projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
xfrm: Add refcount handling to queued policies
[linux-2.6-block.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. */
1121994c 541static u32 xfrm_gen_index(struct net *net, int dir)
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
1da177e4
LT		 551 idx = (idx_generator | dir);
552 idx_generator += 8;
553 if (idx == 0)
554 idx = 8;
1121994c 555 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 556 found = 0;
b67bfe0d 557 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 558 if (p->index == idx) {
559 found = 1;
1da177e4 560 break;
2518c7c2 561 }
1da177e4 562 }
2518c7c2 563 if (!found)
1da177e4
LT		 564 return idx;
565 }
566}
567
2518c7c2
DM		 568static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
569{
570 u32 *p1 = (u32 *) s1;
571 u32 *p2 = (u32 *) s2;
572 int len = sizeof(struct xfrm_selector) / sizeof(u32);
573 int i;
574
575 for (i = 0; i < len; i++) {
576 if (p1[i] != p2[i])
577 return 1;
578 }
579
580 return 0;
581}
582
a0073fe1
SK		 583static void xfrm_policy_requeue(struct xfrm_policy *old,
584 struct xfrm_policy *new)
585{
586 struct xfrm_policy_queue *pq = &old->polq;
587 struct sk_buff_head list;
588
589 __skb_queue_head_init(&list);
590
591 spin_lock_bh(&pq->hold_queue.lock);
592 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 593 if (del_timer(&pq->hold_timer))
594 xfrm_pol_put(old);
a0073fe1
SK		 595 spin_unlock_bh(&pq->hold_queue.lock);
596
597 if (skb_queue_empty(&list))
598 return;
599
600 pq = &new->polq;
601
602 spin_lock_bh(&pq->hold_queue.lock);
603 skb_queue_splice(&list, &pq->hold_queue);
604 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 605 if (!mod_timer(&pq->hold_timer, jiffies))
606 xfrm_pol_hold(new);
a0073fe1
SK		 607 spin_unlock_bh(&pq->hold_queue.lock);
608}
609
7cb8a939
SK		 610static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
611 struct xfrm_policy *pol)
612{
613 u32 mark = policy->mark.v & policy->mark.m;
614
615 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
616 return true;
617
618 if ((mark & pol->mark.m) == pol->mark.v &&
619 policy->priority == pol->priority)
620 return true;
621
622 return false;
623}
624
1da177e4
LT		 625int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
626{
1121994c 627 struct net *net = xp_net(policy);
2518c7c2
DM		 628 struct xfrm_policy *pol;
629 struct xfrm_policy *delpol;
630 struct hlist_head *chain;
b67bfe0d 631 struct hlist_node *newpos;
1da177e4
LT		 632
633 write_lock_bh(&xfrm_policy_lock);
1121994c 634 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 635 delpol = NULL;
636 newpos = NULL;
b67bfe0d 637 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 638 if (pol->type == policy->type &&
2518c7c2 639 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 640 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 641 xfrm_sec_ctx_match(pol->security, policy->security) &&
642 !WARN_ON(delpol)) {
1da177e4
LT		 643 if (excl) {
644 write_unlock_bh(&xfrm_policy_lock);
645 return -EEXIST;
646 }
1da177e4
LT		 647 delpol = pol;
648 if (policy->priority > pol->priority)
649 continue;
650 } else if (policy->priority >= pol->priority) {
a6c7ab55 651 newpos = &pol->bydst;
1da177e4
LT		 652 continue;
653 }
1da177e4
LT		 654 if (delpol)
655 break;
1da177e4
LT		 656 }
657 if (newpos)
2518c7c2
DM		 658 hlist_add_after(newpos, &policy->bydst);
659 else
660 hlist_add_head(&policy->bydst, chain);
1da177e4 661 xfrm_pol_hold(policy);
1121994c 662 net->xfrm.policy_count[dir]++;
1da177e4 663 atomic_inc(&flow_cache_genid);
ca4c3fc2 664
665 /* After previous checking, family can either be AF_INET or AF_INET6 */
666 if (policy->family == AF_INET)
667 rt_genid_bump_ipv4(net);
668 else
669 rt_genid_bump_ipv6(net);
670
a0073fe1
SK		 671 if (delpol) {
672 xfrm_policy_requeue(delpol, policy);
29fa0b30 673 __xfrm_policy_unlink(delpol, dir);
a0073fe1 674 }
1121994c
AD		 675 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
676 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 677 policy->curlft.add_time = get_seconds();
1da177e4
LT		 678 policy->curlft.use_time = 0;
679 if (!mod_timer(&policy->timer, jiffies + HZ))
680 xfrm_pol_hold(policy);
1121994c 681 list_add(&policy->walk.all, &net->xfrm.policy_all);
1da177e4
LT		 682 write_unlock_bh(&xfrm_policy_lock);
683
9b78a82c 684 if (delpol)
1da177e4 685 xfrm_policy_kill(delpol);
1121994c
AD		 686 else if (xfrm_bydst_should_resize(net, dir, NULL))
687 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 688
1da177e4
LT		 689 return 0;
690}
691EXPORT_SYMBOL(xfrm_policy_insert);
692
8ca2e93b
JHS		 693struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
694 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 695 struct xfrm_sec_ctx *ctx, int delete,
696 int *err)
1da177e4 697{
2518c7c2
DM		 698 struct xfrm_policy *pol, *ret;
699 struct hlist_head *chain;
1da177e4 700
ef41aaa0 701 *err = 0;
1da177e4 702 write_lock_bh(&xfrm_policy_lock);
8d1211a6 703 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 704 ret = NULL;
b67bfe0d 705 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 706 if (pol->type == type &&
34f8d884 707 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 708 !selector_cmp(sel, &pol->selector) &&
709 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 710 xfrm_pol_hold(pol);
2518c7c2 711 if (delete) {
03e1ad7b
PM		 712 *err = security_xfrm_policy_delete(
713 pol->security);
ef41aaa0
EP		 714 if (*err) {
715 write_unlock_bh(&xfrm_policy_lock);
716 return pol;
717 }
29fa0b30 718 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 719 }
720 ret = pol;
1da177e4
LT		 721 break;
722 }
723 }
724 write_unlock_bh(&xfrm_policy_lock);
725
fe1a5f03 726 if (ret && delete)
2518c7c2 727 xfrm_policy_kill(ret);
2518c7c2 728 return ret;
1da177e4 729}
df71837d 730EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 731
8ca2e93b
JHS		 732struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
733 int dir, u32 id, int delete, int *err)
1da177e4 734{
2518c7c2
DM		 735 struct xfrm_policy *pol, *ret;
736 struct hlist_head *chain;
1da177e4 737
b5505c6e
HX		 738 *err = -ENOENT;
739 if (xfrm_policy_id2dir(id) != dir)
740 return NULL;
741
ef41aaa0 742 *err = 0;
1da177e4 743 write_lock_bh(&xfrm_policy_lock);
8d1211a6 744 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 745 ret = NULL;
b67bfe0d 746 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 747 if (pol->type == type && pol->index == id &&
748 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 749 xfrm_pol_hold(pol);
2518c7c2 750 if (delete) {
03e1ad7b
PM		 751 *err = security_xfrm_policy_delete(
752 pol->security);
ef41aaa0
EP		 753 if (*err) {
754 write_unlock_bh(&xfrm_policy_lock);
755 return pol;
756 }
29fa0b30 757 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 758 }
759 ret = pol;
1da177e4
LT		 760 break;
761 }
762 }
763 write_unlock_bh(&xfrm_policy_lock);
764
fe1a5f03 765 if (ret && delete)
2518c7c2 766 xfrm_policy_kill(ret);
2518c7c2 767 return ret;
1da177e4
LT		 768}
769EXPORT_SYMBOL(xfrm_policy_byid);
770
4aa2e62c
JL		 771#ifdef CONFIG_SECURITY_NETWORK_XFRM
772static inline int
33ffbbd5 773xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
1da177e4 774{
4aa2e62c
JL		 775 int dir, err = 0;
776
777 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
778 struct xfrm_policy *pol;
4aa2e62c
JL		 779 int i;
780
b67bfe0d 781 hlist_for_each_entry(pol,
33ffbbd5 782 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 783 if (pol->type != type)
784 continue;
03e1ad7b 785 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 786 if (err) {
ab5f5e8b
JL		 787 xfrm_audit_policy_delete(pol, 0,
788 audit_info->loginuid,
2532386f 789 audit_info->sessionid,
ab5f5e8b 790 audit_info->secid);
4aa2e62c
JL		 791 return err;
792 }
7dc12d6d 793 }
33ffbbd5 794 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 795 hlist_for_each_entry(pol,
33ffbbd5 796 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 797 bydst) {
798 if (pol->type != type)
799 continue;
03e1ad7b
PM		 800 err = security_xfrm_policy_delete(
801 pol->security);
4aa2e62c 802 if (err) {
ab5f5e8b
JL		 803 xfrm_audit_policy_delete(pol, 0,
804 audit_info->loginuid,
2532386f 805 audit_info->sessionid,
ab5f5e8b 806 audit_info->secid);
4aa2e62c
JL		 807 return err;
808 }
809 }
810 }
811 }
812 return err;
813}
814#else
815static inline int
33ffbbd5 816xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c
JL		 817{
818 return 0;
819}
820#endif
821
33ffbbd5 822int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c 823{
2f1eb65f 824 int dir, err = 0, cnt = 0;
1da177e4
LT		 825
826 write_lock_bh(&xfrm_policy_lock);
4aa2e62c 827
33ffbbd5 828 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
4aa2e62c
JL		 829 if (err)
830 goto out;
831
1da177e4 832 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 833 struct xfrm_policy *pol;
29fa0b30 834 int i;
2518c7c2
DM		 835
836 again1:
b67bfe0d 837 hlist_for_each_entry(pol,
33ffbbd5 838 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 839 if (pol->type != type)
840 continue;
ea2dea9d 841 __xfrm_policy_unlink(pol, dir);
1da177e4 842 write_unlock_bh(&xfrm_policy_lock);
ea2dea9d 843 cnt++;
1da177e4 844
ab5f5e8b 845 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
2532386f 846 audit_info->sessionid,
ab5f5e8b 847 audit_info->secid);
161a09e7 848
2518c7c2 849 xfrm_policy_kill(pol);
1da177e4
LT		 850
851 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 852 goto again1;
853 }
854
33ffbbd5 855 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 856 again2:
b67bfe0d 857 hlist_for_each_entry(pol,
33ffbbd5 858 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 859 bydst) {
860 if (pol->type != type)
861 continue;
ea2dea9d 862 __xfrm_policy_unlink(pol, dir);
2518c7c2 863 write_unlock_bh(&xfrm_policy_lock);
ea2dea9d 864 cnt++;
2518c7c2 865
ab5f5e8b
JL		 866 xfrm_audit_policy_delete(pol, 1,
867 audit_info->loginuid,
2532386f 868 audit_info->sessionid,
ab5f5e8b 869 audit_info->secid);
2518c7c2
DM		 870 xfrm_policy_kill(pol);
871
872 write_lock_bh(&xfrm_policy_lock);
873 goto again2;
874 }
1da177e4 875 }
2518c7c2 876
1da177e4 877 }
2f1eb65f
JHS		 878 if (!cnt)
879 err = -ESRCH;
4aa2e62c 880out:
1da177e4 881 write_unlock_bh(&xfrm_policy_lock);
4aa2e62c 882 return err;
1da177e4
LT		 883}
884EXPORT_SYMBOL(xfrm_policy_flush);
885
cdcbca7c 886int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 887 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 888 void *data)
889{
12a169e7
HX		 890 struct xfrm_policy *pol;
891 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 892 int error = 0;
893
894 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
895 walk->type != XFRM_POLICY_TYPE_ANY)
896 return -EINVAL;
1da177e4 897
12a169e7 898 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 899 return 0;
900
12a169e7
HX		 901 write_lock_bh(&xfrm_policy_lock);
902 if (list_empty(&walk->walk.all))
cdcbca7c 903 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 904 else
905 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 906 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 907 if (x->dead)
4c563f76 908 continue;
12a169e7
HX		 909 pol = container_of(x, struct xfrm_policy, walk);
910 if (walk->type != XFRM_POLICY_TYPE_ANY &&
911 walk->type != pol->type)
912 continue;
913 error = func(pol, xfrm_policy_id2dir(pol->index),
914 walk->seq, data);
915 if (error) {
916 list_move_tail(&walk->walk.all, &x->all);
917 goto out;
2518c7c2 918 }
12a169e7 919 walk->seq++;
1da177e4 920 }
12a169e7 921 if (walk->seq == 0) {
baf5d743
JHS		 922 error = -ENOENT;
923 goto out;
924 }
12a169e7 925 list_del_init(&walk->walk.all);
1da177e4 926out:
12a169e7 927 write_unlock_bh(&xfrm_policy_lock);
1da177e4
LT		 928 return error;
929}
930EXPORT_SYMBOL(xfrm_policy_walk);
931
12a169e7
HX		 932void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
933{
934 INIT_LIST_HEAD(&walk->walk.all);
935 walk->walk.dead = 1;
936 walk->type = type;
937 walk->seq = 0;
938}
939EXPORT_SYMBOL(xfrm_policy_walk_init);
940
941void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
942{
943 if (list_empty(&walk->walk.all))
944 return;
945
946 write_lock_bh(&xfrm_policy_lock);
947 list_del(&walk->walk.all);
948 write_unlock_bh(&xfrm_policy_lock);
949}
950EXPORT_SYMBOL(xfrm_policy_walk_done);
951
134b0fc5
JM		 952/*
953 * Find policy to apply to this flow.
954 *
955 * Returns 0 if policy found, else an -errno.
956 */
f299d557
DM		 957static int xfrm_policy_match(const struct xfrm_policy *pol,
958 const struct flowi *fl,
2518c7c2 959 u8 type, u16 family, int dir)
1da177e4 960{
f299d557 961 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 962 int ret = -ESRCH;
963 bool match;
1da177e4 964
2518c7c2 965 if (pol->family != family ||
1d28f42c 966 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 967 pol->type != type)
134b0fc5 968 return ret;
1da177e4 969
2518c7c2 970 match = xfrm_selector_match(sel, fl, family);
134b0fc5 971 if (match)
1d28f42c 972 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 973 dir);
2518c7c2 974
134b0fc5 975 return ret;
2518c7c2 976}
1da177e4 977
52479b62 978static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 979 const struct flowi *fl,
2518c7c2
DM		 980 u16 family, u8 dir)
981{
134b0fc5 982 int err;
2518c7c2 983 struct xfrm_policy *pol, *ret;
0b597e7e 984 const xfrm_address_t *daddr, *saddr;
2518c7c2 985 struct hlist_head *chain;
acba48e1 986 u32 priority = ~0U;
df71837d 987
2518c7c2
DM		 988 daddr = xfrm_flowi_daddr(fl, family);
989 saddr = xfrm_flowi_saddr(fl, family);
990 if (unlikely(!daddr || !saddr))
991 return NULL;
992
993 read_lock_bh(&xfrm_policy_lock);
52479b62 994 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 995 ret = NULL;
b67bfe0d 996 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 997 err = xfrm_policy_match(pol, fl, type, family, dir);
998 if (err) {
999 if (err == -ESRCH)
1000 continue;
1001 else {
1002 ret = ERR_PTR(err);
1003 goto fail;
1004 }
1005 } else {
2518c7c2 1006 ret = pol;
acba48e1 1007 priority = ret->priority;
2518c7c2
DM		 1008 break;
1009 }
1010 }
52479b62 1011 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1012 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1013 err = xfrm_policy_match(pol, fl, type, family, dir);
1014 if (err) {
1015 if (err == -ESRCH)
1016 continue;
1017 else {
1018 ret = ERR_PTR(err);
1019 goto fail;
1020 }
1021 } else if (pol->priority < priority) {
acba48e1
DM		 1022 ret = pol;
1023 break;
1da177e4
LT		 1024 }
1025 }
acba48e1
DM		 1026 if (ret)
1027 xfrm_pol_hold(ret);
134b0fc5 1028fail:
1da177e4 1029 read_unlock_bh(&xfrm_policy_lock);
4e81bb83 1030
2518c7c2 1031 return ret;
4e81bb83
MN		 1032}
1033
80c802f3 1034static struct xfrm_policy *
73ff93cd 1035__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1036{
1037#ifdef CONFIG_XFRM_SUB_POLICY
1038 struct xfrm_policy *pol;
1039
1040 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1041 if (pol != NULL)
1042 return pol;
1043#endif
1044 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1045}
1046
b5fb82c4
BZ		 1047static int flow_to_policy_dir(int dir)
1048{
1049 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1050 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1051 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1052 return dir;
1053
1054 switch (dir) {
1055 default:
1056 case FLOW_DIR_IN:
1057 return XFRM_POLICY_IN;
1058 case FLOW_DIR_OUT:
1059 return XFRM_POLICY_OUT;
1060 case FLOW_DIR_FWD:
1061 return XFRM_POLICY_FWD;
1062 }
1063}
1064
fe1a5f03 1065static struct flow_cache_object *
dee9f4bc 1066xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1067 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1068{
1069 struct xfrm_policy *pol;
fe1a5f03
TT		 1070
1071 if (old_obj)
1072 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1073
b5fb82c4 1074 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1075 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1076 return ERR_CAST(pol);
fe1a5f03 1077
fe1a5f03
TT		 1078 /* Resolver returns two references:
1079 * one for cache and one for caller of flow_cache_lookup() */
1080 xfrm_pol_hold(pol);
1081
1082 return &pol->flo;
1da177e4
LT		 1083}
1084
df71837d
TJ		 1085static inline int policy_to_flow_dir(int dir)
1086{
1087 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1088 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1089 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1090 return dir;
1091 switch (dir) {
1092 default:
1093 case XFRM_POLICY_IN:
1094 return FLOW_DIR_IN;
1095 case XFRM_POLICY_OUT:
1096 return FLOW_DIR_OUT;
1097 case XFRM_POLICY_FWD:
1098 return FLOW_DIR_FWD;
3ff50b79 1099 }
df71837d
TJ		 1100}
1101
dee9f4bc
DM		 1102static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1103 const struct flowi *fl)
1da177e4
LT		 1104{
1105 struct xfrm_policy *pol;
1106
1107 read_lock_bh(&xfrm_policy_lock);
1108 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1109 bool match = xfrm_selector_match(&pol->selector, fl,
1110 sk->sk_family);
a716c119 1111 int err = 0;
df71837d 1112
3bccfbc7 1113 if (match) {
34f8d884
JHS		 1114 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1115 pol = NULL;
1116 goto out;
1117 }
03e1ad7b 1118 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1119 fl->flowi_secid,
03e1ad7b 1120 policy_to_flow_dir(dir));
3bccfbc7
VY		 1121 if (!err)
1122 xfrm_pol_hold(pol);
1123 else if (err == -ESRCH)
1124 pol = NULL;
1125 else
1126 pol = ERR_PTR(err);
1127 } else
1da177e4
LT		 1128 pol = NULL;
1129 }
34f8d884 1130out:
1da177e4
LT		 1131 read_unlock_bh(&xfrm_policy_lock);
1132 return pol;
1133}
1134
1135static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1136{
98806f75 1137 struct net *net = xp_net(pol);
1121994c 1138 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1139 pol->family, dir);
4e81bb83 1140
98806f75 1141 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1142 hlist_add_head(&pol->bydst, chain);
e92303f8 1143 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1144 net->xfrm.policy_count[dir]++;
1da177e4 1145 xfrm_pol_hold(pol);
2518c7c2 1146
98806f75
AD		 1147 if (xfrm_bydst_should_resize(net, dir, NULL))
1148 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1149}
1150
1151static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1152 int dir)
1153{
98806f75
AD		 1154 struct net *net = xp_net(pol);
1155
2518c7c2
DM		 1156 if (hlist_unhashed(&pol->bydst))
1157 return NULL;
1da177e4 1158
2518c7c2
DM		 1159 hlist_del(&pol->bydst);
1160 hlist_del(&pol->byidx);
12a169e7 1161 list_del(&pol->walk.all);
98806f75 1162 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1163
1164 return pol;
1da177e4
LT		 1165}
1166
4666faab 1167int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4
LT		 1168{
1169 write_lock_bh(&xfrm_policy_lock);
1170 pol = __xfrm_policy_unlink(pol, dir);
1171 write_unlock_bh(&xfrm_policy_lock);
1172 if (pol) {
1da177e4 1173 xfrm_policy_kill(pol);
4666faab 1174 return 0;
1da177e4 1175 }
4666faab 1176 return -ENOENT;
1da177e4 1177}
a70fcb0b 1178EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1179
1180int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1181{
1121994c 1182 struct net *net = xp_net(pol);
1da177e4
LT		 1183 struct xfrm_policy *old_pol;
1184
4e81bb83
MN		 1185#ifdef CONFIG_XFRM_SUB_POLICY
1186 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1187 return -EINVAL;
1188#endif
1189
1da177e4
LT		 1190 write_lock_bh(&xfrm_policy_lock);
1191 old_pol = sk->sk_policy[dir];
1192 sk->sk_policy[dir] = pol;
1193 if (pol) {
9d729f72 1194 pol->curlft.add_time = get_seconds();
1121994c 1195 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1da177e4
LT		 1196 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1197 }
a0073fe1
SK		 1198 if (old_pol) {
1199 if (pol)
1200 xfrm_policy_requeue(old_pol, pol);
1201
ea2dea9d
TT		 1202 /* Unlinking succeeds always. This is the only function
1203 * allowed to delete or replace socket policy.
1204 */
1da177e4 1205 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
a0073fe1 1206 }
1da177e4
LT		 1207 write_unlock_bh(&xfrm_policy_lock);
1208
1209 if (old_pol) {
1210 xfrm_policy_kill(old_pol);
1211 }
1212 return 0;
1213}
1214
d3e40a9f 1215static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1216{
0331b1f3 1217 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1da177e4
LT		 1218
1219 if (newp) {
1220 newp->selector = old->selector;
03e1ad7b
PM		 1221 if (security_xfrm_policy_clone(old->security,
1222 &newp->security)) {
df71837d
TJ		 1223 kfree(newp);
1224 return NULL; /* ENOMEM */
1225 }
1da177e4
LT		 1226 newp->lft = old->lft;
1227 newp->curlft = old->curlft;
fb977e2c 1228 newp->mark = old->mark;
1da177e4
LT		 1229 newp->action = old->action;
1230 newp->flags = old->flags;
1231 newp->xfrm_nr = old->xfrm_nr;
1232 newp->index = old->index;
4e81bb83 1233 newp->type = old->type;
1da177e4
LT		 1234 memcpy(newp->xfrm_vec, old->xfrm_vec,
1235 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1236 write_lock_bh(&xfrm_policy_lock);
1237 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1238 write_unlock_bh(&xfrm_policy_lock);
1239 xfrm_pol_put(newp);
1240 }
1241 return newp;
1242}
1243
1244int __xfrm_sk_clone_policy(struct sock *sk)
1245{
1246 struct xfrm_policy *p0 = sk->sk_policy[0],
1247 *p1 = sk->sk_policy[1];
1248
1249 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1250 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1251 return -ENOMEM;
1252 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1253 return -ENOMEM;
1254 return 0;
1255}
1256
a1e59abf 1257static int
fbda33b2 1258xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1259 unsigned short family)
1260{
1261 int err;
1262 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1263
1264 if (unlikely(afinfo == NULL))
1265 return -EINVAL;
fbda33b2 1266 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1267 xfrm_policy_put_afinfo(afinfo);
1268 return err;
1269}
1270
1da177e4
LT		 1271/* Resolve list of templates for the flow, given policy. */
1272
1273static int
a6c2e611
DM		 1274xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1275 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1276{
fbda33b2 1277 struct net *net = xp_net(policy);
1da177e4
LT		 1278 int nx;
1279 int i, error;
1280 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1281 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1282 xfrm_address_t tmp;
1da177e4
LT		 1283
1284 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1285 struct xfrm_state *x;
1286 xfrm_address_t *remote = daddr;
1287 xfrm_address_t *local = saddr;
1288 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1289
48b8d783
JK		 1290 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1291 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1292 remote = &tmpl->id.daddr;
1293 local = &tmpl->saddr;
8444cf71
TE		 1294 if (xfrm_addr_any(local, tmpl->encap_family)) {
1295 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1296 if (error)
1297 goto fail;
1298 local = &tmp;
1299 }
1da177e4
LT		 1300 }
1301
1302 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1303
1304 if (x && x->km.state == XFRM_STATE_VALID) {
1305 xfrm[nx++] = x;
1306 daddr = remote;
1307 saddr = local;
1308 continue;
1309 }
1310 if (x) {
1311 error = (x->km.state == XFRM_STATE_ERROR ?
1312 -EINVAL : -EAGAIN);
1313 xfrm_state_put(x);
1314 }
a4322266 1315 else if (error == -ESRCH)
1316 error = -EAGAIN;
1da177e4
LT		 1317
1318 if (!tmpl->optional)
1319 goto fail;
1320 }
1321 return nx;
1322
1323fail:
1324 for (nx--; nx>=0; nx--)
1325 xfrm_state_put(xfrm[nx]);
1326 return error;
1327}
1328
4e81bb83 1329static int
a6c2e611
DM		 1330xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1331 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1332{
41a49cc3
MN		 1333 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1334 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1335 int cnx = 0;
1336 int error;
1337 int ret;
1338 int i;
1339
1340 for (i = 0; i < npols; i++) {
1341 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1342 error = -ENOBUFS;
1343 goto fail;
1344 }
41a49cc3
MN		 1345
1346 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1347 if (ret < 0) {
1348 error = ret;
1349 goto fail;
1350 } else
1351 cnx += ret;
1352 }
1353
41a49cc3
MN		 1354 /* found states are sorted for outbound processing */
1355 if (npols > 1)
1356 xfrm_state_sort(xfrm, tpp, cnx, family);
1357
4e81bb83
MN		 1358 return cnx;
1359
1360 fail:
1361 for (cnx--; cnx>=0; cnx--)
41a49cc3 1362 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1363 return error;
1364
1365}
1366
1da177e4
LT		 1367/* Check that the bundle accepts the flow and its components are
1368 * still valid.
1369 */
1370
05d84025 1371static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1372{
1373 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1374 int tos;
1da177e4 1375
25ee3286
HX		 1376 if (!afinfo)
1377 return -EINVAL;
1378
1379 tos = afinfo->get_tos(fl);
1380
1381 xfrm_policy_put_afinfo(afinfo);
1382
1383 return tos;
1384}
1385
80c802f3
TT		 1386static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1387{
1388 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1389 struct dst_entry *dst = &xdst->u.dst;
1390
1391 if (xdst->route == NULL) {
1392 /* Dummy bundle - if it has xfrms we were not
1393 * able to build bundle as template resolution failed.
1394 * It means we need to try again resolving. */
1395 if (xdst->num_xfrms > 0)
1396 return NULL;
a0073fe1
SK		 1397 } else if (dst->flags & DST_XFRM_QUEUE) {
1398 return NULL;
80c802f3
TT		 1399 } else {
1400 /* Real bundle */
1401 if (stale_bundle(dst))
1402 return NULL;
1403 }
1404
1405 dst_hold(dst);
1406 return flo;
1407}
1408
1409static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1410{
1411 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1412 struct dst_entry *dst = &xdst->u.dst;
1413
1414 if (!xdst->route)
1415 return 0;
1416 if (stale_bundle(dst))
1417 return 0;
1418
1419 return 1;
1420}
1421
1422static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1423{
1424 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1425 struct dst_entry *dst = &xdst->u.dst;
1426
1427 dst_free(dst);
1428}
1429
1430static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1431 .get = xfrm_bundle_flo_get,
1432 .check = xfrm_bundle_flo_check,
1433 .delete = xfrm_bundle_flo_delete,
1434};
1435
d7c7544c 1436static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1437{
1da177e4 1438 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1439 struct dst_ops *dst_ops;
25ee3286
HX		 1440 struct xfrm_dst *xdst;
1441
1442 if (!afinfo)
1443 return ERR_PTR(-EINVAL);
1444
d7c7544c
AD		 1445 switch (family) {
1446 case AF_INET:
1447 dst_ops = &net->xfrm.xfrm4_dst_ops;
1448 break;
dfd56b8b 1449#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1450 case AF_INET6:
1451 dst_ops = &net->xfrm.xfrm6_dst_ops;
1452 break;
1453#endif
1454 default:
1455 BUG();
1456 }
f5b0a874 1457 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1458
d4cae562 1459 if (likely(xdst)) {
141e369d
SK		 1460 struct dst_entry *dst = &xdst->u.dst;
1461
1462 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1463 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1464 if (afinfo->init_dst)
1465 afinfo->init_dst(net, xdst);
d4cae562 1466 } else
0b150932 1467 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1468
d4cae562
MB		 1469 xfrm_policy_put_afinfo(afinfo);
1470
25ee3286
HX		 1471 return xdst;
1472}
1473
a1b05140
MN		 1474static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1475 int nfheader_len)
1476{
1477 struct xfrm_policy_afinfo *afinfo =
1478 xfrm_policy_get_afinfo(dst->ops->family);
1479 int err;
1480
1481 if (!afinfo)
1482 return -EINVAL;
1483
1484 err = afinfo->init_path(path, dst, nfheader_len);
1485
1486 xfrm_policy_put_afinfo(afinfo);
1487
1488 return err;
1489}
1490
87c1e12b 1491static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1492 const struct flowi *fl)
25ee3286
HX		 1493{
1494 struct xfrm_policy_afinfo *afinfo =
1495 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1496 int err;
1497
1498 if (!afinfo)
1da177e4 1499 return -EINVAL;
25ee3286 1500
87c1e12b 1501 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1502
1da177e4 1503 xfrm_policy_put_afinfo(afinfo);
25ee3286 1504
1da177e4
LT		 1505 return err;
1506}
1507
80c802f3 1508
25ee3286
HX		 1509/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1510 * all the metrics... Shortly, bundle a bundle.
1511 */
1512
1513static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1514 struct xfrm_state **xfrm, int nx,
98313ada 1515 const struct flowi *fl,
25ee3286
HX		 1516 struct dst_entry *dst)
1517{
d7c7544c 1518 struct net *net = xp_net(policy);
25ee3286
HX		 1519 unsigned long now = jiffies;
1520 struct net_device *dev;
43a4dea4 1521 struct xfrm_mode *inner_mode;
25ee3286
HX		 1522 struct dst_entry *dst_prev = NULL;
1523 struct dst_entry *dst0 = NULL;
1524 int i = 0;
1525 int err;
1526 int header_len = 0;
a1b05140 1527 int nfheader_len = 0;
25ee3286
HX		 1528 int trailer_len = 0;
1529 int tos;
1530 int family = policy->selector.family;
9bb182a7
YH		 1531 xfrm_address_t saddr, daddr;
1532
1533 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1534
1535 tos = xfrm_get_tos(fl, family);
1536 err = tos;
1537 if (tos < 0)
1538 goto put_states;
1539
1540 dst_hold(dst);
1541
1542 for (; i < nx; i++) {
d7c7544c 1543 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1544 struct dst_entry *dst1 = &xdst->u.dst;
1545
1546 err = PTR_ERR(xdst);
1547 if (IS_ERR(xdst)) {
1548 dst_release(dst);
1549 goto put_states;
1550 }
1551
43a4dea4
SK		 1552 if (xfrm[i]->sel.family == AF_UNSPEC) {
1553 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1554 xfrm_af2proto(family));
1555 if (!inner_mode) {
1556 err = -EAFNOSUPPORT;
1557 dst_release(dst);
1558 goto put_states;
1559 }
1560 } else
1561 inner_mode = xfrm[i]->inner_mode;
1562
25ee3286
HX		 1563 if (!dst_prev)
1564 dst0 = dst1;
1565 else {
1566 dst_prev->child = dst_clone(dst1);
1567 dst1->flags |= DST_NOHASH;
1568 }
1569
1570 xdst->route = dst;
defb3519 1571 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1572
1573 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1574 family = xfrm[i]->props.family;
9bb182a7
YH		 1575 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1576 family);
25ee3286
HX		 1577 err = PTR_ERR(dst);
1578 if (IS_ERR(dst))
1579 goto put_states;
1580 } else
1581 dst_hold(dst);
1582
1583 dst1->xfrm = xfrm[i];
80c802f3 1584 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1585
f5b0a874 1586 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1587 dst1->flags |= DST_HOST;
1588 dst1->lastuse = now;
1589
1590 dst1->input = dst_discard;
43a4dea4 1591 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1592
1593 dst1->next = dst_prev;
1594 dst_prev = dst1;
1595
1596 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1597 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1598 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1599 trailer_len += xfrm[i]->props.trailer_len;
1600 }
1601
1602 dst_prev->child = dst;
1603 dst0->path = dst;
1604
1605 err = -ENODEV;
1606 dev = dst->dev;
1607 if (!dev)
1608 goto free_dst;
1609
a1b05140 1610 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1611 xfrm_init_pmtu(dst_prev);
1612
1613 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1614 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1615
87c1e12b 1616 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1617 if (err)
1618 goto free_dst;
1619
1620 dst_prev->header_len = header_len;
1621 dst_prev->trailer_len = trailer_len;
1622 header_len -= xdst->u.dst.xfrm->props.header_len;
1623 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1624 }
1625
1626out:
1627 return dst0;
1628
1629put_states:
1630 for (; i < nx; i++)
1631 xfrm_state_put(xfrm[i]);
1632free_dst:
1633 if (dst0)
1634 dst_free(dst0);
1635 dst0 = ERR_PTR(err);
1636 goto out;
1637}
1638
157bfc25 1639static int inline
3f0e18fb 1640xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1641{
1642 if (!*target) {
1643 *target = kmalloc(size, GFP_ATOMIC);
1644 if (!*target)
1645 return -ENOMEM;
1646 }
1647 memcpy(*target, src, size);
1648 return 0;
1649}
1650
1651static int inline
1786b389 1652xfrm_dst_update_parent(struct dst_entry *dst, const struct xfrm_selector *sel)
157bfc25
MN		 1653{
1654#ifdef CONFIG_XFRM_SUB_POLICY
1655 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1656 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1657 sel, sizeof(*sel));
1658#else
1659 return 0;
1660#endif
1661}
1662
1663static int inline
3f0e18fb 1664xfrm_dst_update_origin(struct dst_entry *dst, const struct flowi *fl)
157bfc25
MN		 1665{
1666#ifdef CONFIG_XFRM_SUB_POLICY
1667 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1668 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1669#else
1670 return 0;
1671#endif
1672}
1da177e4 1673
73ff93cd 1674static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1675 struct xfrm_policy **pols,
1676 int *num_pols, int *num_xfrms)
1677{
1678 int i;
1679
1680 if (*num_pols == 0 || !pols[0]) {
1681 *num_pols = 0;
1682 *num_xfrms = 0;
1683 return 0;
1684 }
1685 if (IS_ERR(pols[0]))
1686 return PTR_ERR(pols[0]);
1687
1688 *num_xfrms = pols[0]->xfrm_nr;
1689
1690#ifdef CONFIG_XFRM_SUB_POLICY
1691 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1692 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1693 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1694 XFRM_POLICY_TYPE_MAIN,
1695 fl, family,
1696 XFRM_POLICY_OUT);
1697 if (pols[1]) {
1698 if (IS_ERR(pols[1])) {
1699 xfrm_pols_put(pols, *num_pols);
1700 return PTR_ERR(pols[1]);
1701 }
1702 (*num_pols) ++;
1703 (*num_xfrms) += pols[1]->xfrm_nr;
1704 }
1705 }
1706#endif
1707 for (i = 0; i < *num_pols; i++) {
1708 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1709 *num_xfrms = -1;
1710 break;
1711 }
1712 }
1713
1714 return 0;
1715
1716}
1717
1718static struct xfrm_dst *
1719xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1720 const struct flowi *fl, u16 family,
80c802f3
TT		 1721 struct dst_entry *dst_orig)
1722{
1723 struct net *net = xp_net(pols[0]);
1724 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1725 struct dst_entry *dst;
1726 struct xfrm_dst *xdst;
1727 int err;
1728
1729 /* Try to instantiate a bundle */
1730 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1731 if (err <= 0) {
1732 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1733 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1734 return ERR_PTR(err);
1735 }
1736
1737 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1738 if (IS_ERR(dst)) {
1739 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1740 return ERR_CAST(dst);
1741 }
1742
1743 xdst = (struct xfrm_dst *)dst;
1744 xdst->num_xfrms = err;
1745 if (num_pols > 1)
1746 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1747 else
1748 err = xfrm_dst_update_origin(dst, fl);
1749 if (unlikely(err)) {
1750 dst_free(dst);
1751 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1752 return ERR_PTR(err);
1753 }
1754
1755 xdst->num_pols = num_pols;
1756 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1757 xdst->policy_genid = atomic_read(&pols[0]->genid);
1758
1759 return xdst;
1760}
1761
a0073fe1
SK		 1762static void xfrm_policy_queue_process(unsigned long arg)
1763{
1764 int err = 0;
1765 struct sk_buff *skb;
1766 struct sock *sk;
1767 struct dst_entry *dst;
a0073fe1
SK		 1768 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1769 struct xfrm_policy_queue *pq = &pol->polq;
1770 struct flowi fl;
1771 struct sk_buff_head list;
1772
1773 spin_lock(&pq->hold_queue.lock);
1774 skb = skb_peek(&pq->hold_queue);
1775 dst = skb_dst(skb);
1776 sk = skb->sk;
1777 xfrm_decode_session(skb, &fl, dst->ops->family);
1778 spin_unlock(&pq->hold_queue.lock);
1779
1780 dst_hold(dst->path);
1781 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1782 sk, 0);
1783 if (IS_ERR(dst))
1784 goto purge_queue;
1785
1786 if (dst->flags & DST_XFRM_QUEUE) {
1787 dst_release(dst);
1788
1789 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1790 goto purge_queue;
1791
1792 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1793 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1794 xfrm_pol_hold(pol);
1795 goto out;
a0073fe1
SK		 1796 }
1797
1798 dst_release(dst);
1799
1800 __skb_queue_head_init(&list);
1801
1802 spin_lock(&pq->hold_queue.lock);
1803 pq->timeout = 0;
1804 skb_queue_splice_init(&pq->hold_queue, &list);
1805 spin_unlock(&pq->hold_queue.lock);
1806
1807 while (!skb_queue_empty(&list)) {
1808 skb = __skb_dequeue(&list);
1809
1810 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1811 dst_hold(skb_dst(skb)->path);
1812 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1813 &fl, skb->sk, 0);
1814 if (IS_ERR(dst)) {
a0073fe1
SK		 1815 kfree_skb(skb);
1816 continue;
1817 }
1818
1819 nf_reset(skb);
1820 skb_dst_drop(skb);
1821 skb_dst_set(skb, dst);
1822
a0073fe1 1823 err = dst_output(skb);
a0073fe1
SK		 1824 }
1825
e7d8f6cb
SK		 1826out:
1827 xfrm_pol_put(pol);
a0073fe1
SK		 1828 return;
1829
1830purge_queue:
1831 pq->timeout = 0;
1832 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1833 xfrm_pol_put(pol);
a0073fe1
SK		 1834}
1835
1836static int xdst_queue_output(struct sk_buff *skb)
1837{
1838 unsigned long sched_next;
1839 struct dst_entry *dst = skb_dst(skb);
1840 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1841 struct xfrm_policy *pol = xdst->pols[0];
1842 struct xfrm_policy_queue *pq = &pol->polq;
a0073fe1
SK		 1843
1844 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1845 kfree_skb(skb);
1846 return -EAGAIN;
1847 }
1848
1849 skb_dst_force(skb);
a0073fe1
SK		 1850
1851 spin_lock_bh(&pq->hold_queue.lock);
1852
1853 if (!pq->timeout)
1854 pq->timeout = XFRM_QUEUE_TMO_MIN;
1855
1856 sched_next = jiffies + pq->timeout;
1857
1858 if (del_timer(&pq->hold_timer)) {
1859 if (time_before(pq->hold_timer.expires, sched_next))
1860 sched_next = pq->hold_timer.expires;
e7d8f6cb 1861 xfrm_pol_put(pol);
a0073fe1
SK		 1862 }
1863
1864 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1865 if (!mod_timer(&pq->hold_timer, sched_next))
1866 xfrm_pol_hold(pol);
a0073fe1
SK		 1867
1868 spin_unlock_bh(&pq->hold_queue.lock);
1869
1870 return 0;
1871}
1872
1873static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
1874 struct dst_entry *dst,
1875 const struct flowi *fl,
1876 int num_xfrms,
1877 u16 family)
1878{
1879 int err;
1880 struct net_device *dev;
1881 struct dst_entry *dst1;
1882 struct xfrm_dst *xdst;
1883
1884 xdst = xfrm_alloc_dst(net, family);
1885 if (IS_ERR(xdst))
1886 return xdst;
1887
1888 if (net->xfrm.sysctl_larval_drop || num_xfrms <= 0 ||
1889 (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP))
1890 return xdst;
1891
1892 dst1 = &xdst->u.dst;
1893 dst_hold(dst);
1894 xdst->route = dst;
1895
1896 dst_copy_metrics(dst1, dst);
1897
1898 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
1899 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
1900 dst1->lastuse = jiffies;
1901
1902 dst1->input = dst_discard;
1903 dst1->output = xdst_queue_output;
1904
1905 dst_hold(dst);
1906 dst1->child = dst;
1907 dst1->path = dst;
1908
1909 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
1910
1911 err = -ENODEV;
1912 dev = dst->dev;
1913 if (!dev)
1914 goto free_dst;
1915
1916 err = xfrm_fill_dst(xdst, dev, fl);
1917 if (err)
1918 goto free_dst;
1919
1920out:
1921 return xdst;
1922
1923free_dst:
1924 dst_release(dst1);
1925 xdst = ERR_PTR(err);
1926 goto out;
1927}
1928
80c802f3 1929static struct flow_cache_object *
dee9f4bc 1930xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 1931 struct flow_cache_object *oldflo, void *ctx)
1932{
1933 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1934 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1935 struct xfrm_dst *xdst, *new_xdst;
1936 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1937
1938 /* Check if the policies from old bundle are usable */
1939 xdst = NULL;
1940 if (oldflo) {
1941 xdst = container_of(oldflo, struct xfrm_dst, flo);
1942 num_pols = xdst->num_pols;
1943 num_xfrms = xdst->num_xfrms;
1944 pol_dead = 0;
1945 for (i = 0; i < num_pols; i++) {
1946 pols[i] = xdst->pols[i];
1947 pol_dead |= pols[i]->walk.dead;
1948 }
1949 if (pol_dead) {
1950 dst_free(&xdst->u.dst);
1951 xdst = NULL;
1952 num_pols = 0;
1953 num_xfrms = 0;
1954 oldflo = NULL;
1955 }
1956 }
1957
1958 /* Resolve policies to use if we couldn't get them from
1959 * previous cache entry */
1960 if (xdst == NULL) {
1961 num_pols = 1;
b5fb82c4
BZ		 1962 pols[0] = __xfrm_policy_lookup(net, fl, family,
1963 flow_to_policy_dir(dir));
80c802f3
TT		 1964 err = xfrm_expand_policies(fl, family, pols,
1965 &num_pols, &num_xfrms);
1966 if (err < 0)
1967 goto inc_error;
1968 if (num_pols == 0)
1969 return NULL;
1970 if (num_xfrms <= 0)
1971 goto make_dummy_bundle;
1972 }
1973
1974 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1975 if (IS_ERR(new_xdst)) {
1976 err = PTR_ERR(new_xdst);
1977 if (err != -EAGAIN)
1978 goto error;
1979 if (oldflo == NULL)
1980 goto make_dummy_bundle;
1981 dst_hold(&xdst->u.dst);
1982 return oldflo;
d809ec89
TT		 1983 } else if (new_xdst == NULL) {
1984 num_xfrms = 0;
1985 if (oldflo == NULL)
1986 goto make_dummy_bundle;
1987 xdst->num_xfrms = 0;
1988 dst_hold(&xdst->u.dst);
1989 return oldflo;
80c802f3
TT		 1990 }
1991
1992 /* Kill the previous bundle */
1993 if (xdst) {
1994 /* The policies were stolen for newly generated bundle */
1995 xdst->num_pols = 0;
1996 dst_free(&xdst->u.dst);
1997 }
1998
1999 /* Flow cache does not have reference, it dst_free()'s,
2000 * but we do need to return one reference for original caller */
2001 dst_hold(&new_xdst->u.dst);
2002 return &new_xdst->flo;
2003
2004make_dummy_bundle:
2005 /* We found policies, but there's no bundles to instantiate:
2006 * either because the policy blocks, has no transformations or
2007 * we could not build template (no xfrm_states).*/
a0073fe1 2008 xdst = xfrm_create_dummy_bundle(net, dst_orig, fl, num_xfrms, family);
80c802f3
TT		 2009 if (IS_ERR(xdst)) {
2010 xfrm_pols_put(pols, num_pols);
2011 return ERR_CAST(xdst);
2012 }
2013 xdst->num_pols = num_pols;
2014 xdst->num_xfrms = num_xfrms;
2015 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
2016
2017 dst_hold(&xdst->u.dst);
2018 return &xdst->flo;
2019
2020inc_error:
2021 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2022error:
2023 if (xdst != NULL)
2024 dst_free(&xdst->u.dst);
2025 else
2026 xfrm_pols_put(pols, num_pols);
2027 return ERR_PTR(err);
2028}
1da177e4 2029
2774c131
DM		 2030static struct dst_entry *make_blackhole(struct net *net, u16 family,
2031 struct dst_entry *dst_orig)
2032{
2033 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2034 struct dst_entry *ret;
2035
2036 if (!afinfo) {
2037 dst_release(dst_orig);
433a1954 2038 return ERR_PTR(-EINVAL);
2774c131
DM		 2039 } else {
2040 ret = afinfo->blackhole_route(net, dst_orig);
2041 }
2042 xfrm_policy_put_afinfo(afinfo);
2043
2044 return ret;
2045}
2046
1da177e4
LT		 2047/* Main function: finds/creates a bundle for given flow.
2048 *
2049 * At the moment we eat a raw IP route. Mostly to speed up lookups
2050 * on interfaces with disabled IPsec.
2051 */
452edd59
DM		 2052struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2053 const struct flowi *fl,
2054 struct sock *sk, int flags)
1da177e4 2055{
4e81bb83 2056 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2057 struct flow_cache_object *flo;
2058 struct xfrm_dst *xdst;
452edd59 2059 struct dst_entry *dst, *route;
80c802f3 2060 u16 family = dst_orig->ops->family;
df71837d 2061 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2062 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2063
1da177e4 2064restart:
80c802f3
TT		 2065 dst = NULL;
2066 xdst = NULL;
2067 route = NULL;
4e81bb83 2068
f7944fb1 2069 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2070 num_pols = 1;
2071 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2072 err = xfrm_expand_policies(fl, family, pols,
2073 &num_pols, &num_xfrms);
2074 if (err < 0)
75b8c133 2075 goto dropdst;
80c802f3
TT		 2076
2077 if (num_pols) {
2078 if (num_xfrms <= 0) {
2079 drop_pols = num_pols;
2080 goto no_transform;
2081 }
2082
2083 xdst = xfrm_resolve_and_create_bundle(
2084 pols, num_pols, fl,
2085 family, dst_orig);
2086 if (IS_ERR(xdst)) {
2087 xfrm_pols_put(pols, num_pols);
2088 err = PTR_ERR(xdst);
2089 goto dropdst;
d809ec89
TT		 2090 } else if (xdst == NULL) {
2091 num_xfrms = 0;
2092 drop_pols = num_pols;
2093 goto no_transform;
80c802f3
TT		 2094 }
2095
fbd50608
SK		 2096 dst_hold(&xdst->u.dst);
2097
80c802f3
TT		 2098 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2099 xdst->u.dst.next = xfrm_policy_sk_bundles;
2100 xfrm_policy_sk_bundles = &xdst->u.dst;
2101 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2102
2103 route = xdst->route;
0aa64774 2104 }
3bccfbc7 2105 }
1da177e4 2106
80c802f3 2107 if (xdst == NULL) {
1da177e4 2108 /* To accelerate a bit... */
2518c7c2 2109 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2110 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2111 goto nopol;
1da177e4 2112
80c802f3
TT		 2113 flo = flow_cache_lookup(net, fl, family, dir,
2114 xfrm_bundle_lookup, dst_orig);
2115 if (flo == NULL)
2116 goto nopol;
fe1a5f03 2117 if (IS_ERR(flo)) {
80c802f3 2118 err = PTR_ERR(flo);
75b8c133 2119 goto dropdst;
d66e37a9 2120 }
80c802f3
TT		 2121 xdst = container_of(flo, struct xfrm_dst, flo);
2122
2123 num_pols = xdst->num_pols;
2124 num_xfrms = xdst->num_xfrms;
2125 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
2126 route = xdst->route;
2127 }
2128
2129 dst = &xdst->u.dst;
2130 if (route == NULL && num_xfrms > 0) {
2131 /* The only case when xfrm_bundle_lookup() returns a
2132 * bundle with null route, is when the template could
2133 * not be resolved. It means policies are there, but
2134 * bundle could not be created, since we don't yet
2135 * have the xfrm_state's. We need to wait for KM to
2136 * negotiate new SA's or bail out with error.*/
2137 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2138 dst_release(dst);
a1aa3483 2139 xfrm_pols_put(pols, drop_pols);
80c802f3 2140 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2774c131 2141
452edd59 2142 return make_blackhole(net, family, dst_orig);
80c802f3 2143 }
1d28f42c 2144 if (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP) {
80c802f3
TT		 2145 DECLARE_WAITQUEUE(wait, current);
2146
2147 add_wait_queue(&net->xfrm.km_waitq, &wait);
2148 set_current_state(TASK_INTERRUPTIBLE);
2149 schedule();
2150 set_current_state(TASK_RUNNING);
2151 remove_wait_queue(&net->xfrm.km_waitq, &wait);
2152
2153 if (!signal_pending(current)) {
2154 dst_release(dst);
2155 goto restart;
2156 }
2157
2158 err = -ERESTART;
2159 } else
2160 err = -EAGAIN;
2161
2162 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2163 goto error;
1da177e4
LT		 2164 }
2165
80c802f3
TT		 2166no_transform:
2167 if (num_pols == 0)
8b7817f3 2168 goto nopol;
1da177e4 2169
80c802f3
TT		 2170 if ((flags & XFRM_LOOKUP_ICMP) &&
2171 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2172 err = -ENOENT;
8b7817f3 2173 goto error;
80c802f3 2174 }
8b7817f3 2175
80c802f3
TT		 2176 for (i = 0; i < num_pols; i++)
2177 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2178
80c802f3 2179 if (num_xfrms < 0) {
1da177e4 2180 /* Prohibit the flow */
59c9940e 2181 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2182 err = -EPERM;
2183 goto error;
80c802f3
TT		 2184 } else if (num_xfrms > 0) {
2185 /* Flow transformed */
80c802f3
TT		 2186 dst_release(dst_orig);
2187 } else {
2188 /* Flow passes untransformed */
2189 dst_release(dst);
452edd59 2190 dst = dst_orig;
1da177e4 2191 }
80c802f3
TT		 2192ok:
2193 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2194 if (dst && dst->xfrm &&
2195 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2196 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2197 return dst;
1da177e4 2198
80c802f3 2199nopol:
452edd59
DM		 2200 if (!(flags & XFRM_LOOKUP_ICMP)) {
2201 dst = dst_orig;
80c802f3 2202 goto ok;
452edd59 2203 }
80c802f3 2204 err = -ENOENT;
1da177e4 2205error:
80c802f3 2206 dst_release(dst);
75b8c133
HX		 2207dropdst:
2208 dst_release(dst_orig);
80c802f3 2209 xfrm_pols_put(pols, drop_pols);
452edd59 2210 return ERR_PTR(err);
1da177e4
LT		 2211}
2212EXPORT_SYMBOL(xfrm_lookup);
2213
df0ba92a 2214static inline int
8f029de2 2215xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2216{
2217 struct xfrm_state *x;
df0ba92a
MN		 2218
2219 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2220 return 0;
2221 x = skb->sp->xvec[idx];
2222 if (!x->type->reject)
2223 return 0;
1ecafede 2224 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2225}
2226
1da177e4
LT		 2227/* When skb is transformed back to its "native" form, we have to
2228 * check policy restrictions. At the moment we make this in maximally
2229 * stupid way. Shame on me. :-) Of course, connected sockets must
2230 * have policy cached at them.
2231 */
2232
2233static inline int
7db454b9 2234xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2235 unsigned short family)
2236{
2237 if (xfrm_state_kern(x))
928ba416 2238 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2239 return x->id.proto == tmpl->id.proto &&
2240 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2241 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2242 x->props.mode == tmpl->mode &&
c5d18e98 2243 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2244 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2245 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2246 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2247}
2248
df0ba92a
MN		 2249/*
2250 * 0 or more than 0 is returned when validation is succeeded (either bypass
2251 * because of optional transport mode, or next index of the mathced secpath
2252 * state with the template.
2253 * -1 is returned when no matching template is found.
2254 * Otherwise "-2 - errored_index" is returned.
2255 */
1da177e4 2256static inline int
22cccb7e 2257xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2258 unsigned short family)
2259{
2260 int idx = start;
2261
2262 if (tmpl->optional) {
7e49e6de 2263 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2264 return start;
2265 } else
2266 start = -1;
2267 for (; idx < sp->len; idx++) {
dbe5b4aa 2268 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2269 return ++idx;
df0ba92a
MN		 2270 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2271 if (start == -1)
2272 start = -2-idx;
1da177e4 2273 break;
df0ba92a 2274 }
1da177e4
LT		 2275 }
2276 return start;
2277}
2278
d5422efe
HX		 2279int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2280 unsigned int family, int reverse)
1da177e4
LT		 2281{
2282 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2283 int err;
1da177e4
LT		 2284
2285 if (unlikely(afinfo == NULL))
2286 return -EAFNOSUPPORT;
2287
d5422efe 2288 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2289 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2290 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2291 return err;
1da177e4 2292}
d5422efe 2293EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2294
9a7386ec 2295static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2296{
2297 for (; k < sp->len; k++) {
df0ba92a 2298 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2299 *idxp = k;
1da177e4 2300 return 1;
df0ba92a 2301 }
1da177e4
LT		 2302 }
2303
2304 return 0;
2305}
2306
a716c119 2307int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2308 unsigned short family)
2309{
f6e1e25d 2310 struct net *net = dev_net(skb->dev);
1da177e4 2311 struct xfrm_policy *pol;
4e81bb83
MN		 2312 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2313 int npols = 0;
2314 int xfrm_nr;
2315 int pi;
d5422efe 2316 int reverse;
1da177e4 2317 struct flowi fl;
d5422efe 2318 u8 fl_dir;
df0ba92a 2319 int xerr_idx = -1;
1da177e4 2320
d5422efe
HX		 2321 reverse = dir & ~XFRM_POLICY_MASK;
2322 dir &= XFRM_POLICY_MASK;
2323 fl_dir = policy_to_flow_dir(dir);
2324
0aa64774 2325 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2326 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2327 return 0;
0aa64774
MN		 2328 }
2329
eb9c7ebe 2330 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2331
2332 /* First, check used SA against their selectors. */
2333 if (skb->sp) {
2334 int i;
2335
2336 for (i=skb->sp->len-1; i>=0; i--) {
dbe5b4aa 2337 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2338 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2339 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2340 return 0;
0aa64774 2341 }
1da177e4
LT		 2342 }
2343 }
2344
2345 pol = NULL;
3bccfbc7 2346 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2347 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2348 if (IS_ERR(pol)) {
59c9940e 2349 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2350 return 0;
0aa64774 2351 }
3bccfbc7 2352 }
1da177e4 2353
fe1a5f03
TT		 2354 if (!pol) {
2355 struct flow_cache_object *flo;
2356
2357 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2358 xfrm_policy_lookup, NULL);
2359 if (IS_ERR_OR_NULL(flo))
2360 pol = ERR_CAST(flo);
2361 else
2362 pol = container_of(flo, struct xfrm_policy, flo);
2363 }
1da177e4 2364
0aa64774 2365 if (IS_ERR(pol)) {
59c9940e 2366 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2367 return 0;
0aa64774 2368 }
134b0fc5 2369
df0ba92a 2370 if (!pol) {
d1d9facf 2371 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2372 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2373 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2374 return 0;
2375 }
2376 return 1;
2377 }
1da177e4 2378
9d729f72 2379 pol->curlft.use_time = get_seconds();
1da177e4 2380
4e81bb83
MN		 2381 pols[0] = pol;
2382 npols ++;
2383#ifdef CONFIG_XFRM_SUB_POLICY
2384 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2385 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2386 &fl, family,
2387 XFRM_POLICY_IN);
2388 if (pols[1]) {
0aa64774 2389 if (IS_ERR(pols[1])) {
59c9940e 2390 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2391 return 0;
0aa64774 2392 }
9d729f72 2393 pols[1]->curlft.use_time = get_seconds();
4e81bb83
MN		 2394 npols ++;
2395 }
2396 }
2397#endif
2398
1da177e4
LT		 2399 if (pol->action == XFRM_POLICY_ALLOW) {
2400 struct sec_path *sp;
2401 static struct sec_path dummy;
4e81bb83 2402 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2403 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2404 struct xfrm_tmpl **tpp = tp;
2405 int ti = 0;
1da177e4
LT		 2406 int i, k;
2407
2408 if ((sp = skb->sp) == NULL)
2409 sp = &dummy;
2410
4e81bb83
MN		 2411 for (pi = 0; pi < npols; pi++) {
2412 if (pols[pi] != pol &&
0aa64774 2413 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2414 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2415 goto reject;
0aa64774
MN		 2416 }
2417 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2418 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2419 goto reject_error;
0aa64774 2420 }
4e81bb83
MN		 2421 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2422 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2423 }
2424 xfrm_nr = ti;
41a49cc3
MN		 2425 if (npols > 1) {
2426 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2427 tpp = stp;
2428 }
4e81bb83 2429
1da177e4
LT		 2430 /* For each tunnel xfrm, find the first matching tmpl.
2431 * For each tmpl before that, find corresponding xfrm.
2432 * Order is _important_. Later we will implement
2433 * some barriers, but at the moment barriers
2434 * are implied between each two transformations.
2435 */
4e81bb83
MN		 2436 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2437 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2438 if (k < 0) {
d1d9facf
JM		 2439 if (k < -1)
2440 /* "-2 - errored_index" returned */
2441 xerr_idx = -(2+k);
59c9940e 2442 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2443 goto reject;
df0ba92a 2444 }
1da177e4
LT		 2445 }
2446
0aa64774 2447 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2448 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2449 goto reject;
0aa64774 2450 }
1da177e4 2451
4e81bb83 2452 xfrm_pols_put(pols, npols);
1da177e4
LT		 2453 return 1;
2454 }
59c9940e 2455 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2456
2457reject:
df0ba92a 2458 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2459reject_error:
2460 xfrm_pols_put(pols, npols);
1da177e4
LT		 2461 return 0;
2462}
2463EXPORT_SYMBOL(__xfrm_policy_check);
2464
2465int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2466{
99a66657 2467 struct net *net = dev_net(skb->dev);
1da177e4 2468 struct flowi fl;
adf30907 2469 struct dst_entry *dst;
73137147 2470 int res = 1;
1da177e4 2471
0aa64774 2472 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2473 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2474 return 0;
0aa64774 2475 }
1da177e4 2476
fafeeb6c 2477 skb_dst_force(skb);
adf30907 2478
452edd59
DM		 2479 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2480 if (IS_ERR(dst)) {
73137147 2481 res = 0;
452edd59
DM		 2482 dst = NULL;
2483 }
adf30907
ED		 2484 skb_dst_set(skb, dst);
2485 return res;
1da177e4
LT		 2486}
2487EXPORT_SYMBOL(__xfrm_route_forward);
2488
d49c73c7
DM		 2489/* Optimize later using cookies and generation ids. */
2490
1da177e4
LT		 2491static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2492{
d49c73c7 2493 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2494 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2495 * get validated by dst_ops->check on every use. We do this
2496 * because when a normal route referenced by an XFRM dst is
2497 * obsoleted we do not go looking around for all parent
2498 * referencing XFRM dsts so that we can invalidate them. It
2499 * is just too much work. Instead we make the checks here on
2500 * every use. For example:
d49c73c7
DM		 2501 *
2502 * XFRM dst A --> IPv4 dst X
2503 *
2504 * X is the "xdst->route" of A (X is also the "dst->path" of A
2505 * in this example). If X is marked obsolete, "A" will not
2506 * notice. That's what we are validating here via the
2507 * stale_bundle() check.
2508 *
2509 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2510 * dst which causes it's ->obsolete field to be set to
2511 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2512 * this, we want to force a new route lookup.
399c180a 2513 */
d49c73c7
DM		 2514 if (dst->obsolete < 0 && !stale_bundle(dst))
2515 return dst;
2516
1da177e4
LT		 2517 return NULL;
2518}
2519
2520static int stale_bundle(struct dst_entry *dst)
2521{
12fdb4d3 2522 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2523}
2524
aabc9761 2525void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2526{
1da177e4 2527 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2528 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2529 dev_hold(dst->dev);
1da177e4
LT		 2530 dev_put(dev);
2531 }
2532}
aabc9761 2533EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2534
2535static void xfrm_link_failure(struct sk_buff *skb)
2536{
2537 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2538}
2539
2540static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2541{
2542 if (dst) {
2543 if (dst->obsolete) {
2544 dst_release(dst);
2545 dst = NULL;
2546 }
2547 }
2548 return dst;
2549}
2550
80c802f3 2551static void __xfrm_garbage_collect(struct net *net)
1da177e4 2552{
80c802f3 2553 struct dst_entry *head, *next;
1da177e4 2554
80c802f3
TT		 2555 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2556 head = xfrm_policy_sk_bundles;
2557 xfrm_policy_sk_bundles = NULL;
2558 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2518c7c2 2559
80c802f3
TT		 2560 while (head) {
2561 next = head->next;
2562 dst_free(head);
2563 head = next;
1da177e4
LT		 2564 }
2565}
2566
e4c17216 2567void xfrm_garbage_collect(struct net *net)
c0ed1c14
SK		 2568{
2569 flow_cache_flush();
2570 __xfrm_garbage_collect(net);
2571}
e4c17216 2572EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2573
2574static void xfrm_garbage_collect_deferred(struct net *net)
2575{
2576 flow_cache_flush_deferred();
2577 __xfrm_garbage_collect(net);
2578}
2579
25ee3286 2580static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2581{
2582 do {
2583 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2584 u32 pmtu, route_mtu_cached;
2585
2586 pmtu = dst_mtu(dst->child);
2587 xdst->child_mtu_cached = pmtu;
2588
2589 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2590
2591 route_mtu_cached = dst_mtu(xdst->route);
2592 xdst->route_mtu_cached = route_mtu_cached;
2593
2594 if (pmtu > route_mtu_cached)
2595 pmtu = route_mtu_cached;
2596
defb3519 2597 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2598 } while ((dst = dst->next));
2599}
2600
1da177e4
LT		 2601/* Check that the bundle accepts the flow and its components are
2602 * still valid.
2603 */
2604
12fdb4d3 2605static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2606{
2607 struct dst_entry *dst = &first->u.dst;
2608 struct xfrm_dst *last;
2609 u32 mtu;
2610
92d63dec 2611 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2612 (dst->dev && !netif_running(dst->dev)))
2613 return 0;
2614
a0073fe1
SK		 2615 if (dst->flags & DST_XFRM_QUEUE)
2616 return 1;
2617
1da177e4
LT		 2618 last = NULL;
2619
2620 do {
2621 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2622
1da177e4
LT		 2623 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2624 return 0;
80c802f3
TT		 2625 if (xdst->xfrm_genid != dst->xfrm->genid)
2626 return 0;
b1312c89
TT		 2627 if (xdst->num_pols > 0 &&
2628 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2629 return 0;
e53820de 2630
1da177e4
LT		 2631 mtu = dst_mtu(dst->child);
2632 if (xdst->child_mtu_cached != mtu) {
2633 last = xdst;
2634 xdst->child_mtu_cached = mtu;
2635 }
2636
92d63dec 2637 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2638 return 0;
2639 mtu = dst_mtu(xdst->route);
2640 if (xdst->route_mtu_cached != mtu) {
2641 last = xdst;
2642 xdst->route_mtu_cached = mtu;
2643 }
2644
2645 dst = dst->child;
2646 } while (dst->xfrm);
2647
2648 if (likely(!last))
2649 return 1;
2650
2651 mtu = last->child_mtu_cached;
2652 for (;;) {
2653 dst = &last->u.dst;
2654
2655 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2656 if (mtu > last->route_mtu_cached)
2657 mtu = last->route_mtu_cached;
defb3519 2658 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2659
2660 if (last == first)
2661 break;
2662
bd0bf076 2663 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2664 last->child_mtu_cached = mtu;
2665 }
2666
2667 return 1;
2668}
2669
0dbaee3b
DM		 2670static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2671{
2672 return dst_metric_advmss(dst->path);
2673}
2674
ebb762f2 2675static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2676{
618f9bc7
SK		 2677 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2678
2679 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2680}
2681
f894cbf8
DM		 2682static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2683 struct sk_buff *skb,
2684 const void *daddr)
d3aaeb38 2685{
f894cbf8 2686 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2687}
2688
1da177e4
LT		 2689int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2690{
d7c7544c 2691 struct net *net;
1da177e4
LT		 2692 int err = 0;
2693 if (unlikely(afinfo == NULL))
2694 return -EINVAL;
2695 if (unlikely(afinfo->family >= NPROTO))
2696 return -EAFNOSUPPORT;
ef8531b6 2697 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2698 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2699 err = -ENOBUFS;
2700 else {
2701 struct dst_ops *dst_ops = afinfo->dst_ops;
2702 if (likely(dst_ops->kmem_cachep == NULL))
2703 dst_ops->kmem_cachep = xfrm_dst_cache;
2704 if (likely(dst_ops->check == NULL))
2705 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2706 if (likely(dst_ops->default_advmss == NULL))
2707 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2708 if (likely(dst_ops->mtu == NULL))
2709 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2710 if (likely(dst_ops->negative_advice == NULL))
2711 dst_ops->negative_advice = xfrm_negative_advice;
2712 if (likely(dst_ops->link_failure == NULL))
2713 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2714 if (likely(dst_ops->neigh_lookup == NULL))
2715 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2716 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2717 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2718 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2719 }
ef8531b6 2720 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2721
2722 rtnl_lock();
2723 for_each_net(net) {
2724 struct dst_ops *xfrm_dst_ops;
2725
2726 switch (afinfo->family) {
2727 case AF_INET:
2728 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2729 break;
dfd56b8b 2730#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2731 case AF_INET6:
2732 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2733 break;
2734#endif
2735 default:
2736 BUG();
2737 }
2738 *xfrm_dst_ops = *afinfo->dst_ops;
2739 }
2740 rtnl_unlock();
2741
1da177e4
LT		 2742 return err;
2743}
2744EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2745
2746int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2747{
2748 int err = 0;
2749 if (unlikely(afinfo == NULL))
2750 return -EINVAL;
2751 if (unlikely(afinfo->family >= NPROTO))
2752 return -EAFNOSUPPORT;
ef8531b6 2753 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2754 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2755 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2756 err = -EINVAL;
ef8531b6
ED		 2757 else
2758 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2759 NULL);
2760 }
2761 spin_unlock(&xfrm_policy_afinfo_lock);
2762 if (!err) {
2763 struct dst_ops *dst_ops = afinfo->dst_ops;
2764
2765 synchronize_rcu();
2766
2767 dst_ops->kmem_cachep = NULL;
2768 dst_ops->check = NULL;
2769 dst_ops->negative_advice = NULL;
2770 dst_ops->link_failure = NULL;
2771 afinfo->garbage_collect = NULL;
1da177e4 2772 }
1da177e4
LT		 2773 return err;
2774}
2775EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2776
d7c7544c
AD		 2777static void __net_init xfrm_dst_ops_init(struct net *net)
2778{
2779 struct xfrm_policy_afinfo *afinfo;
2780
ef8531b6
ED		 2781 rcu_read_lock();
2782 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2783 if (afinfo)
2784 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2785#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2786 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2787 if (afinfo)
2788 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2789#endif
418a99ac 2790 rcu_read_unlock();
546be240
HX		 2791}
2792
1da177e4
LT		 2793static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2794{
351638e7 2795 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2796
1da177e4
LT		 2797 switch (event) {
2798 case NETDEV_DOWN:
c0ed1c14 2799 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2800 }
2801 return NOTIFY_DONE;
2802}
2803
2804static struct notifier_block xfrm_dev_notifier = {
d5917a35 2805 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2806};
2807
558f82ef 2808#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2809static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2810{
c68cd1a0
AD		 2811 int rv;
2812
7d720c3e 2813 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
1823e4c8
ED		 2814 sizeof(struct linux_xfrm_mib),
2815 __alignof__(struct linux_xfrm_mib)) < 0)
558f82ef 2816 return -ENOMEM;
c68cd1a0
AD		 2817 rv = xfrm_proc_init(net);
2818 if (rv < 0)
7d720c3e 2819 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
c68cd1a0 2820 return rv;
558f82ef 2821}
59c9940e
AD		 2822
2823static void xfrm_statistics_fini(struct net *net)
2824{
c68cd1a0 2825 xfrm_proc_fini(net);
7d720c3e 2826 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
59c9940e
AD		 2827}
2828#else
2829static int __net_init xfrm_statistics_init(struct net *net)
2830{
2831 return 0;
2832}
2833
2834static void xfrm_statistics_fini(struct net *net)
2835{
2836}
558f82ef
MN		 2837#endif
2838
d62ddc21 2839static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2840{
2518c7c2
DM		 2841 unsigned int hmask, sz;
2842 int dir;
2843
d62ddc21
AD		 2844 if (net_eq(net, &init_net))
2845 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2846 sizeof(struct xfrm_dst),
e5d679f3 2847 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2848 NULL);
1da177e4 2849
2518c7c2
DM		 2850 hmask = 8 - 1;
2851 sz = (hmask+1) * sizeof(struct hlist_head);
2852
93b851c1
AD		 2853 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2854 if (!net->xfrm.policy_byidx)
2855 goto out_byidx;
8100bea7 2856 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2857
2858 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2859 struct xfrm_policy_hash *htab;
2860
dc2caba7 2861 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2862 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2863
a35f6c5d 2864 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2865 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2866 if (!htab->table)
a35f6c5d
AD		 2867 goto out_bydst;
2868 htab->hmask = hmask;
2518c7c2
DM		 2869 }
2870
adfcf0b2 2871 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2872 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2873 if (net_eq(net, &init_net))
2874 register_netdevice_notifier(&xfrm_dev_notifier);
2875 return 0;
93b851c1 2876
a35f6c5d
AD		 2877out_bydst:
2878 for (dir--; dir >= 0; dir--) {
2879 struct xfrm_policy_hash *htab;
2880
2881 htab = &net->xfrm.policy_bydst[dir];
2882 xfrm_hash_free(htab->table, sz);
2883 }
2884 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2885out_byidx:
2886 return -ENOMEM;
d62ddc21
AD		 2887}
2888
2889static void xfrm_policy_fini(struct net *net)
2890{
7c2776ee 2891 struct xfrm_audit audit_info;
93b851c1 2892 unsigned int sz;
8b18f8ea 2893 int dir;
93b851c1 2894
7c2776ee
AD		 2895 flush_work(&net->xfrm.policy_hash_work);
2896#ifdef CONFIG_XFRM_SUB_POLICY
e1760bd5 2897 audit_info.loginuid = INVALID_UID;
7c2776ee
AD		 2898 audit_info.sessionid = -1;
2899 audit_info.secid = 0;
2900 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2901#endif
e1760bd5 2902 audit_info.loginuid = INVALID_UID;
7c2776ee
AD		 2903 audit_info.sessionid = -1;
2904 audit_info.secid = 0;
2905 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
7c2776ee 2906
adfcf0b2 2907 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2908
8b18f8ea 2909 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2910 struct xfrm_policy_hash *htab;
2911
8b18f8ea 2912 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2913
2914 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2915 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2916 WARN_ON(!hlist_empty(htab->table));
2917 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2918 }
2919
8100bea7 2920 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2921 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2922 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2923}
2924
d62ddc21
AD		 2925static int __net_init xfrm_net_init(struct net *net)
2926{
2927 int rv;
2928
59c9940e
AD		 2929 rv = xfrm_statistics_init(net);
2930 if (rv < 0)
2931 goto out_statistics;
d62ddc21
AD		 2932 rv = xfrm_state_init(net);
2933 if (rv < 0)
2934 goto out_state;
2935 rv = xfrm_policy_init(net);
2936 if (rv < 0)
2937 goto out_policy;
d7c7544c 2938 xfrm_dst_ops_init(net);
b27aeadb
AD		 2939 rv = xfrm_sysctl_init(net);
2940 if (rv < 0)
2941 goto out_sysctl;
d62ddc21
AD		 2942 return 0;
2943
b27aeadb
AD		 2944out_sysctl:
2945 xfrm_policy_fini(net);
d62ddc21
AD		 2946out_policy:
2947 xfrm_state_fini(net);
2948out_state:
59c9940e
AD		 2949 xfrm_statistics_fini(net);
2950out_statistics:
d62ddc21
AD		 2951 return rv;
2952}
2953
2954static void __net_exit xfrm_net_exit(struct net *net)
2955{
b27aeadb 2956 xfrm_sysctl_fini(net);
d62ddc21
AD		 2957 xfrm_policy_fini(net);
2958 xfrm_state_fini(net);
59c9940e 2959 xfrm_statistics_fini(net);
d62ddc21
AD		 2960}
2961
2962static struct pernet_operations __net_initdata xfrm_net_ops = {
2963 .init = xfrm_net_init,
2964 .exit = xfrm_net_exit,
2965};
2966
1da177e4
LT		 2967void __init xfrm_init(void)
2968{
d62ddc21 2969 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 2970 xfrm_input_init();
2971}
2972
ab5f5e8b 2973#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2974static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2975 struct audit_buffer *audit_buf)
ab5f5e8b 2976{
875179fa
PM		 2977 struct xfrm_sec_ctx *ctx = xp->security;
2978 struct xfrm_selector *sel = &xp->selector;
2979
2980 if (ctx)
ab5f5e8b 2981 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2982 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2983
875179fa 2984 switch(sel->family) {
ab5f5e8b 2985 case AF_INET:
21454aaa 2986 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2987 if (sel->prefixlen_s != 32)
2988 audit_log_format(audit_buf, " src_prefixlen=%d",
2989 sel->prefixlen_s);
21454aaa 2990 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2991 if (sel->prefixlen_d != 32)
2992 audit_log_format(audit_buf, " dst_prefixlen=%d",
2993 sel->prefixlen_d);
ab5f5e8b
JL		 2994 break;
2995 case AF_INET6:
5b095d98 2996 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 2997 if (sel->prefixlen_s != 128)
2998 audit_log_format(audit_buf, " src_prefixlen=%d",
2999 sel->prefixlen_s);
5b095d98 3000 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3001 if (sel->prefixlen_d != 128)
3002 audit_log_format(audit_buf, " dst_prefixlen=%d",
3003 sel->prefixlen_d);
ab5f5e8b
JL		 3004 break;
3005 }
3006}
3007
68277acc 3008void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
e1760bd5 3009 kuid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 3010{
3011 struct audit_buffer *audit_buf;
ab5f5e8b 3012
afeb14b4 3013 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3014 if (audit_buf == NULL)
3015 return;
2532386f 3016 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 3017 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3018 xfrm_audit_common_policyinfo(xp, audit_buf);
3019 audit_log_end(audit_buf);
3020}
3021EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3022
68277acc 3023void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
e1760bd5 3024 kuid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 3025{
3026 struct audit_buffer *audit_buf;
ab5f5e8b 3027
afeb14b4 3028 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3029 if (audit_buf == NULL)
3030 return;
2532386f 3031 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 3032 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3033 xfrm_audit_common_policyinfo(xp, audit_buf);
3034 audit_log_end(audit_buf);
3035}
3036EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3037#endif
3038
80c9abaa 3039#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3040static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3041 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3042{
3043 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3044 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3045 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3046 sel_cmp->family) &&
3047 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3048 sel_cmp->family) &&
80c9abaa
SS		 3049 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3050 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3051 return true;
80c9abaa
SS		 3052 }
3053 } else {
3054 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3055 return true;
80c9abaa
SS		 3056 }
3057 }
bc9b35ad 3058 return false;
80c9abaa
SS		 3059}
3060
b4b7c0b3 3061static struct xfrm_policy * xfrm_migrate_policy_find(const struct xfrm_selector *sel,
80c9abaa
SS		 3062 u8 dir, u8 type)
3063{
3064 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3065 struct hlist_head *chain;
3066 u32 priority = ~0U;
3067
3068 read_lock_bh(&xfrm_policy_lock);
1121994c 3069 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3070 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3071 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3072 pol->type == type) {
3073 ret = pol;
3074 priority = ret->priority;
3075 break;
3076 }
3077 }
8b18f8ea 3078 chain = &init_net.xfrm.policy_inexact[dir];
b67bfe0d 3079 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3080 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3081 pol->type == type &&
3082 pol->priority < priority) {
3083 ret = pol;
3084 break;
3085 }
3086 }
3087
3088 if (ret)
3089 xfrm_pol_hold(ret);
3090
3091 read_unlock_bh(&xfrm_policy_lock);
3092
3093 return ret;
3094}
3095
dd701754 3096static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3097{
3098 int match = 0;
3099
3100 if (t->mode == m->mode && t->id.proto == m->proto &&
3101 (m->reqid == 0 || t->reqid == m->reqid)) {
3102 switch (t->mode) {
3103 case XFRM_MODE_TUNNEL:
3104 case XFRM_MODE_BEET:
70e94e66
YH		 3105 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3106 m->old_family) &&
3107 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3108 m->old_family)) {
80c9abaa
SS		 3109 match = 1;
3110 }
3111 break;
3112 case XFRM_MODE_TRANSPORT:
3113 /* in case of transport mode, template does not store
3114 any IP addresses, hence we just compare mode and
3115 protocol */
3116 match = 1;
3117 break;
3118 default:
3119 break;
3120 }
3121 }
3122 return match;
3123}
3124
3125/* update endpoint address(es) of template(s) */
3126static int xfrm_policy_migrate(struct xfrm_policy *pol,
3127 struct xfrm_migrate *m, int num_migrate)
3128{
3129 struct xfrm_migrate *mp;
80c9abaa
SS		 3130 int i, j, n = 0;
3131
3132 write_lock_bh(&pol->lock);
12a169e7 3133 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3134 /* target policy has been deleted */
3135 write_unlock_bh(&pol->lock);
3136 return -ENOENT;
3137 }
3138
3139 for (i = 0; i < pol->xfrm_nr; i++) {
3140 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3141 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3142 continue;
3143 n++;
1bfcb10f
HX		 3144 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3145 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3146 continue;
3147 /* update endpoints */
3148 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3149 sizeof(pol->xfrm_vec[i].id.daddr));
3150 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3151 sizeof(pol->xfrm_vec[i].saddr));
3152 pol->xfrm_vec[i].encap_family = mp->new_family;
3153 /* flush bundles */
80c802f3 3154 atomic_inc(&pol->genid);
80c9abaa
SS		 3155 }
3156 }
3157
3158 write_unlock_bh(&pol->lock);
3159
3160 if (!n)
3161 return -ENODATA;
3162
3163 return 0;
3164}
3165
dd701754 3166static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3167{
3168 int i, j;
3169
3170 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3171 return -EINVAL;
3172
3173 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3174 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3175 m[i].old_family) &&
3176 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3177 m[i].old_family))
80c9abaa
SS		 3178 return -EINVAL;
3179 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3180 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3181 return -EINVAL;
3182
3183 /* check if there is any duplicated entry */
3184 for (j = i + 1; j < num_migrate; j++) {
3185 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3186 sizeof(m[i].old_daddr)) &&
3187 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3188 sizeof(m[i].old_saddr)) &&
3189 m[i].proto == m[j].proto &&
3190 m[i].mode == m[j].mode &&
3191 m[i].reqid == m[j].reqid &&
3192 m[i].old_family == m[j].old_family)
3193 return -EINVAL;
3194 }
3195 }
3196
3197 return 0;
3198}
3199
b4b7c0b3 3200int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189
AE		 3201 struct xfrm_migrate *m, int num_migrate,
3202 struct xfrm_kmaddress *k)
80c9abaa
SS		 3203{
3204 int i, err, nx_cur = 0, nx_new = 0;
3205 struct xfrm_policy *pol = NULL;
3206 struct xfrm_state *x, *xc;
3207 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3208 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3209 struct xfrm_migrate *mp;
3210
3211 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3212 goto out;
3213
3214 /* Stage 1 - find policy */
3215 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
3216 err = -ENOENT;
3217 goto out;
3218 }
3219
3220 /* Stage 2 - find and update state(s) */
3221 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
3222 if ((x = xfrm_migrate_state_find(mp))) {
3223 x_cur[nx_cur] = x;
3224 nx_cur++;
3225 if ((xc = xfrm_state_migrate(x, mp))) {
3226 x_new[nx_new] = xc;
3227 nx_new++;
3228 } else {
3229 err = -ENODATA;
3230 goto restore_state;
3231 }
3232 }
3233 }
3234
3235 /* Stage 3 - update policy */
3236 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3237 goto restore_state;
3238
3239 /* Stage 4 - delete old state(s) */
3240 if (nx_cur) {
3241 xfrm_states_put(x_cur, nx_cur);
3242 xfrm_states_delete(x_cur, nx_cur);
3243 }
3244
3245 /* Stage 5 - announce */
13c1d189 3246 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3247
3248 xfrm_pol_put(pol);
3249
3250 return 0;
3251out:
3252 return err;
3253
3254restore_state:
3255 if (pol)
3256 xfrm_pol_put(pol);
3257 if (nx_cur)
3258 xfrm_states_put(x_cur, nx_cur);
3259 if (nx_new)
3260 xfrm_states_delete(x_new, nx_new);
3261
3262 return err;
3263}
e610e679 3264EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3265#endif
Linux 6.x block layer and io_uring tree(s)