projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'nfs-for-4.14-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
[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>
ec30d78c 27#include <linux/cpu.h>
68277acc 28#include <linux/audit.h>
25ee3286 29#include <net/dst.h>
6ce74ec7 30#include <net/flow.h>
1da177e4
LT		 31#include <net/xfrm.h>
32#include <net/ip.h>
558f82ef
MN		 33#ifdef CONFIG_XFRM_STATISTICS
34#include <net/snmp.h>
35#endif
1da177e4 36
44e36b42
DM		 37#include "xfrm_hash.h"
38
a0073fe1
SK		 39#define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
40#define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
41#define XFRM_MAX_QUEUE_LEN 100
42
b8c203b2
SK		 43struct xfrm_flo {
44 struct dst_entry *dst_orig;
45 u8 flags;
46};
47
ec30d78c
FW		 48static DEFINE_PER_CPU(struct xfrm_dst *, xfrm_last_dst);
49static struct work_struct *xfrm_pcpu_work __read_mostly;
418a99ac 50static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
37b10383 51static struct xfrm_policy_afinfo const __rcu *xfrm_policy_afinfo[AF_INET6 + 1]
418a99ac 52 __read_mostly;
1da177e4 53
e18b890b 54static struct kmem_cache *xfrm_dst_cache __read_mostly;
30846090 55static __read_mostly seqcount_t xfrm_policy_hash_generation;
1da177e4 56
25ee3286 57static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 58static int stale_bundle(struct dst_entry *dst);
12fdb4d3 59static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 60static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 61
12bfa8bd 62static void __xfrm_policy_link(struct xfrm_policy *pol, int dir);
29fa0b30
WY		 63static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
64 int dir);
65
e37cc8ad
FW		 66static inline bool xfrm_pol_hold_rcu(struct xfrm_policy *policy)
67{
850a6212 68 return refcount_inc_not_zero(&policy->refcnt);
e37cc8ad
FW		 69}
70
bc9b35ad 71static inline bool
200ce96e 72__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 73{
7e1dc7b6
DM		 74 const struct flowi4 *fl4 = &fl->u.ip4;
75
26bff940
AD		 76 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
77 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 78 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
79 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
80 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
81 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 82}
83
bc9b35ad 84static inline bool
200ce96e 85__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 86{
7e1dc7b6
DM		 87 const struct flowi6 *fl6 = &fl->u.ip6;
88
89 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
90 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
91 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
92 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
93 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
94 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 95}
96
bc9b35ad
DM		 97bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
98 unsigned short family)
77681021
AM		 99{
100 switch (family) {
101 case AF_INET:
102 return __xfrm4_selector_match(sel, fl);
103 case AF_INET6:
104 return __xfrm6_selector_match(sel, fl);
105 }
bc9b35ad 106 return false;
77681021
AM		 107}
108
a2817d8b 109static const struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
ef8531b6 110{
a2817d8b 111 const struct xfrm_policy_afinfo *afinfo;
ef8531b6 112
a2817d8b 113 if (unlikely(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
ef8531b6
ED		 114 return NULL;
115 rcu_read_lock();
116 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
117 if (unlikely(!afinfo))
118 rcu_read_unlock();
119 return afinfo;
120}
121
d77e38e6
SK		 122struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos, int oif,
123 const xfrm_address_t *saddr,
124 const xfrm_address_t *daddr,
077fbac4 125 int family, u32 mark)
9bb182a7 126{
37b10383 127 const struct xfrm_policy_afinfo *afinfo;
9bb182a7
YH		 128 struct dst_entry *dst;
129
130 afinfo = xfrm_policy_get_afinfo(family);
131 if (unlikely(afinfo == NULL))
132 return ERR_PTR(-EAFNOSUPPORT);
133
077fbac4 134 dst = afinfo->dst_lookup(net, tos, oif, saddr, daddr, mark);
9bb182a7 135
bdba9fe0 136 rcu_read_unlock();
9bb182a7
YH		 137
138 return dst;
139}
d77e38e6 140EXPORT_SYMBOL(__xfrm_dst_lookup);
9bb182a7 141
42a7b32b
DA		 142static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x,
143 int tos, int oif,
9bb182a7
YH		 144 xfrm_address_t *prev_saddr,
145 xfrm_address_t *prev_daddr,
077fbac4 146 int family, u32 mark)
1da177e4 147{
c5b3cf46 148 struct net *net = xs_net(x);
66cdb3ca
HX		 149 xfrm_address_t *saddr = &x->props.saddr;
150 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 151 struct dst_entry *dst;
152
9bb182a7 153 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 154 saddr = x->coaddr;
9bb182a7
YH		 155 daddr = prev_daddr;
156 }
157 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
158 saddr = prev_saddr;
66cdb3ca 159 daddr = x->coaddr;
9bb182a7 160 }
1da177e4 161
077fbac4 162 dst = __xfrm_dst_lookup(net, tos, oif, saddr, daddr, family, mark);
9bb182a7
YH		 163
164 if (!IS_ERR(dst)) {
165 if (prev_saddr != saddr)
166 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
167 if (prev_daddr != daddr)
168 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
169 }
1da177e4 170
66cdb3ca 171 return dst;
1da177e4 172}
1da177e4 173
1da177e4
LT		 174static inline unsigned long make_jiffies(long secs)
175{
176 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
177 return MAX_SCHEDULE_TIMEOUT-1;
178 else
a716c119 179 return secs*HZ;
1da177e4
LT		 180}
181
182static void xfrm_policy_timer(unsigned long data)
183{
3e94c2dc 184 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 185 unsigned long now = get_seconds();
1da177e4
LT		 186 long next = LONG_MAX;
187 int warn = 0;
188 int dir;
189
190 read_lock(&xp->lock);
191
ea2dea9d 192 if (unlikely(xp->walk.dead))
1da177e4
LT		 193 goto out;
194
77d8d7a6 195 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 196
197 if (xp->lft.hard_add_expires_seconds) {
198 long tmo = xp->lft.hard_add_expires_seconds +
199 xp->curlft.add_time - now;
200 if (tmo <= 0)
201 goto expired;
202 if (tmo < next)
203 next = tmo;
204 }
205 if (xp->lft.hard_use_expires_seconds) {
206 long tmo = xp->lft.hard_use_expires_seconds +
207 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
208 if (tmo <= 0)
209 goto expired;
210 if (tmo < next)
211 next = tmo;
212 }
213 if (xp->lft.soft_add_expires_seconds) {
214 long tmo = xp->lft.soft_add_expires_seconds +
215 xp->curlft.add_time - now;
216 if (tmo <= 0) {
217 warn = 1;
218 tmo = XFRM_KM_TIMEOUT;
219 }
220 if (tmo < next)
221 next = tmo;
222 }
223 if (xp->lft.soft_use_expires_seconds) {
224 long tmo = xp->lft.soft_use_expires_seconds +
225 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
226 if (tmo <= 0) {
227 warn = 1;
228 tmo = XFRM_KM_TIMEOUT;
229 }
230 if (tmo < next)
231 next = tmo;
232 }
233
234 if (warn)
6c5c8ca7 235 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 236 if (next != LONG_MAX &&
237 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
238 xfrm_pol_hold(xp);
239
240out:
241 read_unlock(&xp->lock);
242 xfrm_pol_put(xp);
243 return;
244
245expired:
246 read_unlock(&xp->lock);
4666faab 247 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 248 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 249 xfrm_pol_put(xp);
250}
251
1da177e4
LT		 252/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
253 * SPD calls.
254 */
255
0331b1f3 256struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 257{
258 struct xfrm_policy *policy;
259
0da974f4 260 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 261
262 if (policy) {
0331b1f3 263 write_pnet(&policy->xp_net, net);
12a169e7 264 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 265 INIT_HLIST_NODE(&policy->bydst);
266 INIT_HLIST_NODE(&policy->byidx);
1da177e4 267 rwlock_init(&policy->lock);
850a6212 268 refcount_set(&policy->refcnt, 1);
a0073fe1 269 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 270 setup_timer(&policy->timer, xfrm_policy_timer,
271 (unsigned long)policy);
a0073fe1
SK		 272 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
273 (unsigned long)policy);
1da177e4
LT		 274 }
275 return policy;
276}
277EXPORT_SYMBOL(xfrm_policy_alloc);
278
56f04730
ED		 279static void xfrm_policy_destroy_rcu(struct rcu_head *head)
280{
281 struct xfrm_policy *policy = container_of(head, struct xfrm_policy, rcu);
282
283 security_xfrm_policy_free(policy->security);
284 kfree(policy);
285}
286
1da177e4
LT		 287/* Destroy xfrm_policy: descendant resources must be released to this moment. */
288
64c31b3f 289void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 290{
12a169e7 291 BUG_ON(!policy->walk.dead);
1da177e4 292
0659eea9 293 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 294 BUG();
295
56f04730 296 call_rcu(&policy->rcu, xfrm_policy_destroy_rcu);
1da177e4 297}
64c31b3f 298EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 299
1365e547 300/* Rule must be locked. Release descendant resources, announce
1da177e4
LT		 301 * entry dead. The rule must be unlinked from lists to the moment.
302 */
303
304static void xfrm_policy_kill(struct xfrm_policy *policy)
305{
12a169e7 306 policy->walk.dead = 1;
1da177e4 307
285ead17 308 atomic_inc(&policy->genid);
1da177e4 309
e7d8f6cb
SK		 310 if (del_timer(&policy->polq.hold_timer))
311 xfrm_pol_put(policy);
1ee5e667 312 skb_queue_purge(&policy->polq.hold_queue);
a0073fe1 313
285ead17
TT		 314 if (del_timer(&policy->timer))
315 xfrm_pol_put(policy);
316
317 xfrm_pol_put(policy);
1da177e4
LT		 318}
319
2518c7c2
DM		 320static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
321
e92303f8 322static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 323{
e92303f8 324 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 325}
326
b58555f1
CG		 327/* calculate policy hash thresholds */
328static void __get_hash_thresh(struct net *net,
329 unsigned short family, int dir,
330 u8 *dbits, u8 *sbits)
331{
332 switch (family) {
333 case AF_INET:
334 *dbits = net->xfrm.policy_bydst[dir].dbits4;
335 *sbits = net->xfrm.policy_bydst[dir].sbits4;
336 break;
337
338 case AF_INET6:
339 *dbits = net->xfrm.policy_bydst[dir].dbits6;
340 *sbits = net->xfrm.policy_bydst[dir].sbits6;
341 break;
342
343 default:
344 *dbits = 0;
345 *sbits = 0;
346 }
347}
348
5f803b58
DM		 349static struct hlist_head *policy_hash_bysel(struct net *net,
350 const struct xfrm_selector *sel,
351 unsigned short family, int dir)
2518c7c2 352{
1121994c 353 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 354 unsigned int hash;
355 u8 dbits;
356 u8 sbits;
357
358 __get_hash_thresh(net, family, dir, &dbits, &sbits);
359 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2 360
e1e551bc
FW		 361 if (hash == hmask + 1)
362 return &net->xfrm.policy_inexact[dir];
363
364 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
365 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM		 366}
367
5f803b58
DM		 368static struct hlist_head *policy_hash_direct(struct net *net,
369 const xfrm_address_t *daddr,
370 const xfrm_address_t *saddr,
371 unsigned short family, int dir)
2518c7c2 372{
1121994c 373 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 374 unsigned int hash;
375 u8 dbits;
376 u8 sbits;
377
378 __get_hash_thresh(net, family, dir, &dbits, &sbits);
379 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 380
e1e551bc
FW		 381 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
382 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM		 383}
384
b58555f1
CG		 385static void xfrm_dst_hash_transfer(struct net *net,
386 struct hlist_head *list,
2518c7c2 387 struct hlist_head *ndsttable,
b58555f1
CG		 388 unsigned int nhashmask,
389 int dir)
2518c7c2 390{
b67bfe0d 391 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 392 struct xfrm_policy *pol;
b791160b 393 unsigned int h0 = 0;
b58555f1
CG		 394 u8 dbits;
395 u8 sbits;
2518c7c2 396
b791160b 397redo:
b67bfe0d 398 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 399 unsigned int h;
400
b58555f1 401 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 402 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 403 pol->family, nhashmask, dbits, sbits);
b791160b 404 if (!entry0) {
a5eefc1d
FW		 405 hlist_del_rcu(&pol->bydst);
406 hlist_add_head_rcu(&pol->bydst, ndsttable + h);
b791160b
YH		 407 h0 = h;
408 } else {
409 if (h != h0)
410 continue;
a5eefc1d
FW		 411 hlist_del_rcu(&pol->bydst);
412 hlist_add_behind_rcu(&pol->bydst, entry0);
b791160b 413 }
b67bfe0d 414 entry0 = &pol->bydst;
b791160b
YH		 415 }
416 if (!hlist_empty(list)) {
417 entry0 = NULL;
418 goto redo;
2518c7c2
DM		 419 }
420}
421
422static void xfrm_idx_hash_transfer(struct hlist_head *list,
423 struct hlist_head *nidxtable,
424 unsigned int nhashmask)
425{
b67bfe0d 426 struct hlist_node *tmp;
2518c7c2
DM		 427 struct xfrm_policy *pol;
428
b67bfe0d 429 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 430 unsigned int h;
431
432 h = __idx_hash(pol->index, nhashmask);
433 hlist_add_head(&pol->byidx, nidxtable+h);
434 }
435}
436
437static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
438{
439 return ((old_hmask + 1) << 1) - 1;
440}
441
66caf628 442static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 443{
66caf628 444 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 445 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
446 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
44e36b42 447 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
e1e551bc 448 struct hlist_head *odst;
2518c7c2
DM		 449 int i;
450
451 if (!ndst)
452 return;
453
9d0380df 454 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
30846090
FW		 455 write_seqcount_begin(&xfrm_policy_hash_generation);
456
457 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
458 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
2518c7c2 459
e1e551bc
FW		 460 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
461 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
462
2518c7c2 463 for (i = hmask; i >= 0; i--)
b58555f1 464 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 465
e1e551bc 466 rcu_assign_pointer(net->xfrm.policy_bydst[dir].table, ndst);
66caf628 467 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 468
30846090 469 write_seqcount_end(&xfrm_policy_hash_generation);
9d0380df 470 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 471
e1e551bc
FW		 472 synchronize_rcu();
473
44e36b42 474 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 475}
476
66caf628 477static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 478{
66caf628 479 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 480 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
481 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 482 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 483 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 484 int i;
485
486 if (!nidx)
487 return;
488
9d0380df 489 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 490
491 for (i = hmask; i >= 0; i--)
492 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
493
66caf628
AD		 494 net->xfrm.policy_byidx = nidx;
495 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 496
9d0380df 497 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 498
44e36b42 499 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 500}
501
66caf628 502static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 503{
66caf628
AD		 504 unsigned int cnt = net->xfrm.policy_count[dir];
505 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 506
507 if (total)
508 *total += cnt;
509
510 if ((hmask + 1) < xfrm_policy_hashmax &&
511 cnt > hmask)
512 return 1;
513
514 return 0;
515}
516
66caf628 517static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 518{
66caf628 519 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 520
521 if ((hmask + 1) < xfrm_policy_hashmax &&
522 total > hmask)
523 return 1;
524
525 return 0;
526}
527
e071041b 528void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 529{
e071041b
AD		 530 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
531 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
532 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
533 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
534 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
535 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
536 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 537 si->spdhmcnt = xfrm_policy_hashmax;
ecfd6b18
JHS		 538}
539EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 540
ecfd6b18 541static DEFINE_MUTEX(hash_resize_mutex);
66caf628 542static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 543{
66caf628 544 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 545 int dir, total;
546
547 mutex_lock(&hash_resize_mutex);
548
549 total = 0;
53c2e285 550 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD		 551 if (xfrm_bydst_should_resize(net, dir, &total))
552 xfrm_bydst_resize(net, dir);
2518c7c2 553 }
66caf628
AD		 554 if (xfrm_byidx_should_resize(net, total))
555 xfrm_byidx_resize(net, total);
2518c7c2
DM		 556
557 mutex_unlock(&hash_resize_mutex);
558}
559
880a6fab
CG		 560static void xfrm_hash_rebuild(struct work_struct *work)
561{
562 struct net *net = container_of(work, struct net,
563 xfrm.policy_hthresh.work);
564 unsigned int hmask;
565 struct xfrm_policy *pol;
566 struct xfrm_policy *policy;
567 struct hlist_head *chain;
568 struct hlist_head *odst;
569 struct hlist_node *newpos;
570 int i;
571 int dir;
572 unsigned seq;
573 u8 lbits4, rbits4, lbits6, rbits6;
574
575 mutex_lock(&hash_resize_mutex);
576
577 /* read selector prefixlen thresholds */
578 do {
579 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
580
581 lbits4 = net->xfrm.policy_hthresh.lbits4;
582 rbits4 = net->xfrm.policy_hthresh.rbits4;
583 lbits6 = net->xfrm.policy_hthresh.lbits6;
584 rbits6 = net->xfrm.policy_hthresh.rbits6;
585 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
586
9d0380df 587 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 588
589 /* reset the bydst and inexact table in all directions */
53c2e285 590 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG		 591 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
592 hmask = net->xfrm.policy_bydst[dir].hmask;
593 odst = net->xfrm.policy_bydst[dir].table;
594 for (i = hmask; i >= 0; i--)
595 INIT_HLIST_HEAD(odst + i);
596 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
597 /* dir out => dst = remote, src = local */
598 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
599 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
600 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
601 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
602 } else {
603 /* dir in/fwd => dst = local, src = remote */
604 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
605 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
606 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
607 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
608 }
609 }
610
611 /* re-insert all policies by order of creation */
612 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
6916fb3b
TB		 613 if (xfrm_policy_id2dir(policy->index) >= XFRM_POLICY_MAX) {
614 /* skip socket policies */
615 continue;
616 }
880a6fab
CG		 617 newpos = NULL;
618 chain = policy_hash_bysel(net, &policy->selector,
619 policy->family,
620 xfrm_policy_id2dir(policy->index));
621 hlist_for_each_entry(pol, chain, bydst) {
622 if (policy->priority >= pol->priority)
623 newpos = &pol->bydst;
624 else
625 break;
626 }
627 if (newpos)
628 hlist_add_behind(&policy->bydst, newpos);
629 else
630 hlist_add_head(&policy->bydst, chain);
631 }
632
9d0380df 633 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG		 634
635 mutex_unlock(&hash_resize_mutex);
636}
637
638void xfrm_policy_hash_rebuild(struct net *net)
639{
640 schedule_work(&net->xfrm.policy_hthresh.work);
641}
642EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
643
1da177e4
LT		 644/* Generate new index... KAME seems to generate them ordered by cost
645 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 646static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 647{
1da177e4
LT		 648 static u32 idx_generator;
649
650 for (;;) {
2518c7c2
DM		 651 struct hlist_head *list;
652 struct xfrm_policy *p;
653 u32 idx;
654 int found;
655
e682adf0
FD		 656 if (!index) {
657 idx = (idx_generator | dir);
658 idx_generator += 8;
659 } else {
660 idx = index;
661 index = 0;
662 }
663
1da177e4
LT		 664 if (idx == 0)
665 idx = 8;
1121994c 666 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 667 found = 0;
b67bfe0d 668 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 669 if (p->index == idx) {
670 found = 1;
1da177e4 671 break;
2518c7c2 672 }
1da177e4 673 }
2518c7c2 674 if (!found)
1da177e4
LT		 675 return idx;
676 }
677}
678
2518c7c2
DM		 679static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
680{
681 u32 *p1 = (u32 *) s1;
682 u32 *p2 = (u32 *) s2;
683 int len = sizeof(struct xfrm_selector) / sizeof(u32);
684 int i;
685
686 for (i = 0; i < len; i++) {
687 if (p1[i] != p2[i])
688 return 1;
689 }
690
691 return 0;
692}
693
a0073fe1
SK		 694static void xfrm_policy_requeue(struct xfrm_policy *old,
695 struct xfrm_policy *new)
696{
697 struct xfrm_policy_queue *pq = &old->polq;
698 struct sk_buff_head list;
699
de2ad486
LR		 700 if (skb_queue_empty(&pq->hold_queue))
701 return;
702
a0073fe1
SK		 703 __skb_queue_head_init(&list);
704
705 spin_lock_bh(&pq->hold_queue.lock);
706 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 707 if (del_timer(&pq->hold_timer))
708 xfrm_pol_put(old);
a0073fe1
SK		 709 spin_unlock_bh(&pq->hold_queue.lock);
710
a0073fe1
SK		 711 pq = &new->polq;
712
713 spin_lock_bh(&pq->hold_queue.lock);
714 skb_queue_splice(&list, &pq->hold_queue);
715 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 716 if (!mod_timer(&pq->hold_timer, jiffies))
717 xfrm_pol_hold(new);
a0073fe1
SK		 718 spin_unlock_bh(&pq->hold_queue.lock);
719}
720
7cb8a939
SK		 721static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
722 struct xfrm_policy *pol)
723{
724 u32 mark = policy->mark.v & policy->mark.m;
725
726 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
727 return true;
728
729 if ((mark & pol->mark.m) == pol->mark.v &&
730 policy->priority == pol->priority)
731 return true;
732
733 return false;
734}
735
1da177e4
LT		 736int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
737{
1121994c 738 struct net *net = xp_net(policy);
2518c7c2
DM		 739 struct xfrm_policy *pol;
740 struct xfrm_policy *delpol;
741 struct hlist_head *chain;
b67bfe0d 742 struct hlist_node *newpos;
1da177e4 743
9d0380df 744 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 745 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 746 delpol = NULL;
747 newpos = NULL;
b67bfe0d 748 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 749 if (pol->type == policy->type &&
2518c7c2 750 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 751 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 752 xfrm_sec_ctx_match(pol->security, policy->security) &&
753 !WARN_ON(delpol)) {
1da177e4 754 if (excl) {
9d0380df 755 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 756 return -EEXIST;
757 }
1da177e4
LT		 758 delpol = pol;
759 if (policy->priority > pol->priority)
760 continue;
761 } else if (policy->priority >= pol->priority) {
a6c7ab55 762 newpos = &pol->bydst;
1da177e4
LT		 763 continue;
764 }
1da177e4
LT		 765 if (delpol)
766 break;
1da177e4
LT		 767 }
768 if (newpos)
1d023284 769 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 770 else
771 hlist_add_head(&policy->bydst, chain);
12bfa8bd 772 __xfrm_policy_link(policy, dir);
ca4c3fc2 773
774 /* After previous checking, family can either be AF_INET or AF_INET6 */
775 if (policy->family == AF_INET)
776 rt_genid_bump_ipv4(net);
777 else
778 rt_genid_bump_ipv6(net);
779
a0073fe1
SK		 780 if (delpol) {
781 xfrm_policy_requeue(delpol, policy);
29fa0b30 782 __xfrm_policy_unlink(delpol, dir);
a0073fe1 783 }
e682adf0 784 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 785 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 786 policy->curlft.add_time = get_seconds();
1da177e4
LT		 787 policy->curlft.use_time = 0;
788 if (!mod_timer(&policy->timer, jiffies + HZ))
789 xfrm_pol_hold(policy);
9d0380df 790 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 791
9b78a82c 792 if (delpol)
1da177e4 793 xfrm_policy_kill(delpol);
1121994c
AD		 794 else if (xfrm_bydst_should_resize(net, dir, NULL))
795 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 796
1da177e4
LT		 797 return 0;
798}
799EXPORT_SYMBOL(xfrm_policy_insert);
800
8ca2e93b
JHS		 801struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
802 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 803 struct xfrm_sec_ctx *ctx, int delete,
804 int *err)
1da177e4 805{
2518c7c2
DM		 806 struct xfrm_policy *pol, *ret;
807 struct hlist_head *chain;
1da177e4 808
ef41aaa0 809 *err = 0;
9d0380df 810 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 811 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 812 ret = NULL;
b67bfe0d 813 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 814 if (pol->type == type &&
34f8d884 815 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 816 !selector_cmp(sel, &pol->selector) &&
817 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 818 xfrm_pol_hold(pol);
2518c7c2 819 if (delete) {
03e1ad7b
PM		 820 *err = security_xfrm_policy_delete(
821 pol->security);
ef41aaa0 822 if (*err) {
9d0380df 823 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 824 return pol;
825 }
29fa0b30 826 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 827 }
828 ret = pol;
1da177e4
LT		 829 break;
830 }
831 }
9d0380df 832 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 833
fe1a5f03 834 if (ret && delete)
2518c7c2 835 xfrm_policy_kill(ret);
2518c7c2 836 return ret;
1da177e4 837}
df71837d 838EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 839
8ca2e93b
JHS		 840struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
841 int dir, u32 id, int delete, int *err)
1da177e4 842{
2518c7c2
DM		 843 struct xfrm_policy *pol, *ret;
844 struct hlist_head *chain;
1da177e4 845
b5505c6e
HX		 846 *err = -ENOENT;
847 if (xfrm_policy_id2dir(id) != dir)
848 return NULL;
849
ef41aaa0 850 *err = 0;
9d0380df 851 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 852 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 853 ret = NULL;
b67bfe0d 854 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 855 if (pol->type == type && pol->index == id &&
856 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 857 xfrm_pol_hold(pol);
2518c7c2 858 if (delete) {
03e1ad7b
PM		 859 *err = security_xfrm_policy_delete(
860 pol->security);
ef41aaa0 861 if (*err) {
9d0380df 862 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 863 return pol;
864 }
29fa0b30 865 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 866 }
867 ret = pol;
1da177e4
LT		 868 break;
869 }
870 }
9d0380df 871 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 872
fe1a5f03 873 if (ret && delete)
2518c7c2 874 xfrm_policy_kill(ret);
2518c7c2 875 return ret;
1da177e4
LT		 876}
877EXPORT_SYMBOL(xfrm_policy_byid);
878
4aa2e62c
JL		 879#ifdef CONFIG_SECURITY_NETWORK_XFRM
880static inline int
2e71029e 881xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 882{
4aa2e62c
JL		 883 int dir, err = 0;
884
885 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
886 struct xfrm_policy *pol;
4aa2e62c
JL		 887 int i;
888
b67bfe0d 889 hlist_for_each_entry(pol,
33ffbbd5 890 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 891 if (pol->type != type)
892 continue;
03e1ad7b 893 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 894 if (err) {
2e71029e 895 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 896 return err;
897 }
7dc12d6d 898 }
33ffbbd5 899 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 900 hlist_for_each_entry(pol,
33ffbbd5 901 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 902 bydst) {
903 if (pol->type != type)
904 continue;
03e1ad7b
PM		 905 err = security_xfrm_policy_delete(
906 pol->security);
4aa2e62c 907 if (err) {
ab5f5e8b 908 xfrm_audit_policy_delete(pol, 0,
2e71029e 909 task_valid);
4aa2e62c
JL		 910 return err;
911 }
912 }
913 }
914 }
915 return err;
916}
917#else
918static inline int
2e71029e 919xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 920{
921 return 0;
922}
923#endif
924
2e71029e 925int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 926{
2f1eb65f 927 int dir, err = 0, cnt = 0;
1da177e4 928
9d0380df 929 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 930
2e71029e 931 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 932 if (err)
933 goto out;
934
1da177e4 935 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 936 struct xfrm_policy *pol;
29fa0b30 937 int i;
2518c7c2
DM		 938
939 again1:
b67bfe0d 940 hlist_for_each_entry(pol,
33ffbbd5 941 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 942 if (pol->type != type)
943 continue;
ea2dea9d 944 __xfrm_policy_unlink(pol, dir);
9d0380df 945 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 946 cnt++;
1da177e4 947
2e71029e 948 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 949
2518c7c2 950 xfrm_policy_kill(pol);
1da177e4 951
4141b36a 952 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 953 goto again1;
954 }
955
33ffbbd5 956 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 957 again2:
b67bfe0d 958 hlist_for_each_entry(pol,
33ffbbd5 959 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 960 bydst) {
961 if (pol->type != type)
962 continue;
ea2dea9d 963 __xfrm_policy_unlink(pol, dir);
9d0380df 964 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 965 cnt++;
2518c7c2 966
2e71029e 967 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 968 xfrm_policy_kill(pol);
969
9d0380df 970 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 971 goto again2;
972 }
1da177e4 973 }
2518c7c2 974
1da177e4 975 }
2f1eb65f
JHS		 976 if (!cnt)
977 err = -ESRCH;
ec30d78c
FW		 978 else
979 xfrm_policy_cache_flush();
4aa2e62c 980out:
9d0380df 981 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 982 return err;
1da177e4
LT		 983}
984EXPORT_SYMBOL(xfrm_policy_flush);
985
cdcbca7c 986int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 987 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 988 void *data)
989{
12a169e7
HX		 990 struct xfrm_policy *pol;
991 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 992 int error = 0;
993
994 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
995 walk->type != XFRM_POLICY_TYPE_ANY)
996 return -EINVAL;
1da177e4 997
12a169e7 998 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 999 return 0;
1000
9d0380df 1001 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1002 if (list_empty(&walk->walk.all))
cdcbca7c 1003 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1004 else
80077702
LR		 1005 x = list_first_entry(&walk->walk.all,
1006 struct xfrm_policy_walk_entry, all);
1007
cdcbca7c 1008 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1009 if (x->dead)
4c563f76 1010 continue;
12a169e7
HX		 1011 pol = container_of(x, struct xfrm_policy, walk);
1012 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1013 walk->type != pol->type)
1014 continue;
1015 error = func(pol, xfrm_policy_id2dir(pol->index),
1016 walk->seq, data);
1017 if (error) {
1018 list_move_tail(&walk->walk.all, &x->all);
1019 goto out;
2518c7c2 1020 }
12a169e7 1021 walk->seq++;
1da177e4 1022 }
12a169e7 1023 if (walk->seq == 0) {
baf5d743
JHS		 1024 error = -ENOENT;
1025 goto out;
1026 }
12a169e7 1027 list_del_init(&walk->walk.all);
1da177e4 1028out:
9d0380df 1029 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1030 return error;
1031}
1032EXPORT_SYMBOL(xfrm_policy_walk);
1033
12a169e7
HX		 1034void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1035{
1036 INIT_LIST_HEAD(&walk->walk.all);
1037 walk->walk.dead = 1;
1038 walk->type = type;
1039 walk->seq = 0;
1040}
1041EXPORT_SYMBOL(xfrm_policy_walk_init);
1042
283bc9f3 1043void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1044{
1045 if (list_empty(&walk->walk.all))
1046 return;
1047
9d0380df 1048 spin_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1049 list_del(&walk->walk.all);
9d0380df 1050 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1051}
1052EXPORT_SYMBOL(xfrm_policy_walk_done);
1053
134b0fc5
JM		 1054/*
1055 * Find policy to apply to this flow.
1056 *
1057 * Returns 0 if policy found, else an -errno.
1058 */
f299d557
DM		 1059static int xfrm_policy_match(const struct xfrm_policy *pol,
1060 const struct flowi *fl,
2518c7c2 1061 u8 type, u16 family, int dir)
1da177e4 1062{
f299d557 1063 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1064 int ret = -ESRCH;
1065 bool match;
1da177e4 1066
2518c7c2 1067 if (pol->family != family ||
1d28f42c 1068 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1069 pol->type != type)
134b0fc5 1070 return ret;
1da177e4 1071
2518c7c2 1072 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1073 if (match)
1d28f42c 1074 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1075 dir);
2518c7c2 1076
134b0fc5 1077 return ret;
2518c7c2 1078}
1da177e4 1079
52479b62 1080static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1081 const struct flowi *fl,
2518c7c2
DM		 1082 u16 family, u8 dir)
1083{
134b0fc5 1084 int err;
2518c7c2 1085 struct xfrm_policy *pol, *ret;
0b597e7e 1086 const xfrm_address_t *daddr, *saddr;
2518c7c2 1087 struct hlist_head *chain;
30846090
FW		 1088 unsigned int sequence;
1089 u32 priority;
df71837d 1090
2518c7c2
DM		 1091 daddr = xfrm_flowi_daddr(fl, family);
1092 saddr = xfrm_flowi_saddr(fl, family);
1093 if (unlikely(!daddr || !saddr))
1094 return NULL;
1095
a7c44247 1096 rcu_read_lock();
30846090
FW		 1097 retry:
1098 do {
1099 sequence = read_seqcount_begin(&xfrm_policy_hash_generation);
1100 chain = policy_hash_direct(net, daddr, saddr, family, dir);
1101 } while (read_seqcount_retry(&xfrm_policy_hash_generation, sequence));
1102
1103 priority = ~0U;
2518c7c2 1104 ret = NULL;
a5eefc1d 1105 hlist_for_each_entry_rcu(pol, chain, bydst) {
134b0fc5
JM		 1106 err = xfrm_policy_match(pol, fl, type, family, dir);
1107 if (err) {
1108 if (err == -ESRCH)
1109 continue;
1110 else {
1111 ret = ERR_PTR(err);
1112 goto fail;
1113 }
1114 } else {
2518c7c2 1115 ret = pol;
acba48e1 1116 priority = ret->priority;
2518c7c2
DM		 1117 break;
1118 }
1119 }
52479b62 1120 chain = &net->xfrm.policy_inexact[dir];
a5eefc1d 1121 hlist_for_each_entry_rcu(pol, chain, bydst) {
8faf491e
LR		 1122 if ((pol->priority >= priority) && ret)
1123 break;
1124
134b0fc5
JM		 1125 err = xfrm_policy_match(pol, fl, type, family, dir);
1126 if (err) {
1127 if (err == -ESRCH)
1128 continue;
1129 else {
1130 ret = ERR_PTR(err);
1131 goto fail;
1132 }
8faf491e 1133 } else {
acba48e1
DM		 1134 ret = pol;
1135 break;
1da177e4
LT		 1136 }
1137 }
586f2eb4 1138
30846090
FW		 1139 if (read_seqcount_retry(&xfrm_policy_hash_generation, sequence))
1140 goto retry;
1141
e37cc8ad
FW		 1142 if (ret && !xfrm_pol_hold_rcu(ret))
1143 goto retry;
134b0fc5 1144fail:
a7c44247 1145 rcu_read_unlock();
4e81bb83 1146
2518c7c2 1147 return ret;
4e81bb83
MN		 1148}
1149
80c802f3 1150static struct xfrm_policy *
86dc8ee0 1151xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1152{
1153#ifdef CONFIG_XFRM_SUB_POLICY
1154 struct xfrm_policy *pol;
1155
1156 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1157 if (pol != NULL)
1158 return pol;
1159#endif
1160 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1161}
1162
6f9c9615 1163static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir,
4c86d777 1164 const struct flowi *fl, u16 family)
1da177e4
LT		 1165{
1166 struct xfrm_policy *pol;
1167
d188ba86 1168 rcu_read_lock();
ae33786f 1169 again:
d188ba86
ED		 1170 pol = rcu_dereference(sk->sk_policy[dir]);
1171 if (pol != NULL) {
4c86d777 1172 bool match = xfrm_selector_match(&pol->selector, fl, family);
a716c119 1173 int err = 0;
df71837d 1174
3bccfbc7 1175 if (match) {
34f8d884
JHS		 1176 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1177 pol = NULL;
1178 goto out;
1179 }
03e1ad7b 1180 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1181 fl->flowi_secid,
aff669bc 1182 dir);
330e832a
FW		 1183 if (!err) {
1184 if (!xfrm_pol_hold_rcu(pol))
1185 goto again;
1186 } else if (err == -ESRCH) {
3bccfbc7 1187 pol = NULL;
330e832a 1188 } else {
3bccfbc7 1189 pol = ERR_PTR(err);
330e832a 1190 }
3bccfbc7 1191 } else
1da177e4
LT		 1192 pol = NULL;
1193 }
34f8d884 1194out:
d188ba86 1195 rcu_read_unlock();
1da177e4
LT		 1196 return pol;
1197}
1198
1199static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1200{
98806f75 1201 struct net *net = xp_net(pol);
4e81bb83 1202
98806f75 1203 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1204 net->xfrm.policy_count[dir]++;
1da177e4
LT		 1205 xfrm_pol_hold(pol);
1206}
1207
1208static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1209 int dir)
1210{
98806f75
AD		 1211 struct net *net = xp_net(pol);
1212
53c2e285 1213 if (list_empty(&pol->walk.all))
2518c7c2 1214 return NULL;
1da177e4 1215
53c2e285
HX		 1216 /* Socket policies are not hashed. */
1217 if (!hlist_unhashed(&pol->bydst)) {
a5eefc1d 1218 hlist_del_rcu(&pol->bydst);
53c2e285
HX		 1219 hlist_del(&pol->byidx);
1220 }
1221
1222 list_del_init(&pol->walk.all);
98806f75 1223 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1224
1225 return pol;
1da177e4
LT		 1226}
1227
53c2e285
HX		 1228static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1229{
1230 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1231}
1232
1233static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1234{
1235 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1236}
1237
4666faab 1238int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1239{
283bc9f3
FD		 1240 struct net *net = xp_net(pol);
1241
9d0380df 1242 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1243 pol = __xfrm_policy_unlink(pol, dir);
9d0380df 1244 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1245 if (pol) {
1da177e4 1246 xfrm_policy_kill(pol);
4666faab 1247 return 0;
1da177e4 1248 }
4666faab 1249 return -ENOENT;
1da177e4 1250}
a70fcb0b 1251EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1252
1253int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1254{
1121994c 1255 struct net *net = xp_net(pol);
1da177e4
LT		 1256 struct xfrm_policy *old_pol;
1257
4e81bb83
MN		 1258#ifdef CONFIG_XFRM_SUB_POLICY
1259 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1260 return -EINVAL;
1261#endif
1262
9d0380df 1263 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
d188ba86
ED		 1264 old_pol = rcu_dereference_protected(sk->sk_policy[dir],
1265 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
1da177e4 1266 if (pol) {
9d729f72 1267 pol->curlft.add_time = get_seconds();
e682adf0 1268 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1269 xfrm_sk_policy_link(pol, dir);
1da177e4 1270 }
d188ba86 1271 rcu_assign_pointer(sk->sk_policy[dir], pol);
a0073fe1
SK		 1272 if (old_pol) {
1273 if (pol)
1274 xfrm_policy_requeue(old_pol, pol);
1275
ea2dea9d
TT		 1276 /* Unlinking succeeds always. This is the only function
1277 * allowed to delete or replace socket policy.
1278 */
53c2e285 1279 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1280 }
9d0380df 1281 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1282
1283 if (old_pol) {
1284 xfrm_policy_kill(old_pol);
1285 }
1286 return 0;
1287}
1288
d3e40a9f 1289static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1290{
0331b1f3 1291 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1292 struct net *net = xp_net(old);
1da177e4
LT		 1293
1294 if (newp) {
1295 newp->selector = old->selector;
03e1ad7b
PM		 1296 if (security_xfrm_policy_clone(old->security,
1297 &newp->security)) {
df71837d
TJ		 1298 kfree(newp);
1299 return NULL; /* ENOMEM */
1300 }
1da177e4
LT		 1301 newp->lft = old->lft;
1302 newp->curlft = old->curlft;
fb977e2c 1303 newp->mark = old->mark;
1da177e4
LT		 1304 newp->action = old->action;
1305 newp->flags = old->flags;
1306 newp->xfrm_nr = old->xfrm_nr;
1307 newp->index = old->index;
4e81bb83 1308 newp->type = old->type;
1da177e4
LT		 1309 memcpy(newp->xfrm_vec, old->xfrm_vec,
1310 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
9d0380df 1311 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1312 xfrm_sk_policy_link(newp, dir);
9d0380df 1313 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1314 xfrm_pol_put(newp);
1315 }
1316 return newp;
1317}
1318
d188ba86 1319int __xfrm_sk_clone_policy(struct sock *sk, const struct sock *osk)
1da177e4 1320{
d188ba86
ED		 1321 const struct xfrm_policy *p;
1322 struct xfrm_policy *np;
1323 int i, ret = 0;
1324
1325 rcu_read_lock();
1326 for (i = 0; i < 2; i++) {
1327 p = rcu_dereference(osk->sk_policy[i]);
1328 if (p) {
1329 np = clone_policy(p, i);
1330 if (unlikely(!np)) {
1331 ret = -ENOMEM;
1332 break;
1333 }
1334 rcu_assign_pointer(sk->sk_policy[i], np);
1335 }
1336 }
1337 rcu_read_unlock();
1338 return ret;
1da177e4
LT		 1339}
1340
a1e59abf 1341static int
42a7b32b 1342xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local,
077fbac4 1343 xfrm_address_t *remote, unsigned short family, u32 mark)
a1e59abf
PM		 1344{
1345 int err;
37b10383 1346 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
a1e59abf
PM		 1347
1348 if (unlikely(afinfo == NULL))
1349 return -EINVAL;
077fbac4 1350 err = afinfo->get_saddr(net, oif, local, remote, mark);
bdba9fe0 1351 rcu_read_unlock();
a1e59abf
PM		 1352 return err;
1353}
1354
1da177e4
LT		 1355/* Resolve list of templates for the flow, given policy. */
1356
1357static int
a6c2e611
DM		 1358xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1359 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1360{
fbda33b2 1361 struct net *net = xp_net(policy);
1da177e4
LT		 1362 int nx;
1363 int i, error;
1364 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1365 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1366 xfrm_address_t tmp;
1da177e4 1367
9b7a787d 1368 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1369 struct xfrm_state *x;
1370 xfrm_address_t *remote = daddr;
1371 xfrm_address_t *local = saddr;
1372 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1373
48b8d783
JK		 1374 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1375 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1376 remote = &tmpl->id.daddr;
1377 local = &tmpl->saddr;
8444cf71 1378 if (xfrm_addr_any(local, tmpl->encap_family)) {
42a7b32b
DA		 1379 error = xfrm_get_saddr(net, fl->flowi_oif,
1380 &tmp, remote,
077fbac4 1381 tmpl->encap_family, 0);
a1e59abf
PM		 1382 if (error)
1383 goto fail;
1384 local = &tmp;
1385 }
1da177e4
LT		 1386 }
1387
1388 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1389
1390 if (x && x->km.state == XFRM_STATE_VALID) {
1391 xfrm[nx++] = x;
1392 daddr = remote;
1393 saddr = local;
1394 continue;
1395 }
1396 if (x) {
1397 error = (x->km.state == XFRM_STATE_ERROR ?
1398 -EINVAL : -EAGAIN);
1399 xfrm_state_put(x);
42054569 1400 } else if (error == -ESRCH) {
a4322266 1401 error = -EAGAIN;
42054569 1402 }
1da177e4
LT		 1403
1404 if (!tmpl->optional)
1405 goto fail;
1406 }
1407 return nx;
1408
1409fail:
9b7a787d 1410 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1411 xfrm_state_put(xfrm[nx]);
1412 return error;
1413}
1414
4e81bb83 1415static int
a6c2e611
DM		 1416xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1417 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1418{
41a49cc3
MN		 1419 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1420 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1421 int cnx = 0;
1422 int error;
1423 int ret;
1424 int i;
1425
1426 for (i = 0; i < npols; i++) {
1427 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1428 error = -ENOBUFS;
1429 goto fail;
1430 }
41a49cc3
MN		 1431
1432 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1433 if (ret < 0) {
1434 error = ret;
1435 goto fail;
1436 } else
1437 cnx += ret;
1438 }
1439
41a49cc3
MN		 1440 /* found states are sorted for outbound processing */
1441 if (npols > 1)
1442 xfrm_state_sort(xfrm, tpp, cnx, family);
1443
4e81bb83
MN		 1444 return cnx;
1445
1446 fail:
9b7a787d 1447 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1448 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1449 return error;
1450
1451}
1452
f5e2bb4f 1453static int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286 1454{
37b10383 1455 const struct xfrm_policy_afinfo *afinfo;
f5e2bb4f 1456 int tos = 0;
1da177e4 1457
f5e2bb4f
FW		 1458 afinfo = xfrm_policy_get_afinfo(family);
1459 tos = afinfo ? afinfo->get_tos(fl) : 0;
25ee3286 1460
bdba9fe0 1461 rcu_read_unlock();
25ee3286
HX		 1462
1463 return tos;
1464}
1465
d7c7544c 1466static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1467{
37b10383 1468 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1469 struct dst_ops *dst_ops;
25ee3286
HX		 1470 struct xfrm_dst *xdst;
1471
1472 if (!afinfo)
1473 return ERR_PTR(-EINVAL);
1474
d7c7544c
AD		 1475 switch (family) {
1476 case AF_INET:
1477 dst_ops = &net->xfrm.xfrm4_dst_ops;
1478 break;
dfd56b8b 1479#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1480 case AF_INET6:
1481 dst_ops = &net->xfrm.xfrm6_dst_ops;
1482 break;
1483#endif
1484 default:
1485 BUG();
1486 }
b2a9c0ed 1487 xdst = dst_alloc(dst_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
25ee3286 1488
d4cae562 1489 if (likely(xdst)) {
141e369d
SK		 1490 struct dst_entry *dst = &xdst->u.dst;
1491
1492 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
d4cae562 1493 } else
0b150932 1494 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1495
bdba9fe0 1496 rcu_read_unlock();
d4cae562 1497
25ee3286
HX		 1498 return xdst;
1499}
1500
a1b05140
MN		 1501static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1502 int nfheader_len)
1503{
37b10383 1504 const struct xfrm_policy_afinfo *afinfo =
a1b05140
MN		 1505 xfrm_policy_get_afinfo(dst->ops->family);
1506 int err;
1507
1508 if (!afinfo)
1509 return -EINVAL;
1510
1511 err = afinfo->init_path(path, dst, nfheader_len);
1512
bdba9fe0 1513 rcu_read_unlock();
a1b05140
MN		 1514
1515 return err;
1516}
1517
87c1e12b 1518static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1519 const struct flowi *fl)
25ee3286 1520{
37b10383 1521 const struct xfrm_policy_afinfo *afinfo =
25ee3286
HX		 1522 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1523 int err;
1524
1525 if (!afinfo)
1da177e4 1526 return -EINVAL;
25ee3286 1527
87c1e12b 1528 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1529
bdba9fe0 1530 rcu_read_unlock();
25ee3286 1531
1da177e4
LT		 1532 return err;
1533}
1534
80c802f3 1535
25ee3286
HX		 1536/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1537 * all the metrics... Shortly, bundle a bundle.
1538 */
1539
1540static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1541 struct xfrm_state **xfrm, int nx,
98313ada 1542 const struct flowi *fl,
25ee3286
HX		 1543 struct dst_entry *dst)
1544{
d7c7544c 1545 struct net *net = xp_net(policy);
25ee3286
HX		 1546 unsigned long now = jiffies;
1547 struct net_device *dev;
43a4dea4 1548 struct xfrm_mode *inner_mode;
25ee3286
HX		 1549 struct dst_entry *dst_prev = NULL;
1550 struct dst_entry *dst0 = NULL;
1551 int i = 0;
1552 int err;
1553 int header_len = 0;
a1b05140 1554 int nfheader_len = 0;
25ee3286
HX		 1555 int trailer_len = 0;
1556 int tos;
1557 int family = policy->selector.family;
9bb182a7
YH		 1558 xfrm_address_t saddr, daddr;
1559
1560 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1561
1562 tos = xfrm_get_tos(fl, family);
25ee3286
HX		 1563
1564 dst_hold(dst);
1565
1566 for (; i < nx; i++) {
d7c7544c 1567 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1568 struct dst_entry *dst1 = &xdst->u.dst;
1569
1570 err = PTR_ERR(xdst);
1571 if (IS_ERR(xdst)) {
1572 dst_release(dst);
1573 goto put_states;
1574 }
1575
43a4dea4
SK		 1576 if (xfrm[i]->sel.family == AF_UNSPEC) {
1577 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1578 xfrm_af2proto(family));
1579 if (!inner_mode) {
1580 err = -EAFNOSUPPORT;
1581 dst_release(dst);
1582 goto put_states;
1583 }
1584 } else
1585 inner_mode = xfrm[i]->inner_mode;
1586
25ee3286
HX		 1587 if (!dst_prev)
1588 dst0 = dst1;
52df157f
WW		 1589 else
1590 /* Ref count is taken during xfrm_alloc_dst()
1591 * No need to do dst_clone() on dst1
1592 */
1593 dst_prev->child = dst1;
25ee3286
HX		 1594
1595 xdst->route = dst;
defb3519 1596 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1597
1598 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1599 family = xfrm[i]->props.family;
42a7b32b 1600 dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
077fbac4
LC		 1601 &saddr, &daddr, family,
1602 xfrm[i]->props.output_mark);
25ee3286
HX		 1603 err = PTR_ERR(dst);
1604 if (IS_ERR(dst))
1605 goto put_states;
1606 } else
1607 dst_hold(dst);
1608
1609 dst1->xfrm = xfrm[i];
80c802f3 1610 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1611
f5b0a874 1612 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1613 dst1->flags |= DST_HOST;
1614 dst1->lastuse = now;
1615
1616 dst1->input = dst_discard;
43a4dea4 1617 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1618
1619 dst1->next = dst_prev;
1620 dst_prev = dst1;
1621
1622 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1623 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1624 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1625 trailer_len += xfrm[i]->props.trailer_len;
1626 }
1627
1628 dst_prev->child = dst;
1629 dst0->path = dst;
1630
1631 err = -ENODEV;
1632 dev = dst->dev;
1633 if (!dev)
1634 goto free_dst;
1635
a1b05140 1636 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1637 xfrm_init_pmtu(dst_prev);
1638
1639 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1640 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1641
87c1e12b 1642 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1643 if (err)
1644 goto free_dst;
1645
1646 dst_prev->header_len = header_len;
1647 dst_prev->trailer_len = trailer_len;
1648 header_len -= xdst->u.dst.xfrm->props.header_len;
1649 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1650 }
1651
1652out:
1653 return dst0;
1654
1655put_states:
1656 for (; i < nx; i++)
1657 xfrm_state_put(xfrm[i]);
1658free_dst:
1659 if (dst0)
52df157f 1660 dst_release_immediate(dst0);
25ee3286
HX		 1661 dst0 = ERR_PTR(err);
1662 goto out;
1663}
1664
73ff93cd 1665static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1666 struct xfrm_policy **pols,
1667 int *num_pols, int *num_xfrms)
1668{
1669 int i;
1670
1671 if (*num_pols == 0 || !pols[0]) {
1672 *num_pols = 0;
1673 *num_xfrms = 0;
1674 return 0;
1675 }
1676 if (IS_ERR(pols[0]))
1677 return PTR_ERR(pols[0]);
1678
1679 *num_xfrms = pols[0]->xfrm_nr;
1680
1681#ifdef CONFIG_XFRM_SUB_POLICY
1682 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1683 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1684 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1685 XFRM_POLICY_TYPE_MAIN,
1686 fl, family,
1687 XFRM_POLICY_OUT);
1688 if (pols[1]) {
1689 if (IS_ERR(pols[1])) {
1690 xfrm_pols_put(pols, *num_pols);
1691 return PTR_ERR(pols[1]);
1692 }
02d0892f 1693 (*num_pols)++;
80c802f3
TT		 1694 (*num_xfrms) += pols[1]->xfrm_nr;
1695 }
1696 }
1697#endif
1698 for (i = 0; i < *num_pols; i++) {
1699 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1700 *num_xfrms = -1;
1701 break;
1702 }
1703 }
1704
1705 return 0;
1706
1707}
1708
ec30d78c
FW		 1709static void xfrm_last_dst_update(struct xfrm_dst *xdst, struct xfrm_dst *old)
1710{
1711 this_cpu_write(xfrm_last_dst, xdst);
1712 if (old)
1713 dst_release(&old->u.dst);
1714}
1715
1716static void __xfrm_pcpu_work_fn(void)
1717{
1718 struct xfrm_dst *old;
1719
1720 old = this_cpu_read(xfrm_last_dst);
1721 if (old && !xfrm_bundle_ok(old))
1722 xfrm_last_dst_update(NULL, old);
1723}
1724
1725static void xfrm_pcpu_work_fn(struct work_struct *work)
1726{
1727 local_bh_disable();
1728 rcu_read_lock();
1729 __xfrm_pcpu_work_fn();
1730 rcu_read_unlock();
1731 local_bh_enable();
1732}
1733
1734void xfrm_policy_cache_flush(void)
1735{
1736 struct xfrm_dst *old;
1737 bool found = 0;
1738 int cpu;
1739
1740 local_bh_disable();
1741 rcu_read_lock();
1742 for_each_possible_cpu(cpu) {
1743 old = per_cpu(xfrm_last_dst, cpu);
1744 if (old && !xfrm_bundle_ok(old)) {
1745 if (smp_processor_id() == cpu) {
1746 __xfrm_pcpu_work_fn();
1747 continue;
1748 }
1749 found = true;
1750 break;
1751 }
1752 }
1753
1754 rcu_read_unlock();
1755 local_bh_enable();
1756
1757 if (!found)
1758 return;
1759
1760 get_online_cpus();
1761
1762 for_each_possible_cpu(cpu) {
1763 bool bundle_release;
1764
1765 rcu_read_lock();
1766 old = per_cpu(xfrm_last_dst, cpu);
1767 bundle_release = old && !xfrm_bundle_ok(old);
1768 rcu_read_unlock();
1769
1770 if (!bundle_release)
1771 continue;
1772
1773 if (cpu_online(cpu)) {
1774 schedule_work_on(cpu, &xfrm_pcpu_work[cpu]);
1775 continue;
1776 }
1777
1778 rcu_read_lock();
1779 old = per_cpu(xfrm_last_dst, cpu);
1780 if (old && !xfrm_bundle_ok(old)) {
1781 per_cpu(xfrm_last_dst, cpu) = NULL;
1782 dst_release(&old->u.dst);
1783 }
1784 rcu_read_unlock();
1785 }
1786
1787 put_online_cpus();
1788}
1789
1790static bool xfrm_pol_dead(struct xfrm_dst *xdst)
1791{
1792 unsigned int num_pols = xdst->num_pols;
1793 unsigned int pol_dead = 0, i;
1794
1795 for (i = 0; i < num_pols; i++)
1796 pol_dead |= xdst->pols[i]->walk.dead;
1797
1798 /* Mark DST_OBSOLETE_DEAD to fail the next xfrm_dst_check() */
1799 if (pol_dead)
1800 xdst->u.dst.obsolete = DST_OBSOLETE_DEAD;
1801
1802 return pol_dead;
1803}
1804
80c802f3
TT		 1805static struct xfrm_dst *
1806xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1807 const struct flowi *fl, u16 family,
80c802f3
TT		 1808 struct dst_entry *dst_orig)
1809{
1810 struct net *net = xp_net(pols[0]);
1811 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
ec30d78c 1812 struct xfrm_dst *xdst, *old;
80c802f3 1813 struct dst_entry *dst;
80c802f3
TT		 1814 int err;
1815
ec30d78c
FW		 1816 xdst = this_cpu_read(xfrm_last_dst);
1817 if (xdst &&
1818 xdst->u.dst.dev == dst_orig->dev &&
1819 xdst->num_pols == num_pols &&
1820 !xfrm_pol_dead(xdst) &&
1821 memcmp(xdst->pols, pols,
13ead5c4
FW		 1822 sizeof(struct xfrm_policy *) * num_pols) == 0 &&
1823 xfrm_bundle_ok(xdst)) {
ec30d78c
FW		 1824 dst_hold(&xdst->u.dst);
1825 return xdst;
1826 }
1827
1828 old = xdst;
80c802f3
TT		 1829 /* Try to instantiate a bundle */
1830 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1831 if (err <= 0) {
1832 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1833 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1834 return ERR_PTR(err);
1835 }
1836
1837 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1838 if (IS_ERR(dst)) {
1839 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1840 return ERR_CAST(dst);
1841 }
1842
1843 xdst = (struct xfrm_dst *)dst;
1844 xdst->num_xfrms = err;
80c802f3 1845 xdst->num_pols = num_pols;
3e94c2dc 1846 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1847 xdst->policy_genid = atomic_read(&pols[0]->genid);
1848
ec30d78c
FW		 1849 atomic_set(&xdst->u.dst.__refcnt, 2);
1850 xfrm_last_dst_update(xdst, old);
1851
80c802f3
TT		 1852 return xdst;
1853}
1854
a0073fe1
SK		 1855static void xfrm_policy_queue_process(unsigned long arg)
1856{
a0073fe1
SK		 1857 struct sk_buff *skb;
1858 struct sock *sk;
1859 struct dst_entry *dst;
a0073fe1 1860 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
3f5312ae 1861 struct net *net = xp_net(pol);
a0073fe1
SK		 1862 struct xfrm_policy_queue *pq = &pol->polq;
1863 struct flowi fl;
1864 struct sk_buff_head list;
1865
1866 spin_lock(&pq->hold_queue.lock);
1867 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1868 if (!skb) {
1869 spin_unlock(&pq->hold_queue.lock);
1870 goto out;
1871 }
a0073fe1
SK		 1872 dst = skb_dst(skb);
1873 sk = skb->sk;
1874 xfrm_decode_session(skb, &fl, dst->ops->family);
1875 spin_unlock(&pq->hold_queue.lock);
1876
1877 dst_hold(dst->path);
3f5312ae 1878 dst = xfrm_lookup(net, dst->path, &fl, sk, 0);
a0073fe1
SK		 1879 if (IS_ERR(dst))
1880 goto purge_queue;
1881
1882 if (dst->flags & DST_XFRM_QUEUE) {
1883 dst_release(dst);
1884
1885 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1886 goto purge_queue;
1887
1888 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1889 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1890 xfrm_pol_hold(pol);
1891 goto out;
a0073fe1
SK		 1892 }
1893
1894 dst_release(dst);
1895
1896 __skb_queue_head_init(&list);
1897
1898 spin_lock(&pq->hold_queue.lock);
1899 pq->timeout = 0;
1900 skb_queue_splice_init(&pq->hold_queue, &list);
1901 spin_unlock(&pq->hold_queue.lock);
1902
1903 while (!skb_queue_empty(&list)) {
1904 skb = __skb_dequeue(&list);
1905
1906 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1907 dst_hold(skb_dst(skb)->path);
3f5312ae 1908 dst = xfrm_lookup(net, skb_dst(skb)->path, &fl, skb->sk, 0);
a0073fe1 1909 if (IS_ERR(dst)) {
a0073fe1
SK		 1910 kfree_skb(skb);
1911 continue;
1912 }
1913
1914 nf_reset(skb);
1915 skb_dst_drop(skb);
1916 skb_dst_set(skb, dst);
1917
13206b6b 1918 dst_output(net, skb->sk, skb);
a0073fe1
SK		 1919 }
1920
e7d8f6cb
SK		 1921out:
1922 xfrm_pol_put(pol);
a0073fe1
SK		 1923 return;
1924
1925purge_queue:
1926 pq->timeout = 0;
1ee5e667 1927 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1928 xfrm_pol_put(pol);
a0073fe1
SK		 1929}
1930
ede2059d 1931static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1932{
1933 unsigned long sched_next;
1934 struct dst_entry *dst = skb_dst(skb);
1935 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1936 struct xfrm_policy *pol = xdst->pols[0];
1937 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1938
39bb5e62 1939 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK		 1940 kfree_skb(skb);
1941 return 0;
1942 }
a0073fe1
SK		 1943
1944 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1945 kfree_skb(skb);
1946 return -EAGAIN;
1947 }
1948
1949 skb_dst_force(skb);
a0073fe1
SK		 1950
1951 spin_lock_bh(&pq->hold_queue.lock);
1952
1953 if (!pq->timeout)
1954 pq->timeout = XFRM_QUEUE_TMO_MIN;
1955
1956 sched_next = jiffies + pq->timeout;
1957
1958 if (del_timer(&pq->hold_timer)) {
1959 if (time_before(pq->hold_timer.expires, sched_next))
1960 sched_next = pq->hold_timer.expires;
e7d8f6cb 1961 xfrm_pol_put(pol);
a0073fe1
SK		 1962 }
1963
1964 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1965 if (!mod_timer(&pq->hold_timer, sched_next))
1966 xfrm_pol_hold(pol);
a0073fe1
SK		 1967
1968 spin_unlock_bh(&pq->hold_queue.lock);
1969
1970 return 0;
1971}
1972
1973static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 1974 struct xfrm_flo *xflo,
a0073fe1
SK		 1975 const struct flowi *fl,
1976 int num_xfrms,
1977 u16 family)
1978{
1979 int err;
1980 struct net_device *dev;
b8c203b2 1981 struct dst_entry *dst;
a0073fe1
SK		 1982 struct dst_entry *dst1;
1983 struct xfrm_dst *xdst;
1984
1985 xdst = xfrm_alloc_dst(net, family);
1986 if (IS_ERR(xdst))
1987 return xdst;
1988
b8c203b2
SK		 1989 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
1990 net->xfrm.sysctl_larval_drop ||
1991 num_xfrms <= 0)
a0073fe1
SK		 1992 return xdst;
1993
b8c203b2 1994 dst = xflo->dst_orig;
a0073fe1
SK		 1995 dst1 = &xdst->u.dst;
1996 dst_hold(dst);
1997 xdst->route = dst;
1998
1999 dst_copy_metrics(dst1, dst);
2000
2001 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2002 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2003 dst1->lastuse = jiffies;
2004
2005 dst1->input = dst_discard;
2006 dst1->output = xdst_queue_output;
2007
2008 dst_hold(dst);
2009 dst1->child = dst;
2010 dst1->path = dst;
2011
2012 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2013
2014 err = -ENODEV;
2015 dev = dst->dev;
2016 if (!dev)
2017 goto free_dst;
2018
2019 err = xfrm_fill_dst(xdst, dev, fl);
2020 if (err)
2021 goto free_dst;
2022
2023out:
2024 return xdst;
2025
2026free_dst:
2027 dst_release(dst1);
2028 xdst = ERR_PTR(err);
2029 goto out;
2030}
2031
bd45c539 2032static struct xfrm_dst *
3ca28286 2033xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir, struct xfrm_flo *xflo)
80c802f3 2034{
80c802f3 2035 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
855dad99 2036 int num_pols = 0, num_xfrms = 0, err;
bd45c539 2037 struct xfrm_dst *xdst;
80c802f3 2038
80c802f3
TT		 2039 /* Resolve policies to use if we couldn't get them from
2040 * previous cache entry */
855dad99 2041 num_pols = 1;
86dc8ee0 2042 pols[0] = xfrm_policy_lookup(net, fl, family, dir);
855dad99 2043 err = xfrm_expand_policies(fl, family, pols,
80c802f3 2044 &num_pols, &num_xfrms);
855dad99
FW		 2045 if (err < 0)
2046 goto inc_error;
2047 if (num_pols == 0)
2048 return NULL;
2049 if (num_xfrms <= 0)
2050 goto make_dummy_bundle;
80c802f3 2051
bd45c539 2052 xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
b8c203b2 2053 xflo->dst_orig);
bd45c539
FW		 2054 if (IS_ERR(xdst)) {
2055 err = PTR_ERR(xdst);
80c802f3
TT		 2056 if (err != -EAGAIN)
2057 goto error;
855dad99 2058 goto make_dummy_bundle;
bd45c539 2059 } else if (xdst == NULL) {
d809ec89 2060 num_xfrms = 0;
855dad99 2061 goto make_dummy_bundle;
80c802f3
TT		 2062 }
2063
bd45c539 2064 return xdst;
80c802f3
TT		 2065
2066make_dummy_bundle:
2067 /* We found policies, but there's no bundles to instantiate:
2068 * either because the policy blocks, has no transformations or
2069 * we could not build template (no xfrm_states).*/
b8c203b2 2070 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2071 if (IS_ERR(xdst)) {
2072 xfrm_pols_put(pols, num_pols);
2073 return ERR_CAST(xdst);
2074 }
2075 xdst->num_pols = num_pols;
2076 xdst->num_xfrms = num_xfrms;
3e94c2dc 2077 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2078
2079 dst_hold(&xdst->u.dst);
bd45c539 2080 return xdst;
80c802f3
TT		 2081
2082inc_error:
2083 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2084error:
855dad99 2085 xfrm_pols_put(pols, num_pols);
80c802f3
TT		 2086 return ERR_PTR(err);
2087}
1da177e4 2088
2774c131
DM		 2089static struct dst_entry *make_blackhole(struct net *net, u16 family,
2090 struct dst_entry *dst_orig)
2091{
37b10383 2092 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2774c131
DM		 2093 struct dst_entry *ret;
2094
2095 if (!afinfo) {
2096 dst_release(dst_orig);
433a1954 2097 return ERR_PTR(-EINVAL);
2774c131
DM		 2098 } else {
2099 ret = afinfo->blackhole_route(net, dst_orig);
2100 }
bdba9fe0 2101 rcu_read_unlock();
2774c131
DM		 2102
2103 return ret;
2104}
2105
1da177e4
LT		 2106/* Main function: finds/creates a bundle for given flow.
2107 *
2108 * At the moment we eat a raw IP route. Mostly to speed up lookups
2109 * on interfaces with disabled IPsec.
2110 */
452edd59
DM		 2111struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2112 const struct flowi *fl,
6f9c9615 2113 const struct sock *sk, int flags)
1da177e4 2114{
4e81bb83 2115 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3 2116 struct xfrm_dst *xdst;
452edd59 2117 struct dst_entry *dst, *route;
80c802f3 2118 u16 family = dst_orig->ops->family;
aff669bc 2119 u8 dir = XFRM_POLICY_OUT;
4b021628 2120 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2121
80c802f3
TT		 2122 dst = NULL;
2123 xdst = NULL;
2124 route = NULL;
4e81bb83 2125
bd5eb35f 2126 sk = sk_const_to_full_sk(sk);
f7944fb1 2127 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3 2128 num_pols = 1;
4c86d777 2129 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, family);
80c802f3
TT		 2130 err = xfrm_expand_policies(fl, family, pols,
2131 &num_pols, &num_xfrms);
2132 if (err < 0)
75b8c133 2133 goto dropdst;
80c802f3
TT		 2134
2135 if (num_pols) {
2136 if (num_xfrms <= 0) {
2137 drop_pols = num_pols;
2138 goto no_transform;
2139 }
2140
2141 xdst = xfrm_resolve_and_create_bundle(
2142 pols, num_pols, fl,
2143 family, dst_orig);
2144 if (IS_ERR(xdst)) {
2145 xfrm_pols_put(pols, num_pols);
2146 err = PTR_ERR(xdst);
2147 goto dropdst;
d809ec89
TT		 2148 } else if (xdst == NULL) {
2149 num_xfrms = 0;
2150 drop_pols = num_pols;
2151 goto no_transform;
80c802f3
TT		 2152 }
2153
80c802f3 2154 route = xdst->route;
0aa64774 2155 }
3bccfbc7 2156 }
1da177e4 2157
80c802f3 2158 if (xdst == NULL) {
b8c203b2
SK		 2159 struct xfrm_flo xflo;
2160
2161 xflo.dst_orig = dst_orig;
2162 xflo.flags = flags;
2163
1da177e4 2164 /* To accelerate a bit... */
2518c7c2 2165 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2166 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2167 goto nopol;
1da177e4 2168
bd45c539
FW		 2169 xdst = xfrm_bundle_lookup(net, fl, family, dir, &xflo);
2170 if (xdst == NULL)
80c802f3 2171 goto nopol;
bd45c539
FW		 2172 if (IS_ERR(xdst)) {
2173 err = PTR_ERR(xdst);
75b8c133 2174 goto dropdst;
d66e37a9 2175 }
80c802f3
TT		 2176
2177 num_pols = xdst->num_pols;
2178 num_xfrms = xdst->num_xfrms;
3e94c2dc 2179 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2180 route = xdst->route;
2181 }
2182
2183 dst = &xdst->u.dst;
2184 if (route == NULL && num_xfrms > 0) {
2185 /* The only case when xfrm_bundle_lookup() returns a
2186 * bundle with null route, is when the template could
2187 * not be resolved. It means policies are there, but
2188 * bundle could not be created, since we don't yet
2189 * have the xfrm_state's. We need to wait for KM to
2190 * negotiate new SA's or bail out with error.*/
2191 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2192 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W		 2193 err = -EREMOTE;
2194 goto error;
80c802f3 2195 }
80c802f3 2196
5b8ef341 2197 err = -EAGAIN;
80c802f3
TT		 2198
2199 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2200 goto error;
1da177e4
LT		 2201 }
2202
80c802f3
TT		 2203no_transform:
2204 if (num_pols == 0)
8b7817f3 2205 goto nopol;
1da177e4 2206
80c802f3
TT		 2207 if ((flags & XFRM_LOOKUP_ICMP) &&
2208 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2209 err = -ENOENT;
8b7817f3 2210 goto error;
80c802f3 2211 }
8b7817f3 2212
80c802f3
TT		 2213 for (i = 0; i < num_pols; i++)
2214 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2215
80c802f3 2216 if (num_xfrms < 0) {
1da177e4 2217 /* Prohibit the flow */
59c9940e 2218 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2219 err = -EPERM;
2220 goto error;
80c802f3
TT		 2221 } else if (num_xfrms > 0) {
2222 /* Flow transformed */
80c802f3
TT		 2223 dst_release(dst_orig);
2224 } else {
2225 /* Flow passes untransformed */
2226 dst_release(dst);
452edd59 2227 dst = dst_orig;
1da177e4 2228 }
80c802f3
TT		 2229ok:
2230 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2231 if (dst && dst->xfrm &&
2232 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2233 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2234 return dst;
1da177e4 2235
80c802f3 2236nopol:
452edd59
DM		 2237 if (!(flags & XFRM_LOOKUP_ICMP)) {
2238 dst = dst_orig;
80c802f3 2239 goto ok;
452edd59 2240 }
80c802f3 2241 err = -ENOENT;
1da177e4 2242error:
80c802f3 2243 dst_release(dst);
75b8c133 2244dropdst:
ac37e251
W		 2245 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2246 dst_release(dst_orig);
80c802f3 2247 xfrm_pols_put(pols, drop_pols);
452edd59 2248 return ERR_PTR(err);
1da177e4
LT		 2249}
2250EXPORT_SYMBOL(xfrm_lookup);
2251
f92ee619
SK		 2252/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2253 * Otherwise we may send out blackholed packets.
2254 */
2255struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2256 const struct flowi *fl,
6f9c9615 2257 const struct sock *sk, int flags)
f92ee619 2258{
b8c203b2 2259 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W		 2260 flags | XFRM_LOOKUP_QUEUE |
2261 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK		 2262
2263 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2264 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2265
2266 return dst;
2267}
2268EXPORT_SYMBOL(xfrm_lookup_route);
2269
df0ba92a 2270static inline int
8f029de2 2271xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2272{
2273 struct xfrm_state *x;
df0ba92a
MN		 2274
2275 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2276 return 0;
2277 x = skb->sp->xvec[idx];
2278 if (!x->type->reject)
2279 return 0;
1ecafede 2280 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2281}
2282
1da177e4
LT		 2283/* When skb is transformed back to its "native" form, we have to
2284 * check policy restrictions. At the moment we make this in maximally
2285 * stupid way. Shame on me. :-) Of course, connected sockets must
2286 * have policy cached at them.
2287 */
2288
2289static inline int
7db454b9 2290xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2291 unsigned short family)
2292{
2293 if (xfrm_state_kern(x))
928ba416 2294 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2295 return x->id.proto == tmpl->id.proto &&
2296 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2297 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2298 x->props.mode == tmpl->mode &&
c5d18e98 2299 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2300 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2301 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2302 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2303}
2304
df0ba92a
MN		 2305/*
2306 * 0 or more than 0 is returned when validation is succeeded (either bypass
2307 * because of optional transport mode, or next index of the mathced secpath
2308 * state with the template.
2309 * -1 is returned when no matching template is found.
2310 * Otherwise "-2 - errored_index" is returned.
2311 */
1da177e4 2312static inline int
22cccb7e 2313xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2314 unsigned short family)
2315{
2316 int idx = start;
2317
2318 if (tmpl->optional) {
7e49e6de 2319 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2320 return start;
2321 } else
2322 start = -1;
2323 for (; idx < sp->len; idx++) {
dbe5b4aa 2324 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2325 return ++idx;
df0ba92a
MN		 2326 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2327 if (start == -1)
2328 start = -2-idx;
1da177e4 2329 break;
df0ba92a 2330 }
1da177e4
LT		 2331 }
2332 return start;
2333}
2334
d5422efe
HX		 2335int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2336 unsigned int family, int reverse)
1da177e4 2337{
37b10383 2338 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2339 int err;
1da177e4
LT		 2340
2341 if (unlikely(afinfo == NULL))
2342 return -EAFNOSUPPORT;
2343
d5422efe 2344 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2345 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
bdba9fe0 2346 rcu_read_unlock();
e0d1caa7 2347 return err;
1da177e4 2348}
d5422efe 2349EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2350
9a7386ec 2351static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2352{
2353 for (; k < sp->len; k++) {
df0ba92a 2354 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2355 *idxp = k;
1da177e4 2356 return 1;
df0ba92a 2357 }
1da177e4
LT		 2358 }
2359
2360 return 0;
2361}
2362
a716c119 2363int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2364 unsigned short family)
2365{
f6e1e25d 2366 struct net *net = dev_net(skb->dev);
1da177e4 2367 struct xfrm_policy *pol;
4e81bb83
MN		 2368 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2369 int npols = 0;
2370 int xfrm_nr;
2371 int pi;
d5422efe 2372 int reverse;
1da177e4 2373 struct flowi fl;
df0ba92a 2374 int xerr_idx = -1;
1da177e4 2375
d5422efe
HX		 2376 reverse = dir & ~XFRM_POLICY_MASK;
2377 dir &= XFRM_POLICY_MASK;
d5422efe 2378
0aa64774 2379 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2380 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2381 return 0;
0aa64774
MN		 2382 }
2383
eb9c7ebe 2384 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2385
2386 /* First, check used SA against their selectors. */
2387 if (skb->sp) {
2388 int i;
2389
9b7a787d 2390 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2391 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2392 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2393 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2394 return 0;
0aa64774 2395 }
1da177e4
LT		 2396 }
2397 }
2398
2399 pol = NULL;
bd5eb35f 2400 sk = sk_to_full_sk(sk);
3bccfbc7 2401 if (sk && sk->sk_policy[dir]) {
4c86d777 2402 pol = xfrm_sk_policy_lookup(sk, dir, &fl, family);
0aa64774 2403 if (IS_ERR(pol)) {
59c9940e 2404 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2405 return 0;
0aa64774 2406 }
3bccfbc7 2407 }
1da177e4 2408
86dc8ee0
FW		 2409 if (!pol)
2410 pol = xfrm_policy_lookup(net, &fl, family, dir);
1da177e4 2411
0aa64774 2412 if (IS_ERR(pol)) {
59c9940e 2413 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2414 return 0;
0aa64774 2415 }
134b0fc5 2416
df0ba92a 2417 if (!pol) {
d1d9facf 2418 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2419 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2420 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2421 return 0;
2422 }
2423 return 1;
2424 }
1da177e4 2425
9d729f72 2426 pol->curlft.use_time = get_seconds();
1da177e4 2427
4e81bb83 2428 pols[0] = pol;
02d0892f 2429 npols++;
4e81bb83
MN		 2430#ifdef CONFIG_XFRM_SUB_POLICY
2431 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2432 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2433 &fl, family,
2434 XFRM_POLICY_IN);
2435 if (pols[1]) {
0aa64774 2436 if (IS_ERR(pols[1])) {
59c9940e 2437 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2438 return 0;
0aa64774 2439 }
9d729f72 2440 pols[1]->curlft.use_time = get_seconds();
02d0892f 2441 npols++;
4e81bb83
MN		 2442 }
2443 }
2444#endif
2445
1da177e4
LT		 2446 if (pol->action == XFRM_POLICY_ALLOW) {
2447 struct sec_path *sp;
2448 static struct sec_path dummy;
4e81bb83 2449 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2450 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2451 struct xfrm_tmpl **tpp = tp;
2452 int ti = 0;
1da177e4
LT		 2453 int i, k;
2454
2455 if ((sp = skb->sp) == NULL)
2456 sp = &dummy;
2457
4e81bb83
MN		 2458 for (pi = 0; pi < npols; pi++) {
2459 if (pols[pi] != pol &&
0aa64774 2460 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2461 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2462 goto reject;
0aa64774
MN		 2463 }
2464 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2465 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2466 goto reject_error;
0aa64774 2467 }
4e81bb83
MN		 2468 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2469 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2470 }
2471 xfrm_nr = ti;
41a49cc3 2472 if (npols > 1) {
283bc9f3 2473 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2474 tpp = stp;
2475 }
4e81bb83 2476
1da177e4
LT		 2477 /* For each tunnel xfrm, find the first matching tmpl.
2478 * For each tmpl before that, find corresponding xfrm.
2479 * Order is _important_. Later we will implement
2480 * some barriers, but at the moment barriers
2481 * are implied between each two transformations.
2482 */
4e81bb83
MN		 2483 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2484 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2485 if (k < 0) {
d1d9facf
JM		 2486 if (k < -1)
2487 /* "-2 - errored_index" returned */
2488 xerr_idx = -(2+k);
59c9940e 2489 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2490 goto reject;
df0ba92a 2491 }
1da177e4
LT		 2492 }
2493
0aa64774 2494 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2495 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2496 goto reject;
0aa64774 2497 }
1da177e4 2498
4e81bb83 2499 xfrm_pols_put(pols, npols);
1da177e4
LT		 2500 return 1;
2501 }
59c9940e 2502 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2503
2504reject:
df0ba92a 2505 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2506reject_error:
2507 xfrm_pols_put(pols, npols);
1da177e4
LT		 2508 return 0;
2509}
2510EXPORT_SYMBOL(__xfrm_policy_check);
2511
2512int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2513{
99a66657 2514 struct net *net = dev_net(skb->dev);
1da177e4 2515 struct flowi fl;
adf30907 2516 struct dst_entry *dst;
73137147 2517 int res = 1;
1da177e4 2518
0aa64774 2519 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2520 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2521 return 0;
0aa64774 2522 }
1da177e4 2523
fafeeb6c 2524 skb_dst_force(skb);
adf30907 2525
b8c203b2 2526 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2527 if (IS_ERR(dst)) {
73137147 2528 res = 0;
452edd59
DM		 2529 dst = NULL;
2530 }
adf30907
ED		 2531 skb_dst_set(skb, dst);
2532 return res;
1da177e4
LT		 2533}
2534EXPORT_SYMBOL(__xfrm_route_forward);
2535
d49c73c7
DM		 2536/* Optimize later using cookies and generation ids. */
2537
1da177e4
LT		 2538static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2539{
d49c73c7 2540 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2541 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2542 * get validated by dst_ops->check on every use. We do this
2543 * because when a normal route referenced by an XFRM dst is
2544 * obsoleted we do not go looking around for all parent
2545 * referencing XFRM dsts so that we can invalidate them. It
2546 * is just too much work. Instead we make the checks here on
2547 * every use. For example:
d49c73c7
DM		 2548 *
2549 * XFRM dst A --> IPv4 dst X
2550 *
2551 * X is the "xdst->route" of A (X is also the "dst->path" of A
2552 * in this example). If X is marked obsolete, "A" will not
2553 * notice. That's what we are validating here via the
2554 * stale_bundle() check.
2555 *
52df157f
WW		 2556 * When a dst is removed from the fib tree, DST_OBSOLETE_DEAD will
2557 * be marked on it.
09c75704 2558 * This will force stale_bundle() to fail on any xdst bundle with
52df157f 2559 * this dst linked in it.
399c180a 2560 */
d49c73c7
DM		 2561 if (dst->obsolete < 0 && !stale_bundle(dst))
2562 return dst;
2563
1da177e4
LT		 2564 return NULL;
2565}
2566
2567static int stale_bundle(struct dst_entry *dst)
2568{
12fdb4d3 2569 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2570}
2571
aabc9761 2572void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2573{
1da177e4 2574 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2575 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2576 dev_hold(dst->dev);
1da177e4
LT		 2577 dev_put(dev);
2578 }
2579}
aabc9761 2580EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2581
2582static void xfrm_link_failure(struct sk_buff *skb)
2583{
2584 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2585}
2586
2587static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2588{
2589 if (dst) {
2590 if (dst->obsolete) {
2591 dst_release(dst);
2592 dst = NULL;
2593 }
2594 }
2595 return dst;
2596}
2597
25ee3286 2598static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2599{
2600 do {
2601 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2602 u32 pmtu, route_mtu_cached;
2603
2604 pmtu = dst_mtu(dst->child);
2605 xdst->child_mtu_cached = pmtu;
2606
2607 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2608
2609 route_mtu_cached = dst_mtu(xdst->route);
2610 xdst->route_mtu_cached = route_mtu_cached;
2611
2612 if (pmtu > route_mtu_cached)
2613 pmtu = route_mtu_cached;
2614
defb3519 2615 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2616 } while ((dst = dst->next));
2617}
2618
1da177e4
LT		 2619/* Check that the bundle accepts the flow and its components are
2620 * still valid.
2621 */
2622
12fdb4d3 2623static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2624{
2625 struct dst_entry *dst = &first->u.dst;
2626 struct xfrm_dst *last;
2627 u32 mtu;
2628
92d63dec 2629 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2630 (dst->dev && !netif_running(dst->dev)))
2631 return 0;
2632
a0073fe1
SK		 2633 if (dst->flags & DST_XFRM_QUEUE)
2634 return 1;
2635
1da177e4
LT		 2636 last = NULL;
2637
2638 do {
2639 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2640
1da177e4
LT		 2641 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2642 return 0;
80c802f3
TT		 2643 if (xdst->xfrm_genid != dst->xfrm->genid)
2644 return 0;
b1312c89
TT		 2645 if (xdst->num_pols > 0 &&
2646 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2647 return 0;
e53820de 2648
1da177e4
LT		 2649 mtu = dst_mtu(dst->child);
2650 if (xdst->child_mtu_cached != mtu) {
2651 last = xdst;
2652 xdst->child_mtu_cached = mtu;
2653 }
2654
92d63dec 2655 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2656 return 0;
2657 mtu = dst_mtu(xdst->route);
2658 if (xdst->route_mtu_cached != mtu) {
2659 last = xdst;
2660 xdst->route_mtu_cached = mtu;
2661 }
2662
2663 dst = dst->child;
2664 } while (dst->xfrm);
2665
2666 if (likely(!last))
2667 return 1;
2668
2669 mtu = last->child_mtu_cached;
2670 for (;;) {
2671 dst = &last->u.dst;
2672
2673 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2674 if (mtu > last->route_mtu_cached)
2675 mtu = last->route_mtu_cached;
defb3519 2676 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2677
2678 if (last == first)
2679 break;
2680
bd0bf076 2681 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2682 last->child_mtu_cached = mtu;
2683 }
2684
2685 return 1;
2686}
2687
0dbaee3b
DM		 2688static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2689{
2690 return dst_metric_advmss(dst->path);
2691}
2692
ebb762f2 2693static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2694{
618f9bc7
SK		 2695 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2696
2697 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2698}
2699
1ecc9ad0
JA		 2700static const void *xfrm_get_dst_nexthop(const struct dst_entry *dst,
2701 const void *daddr)
63fca65d
JA		 2702{
2703 const struct dst_entry *path = dst->path;
2704
2705 for (; dst != path; dst = dst->child) {
2706 const struct xfrm_state *xfrm = dst->xfrm;
2707
2708 if (xfrm->props.mode == XFRM_MODE_TRANSPORT)
2709 continue;
2710 if (xfrm->type->flags & XFRM_TYPE_REMOTE_COADDR)
2711 daddr = xfrm->coaddr;
2712 else if (!(xfrm->type->flags & XFRM_TYPE_LOCAL_COADDR))
2713 daddr = &xfrm->id.daddr;
2714 }
1ecc9ad0
JA		 2715 return daddr;
2716}
2717
2718static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2719 struct sk_buff *skb,
2720 const void *daddr)
2721{
2722 const struct dst_entry *path = dst->path;
2723
2724 if (!skb)
2725 daddr = xfrm_get_dst_nexthop(dst, daddr);
2726 return path->ops->neigh_lookup(path, skb, daddr);
2727}
2728
2729static void xfrm_confirm_neigh(const struct dst_entry *dst, const void *daddr)
2730{
2731 const struct dst_entry *path = dst->path;
2732
2733 daddr = xfrm_get_dst_nexthop(dst, daddr);
63fca65d
JA		 2734 path->ops->confirm_neigh(path, daddr);
2735}
2736
a2817d8b 2737int xfrm_policy_register_afinfo(const struct xfrm_policy_afinfo *afinfo, int family)
1da177e4
LT		 2738{
2739 int err = 0;
a2817d8b
FW		 2740
2741 if (WARN_ON(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
1da177e4 2742 return -EAFNOSUPPORT;
a2817d8b 2743
ef8531b6 2744 spin_lock(&xfrm_policy_afinfo_lock);
a2817d8b 2745 if (unlikely(xfrm_policy_afinfo[family] != NULL))
f31e8d4f 2746 err = -EEXIST;
1da177e4
LT		 2747 else {
2748 struct dst_ops *dst_ops = afinfo->dst_ops;
2749 if (likely(dst_ops->kmem_cachep == NULL))
2750 dst_ops->kmem_cachep = xfrm_dst_cache;
2751 if (likely(dst_ops->check == NULL))
2752 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2753 if (likely(dst_ops->default_advmss == NULL))
2754 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2755 if (likely(dst_ops->mtu == NULL))
2756 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2757 if (likely(dst_ops->negative_advice == NULL))
2758 dst_ops->negative_advice = xfrm_negative_advice;
2759 if (likely(dst_ops->link_failure == NULL))
2760 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2761 if (likely(dst_ops->neigh_lookup == NULL))
2762 dst_ops->neigh_lookup = xfrm_neigh_lookup;
63fca65d
JA		 2763 if (likely(!dst_ops->confirm_neigh))
2764 dst_ops->confirm_neigh = xfrm_confirm_neigh;
a2817d8b 2765 rcu_assign_pointer(xfrm_policy_afinfo[family], afinfo);
1da177e4 2766 }
ef8531b6 2767 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c 2768
1da177e4
LT		 2769 return err;
2770}
2771EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2772
a2817d8b 2773void xfrm_policy_unregister_afinfo(const struct xfrm_policy_afinfo *afinfo)
1da177e4 2774{
2b61997a 2775 struct dst_ops *dst_ops = afinfo->dst_ops;
a2817d8b 2776 int i;
2b61997a 2777
a2817d8b
FW		 2778 for (i = 0; i < ARRAY_SIZE(xfrm_policy_afinfo); i++) {
2779 if (xfrm_policy_afinfo[i] != afinfo)
2780 continue;
2781 RCU_INIT_POINTER(xfrm_policy_afinfo[i], NULL);
2782 break;
ef8531b6 2783 }
ef8531b6 2784
2b61997a 2785 synchronize_rcu();
ef8531b6 2786
2b61997a
FW		 2787 dst_ops->kmem_cachep = NULL;
2788 dst_ops->check = NULL;
2789 dst_ops->negative_advice = NULL;
2790 dst_ops->link_failure = NULL;
1da177e4
LT		 2791}
2792EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2793
558f82ef 2794#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2795static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2796{
c68cd1a0 2797 int rv;
698365fa
WC		 2798 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2799 if (!net->mib.xfrm_statistics)
558f82ef 2800 return -ENOMEM;
c68cd1a0
AD		 2801 rv = xfrm_proc_init(net);
2802 if (rv < 0)
698365fa 2803 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2804 return rv;
558f82ef 2805}
59c9940e
AD		 2806
2807static void xfrm_statistics_fini(struct net *net)
2808{
c68cd1a0 2809 xfrm_proc_fini(net);
698365fa 2810 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2811}
2812#else
2813static int __net_init xfrm_statistics_init(struct net *net)
2814{
2815 return 0;
2816}
2817
2818static void xfrm_statistics_fini(struct net *net)
2819{
2820}
558f82ef
MN		 2821#endif
2822
d62ddc21 2823static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2824{
2518c7c2
DM		 2825 unsigned int hmask, sz;
2826 int dir;
2827
d62ddc21
AD		 2828 if (net_eq(net, &init_net))
2829 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2830 sizeof(struct xfrm_dst),
e5d679f3 2831 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2832 NULL);
1da177e4 2833
2518c7c2
DM		 2834 hmask = 8 - 1;
2835 sz = (hmask+1) * sizeof(struct hlist_head);
2836
93b851c1
AD		 2837 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2838 if (!net->xfrm.policy_byidx)
2839 goto out_byidx;
8100bea7 2840 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2841
53c2e285 2842 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 2843 struct xfrm_policy_hash *htab;
2844
dc2caba7 2845 net->xfrm.policy_count[dir] = 0;
53c2e285 2846 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2847 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2848
a35f6c5d 2849 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2850 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2851 if (!htab->table)
a35f6c5d
AD		 2852 goto out_bydst;
2853 htab->hmask = hmask;
b58555f1
CG		 2854 htab->dbits4 = 32;
2855 htab->sbits4 = 32;
2856 htab->dbits6 = 128;
2857 htab->sbits6 = 128;
2518c7c2 2858 }
880a6fab
CG		 2859 net->xfrm.policy_hthresh.lbits4 = 32;
2860 net->xfrm.policy_hthresh.rbits4 = 32;
2861 net->xfrm.policy_hthresh.lbits6 = 128;
2862 net->xfrm.policy_hthresh.rbits6 = 128;
2863
2864 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2865
adfcf0b2 2866 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2867 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2868 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21 2869 if (net_eq(net, &init_net))
21f42cc9 2870 xfrm_dev_init();
d62ddc21 2871 return 0;
93b851c1 2872
a35f6c5d
AD		 2873out_bydst:
2874 for (dir--; dir >= 0; dir--) {
2875 struct xfrm_policy_hash *htab;
2876
2877 htab = &net->xfrm.policy_bydst[dir];
2878 xfrm_hash_free(htab->table, sz);
2879 }
2880 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2881out_byidx:
2882 return -ENOMEM;
d62ddc21
AD		 2883}
2884
2885static void xfrm_policy_fini(struct net *net)
2886{
93b851c1 2887 unsigned int sz;
8b18f8ea 2888 int dir;
93b851c1 2889
7c2776ee
AD		 2890 flush_work(&net->xfrm.policy_hash_work);
2891#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2892 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 2893#endif
2e71029e 2894 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 2895
adfcf0b2 2896 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2897
53c2e285 2898 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD		 2899 struct xfrm_policy_hash *htab;
2900
8b18f8ea 2901 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2902
2903 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2904 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2905 WARN_ON(!hlist_empty(htab->table));
2906 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2907 }
2908
8100bea7 2909 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2910 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2911 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2912}
2913
d62ddc21
AD		 2914static int __net_init xfrm_net_init(struct net *net)
2915{
2916 int rv;
2917
c282222a
FW		 2918 /* Initialize the per-net locks here */
2919 spin_lock_init(&net->xfrm.xfrm_state_lock);
2920 spin_lock_init(&net->xfrm.xfrm_policy_lock);
2921 mutex_init(&net->xfrm.xfrm_cfg_mutex);
2922
59c9940e
AD		 2923 rv = xfrm_statistics_init(net);
2924 if (rv < 0)
2925 goto out_statistics;
d62ddc21
AD		 2926 rv = xfrm_state_init(net);
2927 if (rv < 0)
2928 goto out_state;
2929 rv = xfrm_policy_init(net);
2930 if (rv < 0)
2931 goto out_policy;
b27aeadb
AD		 2932 rv = xfrm_sysctl_init(net);
2933 if (rv < 0)
2934 goto out_sysctl;
283bc9f3 2935
d62ddc21
AD		 2936 return 0;
2937
b27aeadb
AD		 2938out_sysctl:
2939 xfrm_policy_fini(net);
d62ddc21
AD		 2940out_policy:
2941 xfrm_state_fini(net);
2942out_state:
59c9940e
AD		 2943 xfrm_statistics_fini(net);
2944out_statistics:
d62ddc21
AD		 2945 return rv;
2946}
2947
2948static void __net_exit xfrm_net_exit(struct net *net)
2949{
b27aeadb 2950 xfrm_sysctl_fini(net);
d62ddc21
AD		 2951 xfrm_policy_fini(net);
2952 xfrm_state_fini(net);
59c9940e 2953 xfrm_statistics_fini(net);
d62ddc21
AD		 2954}
2955
2956static struct pernet_operations __net_initdata xfrm_net_ops = {
2957 .init = xfrm_net_init,
2958 .exit = xfrm_net_exit,
2959};
2960
1da177e4
LT		 2961void __init xfrm_init(void)
2962{
ec30d78c
FW		 2963 int i;
2964
2965 xfrm_pcpu_work = kmalloc_array(NR_CPUS, sizeof(*xfrm_pcpu_work),
2966 GFP_KERNEL);
2967 BUG_ON(!xfrm_pcpu_work);
2968
2969 for (i = 0; i < NR_CPUS; i++)
2970 INIT_WORK(&xfrm_pcpu_work[i], xfrm_pcpu_work_fn);
2971
d62ddc21 2972 register_pernet_subsys(&xfrm_net_ops);
30846090 2973 seqcount_init(&xfrm_policy_hash_generation);
1da177e4
LT		 2974 xfrm_input_init();
2975}
2976
ab5f5e8b 2977#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2978static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2979 struct audit_buffer *audit_buf)
ab5f5e8b 2980{
875179fa
PM		 2981 struct xfrm_sec_ctx *ctx = xp->security;
2982 struct xfrm_selector *sel = &xp->selector;
2983
2984 if (ctx)
ab5f5e8b 2985 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2986 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2987
9b7a787d 2988 switch (sel->family) {
ab5f5e8b 2989 case AF_INET:
21454aaa 2990 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2991 if (sel->prefixlen_s != 32)
2992 audit_log_format(audit_buf, " src_prefixlen=%d",
2993 sel->prefixlen_s);
21454aaa 2994 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2995 if (sel->prefixlen_d != 32)
2996 audit_log_format(audit_buf, " dst_prefixlen=%d",
2997 sel->prefixlen_d);
ab5f5e8b
JL		 2998 break;
2999 case AF_INET6:
5b095d98 3000 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3001 if (sel->prefixlen_s != 128)
3002 audit_log_format(audit_buf, " src_prefixlen=%d",
3003 sel->prefixlen_s);
5b095d98 3004 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3005 if (sel->prefixlen_d != 128)
3006 audit_log_format(audit_buf, " dst_prefixlen=%d",
3007 sel->prefixlen_d);
ab5f5e8b
JL		 3008 break;
3009 }
3010}
3011
2e71029e 3012void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3013{
3014 struct audit_buffer *audit_buf;
ab5f5e8b 3015
afeb14b4 3016 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3017 if (audit_buf == NULL)
3018 return;
2e71029e 3019 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3020 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3021 xfrm_audit_common_policyinfo(xp, audit_buf);
3022 audit_log_end(audit_buf);
3023}
3024EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3025
68277acc 3026void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3027 bool task_valid)
ab5f5e8b
JL		 3028{
3029 struct audit_buffer *audit_buf;
ab5f5e8b 3030
afeb14b4 3031 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3032 if (audit_buf == NULL)
3033 return;
2e71029e 3034 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3035 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3036 xfrm_audit_common_policyinfo(xp, audit_buf);
3037 audit_log_end(audit_buf);
3038}
3039EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3040#endif
3041
80c9abaa 3042#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3043static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3044 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3045{
3046 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3047 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3048 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3049 sel_cmp->family) &&
3050 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3051 sel_cmp->family) &&
80c9abaa
SS		 3052 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3053 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3054 return true;
80c9abaa
SS		 3055 }
3056 } else {
3057 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3058 return true;
80c9abaa
SS		 3059 }
3060 }
bc9b35ad 3061 return false;
80c9abaa
SS		 3062}
3063
3e94c2dc
WC		 3064static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3065 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3066{
3067 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3068 struct hlist_head *chain;
3069 u32 priority = ~0U;
3070
9d0380df 3071 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d549c4f 3072 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3073 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3074 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3075 pol->type == type) {
3076 ret = pol;
3077 priority = ret->priority;
3078 break;
3079 }
3080 }
8d549c4f 3081 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3082 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR		 3083 if ((pol->priority >= priority) && ret)
3084 break;
3085
80c9abaa 3086 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
8faf491e 3087 pol->type == type) {
80c9abaa
SS		 3088 ret = pol;
3089 break;
3090 }
3091 }
3092
586f2eb4 3093 xfrm_pol_hold(ret);
80c9abaa 3094
9d0380df 3095 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3096
3097 return ret;
3098}
3099
dd701754 3100static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3101{
3102 int match = 0;
3103
3104 if (t->mode == m->mode && t->id.proto == m->proto &&
3105 (m->reqid == 0 || t->reqid == m->reqid)) {
3106 switch (t->mode) {
3107 case XFRM_MODE_TUNNEL:
3108 case XFRM_MODE_BEET:
70e94e66
YH		 3109 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3110 m->old_family) &&
3111 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3112 m->old_family)) {
80c9abaa
SS		 3113 match = 1;
3114 }
3115 break;
3116 case XFRM_MODE_TRANSPORT:
3117 /* in case of transport mode, template does not store
3118 any IP addresses, hence we just compare mode and
3119 protocol */
3120 match = 1;
3121 break;
3122 default:
3123 break;
3124 }
3125 }
3126 return match;
3127}
3128
3129/* update endpoint address(es) of template(s) */
3130static int xfrm_policy_migrate(struct xfrm_policy *pol,
3131 struct xfrm_migrate *m, int num_migrate)
3132{
3133 struct xfrm_migrate *mp;
80c9abaa
SS		 3134 int i, j, n = 0;
3135
3136 write_lock_bh(&pol->lock);
12a169e7 3137 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3138 /* target policy has been deleted */
3139 write_unlock_bh(&pol->lock);
3140 return -ENOENT;
3141 }
3142
3143 for (i = 0; i < pol->xfrm_nr; i++) {
3144 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3145 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3146 continue;
3147 n++;
1bfcb10f
HX		 3148 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3149 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3150 continue;
3151 /* update endpoints */
3152 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3153 sizeof(pol->xfrm_vec[i].id.daddr));
3154 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3155 sizeof(pol->xfrm_vec[i].saddr));
3156 pol->xfrm_vec[i].encap_family = mp->new_family;
3157 /* flush bundles */
80c802f3 3158 atomic_inc(&pol->genid);
80c9abaa
SS		 3159 }
3160 }
3161
3162 write_unlock_bh(&pol->lock);
3163
3164 if (!n)
3165 return -ENODATA;
3166
3167 return 0;
3168}
3169
dd701754 3170static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3171{
3172 int i, j;
3173
3174 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3175 return -EINVAL;
3176
3177 for (i = 0; i < num_migrate; i++) {
80c9abaa
SS		 3178 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3179 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3180 return -EINVAL;
3181
3182 /* check if there is any duplicated entry */
3183 for (j = i + 1; j < num_migrate; j++) {
3184 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3185 sizeof(m[i].old_daddr)) &&
3186 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3187 sizeof(m[i].old_saddr)) &&
3188 m[i].proto == m[j].proto &&
3189 m[i].mode == m[j].mode &&
3190 m[i].reqid == m[j].reqid &&
3191 m[i].old_family == m[j].old_family)
3192 return -EINVAL;
3193 }
3194 }
3195
3196 return 0;
3197}
3198
b4b7c0b3 3199int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3200 struct xfrm_migrate *m, int num_migrate,
4ab47d47
AA		 3201 struct xfrm_kmaddress *k, struct net *net,
3202 struct xfrm_encap_tmpl *encap)
80c9abaa
SS		 3203{
3204 int i, err, nx_cur = 0, nx_new = 0;
3205 struct xfrm_policy *pol = NULL;
3206 struct xfrm_state *x, *xc;
3207 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3208 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3209 struct xfrm_migrate *mp;
3210
7bab0963 3211 /* Stage 0 - sanity checks */
80c9abaa
SS		 3212 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3213 goto out;
3214
7bab0963
VD		 3215 if (dir >= XFRM_POLICY_MAX) {
3216 err = -EINVAL;
3217 goto out;
3218 }
3219
80c9abaa 3220 /* Stage 1 - find policy */
8d549c4f 3221 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3222 err = -ENOENT;
3223 goto out;
3224 }
3225
3226 /* Stage 2 - find and update state(s) */
3227 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3228 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3229 x_cur[nx_cur] = x;
3230 nx_cur++;
4ab47d47
AA		 3231 xc = xfrm_state_migrate(x, mp, encap);
3232 if (xc) {
80c9abaa
SS		 3233 x_new[nx_new] = xc;
3234 nx_new++;
3235 } else {
3236 err = -ENODATA;
3237 goto restore_state;
3238 }
3239 }
3240 }
3241
3242 /* Stage 3 - update policy */
3243 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3244 goto restore_state;
3245
3246 /* Stage 4 - delete old state(s) */
3247 if (nx_cur) {
3248 xfrm_states_put(x_cur, nx_cur);
3249 xfrm_states_delete(x_cur, nx_cur);
3250 }
3251
3252 /* Stage 5 - announce */
8bafd730 3253 km_migrate(sel, dir, type, m, num_migrate, k, encap);
80c9abaa
SS		 3254
3255 xfrm_pol_put(pol);
3256
3257 return 0;
3258out:
3259 return err;
3260
3261restore_state:
3262 if (pol)
3263 xfrm_pol_put(pol);
3264 if (nx_cur)
3265 xfrm_states_put(x_cur, nx_cur);
3266 if (nx_new)
3267 xfrm_states_delete(x_new, nx_new);
3268
3269 return err;
3270}
e610e679 3271EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3272#endif
Linux 6.x block layer and io_uring tree(s)