]> 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: Using the right namespace to migrate key info
[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
4a3e2f71
AV		 42DEFINE_MUTEX(xfrm_cfg_mutex);
43EXPORT_SYMBOL(xfrm_cfg_mutex);
1da177e4 44
80c802f3
TT		 45static DEFINE_SPINLOCK(xfrm_policy_sk_bundle_lock);
46static struct dst_entry *xfrm_policy_sk_bundles;
1da177e4
LT		 47static DEFINE_RWLOCK(xfrm_policy_lock);
48
418a99ac
PJ		 49static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
50static struct xfrm_policy_afinfo __rcu *xfrm_policy_afinfo[NPROTO]
51 __read_mostly;
1da177e4 52
e18b890b 53static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 54
25ee3286 55static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 56static int stale_bundle(struct dst_entry *dst);
12fdb4d3 57static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 58static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 59
29fa0b30
WY		 60static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
61 int dir);
62
bc9b35ad 63static inline bool
200ce96e 64__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 65{
7e1dc7b6
DM		 66 const struct flowi4 *fl4 = &fl->u.ip4;
67
26bff940
AD		 68 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
69 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 70 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
71 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
72 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
73 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 74}
75
bc9b35ad 76static inline bool
200ce96e 77__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 78{
7e1dc7b6
DM		 79 const struct flowi6 *fl6 = &fl->u.ip6;
80
81 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
82 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
83 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
84 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
85 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
86 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 87}
88
bc9b35ad
DM		 89bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
90 unsigned short family)
77681021
AM		 91{
92 switch (family) {
93 case AF_INET:
94 return __xfrm4_selector_match(sel, fl);
95 case AF_INET6:
96 return __xfrm6_selector_match(sel, fl);
97 }
bc9b35ad 98 return false;
77681021
AM		 99}
100
ef8531b6
ED		 101static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
102{
103 struct xfrm_policy_afinfo *afinfo;
104
105 if (unlikely(family >= NPROTO))
106 return NULL;
107 rcu_read_lock();
108 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
109 if (unlikely(!afinfo))
110 rcu_read_unlock();
111 return afinfo;
112}
113
114static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
115{
116 rcu_read_unlock();
117}
118
c5b3cf46 119static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
6418c4e0
DM		 120 const xfrm_address_t *saddr,
121 const xfrm_address_t *daddr,
9bb182a7
YH		 122 int family)
123{
124 struct xfrm_policy_afinfo *afinfo;
125 struct dst_entry *dst;
126
127 afinfo = xfrm_policy_get_afinfo(family);
128 if (unlikely(afinfo == NULL))
129 return ERR_PTR(-EAFNOSUPPORT);
130
c5b3cf46 131 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 132
133 xfrm_policy_put_afinfo(afinfo);
134
135 return dst;
136}
137
25ee3286 138static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 139 xfrm_address_t *prev_saddr,
140 xfrm_address_t *prev_daddr,
25ee3286 141 int family)
1da177e4 142{
c5b3cf46 143 struct net *net = xs_net(x);
66cdb3ca
HX		 144 xfrm_address_t *saddr = &x->props.saddr;
145 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 146 struct dst_entry *dst;
147
9bb182a7 148 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 149 saddr = x->coaddr;
9bb182a7
YH		 150 daddr = prev_daddr;
151 }
152 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
153 saddr = prev_saddr;
66cdb3ca 154 daddr = x->coaddr;
9bb182a7 155 }
1da177e4 156
c5b3cf46 157 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 158
159 if (!IS_ERR(dst)) {
160 if (prev_saddr != saddr)
161 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
162 if (prev_daddr != daddr)
163 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
164 }
1da177e4 165
66cdb3ca 166 return dst;
1da177e4 167}
1da177e4 168
1da177e4
LT		 169static inline unsigned long make_jiffies(long secs)
170{
171 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
172 return MAX_SCHEDULE_TIMEOUT-1;
173 else
a716c119 174 return secs*HZ;
1da177e4
LT		 175}
176
177static void xfrm_policy_timer(unsigned long data)
178{
179 struct xfrm_policy *xp = (struct xfrm_policy*)data;
9d729f72 180 unsigned long now = get_seconds();
1da177e4
LT		 181 long next = LONG_MAX;
182 int warn = 0;
183 int dir;
184
185 read_lock(&xp->lock);
186
ea2dea9d 187 if (unlikely(xp->walk.dead))
1da177e4
LT		 188 goto out;
189
77d8d7a6 190 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 191
192 if (xp->lft.hard_add_expires_seconds) {
193 long tmo = xp->lft.hard_add_expires_seconds +
194 xp->curlft.add_time - now;
195 if (tmo <= 0)
196 goto expired;
197 if (tmo < next)
198 next = tmo;
199 }
200 if (xp->lft.hard_use_expires_seconds) {
201 long tmo = xp->lft.hard_use_expires_seconds +
202 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
203 if (tmo <= 0)
204 goto expired;
205 if (tmo < next)
206 next = tmo;
207 }
208 if (xp->lft.soft_add_expires_seconds) {
209 long tmo = xp->lft.soft_add_expires_seconds +
210 xp->curlft.add_time - now;
211 if (tmo <= 0) {
212 warn = 1;
213 tmo = XFRM_KM_TIMEOUT;
214 }
215 if (tmo < next)
216 next = tmo;
217 }
218 if (xp->lft.soft_use_expires_seconds) {
219 long tmo = xp->lft.soft_use_expires_seconds +
220 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
221 if (tmo <= 0) {
222 warn = 1;
223 tmo = XFRM_KM_TIMEOUT;
224 }
225 if (tmo < next)
226 next = tmo;
227 }
228
229 if (warn)
6c5c8ca7 230 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 231 if (next != LONG_MAX &&
232 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
233 xfrm_pol_hold(xp);
234
235out:
236 read_unlock(&xp->lock);
237 xfrm_pol_put(xp);
238 return;
239
240expired:
241 read_unlock(&xp->lock);
4666faab 242 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 243 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 244 xfrm_pol_put(xp);
245}
246
fe1a5f03
TT		 247static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
248{
249 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
250
251 if (unlikely(pol->walk.dead))
252 flo = NULL;
253 else
254 xfrm_pol_hold(pol);
255
256 return flo;
257}
258
259static int xfrm_policy_flo_check(struct flow_cache_object *flo)
260{
261 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
262
263 return !pol->walk.dead;
264}
265
266static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
267{
268 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
269}
270
271static const struct flow_cache_ops xfrm_policy_fc_ops = {
272 .get = xfrm_policy_flo_get,
273 .check = xfrm_policy_flo_check,
274 .delete = xfrm_policy_flo_delete,
275};
1da177e4
LT		 276
277/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
278 * SPD calls.
279 */
280
0331b1f3 281struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 282{
283 struct xfrm_policy *policy;
284
0da974f4 285 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 286
287 if (policy) {
0331b1f3 288 write_pnet(&policy->xp_net, net);
12a169e7 289 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 290 INIT_HLIST_NODE(&policy->bydst);
291 INIT_HLIST_NODE(&policy->byidx);
1da177e4 292 rwlock_init(&policy->lock);
2518c7c2 293 atomic_set(&policy->refcnt, 1);
a0073fe1 294 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 295 setup_timer(&policy->timer, xfrm_policy_timer,
296 (unsigned long)policy);
a0073fe1
SK		 297 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
298 (unsigned long)policy);
fe1a5f03 299 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 300 }
301 return policy;
302}
303EXPORT_SYMBOL(xfrm_policy_alloc);
304
305/* Destroy xfrm_policy: descendant resources must be released to this moment. */
306
64c31b3f 307void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 308{
12a169e7 309 BUG_ON(!policy->walk.dead);
1da177e4 310
0659eea9 311 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 312 BUG();
313
03e1ad7b 314 security_xfrm_policy_free(policy->security);
1da177e4
LT		 315 kfree(policy);
316}
64c31b3f 317EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 318
a0073fe1
SK		 319static void xfrm_queue_purge(struct sk_buff_head *list)
320{
321 struct sk_buff *skb;
322
302a50bc 323 while ((skb = skb_dequeue(list)) != NULL)
a0073fe1 324 kfree_skb(skb);
a0073fe1
SK		 325}
326
1da177e4
LT		 327/* Rule must be locked. Release descentant resources, announce
328 * entry dead. The rule must be unlinked from lists to the moment.
329 */
330
331static void xfrm_policy_kill(struct xfrm_policy *policy)
332{
12a169e7 333 policy->walk.dead = 1;
1da177e4 334
285ead17 335 atomic_inc(&policy->genid);
1da177e4 336
e7d8f6cb
SK		 337 if (del_timer(&policy->polq.hold_timer))
338 xfrm_pol_put(policy);
a0073fe1
SK		 339 xfrm_queue_purge(&policy->polq.hold_queue);
340
285ead17
TT		 341 if (del_timer(&policy->timer))
342 xfrm_pol_put(policy);
343
344 xfrm_pol_put(policy);
1da177e4
LT		 345}
346
2518c7c2
DM		 347static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
348
e92303f8 349static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 350{
e92303f8 351 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 352}
353
5f803b58
DM		 354static struct hlist_head *policy_hash_bysel(struct net *net,
355 const struct xfrm_selector *sel,
356 unsigned short family, int dir)
2518c7c2 357{
1121994c 358 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 359 unsigned int hash = __sel_hash(sel, family, hmask);
360
361 return (hash == hmask + 1 ?
1121994c
AD		 362 &net->xfrm.policy_inexact[dir] :
363 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 364}
365
5f803b58
DM		 366static struct hlist_head *policy_hash_direct(struct net *net,
367 const xfrm_address_t *daddr,
368 const xfrm_address_t *saddr,
369 unsigned short family, int dir)
2518c7c2 370{
1121994c 371 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 372 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
373
1121994c 374 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 375}
376
2518c7c2
DM		 377static void xfrm_dst_hash_transfer(struct hlist_head *list,
378 struct hlist_head *ndsttable,
379 unsigned int nhashmask)
380{
b67bfe0d 381 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 382 struct xfrm_policy *pol;
b791160b 383 unsigned int h0 = 0;
2518c7c2 384
b791160b 385redo:
b67bfe0d 386 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 387 unsigned int h;
388
389 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
390 pol->family, nhashmask);
b791160b 391 if (!entry0) {
b67bfe0d 392 hlist_del(&pol->bydst);
b791160b
YH		 393 hlist_add_head(&pol->bydst, ndsttable+h);
394 h0 = h;
395 } else {
396 if (h != h0)
397 continue;
b67bfe0d 398 hlist_del(&pol->bydst);
b791160b
YH		 399 hlist_add_after(entry0, &pol->bydst);
400 }
b67bfe0d 401 entry0 = &pol->bydst;
b791160b
YH		 402 }
403 if (!hlist_empty(list)) {
404 entry0 = NULL;
405 goto redo;
2518c7c2
DM		 406 }
407}
408
409static void xfrm_idx_hash_transfer(struct hlist_head *list,
410 struct hlist_head *nidxtable,
411 unsigned int nhashmask)
412{
b67bfe0d 413 struct hlist_node *tmp;
2518c7c2
DM		 414 struct xfrm_policy *pol;
415
b67bfe0d 416 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 417 unsigned int h;
418
419 h = __idx_hash(pol->index, nhashmask);
420 hlist_add_head(&pol->byidx, nidxtable+h);
421 }
422}
423
424static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
425{
426 return ((old_hmask + 1) << 1) - 1;
427}
428
66caf628 429static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 430{
66caf628 431 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 432 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
433 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 434 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 435 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 436 int i;
437
438 if (!ndst)
439 return;
440
441 write_lock_bh(&xfrm_policy_lock);
442
443 for (i = hmask; i >= 0; i--)
444 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
445
66caf628
AD		 446 net->xfrm.policy_bydst[dir].table = ndst;
447 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2
DM		 448
449 write_unlock_bh(&xfrm_policy_lock);
450
44e36b42 451 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 452}
453
66caf628 454static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 455{
66caf628 456 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 457 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
458 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 459 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 460 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 461 int i;
462
463 if (!nidx)
464 return;
465
466 write_lock_bh(&xfrm_policy_lock);
467
468 for (i = hmask; i >= 0; i--)
469 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
470
66caf628
AD		 471 net->xfrm.policy_byidx = nidx;
472 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2
DM		 473
474 write_unlock_bh(&xfrm_policy_lock);
475
44e36b42 476 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 477}
478
66caf628 479static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 480{
66caf628
AD		 481 unsigned int cnt = net->xfrm.policy_count[dir];
482 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 483
484 if (total)
485 *total += cnt;
486
487 if ((hmask + 1) < xfrm_policy_hashmax &&
488 cnt > hmask)
489 return 1;
490
491 return 0;
492}
493
66caf628 494static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 495{
66caf628 496 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 497
498 if ((hmask + 1) < xfrm_policy_hashmax &&
499 total > hmask)
500 return 1;
501
502 return 0;
503}
504
e071041b 505void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18
JHS		 506{
507 read_lock_bh(&xfrm_policy_lock);
e071041b
AD		 508 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
509 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
510 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
511 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
512 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
513 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
514 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18
JHS		 515 si->spdhmcnt = xfrm_policy_hashmax;
516 read_unlock_bh(&xfrm_policy_lock);
517}
518EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 519
ecfd6b18 520static DEFINE_MUTEX(hash_resize_mutex);
66caf628 521static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 522{
66caf628 523 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 524 int dir, total;
525
526 mutex_lock(&hash_resize_mutex);
527
528 total = 0;
529 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 530 if (xfrm_bydst_should_resize(net, dir, &total))
531 xfrm_bydst_resize(net, dir);
2518c7c2 532 }
66caf628
AD		 533 if (xfrm_byidx_should_resize(net, total))
534 xfrm_byidx_resize(net, total);
2518c7c2
DM		 535
536 mutex_unlock(&hash_resize_mutex);
537}
538
1da177e4
LT		 539/* Generate new index... KAME seems to generate them ordered by cost
540 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 541static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 542{
1da177e4
LT		 543 static u32 idx_generator;
544
545 for (;;) {
2518c7c2
DM		 546 struct hlist_head *list;
547 struct xfrm_policy *p;
548 u32 idx;
549 int found;
550
e682adf0
FD		 551 if (!index) {
552 idx = (idx_generator | dir);
553 idx_generator += 8;
554 } else {
555 idx = index;
556 index = 0;
557 }
558
1da177e4
LT		 559 if (idx == 0)
560 idx = 8;
1121994c 561 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 562 found = 0;
b67bfe0d 563 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 564 if (p->index == idx) {
565 found = 1;
1da177e4 566 break;
2518c7c2 567 }
1da177e4 568 }
2518c7c2 569 if (!found)
1da177e4
LT		 570 return idx;
571 }
572}
573
2518c7c2
DM		 574static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
575{
576 u32 *p1 = (u32 *) s1;
577 u32 *p2 = (u32 *) s2;
578 int len = sizeof(struct xfrm_selector) / sizeof(u32);
579 int i;
580
581 for (i = 0; i < len; i++) {
582 if (p1[i] != p2[i])
583 return 1;
584 }
585
586 return 0;
587}
588
a0073fe1
SK		 589static void xfrm_policy_requeue(struct xfrm_policy *old,
590 struct xfrm_policy *new)
591{
592 struct xfrm_policy_queue *pq = &old->polq;
593 struct sk_buff_head list;
594
595 __skb_queue_head_init(&list);
596
597 spin_lock_bh(&pq->hold_queue.lock);
598 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 599 if (del_timer(&pq->hold_timer))
600 xfrm_pol_put(old);
a0073fe1
SK		 601 spin_unlock_bh(&pq->hold_queue.lock);
602
603 if (skb_queue_empty(&list))
604 return;
605
606 pq = &new->polq;
607
608 spin_lock_bh(&pq->hold_queue.lock);
609 skb_queue_splice(&list, &pq->hold_queue);
610 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 611 if (!mod_timer(&pq->hold_timer, jiffies))
612 xfrm_pol_hold(new);
a0073fe1
SK		 613 spin_unlock_bh(&pq->hold_queue.lock);
614}
615
7cb8a939
SK		 616static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
617 struct xfrm_policy *pol)
618{
619 u32 mark = policy->mark.v & policy->mark.m;
620
621 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
622 return true;
623
624 if ((mark & pol->mark.m) == pol->mark.v &&
625 policy->priority == pol->priority)
626 return true;
627
628 return false;
629}
630
1da177e4
LT		 631int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
632{
1121994c 633 struct net *net = xp_net(policy);
2518c7c2
DM		 634 struct xfrm_policy *pol;
635 struct xfrm_policy *delpol;
636 struct hlist_head *chain;
b67bfe0d 637 struct hlist_node *newpos;
1da177e4
LT		 638
639 write_lock_bh(&xfrm_policy_lock);
1121994c 640 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 641 delpol = NULL;
642 newpos = NULL;
b67bfe0d 643 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 644 if (pol->type == policy->type &&
2518c7c2 645 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 646 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 647 xfrm_sec_ctx_match(pol->security, policy->security) &&
648 !WARN_ON(delpol)) {
1da177e4
LT		 649 if (excl) {
650 write_unlock_bh(&xfrm_policy_lock);
651 return -EEXIST;
652 }
1da177e4
LT		 653 delpol = pol;
654 if (policy->priority > pol->priority)
655 continue;
656 } else if (policy->priority >= pol->priority) {
a6c7ab55 657 newpos = &pol->bydst;
1da177e4
LT		 658 continue;
659 }
1da177e4
LT		 660 if (delpol)
661 break;
1da177e4
LT		 662 }
663 if (newpos)
2518c7c2
DM		 664 hlist_add_after(newpos, &policy->bydst);
665 else
666 hlist_add_head(&policy->bydst, chain);
1da177e4 667 xfrm_pol_hold(policy);
1121994c 668 net->xfrm.policy_count[dir]++;
1da177e4 669 atomic_inc(&flow_cache_genid);
ca4c3fc2 670
671 /* After previous checking, family can either be AF_INET or AF_INET6 */
672 if (policy->family == AF_INET)
673 rt_genid_bump_ipv4(net);
674 else
675 rt_genid_bump_ipv6(net);
676
a0073fe1
SK		 677 if (delpol) {
678 xfrm_policy_requeue(delpol, policy);
29fa0b30 679 __xfrm_policy_unlink(delpol, dir);
a0073fe1 680 }
e682adf0 681 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 682 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 683 policy->curlft.add_time = get_seconds();
1da177e4
LT		 684 policy->curlft.use_time = 0;
685 if (!mod_timer(&policy->timer, jiffies + HZ))
686 xfrm_pol_hold(policy);
1121994c 687 list_add(&policy->walk.all, &net->xfrm.policy_all);
1da177e4
LT		 688 write_unlock_bh(&xfrm_policy_lock);
689
9b78a82c 690 if (delpol)
1da177e4 691 xfrm_policy_kill(delpol);
1121994c
AD		 692 else if (xfrm_bydst_should_resize(net, dir, NULL))
693 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 694
1da177e4
LT		 695 return 0;
696}
697EXPORT_SYMBOL(xfrm_policy_insert);
698
8ca2e93b
JHS		 699struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
700 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 701 struct xfrm_sec_ctx *ctx, int delete,
702 int *err)
1da177e4 703{
2518c7c2
DM		 704 struct xfrm_policy *pol, *ret;
705 struct hlist_head *chain;
1da177e4 706
ef41aaa0 707 *err = 0;
1da177e4 708 write_lock_bh(&xfrm_policy_lock);
8d1211a6 709 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 710 ret = NULL;
b67bfe0d 711 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 712 if (pol->type == type &&
34f8d884 713 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 714 !selector_cmp(sel, &pol->selector) &&
715 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 716 xfrm_pol_hold(pol);
2518c7c2 717 if (delete) {
03e1ad7b
PM		 718 *err = security_xfrm_policy_delete(
719 pol->security);
ef41aaa0
EP		 720 if (*err) {
721 write_unlock_bh(&xfrm_policy_lock);
722 return pol;
723 }
29fa0b30 724 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 725 }
726 ret = pol;
1da177e4
LT		 727 break;
728 }
729 }
730 write_unlock_bh(&xfrm_policy_lock);
731
fe1a5f03 732 if (ret && delete)
2518c7c2 733 xfrm_policy_kill(ret);
2518c7c2 734 return ret;
1da177e4 735}
df71837d 736EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 737
8ca2e93b
JHS		 738struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
739 int dir, u32 id, int delete, int *err)
1da177e4 740{
2518c7c2
DM		 741 struct xfrm_policy *pol, *ret;
742 struct hlist_head *chain;
1da177e4 743
b5505c6e
HX		 744 *err = -ENOENT;
745 if (xfrm_policy_id2dir(id) != dir)
746 return NULL;
747
ef41aaa0 748 *err = 0;
1da177e4 749 write_lock_bh(&xfrm_policy_lock);
8d1211a6 750 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 751 ret = NULL;
b67bfe0d 752 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 753 if (pol->type == type && pol->index == id &&
754 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 755 xfrm_pol_hold(pol);
2518c7c2 756 if (delete) {
03e1ad7b
PM		 757 *err = security_xfrm_policy_delete(
758 pol->security);
ef41aaa0
EP		 759 if (*err) {
760 write_unlock_bh(&xfrm_policy_lock);
761 return pol;
762 }
29fa0b30 763 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 764 }
765 ret = pol;
1da177e4
LT		 766 break;
767 }
768 }
769 write_unlock_bh(&xfrm_policy_lock);
770
fe1a5f03 771 if (ret && delete)
2518c7c2 772 xfrm_policy_kill(ret);
2518c7c2 773 return ret;
1da177e4
LT		 774}
775EXPORT_SYMBOL(xfrm_policy_byid);
776
4aa2e62c
JL		 777#ifdef CONFIG_SECURITY_NETWORK_XFRM
778static inline int
33ffbbd5 779xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
1da177e4 780{
4aa2e62c
JL		 781 int dir, err = 0;
782
783 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
784 struct xfrm_policy *pol;
4aa2e62c
JL		 785 int i;
786
b67bfe0d 787 hlist_for_each_entry(pol,
33ffbbd5 788 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 789 if (pol->type != type)
790 continue;
03e1ad7b 791 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 792 if (err) {
ab5f5e8b
JL		 793 xfrm_audit_policy_delete(pol, 0,
794 audit_info->loginuid,
2532386f 795 audit_info->sessionid,
ab5f5e8b 796 audit_info->secid);
4aa2e62c
JL		 797 return err;
798 }
7dc12d6d 799 }
33ffbbd5 800 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 801 hlist_for_each_entry(pol,
33ffbbd5 802 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 803 bydst) {
804 if (pol->type != type)
805 continue;
03e1ad7b
PM		 806 err = security_xfrm_policy_delete(
807 pol->security);
4aa2e62c 808 if (err) {
ab5f5e8b
JL		 809 xfrm_audit_policy_delete(pol, 0,
810 audit_info->loginuid,
2532386f 811 audit_info->sessionid,
ab5f5e8b 812 audit_info->secid);
4aa2e62c
JL		 813 return err;
814 }
815 }
816 }
817 }
818 return err;
819}
820#else
821static inline int
33ffbbd5 822xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c
JL		 823{
824 return 0;
825}
826#endif
827
33ffbbd5 828int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c 829{
2f1eb65f 830 int dir, err = 0, cnt = 0;
1da177e4
LT		 831
832 write_lock_bh(&xfrm_policy_lock);
4aa2e62c 833
33ffbbd5 834 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
4aa2e62c
JL		 835 if (err)
836 goto out;
837
1da177e4 838 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 839 struct xfrm_policy *pol;
29fa0b30 840 int i;
2518c7c2
DM		 841
842 again1:
b67bfe0d 843 hlist_for_each_entry(pol,
33ffbbd5 844 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 845 if (pol->type != type)
846 continue;
ea2dea9d 847 __xfrm_policy_unlink(pol, dir);
1da177e4 848 write_unlock_bh(&xfrm_policy_lock);
ea2dea9d 849 cnt++;
1da177e4 850
ab5f5e8b 851 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
2532386f 852 audit_info->sessionid,
ab5f5e8b 853 audit_info->secid);
161a09e7 854
2518c7c2 855 xfrm_policy_kill(pol);
1da177e4
LT		 856
857 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 858 goto again1;
859 }
860
33ffbbd5 861 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 862 again2:
b67bfe0d 863 hlist_for_each_entry(pol,
33ffbbd5 864 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 865 bydst) {
866 if (pol->type != type)
867 continue;
ea2dea9d 868 __xfrm_policy_unlink(pol, dir);
2518c7c2 869 write_unlock_bh(&xfrm_policy_lock);
ea2dea9d 870 cnt++;
2518c7c2 871
ab5f5e8b
JL		 872 xfrm_audit_policy_delete(pol, 1,
873 audit_info->loginuid,
2532386f 874 audit_info->sessionid,
ab5f5e8b 875 audit_info->secid);
2518c7c2
DM		 876 xfrm_policy_kill(pol);
877
878 write_lock_bh(&xfrm_policy_lock);
879 goto again2;
880 }
1da177e4 881 }
2518c7c2 882
1da177e4 883 }
2f1eb65f
JHS		 884 if (!cnt)
885 err = -ESRCH;
4aa2e62c 886out:
1da177e4 887 write_unlock_bh(&xfrm_policy_lock);
4aa2e62c 888 return err;
1da177e4
LT		 889}
890EXPORT_SYMBOL(xfrm_policy_flush);
891
cdcbca7c 892int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 893 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 894 void *data)
895{
12a169e7
HX		 896 struct xfrm_policy *pol;
897 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 898 int error = 0;
899
900 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
901 walk->type != XFRM_POLICY_TYPE_ANY)
902 return -EINVAL;
1da177e4 903
12a169e7 904 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 905 return 0;
906
12a169e7
HX		 907 write_lock_bh(&xfrm_policy_lock);
908 if (list_empty(&walk->walk.all))
cdcbca7c 909 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 910 else
911 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 912 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 913 if (x->dead)
4c563f76 914 continue;
12a169e7
HX		 915 pol = container_of(x, struct xfrm_policy, walk);
916 if (walk->type != XFRM_POLICY_TYPE_ANY &&
917 walk->type != pol->type)
918 continue;
919 error = func(pol, xfrm_policy_id2dir(pol->index),
920 walk->seq, data);
921 if (error) {
922 list_move_tail(&walk->walk.all, &x->all);
923 goto out;
2518c7c2 924 }
12a169e7 925 walk->seq++;
1da177e4 926 }
12a169e7 927 if (walk->seq == 0) {
baf5d743
JHS		 928 error = -ENOENT;
929 goto out;
930 }
12a169e7 931 list_del_init(&walk->walk.all);
1da177e4 932out:
12a169e7 933 write_unlock_bh(&xfrm_policy_lock);
1da177e4
LT		 934 return error;
935}
936EXPORT_SYMBOL(xfrm_policy_walk);
937
12a169e7
HX		 938void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
939{
940 INIT_LIST_HEAD(&walk->walk.all);
941 walk->walk.dead = 1;
942 walk->type = type;
943 walk->seq = 0;
944}
945EXPORT_SYMBOL(xfrm_policy_walk_init);
946
947void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
948{
949 if (list_empty(&walk->walk.all))
950 return;
951
952 write_lock_bh(&xfrm_policy_lock);
953 list_del(&walk->walk.all);
954 write_unlock_bh(&xfrm_policy_lock);
955}
956EXPORT_SYMBOL(xfrm_policy_walk_done);
957
134b0fc5
JM		 958/*
959 * Find policy to apply to this flow.
960 *
961 * Returns 0 if policy found, else an -errno.
962 */
f299d557
DM		 963static int xfrm_policy_match(const struct xfrm_policy *pol,
964 const struct flowi *fl,
2518c7c2 965 u8 type, u16 family, int dir)
1da177e4 966{
f299d557 967 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 968 int ret = -ESRCH;
969 bool match;
1da177e4 970
2518c7c2 971 if (pol->family != family ||
1d28f42c 972 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 973 pol->type != type)
134b0fc5 974 return ret;
1da177e4 975
2518c7c2 976 match = xfrm_selector_match(sel, fl, family);
134b0fc5 977 if (match)
1d28f42c 978 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 979 dir);
2518c7c2 980
134b0fc5 981 return ret;
2518c7c2 982}
1da177e4 983
52479b62 984static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 985 const struct flowi *fl,
2518c7c2
DM		 986 u16 family, u8 dir)
987{
134b0fc5 988 int err;
2518c7c2 989 struct xfrm_policy *pol, *ret;
0b597e7e 990 const xfrm_address_t *daddr, *saddr;
2518c7c2 991 struct hlist_head *chain;
acba48e1 992 u32 priority = ~0U;
df71837d 993
2518c7c2
DM		 994 daddr = xfrm_flowi_daddr(fl, family);
995 saddr = xfrm_flowi_saddr(fl, family);
996 if (unlikely(!daddr || !saddr))
997 return NULL;
998
999 read_lock_bh(&xfrm_policy_lock);
52479b62 1000 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 1001 ret = NULL;
b67bfe0d 1002 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1003 err = xfrm_policy_match(pol, fl, type, family, dir);
1004 if (err) {
1005 if (err == -ESRCH)
1006 continue;
1007 else {
1008 ret = ERR_PTR(err);
1009 goto fail;
1010 }
1011 } else {
2518c7c2 1012 ret = pol;
acba48e1 1013 priority = ret->priority;
2518c7c2
DM		 1014 break;
1015 }
1016 }
52479b62 1017 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1018 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1019 err = xfrm_policy_match(pol, fl, type, family, dir);
1020 if (err) {
1021 if (err == -ESRCH)
1022 continue;
1023 else {
1024 ret = ERR_PTR(err);
1025 goto fail;
1026 }
1027 } else if (pol->priority < priority) {
acba48e1
DM		 1028 ret = pol;
1029 break;
1da177e4
LT		 1030 }
1031 }
acba48e1
DM		 1032 if (ret)
1033 xfrm_pol_hold(ret);
134b0fc5 1034fail:
1da177e4 1035 read_unlock_bh(&xfrm_policy_lock);
4e81bb83 1036
2518c7c2 1037 return ret;
4e81bb83
MN		 1038}
1039
80c802f3 1040static struct xfrm_policy *
73ff93cd 1041__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1042{
1043#ifdef CONFIG_XFRM_SUB_POLICY
1044 struct xfrm_policy *pol;
1045
1046 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1047 if (pol != NULL)
1048 return pol;
1049#endif
1050 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1051}
1052
b5fb82c4
BZ		 1053static int flow_to_policy_dir(int dir)
1054{
1055 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1056 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1057 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1058 return dir;
1059
1060 switch (dir) {
1061 default:
1062 case FLOW_DIR_IN:
1063 return XFRM_POLICY_IN;
1064 case FLOW_DIR_OUT:
1065 return XFRM_POLICY_OUT;
1066 case FLOW_DIR_FWD:
1067 return XFRM_POLICY_FWD;
1068 }
1069}
1070
fe1a5f03 1071static struct flow_cache_object *
dee9f4bc 1072xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1073 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1074{
1075 struct xfrm_policy *pol;
fe1a5f03
TT		 1076
1077 if (old_obj)
1078 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1079
b5fb82c4 1080 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1081 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1082 return ERR_CAST(pol);
fe1a5f03 1083
fe1a5f03
TT		 1084 /* Resolver returns two references:
1085 * one for cache and one for caller of flow_cache_lookup() */
1086 xfrm_pol_hold(pol);
1087
1088 return &pol->flo;
1da177e4
LT		 1089}
1090
df71837d
TJ		 1091static inline int policy_to_flow_dir(int dir)
1092{
1093 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1094 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1095 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1096 return dir;
1097 switch (dir) {
1098 default:
1099 case XFRM_POLICY_IN:
1100 return FLOW_DIR_IN;
1101 case XFRM_POLICY_OUT:
1102 return FLOW_DIR_OUT;
1103 case XFRM_POLICY_FWD:
1104 return FLOW_DIR_FWD;
3ff50b79 1105 }
df71837d
TJ		 1106}
1107
dee9f4bc
DM		 1108static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1109 const struct flowi *fl)
1da177e4
LT		 1110{
1111 struct xfrm_policy *pol;
1112
1113 read_lock_bh(&xfrm_policy_lock);
1114 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1115 bool match = xfrm_selector_match(&pol->selector, fl,
1116 sk->sk_family);
a716c119 1117 int err = 0;
df71837d 1118
3bccfbc7 1119 if (match) {
34f8d884
JHS		 1120 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1121 pol = NULL;
1122 goto out;
1123 }
03e1ad7b 1124 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1125 fl->flowi_secid,
03e1ad7b 1126 policy_to_flow_dir(dir));
3bccfbc7
VY		 1127 if (!err)
1128 xfrm_pol_hold(pol);
1129 else if (err == -ESRCH)
1130 pol = NULL;
1131 else
1132 pol = ERR_PTR(err);
1133 } else
1da177e4
LT		 1134 pol = NULL;
1135 }
34f8d884 1136out:
1da177e4
LT		 1137 read_unlock_bh(&xfrm_policy_lock);
1138 return pol;
1139}
1140
1141static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1142{
98806f75 1143 struct net *net = xp_net(pol);
1121994c 1144 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1145 pol->family, dir);
4e81bb83 1146
98806f75 1147 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1148 hlist_add_head(&pol->bydst, chain);
e92303f8 1149 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1150 net->xfrm.policy_count[dir]++;
1da177e4 1151 xfrm_pol_hold(pol);
2518c7c2 1152
98806f75
AD		 1153 if (xfrm_bydst_should_resize(net, dir, NULL))
1154 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1155}
1156
1157static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1158 int dir)
1159{
98806f75
AD		 1160 struct net *net = xp_net(pol);
1161
2518c7c2
DM		 1162 if (hlist_unhashed(&pol->bydst))
1163 return NULL;
1da177e4 1164
2518c7c2
DM		 1165 hlist_del(&pol->bydst);
1166 hlist_del(&pol->byidx);
12a169e7 1167 list_del(&pol->walk.all);
98806f75 1168 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1169
1170 return pol;
1da177e4
LT		 1171}
1172
4666faab 1173int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4
LT		 1174{
1175 write_lock_bh(&xfrm_policy_lock);
1176 pol = __xfrm_policy_unlink(pol, dir);
1177 write_unlock_bh(&xfrm_policy_lock);
1178 if (pol) {
1da177e4 1179 xfrm_policy_kill(pol);
4666faab 1180 return 0;
1da177e4 1181 }
4666faab 1182 return -ENOENT;
1da177e4 1183}
a70fcb0b 1184EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1185
1186int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1187{
1121994c 1188 struct net *net = xp_net(pol);
1da177e4
LT		 1189 struct xfrm_policy *old_pol;
1190
4e81bb83
MN		 1191#ifdef CONFIG_XFRM_SUB_POLICY
1192 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1193 return -EINVAL;
1194#endif
1195
1da177e4
LT		 1196 write_lock_bh(&xfrm_policy_lock);
1197 old_pol = sk->sk_policy[dir];
1198 sk->sk_policy[dir] = pol;
1199 if (pol) {
9d729f72 1200 pol->curlft.add_time = get_seconds();
e682adf0 1201 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
1da177e4
LT		 1202 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1203 }
a0073fe1
SK		 1204 if (old_pol) {
1205 if (pol)
1206 xfrm_policy_requeue(old_pol, pol);
1207
ea2dea9d
TT		 1208 /* Unlinking succeeds always. This is the only function
1209 * allowed to delete or replace socket policy.
1210 */
1da177e4 1211 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
a0073fe1 1212 }
1da177e4
LT		 1213 write_unlock_bh(&xfrm_policy_lock);
1214
1215 if (old_pol) {
1216 xfrm_policy_kill(old_pol);
1217 }
1218 return 0;
1219}
1220
d3e40a9f 1221static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1222{
0331b1f3 1223 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1da177e4
LT		 1224
1225 if (newp) {
1226 newp->selector = old->selector;
03e1ad7b
PM		 1227 if (security_xfrm_policy_clone(old->security,
1228 &newp->security)) {
df71837d
TJ		 1229 kfree(newp);
1230 return NULL; /* ENOMEM */
1231 }
1da177e4
LT		 1232 newp->lft = old->lft;
1233 newp->curlft = old->curlft;
fb977e2c 1234 newp->mark = old->mark;
1da177e4
LT		 1235 newp->action = old->action;
1236 newp->flags = old->flags;
1237 newp->xfrm_nr = old->xfrm_nr;
1238 newp->index = old->index;
4e81bb83 1239 newp->type = old->type;
1da177e4
LT		 1240 memcpy(newp->xfrm_vec, old->xfrm_vec,
1241 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1242 write_lock_bh(&xfrm_policy_lock);
1243 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1244 write_unlock_bh(&xfrm_policy_lock);
1245 xfrm_pol_put(newp);
1246 }
1247 return newp;
1248}
1249
1250int __xfrm_sk_clone_policy(struct sock *sk)
1251{
1252 struct xfrm_policy *p0 = sk->sk_policy[0],
1253 *p1 = sk->sk_policy[1];
1254
1255 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1256 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1257 return -ENOMEM;
1258 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1259 return -ENOMEM;
1260 return 0;
1261}
1262
a1e59abf 1263static int
fbda33b2 1264xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1265 unsigned short family)
1266{
1267 int err;
1268 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1269
1270 if (unlikely(afinfo == NULL))
1271 return -EINVAL;
fbda33b2 1272 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1273 xfrm_policy_put_afinfo(afinfo);
1274 return err;
1275}
1276
1da177e4
LT		 1277/* Resolve list of templates for the flow, given policy. */
1278
1279static int
a6c2e611
DM		 1280xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1281 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1282{
fbda33b2 1283 struct net *net = xp_net(policy);
1da177e4
LT		 1284 int nx;
1285 int i, error;
1286 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1287 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1288 xfrm_address_t tmp;
1da177e4
LT		 1289
1290 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1291 struct xfrm_state *x;
1292 xfrm_address_t *remote = daddr;
1293 xfrm_address_t *local = saddr;
1294 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1295
48b8d783
JK		 1296 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1297 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1298 remote = &tmpl->id.daddr;
1299 local = &tmpl->saddr;
8444cf71
TE		 1300 if (xfrm_addr_any(local, tmpl->encap_family)) {
1301 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1302 if (error)
1303 goto fail;
1304 local = &tmp;
1305 }
1da177e4
LT		 1306 }
1307
1308 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1309
1310 if (x && x->km.state == XFRM_STATE_VALID) {
1311 xfrm[nx++] = x;
1312 daddr = remote;
1313 saddr = local;
1314 continue;
1315 }
1316 if (x) {
1317 error = (x->km.state == XFRM_STATE_ERROR ?
1318 -EINVAL : -EAGAIN);
1319 xfrm_state_put(x);
1320 }
a4322266 1321 else if (error == -ESRCH)
1322 error = -EAGAIN;
1da177e4
LT		 1323
1324 if (!tmpl->optional)
1325 goto fail;
1326 }
1327 return nx;
1328
1329fail:
1330 for (nx--; nx>=0; nx--)
1331 xfrm_state_put(xfrm[nx]);
1332 return error;
1333}
1334
4e81bb83 1335static int
a6c2e611
DM		 1336xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1337 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1338{
41a49cc3
MN		 1339 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1340 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1341 int cnx = 0;
1342 int error;
1343 int ret;
1344 int i;
1345
1346 for (i = 0; i < npols; i++) {
1347 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1348 error = -ENOBUFS;
1349 goto fail;
1350 }
41a49cc3
MN		 1351
1352 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1353 if (ret < 0) {
1354 error = ret;
1355 goto fail;
1356 } else
1357 cnx += ret;
1358 }
1359
41a49cc3
MN		 1360 /* found states are sorted for outbound processing */
1361 if (npols > 1)
1362 xfrm_state_sort(xfrm, tpp, cnx, family);
1363
4e81bb83
MN		 1364 return cnx;
1365
1366 fail:
1367 for (cnx--; cnx>=0; cnx--)
41a49cc3 1368 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1369 return error;
1370
1371}
1372
1da177e4
LT		 1373/* Check that the bundle accepts the flow and its components are
1374 * still valid.
1375 */
1376
05d84025 1377static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1378{
1379 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1380 int tos;
1da177e4 1381
25ee3286
HX		 1382 if (!afinfo)
1383 return -EINVAL;
1384
1385 tos = afinfo->get_tos(fl);
1386
1387 xfrm_policy_put_afinfo(afinfo);
1388
1389 return tos;
1390}
1391
80c802f3
TT		 1392static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1393{
1394 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1395 struct dst_entry *dst = &xdst->u.dst;
1396
1397 if (xdst->route == NULL) {
1398 /* Dummy bundle - if it has xfrms we were not
1399 * able to build bundle as template resolution failed.
1400 * It means we need to try again resolving. */
1401 if (xdst->num_xfrms > 0)
1402 return NULL;
a0073fe1
SK		 1403 } else if (dst->flags & DST_XFRM_QUEUE) {
1404 return NULL;
80c802f3
TT		 1405 } else {
1406 /* Real bundle */
1407 if (stale_bundle(dst))
1408 return NULL;
1409 }
1410
1411 dst_hold(dst);
1412 return flo;
1413}
1414
1415static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1416{
1417 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1418 struct dst_entry *dst = &xdst->u.dst;
1419
1420 if (!xdst->route)
1421 return 0;
1422 if (stale_bundle(dst))
1423 return 0;
1424
1425 return 1;
1426}
1427
1428static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1429{
1430 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1431 struct dst_entry *dst = &xdst->u.dst;
1432
1433 dst_free(dst);
1434}
1435
1436static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1437 .get = xfrm_bundle_flo_get,
1438 .check = xfrm_bundle_flo_check,
1439 .delete = xfrm_bundle_flo_delete,
1440};
1441
d7c7544c 1442static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1443{
1da177e4 1444 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1445 struct dst_ops *dst_ops;
25ee3286
HX		 1446 struct xfrm_dst *xdst;
1447
1448 if (!afinfo)
1449 return ERR_PTR(-EINVAL);
1450
d7c7544c
AD		 1451 switch (family) {
1452 case AF_INET:
1453 dst_ops = &net->xfrm.xfrm4_dst_ops;
1454 break;
dfd56b8b 1455#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1456 case AF_INET6:
1457 dst_ops = &net->xfrm.xfrm6_dst_ops;
1458 break;
1459#endif
1460 default:
1461 BUG();
1462 }
f5b0a874 1463 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1464
d4cae562 1465 if (likely(xdst)) {
141e369d
SK		 1466 struct dst_entry *dst = &xdst->u.dst;
1467
1468 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1469 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1470 if (afinfo->init_dst)
1471 afinfo->init_dst(net, xdst);
d4cae562 1472 } else
0b150932 1473 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1474
d4cae562
MB		 1475 xfrm_policy_put_afinfo(afinfo);
1476
25ee3286
HX		 1477 return xdst;
1478}
1479
a1b05140
MN		 1480static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1481 int nfheader_len)
1482{
1483 struct xfrm_policy_afinfo *afinfo =
1484 xfrm_policy_get_afinfo(dst->ops->family);
1485 int err;
1486
1487 if (!afinfo)
1488 return -EINVAL;
1489
1490 err = afinfo->init_path(path, dst, nfheader_len);
1491
1492 xfrm_policy_put_afinfo(afinfo);
1493
1494 return err;
1495}
1496
87c1e12b 1497static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1498 const struct flowi *fl)
25ee3286
HX		 1499{
1500 struct xfrm_policy_afinfo *afinfo =
1501 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1502 int err;
1503
1504 if (!afinfo)
1da177e4 1505 return -EINVAL;
25ee3286 1506
87c1e12b 1507 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1508
1da177e4 1509 xfrm_policy_put_afinfo(afinfo);
25ee3286 1510
1da177e4
LT		 1511 return err;
1512}
1513
80c802f3 1514
25ee3286
HX		 1515/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1516 * all the metrics... Shortly, bundle a bundle.
1517 */
1518
1519static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1520 struct xfrm_state **xfrm, int nx,
98313ada 1521 const struct flowi *fl,
25ee3286
HX		 1522 struct dst_entry *dst)
1523{
d7c7544c 1524 struct net *net = xp_net(policy);
25ee3286
HX		 1525 unsigned long now = jiffies;
1526 struct net_device *dev;
43a4dea4 1527 struct xfrm_mode *inner_mode;
25ee3286
HX		 1528 struct dst_entry *dst_prev = NULL;
1529 struct dst_entry *dst0 = NULL;
1530 int i = 0;
1531 int err;
1532 int header_len = 0;
a1b05140 1533 int nfheader_len = 0;
25ee3286
HX		 1534 int trailer_len = 0;
1535 int tos;
1536 int family = policy->selector.family;
9bb182a7
YH		 1537 xfrm_address_t saddr, daddr;
1538
1539 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1540
1541 tos = xfrm_get_tos(fl, family);
1542 err = tos;
1543 if (tos < 0)
1544 goto put_states;
1545
1546 dst_hold(dst);
1547
1548 for (; i < nx; i++) {
d7c7544c 1549 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1550 struct dst_entry *dst1 = &xdst->u.dst;
1551
1552 err = PTR_ERR(xdst);
1553 if (IS_ERR(xdst)) {
1554 dst_release(dst);
1555 goto put_states;
1556 }
1557
43a4dea4
SK		 1558 if (xfrm[i]->sel.family == AF_UNSPEC) {
1559 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1560 xfrm_af2proto(family));
1561 if (!inner_mode) {
1562 err = -EAFNOSUPPORT;
1563 dst_release(dst);
1564 goto put_states;
1565 }
1566 } else
1567 inner_mode = xfrm[i]->inner_mode;
1568
25ee3286
HX		 1569 if (!dst_prev)
1570 dst0 = dst1;
1571 else {
1572 dst_prev->child = dst_clone(dst1);
1573 dst1->flags |= DST_NOHASH;
1574 }
1575
1576 xdst->route = dst;
defb3519 1577 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1578
1579 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1580 family = xfrm[i]->props.family;
9bb182a7
YH		 1581 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1582 family);
25ee3286
HX		 1583 err = PTR_ERR(dst);
1584 if (IS_ERR(dst))
1585 goto put_states;
1586 } else
1587 dst_hold(dst);
1588
1589 dst1->xfrm = xfrm[i];
80c802f3 1590 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1591
f5b0a874 1592 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1593 dst1->flags |= DST_HOST;
1594 dst1->lastuse = now;
1595
1596 dst1->input = dst_discard;
43a4dea4 1597 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1598
1599 dst1->next = dst_prev;
1600 dst_prev = dst1;
1601
1602 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1603 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1604 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1605 trailer_len += xfrm[i]->props.trailer_len;
1606 }
1607
1608 dst_prev->child = dst;
1609 dst0->path = dst;
1610
1611 err = -ENODEV;
1612 dev = dst->dev;
1613 if (!dev)
1614 goto free_dst;
1615
a1b05140 1616 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1617 xfrm_init_pmtu(dst_prev);
1618
1619 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1620 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1621
87c1e12b 1622 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1623 if (err)
1624 goto free_dst;
1625
1626 dst_prev->header_len = header_len;
1627 dst_prev->trailer_len = trailer_len;
1628 header_len -= xdst->u.dst.xfrm->props.header_len;
1629 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1630 }
1631
1632out:
1633 return dst0;
1634
1635put_states:
1636 for (; i < nx; i++)
1637 xfrm_state_put(xfrm[i]);
1638free_dst:
1639 if (dst0)
1640 dst_free(dst0);
1641 dst0 = ERR_PTR(err);
1642 goto out;
1643}
1644
157bfc25 1645static int inline
3f0e18fb 1646xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1647{
1648 if (!*target) {
1649 *target = kmalloc(size, GFP_ATOMIC);
1650 if (!*target)
1651 return -ENOMEM;
1652 }
1653 memcpy(*target, src, size);
1654 return 0;
1655}
1656
1657static int inline
1786b389 1658xfrm_dst_update_parent(struct dst_entry *dst, const struct xfrm_selector *sel)
157bfc25
MN		 1659{
1660#ifdef CONFIG_XFRM_SUB_POLICY
1661 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1662 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1663 sel, sizeof(*sel));
1664#else
1665 return 0;
1666#endif
1667}
1668
1669static int inline
3f0e18fb 1670xfrm_dst_update_origin(struct dst_entry *dst, const struct flowi *fl)
157bfc25
MN		 1671{
1672#ifdef CONFIG_XFRM_SUB_POLICY
1673 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1674 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1675#else
1676 return 0;
1677#endif
1678}
1da177e4 1679
73ff93cd 1680static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1681 struct xfrm_policy **pols,
1682 int *num_pols, int *num_xfrms)
1683{
1684 int i;
1685
1686 if (*num_pols == 0 || !pols[0]) {
1687 *num_pols = 0;
1688 *num_xfrms = 0;
1689 return 0;
1690 }
1691 if (IS_ERR(pols[0]))
1692 return PTR_ERR(pols[0]);
1693
1694 *num_xfrms = pols[0]->xfrm_nr;
1695
1696#ifdef CONFIG_XFRM_SUB_POLICY
1697 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1698 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1699 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1700 XFRM_POLICY_TYPE_MAIN,
1701 fl, family,
1702 XFRM_POLICY_OUT);
1703 if (pols[1]) {
1704 if (IS_ERR(pols[1])) {
1705 xfrm_pols_put(pols, *num_pols);
1706 return PTR_ERR(pols[1]);
1707 }
1708 (*num_pols) ++;
1709 (*num_xfrms) += pols[1]->xfrm_nr;
1710 }
1711 }
1712#endif
1713 for (i = 0; i < *num_pols; i++) {
1714 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1715 *num_xfrms = -1;
1716 break;
1717 }
1718 }
1719
1720 return 0;
1721
1722}
1723
1724static struct xfrm_dst *
1725xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1726 const struct flowi *fl, u16 family,
80c802f3
TT		 1727 struct dst_entry *dst_orig)
1728{
1729 struct net *net = xp_net(pols[0]);
1730 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1731 struct dst_entry *dst;
1732 struct xfrm_dst *xdst;
1733 int err;
1734
1735 /* Try to instantiate a bundle */
1736 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1737 if (err <= 0) {
1738 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1739 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1740 return ERR_PTR(err);
1741 }
1742
1743 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1744 if (IS_ERR(dst)) {
1745 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1746 return ERR_CAST(dst);
1747 }
1748
1749 xdst = (struct xfrm_dst *)dst;
1750 xdst->num_xfrms = err;
1751 if (num_pols > 1)
1752 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1753 else
1754 err = xfrm_dst_update_origin(dst, fl);
1755 if (unlikely(err)) {
1756 dst_free(dst);
1757 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1758 return ERR_PTR(err);
1759 }
1760
1761 xdst->num_pols = num_pols;
1762 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1763 xdst->policy_genid = atomic_read(&pols[0]->genid);
1764
1765 return xdst;
1766}
1767
a0073fe1
SK		 1768static void xfrm_policy_queue_process(unsigned long arg)
1769{
1770 int err = 0;
1771 struct sk_buff *skb;
1772 struct sock *sk;
1773 struct dst_entry *dst;
a0073fe1
SK		 1774 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1775 struct xfrm_policy_queue *pq = &pol->polq;
1776 struct flowi fl;
1777 struct sk_buff_head list;
1778
1779 spin_lock(&pq->hold_queue.lock);
1780 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1781 if (!skb) {
1782 spin_unlock(&pq->hold_queue.lock);
1783 goto out;
1784 }
a0073fe1
SK		 1785 dst = skb_dst(skb);
1786 sk = skb->sk;
1787 xfrm_decode_session(skb, &fl, dst->ops->family);
1788 spin_unlock(&pq->hold_queue.lock);
1789
1790 dst_hold(dst->path);
1791 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1792 sk, 0);
1793 if (IS_ERR(dst))
1794 goto purge_queue;
1795
1796 if (dst->flags & DST_XFRM_QUEUE) {
1797 dst_release(dst);
1798
1799 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1800 goto purge_queue;
1801
1802 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1803 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1804 xfrm_pol_hold(pol);
1805 goto out;
a0073fe1
SK		 1806 }
1807
1808 dst_release(dst);
1809
1810 __skb_queue_head_init(&list);
1811
1812 spin_lock(&pq->hold_queue.lock);
1813 pq->timeout = 0;
1814 skb_queue_splice_init(&pq->hold_queue, &list);
1815 spin_unlock(&pq->hold_queue.lock);
1816
1817 while (!skb_queue_empty(&list)) {
1818 skb = __skb_dequeue(&list);
1819
1820 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1821 dst_hold(skb_dst(skb)->path);
1822 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1823 &fl, skb->sk, 0);
1824 if (IS_ERR(dst)) {
a0073fe1
SK		 1825 kfree_skb(skb);
1826 continue;
1827 }
1828
1829 nf_reset(skb);
1830 skb_dst_drop(skb);
1831 skb_dst_set(skb, dst);
1832
a0073fe1 1833 err = dst_output(skb);
a0073fe1
SK		 1834 }
1835
e7d8f6cb
SK		 1836out:
1837 xfrm_pol_put(pol);
a0073fe1
SK		 1838 return;
1839
1840purge_queue:
1841 pq->timeout = 0;
1842 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1843 xfrm_pol_put(pol);
a0073fe1
SK		 1844}
1845
1846static int xdst_queue_output(struct sk_buff *skb)
1847{
1848 unsigned long sched_next;
1849 struct dst_entry *dst = skb_dst(skb);
1850 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1851 struct xfrm_policy *pol = xdst->pols[0];
1852 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4
SK		 1853 const struct sk_buff *fclone = skb + 1;
1854
1855 if (unlikely(skb->fclone == SKB_FCLONE_ORIG &&
1856 fclone->fclone == SKB_FCLONE_CLONE)) {
1857 kfree_skb(skb);
1858 return 0;
1859 }
a0073fe1
SK		 1860
1861 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1862 kfree_skb(skb);
1863 return -EAGAIN;
1864 }
1865
1866 skb_dst_force(skb);
a0073fe1
SK		 1867
1868 spin_lock_bh(&pq->hold_queue.lock);
1869
1870 if (!pq->timeout)
1871 pq->timeout = XFRM_QUEUE_TMO_MIN;
1872
1873 sched_next = jiffies + pq->timeout;
1874
1875 if (del_timer(&pq->hold_timer)) {
1876 if (time_before(pq->hold_timer.expires, sched_next))
1877 sched_next = pq->hold_timer.expires;
e7d8f6cb 1878 xfrm_pol_put(pol);
a0073fe1
SK		 1879 }
1880
1881 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1882 if (!mod_timer(&pq->hold_timer, sched_next))
1883 xfrm_pol_hold(pol);
a0073fe1
SK		 1884
1885 spin_unlock_bh(&pq->hold_queue.lock);
1886
1887 return 0;
1888}
1889
1890static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
1891 struct dst_entry *dst,
1892 const struct flowi *fl,
1893 int num_xfrms,
1894 u16 family)
1895{
1896 int err;
1897 struct net_device *dev;
1898 struct dst_entry *dst1;
1899 struct xfrm_dst *xdst;
1900
1901 xdst = xfrm_alloc_dst(net, family);
1902 if (IS_ERR(xdst))
1903 return xdst;
1904
1905 if (net->xfrm.sysctl_larval_drop || num_xfrms <= 0 ||
1906 (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP))
1907 return xdst;
1908
1909 dst1 = &xdst->u.dst;
1910 dst_hold(dst);
1911 xdst->route = dst;
1912
1913 dst_copy_metrics(dst1, dst);
1914
1915 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
1916 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
1917 dst1->lastuse = jiffies;
1918
1919 dst1->input = dst_discard;
1920 dst1->output = xdst_queue_output;
1921
1922 dst_hold(dst);
1923 dst1->child = dst;
1924 dst1->path = dst;
1925
1926 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
1927
1928 err = -ENODEV;
1929 dev = dst->dev;
1930 if (!dev)
1931 goto free_dst;
1932
1933 err = xfrm_fill_dst(xdst, dev, fl);
1934 if (err)
1935 goto free_dst;
1936
1937out:
1938 return xdst;
1939
1940free_dst:
1941 dst_release(dst1);
1942 xdst = ERR_PTR(err);
1943 goto out;
1944}
1945
80c802f3 1946static struct flow_cache_object *
dee9f4bc 1947xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 1948 struct flow_cache_object *oldflo, void *ctx)
1949{
1950 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1951 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1952 struct xfrm_dst *xdst, *new_xdst;
1953 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1954
1955 /* Check if the policies from old bundle are usable */
1956 xdst = NULL;
1957 if (oldflo) {
1958 xdst = container_of(oldflo, struct xfrm_dst, flo);
1959 num_pols = xdst->num_pols;
1960 num_xfrms = xdst->num_xfrms;
1961 pol_dead = 0;
1962 for (i = 0; i < num_pols; i++) {
1963 pols[i] = xdst->pols[i];
1964 pol_dead |= pols[i]->walk.dead;
1965 }
1966 if (pol_dead) {
1967 dst_free(&xdst->u.dst);
1968 xdst = NULL;
1969 num_pols = 0;
1970 num_xfrms = 0;
1971 oldflo = NULL;
1972 }
1973 }
1974
1975 /* Resolve policies to use if we couldn't get them from
1976 * previous cache entry */
1977 if (xdst == NULL) {
1978 num_pols = 1;
b5fb82c4
BZ		 1979 pols[0] = __xfrm_policy_lookup(net, fl, family,
1980 flow_to_policy_dir(dir));
80c802f3
TT		 1981 err = xfrm_expand_policies(fl, family, pols,
1982 &num_pols, &num_xfrms);
1983 if (err < 0)
1984 goto inc_error;
1985 if (num_pols == 0)
1986 return NULL;
1987 if (num_xfrms <= 0)
1988 goto make_dummy_bundle;
1989 }
1990
1991 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1992 if (IS_ERR(new_xdst)) {
1993 err = PTR_ERR(new_xdst);
1994 if (err != -EAGAIN)
1995 goto error;
1996 if (oldflo == NULL)
1997 goto make_dummy_bundle;
1998 dst_hold(&xdst->u.dst);
1999 return oldflo;
d809ec89
TT		 2000 } else if (new_xdst == NULL) {
2001 num_xfrms = 0;
2002 if (oldflo == NULL)
2003 goto make_dummy_bundle;
2004 xdst->num_xfrms = 0;
2005 dst_hold(&xdst->u.dst);
2006 return oldflo;
80c802f3
TT		 2007 }
2008
2009 /* Kill the previous bundle */
2010 if (xdst) {
2011 /* The policies were stolen for newly generated bundle */
2012 xdst->num_pols = 0;
2013 dst_free(&xdst->u.dst);
2014 }
2015
2016 /* Flow cache does not have reference, it dst_free()'s,
2017 * but we do need to return one reference for original caller */
2018 dst_hold(&new_xdst->u.dst);
2019 return &new_xdst->flo;
2020
2021make_dummy_bundle:
2022 /* We found policies, but there's no bundles to instantiate:
2023 * either because the policy blocks, has no transformations or
2024 * we could not build template (no xfrm_states).*/
a0073fe1 2025 xdst = xfrm_create_dummy_bundle(net, dst_orig, fl, num_xfrms, family);
80c802f3
TT		 2026 if (IS_ERR(xdst)) {
2027 xfrm_pols_put(pols, num_pols);
2028 return ERR_CAST(xdst);
2029 }
2030 xdst->num_pols = num_pols;
2031 xdst->num_xfrms = num_xfrms;
2032 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
2033
2034 dst_hold(&xdst->u.dst);
2035 return &xdst->flo;
2036
2037inc_error:
2038 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2039error:
2040 if (xdst != NULL)
2041 dst_free(&xdst->u.dst);
2042 else
2043 xfrm_pols_put(pols, num_pols);
2044 return ERR_PTR(err);
2045}
1da177e4 2046
2774c131
DM		 2047static struct dst_entry *make_blackhole(struct net *net, u16 family,
2048 struct dst_entry *dst_orig)
2049{
2050 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2051 struct dst_entry *ret;
2052
2053 if (!afinfo) {
2054 dst_release(dst_orig);
433a1954 2055 return ERR_PTR(-EINVAL);
2774c131
DM		 2056 } else {
2057 ret = afinfo->blackhole_route(net, dst_orig);
2058 }
2059 xfrm_policy_put_afinfo(afinfo);
2060
2061 return ret;
2062}
2063
1da177e4
LT		 2064/* Main function: finds/creates a bundle for given flow.
2065 *
2066 * At the moment we eat a raw IP route. Mostly to speed up lookups
2067 * on interfaces with disabled IPsec.
2068 */
452edd59
DM		 2069struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2070 const struct flowi *fl,
2071 struct sock *sk, int flags)
1da177e4 2072{
4e81bb83 2073 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2074 struct flow_cache_object *flo;
2075 struct xfrm_dst *xdst;
452edd59 2076 struct dst_entry *dst, *route;
80c802f3 2077 u16 family = dst_orig->ops->family;
df71837d 2078 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2079 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2080
1da177e4 2081restart:
80c802f3
TT		 2082 dst = NULL;
2083 xdst = NULL;
2084 route = NULL;
4e81bb83 2085
f7944fb1 2086 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2087 num_pols = 1;
2088 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2089 err = xfrm_expand_policies(fl, family, pols,
2090 &num_pols, &num_xfrms);
2091 if (err < 0)
75b8c133 2092 goto dropdst;
80c802f3
TT		 2093
2094 if (num_pols) {
2095 if (num_xfrms <= 0) {
2096 drop_pols = num_pols;
2097 goto no_transform;
2098 }
2099
2100 xdst = xfrm_resolve_and_create_bundle(
2101 pols, num_pols, fl,
2102 family, dst_orig);
2103 if (IS_ERR(xdst)) {
2104 xfrm_pols_put(pols, num_pols);
2105 err = PTR_ERR(xdst);
2106 goto dropdst;
d809ec89
TT		 2107 } else if (xdst == NULL) {
2108 num_xfrms = 0;
2109 drop_pols = num_pols;
2110 goto no_transform;
80c802f3
TT		 2111 }
2112
fbd50608
SK		 2113 dst_hold(&xdst->u.dst);
2114
80c802f3
TT		 2115 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2116 xdst->u.dst.next = xfrm_policy_sk_bundles;
2117 xfrm_policy_sk_bundles = &xdst->u.dst;
2118 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2119
2120 route = xdst->route;
0aa64774 2121 }
3bccfbc7 2122 }
1da177e4 2123
80c802f3 2124 if (xdst == NULL) {
1da177e4 2125 /* To accelerate a bit... */
2518c7c2 2126 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2127 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2128 goto nopol;
1da177e4 2129
80c802f3
TT		 2130 flo = flow_cache_lookup(net, fl, family, dir,
2131 xfrm_bundle_lookup, dst_orig);
2132 if (flo == NULL)
2133 goto nopol;
fe1a5f03 2134 if (IS_ERR(flo)) {
80c802f3 2135 err = PTR_ERR(flo);
75b8c133 2136 goto dropdst;
d66e37a9 2137 }
80c802f3
TT		 2138 xdst = container_of(flo, struct xfrm_dst, flo);
2139
2140 num_pols = xdst->num_pols;
2141 num_xfrms = xdst->num_xfrms;
2142 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
2143 route = xdst->route;
2144 }
2145
2146 dst = &xdst->u.dst;
2147 if (route == NULL && num_xfrms > 0) {
2148 /* The only case when xfrm_bundle_lookup() returns a
2149 * bundle with null route, is when the template could
2150 * not be resolved. It means policies are there, but
2151 * bundle could not be created, since we don't yet
2152 * have the xfrm_state's. We need to wait for KM to
2153 * negotiate new SA's or bail out with error.*/
2154 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2155 dst_release(dst);
a1aa3483 2156 xfrm_pols_put(pols, drop_pols);
80c802f3 2157 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2774c131 2158
452edd59 2159 return make_blackhole(net, family, dst_orig);
80c802f3 2160 }
1d28f42c 2161 if (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP) {
80c802f3
TT		 2162 DECLARE_WAITQUEUE(wait, current);
2163
2164 add_wait_queue(&net->xfrm.km_waitq, &wait);
2165 set_current_state(TASK_INTERRUPTIBLE);
2166 schedule();
2167 set_current_state(TASK_RUNNING);
2168 remove_wait_queue(&net->xfrm.km_waitq, &wait);
2169
2170 if (!signal_pending(current)) {
2171 dst_release(dst);
2172 goto restart;
2173 }
2174
2175 err = -ERESTART;
2176 } else
2177 err = -EAGAIN;
2178
2179 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2180 goto error;
1da177e4
LT		 2181 }
2182
80c802f3
TT		 2183no_transform:
2184 if (num_pols == 0)
8b7817f3 2185 goto nopol;
1da177e4 2186
80c802f3
TT		 2187 if ((flags & XFRM_LOOKUP_ICMP) &&
2188 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2189 err = -ENOENT;
8b7817f3 2190 goto error;
80c802f3 2191 }
8b7817f3 2192
80c802f3
TT		 2193 for (i = 0; i < num_pols; i++)
2194 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2195
80c802f3 2196 if (num_xfrms < 0) {
1da177e4 2197 /* Prohibit the flow */
59c9940e 2198 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2199 err = -EPERM;
2200 goto error;
80c802f3
TT		 2201 } else if (num_xfrms > 0) {
2202 /* Flow transformed */
80c802f3
TT		 2203 dst_release(dst_orig);
2204 } else {
2205 /* Flow passes untransformed */
2206 dst_release(dst);
452edd59 2207 dst = dst_orig;
1da177e4 2208 }
80c802f3
TT		 2209ok:
2210 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2211 if (dst && dst->xfrm &&
2212 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2213 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2214 return dst;
1da177e4 2215
80c802f3 2216nopol:
452edd59
DM		 2217 if (!(flags & XFRM_LOOKUP_ICMP)) {
2218 dst = dst_orig;
80c802f3 2219 goto ok;
452edd59 2220 }
80c802f3 2221 err = -ENOENT;
1da177e4 2222error:
80c802f3 2223 dst_release(dst);
75b8c133
HX		 2224dropdst:
2225 dst_release(dst_orig);
80c802f3 2226 xfrm_pols_put(pols, drop_pols);
452edd59 2227 return ERR_PTR(err);
1da177e4
LT		 2228}
2229EXPORT_SYMBOL(xfrm_lookup);
2230
df0ba92a 2231static inline int
8f029de2 2232xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2233{
2234 struct xfrm_state *x;
df0ba92a
MN		 2235
2236 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2237 return 0;
2238 x = skb->sp->xvec[idx];
2239 if (!x->type->reject)
2240 return 0;
1ecafede 2241 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2242}
2243
1da177e4
LT		 2244/* When skb is transformed back to its "native" form, we have to
2245 * check policy restrictions. At the moment we make this in maximally
2246 * stupid way. Shame on me. :-) Of course, connected sockets must
2247 * have policy cached at them.
2248 */
2249
2250static inline int
7db454b9 2251xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2252 unsigned short family)
2253{
2254 if (xfrm_state_kern(x))
928ba416 2255 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2256 return x->id.proto == tmpl->id.proto &&
2257 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2258 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2259 x->props.mode == tmpl->mode &&
c5d18e98 2260 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2261 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2262 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2263 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2264}
2265
df0ba92a
MN		 2266/*
2267 * 0 or more than 0 is returned when validation is succeeded (either bypass
2268 * because of optional transport mode, or next index of the mathced secpath
2269 * state with the template.
2270 * -1 is returned when no matching template is found.
2271 * Otherwise "-2 - errored_index" is returned.
2272 */
1da177e4 2273static inline int
22cccb7e 2274xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2275 unsigned short family)
2276{
2277 int idx = start;
2278
2279 if (tmpl->optional) {
7e49e6de 2280 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2281 return start;
2282 } else
2283 start = -1;
2284 for (; idx < sp->len; idx++) {
dbe5b4aa 2285 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2286 return ++idx;
df0ba92a
MN		 2287 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2288 if (start == -1)
2289 start = -2-idx;
1da177e4 2290 break;
df0ba92a 2291 }
1da177e4
LT		 2292 }
2293 return start;
2294}
2295
d5422efe
HX		 2296int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2297 unsigned int family, int reverse)
1da177e4
LT		 2298{
2299 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2300 int err;
1da177e4
LT		 2301
2302 if (unlikely(afinfo == NULL))
2303 return -EAFNOSUPPORT;
2304
d5422efe 2305 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2306 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2307 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2308 return err;
1da177e4 2309}
d5422efe 2310EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2311
9a7386ec 2312static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2313{
2314 for (; k < sp->len; k++) {
df0ba92a 2315 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2316 *idxp = k;
1da177e4 2317 return 1;
df0ba92a 2318 }
1da177e4
LT		 2319 }
2320
2321 return 0;
2322}
2323
a716c119 2324int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2325 unsigned short family)
2326{
f6e1e25d 2327 struct net *net = dev_net(skb->dev);
1da177e4 2328 struct xfrm_policy *pol;
4e81bb83
MN		 2329 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2330 int npols = 0;
2331 int xfrm_nr;
2332 int pi;
d5422efe 2333 int reverse;
1da177e4 2334 struct flowi fl;
d5422efe 2335 u8 fl_dir;
df0ba92a 2336 int xerr_idx = -1;
1da177e4 2337
d5422efe
HX		 2338 reverse = dir & ~XFRM_POLICY_MASK;
2339 dir &= XFRM_POLICY_MASK;
2340 fl_dir = policy_to_flow_dir(dir);
2341
0aa64774 2342 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2343 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2344 return 0;
0aa64774
MN		 2345 }
2346
eb9c7ebe 2347 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2348
2349 /* First, check used SA against their selectors. */
2350 if (skb->sp) {
2351 int i;
2352
2353 for (i=skb->sp->len-1; i>=0; i--) {
dbe5b4aa 2354 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2355 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2356 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2357 return 0;
0aa64774 2358 }
1da177e4
LT		 2359 }
2360 }
2361
2362 pol = NULL;
3bccfbc7 2363 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2364 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2365 if (IS_ERR(pol)) {
59c9940e 2366 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2367 return 0;
0aa64774 2368 }
3bccfbc7 2369 }
1da177e4 2370
fe1a5f03
TT		 2371 if (!pol) {
2372 struct flow_cache_object *flo;
2373
2374 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2375 xfrm_policy_lookup, NULL);
2376 if (IS_ERR_OR_NULL(flo))
2377 pol = ERR_CAST(flo);
2378 else
2379 pol = container_of(flo, struct xfrm_policy, flo);
2380 }
1da177e4 2381
0aa64774 2382 if (IS_ERR(pol)) {
59c9940e 2383 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2384 return 0;
0aa64774 2385 }
134b0fc5 2386
df0ba92a 2387 if (!pol) {
d1d9facf 2388 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2389 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2390 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2391 return 0;
2392 }
2393 return 1;
2394 }
1da177e4 2395
9d729f72 2396 pol->curlft.use_time = get_seconds();
1da177e4 2397
4e81bb83
MN		 2398 pols[0] = pol;
2399 npols ++;
2400#ifdef CONFIG_XFRM_SUB_POLICY
2401 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2402 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2403 &fl, family,
2404 XFRM_POLICY_IN);
2405 if (pols[1]) {
0aa64774 2406 if (IS_ERR(pols[1])) {
59c9940e 2407 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2408 return 0;
0aa64774 2409 }
9d729f72 2410 pols[1]->curlft.use_time = get_seconds();
4e81bb83
MN		 2411 npols ++;
2412 }
2413 }
2414#endif
2415
1da177e4
LT		 2416 if (pol->action == XFRM_POLICY_ALLOW) {
2417 struct sec_path *sp;
2418 static struct sec_path dummy;
4e81bb83 2419 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2420 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2421 struct xfrm_tmpl **tpp = tp;
2422 int ti = 0;
1da177e4
LT		 2423 int i, k;
2424
2425 if ((sp = skb->sp) == NULL)
2426 sp = &dummy;
2427
4e81bb83
MN		 2428 for (pi = 0; pi < npols; pi++) {
2429 if (pols[pi] != pol &&
0aa64774 2430 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2431 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2432 goto reject;
0aa64774
MN		 2433 }
2434 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2435 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2436 goto reject_error;
0aa64774 2437 }
4e81bb83
MN		 2438 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2439 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2440 }
2441 xfrm_nr = ti;
41a49cc3
MN		 2442 if (npols > 1) {
2443 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2444 tpp = stp;
2445 }
4e81bb83 2446
1da177e4
LT		 2447 /* For each tunnel xfrm, find the first matching tmpl.
2448 * For each tmpl before that, find corresponding xfrm.
2449 * Order is _important_. Later we will implement
2450 * some barriers, but at the moment barriers
2451 * are implied between each two transformations.
2452 */
4e81bb83
MN		 2453 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2454 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2455 if (k < 0) {
d1d9facf
JM		 2456 if (k < -1)
2457 /* "-2 - errored_index" returned */
2458 xerr_idx = -(2+k);
59c9940e 2459 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2460 goto reject;
df0ba92a 2461 }
1da177e4
LT		 2462 }
2463
0aa64774 2464 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2465 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2466 goto reject;
0aa64774 2467 }
1da177e4 2468
4e81bb83 2469 xfrm_pols_put(pols, npols);
1da177e4
LT		 2470 return 1;
2471 }
59c9940e 2472 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2473
2474reject:
df0ba92a 2475 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2476reject_error:
2477 xfrm_pols_put(pols, npols);
1da177e4
LT		 2478 return 0;
2479}
2480EXPORT_SYMBOL(__xfrm_policy_check);
2481
2482int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2483{
99a66657 2484 struct net *net = dev_net(skb->dev);
1da177e4 2485 struct flowi fl;
adf30907 2486 struct dst_entry *dst;
73137147 2487 int res = 1;
1da177e4 2488
0aa64774 2489 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2490 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2491 return 0;
0aa64774 2492 }
1da177e4 2493
fafeeb6c 2494 skb_dst_force(skb);
adf30907 2495
452edd59
DM		 2496 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2497 if (IS_ERR(dst)) {
73137147 2498 res = 0;
452edd59
DM		 2499 dst = NULL;
2500 }
adf30907
ED		 2501 skb_dst_set(skb, dst);
2502 return res;
1da177e4
LT		 2503}
2504EXPORT_SYMBOL(__xfrm_route_forward);
2505
d49c73c7
DM		 2506/* Optimize later using cookies and generation ids. */
2507
1da177e4
LT		 2508static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2509{
d49c73c7 2510 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2511 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2512 * get validated by dst_ops->check on every use. We do this
2513 * because when a normal route referenced by an XFRM dst is
2514 * obsoleted we do not go looking around for all parent
2515 * referencing XFRM dsts so that we can invalidate them. It
2516 * is just too much work. Instead we make the checks here on
2517 * every use. For example:
d49c73c7
DM		 2518 *
2519 * XFRM dst A --> IPv4 dst X
2520 *
2521 * X is the "xdst->route" of A (X is also the "dst->path" of A
2522 * in this example). If X is marked obsolete, "A" will not
2523 * notice. That's what we are validating here via the
2524 * stale_bundle() check.
2525 *
2526 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2527 * dst which causes it's ->obsolete field to be set to
2528 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2529 * this, we want to force a new route lookup.
399c180a 2530 */
d49c73c7
DM		 2531 if (dst->obsolete < 0 && !stale_bundle(dst))
2532 return dst;
2533
1da177e4
LT		 2534 return NULL;
2535}
2536
2537static int stale_bundle(struct dst_entry *dst)
2538{
12fdb4d3 2539 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2540}
2541
aabc9761 2542void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2543{
1da177e4 2544 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2545 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2546 dev_hold(dst->dev);
1da177e4
LT		 2547 dev_put(dev);
2548 }
2549}
aabc9761 2550EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2551
2552static void xfrm_link_failure(struct sk_buff *skb)
2553{
2554 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2555}
2556
2557static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2558{
2559 if (dst) {
2560 if (dst->obsolete) {
2561 dst_release(dst);
2562 dst = NULL;
2563 }
2564 }
2565 return dst;
2566}
2567
80c802f3 2568static void __xfrm_garbage_collect(struct net *net)
1da177e4 2569{
80c802f3 2570 struct dst_entry *head, *next;
1da177e4 2571
80c802f3
TT		 2572 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2573 head = xfrm_policy_sk_bundles;
2574 xfrm_policy_sk_bundles = NULL;
2575 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2518c7c2 2576
80c802f3
TT		 2577 while (head) {
2578 next = head->next;
2579 dst_free(head);
2580 head = next;
1da177e4
LT		 2581 }
2582}
2583
e4c17216 2584void xfrm_garbage_collect(struct net *net)
c0ed1c14
SK		 2585{
2586 flow_cache_flush();
2587 __xfrm_garbage_collect(net);
2588}
e4c17216 2589EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2590
2591static void xfrm_garbage_collect_deferred(struct net *net)
2592{
2593 flow_cache_flush_deferred();
2594 __xfrm_garbage_collect(net);
2595}
2596
25ee3286 2597static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2598{
2599 do {
2600 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2601 u32 pmtu, route_mtu_cached;
2602
2603 pmtu = dst_mtu(dst->child);
2604 xdst->child_mtu_cached = pmtu;
2605
2606 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2607
2608 route_mtu_cached = dst_mtu(xdst->route);
2609 xdst->route_mtu_cached = route_mtu_cached;
2610
2611 if (pmtu > route_mtu_cached)
2612 pmtu = route_mtu_cached;
2613
defb3519 2614 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2615 } while ((dst = dst->next));
2616}
2617
1da177e4
LT		 2618/* Check that the bundle accepts the flow and its components are
2619 * still valid.
2620 */
2621
12fdb4d3 2622static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2623{
2624 struct dst_entry *dst = &first->u.dst;
2625 struct xfrm_dst *last;
2626 u32 mtu;
2627
92d63dec 2628 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2629 (dst->dev && !netif_running(dst->dev)))
2630 return 0;
2631
a0073fe1
SK		 2632 if (dst->flags & DST_XFRM_QUEUE)
2633 return 1;
2634
1da177e4
LT		 2635 last = NULL;
2636
2637 do {
2638 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2639
1da177e4
LT		 2640 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2641 return 0;
80c802f3
TT		 2642 if (xdst->xfrm_genid != dst->xfrm->genid)
2643 return 0;
b1312c89
TT		 2644 if (xdst->num_pols > 0 &&
2645 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2646 return 0;
e53820de 2647
1da177e4
LT		 2648 mtu = dst_mtu(dst->child);
2649 if (xdst->child_mtu_cached != mtu) {
2650 last = xdst;
2651 xdst->child_mtu_cached = mtu;
2652 }
2653
92d63dec 2654 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2655 return 0;
2656 mtu = dst_mtu(xdst->route);
2657 if (xdst->route_mtu_cached != mtu) {
2658 last = xdst;
2659 xdst->route_mtu_cached = mtu;
2660 }
2661
2662 dst = dst->child;
2663 } while (dst->xfrm);
2664
2665 if (likely(!last))
2666 return 1;
2667
2668 mtu = last->child_mtu_cached;
2669 for (;;) {
2670 dst = &last->u.dst;
2671
2672 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2673 if (mtu > last->route_mtu_cached)
2674 mtu = last->route_mtu_cached;
defb3519 2675 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2676
2677 if (last == first)
2678 break;
2679
bd0bf076 2680 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2681 last->child_mtu_cached = mtu;
2682 }
2683
2684 return 1;
2685}
2686
0dbaee3b
DM		 2687static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2688{
2689 return dst_metric_advmss(dst->path);
2690}
2691
ebb762f2 2692static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2693{
618f9bc7
SK		 2694 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2695
2696 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2697}
2698
f894cbf8
DM		 2699static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2700 struct sk_buff *skb,
2701 const void *daddr)
d3aaeb38 2702{
f894cbf8 2703 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2704}
2705
1da177e4
LT		 2706int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2707{
d7c7544c 2708 struct net *net;
1da177e4
LT		 2709 int err = 0;
2710 if (unlikely(afinfo == NULL))
2711 return -EINVAL;
2712 if (unlikely(afinfo->family >= NPROTO))
2713 return -EAFNOSUPPORT;
ef8531b6 2714 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2715 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2716 err = -ENOBUFS;
2717 else {
2718 struct dst_ops *dst_ops = afinfo->dst_ops;
2719 if (likely(dst_ops->kmem_cachep == NULL))
2720 dst_ops->kmem_cachep = xfrm_dst_cache;
2721 if (likely(dst_ops->check == NULL))
2722 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2723 if (likely(dst_ops->default_advmss == NULL))
2724 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2725 if (likely(dst_ops->mtu == NULL))
2726 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2727 if (likely(dst_ops->negative_advice == NULL))
2728 dst_ops->negative_advice = xfrm_negative_advice;
2729 if (likely(dst_ops->link_failure == NULL))
2730 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2731 if (likely(dst_ops->neigh_lookup == NULL))
2732 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2733 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2734 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2735 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2736 }
ef8531b6 2737 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2738
2739 rtnl_lock();
2740 for_each_net(net) {
2741 struct dst_ops *xfrm_dst_ops;
2742
2743 switch (afinfo->family) {
2744 case AF_INET:
2745 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2746 break;
dfd56b8b 2747#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2748 case AF_INET6:
2749 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2750 break;
2751#endif
2752 default:
2753 BUG();
2754 }
2755 *xfrm_dst_ops = *afinfo->dst_ops;
2756 }
2757 rtnl_unlock();
2758
1da177e4
LT		 2759 return err;
2760}
2761EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2762
2763int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2764{
2765 int err = 0;
2766 if (unlikely(afinfo == NULL))
2767 return -EINVAL;
2768 if (unlikely(afinfo->family >= NPROTO))
2769 return -EAFNOSUPPORT;
ef8531b6 2770 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2771 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2772 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2773 err = -EINVAL;
ef8531b6
ED		 2774 else
2775 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2776 NULL);
2777 }
2778 spin_unlock(&xfrm_policy_afinfo_lock);
2779 if (!err) {
2780 struct dst_ops *dst_ops = afinfo->dst_ops;
2781
2782 synchronize_rcu();
2783
2784 dst_ops->kmem_cachep = NULL;
2785 dst_ops->check = NULL;
2786 dst_ops->negative_advice = NULL;
2787 dst_ops->link_failure = NULL;
2788 afinfo->garbage_collect = NULL;
1da177e4 2789 }
1da177e4
LT		 2790 return err;
2791}
2792EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2793
d7c7544c
AD		 2794static void __net_init xfrm_dst_ops_init(struct net *net)
2795{
2796 struct xfrm_policy_afinfo *afinfo;
2797
ef8531b6
ED		 2798 rcu_read_lock();
2799 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2800 if (afinfo)
2801 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2802#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2803 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2804 if (afinfo)
2805 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2806#endif
418a99ac 2807 rcu_read_unlock();
546be240
HX		 2808}
2809
1da177e4
LT		 2810static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2811{
351638e7 2812 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2813
1da177e4
LT		 2814 switch (event) {
2815 case NETDEV_DOWN:
c0ed1c14 2816 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2817 }
2818 return NOTIFY_DONE;
2819}
2820
2821static struct notifier_block xfrm_dev_notifier = {
d5917a35 2822 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2823};
2824
558f82ef 2825#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2826static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2827{
c68cd1a0
AD		 2828 int rv;
2829
7d720c3e 2830 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
1823e4c8
ED		 2831 sizeof(struct linux_xfrm_mib),
2832 __alignof__(struct linux_xfrm_mib)) < 0)
558f82ef 2833 return -ENOMEM;
c68cd1a0
AD		 2834 rv = xfrm_proc_init(net);
2835 if (rv < 0)
7d720c3e 2836 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
c68cd1a0 2837 return rv;
558f82ef 2838}
59c9940e
AD		 2839
2840static void xfrm_statistics_fini(struct net *net)
2841{
c68cd1a0 2842 xfrm_proc_fini(net);
7d720c3e 2843 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
59c9940e
AD		 2844}
2845#else
2846static int __net_init xfrm_statistics_init(struct net *net)
2847{
2848 return 0;
2849}
2850
2851static void xfrm_statistics_fini(struct net *net)
2852{
2853}
558f82ef
MN		 2854#endif
2855
d62ddc21 2856static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2857{
2518c7c2
DM		 2858 unsigned int hmask, sz;
2859 int dir;
2860
d62ddc21
AD		 2861 if (net_eq(net, &init_net))
2862 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2863 sizeof(struct xfrm_dst),
e5d679f3 2864 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2865 NULL);
1da177e4 2866
2518c7c2
DM		 2867 hmask = 8 - 1;
2868 sz = (hmask+1) * sizeof(struct hlist_head);
2869
93b851c1
AD		 2870 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2871 if (!net->xfrm.policy_byidx)
2872 goto out_byidx;
8100bea7 2873 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2874
2875 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2876 struct xfrm_policy_hash *htab;
2877
dc2caba7 2878 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2879 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2880
a35f6c5d 2881 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2882 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2883 if (!htab->table)
a35f6c5d
AD		 2884 goto out_bydst;
2885 htab->hmask = hmask;
2518c7c2
DM		 2886 }
2887
adfcf0b2 2888 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2889 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2890 if (net_eq(net, &init_net))
2891 register_netdevice_notifier(&xfrm_dev_notifier);
2892 return 0;
93b851c1 2893
a35f6c5d
AD		 2894out_bydst:
2895 for (dir--; dir >= 0; dir--) {
2896 struct xfrm_policy_hash *htab;
2897
2898 htab = &net->xfrm.policy_bydst[dir];
2899 xfrm_hash_free(htab->table, sz);
2900 }
2901 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2902out_byidx:
2903 return -ENOMEM;
d62ddc21
AD		 2904}
2905
2906static void xfrm_policy_fini(struct net *net)
2907{
7c2776ee 2908 struct xfrm_audit audit_info;
93b851c1 2909 unsigned int sz;
8b18f8ea 2910 int dir;
93b851c1 2911
7c2776ee
AD		 2912 flush_work(&net->xfrm.policy_hash_work);
2913#ifdef CONFIG_XFRM_SUB_POLICY
e1760bd5 2914 audit_info.loginuid = INVALID_UID;
7c2776ee
AD		 2915 audit_info.sessionid = -1;
2916 audit_info.secid = 0;
2917 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2918#endif
e1760bd5 2919 audit_info.loginuid = INVALID_UID;
7c2776ee
AD		 2920 audit_info.sessionid = -1;
2921 audit_info.secid = 0;
2922 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
7c2776ee 2923
adfcf0b2 2924 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2925
8b18f8ea 2926 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2927 struct xfrm_policy_hash *htab;
2928
8b18f8ea 2929 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2930
2931 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2932 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2933 WARN_ON(!hlist_empty(htab->table));
2934 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2935 }
2936
8100bea7 2937 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2938 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2939 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2940}
2941
d62ddc21
AD		 2942static int __net_init xfrm_net_init(struct net *net)
2943{
2944 int rv;
2945
59c9940e
AD		 2946 rv = xfrm_statistics_init(net);
2947 if (rv < 0)
2948 goto out_statistics;
d62ddc21
AD		 2949 rv = xfrm_state_init(net);
2950 if (rv < 0)
2951 goto out_state;
2952 rv = xfrm_policy_init(net);
2953 if (rv < 0)
2954 goto out_policy;
d7c7544c 2955 xfrm_dst_ops_init(net);
b27aeadb
AD		 2956 rv = xfrm_sysctl_init(net);
2957 if (rv < 0)
2958 goto out_sysctl;
d62ddc21
AD		 2959 return 0;
2960
b27aeadb
AD		 2961out_sysctl:
2962 xfrm_policy_fini(net);
d62ddc21
AD		 2963out_policy:
2964 xfrm_state_fini(net);
2965out_state:
59c9940e
AD		 2966 xfrm_statistics_fini(net);
2967out_statistics:
d62ddc21
AD		 2968 return rv;
2969}
2970
2971static void __net_exit xfrm_net_exit(struct net *net)
2972{
b27aeadb 2973 xfrm_sysctl_fini(net);
d62ddc21
AD		 2974 xfrm_policy_fini(net);
2975 xfrm_state_fini(net);
59c9940e 2976 xfrm_statistics_fini(net);
d62ddc21
AD		 2977}
2978
2979static struct pernet_operations __net_initdata xfrm_net_ops = {
2980 .init = xfrm_net_init,
2981 .exit = xfrm_net_exit,
2982};
2983
1da177e4
LT		 2984void __init xfrm_init(void)
2985{
d62ddc21 2986 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 2987 xfrm_input_init();
2988}
2989
ab5f5e8b 2990#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2991static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2992 struct audit_buffer *audit_buf)
ab5f5e8b 2993{
875179fa
PM		 2994 struct xfrm_sec_ctx *ctx = xp->security;
2995 struct xfrm_selector *sel = &xp->selector;
2996
2997 if (ctx)
ab5f5e8b 2998 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2999 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3000
875179fa 3001 switch(sel->family) {
ab5f5e8b 3002 case AF_INET:
21454aaa 3003 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3004 if (sel->prefixlen_s != 32)
3005 audit_log_format(audit_buf, " src_prefixlen=%d",
3006 sel->prefixlen_s);
21454aaa 3007 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3008 if (sel->prefixlen_d != 32)
3009 audit_log_format(audit_buf, " dst_prefixlen=%d",
3010 sel->prefixlen_d);
ab5f5e8b
JL		 3011 break;
3012 case AF_INET6:
5b095d98 3013 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3014 if (sel->prefixlen_s != 128)
3015 audit_log_format(audit_buf, " src_prefixlen=%d",
3016 sel->prefixlen_s);
5b095d98 3017 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3018 if (sel->prefixlen_d != 128)
3019 audit_log_format(audit_buf, " dst_prefixlen=%d",
3020 sel->prefixlen_d);
ab5f5e8b
JL		 3021 break;
3022 }
3023}
3024
68277acc 3025void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
e1760bd5 3026 kuid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 3027{
3028 struct audit_buffer *audit_buf;
ab5f5e8b 3029
afeb14b4 3030 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3031 if (audit_buf == NULL)
3032 return;
2532386f 3033 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 3034 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3035 xfrm_audit_common_policyinfo(xp, audit_buf);
3036 audit_log_end(audit_buf);
3037}
3038EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3039
68277acc 3040void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
e1760bd5 3041 kuid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 3042{
3043 struct audit_buffer *audit_buf;
ab5f5e8b 3044
afeb14b4 3045 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3046 if (audit_buf == NULL)
3047 return;
2532386f 3048 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 3049 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3050 xfrm_audit_common_policyinfo(xp, audit_buf);
3051 audit_log_end(audit_buf);
3052}
3053EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3054#endif
3055
80c9abaa 3056#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3057static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3058 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3059{
3060 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3061 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3062 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3063 sel_cmp->family) &&
3064 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3065 sel_cmp->family) &&
80c9abaa
SS		 3066 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3067 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3068 return true;
80c9abaa
SS		 3069 }
3070 } else {
3071 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3072 return true;
80c9abaa
SS		 3073 }
3074 }
bc9b35ad 3075 return false;
80c9abaa
SS		 3076}
3077
b4b7c0b3 3078static struct xfrm_policy * xfrm_migrate_policy_find(const struct xfrm_selector *sel,
8d549c4f 3079 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3080{
3081 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3082 struct hlist_head *chain;
3083 u32 priority = ~0U;
3084
3085 read_lock_bh(&xfrm_policy_lock);
8d549c4f 3086 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3087 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3088 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3089 pol->type == type) {
3090 ret = pol;
3091 priority = ret->priority;
3092 break;
3093 }
3094 }
8d549c4f 3095 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3096 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3097 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3098 pol->type == type &&
3099 pol->priority < priority) {
3100 ret = pol;
3101 break;
3102 }
3103 }
3104
3105 if (ret)
3106 xfrm_pol_hold(ret);
3107
3108 read_unlock_bh(&xfrm_policy_lock);
3109
3110 return ret;
3111}
3112
dd701754 3113static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3114{
3115 int match = 0;
3116
3117 if (t->mode == m->mode && t->id.proto == m->proto &&
3118 (m->reqid == 0 || t->reqid == m->reqid)) {
3119 switch (t->mode) {
3120 case XFRM_MODE_TUNNEL:
3121 case XFRM_MODE_BEET:
70e94e66
YH		 3122 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3123 m->old_family) &&
3124 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3125 m->old_family)) {
80c9abaa
SS		 3126 match = 1;
3127 }
3128 break;
3129 case XFRM_MODE_TRANSPORT:
3130 /* in case of transport mode, template does not store
3131 any IP addresses, hence we just compare mode and
3132 protocol */
3133 match = 1;
3134 break;
3135 default:
3136 break;
3137 }
3138 }
3139 return match;
3140}
3141
3142/* update endpoint address(es) of template(s) */
3143static int xfrm_policy_migrate(struct xfrm_policy *pol,
3144 struct xfrm_migrate *m, int num_migrate)
3145{
3146 struct xfrm_migrate *mp;
80c9abaa
SS		 3147 int i, j, n = 0;
3148
3149 write_lock_bh(&pol->lock);
12a169e7 3150 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3151 /* target policy has been deleted */
3152 write_unlock_bh(&pol->lock);
3153 return -ENOENT;
3154 }
3155
3156 for (i = 0; i < pol->xfrm_nr; i++) {
3157 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3158 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3159 continue;
3160 n++;
1bfcb10f
HX		 3161 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3162 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3163 continue;
3164 /* update endpoints */
3165 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3166 sizeof(pol->xfrm_vec[i].id.daddr));
3167 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3168 sizeof(pol->xfrm_vec[i].saddr));
3169 pol->xfrm_vec[i].encap_family = mp->new_family;
3170 /* flush bundles */
80c802f3 3171 atomic_inc(&pol->genid);
80c9abaa
SS		 3172 }
3173 }
3174
3175 write_unlock_bh(&pol->lock);
3176
3177 if (!n)
3178 return -ENODATA;
3179
3180 return 0;
3181}
3182
dd701754 3183static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3184{
3185 int i, j;
3186
3187 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3188 return -EINVAL;
3189
3190 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3191 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3192 m[i].old_family) &&
3193 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3194 m[i].old_family))
80c9abaa
SS		 3195 return -EINVAL;
3196 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3197 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3198 return -EINVAL;
3199
3200 /* check if there is any duplicated entry */
3201 for (j = i + 1; j < num_migrate; j++) {
3202 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3203 sizeof(m[i].old_daddr)) &&
3204 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3205 sizeof(m[i].old_saddr)) &&
3206 m[i].proto == m[j].proto &&
3207 m[i].mode == m[j].mode &&
3208 m[i].reqid == m[j].reqid &&
3209 m[i].old_family == m[j].old_family)
3210 return -EINVAL;
3211 }
3212 }
3213
3214 return 0;
3215}
3216
b4b7c0b3 3217int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3218 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3219 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3220{
3221 int i, err, nx_cur = 0, nx_new = 0;
3222 struct xfrm_policy *pol = NULL;
3223 struct xfrm_state *x, *xc;
3224 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3225 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3226 struct xfrm_migrate *mp;
3227
3228 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3229 goto out;
3230
3231 /* Stage 1 - find policy */
8d549c4f 3232 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3233 err = -ENOENT;
3234 goto out;
3235 }
3236
3237 /* Stage 2 - find and update state(s) */
3238 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
3239 if ((x = xfrm_migrate_state_find(mp))) {
3240 x_cur[nx_cur] = x;
3241 nx_cur++;
3242 if ((xc = xfrm_state_migrate(x, mp))) {
3243 x_new[nx_new] = xc;
3244 nx_new++;
3245 } else {
3246 err = -ENODATA;
3247 goto restore_state;
3248 }
3249 }
3250 }
3251
3252 /* Stage 3 - update policy */
3253 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3254 goto restore_state;
3255
3256 /* Stage 4 - delete old state(s) */
3257 if (nx_cur) {
3258 xfrm_states_put(x_cur, nx_cur);
3259 xfrm_states_delete(x_cur, nx_cur);
3260 }
3261
3262 /* Stage 5 - announce */
13c1d189 3263 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3264
3265 xfrm_pol_put(pol);
3266
3267 return 0;
3268out:
3269 return err;
3270
3271restore_state:
3272 if (pol)
3273 xfrm_pol_put(pol);
3274 if (nx_cur)
3275 xfrm_states_put(x_cur, nx_cur);
3276 if (nx_new)
3277 xfrm_states_delete(x_new, nx_new);
3278
3279 return err;
3280}
e610e679 3281EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3282#endif
Ubuntu Artful kernel mirror