projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branches 'acpi-hotplug', 'acpi-sysfs' and 'acpi-sleep'
[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);
2bb53e25
SK		 1775 if (!skb) {
1776 spin_unlock(&pq->hold_queue.lock);
1777 goto out;
1778 }
a0073fe1
SK		 1779 dst = skb_dst(skb);
1780 sk = skb->sk;
1781 xfrm_decode_session(skb, &fl, dst->ops->family);
1782 spin_unlock(&pq->hold_queue.lock);
1783
1784 dst_hold(dst->path);
1785 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1786 sk, 0);
1787 if (IS_ERR(dst))
1788 goto purge_queue;
1789
1790 if (dst->flags & DST_XFRM_QUEUE) {
1791 dst_release(dst);
1792
1793 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1794 goto purge_queue;
1795
1796 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1797 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1798 xfrm_pol_hold(pol);
1799 goto out;
a0073fe1
SK		 1800 }
1801
1802 dst_release(dst);
1803
1804 __skb_queue_head_init(&list);
1805
1806 spin_lock(&pq->hold_queue.lock);
1807 pq->timeout = 0;
1808 skb_queue_splice_init(&pq->hold_queue, &list);
1809 spin_unlock(&pq->hold_queue.lock);
1810
1811 while (!skb_queue_empty(&list)) {
1812 skb = __skb_dequeue(&list);
1813
1814 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1815 dst_hold(skb_dst(skb)->path);
1816 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1817 &fl, skb->sk, 0);
1818 if (IS_ERR(dst)) {
a0073fe1
SK		 1819 kfree_skb(skb);
1820 continue;
1821 }
1822
1823 nf_reset(skb);
1824 skb_dst_drop(skb);
1825 skb_dst_set(skb, dst);
1826
a0073fe1 1827 err = dst_output(skb);
a0073fe1
SK		 1828 }
1829
e7d8f6cb
SK		 1830out:
1831 xfrm_pol_put(pol);
a0073fe1
SK		 1832 return;
1833
1834purge_queue:
1835 pq->timeout = 0;
1836 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1837 xfrm_pol_put(pol);
a0073fe1
SK		 1838}
1839
1840static int xdst_queue_output(struct sk_buff *skb)
1841{
1842 unsigned long sched_next;
1843 struct dst_entry *dst = skb_dst(skb);
1844 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1845 struct xfrm_policy *pol = xdst->pols[0];
1846 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4
SK		 1847 const struct sk_buff *fclone = skb + 1;
1848
1849 if (unlikely(skb->fclone == SKB_FCLONE_ORIG &&
1850 fclone->fclone == SKB_FCLONE_CLONE)) {
1851 kfree_skb(skb);
1852 return 0;
1853 }
a0073fe1
SK		 1854
1855 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1856 kfree_skb(skb);
1857 return -EAGAIN;
1858 }
1859
1860 skb_dst_force(skb);
a0073fe1
SK		 1861
1862 spin_lock_bh(&pq->hold_queue.lock);
1863
1864 if (!pq->timeout)
1865 pq->timeout = XFRM_QUEUE_TMO_MIN;
1866
1867 sched_next = jiffies + pq->timeout;
1868
1869 if (del_timer(&pq->hold_timer)) {
1870 if (time_before(pq->hold_timer.expires, sched_next))
1871 sched_next = pq->hold_timer.expires;
e7d8f6cb 1872 xfrm_pol_put(pol);
a0073fe1
SK		 1873 }
1874
1875 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1876 if (!mod_timer(&pq->hold_timer, sched_next))
1877 xfrm_pol_hold(pol);
a0073fe1
SK		 1878
1879 spin_unlock_bh(&pq->hold_queue.lock);
1880
1881 return 0;
1882}
1883
1884static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
1885 struct dst_entry *dst,
1886 const struct flowi *fl,
1887 int num_xfrms,
1888 u16 family)
1889{
1890 int err;
1891 struct net_device *dev;
1892 struct dst_entry *dst1;
1893 struct xfrm_dst *xdst;
1894
1895 xdst = xfrm_alloc_dst(net, family);
1896 if (IS_ERR(xdst))
1897 return xdst;
1898
1899 if (net->xfrm.sysctl_larval_drop || num_xfrms <= 0 ||
1900 (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP))
1901 return xdst;
1902
1903 dst1 = &xdst->u.dst;
1904 dst_hold(dst);
1905 xdst->route = dst;
1906
1907 dst_copy_metrics(dst1, dst);
1908
1909 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
1910 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
1911 dst1->lastuse = jiffies;
1912
1913 dst1->input = dst_discard;
1914 dst1->output = xdst_queue_output;
1915
1916 dst_hold(dst);
1917 dst1->child = dst;
1918 dst1->path = dst;
1919
1920 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
1921
1922 err = -ENODEV;
1923 dev = dst->dev;
1924 if (!dev)
1925 goto free_dst;
1926
1927 err = xfrm_fill_dst(xdst, dev, fl);
1928 if (err)
1929 goto free_dst;
1930
1931out:
1932 return xdst;
1933
1934free_dst:
1935 dst_release(dst1);
1936 xdst = ERR_PTR(err);
1937 goto out;
1938}
1939
80c802f3 1940static struct flow_cache_object *
dee9f4bc 1941xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 1942 struct flow_cache_object *oldflo, void *ctx)
1943{
1944 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1945 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1946 struct xfrm_dst *xdst, *new_xdst;
1947 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1948
1949 /* Check if the policies from old bundle are usable */
1950 xdst = NULL;
1951 if (oldflo) {
1952 xdst = container_of(oldflo, struct xfrm_dst, flo);
1953 num_pols = xdst->num_pols;
1954 num_xfrms = xdst->num_xfrms;
1955 pol_dead = 0;
1956 for (i = 0; i < num_pols; i++) {
1957 pols[i] = xdst->pols[i];
1958 pol_dead |= pols[i]->walk.dead;
1959 }
1960 if (pol_dead) {
1961 dst_free(&xdst->u.dst);
1962 xdst = NULL;
1963 num_pols = 0;
1964 num_xfrms = 0;
1965 oldflo = NULL;
1966 }
1967 }
1968
1969 /* Resolve policies to use if we couldn't get them from
1970 * previous cache entry */
1971 if (xdst == NULL) {
1972 num_pols = 1;
b5fb82c4
BZ		 1973 pols[0] = __xfrm_policy_lookup(net, fl, family,
1974 flow_to_policy_dir(dir));
80c802f3
TT		 1975 err = xfrm_expand_policies(fl, family, pols,
1976 &num_pols, &num_xfrms);
1977 if (err < 0)
1978 goto inc_error;
1979 if (num_pols == 0)
1980 return NULL;
1981 if (num_xfrms <= 0)
1982 goto make_dummy_bundle;
1983 }
1984
1985 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1986 if (IS_ERR(new_xdst)) {
1987 err = PTR_ERR(new_xdst);
1988 if (err != -EAGAIN)
1989 goto error;
1990 if (oldflo == NULL)
1991 goto make_dummy_bundle;
1992 dst_hold(&xdst->u.dst);
1993 return oldflo;
d809ec89
TT		 1994 } else if (new_xdst == NULL) {
1995 num_xfrms = 0;
1996 if (oldflo == NULL)
1997 goto make_dummy_bundle;
1998 xdst->num_xfrms = 0;
1999 dst_hold(&xdst->u.dst);
2000 return oldflo;
80c802f3
TT		 2001 }
2002
2003 /* Kill the previous bundle */
2004 if (xdst) {
2005 /* The policies were stolen for newly generated bundle */
2006 xdst->num_pols = 0;
2007 dst_free(&xdst->u.dst);
2008 }
2009
2010 /* Flow cache does not have reference, it dst_free()'s,
2011 * but we do need to return one reference for original caller */
2012 dst_hold(&new_xdst->u.dst);
2013 return &new_xdst->flo;
2014
2015make_dummy_bundle:
2016 /* We found policies, but there's no bundles to instantiate:
2017 * either because the policy blocks, has no transformations or
2018 * we could not build template (no xfrm_states).*/
a0073fe1 2019 xdst = xfrm_create_dummy_bundle(net, dst_orig, fl, num_xfrms, family);
80c802f3
TT		 2020 if (IS_ERR(xdst)) {
2021 xfrm_pols_put(pols, num_pols);
2022 return ERR_CAST(xdst);
2023 }
2024 xdst->num_pols = num_pols;
2025 xdst->num_xfrms = num_xfrms;
2026 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
2027
2028 dst_hold(&xdst->u.dst);
2029 return &xdst->flo;
2030
2031inc_error:
2032 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2033error:
2034 if (xdst != NULL)
2035 dst_free(&xdst->u.dst);
2036 else
2037 xfrm_pols_put(pols, num_pols);
2038 return ERR_PTR(err);
2039}
1da177e4 2040
2774c131
DM		 2041static struct dst_entry *make_blackhole(struct net *net, u16 family,
2042 struct dst_entry *dst_orig)
2043{
2044 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2045 struct dst_entry *ret;
2046
2047 if (!afinfo) {
2048 dst_release(dst_orig);
433a1954 2049 return ERR_PTR(-EINVAL);
2774c131
DM		 2050 } else {
2051 ret = afinfo->blackhole_route(net, dst_orig);
2052 }
2053 xfrm_policy_put_afinfo(afinfo);
2054
2055 return ret;
2056}
2057
1da177e4
LT		 2058/* Main function: finds/creates a bundle for given flow.
2059 *
2060 * At the moment we eat a raw IP route. Mostly to speed up lookups
2061 * on interfaces with disabled IPsec.
2062 */
452edd59
DM		 2063struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2064 const struct flowi *fl,
2065 struct sock *sk, int flags)
1da177e4 2066{
4e81bb83 2067 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2068 struct flow_cache_object *flo;
2069 struct xfrm_dst *xdst;
452edd59 2070 struct dst_entry *dst, *route;
80c802f3 2071 u16 family = dst_orig->ops->family;
df71837d 2072 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2073 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2074
1da177e4 2075restart:
80c802f3
TT		 2076 dst = NULL;
2077 xdst = NULL;
2078 route = NULL;
4e81bb83 2079
f7944fb1 2080 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2081 num_pols = 1;
2082 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2083 err = xfrm_expand_policies(fl, family, pols,
2084 &num_pols, &num_xfrms);
2085 if (err < 0)
75b8c133 2086 goto dropdst;
80c802f3
TT		 2087
2088 if (num_pols) {
2089 if (num_xfrms <= 0) {
2090 drop_pols = num_pols;
2091 goto no_transform;
2092 }
2093
2094 xdst = xfrm_resolve_and_create_bundle(
2095 pols, num_pols, fl,
2096 family, dst_orig);
2097 if (IS_ERR(xdst)) {
2098 xfrm_pols_put(pols, num_pols);
2099 err = PTR_ERR(xdst);
2100 goto dropdst;
d809ec89
TT		 2101 } else if (xdst == NULL) {
2102 num_xfrms = 0;
2103 drop_pols = num_pols;
2104 goto no_transform;
80c802f3
TT		 2105 }
2106
fbd50608
SK		 2107 dst_hold(&xdst->u.dst);
2108
80c802f3
TT		 2109 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2110 xdst->u.dst.next = xfrm_policy_sk_bundles;
2111 xfrm_policy_sk_bundles = &xdst->u.dst;
2112 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2113
2114 route = xdst->route;
0aa64774 2115 }
3bccfbc7 2116 }
1da177e4 2117
80c802f3 2118 if (xdst == NULL) {
1da177e4 2119 /* To accelerate a bit... */
2518c7c2 2120 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2121 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2122 goto nopol;
1da177e4 2123
80c802f3
TT		 2124 flo = flow_cache_lookup(net, fl, family, dir,
2125 xfrm_bundle_lookup, dst_orig);
2126 if (flo == NULL)
2127 goto nopol;
fe1a5f03 2128 if (IS_ERR(flo)) {
80c802f3 2129 err = PTR_ERR(flo);
75b8c133 2130 goto dropdst;
d66e37a9 2131 }
80c802f3
TT		 2132 xdst = container_of(flo, struct xfrm_dst, flo);
2133
2134 num_pols = xdst->num_pols;
2135 num_xfrms = xdst->num_xfrms;
2136 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
2137 route = xdst->route;
2138 }
2139
2140 dst = &xdst->u.dst;
2141 if (route == NULL && num_xfrms > 0) {
2142 /* The only case when xfrm_bundle_lookup() returns a
2143 * bundle with null route, is when the template could
2144 * not be resolved. It means policies are there, but
2145 * bundle could not be created, since we don't yet
2146 * have the xfrm_state's. We need to wait for KM to
2147 * negotiate new SA's or bail out with error.*/
2148 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2149 dst_release(dst);
a1aa3483 2150 xfrm_pols_put(pols, drop_pols);
80c802f3 2151 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2774c131 2152
452edd59 2153 return make_blackhole(net, family, dst_orig);
80c802f3 2154 }
1d28f42c 2155 if (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP) {
80c802f3
TT		 2156 DECLARE_WAITQUEUE(wait, current);
2157
2158 add_wait_queue(&net->xfrm.km_waitq, &wait);
2159 set_current_state(TASK_INTERRUPTIBLE);
2160 schedule();
2161 set_current_state(TASK_RUNNING);
2162 remove_wait_queue(&net->xfrm.km_waitq, &wait);
2163
2164 if (!signal_pending(current)) {
2165 dst_release(dst);
2166 goto restart;
2167 }
2168
2169 err = -ERESTART;
2170 } else
2171 err = -EAGAIN;
2172
2173 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2174 goto error;
1da177e4
LT		 2175 }
2176
80c802f3
TT		 2177no_transform:
2178 if (num_pols == 0)
8b7817f3 2179 goto nopol;
1da177e4 2180
80c802f3
TT		 2181 if ((flags & XFRM_LOOKUP_ICMP) &&
2182 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2183 err = -ENOENT;
8b7817f3 2184 goto error;
80c802f3 2185 }
8b7817f3 2186
80c802f3
TT		 2187 for (i = 0; i < num_pols; i++)
2188 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2189
80c802f3 2190 if (num_xfrms < 0) {
1da177e4 2191 /* Prohibit the flow */
59c9940e 2192 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2193 err = -EPERM;
2194 goto error;
80c802f3
TT		 2195 } else if (num_xfrms > 0) {
2196 /* Flow transformed */
80c802f3
TT		 2197 dst_release(dst_orig);
2198 } else {
2199 /* Flow passes untransformed */
2200 dst_release(dst);
452edd59 2201 dst = dst_orig;
1da177e4 2202 }
80c802f3
TT		 2203ok:
2204 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2205 if (dst && dst->xfrm &&
2206 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2207 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2208 return dst;
1da177e4 2209
80c802f3 2210nopol:
452edd59
DM		 2211 if (!(flags & XFRM_LOOKUP_ICMP)) {
2212 dst = dst_orig;
80c802f3 2213 goto ok;
452edd59 2214 }
80c802f3 2215 err = -ENOENT;
1da177e4 2216error:
80c802f3 2217 dst_release(dst);
75b8c133
HX		 2218dropdst:
2219 dst_release(dst_orig);
80c802f3 2220 xfrm_pols_put(pols, drop_pols);
452edd59 2221 return ERR_PTR(err);
1da177e4
LT		 2222}
2223EXPORT_SYMBOL(xfrm_lookup);
2224
df0ba92a 2225static inline int
8f029de2 2226xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2227{
2228 struct xfrm_state *x;
df0ba92a
MN		 2229
2230 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2231 return 0;
2232 x = skb->sp->xvec[idx];
2233 if (!x->type->reject)
2234 return 0;
1ecafede 2235 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2236}
2237
1da177e4
LT		 2238/* When skb is transformed back to its "native" form, we have to
2239 * check policy restrictions. At the moment we make this in maximally
2240 * stupid way. Shame on me. :-) Of course, connected sockets must
2241 * have policy cached at them.
2242 */
2243
2244static inline int
7db454b9 2245xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2246 unsigned short family)
2247{
2248 if (xfrm_state_kern(x))
928ba416 2249 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2250 return x->id.proto == tmpl->id.proto &&
2251 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2252 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2253 x->props.mode == tmpl->mode &&
c5d18e98 2254 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2255 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2256 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2257 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2258}
2259
df0ba92a
MN		 2260/*
2261 * 0 or more than 0 is returned when validation is succeeded (either bypass
2262 * because of optional transport mode, or next index of the mathced secpath
2263 * state with the template.
2264 * -1 is returned when no matching template is found.
2265 * Otherwise "-2 - errored_index" is returned.
2266 */
1da177e4 2267static inline int
22cccb7e 2268xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2269 unsigned short family)
2270{
2271 int idx = start;
2272
2273 if (tmpl->optional) {
7e49e6de 2274 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2275 return start;
2276 } else
2277 start = -1;
2278 for (; idx < sp->len; idx++) {
dbe5b4aa 2279 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2280 return ++idx;
df0ba92a
MN		 2281 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2282 if (start == -1)
2283 start = -2-idx;
1da177e4 2284 break;
df0ba92a 2285 }
1da177e4
LT		 2286 }
2287 return start;
2288}
2289
d5422efe
HX		 2290int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2291 unsigned int family, int reverse)
1da177e4
LT		 2292{
2293 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2294 int err;
1da177e4
LT		 2295
2296 if (unlikely(afinfo == NULL))
2297 return -EAFNOSUPPORT;
2298
d5422efe 2299 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2300 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2301 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2302 return err;
1da177e4 2303}
d5422efe 2304EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2305
9a7386ec 2306static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2307{
2308 for (; k < sp->len; k++) {
df0ba92a 2309 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2310 *idxp = k;
1da177e4 2311 return 1;
df0ba92a 2312 }
1da177e4
LT		 2313 }
2314
2315 return 0;
2316}
2317
a716c119 2318int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2319 unsigned short family)
2320{
f6e1e25d 2321 struct net *net = dev_net(skb->dev);
1da177e4 2322 struct xfrm_policy *pol;
4e81bb83
MN		 2323 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2324 int npols = 0;
2325 int xfrm_nr;
2326 int pi;
d5422efe 2327 int reverse;
1da177e4 2328 struct flowi fl;
d5422efe 2329 u8 fl_dir;
df0ba92a 2330 int xerr_idx = -1;
1da177e4 2331
d5422efe
HX		 2332 reverse = dir & ~XFRM_POLICY_MASK;
2333 dir &= XFRM_POLICY_MASK;
2334 fl_dir = policy_to_flow_dir(dir);
2335
0aa64774 2336 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2337 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2338 return 0;
0aa64774
MN		 2339 }
2340
eb9c7ebe 2341 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2342
2343 /* First, check used SA against their selectors. */
2344 if (skb->sp) {
2345 int i;
2346
2347 for (i=skb->sp->len-1; i>=0; i--) {
dbe5b4aa 2348 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2349 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2350 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2351 return 0;
0aa64774 2352 }
1da177e4
LT		 2353 }
2354 }
2355
2356 pol = NULL;
3bccfbc7 2357 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2358 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2359 if (IS_ERR(pol)) {
59c9940e 2360 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2361 return 0;
0aa64774 2362 }
3bccfbc7 2363 }
1da177e4 2364
fe1a5f03
TT		 2365 if (!pol) {
2366 struct flow_cache_object *flo;
2367
2368 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2369 xfrm_policy_lookup, NULL);
2370 if (IS_ERR_OR_NULL(flo))
2371 pol = ERR_CAST(flo);
2372 else
2373 pol = container_of(flo, struct xfrm_policy, flo);
2374 }
1da177e4 2375
0aa64774 2376 if (IS_ERR(pol)) {
59c9940e 2377 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2378 return 0;
0aa64774 2379 }
134b0fc5 2380
df0ba92a 2381 if (!pol) {
d1d9facf 2382 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2383 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2384 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2385 return 0;
2386 }
2387 return 1;
2388 }
1da177e4 2389
9d729f72 2390 pol->curlft.use_time = get_seconds();
1da177e4 2391
4e81bb83
MN		 2392 pols[0] = pol;
2393 npols ++;
2394#ifdef CONFIG_XFRM_SUB_POLICY
2395 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2396 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2397 &fl, family,
2398 XFRM_POLICY_IN);
2399 if (pols[1]) {
0aa64774 2400 if (IS_ERR(pols[1])) {
59c9940e 2401 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2402 return 0;
0aa64774 2403 }
9d729f72 2404 pols[1]->curlft.use_time = get_seconds();
4e81bb83
MN		 2405 npols ++;
2406 }
2407 }
2408#endif
2409
1da177e4
LT		 2410 if (pol->action == XFRM_POLICY_ALLOW) {
2411 struct sec_path *sp;
2412 static struct sec_path dummy;
4e81bb83 2413 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2414 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2415 struct xfrm_tmpl **tpp = tp;
2416 int ti = 0;
1da177e4
LT		 2417 int i, k;
2418
2419 if ((sp = skb->sp) == NULL)
2420 sp = &dummy;
2421
4e81bb83
MN		 2422 for (pi = 0; pi < npols; pi++) {
2423 if (pols[pi] != pol &&
0aa64774 2424 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2425 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2426 goto reject;
0aa64774
MN		 2427 }
2428 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2429 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2430 goto reject_error;
0aa64774 2431 }
4e81bb83
MN		 2432 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2433 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2434 }
2435 xfrm_nr = ti;
41a49cc3
MN		 2436 if (npols > 1) {
2437 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2438 tpp = stp;
2439 }
4e81bb83 2440
1da177e4
LT		 2441 /* For each tunnel xfrm, find the first matching tmpl.
2442 * For each tmpl before that, find corresponding xfrm.
2443 * Order is _important_. Later we will implement
2444 * some barriers, but at the moment barriers
2445 * are implied between each two transformations.
2446 */
4e81bb83
MN		 2447 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2448 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2449 if (k < 0) {
d1d9facf
JM		 2450 if (k < -1)
2451 /* "-2 - errored_index" returned */
2452 xerr_idx = -(2+k);
59c9940e 2453 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2454 goto reject;
df0ba92a 2455 }
1da177e4
LT		 2456 }
2457
0aa64774 2458 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2459 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2460 goto reject;
0aa64774 2461 }
1da177e4 2462
4e81bb83 2463 xfrm_pols_put(pols, npols);
1da177e4
LT		 2464 return 1;
2465 }
59c9940e 2466 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2467
2468reject:
df0ba92a 2469 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2470reject_error:
2471 xfrm_pols_put(pols, npols);
1da177e4
LT		 2472 return 0;
2473}
2474EXPORT_SYMBOL(__xfrm_policy_check);
2475
2476int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2477{
99a66657 2478 struct net *net = dev_net(skb->dev);
1da177e4 2479 struct flowi fl;
adf30907 2480 struct dst_entry *dst;
73137147 2481 int res = 1;
1da177e4 2482
0aa64774 2483 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2484 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2485 return 0;
0aa64774 2486 }
1da177e4 2487
fafeeb6c 2488 skb_dst_force(skb);
adf30907 2489
452edd59
DM		 2490 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2491 if (IS_ERR(dst)) {
73137147 2492 res = 0;
452edd59
DM		 2493 dst = NULL;
2494 }
adf30907
ED		 2495 skb_dst_set(skb, dst);
2496 return res;
1da177e4
LT		 2497}
2498EXPORT_SYMBOL(__xfrm_route_forward);
2499
d49c73c7
DM		 2500/* Optimize later using cookies and generation ids. */
2501
1da177e4
LT		 2502static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2503{
d49c73c7 2504 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2505 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2506 * get validated by dst_ops->check on every use. We do this
2507 * because when a normal route referenced by an XFRM dst is
2508 * obsoleted we do not go looking around for all parent
2509 * referencing XFRM dsts so that we can invalidate them. It
2510 * is just too much work. Instead we make the checks here on
2511 * every use. For example:
d49c73c7
DM		 2512 *
2513 * XFRM dst A --> IPv4 dst X
2514 *
2515 * X is the "xdst->route" of A (X is also the "dst->path" of A
2516 * in this example). If X is marked obsolete, "A" will not
2517 * notice. That's what we are validating here via the
2518 * stale_bundle() check.
2519 *
2520 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2521 * dst which causes it's ->obsolete field to be set to
2522 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2523 * this, we want to force a new route lookup.
399c180a 2524 */
d49c73c7
DM		 2525 if (dst->obsolete < 0 && !stale_bundle(dst))
2526 return dst;
2527
1da177e4
LT		 2528 return NULL;
2529}
2530
2531static int stale_bundle(struct dst_entry *dst)
2532{
12fdb4d3 2533 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2534}
2535
aabc9761 2536void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2537{
1da177e4 2538 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2539 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2540 dev_hold(dst->dev);
1da177e4
LT		 2541 dev_put(dev);
2542 }
2543}
aabc9761 2544EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2545
2546static void xfrm_link_failure(struct sk_buff *skb)
2547{
2548 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2549}
2550
2551static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2552{
2553 if (dst) {
2554 if (dst->obsolete) {
2555 dst_release(dst);
2556 dst = NULL;
2557 }
2558 }
2559 return dst;
2560}
2561
80c802f3 2562static void __xfrm_garbage_collect(struct net *net)
1da177e4 2563{
80c802f3 2564 struct dst_entry *head, *next;
1da177e4 2565
80c802f3
TT		 2566 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2567 head = xfrm_policy_sk_bundles;
2568 xfrm_policy_sk_bundles = NULL;
2569 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2518c7c2 2570
80c802f3
TT		 2571 while (head) {
2572 next = head->next;
2573 dst_free(head);
2574 head = next;
1da177e4
LT		 2575 }
2576}
2577
e4c17216 2578void xfrm_garbage_collect(struct net *net)
c0ed1c14
SK		 2579{
2580 flow_cache_flush();
2581 __xfrm_garbage_collect(net);
2582}
e4c17216 2583EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2584
2585static void xfrm_garbage_collect_deferred(struct net *net)
2586{
2587 flow_cache_flush_deferred();
2588 __xfrm_garbage_collect(net);
2589}
2590
25ee3286 2591static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2592{
2593 do {
2594 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2595 u32 pmtu, route_mtu_cached;
2596
2597 pmtu = dst_mtu(dst->child);
2598 xdst->child_mtu_cached = pmtu;
2599
2600 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2601
2602 route_mtu_cached = dst_mtu(xdst->route);
2603 xdst->route_mtu_cached = route_mtu_cached;
2604
2605 if (pmtu > route_mtu_cached)
2606 pmtu = route_mtu_cached;
2607
defb3519 2608 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2609 } while ((dst = dst->next));
2610}
2611
1da177e4
LT		 2612/* Check that the bundle accepts the flow and its components are
2613 * still valid.
2614 */
2615
12fdb4d3 2616static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2617{
2618 struct dst_entry *dst = &first->u.dst;
2619 struct xfrm_dst *last;
2620 u32 mtu;
2621
92d63dec 2622 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2623 (dst->dev && !netif_running(dst->dev)))
2624 return 0;
2625
a0073fe1
SK		 2626 if (dst->flags & DST_XFRM_QUEUE)
2627 return 1;
2628
1da177e4
LT		 2629 last = NULL;
2630
2631 do {
2632 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2633
1da177e4
LT		 2634 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2635 return 0;
80c802f3
TT		 2636 if (xdst->xfrm_genid != dst->xfrm->genid)
2637 return 0;
b1312c89
TT		 2638 if (xdst->num_pols > 0 &&
2639 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2640 return 0;
e53820de 2641
1da177e4
LT		 2642 mtu = dst_mtu(dst->child);
2643 if (xdst->child_mtu_cached != mtu) {
2644 last = xdst;
2645 xdst->child_mtu_cached = mtu;
2646 }
2647
92d63dec 2648 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2649 return 0;
2650 mtu = dst_mtu(xdst->route);
2651 if (xdst->route_mtu_cached != mtu) {
2652 last = xdst;
2653 xdst->route_mtu_cached = mtu;
2654 }
2655
2656 dst = dst->child;
2657 } while (dst->xfrm);
2658
2659 if (likely(!last))
2660 return 1;
2661
2662 mtu = last->child_mtu_cached;
2663 for (;;) {
2664 dst = &last->u.dst;
2665
2666 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2667 if (mtu > last->route_mtu_cached)
2668 mtu = last->route_mtu_cached;
defb3519 2669 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2670
2671 if (last == first)
2672 break;
2673
bd0bf076 2674 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2675 last->child_mtu_cached = mtu;
2676 }
2677
2678 return 1;
2679}
2680
0dbaee3b
DM		 2681static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2682{
2683 return dst_metric_advmss(dst->path);
2684}
2685
ebb762f2 2686static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2687{
618f9bc7
SK		 2688 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2689
2690 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2691}
2692
f894cbf8
DM		 2693static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2694 struct sk_buff *skb,
2695 const void *daddr)
d3aaeb38 2696{
f894cbf8 2697 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2698}
2699
1da177e4
LT		 2700int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2701{
d7c7544c 2702 struct net *net;
1da177e4
LT		 2703 int err = 0;
2704 if (unlikely(afinfo == NULL))
2705 return -EINVAL;
2706 if (unlikely(afinfo->family >= NPROTO))
2707 return -EAFNOSUPPORT;
ef8531b6 2708 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2709 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2710 err = -ENOBUFS;
2711 else {
2712 struct dst_ops *dst_ops = afinfo->dst_ops;
2713 if (likely(dst_ops->kmem_cachep == NULL))
2714 dst_ops->kmem_cachep = xfrm_dst_cache;
2715 if (likely(dst_ops->check == NULL))
2716 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2717 if (likely(dst_ops->default_advmss == NULL))
2718 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2719 if (likely(dst_ops->mtu == NULL))
2720 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2721 if (likely(dst_ops->negative_advice == NULL))
2722 dst_ops->negative_advice = xfrm_negative_advice;
2723 if (likely(dst_ops->link_failure == NULL))
2724 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2725 if (likely(dst_ops->neigh_lookup == NULL))
2726 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2727 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2728 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2729 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2730 }
ef8531b6 2731 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2732
2733 rtnl_lock();
2734 for_each_net(net) {
2735 struct dst_ops *xfrm_dst_ops;
2736
2737 switch (afinfo->family) {
2738 case AF_INET:
2739 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2740 break;
dfd56b8b 2741#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2742 case AF_INET6:
2743 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2744 break;
2745#endif
2746 default:
2747 BUG();
2748 }
2749 *xfrm_dst_ops = *afinfo->dst_ops;
2750 }
2751 rtnl_unlock();
2752
1da177e4
LT		 2753 return err;
2754}
2755EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2756
2757int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2758{
2759 int err = 0;
2760 if (unlikely(afinfo == NULL))
2761 return -EINVAL;
2762 if (unlikely(afinfo->family >= NPROTO))
2763 return -EAFNOSUPPORT;
ef8531b6 2764 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2765 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2766 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2767 err = -EINVAL;
ef8531b6
ED		 2768 else
2769 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2770 NULL);
2771 }
2772 spin_unlock(&xfrm_policy_afinfo_lock);
2773 if (!err) {
2774 struct dst_ops *dst_ops = afinfo->dst_ops;
2775
2776 synchronize_rcu();
2777
2778 dst_ops->kmem_cachep = NULL;
2779 dst_ops->check = NULL;
2780 dst_ops->negative_advice = NULL;
2781 dst_ops->link_failure = NULL;
2782 afinfo->garbage_collect = NULL;
1da177e4 2783 }
1da177e4
LT		 2784 return err;
2785}
2786EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2787
d7c7544c
AD		 2788static void __net_init xfrm_dst_ops_init(struct net *net)
2789{
2790 struct xfrm_policy_afinfo *afinfo;
2791
ef8531b6
ED		 2792 rcu_read_lock();
2793 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2794 if (afinfo)
2795 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2796#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2797 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2798 if (afinfo)
2799 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2800#endif
418a99ac 2801 rcu_read_unlock();
546be240
HX		 2802}
2803
1da177e4
LT		 2804static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2805{
351638e7 2806 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2807
1da177e4
LT		 2808 switch (event) {
2809 case NETDEV_DOWN:
c0ed1c14 2810 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2811 }
2812 return NOTIFY_DONE;
2813}
2814
2815static struct notifier_block xfrm_dev_notifier = {
d5917a35 2816 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2817};
2818
558f82ef 2819#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2820static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2821{
c68cd1a0
AD		 2822 int rv;
2823
7d720c3e 2824 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
1823e4c8
ED		 2825 sizeof(struct linux_xfrm_mib),
2826 __alignof__(struct linux_xfrm_mib)) < 0)
558f82ef 2827 return -ENOMEM;
c68cd1a0
AD		 2828 rv = xfrm_proc_init(net);
2829 if (rv < 0)
7d720c3e 2830 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
c68cd1a0 2831 return rv;
558f82ef 2832}
59c9940e
AD		 2833
2834static void xfrm_statistics_fini(struct net *net)
2835{
c68cd1a0 2836 xfrm_proc_fini(net);
7d720c3e 2837 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
59c9940e
AD		 2838}
2839#else
2840static int __net_init xfrm_statistics_init(struct net *net)
2841{
2842 return 0;
2843}
2844
2845static void xfrm_statistics_fini(struct net *net)
2846{
2847}
558f82ef
MN		 2848#endif
2849
d62ddc21 2850static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2851{
2518c7c2
DM		 2852 unsigned int hmask, sz;
2853 int dir;
2854
d62ddc21
AD		 2855 if (net_eq(net, &init_net))
2856 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2857 sizeof(struct xfrm_dst),
e5d679f3 2858 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2859 NULL);
1da177e4 2860
2518c7c2
DM		 2861 hmask = 8 - 1;
2862 sz = (hmask+1) * sizeof(struct hlist_head);
2863
93b851c1
AD		 2864 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2865 if (!net->xfrm.policy_byidx)
2866 goto out_byidx;
8100bea7 2867 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2868
2869 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2870 struct xfrm_policy_hash *htab;
2871
dc2caba7 2872 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2873 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2874
a35f6c5d 2875 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2876 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2877 if (!htab->table)
a35f6c5d
AD		 2878 goto out_bydst;
2879 htab->hmask = hmask;
2518c7c2
DM		 2880 }
2881
adfcf0b2 2882 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2883 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2884 if (net_eq(net, &init_net))
2885 register_netdevice_notifier(&xfrm_dev_notifier);
2886 return 0;
93b851c1 2887
a35f6c5d
AD		 2888out_bydst:
2889 for (dir--; dir >= 0; dir--) {
2890 struct xfrm_policy_hash *htab;
2891
2892 htab = &net->xfrm.policy_bydst[dir];
2893 xfrm_hash_free(htab->table, sz);
2894 }
2895 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2896out_byidx:
2897 return -ENOMEM;
d62ddc21
AD		 2898}
2899
2900static void xfrm_policy_fini(struct net *net)
2901{
7c2776ee 2902 struct xfrm_audit audit_info;
93b851c1 2903 unsigned int sz;
8b18f8ea 2904 int dir;
93b851c1 2905
7c2776ee
AD		 2906 flush_work(&net->xfrm.policy_hash_work);
2907#ifdef CONFIG_XFRM_SUB_POLICY
e1760bd5 2908 audit_info.loginuid = INVALID_UID;
7c2776ee
AD		 2909 audit_info.sessionid = -1;
2910 audit_info.secid = 0;
2911 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2912#endif
e1760bd5 2913 audit_info.loginuid = INVALID_UID;
7c2776ee
AD		 2914 audit_info.sessionid = -1;
2915 audit_info.secid = 0;
2916 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
7c2776ee 2917
adfcf0b2 2918 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2919
8b18f8ea 2920 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2921 struct xfrm_policy_hash *htab;
2922
8b18f8ea 2923 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2924
2925 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2926 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2927 WARN_ON(!hlist_empty(htab->table));
2928 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2929 }
2930
8100bea7 2931 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2932 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2933 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2934}
2935
d62ddc21
AD		 2936static int __net_init xfrm_net_init(struct net *net)
2937{
2938 int rv;
2939
59c9940e
AD		 2940 rv = xfrm_statistics_init(net);
2941 if (rv < 0)
2942 goto out_statistics;
d62ddc21
AD		 2943 rv = xfrm_state_init(net);
2944 if (rv < 0)
2945 goto out_state;
2946 rv = xfrm_policy_init(net);
2947 if (rv < 0)
2948 goto out_policy;
d7c7544c 2949 xfrm_dst_ops_init(net);
b27aeadb
AD		 2950 rv = xfrm_sysctl_init(net);
2951 if (rv < 0)
2952 goto out_sysctl;
d62ddc21
AD		 2953 return 0;
2954
b27aeadb
AD		 2955out_sysctl:
2956 xfrm_policy_fini(net);
d62ddc21
AD		 2957out_policy:
2958 xfrm_state_fini(net);
2959out_state:
59c9940e
AD		 2960 xfrm_statistics_fini(net);
2961out_statistics:
d62ddc21
AD		 2962 return rv;
2963}
2964
2965static void __net_exit xfrm_net_exit(struct net *net)
2966{
b27aeadb 2967 xfrm_sysctl_fini(net);
d62ddc21
AD		 2968 xfrm_policy_fini(net);
2969 xfrm_state_fini(net);
59c9940e 2970 xfrm_statistics_fini(net);
d62ddc21
AD		 2971}
2972
2973static struct pernet_operations __net_initdata xfrm_net_ops = {
2974 .init = xfrm_net_init,
2975 .exit = xfrm_net_exit,
2976};
2977
1da177e4
LT		 2978void __init xfrm_init(void)
2979{
d62ddc21 2980 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 2981 xfrm_input_init();
2982}
2983
ab5f5e8b 2984#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2985static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2986 struct audit_buffer *audit_buf)
ab5f5e8b 2987{
875179fa
PM		 2988 struct xfrm_sec_ctx *ctx = xp->security;
2989 struct xfrm_selector *sel = &xp->selector;
2990
2991 if (ctx)
ab5f5e8b 2992 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2993 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2994
875179fa 2995 switch(sel->family) {
ab5f5e8b 2996 case AF_INET:
21454aaa 2997 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2998 if (sel->prefixlen_s != 32)
2999 audit_log_format(audit_buf, " src_prefixlen=%d",
3000 sel->prefixlen_s);
21454aaa 3001 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3002 if (sel->prefixlen_d != 32)
3003 audit_log_format(audit_buf, " dst_prefixlen=%d",
3004 sel->prefixlen_d);
ab5f5e8b
JL		 3005 break;
3006 case AF_INET6:
5b095d98 3007 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3008 if (sel->prefixlen_s != 128)
3009 audit_log_format(audit_buf, " src_prefixlen=%d",
3010 sel->prefixlen_s);
5b095d98 3011 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3012 if (sel->prefixlen_d != 128)
3013 audit_log_format(audit_buf, " dst_prefixlen=%d",
3014 sel->prefixlen_d);
ab5f5e8b
JL		 3015 break;
3016 }
3017}
3018
68277acc 3019void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
e1760bd5 3020 kuid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 3021{
3022 struct audit_buffer *audit_buf;
ab5f5e8b 3023
afeb14b4 3024 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3025 if (audit_buf == NULL)
3026 return;
2532386f 3027 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 3028 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3029 xfrm_audit_common_policyinfo(xp, audit_buf);
3030 audit_log_end(audit_buf);
3031}
3032EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3033
68277acc 3034void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
e1760bd5 3035 kuid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 3036{
3037 struct audit_buffer *audit_buf;
ab5f5e8b 3038
afeb14b4 3039 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3040 if (audit_buf == NULL)
3041 return;
2532386f 3042 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 3043 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3044 xfrm_audit_common_policyinfo(xp, audit_buf);
3045 audit_log_end(audit_buf);
3046}
3047EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3048#endif
3049
80c9abaa 3050#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3051static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3052 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3053{
3054 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3055 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3056 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3057 sel_cmp->family) &&
3058 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3059 sel_cmp->family) &&
80c9abaa
SS		 3060 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3061 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3062 return true;
80c9abaa
SS		 3063 }
3064 } else {
3065 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3066 return true;
80c9abaa
SS		 3067 }
3068 }
bc9b35ad 3069 return false;
80c9abaa
SS		 3070}
3071
b4b7c0b3 3072static struct xfrm_policy * xfrm_migrate_policy_find(const struct xfrm_selector *sel,
80c9abaa
SS		 3073 u8 dir, u8 type)
3074{
3075 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3076 struct hlist_head *chain;
3077 u32 priority = ~0U;
3078
3079 read_lock_bh(&xfrm_policy_lock);
1121994c 3080 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3081 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3082 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3083 pol->type == type) {
3084 ret = pol;
3085 priority = ret->priority;
3086 break;
3087 }
3088 }
8b18f8ea 3089 chain = &init_net.xfrm.policy_inexact[dir];
b67bfe0d 3090 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3091 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3092 pol->type == type &&
3093 pol->priority < priority) {
3094 ret = pol;
3095 break;
3096 }
3097 }
3098
3099 if (ret)
3100 xfrm_pol_hold(ret);
3101
3102 read_unlock_bh(&xfrm_policy_lock);
3103
3104 return ret;
3105}
3106
dd701754 3107static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3108{
3109 int match = 0;
3110
3111 if (t->mode == m->mode && t->id.proto == m->proto &&
3112 (m->reqid == 0 || t->reqid == m->reqid)) {
3113 switch (t->mode) {
3114 case XFRM_MODE_TUNNEL:
3115 case XFRM_MODE_BEET:
70e94e66
YH		 3116 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3117 m->old_family) &&
3118 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3119 m->old_family)) {
80c9abaa
SS		 3120 match = 1;
3121 }
3122 break;
3123 case XFRM_MODE_TRANSPORT:
3124 /* in case of transport mode, template does not store
3125 any IP addresses, hence we just compare mode and
3126 protocol */
3127 match = 1;
3128 break;
3129 default:
3130 break;
3131 }
3132 }
3133 return match;
3134}
3135
3136/* update endpoint address(es) of template(s) */
3137static int xfrm_policy_migrate(struct xfrm_policy *pol,
3138 struct xfrm_migrate *m, int num_migrate)
3139{
3140 struct xfrm_migrate *mp;
80c9abaa
SS		 3141 int i, j, n = 0;
3142
3143 write_lock_bh(&pol->lock);
12a169e7 3144 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3145 /* target policy has been deleted */
3146 write_unlock_bh(&pol->lock);
3147 return -ENOENT;
3148 }
3149
3150 for (i = 0; i < pol->xfrm_nr; i++) {
3151 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3152 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3153 continue;
3154 n++;
1bfcb10f
HX		 3155 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3156 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3157 continue;
3158 /* update endpoints */
3159 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3160 sizeof(pol->xfrm_vec[i].id.daddr));
3161 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3162 sizeof(pol->xfrm_vec[i].saddr));
3163 pol->xfrm_vec[i].encap_family = mp->new_family;
3164 /* flush bundles */
80c802f3 3165 atomic_inc(&pol->genid);
80c9abaa
SS		 3166 }
3167 }
3168
3169 write_unlock_bh(&pol->lock);
3170
3171 if (!n)
3172 return -ENODATA;
3173
3174 return 0;
3175}
3176
dd701754 3177static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3178{
3179 int i, j;
3180
3181 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3182 return -EINVAL;
3183
3184 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3185 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3186 m[i].old_family) &&
3187 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3188 m[i].old_family))
80c9abaa
SS		 3189 return -EINVAL;
3190 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3191 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3192 return -EINVAL;
3193
3194 /* check if there is any duplicated entry */
3195 for (j = i + 1; j < num_migrate; j++) {
3196 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3197 sizeof(m[i].old_daddr)) &&
3198 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3199 sizeof(m[i].old_saddr)) &&
3200 m[i].proto == m[j].proto &&
3201 m[i].mode == m[j].mode &&
3202 m[i].reqid == m[j].reqid &&
3203 m[i].old_family == m[j].old_family)
3204 return -EINVAL;
3205 }
3206 }
3207
3208 return 0;
3209}
3210
b4b7c0b3 3211int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189
AE		 3212 struct xfrm_migrate *m, int num_migrate,
3213 struct xfrm_kmaddress *k)
80c9abaa
SS		 3214{
3215 int i, err, nx_cur = 0, nx_new = 0;
3216 struct xfrm_policy *pol = NULL;
3217 struct xfrm_state *x, *xc;
3218 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3219 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3220 struct xfrm_migrate *mp;
3221
3222 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3223 goto out;
3224
3225 /* Stage 1 - find policy */
3226 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
3227 err = -ENOENT;
3228 goto out;
3229 }
3230
3231 /* Stage 2 - find and update state(s) */
3232 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
3233 if ((x = xfrm_migrate_state_find(mp))) {
3234 x_cur[nx_cur] = x;
3235 nx_cur++;
3236 if ((xc = xfrm_state_migrate(x, mp))) {
3237 x_new[nx_new] = xc;
3238 nx_new++;
3239 } else {
3240 err = -ENODATA;
3241 goto restore_state;
3242 }
3243 }
3244 }
3245
3246 /* Stage 3 - update policy */
3247 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3248 goto restore_state;
3249
3250 /* Stage 4 - delete old state(s) */
3251 if (nx_cur) {
3252 xfrm_states_put(x_cur, nx_cur);
3253 xfrm_states_delete(x_cur, nx_cur);
3254 }
3255
3256 /* Stage 5 - announce */
13c1d189 3257 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3258
3259 xfrm_pol_put(pol);
3260
3261 return 0;
3262out:
3263 return err;
3264
3265restore_state:
3266 if (pol)
3267 xfrm_pol_put(pol);
3268 if (nx_cur)
3269 xfrm_states_put(x_cur, nx_cur);
3270 if (nx_new)
3271 xfrm_states_delete(x_new, nx_new);
3272
3273 return err;
3274}
e610e679 3275EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3276#endif
Linux 6.x block layer and io_uring tree(s)