1 is returned when no matching template is found.
2314 * Otherwise "-2 - errored_index" is returned.
2315 */
1da177e4 2316static inline int
22cccb7e 2317xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2318 unsigned short family)
2319{
2320 int idx = start;
2321
2322 if (tmpl->optional) {
7e49e6de 2323 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2324 return start;
2325 } else
2326 start = -1;
2327 for (; idx < sp->len; idx++) {
dbe5b4aa 2328 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2329 return ++idx;
df0ba92a
MN		 2330 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2331 if (start == -1)
2332 start = -2-idx;
1da177e4 2333 break;
df0ba92a 2334 }
1da177e4
LT		 2335 }
2336 return start;
2337}
2338
d5422efe
HX		 2339int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2340 unsigned int family, int reverse)
1da177e4 2341{
37b10383 2342 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2343 int err;
1da177e4
LT		 2344
2345 if (unlikely(afinfo == NULL))
2346 return -EAFNOSUPPORT;
2347
d5422efe 2348 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2349 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
bdba9fe0 2350 rcu_read_unlock();
e0d1caa7 2351 return err;
1da177e4 2352}
d5422efe 2353EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2354
9a7386ec 2355static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2356{
2357 for (; k < sp->len; k++) {
df0ba92a 2358 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2359 *idxp = k;
1da177e4 2360 return 1;
df0ba92a 2361 }
1da177e4
LT		 2362 }
2363
2364 return 0;
2365}
2366
a716c119 2367int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2368 unsigned short family)
2369{
f6e1e25d 2370 struct net *net = dev_net(skb->dev);
1da177e4 2371 struct xfrm_policy *pol;
4e81bb83
MN		 2372 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2373 int npols = 0;
2374 int xfrm_nr;
2375 int pi;
d5422efe 2376 int reverse;
1da177e4 2377 struct flowi fl;
df0ba92a 2378 int xerr_idx = -1;
1da177e4 2379
d5422efe
HX		 2380 reverse = dir & ~XFRM_POLICY_MASK;
2381 dir &= XFRM_POLICY_MASK;
d5422efe 2382
0aa64774 2383 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2384 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2385 return 0;
0aa64774
MN		 2386 }
2387
eb9c7ebe 2388 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2389
2390 /* First, check used SA against their selectors. */
2391 if (skb->sp) {
2392 int i;
2393
9b7a787d 2394 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2395 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2396 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2397 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2398 return 0;
0aa64774 2399 }
1da177e4
LT		 2400 }
2401 }
2402
2403 pol = NULL;
bd5eb35f 2404 sk = sk_to_full_sk(sk);
3bccfbc7 2405 if (sk && sk->sk_policy[dir]) {
4c86d777 2406 pol = xfrm_sk_policy_lookup(sk, dir, &fl, family);
0aa64774 2407 if (IS_ERR(pol)) {
59c9940e 2408 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2409 return 0;
0aa64774 2410 }
3bccfbc7 2411 }
1da177e4 2412
86dc8ee0
FW		 2413 if (!pol)
2414 pol = xfrm_policy_lookup(net, &fl, family, dir);
1da177e4 2415
0aa64774 2416 if (IS_ERR(pol)) {
59c9940e 2417 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2418 return 0;
0aa64774 2419 }
134b0fc5 2420
df0ba92a 2421 if (!pol) {
d1d9facf 2422 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2423 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2424 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2425 return 0;
2426 }
2427 return 1;
2428 }
1da177e4 2429
9d729f72 2430 pol->curlft.use_time = get_seconds();
1da177e4 2431
4e81bb83 2432 pols[0] = pol;
02d0892f 2433 npols++;
4e81bb83
MN		 2434#ifdef CONFIG_XFRM_SUB_POLICY
2435 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2436 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2437 &fl, family,
2438 XFRM_POLICY_IN);
2439 if (pols[1]) {
0aa64774 2440 if (IS_ERR(pols[1])) {
59c9940e 2441 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2442 return 0;
0aa64774 2443 }
9d729f72 2444 pols[1]->curlft.use_time = get_seconds();
02d0892f 2445 npols++;
4e81bb83
MN		 2446 }
2447 }
2448#endif
2449
1da177e4
LT		 2450 if (pol->action == XFRM_POLICY_ALLOW) {
2451 struct sec_path *sp;
2452 static struct sec_path dummy;
4e81bb83 2453 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2454 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2455 struct xfrm_tmpl **tpp = tp;
2456 int ti = 0;
1da177e4
LT		 2457 int i, k;
2458
2459 if ((sp = skb->sp) == NULL)
2460 sp = &dummy;
2461
4e81bb83
MN		 2462 for (pi = 0; pi < npols; pi++) {
2463 if (pols[pi] != pol &&
0aa64774 2464 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2465 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2466 goto reject;
0aa64774
MN		 2467 }
2468 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2469 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2470 goto reject_error;
0aa64774 2471 }
4e81bb83
MN		 2472 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2473 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2474 }
2475 xfrm_nr = ti;
41a49cc3 2476 if (npols > 1) {
283bc9f3 2477 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2478 tpp = stp;
2479 }
4e81bb83 2480
1da177e4
LT		 2481 /* For each tunnel xfrm, find the first matching tmpl.
2482 * For each tmpl before that, find corresponding xfrm.
2483 * Order is _important_. Later we will implement
2484 * some barriers, but at the moment barriers
2485 * are implied between each two transformations.
2486 */
4e81bb83
MN		 2487 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2488 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2489 if (k < 0) {
d1d9facf
JM		 2490 if (k < -1)
2491 /* "-2 - errored_index" returned */
2492 xerr_idx = -(2+k);
59c9940e 2493 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2494 goto reject;
df0ba92a 2495 }
1da177e4
LT		 2496 }
2497
0aa64774 2498 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2499 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2500 goto reject;
0aa64774 2501 }
1da177e4 2502
4e81bb83 2503 xfrm_pols_put(pols, npols);
1da177e4
LT		 2504 return 1;
2505 }
59c9940e 2506 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2507
2508reject:
df0ba92a 2509 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2510reject_error:
2511 xfrm_pols_put(pols, npols);
1da177e4
LT		 2512 return 0;
2513}
2514EXPORT_SYMBOL(__xfrm_policy_check);
2515
2516int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2517{
99a66657 2518 struct net *net = dev_net(skb->dev);
1da177e4 2519 struct flowi fl;
adf30907 2520 struct dst_entry *dst;
73137147 2521 int res = 1;
1da177e4 2522
0aa64774 2523 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2524 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2525 return 0;
0aa64774 2526 }
1da177e4 2527
fafeeb6c 2528 skb_dst_force(skb);
adf30907 2529
b8c203b2 2530 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2531 if (IS_ERR(dst)) {
73137147 2532 res = 0;
452edd59
DM		 2533 dst = NULL;
2534 }
adf30907
ED		 2535 skb_dst_set(skb, dst);
2536 return res;
1da177e4
LT		 2537}
2538EXPORT_SYMBOL(__xfrm_route_forward);
2539
d49c73c7
DM		 2540/* Optimize later using cookies and generation ids. */
2541
1da177e4
LT		 2542static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2543{
d49c73c7 2544 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2545 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2546 * get validated by dst_ops->check on every use. We do this
2547 * because when a normal route referenced by an XFRM dst is
2548 * obsoleted we do not go looking around for all parent
2549 * referencing XFRM dsts so that we can invalidate them. It
2550 * is just too much work. Instead we make the checks here on
2551 * every use. For example:
d49c73c7
DM		 2552 *
2553 * XFRM dst A --> IPv4 dst X
2554 *
2555 * X is the "xdst->route" of A (X is also the "dst->path" of A
2556 * in this example). If X is marked obsolete, "A" will not
2557 * notice. That's what we are validating here via the
2558 * stale_bundle() check.
2559 *
52df157f
WW		 2560 * When a dst is removed from the fib tree, DST_OBSOLETE_DEAD will
2561 * be marked on it.
09c75704 2562 * This will force stale_bundle() to fail on any xdst bundle with
52df157f 2563 * this dst linked in it.
399c180a 2564 */
d49c73c7
DM		 2565 if (dst->obsolete < 0 && !stale_bundle(dst))
2566 return dst;
2567
1da177e4
LT		 2568 return NULL;
2569}
2570
2571static int stale_bundle(struct dst_entry *dst)
2572{
12fdb4d3 2573 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2574}
2575
aabc9761 2576void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2577{
b92cf4aa 2578 while ((dst = xfrm_dst_child(dst)) && dst->xfrm && dst->dev == dev) {
c346dca1 2579 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2580 dev_hold(dst->dev);
1da177e4
LT		 2581 dev_put(dev);
2582 }
2583}
aabc9761 2584EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2585
2586static void xfrm_link_failure(struct sk_buff *skb)
2587{
2588 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2589}
2590
2591static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2592{
2593 if (dst) {
2594 if (dst->obsolete) {
2595 dst_release(dst);
2596 dst = NULL;
2597 }
2598 }
2599 return dst;
2600}
2601
25ee3286 2602static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2603{
2604 do {
2605 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2606 u32 pmtu, route_mtu_cached;
2607
b92cf4aa 2608 pmtu = dst_mtu(xfrm_dst_child(dst));
1da177e4
LT		 2609 xdst->child_mtu_cached = pmtu;
2610
2611 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2612
2613 route_mtu_cached = dst_mtu(xdst->route);
2614 xdst->route_mtu_cached = route_mtu_cached;
2615
2616 if (pmtu > route_mtu_cached)
2617 pmtu = route_mtu_cached;
2618
defb3519 2619 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2620 } while ((dst = dst->next));
2621}
2622
1da177e4
LT		 2623/* Check that the bundle accepts the flow and its components are
2624 * still valid.
2625 */
2626
12fdb4d3 2627static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2628{
2629 struct dst_entry *dst = &first->u.dst;
2630 struct xfrm_dst *last;
2631 u32 mtu;
2632
0f6c480f 2633 if (!dst_check(xfrm_dst_path(dst), ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2634 (dst->dev && !netif_running(dst->dev)))
2635 return 0;
2636
a0073fe1
SK		 2637 if (dst->flags & DST_XFRM_QUEUE)
2638 return 1;
2639
1da177e4
LT		 2640 last = NULL;
2641
2642 do {
2643 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2644
1da177e4
LT		 2645 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2646 return 0;
80c802f3
TT		 2647 if (xdst->xfrm_genid != dst->xfrm->genid)
2648 return 0;
b1312c89
TT		 2649 if (xdst->num_pols > 0 &&
2650 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2651 return 0;
e53820de 2652
b92cf4aa 2653 mtu = dst_mtu(xfrm_dst_child(dst));
1da177e4
LT		 2654 if (xdst->child_mtu_cached != mtu) {
2655 last = xdst;
2656 xdst->child_mtu_cached = mtu;
2657 }
2658
92d63dec 2659 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2660 return 0;
2661 mtu = dst_mtu(xdst->route);
2662 if (xdst->route_mtu_cached != mtu) {
2663 last = xdst;
2664 xdst->route_mtu_cached = mtu;
2665 }
2666
b92cf4aa 2667 dst = xfrm_dst_child(dst);
1da177e4
LT		 2668 } while (dst->xfrm);
2669
2670 if (likely(!last))
2671 return 1;
2672
2673 mtu = last->child_mtu_cached;
2674 for (;;) {
2675 dst = &last->u.dst;
2676
2677 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2678 if (mtu > last->route_mtu_cached)
2679 mtu = last->route_mtu_cached;
defb3519 2680 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2681
2682 if (last == first)
2683 break;
2684
bd0bf076 2685 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2686 last->child_mtu_cached = mtu;
2687 }
2688
2689 return 1;
2690}
2691
0dbaee3b
DM		 2692static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2693{
0f6c480f 2694 return dst_metric_advmss(xfrm_dst_path(dst));
0dbaee3b
DM		 2695}
2696
ebb762f2 2697static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2698{
618f9bc7
SK		 2699 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2700
0f6c480f 2701 return mtu ? : dst_mtu(xfrm_dst_path(dst));
d33e4553
DM		 2702}
2703
1ecc9ad0
JA		 2704static const void *xfrm_get_dst_nexthop(const struct dst_entry *dst,
2705 const void *daddr)
63fca65d 2706{
0f6c480f 2707 while (dst->xfrm) {
63fca65d
JA		 2708 const struct xfrm_state *xfrm = dst->xfrm;
2709
2710 if (xfrm->props.mode == XFRM_MODE_TRANSPORT)
2711 continue;
2712 if (xfrm->type->flags & XFRM_TYPE_REMOTE_COADDR)
2713 daddr = xfrm->coaddr;
2714 else if (!(xfrm->type->flags & XFRM_TYPE_LOCAL_COADDR))
2715 daddr = &xfrm->id.daddr;
0f6c480f
DM		 2716
2717 dst = xfrm_dst_child(dst);
63fca65d 2718 }
1ecc9ad0
JA		 2719 return daddr;
2720}
2721
2722static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2723 struct sk_buff *skb,
2724 const void *daddr)
2725{
0f6c480f 2726 const struct dst_entry *path = xfrm_dst_path(dst);
1ecc9ad0
JA		 2727
2728 if (!skb)
2729 daddr = xfrm_get_dst_nexthop(dst, daddr);
2730 return path->ops->neigh_lookup(path, skb, daddr);
2731}
2732
2733static void xfrm_confirm_neigh(const struct dst_entry *dst, const void *daddr)
2734{
0f6c480f 2735 const struct dst_entry *path = xfrm_dst_path(dst);
1ecc9ad0
JA		 2736
2737 daddr = xfrm_get_dst_nexthop(dst, daddr);
63fca65d
JA		 2738 path->ops->confirm_neigh(path, daddr);
2739}
2740
a2817d8b 2741int xfrm_policy_register_afinfo(const struct xfrm_policy_afinfo *afinfo, int family)
1da177e4
LT		 2742{
2743 int err = 0;
a2817d8b
FW		 2744
2745 if (WARN_ON(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
1da177e4 2746 return -EAFNOSUPPORT;
a2817d8b 2747
ef8531b6 2748 spin_lock(&xfrm_policy_afinfo_lock);
a2817d8b 2749 if (unlikely(xfrm_policy_afinfo[family] != NULL))
f31e8d4f 2750 err = -EEXIST;
1da177e4
LT		 2751 else {
2752 struct dst_ops *dst_ops = afinfo->dst_ops;
2753 if (likely(dst_ops->kmem_cachep == NULL))
2754 dst_ops->kmem_cachep = xfrm_dst_cache;
2755 if (likely(dst_ops->check == NULL))
2756 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2757 if (likely(dst_ops->default_advmss == NULL))
2758 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2759 if (likely(dst_ops->mtu == NULL))
2760 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2761 if (likely(dst_ops->negative_advice == NULL))
2762 dst_ops->negative_advice = xfrm_negative_advice;
2763 if (likely(dst_ops->link_failure == NULL))
2764 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2765 if (likely(dst_ops->neigh_lookup == NULL))
2766 dst_ops->neigh_lookup = xfrm_neigh_lookup;
63fca65d
JA		 2767 if (likely(!dst_ops->confirm_neigh))
2768 dst_ops->confirm_neigh = xfrm_confirm_neigh;
a2817d8b 2769 rcu_assign_pointer(xfrm_policy_afinfo[family], afinfo);
1da177e4 2770 }
ef8531b6 2771 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c 2772
1da177e4
LT		 2773 return err;
2774}
2775EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2776
a2817d8b 2777void xfrm_policy_unregister_afinfo(const struct xfrm_policy_afinfo *afinfo)
1da177e4 2778{
2b61997a 2779 struct dst_ops *dst_ops = afinfo->dst_ops;
a2817d8b 2780 int i;
2b61997a 2781
a2817d8b
FW		 2782 for (i = 0; i < ARRAY_SIZE(xfrm_policy_afinfo); i++) {
2783 if (xfrm_policy_afinfo[i] != afinfo)
2784 continue;
2785 RCU_INIT_POINTER(xfrm_policy_afinfo[i], NULL);
2786 break;
ef8531b6 2787 }
ef8531b6 2788
2b61997a 2789 synchronize_rcu();
ef8531b6 2790
2b61997a
FW		 2791 dst_ops->kmem_cachep = NULL;
2792 dst_ops->check = NULL;
2793 dst_ops->negative_advice = NULL;
2794 dst_ops->link_failure = NULL;
1da177e4
LT		 2795}
2796EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2797
558f82ef 2798#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2799static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2800{
c68cd1a0 2801 int rv;
698365fa
WC		 2802 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2803 if (!net->mib.xfrm_statistics)
558f82ef 2804 return -ENOMEM;
c68cd1a0
AD		 2805 rv = xfrm_proc_init(net);
2806 if (rv < 0)
698365fa 2807 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2808 return rv;
558f82ef 2809}
59c9940e
AD		 2810
2811static void xfrm_statistics_fini(struct net *net)
2812{
c68cd1a0 2813 xfrm_proc_fini(net);
698365fa 2814 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2815}
2816#else
2817static int __net_init xfrm_statistics_init(struct net *net)
2818{
2819 return 0;
2820}
2821
2822static void xfrm_statistics_fini(struct net *net)
2823{
2824}
558f82ef
MN		 2825#endif
2826
d62ddc21 2827static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2828{
2518c7c2
DM		 2829 unsigned int hmask, sz;
2830 int dir;
2831
d62ddc21
AD		 2832 if (net_eq(net, &init_net))
2833 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2834 sizeof(struct xfrm_dst),
e5d679f3 2835 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2836 NULL);
1da177e4 2837
2518c7c2
DM		 2838 hmask = 8 - 1;
2839 sz = (hmask+1) * sizeof(struct hlist_head);
2840
93b851c1
AD		 2841 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2842 if (!net->xfrm.policy_byidx)
2843 goto out_byidx;
8100bea7 2844 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2845
53c2e285 2846 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2847 struct xfrm_policy_hash *htab;
2848
dc2caba7 2849 net->xfrm.policy_count[dir] = 0;
53c2e285 2850 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2851 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2852
a35f6c5d 2853 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2854 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2855 if (!htab->table)
a35f6c5d
AD		 2856 goto out_bydst;
2857 htab->hmask = hmask;
b58555f1
CG		 2858 htab->dbits4 = 32;
2859 htab->sbits4 = 32;
2860 htab->dbits6 = 128;
2861 htab->sbits6 = 128;
2518c7c2 2862 }
880a6fab
CG		 2863 net->xfrm.policy_hthresh.lbits4 = 32;
2864 net->xfrm.policy_hthresh.rbits4 = 32;
2865 net->xfrm.policy_hthresh.lbits6 = 128;
2866 net->xfrm.policy_hthresh.rbits6 = 128;
2867
2868 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2869
adfcf0b2 2870 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2871 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2872 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21 2873 if (net_eq(net, &init_net))
21f42cc9 2874 xfrm_dev_init();
d62ddc21 2875 return 0;
93b851c1 2876
a35f6c5d
AD		 2877out_bydst:
2878 for (dir--; dir >= 0; dir--) {
2879 struct xfrm_policy_hash *htab;
2880
2881 htab = &net->xfrm.policy_bydst[dir];
2882 xfrm_hash_free(htab->table, sz);
2883 }
2884 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2885out_byidx:
2886 return -ENOMEM;
d62ddc21
AD		 2887}
2888
2889static void xfrm_policy_fini(struct net *net)
2890{
93b851c1 2891 unsigned int sz;
8b18f8ea 2892 int dir;
93b851c1 2893
7c2776ee
AD		 2894 flush_work(&net->xfrm.policy_hash_work);
2895#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2896 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 2897#endif
2e71029e 2898 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 2899
adfcf0b2 2900 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2901
53c2e285 2902 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 2903 struct xfrm_policy_hash *htab;
2904
8b18f8ea 2905 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2906
2907 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2908 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2909 WARN_ON(!hlist_empty(htab->table));
2910 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2911 }
2912
8100bea7 2913 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2914 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2915 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2916}
2917
d62ddc21
AD		 2918static int __net_init xfrm_net_init(struct net *net)
2919{
2920 int rv;
2921
c282222a
FW		 2922 /* Initialize the per-net locks here */
2923 spin_lock_init(&net->xfrm.xfrm_state_lock);
2924 spin_lock_init(&net->xfrm.xfrm_policy_lock);
2925 mutex_init(&net->xfrm.xfrm_cfg_mutex);
2926
59c9940e
AD		 2927 rv = xfrm_statistics_init(net);
2928 if (rv < 0)
2929 goto out_statistics;
d62ddc21
AD		 2930 rv = xfrm_state_init(net);
2931 if (rv < 0)
2932 goto out_state;
2933 rv = xfrm_policy_init(net);
2934 if (rv < 0)
2935 goto out_policy;
b27aeadb
AD		 2936 rv = xfrm_sysctl_init(net);
2937 if (rv < 0)
2938 goto out_sysctl;
283bc9f3 2939
d62ddc21
AD		 2940 return 0;
2941
b27aeadb
AD		 2942out_sysctl:
2943 xfrm_policy_fini(net);
d62ddc21
AD		 2944out_policy:
2945 xfrm_state_fini(net);
2946out_state:
59c9940e
AD		 2947 xfrm_statistics_fini(net);
2948out_statistics:
d62ddc21
AD		 2949 return rv;
2950}
2951
2952static void __net_exit xfrm_net_exit(struct net *net)
2953{
b27aeadb 2954 xfrm_sysctl_fini(net);
d62ddc21
AD		 2955 xfrm_policy_fini(net);
2956 xfrm_state_fini(net);
59c9940e 2957 xfrm_statistics_fini(net);
d62ddc21
AD		 2958}
2959
2960static struct pernet_operations __net_initdata xfrm_net_ops = {
2961 .init = xfrm_net_init,
2962 .exit = xfrm_net_exit,
2963};
2964
1da177e4
LT		 2965void __init xfrm_init(void)
2966{
ec30d78c
FW		 2967 int i;
2968
2969 xfrm_pcpu_work = kmalloc_array(NR_CPUS, sizeof(*xfrm_pcpu_work),
2970 GFP_KERNEL);
2971 BUG_ON(!xfrm_pcpu_work);
2972
2973 for (i = 0; i < NR_CPUS; i++)
2974 INIT_WORK(&xfrm_pcpu_work[i], xfrm_pcpu_work_fn);
2975
d62ddc21 2976 register_pernet_subsys(&xfrm_net_ops);
30846090 2977 seqcount_init(&xfrm_policy_hash_generation);
1da177e4
LT		 2978 xfrm_input_init();
2979}
2980
ab5f5e8b 2981#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2982static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2983 struct audit_buffer *audit_buf)
ab5f5e8b 2984{
875179fa
PM		 2985 struct xfrm_sec_ctx *ctx = xp->security;
2986 struct xfrm_selector *sel = &xp->selector;
2987
2988 if (ctx)
ab5f5e8b 2989 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2990 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2991
9b7a787d 2992 switch (sel->family) {
ab5f5e8b 2993 case AF_INET:
21454aaa 2994 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2995 if (sel->prefixlen_s != 32)
2996 audit_log_format(audit_buf, " src_prefixlen=%d",
2997 sel->prefixlen_s);
21454aaa 2998 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2999 if (sel->prefixlen_d != 32)
3000 audit_log_format(audit_buf, " dst_prefixlen=%d",
3001 sel->prefixlen_d);
ab5f5e8b
JL		 3002 break;
3003 case AF_INET6:
5b095d98 3004 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3005 if (sel->prefixlen_s != 128)
3006 audit_log_format(audit_buf, " src_prefixlen=%d",
3007 sel->prefixlen_s);
5b095d98 3008 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3009 if (sel->prefixlen_d != 128)
3010 audit_log_format(audit_buf, " dst_prefixlen=%d",
3011 sel->prefixlen_d);
ab5f5e8b
JL		 3012 break;
3013 }
3014}
3015
2e71029e 3016void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3017{
3018 struct audit_buffer *audit_buf;
ab5f5e8b 3019
afeb14b4 3020 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3021 if (audit_buf == NULL)
3022 return;
2e71029e 3023 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3024 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3025 xfrm_audit_common_policyinfo(xp, audit_buf);
3026 audit_log_end(audit_buf);
3027}
3028EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3029
68277acc 3030void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3031 bool task_valid)
ab5f5e8b
JL		 3032{
3033 struct audit_buffer *audit_buf;
ab5f5e8b 3034
afeb14b4 3035 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3036 if (audit_buf == NULL)
3037 return;
2e71029e 3038 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3039 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3040 xfrm_audit_common_policyinfo(xp, audit_buf);
3041 audit_log_end(audit_buf);
3042}
3043EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3044#endif
3045
80c9abaa 3046#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3047static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3048 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3049{
3050 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3051 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3052 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3053 sel_cmp->family) &&
3054 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3055 sel_cmp->family) &&
80c9abaa
SS		 3056 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3057 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3058 return true;
80c9abaa
SS		 3059 }
3060 } else {
3061 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3062 return true;
80c9abaa
SS		 3063 }
3064 }
bc9b35ad 3065 return false;
80c9abaa
SS		 3066}
3067
3e94c2dc
WC		 3068static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3069 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3070{
3071 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3072 struct hlist_head *chain;
3073 u32 priority = ~0U;
3074
9d0380df 3075 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d549c4f 3076 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3077 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3078 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3079 pol->type == type) {
3080 ret = pol;
3081 priority = ret->priority;
3082 break;
3083 }
3084 }
8d549c4f 3085 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3086 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR		 3087 if ((pol->priority >= priority) && ret)
3088 break;
3089
80c9abaa 3090 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
8faf491e 3091 pol->type == type) {
80c9abaa
SS		 3092 ret = pol;
3093 break;
3094 }
3095 }
3096
586f2eb4 3097 xfrm_pol_hold(ret);
80c9abaa 3098
9d0380df 3099 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3100
3101 return ret;
3102}
3103
dd701754 3104static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3105{
3106 int match = 0;
3107
3108 if (t->mode == m->mode && t->id.proto == m->proto &&
3109 (m->reqid == 0 || t->reqid == m->reqid)) {
3110 switch (t->mode) {
3111 case XFRM_MODE_TUNNEL:
3112 case XFRM_MODE_BEET:
70e94e66
YH		 3113 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3114 m->old_family) &&
3115 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3116 m->old_family)) {
80c9abaa
SS		 3117 match = 1;
3118 }
3119 break;
3120 case XFRM_MODE_TRANSPORT:
3121 /* in case of transport mode, template does not store
3122 any IP addresses, hence we just compare mode and
3123 protocol */
3124 match = 1;
3125 break;
3126 default:
3127 break;
3128 }
3129 }
3130 return match;
3131}
3132
3133/* update endpoint address(es) of template(s) */
3134static int xfrm_policy_migrate(struct xfrm_policy *pol,
3135 struct xfrm_migrate *m, int num_migrate)
3136{
3137 struct xfrm_migrate *mp;
80c9abaa
SS		 3138 int i, j, n = 0;
3139
3140 write_lock_bh(&pol->lock);
12a169e7 3141 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3142 /* target policy has been deleted */
3143 write_unlock_bh(&pol->lock);
3144 return -ENOENT;
3145 }
3146
3147 for (i = 0; i < pol->xfrm_nr; i++) {
3148 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3149 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3150 continue;
3151 n++;
1bfcb10f
HX		 3152 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3153 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3154 continue;
3155 /* update endpoints */
3156 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3157 sizeof(pol->xfrm_vec[i].id.daddr));
3158 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3159 sizeof(pol->xfrm_vec[i].saddr));
3160 pol->xfrm_vec[i].encap_family = mp->new_family;
3161 /* flush bundles */
80c802f3 3162 atomic_inc(&pol->genid);
80c9abaa
SS		 3163 }
3164 }
3165
3166 write_unlock_bh(&pol->lock);
3167
3168 if (!n)
3169 return -ENODATA;
3170
3171 return 0;
3172}
3173
dd701754 3174static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3175{
3176 int i, j;
3177
3178 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3179 return -EINVAL;
3180
3181 for (i = 0; i < num_migrate; i++) {
80c9abaa
SS		 3182 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3183 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3184 return -EINVAL;
3185
3186 /* check if there is any duplicated entry */
3187 for (j = i + 1; j < num_migrate; j++) {
3188 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3189 sizeof(m[i].old_daddr)) &&
3190 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3191 sizeof(m[i].old_saddr)) &&
3192 m[i].proto == m[j].proto &&
3193 m[i].mode == m[j].mode &&
3194 m[i].reqid == m[j].reqid &&
3195 m[i].old_family == m[j].old_family)
3196 return -EINVAL;
3197 }
3198 }
3199
3200 return 0;
3201}
3202
b4b7c0b3 3203int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3204 struct xfrm_migrate *m, int num_migrate,
4ab47d47
AA		 3205 struct xfrm_kmaddress *k, struct net *net,
3206 struct xfrm_encap_tmpl *encap)
80c9abaa
SS		 3207{
3208 int i, err, nx_cur = 0, nx_new = 0;
3209 struct xfrm_policy *pol = NULL;
3210 struct xfrm_state *x, *xc;
3211 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3212 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3213 struct xfrm_migrate *mp;
3214
7bab0963 3215 /* Stage 0 - sanity checks */
80c9abaa
SS		 3216 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3217 goto out;
3218
7bab0963
VD		 3219 if (dir >= XFRM_POLICY_MAX) {
3220 err = -EINVAL;
3221 goto out;
3222 }
3223
80c9abaa 3224 /* Stage 1 - find policy */
8d549c4f 3225 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3226 err = -ENOENT;
3227 goto out;
3228 }
3229
3230 /* Stage 2 - find and update state(s) */
3231 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3232 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3233 x_cur[nx_cur] = x;
3234 nx_cur++;
4ab47d47
AA		 3235 xc = xfrm_state_migrate(x, mp, encap);
3236 if (xc) {
80c9abaa
SS		 3237 x_new[nx_new] = xc;
3238 nx_new++;
3239 } else {
3240 err = -ENODATA;
3241 goto restore_state;
3242 }
3243 }
3244 }
3245
3246 /* Stage 3 - update policy */
3247 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3248 goto restore_state;
3249
3250 /* Stage 4 - delete old state(s) */
3251 if (nx_cur) {
3252 xfrm_states_put(x_cur, nx_cur);
3253 xfrm_states_delete(x_cur, nx_cur);
3254 }
3255
3256 /* Stage 5 - announce */
8bafd730 3257 km_migrate(sel, dir, type, m, num_migrate, k, encap);
80c9abaa
SS		 3258
3259 xfrm_pol_put(pol);
3260
3261 return 0;
3262out:
3263 return err;
3264
3265restore_state:
3266 if (pol)
3267 xfrm_pol_put(pol);
3268 if (nx_cur)
3269 xfrm_states_put(x_cur, nx_cur);
3270 if (nx_new)
3271 xfrm_states_delete(x_new, nx_new);
3272
3273 return err;
3274}
e610e679 3275EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3276#endif
Linux 6.x block layer and io_uring tree(s)