projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-linus-1' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
[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
c5b3cf46 118static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
6418c4e0
DM		 119 const xfrm_address_t *saddr,
120 const xfrm_address_t *daddr,
9bb182a7
YH		 121 int family)
122{
123 struct xfrm_policy_afinfo *afinfo;
124 struct dst_entry *dst;
125
126 afinfo = xfrm_policy_get_afinfo(family);
127 if (unlikely(afinfo == NULL))
128 return ERR_PTR(-EAFNOSUPPORT);
129
c5b3cf46 130 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 131
132 xfrm_policy_put_afinfo(afinfo);
133
134 return dst;
135}
136
25ee3286 137static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 138 xfrm_address_t *prev_saddr,
139 xfrm_address_t *prev_daddr,
25ee3286 140 int family)
1da177e4 141{
c5b3cf46 142 struct net *net = xs_net(x);
66cdb3ca
HX		 143 xfrm_address_t *saddr = &x->props.saddr;
144 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 145 struct dst_entry *dst;
146
9bb182a7 147 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 148 saddr = x->coaddr;
9bb182a7
YH		 149 daddr = prev_daddr;
150 }
151 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
152 saddr = prev_saddr;
66cdb3ca 153 daddr = x->coaddr;
9bb182a7 154 }
1da177e4 155
c5b3cf46 156 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 157
158 if (!IS_ERR(dst)) {
159 if (prev_saddr != saddr)
160 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
161 if (prev_daddr != daddr)
162 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
163 }
1da177e4 164
66cdb3ca 165 return dst;
1da177e4 166}
1da177e4 167
1da177e4
LT		 168static inline unsigned long make_jiffies(long secs)
169{
170 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
171 return MAX_SCHEDULE_TIMEOUT-1;
172 else
a716c119 173 return secs*HZ;
1da177e4
LT		 174}
175
176static void xfrm_policy_timer(unsigned long data)
177{
3e94c2dc 178 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 179 unsigned long now = get_seconds();
1da177e4
LT		 180 long next = LONG_MAX;
181 int warn = 0;
182 int dir;
183
184 read_lock(&xp->lock);
185
ea2dea9d 186 if (unlikely(xp->walk.dead))
1da177e4
LT		 187 goto out;
188
77d8d7a6 189 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 190
191 if (xp->lft.hard_add_expires_seconds) {
192 long tmo = xp->lft.hard_add_expires_seconds +
193 xp->curlft.add_time - now;
194 if (tmo <= 0)
195 goto expired;
196 if (tmo < next)
197 next = tmo;
198 }
199 if (xp->lft.hard_use_expires_seconds) {
200 long tmo = xp->lft.hard_use_expires_seconds +
201 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
202 if (tmo <= 0)
203 goto expired;
204 if (tmo < next)
205 next = tmo;
206 }
207 if (xp->lft.soft_add_expires_seconds) {
208 long tmo = xp->lft.soft_add_expires_seconds +
209 xp->curlft.add_time - now;
210 if (tmo <= 0) {
211 warn = 1;
212 tmo = XFRM_KM_TIMEOUT;
213 }
214 if (tmo < next)
215 next = tmo;
216 }
217 if (xp->lft.soft_use_expires_seconds) {
218 long tmo = xp->lft.soft_use_expires_seconds +
219 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
220 if (tmo <= 0) {
221 warn = 1;
222 tmo = XFRM_KM_TIMEOUT;
223 }
224 if (tmo < next)
225 next = tmo;
226 }
227
228 if (warn)
6c5c8ca7 229 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 230 if (next != LONG_MAX &&
231 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
232 xfrm_pol_hold(xp);
233
234out:
235 read_unlock(&xp->lock);
236 xfrm_pol_put(xp);
237 return;
238
239expired:
240 read_unlock(&xp->lock);
4666faab 241 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 242 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 243 xfrm_pol_put(xp);
244}
245
fe1a5f03
TT		 246static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
247{
248 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
249
250 if (unlikely(pol->walk.dead))
251 flo = NULL;
252 else
253 xfrm_pol_hold(pol);
254
255 return flo;
256}
257
258static int xfrm_policy_flo_check(struct flow_cache_object *flo)
259{
260 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
261
262 return !pol->walk.dead;
263}
264
265static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
266{
267 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
268}
269
270static const struct flow_cache_ops xfrm_policy_fc_ops = {
271 .get = xfrm_policy_flo_get,
272 .check = xfrm_policy_flo_check,
273 .delete = xfrm_policy_flo_delete,
274};
1da177e4
LT		 275
276/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
277 * SPD calls.
278 */
279
0331b1f3 280struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 281{
282 struct xfrm_policy *policy;
283
0da974f4 284 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 285
286 if (policy) {
0331b1f3 287 write_pnet(&policy->xp_net, net);
12a169e7 288 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 289 INIT_HLIST_NODE(&policy->bydst);
290 INIT_HLIST_NODE(&policy->byidx);
1da177e4 291 rwlock_init(&policy->lock);
2518c7c2 292 atomic_set(&policy->refcnt, 1);
a0073fe1 293 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 294 setup_timer(&policy->timer, xfrm_policy_timer,
295 (unsigned long)policy);
a0073fe1
SK		 296 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
297 (unsigned long)policy);
fe1a5f03 298 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 299 }
300 return policy;
301}
302EXPORT_SYMBOL(xfrm_policy_alloc);
303
304/* Destroy xfrm_policy: descendant resources must be released to this moment. */
305
64c31b3f 306void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 307{
12a169e7 308 BUG_ON(!policy->walk.dead);
1da177e4 309
0659eea9 310 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 311 BUG();
312
03e1ad7b 313 security_xfrm_policy_free(policy->security);
1da177e4
LT		 314 kfree(policy);
315}
64c31b3f 316EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 317
a0073fe1
SK		 318static void xfrm_queue_purge(struct sk_buff_head *list)
319{
320 struct sk_buff *skb;
321
302a50bc 322 while ((skb = skb_dequeue(list)) != NULL)
a0073fe1 323 kfree_skb(skb);
a0073fe1
SK		 324}
325
1da177e4
LT		 326/* Rule must be locked. Release descentant resources, announce
327 * entry dead. The rule must be unlinked from lists to the moment.
328 */
329
330static void xfrm_policy_kill(struct xfrm_policy *policy)
331{
12a169e7 332 policy->walk.dead = 1;
1da177e4 333
285ead17 334 atomic_inc(&policy->genid);
1da177e4 335
e7d8f6cb
SK		 336 if (del_timer(&policy->polq.hold_timer))
337 xfrm_pol_put(policy);
a0073fe1
SK		 338 xfrm_queue_purge(&policy->polq.hold_queue);
339
285ead17
TT		 340 if (del_timer(&policy->timer))
341 xfrm_pol_put(policy);
342
343 xfrm_pol_put(policy);
1da177e4
LT		 344}
345
2518c7c2
DM		 346static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
347
e92303f8 348static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 349{
e92303f8 350 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 351}
352
b58555f1
CG		 353/* calculate policy hash thresholds */
354static void __get_hash_thresh(struct net *net,
355 unsigned short family, int dir,
356 u8 *dbits, u8 *sbits)
357{
358 switch (family) {
359 case AF_INET:
360 *dbits = net->xfrm.policy_bydst[dir].dbits4;
361 *sbits = net->xfrm.policy_bydst[dir].sbits4;
362 break;
363
364 case AF_INET6:
365 *dbits = net->xfrm.policy_bydst[dir].dbits6;
366 *sbits = net->xfrm.policy_bydst[dir].sbits6;
367 break;
368
369 default:
370 *dbits = 0;
371 *sbits = 0;
372 }
373}
374
5f803b58
DM		 375static struct hlist_head *policy_hash_bysel(struct net *net,
376 const struct xfrm_selector *sel,
377 unsigned short family, int dir)
2518c7c2 378{
1121994c 379 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 380 unsigned int hash;
381 u8 dbits;
382 u8 sbits;
383
384 __get_hash_thresh(net, family, dir, &dbits, &sbits);
385 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2
DM		 386
387 return (hash == hmask + 1 ?
1121994c
AD		 388 &net->xfrm.policy_inexact[dir] :
389 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 390}
391
5f803b58
DM		 392static struct hlist_head *policy_hash_direct(struct net *net,
393 const xfrm_address_t *daddr,
394 const xfrm_address_t *saddr,
395 unsigned short family, int dir)
2518c7c2 396{
1121994c 397 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 398 unsigned int hash;
399 u8 dbits;
400 u8 sbits;
401
402 __get_hash_thresh(net, family, dir, &dbits, &sbits);
403 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 404
1121994c 405 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 406}
407
b58555f1
CG		 408static void xfrm_dst_hash_transfer(struct net *net,
409 struct hlist_head *list,
2518c7c2 410 struct hlist_head *ndsttable,
b58555f1
CG		 411 unsigned int nhashmask,
412 int dir)
2518c7c2 413{
b67bfe0d 414 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 415 struct xfrm_policy *pol;
b791160b 416 unsigned int h0 = 0;
b58555f1
CG		 417 u8 dbits;
418 u8 sbits;
2518c7c2 419
b791160b 420redo:
b67bfe0d 421 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 422 unsigned int h;
423
b58555f1 424 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 425 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 426 pol->family, nhashmask, dbits, sbits);
b791160b 427 if (!entry0) {
b67bfe0d 428 hlist_del(&pol->bydst);
b791160b
YH		 429 hlist_add_head(&pol->bydst, ndsttable+h);
430 h0 = h;
431 } else {
432 if (h != h0)
433 continue;
b67bfe0d 434 hlist_del(&pol->bydst);
1d023284 435 hlist_add_behind(&pol->bydst, entry0);
b791160b 436 }
b67bfe0d 437 entry0 = &pol->bydst;
b791160b
YH		 438 }
439 if (!hlist_empty(list)) {
440 entry0 = NULL;
441 goto redo;
2518c7c2
DM		 442 }
443}
444
445static void xfrm_idx_hash_transfer(struct hlist_head *list,
446 struct hlist_head *nidxtable,
447 unsigned int nhashmask)
448{
b67bfe0d 449 struct hlist_node *tmp;
2518c7c2
DM		 450 struct xfrm_policy *pol;
451
b67bfe0d 452 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 453 unsigned int h;
454
455 h = __idx_hash(pol->index, nhashmask);
456 hlist_add_head(&pol->byidx, nidxtable+h);
457 }
458}
459
460static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
461{
462 return ((old_hmask + 1) << 1) - 1;
463}
464
66caf628 465static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 466{
66caf628 467 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 468 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
469 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 470 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 471 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 472 int i;
473
474 if (!ndst)
475 return;
476
283bc9f3 477 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 478
479 for (i = hmask; i >= 0; i--)
b58555f1 480 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 481
66caf628
AD		 482 net->xfrm.policy_bydst[dir].table = ndst;
483 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 484
283bc9f3 485 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 486
44e36b42 487 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 488}
489
66caf628 490static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 491{
66caf628 492 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 493 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
494 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 495 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 496 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 497 int i;
498
499 if (!nidx)
500 return;
501
283bc9f3 502 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 503
504 for (i = hmask; i >= 0; i--)
505 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
506
66caf628
AD		 507 net->xfrm.policy_byidx = nidx;
508 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 509
283bc9f3 510 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 511
44e36b42 512 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 513}
514
66caf628 515static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 516{
66caf628
AD		 517 unsigned int cnt = net->xfrm.policy_count[dir];
518 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 519
520 if (total)
521 *total += cnt;
522
523 if ((hmask + 1) < xfrm_policy_hashmax &&
524 cnt > hmask)
525 return 1;
526
527 return 0;
528}
529
66caf628 530static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 531{
66caf628 532 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 533
534 if ((hmask + 1) < xfrm_policy_hashmax &&
535 total > hmask)
536 return 1;
537
538 return 0;
539}
540
e071041b 541void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 542{
283bc9f3 543 read_lock_bh(&net->xfrm.xfrm_policy_lock);
e071041b
AD		 544 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
545 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
546 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
547 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
548 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
549 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
550 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 551 si->spdhmcnt = xfrm_policy_hashmax;
283bc9f3 552 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
ecfd6b18
JHS		 553}
554EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 555
ecfd6b18 556static DEFINE_MUTEX(hash_resize_mutex);
66caf628 557static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 558{
66caf628 559 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 560 int dir, total;
561
562 mutex_lock(&hash_resize_mutex);
563
564 total = 0;
53c2e285 565 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD		 566 if (xfrm_bydst_should_resize(net, dir, &total))
567 xfrm_bydst_resize(net, dir);
2518c7c2 568 }
66caf628
AD		 569 if (xfrm_byidx_should_resize(net, total))
570 xfrm_byidx_resize(net, total);
2518c7c2
DM		 571
572 mutex_unlock(&hash_resize_mutex);
573}
574
880a6fab
CG		 575static void xfrm_hash_rebuild(struct work_struct *work)
576{
577 struct net *net = container_of(work, struct net,
578 xfrm.policy_hthresh.work);
579 unsigned int hmask;
580 struct xfrm_policy *pol;
581 struct xfrm_policy *policy;
582 struct hlist_head *chain;
583 struct hlist_head *odst;
584 struct hlist_node *newpos;
585 int i;
586 int dir;
587 unsigned seq;
588 u8 lbits4, rbits4, lbits6, rbits6;
589
590 mutex_lock(&hash_resize_mutex);
591
592 /* read selector prefixlen thresholds */
593 do {
594 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
595
596 lbits4 = net->xfrm.policy_hthresh.lbits4;
597 rbits4 = net->xfrm.policy_hthresh.rbits4;
598 lbits6 = net->xfrm.policy_hthresh.lbits6;
599 rbits6 = net->xfrm.policy_hthresh.rbits6;
600 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
601
602 write_lock_bh(&net->xfrm.xfrm_policy_lock);
603
604 /* reset the bydst and inexact table in all directions */
53c2e285 605 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG		 606 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
607 hmask = net->xfrm.policy_bydst[dir].hmask;
608 odst = net->xfrm.policy_bydst[dir].table;
609 for (i = hmask; i >= 0; i--)
610 INIT_HLIST_HEAD(odst + i);
611 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
612 /* dir out => dst = remote, src = local */
613 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
614 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
615 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
616 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
617 } else {
618 /* dir in/fwd => dst = local, src = remote */
619 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
620 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
621 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
622 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
623 }
624 }
625
626 /* re-insert all policies by order of creation */
627 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
628 newpos = NULL;
629 chain = policy_hash_bysel(net, &policy->selector,
630 policy->family,
631 xfrm_policy_id2dir(policy->index));
632 hlist_for_each_entry(pol, chain, bydst) {
633 if (policy->priority >= pol->priority)
634 newpos = &pol->bydst;
635 else
636 break;
637 }
638 if (newpos)
639 hlist_add_behind(&policy->bydst, newpos);
640 else
641 hlist_add_head(&policy->bydst, chain);
642 }
643
644 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
645
646 mutex_unlock(&hash_resize_mutex);
647}
648
649void xfrm_policy_hash_rebuild(struct net *net)
650{
651 schedule_work(&net->xfrm.policy_hthresh.work);
652}
653EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
654
1da177e4
LT		 655/* Generate new index... KAME seems to generate them ordered by cost
656 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 657static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 658{
1da177e4
LT		 659 static u32 idx_generator;
660
661 for (;;) {
2518c7c2
DM		 662 struct hlist_head *list;
663 struct xfrm_policy *p;
664 u32 idx;
665 int found;
666
e682adf0
FD		 667 if (!index) {
668 idx = (idx_generator | dir);
669 idx_generator += 8;
670 } else {
671 idx = index;
672 index = 0;
673 }
674
1da177e4
LT		 675 if (idx == 0)
676 idx = 8;
1121994c 677 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 678 found = 0;
b67bfe0d 679 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 680 if (p->index == idx) {
681 found = 1;
1da177e4 682 break;
2518c7c2 683 }
1da177e4 684 }
2518c7c2 685 if (!found)
1da177e4
LT		 686 return idx;
687 }
688}
689
2518c7c2
DM		 690static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
691{
692 u32 *p1 = (u32 *) s1;
693 u32 *p2 = (u32 *) s2;
694 int len = sizeof(struct xfrm_selector) / sizeof(u32);
695 int i;
696
697 for (i = 0; i < len; i++) {
698 if (p1[i] != p2[i])
699 return 1;
700 }
701
702 return 0;
703}
704
a0073fe1
SK		 705static void xfrm_policy_requeue(struct xfrm_policy *old,
706 struct xfrm_policy *new)
707{
708 struct xfrm_policy_queue *pq = &old->polq;
709 struct sk_buff_head list;
710
711 __skb_queue_head_init(&list);
712
713 spin_lock_bh(&pq->hold_queue.lock);
714 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 715 if (del_timer(&pq->hold_timer))
716 xfrm_pol_put(old);
a0073fe1
SK		 717 spin_unlock_bh(&pq->hold_queue.lock);
718
719 if (skb_queue_empty(&list))
720 return;
721
722 pq = &new->polq;
723
724 spin_lock_bh(&pq->hold_queue.lock);
725 skb_queue_splice(&list, &pq->hold_queue);
726 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 727 if (!mod_timer(&pq->hold_timer, jiffies))
728 xfrm_pol_hold(new);
a0073fe1
SK		 729 spin_unlock_bh(&pq->hold_queue.lock);
730}
731
7cb8a939
SK		 732static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
733 struct xfrm_policy *pol)
734{
735 u32 mark = policy->mark.v & policy->mark.m;
736
737 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
738 return true;
739
740 if ((mark & pol->mark.m) == pol->mark.v &&
741 policy->priority == pol->priority)
742 return true;
743
744 return false;
745}
746
1da177e4
LT		 747int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
748{
1121994c 749 struct net *net = xp_net(policy);
2518c7c2
DM		 750 struct xfrm_policy *pol;
751 struct xfrm_policy *delpol;
752 struct hlist_head *chain;
b67bfe0d 753 struct hlist_node *newpos;
1da177e4 754
283bc9f3 755 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 756 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 757 delpol = NULL;
758 newpos = NULL;
b67bfe0d 759 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 760 if (pol->type == policy->type &&
2518c7c2 761 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 762 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 763 xfrm_sec_ctx_match(pol->security, policy->security) &&
764 !WARN_ON(delpol)) {
1da177e4 765 if (excl) {
283bc9f3 766 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 767 return -EEXIST;
768 }
1da177e4
LT		 769 delpol = pol;
770 if (policy->priority > pol->priority)
771 continue;
772 } else if (policy->priority >= pol->priority) {
a6c7ab55 773 newpos = &pol->bydst;
1da177e4
LT		 774 continue;
775 }
1da177e4
LT		 776 if (delpol)
777 break;
1da177e4
LT		 778 }
779 if (newpos)
1d023284 780 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 781 else
782 hlist_add_head(&policy->bydst, chain);
12bfa8bd 783 __xfrm_policy_link(policy, dir);
ca925cf1 784 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 785
786 /* After previous checking, family can either be AF_INET or AF_INET6 */
787 if (policy->family == AF_INET)
788 rt_genid_bump_ipv4(net);
789 else
790 rt_genid_bump_ipv6(net);
791
a0073fe1
SK		 792 if (delpol) {
793 xfrm_policy_requeue(delpol, policy);
29fa0b30 794 __xfrm_policy_unlink(delpol, dir);
a0073fe1 795 }
e682adf0 796 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 797 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 798 policy->curlft.add_time = get_seconds();
1da177e4
LT		 799 policy->curlft.use_time = 0;
800 if (!mod_timer(&policy->timer, jiffies + HZ))
801 xfrm_pol_hold(policy);
283bc9f3 802 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 803
9b78a82c 804 if (delpol)
1da177e4 805 xfrm_policy_kill(delpol);
1121994c
AD		 806 else if (xfrm_bydst_should_resize(net, dir, NULL))
807 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 808
1da177e4
LT		 809 return 0;
810}
811EXPORT_SYMBOL(xfrm_policy_insert);
812
8ca2e93b
JHS		 813struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
814 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 815 struct xfrm_sec_ctx *ctx, int delete,
816 int *err)
1da177e4 817{
2518c7c2
DM		 818 struct xfrm_policy *pol, *ret;
819 struct hlist_head *chain;
1da177e4 820
ef41aaa0 821 *err = 0;
283bc9f3 822 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 823 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 824 ret = NULL;
b67bfe0d 825 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 826 if (pol->type == type &&
34f8d884 827 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 828 !selector_cmp(sel, &pol->selector) &&
829 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 830 xfrm_pol_hold(pol);
2518c7c2 831 if (delete) {
03e1ad7b
PM		 832 *err = security_xfrm_policy_delete(
833 pol->security);
ef41aaa0 834 if (*err) {
283bc9f3 835 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 836 return pol;
837 }
29fa0b30 838 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 839 }
840 ret = pol;
1da177e4
LT		 841 break;
842 }
843 }
283bc9f3 844 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 845
fe1a5f03 846 if (ret && delete)
2518c7c2 847 xfrm_policy_kill(ret);
2518c7c2 848 return ret;
1da177e4 849}
df71837d 850EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 851
8ca2e93b
JHS		 852struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
853 int dir, u32 id, int delete, int *err)
1da177e4 854{
2518c7c2
DM		 855 struct xfrm_policy *pol, *ret;
856 struct hlist_head *chain;
1da177e4 857
b5505c6e
HX		 858 *err = -ENOENT;
859 if (xfrm_policy_id2dir(id) != dir)
860 return NULL;
861
ef41aaa0 862 *err = 0;
283bc9f3 863 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 864 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 865 ret = NULL;
b67bfe0d 866 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 867 if (pol->type == type && pol->index == id &&
868 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 869 xfrm_pol_hold(pol);
2518c7c2 870 if (delete) {
03e1ad7b
PM		 871 *err = security_xfrm_policy_delete(
872 pol->security);
ef41aaa0 873 if (*err) {
283bc9f3 874 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 875 return pol;
876 }
29fa0b30 877 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 878 }
879 ret = pol;
1da177e4
LT		 880 break;
881 }
882 }
283bc9f3 883 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 884
fe1a5f03 885 if (ret && delete)
2518c7c2 886 xfrm_policy_kill(ret);
2518c7c2 887 return ret;
1da177e4
LT		 888}
889EXPORT_SYMBOL(xfrm_policy_byid);
890
4aa2e62c
JL		 891#ifdef CONFIG_SECURITY_NETWORK_XFRM
892static inline int
2e71029e 893xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 894{
4aa2e62c
JL		 895 int dir, err = 0;
896
897 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
898 struct xfrm_policy *pol;
4aa2e62c
JL		 899 int i;
900
b67bfe0d 901 hlist_for_each_entry(pol,
33ffbbd5 902 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 903 if (pol->type != type)
904 continue;
03e1ad7b 905 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 906 if (err) {
2e71029e 907 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 908 return err;
909 }
7dc12d6d 910 }
33ffbbd5 911 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 912 hlist_for_each_entry(pol,
33ffbbd5 913 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 914 bydst) {
915 if (pol->type != type)
916 continue;
03e1ad7b
PM		 917 err = security_xfrm_policy_delete(
918 pol->security);
4aa2e62c 919 if (err) {
ab5f5e8b 920 xfrm_audit_policy_delete(pol, 0,
2e71029e 921 task_valid);
4aa2e62c
JL		 922 return err;
923 }
924 }
925 }
926 }
927 return err;
928}
929#else
930static inline int
2e71029e 931xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 932{
933 return 0;
934}
935#endif
936
2e71029e 937int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 938{
2f1eb65f 939 int dir, err = 0, cnt = 0;
1da177e4 940
283bc9f3 941 write_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 942
2e71029e 943 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 944 if (err)
945 goto out;
946
1da177e4 947 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 948 struct xfrm_policy *pol;
29fa0b30 949 int i;
2518c7c2
DM		 950
951 again1:
b67bfe0d 952 hlist_for_each_entry(pol,
33ffbbd5 953 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 954 if (pol->type != type)
955 continue;
ea2dea9d 956 __xfrm_policy_unlink(pol, dir);
283bc9f3 957 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 958 cnt++;
1da177e4 959
2e71029e 960 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 961
2518c7c2 962 xfrm_policy_kill(pol);
1da177e4 963
283bc9f3 964 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 965 goto again1;
966 }
967
33ffbbd5 968 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 969 again2:
b67bfe0d 970 hlist_for_each_entry(pol,
33ffbbd5 971 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 972 bydst) {
973 if (pol->type != type)
974 continue;
ea2dea9d 975 __xfrm_policy_unlink(pol, dir);
283bc9f3 976 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 977 cnt++;
2518c7c2 978
2e71029e 979 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 980 xfrm_policy_kill(pol);
981
283bc9f3 982 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 983 goto again2;
984 }
1da177e4 985 }
2518c7c2 986
1da177e4 987 }
2f1eb65f
JHS		 988 if (!cnt)
989 err = -ESRCH;
4aa2e62c 990out:
283bc9f3 991 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 992 return err;
1da177e4
LT		 993}
994EXPORT_SYMBOL(xfrm_policy_flush);
995
cdcbca7c 996int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 997 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 998 void *data)
999{
12a169e7
HX		 1000 struct xfrm_policy *pol;
1001 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 1002 int error = 0;
1003
1004 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
1005 walk->type != XFRM_POLICY_TYPE_ANY)
1006 return -EINVAL;
1da177e4 1007
12a169e7 1008 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 1009 return 0;
1010
283bc9f3 1011 write_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1012 if (list_empty(&walk->walk.all))
cdcbca7c 1013 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 1014 else
1015 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 1016 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1017 if (x->dead)
4c563f76 1018 continue;
12a169e7
HX		 1019 pol = container_of(x, struct xfrm_policy, walk);
1020 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1021 walk->type != pol->type)
1022 continue;
1023 error = func(pol, xfrm_policy_id2dir(pol->index),
1024 walk->seq, data);
1025 if (error) {
1026 list_move_tail(&walk->walk.all, &x->all);
1027 goto out;
2518c7c2 1028 }
12a169e7 1029 walk->seq++;
1da177e4 1030 }
12a169e7 1031 if (walk->seq == 0) {
baf5d743
JHS		 1032 error = -ENOENT;
1033 goto out;
1034 }
12a169e7 1035 list_del_init(&walk->walk.all);
1da177e4 1036out:
283bc9f3 1037 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1038 return error;
1039}
1040EXPORT_SYMBOL(xfrm_policy_walk);
1041
12a169e7
HX		 1042void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1043{
1044 INIT_LIST_HEAD(&walk->walk.all);
1045 walk->walk.dead = 1;
1046 walk->type = type;
1047 walk->seq = 0;
1048}
1049EXPORT_SYMBOL(xfrm_policy_walk_init);
1050
283bc9f3 1051void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1052{
1053 if (list_empty(&walk->walk.all))
1054 return;
1055
283bc9f3 1056 write_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1057 list_del(&walk->walk.all);
283bc9f3 1058 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1059}
1060EXPORT_SYMBOL(xfrm_policy_walk_done);
1061
134b0fc5
JM		 1062/*
1063 * Find policy to apply to this flow.
1064 *
1065 * Returns 0 if policy found, else an -errno.
1066 */
f299d557
DM		 1067static int xfrm_policy_match(const struct xfrm_policy *pol,
1068 const struct flowi *fl,
2518c7c2 1069 u8 type, u16 family, int dir)
1da177e4 1070{
f299d557 1071 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1072 int ret = -ESRCH;
1073 bool match;
1da177e4 1074
2518c7c2 1075 if (pol->family != family ||
1d28f42c 1076 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1077 pol->type != type)
134b0fc5 1078 return ret;
1da177e4 1079
2518c7c2 1080 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1081 if (match)
1d28f42c 1082 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1083 dir);
2518c7c2 1084
134b0fc5 1085 return ret;
2518c7c2 1086}
1da177e4 1087
52479b62 1088static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1089 const struct flowi *fl,
2518c7c2
DM		 1090 u16 family, u8 dir)
1091{
134b0fc5 1092 int err;
2518c7c2 1093 struct xfrm_policy *pol, *ret;
0b597e7e 1094 const xfrm_address_t *daddr, *saddr;
2518c7c2 1095 struct hlist_head *chain;
acba48e1 1096 u32 priority = ~0U;
df71837d 1097
2518c7c2
DM		 1098 daddr = xfrm_flowi_daddr(fl, family);
1099 saddr = xfrm_flowi_saddr(fl, family);
1100 if (unlikely(!daddr || !saddr))
1101 return NULL;
1102
283bc9f3 1103 read_lock_bh(&net->xfrm.xfrm_policy_lock);
52479b62 1104 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 1105 ret = NULL;
b67bfe0d 1106 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1107 err = xfrm_policy_match(pol, fl, type, family, dir);
1108 if (err) {
1109 if (err == -ESRCH)
1110 continue;
1111 else {
1112 ret = ERR_PTR(err);
1113 goto fail;
1114 }
1115 } else {
2518c7c2 1116 ret = pol;
acba48e1 1117 priority = ret->priority;
2518c7c2
DM		 1118 break;
1119 }
1120 }
52479b62 1121 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1122 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1123 err = xfrm_policy_match(pol, fl, type, family, dir);
1124 if (err) {
1125 if (err == -ESRCH)
1126 continue;
1127 else {
1128 ret = ERR_PTR(err);
1129 goto fail;
1130 }
1131 } else if (pol->priority < priority) {
acba48e1
DM		 1132 ret = pol;
1133 break;
1da177e4
LT		 1134 }
1135 }
acba48e1
DM		 1136 if (ret)
1137 xfrm_pol_hold(ret);
134b0fc5 1138fail:
283bc9f3 1139 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
4e81bb83 1140
2518c7c2 1141 return ret;
4e81bb83
MN		 1142}
1143
80c802f3 1144static struct xfrm_policy *
73ff93cd 1145__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1146{
1147#ifdef CONFIG_XFRM_SUB_POLICY
1148 struct xfrm_policy *pol;
1149
1150 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1151 if (pol != NULL)
1152 return pol;
1153#endif
1154 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1155}
1156
b5fb82c4
BZ		 1157static int flow_to_policy_dir(int dir)
1158{
1159 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1160 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1161 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1162 return dir;
1163
1164 switch (dir) {
1165 default:
1166 case FLOW_DIR_IN:
1167 return XFRM_POLICY_IN;
1168 case FLOW_DIR_OUT:
1169 return XFRM_POLICY_OUT;
1170 case FLOW_DIR_FWD:
1171 return XFRM_POLICY_FWD;
1172 }
1173}
1174
fe1a5f03 1175static struct flow_cache_object *
dee9f4bc 1176xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1177 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1178{
1179 struct xfrm_policy *pol;
fe1a5f03
TT		 1180
1181 if (old_obj)
1182 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1183
b5fb82c4 1184 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1185 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1186 return ERR_CAST(pol);
fe1a5f03 1187
fe1a5f03
TT		 1188 /* Resolver returns two references:
1189 * one for cache and one for caller of flow_cache_lookup() */
1190 xfrm_pol_hold(pol);
1191
1192 return &pol->flo;
1da177e4
LT		 1193}
1194
df71837d
TJ		 1195static inline int policy_to_flow_dir(int dir)
1196{
1197 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1198 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1199 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1200 return dir;
1201 switch (dir) {
1202 default:
1203 case XFRM_POLICY_IN:
1204 return FLOW_DIR_IN;
1205 case XFRM_POLICY_OUT:
1206 return FLOW_DIR_OUT;
1207 case XFRM_POLICY_FWD:
1208 return FLOW_DIR_FWD;
3ff50b79 1209 }
df71837d
TJ		 1210}
1211
dee9f4bc
DM		 1212static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1213 const struct flowi *fl)
1da177e4
LT		 1214{
1215 struct xfrm_policy *pol;
283bc9f3 1216 struct net *net = sock_net(sk);
1da177e4 1217
283bc9f3 1218 read_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1219 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1220 bool match = xfrm_selector_match(&pol->selector, fl,
1221 sk->sk_family);
a716c119 1222 int err = 0;
df71837d 1223
3bccfbc7 1224 if (match) {
34f8d884
JHS		 1225 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1226 pol = NULL;
1227 goto out;
1228 }
03e1ad7b 1229 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1230 fl->flowi_secid,
03e1ad7b 1231 policy_to_flow_dir(dir));
3bccfbc7
VY		 1232 if (!err)
1233 xfrm_pol_hold(pol);
1234 else if (err == -ESRCH)
1235 pol = NULL;
1236 else
1237 pol = ERR_PTR(err);
1238 } else
1da177e4
LT		 1239 pol = NULL;
1240 }
34f8d884 1241out:
283bc9f3 1242 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1243 return pol;
1244}
1245
1246static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1247{
98806f75 1248 struct net *net = xp_net(pol);
4e81bb83 1249
98806f75 1250 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1251 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1252 xfrm_pol_hold(pol);
1253}
1254
1255static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1256 int dir)
1257{
98806f75
AD		 1258 struct net *net = xp_net(pol);
1259
53c2e285 1260 if (list_empty(&pol->walk.all))
2518c7c2 1261 return NULL;
1da177e4 1262
53c2e285
HX		 1263 /* Socket policies are not hashed. */
1264 if (!hlist_unhashed(&pol->bydst)) {
1265 hlist_del(&pol->bydst);
1266 hlist_del(&pol->byidx);
1267 }
1268
1269 list_del_init(&pol->walk.all);
98806f75 1270 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1271
1272 return pol;
1da177e4
LT		 1273}
1274
53c2e285
HX		 1275static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1276{
1277 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1278}
1279
1280static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1281{
1282 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1283}
1284
4666faab 1285int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1286{
283bc9f3
FD		 1287 struct net *net = xp_net(pol);
1288
1289 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1290 pol = __xfrm_policy_unlink(pol, dir);
283bc9f3 1291 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1292 if (pol) {
1da177e4 1293 xfrm_policy_kill(pol);
4666faab 1294 return 0;
1da177e4 1295 }
4666faab 1296 return -ENOENT;
1da177e4 1297}
a70fcb0b 1298EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1299
1300int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1301{
1121994c 1302 struct net *net = xp_net(pol);
1da177e4
LT		 1303 struct xfrm_policy *old_pol;
1304
4e81bb83
MN		 1305#ifdef CONFIG_XFRM_SUB_POLICY
1306 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1307 return -EINVAL;
1308#endif
1309
283bc9f3 1310 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1311 old_pol = sk->sk_policy[dir];
1312 sk->sk_policy[dir] = pol;
1313 if (pol) {
9d729f72 1314 pol->curlft.add_time = get_seconds();
e682adf0 1315 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1316 xfrm_sk_policy_link(pol, dir);
1da177e4 1317 }
a0073fe1
SK		 1318 if (old_pol) {
1319 if (pol)
1320 xfrm_policy_requeue(old_pol, pol);
1321
ea2dea9d
TT		 1322 /* Unlinking succeeds always. This is the only function
1323 * allowed to delete or replace socket policy.
1324 */
53c2e285 1325 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1326 }
283bc9f3 1327 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1328
1329 if (old_pol) {
1330 xfrm_policy_kill(old_pol);
1331 }
1332 return 0;
1333}
1334
d3e40a9f 1335static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1336{
0331b1f3 1337 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1338 struct net *net = xp_net(old);
1da177e4
LT		 1339
1340 if (newp) {
1341 newp->selector = old->selector;
03e1ad7b
PM		 1342 if (security_xfrm_policy_clone(old->security,
1343 &newp->security)) {
df71837d
TJ		 1344 kfree(newp);
1345 return NULL; /* ENOMEM */
1346 }
1da177e4
LT		 1347 newp->lft = old->lft;
1348 newp->curlft = old->curlft;
fb977e2c 1349 newp->mark = old->mark;
1da177e4
LT		 1350 newp->action = old->action;
1351 newp->flags = old->flags;
1352 newp->xfrm_nr = old->xfrm_nr;
1353 newp->index = old->index;
4e81bb83 1354 newp->type = old->type;
1da177e4
LT		 1355 memcpy(newp->xfrm_vec, old->xfrm_vec,
1356 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
283bc9f3 1357 write_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1358 xfrm_sk_policy_link(newp, dir);
283bc9f3 1359 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1360 xfrm_pol_put(newp);
1361 }
1362 return newp;
1363}
1364
1365int __xfrm_sk_clone_policy(struct sock *sk)
1366{
1367 struct xfrm_policy *p0 = sk->sk_policy[0],
1368 *p1 = sk->sk_policy[1];
1369
1370 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1371 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1372 return -ENOMEM;
1373 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1374 return -ENOMEM;
1375 return 0;
1376}
1377
a1e59abf 1378static int
fbda33b2 1379xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1380 unsigned short family)
1381{
1382 int err;
1383 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1384
1385 if (unlikely(afinfo == NULL))
1386 return -EINVAL;
fbda33b2 1387 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1388 xfrm_policy_put_afinfo(afinfo);
1389 return err;
1390}
1391
1da177e4
LT		 1392/* Resolve list of templates for the flow, given policy. */
1393
1394static int
a6c2e611
DM		 1395xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1396 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1397{
fbda33b2 1398 struct net *net = xp_net(policy);
1da177e4
LT		 1399 int nx;
1400 int i, error;
1401 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1402 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1403 xfrm_address_t tmp;
1da177e4 1404
9b7a787d 1405 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1406 struct xfrm_state *x;
1407 xfrm_address_t *remote = daddr;
1408 xfrm_address_t *local = saddr;
1409 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1410
48b8d783
JK		 1411 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1412 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1413 remote = &tmpl->id.daddr;
1414 local = &tmpl->saddr;
8444cf71
TE		 1415 if (xfrm_addr_any(local, tmpl->encap_family)) {
1416 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1417 if (error)
1418 goto fail;
1419 local = &tmp;
1420 }
1da177e4
LT		 1421 }
1422
1423 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1424
1425 if (x && x->km.state == XFRM_STATE_VALID) {
1426 xfrm[nx++] = x;
1427 daddr = remote;
1428 saddr = local;
1429 continue;
1430 }
1431 if (x) {
1432 error = (x->km.state == XFRM_STATE_ERROR ?
1433 -EINVAL : -EAGAIN);
1434 xfrm_state_put(x);
42054569 1435 } else if (error == -ESRCH) {
a4322266 1436 error = -EAGAIN;
42054569 1437 }
1da177e4
LT		 1438
1439 if (!tmpl->optional)
1440 goto fail;
1441 }
1442 return nx;
1443
1444fail:
9b7a787d 1445 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1446 xfrm_state_put(xfrm[nx]);
1447 return error;
1448}
1449
4e81bb83 1450static int
a6c2e611
DM		 1451xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1452 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1453{
41a49cc3
MN		 1454 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1455 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1456 int cnx = 0;
1457 int error;
1458 int ret;
1459 int i;
1460
1461 for (i = 0; i < npols; i++) {
1462 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1463 error = -ENOBUFS;
1464 goto fail;
1465 }
41a49cc3
MN		 1466
1467 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1468 if (ret < 0) {
1469 error = ret;
1470 goto fail;
1471 } else
1472 cnx += ret;
1473 }
1474
41a49cc3
MN		 1475 /* found states are sorted for outbound processing */
1476 if (npols > 1)
1477 xfrm_state_sort(xfrm, tpp, cnx, family);
1478
4e81bb83
MN		 1479 return cnx;
1480
1481 fail:
9b7a787d 1482 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1483 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1484 return error;
1485
1486}
1487
1da177e4
LT		 1488/* Check that the bundle accepts the flow and its components are
1489 * still valid.
1490 */
1491
05d84025 1492static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1493{
1494 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1495 int tos;
1da177e4 1496
25ee3286
HX		 1497 if (!afinfo)
1498 return -EINVAL;
1499
1500 tos = afinfo->get_tos(fl);
1501
1502 xfrm_policy_put_afinfo(afinfo);
1503
1504 return tos;
1505}
1506
80c802f3
TT		 1507static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1508{
1509 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1510 struct dst_entry *dst = &xdst->u.dst;
1511
1512 if (xdst->route == NULL) {
1513 /* Dummy bundle - if it has xfrms we were not
1514 * able to build bundle as template resolution failed.
1515 * It means we need to try again resolving. */
1516 if (xdst->num_xfrms > 0)
1517 return NULL;
a0073fe1
SK		 1518 } else if (dst->flags & DST_XFRM_QUEUE) {
1519 return NULL;
80c802f3
TT		 1520 } else {
1521 /* Real bundle */
1522 if (stale_bundle(dst))
1523 return NULL;
1524 }
1525
1526 dst_hold(dst);
1527 return flo;
1528}
1529
1530static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1531{
1532 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1533 struct dst_entry *dst = &xdst->u.dst;
1534
1535 if (!xdst->route)
1536 return 0;
1537 if (stale_bundle(dst))
1538 return 0;
1539
1540 return 1;
1541}
1542
1543static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1544{
1545 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1546 struct dst_entry *dst = &xdst->u.dst;
1547
1548 dst_free(dst);
1549}
1550
1551static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1552 .get = xfrm_bundle_flo_get,
1553 .check = xfrm_bundle_flo_check,
1554 .delete = xfrm_bundle_flo_delete,
1555};
1556
d7c7544c 1557static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1558{
1da177e4 1559 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1560 struct dst_ops *dst_ops;
25ee3286
HX		 1561 struct xfrm_dst *xdst;
1562
1563 if (!afinfo)
1564 return ERR_PTR(-EINVAL);
1565
d7c7544c
AD		 1566 switch (family) {
1567 case AF_INET:
1568 dst_ops = &net->xfrm.xfrm4_dst_ops;
1569 break;
dfd56b8b 1570#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1571 case AF_INET6:
1572 dst_ops = &net->xfrm.xfrm6_dst_ops;
1573 break;
1574#endif
1575 default:
1576 BUG();
1577 }
f5b0a874 1578 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1579
d4cae562 1580 if (likely(xdst)) {
141e369d
SK		 1581 struct dst_entry *dst = &xdst->u.dst;
1582
1583 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1584 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1585 if (afinfo->init_dst)
1586 afinfo->init_dst(net, xdst);
d4cae562 1587 } else
0b150932 1588 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1589
d4cae562
MB		 1590 xfrm_policy_put_afinfo(afinfo);
1591
25ee3286
HX		 1592 return xdst;
1593}
1594
a1b05140
MN		 1595static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1596 int nfheader_len)
1597{
1598 struct xfrm_policy_afinfo *afinfo =
1599 xfrm_policy_get_afinfo(dst->ops->family);
1600 int err;
1601
1602 if (!afinfo)
1603 return -EINVAL;
1604
1605 err = afinfo->init_path(path, dst, nfheader_len);
1606
1607 xfrm_policy_put_afinfo(afinfo);
1608
1609 return err;
1610}
1611
87c1e12b 1612static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1613 const struct flowi *fl)
25ee3286
HX		 1614{
1615 struct xfrm_policy_afinfo *afinfo =
1616 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1617 int err;
1618
1619 if (!afinfo)
1da177e4 1620 return -EINVAL;
25ee3286 1621
87c1e12b 1622 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1623
1da177e4 1624 xfrm_policy_put_afinfo(afinfo);
25ee3286 1625
1da177e4
LT		 1626 return err;
1627}
1628
80c802f3 1629
25ee3286
HX		 1630/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1631 * all the metrics... Shortly, bundle a bundle.
1632 */
1633
1634static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1635 struct xfrm_state **xfrm, int nx,
98313ada 1636 const struct flowi *fl,
25ee3286
HX		 1637 struct dst_entry *dst)
1638{
d7c7544c 1639 struct net *net = xp_net(policy);
25ee3286
HX		 1640 unsigned long now = jiffies;
1641 struct net_device *dev;
43a4dea4 1642 struct xfrm_mode *inner_mode;
25ee3286
HX		 1643 struct dst_entry *dst_prev = NULL;
1644 struct dst_entry *dst0 = NULL;
1645 int i = 0;
1646 int err;
1647 int header_len = 0;
a1b05140 1648 int nfheader_len = 0;
25ee3286
HX		 1649 int trailer_len = 0;
1650 int tos;
1651 int family = policy->selector.family;
9bb182a7
YH		 1652 xfrm_address_t saddr, daddr;
1653
1654 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1655
1656 tos = xfrm_get_tos(fl, family);
1657 err = tos;
1658 if (tos < 0)
1659 goto put_states;
1660
1661 dst_hold(dst);
1662
1663 for (; i < nx; i++) {
d7c7544c 1664 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1665 struct dst_entry *dst1 = &xdst->u.dst;
1666
1667 err = PTR_ERR(xdst);
1668 if (IS_ERR(xdst)) {
1669 dst_release(dst);
1670 goto put_states;
1671 }
1672
43a4dea4
SK		 1673 if (xfrm[i]->sel.family == AF_UNSPEC) {
1674 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1675 xfrm_af2proto(family));
1676 if (!inner_mode) {
1677 err = -EAFNOSUPPORT;
1678 dst_release(dst);
1679 goto put_states;
1680 }
1681 } else
1682 inner_mode = xfrm[i]->inner_mode;
1683
25ee3286
HX		 1684 if (!dst_prev)
1685 dst0 = dst1;
1686 else {
1687 dst_prev->child = dst_clone(dst1);
1688 dst1->flags |= DST_NOHASH;
1689 }
1690
1691 xdst->route = dst;
defb3519 1692 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1693
1694 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1695 family = xfrm[i]->props.family;
9bb182a7
YH		 1696 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1697 family);
25ee3286
HX		 1698 err = PTR_ERR(dst);
1699 if (IS_ERR(dst))
1700 goto put_states;
1701 } else
1702 dst_hold(dst);
1703
1704 dst1->xfrm = xfrm[i];
80c802f3 1705 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1706
f5b0a874 1707 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1708 dst1->flags |= DST_HOST;
1709 dst1->lastuse = now;
1710
1711 dst1->input = dst_discard;
43a4dea4 1712 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1713
1714 dst1->next = dst_prev;
1715 dst_prev = dst1;
1716
1717 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1718 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1719 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1720 trailer_len += xfrm[i]->props.trailer_len;
1721 }
1722
1723 dst_prev->child = dst;
1724 dst0->path = dst;
1725
1726 err = -ENODEV;
1727 dev = dst->dev;
1728 if (!dev)
1729 goto free_dst;
1730
a1b05140 1731 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1732 xfrm_init_pmtu(dst_prev);
1733
1734 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1735 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1736
87c1e12b 1737 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1738 if (err)
1739 goto free_dst;
1740
1741 dst_prev->header_len = header_len;
1742 dst_prev->trailer_len = trailer_len;
1743 header_len -= xdst->u.dst.xfrm->props.header_len;
1744 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1745 }
1746
1747out:
1748 return dst0;
1749
1750put_states:
1751 for (; i < nx; i++)
1752 xfrm_state_put(xfrm[i]);
1753free_dst:
1754 if (dst0)
1755 dst_free(dst0);
1756 dst0 = ERR_PTR(err);
1757 goto out;
1758}
1759
da7c224b 1760#ifdef CONFIG_XFRM_SUB_POLICY
be7928d2 1761static int xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1762{
1763 if (!*target) {
1764 *target = kmalloc(size, GFP_ATOMIC);
1765 if (!*target)
1766 return -ENOMEM;
1767 }
be7928d2 1768
157bfc25
MN		 1769 memcpy(*target, src, size);
1770 return 0;
1771}
da7c224b 1772#endif
157bfc25 1773
be7928d2
DB		 1774static int xfrm_dst_update_parent(struct dst_entry *dst,
1775 const struct xfrm_selector *sel)
157bfc25
MN		 1776{
1777#ifdef CONFIG_XFRM_SUB_POLICY
1778 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1779 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1780 sel, sizeof(*sel));
1781#else
1782 return 0;
1783#endif
1784}
1785
be7928d2
DB		 1786static int xfrm_dst_update_origin(struct dst_entry *dst,
1787 const struct flowi *fl)
157bfc25
MN		 1788{
1789#ifdef CONFIG_XFRM_SUB_POLICY
1790 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1791 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1792#else
1793 return 0;
1794#endif
1795}
1da177e4 1796
73ff93cd 1797static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1798 struct xfrm_policy **pols,
1799 int *num_pols, int *num_xfrms)
1800{
1801 int i;
1802
1803 if (*num_pols == 0 || !pols[0]) {
1804 *num_pols = 0;
1805 *num_xfrms = 0;
1806 return 0;
1807 }
1808 if (IS_ERR(pols[0]))
1809 return PTR_ERR(pols[0]);
1810
1811 *num_xfrms = pols[0]->xfrm_nr;
1812
1813#ifdef CONFIG_XFRM_SUB_POLICY
1814 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1815 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1816 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1817 XFRM_POLICY_TYPE_MAIN,
1818 fl, family,
1819 XFRM_POLICY_OUT);
1820 if (pols[1]) {
1821 if (IS_ERR(pols[1])) {
1822 xfrm_pols_put(pols, *num_pols);
1823 return PTR_ERR(pols[1]);
1824 }
02d0892f 1825 (*num_pols)++;
80c802f3
TT		 1826 (*num_xfrms) += pols[1]->xfrm_nr;
1827 }
1828 }
1829#endif
1830 for (i = 0; i < *num_pols; i++) {
1831 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1832 *num_xfrms = -1;
1833 break;
1834 }
1835 }
1836
1837 return 0;
1838
1839}
1840
1841static struct xfrm_dst *
1842xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1843 const struct flowi *fl, u16 family,
80c802f3
TT		 1844 struct dst_entry *dst_orig)
1845{
1846 struct net *net = xp_net(pols[0]);
1847 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1848 struct dst_entry *dst;
1849 struct xfrm_dst *xdst;
1850 int err;
1851
1852 /* Try to instantiate a bundle */
1853 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1854 if (err <= 0) {
1855 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1856 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1857 return ERR_PTR(err);
1858 }
1859
1860 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1861 if (IS_ERR(dst)) {
1862 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1863 return ERR_CAST(dst);
1864 }
1865
1866 xdst = (struct xfrm_dst *)dst;
1867 xdst->num_xfrms = err;
1868 if (num_pols > 1)
1869 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1870 else
1871 err = xfrm_dst_update_origin(dst, fl);
1872 if (unlikely(err)) {
1873 dst_free(dst);
1874 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1875 return ERR_PTR(err);
1876 }
1877
1878 xdst->num_pols = num_pols;
3e94c2dc 1879 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1880 xdst->policy_genid = atomic_read(&pols[0]->genid);
1881
1882 return xdst;
1883}
1884
a0073fe1
SK		 1885static void xfrm_policy_queue_process(unsigned long arg)
1886{
a0073fe1
SK		 1887 struct sk_buff *skb;
1888 struct sock *sk;
1889 struct dst_entry *dst;
a0073fe1
SK		 1890 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1891 struct xfrm_policy_queue *pq = &pol->polq;
1892 struct flowi fl;
1893 struct sk_buff_head list;
1894
1895 spin_lock(&pq->hold_queue.lock);
1896 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1897 if (!skb) {
1898 spin_unlock(&pq->hold_queue.lock);
1899 goto out;
1900 }
a0073fe1
SK		 1901 dst = skb_dst(skb);
1902 sk = skb->sk;
1903 xfrm_decode_session(skb, &fl, dst->ops->family);
1904 spin_unlock(&pq->hold_queue.lock);
1905
1906 dst_hold(dst->path);
1907 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1908 sk, 0);
1909 if (IS_ERR(dst))
1910 goto purge_queue;
1911
1912 if (dst->flags & DST_XFRM_QUEUE) {
1913 dst_release(dst);
1914
1915 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1916 goto purge_queue;
1917
1918 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1919 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1920 xfrm_pol_hold(pol);
1921 goto out;
a0073fe1
SK		 1922 }
1923
1924 dst_release(dst);
1925
1926 __skb_queue_head_init(&list);
1927
1928 spin_lock(&pq->hold_queue.lock);
1929 pq->timeout = 0;
1930 skb_queue_splice_init(&pq->hold_queue, &list);
1931 spin_unlock(&pq->hold_queue.lock);
1932
1933 while (!skb_queue_empty(&list)) {
1934 skb = __skb_dequeue(&list);
1935
1936 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1937 dst_hold(skb_dst(skb)->path);
1938 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1939 &fl, skb->sk, 0);
1940 if (IS_ERR(dst)) {
a0073fe1
SK		 1941 kfree_skb(skb);
1942 continue;
1943 }
1944
1945 nf_reset(skb);
1946 skb_dst_drop(skb);
1947 skb_dst_set(skb, dst);
1948
5c1e9f2c 1949 dst_output(skb);
a0073fe1
SK		 1950 }
1951
e7d8f6cb
SK		 1952out:
1953 xfrm_pol_put(pol);
a0073fe1
SK		 1954 return;
1955
1956purge_queue:
1957 pq->timeout = 0;
1958 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1959 xfrm_pol_put(pol);
a0073fe1
SK		 1960}
1961
aad88724 1962static int xdst_queue_output(struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1963{
1964 unsigned long sched_next;
1965 struct dst_entry *dst = skb_dst(skb);
1966 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1967 struct xfrm_policy *pol = xdst->pols[0];
1968 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1969
39bb5e62 1970 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1971 kfree_skb(skb);
1972 return 0;
1973 }
a0073fe1
SK		 1974
1975 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1976 kfree_skb(skb);
1977 return -EAGAIN;
1978 }
1979
1980 skb_dst_force(skb);
a0073fe1
SK		 1981
1982 spin_lock_bh(&pq->hold_queue.lock);
1983
1984 if (!pq->timeout)
1985 pq->timeout = XFRM_QUEUE_TMO_MIN;
1986
1987 sched_next = jiffies + pq->timeout;
1988
1989 if (del_timer(&pq->hold_timer)) {
1990 if (time_before(pq->hold_timer.expires, sched_next))
1991 sched_next = pq->hold_timer.expires;
e7d8f6cb 1992 xfrm_pol_put(pol);
a0073fe1
SK		 1993 }
1994
1995 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1996 if (!mod_timer(&pq->hold_timer, sched_next))
1997 xfrm_pol_hold(pol);
a0073fe1
SK		 1998
1999 spin_unlock_bh(&pq->hold_queue.lock);
2000
2001 return 0;
2002}
2003
2004static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 2005 struct xfrm_flo *xflo,
a0073fe1
SK		 2006 const struct flowi *fl,
2007 int num_xfrms,
2008 u16 family)
2009{
2010 int err;
2011 struct net_device *dev;
b8c203b2 2012 struct dst_entry *dst;
a0073fe1
SK		 2013 struct dst_entry *dst1;
2014 struct xfrm_dst *xdst;
2015
2016 xdst = xfrm_alloc_dst(net, family);
2017 if (IS_ERR(xdst))
2018 return xdst;
2019
b8c203b2
SK		 2020 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2021 net->xfrm.sysctl_larval_drop ||
2022 num_xfrms <= 0)
a0073fe1
SK		 2023 return xdst;
2024
b8c203b2 2025 dst = xflo->dst_orig;
a0073fe1
SK		 2026 dst1 = &xdst->u.dst;
2027 dst_hold(dst);
2028 xdst->route = dst;
2029
2030 dst_copy_metrics(dst1, dst);
2031
2032 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2033 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2034 dst1->lastuse = jiffies;
2035
2036 dst1->input = dst_discard;
2037 dst1->output = xdst_queue_output;
2038
2039 dst_hold(dst);
2040 dst1->child = dst;
2041 dst1->path = dst;
2042
2043 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2044
2045 err = -ENODEV;
2046 dev = dst->dev;
2047 if (!dev)
2048 goto free_dst;
2049
2050 err = xfrm_fill_dst(xdst, dev, fl);
2051 if (err)
2052 goto free_dst;
2053
2054out:
2055 return xdst;
2056
2057free_dst:
2058 dst_release(dst1);
2059 xdst = ERR_PTR(err);
2060 goto out;
2061}
2062
80c802f3 2063static struct flow_cache_object *
dee9f4bc 2064xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 2065 struct flow_cache_object *oldflo, void *ctx)
2066{
b8c203b2 2067 struct xfrm_flo *xflo = (struct xfrm_flo *)ctx;
80c802f3
TT		 2068 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2069 struct xfrm_dst *xdst, *new_xdst;
2070 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
2071
2072 /* Check if the policies from old bundle are usable */
2073 xdst = NULL;
2074 if (oldflo) {
2075 xdst = container_of(oldflo, struct xfrm_dst, flo);
2076 num_pols = xdst->num_pols;
2077 num_xfrms = xdst->num_xfrms;
2078 pol_dead = 0;
2079 for (i = 0; i < num_pols; i++) {
2080 pols[i] = xdst->pols[i];
2081 pol_dead |= pols[i]->walk.dead;
2082 }
2083 if (pol_dead) {
2084 dst_free(&xdst->u.dst);
2085 xdst = NULL;
2086 num_pols = 0;
2087 num_xfrms = 0;
2088 oldflo = NULL;
2089 }
2090 }
2091
2092 /* Resolve policies to use if we couldn't get them from
2093 * previous cache entry */
2094 if (xdst == NULL) {
2095 num_pols = 1;
b5fb82c4
BZ		 2096 pols[0] = __xfrm_policy_lookup(net, fl, family,
2097 flow_to_policy_dir(dir));
80c802f3
TT		 2098 err = xfrm_expand_policies(fl, family, pols,
2099 &num_pols, &num_xfrms);
2100 if (err < 0)
2101 goto inc_error;
2102 if (num_pols == 0)
2103 return NULL;
2104 if (num_xfrms <= 0)
2105 goto make_dummy_bundle;
2106 }
2107
b8c203b2
SK		 2108 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
2109 xflo->dst_orig);
80c802f3
TT		 2110 if (IS_ERR(new_xdst)) {
2111 err = PTR_ERR(new_xdst);
2112 if (err != -EAGAIN)
2113 goto error;
2114 if (oldflo == NULL)
2115 goto make_dummy_bundle;
2116 dst_hold(&xdst->u.dst);
2117 return oldflo;
d809ec89
TT		 2118 } else if (new_xdst == NULL) {
2119 num_xfrms = 0;
2120 if (oldflo == NULL)
2121 goto make_dummy_bundle;
2122 xdst->num_xfrms = 0;
2123 dst_hold(&xdst->u.dst);
2124 return oldflo;
80c802f3
TT		 2125 }
2126
2127 /* Kill the previous bundle */
2128 if (xdst) {
2129 /* The policies were stolen for newly generated bundle */
2130 xdst->num_pols = 0;
2131 dst_free(&xdst->u.dst);
2132 }
2133
2134 /* Flow cache does not have reference, it dst_free()'s,
2135 * but we do need to return one reference for original caller */
2136 dst_hold(&new_xdst->u.dst);
2137 return &new_xdst->flo;
2138
2139make_dummy_bundle:
2140 /* We found policies, but there's no bundles to instantiate:
2141 * either because the policy blocks, has no transformations or
2142 * we could not build template (no xfrm_states).*/
b8c203b2 2143 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2144 if (IS_ERR(xdst)) {
2145 xfrm_pols_put(pols, num_pols);
2146 return ERR_CAST(xdst);
2147 }
2148 xdst->num_pols = num_pols;
2149 xdst->num_xfrms = num_xfrms;
3e94c2dc 2150 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2151
2152 dst_hold(&xdst->u.dst);
2153 return &xdst->flo;
2154
2155inc_error:
2156 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2157error:
2158 if (xdst != NULL)
2159 dst_free(&xdst->u.dst);
2160 else
2161 xfrm_pols_put(pols, num_pols);
2162 return ERR_PTR(err);
2163}
1da177e4 2164
2774c131
DM		 2165static struct dst_entry *make_blackhole(struct net *net, u16 family,
2166 struct dst_entry *dst_orig)
2167{
2168 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2169 struct dst_entry *ret;
2170
2171 if (!afinfo) {
2172 dst_release(dst_orig);
433a1954 2173 return ERR_PTR(-EINVAL);
2774c131
DM		 2174 } else {
2175 ret = afinfo->blackhole_route(net, dst_orig);
2176 }
2177 xfrm_policy_put_afinfo(afinfo);
2178
2179 return ret;
2180}
2181
1da177e4
LT		 2182/* Main function: finds/creates a bundle for given flow.
2183 *
2184 * At the moment we eat a raw IP route. Mostly to speed up lookups
2185 * on interfaces with disabled IPsec.
2186 */
452edd59
DM		 2187struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2188 const struct flowi *fl,
2189 struct sock *sk, int flags)
1da177e4 2190{
4e81bb83 2191 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2192 struct flow_cache_object *flo;
2193 struct xfrm_dst *xdst;
452edd59 2194 struct dst_entry *dst, *route;
80c802f3 2195 u16 family = dst_orig->ops->family;
df71837d 2196 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2197 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2198
80c802f3
TT		 2199 dst = NULL;
2200 xdst = NULL;
2201 route = NULL;
4e81bb83 2202
f7944fb1 2203 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2204 num_pols = 1;
2205 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2206 err = xfrm_expand_policies(fl, family, pols,
2207 &num_pols, &num_xfrms);
2208 if (err < 0)
75b8c133 2209 goto dropdst;
80c802f3
TT		 2210
2211 if (num_pols) {
2212 if (num_xfrms <= 0) {
2213 drop_pols = num_pols;
2214 goto no_transform;
2215 }
2216
2217 xdst = xfrm_resolve_and_create_bundle(
2218 pols, num_pols, fl,
2219 family, dst_orig);
2220 if (IS_ERR(xdst)) {
2221 xfrm_pols_put(pols, num_pols);
2222 err = PTR_ERR(xdst);
2223 goto dropdst;
d809ec89
TT		 2224 } else if (xdst == NULL) {
2225 num_xfrms = 0;
2226 drop_pols = num_pols;
2227 goto no_transform;
80c802f3
TT		 2228 }
2229
b7eea454
SK		 2230 dst_hold(&xdst->u.dst);
2231 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2232 route = xdst->route;
0aa64774 2233 }
3bccfbc7 2234 }
1da177e4 2235
80c802f3 2236 if (xdst == NULL) {
b8c203b2
SK		 2237 struct xfrm_flo xflo;
2238
2239 xflo.dst_orig = dst_orig;
2240 xflo.flags = flags;
2241
1da177e4 2242 /* To accelerate a bit... */
2518c7c2 2243 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2244 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2245 goto nopol;
1da177e4 2246
80c802f3 2247 flo = flow_cache_lookup(net, fl, family, dir,
b8c203b2 2248 xfrm_bundle_lookup, &xflo);
80c802f3
TT		 2249 if (flo == NULL)
2250 goto nopol;
fe1a5f03 2251 if (IS_ERR(flo)) {
80c802f3 2252 err = PTR_ERR(flo);
75b8c133 2253 goto dropdst;
d66e37a9 2254 }
80c802f3
TT		 2255 xdst = container_of(flo, struct xfrm_dst, flo);
2256
2257 num_pols = xdst->num_pols;
2258 num_xfrms = xdst->num_xfrms;
3e94c2dc 2259 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2260 route = xdst->route;
2261 }
2262
2263 dst = &xdst->u.dst;
2264 if (route == NULL && num_xfrms > 0) {
2265 /* The only case when xfrm_bundle_lookup() returns a
2266 * bundle with null route, is when the template could
2267 * not be resolved. It means policies are there, but
2268 * bundle could not be created, since we don't yet
2269 * have the xfrm_state's. We need to wait for KM to
2270 * negotiate new SA's or bail out with error.*/
2271 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2272 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2273 err = -EREMOTE;
2274 goto error;
80c802f3 2275 }
80c802f3 2276
5b8ef341 2277 err = -EAGAIN;
80c802f3
TT		 2278
2279 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2280 goto error;
1da177e4
LT		 2281 }
2282
80c802f3
TT		 2283no_transform:
2284 if (num_pols == 0)
8b7817f3 2285 goto nopol;
1da177e4 2286
80c802f3
TT		 2287 if ((flags & XFRM_LOOKUP_ICMP) &&
2288 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2289 err = -ENOENT;
8b7817f3 2290 goto error;
80c802f3 2291 }
8b7817f3 2292
80c802f3
TT		 2293 for (i = 0; i < num_pols; i++)
2294 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2295
80c802f3 2296 if (num_xfrms < 0) {
1da177e4 2297 /* Prohibit the flow */
59c9940e 2298 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2299 err = -EPERM;
2300 goto error;
80c802f3
TT		 2301 } else if (num_xfrms > 0) {
2302 /* Flow transformed */
80c802f3
TT		 2303 dst_release(dst_orig);
2304 } else {
2305 /* Flow passes untransformed */
2306 dst_release(dst);
452edd59 2307 dst = dst_orig;
1da177e4 2308 }
80c802f3
TT		 2309ok:
2310 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2311 if (dst && dst->xfrm &&
2312 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2313 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2314 return dst;
1da177e4 2315
80c802f3 2316nopol:
452edd59
DM		 2317 if (!(flags & XFRM_LOOKUP_ICMP)) {
2318 dst = dst_orig;
80c802f3 2319 goto ok;
452edd59 2320 }
80c802f3 2321 err = -ENOENT;
1da177e4 2322error:
80c802f3 2323 dst_release(dst);
75b8c133 2324dropdst:
ac37e251
W		 2325 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2326 dst_release(dst_orig);
80c802f3 2327 xfrm_pols_put(pols, drop_pols);
452edd59 2328 return ERR_PTR(err);
1da177e4
LT		 2329}
2330EXPORT_SYMBOL(xfrm_lookup);
2331
f92ee619
SK		 2332/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2333 * Otherwise we may send out blackholed packets.
2334 */
2335struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2336 const struct flowi *fl,
2337 struct sock *sk, int flags)
2338{
b8c203b2 2339 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2340 flags | XFRM_LOOKUP_QUEUE |
2341 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2342
2343 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2344 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2345
2346 return dst;
2347}
2348EXPORT_SYMBOL(xfrm_lookup_route);
2349
df0ba92a 2350static inline int
8f029de2 2351xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2352{
2353 struct xfrm_state *x;
df0ba92a
MN		 2354
2355 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2356 return 0;
2357 x = skb->sp->xvec[idx];
2358 if (!x->type->reject)
2359 return 0;
1ecafede 2360 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2361}
2362
1da177e4
LT		 2363/* When skb is transformed back to its "native" form, we have to
2364 * check policy restrictions. At the moment we make this in maximally
2365 * stupid way. Shame on me. :-) Of course, connected sockets must
2366 * have policy cached at them.
2367 */
2368
2369static inline int
7db454b9 2370xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2371 unsigned short family)
2372{
2373 if (xfrm_state_kern(x))
928ba416 2374 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2375 return x->id.proto == tmpl->id.proto &&
2376 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2377 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2378 x->props.mode == tmpl->mode &&
c5d18e98 2379 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2380 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2381 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2382 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2383}
2384
df0ba92a
MN		 2385/*
2386 * 0 or more than 0 is returned when validation is succeeded (either bypass
2387 * because of optional transport mode, or next index of the mathced secpath
2388 * state with the template.
2389 * -1 is returned when no matching template is found.
2390 * Otherwise "-2 - errored_index" is returned.
2391 */
1da177e4 2392static inline int
22cccb7e 2393xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2394 unsigned short family)
2395{
2396 int idx = start;
2397
2398 if (tmpl->optional) {
7e49e6de 2399 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2400 return start;
2401 } else
2402 start = -1;
2403 for (; idx < sp->len; idx++) {
dbe5b4aa 2404 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2405 return ++idx;
df0ba92a
MN		 2406 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2407 if (start == -1)
2408 start = -2-idx;
1da177e4 2409 break;
df0ba92a 2410 }
1da177e4
LT		 2411 }
2412 return start;
2413}
2414
d5422efe
HX		 2415int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2416 unsigned int family, int reverse)
1da177e4
LT		 2417{
2418 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2419 int err;
1da177e4
LT		 2420
2421 if (unlikely(afinfo == NULL))
2422 return -EAFNOSUPPORT;
2423
d5422efe 2424 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2425 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2426 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2427 return err;
1da177e4 2428}
d5422efe 2429EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2430
9a7386ec 2431static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2432{
2433 for (; k < sp->len; k++) {
df0ba92a 2434 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2435 *idxp = k;
1da177e4 2436 return 1;
df0ba92a 2437 }
1da177e4
LT		 2438 }
2439
2440 return 0;
2441}
2442
a716c119 2443int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2444 unsigned short family)
2445{
f6e1e25d 2446 struct net *net = dev_net(skb->dev);
1da177e4 2447 struct xfrm_policy *pol;
4e81bb83
MN		 2448 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2449 int npols = 0;
2450 int xfrm_nr;
2451 int pi;
d5422efe 2452 int reverse;
1da177e4 2453 struct flowi fl;
d5422efe 2454 u8 fl_dir;
df0ba92a 2455 int xerr_idx = -1;
1da177e4 2456
d5422efe
HX		 2457 reverse = dir & ~XFRM_POLICY_MASK;
2458 dir &= XFRM_POLICY_MASK;
2459 fl_dir = policy_to_flow_dir(dir);
2460
0aa64774 2461 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2462 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2463 return 0;
0aa64774
MN		 2464 }
2465
eb9c7ebe 2466 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2467
2468 /* First, check used SA against their selectors. */
2469 if (skb->sp) {
2470 int i;
2471
9b7a787d 2472 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2473 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2474 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2475 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2476 return 0;
0aa64774 2477 }
1da177e4
LT		 2478 }
2479 }
2480
2481 pol = NULL;
3bccfbc7 2482 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2483 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2484 if (IS_ERR(pol)) {
59c9940e 2485 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2486 return 0;
0aa64774 2487 }
3bccfbc7 2488 }
1da177e4 2489
fe1a5f03
TT		 2490 if (!pol) {
2491 struct flow_cache_object *flo;
2492
2493 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2494 xfrm_policy_lookup, NULL);
2495 if (IS_ERR_OR_NULL(flo))
2496 pol = ERR_CAST(flo);
2497 else
2498 pol = container_of(flo, struct xfrm_policy, flo);
2499 }
1da177e4 2500
0aa64774 2501 if (IS_ERR(pol)) {
59c9940e 2502 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2503 return 0;
0aa64774 2504 }
134b0fc5 2505
df0ba92a 2506 if (!pol) {
d1d9facf 2507 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2508 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2509 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2510 return 0;
2511 }
2512 return 1;
2513 }
1da177e4 2514
9d729f72 2515 pol->curlft.use_time = get_seconds();
1da177e4 2516
4e81bb83 2517 pols[0] = pol;
02d0892f 2518 npols++;
4e81bb83
MN		 2519#ifdef CONFIG_XFRM_SUB_POLICY
2520 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2521 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2522 &fl, family,
2523 XFRM_POLICY_IN);
2524 if (pols[1]) {
0aa64774 2525 if (IS_ERR(pols[1])) {
59c9940e 2526 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2527 return 0;
0aa64774 2528 }
9d729f72 2529 pols[1]->curlft.use_time = get_seconds();
02d0892f 2530 npols++;
4e81bb83
MN		 2531 }
2532 }
2533#endif
2534
1da177e4
LT		 2535 if (pol->action == XFRM_POLICY_ALLOW) {
2536 struct sec_path *sp;
2537 static struct sec_path dummy;
4e81bb83 2538 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2539 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2540 struct xfrm_tmpl **tpp = tp;
2541 int ti = 0;
1da177e4
LT		 2542 int i, k;
2543
2544 if ((sp = skb->sp) == NULL)
2545 sp = &dummy;
2546
4e81bb83
MN		 2547 for (pi = 0; pi < npols; pi++) {
2548 if (pols[pi] != pol &&
0aa64774 2549 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2550 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2551 goto reject;
0aa64774
MN		 2552 }
2553 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2554 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2555 goto reject_error;
0aa64774 2556 }
4e81bb83
MN		 2557 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2558 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2559 }
2560 xfrm_nr = ti;
41a49cc3 2561 if (npols > 1) {
283bc9f3 2562 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2563 tpp = stp;
2564 }
4e81bb83 2565
1da177e4
LT		 2566 /* For each tunnel xfrm, find the first matching tmpl.
2567 * For each tmpl before that, find corresponding xfrm.
2568 * Order is _important_. Later we will implement
2569 * some barriers, but at the moment barriers
2570 * are implied between each two transformations.
2571 */
4e81bb83
MN		 2572 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2573 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2574 if (k < 0) {
d1d9facf
JM		 2575 if (k < -1)
2576 /* "-2 - errored_index" returned */
2577 xerr_idx = -(2+k);
59c9940e 2578 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2579 goto reject;
df0ba92a 2580 }
1da177e4
LT		 2581 }
2582
0aa64774 2583 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2584 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2585 goto reject;
0aa64774 2586 }
1da177e4 2587
4e81bb83 2588 xfrm_pols_put(pols, npols);
1da177e4
LT		 2589 return 1;
2590 }
59c9940e 2591 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2592
2593reject:
df0ba92a 2594 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2595reject_error:
2596 xfrm_pols_put(pols, npols);
1da177e4
LT		 2597 return 0;
2598}
2599EXPORT_SYMBOL(__xfrm_policy_check);
2600
2601int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2602{
99a66657 2603 struct net *net = dev_net(skb->dev);
1da177e4 2604 struct flowi fl;
adf30907 2605 struct dst_entry *dst;
73137147 2606 int res = 1;
1da177e4 2607
0aa64774 2608 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2609 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2610 return 0;
0aa64774 2611 }
1da177e4 2612
fafeeb6c 2613 skb_dst_force(skb);
adf30907 2614
b8c203b2 2615 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2616 if (IS_ERR(dst)) {
73137147 2617 res = 0;
452edd59
DM		 2618 dst = NULL;
2619 }
adf30907
ED		 2620 skb_dst_set(skb, dst);
2621 return res;
1da177e4
LT		 2622}
2623EXPORT_SYMBOL(__xfrm_route_forward);
2624
d49c73c7
DM		 2625/* Optimize later using cookies and generation ids. */
2626
1da177e4
LT		 2627static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2628{
d49c73c7 2629 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2630 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2631 * get validated by dst_ops->check on every use. We do this
2632 * because when a normal route referenced by an XFRM dst is
2633 * obsoleted we do not go looking around for all parent
2634 * referencing XFRM dsts so that we can invalidate them. It
2635 * is just too much work. Instead we make the checks here on
2636 * every use. For example:
d49c73c7
DM		 2637 *
2638 * XFRM dst A --> IPv4 dst X
2639 *
2640 * X is the "xdst->route" of A (X is also the "dst->path" of A
2641 * in this example). If X is marked obsolete, "A" will not
2642 * notice. That's what we are validating here via the
2643 * stale_bundle() check.
2644 *
2645 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2646 * dst which causes it's ->obsolete field to be set to
2647 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2648 * this, we want to force a new route lookup.
399c180a 2649 */
d49c73c7
DM		 2650 if (dst->obsolete < 0 && !stale_bundle(dst))
2651 return dst;
2652
1da177e4
LT		 2653 return NULL;
2654}
2655
2656static int stale_bundle(struct dst_entry *dst)
2657{
12fdb4d3 2658 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2659}
2660
aabc9761 2661void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2662{
1da177e4 2663 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2664 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2665 dev_hold(dst->dev);
1da177e4
LT		 2666 dev_put(dev);
2667 }
2668}
aabc9761 2669EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2670
2671static void xfrm_link_failure(struct sk_buff *skb)
2672{
2673 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2674}
2675
2676static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2677{
2678 if (dst) {
2679 if (dst->obsolete) {
2680 dst_release(dst);
2681 dst = NULL;
2682 }
2683 }
2684 return dst;
2685}
2686
e4c17216 2687void xfrm_garbage_collect(struct net *net)
c0ed1c14 2688{
ca925cf1 2689 flow_cache_flush(net);
c0ed1c14 2690}
e4c17216 2691EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2692
2693static void xfrm_garbage_collect_deferred(struct net *net)
2694{
ca925cf1 2695 flow_cache_flush_deferred(net);
c0ed1c14
SK		 2696}
2697
25ee3286 2698static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2699{
2700 do {
2701 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2702 u32 pmtu, route_mtu_cached;
2703
2704 pmtu = dst_mtu(dst->child);
2705 xdst->child_mtu_cached = pmtu;
2706
2707 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2708
2709 route_mtu_cached = dst_mtu(xdst->route);
2710 xdst->route_mtu_cached = route_mtu_cached;
2711
2712 if (pmtu > route_mtu_cached)
2713 pmtu = route_mtu_cached;
2714
defb3519 2715 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2716 } while ((dst = dst->next));
2717}
2718
1da177e4
LT		 2719/* Check that the bundle accepts the flow and its components are
2720 * still valid.
2721 */
2722
12fdb4d3 2723static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2724{
2725 struct dst_entry *dst = &first->u.dst;
2726 struct xfrm_dst *last;
2727 u32 mtu;
2728
92d63dec 2729 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2730 (dst->dev && !netif_running(dst->dev)))
2731 return 0;
2732
a0073fe1
SK		 2733 if (dst->flags & DST_XFRM_QUEUE)
2734 return 1;
2735
1da177e4
LT		 2736 last = NULL;
2737
2738 do {
2739 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2740
1da177e4
LT		 2741 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2742 return 0;
80c802f3
TT		 2743 if (xdst->xfrm_genid != dst->xfrm->genid)
2744 return 0;
b1312c89
TT		 2745 if (xdst->num_pols > 0 &&
2746 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2747 return 0;
e53820de 2748
1da177e4
LT		 2749 mtu = dst_mtu(dst->child);
2750 if (xdst->child_mtu_cached != mtu) {
2751 last = xdst;
2752 xdst->child_mtu_cached = mtu;
2753 }
2754
92d63dec 2755 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2756 return 0;
2757 mtu = dst_mtu(xdst->route);
2758 if (xdst->route_mtu_cached != mtu) {
2759 last = xdst;
2760 xdst->route_mtu_cached = mtu;
2761 }
2762
2763 dst = dst->child;
2764 } while (dst->xfrm);
2765
2766 if (likely(!last))
2767 return 1;
2768
2769 mtu = last->child_mtu_cached;
2770 for (;;) {
2771 dst = &last->u.dst;
2772
2773 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2774 if (mtu > last->route_mtu_cached)
2775 mtu = last->route_mtu_cached;
defb3519 2776 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2777
2778 if (last == first)
2779 break;
2780
bd0bf076 2781 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2782 last->child_mtu_cached = mtu;
2783 }
2784
2785 return 1;
2786}
2787
0dbaee3b
DM		 2788static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2789{
2790 return dst_metric_advmss(dst->path);
2791}
2792
ebb762f2 2793static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2794{
618f9bc7
SK		 2795 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2796
2797 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2798}
2799
f894cbf8
DM		 2800static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2801 struct sk_buff *skb,
2802 const void *daddr)
d3aaeb38 2803{
f894cbf8 2804 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2805}
2806
1da177e4
LT		 2807int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2808{
d7c7544c 2809 struct net *net;
1da177e4
LT		 2810 int err = 0;
2811 if (unlikely(afinfo == NULL))
2812 return -EINVAL;
2813 if (unlikely(afinfo->family >= NPROTO))
2814 return -EAFNOSUPPORT;
ef8531b6 2815 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2816 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2817 err = -ENOBUFS;
2818 else {
2819 struct dst_ops *dst_ops = afinfo->dst_ops;
2820 if (likely(dst_ops->kmem_cachep == NULL))
2821 dst_ops->kmem_cachep = xfrm_dst_cache;
2822 if (likely(dst_ops->check == NULL))
2823 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2824 if (likely(dst_ops->default_advmss == NULL))
2825 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2826 if (likely(dst_ops->mtu == NULL))
2827 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2828 if (likely(dst_ops->negative_advice == NULL))
2829 dst_ops->negative_advice = xfrm_negative_advice;
2830 if (likely(dst_ops->link_failure == NULL))
2831 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2832 if (likely(dst_ops->neigh_lookup == NULL))
2833 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2834 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2835 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2836 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2837 }
ef8531b6 2838 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2839
2840 rtnl_lock();
2841 for_each_net(net) {
2842 struct dst_ops *xfrm_dst_ops;
2843
2844 switch (afinfo->family) {
2845 case AF_INET:
2846 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2847 break;
dfd56b8b 2848#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2849 case AF_INET6:
2850 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2851 break;
2852#endif
2853 default:
2854 BUG();
2855 }
2856 *xfrm_dst_ops = *afinfo->dst_ops;
2857 }
2858 rtnl_unlock();
2859
1da177e4
LT		 2860 return err;
2861}
2862EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2863
2864int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2865{
2866 int err = 0;
2867 if (unlikely(afinfo == NULL))
2868 return -EINVAL;
2869 if (unlikely(afinfo->family >= NPROTO))
2870 return -EAFNOSUPPORT;
ef8531b6 2871 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2872 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2873 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2874 err = -EINVAL;
ef8531b6
ED		 2875 else
2876 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2877 NULL);
2878 }
2879 spin_unlock(&xfrm_policy_afinfo_lock);
2880 if (!err) {
2881 struct dst_ops *dst_ops = afinfo->dst_ops;
2882
2883 synchronize_rcu();
2884
2885 dst_ops->kmem_cachep = NULL;
2886 dst_ops->check = NULL;
2887 dst_ops->negative_advice = NULL;
2888 dst_ops->link_failure = NULL;
2889 afinfo->garbage_collect = NULL;
1da177e4 2890 }
1da177e4
LT		 2891 return err;
2892}
2893EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2894
d7c7544c
AD		 2895static void __net_init xfrm_dst_ops_init(struct net *net)
2896{
2897 struct xfrm_policy_afinfo *afinfo;
2898
ef8531b6
ED		 2899 rcu_read_lock();
2900 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2901 if (afinfo)
2902 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2903#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2904 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2905 if (afinfo)
2906 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2907#endif
418a99ac 2908 rcu_read_unlock();
546be240
HX		 2909}
2910
1da177e4
LT		 2911static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2912{
351638e7 2913 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2914
1da177e4
LT		 2915 switch (event) {
2916 case NETDEV_DOWN:
c0ed1c14 2917 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2918 }
2919 return NOTIFY_DONE;
2920}
2921
2922static struct notifier_block xfrm_dev_notifier = {
d5917a35 2923 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2924};
2925
558f82ef 2926#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2927static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2928{
c68cd1a0 2929 int rv;
698365fa
WC		 2930 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2931 if (!net->mib.xfrm_statistics)
558f82ef 2932 return -ENOMEM;
c68cd1a0
AD		 2933 rv = xfrm_proc_init(net);
2934 if (rv < 0)
698365fa 2935 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2936 return rv;
558f82ef 2937}
59c9940e
AD		 2938
2939static void xfrm_statistics_fini(struct net *net)
2940{
c68cd1a0 2941 xfrm_proc_fini(net);
698365fa 2942 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2943}
2944#else
2945static int __net_init xfrm_statistics_init(struct net *net)
2946{
2947 return 0;
2948}
2949
2950static void xfrm_statistics_fini(struct net *net)
2951{
2952}
558f82ef
MN		 2953#endif
2954
d62ddc21 2955static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2956{
2518c7c2
DM		 2957 unsigned int hmask, sz;
2958 int dir;
2959
d62ddc21
AD		 2960 if (net_eq(net, &init_net))
2961 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2962 sizeof(struct xfrm_dst),
e5d679f3 2963 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2964 NULL);
1da177e4 2965
2518c7c2
DM		 2966 hmask = 8 - 1;
2967 sz = (hmask+1) * sizeof(struct hlist_head);
2968
93b851c1
AD		 2969 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2970 if (!net->xfrm.policy_byidx)
2971 goto out_byidx;
8100bea7 2972 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2973
53c2e285 2974 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2975 struct xfrm_policy_hash *htab;
2976
dc2caba7 2977 net->xfrm.policy_count[dir] = 0;
53c2e285 2978 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2979 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2980
a35f6c5d 2981 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2982 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2983 if (!htab->table)
a35f6c5d
AD		 2984 goto out_bydst;
2985 htab->hmask = hmask;
b58555f1
CG		 2986 htab->dbits4 = 32;
2987 htab->sbits4 = 32;
2988 htab->dbits6 = 128;
2989 htab->sbits6 = 128;
2518c7c2 2990 }
880a6fab
CG		 2991 net->xfrm.policy_hthresh.lbits4 = 32;
2992 net->xfrm.policy_hthresh.rbits4 = 32;
2993 net->xfrm.policy_hthresh.lbits6 = 128;
2994 net->xfrm.policy_hthresh.rbits6 = 128;
2995
2996 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2997
adfcf0b2 2998 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2999 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 3000 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21
AD		 3001 if (net_eq(net, &init_net))
3002 register_netdevice_notifier(&xfrm_dev_notifier);
3003 return 0;
93b851c1 3004
a35f6c5d
AD		 3005out_bydst:
3006 for (dir--; dir >= 0; dir--) {
3007 struct xfrm_policy_hash *htab;
3008
3009 htab = &net->xfrm.policy_bydst[dir];
3010 xfrm_hash_free(htab->table, sz);
3011 }
3012 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 3013out_byidx:
3014 return -ENOMEM;
d62ddc21
AD		 3015}
3016
3017static void xfrm_policy_fini(struct net *net)
3018{
93b851c1 3019 unsigned int sz;
8b18f8ea 3020 int dir;
93b851c1 3021
7c2776ee
AD		 3022 flush_work(&net->xfrm.policy_hash_work);
3023#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 3024 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 3025#endif
2e71029e 3026 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 3027
adfcf0b2 3028 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 3029
53c2e285 3030 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 3031 struct xfrm_policy_hash *htab;
3032
8b18f8ea 3033 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 3034
3035 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 3036 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 3037 WARN_ON(!hlist_empty(htab->table));
3038 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 3039 }
3040
8100bea7 3041 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 3042 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
3043 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 3044}
3045
d62ddc21
AD		 3046static int __net_init xfrm_net_init(struct net *net)
3047{
3048 int rv;
3049
59c9940e
AD		 3050 rv = xfrm_statistics_init(net);
3051 if (rv < 0)
3052 goto out_statistics;
d62ddc21
AD		 3053 rv = xfrm_state_init(net);
3054 if (rv < 0)
3055 goto out_state;
3056 rv = xfrm_policy_init(net);
3057 if (rv < 0)
3058 goto out_policy;
d7c7544c 3059 xfrm_dst_ops_init(net);
b27aeadb
AD		 3060 rv = xfrm_sysctl_init(net);
3061 if (rv < 0)
3062 goto out_sysctl;
4a93f509
SK		 3063 rv = flow_cache_init(net);
3064 if (rv < 0)
3065 goto out;
283bc9f3
FD		 3066
3067 /* Initialize the per-net locks here */
3068 spin_lock_init(&net->xfrm.xfrm_state_lock);
3069 rwlock_init(&net->xfrm.xfrm_policy_lock);
283bc9f3
FD		 3070 mutex_init(&net->xfrm.xfrm_cfg_mutex);
3071
d62ddc21
AD		 3072 return 0;
3073
4a93f509
SK		 3074out:
3075 xfrm_sysctl_fini(net);
b27aeadb
AD		 3076out_sysctl:
3077 xfrm_policy_fini(net);
d62ddc21
AD		 3078out_policy:
3079 xfrm_state_fini(net);
3080out_state:
59c9940e
AD		 3081 xfrm_statistics_fini(net);
3082out_statistics:
d62ddc21
AD		 3083 return rv;
3084}
3085
3086static void __net_exit xfrm_net_exit(struct net *net)
3087{
4a93f509 3088 flow_cache_fini(net);
b27aeadb 3089 xfrm_sysctl_fini(net);
d62ddc21
AD		 3090 xfrm_policy_fini(net);
3091 xfrm_state_fini(net);
59c9940e 3092 xfrm_statistics_fini(net);
d62ddc21
AD		 3093}
3094
3095static struct pernet_operations __net_initdata xfrm_net_ops = {
3096 .init = xfrm_net_init,
3097 .exit = xfrm_net_exit,
3098};
3099
1da177e4
LT		 3100void __init xfrm_init(void)
3101{
d62ddc21 3102 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 3103 xfrm_input_init();
3104}
3105
ab5f5e8b 3106#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 3107static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
3108 struct audit_buffer *audit_buf)
ab5f5e8b 3109{
875179fa
PM		 3110 struct xfrm_sec_ctx *ctx = xp->security;
3111 struct xfrm_selector *sel = &xp->selector;
3112
3113 if (ctx)
ab5f5e8b 3114 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3115 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3116
9b7a787d 3117 switch (sel->family) {
ab5f5e8b 3118 case AF_INET:
21454aaa 3119 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3120 if (sel->prefixlen_s != 32)
3121 audit_log_format(audit_buf, " src_prefixlen=%d",
3122 sel->prefixlen_s);
21454aaa 3123 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3124 if (sel->prefixlen_d != 32)
3125 audit_log_format(audit_buf, " dst_prefixlen=%d",
3126 sel->prefixlen_d);
ab5f5e8b
JL		 3127 break;
3128 case AF_INET6:
5b095d98 3129 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3130 if (sel->prefixlen_s != 128)
3131 audit_log_format(audit_buf, " src_prefixlen=%d",
3132 sel->prefixlen_s);
5b095d98 3133 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3134 if (sel->prefixlen_d != 128)
3135 audit_log_format(audit_buf, " dst_prefixlen=%d",
3136 sel->prefixlen_d);
ab5f5e8b
JL		 3137 break;
3138 }
3139}
3140
2e71029e 3141void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3142{
3143 struct audit_buffer *audit_buf;
ab5f5e8b 3144
afeb14b4 3145 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3146 if (audit_buf == NULL)
3147 return;
2e71029e 3148 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3149 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3150 xfrm_audit_common_policyinfo(xp, audit_buf);
3151 audit_log_end(audit_buf);
3152}
3153EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3154
68277acc 3155void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3156 bool task_valid)
ab5f5e8b
JL		 3157{
3158 struct audit_buffer *audit_buf;
ab5f5e8b 3159
afeb14b4 3160 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3161 if (audit_buf == NULL)
3162 return;
2e71029e 3163 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3164 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3165 xfrm_audit_common_policyinfo(xp, audit_buf);
3166 audit_log_end(audit_buf);
3167}
3168EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3169#endif
3170
80c9abaa 3171#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3172static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3173 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3174{
3175 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3176 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3177 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3178 sel_cmp->family) &&
3179 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3180 sel_cmp->family) &&
80c9abaa
SS		 3181 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3182 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3183 return true;
80c9abaa
SS		 3184 }
3185 } else {
3186 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3187 return true;
80c9abaa
SS		 3188 }
3189 }
bc9b35ad 3190 return false;
80c9abaa
SS		 3191}
3192
3e94c2dc
WC		 3193static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3194 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3195{
3196 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3197 struct hlist_head *chain;
3198 u32 priority = ~0U;
3199
283bc9f3 3200 read_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME*/
8d549c4f 3201 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3202 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3203 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3204 pol->type == type) {
3205 ret = pol;
3206 priority = ret->priority;
3207 break;
3208 }
3209 }
8d549c4f 3210 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3211 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3212 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3213 pol->type == type &&
3214 pol->priority < priority) {
3215 ret = pol;
3216 break;
3217 }
3218 }
3219
3220 if (ret)
3221 xfrm_pol_hold(ret);
3222
283bc9f3 3223 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3224
3225 return ret;
3226}
3227
dd701754 3228static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3229{
3230 int match = 0;
3231
3232 if (t->mode == m->mode && t->id.proto == m->proto &&
3233 (m->reqid == 0 || t->reqid == m->reqid)) {
3234 switch (t->mode) {
3235 case XFRM_MODE_TUNNEL:
3236 case XFRM_MODE_BEET:
70e94e66
YH		 3237 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3238 m->old_family) &&
3239 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3240 m->old_family)) {
80c9abaa
SS		 3241 match = 1;
3242 }
3243 break;
3244 case XFRM_MODE_TRANSPORT:
3245 /* in case of transport mode, template does not store
3246 any IP addresses, hence we just compare mode and
3247 protocol */
3248 match = 1;
3249 break;
3250 default:
3251 break;
3252 }
3253 }
3254 return match;
3255}
3256
3257/* update endpoint address(es) of template(s) */
3258static int xfrm_policy_migrate(struct xfrm_policy *pol,
3259 struct xfrm_migrate *m, int num_migrate)
3260{
3261 struct xfrm_migrate *mp;
80c9abaa
SS		 3262 int i, j, n = 0;
3263
3264 write_lock_bh(&pol->lock);
12a169e7 3265 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3266 /* target policy has been deleted */
3267 write_unlock_bh(&pol->lock);
3268 return -ENOENT;
3269 }
3270
3271 for (i = 0; i < pol->xfrm_nr; i++) {
3272 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3273 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3274 continue;
3275 n++;
1bfcb10f
HX		 3276 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3277 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3278 continue;
3279 /* update endpoints */
3280 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3281 sizeof(pol->xfrm_vec[i].id.daddr));
3282 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3283 sizeof(pol->xfrm_vec[i].saddr));
3284 pol->xfrm_vec[i].encap_family = mp->new_family;
3285 /* flush bundles */
80c802f3 3286 atomic_inc(&pol->genid);
80c9abaa
SS		 3287 }
3288 }
3289
3290 write_unlock_bh(&pol->lock);
3291
3292 if (!n)
3293 return -ENODATA;
3294
3295 return 0;
3296}
3297
dd701754 3298static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3299{
3300 int i, j;
3301
3302 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3303 return -EINVAL;
3304
3305 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3306 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3307 m[i].old_family) &&
3308 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3309 m[i].old_family))
80c9abaa
SS		 3310 return -EINVAL;
3311 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3312 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3313 return -EINVAL;
3314
3315 /* check if there is any duplicated entry */
3316 for (j = i + 1; j < num_migrate; j++) {
3317 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3318 sizeof(m[i].old_daddr)) &&
3319 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3320 sizeof(m[i].old_saddr)) &&
3321 m[i].proto == m[j].proto &&
3322 m[i].mode == m[j].mode &&
3323 m[i].reqid == m[j].reqid &&
3324 m[i].old_family == m[j].old_family)
3325 return -EINVAL;
3326 }
3327 }
3328
3329 return 0;
3330}
3331
b4b7c0b3 3332int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3333 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3334 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3335{
3336 int i, err, nx_cur = 0, nx_new = 0;
3337 struct xfrm_policy *pol = NULL;
3338 struct xfrm_state *x, *xc;
3339 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3340 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3341 struct xfrm_migrate *mp;
3342
3343 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3344 goto out;
3345
3346 /* Stage 1 - find policy */
8d549c4f 3347 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3348 err = -ENOENT;
3349 goto out;
3350 }
3351
3352 /* Stage 2 - find and update state(s) */
3353 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3354 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3355 x_cur[nx_cur] = x;
3356 nx_cur++;
3357 if ((xc = xfrm_state_migrate(x, mp))) {
3358 x_new[nx_new] = xc;
3359 nx_new++;
3360 } else {
3361 err = -ENODATA;
3362 goto restore_state;
3363 }
3364 }
3365 }
3366
3367 /* Stage 3 - update policy */
3368 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3369 goto restore_state;
3370
3371 /* Stage 4 - delete old state(s) */
3372 if (nx_cur) {
3373 xfrm_states_put(x_cur, nx_cur);
3374 xfrm_states_delete(x_cur, nx_cur);
3375 }
3376
3377 /* Stage 5 - announce */
13c1d189 3378 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3379
3380 xfrm_pol_put(pol);
3381
3382 return 0;
3383out:
3384 return err;
3385
3386restore_state:
3387 if (pol)
3388 xfrm_pol_put(pol);
3389 if (nx_cur)
3390 xfrm_states_put(x_cur, nx_cur);
3391 if (nx_new)
3392 xfrm_states_delete(x_new, nx_new);
3393
3394 return err;
3395}
e610e679 3396EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3397#endif
Linux kernel - Deliverables
This page took 1.742065 seconds and 5 git commands to generate.