]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - net/xfrm/xfrm_policy.c
net: xfrm: allow clearing socket xfrm policies.
[mirror_ubuntu-bionic-kernel.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT
2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT
14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT
17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
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);
c3aed709 60static void xfrm_policy_queue_process(struct timer_list *t);
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
c3aed709 182static void xfrm_policy_timer(struct timer_list *t)
1da177e4 183{
c3aed709 184 struct xfrm_policy *xp = from_timer(xp, t, timer);
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);
c3aed709
KC
270 timer_setup(&policy->timer, xfrm_policy_timer, 0);
271 timer_setup(&policy->polq.hold_timer,
272 xfrm_policy_queue_process, 0);
1da177e4
LT
273 }
274 return policy;
275}
276EXPORT_SYMBOL(xfrm_policy_alloc);
277
56f04730
ED
278static void xfrm_policy_destroy_rcu(struct rcu_head *head)
279{
280 struct xfrm_policy *policy = container_of(head, struct xfrm_policy, rcu);
281
282 security_xfrm_policy_free(policy->security);
283 kfree(policy);
284}
285
1da177e4
LT
286/* Destroy xfrm_policy: descendant resources must be released to this moment. */
287
64c31b3f 288void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 289{
12a169e7 290 BUG_ON(!policy->walk.dead);
1da177e4 291
0659eea9 292 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT
293 BUG();
294
56f04730 295 call_rcu(&policy->rcu, xfrm_policy_destroy_rcu);
1da177e4 296}
64c31b3f 297EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 298
1365e547 299/* Rule must be locked. Release descendant resources, announce
1da177e4
LT
300 * entry dead. The rule must be unlinked from lists to the moment.
301 */
302
303static void xfrm_policy_kill(struct xfrm_policy *policy)
304{
12a169e7 305 policy->walk.dead = 1;
1da177e4 306
285ead17 307 atomic_inc(&policy->genid);
1da177e4 308
e7d8f6cb
SK
309 if (del_timer(&policy->polq.hold_timer))
310 xfrm_pol_put(policy);
1ee5e667 311 skb_queue_purge(&policy->polq.hold_queue);
a0073fe1 312
285ead17
TT
313 if (del_timer(&policy->timer))
314 xfrm_pol_put(policy);
315
316 xfrm_pol_put(policy);
1da177e4
LT
317}
318
2518c7c2
DM
319static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
320
e92303f8 321static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 322{
e92303f8 323 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM
324}
325
b58555f1
CG
326/* calculate policy hash thresholds */
327static void __get_hash_thresh(struct net *net,
328 unsigned short family, int dir,
329 u8 *dbits, u8 *sbits)
330{
331 switch (family) {
332 case AF_INET:
333 *dbits = net->xfrm.policy_bydst[dir].dbits4;
334 *sbits = net->xfrm.policy_bydst[dir].sbits4;
335 break;
336
337 case AF_INET6:
338 *dbits = net->xfrm.policy_bydst[dir].dbits6;
339 *sbits = net->xfrm.policy_bydst[dir].sbits6;
340 break;
341
342 default:
343 *dbits = 0;
344 *sbits = 0;
345 }
346}
347
5f803b58
DM
348static struct hlist_head *policy_hash_bysel(struct net *net,
349 const struct xfrm_selector *sel,
350 unsigned short family, int dir)
2518c7c2 351{
1121994c 352 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG
353 unsigned int hash;
354 u8 dbits;
355 u8 sbits;
356
357 __get_hash_thresh(net, family, dir, &dbits, &sbits);
358 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2 359
e1e551bc
FW
360 if (hash == hmask + 1)
361 return &net->xfrm.policy_inexact[dir];
362
363 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
364 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM
365}
366
5f803b58
DM
367static struct hlist_head *policy_hash_direct(struct net *net,
368 const xfrm_address_t *daddr,
369 const xfrm_address_t *saddr,
370 unsigned short family, int dir)
2518c7c2 371{
1121994c 372 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG
373 unsigned int hash;
374 u8 dbits;
375 u8 sbits;
376
377 __get_hash_thresh(net, family, dir, &dbits, &sbits);
378 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 379
e1e551bc
FW
380 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table,
381 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash;
2518c7c2
DM
382}
383
b58555f1
CG
384static void xfrm_dst_hash_transfer(struct net *net,
385 struct hlist_head *list,
2518c7c2 386 struct hlist_head *ndsttable,
b58555f1
CG
387 unsigned int nhashmask,
388 int dir)
2518c7c2 389{
b67bfe0d 390 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 391 struct xfrm_policy *pol;
b791160b 392 unsigned int h0 = 0;
b58555f1
CG
393 u8 dbits;
394 u8 sbits;
2518c7c2 395
b791160b 396redo:
b67bfe0d 397 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM
398 unsigned int h;
399
b58555f1 400 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 401 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 402 pol->family, nhashmask, dbits, sbits);
b791160b 403 if (!entry0) {
a5eefc1d
FW
404 hlist_del_rcu(&pol->bydst);
405 hlist_add_head_rcu(&pol->bydst, ndsttable + h);
b791160b
YH
406 h0 = h;
407 } else {
408 if (h != h0)
409 continue;
a5eefc1d
FW
410 hlist_del_rcu(&pol->bydst);
411 hlist_add_behind_rcu(&pol->bydst, entry0);
b791160b 412 }
b67bfe0d 413 entry0 = &pol->bydst;
b791160b
YH
414 }
415 if (!hlist_empty(list)) {
416 entry0 = NULL;
417 goto redo;
2518c7c2
DM
418 }
419}
420
421static void xfrm_idx_hash_transfer(struct hlist_head *list,
422 struct hlist_head *nidxtable,
423 unsigned int nhashmask)
424{
b67bfe0d 425 struct hlist_node *tmp;
2518c7c2
DM
426 struct xfrm_policy *pol;
427
b67bfe0d 428 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM
429 unsigned int h;
430
431 h = __idx_hash(pol->index, nhashmask);
432 hlist_add_head(&pol->byidx, nidxtable+h);
433 }
434}
435
436static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
437{
438 return ((old_hmask + 1) << 1) - 1;
439}
440
66caf628 441static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 442{
66caf628 443 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM
444 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
445 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
44e36b42 446 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
e1e551bc 447 struct hlist_head *odst;
2518c7c2
DM
448 int i;
449
450 if (!ndst)
451 return;
452
9d0380df 453 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
30846090
FW
454 write_seqcount_begin(&xfrm_policy_hash_generation);
455
456 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
457 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
2518c7c2 458
e1e551bc
FW
459 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table,
460 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
461
2518c7c2 462 for (i = hmask; i >= 0; i--)
b58555f1 463 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 464
e1e551bc 465 rcu_assign_pointer(net->xfrm.policy_bydst[dir].table, ndst);
66caf628 466 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 467
30846090 468 write_seqcount_end(&xfrm_policy_hash_generation);
9d0380df 469 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 470
e1e551bc
FW
471 synchronize_rcu();
472
44e36b42 473 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM
474}
475
66caf628 476static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 477{
66caf628 478 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM
479 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
480 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 481 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 482 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM
483 int i;
484
485 if (!nidx)
486 return;
487
9d0380df 488 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM
489
490 for (i = hmask; i >= 0; i--)
491 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
492
66caf628
AD
493 net->xfrm.policy_byidx = nidx;
494 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 495
9d0380df 496 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 497
44e36b42 498 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM
499}
500
66caf628 501static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 502{
66caf628
AD
503 unsigned int cnt = net->xfrm.policy_count[dir];
504 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM
505
506 if (total)
507 *total += cnt;
508
509 if ((hmask + 1) < xfrm_policy_hashmax &&
510 cnt > hmask)
511 return 1;
512
513 return 0;
514}
515
66caf628 516static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 517{
66caf628 518 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM
519
520 if ((hmask + 1) < xfrm_policy_hashmax &&
521 total > hmask)
522 return 1;
523
524 return 0;
525}
526
e071041b 527void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 528{
e071041b
AD
529 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
530 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
531 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
532 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
533 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
534 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
535 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 536 si->spdhmcnt = xfrm_policy_hashmax;
ecfd6b18
JHS
537}
538EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 539
ecfd6b18 540static DEFINE_MUTEX(hash_resize_mutex);
66caf628 541static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 542{
66caf628 543 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM
544 int dir, total;
545
546 mutex_lock(&hash_resize_mutex);
547
548 total = 0;
53c2e285 549 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
66caf628
AD
550 if (xfrm_bydst_should_resize(net, dir, &total))
551 xfrm_bydst_resize(net, dir);
2518c7c2 552 }
66caf628
AD
553 if (xfrm_byidx_should_resize(net, total))
554 xfrm_byidx_resize(net, total);
2518c7c2
DM
555
556 mutex_unlock(&hash_resize_mutex);
557}
558
880a6fab
CG
559static void xfrm_hash_rebuild(struct work_struct *work)
560{
561 struct net *net = container_of(work, struct net,
562 xfrm.policy_hthresh.work);
563 unsigned int hmask;
564 struct xfrm_policy *pol;
565 struct xfrm_policy *policy;
566 struct hlist_head *chain;
567 struct hlist_head *odst;
568 struct hlist_node *newpos;
569 int i;
570 int dir;
571 unsigned seq;
572 u8 lbits4, rbits4, lbits6, rbits6;
573
574 mutex_lock(&hash_resize_mutex);
575
576 /* read selector prefixlen thresholds */
577 do {
578 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
579
580 lbits4 = net->xfrm.policy_hthresh.lbits4;
581 rbits4 = net->xfrm.policy_hthresh.rbits4;
582 lbits6 = net->xfrm.policy_hthresh.lbits6;
583 rbits6 = net->xfrm.policy_hthresh.rbits6;
584 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
585
9d0380df 586 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
880a6fab
CG
587
588 /* reset the bydst and inexact table in all directions */
53c2e285 589 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
880a6fab
CG
590 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
591 hmask = net->xfrm.policy_bydst[dir].hmask;
592 odst = net->xfrm.policy_bydst[dir].table;
593 for (i = hmask; i >= 0; i--)
594 INIT_HLIST_HEAD(odst + i);
595 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
596 /* dir out => dst = remote, src = local */
597 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
598 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
599 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
600 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
601 } else {
602 /* dir in/fwd => dst = local, src = remote */
603 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
604 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
605 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
606 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
607 }
608 }
609
610 /* re-insert all policies by order of creation */
611 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
862591bf
FW
612 if (policy->walk.dead ||
613 xfrm_policy_id2dir(policy->index) >= XFRM_POLICY_MAX) {
6916fb3b
TB
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;
4aa2e62c 978out:
9d0380df 979 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 980 return err;
1da177e4
LT
981}
982EXPORT_SYMBOL(xfrm_policy_flush);
983
cdcbca7c 984int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 985 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT
986 void *data)
987{
12a169e7
HX
988 struct xfrm_policy *pol;
989 struct xfrm_policy_walk_entry *x;
4c563f76
TT
990 int error = 0;
991
992 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
993 walk->type != XFRM_POLICY_TYPE_ANY)
994 return -EINVAL;
1da177e4 995
12a169e7 996 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT
997 return 0;
998
9d0380df 999 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1000 if (list_empty(&walk->walk.all))
cdcbca7c 1001 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7 1002 else
80077702
LR
1003 x = list_first_entry(&walk->walk.all,
1004 struct xfrm_policy_walk_entry, all);
1005
cdcbca7c 1006 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1007 if (x->dead)
4c563f76 1008 continue;
12a169e7
HX
1009 pol = container_of(x, struct xfrm_policy, walk);
1010 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1011 walk->type != pol->type)
1012 continue;
1013 error = func(pol, xfrm_policy_id2dir(pol->index),
1014 walk->seq, data);
1015 if (error) {
1016 list_move_tail(&walk->walk.all, &x->all);
1017 goto out;
2518c7c2 1018 }
12a169e7 1019 walk->seq++;
1da177e4 1020 }
12a169e7 1021 if (walk->seq == 0) {
baf5d743
JHS
1022 error = -ENOENT;
1023 goto out;
1024 }
12a169e7 1025 list_del_init(&walk->walk.all);
1da177e4 1026out:
9d0380df 1027 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT
1028 return error;
1029}
1030EXPORT_SYMBOL(xfrm_policy_walk);
1031
12a169e7
HX
1032void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1033{
1034 INIT_LIST_HEAD(&walk->walk.all);
1035 walk->walk.dead = 1;
1036 walk->type = type;
1037 walk->seq = 0;
1038}
1039EXPORT_SYMBOL(xfrm_policy_walk_init);
1040
283bc9f3 1041void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX
1042{
1043 if (list_empty(&walk->walk.all))
1044 return;
1045
9d0380df 1046 spin_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1047 list_del(&walk->walk.all);
9d0380df 1048 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX
1049}
1050EXPORT_SYMBOL(xfrm_policy_walk_done);
1051
134b0fc5
JM
1052/*
1053 * Find policy to apply to this flow.
1054 *
1055 * Returns 0 if policy found, else an -errno.
1056 */
f299d557
DM
1057static int xfrm_policy_match(const struct xfrm_policy *pol,
1058 const struct flowi *fl,
2518c7c2 1059 u8 type, u16 family, int dir)
1da177e4 1060{
f299d557 1061 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM
1062 int ret = -ESRCH;
1063 bool match;
1da177e4 1064
2518c7c2 1065 if (pol->family != family ||
1d28f42c 1066 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1067 pol->type != type)
134b0fc5 1068 return ret;
1da177e4 1069
2518c7c2 1070 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1071 if (match)
1d28f42c 1072 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1073 dir);
2518c7c2 1074
134b0fc5 1075 return ret;
2518c7c2 1076}
1da177e4 1077
52479b62 1078static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1079 const struct flowi *fl,
2518c7c2
DM
1080 u16 family, u8 dir)
1081{
134b0fc5 1082 int err;
2518c7c2 1083 struct xfrm_policy *pol, *ret;
0b597e7e 1084 const xfrm_address_t *daddr, *saddr;
2518c7c2 1085 struct hlist_head *chain;
30846090
FW
1086 unsigned int sequence;
1087 u32 priority;
df71837d 1088
2518c7c2
DM
1089 daddr = xfrm_flowi_daddr(fl, family);
1090 saddr = xfrm_flowi_saddr(fl, family);
1091 if (unlikely(!daddr || !saddr))
1092 return NULL;
1093
a7c44247 1094 rcu_read_lock();
30846090
FW
1095 retry:
1096 do {
1097 sequence = read_seqcount_begin(&xfrm_policy_hash_generation);
1098 chain = policy_hash_direct(net, daddr, saddr, family, dir);
1099 } while (read_seqcount_retry(&xfrm_policy_hash_generation, sequence));
1100
1101 priority = ~0U;
2518c7c2 1102 ret = NULL;
a5eefc1d 1103 hlist_for_each_entry_rcu(pol, chain, bydst) {
134b0fc5
JM
1104 err = xfrm_policy_match(pol, fl, type, family, dir);
1105 if (err) {
1106 if (err == -ESRCH)
1107 continue;
1108 else {
1109 ret = ERR_PTR(err);
1110 goto fail;
1111 }
1112 } else {
2518c7c2 1113 ret = pol;
acba48e1 1114 priority = ret->priority;
2518c7c2
DM
1115 break;
1116 }
1117 }
52479b62 1118 chain = &net->xfrm.policy_inexact[dir];
a5eefc1d 1119 hlist_for_each_entry_rcu(pol, chain, bydst) {
8faf491e
LR
1120 if ((pol->priority >= priority) && ret)
1121 break;
1122
134b0fc5
JM
1123 err = xfrm_policy_match(pol, fl, type, family, dir);
1124 if (err) {
1125 if (err == -ESRCH)
1126 continue;
1127 else {
1128 ret = ERR_PTR(err);
1129 goto fail;
1130 }
8faf491e 1131 } else {
acba48e1
DM
1132 ret = pol;
1133 break;
1da177e4
LT
1134 }
1135 }
586f2eb4 1136
30846090
FW
1137 if (read_seqcount_retry(&xfrm_policy_hash_generation, sequence))
1138 goto retry;
1139
e37cc8ad
FW
1140 if (ret && !xfrm_pol_hold_rcu(ret))
1141 goto retry;
134b0fc5 1142fail:
a7c44247 1143 rcu_read_unlock();
4e81bb83 1144
2518c7c2 1145 return ret;
4e81bb83
MN
1146}
1147
80c802f3 1148static struct xfrm_policy *
86dc8ee0 1149xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT
1150{
1151#ifdef CONFIG_XFRM_SUB_POLICY
1152 struct xfrm_policy *pol;
1153
1154 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1155 if (pol != NULL)
1156 return pol;
1157#endif
1158 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1159}
1160
6f9c9615 1161static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir,
4c86d777 1162 const struct flowi *fl, u16 family)
1da177e4
LT
1163{
1164 struct xfrm_policy *pol;
1165
d188ba86 1166 rcu_read_lock();
ae33786f 1167 again:
d188ba86
ED
1168 pol = rcu_dereference(sk->sk_policy[dir]);
1169 if (pol != NULL) {
ddc47e44 1170 bool match;
a716c119 1171 int err = 0;
df71837d 1172
ddc47e44
SK
1173 if (pol->family != family) {
1174 pol = NULL;
1175 goto out;
1176 }
1177
1178 match = xfrm_selector_match(&pol->selector, fl, family);
3bccfbc7 1179 if (match) {
34f8d884
JHS
1180 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1181 pol = NULL;
1182 goto out;
1183 }
03e1ad7b 1184 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1185 fl->flowi_secid,
aff669bc 1186 dir);
330e832a
FW
1187 if (!err) {
1188 if (!xfrm_pol_hold_rcu(pol))
1189 goto again;
1190 } else if (err == -ESRCH) {
3bccfbc7 1191 pol = NULL;
330e832a 1192 } else {
3bccfbc7 1193 pol = ERR_PTR(err);
330e832a 1194 }
3bccfbc7 1195 } else
1da177e4
LT
1196 pol = NULL;
1197 }
34f8d884 1198out:
d188ba86 1199 rcu_read_unlock();
1da177e4
LT
1200 return pol;
1201}
1202
1203static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1204{
98806f75 1205 struct net *net = xp_net(pol);
4e81bb83 1206
98806f75 1207 list_add(&pol->walk.all, &net->xfrm.policy_all);
98806f75 1208 net->xfrm.policy_count[dir]++;
1da177e4
LT
1209 xfrm_pol_hold(pol);
1210}
1211
1212static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1213 int dir)
1214{
98806f75
AD
1215 struct net *net = xp_net(pol);
1216
53c2e285 1217 if (list_empty(&pol->walk.all))
2518c7c2 1218 return NULL;
1da177e4 1219
53c2e285
HX
1220 /* Socket policies are not hashed. */
1221 if (!hlist_unhashed(&pol->bydst)) {
a5eefc1d 1222 hlist_del_rcu(&pol->bydst);
53c2e285
HX
1223 hlist_del(&pol->byidx);
1224 }
1225
1226 list_del_init(&pol->walk.all);
98806f75 1227 net->xfrm.policy_count[dir]--;
2518c7c2
DM
1228
1229 return pol;
1da177e4
LT
1230}
1231
53c2e285
HX
1232static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir)
1233{
1234 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir);
1235}
1236
1237static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir)
1238{
1239 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir);
1240}
1241
4666faab 1242int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1243{
283bc9f3
FD
1244 struct net *net = xp_net(pol);
1245
9d0380df 1246 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1247 pol = __xfrm_policy_unlink(pol, dir);
9d0380df 1248 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1249 if (pol) {
1da177e4 1250 xfrm_policy_kill(pol);
4666faab 1251 return 0;
1da177e4 1252 }
4666faab 1253 return -ENOENT;
1da177e4 1254}
a70fcb0b 1255EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT
1256
1257int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1258{
acb2e0cd 1259 struct net *net = sock_net(sk);
1da177e4
LT
1260 struct xfrm_policy *old_pol;
1261
4e81bb83
MN
1262#ifdef CONFIG_XFRM_SUB_POLICY
1263 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1264 return -EINVAL;
1265#endif
1266
9d0380df 1267 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
d188ba86
ED
1268 old_pol = rcu_dereference_protected(sk->sk_policy[dir],
1269 lockdep_is_held(&net->xfrm.xfrm_policy_lock));
1da177e4 1270 if (pol) {
9d729f72 1271 pol->curlft.add_time = get_seconds();
e682adf0 1272 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
53c2e285 1273 xfrm_sk_policy_link(pol, dir);
1da177e4 1274 }
d188ba86 1275 rcu_assign_pointer(sk->sk_policy[dir], pol);
a0073fe1
SK
1276 if (old_pol) {
1277 if (pol)
1278 xfrm_policy_requeue(old_pol, pol);
1279
ea2dea9d
TT
1280 /* Unlinking succeeds always. This is the only function
1281 * allowed to delete or replace socket policy.
1282 */
53c2e285 1283 xfrm_sk_policy_unlink(old_pol, dir);
a0073fe1 1284 }
9d0380df 1285 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT
1286
1287 if (old_pol) {
1288 xfrm_policy_kill(old_pol);
1289 }
1290 return 0;
1291}
1292
d3e40a9f 1293static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1294{
0331b1f3 1295 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1296 struct net *net = xp_net(old);
1da177e4
LT
1297
1298 if (newp) {
1299 newp->selector = old->selector;
03e1ad7b
PM
1300 if (security_xfrm_policy_clone(old->security,
1301 &newp->security)) {
df71837d
TJ
1302 kfree(newp);
1303 return NULL; /* ENOMEM */
1304 }
1da177e4
LT
1305 newp->lft = old->lft;
1306 newp->curlft = old->curlft;
fb977e2c 1307 newp->mark = old->mark;
1da177e4
LT
1308 newp->action = old->action;
1309 newp->flags = old->flags;
1310 newp->xfrm_nr = old->xfrm_nr;
1311 newp->index = old->index;
4e81bb83 1312 newp->type = old->type;
0e74aa1d 1313 newp->family = old->family;
1da177e4
LT
1314 memcpy(newp->xfrm_vec, old->xfrm_vec,
1315 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
9d0380df 1316 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
53c2e285 1317 xfrm_sk_policy_link(newp, dir);
9d0380df 1318 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT
1319 xfrm_pol_put(newp);
1320 }
1321 return newp;
1322}
1323
d188ba86 1324int __xfrm_sk_clone_policy(struct sock *sk, const struct sock *osk)
1da177e4 1325{
d188ba86
ED
1326 const struct xfrm_policy *p;
1327 struct xfrm_policy *np;
1328 int i, ret = 0;
1329
1330 rcu_read_lock();
1331 for (i = 0; i < 2; i++) {
1332 p = rcu_dereference(osk->sk_policy[i]);
1333 if (p) {
1334 np = clone_policy(p, i);
1335 if (unlikely(!np)) {
1336 ret = -ENOMEM;
1337 break;
1338 }
1339 rcu_assign_pointer(sk->sk_policy[i], np);
1340 }
1341 }
1342 rcu_read_unlock();
1343 return ret;
1da177e4
LT
1344}
1345
a1e59abf 1346static int
42a7b32b 1347xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local,
077fbac4 1348 xfrm_address_t *remote, unsigned short family, u32 mark)
a1e59abf
PM
1349{
1350 int err;
37b10383 1351 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
a1e59abf
PM
1352
1353 if (unlikely(afinfo == NULL))
1354 return -EINVAL;
077fbac4 1355 err = afinfo->get_saddr(net, oif, local, remote, mark);
bdba9fe0 1356 rcu_read_unlock();
a1e59abf
PM
1357 return err;
1358}
1359
1da177e4
LT
1360/* Resolve list of templates for the flow, given policy. */
1361
1362static int
a6c2e611
DM
1363xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1364 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1365{
fbda33b2 1366 struct net *net = xp_net(policy);
1da177e4
LT
1367 int nx;
1368 int i, error;
94802151
SK
1369 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1370 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1371 xfrm_address_t tmp;
1da177e4 1372
9b7a787d 1373 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4 1374 struct xfrm_state *x;
94802151
SK
1375 xfrm_address_t *remote = daddr;
1376 xfrm_address_t *local = saddr;
1da177e4
LT
1377 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1378
94802151
SK
1379 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1380 tmpl->mode == XFRM_MODE_BEET) {
1381 remote = &tmpl->id.daddr;
1382 local = &tmpl->saddr;
1383 if (xfrm_addr_any(local, tmpl->encap_family)) {
1384 error = xfrm_get_saddr(net, fl->flowi_oif,
1385 &tmp, remote,
1386 tmpl->encap_family, 0);
1387 if (error)
1388 goto fail;
1389 local = &tmp;
1390 }
1da177e4
LT
1391 }
1392
1393 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1394
1395 if (x && x->km.state == XFRM_STATE_VALID) {
1396 xfrm[nx++] = x;
94802151
SK
1397 daddr = remote;
1398 saddr = local;
1da177e4
LT
1399 continue;
1400 }
1401 if (x) {
1402 error = (x->km.state == XFRM_STATE_ERROR ?
1403 -EINVAL : -EAGAIN);
1404 xfrm_state_put(x);
42054569 1405 } else if (error == -ESRCH) {
a4322266 1406 error = -EAGAIN;
42054569 1407 }
1da177e4
LT
1408
1409 if (!tmpl->optional)
1410 goto fail;
1411 }
1412 return nx;
1413
1414fail:
9b7a787d 1415 for (nx--; nx >= 0; nx--)
1da177e4
LT
1416 xfrm_state_put(xfrm[nx]);
1417 return error;
1418}
1419
4e81bb83 1420static int
a6c2e611
DM
1421xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1422 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1423{
41a49cc3
MN
1424 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1425 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN
1426 int cnx = 0;
1427 int error;
1428 int ret;
1429 int i;
1430
1431 for (i = 0; i < npols; i++) {
1432 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1433 error = -ENOBUFS;
1434 goto fail;
1435 }
41a49cc3
MN
1436
1437 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN
1438 if (ret < 0) {
1439 error = ret;
1440 goto fail;
1441 } else
1442 cnx += ret;
1443 }
1444
41a49cc3
MN
1445 /* found states are sorted for outbound processing */
1446 if (npols > 1)
1447 xfrm_state_sort(xfrm, tpp, cnx, family);
1448
4e81bb83
MN
1449 return cnx;
1450
1451 fail:
9b7a787d 1452 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1453 xfrm_state_put(tpp[cnx]);
4e81bb83
MN
1454 return error;
1455
1456}
1457
f5e2bb4f 1458static int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286 1459{
37b10383 1460 const struct xfrm_policy_afinfo *afinfo;
f5e2bb4f 1461 int tos = 0;
1da177e4 1462
f5e2bb4f
FW
1463 afinfo = xfrm_policy_get_afinfo(family);
1464 tos = afinfo ? afinfo->get_tos(fl) : 0;
25ee3286 1465
bdba9fe0 1466 rcu_read_unlock();
25ee3286
HX
1467
1468 return tos;
1469}
1470
d7c7544c 1471static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1472{
37b10383 1473 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1474 struct dst_ops *dst_ops;
25ee3286
HX
1475 struct xfrm_dst *xdst;
1476
1477 if (!afinfo)
1478 return ERR_PTR(-EINVAL);
1479
d7c7544c
AD
1480 switch (family) {
1481 case AF_INET:
1482 dst_ops = &net->xfrm.xfrm4_dst_ops;
1483 break;
dfd56b8b 1484#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD
1485 case AF_INET6:
1486 dst_ops = &net->xfrm.xfrm6_dst_ops;
1487 break;
1488#endif
1489 default:
1490 BUG();
1491 }
b2a9c0ed 1492 xdst = dst_alloc(dst_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
25ee3286 1493
d4cae562 1494 if (likely(xdst)) {
141e369d
SK
1495 struct dst_entry *dst = &xdst->u.dst;
1496
1497 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
d4cae562 1498 } else
0b150932 1499 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1500
bdba9fe0 1501 rcu_read_unlock();
d4cae562 1502
25ee3286
HX
1503 return xdst;
1504}
1505
a1b05140
MN
1506static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1507 int nfheader_len)
1508{
37b10383 1509 const struct xfrm_policy_afinfo *afinfo =
a1b05140
MN
1510 xfrm_policy_get_afinfo(dst->ops->family);
1511 int err;
1512
1513 if (!afinfo)
1514 return -EINVAL;
1515
1516 err = afinfo->init_path(path, dst, nfheader_len);
1517
bdba9fe0 1518 rcu_read_unlock();
a1b05140
MN
1519
1520 return err;
1521}
1522
87c1e12b 1523static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1524 const struct flowi *fl)
25ee3286 1525{
37b10383 1526 const struct xfrm_policy_afinfo *afinfo =
25ee3286
HX
1527 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1528 int err;
1529
1530 if (!afinfo)
1da177e4 1531 return -EINVAL;
25ee3286 1532
87c1e12b 1533 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1534
bdba9fe0 1535 rcu_read_unlock();
25ee3286 1536
1da177e4
LT
1537 return err;
1538}
1539
80c802f3 1540
25ee3286
HX
1541/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1542 * all the metrics... Shortly, bundle a bundle.
1543 */
1544
1545static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1546 struct xfrm_state **xfrm, int nx,
98313ada 1547 const struct flowi *fl,
25ee3286
HX
1548 struct dst_entry *dst)
1549{
d7c7544c 1550 struct net *net = xp_net(policy);
25ee3286
HX
1551 unsigned long now = jiffies;
1552 struct net_device *dev;
43a4dea4 1553 struct xfrm_mode *inner_mode;
25ee3286
HX
1554 struct dst_entry *dst_prev = NULL;
1555 struct dst_entry *dst0 = NULL;
1556 int i = 0;
1557 int err;
1558 int header_len = 0;
a1b05140 1559 int nfheader_len = 0;
25ee3286
HX
1560 int trailer_len = 0;
1561 int tos;
1562 int family = policy->selector.family;
9bb182a7
YH
1563 xfrm_address_t saddr, daddr;
1564
1565 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX
1566
1567 tos = xfrm_get_tos(fl, family);
25ee3286
HX
1568
1569 dst_hold(dst);
1570
1571 for (; i < nx; i++) {
d7c7544c 1572 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX
1573 struct dst_entry *dst1 = &xdst->u.dst;
1574
1575 err = PTR_ERR(xdst);
1576 if (IS_ERR(xdst)) {
1577 dst_release(dst);
1578 goto put_states;
1579 }
1580
10a7ef33
DM
1581 if (!dst_prev)
1582 dst0 = dst1;
1583 else
1584 /* Ref count is taken during xfrm_alloc_dst()
1585 * No need to do dst_clone() on dst1
1586 */
1587 dst_prev->child = dst1;
1588
43a4dea4
SK
1589 if (xfrm[i]->sel.family == AF_UNSPEC) {
1590 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1591 xfrm_af2proto(family));
1592 if (!inner_mode) {
1593 err = -EAFNOSUPPORT;
1594 dst_release(dst);
1595 goto put_states;
1596 }
1597 } else
1598 inner_mode = xfrm[i]->inner_mode;
1599
25ee3286 1600 xdst->route = dst;
defb3519 1601 dst_copy_metrics(dst1, dst);
25ee3286
HX
1602
1603 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1604 family = xfrm[i]->props.family;
42a7b32b 1605 dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
077fbac4
LC
1606 &saddr, &daddr, family,
1607 xfrm[i]->props.output_mark);
25ee3286
HX
1608 err = PTR_ERR(dst);
1609 if (IS_ERR(dst))
1610 goto put_states;
1611 } else
1612 dst_hold(dst);
1613
1614 dst1->xfrm = xfrm[i];
80c802f3 1615 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1616
f5b0a874 1617 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX
1618 dst1->flags |= DST_HOST;
1619 dst1->lastuse = now;
1620
1621 dst1->input = dst_discard;
43a4dea4 1622 dst1->output = inner_mode->afinfo->output;
25ee3286
HX
1623
1624 dst1->next = dst_prev;
1625 dst_prev = dst1;
1626
1627 header_len += xfrm[i]->props.header_len;
a1b05140
MN
1628 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1629 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX
1630 trailer_len += xfrm[i]->props.trailer_len;
1631 }
1632
1633 dst_prev->child = dst;
1634 dst0->path = dst;
1635
1636 err = -ENODEV;
1637 dev = dst->dev;
1638 if (!dev)
1639 goto free_dst;
1640
a1b05140 1641 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX
1642 xfrm_init_pmtu(dst_prev);
1643
1644 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1645 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1646
87c1e12b 1647 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX
1648 if (err)
1649 goto free_dst;
1650
1651 dst_prev->header_len = header_len;
1652 dst_prev->trailer_len = trailer_len;
1653 header_len -= xdst->u.dst.xfrm->props.header_len;
1654 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1655 }
1656
1657out:
1658 return dst0;
1659
1660put_states:
1661 for (; i < nx; i++)
1662 xfrm_state_put(xfrm[i]);
1663free_dst:
1664 if (dst0)
52df157f 1665 dst_release_immediate(dst0);
25ee3286
HX
1666 dst0 = ERR_PTR(err);
1667 goto out;
1668}
1669
73ff93cd 1670static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT
1671 struct xfrm_policy **pols,
1672 int *num_pols, int *num_xfrms)
1673{
1674 int i;
1675
1676 if (*num_pols == 0 || !pols[0]) {
1677 *num_pols = 0;
1678 *num_xfrms = 0;
1679 return 0;
1680 }
1681 if (IS_ERR(pols[0]))
1682 return PTR_ERR(pols[0]);
1683
1684 *num_xfrms = pols[0]->xfrm_nr;
1685
1686#ifdef CONFIG_XFRM_SUB_POLICY
1687 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1688 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1689 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1690 XFRM_POLICY_TYPE_MAIN,
1691 fl, family,
1692 XFRM_POLICY_OUT);
1693 if (pols[1]) {
1694 if (IS_ERR(pols[1])) {
1695 xfrm_pols_put(pols, *num_pols);
1696 return PTR_ERR(pols[1]);
1697 }
02d0892f 1698 (*num_pols)++;
80c802f3
TT
1699 (*num_xfrms) += pols[1]->xfrm_nr;
1700 }
1701 }
1702#endif
1703 for (i = 0; i < *num_pols; i++) {
1704 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1705 *num_xfrms = -1;
1706 break;
1707 }
1708 }
1709
1710 return 0;
1711
1712}
1713
ec30d78c
FW
1714static void xfrm_last_dst_update(struct xfrm_dst *xdst, struct xfrm_dst *old)
1715{
1716 this_cpu_write(xfrm_last_dst, xdst);
1717 if (old)
1718 dst_release(&old->u.dst);
1719}
1720
1721static void __xfrm_pcpu_work_fn(void)
1722{
1723 struct xfrm_dst *old;
1724
1725 old = this_cpu_read(xfrm_last_dst);
1726 if (old && !xfrm_bundle_ok(old))
1727 xfrm_last_dst_update(NULL, old);
1728}
1729
1730static void xfrm_pcpu_work_fn(struct work_struct *work)
1731{
1732 local_bh_disable();
1733 rcu_read_lock();
1734 __xfrm_pcpu_work_fn();
1735 rcu_read_unlock();
1736 local_bh_enable();
1737}
1738
1739void xfrm_policy_cache_flush(void)
1740{
1741 struct xfrm_dst *old;
1742 bool found = 0;
1743 int cpu;
1744
b1bdcb59
FW
1745 might_sleep();
1746
ec30d78c
FW
1747 local_bh_disable();
1748 rcu_read_lock();
1749 for_each_possible_cpu(cpu) {
1750 old = per_cpu(xfrm_last_dst, cpu);
1751 if (old && !xfrm_bundle_ok(old)) {
1752 if (smp_processor_id() == cpu) {
1753 __xfrm_pcpu_work_fn();
1754 continue;
1755 }
1756 found = true;
1757 break;
1758 }
1759 }
1760
1761 rcu_read_unlock();
1762 local_bh_enable();
1763
1764 if (!found)
1765 return;
1766
1767 get_online_cpus();
1768
1769 for_each_possible_cpu(cpu) {
1770 bool bundle_release;
1771
1772 rcu_read_lock();
1773 old = per_cpu(xfrm_last_dst, cpu);
1774 bundle_release = old && !xfrm_bundle_ok(old);
1775 rcu_read_unlock();
1776
1777 if (!bundle_release)
1778 continue;
1779
1780 if (cpu_online(cpu)) {
1781 schedule_work_on(cpu, &xfrm_pcpu_work[cpu]);
1782 continue;
1783 }
1784
1785 rcu_read_lock();
1786 old = per_cpu(xfrm_last_dst, cpu);
1787 if (old && !xfrm_bundle_ok(old)) {
1788 per_cpu(xfrm_last_dst, cpu) = NULL;
1789 dst_release(&old->u.dst);
1790 }
1791 rcu_read_unlock();
1792 }
1793
1794 put_online_cpus();
1795}
1796
cf379667
FW
1797static bool xfrm_xdst_can_reuse(struct xfrm_dst *xdst,
1798 struct xfrm_state * const xfrm[],
1799 int num)
ec30d78c 1800{
cf379667
FW
1801 const struct dst_entry *dst = &xdst->u.dst;
1802 int i;
ec30d78c 1803
cf379667
FW
1804 if (xdst->num_xfrms != num)
1805 return false;
ec30d78c 1806
cf379667
FW
1807 for (i = 0; i < num; i++) {
1808 if (!dst || dst->xfrm != xfrm[i])
1809 return false;
1810 dst = dst->child;
1811 }
ec30d78c 1812
cf379667 1813 return xfrm_bundle_ok(xdst);
ec30d78c
FW
1814}
1815
80c802f3
TT
1816static struct xfrm_dst *
1817xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1818 const struct flowi *fl, u16 family,
80c802f3
TT
1819 struct dst_entry *dst_orig)
1820{
1821 struct net *net = xp_net(pols[0]);
1822 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
ec30d78c 1823 struct xfrm_dst *xdst, *old;
80c802f3 1824 struct dst_entry *dst;
80c802f3
TT
1825 int err;
1826
cf379667
FW
1827 /* Try to instantiate a bundle */
1828 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1829 if (err <= 0) {
1830 if (err != 0 && err != -EAGAIN)
1831 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1832 return ERR_PTR(err);
1833 }
1834
ec30d78c
FW
1835 xdst = this_cpu_read(xfrm_last_dst);
1836 if (xdst &&
1837 xdst->u.dst.dev == dst_orig->dev &&
1838 xdst->num_pols == num_pols &&
ec30d78c 1839 memcmp(xdst->pols, pols,
13ead5c4 1840 sizeof(struct xfrm_policy *) * num_pols) == 0 &&
cf379667 1841 xfrm_xdst_can_reuse(xdst, xfrm, err)) {
ec30d78c 1842 dst_hold(&xdst->u.dst);
d2950278 1843 xfrm_pols_put(pols, num_pols);
cf379667
FW
1844 while (err > 0)
1845 xfrm_state_put(xfrm[--err]);
ec30d78c
FW
1846 return xdst;
1847 }
1848
1849 old = xdst;
80c802f3
TT
1850
1851 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1852 if (IS_ERR(dst)) {
1853 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1854 return ERR_CAST(dst);
1855 }
1856
1857 xdst = (struct xfrm_dst *)dst;
1858 xdst->num_xfrms = err;
80c802f3 1859 xdst->num_pols = num_pols;
3e94c2dc 1860 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT
1861 xdst->policy_genid = atomic_read(&pols[0]->genid);
1862
ec30d78c
FW
1863 atomic_set(&xdst->u.dst.__refcnt, 2);
1864 xfrm_last_dst_update(xdst, old);
1865
80c802f3
TT
1866 return xdst;
1867}
1868
c3aed709 1869static void xfrm_policy_queue_process(struct timer_list *t)
a0073fe1 1870{
a0073fe1
SK
1871 struct sk_buff *skb;
1872 struct sock *sk;
1873 struct dst_entry *dst;
c3aed709 1874 struct xfrm_policy *pol = from_timer(pol, t, polq.hold_timer);
3f5312ae 1875 struct net *net = xp_net(pol);
a0073fe1
SK
1876 struct xfrm_policy_queue *pq = &pol->polq;
1877 struct flowi fl;
1878 struct sk_buff_head list;
1879
1880 spin_lock(&pq->hold_queue.lock);
1881 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK
1882 if (!skb) {
1883 spin_unlock(&pq->hold_queue.lock);
1884 goto out;
1885 }
a0073fe1
SK
1886 dst = skb_dst(skb);
1887 sk = skb->sk;
1888 xfrm_decode_session(skb, &fl, dst->ops->family);
1889 spin_unlock(&pq->hold_queue.lock);
1890
1891 dst_hold(dst->path);
3f5312ae 1892 dst = xfrm_lookup(net, dst->path, &fl, sk, 0);
a0073fe1
SK
1893 if (IS_ERR(dst))
1894 goto purge_queue;
1895
1896 if (dst->flags & DST_XFRM_QUEUE) {
1897 dst_release(dst);
1898
1899 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1900 goto purge_queue;
1901
1902 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK
1903 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1904 xfrm_pol_hold(pol);
1905 goto out;
a0073fe1
SK
1906 }
1907
1908 dst_release(dst);
1909
1910 __skb_queue_head_init(&list);
1911
1912 spin_lock(&pq->hold_queue.lock);
1913 pq->timeout = 0;
1914 skb_queue_splice_init(&pq->hold_queue, &list);
1915 spin_unlock(&pq->hold_queue.lock);
1916
1917 while (!skb_queue_empty(&list)) {
1918 skb = __skb_dequeue(&list);
1919
1920 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1921 dst_hold(skb_dst(skb)->path);
3f5312ae 1922 dst = xfrm_lookup(net, skb_dst(skb)->path, &fl, skb->sk, 0);
a0073fe1 1923 if (IS_ERR(dst)) {
a0073fe1
SK
1924 kfree_skb(skb);
1925 continue;
1926 }
1927
1928 nf_reset(skb);
1929 skb_dst_drop(skb);
1930 skb_dst_set(skb, dst);
1931
13206b6b 1932 dst_output(net, skb->sk, skb);
a0073fe1
SK
1933 }
1934
e7d8f6cb
SK
1935out:
1936 xfrm_pol_put(pol);
a0073fe1
SK
1937 return;
1938
1939purge_queue:
1940 pq->timeout = 0;
1ee5e667 1941 skb_queue_purge(&pq->hold_queue);
e7d8f6cb 1942 xfrm_pol_put(pol);
a0073fe1
SK
1943}
1944
ede2059d 1945static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb)
a0073fe1
SK
1946{
1947 unsigned long sched_next;
1948 struct dst_entry *dst = skb_dst(skb);
1949 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK
1950 struct xfrm_policy *pol = xdst->pols[0];
1951 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1952
39bb5e62 1953 if (unlikely(skb_fclone_busy(sk, skb))) {
4d53eff4
SK
1954 kfree_skb(skb);
1955 return 0;
1956 }
a0073fe1
SK
1957
1958 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1959 kfree_skb(skb);
1960 return -EAGAIN;
1961 }
1962
1963 skb_dst_force(skb);
a0073fe1
SK
1964
1965 spin_lock_bh(&pq->hold_queue.lock);
1966
1967 if (!pq->timeout)
1968 pq->timeout = XFRM_QUEUE_TMO_MIN;
1969
1970 sched_next = jiffies + pq->timeout;
1971
1972 if (del_timer(&pq->hold_timer)) {
1973 if (time_before(pq->hold_timer.expires, sched_next))
1974 sched_next = pq->hold_timer.expires;
e7d8f6cb 1975 xfrm_pol_put(pol);
a0073fe1
SK
1976 }
1977
1978 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK
1979 if (!mod_timer(&pq->hold_timer, sched_next))
1980 xfrm_pol_hold(pol);
a0073fe1
SK
1981
1982 spin_unlock_bh(&pq->hold_queue.lock);
1983
1984 return 0;
1985}
1986
1987static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 1988 struct xfrm_flo *xflo,
a0073fe1
SK
1989 const struct flowi *fl,
1990 int num_xfrms,
1991 u16 family)
1992{
1993 int err;
1994 struct net_device *dev;
b8c203b2 1995 struct dst_entry *dst;
a0073fe1
SK
1996 struct dst_entry *dst1;
1997 struct xfrm_dst *xdst;
1998
1999 xdst = xfrm_alloc_dst(net, family);
2000 if (IS_ERR(xdst))
2001 return xdst;
2002
b8c203b2
SK
2003 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2004 net->xfrm.sysctl_larval_drop ||
2005 num_xfrms <= 0)
a0073fe1
SK
2006 return xdst;
2007
b8c203b2 2008 dst = xflo->dst_orig;
a0073fe1
SK
2009 dst1 = &xdst->u.dst;
2010 dst_hold(dst);
2011 xdst->route = dst;
2012
2013 dst_copy_metrics(dst1, dst);
2014
2015 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2016 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2017 dst1->lastuse = jiffies;
2018
2019 dst1->input = dst_discard;
2020 dst1->output = xdst_queue_output;
2021
2022 dst_hold(dst);
2023 dst1->child = dst;
2024 dst1->path = dst;
2025
2026 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2027
2028 err = -ENODEV;
2029 dev = dst->dev;
2030 if (!dev)
2031 goto free_dst;
2032
2033 err = xfrm_fill_dst(xdst, dev, fl);
2034 if (err)
2035 goto free_dst;
2036
2037out:
2038 return xdst;
2039
2040free_dst:
2041 dst_release(dst1);
2042 xdst = ERR_PTR(err);
2043 goto out;
2044}
2045
bd45c539 2046static struct xfrm_dst *
3ca28286 2047xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir, struct xfrm_flo *xflo)
80c802f3 2048{
80c802f3 2049 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
855dad99 2050 int num_pols = 0, num_xfrms = 0, err;
bd45c539 2051 struct xfrm_dst *xdst;
80c802f3 2052
80c802f3
TT
2053 /* Resolve policies to use if we couldn't get them from
2054 * previous cache entry */
855dad99 2055 num_pols = 1;
86dc8ee0 2056 pols[0] = xfrm_policy_lookup(net, fl, family, dir);
855dad99 2057 err = xfrm_expand_policies(fl, family, pols,
80c802f3 2058 &num_pols, &num_xfrms);
855dad99
FW
2059 if (err < 0)
2060 goto inc_error;
2061 if (num_pols == 0)
2062 return NULL;
2063 if (num_xfrms <= 0)
2064 goto make_dummy_bundle;
80c802f3 2065
76a42011 2066 local_bh_disable();
bd45c539 2067 xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
76a42011
SK
2068 xflo->dst_orig);
2069 local_bh_enable();
2070
bd45c539
FW
2071 if (IS_ERR(xdst)) {
2072 err = PTR_ERR(xdst);
80c802f3
TT
2073 if (err != -EAGAIN)
2074 goto error;
855dad99 2075 goto make_dummy_bundle;
bd45c539 2076 } else if (xdst == NULL) {
d809ec89 2077 num_xfrms = 0;
855dad99 2078 goto make_dummy_bundle;
80c802f3
TT
2079 }
2080
bd45c539 2081 return xdst;
80c802f3
TT
2082
2083make_dummy_bundle:
2084 /* We found policies, but there's no bundles to instantiate:
2085 * either because the policy blocks, has no transformations or
2086 * we could not build template (no xfrm_states).*/
b8c203b2 2087 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT
2088 if (IS_ERR(xdst)) {
2089 xfrm_pols_put(pols, num_pols);
2090 return ERR_CAST(xdst);
2091 }
2092 xdst->num_pols = num_pols;
2093 xdst->num_xfrms = num_xfrms;
3e94c2dc 2094 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3 2095
bd45c539 2096 return xdst;
80c802f3
TT
2097
2098inc_error:
2099 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2100error:
855dad99 2101 xfrm_pols_put(pols, num_pols);
80c802f3
TT
2102 return ERR_PTR(err);
2103}
1da177e4 2104
2774c131
DM
2105static struct dst_entry *make_blackhole(struct net *net, u16 family,
2106 struct dst_entry *dst_orig)
2107{
37b10383 2108 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2774c131
DM
2109 struct dst_entry *ret;
2110
2111 if (!afinfo) {
2112 dst_release(dst_orig);
433a1954 2113 return ERR_PTR(-EINVAL);
2774c131
DM
2114 } else {
2115 ret = afinfo->blackhole_route(net, dst_orig);
2116 }
bdba9fe0 2117 rcu_read_unlock();
2774c131
DM
2118
2119 return ret;
2120}
2121
1da177e4
LT
2122/* Main function: finds/creates a bundle for given flow.
2123 *
2124 * At the moment we eat a raw IP route. Mostly to speed up lookups
2125 * on interfaces with disabled IPsec.
2126 */
452edd59
DM
2127struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2128 const struct flowi *fl,
6f9c9615 2129 const struct sock *sk, int flags)
1da177e4 2130{
4e81bb83 2131 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3 2132 struct xfrm_dst *xdst;
452edd59 2133 struct dst_entry *dst, *route;
80c802f3 2134 u16 family = dst_orig->ops->family;
aff669bc 2135 u8 dir = XFRM_POLICY_OUT;
4b021628 2136 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2137
80c802f3
TT
2138 dst = NULL;
2139 xdst = NULL;
2140 route = NULL;
4e81bb83 2141
bd5eb35f 2142 sk = sk_const_to_full_sk(sk);
f7944fb1 2143 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3 2144 num_pols = 1;
4c86d777 2145 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, family);
80c802f3
TT
2146 err = xfrm_expand_policies(fl, family, pols,
2147 &num_pols, &num_xfrms);
2148 if (err < 0)
75b8c133 2149 goto dropdst;
80c802f3
TT
2150
2151 if (num_pols) {
2152 if (num_xfrms <= 0) {
2153 drop_pols = num_pols;
2154 goto no_transform;
2155 }
2156
76a42011 2157 local_bh_disable();
80c802f3
TT
2158 xdst = xfrm_resolve_and_create_bundle(
2159 pols, num_pols, fl,
2160 family, dst_orig);
76a42011
SK
2161 local_bh_enable();
2162
80c802f3
TT
2163 if (IS_ERR(xdst)) {
2164 xfrm_pols_put(pols, num_pols);
2165 err = PTR_ERR(xdst);
2166 goto dropdst;
d809ec89
TT
2167 } else if (xdst == NULL) {
2168 num_xfrms = 0;
2169 drop_pols = num_pols;
2170 goto no_transform;
80c802f3
TT
2171 }
2172
80c802f3 2173 route = xdst->route;
0aa64774 2174 }
3bccfbc7 2175 }
1da177e4 2176
80c802f3 2177 if (xdst == NULL) {
b8c203b2
SK
2178 struct xfrm_flo xflo;
2179
2180 xflo.dst_orig = dst_orig;
2181 xflo.flags = flags;
2182
1da177e4 2183 /* To accelerate a bit... */
2518c7c2 2184 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2185 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2186 goto nopol;
1da177e4 2187
bd45c539
FW
2188 xdst = xfrm_bundle_lookup(net, fl, family, dir, &xflo);
2189 if (xdst == NULL)
80c802f3 2190 goto nopol;
bd45c539
FW
2191 if (IS_ERR(xdst)) {
2192 err = PTR_ERR(xdst);
75b8c133 2193 goto dropdst;
d66e37a9 2194 }
80c802f3
TT
2195
2196 num_pols = xdst->num_pols;
2197 num_xfrms = xdst->num_xfrms;
3e94c2dc 2198 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT
2199 route = xdst->route;
2200 }
2201
2202 dst = &xdst->u.dst;
2203 if (route == NULL && num_xfrms > 0) {
2204 /* The only case when xfrm_bundle_lookup() returns a
2205 * bundle with null route, is when the template could
2206 * not be resolved. It means policies are there, but
2207 * bundle could not be created, since we don't yet
2208 * have the xfrm_state's. We need to wait for KM to
2209 * negotiate new SA's or bail out with error.*/
2210 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2211 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
ac37e251
W
2212 err = -EREMOTE;
2213 goto error;
80c802f3 2214 }
80c802f3 2215
5b8ef341 2216 err = -EAGAIN;
80c802f3
TT
2217
2218 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2219 goto error;
1da177e4
LT
2220 }
2221
80c802f3
TT
2222no_transform:
2223 if (num_pols == 0)
8b7817f3 2224 goto nopol;
1da177e4 2225
80c802f3
TT
2226 if ((flags & XFRM_LOOKUP_ICMP) &&
2227 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2228 err = -ENOENT;
8b7817f3 2229 goto error;
80c802f3 2230 }
8b7817f3 2231
80c802f3
TT
2232 for (i = 0; i < num_pols; i++)
2233 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2234
80c802f3 2235 if (num_xfrms < 0) {
1da177e4 2236 /* Prohibit the flow */
59c9940e 2237 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM
2238 err = -EPERM;
2239 goto error;
80c802f3
TT
2240 } else if (num_xfrms > 0) {
2241 /* Flow transformed */
80c802f3
TT
2242 dst_release(dst_orig);
2243 } else {
2244 /* Flow passes untransformed */
2245 dst_release(dst);
452edd59 2246 dst = dst_orig;
1da177e4 2247 }
80c802f3
TT
2248ok:
2249 xfrm_pols_put(pols, drop_pols);
0c183379
G
2250 if (dst && dst->xfrm &&
2251 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2252 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2253 return dst;
1da177e4 2254
80c802f3 2255nopol:
452edd59
DM
2256 if (!(flags & XFRM_LOOKUP_ICMP)) {
2257 dst = dst_orig;
80c802f3 2258 goto ok;
452edd59 2259 }
80c802f3 2260 err = -ENOENT;
1da177e4 2261error:
80c802f3 2262 dst_release(dst);
75b8c133 2263dropdst:
ac37e251
W
2264 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF))
2265 dst_release(dst_orig);
80c802f3 2266 xfrm_pols_put(pols, drop_pols);
452edd59 2267 return ERR_PTR(err);
1da177e4
LT
2268}
2269EXPORT_SYMBOL(xfrm_lookup);
2270
f92ee619
SK
2271/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2272 * Otherwise we may send out blackholed packets.
2273 */
2274struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2275 const struct flowi *fl,
6f9c9615 2276 const struct sock *sk, int flags)
f92ee619 2277{
b8c203b2 2278 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
ac37e251
W
2279 flags | XFRM_LOOKUP_QUEUE |
2280 XFRM_LOOKUP_KEEP_DST_REF);
f92ee619
SK
2281
2282 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2283 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2284
2285 return dst;
2286}
2287EXPORT_SYMBOL(xfrm_lookup_route);
2288
df0ba92a 2289static inline int
8f029de2 2290xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN
2291{
2292 struct xfrm_state *x;
df0ba92a
MN
2293
2294 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2295 return 0;
2296 x = skb->sp->xvec[idx];
2297 if (!x->type->reject)
2298 return 0;
1ecafede 2299 return x->type->reject(x, skb, fl);
df0ba92a
MN
2300}
2301
1da177e4
LT
2302/* When skb is transformed back to its "native" form, we have to
2303 * check policy restrictions. At the moment we make this in maximally
2304 * stupid way. Shame on me. :-) Of course, connected sockets must
2305 * have policy cached at them.
2306 */
2307
2308static inline int
7db454b9 2309xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT
2310 unsigned short family)
2311{
2312 if (xfrm_state_kern(x))
928ba416 2313 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT
2314 return x->id.proto == tmpl->id.proto &&
2315 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2316 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2317 x->props.mode == tmpl->mode &&
c5d18e98 2318 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2319 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN
2320 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2321 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT
2322}
2323
df0ba92a
MN
2324/*
2325 * 0 or more than 0 is returned when validation is succeeded (either bypass
2326 * because of optional transport mode, or next index of the mathced secpath
2327 * state with the template.
2328 * -1 is returned when no matching template is found.
2329 * Otherwise "-2 - errored_index" is returned.
2330 */
1da177e4 2331static inline int
22cccb7e 2332xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT
2333 unsigned short family)
2334{
2335 int idx = start;
2336
2337 if (tmpl->optional) {
7e49e6de 2338 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT
2339 return start;
2340 } else
2341 start = -1;
2342 for (; idx < sp->len; idx++) {
dbe5b4aa 2343 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2344 return ++idx;
df0ba92a
MN
2345 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2346 if (start == -1)
2347 start = -2-idx;
1da177e4 2348 break;
df0ba92a 2349 }
1da177e4
LT
2350 }
2351 return start;
2352}
2353
d5422efe
HX
2354int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2355 unsigned int family, int reverse)
1da177e4 2356{
37b10383 2357 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2358 int err;
1da177e4
LT
2359
2360 if (unlikely(afinfo == NULL))
2361 return -EAFNOSUPPORT;
2362
d5422efe 2363 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2364 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
bdba9fe0 2365 rcu_read_unlock();
e0d1caa7 2366 return err;
1da177e4 2367}
d5422efe 2368EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2369
9a7386ec 2370static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT
2371{
2372 for (; k < sp->len; k++) {
df0ba92a 2373 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2374 *idxp = k;
1da177e4 2375 return 1;
df0ba92a 2376 }
1da177e4
LT
2377 }
2378
2379 return 0;
2380}
2381
a716c119 2382int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT
2383 unsigned short family)
2384{
f6e1e25d 2385 struct net *net = dev_net(skb->dev);
1da177e4 2386 struct xfrm_policy *pol;
4e81bb83
MN
2387 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2388 int npols = 0;
2389 int xfrm_nr;
2390 int pi;
d5422efe 2391 int reverse;
1da177e4 2392 struct flowi fl;
df0ba92a 2393 int xerr_idx = -1;
1da177e4 2394
d5422efe
HX
2395 reverse = dir & ~XFRM_POLICY_MASK;
2396 dir &= XFRM_POLICY_MASK;
d5422efe 2397
0aa64774 2398 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2399 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2400 return 0;
0aa64774
MN
2401 }
2402
eb9c7ebe 2403 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT
2404
2405 /* First, check used SA against their selectors. */
2406 if (skb->sp) {
2407 int i;
2408
9b7a787d 2409 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2410 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2411 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2412 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2413 return 0;
0aa64774 2414 }
1da177e4
LT
2415 }
2416 }
2417
2418 pol = NULL;
bd5eb35f 2419 sk = sk_to_full_sk(sk);
3bccfbc7 2420 if (sk && sk->sk_policy[dir]) {
4c86d777 2421 pol = xfrm_sk_policy_lookup(sk, dir, &fl, family);
0aa64774 2422 if (IS_ERR(pol)) {
59c9940e 2423 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2424 return 0;
0aa64774 2425 }
3bccfbc7 2426 }
1da177e4 2427
86dc8ee0
FW
2428 if (!pol)
2429 pol = xfrm_policy_lookup(net, &fl, family, dir);
1da177e4 2430
0aa64774 2431 if (IS_ERR(pol)) {
59c9940e 2432 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2433 return 0;
0aa64774 2434 }
134b0fc5 2435
df0ba92a 2436 if (!pol) {
d1d9facf 2437 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2438 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2439 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN
2440 return 0;
2441 }
2442 return 1;
2443 }
1da177e4 2444
9d729f72 2445 pol->curlft.use_time = get_seconds();
1da177e4 2446
4e81bb83 2447 pols[0] = pol;
02d0892f 2448 npols++;
4e81bb83
MN
2449#ifdef CONFIG_XFRM_SUB_POLICY
2450 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2451 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN
2452 &fl, family,
2453 XFRM_POLICY_IN);
2454 if (pols[1]) {
0aa64774 2455 if (IS_ERR(pols[1])) {
59c9940e 2456 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2457 return 0;
0aa64774 2458 }
9d729f72 2459 pols[1]->curlft.use_time = get_seconds();
02d0892f 2460 npols++;
4e81bb83
MN
2461 }
2462 }
2463#endif
2464
1da177e4
LT
2465 if (pol->action == XFRM_POLICY_ALLOW) {
2466 struct sec_path *sp;
2467 static struct sec_path dummy;
4e81bb83 2468 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2469 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN
2470 struct xfrm_tmpl **tpp = tp;
2471 int ti = 0;
1da177e4
LT
2472 int i, k;
2473
2474 if ((sp = skb->sp) == NULL)
2475 sp = &dummy;
2476
4e81bb83
MN
2477 for (pi = 0; pi < npols; pi++) {
2478 if (pols[pi] != pol &&
0aa64774 2479 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2480 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2481 goto reject;
0aa64774
MN
2482 }
2483 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2484 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2485 goto reject_error;
0aa64774 2486 }
4e81bb83
MN
2487 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2488 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2489 }
2490 xfrm_nr = ti;
41a49cc3 2491 if (npols > 1) {
283bc9f3 2492 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN
2493 tpp = stp;
2494 }
4e81bb83 2495
1da177e4
LT
2496 /* For each tunnel xfrm, find the first matching tmpl.
2497 * For each tmpl before that, find corresponding xfrm.
2498 * Order is _important_. Later we will implement
2499 * some barriers, but at the moment barriers
2500 * are implied between each two transformations.
2501 */
4e81bb83
MN
2502 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2503 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2504 if (k < 0) {
d1d9facf
JM
2505 if (k < -1)
2506 /* "-2 - errored_index" returned */
2507 xerr_idx = -(2+k);
59c9940e 2508 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2509 goto reject;
df0ba92a 2510 }
1da177e4
LT
2511 }
2512
0aa64774 2513 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2514 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2515 goto reject;
0aa64774 2516 }
1da177e4 2517
4e81bb83 2518 xfrm_pols_put(pols, npols);
1da177e4
LT
2519 return 1;
2520 }
59c9940e 2521 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT
2522
2523reject:
df0ba92a 2524 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN
2525reject_error:
2526 xfrm_pols_put(pols, npols);
1da177e4
LT
2527 return 0;
2528}
2529EXPORT_SYMBOL(__xfrm_policy_check);
2530
2531int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2532{
99a66657 2533 struct net *net = dev_net(skb->dev);
1da177e4 2534 struct flowi fl;
adf30907 2535 struct dst_entry *dst;
73137147 2536 int res = 1;
1da177e4 2537
0aa64774 2538 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2539 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2540 return 0;
0aa64774 2541 }
1da177e4 2542
fafeeb6c 2543 skb_dst_force(skb);
adf30907 2544
b8c203b2 2545 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2546 if (IS_ERR(dst)) {
73137147 2547 res = 0;
452edd59
DM
2548 dst = NULL;
2549 }
adf30907
ED
2550 skb_dst_set(skb, dst);
2551 return res;
1da177e4
LT
2552}
2553EXPORT_SYMBOL(__xfrm_route_forward);
2554
d49c73c7
DM
2555/* Optimize later using cookies and generation ids. */
2556
1da177e4
LT
2557static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2558{
d49c73c7 2559 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM
2560 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2561 * get validated by dst_ops->check on every use. We do this
2562 * because when a normal route referenced by an XFRM dst is
2563 * obsoleted we do not go looking around for all parent
2564 * referencing XFRM dsts so that we can invalidate them. It
2565 * is just too much work. Instead we make the checks here on
2566 * every use. For example:
d49c73c7
DM
2567 *
2568 * XFRM dst A --> IPv4 dst X
2569 *
2570 * X is the "xdst->route" of A (X is also the "dst->path" of A
2571 * in this example). If X is marked obsolete, "A" will not
2572 * notice. That's what we are validating here via the
2573 * stale_bundle() check.
2574 *
52df157f
WW
2575 * When a dst is removed from the fib tree, DST_OBSOLETE_DEAD will
2576 * be marked on it.
09c75704 2577 * This will force stale_bundle() to fail on any xdst bundle with
52df157f 2578 * this dst linked in it.
399c180a 2579 */
d49c73c7
DM
2580 if (dst->obsolete < 0 && !stale_bundle(dst))
2581 return dst;
2582
1da177e4
LT
2583 return NULL;
2584}
2585
2586static int stale_bundle(struct dst_entry *dst)
2587{
12fdb4d3 2588 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT
2589}
2590
aabc9761 2591void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2592{
1da177e4 2593 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2594 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2595 dev_hold(dst->dev);
1da177e4
LT
2596 dev_put(dev);
2597 }
2598}
aabc9761 2599EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT
2600
2601static void xfrm_link_failure(struct sk_buff *skb)
2602{
2603 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT
2604}
2605
2606static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2607{
2608 if (dst) {
2609 if (dst->obsolete) {
2610 dst_release(dst);
2611 dst = NULL;
2612 }
2613 }
2614 return dst;
2615}
2616
25ee3286 2617static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT
2618{
2619 do {
2620 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2621 u32 pmtu, route_mtu_cached;
2622
2623 pmtu = dst_mtu(dst->child);
2624 xdst->child_mtu_cached = pmtu;
2625
2626 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2627
2628 route_mtu_cached = dst_mtu(xdst->route);
2629 xdst->route_mtu_cached = route_mtu_cached;
2630
2631 if (pmtu > route_mtu_cached)
2632 pmtu = route_mtu_cached;
2633
defb3519 2634 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT
2635 } while ((dst = dst->next));
2636}
2637
1da177e4
LT
2638/* Check that the bundle accepts the flow and its components are
2639 * still valid.
2640 */
2641
12fdb4d3 2642static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT
2643{
2644 struct dst_entry *dst = &first->u.dst;
2645 struct xfrm_dst *last;
2646 u32 mtu;
2647
92d63dec 2648 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT
2649 (dst->dev && !netif_running(dst->dev)))
2650 return 0;
2651
a0073fe1
SK
2652 if (dst->flags & DST_XFRM_QUEUE)
2653 return 1;
2654
1da177e4
LT
2655 last = NULL;
2656
2657 do {
2658 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2659
1da177e4
LT
2660 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2661 return 0;
80c802f3
TT
2662 if (xdst->xfrm_genid != dst->xfrm->genid)
2663 return 0;
b1312c89
TT
2664 if (xdst->num_pols > 0 &&
2665 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2666 return 0;
e53820de 2667
1da177e4
LT
2668 mtu = dst_mtu(dst->child);
2669 if (xdst->child_mtu_cached != mtu) {
2670 last = xdst;
2671 xdst->child_mtu_cached = mtu;
2672 }
2673
92d63dec 2674 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT
2675 return 0;
2676 mtu = dst_mtu(xdst->route);
2677 if (xdst->route_mtu_cached != mtu) {
2678 last = xdst;
2679 xdst->route_mtu_cached = mtu;
2680 }
2681
2682 dst = dst->child;
2683 } while (dst->xfrm);
2684
2685 if (likely(!last))
2686 return 1;
2687
2688 mtu = last->child_mtu_cached;
2689 for (;;) {
2690 dst = &last->u.dst;
2691
2692 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2693 if (mtu > last->route_mtu_cached)
2694 mtu = last->route_mtu_cached;
defb3519 2695 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT
2696
2697 if (last == first)
2698 break;
2699
bd0bf076 2700 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT
2701 last->child_mtu_cached = mtu;
2702 }
2703
2704 return 1;
2705}
2706
0dbaee3b
DM
2707static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2708{
2709 return dst_metric_advmss(dst->path);
2710}
2711
ebb762f2 2712static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2713{
618f9bc7
SK
2714 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2715
2716 return mtu ? : dst_mtu(dst->path);
d33e4553
DM
2717}
2718
1ecc9ad0
JA
2719static const void *xfrm_get_dst_nexthop(const struct dst_entry *dst,
2720 const void *daddr)
63fca65d
JA
2721{
2722 const struct dst_entry *path = dst->path;
2723
2724 for (; dst != path; dst = dst->child) {
2725 const struct xfrm_state *xfrm = dst->xfrm;
2726
2727 if (xfrm->props.mode == XFRM_MODE_TRANSPORT)
2728 continue;
2729 if (xfrm->type->flags & XFRM_TYPE_REMOTE_COADDR)
2730 daddr = xfrm->coaddr;
2731 else if (!(xfrm->type->flags & XFRM_TYPE_LOCAL_COADDR))
2732 daddr = &xfrm->id.daddr;
2733 }
1ecc9ad0
JA
2734 return daddr;
2735}
2736
2737static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2738 struct sk_buff *skb,
2739 const void *daddr)
2740{
2741 const struct dst_entry *path = dst->path;
2742
2743 if (!skb)
2744 daddr = xfrm_get_dst_nexthop(dst, daddr);
2745 return path->ops->neigh_lookup(path, skb, daddr);
2746}
2747
2748static void xfrm_confirm_neigh(const struct dst_entry *dst, const void *daddr)
2749{
2750 const struct dst_entry *path = dst->path;
2751
2752 daddr = xfrm_get_dst_nexthop(dst, daddr);
63fca65d
JA
2753 path->ops->confirm_neigh(path, daddr);
2754}
2755
a2817d8b 2756int xfrm_policy_register_afinfo(const struct xfrm_policy_afinfo *afinfo, int family)
1da177e4
LT
2757{
2758 int err = 0;
a2817d8b
FW
2759
2760 if (WARN_ON(family >= ARRAY_SIZE(xfrm_policy_afinfo)))
1da177e4 2761 return -EAFNOSUPPORT;
a2817d8b 2762
ef8531b6 2763 spin_lock(&xfrm_policy_afinfo_lock);
a2817d8b 2764 if (unlikely(xfrm_policy_afinfo[family] != NULL))
f31e8d4f 2765 err = -EEXIST;
1da177e4
LT
2766 else {
2767 struct dst_ops *dst_ops = afinfo->dst_ops;
2768 if (likely(dst_ops->kmem_cachep == NULL))
2769 dst_ops->kmem_cachep = xfrm_dst_cache;
2770 if (likely(dst_ops->check == NULL))
2771 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM
2772 if (likely(dst_ops->default_advmss == NULL))
2773 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK
2774 if (likely(dst_ops->mtu == NULL))
2775 dst_ops->mtu = xfrm_mtu;
1da177e4
LT
2776 if (likely(dst_ops->negative_advice == NULL))
2777 dst_ops->negative_advice = xfrm_negative_advice;
2778 if (likely(dst_ops->link_failure == NULL))
2779 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM
2780 if (likely(dst_ops->neigh_lookup == NULL))
2781 dst_ops->neigh_lookup = xfrm_neigh_lookup;
63fca65d
JA
2782 if (likely(!dst_ops->confirm_neigh))
2783 dst_ops->confirm_neigh = xfrm_confirm_neigh;
a2817d8b 2784 rcu_assign_pointer(xfrm_policy_afinfo[family], afinfo);
1da177e4 2785 }
ef8531b6 2786 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c 2787
1da177e4
LT
2788 return err;
2789}
2790EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2791
a2817d8b 2792void xfrm_policy_unregister_afinfo(const struct xfrm_policy_afinfo *afinfo)
1da177e4 2793{
2b61997a 2794 struct dst_ops *dst_ops = afinfo->dst_ops;
a2817d8b 2795 int i;
2b61997a 2796
a2817d8b
FW
2797 for (i = 0; i < ARRAY_SIZE(xfrm_policy_afinfo); i++) {
2798 if (xfrm_policy_afinfo[i] != afinfo)
2799 continue;
2800 RCU_INIT_POINTER(xfrm_policy_afinfo[i], NULL);
2801 break;
ef8531b6 2802 }
ef8531b6 2803
2b61997a 2804 synchronize_rcu();
ef8531b6 2805
2b61997a
FW
2806 dst_ops->kmem_cachep = NULL;
2807 dst_ops->check = NULL;
2808 dst_ops->negative_advice = NULL;
2809 dst_ops->link_failure = NULL;
1da177e4
LT
2810}
2811EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2812
558f82ef 2813#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2814static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2815{
c68cd1a0 2816 int rv;
698365fa
WC
2817 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2818 if (!net->mib.xfrm_statistics)
558f82ef 2819 return -ENOMEM;
c68cd1a0
AD
2820 rv = xfrm_proc_init(net);
2821 if (rv < 0)
698365fa 2822 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2823 return rv;
558f82ef 2824}
59c9940e
AD
2825
2826static void xfrm_statistics_fini(struct net *net)
2827{
c68cd1a0 2828 xfrm_proc_fini(net);
698365fa 2829 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD
2830}
2831#else
2832static int __net_init xfrm_statistics_init(struct net *net)
2833{
2834 return 0;
2835}
2836
2837static void xfrm_statistics_fini(struct net *net)
2838{
2839}
558f82ef
MN
2840#endif
2841
d62ddc21 2842static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2843{
2518c7c2
DM
2844 unsigned int hmask, sz;
2845 int dir;
2846
d62ddc21
AD
2847 if (net_eq(net, &init_net))
2848 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2849 sizeof(struct xfrm_dst),
e5d679f3 2850 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2851 NULL);
1da177e4 2852
2518c7c2
DM
2853 hmask = 8 - 1;
2854 sz = (hmask+1) * sizeof(struct hlist_head);
2855
93b851c1
AD
2856 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2857 if (!net->xfrm.policy_byidx)
2858 goto out_byidx;
8100bea7 2859 net->xfrm.policy_idx_hmask = hmask;
2518c7c2 2860
53c2e285 2861 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM
2862 struct xfrm_policy_hash *htab;
2863
dc2caba7 2864 net->xfrm.policy_count[dir] = 0;
53c2e285 2865 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0;
8b18f8ea 2866 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2867
a35f6c5d 2868 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2869 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2870 if (!htab->table)
a35f6c5d
AD
2871 goto out_bydst;
2872 htab->hmask = hmask;
b58555f1
CG
2873 htab->dbits4 = 32;
2874 htab->sbits4 = 32;
2875 htab->dbits6 = 128;
2876 htab->sbits6 = 128;
2518c7c2 2877 }
880a6fab
CG
2878 net->xfrm.policy_hthresh.lbits4 = 32;
2879 net->xfrm.policy_hthresh.rbits4 = 32;
2880 net->xfrm.policy_hthresh.lbits6 = 128;
2881 net->xfrm.policy_hthresh.rbits6 = 128;
2882
2883 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2884
adfcf0b2 2885 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2886 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2887 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21 2888 if (net_eq(net, &init_net))
21f42cc9 2889 xfrm_dev_init();
d62ddc21 2890 return 0;
93b851c1 2891
a35f6c5d
AD
2892out_bydst:
2893 for (dir--; dir >= 0; dir--) {
2894 struct xfrm_policy_hash *htab;
2895
2896 htab = &net->xfrm.policy_bydst[dir];
2897 xfrm_hash_free(htab->table, sz);
2898 }
2899 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD
2900out_byidx:
2901 return -ENOMEM;
d62ddc21
AD
2902}
2903
2904static void xfrm_policy_fini(struct net *net)
2905{
93b851c1 2906 unsigned int sz;
8b18f8ea 2907 int dir;
93b851c1 2908
7c2776ee
AD
2909 flush_work(&net->xfrm.policy_hash_work);
2910#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2911 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 2912#endif
2e71029e 2913 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 2914
adfcf0b2 2915 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2916
53c2e285 2917 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
a35f6c5d
AD
2918 struct xfrm_policy_hash *htab;
2919
8b18f8ea 2920 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD
2921
2922 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2923 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD
2924 WARN_ON(!hlist_empty(htab->table));
2925 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD
2926 }
2927
8100bea7 2928 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD
2929 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2930 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT
2931}
2932
d62ddc21
AD
2933static int __net_init xfrm_net_init(struct net *net)
2934{
2935 int rv;
2936
c282222a
FW
2937 /* Initialize the per-net locks here */
2938 spin_lock_init(&net->xfrm.xfrm_state_lock);
2939 spin_lock_init(&net->xfrm.xfrm_policy_lock);
2940 mutex_init(&net->xfrm.xfrm_cfg_mutex);
2941
59c9940e
AD
2942 rv = xfrm_statistics_init(net);
2943 if (rv < 0)
2944 goto out_statistics;
d62ddc21
AD
2945 rv = xfrm_state_init(net);
2946 if (rv < 0)
2947 goto out_state;
2948 rv = xfrm_policy_init(net);
2949 if (rv < 0)
2950 goto out_policy;
b27aeadb
AD
2951 rv = xfrm_sysctl_init(net);
2952 if (rv < 0)
2953 goto out_sysctl;
283bc9f3 2954
d62ddc21
AD
2955 return 0;
2956
b27aeadb
AD
2957out_sysctl:
2958 xfrm_policy_fini(net);
d62ddc21
AD
2959out_policy:
2960 xfrm_state_fini(net);
2961out_state:
59c9940e
AD
2962 xfrm_statistics_fini(net);
2963out_statistics:
d62ddc21
AD
2964 return rv;
2965}
2966
2967static void __net_exit xfrm_net_exit(struct net *net)
2968{
b27aeadb 2969 xfrm_sysctl_fini(net);
d62ddc21
AD
2970 xfrm_policy_fini(net);
2971 xfrm_state_fini(net);
59c9940e 2972 xfrm_statistics_fini(net);
d62ddc21
AD
2973}
2974
2975static struct pernet_operations __net_initdata xfrm_net_ops = {
2976 .init = xfrm_net_init,
2977 .exit = xfrm_net_exit,
2978};
2979
1da177e4
LT
2980void __init xfrm_init(void)
2981{
ec30d78c
FW
2982 int i;
2983
2984 xfrm_pcpu_work = kmalloc_array(NR_CPUS, sizeof(*xfrm_pcpu_work),
2985 GFP_KERNEL);
2986 BUG_ON(!xfrm_pcpu_work);
2987
2988 for (i = 0; i < NR_CPUS; i++)
2989 INIT_WORK(&xfrm_pcpu_work[i], xfrm_pcpu_work_fn);
2990
d62ddc21 2991 register_pernet_subsys(&xfrm_net_ops);
30846090 2992 seqcount_init(&xfrm_policy_hash_generation);
1da177e4
LT
2993 xfrm_input_init();
2994}
2995
ab5f5e8b 2996#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ
2997static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2998 struct audit_buffer *audit_buf)
ab5f5e8b 2999{
875179fa
PM
3000 struct xfrm_sec_ctx *ctx = xp->security;
3001 struct xfrm_selector *sel = &xp->selector;
3002
3003 if (ctx)
ab5f5e8b 3004 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3005 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3006
9b7a787d 3007 switch (sel->family) {
ab5f5e8b 3008 case AF_INET:
21454aaa 3009 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM
3010 if (sel->prefixlen_s != 32)
3011 audit_log_format(audit_buf, " src_prefixlen=%d",
3012 sel->prefixlen_s);
21454aaa 3013 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM
3014 if (sel->prefixlen_d != 32)
3015 audit_log_format(audit_buf, " dst_prefixlen=%d",
3016 sel->prefixlen_d);
ab5f5e8b
JL
3017 break;
3018 case AF_INET6:
5b095d98 3019 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM
3020 if (sel->prefixlen_s != 128)
3021 audit_log_format(audit_buf, " src_prefixlen=%d",
3022 sel->prefixlen_s);
5b095d98 3023 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM
3024 if (sel->prefixlen_d != 128)
3025 audit_log_format(audit_buf, " dst_prefixlen=%d",
3026 sel->prefixlen_d);
ab5f5e8b
JL
3027 break;
3028 }
3029}
3030
2e71029e 3031void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL
3032{
3033 struct audit_buffer *audit_buf;
ab5f5e8b 3034
afeb14b4 3035 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL
3036 if (audit_buf == NULL)
3037 return;
2e71029e 3038 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3039 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL
3040 xfrm_audit_common_policyinfo(xp, audit_buf);
3041 audit_log_end(audit_buf);
3042}
3043EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3044
68277acc 3045void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3046 bool task_valid)
ab5f5e8b
JL
3047{
3048 struct audit_buffer *audit_buf;
ab5f5e8b 3049
afeb14b4 3050 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL
3051 if (audit_buf == NULL)
3052 return;
2e71029e 3053 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3054 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL
3055 xfrm_audit_common_policyinfo(xp, audit_buf);
3056 audit_log_end(audit_buf);
3057}
3058EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3059#endif
3060
80c9abaa 3061#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM
3062static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3063 const struct xfrm_selector *sel_tgt)
80c9abaa
SS
3064{
3065 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3066 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH
3067 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3068 sel_cmp->family) &&
3069 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3070 sel_cmp->family) &&
80c9abaa
SS
3071 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3072 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3073 return true;
80c9abaa
SS
3074 }
3075 } else {
3076 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3077 return true;
80c9abaa
SS
3078 }
3079 }
bc9b35ad 3080 return false;
80c9abaa
SS
3081}
3082
3e94c2dc
WC
3083static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3084 u8 dir, u8 type, struct net *net)
80c9abaa
SS
3085{
3086 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS
3087 struct hlist_head *chain;
3088 u32 priority = ~0U;
3089
9d0380df 3090 spin_lock_bh(&net->xfrm.xfrm_policy_lock);
8d549c4f 3091 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3092 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS
3093 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3094 pol->type == type) {
3095 ret = pol;
3096 priority = ret->priority;
3097 break;
3098 }
3099 }
8d549c4f 3100 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3101 hlist_for_each_entry(pol, chain, bydst) {
8faf491e
LR
3102 if ((pol->priority >= priority) && ret)
3103 break;
3104
80c9abaa 3105 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
8faf491e 3106 pol->type == type) {
80c9abaa
SS
3107 ret = pol;
3108 break;
3109 }
3110 }
3111
586f2eb4 3112 xfrm_pol_hold(ret);
80c9abaa 3113
9d0380df 3114 spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS
3115
3116 return ret;
3117}
3118
dd701754 3119static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS
3120{
3121 int match = 0;
3122
3123 if (t->mode == m->mode && t->id.proto == m->proto &&
3124 (m->reqid == 0 || t->reqid == m->reqid)) {
3125 switch (t->mode) {
3126 case XFRM_MODE_TUNNEL:
3127 case XFRM_MODE_BEET:
70e94e66
YH
3128 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3129 m->old_family) &&
3130 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3131 m->old_family)) {
80c9abaa
SS
3132 match = 1;
3133 }
3134 break;
3135 case XFRM_MODE_TRANSPORT:
3136 /* in case of transport mode, template does not store
3137 any IP addresses, hence we just compare mode and
3138 protocol */
3139 match = 1;
3140 break;
3141 default:
3142 break;
3143 }
3144 }
3145 return match;
3146}
3147
3148/* update endpoint address(es) of template(s) */
3149static int xfrm_policy_migrate(struct xfrm_policy *pol,
3150 struct xfrm_migrate *m, int num_migrate)
3151{
3152 struct xfrm_migrate *mp;
80c9abaa
SS
3153 int i, j, n = 0;
3154
3155 write_lock_bh(&pol->lock);
12a169e7 3156 if (unlikely(pol->walk.dead)) {
80c9abaa
SS
3157 /* target policy has been deleted */
3158 write_unlock_bh(&pol->lock);
3159 return -ENOENT;
3160 }
3161
3162 for (i = 0; i < pol->xfrm_nr; i++) {
3163 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3164 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3165 continue;
3166 n++;
1bfcb10f
HX
3167 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3168 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS
3169 continue;
3170 /* update endpoints */
3171 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3172 sizeof(pol->xfrm_vec[i].id.daddr));
3173 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3174 sizeof(pol->xfrm_vec[i].saddr));
3175 pol->xfrm_vec[i].encap_family = mp->new_family;
3176 /* flush bundles */
80c802f3 3177 atomic_inc(&pol->genid);
80c9abaa
SS
3178 }
3179 }
3180
3181 write_unlock_bh(&pol->lock);
3182
3183 if (!n)
3184 return -ENODATA;
3185
3186 return 0;
3187}
3188
dd701754 3189static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS
3190{
3191 int i, j;
3192
3193 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3194 return -EINVAL;
3195
3196 for (i = 0; i < num_migrate; i++) {
80c9abaa
SS
3197 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3198 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3199 return -EINVAL;
3200
3201 /* check if there is any duplicated entry */
3202 for (j = i + 1; j < num_migrate; j++) {
3203 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3204 sizeof(m[i].old_daddr)) &&
3205 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3206 sizeof(m[i].old_saddr)) &&
3207 m[i].proto == m[j].proto &&
3208 m[i].mode == m[j].mode &&
3209 m[i].reqid == m[j].reqid &&
3210 m[i].old_family == m[j].old_family)
3211 return -EINVAL;
3212 }
3213 }
3214
3215 return 0;
3216}
3217
b4b7c0b3 3218int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3219 struct xfrm_migrate *m, int num_migrate,
4ab47d47
AA
3220 struct xfrm_kmaddress *k, struct net *net,
3221 struct xfrm_encap_tmpl *encap)
80c9abaa
SS
3222{
3223 int i, err, nx_cur = 0, nx_new = 0;
3224 struct xfrm_policy *pol = NULL;
3225 struct xfrm_state *x, *xc;
3226 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3227 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3228 struct xfrm_migrate *mp;
3229
7bab0963 3230 /* Stage 0 - sanity checks */
80c9abaa
SS
3231 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3232 goto out;
3233
7bab0963
VD
3234 if (dir >= XFRM_POLICY_MAX) {
3235 err = -EINVAL;
3236 goto out;
3237 }
3238
80c9abaa 3239 /* Stage 1 - find policy */
8d549c4f 3240 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS
3241 err = -ENOENT;
3242 goto out;
3243 }
3244
3245 /* Stage 2 - find and update state(s) */
3246 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3247 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS
3248 x_cur[nx_cur] = x;
3249 nx_cur++;
4ab47d47
AA
3250 xc = xfrm_state_migrate(x, mp, encap);
3251 if (xc) {
80c9abaa
SS
3252 x_new[nx_new] = xc;
3253 nx_new++;
3254 } else {
3255 err = -ENODATA;
3256 goto restore_state;
3257 }
3258 }
3259 }
3260
3261 /* Stage 3 - update policy */
3262 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3263 goto restore_state;
3264
3265 /* Stage 4 - delete old state(s) */
3266 if (nx_cur) {
3267 xfrm_states_put(x_cur, nx_cur);
3268 xfrm_states_delete(x_cur, nx_cur);
3269 }
3270
3271 /* Stage 5 - announce */
8bafd730 3272 km_migrate(sel, dir, type, m, num_migrate, k, encap);
80c9abaa
SS
3273
3274 xfrm_pol_put(pol);
3275
3276 return 0;
3277out:
3278 return err;
3279
3280restore_state:
3281 if (pol)
3282 xfrm_pol_put(pol);
3283 if (nx_cur)
3284 xfrm_states_put(x_cur, nx_cur);
3285 if (nx_new)
3286 xfrm_states_delete(x_new, nx_new);
3287
3288 return err;
3289}
e610e679 3290EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3291#endif