]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - net/xfrm/xfrm_policy.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
xfrm: remove the unnecessary checking before call xfrm_pol_hold
[mirror_ubuntu-artful-kernel.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
1da177e4
LT		 318/* Rule must be locked. Release descentant resources, announce
319 * entry dead. The rule must be unlinked from lists to the moment.
320 */
321
322static void xfrm_policy_kill(struct xfrm_policy *policy)
323{
12a169e7 324 policy->walk.dead = 1;
1da177e4 325
285ead17 326 atomic_inc(&policy->genid);
1da177e4 327
e7d8f6cb
SK		 328 if (del_timer(&policy->polq.hold_timer))
329 xfrm_pol_put(policy);
1ee5e667 330 skb_queue_purge(&policy->polq.hold_queue);
a0073fe1 331
285ead17
TT		 332 if (del_timer(&policy->timer))
333 xfrm_pol_put(policy);
334
335 xfrm_pol_put(policy);
1da177e4
LT		 336}
337
2518c7c2
DM		 338static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
339
e92303f8 340static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 341{
e92303f8 342 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 343}
344
b58555f1
CG		 345/* calculate policy hash thresholds */
346static void __get_hash_thresh(struct net *net,
347 unsigned short family, int dir,
348 u8 *dbits, u8 *sbits)
349{
350 switch (family) {
351 case AF_INET:
352 *dbits = net->xfrm.policy_bydst[dir].dbits4;
353 *sbits = net->xfrm.policy_bydst[dir].sbits4;
354 break;
355
356 case AF_INET6:
357 *dbits = net->xfrm.policy_bydst[dir].dbits6;
358 *sbits = net->xfrm.policy_bydst[dir].sbits6;
359 break;
360
361 default:
362 *dbits = 0;
363 *sbits = 0;
364 }
365}
366
5f803b58
DM		 367static struct hlist_head *policy_hash_bysel(struct net *net,
368 const struct xfrm_selector *sel,
369 unsigned short family, int dir)
2518c7c2 370{
1121994c 371 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 372 unsigned int hash;
373 u8 dbits;
374 u8 sbits;
375
376 __get_hash_thresh(net, family, dir, &dbits, &sbits);
377 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2
DM		 378
379 return (hash == hmask + 1 ?
1121994c
AD		 380 &net->xfrm.policy_inexact[dir] :
381 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 382}
383
5f803b58
DM		 384static struct hlist_head *policy_hash_direct(struct net *net,
385 const xfrm_address_t *daddr,
386 const xfrm_address_t *saddr,
387 unsigned short family, int dir)
2518c7c2 388{
1121994c 389 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 390 unsigned int hash;
391 u8 dbits;
392 u8 sbits;
393
394 __get_hash_thresh(net, family, dir, &dbits, &sbits);
395 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 396
1121994c 397 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 398}
399
b58555f1
CG		 400static void xfrm_dst_hash_transfer(struct net *net,
401 struct hlist_head *list,
2518c7c2 402 struct hlist_head *ndsttable,
b58555f1
CG		 403 unsigned int nhashmask,
404 int dir)
2518c7c2 405{
b67bfe0d 406 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 407 struct xfrm_policy *pol;
b791160b 408 unsigned int h0 = 0;
b58555f1
CG		 409 u8 dbits;
410 u8 sbits;
2518c7c2 411
b791160b 412redo:
b67bfe0d 413 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 414 unsigned int h;
415
b58555f1 416 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 417 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 418 pol->family, nhashmask, dbits, sbits);
b791160b 419 if (!entry0) {
b67bfe0d 420 hlist_del(&pol->bydst);
b791160b
YH		 421 hlist_add_head(&pol->bydst, ndsttable+h);
422 h0 = h;
423 } else {
424 if (h != h0)
425 continue;
b67bfe0d 426 hlist_del(&pol->bydst);
1d023284 427 hlist_add_behind(&pol->bydst, entry0);
b791160b 428 }
b67bfe0d 429 entry0 = &pol->bydst;
b791160b
YH		 430 }
431 if (!hlist_empty(list)) {
432 entry0 = NULL;
433 goto redo;
2518c7c2
DM		 434 }
435}
436
437static void xfrm_idx_hash_transfer(struct hlist_head *list,
438 struct hlist_head *nidxtable,
439 unsigned int nhashmask)
440{
b67bfe0d 441 struct hlist_node *tmp;
2518c7c2
DM		 442 struct xfrm_policy *pol;
443
b67bfe0d 444 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 445 unsigned int h;
446
447 h = __idx_hash(pol->index, nhashmask);
448 hlist_add_head(&pol->byidx, nidxtable+h);
449 }
450}
451
452static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
453{
454 return ((old_hmask + 1) << 1) - 1;
455}
456
66caf628 457static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 458{
66caf628 459 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 460 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
461 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 462 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 463 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 464 int i;
465
466 if (!ndst)
467 return;
468
283bc9f3 469 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 470
471 for (i = hmask; i >= 0; i--)
b58555f1 472 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 473
66caf628
AD		 474 net->xfrm.policy_bydst[dir].table = ndst;
475 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 476
283bc9f3 477 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 478
44e36b42 479 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 480}
481
66caf628 482static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 483{
66caf628 484 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 485 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
486 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 487 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 488 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 489 int i;
490
491 if (!nidx)
492 return;
493
283bc9f3 494 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 495
496 for (i = hmask; i >= 0; i--)
497 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
498
66caf628
AD		 499 net->xfrm.policy_byidx = nidx;
500 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 501
283bc9f3 502 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 503
44e36b42 504 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 505}
506
66caf628 507static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 508{
66caf628
AD		 509 unsigned int cnt = net->xfrm.policy_count[dir];
510 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 511
512 if (total)
513 *total += cnt;
514
515 if ((hmask + 1) < xfrm_policy_hashmax &&
516 cnt > hmask)
517 return 1;
518
519 return 0;
520}
521
66caf628 522static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 523{
66caf628 524 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 525
526 if ((hmask + 1) < xfrm_policy_hashmax &&
527 total > hmask)
528 return 1;
529
530 return 0;
531}
532
e071041b 533void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 534{
283bc9f3 535 read_lock_bh(&net->xfrm.xfrm_policy_lock);
e071041b
AD		 536 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
537 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
538 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
539 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
540 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
541 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
542 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 543 si->spdhmcnt = xfrm_policy_hashmax;
283bc9f3 544 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
ecfd6b18
JHS		 545}
546EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 547
ecfd6b18 548static DEFINE_MUTEX(hash_resize_mutex);
66caf628 549static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 550{
66caf628 551 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 552 int dir, total;
553
554 mutex_lock(&hash_resize_mutex);
555
556 total = 0;
53c2e285 557 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD		 558 if (xfrm_bydst_should_resize(net, dir, &total))
559 xfrm_bydst_resize(net, dir);
2518c7c2 560 }
66caf628
AD		 561 if (xfrm_byidx_should_resize(net, total))
562 xfrm_byidx_resize(net, total);
2518c7c2
DM		 563
564 mutex_unlock(&hash_resize_mutex);
565}
566
880a6fab
CG		 567static void xfrm_hash_rebuild(struct work_struct *work)
568{
569 struct net *net = container_of(work, struct net,
570 xfrm.policy_hthresh.work);
571 unsigned int hmask;
572 struct xfrm_policy *pol;
573 struct xfrm_policy *policy;
574 struct hlist_head *chain;
575 struct hlist_head *odst;
576 struct hlist_node *newpos;
577 int i;
578 int dir;
579 unsigned seq;
580 u8 lbits4, rbits4, lbits6, rbits6;
581
582 mutex_lock(&hash_resize_mutex);
583
584 /* read selector prefixlen thresholds */
585 do {
586 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
587
588 lbits4 = net->xfrm.policy_hthresh.lbits4;
589 rbits4 = net->xfrm.policy_hthresh.rbits4;
590 lbits6 = net->xfrm.policy_hthresh.lbits6;
591 rbits6 = net->xfrm.policy_hthresh.rbits6;
592 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
593
594 write_lock_bh(&net->xfrm.xfrm_policy_lock);
595
596 /* reset the bydst and inexact table in all directions */
53c2e285 597 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG		 598 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
599 hmask = net->xfrm.policy_bydst[dir].hmask;
600 odst = net->xfrm.policy_bydst[dir].table;
601 for (i = hmask; i >= 0; i--)
602 INIT_HLIST_HEAD(odst + i);
603 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
604 /* dir out => dst = remote, src = local */
605 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
606 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
607 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
608 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
609 } else {
610 /* dir in/fwd => dst = local, src = remote */
611 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
612 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
613 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
614 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
615 }
616 }
617
618 /* re-insert all policies by order of creation */
619 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
620 newpos = NULL;
621 chain = policy_hash_bysel(net, &policy->selector,
622 policy->family,
623 xfrm_policy_id2dir(policy->index));
624 hlist_for_each_entry(pol, chain, bydst) {
625 if (policy->priority >= pol->priority)
626 newpos = &pol->bydst;
627 else
628 break;
629 }
630 if (newpos)
631 hlist_add_behind(&policy->bydst, newpos);
632 else
633 hlist_add_head(&policy->bydst, chain);
634 }
635
636 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
637
638 mutex_unlock(&hash_resize_mutex);
639}
640
641void xfrm_policy_hash_rebuild(struct net *net)
642{
643 schedule_work(&net->xfrm.policy_hthresh.work);
644}
645EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
646
1da177e4
LT		 647/* Generate new index... KAME seems to generate them ordered by cost
648 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 649static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 650{
1da177e4
LT		 651 static u32 idx_generator;
652
653 for (;;) {
2518c7c2
DM		 654 struct hlist_head *list;
655 struct xfrm_policy *p;
656 u32 idx;
657 int found;
658
e682adf0
FD		 659 if (!index) {
660 idx = (idx_generator | dir);
661 idx_generator += 8;
662 } else {
663 idx = index;
664 index = 0;
665 }
666
1da177e4
LT		 667 if (idx == 0)
668 idx = 8;
1121994c 669 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 670 found = 0;
b67bfe0d 671 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 672 if (p->index == idx) {
673 found = 1;
1da177e4 674 break;
2518c7c2 675 }
1da177e4 676 }
2518c7c2 677 if (!found)
1da177e4
LT		 678 return idx;
679 }
680}
681
2518c7c2
DM		 682static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
683{
684 u32 *p1 = (u32 *) s1;
685 u32 *p2 = (u32 *) s2;
686 int len = sizeof(struct xfrm_selector) / sizeof(u32);
687 int i;
688
689 for (i = 0; i < len; i++) {
690 if (p1[i] != p2[i])
691 return 1;
692 }
693
694 return 0;
695}
696
a0073fe1
SK		 697static void xfrm_policy_requeue(struct xfrm_policy *old,
698 struct xfrm_policy *new)
699{
700 struct xfrm_policy_queue *pq = &old->polq;
701 struct sk_buff_head list;
702
703 __skb_queue_head_init(&list);
704
705 spin_lock_bh(&pq->hold_queue.lock);
706 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 707 if (del_timer(&pq->hold_timer))
708 xfrm_pol_put(old);
a0073fe1
SK		 709 spin_unlock_bh(&pq->hold_queue.lock);
710
711 if (skb_queue_empty(&list))
712 return;
713
714 pq = &new->polq;
715
716 spin_lock_bh(&pq->hold_queue.lock);
717 skb_queue_splice(&list, &pq->hold_queue);
718 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 719 if (!mod_timer(&pq->hold_timer, jiffies))
720 xfrm_pol_hold(new);
a0073fe1
SK		 721 spin_unlock_bh(&pq->hold_queue.lock);
722}
723
7cb8a939
SK		 724static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
725 struct xfrm_policy *pol)
726{
727 u32 mark = policy->mark.v & policy->mark.m;
728
729 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
730 return true;
731
732 if ((mark & pol->mark.m) == pol->mark.v &&
733 policy->priority == pol->priority)
734 return true;
735
736 return false;
737}
738
1da177e4
LT		 739int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
740{
1121994c 741 struct net *net = xp_net(policy);
2518c7c2
DM		 742 struct xfrm_policy *pol;
743 struct xfrm_policy *delpol;
744 struct hlist_head *chain;
b67bfe0d 745 struct hlist_node *newpos;
1da177e4 746
283bc9f3 747 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 748 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 749 delpol = NULL;
750 newpos = NULL;
b67bfe0d 751 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 752 if (pol->type == policy->type &&
2518c7c2 753 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 754 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 755 xfrm_sec_ctx_match(pol->security, policy->security) &&
756 !WARN_ON(delpol)) {
1da177e4 757 if (excl) {
283bc9f3 758 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 759 return -EEXIST;
760 }
1da177e4
LT		 761 delpol = pol;
762 if (policy->priority > pol->priority)
763 continue;
764 } else if (policy->priority >= pol->priority) {
a6c7ab55 765 newpos = &pol->bydst;
1da177e4
LT		 766 continue;
767 }
1da177e4
LT		 768 if (delpol)
769 break;
1da177e4
LT		 770 }
771 if (newpos)
1d023284 772 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 773 else
774 hlist_add_head(&policy->bydst, chain);
12bfa8bd 775 __xfrm_policy_link(policy, dir);
ca925cf1 776 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 777
778 /* After previous checking, family can either be AF_INET or AF_INET6 */
779 if (policy->family == AF_INET)
780 rt_genid_bump_ipv4(net);
781 else
782 rt_genid_bump_ipv6(net);
783
a0073fe1
SK		 784 if (delpol) {
785 xfrm_policy_requeue(delpol, policy);
29fa0b30 786 __xfrm_policy_unlink(delpol, dir);
a0073fe1 787 }
e682adf0 788 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 789 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 790 policy->curlft.add_time = get_seconds();
1da177e4
LT		 791 policy->curlft.use_time = 0;
792 if (!mod_timer(&policy->timer, jiffies + HZ))
793 xfrm_pol_hold(policy);
283bc9f3 794 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 795
9b78a82c 796 if (delpol)
1da177e4 797 xfrm_policy_kill(delpol);
1121994c
AD		 798 else if (xfrm_bydst_should_resize(net, dir, NULL))
799 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 800
1da177e4
LT		 801 return 0;
802}
803EXPORT_SYMBOL(xfrm_policy_insert);
804
8ca2e93b
JHS		 805struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
806 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 807 struct xfrm_sec_ctx *ctx, int delete,
808 int *err)
1da177e4 809{
2518c7c2
DM		 810 struct xfrm_policy *pol, *ret;
811 struct hlist_head *chain;
1da177e4 812
ef41aaa0 813 *err = 0;
283bc9f3 814 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 815 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 816 ret = NULL;
b67bfe0d 817 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 818 if (pol->type == type &&
34f8d884 819 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 820 !selector_cmp(sel, &pol->selector) &&
821 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 822 xfrm_pol_hold(pol);
2518c7c2 823 if (delete) {
03e1ad7b
PM		 824 *err = security_xfrm_policy_delete(
825 pol->security);
ef41aaa0 826 if (*err) {
283bc9f3 827 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 828 return pol;
829 }
29fa0b30 830 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 831 }
832 ret = pol;
1da177e4
LT		 833 break;
834 }
835 }
283bc9f3 836 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 837
fe1a5f03 838 if (ret && delete)
2518c7c2 839 xfrm_policy_kill(ret);
2518c7c2 840 return ret;
1da177e4 841}
df71837d 842EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 843
8ca2e93b
JHS		 844struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
845 int dir, u32 id, int delete, int *err)
1da177e4 846{
2518c7c2
DM		 847 struct xfrm_policy *pol, *ret;
848 struct hlist_head *chain;
1da177e4 849
b5505c6e
HX		 850 *err = -ENOENT;
851 if (xfrm_policy_id2dir(id) != dir)
852 return NULL;
853
ef41aaa0 854 *err = 0;
283bc9f3 855 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 856 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 857 ret = NULL;
b67bfe0d 858 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 859 if (pol->type == type && pol->index == id &&
860 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 861 xfrm_pol_hold(pol);
2518c7c2 862 if (delete) {
03e1ad7b
PM		 863 *err = security_xfrm_policy_delete(
864 pol->security);
ef41aaa0 865 if (*err) {
283bc9f3 866 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 867 return pol;
868 }
29fa0b30 869 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 870 }
871 ret = pol;
1da177e4
LT		 872 break;
873 }
874 }
283bc9f3 875 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 876
fe1a5f03 877 if (ret && delete)
2518c7c2 878 xfrm_policy_kill(ret);
2518c7c2 879 return ret;
1da177e4
LT		 880}
881EXPORT_SYMBOL(xfrm_policy_byid);
882
4aa2e62c
JL		 883#ifdef CONFIG_SECURITY_NETWORK_XFRM
884static inline int
2e71029e 885xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 886{
4aa2e62c
JL		 887 int dir, err = 0;
888
889 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
890 struct xfrm_policy *pol;
4aa2e62c
JL		 891 int i;
892
b67bfe0d 893 hlist_for_each_entry(pol,
33ffbbd5 894 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 895 if (pol->type != type)
896 continue;
03e1ad7b 897 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 898 if (err) {
2e71029e 899 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 900 return err;
901 }
7dc12d6d 902 }
33ffbbd5 903 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 904 hlist_for_each_entry(pol,
33ffbbd5 905 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 906 bydst) {
907 if (pol->type != type)
908 continue;
03e1ad7b
PM		 909 err = security_xfrm_policy_delete(
910 pol->security);
4aa2e62c 911 if (err) {
ab5f5e8b 912 xfrm_audit_policy_delete(pol, 0,
2e71029e 913 task_valid);
4aa2e62c
JL		 914 return err;
915 }
916 }
917 }
918 }
919 return err;
920}
921#else
922static inline int
2e71029e 923xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 924{
925 return 0;
926}
927#endif
928
2e71029e 929int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 930{
2f1eb65f 931 int dir, err = 0, cnt = 0;
1da177e4 932
283bc9f3 933 write_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 934
2e71029e 935 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 936 if (err)
937 goto out;
938
1da177e4 939 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 940 struct xfrm_policy *pol;
29fa0b30 941 int i;
2518c7c2
DM		 942
943 again1:
b67bfe0d 944 hlist_for_each_entry(pol,
33ffbbd5 945 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 946 if (pol->type != type)
947 continue;
ea2dea9d 948 __xfrm_policy_unlink(pol, dir);
283bc9f3 949 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 950 cnt++;
1da177e4 951
2e71029e 952 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 953
2518c7c2 954 xfrm_policy_kill(pol);
1da177e4 955
283bc9f3 956 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 957 goto again1;
958 }
959
33ffbbd5 960 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 961 again2:
b67bfe0d 962 hlist_for_each_entry(pol,
33ffbbd5 963 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 964 bydst) {
965 if (pol->type != type)
966 continue;
ea2dea9d 967 __xfrm_policy_unlink(pol, dir);
283bc9f3 968 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 969 cnt++;
2518c7c2 970
2e71029e 971 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 972 xfrm_policy_kill(pol);
973
283bc9f3 974 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 975 goto again2;
976 }
1da177e4 977 }
2518c7c2 978
1da177e4 979 }
2f1eb65f
JHS		 980 if (!cnt)
981 err = -ESRCH;
4aa2e62c 982out:
283bc9f3 983 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 984 return err;
1da177e4
LT		 985}
986EXPORT_SYMBOL(xfrm_policy_flush);
987
cdcbca7c 988int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 989 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 990 void *data)
991{
12a169e7
HX		 992 struct xfrm_policy *pol;
993 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 994 int error = 0;
995
996 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
997 walk->type != XFRM_POLICY_TYPE_ANY)
998 return -EINVAL;
1da177e4 999
12a169e7 1000 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 1001 return 0;
1002
283bc9f3 1003 write_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1004 if (list_empty(&walk->walk.all))
cdcbca7c 1005 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1006 else
80077702
LR		 1007 x = list_first_entry(&walk->walk.all,
1008 struct xfrm_policy_walk_entry, all);
1009
cdcbca7c 1010 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1011 if (x->dead)
4c563f76 1012 continue;
12a169e7
HX		 1013 pol = container_of(x, struct xfrm_policy, walk);
1014 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1015 walk->type != pol->type)
1016 continue;
1017 error = func(pol, xfrm_policy_id2dir(pol->index),
1018 walk->seq, data);
1019 if (error) {
1020 list_move_tail(&walk->walk.all, &x->all);
1021 goto out;
2518c7c2 1022 }
12a169e7 1023 walk->seq++;
1da177e4 1024 }
12a169e7 1025 if (walk->seq == 0) {
baf5d743
JHS		 1026 error = -ENOENT;
1027 goto out;
1028 }
12a169e7 1029 list_del_init(&walk->walk.all);
1da177e4 1030out:
283bc9f3 1031 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1032 return error;
1033}
1034EXPORT_SYMBOL(xfrm_policy_walk);
1035
12a169e7
HX		 1036void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1037{
1038 INIT_LIST_HEAD(&walk->walk.all);
1039 walk->walk.dead = 1;
1040 walk->type = type;
1041 walk->seq = 0;
1042}
1043EXPORT_SYMBOL(xfrm_policy_walk_init);
1044
283bc9f3 1045void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1046{
1047 if (list_empty(&walk->walk.all))
1048 return;
1049
283bc9f3 1050 write_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1051 list_del(&walk->walk.all);
283bc9f3 1052 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1053}
1054EXPORT_SYMBOL(xfrm_policy_walk_done);
1055
134b0fc5
JM		 1056/*
1057 * Find policy to apply to this flow.
1058 *
1059 * Returns 0 if policy found, else an -errno.
1060 */
f299d557
DM		 1061static int xfrm_policy_match(const struct xfrm_policy *pol,
1062 const struct flowi *fl,
2518c7c2 1063 u8 type, u16 family, int dir)
1da177e4 1064{
f299d557 1065 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1066 int ret = -ESRCH;
1067 bool match;
1da177e4 1068
2518c7c2 1069 if (pol->family != family ||
1d28f42c 1070 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1071 pol->type != type)
134b0fc5 1072 return ret;
1da177e4 1073
2518c7c2 1074 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1075 if (match)
1d28f42c 1076 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1077 dir);
2518c7c2 1078
134b0fc5 1079 return ret;
2518c7c2 1080}
1da177e4 1081
52479b62 1082static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1083 const struct flowi *fl,
2518c7c2
DM		 1084 u16 family, u8 dir)
1085{
134b0fc5 1086 int err;
2518c7c2 1087 struct xfrm_policy *pol, *ret;
0b597e7e 1088 const xfrm_address_t *daddr, *saddr;
2518c7c2 1089 struct hlist_head *chain;
acba48e1 1090 u32 priority = ~0U;
df71837d 1091
2518c7c2
DM		 1092 daddr = xfrm_flowi_daddr(fl, family);
1093 saddr = xfrm_flowi_saddr(fl, family);
1094 if (unlikely(!daddr || !saddr))
1095 return NULL;
1096
283bc9f3 1097 read_lock_bh(&net->xfrm.xfrm_policy_lock);
52479b62 1098 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 1099 ret = NULL;
b67bfe0d 1100 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1101 err = xfrm_policy_match(pol, fl, type, family, dir);
1102 if (err) {
1103 if (err == -ESRCH)
1104 continue;
1105 else {
1106 ret = ERR_PTR(err);
1107 goto fail;
1108 }
1109 } else {
2518c7c2 1110 ret = pol;
acba48e1 1111 priority = ret->priority;
2518c7c2
DM		 1112 break;
1113 }
1114 }
52479b62 1115 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1116 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1117 err = xfrm_policy_match(pol, fl, type, family, dir);
1118 if (err) {
1119 if (err == -ESRCH)
1120 continue;
1121 else {
1122 ret = ERR_PTR(err);
1123 goto fail;
1124 }
1125 } else if (pol->priority < priority) {
acba48e1
DM		 1126 ret = pol;
1127 break;
1da177e4
LT		 1128 }
1129 }
586f2eb4
LR		 1130
1131 xfrm_pol_hold(ret);
134b0fc5 1132fail:
283bc9f3 1133 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
4e81bb83 1134
2518c7c2 1135 return ret;
4e81bb83
MN		 1136}
1137
80c802f3 1138static struct xfrm_policy *
73ff93cd 1139__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1140{
1141#ifdef CONFIG_XFRM_SUB_POLICY
1142 struct xfrm_policy *pol;
1143
1144 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1145 if (pol != NULL)
1146 return pol;
1147#endif
1148 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1149}
1150
b5fb82c4
BZ		 1151static int flow_to_policy_dir(int dir)
1152{
1153 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1154 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1155 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1156 return dir;
1157
1158 switch (dir) {
1159 default:
1160 case FLOW_DIR_IN:
1161 return XFRM_POLICY_IN;
1162 case FLOW_DIR_OUT:
1163 return XFRM_POLICY_OUT;
1164 case FLOW_DIR_FWD:
1165 return XFRM_POLICY_FWD;
1166 }
1167}
1168
fe1a5f03 1169static struct flow_cache_object *
dee9f4bc 1170xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1171 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1172{
1173 struct xfrm_policy *pol;
fe1a5f03
TT		 1174
1175 if (old_obj)
1176 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1177
b5fb82c4 1178 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1179 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1180 return ERR_CAST(pol);
fe1a5f03 1181
fe1a5f03
TT		 1182 /* Resolver returns two references:
1183 * one for cache and one for caller of flow_cache_lookup() */
1184 xfrm_pol_hold(pol);
1185
1186 return &pol->flo;
1da177e4
LT		 1187}
1188
df71837d
TJ		 1189static inline int policy_to_flow_dir(int dir)
1190{
1191 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1192 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1193 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1194 return dir;
1195 switch (dir) {
1196 default:
1197 case XFRM_POLICY_IN:
1198 return FLOW_DIR_IN;
1199 case XFRM_POLICY_OUT:
1200 return FLOW_DIR_OUT;
1201 case XFRM_POLICY_FWD:
1202 return FLOW_DIR_FWD;
3ff50b79 1203 }
df71837d
TJ		 1204}
1205
dee9f4bc
DM		 1206static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1207 const struct flowi *fl)
1da177e4
LT		 1208{
1209 struct xfrm_policy *pol;
283bc9f3 1210 struct net *net = sock_net(sk);
1da177e4 1211
283bc9f3 1212 read_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1213 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1214 bool match = xfrm_selector_match(&pol->selector, fl,
1215 sk->sk_family);
a716c119 1216 int err = 0;
df71837d 1217
3bccfbc7 1218 if (match) {
34f8d884
JHS		 1219 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1220 pol = NULL;
1221 goto out;
1222 }
03e1ad7b 1223 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1224 fl->flowi_secid,
03e1ad7b 1225 policy_to_flow_dir(dir));
3bccfbc7
VY		 1226 if (!err)
1227 xfrm_pol_hold(pol);
1228 else if (err == -ESRCH)
1229 pol = NULL;
1230 else
1231 pol = ERR_PTR(err);
1232 } else
1da177e4
LT		 1233 pol = NULL;
1234 }
34f8d884 1235out:
283bc9f3 1236 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1237 return pol;
1238}
1239
1240static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1241{
98806f75 1242 struct net *net = xp_net(pol);
4e81bb83 1243
98806f75 1244 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1245 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1246 xfrm_pol_hold(pol);
1247}
1248
1249static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1250 int dir)
1251{
98806f75
AD		 1252 struct net *net = xp_net(pol);
1253
53c2e285 1254 if (list_empty(&pol->walk.all))
2518c7c2 1255 return NULL;
1da177e4 1256
53c2e285
HX		 1257 /* Socket policies are not hashed. */
1258 if (!hlist_unhashed(&pol->bydst)) {
1259 hlist_del(&pol->bydst);
1260 hlist_del(&pol->byidx);
1261 }
1262
1263 list_del_init(&pol->walk.all);
98806f75 1264 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1265
1266 return pol;
1da177e4
LT		 1267}
1268
53c2e285
HX		 1269static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1270{
1271 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1272}
1273
1274static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1275{
1276 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1277}
1278
4666faab 1279int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1280{
283bc9f3
FD		 1281 struct net *net = xp_net(pol);
1282
1283 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1284 pol = __xfrm_policy_unlink(pol, dir);
283bc9f3 1285 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1286 if (pol) {
1da177e4 1287 xfrm_policy_kill(pol);
4666faab 1288 return 0;
1da177e4 1289 }
4666faab 1290 return -ENOENT;
1da177e4 1291}
a70fcb0b 1292EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1293
1294int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1295{
1121994c 1296 struct net *net = xp_net(pol);
1da177e4
LT		 1297 struct xfrm_policy *old_pol;
1298
4e81bb83
MN		 1299#ifdef CONFIG_XFRM_SUB_POLICY
1300 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1301 return -EINVAL;
1302#endif
1303
283bc9f3 1304 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1305 old_pol = sk->sk_policy[dir];
1306 sk->sk_policy[dir] = pol;
1307 if (pol) {
9d729f72 1308 pol->curlft.add_time = get_seconds();
e682adf0 1309 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1310 xfrm_sk_policy_link(pol, dir);
1da177e4 1311 }
a0073fe1
SK		 1312 if (old_pol) {
1313 if (pol)
1314 xfrm_policy_requeue(old_pol, pol);
1315
ea2dea9d
TT		 1316 /* Unlinking succeeds always. This is the only function
1317 * allowed to delete or replace socket policy.
1318 */
53c2e285 1319 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1320 }
283bc9f3 1321 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1322
1323 if (old_pol) {
1324 xfrm_policy_kill(old_pol);
1325 }
1326 return 0;
1327}
1328
d3e40a9f 1329static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1330{
0331b1f3 1331 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1332 struct net *net = xp_net(old);
1da177e4
LT		 1333
1334 if (newp) {
1335 newp->selector = old->selector;
03e1ad7b
PM		 1336 if (security_xfrm_policy_clone(old->security,
1337 &newp->security)) {
df71837d
TJ		 1338 kfree(newp);
1339 return NULL; /* ENOMEM */
1340 }
1da177e4
LT		 1341 newp->lft = old->lft;
1342 newp->curlft = old->curlft;
fb977e2c 1343 newp->mark = old->mark;
1da177e4
LT		 1344 newp->action = old->action;
1345 newp->flags = old->flags;
1346 newp->xfrm_nr = old->xfrm_nr;
1347 newp->index = old->index;
4e81bb83 1348 newp->type = old->type;
1da177e4
LT		 1349 memcpy(newp->xfrm_vec, old->xfrm_vec,
1350 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
283bc9f3 1351 write_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1352 xfrm_sk_policy_link(newp, dir);
283bc9f3 1353 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1354 xfrm_pol_put(newp);
1355 }
1356 return newp;
1357}
1358
1359int __xfrm_sk_clone_policy(struct sock *sk)
1360{
1361 struct xfrm_policy *p0 = sk->sk_policy[0],
1362 *p1 = sk->sk_policy[1];
1363
1364 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1365 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1366 return -ENOMEM;
1367 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1368 return -ENOMEM;
1369 return 0;
1370}
1371
a1e59abf 1372static int
fbda33b2 1373xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1374 unsigned short family)
1375{
1376 int err;
1377 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1378
1379 if (unlikely(afinfo == NULL))
1380 return -EINVAL;
fbda33b2 1381 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1382 xfrm_policy_put_afinfo(afinfo);
1383 return err;
1384}
1385
1da177e4
LT		 1386/* Resolve list of templates for the flow, given policy. */
1387
1388static int
a6c2e611
DM		 1389xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1390 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1391{
fbda33b2 1392 struct net *net = xp_net(policy);
1da177e4
LT		 1393 int nx;
1394 int i, error;
1395 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1396 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1397 xfrm_address_t tmp;
1da177e4 1398
9b7a787d 1399 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1400 struct xfrm_state *x;
1401 xfrm_address_t *remote = daddr;
1402 xfrm_address_t *local = saddr;
1403 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1404
48b8d783
JK		 1405 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1406 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1407 remote = &tmpl->id.daddr;
1408 local = &tmpl->saddr;
8444cf71
TE		 1409 if (xfrm_addr_any(local, tmpl->encap_family)) {
1410 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1411 if (error)
1412 goto fail;
1413 local = &tmp;
1414 }
1da177e4
LT		 1415 }
1416
1417 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1418
1419 if (x && x->km.state == XFRM_STATE_VALID) {
1420 xfrm[nx++] = x;
1421 daddr = remote;
1422 saddr = local;
1423 continue;
1424 }
1425 if (x) {
1426 error = (x->km.state == XFRM_STATE_ERROR ?
1427 -EINVAL : -EAGAIN);
1428 xfrm_state_put(x);
42054569 1429 } else if (error == -ESRCH) {
a4322266 1430 error = -EAGAIN;
42054569 1431 }
1da177e4
LT		 1432
1433 if (!tmpl->optional)
1434 goto fail;
1435 }
1436 return nx;
1437
1438fail:
9b7a787d 1439 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1440 xfrm_state_put(xfrm[nx]);
1441 return error;
1442}
1443
4e81bb83 1444static int
a6c2e611
DM		 1445xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1446 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1447{
41a49cc3
MN		 1448 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1449 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1450 int cnx = 0;
1451 int error;
1452 int ret;
1453 int i;
1454
1455 for (i = 0; i < npols; i++) {
1456 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1457 error = -ENOBUFS;
1458 goto fail;
1459 }
41a49cc3
MN		 1460
1461 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1462 if (ret < 0) {
1463 error = ret;
1464 goto fail;
1465 } else
1466 cnx += ret;
1467 }
1468
41a49cc3
MN		 1469 /* found states are sorted for outbound processing */
1470 if (npols > 1)
1471 xfrm_state_sort(xfrm, tpp, cnx, family);
1472
4e81bb83
MN		 1473 return cnx;
1474
1475 fail:
9b7a787d 1476 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1477 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1478 return error;
1479
1480}
1481
1da177e4
LT		 1482/* Check that the bundle accepts the flow and its components are
1483 * still valid.
1484 */
1485
05d84025 1486static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1487{
1488 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1489 int tos;
1da177e4 1490
25ee3286
HX		 1491 if (!afinfo)
1492 return -EINVAL;
1493
1494 tos = afinfo->get_tos(fl);
1495
1496 xfrm_policy_put_afinfo(afinfo);
1497
1498 return tos;
1499}
1500
80c802f3
TT		 1501static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1502{
1503 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1504 struct dst_entry *dst = &xdst->u.dst;
1505
1506 if (xdst->route == NULL) {
1507 /* Dummy bundle - if it has xfrms we were not
1508 * able to build bundle as template resolution failed.
1509 * It means we need to try again resolving. */
1510 if (xdst->num_xfrms > 0)
1511 return NULL;
a0073fe1
SK		 1512 } else if (dst->flags & DST_XFRM_QUEUE) {
1513 return NULL;
80c802f3
TT		 1514 } else {
1515 /* Real bundle */
1516 if (stale_bundle(dst))
1517 return NULL;
1518 }
1519
1520 dst_hold(dst);
1521 return flo;
1522}
1523
1524static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1525{
1526 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1527 struct dst_entry *dst = &xdst->u.dst;
1528
1529 if (!xdst->route)
1530 return 0;
1531 if (stale_bundle(dst))
1532 return 0;
1533
1534 return 1;
1535}
1536
1537static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1538{
1539 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1540 struct dst_entry *dst = &xdst->u.dst;
1541
1542 dst_free(dst);
1543}
1544
1545static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1546 .get = xfrm_bundle_flo_get,
1547 .check = xfrm_bundle_flo_check,
1548 .delete = xfrm_bundle_flo_delete,
1549};
1550
d7c7544c 1551static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1552{
1da177e4 1553 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1554 struct dst_ops *dst_ops;
25ee3286
HX		 1555 struct xfrm_dst *xdst;
1556
1557 if (!afinfo)
1558 return ERR_PTR(-EINVAL);
1559
d7c7544c
AD		 1560 switch (family) {
1561 case AF_INET:
1562 dst_ops = &net->xfrm.xfrm4_dst_ops;
1563 break;
dfd56b8b 1564#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1565 case AF_INET6:
1566 dst_ops = &net->xfrm.xfrm6_dst_ops;
1567 break;
1568#endif
1569 default:
1570 BUG();
1571 }
f5b0a874 1572 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1573
d4cae562 1574 if (likely(xdst)) {
141e369d
SK		 1575 struct dst_entry *dst = &xdst->u.dst;
1576
1577 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1578 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1579 if (afinfo->init_dst)
1580 afinfo->init_dst(net, xdst);
d4cae562 1581 } else
0b150932 1582 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1583
d4cae562
MB		 1584 xfrm_policy_put_afinfo(afinfo);
1585
25ee3286
HX		 1586 return xdst;
1587}
1588
a1b05140
MN		 1589static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1590 int nfheader_len)
1591{
1592 struct xfrm_policy_afinfo *afinfo =
1593 xfrm_policy_get_afinfo(dst->ops->family);
1594 int err;
1595
1596 if (!afinfo)
1597 return -EINVAL;
1598
1599 err = afinfo->init_path(path, dst, nfheader_len);
1600
1601 xfrm_policy_put_afinfo(afinfo);
1602
1603 return err;
1604}
1605
87c1e12b 1606static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1607 const struct flowi *fl)
25ee3286
HX		 1608{
1609 struct xfrm_policy_afinfo *afinfo =
1610 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1611 int err;
1612
1613 if (!afinfo)
1da177e4 1614 return -EINVAL;
25ee3286 1615
87c1e12b 1616 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1617
1da177e4 1618 xfrm_policy_put_afinfo(afinfo);
25ee3286 1619
1da177e4
LT		 1620 return err;
1621}
1622
80c802f3 1623
25ee3286
HX		 1624/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1625 * all the metrics... Shortly, bundle a bundle.
1626 */
1627
1628static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1629 struct xfrm_state **xfrm, int nx,
98313ada 1630 const struct flowi *fl,
25ee3286
HX		 1631 struct dst_entry *dst)
1632{
d7c7544c 1633 struct net *net = xp_net(policy);
25ee3286
HX		 1634 unsigned long now = jiffies;
1635 struct net_device *dev;
43a4dea4 1636 struct xfrm_mode *inner_mode;
25ee3286
HX		 1637 struct dst_entry *dst_prev = NULL;
1638 struct dst_entry *dst0 = NULL;
1639 int i = 0;
1640 int err;
1641 int header_len = 0;
a1b05140 1642 int nfheader_len = 0;
25ee3286
HX		 1643 int trailer_len = 0;
1644 int tos;
1645 int family = policy->selector.family;
9bb182a7
YH		 1646 xfrm_address_t saddr, daddr;
1647
1648 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1649
1650 tos = xfrm_get_tos(fl, family);
1651 err = tos;
1652 if (tos < 0)
1653 goto put_states;
1654
1655 dst_hold(dst);
1656
1657 for (; i < nx; i++) {
d7c7544c 1658 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1659 struct dst_entry *dst1 = &xdst->u.dst;
1660
1661 err = PTR_ERR(xdst);
1662 if (IS_ERR(xdst)) {
1663 dst_release(dst);
1664 goto put_states;
1665 }
1666
43a4dea4
SK		 1667 if (xfrm[i]->sel.family == AF_UNSPEC) {
1668 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1669 xfrm_af2proto(family));
1670 if (!inner_mode) {
1671 err = -EAFNOSUPPORT;
1672 dst_release(dst);
1673 goto put_states;
1674 }
1675 } else
1676 inner_mode = xfrm[i]->inner_mode;
1677
25ee3286
HX		 1678 if (!dst_prev)
1679 dst0 = dst1;
1680 else {
1681 dst_prev->child = dst_clone(dst1);
1682 dst1->flags |= DST_NOHASH;
1683 }
1684
1685 xdst->route = dst;
defb3519 1686 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1687
1688 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1689 family = xfrm[i]->props.family;
9bb182a7
YH		 1690 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1691 family);
25ee3286
HX		 1692 err = PTR_ERR(dst);
1693 if (IS_ERR(dst))
1694 goto put_states;
1695 } else
1696 dst_hold(dst);
1697
1698 dst1->xfrm = xfrm[i];
80c802f3 1699 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1700
f5b0a874 1701 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1702 dst1->flags |= DST_HOST;
1703 dst1->lastuse = now;
1704
1705 dst1->input = dst_discard;
43a4dea4 1706 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1707
1708 dst1->next = dst_prev;
1709 dst_prev = dst1;
1710
1711 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1712 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1713 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1714 trailer_len += xfrm[i]->props.trailer_len;
1715 }
1716
1717 dst_prev->child = dst;
1718 dst0->path = dst;
1719
1720 err = -ENODEV;
1721 dev = dst->dev;
1722 if (!dev)
1723 goto free_dst;
1724
a1b05140 1725 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1726 xfrm_init_pmtu(dst_prev);
1727
1728 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1729 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1730
87c1e12b 1731 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1732 if (err)
1733 goto free_dst;
1734
1735 dst_prev->header_len = header_len;
1736 dst_prev->trailer_len = trailer_len;
1737 header_len -= xdst->u.dst.xfrm->props.header_len;
1738 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1739 }
1740
1741out:
1742 return dst0;
1743
1744put_states:
1745 for (; i < nx; i++)
1746 xfrm_state_put(xfrm[i]);
1747free_dst:
1748 if (dst0)
1749 dst_free(dst0);
1750 dst0 = ERR_PTR(err);
1751 goto out;
1752}
1753
da7c224b 1754#ifdef CONFIG_XFRM_SUB_POLICY
be7928d2 1755static int xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1756{
1757 if (!*target) {
1758 *target = kmalloc(size, GFP_ATOMIC);
1759 if (!*target)
1760 return -ENOMEM;
1761 }
be7928d2 1762
157bfc25
MN		 1763 memcpy(*target, src, size);
1764 return 0;
1765}
da7c224b 1766#endif
157bfc25 1767
be7928d2
DB		 1768static int xfrm_dst_update_parent(struct dst_entry *dst,
1769 const struct xfrm_selector *sel)
157bfc25
MN		 1770{
1771#ifdef CONFIG_XFRM_SUB_POLICY
1772 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1773 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1774 sel, sizeof(*sel));
1775#else
1776 return 0;
1777#endif
1778}
1779
be7928d2
DB		 1780static int xfrm_dst_update_origin(struct dst_entry *dst,
1781 const struct flowi *fl)
157bfc25
MN		 1782{
1783#ifdef CONFIG_XFRM_SUB_POLICY
1784 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1785 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1786#else
1787 return 0;
1788#endif
1789}
1da177e4 1790
73ff93cd 1791static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1792 struct xfrm_policy **pols,
1793 int *num_pols, int *num_xfrms)
1794{
1795 int i;
1796
1797 if (*num_pols == 0 || !pols[0]) {
1798 *num_pols = 0;
1799 *num_xfrms = 0;
1800 return 0;
1801 }
1802 if (IS_ERR(pols[0]))
1803 return PTR_ERR(pols[0]);
1804
1805 *num_xfrms = pols[0]->xfrm_nr;
1806
1807#ifdef CONFIG_XFRM_SUB_POLICY
1808 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1809 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1810 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1811 XFRM_POLICY_TYPE_MAIN,
1812 fl, family,
1813 XFRM_POLICY_OUT);
1814 if (pols[1]) {
1815 if (IS_ERR(pols[1])) {
1816 xfrm_pols_put(pols, *num_pols);
1817 return PTR_ERR(pols[1]);
1818 }
02d0892f 1819 (*num_pols)++;
80c802f3
TT		 1820 (*num_xfrms) += pols[1]->xfrm_nr;
1821 }
1822 }
1823#endif
1824 for (i = 0; i < *num_pols; i++) {
1825 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1826 *num_xfrms = -1;
1827 break;
1828 }
1829 }
1830
1831 return 0;
1832
1833}
1834
1835static struct xfrm_dst *
1836xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1837 const struct flowi *fl, u16 family,
80c802f3
TT		 1838 struct dst_entry *dst_orig)
1839{
1840 struct net *net = xp_net(pols[0]);
1841 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1842 struct dst_entry *dst;
1843 struct xfrm_dst *xdst;
1844 int err;
1845
1846 /* Try to instantiate a bundle */
1847 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1848 if (err <= 0) {
1849 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1850 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1851 return ERR_PTR(err);
1852 }
1853
1854 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1855 if (IS_ERR(dst)) {
1856 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1857 return ERR_CAST(dst);
1858 }
1859
1860 xdst = (struct xfrm_dst *)dst;
1861 xdst->num_xfrms = err;
1862 if (num_pols > 1)
1863 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1864 else
1865 err = xfrm_dst_update_origin(dst, fl);
1866 if (unlikely(err)) {
1867 dst_free(dst);
1868 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1869 return ERR_PTR(err);
1870 }
1871
1872 xdst->num_pols = num_pols;
3e94c2dc 1873 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1874 xdst->policy_genid = atomic_read(&pols[0]->genid);
1875
1876 return xdst;
1877}
1878
a0073fe1
SK		 1879static void xfrm_policy_queue_process(unsigned long arg)
1880{
a0073fe1
SK		 1881 struct sk_buff *skb;
1882 struct sock *sk;
1883 struct dst_entry *dst;
a0073fe1
SK		 1884 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1885 struct xfrm_policy_queue *pq = &pol->polq;
1886 struct flowi fl;
1887 struct sk_buff_head list;
1888
1889 spin_lock(&pq->hold_queue.lock);
1890 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1891 if (!skb) {
1892 spin_unlock(&pq->hold_queue.lock);
1893 goto out;
1894 }
a0073fe1
SK		 1895 dst = skb_dst(skb);
1896 sk = skb->sk;
1897 xfrm_decode_session(skb, &fl, dst->ops->family);
1898 spin_unlock(&pq->hold_queue.lock);
1899
1900 dst_hold(dst->path);
1901 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1902 sk, 0);
1903 if (IS_ERR(dst))
1904 goto purge_queue;
1905
1906 if (dst->flags & DST_XFRM_QUEUE) {
1907 dst_release(dst);
1908
1909 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1910 goto purge_queue;
1911
1912 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1913 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1914 xfrm_pol_hold(pol);
1915 goto out;
a0073fe1
SK		 1916 }
1917
1918 dst_release(dst);
1919
1920 __skb_queue_head_init(&list);
1921
1922 spin_lock(&pq->hold_queue.lock);
1923 pq->timeout = 0;
1924 skb_queue_splice_init(&pq->hold_queue, &list);
1925 spin_unlock(&pq->hold_queue.lock);
1926
1927 while (!skb_queue_empty(&list)) {
1928 skb = __skb_dequeue(&list);
1929
1930 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1931 dst_hold(skb_dst(skb)->path);
1932 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1933 &fl, skb->sk, 0);
1934 if (IS_ERR(dst)) {
a0073fe1
SK		 1935 kfree_skb(skb);
1936 continue;
1937 }
1938
1939 nf_reset(skb);
1940 skb_dst_drop(skb);
1941 skb_dst_set(skb, dst);
1942
5c1e9f2c 1943 dst_output(skb);
a0073fe1
SK		 1944 }
1945
e7d8f6cb
SK		 1946out:
1947 xfrm_pol_put(pol);
a0073fe1
SK		 1948 return;
1949
1950purge_queue:
1951 pq->timeout = 0;
1ee5e667 1952 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1953 xfrm_pol_put(pol);
a0073fe1
SK		 1954}
1955
aad88724 1956static int xdst_queue_output(struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1957{
1958 unsigned long sched_next;
1959 struct dst_entry *dst = skb_dst(skb);
1960 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1961 struct xfrm_policy *pol = xdst->pols[0];
1962 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1963
39bb5e62 1964 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1965 kfree_skb(skb);
1966 return 0;
1967 }
a0073fe1
SK		 1968
1969 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1970 kfree_skb(skb);
1971 return -EAGAIN;
1972 }
1973
1974 skb_dst_force(skb);
a0073fe1
SK		 1975
1976 spin_lock_bh(&pq->hold_queue.lock);
1977
1978 if (!pq->timeout)
1979 pq->timeout = XFRM_QUEUE_TMO_MIN;
1980
1981 sched_next = jiffies + pq->timeout;
1982
1983 if (del_timer(&pq->hold_timer)) {
1984 if (time_before(pq->hold_timer.expires, sched_next))
1985 sched_next = pq->hold_timer.expires;
e7d8f6cb 1986 xfrm_pol_put(pol);
a0073fe1
SK		 1987 }
1988
1989 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1990 if (!mod_timer(&pq->hold_timer, sched_next))
1991 xfrm_pol_hold(pol);
a0073fe1
SK		 1992
1993 spin_unlock_bh(&pq->hold_queue.lock);
1994
1995 return 0;
1996}
1997
1998static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 1999 struct xfrm_flo *xflo,
a0073fe1
SK		 2000 const struct flowi *fl,
2001 int num_xfrms,
2002 u16 family)
2003{
2004 int err;
2005 struct net_device *dev;
b8c203b2 2006 struct dst_entry *dst;
a0073fe1
SK		 2007 struct dst_entry *dst1;
2008 struct xfrm_dst *xdst;
2009
2010 xdst = xfrm_alloc_dst(net, family);
2011 if (IS_ERR(xdst))
2012 return xdst;
2013
b8c203b2
SK		 2014 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2015 net->xfrm.sysctl_larval_drop ||
2016 num_xfrms <= 0)
a0073fe1
SK		 2017 return xdst;
2018
b8c203b2 2019 dst = xflo->dst_orig;
a0073fe1
SK		 2020 dst1 = &xdst->u.dst;
2021 dst_hold(dst);
2022 xdst->route = dst;
2023
2024 dst_copy_metrics(dst1, dst);
2025
2026 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2027 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2028 dst1->lastuse = jiffies;
2029
2030 dst1->input = dst_discard;
2031 dst1->output = xdst_queue_output;
2032
2033 dst_hold(dst);
2034 dst1->child = dst;
2035 dst1->path = dst;
2036
2037 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2038
2039 err = -ENODEV;
2040 dev = dst->dev;
2041 if (!dev)
2042 goto free_dst;
2043
2044 err = xfrm_fill_dst(xdst, dev, fl);
2045 if (err)
2046 goto free_dst;
2047
2048out:
2049 return xdst;
2050
2051free_dst:
2052 dst_release(dst1);
2053 xdst = ERR_PTR(err);
2054 goto out;
2055}
2056
80c802f3 2057static struct flow_cache_object *
dee9f4bc 2058xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 2059 struct flow_cache_object *oldflo, void *ctx)
2060{
b8c203b2 2061 struct xfrm_flo *xflo = (struct xfrm_flo *)ctx;
80c802f3
TT		 2062 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2063 struct xfrm_dst *xdst, *new_xdst;
2064 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
2065
2066 /* Check if the policies from old bundle are usable */
2067 xdst = NULL;
2068 if (oldflo) {
2069 xdst = container_of(oldflo, struct xfrm_dst, flo);
2070 num_pols = xdst->num_pols;
2071 num_xfrms = xdst->num_xfrms;
2072 pol_dead = 0;
2073 for (i = 0; i < num_pols; i++) {
2074 pols[i] = xdst->pols[i];
2075 pol_dead |= pols[i]->walk.dead;
2076 }
2077 if (pol_dead) {
2078 dst_free(&xdst->u.dst);
2079 xdst = NULL;
2080 num_pols = 0;
2081 num_xfrms = 0;
2082 oldflo = NULL;
2083 }
2084 }
2085
2086 /* Resolve policies to use if we couldn't get them from
2087 * previous cache entry */
2088 if (xdst == NULL) {
2089 num_pols = 1;
b5fb82c4
BZ		 2090 pols[0] = __xfrm_policy_lookup(net, fl, family,
2091 flow_to_policy_dir(dir));
80c802f3
TT		 2092 err = xfrm_expand_policies(fl, family, pols,
2093 &num_pols, &num_xfrms);
2094 if (err < 0)
2095 goto inc_error;
2096 if (num_pols == 0)
2097 return NULL;
2098 if (num_xfrms <= 0)
2099 goto make_dummy_bundle;
2100 }
2101
b8c203b2
SK		 2102 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
2103 xflo->dst_orig);
80c802f3
TT		 2104 if (IS_ERR(new_xdst)) {
2105 err = PTR_ERR(new_xdst);
2106 if (err != -EAGAIN)
2107 goto error;
2108 if (oldflo == NULL)
2109 goto make_dummy_bundle;
2110 dst_hold(&xdst->u.dst);
2111 return oldflo;
d809ec89
TT		 2112 } else if (new_xdst == NULL) {
2113 num_xfrms = 0;
2114 if (oldflo == NULL)
2115 goto make_dummy_bundle;
2116 xdst->num_xfrms = 0;
2117 dst_hold(&xdst->u.dst);
2118 return oldflo;
80c802f3
TT		 2119 }
2120
2121 /* Kill the previous bundle */
2122 if (xdst) {
2123 /* The policies were stolen for newly generated bundle */
2124 xdst->num_pols = 0;
2125 dst_free(&xdst->u.dst);
2126 }
2127
2128 /* Flow cache does not have reference, it dst_free()'s,
2129 * but we do need to return one reference for original caller */
2130 dst_hold(&new_xdst->u.dst);
2131 return &new_xdst->flo;
2132
2133make_dummy_bundle:
2134 /* We found policies, but there's no bundles to instantiate:
2135 * either because the policy blocks, has no transformations or
2136 * we could not build template (no xfrm_states).*/
b8c203b2 2137 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2138 if (IS_ERR(xdst)) {
2139 xfrm_pols_put(pols, num_pols);
2140 return ERR_CAST(xdst);
2141 }
2142 xdst->num_pols = num_pols;
2143 xdst->num_xfrms = num_xfrms;
3e94c2dc 2144 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2145
2146 dst_hold(&xdst->u.dst);
2147 return &xdst->flo;
2148
2149inc_error:
2150 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2151error:
2152 if (xdst != NULL)
2153 dst_free(&xdst->u.dst);
2154 else
2155 xfrm_pols_put(pols, num_pols);
2156 return ERR_PTR(err);
2157}
1da177e4 2158
2774c131
DM		 2159static struct dst_entry *make_blackhole(struct net *net, u16 family,
2160 struct dst_entry *dst_orig)
2161{
2162 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2163 struct dst_entry *ret;
2164
2165 if (!afinfo) {
2166 dst_release(dst_orig);
433a1954 2167 return ERR_PTR(-EINVAL);
2774c131
DM		 2168 } else {
2169 ret = afinfo->blackhole_route(net, dst_orig);
2170 }
2171 xfrm_policy_put_afinfo(afinfo);
2172
2173 return ret;
2174}
2175
1da177e4
LT		 2176/* Main function: finds/creates a bundle for given flow.
2177 *
2178 * At the moment we eat a raw IP route. Mostly to speed up lookups
2179 * on interfaces with disabled IPsec.
2180 */
452edd59
DM		 2181struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2182 const struct flowi *fl,
2183 struct sock *sk, int flags)
1da177e4 2184{
4e81bb83 2185 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2186 struct flow_cache_object *flo;
2187 struct xfrm_dst *xdst;
452edd59 2188 struct dst_entry *dst, *route;
80c802f3 2189 u16 family = dst_orig->ops->family;
df71837d 2190 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2191 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2192
80c802f3
TT		 2193 dst = NULL;
2194 xdst = NULL;
2195 route = NULL;
4e81bb83 2196
f7944fb1 2197 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2198 num_pols = 1;
2199 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2200 err = xfrm_expand_policies(fl, family, pols,
2201 &num_pols, &num_xfrms);
2202 if (err < 0)
75b8c133 2203 goto dropdst;
80c802f3
TT		 2204
2205 if (num_pols) {
2206 if (num_xfrms <= 0) {
2207 drop_pols = num_pols;
2208 goto no_transform;
2209 }
2210
2211 xdst = xfrm_resolve_and_create_bundle(
2212 pols, num_pols, fl,
2213 family, dst_orig);
2214 if (IS_ERR(xdst)) {
2215 xfrm_pols_put(pols, num_pols);
2216 err = PTR_ERR(xdst);
2217 goto dropdst;
d809ec89
TT		 2218 } else if (xdst == NULL) {
2219 num_xfrms = 0;
2220 drop_pols = num_pols;
2221 goto no_transform;
80c802f3
TT		 2222 }
2223
b7eea454
SK		 2224 dst_hold(&xdst->u.dst);
2225 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2226 route = xdst->route;
0aa64774 2227 }
3bccfbc7 2228 }
1da177e4 2229
80c802f3 2230 if (xdst == NULL) {
b8c203b2
SK		 2231 struct xfrm_flo xflo;
2232
2233 xflo.dst_orig = dst_orig;
2234 xflo.flags = flags;
2235
1da177e4 2236 /* To accelerate a bit... */
2518c7c2 2237 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2238 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2239 goto nopol;
1da177e4 2240
80c802f3 2241 flo = flow_cache_lookup(net, fl, family, dir,
b8c203b2 2242 xfrm_bundle_lookup, &xflo);
80c802f3
TT		 2243 if (flo == NULL)
2244 goto nopol;
fe1a5f03 2245 if (IS_ERR(flo)) {
80c802f3 2246 err = PTR_ERR(flo);
75b8c133 2247 goto dropdst;
d66e37a9 2248 }
80c802f3
TT		 2249 xdst = container_of(flo, struct xfrm_dst, flo);
2250
2251 num_pols = xdst->num_pols;
2252 num_xfrms = xdst->num_xfrms;
3e94c2dc 2253 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2254 route = xdst->route;
2255 }
2256
2257 dst = &xdst->u.dst;
2258 if (route == NULL && num_xfrms > 0) {
2259 /* The only case when xfrm_bundle_lookup() returns a
2260 * bundle with null route, is when the template could
2261 * not be resolved. It means policies are there, but
2262 * bundle could not be created, since we don't yet
2263 * have the xfrm_state's. We need to wait for KM to
2264 * negotiate new SA's or bail out with error.*/
2265 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2266 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2267 err = -EREMOTE;
2268 goto error;
80c802f3 2269 }
80c802f3 2270
5b8ef341 2271 err = -EAGAIN;
80c802f3
TT		 2272
2273 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2274 goto error;
1da177e4
LT		 2275 }
2276
80c802f3
TT		 2277no_transform:
2278 if (num_pols == 0)
8b7817f3 2279 goto nopol;
1da177e4 2280
80c802f3
TT		 2281 if ((flags & XFRM_LOOKUP_ICMP) &&
2282 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2283 err = -ENOENT;
8b7817f3 2284 goto error;
80c802f3 2285 }
8b7817f3 2286
80c802f3
TT		 2287 for (i = 0; i < num_pols; i++)
2288 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2289
80c802f3 2290 if (num_xfrms < 0) {
1da177e4 2291 /* Prohibit the flow */
59c9940e 2292 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2293 err = -EPERM;
2294 goto error;
80c802f3
TT		 2295 } else if (num_xfrms > 0) {
2296 /* Flow transformed */
80c802f3
TT		 2297 dst_release(dst_orig);
2298 } else {
2299 /* Flow passes untransformed */
2300 dst_release(dst);
452edd59 2301 dst = dst_orig;
1da177e4 2302 }
80c802f3
TT		 2303ok:
2304 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2305 if (dst && dst->xfrm &&
2306 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2307 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2308 return dst;
1da177e4 2309
80c802f3 2310nopol:
452edd59
DM		 2311 if (!(flags & XFRM_LOOKUP_ICMP)) {
2312 dst = dst_orig;
80c802f3 2313 goto ok;
452edd59 2314 }
80c802f3 2315 err = -ENOENT;
1da177e4 2316error:
80c802f3 2317 dst_release(dst);
75b8c133 2318dropdst:
ac37e251
W		 2319 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2320 dst_release(dst_orig);
80c802f3 2321 xfrm_pols_put(pols, drop_pols);
452edd59 2322 return ERR_PTR(err);
1da177e4
LT		 2323}
2324EXPORT_SYMBOL(xfrm_lookup);
2325
f92ee619
SK		 2326/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2327 * Otherwise we may send out blackholed packets.
2328 */
2329struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2330 const struct flowi *fl,
2331 struct sock *sk, int flags)
2332{
b8c203b2 2333 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2334 flags | XFRM_LOOKUP_QUEUE |
2335 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2336
2337 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2338 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2339
2340 return dst;
2341}
2342EXPORT_SYMBOL(xfrm_lookup_route);
2343
df0ba92a 2344static inline int
8f029de2 2345xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2346{
2347 struct xfrm_state *x;
df0ba92a
MN		 2348
2349 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2350 return 0;
2351 x = skb->sp->xvec[idx];
2352 if (!x->type->reject)
2353 return 0;
1ecafede 2354 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2355}
2356
1da177e4
LT		 2357/* When skb is transformed back to its "native" form, we have to
2358 * check policy restrictions. At the moment we make this in maximally
2359 * stupid way. Shame on me. :-) Of course, connected sockets must
2360 * have policy cached at them.
2361 */
2362
2363static inline int
7db454b9 2364xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2365 unsigned short family)
2366{
2367 if (xfrm_state_kern(x))
928ba416 2368 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2369 return x->id.proto == tmpl->id.proto &&
2370 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2371 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2372 x->props.mode == tmpl->mode &&
c5d18e98 2373 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2374 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2375 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2376 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2377}
2378
df0ba92a
MN		 2379/*
2380 * 0 or more than 0 is returned when validation is succeeded (either bypass
2381 * because of optional transport mode, or next index of the mathced secpath
2382 * state with the template.
2383 * -1 is returned when no matching template is found.
2384 * Otherwise "-2 - errored_index" is returned.
2385 */
1da177e4 2386static inline int
22cccb7e 2387xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2388 unsigned short family)
2389{
2390 int idx = start;
2391
2392 if (tmpl->optional) {
7e49e6de 2393 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2394 return start;
2395 } else
2396 start = -1;
2397 for (; idx < sp->len; idx++) {
dbe5b4aa 2398 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2399 return ++idx;
df0ba92a
MN		 2400 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2401 if (start == -1)
2402 start = -2-idx;
1da177e4 2403 break;
df0ba92a 2404 }
1da177e4
LT		 2405 }
2406 return start;
2407}
2408
d5422efe
HX		 2409int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2410 unsigned int family, int reverse)
1da177e4
LT		 2411{
2412 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2413 int err;
1da177e4
LT		 2414
2415 if (unlikely(afinfo == NULL))
2416 return -EAFNOSUPPORT;
2417
d5422efe 2418 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2419 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2420 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2421 return err;
1da177e4 2422}
d5422efe 2423EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2424
9a7386ec 2425static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2426{
2427 for (; k < sp->len; k++) {
df0ba92a 2428 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2429 *idxp = k;
1da177e4 2430 return 1;
df0ba92a 2431 }
1da177e4
LT		 2432 }
2433
2434 return 0;
2435}
2436
a716c119 2437int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2438 unsigned short family)
2439{
f6e1e25d 2440 struct net *net = dev_net(skb->dev);
1da177e4 2441 struct xfrm_policy *pol;
4e81bb83
MN		 2442 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2443 int npols = 0;
2444 int xfrm_nr;
2445 int pi;
d5422efe 2446 int reverse;
1da177e4 2447 struct flowi fl;
d5422efe 2448 u8 fl_dir;
df0ba92a 2449 int xerr_idx = -1;
1da177e4 2450
d5422efe
HX		 2451 reverse = dir & ~XFRM_POLICY_MASK;
2452 dir &= XFRM_POLICY_MASK;
2453 fl_dir = policy_to_flow_dir(dir);
2454
0aa64774 2455 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2456 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2457 return 0;
0aa64774
MN		 2458 }
2459
eb9c7ebe 2460 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2461
2462 /* First, check used SA against their selectors. */
2463 if (skb->sp) {
2464 int i;
2465
9b7a787d 2466 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2467 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2468 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2469 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2470 return 0;
0aa64774 2471 }
1da177e4
LT		 2472 }
2473 }
2474
2475 pol = NULL;
3bccfbc7 2476 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2477 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2478 if (IS_ERR(pol)) {
59c9940e 2479 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2480 return 0;
0aa64774 2481 }
3bccfbc7 2482 }
1da177e4 2483
fe1a5f03
TT		 2484 if (!pol) {
2485 struct flow_cache_object *flo;
2486
2487 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2488 xfrm_policy_lookup, NULL);
2489 if (IS_ERR_OR_NULL(flo))
2490 pol = ERR_CAST(flo);
2491 else
2492 pol = container_of(flo, struct xfrm_policy, flo);
2493 }
1da177e4 2494
0aa64774 2495 if (IS_ERR(pol)) {
59c9940e 2496 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2497 return 0;
0aa64774 2498 }
134b0fc5 2499
df0ba92a 2500 if (!pol) {
d1d9facf 2501 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2502 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2503 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2504 return 0;
2505 }
2506 return 1;
2507 }
1da177e4 2508
9d729f72 2509 pol->curlft.use_time = get_seconds();
1da177e4 2510
4e81bb83 2511 pols[0] = pol;
02d0892f 2512 npols++;
4e81bb83
MN		 2513#ifdef CONFIG_XFRM_SUB_POLICY
2514 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2515 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2516 &fl, family,
2517 XFRM_POLICY_IN);
2518 if (pols[1]) {
0aa64774 2519 if (IS_ERR(pols[1])) {
59c9940e 2520 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2521 return 0;
0aa64774 2522 }
9d729f72 2523 pols[1]->curlft.use_time = get_seconds();
02d0892f 2524 npols++;
4e81bb83
MN		 2525 }
2526 }
2527#endif
2528
1da177e4
LT		 2529 if (pol->action == XFRM_POLICY_ALLOW) {
2530 struct sec_path *sp;
2531 static struct sec_path dummy;
4e81bb83 2532 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2533 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2534 struct xfrm_tmpl **tpp = tp;
2535 int ti = 0;
1da177e4
LT		 2536 int i, k;
2537
2538 if ((sp = skb->sp) == NULL)
2539 sp = &dummy;
2540
4e81bb83
MN		 2541 for (pi = 0; pi < npols; pi++) {
2542 if (pols[pi] != pol &&
0aa64774 2543 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2544 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2545 goto reject;
0aa64774
MN		 2546 }
2547 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2548 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2549 goto reject_error;
0aa64774 2550 }
4e81bb83
MN		 2551 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2552 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2553 }
2554 xfrm_nr = ti;
41a49cc3 2555 if (npols > 1) {
283bc9f3 2556 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2557 tpp = stp;
2558 }
4e81bb83 2559
1da177e4
LT		 2560 /* For each tunnel xfrm, find the first matching tmpl.
2561 * For each tmpl before that, find corresponding xfrm.
2562 * Order is _important_. Later we will implement
2563 * some barriers, but at the moment barriers
2564 * are implied between each two transformations.
2565 */
4e81bb83
MN		 2566 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2567 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2568 if (k < 0) {
d1d9facf
JM		 2569 if (k < -1)
2570 /* "-2 - errored_index" returned */
2571 xerr_idx = -(2+k);
59c9940e 2572 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2573 goto reject;
df0ba92a 2574 }
1da177e4
LT		 2575 }
2576
0aa64774 2577 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2578 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2579 goto reject;
0aa64774 2580 }
1da177e4 2581
4e81bb83 2582 xfrm_pols_put(pols, npols);
1da177e4
LT		 2583 return 1;
2584 }
59c9940e 2585 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2586
2587reject:
df0ba92a 2588 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2589reject_error:
2590 xfrm_pols_put(pols, npols);
1da177e4
LT		 2591 return 0;
2592}
2593EXPORT_SYMBOL(__xfrm_policy_check);
2594
2595int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2596{
99a66657 2597 struct net *net = dev_net(skb->dev);
1da177e4 2598 struct flowi fl;
adf30907 2599 struct dst_entry *dst;
73137147 2600 int res = 1;
1da177e4 2601
0aa64774 2602 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2603 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2604 return 0;
0aa64774 2605 }
1da177e4 2606
fafeeb6c 2607 skb_dst_force(skb);
adf30907 2608
b8c203b2 2609 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2610 if (IS_ERR(dst)) {
73137147 2611 res = 0;
452edd59
DM		 2612 dst = NULL;
2613 }
adf30907
ED		 2614 skb_dst_set(skb, dst);
2615 return res;
1da177e4
LT		 2616}
2617EXPORT_SYMBOL(__xfrm_route_forward);
2618
d49c73c7
DM		 2619/* Optimize later using cookies and generation ids. */
2620
1da177e4
LT		 2621static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2622{
d49c73c7 2623 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2624 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2625 * get validated by dst_ops->check on every use. We do this
2626 * because when a normal route referenced by an XFRM dst is
2627 * obsoleted we do not go looking around for all parent
2628 * referencing XFRM dsts so that we can invalidate them. It
2629 * is just too much work. Instead we make the checks here on
2630 * every use. For example:
d49c73c7
DM		 2631 *
2632 * XFRM dst A --> IPv4 dst X
2633 *
2634 * X is the "xdst->route" of A (X is also the "dst->path" of A
2635 * in this example). If X is marked obsolete, "A" will not
2636 * notice. That's what we are validating here via the
2637 * stale_bundle() check.
2638 *
2639 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2640 * dst which causes it's ->obsolete field to be set to
2641 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2642 * this, we want to force a new route lookup.
399c180a 2643 */
d49c73c7
DM		 2644 if (dst->obsolete < 0 && !stale_bundle(dst))
2645 return dst;
2646
1da177e4
LT		 2647 return NULL;
2648}
2649
2650static int stale_bundle(struct dst_entry *dst)
2651{
12fdb4d3 2652 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2653}
2654
aabc9761 2655void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2656{
1da177e4 2657 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2658 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2659 dev_hold(dst->dev);
1da177e4
LT		 2660 dev_put(dev);
2661 }
2662}
aabc9761 2663EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2664
2665static void xfrm_link_failure(struct sk_buff *skb)
2666{
2667 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2668}
2669
2670static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2671{
2672 if (dst) {
2673 if (dst->obsolete) {
2674 dst_release(dst);
2675 dst = NULL;
2676 }
2677 }
2678 return dst;
2679}
2680
e4c17216 2681void xfrm_garbage_collect(struct net *net)
c0ed1c14 2682{
ca925cf1 2683 flow_cache_flush(net);
c0ed1c14 2684}
e4c17216 2685EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2686
2687static void xfrm_garbage_collect_deferred(struct net *net)
2688{
ca925cf1 2689 flow_cache_flush_deferred(net);
c0ed1c14
SK		 2690}
2691
25ee3286 2692static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2693{
2694 do {
2695 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2696 u32 pmtu, route_mtu_cached;
2697
2698 pmtu = dst_mtu(dst->child);
2699 xdst->child_mtu_cached = pmtu;
2700
2701 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2702
2703 route_mtu_cached = dst_mtu(xdst->route);
2704 xdst->route_mtu_cached = route_mtu_cached;
2705
2706 if (pmtu > route_mtu_cached)
2707 pmtu = route_mtu_cached;
2708
defb3519 2709 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2710 } while ((dst = dst->next));
2711}
2712
1da177e4
LT		 2713/* Check that the bundle accepts the flow and its components are
2714 * still valid.
2715 */
2716
12fdb4d3 2717static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2718{
2719 struct dst_entry *dst = &first->u.dst;
2720 struct xfrm_dst *last;
2721 u32 mtu;
2722
92d63dec 2723 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2724 (dst->dev && !netif_running(dst->dev)))
2725 return 0;
2726
a0073fe1
SK		 2727 if (dst->flags & DST_XFRM_QUEUE)
2728 return 1;
2729
1da177e4
LT		 2730 last = NULL;
2731
2732 do {
2733 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2734
1da177e4
LT		 2735 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2736 return 0;
80c802f3
TT		 2737 if (xdst->xfrm_genid != dst->xfrm->genid)
2738 return 0;
b1312c89
TT		 2739 if (xdst->num_pols > 0 &&
2740 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2741 return 0;
e53820de 2742
1da177e4
LT		 2743 mtu = dst_mtu(dst->child);
2744 if (xdst->child_mtu_cached != mtu) {
2745 last = xdst;
2746 xdst->child_mtu_cached = mtu;
2747 }
2748
92d63dec 2749 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2750 return 0;
2751 mtu = dst_mtu(xdst->route);
2752 if (xdst->route_mtu_cached != mtu) {
2753 last = xdst;
2754 xdst->route_mtu_cached = mtu;
2755 }
2756
2757 dst = dst->child;
2758 } while (dst->xfrm);
2759
2760 if (likely(!last))
2761 return 1;
2762
2763 mtu = last->child_mtu_cached;
2764 for (;;) {
2765 dst = &last->u.dst;
2766
2767 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2768 if (mtu > last->route_mtu_cached)
2769 mtu = last->route_mtu_cached;
defb3519 2770 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2771
2772 if (last == first)
2773 break;
2774
bd0bf076 2775 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2776 last->child_mtu_cached = mtu;
2777 }
2778
2779 return 1;
2780}
2781
0dbaee3b
DM		 2782static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2783{
2784 return dst_metric_advmss(dst->path);
2785}
2786
ebb762f2 2787static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2788{
618f9bc7
SK		 2789 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2790
2791 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2792}
2793
f894cbf8
DM		 2794static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2795 struct sk_buff *skb,
2796 const void *daddr)
d3aaeb38 2797{
f894cbf8 2798 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2799}
2800
1da177e4
LT		 2801int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2802{
d7c7544c 2803 struct net *net;
1da177e4
LT		 2804 int err = 0;
2805 if (unlikely(afinfo == NULL))
2806 return -EINVAL;
2807 if (unlikely(afinfo->family >= NPROTO))
2808 return -EAFNOSUPPORT;
ef8531b6 2809 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4 2810 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
f31e8d4f 2811 err = -EEXIST;
1da177e4
LT		 2812 else {
2813 struct dst_ops *dst_ops = afinfo->dst_ops;
2814 if (likely(dst_ops->kmem_cachep == NULL))
2815 dst_ops->kmem_cachep = xfrm_dst_cache;
2816 if (likely(dst_ops->check == NULL))
2817 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2818 if (likely(dst_ops->default_advmss == NULL))
2819 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2820 if (likely(dst_ops->mtu == NULL))
2821 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2822 if (likely(dst_ops->negative_advice == NULL))
2823 dst_ops->negative_advice = xfrm_negative_advice;
2824 if (likely(dst_ops->link_failure == NULL))
2825 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2826 if (likely(dst_ops->neigh_lookup == NULL))
2827 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2828 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2829 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2830 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2831 }
ef8531b6 2832 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2833
2834 rtnl_lock();
2835 for_each_net(net) {
2836 struct dst_ops *xfrm_dst_ops;
2837
2838 switch (afinfo->family) {
2839 case AF_INET:
2840 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2841 break;
dfd56b8b 2842#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2843 case AF_INET6:
2844 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2845 break;
2846#endif
2847 default:
2848 BUG();
2849 }
2850 *xfrm_dst_ops = *afinfo->dst_ops;
2851 }
2852 rtnl_unlock();
2853
1da177e4
LT		 2854 return err;
2855}
2856EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2857
2858int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2859{
2860 int err = 0;
2861 if (unlikely(afinfo == NULL))
2862 return -EINVAL;
2863 if (unlikely(afinfo->family >= NPROTO))
2864 return -EAFNOSUPPORT;
ef8531b6 2865 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2866 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2867 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2868 err = -EINVAL;
ef8531b6
ED		 2869 else
2870 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2871 NULL);
2872 }
2873 spin_unlock(&xfrm_policy_afinfo_lock);
2874 if (!err) {
2875 struct dst_ops *dst_ops = afinfo->dst_ops;
2876
2877 synchronize_rcu();
2878
2879 dst_ops->kmem_cachep = NULL;
2880 dst_ops->check = NULL;
2881 dst_ops->negative_advice = NULL;
2882 dst_ops->link_failure = NULL;
2883 afinfo->garbage_collect = NULL;
1da177e4 2884 }
1da177e4
LT		 2885 return err;
2886}
2887EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2888
d7c7544c
AD		 2889static void __net_init xfrm_dst_ops_init(struct net *net)
2890{
2891 struct xfrm_policy_afinfo *afinfo;
2892
ef8531b6
ED		 2893 rcu_read_lock();
2894 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2895 if (afinfo)
2896 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2897#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2898 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2899 if (afinfo)
2900 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2901#endif
418a99ac 2902 rcu_read_unlock();
546be240
HX		 2903}
2904
1da177e4
LT		 2905static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2906{
351638e7 2907 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2908
1da177e4
LT		 2909 switch (event) {
2910 case NETDEV_DOWN:
c0ed1c14 2911 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2912 }
2913 return NOTIFY_DONE;
2914}
2915
2916static struct notifier_block xfrm_dev_notifier = {
d5917a35 2917 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2918};
2919
558f82ef 2920#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2921static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2922{
c68cd1a0 2923 int rv;
698365fa
WC		 2924 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2925 if (!net->mib.xfrm_statistics)
558f82ef 2926 return -ENOMEM;
c68cd1a0
AD		 2927 rv = xfrm_proc_init(net);
2928 if (rv < 0)
698365fa 2929 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2930 return rv;
558f82ef 2931}
59c9940e
AD		 2932
2933static void xfrm_statistics_fini(struct net *net)
2934{
c68cd1a0 2935 xfrm_proc_fini(net);
698365fa 2936 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2937}
2938#else
2939static int __net_init xfrm_statistics_init(struct net *net)
2940{
2941 return 0;
2942}
2943
2944static void xfrm_statistics_fini(struct net *net)
2945{
2946}
558f82ef
MN		 2947#endif
2948
d62ddc21 2949static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2950{
2518c7c2
DM		 2951 unsigned int hmask, sz;
2952 int dir;
2953
d62ddc21
AD		 2954 if (net_eq(net, &init_net))
2955 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2956 sizeof(struct xfrm_dst),
e5d679f3 2957 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2958 NULL);
1da177e4 2959
2518c7c2
DM		 2960 hmask = 8 - 1;
2961 sz = (hmask+1) * sizeof(struct hlist_head);
2962
93b851c1
AD		 2963 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2964 if (!net->xfrm.policy_byidx)
2965 goto out_byidx;
8100bea7 2966 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2967
53c2e285 2968 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2969 struct xfrm_policy_hash *htab;
2970
dc2caba7 2971 net->xfrm.policy_count[dir] = 0;
53c2e285 2972 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2973 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2974
a35f6c5d 2975 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2976 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2977 if (!htab->table)
a35f6c5d
AD		 2978 goto out_bydst;
2979 htab->hmask = hmask;
b58555f1
CG		 2980 htab->dbits4 = 32;
2981 htab->sbits4 = 32;
2982 htab->dbits6 = 128;
2983 htab->sbits6 = 128;
2518c7c2 2984 }
880a6fab
CG		 2985 net->xfrm.policy_hthresh.lbits4 = 32;
2986 net->xfrm.policy_hthresh.rbits4 = 32;
2987 net->xfrm.policy_hthresh.lbits6 = 128;
2988 net->xfrm.policy_hthresh.rbits6 = 128;
2989
2990 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2991
adfcf0b2 2992 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2993 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2994 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21
AD		 2995 if (net_eq(net, &init_net))
2996 register_netdevice_notifier(&xfrm_dev_notifier);
2997 return 0;
93b851c1 2998
a35f6c5d
AD		 2999out_bydst:
3000 for (dir--; dir >= 0; dir--) {
3001 struct xfrm_policy_hash *htab;
3002
3003 htab = &net->xfrm.policy_bydst[dir];
3004 xfrm_hash_free(htab->table, sz);
3005 }
3006 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 3007out_byidx:
3008 return -ENOMEM;
d62ddc21
AD		 3009}
3010
3011static void xfrm_policy_fini(struct net *net)
3012{
93b851c1 3013 unsigned int sz;
8b18f8ea 3014 int dir;
93b851c1 3015
7c2776ee
AD		 3016 flush_work(&net->xfrm.policy_hash_work);
3017#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 3018 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 3019#endif
2e71029e 3020 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 3021
adfcf0b2 3022 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 3023
53c2e285 3024 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 3025 struct xfrm_policy_hash *htab;
3026
8b18f8ea 3027 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 3028
3029 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 3030 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 3031 WARN_ON(!hlist_empty(htab->table));
3032 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 3033 }
3034
8100bea7 3035 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 3036 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
3037 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 3038}
3039
d62ddc21
AD		 3040static int __net_init xfrm_net_init(struct net *net)
3041{
3042 int rv;
3043
59c9940e
AD		 3044 rv = xfrm_statistics_init(net);
3045 if (rv < 0)
3046 goto out_statistics;
d62ddc21
AD		 3047 rv = xfrm_state_init(net);
3048 if (rv < 0)
3049 goto out_state;
3050 rv = xfrm_policy_init(net);
3051 if (rv < 0)
3052 goto out_policy;
d7c7544c 3053 xfrm_dst_ops_init(net);
b27aeadb
AD		 3054 rv = xfrm_sysctl_init(net);
3055 if (rv < 0)
3056 goto out_sysctl;
4a93f509
SK		 3057 rv = flow_cache_init(net);
3058 if (rv < 0)
3059 goto out;
283bc9f3
FD		 3060
3061 /* Initialize the per-net locks here */
3062 spin_lock_init(&net->xfrm.xfrm_state_lock);
3063 rwlock_init(&net->xfrm.xfrm_policy_lock);
283bc9f3
FD		 3064 mutex_init(&net->xfrm.xfrm_cfg_mutex);
3065
d62ddc21
AD		 3066 return 0;
3067
4a93f509
SK		 3068out:
3069 xfrm_sysctl_fini(net);
b27aeadb
AD		 3070out_sysctl:
3071 xfrm_policy_fini(net);
d62ddc21
AD		 3072out_policy:
3073 xfrm_state_fini(net);
3074out_state:
59c9940e
AD		 3075 xfrm_statistics_fini(net);
3076out_statistics:
d62ddc21
AD		 3077 return rv;
3078}
3079
3080static void __net_exit xfrm_net_exit(struct net *net)
3081{
4a93f509 3082 flow_cache_fini(net);
b27aeadb 3083 xfrm_sysctl_fini(net);
d62ddc21
AD		 3084 xfrm_policy_fini(net);
3085 xfrm_state_fini(net);
59c9940e 3086 xfrm_statistics_fini(net);
d62ddc21
AD		 3087}
3088
3089static struct pernet_operations __net_initdata xfrm_net_ops = {
3090 .init = xfrm_net_init,
3091 .exit = xfrm_net_exit,
3092};
3093
1da177e4
LT		 3094void __init xfrm_init(void)
3095{
d62ddc21 3096 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 3097 xfrm_input_init();
3098}
3099
ab5f5e8b 3100#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 3101static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
3102 struct audit_buffer *audit_buf)
ab5f5e8b 3103{
875179fa
PM		 3104 struct xfrm_sec_ctx *ctx = xp->security;
3105 struct xfrm_selector *sel = &xp->selector;
3106
3107 if (ctx)
ab5f5e8b 3108 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3109 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3110
9b7a787d 3111 switch (sel->family) {
ab5f5e8b 3112 case AF_INET:
21454aaa 3113 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3114 if (sel->prefixlen_s != 32)
3115 audit_log_format(audit_buf, " src_prefixlen=%d",
3116 sel->prefixlen_s);
21454aaa 3117 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3118 if (sel->prefixlen_d != 32)
3119 audit_log_format(audit_buf, " dst_prefixlen=%d",
3120 sel->prefixlen_d);
ab5f5e8b
JL		 3121 break;
3122 case AF_INET6:
5b095d98 3123 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3124 if (sel->prefixlen_s != 128)
3125 audit_log_format(audit_buf, " src_prefixlen=%d",
3126 sel->prefixlen_s);
5b095d98 3127 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3128 if (sel->prefixlen_d != 128)
3129 audit_log_format(audit_buf, " dst_prefixlen=%d",
3130 sel->prefixlen_d);
ab5f5e8b
JL		 3131 break;
3132 }
3133}
3134
2e71029e 3135void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3136{
3137 struct audit_buffer *audit_buf;
ab5f5e8b 3138
afeb14b4 3139 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3140 if (audit_buf == NULL)
3141 return;
2e71029e 3142 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3143 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3144 xfrm_audit_common_policyinfo(xp, audit_buf);
3145 audit_log_end(audit_buf);
3146}
3147EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3148
68277acc 3149void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3150 bool task_valid)
ab5f5e8b
JL		 3151{
3152 struct audit_buffer *audit_buf;
ab5f5e8b 3153
afeb14b4 3154 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3155 if (audit_buf == NULL)
3156 return;
2e71029e 3157 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3158 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3159 xfrm_audit_common_policyinfo(xp, audit_buf);
3160 audit_log_end(audit_buf);
3161}
3162EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3163#endif
3164
80c9abaa 3165#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3166static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3167 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3168{
3169 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3170 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3171 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3172 sel_cmp->family) &&
3173 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3174 sel_cmp->family) &&
80c9abaa
SS		 3175 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3176 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3177 return true;
80c9abaa
SS		 3178 }
3179 } else {
3180 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3181 return true;
80c9abaa
SS		 3182 }
3183 }
bc9b35ad 3184 return false;
80c9abaa
SS		 3185}
3186
3e94c2dc
WC		 3187static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3188 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3189{
3190 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3191 struct hlist_head *chain;
3192 u32 priority = ~0U;
3193
283bc9f3 3194 read_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME*/
8d549c4f 3195 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3196 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3197 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3198 pol->type == type) {
3199 ret = pol;
3200 priority = ret->priority;
3201 break;
3202 }
3203 }
8d549c4f 3204 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3205 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3206 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3207 pol->type == type &&
3208 pol->priority < priority) {
3209 ret = pol;
3210 break;
3211 }
3212 }
3213
586f2eb4 3214 xfrm_pol_hold(ret);
80c9abaa 3215
283bc9f3 3216 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3217
3218 return ret;
3219}
3220
dd701754 3221static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3222{
3223 int match = 0;
3224
3225 if (t->mode == m->mode && t->id.proto == m->proto &&
3226 (m->reqid == 0 || t->reqid == m->reqid)) {
3227 switch (t->mode) {
3228 case XFRM_MODE_TUNNEL:
3229 case XFRM_MODE_BEET:
70e94e66
YH		 3230 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3231 m->old_family) &&
3232 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3233 m->old_family)) {
80c9abaa
SS		 3234 match = 1;
3235 }
3236 break;
3237 case XFRM_MODE_TRANSPORT:
3238 /* in case of transport mode, template does not store
3239 any IP addresses, hence we just compare mode and
3240 protocol */
3241 match = 1;
3242 break;
3243 default:
3244 break;
3245 }
3246 }
3247 return match;
3248}
3249
3250/* update endpoint address(es) of template(s) */
3251static int xfrm_policy_migrate(struct xfrm_policy *pol,
3252 struct xfrm_migrate *m, int num_migrate)
3253{
3254 struct xfrm_migrate *mp;
80c9abaa
SS		 3255 int i, j, n = 0;
3256
3257 write_lock_bh(&pol->lock);
12a169e7 3258 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3259 /* target policy has been deleted */
3260 write_unlock_bh(&pol->lock);
3261 return -ENOENT;
3262 }
3263
3264 for (i = 0; i < pol->xfrm_nr; i++) {
3265 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3266 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3267 continue;
3268 n++;
1bfcb10f
HX		 3269 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3270 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3271 continue;
3272 /* update endpoints */
3273 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3274 sizeof(pol->xfrm_vec[i].id.daddr));
3275 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3276 sizeof(pol->xfrm_vec[i].saddr));
3277 pol->xfrm_vec[i].encap_family = mp->new_family;
3278 /* flush bundles */
80c802f3 3279 atomic_inc(&pol->genid);
80c9abaa
SS		 3280 }
3281 }
3282
3283 write_unlock_bh(&pol->lock);
3284
3285 if (!n)
3286 return -ENODATA;
3287
3288 return 0;
3289}
3290
dd701754 3291static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3292{
3293 int i, j;
3294
3295 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3296 return -EINVAL;
3297
3298 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3299 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3300 m[i].old_family) &&
3301 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3302 m[i].old_family))
80c9abaa
SS		 3303 return -EINVAL;
3304 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3305 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3306 return -EINVAL;
3307
3308 /* check if there is any duplicated entry */
3309 for (j = i + 1; j < num_migrate; j++) {
3310 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3311 sizeof(m[i].old_daddr)) &&
3312 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3313 sizeof(m[i].old_saddr)) &&
3314 m[i].proto == m[j].proto &&
3315 m[i].mode == m[j].mode &&
3316 m[i].reqid == m[j].reqid &&
3317 m[i].old_family == m[j].old_family)
3318 return -EINVAL;
3319 }
3320 }
3321
3322 return 0;
3323}
3324
b4b7c0b3 3325int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3326 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3327 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3328{
3329 int i, err, nx_cur = 0, nx_new = 0;
3330 struct xfrm_policy *pol = NULL;
3331 struct xfrm_state *x, *xc;
3332 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3333 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3334 struct xfrm_migrate *mp;
3335
3336 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3337 goto out;
3338
3339 /* Stage 1 - find policy */
8d549c4f 3340 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3341 err = -ENOENT;
3342 goto out;
3343 }
3344
3345 /* Stage 2 - find and update state(s) */
3346 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3347 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3348 x_cur[nx_cur] = x;
3349 nx_cur++;
3350 if ((xc = xfrm_state_migrate(x, mp))) {
3351 x_new[nx_new] = xc;
3352 nx_new++;
3353 } else {
3354 err = -ENODATA;
3355 goto restore_state;
3356 }
3357 }
3358 }
3359
3360 /* Stage 3 - update policy */
3361 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3362 goto restore_state;
3363
3364 /* Stage 4 - delete old state(s) */
3365 if (nx_cur) {
3366 xfrm_states_put(x_cur, nx_cur);
3367 xfrm_states_delete(x_cur, nx_cur);
3368 }
3369
3370 /* Stage 5 - announce */
13c1d189 3371 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3372
3373 xfrm_pol_put(pol);
3374
3375 return 0;
3376out:
3377 return err;
3378
3379restore_state:
3380 if (pol)
3381 xfrm_pol_put(pol);
3382 if (nx_cur)
3383 xfrm_states_put(x_cur, nx_cur);
3384 if (nx_new)
3385 xfrm_states_delete(x_new, nx_new);
3386
3387 return err;
3388}
e610e679 3389EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3390#endif
Ubuntu Artful kernel mirror