projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
xfrm: add rcu grace period in xfrm_policy_destroy()
[deliverable/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 <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
68277acc 27#include <linux/audit.h>
25ee3286 28#include <net/dst.h>
6ce74ec7 29#include <net/flow.h>
1da177e4
LT		 30#include <net/xfrm.h>
31#include <net/ip.h>
558f82ef
MN		 32#ifdef CONFIG_XFRM_STATISTICS
33#include <net/snmp.h>
34#endif
1da177e4 35
44e36b42
DM		 36#include "xfrm_hash.h"
37
a0073fe1
SK		 38#define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
39#define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
40#define XFRM_MAX_QUEUE_LEN 100
41
b8c203b2
SK		 42struct xfrm_flo {
43 struct dst_entry *dst_orig;
44 u8 flags;
45};
46
418a99ac
PJ		 47static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
48static struct xfrm_policy_afinfo __rcu *xfrm_policy_afinfo[NPROTO]
49 __read_mostly;
1da177e4 50
e18b890b 51static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 52
25ee3286 53static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 54static int stale_bundle(struct dst_entry *dst);
12fdb4d3 55static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 56static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 57
12bfa8bd 58static void __xfrm_policy_link(struct xfrm_policy *pol, int dir);
29fa0b30
WY		 59static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
60 int dir);
61
bc9b35ad 62static inline bool
200ce96e 63__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 64{
7e1dc7b6
DM		 65 const struct flowi4 *fl4 = &fl->u.ip4;
66
26bff940
AD		 67 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
68 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 69 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
70 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
71 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
72 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 73}
74
bc9b35ad 75static inline bool
200ce96e 76__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 77{
7e1dc7b6
DM		 78 const struct flowi6 *fl6 = &fl->u.ip6;
79
80 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
81 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
82 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
83 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
84 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
85 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 86}
87
bc9b35ad
DM		 88bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
89 unsigned short family)
77681021
AM		 90{
91 switch (family) {
92 case AF_INET:
93 return __xfrm4_selector_match(sel, fl);
94 case AF_INET6:
95 return __xfrm6_selector_match(sel, fl);
96 }
bc9b35ad 97 return false;
77681021
AM		 98}
99
ef8531b6
ED		 100static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
101{
102 struct xfrm_policy_afinfo *afinfo;
103
104 if (unlikely(family >= NPROTO))
105 return NULL;
106 rcu_read_lock();
107 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
108 if (unlikely(!afinfo))
109 rcu_read_unlock();
110 return afinfo;
111}
112
113static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
114{
115 rcu_read_unlock();
116}
117
42a7b32b
DA		 118static inline struct dst_entry *__xfrm_dst_lookup(struct net *net,
119 int tos, int oif,
6418c4e0
DM		 120 const xfrm_address_t *saddr,
121 const xfrm_address_t *daddr,
9bb182a7
YH		 122 int family)
123{
124 struct xfrm_policy_afinfo *afinfo;
125 struct dst_entry *dst;
126
127 afinfo = xfrm_policy_get_afinfo(family);
128 if (unlikely(afinfo == NULL))
129 return ERR_PTR(-EAFNOSUPPORT);
130
42a7b32b 131 dst = afinfo->dst_lookup(net, tos, oif, saddr, daddr);
9bb182a7
YH		 132
133 xfrm_policy_put_afinfo(afinfo);
134
135 return dst;
136}
137
42a7b32b
DA		 138static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
139 int tos, int oif,
9bb182a7
YH		 140 xfrm_address_t *prev_saddr,
141 xfrm_address_t *prev_daddr,
25ee3286 142 int family)
1da177e4 143{
c5b3cf46 144 struct net *net = xs_net(x);
66cdb3ca
HX		 145 xfrm_address_t *saddr = &x->props.saddr;
146 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 147 struct dst_entry *dst;
148
9bb182a7 149 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 150 saddr = x->coaddr;
9bb182a7
YH		 151 daddr = prev_daddr;
152 }
153 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
154 saddr = prev_saddr;
66cdb3ca 155 daddr = x->coaddr;
9bb182a7 156 }
1da177e4 157
42a7b32b 158 dst = __xfrm_dst_lookup(net, tos, oif, saddr, daddr, family);
9bb182a7
YH		 159
160 if (!IS_ERR(dst)) {
161 if (prev_saddr != saddr)
162 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
163 if (prev_daddr != daddr)
164 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
165 }
1da177e4 166
66cdb3ca 167 return dst;
1da177e4 168}
1da177e4 169
1da177e4
LT		 170static inline unsigned long make_jiffies(long secs)
171{
172 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
173 return MAX_SCHEDULE_TIMEOUT-1;
174 else
a716c119 175 return secs*HZ;
1da177e4
LT		 176}
177
178static void xfrm_policy_timer(unsigned long data)
179{
3e94c2dc 180 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 181 unsigned long now = get_seconds();
1da177e4
LT		 182 long next = LONG_MAX;
183 int warn = 0;
184 int dir;
185
186 read_lock(&xp->lock);
187
ea2dea9d 188 if (unlikely(xp->walk.dead))
1da177e4
LT		 189 goto out;
190
77d8d7a6 191 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 192
193 if (xp->lft.hard_add_expires_seconds) {
194 long tmo = xp->lft.hard_add_expires_seconds +
195 xp->curlft.add_time - now;
196 if (tmo <= 0)
197 goto expired;
198 if (tmo < next)
199 next = tmo;
200 }
201 if (xp->lft.hard_use_expires_seconds) {
202 long tmo = xp->lft.hard_use_expires_seconds +
203 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
204 if (tmo <= 0)
205 goto expired;
206 if (tmo < next)
207 next = tmo;
208 }
209 if (xp->lft.soft_add_expires_seconds) {
210 long tmo = xp->lft.soft_add_expires_seconds +
211 xp->curlft.add_time - now;
212 if (tmo <= 0) {
213 warn = 1;
214 tmo = XFRM_KM_TIMEOUT;
215 }
216 if (tmo < next)
217 next = tmo;
218 }
219 if (xp->lft.soft_use_expires_seconds) {
220 long tmo = xp->lft.soft_use_expires_seconds +
221 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
222 if (tmo <= 0) {
223 warn = 1;
224 tmo = XFRM_KM_TIMEOUT;
225 }
226 if (tmo < next)
227 next = tmo;
228 }
229
230 if (warn)
6c5c8ca7 231 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 232 if (next != LONG_MAX &&
233 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
234 xfrm_pol_hold(xp);
235
236out:
237 read_unlock(&xp->lock);
238 xfrm_pol_put(xp);
239 return;
240
241expired:
242 read_unlock(&xp->lock);
4666faab 243 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 244 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 245 xfrm_pol_put(xp);
246}
247
fe1a5f03
TT		 248static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
249{
250 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
251
252 if (unlikely(pol->walk.dead))
253 flo = NULL;
254 else
255 xfrm_pol_hold(pol);
256
257 return flo;
258}
259
260static int xfrm_policy_flo_check(struct flow_cache_object *flo)
261{
262 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
263
264 return !pol->walk.dead;
265}
266
267static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
268{
269 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
270}
271
272static const struct flow_cache_ops xfrm_policy_fc_ops = {
273 .get = xfrm_policy_flo_get,
274 .check = xfrm_policy_flo_check,
275 .delete = xfrm_policy_flo_delete,
276};
1da177e4
LT		 277
278/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
279 * SPD calls.
280 */
281
0331b1f3 282struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 283{
284 struct xfrm_policy *policy;
285
0da974f4 286 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 287
288 if (policy) {
0331b1f3 289 write_pnet(&policy->xp_net, net);
12a169e7 290 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 291 INIT_HLIST_NODE(&policy->bydst);
292 INIT_HLIST_NODE(&policy->byidx);
1da177e4 293 rwlock_init(&policy->lock);
2518c7c2 294 atomic_set(&policy->refcnt, 1);
a0073fe1 295 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 296 setup_timer(&policy->timer, xfrm_policy_timer,
297 (unsigned long)policy);
a0073fe1
SK		 298 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
299 (unsigned long)policy);
fe1a5f03 300 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 301 }
302 return policy;
303}
304EXPORT_SYMBOL(xfrm_policy_alloc);
305
56f04730
ED		 306static void xfrm_policy_destroy_rcu(struct rcu_head *head)
307{
308 struct xfrm_policy *policy = container_of(head, struct xfrm_policy, rcu);
309
310 security_xfrm_policy_free(policy->security);
311 kfree(policy);
312}
313
1da177e4
LT		 314/* Destroy xfrm_policy: descendant resources must be released to this moment. */
315
64c31b3f 316void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 317{
12a169e7 318 BUG_ON(!policy->walk.dead);
1da177e4 319
0659eea9 320 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 321 BUG();
322
56f04730 323 call_rcu(&policy->rcu, xfrm_policy_destroy_rcu);
1da177e4 324}
64c31b3f 325EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 326
1da177e4
LT		 327/* Rule must be locked. Release descentant resources, announce
328 * entry dead. The rule must be unlinked from lists to the moment.
329 */
330
331static void xfrm_policy_kill(struct xfrm_policy *policy)
332{
12a169e7 333 policy->walk.dead = 1;
1da177e4 334
285ead17 335 atomic_inc(&policy->genid);
1da177e4 336
e7d8f6cb
SK		 337 if (del_timer(&policy->polq.hold_timer))
338 xfrm_pol_put(policy);
1ee5e667 339 skb_queue_purge(&policy->polq.hold_queue);
a0073fe1 340
285ead17
TT		 341 if (del_timer(&policy->timer))
342 xfrm_pol_put(policy);
343
344 xfrm_pol_put(policy);
1da177e4
LT		 345}
346
2518c7c2
DM		 347static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
348
e92303f8 349static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 350{
e92303f8 351 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 352}
353
b58555f1
CG		 354/* calculate policy hash thresholds */
355static void __get_hash_thresh(struct net *net,
356 unsigned short family, int dir,
357 u8 *dbits, u8 *sbits)
358{
359 switch (family) {
360 case AF_INET:
361 *dbits = net->xfrm.policy_bydst[dir].dbits4;
362 *sbits = net->xfrm.policy_bydst[dir].sbits4;
363 break;
364
365 case AF_INET6:
366 *dbits = net->xfrm.policy_bydst[dir].dbits6;
367 *sbits = net->xfrm.policy_bydst[dir].sbits6;
368 break;
369
370 default:
371 *dbits = 0;
372 *sbits = 0;
373 }
374}
375
5f803b58
DM		 376static struct hlist_head *policy_hash_bysel(struct net *net,
377 const struct xfrm_selector *sel,
378 unsigned short family, int dir)
2518c7c2 379{
1121994c 380 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 381 unsigned int hash;
382 u8 dbits;
383 u8 sbits;
384
385 __get_hash_thresh(net, family, dir, &dbits, &sbits);
386 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2
DM		 387
388 return (hash == hmask + 1 ?
1121994c
AD		 389 &net->xfrm.policy_inexact[dir] :
390 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 391}
392
5f803b58
DM		 393static struct hlist_head *policy_hash_direct(struct net *net,
394 const xfrm_address_t *daddr,
395 const xfrm_address_t *saddr,
396 unsigned short family, int dir)
2518c7c2 397{
1121994c 398 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 399 unsigned int hash;
400 u8 dbits;
401 u8 sbits;
402
403 __get_hash_thresh(net, family, dir, &dbits, &sbits);
404 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 405
1121994c 406 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 407}
408
b58555f1
CG		 409static void xfrm_dst_hash_transfer(struct net *net,
410 struct hlist_head *list,
2518c7c2 411 struct hlist_head *ndsttable,
b58555f1
CG		 412 unsigned int nhashmask,
413 int dir)
2518c7c2 414{
b67bfe0d 415 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 416 struct xfrm_policy *pol;
b791160b 417 unsigned int h0 = 0;
b58555f1
CG		 418 u8 dbits;
419 u8 sbits;
2518c7c2 420
b791160b 421redo:
b67bfe0d 422 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 423 unsigned int h;
424
b58555f1 425 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 426 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 427 pol->family, nhashmask, dbits, sbits);
b791160b 428 if (!entry0) {
b67bfe0d 429 hlist_del(&pol->bydst);
b791160b
YH		 430 hlist_add_head(&pol->bydst, ndsttable+h);
431 h0 = h;
432 } else {
433 if (h != h0)
434 continue;
b67bfe0d 435 hlist_del(&pol->bydst);
1d023284 436 hlist_add_behind(&pol->bydst, entry0);
b791160b 437 }
b67bfe0d 438 entry0 = &pol->bydst;
b791160b
YH		 439 }
440 if (!hlist_empty(list)) {
441 entry0 = NULL;
442 goto redo;
2518c7c2
DM		 443 }
444}
445
446static void xfrm_idx_hash_transfer(struct hlist_head *list,
447 struct hlist_head *nidxtable,
448 unsigned int nhashmask)
449{
b67bfe0d 450 struct hlist_node *tmp;
2518c7c2
DM		 451 struct xfrm_policy *pol;
452
b67bfe0d 453 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 454 unsigned int h;
455
456 h = __idx_hash(pol->index, nhashmask);
457 hlist_add_head(&pol->byidx, nidxtable+h);
458 }
459}
460
461static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
462{
463 return ((old_hmask + 1) << 1) - 1;
464}
465
66caf628 466static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 467{
66caf628 468 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 469 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
470 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 471 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 472 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 473 int i;
474
475 if (!ndst)
476 return;
477
283bc9f3 478 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 479
480 for (i = hmask; i >= 0; i--)
b58555f1 481 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 482
66caf628
AD		 483 net->xfrm.policy_bydst[dir].table = ndst;
484 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 485
283bc9f3 486 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 487
44e36b42 488 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 489}
490
66caf628 491static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 492{
66caf628 493 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 494 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
495 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 496 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 497 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 498 int i;
499
500 if (!nidx)
501 return;
502
283bc9f3 503 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 504
505 for (i = hmask; i >= 0; i--)
506 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
507
66caf628
AD		 508 net->xfrm.policy_byidx = nidx;
509 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 510
283bc9f3 511 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 512
44e36b42 513 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 514}
515
66caf628 516static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 517{
66caf628
AD		 518 unsigned int cnt = net->xfrm.policy_count[dir];
519 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 520
521 if (total)
522 *total += cnt;
523
524 if ((hmask + 1) < xfrm_policy_hashmax &&
525 cnt > hmask)
526 return 1;
527
528 return 0;
529}
530
66caf628 531static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 532{
66caf628 533 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 534
535 if ((hmask + 1) < xfrm_policy_hashmax &&
536 total > hmask)
537 return 1;
538
539 return 0;
540}
541
e071041b 542void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 543{
283bc9f3 544 read_lock_bh(&net->xfrm.xfrm_policy_lock);
e071041b
AD		 545 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
546 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
547 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
548 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
549 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
550 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
551 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 552 si->spdhmcnt = xfrm_policy_hashmax;
283bc9f3 553 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
ecfd6b18
JHS		 554}
555EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 556
ecfd6b18 557static DEFINE_MUTEX(hash_resize_mutex);
66caf628 558static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 559{
66caf628 560 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 561 int dir, total;
562
563 mutex_lock(&hash_resize_mutex);
564
565 total = 0;
53c2e285 566 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD		 567 if (xfrm_bydst_should_resize(net, dir, &total))
568 xfrm_bydst_resize(net, dir);
2518c7c2 569 }
66caf628
AD		 570 if (xfrm_byidx_should_resize(net, total))
571 xfrm_byidx_resize(net, total);
2518c7c2
DM		 572
573 mutex_unlock(&hash_resize_mutex);
574}
575
880a6fab
CG		 576static void xfrm_hash_rebuild(struct work_struct *work)
577{
578 struct net *net = container_of(work, struct net,
579 xfrm.policy_hthresh.work);
580 unsigned int hmask;
581 struct xfrm_policy *pol;
582 struct xfrm_policy *policy;
583 struct hlist_head *chain;
584 struct hlist_head *odst;
585 struct hlist_node *newpos;
586 int i;
587 int dir;
588 unsigned seq;
589 u8 lbits4, rbits4, lbits6, rbits6;
590
591 mutex_lock(&hash_resize_mutex);
592
593 /* read selector prefixlen thresholds */
594 do {
595 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
596
597 lbits4 = net->xfrm.policy_hthresh.lbits4;
598 rbits4 = net->xfrm.policy_hthresh.rbits4;
599 lbits6 = net->xfrm.policy_hthresh.lbits6;
600 rbits6 = net->xfrm.policy_hthresh.rbits6;
601 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
602
603 write_lock_bh(&net->xfrm.xfrm_policy_lock);
604
605 /* reset the bydst and inexact table in all directions */
53c2e285 606 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG		 607 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
608 hmask = net->xfrm.policy_bydst[dir].hmask;
609 odst = net->xfrm.policy_bydst[dir].table;
610 for (i = hmask; i >= 0; i--)
611 INIT_HLIST_HEAD(odst + i);
612 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
613 /* dir out => dst = remote, src = local */
614 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
615 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
616 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
617 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
618 } else {
619 /* dir in/fwd => dst = local, src = remote */
620 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
621 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
622 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
623 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
624 }
625 }
626
627 /* re-insert all policies by order of creation */
628 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
629 newpos = NULL;
630 chain = policy_hash_bysel(net, &policy->selector,
631 policy->family,
632 xfrm_policy_id2dir(policy->index));
633 hlist_for_each_entry(pol, chain, bydst) {
634 if (policy->priority >= pol->priority)
635 newpos = &pol->bydst;
636 else
637 break;
638 }
639 if (newpos)
640 hlist_add_behind(&policy->bydst, newpos);
641 else
642 hlist_add_head(&policy->bydst, chain);
643 }
644
645 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
646
647 mutex_unlock(&hash_resize_mutex);
648}
649
650void xfrm_policy_hash_rebuild(struct net *net)
651{
652 schedule_work(&net->xfrm.policy_hthresh.work);
653}
654EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
655
1da177e4
LT		 656/* Generate new index... KAME seems to generate them ordered by cost
657 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 658static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 659{
1da177e4
LT		 660 static u32 idx_generator;
661
662 for (;;) {
2518c7c2
DM		 663 struct hlist_head *list;
664 struct xfrm_policy *p;
665 u32 idx;
666 int found;
667
e682adf0
FD		 668 if (!index) {
669 idx = (idx_generator | dir);
670 idx_generator += 8;
671 } else {
672 idx = index;
673 index = 0;
674 }
675
1da177e4
LT		 676 if (idx == 0)
677 idx = 8;
1121994c 678 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 679 found = 0;
b67bfe0d 680 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 681 if (p->index == idx) {
682 found = 1;
1da177e4 683 break;
2518c7c2 684 }
1da177e4 685 }
2518c7c2 686 if (!found)
1da177e4
LT		 687 return idx;
688 }
689}
690
2518c7c2
DM		 691static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
692{
693 u32 *p1 = (u32 *) s1;
694 u32 *p2 = (u32 *) s2;
695 int len = sizeof(struct xfrm_selector) / sizeof(u32);
696 int i;
697
698 for (i = 0; i < len; i++) {
699 if (p1[i] != p2[i])
700 return 1;
701 }
702
703 return 0;
704}
705
a0073fe1
SK		 706static void xfrm_policy_requeue(struct xfrm_policy *old,
707 struct xfrm_policy *new)
708{
709 struct xfrm_policy_queue *pq = &old->polq;
710 struct sk_buff_head list;
711
de2ad486
LR		 712 if (skb_queue_empty(&pq->hold_queue))
713 return;
714
a0073fe1
SK		 715 __skb_queue_head_init(&list);
716
717 spin_lock_bh(&pq->hold_queue.lock);
718 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 719 if (del_timer(&pq->hold_timer))
720 xfrm_pol_put(old);
a0073fe1
SK		 721 spin_unlock_bh(&pq->hold_queue.lock);
722
a0073fe1
SK		 723 pq = &new->polq;
724
725 spin_lock_bh(&pq->hold_queue.lock);
726 skb_queue_splice(&list, &pq->hold_queue);
727 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 728 if (!mod_timer(&pq->hold_timer, jiffies))
729 xfrm_pol_hold(new);
a0073fe1
SK		 730 spin_unlock_bh(&pq->hold_queue.lock);
731}
732
7cb8a939
SK		 733static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
734 struct xfrm_policy *pol)
735{
736 u32 mark = policy->mark.v & policy->mark.m;
737
738 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
739 return true;
740
741 if ((mark & pol->mark.m) == pol->mark.v &&
742 policy->priority == pol->priority)
743 return true;
744
745 return false;
746}
747
1da177e4
LT		 748int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
749{
1121994c 750 struct net *net = xp_net(policy);
2518c7c2
DM		 751 struct xfrm_policy *pol;
752 struct xfrm_policy *delpol;
753 struct hlist_head *chain;
b67bfe0d 754 struct hlist_node *newpos;
1da177e4 755
283bc9f3 756 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 757 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 758 delpol = NULL;
759 newpos = NULL;
b67bfe0d 760 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 761 if (pol->type == policy->type &&
2518c7c2 762 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 763 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 764 xfrm_sec_ctx_match(pol->security, policy->security) &&
765 !WARN_ON(delpol)) {
1da177e4 766 if (excl) {
283bc9f3 767 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 768 return -EEXIST;
769 }
1da177e4
LT		 770 delpol = pol;
771 if (policy->priority > pol->priority)
772 continue;
773 } else if (policy->priority >= pol->priority) {
a6c7ab55 774 newpos = &pol->bydst;
1da177e4
LT		 775 continue;
776 }
1da177e4
LT		 777 if (delpol)
778 break;
1da177e4
LT		 779 }
780 if (newpos)
1d023284 781 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 782 else
783 hlist_add_head(&policy->bydst, chain);
12bfa8bd 784 __xfrm_policy_link(policy, dir);
ca925cf1 785 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 786
787 /* After previous checking, family can either be AF_INET or AF_INET6 */
788 if (policy->family == AF_INET)
789 rt_genid_bump_ipv4(net);
790 else
791 rt_genid_bump_ipv6(net);
792
a0073fe1
SK		 793 if (delpol) {
794 xfrm_policy_requeue(delpol, policy);
29fa0b30 795 __xfrm_policy_unlink(delpol, dir);
a0073fe1 796 }
e682adf0 797 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 798 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 799 policy->curlft.add_time = get_seconds();
1da177e4
LT		 800 policy->curlft.use_time = 0;
801 if (!mod_timer(&policy->timer, jiffies + HZ))
802 xfrm_pol_hold(policy);
283bc9f3 803 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 804
9b78a82c 805 if (delpol)
1da177e4 806 xfrm_policy_kill(delpol);
1121994c
AD		 807 else if (xfrm_bydst_should_resize(net, dir, NULL))
808 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 809
1da177e4
LT		 810 return 0;
811}
812EXPORT_SYMBOL(xfrm_policy_insert);
813
8ca2e93b
JHS		 814struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
815 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 816 struct xfrm_sec_ctx *ctx, int delete,
817 int *err)
1da177e4 818{
2518c7c2
DM		 819 struct xfrm_policy *pol, *ret;
820 struct hlist_head *chain;
1da177e4 821
ef41aaa0 822 *err = 0;
283bc9f3 823 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 824 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 825 ret = NULL;
b67bfe0d 826 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 827 if (pol->type == type &&
34f8d884 828 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 829 !selector_cmp(sel, &pol->selector) &&
830 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 831 xfrm_pol_hold(pol);
2518c7c2 832 if (delete) {
03e1ad7b
PM		 833 *err = security_xfrm_policy_delete(
834 pol->security);
ef41aaa0 835 if (*err) {
283bc9f3 836 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 837 return pol;
838 }
29fa0b30 839 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 840 }
841 ret = pol;
1da177e4
LT		 842 break;
843 }
844 }
283bc9f3 845 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 846
fe1a5f03 847 if (ret && delete)
2518c7c2 848 xfrm_policy_kill(ret);
2518c7c2 849 return ret;
1da177e4 850}
df71837d 851EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 852
8ca2e93b
JHS		 853struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
854 int dir, u32 id, int delete, int *err)
1da177e4 855{
2518c7c2
DM		 856 struct xfrm_policy *pol, *ret;
857 struct hlist_head *chain;
1da177e4 858
b5505c6e
HX		 859 *err = -ENOENT;
860 if (xfrm_policy_id2dir(id) != dir)
861 return NULL;
862
ef41aaa0 863 *err = 0;
283bc9f3 864 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 865 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 866 ret = NULL;
b67bfe0d 867 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 868 if (pol->type == type && pol->index == id &&
869 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 870 xfrm_pol_hold(pol);
2518c7c2 871 if (delete) {
03e1ad7b
PM		 872 *err = security_xfrm_policy_delete(
873 pol->security);
ef41aaa0 874 if (*err) {
283bc9f3 875 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 876 return pol;
877 }
29fa0b30 878 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 879 }
880 ret = pol;
1da177e4
LT		 881 break;
882 }
883 }
283bc9f3 884 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 885
fe1a5f03 886 if (ret && delete)
2518c7c2 887 xfrm_policy_kill(ret);
2518c7c2 888 return ret;
1da177e4
LT		 889}
890EXPORT_SYMBOL(xfrm_policy_byid);
891
4aa2e62c
JL		 892#ifdef CONFIG_SECURITY_NETWORK_XFRM
893static inline int
2e71029e 894xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 895{
4aa2e62c
JL		 896 int dir, err = 0;
897
898 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
899 struct xfrm_policy *pol;
4aa2e62c
JL		 900 int i;
901
b67bfe0d 902 hlist_for_each_entry(pol,
33ffbbd5 903 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 904 if (pol->type != type)
905 continue;
03e1ad7b 906 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 907 if (err) {
2e71029e 908 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 909 return err;
910 }
7dc12d6d 911 }
33ffbbd5 912 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 913 hlist_for_each_entry(pol,
33ffbbd5 914 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 915 bydst) {
916 if (pol->type != type)
917 continue;
03e1ad7b
PM		 918 err = security_xfrm_policy_delete(
919 pol->security);
4aa2e62c 920 if (err) {
ab5f5e8b 921 xfrm_audit_policy_delete(pol, 0,
2e71029e 922 task_valid);
4aa2e62c
JL		 923 return err;
924 }
925 }
926 }
927 }
928 return err;
929}
930#else
931static inline int
2e71029e 932xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 933{
934 return 0;
935}
936#endif
937
2e71029e 938int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 939{
2f1eb65f 940 int dir, err = 0, cnt = 0;
1da177e4 941
283bc9f3 942 write_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 943
2e71029e 944 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 945 if (err)
946 goto out;
947
1da177e4 948 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 949 struct xfrm_policy *pol;
29fa0b30 950 int i;
2518c7c2
DM		 951
952 again1:
b67bfe0d 953 hlist_for_each_entry(pol,
33ffbbd5 954 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 955 if (pol->type != type)
956 continue;
ea2dea9d 957 __xfrm_policy_unlink(pol, dir);
283bc9f3 958 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 959 cnt++;
1da177e4 960
2e71029e 961 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 962
2518c7c2 963 xfrm_policy_kill(pol);
1da177e4 964
283bc9f3 965 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 966 goto again1;
967 }
968
33ffbbd5 969 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 970 again2:
b67bfe0d 971 hlist_for_each_entry(pol,
33ffbbd5 972 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 973 bydst) {
974 if (pol->type != type)
975 continue;
ea2dea9d 976 __xfrm_policy_unlink(pol, dir);
283bc9f3 977 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 978 cnt++;
2518c7c2 979
2e71029e 980 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 981 xfrm_policy_kill(pol);
982
283bc9f3 983 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 984 goto again2;
985 }
1da177e4 986 }
2518c7c2 987
1da177e4 988 }
2f1eb65f
JHS		 989 if (!cnt)
990 err = -ESRCH;
4aa2e62c 991out:
283bc9f3 992 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 993 return err;
1da177e4
LT		 994}
995EXPORT_SYMBOL(xfrm_policy_flush);
996
cdcbca7c 997int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 998 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 999 void *data)
1000{
12a169e7
HX		 1001 struct xfrm_policy *pol;
1002 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 1003 int error = 0;
1004
1005 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
1006 walk->type != XFRM_POLICY_TYPE_ANY)
1007 return -EINVAL;
1da177e4 1008
12a169e7 1009 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 1010 return 0;
1011
283bc9f3 1012 write_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1013 if (list_empty(&walk->walk.all))
cdcbca7c 1014 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1015 else
80077702
LR		 1016 x = list_first_entry(&walk->walk.all,
1017 struct xfrm_policy_walk_entry, all);
1018
cdcbca7c 1019 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1020 if (x->dead)
4c563f76 1021 continue;
12a169e7
HX		 1022 pol = container_of(x, struct xfrm_policy, walk);
1023 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1024 walk->type != pol->type)
1025 continue;
1026 error = func(pol, xfrm_policy_id2dir(pol->index),
1027 walk->seq, data);
1028 if (error) {
1029 list_move_tail(&walk->walk.all, &x->all);
1030 goto out;
2518c7c2 1031 }
12a169e7 1032 walk->seq++;
1da177e4 1033 }
12a169e7 1034 if (walk->seq == 0) {
baf5d743
JHS		 1035 error = -ENOENT;
1036 goto out;
1037 }
12a169e7 1038 list_del_init(&walk->walk.all);
1da177e4 1039out:
283bc9f3 1040 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1041 return error;
1042}
1043EXPORT_SYMBOL(xfrm_policy_walk);
1044
12a169e7
HX		 1045void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1046{
1047 INIT_LIST_HEAD(&walk->walk.all);
1048 walk->walk.dead = 1;
1049 walk->type = type;
1050 walk->seq = 0;
1051}
1052EXPORT_SYMBOL(xfrm_policy_walk_init);
1053
283bc9f3 1054void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1055{
1056 if (list_empty(&walk->walk.all))
1057 return;
1058
283bc9f3 1059 write_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1060 list_del(&walk->walk.all);
283bc9f3 1061 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1062}
1063EXPORT_SYMBOL(xfrm_policy_walk_done);
1064
134b0fc5
JM		 1065/*
1066 * Find policy to apply to this flow.
1067 *
1068 * Returns 0 if policy found, else an -errno.
1069 */
f299d557
DM		 1070static int xfrm_policy_match(const struct xfrm_policy *pol,
1071 const struct flowi *fl,
2518c7c2 1072 u8 type, u16 family, int dir)
1da177e4 1073{
f299d557 1074 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1075 int ret = -ESRCH;
1076 bool match;
1da177e4 1077
2518c7c2 1078 if (pol->family != family ||
1d28f42c 1079 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1080 pol->type != type)
134b0fc5 1081 return ret;
1da177e4 1082
2518c7c2 1083 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1084 if (match)
1d28f42c 1085 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1086 dir);
2518c7c2 1087
134b0fc5 1088 return ret;
2518c7c2 1089}
1da177e4 1090
52479b62 1091static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1092 const struct flowi *fl,
2518c7c2
DM		 1093 u16 family, u8 dir)
1094{
134b0fc5 1095 int err;
2518c7c2 1096 struct xfrm_policy *pol, *ret;
0b597e7e 1097 const xfrm_address_t *daddr, *saddr;
2518c7c2 1098 struct hlist_head *chain;
acba48e1 1099 u32 priority = ~0U;
df71837d 1100
2518c7c2
DM		 1101 daddr = xfrm_flowi_daddr(fl, family);
1102 saddr = xfrm_flowi_saddr(fl, family);
1103 if (unlikely(!daddr || !saddr))
1104 return NULL;
1105
283bc9f3 1106 read_lock_bh(&net->xfrm.xfrm_policy_lock);
52479b62 1107 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 1108 ret = NULL;
b67bfe0d 1109 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1110 err = xfrm_policy_match(pol, fl, type, family, dir);
1111 if (err) {
1112 if (err == -ESRCH)
1113 continue;
1114 else {
1115 ret = ERR_PTR(err);
1116 goto fail;
1117 }
1118 } else {
2518c7c2 1119 ret = pol;
acba48e1 1120 priority = ret->priority;
2518c7c2
DM		 1121 break;
1122 }
1123 }
52479b62 1124 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1125 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR		 1126 if ((pol->priority >= priority) && ret)
1127 break;
1128
134b0fc5
JM		 1129 err = xfrm_policy_match(pol, fl, type, family, dir);
1130 if (err) {
1131 if (err == -ESRCH)
1132 continue;
1133 else {
1134 ret = ERR_PTR(err);
1135 goto fail;
1136 }
8faf491e 1137 } else {
acba48e1
DM		 1138 ret = pol;
1139 break;
1da177e4
LT		 1140 }
1141 }
586f2eb4
LR		 1142
1143 xfrm_pol_hold(ret);
134b0fc5 1144fail:
283bc9f3 1145 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
4e81bb83 1146
2518c7c2 1147 return ret;
4e81bb83
MN		 1148}
1149
80c802f3 1150static struct xfrm_policy *
73ff93cd 1151__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1152{
1153#ifdef CONFIG_XFRM_SUB_POLICY
1154 struct xfrm_policy *pol;
1155
1156 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1157 if (pol != NULL)
1158 return pol;
1159#endif
1160 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1161}
1162
b5fb82c4
BZ		 1163static int flow_to_policy_dir(int dir)
1164{
1165 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1166 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1167 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1168 return dir;
1169
1170 switch (dir) {
1171 default:
1172 case FLOW_DIR_IN:
1173 return XFRM_POLICY_IN;
1174 case FLOW_DIR_OUT:
1175 return XFRM_POLICY_OUT;
1176 case FLOW_DIR_FWD:
1177 return XFRM_POLICY_FWD;
1178 }
1179}
1180
fe1a5f03 1181static struct flow_cache_object *
dee9f4bc 1182xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1183 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1184{
1185 struct xfrm_policy *pol;
fe1a5f03
TT		 1186
1187 if (old_obj)
1188 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1189
b5fb82c4 1190 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1191 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1192 return ERR_CAST(pol);
fe1a5f03 1193
fe1a5f03
TT		 1194 /* Resolver returns two references:
1195 * one for cache and one for caller of flow_cache_lookup() */
1196 xfrm_pol_hold(pol);
1197
1198 return &pol->flo;
1da177e4
LT		 1199}
1200
df71837d
TJ		 1201static inline int policy_to_flow_dir(int dir)
1202{
1203 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1204 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1205 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1206 return dir;
1207 switch (dir) {
1208 default:
1209 case XFRM_POLICY_IN:
1210 return FLOW_DIR_IN;
1211 case XFRM_POLICY_OUT:
1212 return FLOW_DIR_OUT;
1213 case XFRM_POLICY_FWD:
1214 return FLOW_DIR_FWD;
3ff50b79 1215 }
df71837d
TJ		 1216}
1217
6f9c9615 1218static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir,
dee9f4bc 1219 const struct flowi *fl)
1da177e4
LT		 1220{
1221 struct xfrm_policy *pol;
283bc9f3 1222 struct net *net = sock_net(sk);
1da177e4 1223
283bc9f3 1224 read_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1225 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1226 bool match = xfrm_selector_match(&pol->selector, fl,
1227 sk->sk_family);
a716c119 1228 int err = 0;
df71837d 1229
3bccfbc7 1230 if (match) {
34f8d884
JHS		 1231 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1232 pol = NULL;
1233 goto out;
1234 }
03e1ad7b 1235 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1236 fl->flowi_secid,
03e1ad7b 1237 policy_to_flow_dir(dir));
3bccfbc7
VY		 1238 if (!err)
1239 xfrm_pol_hold(pol);
1240 else if (err == -ESRCH)
1241 pol = NULL;
1242 else
1243 pol = ERR_PTR(err);
1244 } else
1da177e4
LT		 1245 pol = NULL;
1246 }
34f8d884 1247out:
283bc9f3 1248 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1249 return pol;
1250}
1251
1252static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1253{
98806f75 1254 struct net *net = xp_net(pol);
4e81bb83 1255
98806f75 1256 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1257 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1258 xfrm_pol_hold(pol);
1259}
1260
1261static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1262 int dir)
1263{
98806f75
AD		 1264 struct net *net = xp_net(pol);
1265
53c2e285 1266 if (list_empty(&pol->walk.all))
2518c7c2 1267 return NULL;
1da177e4 1268
53c2e285
HX		 1269 /* Socket policies are not hashed. */
1270 if (!hlist_unhashed(&pol->bydst)) {
1271 hlist_del(&pol->bydst);
1272 hlist_del(&pol->byidx);
1273 }
1274
1275 list_del_init(&pol->walk.all);
98806f75 1276 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1277
1278 return pol;
1da177e4
LT		 1279}
1280
53c2e285
HX		 1281static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1282{
1283 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1284}
1285
1286static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1287{
1288 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1289}
1290
4666faab 1291int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1292{
283bc9f3
FD		 1293 struct net *net = xp_net(pol);
1294
1295 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1296 pol = __xfrm_policy_unlink(pol, dir);
283bc9f3 1297 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1298 if (pol) {
1da177e4 1299 xfrm_policy_kill(pol);
4666faab 1300 return 0;
1da177e4 1301 }
4666faab 1302 return -ENOENT;
1da177e4 1303}
a70fcb0b 1304EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1305
1306int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1307{
1121994c 1308 struct net *net = xp_net(pol);
1da177e4
LT		 1309 struct xfrm_policy *old_pol;
1310
4e81bb83
MN		 1311#ifdef CONFIG_XFRM_SUB_POLICY
1312 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1313 return -EINVAL;
1314#endif
1315
283bc9f3 1316 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1317 old_pol = sk->sk_policy[dir];
1318 sk->sk_policy[dir] = pol;
1319 if (pol) {
9d729f72 1320 pol->curlft.add_time = get_seconds();
e682adf0 1321 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1322 xfrm_sk_policy_link(pol, dir);
1da177e4 1323 }
a0073fe1
SK		 1324 if (old_pol) {
1325 if (pol)
1326 xfrm_policy_requeue(old_pol, pol);
1327
ea2dea9d
TT		 1328 /* Unlinking succeeds always. This is the only function
1329 * allowed to delete or replace socket policy.
1330 */
53c2e285 1331 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1332 }
283bc9f3 1333 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1334
1335 if (old_pol) {
1336 xfrm_policy_kill(old_pol);
1337 }
1338 return 0;
1339}
1340
d3e40a9f 1341static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1342{
0331b1f3 1343 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1344 struct net *net = xp_net(old);
1da177e4
LT		 1345
1346 if (newp) {
1347 newp->selector = old->selector;
03e1ad7b
PM		 1348 if (security_xfrm_policy_clone(old->security,
1349 &newp->security)) {
df71837d
TJ		 1350 kfree(newp);
1351 return NULL; /* ENOMEM */
1352 }
1da177e4
LT		 1353 newp->lft = old->lft;
1354 newp->curlft = old->curlft;
fb977e2c 1355 newp->mark = old->mark;
1da177e4
LT		 1356 newp->action = old->action;
1357 newp->flags = old->flags;
1358 newp->xfrm_nr = old->xfrm_nr;
1359 newp->index = old->index;
4e81bb83 1360 newp->type = old->type;
1da177e4
LT		 1361 memcpy(newp->xfrm_vec, old->xfrm_vec,
1362 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
283bc9f3 1363 write_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1364 xfrm_sk_policy_link(newp, dir);
283bc9f3 1365 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1366 xfrm_pol_put(newp);
1367 }
1368 return newp;
1369}
1370
1371int __xfrm_sk_clone_policy(struct sock *sk)
1372{
1373 struct xfrm_policy *p0 = sk->sk_policy[0],
1374 *p1 = sk->sk_policy[1];
1375
1376 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1377 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1378 return -ENOMEM;
1379 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1380 return -ENOMEM;
1381 return 0;
1382}
1383
a1e59abf 1384static int
42a7b32b
DA		 1385xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local,
1386 xfrm_address_t *remote, unsigned short family)
a1e59abf
PM		 1387{
1388 int err;
1389 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1390
1391 if (unlikely(afinfo == NULL))
1392 return -EINVAL;
42a7b32b 1393 err = afinfo->get_saddr(net, oif, local, remote);
a1e59abf
PM		 1394 xfrm_policy_put_afinfo(afinfo);
1395 return err;
1396}
1397
1da177e4
LT		 1398/* Resolve list of templates for the flow, given policy. */
1399
1400static int
a6c2e611
DM		 1401xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1402 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1403{
fbda33b2 1404 struct net *net = xp_net(policy);
1da177e4
LT		 1405 int nx;
1406 int i, error;
1407 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1408 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1409 xfrm_address_t tmp;
1da177e4 1410
9b7a787d 1411 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1412 struct xfrm_state *x;
1413 xfrm_address_t *remote = daddr;
1414 xfrm_address_t *local = saddr;
1415 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1416
48b8d783
JK		 1417 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1418 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1419 remote = &tmpl->id.daddr;
1420 local = &tmpl->saddr;
8444cf71 1421 if (xfrm_addr_any(local, tmpl->encap_family)) {
42a7b32b
DA		 1422 error = xfrm_get_saddr(net, fl->flowi_oif,
1423 &tmp, remote,
1424 tmpl->encap_family);
a1e59abf
PM		 1425 if (error)
1426 goto fail;
1427 local = &tmp;
1428 }
1da177e4
LT		 1429 }
1430
1431 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1432
1433 if (x && x->km.state == XFRM_STATE_VALID) {
1434 xfrm[nx++] = x;
1435 daddr = remote;
1436 saddr = local;
1437 continue;
1438 }
1439 if (x) {
1440 error = (x->km.state == XFRM_STATE_ERROR ?
1441 -EINVAL : -EAGAIN);
1442 xfrm_state_put(x);
42054569 1443 } else if (error == -ESRCH) {
a4322266 1444 error = -EAGAIN;
42054569 1445 }
1da177e4
LT		 1446
1447 if (!tmpl->optional)
1448 goto fail;
1449 }
1450 return nx;
1451
1452fail:
9b7a787d 1453 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1454 xfrm_state_put(xfrm[nx]);
1455 return error;
1456}
1457
4e81bb83 1458static int
a6c2e611
DM		 1459xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1460 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1461{
41a49cc3
MN		 1462 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1463 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1464 int cnx = 0;
1465 int error;
1466 int ret;
1467 int i;
1468
1469 for (i = 0; i < npols; i++) {
1470 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1471 error = -ENOBUFS;
1472 goto fail;
1473 }
41a49cc3
MN		 1474
1475 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1476 if (ret < 0) {
1477 error = ret;
1478 goto fail;
1479 } else
1480 cnx += ret;
1481 }
1482
41a49cc3
MN		 1483 /* found states are sorted for outbound processing */
1484 if (npols > 1)
1485 xfrm_state_sort(xfrm, tpp, cnx, family);
1486
4e81bb83
MN		 1487 return cnx;
1488
1489 fail:
9b7a787d 1490 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1491 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1492 return error;
1493
1494}
1495
1da177e4
LT		 1496/* Check that the bundle accepts the flow and its components are
1497 * still valid.
1498 */
1499
05d84025 1500static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1501{
1502 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1503 int tos;
1da177e4 1504
25ee3286
HX		 1505 if (!afinfo)
1506 return -EINVAL;
1507
1508 tos = afinfo->get_tos(fl);
1509
1510 xfrm_policy_put_afinfo(afinfo);
1511
1512 return tos;
1513}
1514
80c802f3
TT		 1515static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1516{
1517 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1518 struct dst_entry *dst = &xdst->u.dst;
1519
1520 if (xdst->route == NULL) {
1521 /* Dummy bundle - if it has xfrms we were not
1522 * able to build bundle as template resolution failed.
1523 * It means we need to try again resolving. */
1524 if (xdst->num_xfrms > 0)
1525 return NULL;
a0073fe1
SK		 1526 } else if (dst->flags & DST_XFRM_QUEUE) {
1527 return NULL;
80c802f3
TT		 1528 } else {
1529 /* Real bundle */
1530 if (stale_bundle(dst))
1531 return NULL;
1532 }
1533
1534 dst_hold(dst);
1535 return flo;
1536}
1537
1538static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1539{
1540 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1541 struct dst_entry *dst = &xdst->u.dst;
1542
1543 if (!xdst->route)
1544 return 0;
1545 if (stale_bundle(dst))
1546 return 0;
1547
1548 return 1;
1549}
1550
1551static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1552{
1553 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1554 struct dst_entry *dst = &xdst->u.dst;
1555
1556 dst_free(dst);
1557}
1558
1559static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1560 .get = xfrm_bundle_flo_get,
1561 .check = xfrm_bundle_flo_check,
1562 .delete = xfrm_bundle_flo_delete,
1563};
1564
d7c7544c 1565static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1566{
1da177e4 1567 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1568 struct dst_ops *dst_ops;
25ee3286
HX		 1569 struct xfrm_dst *xdst;
1570
1571 if (!afinfo)
1572 return ERR_PTR(-EINVAL);
1573
d7c7544c
AD		 1574 switch (family) {
1575 case AF_INET:
1576 dst_ops = &net->xfrm.xfrm4_dst_ops;
1577 break;
dfd56b8b 1578#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1579 case AF_INET6:
1580 dst_ops = &net->xfrm.xfrm6_dst_ops;
1581 break;
1582#endif
1583 default:
1584 BUG();
1585 }
f5b0a874 1586 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1587
d4cae562 1588 if (likely(xdst)) {
141e369d
SK		 1589 struct dst_entry *dst = &xdst->u.dst;
1590
1591 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1592 xdst->flo.ops = &xfrm_bundle_fc_ops;
d4cae562 1593 } else
0b150932 1594 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1595
d4cae562
MB		 1596 xfrm_policy_put_afinfo(afinfo);
1597
25ee3286
HX		 1598 return xdst;
1599}
1600
a1b05140
MN		 1601static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1602 int nfheader_len)
1603{
1604 struct xfrm_policy_afinfo *afinfo =
1605 xfrm_policy_get_afinfo(dst->ops->family);
1606 int err;
1607
1608 if (!afinfo)
1609 return -EINVAL;
1610
1611 err = afinfo->init_path(path, dst, nfheader_len);
1612
1613 xfrm_policy_put_afinfo(afinfo);
1614
1615 return err;
1616}
1617
87c1e12b 1618static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1619 const struct flowi *fl)
25ee3286
HX		 1620{
1621 struct xfrm_policy_afinfo *afinfo =
1622 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1623 int err;
1624
1625 if (!afinfo)
1da177e4 1626 return -EINVAL;
25ee3286 1627
87c1e12b 1628 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1629
1da177e4 1630 xfrm_policy_put_afinfo(afinfo);
25ee3286 1631
1da177e4
LT		 1632 return err;
1633}
1634
80c802f3 1635
25ee3286
HX		 1636/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1637 * all the metrics... Shortly, bundle a bundle.
1638 */
1639
1640static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1641 struct xfrm_state **xfrm, int nx,
98313ada 1642 const struct flowi *fl,
25ee3286
HX		 1643 struct dst_entry *dst)
1644{
d7c7544c 1645 struct net *net = xp_net(policy);
25ee3286
HX		 1646 unsigned long now = jiffies;
1647 struct net_device *dev;
43a4dea4 1648 struct xfrm_mode *inner_mode;
25ee3286
HX		 1649 struct dst_entry *dst_prev = NULL;
1650 struct dst_entry *dst0 = NULL;
1651 int i = 0;
1652 int err;
1653 int header_len = 0;
a1b05140 1654 int nfheader_len = 0;
25ee3286
HX		 1655 int trailer_len = 0;
1656 int tos;
1657 int family = policy->selector.family;
9bb182a7
YH		 1658 xfrm_address_t saddr, daddr;
1659
1660 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1661
1662 tos = xfrm_get_tos(fl, family);
1663 err = tos;
1664 if (tos < 0)
1665 goto put_states;
1666
1667 dst_hold(dst);
1668
1669 for (; i < nx; i++) {
d7c7544c 1670 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1671 struct dst_entry *dst1 = &xdst->u.dst;
1672
1673 err = PTR_ERR(xdst);
1674 if (IS_ERR(xdst)) {
1675 dst_release(dst);
1676 goto put_states;
1677 }
1678
43a4dea4
SK		 1679 if (xfrm[i]->sel.family == AF_UNSPEC) {
1680 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1681 xfrm_af2proto(family));
1682 if (!inner_mode) {
1683 err = -EAFNOSUPPORT;
1684 dst_release(dst);
1685 goto put_states;
1686 }
1687 } else
1688 inner_mode = xfrm[i]->inner_mode;
1689
25ee3286
HX		 1690 if (!dst_prev)
1691 dst0 = dst1;
1692 else {
1693 dst_prev->child = dst_clone(dst1);
1694 dst1->flags |= DST_NOHASH;
1695 }
1696
1697 xdst->route = dst;
defb3519 1698 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1699
1700 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1701 family = xfrm[i]->props.family;
42a7b32b
DA		 1702 dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
1703 &saddr, &daddr, family);
25ee3286
HX		 1704 err = PTR_ERR(dst);
1705 if (IS_ERR(dst))
1706 goto put_states;
1707 } else
1708 dst_hold(dst);
1709
1710 dst1->xfrm = xfrm[i];
80c802f3 1711 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1712
f5b0a874 1713 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1714 dst1->flags |= DST_HOST;
1715 dst1->lastuse = now;
1716
1717 dst1->input = dst_discard;
43a4dea4 1718 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1719
1720 dst1->next = dst_prev;
1721 dst_prev = dst1;
1722
1723 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1724 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1725 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1726 trailer_len += xfrm[i]->props.trailer_len;
1727 }
1728
1729 dst_prev->child = dst;
1730 dst0->path = dst;
1731
1732 err = -ENODEV;
1733 dev = dst->dev;
1734 if (!dev)
1735 goto free_dst;
1736
a1b05140 1737 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1738 xfrm_init_pmtu(dst_prev);
1739
1740 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1741 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1742
87c1e12b 1743 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1744 if (err)
1745 goto free_dst;
1746
1747 dst_prev->header_len = header_len;
1748 dst_prev->trailer_len = trailer_len;
1749 header_len -= xdst->u.dst.xfrm->props.header_len;
1750 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1751 }
1752
1753out:
1754 return dst0;
1755
1756put_states:
1757 for (; i < nx; i++)
1758 xfrm_state_put(xfrm[i]);
1759free_dst:
1760 if (dst0)
1761 dst_free(dst0);
1762 dst0 = ERR_PTR(err);
1763 goto out;
1764}
1765
da7c224b 1766#ifdef CONFIG_XFRM_SUB_POLICY
be7928d2 1767static int xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1768{
1769 if (!*target) {
1770 *target = kmalloc(size, GFP_ATOMIC);
1771 if (!*target)
1772 return -ENOMEM;
1773 }
be7928d2 1774
157bfc25
MN		 1775 memcpy(*target, src, size);
1776 return 0;
1777}
da7c224b 1778#endif
157bfc25 1779
be7928d2
DB		 1780static int xfrm_dst_update_parent(struct dst_entry *dst,
1781 const struct xfrm_selector *sel)
157bfc25
MN		 1782{
1783#ifdef CONFIG_XFRM_SUB_POLICY
1784 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1785 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1786 sel, sizeof(*sel));
1787#else
1788 return 0;
1789#endif
1790}
1791
be7928d2
DB		 1792static int xfrm_dst_update_origin(struct dst_entry *dst,
1793 const struct flowi *fl)
157bfc25
MN		 1794{
1795#ifdef CONFIG_XFRM_SUB_POLICY
1796 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1797 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1798#else
1799 return 0;
1800#endif
1801}
1da177e4 1802
73ff93cd 1803static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1804 struct xfrm_policy **pols,
1805 int *num_pols, int *num_xfrms)
1806{
1807 int i;
1808
1809 if (*num_pols == 0 || !pols[0]) {
1810 *num_pols = 0;
1811 *num_xfrms = 0;
1812 return 0;
1813 }
1814 if (IS_ERR(pols[0]))
1815 return PTR_ERR(pols[0]);
1816
1817 *num_xfrms = pols[0]->xfrm_nr;
1818
1819#ifdef CONFIG_XFRM_SUB_POLICY
1820 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1821 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1822 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1823 XFRM_POLICY_TYPE_MAIN,
1824 fl, family,
1825 XFRM_POLICY_OUT);
1826 if (pols[1]) {
1827 if (IS_ERR(pols[1])) {
1828 xfrm_pols_put(pols, *num_pols);
1829 return PTR_ERR(pols[1]);
1830 }
02d0892f 1831 (*num_pols)++;
80c802f3
TT		 1832 (*num_xfrms) += pols[1]->xfrm_nr;
1833 }
1834 }
1835#endif
1836 for (i = 0; i < *num_pols; i++) {
1837 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1838 *num_xfrms = -1;
1839 break;
1840 }
1841 }
1842
1843 return 0;
1844
1845}
1846
1847static struct xfrm_dst *
1848xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1849 const struct flowi *fl, u16 family,
80c802f3
TT		 1850 struct dst_entry *dst_orig)
1851{
1852 struct net *net = xp_net(pols[0]);
1853 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1854 struct dst_entry *dst;
1855 struct xfrm_dst *xdst;
1856 int err;
1857
1858 /* Try to instantiate a bundle */
1859 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1860 if (err <= 0) {
1861 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1862 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1863 return ERR_PTR(err);
1864 }
1865
1866 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1867 if (IS_ERR(dst)) {
1868 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1869 return ERR_CAST(dst);
1870 }
1871
1872 xdst = (struct xfrm_dst *)dst;
1873 xdst->num_xfrms = err;
1874 if (num_pols > 1)
1875 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1876 else
1877 err = xfrm_dst_update_origin(dst, fl);
1878 if (unlikely(err)) {
1879 dst_free(dst);
1880 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1881 return ERR_PTR(err);
1882 }
1883
1884 xdst->num_pols = num_pols;
3e94c2dc 1885 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1886 xdst->policy_genid = atomic_read(&pols[0]->genid);
1887
1888 return xdst;
1889}
1890
a0073fe1
SK		 1891static void xfrm_policy_queue_process(unsigned long arg)
1892{
a0073fe1
SK		 1893 struct sk_buff *skb;
1894 struct sock *sk;
1895 struct dst_entry *dst;
a0073fe1 1896 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
3f5312ae 1897 struct net *net = xp_net(pol);
a0073fe1
SK		 1898 struct xfrm_policy_queue *pq = &pol->polq;
1899 struct flowi fl;
1900 struct sk_buff_head list;
1901
1902 spin_lock(&pq->hold_queue.lock);
1903 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1904 if (!skb) {
1905 spin_unlock(&pq->hold_queue.lock);
1906 goto out;
1907 }
a0073fe1
SK		 1908 dst = skb_dst(skb);
1909 sk = skb->sk;
1910 xfrm_decode_session(skb, &fl, dst->ops->family);
1911 spin_unlock(&pq->hold_queue.lock);
1912
1913 dst_hold(dst->path);
3f5312ae 1914 dst = xfrm_lookup(net, dst->path, &fl, sk, 0);
a0073fe1
SK		 1915 if (IS_ERR(dst))
1916 goto purge_queue;
1917
1918 if (dst->flags & DST_XFRM_QUEUE) {
1919 dst_release(dst);
1920
1921 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1922 goto purge_queue;
1923
1924 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1925 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1926 xfrm_pol_hold(pol);
1927 goto out;
a0073fe1
SK		 1928 }
1929
1930 dst_release(dst);
1931
1932 __skb_queue_head_init(&list);
1933
1934 spin_lock(&pq->hold_queue.lock);
1935 pq->timeout = 0;
1936 skb_queue_splice_init(&pq->hold_queue, &list);
1937 spin_unlock(&pq->hold_queue.lock);
1938
1939 while (!skb_queue_empty(&list)) {
1940 skb = __skb_dequeue(&list);
1941
1942 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1943 dst_hold(skb_dst(skb)->path);
3f5312ae 1944 dst = xfrm_lookup(net, skb_dst(skb)->path, &fl, skb->sk, 0);
a0073fe1 1945 if (IS_ERR(dst)) {
a0073fe1
SK		 1946 kfree_skb(skb);
1947 continue;
1948 }
1949
1950 nf_reset(skb);
1951 skb_dst_drop(skb);
1952 skb_dst_set(skb, dst);
1953
13206b6b 1954 dst_output(net, skb->sk, skb);
a0073fe1
SK		 1955 }
1956
e7d8f6cb
SK		 1957out:
1958 xfrm_pol_put(pol);
a0073fe1
SK		 1959 return;
1960
1961purge_queue:
1962 pq->timeout = 0;
1ee5e667 1963 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1964 xfrm_pol_put(pol);
a0073fe1
SK		 1965}
1966
ede2059d 1967static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1968{
1969 unsigned long sched_next;
1970 struct dst_entry *dst = skb_dst(skb);
1971 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1972 struct xfrm_policy *pol = xdst->pols[0];
1973 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1974
39bb5e62 1975 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1976 kfree_skb(skb);
1977 return 0;
1978 }
a0073fe1
SK		 1979
1980 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1981 kfree_skb(skb);
1982 return -EAGAIN;
1983 }
1984
1985 skb_dst_force(skb);
a0073fe1
SK		 1986
1987 spin_lock_bh(&pq->hold_queue.lock);
1988
1989 if (!pq->timeout)
1990 pq->timeout = XFRM_QUEUE_TMO_MIN;
1991
1992 sched_next = jiffies + pq->timeout;
1993
1994 if (del_timer(&pq->hold_timer)) {
1995 if (time_before(pq->hold_timer.expires, sched_next))
1996 sched_next = pq->hold_timer.expires;
e7d8f6cb 1997 xfrm_pol_put(pol);
a0073fe1
SK		 1998 }
1999
2000 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 2001 if (!mod_timer(&pq->hold_timer, sched_next))
2002 xfrm_pol_hold(pol);
a0073fe1
SK		 2003
2004 spin_unlock_bh(&pq->hold_queue.lock);
2005
2006 return 0;
2007}
2008
2009static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 2010 struct xfrm_flo *xflo,
a0073fe1
SK		 2011 const struct flowi *fl,
2012 int num_xfrms,
2013 u16 family)
2014{
2015 int err;
2016 struct net_device *dev;
b8c203b2 2017 struct dst_entry *dst;
a0073fe1
SK		 2018 struct dst_entry *dst1;
2019 struct xfrm_dst *xdst;
2020
2021 xdst = xfrm_alloc_dst(net, family);
2022 if (IS_ERR(xdst))
2023 return xdst;
2024
b8c203b2
SK		 2025 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2026 net->xfrm.sysctl_larval_drop ||
2027 num_xfrms <= 0)
a0073fe1
SK		 2028 return xdst;
2029
b8c203b2 2030 dst = xflo->dst_orig;
a0073fe1
SK		 2031 dst1 = &xdst->u.dst;
2032 dst_hold(dst);
2033 xdst->route = dst;
2034
2035 dst_copy_metrics(dst1, dst);
2036
2037 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2038 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2039 dst1->lastuse = jiffies;
2040
2041 dst1->input = dst_discard;
2042 dst1->output = xdst_queue_output;
2043
2044 dst_hold(dst);
2045 dst1->child = dst;
2046 dst1->path = dst;
2047
2048 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2049
2050 err = -ENODEV;
2051 dev = dst->dev;
2052 if (!dev)
2053 goto free_dst;
2054
2055 err = xfrm_fill_dst(xdst, dev, fl);
2056 if (err)
2057 goto free_dst;
2058
2059out:
2060 return xdst;
2061
2062free_dst:
2063 dst_release(dst1);
2064 xdst = ERR_PTR(err);
2065 goto out;
2066}
2067
80c802f3 2068static struct flow_cache_object *
dee9f4bc 2069xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 2070 struct flow_cache_object *oldflo, void *ctx)
2071{
b8c203b2 2072 struct xfrm_flo *xflo = (struct xfrm_flo *)ctx;
80c802f3
TT		 2073 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2074 struct xfrm_dst *xdst, *new_xdst;
2075 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
2076
2077 /* Check if the policies from old bundle are usable */
2078 xdst = NULL;
2079 if (oldflo) {
2080 xdst = container_of(oldflo, struct xfrm_dst, flo);
2081 num_pols = xdst->num_pols;
2082 num_xfrms = xdst->num_xfrms;
2083 pol_dead = 0;
2084 for (i = 0; i < num_pols; i++) {
2085 pols[i] = xdst->pols[i];
2086 pol_dead |= pols[i]->walk.dead;
2087 }
2088 if (pol_dead) {
2089 dst_free(&xdst->u.dst);
2090 xdst = NULL;
2091 num_pols = 0;
2092 num_xfrms = 0;
2093 oldflo = NULL;
2094 }
2095 }
2096
2097 /* Resolve policies to use if we couldn't get them from
2098 * previous cache entry */
2099 if (xdst == NULL) {
2100 num_pols = 1;
b5fb82c4
BZ		 2101 pols[0] = __xfrm_policy_lookup(net, fl, family,
2102 flow_to_policy_dir(dir));
80c802f3
TT		 2103 err = xfrm_expand_policies(fl, family, pols,
2104 &num_pols, &num_xfrms);
2105 if (err < 0)
2106 goto inc_error;
2107 if (num_pols == 0)
2108 return NULL;
2109 if (num_xfrms <= 0)
2110 goto make_dummy_bundle;
2111 }
2112
b8c203b2
SK		 2113 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
2114 xflo->dst_orig);
80c802f3
TT		 2115 if (IS_ERR(new_xdst)) {
2116 err = PTR_ERR(new_xdst);
2117 if (err != -EAGAIN)
2118 goto error;
2119 if (oldflo == NULL)
2120 goto make_dummy_bundle;
2121 dst_hold(&xdst->u.dst);
2122 return oldflo;
d809ec89
TT		 2123 } else if (new_xdst == NULL) {
2124 num_xfrms = 0;
2125 if (oldflo == NULL)
2126 goto make_dummy_bundle;
2127 xdst->num_xfrms = 0;
2128 dst_hold(&xdst->u.dst);
2129 return oldflo;
80c802f3
TT		 2130 }
2131
2132 /* Kill the previous bundle */
2133 if (xdst) {
2134 /* The policies were stolen for newly generated bundle */
2135 xdst->num_pols = 0;
2136 dst_free(&xdst->u.dst);
2137 }
2138
2139 /* Flow cache does not have reference, it dst_free()'s,
2140 * but we do need to return one reference for original caller */
2141 dst_hold(&new_xdst->u.dst);
2142 return &new_xdst->flo;
2143
2144make_dummy_bundle:
2145 /* We found policies, but there's no bundles to instantiate:
2146 * either because the policy blocks, has no transformations or
2147 * we could not build template (no xfrm_states).*/
b8c203b2 2148 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2149 if (IS_ERR(xdst)) {
2150 xfrm_pols_put(pols, num_pols);
2151 return ERR_CAST(xdst);
2152 }
2153 xdst->num_pols = num_pols;
2154 xdst->num_xfrms = num_xfrms;
3e94c2dc 2155 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2156
2157 dst_hold(&xdst->u.dst);
2158 return &xdst->flo;
2159
2160inc_error:
2161 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2162error:
2163 if (xdst != NULL)
2164 dst_free(&xdst->u.dst);
2165 else
2166 xfrm_pols_put(pols, num_pols);
2167 return ERR_PTR(err);
2168}
1da177e4 2169
2774c131
DM		 2170static struct dst_entry *make_blackhole(struct net *net, u16 family,
2171 struct dst_entry *dst_orig)
2172{
2173 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2174 struct dst_entry *ret;
2175
2176 if (!afinfo) {
2177 dst_release(dst_orig);
433a1954 2178 return ERR_PTR(-EINVAL);
2774c131
DM		 2179 } else {
2180 ret = afinfo->blackhole_route(net, dst_orig);
2181 }
2182 xfrm_policy_put_afinfo(afinfo);
2183
2184 return ret;
2185}
2186
1da177e4
LT		 2187/* Main function: finds/creates a bundle for given flow.
2188 *
2189 * At the moment we eat a raw IP route. Mostly to speed up lookups
2190 * on interfaces with disabled IPsec.
2191 */
452edd59
DM		 2192struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2193 const struct flowi *fl,
6f9c9615 2194 const struct sock *sk, int flags)
1da177e4 2195{
4e81bb83 2196 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2197 struct flow_cache_object *flo;
2198 struct xfrm_dst *xdst;
452edd59 2199 struct dst_entry *dst, *route;
80c802f3 2200 u16 family = dst_orig->ops->family;
df71837d 2201 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2202 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2203
80c802f3
TT		 2204 dst = NULL;
2205 xdst = NULL;
2206 route = NULL;
4e81bb83 2207
bd5eb35f 2208 sk = sk_const_to_full_sk(sk);
f7944fb1 2209 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2210 num_pols = 1;
2211 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2212 err = xfrm_expand_policies(fl, family, pols,
2213 &num_pols, &num_xfrms);
2214 if (err < 0)
75b8c133 2215 goto dropdst;
80c802f3
TT		 2216
2217 if (num_pols) {
2218 if (num_xfrms <= 0) {
2219 drop_pols = num_pols;
2220 goto no_transform;
2221 }
2222
2223 xdst = xfrm_resolve_and_create_bundle(
2224 pols, num_pols, fl,
2225 family, dst_orig);
2226 if (IS_ERR(xdst)) {
2227 xfrm_pols_put(pols, num_pols);
2228 err = PTR_ERR(xdst);
2229 goto dropdst;
d809ec89
TT		 2230 } else if (xdst == NULL) {
2231 num_xfrms = 0;
2232 drop_pols = num_pols;
2233 goto no_transform;
80c802f3
TT		 2234 }
2235
b7eea454
SK		 2236 dst_hold(&xdst->u.dst);
2237 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2238 route = xdst->route;
0aa64774 2239 }
3bccfbc7 2240 }
1da177e4 2241
80c802f3 2242 if (xdst == NULL) {
b8c203b2
SK		 2243 struct xfrm_flo xflo;
2244
2245 xflo.dst_orig = dst_orig;
2246 xflo.flags = flags;
2247
1da177e4 2248 /* To accelerate a bit... */
2518c7c2 2249 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2250 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2251 goto nopol;
1da177e4 2252
80c802f3 2253 flo = flow_cache_lookup(net, fl, family, dir,
b8c203b2 2254 xfrm_bundle_lookup, &xflo);
80c802f3
TT		 2255 if (flo == NULL)
2256 goto nopol;
fe1a5f03 2257 if (IS_ERR(flo)) {
80c802f3 2258 err = PTR_ERR(flo);
75b8c133 2259 goto dropdst;
d66e37a9 2260 }
80c802f3
TT		 2261 xdst = container_of(flo, struct xfrm_dst, flo);
2262
2263 num_pols = xdst->num_pols;
2264 num_xfrms = xdst->num_xfrms;
3e94c2dc 2265 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2266 route = xdst->route;
2267 }
2268
2269 dst = &xdst->u.dst;
2270 if (route == NULL && num_xfrms > 0) {
2271 /* The only case when xfrm_bundle_lookup() returns a
2272 * bundle with null route, is when the template could
2273 * not be resolved. It means policies are there, but
2274 * bundle could not be created, since we don't yet
2275 * have the xfrm_state's. We need to wait for KM to
2276 * negotiate new SA's or bail out with error.*/
2277 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2278 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2279 err = -EREMOTE;
2280 goto error;
80c802f3 2281 }
80c802f3 2282
5b8ef341 2283 err = -EAGAIN;
80c802f3
TT		 2284
2285 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2286 goto error;
1da177e4
LT		 2287 }
2288
80c802f3
TT		 2289no_transform:
2290 if (num_pols == 0)
8b7817f3 2291 goto nopol;
1da177e4 2292
80c802f3
TT		 2293 if ((flags & XFRM_LOOKUP_ICMP) &&
2294 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2295 err = -ENOENT;
8b7817f3 2296 goto error;
80c802f3 2297 }
8b7817f3 2298
80c802f3
TT		 2299 for (i = 0; i < num_pols; i++)
2300 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2301
80c802f3 2302 if (num_xfrms < 0) {
1da177e4 2303 /* Prohibit the flow */
59c9940e 2304 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2305 err = -EPERM;
2306 goto error;
80c802f3
TT		 2307 } else if (num_xfrms > 0) {
2308 /* Flow transformed */
80c802f3
TT		 2309 dst_release(dst_orig);
2310 } else {
2311 /* Flow passes untransformed */
2312 dst_release(dst);
452edd59 2313 dst = dst_orig;
1da177e4 2314 }
80c802f3
TT		 2315ok:
2316 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2317 if (dst && dst->xfrm &&
2318 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2319 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2320 return dst;
1da177e4 2321
80c802f3 2322nopol:
452edd59
DM		 2323 if (!(flags & XFRM_LOOKUP_ICMP)) {
2324 dst = dst_orig;
80c802f3 2325 goto ok;
452edd59 2326 }
80c802f3 2327 err = -ENOENT;
1da177e4 2328error:
80c802f3 2329 dst_release(dst);
75b8c133 2330dropdst:
ac37e251
W		 2331 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2332 dst_release(dst_orig);
80c802f3 2333 xfrm_pols_put(pols, drop_pols);
452edd59 2334 return ERR_PTR(err);
1da177e4
LT		 2335}
2336EXPORT_SYMBOL(xfrm_lookup);
2337
f92ee619
SK		 2338/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2339 * Otherwise we may send out blackholed packets.
2340 */
2341struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2342 const struct flowi *fl,
6f9c9615 2343 const struct sock *sk, int flags)
f92ee619 2344{
b8c203b2 2345 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2346 flags | XFRM_LOOKUP_QUEUE |
2347 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2348
2349 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2350 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2351
2352 return dst;
2353}
2354EXPORT_SYMBOL(xfrm_lookup_route);
2355
df0ba92a 2356static inline int
8f029de2 2357xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2358{
2359 struct xfrm_state *x;
df0ba92a
MN		 2360
2361 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2362 return 0;
2363 x = skb->sp->xvec[idx];
2364 if (!x->type->reject)
2365 return 0;
1ecafede 2366 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2367}
2368
1da177e4
LT		 2369/* When skb is transformed back to its "native" form, we have to
2370 * check policy restrictions. At the moment we make this in maximally
2371 * stupid way. Shame on me. :-) Of course, connected sockets must
2372 * have policy cached at them.
2373 */
2374
2375static inline int
7db454b9 2376xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2377 unsigned short family)
2378{
2379 if (xfrm_state_kern(x))
928ba416 2380 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2381 return x->id.proto == tmpl->id.proto &&
2382 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2383 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2384 x->props.mode == tmpl->mode &&
c5d18e98 2385 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2386 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2387 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2388 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2389}
2390
df0ba92a
MN		 2391/*
2392 * 0 or more than 0 is returned when validation is succeeded (either bypass
2393 * because of optional transport mode, or next index of the mathced secpath
2394 * state with the template.
2395 * -1 is returned when no matching template is found.
2396 * Otherwise "-2 - errored_index" is returned.
2397 */
1da177e4 2398static inline int
22cccb7e 2399xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2400 unsigned short family)
2401{
2402 int idx = start;
2403
2404 if (tmpl->optional) {
7e49e6de 2405 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2406 return start;
2407 } else
2408 start = -1;
2409 for (; idx < sp->len; idx++) {
dbe5b4aa 2410 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2411 return ++idx;
df0ba92a
MN		 2412 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2413 if (start == -1)
2414 start = -2-idx;
1da177e4 2415 break;
df0ba92a 2416 }
1da177e4
LT		 2417 }
2418 return start;
2419}
2420
d5422efe
HX		 2421int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2422 unsigned int family, int reverse)
1da177e4
LT		 2423{
2424 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2425 int err;
1da177e4
LT		 2426
2427 if (unlikely(afinfo == NULL))
2428 return -EAFNOSUPPORT;
2429
d5422efe 2430 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2431 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2432 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2433 return err;
1da177e4 2434}
d5422efe 2435EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2436
9a7386ec 2437static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2438{
2439 for (; k < sp->len; k++) {
df0ba92a 2440 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2441 *idxp = k;
1da177e4 2442 return 1;
df0ba92a 2443 }
1da177e4
LT		 2444 }
2445
2446 return 0;
2447}
2448
a716c119 2449int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2450 unsigned short family)
2451{
f6e1e25d 2452 struct net *net = dev_net(skb->dev);
1da177e4 2453 struct xfrm_policy *pol;
4e81bb83
MN		 2454 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2455 int npols = 0;
2456 int xfrm_nr;
2457 int pi;
d5422efe 2458 int reverse;
1da177e4 2459 struct flowi fl;
d5422efe 2460 u8 fl_dir;
df0ba92a 2461 int xerr_idx = -1;
1da177e4 2462
d5422efe
HX		 2463 reverse = dir & ~XFRM_POLICY_MASK;
2464 dir &= XFRM_POLICY_MASK;
2465 fl_dir = policy_to_flow_dir(dir);
2466
0aa64774 2467 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2468 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2469 return 0;
0aa64774
MN		 2470 }
2471
eb9c7ebe 2472 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2473
2474 /* First, check used SA against their selectors. */
2475 if (skb->sp) {
2476 int i;
2477
9b7a787d 2478 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2479 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2480 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2481 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2482 return 0;
0aa64774 2483 }
1da177e4
LT		 2484 }
2485 }
2486
2487 pol = NULL;
bd5eb35f 2488 sk = sk_to_full_sk(sk);
3bccfbc7 2489 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2490 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2491 if (IS_ERR(pol)) {
59c9940e 2492 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2493 return 0;
0aa64774 2494 }
3bccfbc7 2495 }
1da177e4 2496
fe1a5f03
TT		 2497 if (!pol) {
2498 struct flow_cache_object *flo;
2499
2500 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2501 xfrm_policy_lookup, NULL);
2502 if (IS_ERR_OR_NULL(flo))
2503 pol = ERR_CAST(flo);
2504 else
2505 pol = container_of(flo, struct xfrm_policy, flo);
2506 }
1da177e4 2507
0aa64774 2508 if (IS_ERR(pol)) {
59c9940e 2509 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2510 return 0;
0aa64774 2511 }
134b0fc5 2512
df0ba92a 2513 if (!pol) {
d1d9facf 2514 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2515 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2516 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2517 return 0;
2518 }
2519 return 1;
2520 }
1da177e4 2521
9d729f72 2522 pol->curlft.use_time = get_seconds();
1da177e4 2523
4e81bb83 2524 pols[0] = pol;
02d0892f 2525 npols++;
4e81bb83
MN		 2526#ifdef CONFIG_XFRM_SUB_POLICY
2527 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2528 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2529 &fl, family,
2530 XFRM_POLICY_IN);
2531 if (pols[1]) {
0aa64774 2532 if (IS_ERR(pols[1])) {
59c9940e 2533 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2534 return 0;
0aa64774 2535 }
9d729f72 2536 pols[1]->curlft.use_time = get_seconds();
02d0892f 2537 npols++;
4e81bb83
MN		 2538 }
2539 }
2540#endif
2541
1da177e4
LT		 2542 if (pol->action == XFRM_POLICY_ALLOW) {
2543 struct sec_path *sp;
2544 static struct sec_path dummy;
4e81bb83 2545 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2546 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2547 struct xfrm_tmpl **tpp = tp;
2548 int ti = 0;
1da177e4
LT		 2549 int i, k;
2550
2551 if ((sp = skb->sp) == NULL)
2552 sp = &dummy;
2553
4e81bb83
MN		 2554 for (pi = 0; pi < npols; pi++) {
2555 if (pols[pi] != pol &&
0aa64774 2556 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2557 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2558 goto reject;
0aa64774
MN		 2559 }
2560 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2561 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2562 goto reject_error;
0aa64774 2563 }
4e81bb83
MN		 2564 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2565 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2566 }
2567 xfrm_nr = ti;
41a49cc3 2568 if (npols > 1) {
283bc9f3 2569 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2570 tpp = stp;
2571 }
4e81bb83 2572
1da177e4
LT		 2573 /* For each tunnel xfrm, find the first matching tmpl.
2574 * For each tmpl before that, find corresponding xfrm.
2575 * Order is _important_. Later we will implement
2576 * some barriers, but at the moment barriers
2577 * are implied between each two transformations.
2578 */
4e81bb83
MN		 2579 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2580 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2581 if (k < 0) {
d1d9facf
JM		 2582 if (k < -1)
2583 /* "-2 - errored_index" returned */
2584 xerr_idx = -(2+k);
59c9940e 2585 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2586 goto reject;
df0ba92a 2587 }
1da177e4
LT		 2588 }
2589
0aa64774 2590 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2591 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2592 goto reject;
0aa64774 2593 }
1da177e4 2594
4e81bb83 2595 xfrm_pols_put(pols, npols);
1da177e4
LT		 2596 return 1;
2597 }
59c9940e 2598 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2599
2600reject:
df0ba92a 2601 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2602reject_error:
2603 xfrm_pols_put(pols, npols);
1da177e4
LT		 2604 return 0;
2605}
2606EXPORT_SYMBOL(__xfrm_policy_check);
2607
2608int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2609{
99a66657 2610 struct net *net = dev_net(skb->dev);
1da177e4 2611 struct flowi fl;
adf30907 2612 struct dst_entry *dst;
73137147 2613 int res = 1;
1da177e4 2614
0aa64774 2615 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2616 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2617 return 0;
0aa64774 2618 }
1da177e4 2619
fafeeb6c 2620 skb_dst_force(skb);
adf30907 2621
b8c203b2 2622 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2623 if (IS_ERR(dst)) {
73137147 2624 res = 0;
452edd59
DM		 2625 dst = NULL;
2626 }
adf30907
ED		 2627 skb_dst_set(skb, dst);
2628 return res;
1da177e4
LT		 2629}
2630EXPORT_SYMBOL(__xfrm_route_forward);
2631
d49c73c7
DM		 2632/* Optimize later using cookies and generation ids. */
2633
1da177e4
LT		 2634static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2635{
d49c73c7 2636 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2637 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2638 * get validated by dst_ops->check on every use. We do this
2639 * because when a normal route referenced by an XFRM dst is
2640 * obsoleted we do not go looking around for all parent
2641 * referencing XFRM dsts so that we can invalidate them. It
2642 * is just too much work. Instead we make the checks here on
2643 * every use. For example:
d49c73c7
DM		 2644 *
2645 * XFRM dst A --> IPv4 dst X
2646 *
2647 * X is the "xdst->route" of A (X is also the "dst->path" of A
2648 * in this example). If X is marked obsolete, "A" will not
2649 * notice. That's what we are validating here via the
2650 * stale_bundle() check.
2651 *
2652 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2653 * dst which causes it's ->obsolete field to be set to
2654 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2655 * this, we want to force a new route lookup.
399c180a 2656 */
d49c73c7
DM		 2657 if (dst->obsolete < 0 && !stale_bundle(dst))
2658 return dst;
2659
1da177e4
LT		 2660 return NULL;
2661}
2662
2663static int stale_bundle(struct dst_entry *dst)
2664{
12fdb4d3 2665 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2666}
2667
aabc9761 2668void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2669{
1da177e4 2670 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2671 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2672 dev_hold(dst->dev);
1da177e4
LT		 2673 dev_put(dev);
2674 }
2675}
aabc9761 2676EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2677
2678static void xfrm_link_failure(struct sk_buff *skb)
2679{
2680 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2681}
2682
2683static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2684{
2685 if (dst) {
2686 if (dst->obsolete) {
2687 dst_release(dst);
2688 dst = NULL;
2689 }
2690 }
2691 return dst;
2692}
2693
e4c17216 2694void xfrm_garbage_collect(struct net *net)
c0ed1c14 2695{
ca925cf1 2696 flow_cache_flush(net);
c0ed1c14 2697}
e4c17216 2698EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2699
2700static void xfrm_garbage_collect_deferred(struct net *net)
2701{
ca925cf1 2702 flow_cache_flush_deferred(net);
c0ed1c14
SK		 2703}
2704
25ee3286 2705static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2706{
2707 do {
2708 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2709 u32 pmtu, route_mtu_cached;
2710
2711 pmtu = dst_mtu(dst->child);
2712 xdst->child_mtu_cached = pmtu;
2713
2714 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2715
2716 route_mtu_cached = dst_mtu(xdst->route);
2717 xdst->route_mtu_cached = route_mtu_cached;
2718
2719 if (pmtu > route_mtu_cached)
2720 pmtu = route_mtu_cached;
2721
defb3519 2722 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2723 } while ((dst = dst->next));
2724}
2725
1da177e4
LT		 2726/* Check that the bundle accepts the flow and its components are
2727 * still valid.
2728 */
2729
12fdb4d3 2730static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2731{
2732 struct dst_entry *dst = &first->u.dst;
2733 struct xfrm_dst *last;
2734 u32 mtu;
2735
92d63dec 2736 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2737 (dst->dev && !netif_running(dst->dev)))
2738 return 0;
2739
a0073fe1
SK		 2740 if (dst->flags & DST_XFRM_QUEUE)
2741 return 1;
2742
1da177e4
LT		 2743 last = NULL;
2744
2745 do {
2746 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2747
1da177e4
LT		 2748 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2749 return 0;
80c802f3
TT		 2750 if (xdst->xfrm_genid != dst->xfrm->genid)
2751 return 0;
b1312c89
TT		 2752 if (xdst->num_pols > 0 &&
2753 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2754 return 0;
e53820de 2755
1da177e4
LT		 2756 mtu = dst_mtu(dst->child);
2757 if (xdst->child_mtu_cached != mtu) {
2758 last = xdst;
2759 xdst->child_mtu_cached = mtu;
2760 }
2761
92d63dec 2762 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2763 return 0;
2764 mtu = dst_mtu(xdst->route);
2765 if (xdst->route_mtu_cached != mtu) {
2766 last = xdst;
2767 xdst->route_mtu_cached = mtu;
2768 }
2769
2770 dst = dst->child;
2771 } while (dst->xfrm);
2772
2773 if (likely(!last))
2774 return 1;
2775
2776 mtu = last->child_mtu_cached;
2777 for (;;) {
2778 dst = &last->u.dst;
2779
2780 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2781 if (mtu > last->route_mtu_cached)
2782 mtu = last->route_mtu_cached;
defb3519 2783 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2784
2785 if (last == first)
2786 break;
2787
bd0bf076 2788 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2789 last->child_mtu_cached = mtu;
2790 }
2791
2792 return 1;
2793}
2794
0dbaee3b
DM		 2795static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2796{
2797 return dst_metric_advmss(dst->path);
2798}
2799
ebb762f2 2800static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2801{
618f9bc7
SK		 2802 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2803
2804 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2805}
2806
f894cbf8
DM		 2807static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2808 struct sk_buff *skb,
2809 const void *daddr)
d3aaeb38 2810{
f894cbf8 2811 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2812}
2813
1da177e4
LT		 2814int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2815{
d7c7544c 2816 struct net *net;
1da177e4
LT		 2817 int err = 0;
2818 if (unlikely(afinfo == NULL))
2819 return -EINVAL;
2820 if (unlikely(afinfo->family >= NPROTO))
2821 return -EAFNOSUPPORT;
ef8531b6 2822 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4 2823 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
f31e8d4f 2824 err = -EEXIST;
1da177e4
LT		 2825 else {
2826 struct dst_ops *dst_ops = afinfo->dst_ops;
2827 if (likely(dst_ops->kmem_cachep == NULL))
2828 dst_ops->kmem_cachep = xfrm_dst_cache;
2829 if (likely(dst_ops->check == NULL))
2830 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2831 if (likely(dst_ops->default_advmss == NULL))
2832 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2833 if (likely(dst_ops->mtu == NULL))
2834 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2835 if (likely(dst_ops->negative_advice == NULL))
2836 dst_ops->negative_advice = xfrm_negative_advice;
2837 if (likely(dst_ops->link_failure == NULL))
2838 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2839 if (likely(dst_ops->neigh_lookup == NULL))
2840 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2841 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2842 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2843 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2844 }
ef8531b6 2845 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2846
2847 rtnl_lock();
2848 for_each_net(net) {
2849 struct dst_ops *xfrm_dst_ops;
2850
2851 switch (afinfo->family) {
2852 case AF_INET:
2853 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2854 break;
dfd56b8b 2855#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2856 case AF_INET6:
2857 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2858 break;
2859#endif
2860 default:
2861 BUG();
2862 }
2863 *xfrm_dst_ops = *afinfo->dst_ops;
2864 }
2865 rtnl_unlock();
2866
1da177e4
LT		 2867 return err;
2868}
2869EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2870
2871int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2872{
2873 int err = 0;
2874 if (unlikely(afinfo == NULL))
2875 return -EINVAL;
2876 if (unlikely(afinfo->family >= NPROTO))
2877 return -EAFNOSUPPORT;
ef8531b6 2878 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2879 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2880 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2881 err = -EINVAL;
ef8531b6
ED		 2882 else
2883 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2884 NULL);
2885 }
2886 spin_unlock(&xfrm_policy_afinfo_lock);
2887 if (!err) {
2888 struct dst_ops *dst_ops = afinfo->dst_ops;
2889
2890 synchronize_rcu();
2891
2892 dst_ops->kmem_cachep = NULL;
2893 dst_ops->check = NULL;
2894 dst_ops->negative_advice = NULL;
2895 dst_ops->link_failure = NULL;
2896 afinfo->garbage_collect = NULL;
1da177e4 2897 }
1da177e4
LT		 2898 return err;
2899}
2900EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2901
d7c7544c
AD		 2902static void __net_init xfrm_dst_ops_init(struct net *net)
2903{
2904 struct xfrm_policy_afinfo *afinfo;
2905
ef8531b6
ED		 2906 rcu_read_lock();
2907 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2908 if (afinfo)
2909 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2910#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2911 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2912 if (afinfo)
2913 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2914#endif
418a99ac 2915 rcu_read_unlock();
546be240
HX		 2916}
2917
1da177e4
LT		 2918static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2919{
351638e7 2920 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2921
1da177e4
LT		 2922 switch (event) {
2923 case NETDEV_DOWN:
c0ed1c14 2924 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2925 }
2926 return NOTIFY_DONE;
2927}
2928
2929static struct notifier_block xfrm_dev_notifier = {
d5917a35 2930 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2931};
2932
558f82ef 2933#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2934static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2935{
c68cd1a0 2936 int rv;
698365fa
WC		 2937 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2938 if (!net->mib.xfrm_statistics)
558f82ef 2939 return -ENOMEM;
c68cd1a0
AD		 2940 rv = xfrm_proc_init(net);
2941 if (rv < 0)
698365fa 2942 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2943 return rv;
558f82ef 2944}
59c9940e
AD		 2945
2946static void xfrm_statistics_fini(struct net *net)
2947{
c68cd1a0 2948 xfrm_proc_fini(net);
698365fa 2949 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2950}
2951#else
2952static int __net_init xfrm_statistics_init(struct net *net)
2953{
2954 return 0;
2955}
2956
2957static void xfrm_statistics_fini(struct net *net)
2958{
2959}
558f82ef
MN		 2960#endif
2961
d62ddc21 2962static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2963{
2518c7c2
DM		 2964 unsigned int hmask, sz;
2965 int dir;
2966
d62ddc21
AD		 2967 if (net_eq(net, &init_net))
2968 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2969 sizeof(struct xfrm_dst),
e5d679f3 2970 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2971 NULL);
1da177e4 2972
2518c7c2
DM		 2973 hmask = 8 - 1;
2974 sz = (hmask+1) * sizeof(struct hlist_head);
2975
93b851c1
AD		 2976 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2977 if (!net->xfrm.policy_byidx)
2978 goto out_byidx;
8100bea7 2979 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2980
53c2e285 2981 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2982 struct xfrm_policy_hash *htab;
2983
dc2caba7 2984 net->xfrm.policy_count[dir] = 0;
53c2e285 2985 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2986 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2987
a35f6c5d 2988 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2989 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2990 if (!htab->table)
a35f6c5d
AD		 2991 goto out_bydst;
2992 htab->hmask = hmask;
b58555f1
CG		 2993 htab->dbits4 = 32;
2994 htab->sbits4 = 32;
2995 htab->dbits6 = 128;
2996 htab->sbits6 = 128;
2518c7c2 2997 }
880a6fab
CG		 2998 net->xfrm.policy_hthresh.lbits4 = 32;
2999 net->xfrm.policy_hthresh.rbits4 = 32;
3000 net->xfrm.policy_hthresh.lbits6 = 128;
3001 net->xfrm.policy_hthresh.rbits6 = 128;
3002
3003 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 3004
adfcf0b2 3005 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 3006 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 3007 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21
AD		 3008 if (net_eq(net, &init_net))
3009 register_netdevice_notifier(&xfrm_dev_notifier);
3010 return 0;
93b851c1 3011
a35f6c5d
AD		 3012out_bydst:
3013 for (dir--; dir >= 0; dir--) {
3014 struct xfrm_policy_hash *htab;
3015
3016 htab = &net->xfrm.policy_bydst[dir];
3017 xfrm_hash_free(htab->table, sz);
3018 }
3019 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 3020out_byidx:
3021 return -ENOMEM;
d62ddc21
AD		 3022}
3023
3024static void xfrm_policy_fini(struct net *net)
3025{
93b851c1 3026 unsigned int sz;
8b18f8ea 3027 int dir;
93b851c1 3028
7c2776ee
AD		 3029 flush_work(&net->xfrm.policy_hash_work);
3030#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 3031 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 3032#endif
2e71029e 3033 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 3034
adfcf0b2 3035 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 3036
53c2e285 3037 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 3038 struct xfrm_policy_hash *htab;
3039
8b18f8ea 3040 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 3041
3042 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 3043 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 3044 WARN_ON(!hlist_empty(htab->table));
3045 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 3046 }
3047
8100bea7 3048 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 3049 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
3050 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 3051}
3052
d62ddc21
AD		 3053static int __net_init xfrm_net_init(struct net *net)
3054{
3055 int rv;
3056
59c9940e
AD		 3057 rv = xfrm_statistics_init(net);
3058 if (rv < 0)
3059 goto out_statistics;
d62ddc21
AD		 3060 rv = xfrm_state_init(net);
3061 if (rv < 0)
3062 goto out_state;
3063 rv = xfrm_policy_init(net);
3064 if (rv < 0)
3065 goto out_policy;
d7c7544c 3066 xfrm_dst_ops_init(net);
b27aeadb
AD		 3067 rv = xfrm_sysctl_init(net);
3068 if (rv < 0)
3069 goto out_sysctl;
4a93f509
SK		 3070 rv = flow_cache_init(net);
3071 if (rv < 0)
3072 goto out;
283bc9f3
FD		 3073
3074 /* Initialize the per-net locks here */
3075 spin_lock_init(&net->xfrm.xfrm_state_lock);
3076 rwlock_init(&net->xfrm.xfrm_policy_lock);
283bc9f3
FD		 3077 mutex_init(&net->xfrm.xfrm_cfg_mutex);
3078
d62ddc21
AD		 3079 return 0;
3080
4a93f509
SK		 3081out:
3082 xfrm_sysctl_fini(net);
b27aeadb
AD		 3083out_sysctl:
3084 xfrm_policy_fini(net);
d62ddc21
AD		 3085out_policy:
3086 xfrm_state_fini(net);
3087out_state:
59c9940e
AD		 3088 xfrm_statistics_fini(net);
3089out_statistics:
d62ddc21
AD		 3090 return rv;
3091}
3092
3093static void __net_exit xfrm_net_exit(struct net *net)
3094{
4a93f509 3095 flow_cache_fini(net);
b27aeadb 3096 xfrm_sysctl_fini(net);
d62ddc21
AD		 3097 xfrm_policy_fini(net);
3098 xfrm_state_fini(net);
59c9940e 3099 xfrm_statistics_fini(net);
d62ddc21
AD		 3100}
3101
3102static struct pernet_operations __net_initdata xfrm_net_ops = {
3103 .init = xfrm_net_init,
3104 .exit = xfrm_net_exit,
3105};
3106
1da177e4
LT		 3107void __init xfrm_init(void)
3108{
d62ddc21 3109 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 3110 xfrm_input_init();
3111}
3112
ab5f5e8b 3113#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 3114static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
3115 struct audit_buffer *audit_buf)
ab5f5e8b 3116{
875179fa
PM		 3117 struct xfrm_sec_ctx *ctx = xp->security;
3118 struct xfrm_selector *sel = &xp->selector;
3119
3120 if (ctx)
ab5f5e8b 3121 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3122 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3123
9b7a787d 3124 switch (sel->family) {
ab5f5e8b 3125 case AF_INET:
21454aaa 3126 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3127 if (sel->prefixlen_s != 32)
3128 audit_log_format(audit_buf, " src_prefixlen=%d",
3129 sel->prefixlen_s);
21454aaa 3130 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3131 if (sel->prefixlen_d != 32)
3132 audit_log_format(audit_buf, " dst_prefixlen=%d",
3133 sel->prefixlen_d);
ab5f5e8b
JL		 3134 break;
3135 case AF_INET6:
5b095d98 3136 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3137 if (sel->prefixlen_s != 128)
3138 audit_log_format(audit_buf, " src_prefixlen=%d",
3139 sel->prefixlen_s);
5b095d98 3140 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3141 if (sel->prefixlen_d != 128)
3142 audit_log_format(audit_buf, " dst_prefixlen=%d",
3143 sel->prefixlen_d);
ab5f5e8b
JL		 3144 break;
3145 }
3146}
3147
2e71029e 3148void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3149{
3150 struct audit_buffer *audit_buf;
ab5f5e8b 3151
afeb14b4 3152 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3153 if (audit_buf == NULL)
3154 return;
2e71029e 3155 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3156 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3157 xfrm_audit_common_policyinfo(xp, audit_buf);
3158 audit_log_end(audit_buf);
3159}
3160EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3161
68277acc 3162void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3163 bool task_valid)
ab5f5e8b
JL		 3164{
3165 struct audit_buffer *audit_buf;
ab5f5e8b 3166
afeb14b4 3167 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3168 if (audit_buf == NULL)
3169 return;
2e71029e 3170 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3171 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3172 xfrm_audit_common_policyinfo(xp, audit_buf);
3173 audit_log_end(audit_buf);
3174}
3175EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3176#endif
3177
80c9abaa 3178#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3179static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3180 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3181{
3182 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3183 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3184 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3185 sel_cmp->family) &&
3186 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3187 sel_cmp->family) &&
80c9abaa
SS		 3188 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3189 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3190 return true;
80c9abaa
SS		 3191 }
3192 } else {
3193 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3194 return true;
80c9abaa
SS		 3195 }
3196 }
bc9b35ad 3197 return false;
80c9abaa
SS		 3198}
3199
3e94c2dc
WC		 3200static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3201 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3202{
3203 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3204 struct hlist_head *chain;
3205 u32 priority = ~0U;
3206
283bc9f3 3207 read_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME*/
8d549c4f 3208 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3209 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3210 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3211 pol->type == type) {
3212 ret = pol;
3213 priority = ret->priority;
3214 break;
3215 }
3216 }
8d549c4f 3217 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3218 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR		 3219 if ((pol->priority >= priority) && ret)
3220 break;
3221
80c9abaa 3222 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
8faf491e 3223 pol->type == type) {
80c9abaa
SS		 3224 ret = pol;
3225 break;
3226 }
3227 }
3228
586f2eb4 3229 xfrm_pol_hold(ret);
80c9abaa 3230
283bc9f3 3231 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3232
3233 return ret;
3234}
3235
dd701754 3236static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3237{
3238 int match = 0;
3239
3240 if (t->mode == m->mode && t->id.proto == m->proto &&
3241 (m->reqid == 0 || t->reqid == m->reqid)) {
3242 switch (t->mode) {
3243 case XFRM_MODE_TUNNEL:
3244 case XFRM_MODE_BEET:
70e94e66
YH		 3245 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3246 m->old_family) &&
3247 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3248 m->old_family)) {
80c9abaa
SS		 3249 match = 1;
3250 }
3251 break;
3252 case XFRM_MODE_TRANSPORT:
3253 /* in case of transport mode, template does not store
3254 any IP addresses, hence we just compare mode and
3255 protocol */
3256 match = 1;
3257 break;
3258 default:
3259 break;
3260 }
3261 }
3262 return match;
3263}
3264
3265/* update endpoint address(es) of template(s) */
3266static int xfrm_policy_migrate(struct xfrm_policy *pol,
3267 struct xfrm_migrate *m, int num_migrate)
3268{
3269 struct xfrm_migrate *mp;
80c9abaa
SS		 3270 int i, j, n = 0;
3271
3272 write_lock_bh(&pol->lock);
12a169e7 3273 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3274 /* target policy has been deleted */
3275 write_unlock_bh(&pol->lock);
3276 return -ENOENT;
3277 }
3278
3279 for (i = 0; i < pol->xfrm_nr; i++) {
3280 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3281 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3282 continue;
3283 n++;
1bfcb10f
HX		 3284 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3285 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3286 continue;
3287 /* update endpoints */
3288 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3289 sizeof(pol->xfrm_vec[i].id.daddr));
3290 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3291 sizeof(pol->xfrm_vec[i].saddr));
3292 pol->xfrm_vec[i].encap_family = mp->new_family;
3293 /* flush bundles */
80c802f3 3294 atomic_inc(&pol->genid);
80c9abaa
SS		 3295 }
3296 }
3297
3298 write_unlock_bh(&pol->lock);
3299
3300 if (!n)
3301 return -ENODATA;
3302
3303 return 0;
3304}
3305
dd701754 3306static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3307{
3308 int i, j;
3309
3310 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3311 return -EINVAL;
3312
3313 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3314 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3315 m[i].old_family) &&
3316 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3317 m[i].old_family))
80c9abaa
SS		 3318 return -EINVAL;
3319 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3320 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3321 return -EINVAL;
3322
3323 /* check if there is any duplicated entry */
3324 for (j = i + 1; j < num_migrate; j++) {
3325 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3326 sizeof(m[i].old_daddr)) &&
3327 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3328 sizeof(m[i].old_saddr)) &&
3329 m[i].proto == m[j].proto &&
3330 m[i].mode == m[j].mode &&
3331 m[i].reqid == m[j].reqid &&
3332 m[i].old_family == m[j].old_family)
3333 return -EINVAL;
3334 }
3335 }
3336
3337 return 0;
3338}
3339
b4b7c0b3 3340int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3341 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3342 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3343{
3344 int i, err, nx_cur = 0, nx_new = 0;
3345 struct xfrm_policy *pol = NULL;
3346 struct xfrm_state *x, *xc;
3347 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3348 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3349 struct xfrm_migrate *mp;
3350
3351 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3352 goto out;
3353
3354 /* Stage 1 - find policy */
8d549c4f 3355 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3356 err = -ENOENT;
3357 goto out;
3358 }
3359
3360 /* Stage 2 - find and update state(s) */
3361 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3362 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3363 x_cur[nx_cur] = x;
3364 nx_cur++;
3365 if ((xc = xfrm_state_migrate(x, mp))) {
3366 x_new[nx_new] = xc;
3367 nx_new++;
3368 } else {
3369 err = -ENODATA;
3370 goto restore_state;
3371 }
3372 }
3373 }
3374
3375 /* Stage 3 - update policy */
3376 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3377 goto restore_state;
3378
3379 /* Stage 4 - delete old state(s) */
3380 if (nx_cur) {
3381 xfrm_states_put(x_cur, nx_cur);
3382 xfrm_states_delete(x_cur, nx_cur);
3383 }
3384
3385 /* Stage 5 - announce */
13c1d189 3386 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3387
3388 xfrm_pol_put(pol);
3389
3390 return 0;
3391out:
3392 return err;
3393
3394restore_state:
3395 if (pol)
3396 xfrm_pol_put(pol);
3397 if (nx_cur)
3398 xfrm_states_put(x_cur, nx_cur);
3399 if (nx_new)
3400 xfrm_states_delete(x_new, nx_new);
3401
3402 return err;
3403}
e610e679 3404EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3405#endif
Linux kernel - Deliverables
This page took 1.151917 seconds and 5 git commands to generate.