]> Git Repo - linux.git/blame - net/xfrm/xfrm_policy.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'r8169-enable-ASPM-on-RTL8168E-VL'
[linux.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 <[email protected]>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <[email protected]> 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
f8c3d0dd 54static struct kmem_cache *xfrm_dst_cache __ro_after_init;
30846090 55static __read_mostly seqcount_t xfrm_policy_hash_generation;
1da177e4 56
5492093d 57static void xfrm_init_pmtu(struct xfrm_dst **bundle, int nr);
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) {
862591bf
FW		 612 if (policy->walk.dead ||
613 xfrm_policy_id2dir(policy->index) >= XFRM_POLICY_MAX) {
6916fb3b
TB		 614 /* skip socket policies */
615 continue;
616 }
880a6fab
CG		 617 newpos = NULL;
618 chain = policy_hash_bysel(net, &policy->selector,
619 policy->family,
620 xfrm_policy_id2dir(policy->index));
621 hlist_for_each_entry(pol, chain, bydst) {
622 if (policy->priority >= pol->priority)
623 newpos = &pol->bydst;
624 else
625 break;
626 }
627 if (newpos)
628 hlist_add_behind(&policy->bydst, newpos);
629 else
630 hlist_add_head(&policy->bydst, chain);
631 }
632
9d0380df 633 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 634
635 mutex_unlock(&hash_resize_mutex);
636}
637
638void xfrm_policy_hash_rebuild(struct net *net)
639{
640 schedule_work(&net->xfrm.policy_hthresh.work);
641}
642EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
643
1da177e4
LT		 644/* Generate new index... KAME seems to generate them ordered by cost
645 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 646static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 647{
1da177e4
LT		 648 static u32 idx_generator;
649
650 for (;;) {
2518c7c2
DM		 651 struct hlist_head *list;
652 struct xfrm_policy *p;
653 u32 idx;
654 int found;
655
e682adf0
FD		 656 if (!index) {
657 idx = (idx_generator | dir);
658 idx_generator += 8;
659 } else {
660 idx = index;
661 index = 0;
662 }
663
1da177e4
LT		 664 if (idx == 0)
665 idx = 8;
1121994c 666 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 667 found = 0;
b67bfe0d 668 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 669 if (p->index == idx) {
670 found = 1;
1da177e4 671 break;
2518c7c2 672 }
1da177e4 673 }
2518c7c2 674 if (!found)
1da177e4
LT		 675 return idx;
676 }
677}
678
2518c7c2
DM		 679static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
680{
681 u32 *p1 = (u32 *) s1;
682 u32 *p2 = (u32 *) s2;
683 int len = sizeof(struct xfrm_selector) / sizeof(u32);
684 int i;
685
686 for (i = 0; i < len; i++) {
687 if (p1[i] != p2[i])
688 return 1;
689 }
690
691 return 0;
692}
693
a0073fe1
SK		 694static void xfrm_policy_requeue(struct xfrm_policy *old,
695 struct xfrm_policy *new)
696{
697 struct xfrm_policy_queue *pq = &old->polq;
698 struct sk_buff_head list;
699
de2ad486
LR		 700 if (skb_queue_empty(&pq->hold_queue))
701 return;
702
a0073fe1
SK		 703 __skb_queue_head_init(&list);
704
705 spin_lock_bh(&pq->hold_queue.lock);
706 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 707 if (del_timer(&pq->hold_timer))
708 xfrm_pol_put(old);
a0073fe1
SK		 709 spin_unlock_bh(&pq->hold_queue.lock);
710
a0073fe1
SK		 711 pq = &new->polq;
712
713 spin_lock_bh(&pq->hold_queue.lock);
714 skb_queue_splice(&list, &pq->hold_queue);
715 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 716 if (!mod_timer(&pq->hold_timer, jiffies))
717 xfrm_pol_hold(new);
a0073fe1
SK		 718 spin_unlock_bh(&pq->hold_queue.lock);
719}
720
7cb8a939
SK		 721static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
722 struct xfrm_policy *pol)
723{
724 u32 mark = policy->mark.v & policy->mark.m;
725
726 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
727 return true;
728
729 if ((mark & pol->mark.m) == pol->mark.v &&
730 policy->priority == pol->priority)
731 return true;
732
733 return false;
734}
735
1da177e4
LT		 736int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
737{
1121994c 738 struct net *net = xp_net(policy);
2518c7c2
DM		 739 struct xfrm_policy *pol;
740 struct xfrm_policy *delpol;
741 struct hlist_head *chain;
b67bfe0d 742 struct hlist_node *newpos;
1da177e4 743
9d0380df 744 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 745 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 746 delpol = NULL;
747 newpos = NULL;
b67bfe0d 748 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 749 if (pol->type == policy->type &&
2518c7c2 750 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 751 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 752 xfrm_sec_ctx_match(pol->security, policy->security) &&
753 !WARN_ON(delpol)) {
1da177e4 754 if (excl) {
9d0380df 755 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 756 return -EEXIST;
757 }
1da177e4
LT		 758 delpol = pol;
759 if (policy->priority > pol->priority)
760 continue;
761 } else if (policy->priority >= pol->priority) {
a6c7ab55 762 newpos = &pol->bydst;
1da177e4
LT		 763 continue;
764 }
1da177e4
LT		 765 if (delpol)
766 break;
1da177e4
LT		 767 }
768 if (newpos)
1d023284 769 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 770 else
771 hlist_add_head(&policy->bydst, chain);
12bfa8bd 772 __xfrm_policy_link(policy, dir);
ca4c3fc2 773
774 /* After previous checking, family can either be AF_INET or AF_INET6 */
775 if (policy->family == AF_INET)
776 rt_genid_bump_ipv4(net);
777 else
778 rt_genid_bump_ipv6(net);
779
a0073fe1
SK		 780 if (delpol) {
781 xfrm_policy_requeue(delpol, policy);
29fa0b30 782 __xfrm_policy_unlink(delpol, dir);
a0073fe1 783 }
e682adf0 784 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 785 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 786 policy->curlft.add_time = get_seconds();
1da177e4
LT		 787 policy->curlft.use_time = 0;
788 if (!mod_timer(&policy->timer, jiffies + HZ))
789 xfrm_pol_hold(policy);
9d0380df 790 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 791
9b78a82c 792 if (delpol)
1da177e4 793 xfrm_policy_kill(delpol);
1121994c
AD		 794 else if (xfrm_bydst_should_resize(net, dir, NULL))
795 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 796
1da177e4
LT		 797 return 0;
798}
799EXPORT_SYMBOL(xfrm_policy_insert);
800
8ca2e93b
JHS		 801struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
802 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 803 struct xfrm_sec_ctx *ctx, int delete,
804 int *err)
1da177e4 805{
2518c7c2
DM		 806 struct xfrm_policy *pol, *ret;
807 struct hlist_head *chain;
1da177e4 808
ef41aaa0 809 *err = 0;
9d0380df 810 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 811 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 812 ret = NULL;
b67bfe0d 813 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 814 if (pol->type == type &&
34f8d884 815 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 816 !selector_cmp(sel, &pol->selector) &&
817 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 818 xfrm_pol_hold(pol);
2518c7c2 819 if (delete) {
03e1ad7b
PM		 820 *err = security_xfrm_policy_delete(
821 pol->security);
ef41aaa0 822 if (*err) {
9d0380df 823 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 824 return pol;
825 }
29fa0b30 826 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 827 }
828 ret = pol;
1da177e4
LT		 829 break;
830 }
831 }
9d0380df 832 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 833
fe1a5f03 834 if (ret && delete)
2518c7c2 835 xfrm_policy_kill(ret);
2518c7c2 836 return ret;
1da177e4 837}
df71837d 838EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 839
8ca2e93b
JHS		 840struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
841 int dir, u32 id, int delete, int *err)
1da177e4 842{
2518c7c2
DM		 843 struct xfrm_policy *pol, *ret;
844 struct hlist_head *chain;
1da177e4 845
b5505c6e
HX		 846 *err = -ENOENT;
847 if (xfrm_policy_id2dir(id) != dir)
848 return NULL;
849
ef41aaa0 850 *err = 0;
9d0380df 851 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 852 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 853 ret = NULL;
b67bfe0d 854 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 855 if (pol->type == type && pol->index == id &&
856 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 857 xfrm_pol_hold(pol);
2518c7c2 858 if (delete) {
03e1ad7b
PM		 859 *err = security_xfrm_policy_delete(
860 pol->security);
ef41aaa0 861 if (*err) {
9d0380df 862 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 863 return pol;
864 }
29fa0b30 865 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 866 }
867 ret = pol;
1da177e4
LT		 868 break;
869 }
870 }
9d0380df 871 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 872
fe1a5f03 873 if (ret && delete)
2518c7c2 874 xfrm_policy_kill(ret);
2518c7c2 875 return ret;
1da177e4
LT		 876}
877EXPORT_SYMBOL(xfrm_policy_byid);
878
4aa2e62c
JL		 879#ifdef CONFIG_SECURITY_NETWORK_XFRM
880static inline int
2e71029e 881xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 882{
4aa2e62c
JL		 883 int dir, err = 0;
884
885 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
886 struct xfrm_policy *pol;
4aa2e62c
JL		 887 int i;
888
b67bfe0d 889 hlist_for_each_entry(pol,
33ffbbd5 890 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 891 if (pol->type != type)
892 continue;
03e1ad7b 893 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 894 if (err) {
2e71029e 895 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 896 return err;
897 }
7dc12d6d 898 }
33ffbbd5 899 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 900 hlist_for_each_entry(pol,
33ffbbd5 901 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 902 bydst) {
903 if (pol->type != type)
904 continue;
03e1ad7b
PM		 905 err = security_xfrm_policy_delete(
906 pol->security);
4aa2e62c 907 if (err) {
ab5f5e8b 908 xfrm_audit_policy_delete(pol, 0,
2e71029e 909 task_valid);
4aa2e62c
JL		 910 return err;
911 }
912 }
913 }
914 }
915 return err;
916}
917#else
918static inline int
2e71029e 919xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 920{
921 return 0;
922}
923#endif
924
2e71029e 925int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 926{
2f1eb65f 927 int dir, err = 0, cnt = 0;
1da177e4 928
9d0380df 929 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 930
2e71029e 931 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 932 if (err)
933 goto out;
934
1da177e4 935 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 936 struct xfrm_policy *pol;
29fa0b30 937 int i;
2518c7c2
DM		 938
939 again1:
b67bfe0d 940 hlist_for_each_entry(pol,
33ffbbd5 941 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 942 if (pol->type != type)
943 continue;
ea2dea9d 944 __xfrm_policy_unlink(pol, dir);
9d0380df 945 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 946 cnt++;
1da177e4 947
2e71029e 948 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 949
2518c7c2 950 xfrm_policy_kill(pol);
1da177e4 951
4141b36a 952 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 953 goto again1;
954 }
955
33ffbbd5 956 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 957 again2:
b67bfe0d 958 hlist_for_each_entry(pol,
33ffbbd5 959 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 960 bydst) {
961 if (pol->type != type)
962 continue;
ea2dea9d 963 __xfrm_policy_unlink(pol, dir);
9d0380df 964 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 965 cnt++;
2518c7c2 966
2e71029e 967 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 968 xfrm_policy_kill(pol);
969
9d0380df 970 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 971 goto again2;
972 }
1da177e4 973 }
2518c7c2 974
1da177e4 975 }
2f1eb65f
JHS		 976 if (!cnt)
977 err = -ESRCH;
4aa2e62c 978out:
9d0380df 979 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 980 return err;
1da177e4
LT		 981}
982EXPORT_SYMBOL(xfrm_policy_flush);
983
cdcbca7c 984int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 985 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 986 void *data)
987{
12a169e7
HX		 988 struct xfrm_policy *pol;
989 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 990 int error = 0;
991
992 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
993 walk->type != XFRM_POLICY_TYPE_ANY)
994 return -EINVAL;
1da177e4 995
12a169e7 996 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 997 return 0;
998
9d0380df 999 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1000 if (list_empty(&walk->walk.all))
cdcbca7c 1001 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1002 else
80077702
LR		 1003 x = list_first_entry(&walk->walk.all,
1004 struct xfrm_policy_walk_entry, all);
1005
cdcbca7c 1006 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1007 if (x->dead)
4c563f76 1008 continue;
12a169e7
HX		 1009 pol = container_of(x, struct xfrm_policy, walk);
1010 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1011 walk->type != pol->type)
1012 continue;
1013 error = func(pol, xfrm_policy_id2dir(pol->index),
1014 walk->seq, data);
1015 if (error) {
1016 list_move_tail(&walk->walk.all, &x->all);
1017 goto out;
2518c7c2 1018 }
12a169e7 1019 walk->seq++;
1da177e4 1020 }
12a169e7 1021 if (walk->seq == 0) {
baf5d743
JHS		 1022 error = -ENOENT;
1023 goto out;
1024 }
12a169e7 1025 list_del_init(&walk->walk.all);
1da177e4 1026out:
9d0380df 1027 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1028 return error;
1029}
1030EXPORT_SYMBOL(xfrm_policy_walk);
1031
12a169e7
HX		 1032void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1033{
1034 INIT_LIST_HEAD(&walk->walk.all);
1035 walk->walk.dead = 1;
1036 walk->type = type;
1037 walk->seq = 0;
1038}
1039EXPORT_SYMBOL(xfrm_policy_walk_init);
1040
283bc9f3 1041void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1042{
1043 if (list_empty(&walk->walk.all))
1044 return;
1045
9d0380df 1046 spin_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1047 list_del(&walk->walk.all);
9d0380df 1048 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1049}
1050EXPORT_SYMBOL(xfrm_policy_walk_done);
1051
134b0fc5
JM		 1052/*
1053 * Find policy to apply to this flow.
1054 *
1055 * Returns 0 if policy found, else an -errno.
1056 */
f299d557
DM		 1057static int xfrm_policy_match(const struct xfrm_policy *pol,
1058 const struct flowi *fl,
2518c7c2 1059 u8 type, u16 family, int dir)
1da177e4 1060{
f299d557 1061 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1062 int ret = -ESRCH;
1063 bool match;
1da177e4 1064
2518c7c2 1065 if (pol->family != family ||
1d28f42c 1066 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1067 pol->type != type)
134b0fc5 1068 return ret;
1da177e4 1069
2518c7c2 1070 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1071 if (match)
1d28f42c 1072 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1073 dir);
2518c7c2 1074
134b0fc5 1075 return ret;
2518c7c2 1076}
1da177e4 1077
52479b62 1078static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1079 const struct flowi *fl,
2518c7c2
DM		 1080 u16 family, u8 dir)
1081{
134b0fc5 1082 int err;
2518c7c2 1083 struct xfrm_policy *pol, *ret;
0b597e7e 1084 const xfrm_address_t *daddr, *saddr;
2518c7c2 1085 struct hlist_head *chain;
30846090
FW		 1086 unsigned int sequence;
1087 u32 priority;
df71837d 1088
2518c7c2
DM		 1089 daddr = xfrm_flowi_daddr(fl, family);
1090 saddr = xfrm_flowi_saddr(fl, family);
1091 if (unlikely(!daddr || !saddr))
1092 return NULL;
1093
a7c44247 1094 rcu_read_lock();
30846090
FW		 1095 retry:
1096 do {
1097 sequence = read_seqcount_begin(&xfrm_policy_hash_generation);
1098 chain = policy_hash_direct(net, daddr, saddr, family, dir);
1099 } while (read_seqcount_retry(&xfrm_policy_hash_generation, sequence));
1100
1101 priority = ~0U;
2518c7c2 1102 ret = NULL;
a5eefc1d 1103 hlist_for_each_entry_rcu(pol, chain, bydst) {
134b0fc5
JM		 1104 err = xfrm_policy_match(pol, fl, type, family, dir);
1105 if (err) {
1106 if (err == -ESRCH)
1107 continue;
1108 else {
1109 ret = ERR_PTR(err);
1110 goto fail;
1111 }
1112 } else {
2518c7c2 1113 ret = pol;
acba48e1 1114 priority = ret->priority;
2518c7c2
DM		 1115 break;
1116 }
1117 }
52479b62 1118 chain = &net->xfrm.policy_inexact[dir];
a5eefc1d 1119 hlist_for_each_entry_rcu(pol, chain, bydst) {
8faf491e
LR		 1120 if ((pol->priority >= priority) && ret)
1121 break;
1122
134b0fc5
JM		 1123 err = xfrm_policy_match(pol, fl, type, family, dir);
1124 if (err) {
1125 if (err == -ESRCH)
1126 continue;
1127 else {
1128 ret = ERR_PTR(err);
1129 goto fail;
1130 }
8faf491e 1131 } else {
acba48e1
DM		 1132 ret = pol;
1133 break;
1da177e4
LT		 1134 }
1135 }
586f2eb4 1136
30846090
FW		 1137 if (read_seqcount_retry(&xfrm_policy_hash_generation, sequence))
1138 goto retry;
1139
e37cc8ad
FW		 1140 if (ret && !xfrm_pol_hold_rcu(ret))
1141 goto retry;
134b0fc5 1142fail:
a7c44247 1143 rcu_read_unlock();
4e81bb83 1144
2518c7c2 1145 return ret;
4e81bb83
MN		 1146}
1147
80c802f3 1148static struct xfrm_policy *
86dc8ee0 1149xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1150{
1151#ifdef CONFIG_XFRM_SUB_POLICY
1152 struct xfrm_policy *pol;
1153
1154 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1155 if (pol != NULL)
1156 return pol;
1157#endif
1158 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1159}
1160
6f9c9615 1161static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir,
4c86d777 1162 const struct flowi *fl, u16 family)
1da177e4
LT		 1163{
1164 struct xfrm_policy *pol;
1165
d188ba86 1166 rcu_read_lock();
ae33786f 1167 again:
d188ba86
ED		 1168 pol = rcu_dereference(sk->sk_policy[dir]);
1169 if (pol != NULL) {
ddc47e44 1170 bool match;
a716c119 1171 int err = 0;
df71837d 1172
ddc47e44
SK		 1173 if (pol->family != family) {
1174 pol = NULL;
1175 goto out;
1176 }
1177
1178 match = xfrm_selector_match(&pol->selector, fl, family);
3bccfbc7 1179 if (match) {
34f8d884
JHS		 1180 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1181 pol = NULL;
1182 goto out;
1183 }
03e1ad7b 1184 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1185 fl->flowi_secid,
aff669bc 1186 dir);
330e832a
FW		 1187 if (!err) {
1188 if (!xfrm_pol_hold_rcu(pol))
1189 goto again;
1190 } else if (err == -ESRCH) {
3bccfbc7 1191 pol = NULL;
330e832a 1192 } else {
3bccfbc7 1193 pol = ERR_PTR(err);
330e832a 1194 }
3bccfbc7 1195 } else
1da177e4
LT		 1196 pol = NULL;
1197 }
34f8d884 1198out:
d188ba86 1199 rcu_read_unlock();
1da177e4
LT		 1200 return pol;
1201}
1202
1203static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1204{
98806f75 1205 struct net *net = xp_net(pol);
4e81bb83 1206
98806f75 1207 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1208 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1209 xfrm_pol_hold(pol);
1210}
1211
1212static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1213 int dir)
1214{
98806f75
AD		 1215 struct net *net = xp_net(pol);
1216
53c2e285 1217 if (list_empty(&pol->walk.all))
2518c7c2 1218 return NULL;
1da177e4 1219
53c2e285
HX		 1220 /* Socket policies are not hashed. */
1221 if (!hlist_unhashed(&pol->bydst)) {
a5eefc1d 1222 hlist_del_rcu(&pol->bydst);
53c2e285
HX		 1223 hlist_del(&pol->byidx);
1224 }
1225
1226 list_del_init(&pol->walk.all);
98806f75 1227 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1228
1229 return pol;
1da177e4
LT		 1230}
1231
53c2e285
HX		 1232static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1233{
1234 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1235}
1236
1237static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1238{
1239 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1240}
1241
4666faab 1242int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1243{
283bc9f3
FD		 1244 struct net *net = xp_net(pol);
1245
9d0380df 1246 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1247 pol = __xfrm_policy_unlink(pol, dir);
9d0380df 1248 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1249 if (pol) {
1da177e4 1250 xfrm_policy_kill(pol);
4666faab 1251 return 0;
1da177e4 1252 }
4666faab 1253 return -ENOENT;
1da177e4 1254}
a70fcb0b 1255EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1256
1257int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1258{
be8f8284 1259 struct net *net = sock_net(sk);
1da177e4
LT		 1260 struct xfrm_policy *old_pol;
1261
4e81bb83
MN		 1262#ifdef CONFIG_XFRM_SUB_POLICY
1263 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1264 return -EINVAL;
1265#endif
1266
9d0380df 1267 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
d188ba86
ED		 1268 old_pol = rcu_dereference_protected(sk->sk_policy[dir],
1269 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
1da177e4 1270 if (pol) {
9d729f72 1271 pol->curlft.add_time = get_seconds();
e682adf0 1272 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1273 xfrm_sk_policy_link(pol, dir);
1da177e4 1274 }
d188ba86 1275 rcu_assign_pointer(sk->sk_policy[dir], pol);
a0073fe1
SK		 1276 if (old_pol) {
1277 if (pol)
1278 xfrm_policy_requeue(old_pol, pol);
1279
ea2dea9d
TT		 1280 /* Unlinking succeeds always. This is the only function
1281 * allowed to delete or replace socket policy.
1282 */
53c2e285 1283 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1284 }
9d0380df 1285 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1286
1287 if (old_pol) {
1288 xfrm_policy_kill(old_pol);
1289 }
1290 return 0;
1291}
1292
d3e40a9f 1293static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1294{
0331b1f3 1295 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1296 struct net *net = xp_net(old);
1da177e4
LT		 1297
1298 if (newp) {
1299 newp->selector = old->selector;
03e1ad7b
PM		 1300 if (security_xfrm_policy_clone(old->security,
1301 &newp->security)) {
df71837d
TJ		 1302 kfree(newp);
1303 return NULL; /* ENOMEM */
1304 }
1da177e4
LT		 1305 newp->lft = old->lft;
1306 newp->curlft = old->curlft;
fb977e2c 1307 newp->mark = old->mark;
1da177e4
LT		 1308 newp->action = old->action;
1309 newp->flags = old->flags;
1310 newp->xfrm_nr = old->xfrm_nr;
1311 newp->index = old->index;
4e81bb83 1312 newp->type = old->type;
0e74aa1d 1313 newp->family = old->family;
1da177e4
LT		 1314 memcpy(newp->xfrm_vec, old->xfrm_vec,
1315 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
9d0380df 1316 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1317 xfrm_sk_policy_link(newp, dir);
9d0380df 1318 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1319 xfrm_pol_put(newp);
1320 }
1321 return newp;
1322}
1323
d188ba86 1324int __xfrm_sk_clone_policy(struct sock *sk, const struct sock *osk)
1da177e4 1325{
d188ba86
ED		 1326 const struct xfrm_policy *p;
1327 struct xfrm_policy *np;
1328 int i, ret = 0;
1329
1330 rcu_read_lock();
1331 for (i = 0; i < 2; i++) {
1332 p = rcu_dereference(osk->sk_policy[i]);
1333 if (p) {
1334 np = clone_policy(p, i);
1335 if (unlikely(!np)) {
1336 ret = -ENOMEM;
1337 break;
1338 }
1339 rcu_assign_pointer(sk->sk_policy[i], np);
1340 }
1341 }
1342 rcu_read_unlock();
1343 return ret;
1da177e4
LT		 1344}
1345
a1e59abf 1346static int
42a7b32b 1347xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local,
077fbac4 1348 xfrm_address_t *remote, unsigned short family, u32 mark)
a1e59abf
PM		 1349{
1350 int err;
37b10383 1351 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
a1e59abf
PM		 1352
1353 if (unlikely(afinfo == NULL))
1354 return -EINVAL;
077fbac4 1355 err = afinfo->get_saddr(net, oif, local, remote, mark);
bdba9fe0 1356 rcu_read_unlock();
a1e59abf
PM		 1357 return err;
1358}
1359
1da177e4
LT		 1360/* Resolve list of templates for the flow, given policy. */
1361
1362static int
a6c2e611
DM		 1363xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1364 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1365{
fbda33b2 1366 struct net *net = xp_net(policy);
1da177e4
LT		 1367 int nx;
1368 int i, error;
94802151
SK		 1369 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1370 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1371 xfrm_address_t tmp;
1da177e4 1372
9b7a787d 1373 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4 1374 struct xfrm_state *x;
94802151
SK		 1375 xfrm_address_t *remote = daddr;
1376 xfrm_address_t *local = saddr;
1da177e4
LT		 1377 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1378
94802151
SK		 1379 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1380 tmpl->mode == XFRM_MODE_BEET) {
1381 remote = &tmpl->id.daddr;
1382 local = &tmpl->saddr;
1383 if (xfrm_addr_any(local, tmpl->encap_family)) {
1384 error = xfrm_get_saddr(net, fl->flowi_oif,
1385 &tmp, remote,
1386 tmpl->encap_family, 0);
1387 if (error)
1388 goto fail;
1389 local = &tmp;
1390 }
1da177e4
LT		 1391 }
1392
1393 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1394
1395 if (x && x->km.state == XFRM_STATE_VALID) {
1396 xfrm[nx++] = x;
94802151
SK		 1397 daddr = remote;
1398 saddr = local;
1da177e4
LT		 1399 continue;
1400 }
1401 if (x) {
1402 error = (x->km.state == XFRM_STATE_ERROR ?
1403 -EINVAL : -EAGAIN);
1404 xfrm_state_put(x);
42054569 1405 } else if (error == -ESRCH) {
a4322266 1406 error = -EAGAIN;
42054569 1407 }
1da177e4
LT		 1408
1409 if (!tmpl->optional)
1410 goto fail;
1411 }
1412 return nx;
1413
1414fail:
9b7a787d 1415 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1416 xfrm_state_put(xfrm[nx]);
1417 return error;
1418}
1419
4e81bb83 1420static int
a6c2e611
DM		 1421xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1422 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1423{
41a49cc3
MN		 1424 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1425 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1426 int cnx = 0;
1427 int error;
1428 int ret;
1429 int i;
1430
1431 for (i = 0; i < npols; i++) {
1432 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1433 error = -ENOBUFS;
1434 goto fail;
1435 }
41a49cc3
MN		 1436
1437 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1438 if (ret < 0) {
1439 error = ret;
1440 goto fail;
1441 } else
1442 cnx += ret;
1443 }
1444
41a49cc3
MN		 1445 /* found states are sorted for outbound processing */
1446 if (npols > 1)
1447 xfrm_state_sort(xfrm, tpp, cnx, family);
1448
4e81bb83
MN		 1449 return cnx;
1450
1451 fail:
9b7a787d 1452 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1453 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1454 return error;
1455
1456}
1457
f5e2bb4f 1458static int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286 1459{
37b10383 1460 const struct xfrm_policy_afinfo *afinfo;
143a4454 1461 int tos;
1da177e4 1462
f5e2bb4f 1463 afinfo = xfrm_policy_get_afinfo(family);
143a4454
XL		 1464 if (!afinfo)
1465 return 0;
1466
1467 tos = afinfo->get_tos(fl);
25ee3286 1468
bdba9fe0 1469 rcu_read_unlock();
25ee3286
HX		 1470
1471 return tos;
1472}
1473
d7c7544c 1474static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1475{
37b10383 1476 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1477 struct dst_ops *dst_ops;
25ee3286
HX		 1478 struct xfrm_dst *xdst;
1479
1480 if (!afinfo)
1481 return ERR_PTR(-EINVAL);
1482
d7c7544c
AD		 1483 switch (family) {
1484 case AF_INET:
1485 dst_ops = &net->xfrm.xfrm4_dst_ops;
1486 break;
dfd56b8b 1487#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1488 case AF_INET6:
1489 dst_ops = &net->xfrm.xfrm6_dst_ops;
1490 break;
1491#endif
1492 default:
1493 BUG();
1494 }
b2a9c0ed 1495 xdst = dst_alloc(dst_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
25ee3286 1496
d4cae562 1497 if (likely(xdst)) {
141e369d
SK		 1498 struct dst_entry *dst = &xdst->u.dst;
1499
1500 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
d4cae562 1501 } else
0b150932 1502 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1503
bdba9fe0 1504 rcu_read_unlock();
d4cae562 1505
25ee3286
HX		 1506 return xdst;
1507}
1508
a1b05140
MN		 1509static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1510 int nfheader_len)
1511{
37b10383 1512 const struct xfrm_policy_afinfo *afinfo =
a1b05140
MN		 1513 xfrm_policy_get_afinfo(dst->ops->family);
1514 int err;
1515
1516 if (!afinfo)
1517 return -EINVAL;
1518
1519 err = afinfo->init_path(path, dst, nfheader_len);
1520
bdba9fe0 1521 rcu_read_unlock();
a1b05140
MN		 1522
1523 return err;
1524}
1525
87c1e12b 1526static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1527 const struct flowi *fl)
25ee3286 1528{
37b10383 1529 const struct xfrm_policy_afinfo *afinfo =
25ee3286
HX		 1530 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1531 int err;
1532
1533 if (!afinfo)
1da177e4 1534 return -EINVAL;
25ee3286 1535
87c1e12b 1536 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1537
bdba9fe0 1538 rcu_read_unlock();
25ee3286 1539
1da177e4
LT		 1540 return err;
1541}
1542
80c802f3 1543
25ee3286
HX		 1544/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1545 * all the metrics... Shortly, bundle a bundle.
1546 */
1547
1548static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
5492093d
DM		 1549 struct xfrm_state **xfrm,
1550 struct xfrm_dst **bundle,
1551 int nx,
98313ada 1552 const struct flowi *fl,
25ee3286
HX		 1553 struct dst_entry *dst)
1554{
d7c7544c 1555 struct net *net = xp_net(policy);
25ee3286
HX		 1556 unsigned long now = jiffies;
1557 struct net_device *dev;
43a4dea4 1558 struct xfrm_mode *inner_mode;
45b018be
DM		 1559 struct xfrm_dst *xdst_prev = NULL;
1560 struct xfrm_dst *xdst0 = NULL;
25ee3286
HX		 1561 int i = 0;
1562 int err;
1563 int header_len = 0;
a1b05140 1564 int nfheader_len = 0;
25ee3286
HX		 1565 int trailer_len = 0;
1566 int tos;
1567 int family = policy->selector.family;
9bb182a7
YH		 1568 xfrm_address_t saddr, daddr;
1569
1570 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1571
1572 tos = xfrm_get_tos(fl, family);
25ee3286
HX		 1573
1574 dst_hold(dst);
1575
1576 for (; i < nx; i++) {
d7c7544c 1577 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1578 struct dst_entry *dst1 = &xdst->u.dst;
1579
1580 err = PTR_ERR(xdst);
1581 if (IS_ERR(xdst)) {
1582 dst_release(dst);
1583 goto put_states;
1584 }
1585
5492093d 1586 bundle[i] = xdst;
45b018be
DM		 1587 if (!xdst_prev)
1588 xdst0 = xdst;
10a7ef33
DM		 1589 else
1590 /* Ref count is taken during xfrm_alloc_dst()
1591 * No need to do dst_clone() on dst1
1592 */
45b018be 1593 xfrm_dst_set_child(xdst_prev, &xdst->u.dst);
10a7ef33 1594
43a4dea4
SK		 1595 if (xfrm[i]->sel.family == AF_UNSPEC) {
1596 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1597 xfrm_af2proto(family));
1598 if (!inner_mode) {
1599 err = -EAFNOSUPPORT;
1600 dst_release(dst);
1601 goto put_states;
1602 }
1603 } else
1604 inner_mode = xfrm[i]->inner_mode;
1605
25ee3286 1606 xdst->route = dst;
defb3519 1607 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1608
1609 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1610 family = xfrm[i]->props.family;
42a7b32b 1611 dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
077fbac4
LC		 1612 &saddr, &daddr, family,
1613 xfrm[i]->props.output_mark);
25ee3286
HX		 1614 err = PTR_ERR(dst);
1615 if (IS_ERR(dst))
1616 goto put_states;
1617 } else
1618 dst_hold(dst);
1619
1620 dst1->xfrm = xfrm[i];
80c802f3 1621 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1622
f5b0a874 1623 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1624 dst1->flags |= DST_HOST;
1625 dst1->lastuse = now;
1626
1627 dst1->input = dst_discard;
43a4dea4 1628 dst1->output = inner_mode->afinfo->output;
25ee3286 1629
45b018be 1630 xdst_prev = xdst;
25ee3286
HX		 1631
1632 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1633 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1634 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1635 trailer_len += xfrm[i]->props.trailer_len;
1636 }
1637
45b018be 1638 xfrm_dst_set_child(xdst_prev, dst);
0f6c480f 1639 xdst0->path = dst;
25ee3286
HX		 1640
1641 err = -ENODEV;
1642 dev = dst->dev;
1643 if (!dev)
1644 goto free_dst;
1645
45b018be 1646 xfrm_init_path(xdst0, dst, nfheader_len);
5492093d 1647 xfrm_init_pmtu(bundle, nx);
25ee3286 1648
45b018be
DM		 1649 for (xdst_prev = xdst0; xdst_prev != (struct xfrm_dst *)dst;
1650 xdst_prev = (struct xfrm_dst *) xfrm_dst_child(&xdst_prev->u.dst)) {
1651 err = xfrm_fill_dst(xdst_prev, dev, fl);
25ee3286
HX		 1652 if (err)
1653 goto free_dst;
1654
45b018be
DM		 1655 xdst_prev->u.dst.header_len = header_len;
1656 xdst_prev->u.dst.trailer_len = trailer_len;
1657 header_len -= xdst_prev->u.dst.xfrm->props.header_len;
1658 trailer_len -= xdst_prev->u.dst.xfrm->props.trailer_len;
25ee3286
HX		 1659 }
1660
45b018be 1661 return &xdst0->u.dst;
25ee3286
HX		 1662
1663put_states:
1664 for (; i < nx; i++)
1665 xfrm_state_put(xfrm[i]);
1666free_dst:
45b018be
DM		 1667 if (xdst0)
1668 dst_release_immediate(&xdst0->u.dst);
38369f54
SK		 1669
1670 return ERR_PTR(err);
25ee3286
HX		 1671}
1672
73ff93cd 1673static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1674 struct xfrm_policy **pols,
1675 int *num_pols, int *num_xfrms)
1676{
1677 int i;
1678
1679 if (*num_pols == 0 || !pols[0]) {
1680 *num_pols = 0;
1681 *num_xfrms = 0;
1682 return 0;
1683 }
1684 if (IS_ERR(pols[0]))
1685 return PTR_ERR(pols[0]);
1686
1687 *num_xfrms = pols[0]->xfrm_nr;
1688
1689#ifdef CONFIG_XFRM_SUB_POLICY
1690 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1691 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1692 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1693 XFRM_POLICY_TYPE_MAIN,
1694 fl, family,
1695 XFRM_POLICY_OUT);
1696 if (pols[1]) {
1697 if (IS_ERR(pols[1])) {
1698 xfrm_pols_put(pols, *num_pols);
1699 return PTR_ERR(pols[1]);
1700 }
02d0892f 1701 (*num_pols)++;
80c802f3
TT		 1702 (*num_xfrms) += pols[1]->xfrm_nr;
1703 }
1704 }
1705#endif
1706 for (i = 0; i < *num_pols; i++) {
1707 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1708 *num_xfrms = -1;
1709 break;
1710 }
1711 }
1712
1713 return 0;
1714
1715}
1716
ec30d78c
FW		 1717static void xfrm_last_dst_update(struct xfrm_dst *xdst, struct xfrm_dst *old)
1718{
1719 this_cpu_write(xfrm_last_dst, xdst);
1720 if (old)
1721 dst_release(&old->u.dst);
1722}
1723
1724static void __xfrm_pcpu_work_fn(void)
1725{
1726 struct xfrm_dst *old;
1727
1728 old = this_cpu_read(xfrm_last_dst);
1729 if (old && !xfrm_bundle_ok(old))
1730 xfrm_last_dst_update(NULL, old);
1731}
1732
1733static void xfrm_pcpu_work_fn(struct work_struct *work)
1734{
1735 local_bh_disable();
1736 rcu_read_lock();
1737 __xfrm_pcpu_work_fn();
1738 rcu_read_unlock();
1739 local_bh_enable();
1740}
1741
1742void xfrm_policy_cache_flush(void)
1743{
1744 struct xfrm_dst *old;
415a1329 1745 bool found = false;
ec30d78c
FW		 1746 int cpu;
1747
b1bdcb59
FW		 1748 might_sleep();
1749
ec30d78c
FW		 1750 local_bh_disable();
1751 rcu_read_lock();
1752 for_each_possible_cpu(cpu) {
1753 old = per_cpu(xfrm_last_dst, cpu);
1754 if (old && !xfrm_bundle_ok(old)) {
1755 if (smp_processor_id() == cpu) {
1756 __xfrm_pcpu_work_fn();
1757 continue;
1758 }
1759 found = true;
1760 break;
1761 }
1762 }
1763
1764 rcu_read_unlock();
1765 local_bh_enable();
1766
1767 if (!found)
1768 return;
1769
1770 get_online_cpus();
1771
1772 for_each_possible_cpu(cpu) {
1773 bool bundle_release;
1774
1775 rcu_read_lock();
1776 old = per_cpu(xfrm_last_dst, cpu);
1777 bundle_release = old && !xfrm_bundle_ok(old);
1778 rcu_read_unlock();
1779
1780 if (!bundle_release)
1781 continue;
1782
1783 if (cpu_online(cpu)) {
1784 schedule_work_on(cpu, &xfrm_pcpu_work[cpu]);
1785 continue;
1786 }
1787
1788 rcu_read_lock();
1789 old = per_cpu(xfrm_last_dst, cpu);
1790 if (old && !xfrm_bundle_ok(old)) {
1791 per_cpu(xfrm_last_dst, cpu) = NULL;
1792 dst_release(&old->u.dst);
1793 }
1794 rcu_read_unlock();
1795 }
1796
1797 put_online_cpus();
1798}
1799
cf379667
FW		 1800static bool xfrm_xdst_can_reuse(struct xfrm_dst *xdst,
1801 struct xfrm_state * const xfrm[],
1802 int num)
ec30d78c 1803{
cf379667
FW		 1804 const struct dst_entry *dst = &xdst->u.dst;
1805 int i;
ec30d78c 1806
cf379667
FW		 1807 if (xdst->num_xfrms != num)
1808 return false;
ec30d78c 1809
cf379667
FW		 1810 for (i = 0; i < num; i++) {
1811 if (!dst || dst->xfrm != xfrm[i])
1812 return false;
b92cf4aa 1813 dst = xfrm_dst_child(dst);
cf379667 1814 }
ec30d78c 1815
cf379667 1816 return xfrm_bundle_ok(xdst);
ec30d78c
FW		 1817}
1818
80c802f3
TT		 1819static struct xfrm_dst *
1820xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1821 const struct flowi *fl, u16 family,
80c802f3
TT		 1822 struct dst_entry *dst_orig)
1823{
1824 struct net *net = xp_net(pols[0]);
1825 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
5492093d 1826 struct xfrm_dst *bundle[XFRM_MAX_DEPTH];
ec30d78c 1827 struct xfrm_dst *xdst, *old;
80c802f3 1828 struct dst_entry *dst;
80c802f3
TT		 1829 int err;
1830
cf379667
FW		 1831 /* Try to instantiate a bundle */
1832 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1833 if (err <= 0) {
1834 if (err != 0 && err != -EAGAIN)
1835 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1836 return ERR_PTR(err);
1837 }
1838
ec30d78c
FW		 1839 xdst = this_cpu_read(xfrm_last_dst);
1840 if (xdst &&
1841 xdst->u.dst.dev == dst_orig->dev &&
1842 xdst->num_pols == num_pols &&
ec30d78c 1843 memcmp(xdst->pols, pols,
13ead5c4 1844 sizeof(struct xfrm_policy *) * num_pols) == 0 &&
cf379667 1845 xfrm_xdst_can_reuse(xdst, xfrm, err)) {
ec30d78c 1846 dst_hold(&xdst->u.dst);
d2950278 1847 xfrm_pols_put(pols, num_pols);
cf379667
FW		 1848 while (err > 0)
1849 xfrm_state_put(xfrm[--err]);
ec30d78c
FW		 1850 return xdst;
1851 }
1852
1853 old = xdst;
80c802f3 1854
5492093d 1855 dst = xfrm_bundle_create(pols[0], xfrm, bundle, err, fl, dst_orig);
80c802f3
TT		 1856 if (IS_ERR(dst)) {
1857 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1858 return ERR_CAST(dst);
1859 }
1860
1861 xdst = (struct xfrm_dst *)dst;
1862 xdst->num_xfrms = err;
80c802f3 1863 xdst->num_pols = num_pols;
3e94c2dc 1864 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1865 xdst->policy_genid = atomic_read(&pols[0]->genid);
1866
ec30d78c
FW		 1867 atomic_set(&xdst->u.dst.__refcnt, 2);
1868 xfrm_last_dst_update(xdst, old);
1869
80c802f3
TT		 1870 return xdst;
1871}
1872
c3aed709 1873static void xfrm_policy_queue_process(struct timer_list *t)
a0073fe1 1874{
a0073fe1
SK		 1875 struct sk_buff *skb;
1876 struct sock *sk;
1877 struct dst_entry *dst;
c3aed709 1878 struct xfrm_policy *pol = from_timer(pol, t, polq.hold_timer);
3f5312ae 1879 struct net *net = xp_net(pol);
a0073fe1
SK		 1880 struct xfrm_policy_queue *pq = &pol->polq;
1881 struct flowi fl;
1882 struct sk_buff_head list;
1883
1884 spin_lock(&pq->hold_queue.lock);
1885 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1886 if (!skb) {
1887 spin_unlock(&pq->hold_queue.lock);
1888 goto out;
1889 }
a0073fe1
SK		 1890 dst = skb_dst(skb);
1891 sk = skb->sk;
1892 xfrm_decode_session(skb, &fl, dst->ops->family);
1893 spin_unlock(&pq->hold_queue.lock);
1894
0f6c480f 1895 dst_hold(xfrm_dst_path(dst));
2471c981 1896 dst = xfrm_lookup(net, xfrm_dst_path(dst), &fl, sk, XFRM_LOOKUP_QUEUE);
a0073fe1
SK		 1897 if (IS_ERR(dst))
1898 goto purge_queue;
1899
1900 if (dst->flags & DST_XFRM_QUEUE) {
1901 dst_release(dst);
1902
1903 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1904 goto purge_queue;
1905
1906 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1907 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1908 xfrm_pol_hold(pol);
1909 goto out;
a0073fe1
SK		 1910 }
1911
1912 dst_release(dst);
1913
1914 __skb_queue_head_init(&list);
1915
1916 spin_lock(&pq->hold_queue.lock);
1917 pq->timeout = 0;
1918 skb_queue_splice_init(&pq->hold_queue, &list);
1919 spin_unlock(&pq->hold_queue.lock);
1920
1921 while (!skb_queue_empty(&list)) {
1922 skb = __skb_dequeue(&list);
1923
1924 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
0f6c480f
DM		 1925 dst_hold(xfrm_dst_path(skb_dst(skb)));
1926 dst = xfrm_lookup(net, xfrm_dst_path(skb_dst(skb)), &fl, skb->sk, 0);
a0073fe1 1927 if (IS_ERR(dst)) {
a0073fe1
SK		 1928 kfree_skb(skb);
1929 continue;
1930 }
1931
1932 nf_reset(skb);
1933 skb_dst_drop(skb);
1934 skb_dst_set(skb, dst);
1935
13206b6b 1936 dst_output(net, skb->sk, skb);
a0073fe1
SK		 1937 }
1938
e7d8f6cb
SK		 1939out:
1940 xfrm_pol_put(pol);
a0073fe1
SK		 1941 return;
1942
1943purge_queue:
1944 pq->timeout = 0;
1ee5e667 1945 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1946 xfrm_pol_put(pol);
a0073fe1
SK		 1947}
1948
ede2059d 1949static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1950{
1951 unsigned long sched_next;
1952 struct dst_entry *dst = skb_dst(skb);
1953 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1954 struct xfrm_policy *pol = xdst->pols[0];
1955 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1956
39bb5e62 1957 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1958 kfree_skb(skb);
1959 return 0;
1960 }
a0073fe1
SK		 1961
1962 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1963 kfree_skb(skb);
1964 return -EAGAIN;
1965 }
1966
1967 skb_dst_force(skb);
a0073fe1
SK		 1968
1969 spin_lock_bh(&pq->hold_queue.lock);
1970
1971 if (!pq->timeout)
1972 pq->timeout = XFRM_QUEUE_TMO_MIN;
1973
1974 sched_next = jiffies + pq->timeout;
1975
1976 if (del_timer(&pq->hold_timer)) {
1977 if (time_before(pq->hold_timer.expires, sched_next))
1978 sched_next = pq->hold_timer.expires;
e7d8f6cb 1979 xfrm_pol_put(pol);
a0073fe1
SK		 1980 }
1981
1982 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1983 if (!mod_timer(&pq->hold_timer, sched_next))
1984 xfrm_pol_hold(pol);
a0073fe1
SK		 1985
1986 spin_unlock_bh(&pq->hold_queue.lock);
1987
1988 return 0;
1989}
1990
1991static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 1992 struct xfrm_flo *xflo,
a0073fe1
SK		 1993 const struct flowi *fl,
1994 int num_xfrms,
1995 u16 family)
1996{
1997 int err;
1998 struct net_device *dev;
b8c203b2 1999 struct dst_entry *dst;
a0073fe1
SK		 2000 struct dst_entry *dst1;
2001 struct xfrm_dst *xdst;
2002
2003 xdst = xfrm_alloc_dst(net, family);
2004 if (IS_ERR(xdst))
2005 return xdst;
2006
b8c203b2
SK		 2007 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2008 net->xfrm.sysctl_larval_drop ||
2009 num_xfrms <= 0)
a0073fe1
SK		 2010 return xdst;
2011
b8c203b2 2012 dst = xflo->dst_orig;
a0073fe1
SK		 2013 dst1 = &xdst->u.dst;
2014 dst_hold(dst);
2015 xdst->route = dst;
2016
2017 dst_copy_metrics(dst1, dst);
2018
2019 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2020 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2021 dst1->lastuse = jiffies;
2022
2023 dst1->input = dst_discard;
2024 dst1->output = xdst_queue_output;
2025
2026 dst_hold(dst);
45b018be 2027 xfrm_dst_set_child(xdst, dst);
0f6c480f 2028 xdst->path = dst;
a0073fe1
SK		 2029
2030 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2031
2032 err = -ENODEV;
2033 dev = dst->dev;
2034 if (!dev)
2035 goto free_dst;
2036
2037 err = xfrm_fill_dst(xdst, dev, fl);
2038 if (err)
2039 goto free_dst;
2040
2041out:
2042 return xdst;
2043
2044free_dst:
2045 dst_release(dst1);
2046 xdst = ERR_PTR(err);
2047 goto out;
2048}
2049
bd45c539 2050static struct xfrm_dst *
3ca28286 2051xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir, struct xfrm_flo *xflo)
80c802f3 2052{
80c802f3 2053 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
855dad99 2054 int num_pols = 0, num_xfrms = 0, err;
bd45c539 2055 struct xfrm_dst *xdst;
80c802f3 2056
80c802f3
TT		 2057 /* Resolve policies to use if we couldn't get them from
2058 * previous cache entry */
855dad99 2059 num_pols = 1;
86dc8ee0 2060 pols[0] = xfrm_policy_lookup(net, fl, family, dir);
855dad99 2061 err = xfrm_expand_policies(fl, family, pols,
80c802f3 2062 &num_pols, &num_xfrms);
855dad99
FW		 2063 if (err < 0)
2064 goto inc_error;
2065 if (num_pols == 0)
2066 return NULL;
2067 if (num_xfrms <= 0)
2068 goto make_dummy_bundle;
80c802f3 2069
76a42011 2070 local_bh_disable();
bd45c539 2071 xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
76a42011
SK		 2072 xflo->dst_orig);
2073 local_bh_enable();
2074
bd45c539
FW		 2075 if (IS_ERR(xdst)) {
2076 err = PTR_ERR(xdst);
80c802f3
TT		 2077 if (err != -EAGAIN)
2078 goto error;
855dad99 2079 goto make_dummy_bundle;
bd45c539 2080 } else if (xdst == NULL) {
d809ec89 2081 num_xfrms = 0;
855dad99 2082 goto make_dummy_bundle;
80c802f3
TT		 2083 }
2084
bd45c539 2085 return xdst;
80c802f3
TT		 2086
2087make_dummy_bundle:
2088 /* We found policies, but there's no bundles to instantiate:
2089 * either because the policy blocks, has no transformations or
2090 * we could not build template (no xfrm_states).*/
b8c203b2 2091 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2092 if (IS_ERR(xdst)) {
2093 xfrm_pols_put(pols, num_pols);
2094 return ERR_CAST(xdst);
2095 }
2096 xdst->num_pols = num_pols;
2097 xdst->num_xfrms = num_xfrms;
3e94c2dc 2098 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3 2099
bd45c539 2100 return xdst;
80c802f3
TT		 2101
2102inc_error:
2103 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2104error:
855dad99 2105 xfrm_pols_put(pols, num_pols);
80c802f3
TT		 2106 return ERR_PTR(err);
2107}
1da177e4 2108
2774c131
DM		 2109static struct dst_entry *make_blackhole(struct net *net, u16 family,
2110 struct dst_entry *dst_orig)
2111{
37b10383 2112 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2774c131
DM		 2113 struct dst_entry *ret;
2114
2115 if (!afinfo) {
2116 dst_release(dst_orig);
433a1954 2117 return ERR_PTR(-EINVAL);
2774c131
DM		 2118 } else {
2119 ret = afinfo->blackhole_route(net, dst_orig);
2120 }
bdba9fe0 2121 rcu_read_unlock();
2774c131
DM		 2122
2123 return ret;
2124}
2125
1da177e4
LT		 2126/* Main function: finds/creates a bundle for given flow.
2127 *
2128 * At the moment we eat a raw IP route. Mostly to speed up lookups
2129 * on interfaces with disabled IPsec.
2130 */
452edd59
DM		 2131struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2132 const struct flowi *fl,
6f9c9615 2133 const struct sock *sk, int flags)
1da177e4 2134{
4e81bb83 2135 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3 2136 struct xfrm_dst *xdst;
452edd59 2137 struct dst_entry *dst, *route;
80c802f3 2138 u16 family = dst_orig->ops->family;
aff669bc 2139 u8 dir = XFRM_POLICY_OUT;
4b021628 2140 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2141
80c802f3
TT		 2142 dst = NULL;
2143 xdst = NULL;
2144 route = NULL;
4e81bb83 2145
bd5eb35f 2146 sk = sk_const_to_full_sk(sk);
f7944fb1 2147 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3 2148 num_pols = 1;
4c86d777 2149 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, family);
80c802f3
TT		 2150 err = xfrm_expand_policies(fl, family, pols,
2151 &num_pols, &num_xfrms);
2152 if (err < 0)
75b8c133 2153 goto dropdst;
80c802f3
TT		 2154
2155 if (num_pols) {
2156 if (num_xfrms <= 0) {
2157 drop_pols = num_pols;
2158 goto no_transform;
2159 }
2160
76a42011 2161 local_bh_disable();
80c802f3
TT		 2162 xdst = xfrm_resolve_and_create_bundle(
2163 pols, num_pols, fl,
2164 family, dst_orig);
76a42011
SK		 2165 local_bh_enable();
2166
80c802f3
TT		 2167 if (IS_ERR(xdst)) {
2168 xfrm_pols_put(pols, num_pols);
2169 err = PTR_ERR(xdst);
2170 goto dropdst;
d809ec89
TT		 2171 } else if (xdst == NULL) {
2172 num_xfrms = 0;
2173 drop_pols = num_pols;
2174 goto no_transform;
80c802f3
TT		 2175 }
2176
80c802f3 2177 route = xdst->route;
0aa64774 2178 }
3bccfbc7 2179 }
1da177e4 2180
80c802f3 2181 if (xdst == NULL) {
b8c203b2
SK		 2182 struct xfrm_flo xflo;
2183
2184 xflo.dst_orig = dst_orig;
2185 xflo.flags = flags;
2186
1da177e4 2187 /* To accelerate a bit... */
2518c7c2 2188 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2189 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2190 goto nopol;
1da177e4 2191
bd45c539
FW		 2192 xdst = xfrm_bundle_lookup(net, fl, family, dir, &xflo);
2193 if (xdst == NULL)
80c802f3 2194 goto nopol;
bd45c539
FW		 2195 if (IS_ERR(xdst)) {
2196 err = PTR_ERR(xdst);
75b8c133 2197 goto dropdst;
d66e37a9 2198 }
80c802f3
TT		 2199
2200 num_pols = xdst->num_pols;
2201 num_xfrms = xdst->num_xfrms;
3e94c2dc 2202 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2203 route = xdst->route;
2204 }
2205
2206 dst = &xdst->u.dst;
2207 if (route == NULL && num_xfrms > 0) {
2208 /* The only case when xfrm_bundle_lookup() returns a
2209 * bundle with null route, is when the template could
2210 * not be resolved. It means policies are there, but
2211 * bundle could not be created, since we don't yet
2212 * have the xfrm_state's. We need to wait for KM to
2213 * negotiate new SA's or bail out with error.*/
2214 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2215 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2216 err = -EREMOTE;
2217 goto error;
80c802f3 2218 }
80c802f3 2219
5b8ef341 2220 err = -EAGAIN;
80c802f3
TT		 2221
2222 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2223 goto error;
1da177e4
LT		 2224 }
2225
80c802f3
TT		 2226no_transform:
2227 if (num_pols == 0)
8b7817f3 2228 goto nopol;
1da177e4 2229
80c802f3
TT		 2230 if ((flags & XFRM_LOOKUP_ICMP) &&
2231 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2232 err = -ENOENT;
8b7817f3 2233 goto error;
80c802f3 2234 }
8b7817f3 2235
80c802f3
TT		 2236 for (i = 0; i < num_pols; i++)
2237 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2238
80c802f3 2239 if (num_xfrms < 0) {
1da177e4 2240 /* Prohibit the flow */
59c9940e 2241 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2242 err = -EPERM;
2243 goto error;
80c802f3
TT		 2244 } else if (num_xfrms > 0) {
2245 /* Flow transformed */
80c802f3
TT		 2246 dst_release(dst_orig);
2247 } else {
2248 /* Flow passes untransformed */
2249 dst_release(dst);
452edd59 2250 dst = dst_orig;
1da177e4 2251 }
80c802f3
TT		 2252ok:
2253 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2254 if (dst && dst->xfrm &&
2255 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2256 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2257 return dst;
1da177e4 2258
80c802f3 2259nopol:
452edd59
DM		 2260 if (!(flags & XFRM_LOOKUP_ICMP)) {
2261 dst = dst_orig;
80c802f3 2262 goto ok;
452edd59 2263 }
80c802f3 2264 err = -ENOENT;
1da177e4 2265error:
80c802f3 2266 dst_release(dst);
75b8c133 2267dropdst:
ac37e251
W		 2268 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2269 dst_release(dst_orig);
80c802f3 2270 xfrm_pols_put(pols, drop_pols);
452edd59 2271 return ERR_PTR(err);
1da177e4
LT		 2272}
2273EXPORT_SYMBOL(xfrm_lookup);
2274
f92ee619
SK		 2275/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2276 * Otherwise we may send out blackholed packets.
2277 */
2278struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2279 const struct flowi *fl,
6f9c9615 2280 const struct sock *sk, int flags)
f92ee619 2281{
b8c203b2 2282 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2283 flags | XFRM_LOOKUP_QUEUE |
2284 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2285
2286 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2287 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2288
2289 return dst;
2290}
2291EXPORT_SYMBOL(xfrm_lookup_route);
2292
df0ba92a 2293static inline int
8f029de2 2294xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2295{
2296 struct xfrm_state *x;
df0ba92a
MN		 2297
2298 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2299 return 0;
2300 x = skb->sp->xvec[idx];
2301 if (!x->type->reject)
2302 return 0;
1ecafede 2303 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2304}
2305
1da177e4
LT		 2306/* When skb is transformed back to its "native" form, we have to
2307 * check policy restrictions. At the moment we make this in maximally
2308 * stupid way. Shame on me. :-) Of course, connected sockets must
2309 * have policy cached at them.
2310 */
2311
2312static inline int
7db454b9 2313xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2314 unsigned short family)
2315{
2316 if (xfrm_state_kern(x))
928ba416 2317 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2318 return x->id.proto == tmpl->id.proto &&
2319 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2320 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2321 x->props.mode == tmpl->mode &&
c5d18e98 2322 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2323 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2324 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2325 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2326}
2327
df0ba92a
MN		 2328/*
2329 * 0 or more than 0 is returned when validation is succeeded (either bypass
2330 * because of optional transport mode, or next index of the mathced secpath
2331 * state with the template.
2332 * -1 is returned when no matching template is found.
2333 * Otherwise "-2 - errored_index" is returned.
2334 */
1da177e4 2335static inline int
22cccb7e 2336xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2337 unsigned short family)
2338{
2339 int idx = start;
2340
2341 if (tmpl->optional) {
7e49e6de 2342 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2343 return start;
2344 } else
2345 start = -1;
2346 for (; idx < sp->len; idx++) {
dbe5b4aa 2347 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2348 return ++idx;
df0ba92a
MN		 2349 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2350 if (start == -1)
2351 start = -2-idx;
1da177e4 2352 break;
df0ba92a 2353 }
1da177e4
LT		 2354 }
2355 return start;
2356}
2357
d5422efe
HX		 2358int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2359 unsigned int family, int reverse)
1da177e4 2360{
37b10383 2361 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2362 int err;
1da177e4
LT		 2363
2364 if (unlikely(afinfo == NULL))
2365 return -EAFNOSUPPORT;
2366
d5422efe 2367 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2368 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
bdba9fe0 2369 rcu_read_unlock();
e0d1caa7 2370 return err;
1da177e4 2371}
d5422efe 2372EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2373
9a7386ec 2374static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2375{
2376 for (; k < sp->len; k++) {
df0ba92a 2377 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2378 *idxp = k;
1da177e4 2379 return 1;
df0ba92a 2380 }
1da177e4
LT		 2381 }
2382
2383 return 0;
2384}
2385
a716c119 2386int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2387 unsigned short family)
2388{
f6e1e25d 2389 struct net *net = dev_net(skb->dev);
1da177e4 2390 struct xfrm_policy *pol;
4e81bb83
MN		 2391 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2392 int npols = 0;
2393 int xfrm_nr;
2394 int pi;
d5422efe 2395 int reverse;
1da177e4 2396 struct flowi fl;
df0ba92a 2397 int xerr_idx = -1;
1da177e4 2398
d5422efe
HX		 2399 reverse = dir & ~XFRM_POLICY_MASK;
2400 dir &= XFRM_POLICY_MASK;
d5422efe 2401
0aa64774 2402 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2403 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2404 return 0;
0aa64774
MN		 2405 }
2406
eb9c7ebe 2407 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2408
2409 /* First, check used SA against their selectors. */
2410 if (skb->sp) {
2411 int i;
2412
9b7a787d 2413 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2414 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2415 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2416 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2417 return 0;
0aa64774 2418 }
1da177e4
LT		 2419 }
2420 }
2421
2422 pol = NULL;
bd5eb35f 2423 sk = sk_to_full_sk(sk);
3bccfbc7 2424 if (sk && sk->sk_policy[dir]) {
4c86d777 2425 pol = xfrm_sk_policy_lookup(sk, dir, &fl, family);
0aa64774 2426 if (IS_ERR(pol)) {
59c9940e 2427 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2428 return 0;
0aa64774 2429 }
3bccfbc7 2430 }
1da177e4 2431
86dc8ee0
FW		 2432 if (!pol)
2433 pol = xfrm_policy_lookup(net, &fl, family, dir);
1da177e4 2434
0aa64774 2435 if (IS_ERR(pol)) {
59c9940e 2436 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2437 return 0;
0aa64774 2438 }
134b0fc5 2439
df0ba92a 2440 if (!pol) {
d1d9facf 2441 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2442 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2443 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2444 return 0;
2445 }
2446 return 1;
2447 }
1da177e4 2448
9d729f72 2449 pol->curlft.use_time = get_seconds();
1da177e4 2450
4e81bb83 2451 pols[0] = pol;
02d0892f 2452 npols++;
4e81bb83
MN		 2453#ifdef CONFIG_XFRM_SUB_POLICY
2454 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2455 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2456 &fl, family,
2457 XFRM_POLICY_IN);
2458 if (pols[1]) {
0aa64774 2459 if (IS_ERR(pols[1])) {
59c9940e 2460 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2461 return 0;
0aa64774 2462 }
9d729f72 2463 pols[1]->curlft.use_time = get_seconds();
02d0892f 2464 npols++;
4e81bb83
MN		 2465 }
2466 }
2467#endif
2468
1da177e4
LT		 2469 if (pol->action == XFRM_POLICY_ALLOW) {
2470 struct sec_path *sp;
2471 static struct sec_path dummy;
4e81bb83 2472 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2473 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2474 struct xfrm_tmpl **tpp = tp;
2475 int ti = 0;
1da177e4
LT		 2476 int i, k;
2477
2478 if ((sp = skb->sp) == NULL)
2479 sp = &dummy;
2480
4e81bb83
MN		 2481 for (pi = 0; pi < npols; pi++) {
2482 if (pols[pi] != pol &&
0aa64774 2483 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2484 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2485 goto reject;
0aa64774
MN		 2486 }
2487 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2488 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2489 goto reject_error;
0aa64774 2490 }
4e81bb83
MN		 2491 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2492 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2493 }
2494 xfrm_nr = ti;
41a49cc3 2495 if (npols > 1) {
283bc9f3 2496 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2497 tpp = stp;
2498 }
4e81bb83 2499
1da177e4
LT		 2500 /* For each tunnel xfrm, find the first matching tmpl.
2501 * For each tmpl before that, find corresponding xfrm.
2502 * Order is _important_. Later we will implement
2503 * some barriers, but at the moment barriers
2504 * are implied between each two transformations.
2505 */
4e81bb83
MN		 2506 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2507 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2508 if (k < 0) {
d1d9facf
JM		 2509 if (k < -1)
2510 /* "-2 - errored_index" returned */
2511 xerr_idx = -(2+k);
59c9940e 2512 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2513 goto reject;
df0ba92a 2514 }
1da177e4
LT		 2515 }
2516
0aa64774 2517 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2518 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2519 goto reject;
0aa64774 2520 }
1da177e4 2521
4e81bb83 2522 xfrm_pols_put(pols, npols);
1da177e4
LT		 2523 return 1;
2524 }
59c9940e 2525 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2526
2527reject:
df0ba92a 2528 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2529reject_error:
2530 xfrm_pols_put(pols, npols);
1da177e4
LT		 2531 return 0;
2532}
2533EXPORT_SYMBOL(__xfrm_policy_check);
2534
2535int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2536{
99a66657 2537 struct net *net = dev_net(skb->dev);
1da177e4 2538 struct flowi fl;
adf30907 2539 struct dst_entry *dst;
73137147 2540 int res = 1;
1da177e4 2541
0aa64774 2542 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2543 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2544 return 0;
0aa64774 2545 }
1da177e4 2546
fafeeb6c 2547 skb_dst_force(skb);
adf30907 2548
b8c203b2 2549 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2550 if (IS_ERR(dst)) {
73137147 2551 res = 0;
452edd59
DM		 2552 dst = NULL;
2553 }
adf30907
ED		 2554 skb_dst_set(skb, dst);
2555 return res;
1da177e4
LT		 2556}
2557EXPORT_SYMBOL(__xfrm_route_forward);
2558
d49c73c7
DM		 2559/* Optimize later using cookies and generation ids. */
2560
1da177e4
LT		 2561static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2562{
d49c73c7 2563 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2564 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2565 * get validated by dst_ops->check on every use. We do this
2566 * because when a normal route referenced by an XFRM dst is
2567 * obsoleted we do not go looking around for all parent
2568 * referencing XFRM dsts so that we can invalidate them. It
2569 * is just too much work. Instead we make the checks here on
2570 * every use. For example:
d49c73c7
DM		 2571 *
2572 * XFRM dst A --> IPv4 dst X
2573 *
2574 * X is the "xdst->route" of A (X is also the "dst->path" of A
2575 * in this example). If X is marked obsolete, "A" will not
2576 * notice. That's what we are validating here via the
2577 * stale_bundle() check.
2578 *
52df157f
WW		 2579 * When a dst is removed from the fib tree, DST_OBSOLETE_DEAD will
2580 * be marked on it.
09c75704 2581 * This will force stale_bundle() to fail on any xdst bundle with
52df157f 2582 * this dst linked in it.
399c180a 2583 */
d49c73c7
DM		 2584 if (dst->obsolete < 0 && !stale_bundle(dst))
2585 return dst;
2586
1da177e4
LT		 2587 return NULL;
2588}
2589
2590static int stale_bundle(struct dst_entry *dst)
2591{
12fdb4d3 2592 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2593}
2594
aabc9761 2595void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2596{
b92cf4aa 2597 while ((dst = xfrm_dst_child(dst)) && dst->xfrm && dst->dev == dev) {
c346dca1 2598 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2599 dev_hold(dst->dev);
1da177e4
LT		 2600 dev_put(dev);
2601 }
2602}
aabc9761 2603EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2604
2605static void xfrm_link_failure(struct sk_buff *skb)
2606{
2607 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2608}
2609
2610static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2611{
2612 if (dst) {
2613 if (dst->obsolete) {
2614 dst_release(dst);
2615 dst = NULL;
2616 }
2617 }
2618 return dst;
2619}
2620
5492093d 2621static void xfrm_init_pmtu(struct xfrm_dst **bundle, int nr)
1da177e4 2622{
5492093d
DM		 2623 while (nr--) {
2624 struct xfrm_dst *xdst = bundle[nr];
1da177e4 2625 u32 pmtu, route_mtu_cached;
5492093d 2626 struct dst_entry *dst;
1da177e4 2627
5492093d 2628 dst = &xdst->u.dst;
b92cf4aa 2629 pmtu = dst_mtu(xfrm_dst_child(dst));
1da177e4
LT		 2630 xdst->child_mtu_cached = pmtu;
2631
2632 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2633
2634 route_mtu_cached = dst_mtu(xdst->route);
2635 xdst->route_mtu_cached = route_mtu_cached;
2636
2637 if (pmtu > route_mtu_cached)
2638 pmtu = route_mtu_cached;
2639
defb3519 2640 dst_metric_set(dst, RTAX_MTU, pmtu);
5492093d 2641 }
1da177e4
LT		 2642}
2643
1da177e4
LT		 2644/* Check that the bundle accepts the flow and its components are
2645 * still valid.
2646 */
2647
12fdb4d3 2648static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4 2649{
5492093d 2650 struct xfrm_dst *bundle[XFRM_MAX_DEPTH];
1da177e4 2651 struct dst_entry *dst = &first->u.dst;
5492093d
DM		 2652 struct xfrm_dst *xdst;
2653 int start_from, nr;
1da177e4
LT		 2654 u32 mtu;
2655
0f6c480f 2656 if (!dst_check(xfrm_dst_path(dst), ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2657 (dst->dev && !netif_running(dst->dev)))
2658 return 0;
2659
a0073fe1
SK		 2660 if (dst->flags & DST_XFRM_QUEUE)
2661 return 1;
2662
5492093d 2663 start_from = nr = 0;
1da177e4
LT		 2664 do {
2665 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2666
1da177e4
LT		 2667 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2668 return 0;
80c802f3
TT		 2669 if (xdst->xfrm_genid != dst->xfrm->genid)
2670 return 0;
b1312c89
TT		 2671 if (xdst->num_pols > 0 &&
2672 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2673 return 0;
e53820de 2674
5492093d
DM		 2675 bundle[nr++] = xdst;
2676
b92cf4aa 2677 mtu = dst_mtu(xfrm_dst_child(dst));
1da177e4 2678 if (xdst->child_mtu_cached != mtu) {
5492093d 2679 start_from = nr;
1da177e4
LT		 2680 xdst->child_mtu_cached = mtu;
2681 }
2682
92d63dec 2683 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2684 return 0;
2685 mtu = dst_mtu(xdst->route);
2686 if (xdst->route_mtu_cached != mtu) {
5492093d 2687 start_from = nr;
1da177e4
LT		 2688 xdst->route_mtu_cached = mtu;
2689 }
2690
b92cf4aa 2691 dst = xfrm_dst_child(dst);
1da177e4
LT		 2692 } while (dst->xfrm);
2693
5492093d 2694 if (likely(!start_from))
1da177e4
LT		 2695 return 1;
2696
5492093d
DM		 2697 xdst = bundle[start_from - 1];
2698 mtu = xdst->child_mtu_cached;
2699 while (start_from--) {
2700 dst = &xdst->u.dst;
1da177e4
LT		 2701
2702 mtu = xfrm_state_mtu(dst->xfrm, mtu);
5492093d
DM		 2703 if (mtu > xdst->route_mtu_cached)
2704 mtu = xdst->route_mtu_cached;
defb3519 2705 dst_metric_set(dst, RTAX_MTU, mtu);
5492093d 2706 if (!start_from)
1da177e4
LT		 2707 break;
2708
5492093d
DM		 2709 xdst = bundle[start_from - 1];
2710 xdst->child_mtu_cached = mtu;
1da177e4
LT		 2711 }
2712
2713 return 1;
2714}
2715
0dbaee3b
DM		 2716static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2717{
0f6c480f 2718 return dst_metric_advmss(xfrm_dst_path(dst));
0dbaee3b
DM		 2719}
2720
ebb762f2 2721static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2722{
618f9bc7
SK		 2723 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2724
0f6c480f 2725 return mtu ? : dst_mtu(xfrm_dst_path(dst));
d33e4553
DM		 2726}
2727
1ecc9ad0
JA		 2728static const void *xfrm_get_dst_nexthop(const struct dst_entry *dst,
2729 const void *daddr)
63fca65d 2730{
0f6c480f 2731 while (dst->xfrm) {
63fca65d
JA		 2732 const struct xfrm_state *xfrm = dst->xfrm;
2733
013cb81e
SK		 2734 dst = xfrm_dst_child(dst);
2735
63fca65d
JA		 2736 if (xfrm->props.mode == XFRM_MODE_TRANSPORT)
2737 continue;
2738 if (xfrm->type->flags & XFRM_TYPE_REMOTE_COADDR)
2739 daddr = xfrm->coaddr;
2740 else if (!(xfrm->type->flags & XFRM_TYPE_LOCAL_COADDR))
2741 daddr = &xfrm->id.daddr;
2742 }
1ecc9ad0
JA		 2743 return daddr;
2744}
2745
2746static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2747 struct sk_buff *skb,
2748 const void *daddr)
2749{
0f6c480f 2750 const struct dst_entry *path = xfrm_dst_path(dst);
1ecc9ad0
JA		 2751
2752 if (!skb)
2753 daddr = xfrm_get_dst_nexthop(dst, daddr);
2754 return path->ops->neigh_lookup(path, skb, daddr);
2755}
2756
2757static void xfrm_confirm_neigh(const struct dst_entry *dst, const void *daddr)
2758{
0f6c480f 2759 const struct dst_entry *path = xfrm_dst_path(dst);
1ecc9ad0
JA		 2760
2761 daddr = xfrm_get_dst_nexthop(dst, daddr);
63fca65d
JA		 2762 path->ops->confirm_neigh(path, daddr);
2763}
2764
a2817d8b 2765int xfrm_policy_register_afinfo(const struct xfrm_policy_afinfo *afinfo, int family)
1da177e4
LT		 2766{
2767 int err = 0;
a2817d8b
FW		 2768
2769 if (WARN_ON(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
1da177e4 2770 return -EAFNOSUPPORT;
a2817d8b 2771
ef8531b6 2772 spin_lock(&xfrm_policy_afinfo_lock);
a2817d8b 2773 if (unlikely(xfrm_policy_afinfo[family] != NULL))
f31e8d4f 2774 err = -EEXIST;
1da177e4
LT		 2775 else {
2776 struct dst_ops *dst_ops = afinfo->dst_ops;
2777 if (likely(dst_ops->kmem_cachep == NULL))
2778 dst_ops->kmem_cachep = xfrm_dst_cache;
2779 if (likely(dst_ops->check == NULL))
2780 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2781 if (likely(dst_ops->default_advmss == NULL))
2782 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2783 if (likely(dst_ops->mtu == NULL))
2784 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2785 if (likely(dst_ops->negative_advice == NULL))
2786 dst_ops->negative_advice = xfrm_negative_advice;
2787 if (likely(dst_ops->link_failure == NULL))
2788 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2789 if (likely(dst_ops->neigh_lookup == NULL))
2790 dst_ops->neigh_lookup = xfrm_neigh_lookup;
63fca65d
JA		 2791 if (likely(!dst_ops->confirm_neigh))
2792 dst_ops->confirm_neigh = xfrm_confirm_neigh;
a2817d8b 2793 rcu_assign_pointer(xfrm_policy_afinfo[family], afinfo);
1da177e4 2794 }
ef8531b6 2795 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c 2796
1da177e4
LT		 2797 return err;
2798}
2799EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2800
a2817d8b 2801void xfrm_policy_unregister_afinfo(const struct xfrm_policy_afinfo *afinfo)
1da177e4 2802{
2b61997a 2803 struct dst_ops *dst_ops = afinfo->dst_ops;
a2817d8b 2804 int i;
2b61997a 2805
a2817d8b
FW		 2806 for (i = 0; i < ARRAY_SIZE(xfrm_policy_afinfo); i++) {
2807 if (xfrm_policy_afinfo[i] != afinfo)
2808 continue;
2809 RCU_INIT_POINTER(xfrm_policy_afinfo[i], NULL);
2810 break;
ef8531b6 2811 }
ef8531b6 2812
2b61997a 2813 synchronize_rcu();
ef8531b6 2814
2b61997a
FW		 2815 dst_ops->kmem_cachep = NULL;
2816 dst_ops->check = NULL;
2817 dst_ops->negative_advice = NULL;
2818 dst_ops->link_failure = NULL;
1da177e4
LT		 2819}
2820EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2821
558f82ef 2822#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2823static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2824{
c68cd1a0 2825 int rv;
698365fa
WC		 2826 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2827 if (!net->mib.xfrm_statistics)
558f82ef 2828 return -ENOMEM;
c68cd1a0
AD		 2829 rv = xfrm_proc_init(net);
2830 if (rv < 0)
698365fa 2831 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2832 return rv;
558f82ef 2833}
59c9940e
AD		 2834
2835static void xfrm_statistics_fini(struct net *net)
2836{
c68cd1a0 2837 xfrm_proc_fini(net);
698365fa 2838 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2839}
2840#else
2841static int __net_init xfrm_statistics_init(struct net *net)
2842{
2843 return 0;
2844}
2845
2846static void xfrm_statistics_fini(struct net *net)
2847{
2848}
558f82ef
MN		 2849#endif
2850
d62ddc21 2851static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2852{
2518c7c2
DM		 2853 unsigned int hmask, sz;
2854 int dir;
2855
d62ddc21
AD		 2856 if (net_eq(net, &init_net))
2857 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2858 sizeof(struct xfrm_dst),
e5d679f3 2859 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2860 NULL);
1da177e4 2861
2518c7c2
DM		 2862 hmask = 8 - 1;
2863 sz = (hmask+1) * sizeof(struct hlist_head);
2864
93b851c1
AD		 2865 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2866 if (!net->xfrm.policy_byidx)
2867 goto out_byidx;
8100bea7 2868 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2869
53c2e285 2870 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2871 struct xfrm_policy_hash *htab;
2872
dc2caba7 2873 net->xfrm.policy_count[dir] = 0;
53c2e285 2874 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2875 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2876
a35f6c5d 2877 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2878 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2879 if (!htab->table)
a35f6c5d
AD		 2880 goto out_bydst;
2881 htab->hmask = hmask;
b58555f1
CG		 2882 htab->dbits4 = 32;
2883 htab->sbits4 = 32;
2884 htab->dbits6 = 128;
2885 htab->sbits6 = 128;
2518c7c2 2886 }
880a6fab
CG		 2887 net->xfrm.policy_hthresh.lbits4 = 32;
2888 net->xfrm.policy_hthresh.rbits4 = 32;
2889 net->xfrm.policy_hthresh.lbits6 = 128;
2890 net->xfrm.policy_hthresh.rbits6 = 128;
2891
2892 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2893
adfcf0b2 2894 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2895 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2896 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21 2897 return 0;
93b851c1 2898
a35f6c5d
AD		 2899out_bydst:
2900 for (dir--; dir >= 0; dir--) {
2901 struct xfrm_policy_hash *htab;
2902
2903 htab = &net->xfrm.policy_bydst[dir];
2904 xfrm_hash_free(htab->table, sz);
2905 }
2906 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2907out_byidx:
2908 return -ENOMEM;
d62ddc21
AD		 2909}
2910
2911static void xfrm_policy_fini(struct net *net)
2912{
93b851c1 2913 unsigned int sz;
8b18f8ea 2914 int dir;
93b851c1 2915
7c2776ee
AD		 2916 flush_work(&net->xfrm.policy_hash_work);
2917#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2918 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 2919#endif
2e71029e 2920 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 2921
adfcf0b2 2922 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2923
53c2e285 2924 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 2925 struct xfrm_policy_hash *htab;
2926
8b18f8ea 2927 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2928
2929 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2930 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2931 WARN_ON(!hlist_empty(htab->table));
2932 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2933 }
2934
8100bea7 2935 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2936 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2937 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2938}
2939
d62ddc21
AD		 2940static int __net_init xfrm_net_init(struct net *net)
2941{
2942 int rv;
2943
c282222a
FW		 2944 /* Initialize the per-net locks here */
2945 spin_lock_init(&net->xfrm.xfrm_state_lock);
2946 spin_lock_init(&net->xfrm.xfrm_policy_lock);
2947 mutex_init(&net->xfrm.xfrm_cfg_mutex);
2948
59c9940e
AD		 2949 rv = xfrm_statistics_init(net);
2950 if (rv < 0)
2951 goto out_statistics;
d62ddc21
AD		 2952 rv = xfrm_state_init(net);
2953 if (rv < 0)
2954 goto out_state;
2955 rv = xfrm_policy_init(net);
2956 if (rv < 0)
2957 goto out_policy;
b27aeadb
AD		 2958 rv = xfrm_sysctl_init(net);
2959 if (rv < 0)
2960 goto out_sysctl;
283bc9f3 2961
d62ddc21
AD		 2962 return 0;
2963
b27aeadb
AD		 2964out_sysctl:
2965 xfrm_policy_fini(net);
d62ddc21
AD		 2966out_policy:
2967 xfrm_state_fini(net);
2968out_state:
59c9940e
AD		 2969 xfrm_statistics_fini(net);
2970out_statistics:
d62ddc21
AD		 2971 return rv;
2972}
2973
2974static void __net_exit xfrm_net_exit(struct net *net)
2975{
b27aeadb 2976 xfrm_sysctl_fini(net);
d62ddc21
AD		 2977 xfrm_policy_fini(net);
2978 xfrm_state_fini(net);
59c9940e 2979 xfrm_statistics_fini(net);
d62ddc21
AD		 2980}
2981
2982static struct pernet_operations __net_initdata xfrm_net_ops = {
2983 .init = xfrm_net_init,
2984 .exit = xfrm_net_exit,
2985};
2986
1da177e4
LT		 2987void __init xfrm_init(void)
2988{
ec30d78c
FW		 2989 int i;
2990
2991 xfrm_pcpu_work = kmalloc_array(NR_CPUS, sizeof(*xfrm_pcpu_work),
2992 GFP_KERNEL);
2993 BUG_ON(!xfrm_pcpu_work);
2994
2995 for (i = 0; i < NR_CPUS; i++)
2996 INIT_WORK(&xfrm_pcpu_work[i], xfrm_pcpu_work_fn);
2997
d62ddc21 2998 register_pernet_subsys(&xfrm_net_ops);
e9a441b6 2999 xfrm_dev_init();
30846090 3000 seqcount_init(&xfrm_policy_hash_generation);
1da177e4
LT		 3001 xfrm_input_init();
3002}
3003
ab5f5e8b 3004#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 3005static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
3006 struct audit_buffer *audit_buf)
ab5f5e8b 3007{
875179fa
PM		 3008 struct xfrm_sec_ctx *ctx = xp->security;
3009 struct xfrm_selector *sel = &xp->selector;
3010
3011 if (ctx)
ab5f5e8b 3012 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3013 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3014
9b7a787d 3015 switch (sel->family) {
ab5f5e8b 3016 case AF_INET:
21454aaa 3017 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3018 if (sel->prefixlen_s != 32)
3019 audit_log_format(audit_buf, " src_prefixlen=%d",
3020 sel->prefixlen_s);
21454aaa 3021 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3022 if (sel->prefixlen_d != 32)
3023 audit_log_format(audit_buf, " dst_prefixlen=%d",
3024 sel->prefixlen_d);
ab5f5e8b
JL		 3025 break;
3026 case AF_INET6:
5b095d98 3027 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3028 if (sel->prefixlen_s != 128)
3029 audit_log_format(audit_buf, " src_prefixlen=%d",
3030 sel->prefixlen_s);
5b095d98 3031 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3032 if (sel->prefixlen_d != 128)
3033 audit_log_format(audit_buf, " dst_prefixlen=%d",
3034 sel->prefixlen_d);
ab5f5e8b
JL		 3035 break;
3036 }
3037}
3038
2e71029e 3039void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3040{
3041 struct audit_buffer *audit_buf;
ab5f5e8b 3042
afeb14b4 3043 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3044 if (audit_buf == NULL)
3045 return;
2e71029e 3046 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3047 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3048 xfrm_audit_common_policyinfo(xp, audit_buf);
3049 audit_log_end(audit_buf);
3050}
3051EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3052
68277acc 3053void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3054 bool task_valid)
ab5f5e8b
JL		 3055{
3056 struct audit_buffer *audit_buf;
ab5f5e8b 3057
afeb14b4 3058 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3059 if (audit_buf == NULL)
3060 return;
2e71029e 3061 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3062 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3063 xfrm_audit_common_policyinfo(xp, audit_buf);
3064 audit_log_end(audit_buf);
3065}
3066EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3067#endif
3068
80c9abaa 3069#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3070static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3071 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3072{
3073 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3074 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3075 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3076 sel_cmp->family) &&
3077 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3078 sel_cmp->family) &&
80c9abaa
SS		 3079 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3080 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3081 return true;
80c9abaa
SS		 3082 }
3083 } else {
3084 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3085 return true;
80c9abaa
SS		 3086 }
3087 }
bc9b35ad 3088 return false;
80c9abaa
SS		 3089}
3090
3e94c2dc
WC		 3091static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3092 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3093{
3094 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3095 struct hlist_head *chain;
3096 u32 priority = ~0U;
3097
9d0380df 3098 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d549c4f 3099 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3100 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3101 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3102 pol->type == type) {
3103 ret = pol;
3104 priority = ret->priority;
3105 break;
3106 }
3107 }
8d549c4f 3108 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3109 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR		 3110 if ((pol->priority >= priority) && ret)
3111 break;
3112
80c9abaa 3113 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
8faf491e 3114 pol->type == type) {
80c9abaa
SS		 3115 ret = pol;
3116 break;
3117 }
3118 }
3119
586f2eb4 3120 xfrm_pol_hold(ret);
80c9abaa 3121
9d0380df 3122 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3123
3124 return ret;
3125}
3126
dd701754 3127static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3128{
3129 int match = 0;
3130
3131 if (t->mode == m->mode && t->id.proto == m->proto &&
3132 (m->reqid == 0 || t->reqid == m->reqid)) {
3133 switch (t->mode) {
3134 case XFRM_MODE_TUNNEL:
3135 case XFRM_MODE_BEET:
70e94e66
YH		 3136 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3137 m->old_family) &&
3138 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3139 m->old_family)) {
80c9abaa
SS		 3140 match = 1;
3141 }
3142 break;
3143 case XFRM_MODE_TRANSPORT:
3144 /* in case of transport mode, template does not store
3145 any IP addresses, hence we just compare mode and
3146 protocol */
3147 match = 1;
3148 break;
3149 default:
3150 break;
3151 }
3152 }
3153 return match;
3154}
3155
3156/* update endpoint address(es) of template(s) */
3157static int xfrm_policy_migrate(struct xfrm_policy *pol,
3158 struct xfrm_migrate *m, int num_migrate)
3159{
3160 struct xfrm_migrate *mp;
80c9abaa
SS		 3161 int i, j, n = 0;
3162
3163 write_lock_bh(&pol->lock);
12a169e7 3164 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3165 /* target policy has been deleted */
3166 write_unlock_bh(&pol->lock);
3167 return -ENOENT;
3168 }
3169
3170 for (i = 0; i < pol->xfrm_nr; i++) {
3171 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3172 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3173 continue;
3174 n++;
1bfcb10f
HX		 3175 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3176 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3177 continue;
3178 /* update endpoints */
3179 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3180 sizeof(pol->xfrm_vec[i].id.daddr));
3181 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3182 sizeof(pol->xfrm_vec[i].saddr));
3183 pol->xfrm_vec[i].encap_family = mp->new_family;
3184 /* flush bundles */
80c802f3 3185 atomic_inc(&pol->genid);
80c9abaa
SS		 3186 }
3187 }
3188
3189 write_unlock_bh(&pol->lock);
3190
3191 if (!n)
3192 return -ENODATA;
3193
3194 return 0;
3195}
3196
dd701754 3197static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3198{
3199 int i, j;
3200
3201 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3202 return -EINVAL;
3203
3204 for (i = 0; i < num_migrate; i++) {
80c9abaa
SS		 3205 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3206 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3207 return -EINVAL;
3208
3209 /* check if there is any duplicated entry */
3210 for (j = i + 1; j < num_migrate; j++) {
3211 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3212 sizeof(m[i].old_daddr)) &&
3213 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3214 sizeof(m[i].old_saddr)) &&
3215 m[i].proto == m[j].proto &&
3216 m[i].mode == m[j].mode &&
3217 m[i].reqid == m[j].reqid &&
3218 m[i].old_family == m[j].old_family)
3219 return -EINVAL;
3220 }
3221 }
3222
3223 return 0;
3224}
3225
b4b7c0b3 3226int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3227 struct xfrm_migrate *m, int num_migrate,
4ab47d47
AA		 3228 struct xfrm_kmaddress *k, struct net *net,
3229 struct xfrm_encap_tmpl *encap)
80c9abaa
SS		 3230{
3231 int i, err, nx_cur = 0, nx_new = 0;
3232 struct xfrm_policy *pol = NULL;
3233 struct xfrm_state *x, *xc;
3234 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3235 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3236 struct xfrm_migrate *mp;
3237
7bab0963 3238 /* Stage 0 - sanity checks */
80c9abaa
SS		 3239 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3240 goto out;
3241
7bab0963
VD		 3242 if (dir >= XFRM_POLICY_MAX) {
3243 err = -EINVAL;
3244 goto out;
3245 }
3246
80c9abaa 3247 /* Stage 1 - find policy */
8d549c4f 3248 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3249 err = -ENOENT;
3250 goto out;
3251 }
3252
3253 /* Stage 2 - find and update state(s) */
3254 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3255 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3256 x_cur[nx_cur] = x;
3257 nx_cur++;
4ab47d47
AA		 3258 xc = xfrm_state_migrate(x, mp, encap);
3259 if (xc) {
80c9abaa
SS		 3260 x_new[nx_new] = xc;
3261 nx_new++;
3262 } else {
3263 err = -ENODATA;
3264 goto restore_state;
3265 }
3266 }
3267 }
3268
3269 /* Stage 3 - update policy */
3270 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3271 goto restore_state;
3272
3273 /* Stage 4 - delete old state(s) */
3274 if (nx_cur) {
3275 xfrm_states_put(x_cur, nx_cur);
3276 xfrm_states_delete(x_cur, nx_cur);
3277 }
3278
3279 /* Stage 5 - announce */
8bafd730 3280 km_migrate(sel, dir, type, m, num_migrate, k, encap);
80c9abaa
SS		 3281
3282 xfrm_pol_put(pol);
3283
3284 return 0;
3285out:
3286 return err;
3287
3288restore_state:
3289 if (pol)
3290 xfrm_pol_put(pol);
3291 if (nx_cur)
3292 xfrm_states_put(x_cur, nx_cur);
3293 if (nx_new)
3294 xfrm_states_delete(x_new, nx_new);
3295
3296 return err;
3297}
e610e679 3298EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3299#endif
Empty description
This page took 1.839313 seconds and 4 git commands to generate.