projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb...
[deliverable/linux.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT		 2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
68277acc 27#include <linux/audit.h>
25ee3286 28#include <net/dst.h>
6ce74ec7 29#include <net/flow.h>
1da177e4
LT		 30#include <net/xfrm.h>
31#include <net/ip.h>
558f82ef
MN		 32#ifdef CONFIG_XFRM_STATISTICS
33#include <net/snmp.h>
34#endif
1da177e4 35
44e36b42
DM		 36#include "xfrm_hash.h"
37
a0073fe1
SK		 38#define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
39#define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
40#define XFRM_MAX_QUEUE_LEN 100
41
b8c203b2
SK		 42struct xfrm_flo {
43 struct dst_entry *dst_orig;
44 u8 flags;
45};
46
418a99ac
PJ		 47static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
48static struct xfrm_policy_afinfo __rcu *xfrm_policy_afinfo[NPROTO]
49 __read_mostly;
1da177e4 50
e18b890b 51static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 52
25ee3286 53static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 54static int stale_bundle(struct dst_entry *dst);
12fdb4d3 55static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 56static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 57
29fa0b30
WY		 58static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
59 int dir);
60
bc9b35ad 61static inline bool
200ce96e 62__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 63{
7e1dc7b6
DM		 64 const struct flowi4 *fl4 = &fl->u.ip4;
65
26bff940
AD		 66 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
67 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 68 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
69 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
70 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
71 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 72}
73
bc9b35ad 74static inline bool
200ce96e 75__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 76{
7e1dc7b6
DM		 77 const struct flowi6 *fl6 = &fl->u.ip6;
78
79 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
80 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
81 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
82 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
83 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
84 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 85}
86
bc9b35ad
DM		 87bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
88 unsigned short family)
77681021
AM		 89{
90 switch (family) {
91 case AF_INET:
92 return __xfrm4_selector_match(sel, fl);
93 case AF_INET6:
94 return __xfrm6_selector_match(sel, fl);
95 }
bc9b35ad 96 return false;
77681021
AM		 97}
98
ef8531b6
ED		 99static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
100{
101 struct xfrm_policy_afinfo *afinfo;
102
103 if (unlikely(family >= NPROTO))
104 return NULL;
105 rcu_read_lock();
106 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
107 if (unlikely(!afinfo))
108 rcu_read_unlock();
109 return afinfo;
110}
111
112static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
113{
114 rcu_read_unlock();
115}
116
c5b3cf46 117static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
6418c4e0
DM		 118 const xfrm_address_t *saddr,
119 const xfrm_address_t *daddr,
9bb182a7
YH		 120 int family)
121{
122 struct xfrm_policy_afinfo *afinfo;
123 struct dst_entry *dst;
124
125 afinfo = xfrm_policy_get_afinfo(family);
126 if (unlikely(afinfo == NULL))
127 return ERR_PTR(-EAFNOSUPPORT);
128
c5b3cf46 129 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 130
131 xfrm_policy_put_afinfo(afinfo);
132
133 return dst;
134}
135
25ee3286 136static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 137 xfrm_address_t *prev_saddr,
138 xfrm_address_t *prev_daddr,
25ee3286 139 int family)
1da177e4 140{
c5b3cf46 141 struct net *net = xs_net(x);
66cdb3ca
HX		 142 xfrm_address_t *saddr = &x->props.saddr;
143 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 144 struct dst_entry *dst;
145
9bb182a7 146 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 147 saddr = x->coaddr;
9bb182a7
YH		 148 daddr = prev_daddr;
149 }
150 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
151 saddr = prev_saddr;
66cdb3ca 152 daddr = x->coaddr;
9bb182a7 153 }
1da177e4 154
c5b3cf46 155 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 156
157 if (!IS_ERR(dst)) {
158 if (prev_saddr != saddr)
159 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
160 if (prev_daddr != daddr)
161 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
162 }
1da177e4 163
66cdb3ca 164 return dst;
1da177e4 165}
1da177e4 166
1da177e4
LT		 167static inline unsigned long make_jiffies(long secs)
168{
169 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
170 return MAX_SCHEDULE_TIMEOUT-1;
171 else
a716c119 172 return secs*HZ;
1da177e4
LT		 173}
174
175static void xfrm_policy_timer(unsigned long data)
176{
3e94c2dc 177 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 178 unsigned long now = get_seconds();
1da177e4
LT		 179 long next = LONG_MAX;
180 int warn = 0;
181 int dir;
182
183 read_lock(&xp->lock);
184
ea2dea9d 185 if (unlikely(xp->walk.dead))
1da177e4
LT		 186 goto out;
187
77d8d7a6 188 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 189
190 if (xp->lft.hard_add_expires_seconds) {
191 long tmo = xp->lft.hard_add_expires_seconds +
192 xp->curlft.add_time - now;
193 if (tmo <= 0)
194 goto expired;
195 if (tmo < next)
196 next = tmo;
197 }
198 if (xp->lft.hard_use_expires_seconds) {
199 long tmo = xp->lft.hard_use_expires_seconds +
200 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
201 if (tmo <= 0)
202 goto expired;
203 if (tmo < next)
204 next = tmo;
205 }
206 if (xp->lft.soft_add_expires_seconds) {
207 long tmo = xp->lft.soft_add_expires_seconds +
208 xp->curlft.add_time - now;
209 if (tmo <= 0) {
210 warn = 1;
211 tmo = XFRM_KM_TIMEOUT;
212 }
213 if (tmo < next)
214 next = tmo;
215 }
216 if (xp->lft.soft_use_expires_seconds) {
217 long tmo = xp->lft.soft_use_expires_seconds +
218 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
219 if (tmo <= 0) {
220 warn = 1;
221 tmo = XFRM_KM_TIMEOUT;
222 }
223 if (tmo < next)
224 next = tmo;
225 }
226
227 if (warn)
6c5c8ca7 228 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 229 if (next != LONG_MAX &&
230 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
231 xfrm_pol_hold(xp);
232
233out:
234 read_unlock(&xp->lock);
235 xfrm_pol_put(xp);
236 return;
237
238expired:
239 read_unlock(&xp->lock);
4666faab 240 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 241 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 242 xfrm_pol_put(xp);
243}
244
fe1a5f03
TT		 245static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
246{
247 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
248
249 if (unlikely(pol->walk.dead))
250 flo = NULL;
251 else
252 xfrm_pol_hold(pol);
253
254 return flo;
255}
256
257static int xfrm_policy_flo_check(struct flow_cache_object *flo)
258{
259 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
260
261 return !pol->walk.dead;
262}
263
264static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
265{
266 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
267}
268
269static const struct flow_cache_ops xfrm_policy_fc_ops = {
270 .get = xfrm_policy_flo_get,
271 .check = xfrm_policy_flo_check,
272 .delete = xfrm_policy_flo_delete,
273};
1da177e4
LT		 274
275/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
276 * SPD calls.
277 */
278
0331b1f3 279struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 280{
281 struct xfrm_policy *policy;
282
0da974f4 283 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 284
285 if (policy) {
0331b1f3 286 write_pnet(&policy->xp_net, net);
12a169e7 287 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 288 INIT_HLIST_NODE(&policy->bydst);
289 INIT_HLIST_NODE(&policy->byidx);
1da177e4 290 rwlock_init(&policy->lock);
2518c7c2 291 atomic_set(&policy->refcnt, 1);
a0073fe1 292 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 293 setup_timer(&policy->timer, xfrm_policy_timer,
294 (unsigned long)policy);
a0073fe1
SK		 295 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
296 (unsigned long)policy);
fe1a5f03 297 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 298 }
299 return policy;
300}
301EXPORT_SYMBOL(xfrm_policy_alloc);
302
303/* Destroy xfrm_policy: descendant resources must be released to this moment. */
304
64c31b3f 305void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 306{
12a169e7 307 BUG_ON(!policy->walk.dead);
1da177e4 308
0659eea9 309 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 310 BUG();
311
03e1ad7b 312 security_xfrm_policy_free(policy->security);
1da177e4
LT		 313 kfree(policy);
314}
64c31b3f 315EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 316
a0073fe1
SK		 317static void xfrm_queue_purge(struct sk_buff_head *list)
318{
319 struct sk_buff *skb;
320
302a50bc 321 while ((skb = skb_dequeue(list)) != NULL)
a0073fe1 322 kfree_skb(skb);
a0073fe1
SK		 323}
324
1da177e4
LT		 325/* Rule must be locked. Release descentant resources, announce
326 * entry dead. The rule must be unlinked from lists to the moment.
327 */
328
329static void xfrm_policy_kill(struct xfrm_policy *policy)
330{
12a169e7 331 policy->walk.dead = 1;
1da177e4 332
285ead17 333 atomic_inc(&policy->genid);
1da177e4 334
e7d8f6cb
SK		 335 if (del_timer(&policy->polq.hold_timer))
336 xfrm_pol_put(policy);
a0073fe1
SK		 337 xfrm_queue_purge(&policy->polq.hold_queue);
338
285ead17
TT		 339 if (del_timer(&policy->timer))
340 xfrm_pol_put(policy);
341
342 xfrm_pol_put(policy);
1da177e4
LT		 343}
344
2518c7c2
DM		 345static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
346
e92303f8 347static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 348{
e92303f8 349 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 350}
351
5f803b58
DM		 352static struct hlist_head *policy_hash_bysel(struct net *net,
353 const struct xfrm_selector *sel,
354 unsigned short family, int dir)
2518c7c2 355{
1121994c 356 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 357 unsigned int hash = __sel_hash(sel, family, hmask);
358
359 return (hash == hmask + 1 ?
1121994c
AD		 360 &net->xfrm.policy_inexact[dir] :
361 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 362}
363
5f803b58
DM		 364static struct hlist_head *policy_hash_direct(struct net *net,
365 const xfrm_address_t *daddr,
366 const xfrm_address_t *saddr,
367 unsigned short family, int dir)
2518c7c2 368{
1121994c 369 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 370 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
371
1121994c 372 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 373}
374
2518c7c2
DM		 375static void xfrm_dst_hash_transfer(struct hlist_head *list,
376 struct hlist_head *ndsttable,
377 unsigned int nhashmask)
378{
b67bfe0d 379 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 380 struct xfrm_policy *pol;
b791160b 381 unsigned int h0 = 0;
2518c7c2 382
b791160b 383redo:
b67bfe0d 384 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 385 unsigned int h;
386
387 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
388 pol->family, nhashmask);
b791160b 389 if (!entry0) {
b67bfe0d 390 hlist_del(&pol->bydst);
b791160b
YH		 391 hlist_add_head(&pol->bydst, ndsttable+h);
392 h0 = h;
393 } else {
394 if (h != h0)
395 continue;
b67bfe0d 396 hlist_del(&pol->bydst);
1d023284 397 hlist_add_behind(&pol->bydst, entry0);
b791160b 398 }
b67bfe0d 399 entry0 = &pol->bydst;
b791160b
YH		 400 }
401 if (!hlist_empty(list)) {
402 entry0 = NULL;
403 goto redo;
2518c7c2
DM		 404 }
405}
406
407static void xfrm_idx_hash_transfer(struct hlist_head *list,
408 struct hlist_head *nidxtable,
409 unsigned int nhashmask)
410{
b67bfe0d 411 struct hlist_node *tmp;
2518c7c2
DM		 412 struct xfrm_policy *pol;
413
b67bfe0d 414 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 415 unsigned int h;
416
417 h = __idx_hash(pol->index, nhashmask);
418 hlist_add_head(&pol->byidx, nidxtable+h);
419 }
420}
421
422static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
423{
424 return ((old_hmask + 1) << 1) - 1;
425}
426
66caf628 427static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 428{
66caf628 429 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 430 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
431 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 432 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 433 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 434 int i;
435
436 if (!ndst)
437 return;
438
283bc9f3 439 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 440
441 for (i = hmask; i >= 0; i--)
442 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
443
66caf628
AD		 444 net->xfrm.policy_bydst[dir].table = ndst;
445 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 446
283bc9f3 447 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 448
44e36b42 449 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 450}
451
66caf628 452static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 453{
66caf628 454 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 455 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
456 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 457 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 458 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 459 int i;
460
461 if (!nidx)
462 return;
463
283bc9f3 464 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 465
466 for (i = hmask; i >= 0; i--)
467 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
468
66caf628
AD		 469 net->xfrm.policy_byidx = nidx;
470 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 471
283bc9f3 472 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 473
44e36b42 474 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 475}
476
66caf628 477static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 478{
66caf628
AD		 479 unsigned int cnt = net->xfrm.policy_count[dir];
480 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 481
482 if (total)
483 *total += cnt;
484
485 if ((hmask + 1) < xfrm_policy_hashmax &&
486 cnt > hmask)
487 return 1;
488
489 return 0;
490}
491
66caf628 492static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 493{
66caf628 494 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 495
496 if ((hmask + 1) < xfrm_policy_hashmax &&
497 total > hmask)
498 return 1;
499
500 return 0;
501}
502
e071041b 503void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 504{
283bc9f3 505 read_lock_bh(&net->xfrm.xfrm_policy_lock);
e071041b
AD		 506 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
507 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
508 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
509 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
510 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
511 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
512 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 513 si->spdhmcnt = xfrm_policy_hashmax;
283bc9f3 514 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
ecfd6b18
JHS		 515}
516EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 517
ecfd6b18 518static DEFINE_MUTEX(hash_resize_mutex);
66caf628 519static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 520{
66caf628 521 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 522 int dir, total;
523
524 mutex_lock(&hash_resize_mutex);
525
526 total = 0;
527 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 528 if (xfrm_bydst_should_resize(net, dir, &total))
529 xfrm_bydst_resize(net, dir);
2518c7c2 530 }
66caf628
AD		 531 if (xfrm_byidx_should_resize(net, total))
532 xfrm_byidx_resize(net, total);
2518c7c2
DM		 533
534 mutex_unlock(&hash_resize_mutex);
535}
536
1da177e4
LT		 537/* Generate new index... KAME seems to generate them ordered by cost
538 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 539static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 540{
1da177e4
LT		 541 static u32 idx_generator;
542
543 for (;;) {
2518c7c2
DM		 544 struct hlist_head *list;
545 struct xfrm_policy *p;
546 u32 idx;
547 int found;
548
e682adf0
FD		 549 if (!index) {
550 idx = (idx_generator | dir);
551 idx_generator += 8;
552 } else {
553 idx = index;
554 index = 0;
555 }
556
1da177e4
LT		 557 if (idx == 0)
558 idx = 8;
1121994c 559 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 560 found = 0;
b67bfe0d 561 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 562 if (p->index == idx) {
563 found = 1;
1da177e4 564 break;
2518c7c2 565 }
1da177e4 566 }
2518c7c2 567 if (!found)
1da177e4
LT		 568 return idx;
569 }
570}
571
2518c7c2
DM		 572static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
573{
574 u32 *p1 = (u32 *) s1;
575 u32 *p2 = (u32 *) s2;
576 int len = sizeof(struct xfrm_selector) / sizeof(u32);
577 int i;
578
579 for (i = 0; i < len; i++) {
580 if (p1[i] != p2[i])
581 return 1;
582 }
583
584 return 0;
585}
586
a0073fe1
SK		 587static void xfrm_policy_requeue(struct xfrm_policy *old,
588 struct xfrm_policy *new)
589{
590 struct xfrm_policy_queue *pq = &old->polq;
591 struct sk_buff_head list;
592
593 __skb_queue_head_init(&list);
594
595 spin_lock_bh(&pq->hold_queue.lock);
596 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 597 if (del_timer(&pq->hold_timer))
598 xfrm_pol_put(old);
a0073fe1
SK		 599 spin_unlock_bh(&pq->hold_queue.lock);
600
601 if (skb_queue_empty(&list))
602 return;
603
604 pq = &new->polq;
605
606 spin_lock_bh(&pq->hold_queue.lock);
607 skb_queue_splice(&list, &pq->hold_queue);
608 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 609 if (!mod_timer(&pq->hold_timer, jiffies))
610 xfrm_pol_hold(new);
a0073fe1
SK		 611 spin_unlock_bh(&pq->hold_queue.lock);
612}
613
7cb8a939
SK		 614static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
615 struct xfrm_policy *pol)
616{
617 u32 mark = policy->mark.v & policy->mark.m;
618
619 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
620 return true;
621
622 if ((mark & pol->mark.m) == pol->mark.v &&
623 policy->priority == pol->priority)
624 return true;
625
626 return false;
627}
628
1da177e4
LT		 629int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
630{
1121994c 631 struct net *net = xp_net(policy);
2518c7c2
DM		 632 struct xfrm_policy *pol;
633 struct xfrm_policy *delpol;
634 struct hlist_head *chain;
b67bfe0d 635 struct hlist_node *newpos;
1da177e4 636
283bc9f3 637 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 638 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 639 delpol = NULL;
640 newpos = NULL;
b67bfe0d 641 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 642 if (pol->type == policy->type &&
2518c7c2 643 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 644 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 645 xfrm_sec_ctx_match(pol->security, policy->security) &&
646 !WARN_ON(delpol)) {
1da177e4 647 if (excl) {
283bc9f3 648 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 649 return -EEXIST;
650 }
1da177e4
LT		 651 delpol = pol;
652 if (policy->priority > pol->priority)
653 continue;
654 } else if (policy->priority >= pol->priority) {
a6c7ab55 655 newpos = &pol->bydst;
1da177e4
LT		 656 continue;
657 }
1da177e4
LT		 658 if (delpol)
659 break;
1da177e4
LT		 660 }
661 if (newpos)
1d023284 662 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 663 else
664 hlist_add_head(&policy->bydst, chain);
1da177e4 665 xfrm_pol_hold(policy);
1121994c 666 net->xfrm.policy_count[dir]++;
ca925cf1 667 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 668
669 /* After previous checking, family can either be AF_INET or AF_INET6 */
670 if (policy->family == AF_INET)
671 rt_genid_bump_ipv4(net);
672 else
673 rt_genid_bump_ipv6(net);
674
a0073fe1
SK		 675 if (delpol) {
676 xfrm_policy_requeue(delpol, policy);
29fa0b30 677 __xfrm_policy_unlink(delpol, dir);
a0073fe1 678 }
e682adf0 679 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 680 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 681 policy->curlft.add_time = get_seconds();
1da177e4
LT		 682 policy->curlft.use_time = 0;
683 if (!mod_timer(&policy->timer, jiffies + HZ))
684 xfrm_pol_hold(policy);
1121994c 685 list_add(&policy->walk.all, &net->xfrm.policy_all);
283bc9f3 686 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 687
9b78a82c 688 if (delpol)
1da177e4 689 xfrm_policy_kill(delpol);
1121994c
AD		 690 else if (xfrm_bydst_should_resize(net, dir, NULL))
691 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 692
1da177e4
LT		 693 return 0;
694}
695EXPORT_SYMBOL(xfrm_policy_insert);
696
8ca2e93b
JHS		 697struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
698 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 699 struct xfrm_sec_ctx *ctx, int delete,
700 int *err)
1da177e4 701{
2518c7c2
DM		 702 struct xfrm_policy *pol, *ret;
703 struct hlist_head *chain;
1da177e4 704
ef41aaa0 705 *err = 0;
283bc9f3 706 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 707 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 708 ret = NULL;
b67bfe0d 709 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 710 if (pol->type == type &&
34f8d884 711 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 712 !selector_cmp(sel, &pol->selector) &&
713 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 714 xfrm_pol_hold(pol);
2518c7c2 715 if (delete) {
03e1ad7b
PM		 716 *err = security_xfrm_policy_delete(
717 pol->security);
ef41aaa0 718 if (*err) {
283bc9f3 719 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 720 return pol;
721 }
29fa0b30 722 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 723 }
724 ret = pol;
1da177e4
LT		 725 break;
726 }
727 }
283bc9f3 728 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 729
fe1a5f03 730 if (ret && delete)
2518c7c2 731 xfrm_policy_kill(ret);
2518c7c2 732 return ret;
1da177e4 733}
df71837d 734EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 735
8ca2e93b
JHS		 736struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
737 int dir, u32 id, int delete, int *err)
1da177e4 738{
2518c7c2
DM		 739 struct xfrm_policy *pol, *ret;
740 struct hlist_head *chain;
1da177e4 741
b5505c6e
HX		 742 *err = -ENOENT;
743 if (xfrm_policy_id2dir(id) != dir)
744 return NULL;
745
ef41aaa0 746 *err = 0;
283bc9f3 747 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 748 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 749 ret = NULL;
b67bfe0d 750 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 751 if (pol->type == type && pol->index == id &&
752 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 753 xfrm_pol_hold(pol);
2518c7c2 754 if (delete) {
03e1ad7b
PM		 755 *err = security_xfrm_policy_delete(
756 pol->security);
ef41aaa0 757 if (*err) {
283bc9f3 758 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 759 return pol;
760 }
29fa0b30 761 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 762 }
763 ret = pol;
1da177e4
LT		 764 break;
765 }
766 }
283bc9f3 767 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 768
fe1a5f03 769 if (ret && delete)
2518c7c2 770 xfrm_policy_kill(ret);
2518c7c2 771 return ret;
1da177e4
LT		 772}
773EXPORT_SYMBOL(xfrm_policy_byid);
774
4aa2e62c
JL		 775#ifdef CONFIG_SECURITY_NETWORK_XFRM
776static inline int
2e71029e 777xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 778{
4aa2e62c
JL		 779 int dir, err = 0;
780
781 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
782 struct xfrm_policy *pol;
4aa2e62c
JL		 783 int i;
784
b67bfe0d 785 hlist_for_each_entry(pol,
33ffbbd5 786 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 787 if (pol->type != type)
788 continue;
03e1ad7b 789 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 790 if (err) {
2e71029e 791 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 792 return err;
793 }
7dc12d6d 794 }
33ffbbd5 795 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 796 hlist_for_each_entry(pol,
33ffbbd5 797 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 798 bydst) {
799 if (pol->type != type)
800 continue;
03e1ad7b
PM		 801 err = security_xfrm_policy_delete(
802 pol->security);
4aa2e62c 803 if (err) {
ab5f5e8b 804 xfrm_audit_policy_delete(pol, 0,
2e71029e 805 task_valid);
4aa2e62c
JL		 806 return err;
807 }
808 }
809 }
810 }
811 return err;
812}
813#else
814static inline int
2e71029e 815xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 816{
817 return 0;
818}
819#endif
820
2e71029e 821int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 822{
2f1eb65f 823 int dir, err = 0, cnt = 0;
1da177e4 824
283bc9f3 825 write_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 826
2e71029e 827 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 828 if (err)
829 goto out;
830
1da177e4 831 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 832 struct xfrm_policy *pol;
29fa0b30 833 int i;
2518c7c2
DM		 834
835 again1:
b67bfe0d 836 hlist_for_each_entry(pol,
33ffbbd5 837 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 838 if (pol->type != type)
839 continue;
ea2dea9d 840 __xfrm_policy_unlink(pol, dir);
283bc9f3 841 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 842 cnt++;
1da177e4 843
2e71029e 844 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 845
2518c7c2 846 xfrm_policy_kill(pol);
1da177e4 847
283bc9f3 848 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 849 goto again1;
850 }
851
33ffbbd5 852 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 853 again2:
b67bfe0d 854 hlist_for_each_entry(pol,
33ffbbd5 855 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 856 bydst) {
857 if (pol->type != type)
858 continue;
ea2dea9d 859 __xfrm_policy_unlink(pol, dir);
283bc9f3 860 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 861 cnt++;
2518c7c2 862
2e71029e 863 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 864 xfrm_policy_kill(pol);
865
283bc9f3 866 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 867 goto again2;
868 }
1da177e4 869 }
2518c7c2 870
1da177e4 871 }
2f1eb65f
JHS		 872 if (!cnt)
873 err = -ESRCH;
4aa2e62c 874out:
283bc9f3 875 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 876 return err;
1da177e4
LT		 877}
878EXPORT_SYMBOL(xfrm_policy_flush);
879
cdcbca7c 880int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 881 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 882 void *data)
883{
12a169e7
HX		 884 struct xfrm_policy *pol;
885 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 886 int error = 0;
887
888 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
889 walk->type != XFRM_POLICY_TYPE_ANY)
890 return -EINVAL;
1da177e4 891
12a169e7 892 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 893 return 0;
894
283bc9f3 895 write_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 896 if (list_empty(&walk->walk.all))
cdcbca7c 897 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 898 else
899 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 900 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 901 if (x->dead)
4c563f76 902 continue;
12a169e7
HX		 903 pol = container_of(x, struct xfrm_policy, walk);
904 if (walk->type != XFRM_POLICY_TYPE_ANY &&
905 walk->type != pol->type)
906 continue;
907 error = func(pol, xfrm_policy_id2dir(pol->index),
908 walk->seq, data);
909 if (error) {
910 list_move_tail(&walk->walk.all, &x->all);
911 goto out;
2518c7c2 912 }
12a169e7 913 walk->seq++;
1da177e4 914 }
12a169e7 915 if (walk->seq == 0) {
baf5d743
JHS		 916 error = -ENOENT;
917 goto out;
918 }
12a169e7 919 list_del_init(&walk->walk.all);
1da177e4 920out:
283bc9f3 921 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 922 return error;
923}
924EXPORT_SYMBOL(xfrm_policy_walk);
925
12a169e7
HX		 926void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
927{
928 INIT_LIST_HEAD(&walk->walk.all);
929 walk->walk.dead = 1;
930 walk->type = type;
931 walk->seq = 0;
932}
933EXPORT_SYMBOL(xfrm_policy_walk_init);
934
283bc9f3 935void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 936{
937 if (list_empty(&walk->walk.all))
938 return;
939
283bc9f3 940 write_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 941 list_del(&walk->walk.all);
283bc9f3 942 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 943}
944EXPORT_SYMBOL(xfrm_policy_walk_done);
945
134b0fc5
JM		 946/*
947 * Find policy to apply to this flow.
948 *
949 * Returns 0 if policy found, else an -errno.
950 */
f299d557
DM		 951static int xfrm_policy_match(const struct xfrm_policy *pol,
952 const struct flowi *fl,
2518c7c2 953 u8 type, u16 family, int dir)
1da177e4 954{
f299d557 955 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 956 int ret = -ESRCH;
957 bool match;
1da177e4 958
2518c7c2 959 if (pol->family != family ||
1d28f42c 960 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 961 pol->type != type)
134b0fc5 962 return ret;
1da177e4 963
2518c7c2 964 match = xfrm_selector_match(sel, fl, family);
134b0fc5 965 if (match)
1d28f42c 966 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 967 dir);
2518c7c2 968
134b0fc5 969 return ret;
2518c7c2 970}
1da177e4 971
52479b62 972static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 973 const struct flowi *fl,
2518c7c2
DM		 974 u16 family, u8 dir)
975{
134b0fc5 976 int err;
2518c7c2 977 struct xfrm_policy *pol, *ret;
0b597e7e 978 const xfrm_address_t *daddr, *saddr;
2518c7c2 979 struct hlist_head *chain;
acba48e1 980 u32 priority = ~0U;
df71837d 981
2518c7c2
DM		 982 daddr = xfrm_flowi_daddr(fl, family);
983 saddr = xfrm_flowi_saddr(fl, family);
984 if (unlikely(!daddr || !saddr))
985 return NULL;
986
283bc9f3 987 read_lock_bh(&net->xfrm.xfrm_policy_lock);
52479b62 988 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 989 ret = NULL;
b67bfe0d 990 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 991 err = xfrm_policy_match(pol, fl, type, family, dir);
992 if (err) {
993 if (err == -ESRCH)
994 continue;
995 else {
996 ret = ERR_PTR(err);
997 goto fail;
998 }
999 } else {
2518c7c2 1000 ret = pol;
acba48e1 1001 priority = ret->priority;
2518c7c2
DM		 1002 break;
1003 }
1004 }
52479b62 1005 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1006 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1007 err = xfrm_policy_match(pol, fl, type, family, dir);
1008 if (err) {
1009 if (err == -ESRCH)
1010 continue;
1011 else {
1012 ret = ERR_PTR(err);
1013 goto fail;
1014 }
1015 } else if (pol->priority < priority) {
acba48e1
DM		 1016 ret = pol;
1017 break;
1da177e4
LT		 1018 }
1019 }
acba48e1
DM		 1020 if (ret)
1021 xfrm_pol_hold(ret);
134b0fc5 1022fail:
283bc9f3 1023 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
4e81bb83 1024
2518c7c2 1025 return ret;
4e81bb83
MN		 1026}
1027
80c802f3 1028static struct xfrm_policy *
73ff93cd 1029__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1030{
1031#ifdef CONFIG_XFRM_SUB_POLICY
1032 struct xfrm_policy *pol;
1033
1034 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1035 if (pol != NULL)
1036 return pol;
1037#endif
1038 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1039}
1040
b5fb82c4
BZ		 1041static int flow_to_policy_dir(int dir)
1042{
1043 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1044 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1045 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1046 return dir;
1047
1048 switch (dir) {
1049 default:
1050 case FLOW_DIR_IN:
1051 return XFRM_POLICY_IN;
1052 case FLOW_DIR_OUT:
1053 return XFRM_POLICY_OUT;
1054 case FLOW_DIR_FWD:
1055 return XFRM_POLICY_FWD;
1056 }
1057}
1058
fe1a5f03 1059static struct flow_cache_object *
dee9f4bc 1060xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1061 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1062{
1063 struct xfrm_policy *pol;
fe1a5f03
TT		 1064
1065 if (old_obj)
1066 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1067
b5fb82c4 1068 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1069 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1070 return ERR_CAST(pol);
fe1a5f03 1071
fe1a5f03
TT		 1072 /* Resolver returns two references:
1073 * one for cache and one for caller of flow_cache_lookup() */
1074 xfrm_pol_hold(pol);
1075
1076 return &pol->flo;
1da177e4
LT		 1077}
1078
df71837d
TJ		 1079static inline int policy_to_flow_dir(int dir)
1080{
1081 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1082 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1083 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1084 return dir;
1085 switch (dir) {
1086 default:
1087 case XFRM_POLICY_IN:
1088 return FLOW_DIR_IN;
1089 case XFRM_POLICY_OUT:
1090 return FLOW_DIR_OUT;
1091 case XFRM_POLICY_FWD:
1092 return FLOW_DIR_FWD;
3ff50b79 1093 }
df71837d
TJ		 1094}
1095
dee9f4bc
DM		 1096static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1097 const struct flowi *fl)
1da177e4
LT		 1098{
1099 struct xfrm_policy *pol;
283bc9f3 1100 struct net *net = sock_net(sk);
1da177e4 1101
283bc9f3 1102 read_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1103 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1104 bool match = xfrm_selector_match(&pol->selector, fl,
1105 sk->sk_family);
a716c119 1106 int err = 0;
df71837d 1107
3bccfbc7 1108 if (match) {
34f8d884
JHS		 1109 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1110 pol = NULL;
1111 goto out;
1112 }
03e1ad7b 1113 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1114 fl->flowi_secid,
03e1ad7b 1115 policy_to_flow_dir(dir));
3bccfbc7
VY		 1116 if (!err)
1117 xfrm_pol_hold(pol);
1118 else if (err == -ESRCH)
1119 pol = NULL;
1120 else
1121 pol = ERR_PTR(err);
1122 } else
1da177e4
LT		 1123 pol = NULL;
1124 }
34f8d884 1125out:
283bc9f3 1126 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1127 return pol;
1128}
1129
1130static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1131{
98806f75 1132 struct net *net = xp_net(pol);
1121994c 1133 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1134 pol->family, dir);
4e81bb83 1135
98806f75 1136 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1137 hlist_add_head(&pol->bydst, chain);
e92303f8 1138 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1139 net->xfrm.policy_count[dir]++;
1da177e4 1140 xfrm_pol_hold(pol);
2518c7c2 1141
98806f75
AD		 1142 if (xfrm_bydst_should_resize(net, dir, NULL))
1143 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1144}
1145
1146static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1147 int dir)
1148{
98806f75
AD		 1149 struct net *net = xp_net(pol);
1150
2518c7c2
DM		 1151 if (hlist_unhashed(&pol->bydst))
1152 return NULL;
1da177e4 1153
3a9016f9 1154 hlist_del_init(&pol->bydst);
2518c7c2 1155 hlist_del(&pol->byidx);
12a169e7 1156 list_del(&pol->walk.all);
98806f75 1157 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1158
1159 return pol;
1da177e4
LT		 1160}
1161
4666faab 1162int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1163{
283bc9f3
FD		 1164 struct net *net = xp_net(pol);
1165
1166 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1167 pol = __xfrm_policy_unlink(pol, dir);
283bc9f3 1168 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1169 if (pol) {
1da177e4 1170 xfrm_policy_kill(pol);
4666faab 1171 return 0;
1da177e4 1172 }
4666faab 1173 return -ENOENT;
1da177e4 1174}
a70fcb0b 1175EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1176
1177int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1178{
1121994c 1179 struct net *net = xp_net(pol);
1da177e4
LT		 1180 struct xfrm_policy *old_pol;
1181
4e81bb83
MN		 1182#ifdef CONFIG_XFRM_SUB_POLICY
1183 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1184 return -EINVAL;
1185#endif
1186
283bc9f3 1187 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1188 old_pol = sk->sk_policy[dir];
1189 sk->sk_policy[dir] = pol;
1190 if (pol) {
9d729f72 1191 pol->curlft.add_time = get_seconds();
e682adf0 1192 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
1da177e4
LT		 1193 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1194 }
a0073fe1
SK		 1195 if (old_pol) {
1196 if (pol)
1197 xfrm_policy_requeue(old_pol, pol);
1198
ea2dea9d
TT		 1199 /* Unlinking succeeds always. This is the only function
1200 * allowed to delete or replace socket policy.
1201 */
1da177e4 1202 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
a0073fe1 1203 }
283bc9f3 1204 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1205
1206 if (old_pol) {
1207 xfrm_policy_kill(old_pol);
1208 }
1209 return 0;
1210}
1211
d3e40a9f 1212static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1213{
0331b1f3 1214 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1215 struct net *net = xp_net(old);
1da177e4
LT		 1216
1217 if (newp) {
1218 newp->selector = old->selector;
03e1ad7b
PM		 1219 if (security_xfrm_policy_clone(old->security,
1220 &newp->security)) {
df71837d
TJ		 1221 kfree(newp);
1222 return NULL; /* ENOMEM */
1223 }
1da177e4
LT		 1224 newp->lft = old->lft;
1225 newp->curlft = old->curlft;
fb977e2c 1226 newp->mark = old->mark;
1da177e4
LT		 1227 newp->action = old->action;
1228 newp->flags = old->flags;
1229 newp->xfrm_nr = old->xfrm_nr;
1230 newp->index = old->index;
4e81bb83 1231 newp->type = old->type;
1da177e4
LT		 1232 memcpy(newp->xfrm_vec, old->xfrm_vec,
1233 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
283bc9f3 1234 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1235 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
283bc9f3 1236 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1237 xfrm_pol_put(newp);
1238 }
1239 return newp;
1240}
1241
1242int __xfrm_sk_clone_policy(struct sock *sk)
1243{
1244 struct xfrm_policy *p0 = sk->sk_policy[0],
1245 *p1 = sk->sk_policy[1];
1246
1247 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1248 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1249 return -ENOMEM;
1250 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1251 return -ENOMEM;
1252 return 0;
1253}
1254
a1e59abf 1255static int
fbda33b2 1256xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1257 unsigned short family)
1258{
1259 int err;
1260 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1261
1262 if (unlikely(afinfo == NULL))
1263 return -EINVAL;
fbda33b2 1264 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1265 xfrm_policy_put_afinfo(afinfo);
1266 return err;
1267}
1268
1da177e4
LT		 1269/* Resolve list of templates for the flow, given policy. */
1270
1271static int
a6c2e611
DM		 1272xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1273 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1274{
fbda33b2 1275 struct net *net = xp_net(policy);
1da177e4
LT		 1276 int nx;
1277 int i, error;
1278 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1279 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1280 xfrm_address_t tmp;
1da177e4 1281
9b7a787d 1282 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1283 struct xfrm_state *x;
1284 xfrm_address_t *remote = daddr;
1285 xfrm_address_t *local = saddr;
1286 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1287
48b8d783
JK		 1288 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1289 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1290 remote = &tmpl->id.daddr;
1291 local = &tmpl->saddr;
8444cf71
TE		 1292 if (xfrm_addr_any(local, tmpl->encap_family)) {
1293 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1294 if (error)
1295 goto fail;
1296 local = &tmp;
1297 }
1da177e4
LT		 1298 }
1299
1300 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1301
1302 if (x && x->km.state == XFRM_STATE_VALID) {
1303 xfrm[nx++] = x;
1304 daddr = remote;
1305 saddr = local;
1306 continue;
1307 }
1308 if (x) {
1309 error = (x->km.state == XFRM_STATE_ERROR ?
1310 -EINVAL : -EAGAIN);
1311 xfrm_state_put(x);
42054569 1312 } else if (error == -ESRCH) {
a4322266 1313 error = -EAGAIN;
42054569 1314 }
1da177e4
LT		 1315
1316 if (!tmpl->optional)
1317 goto fail;
1318 }
1319 return nx;
1320
1321fail:
9b7a787d 1322 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1323 xfrm_state_put(xfrm[nx]);
1324 return error;
1325}
1326
4e81bb83 1327static int
a6c2e611
DM		 1328xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1329 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1330{
41a49cc3
MN		 1331 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1332 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1333 int cnx = 0;
1334 int error;
1335 int ret;
1336 int i;
1337
1338 for (i = 0; i < npols; i++) {
1339 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1340 error = -ENOBUFS;
1341 goto fail;
1342 }
41a49cc3
MN		 1343
1344 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1345 if (ret < 0) {
1346 error = ret;
1347 goto fail;
1348 } else
1349 cnx += ret;
1350 }
1351
41a49cc3
MN		 1352 /* found states are sorted for outbound processing */
1353 if (npols > 1)
1354 xfrm_state_sort(xfrm, tpp, cnx, family);
1355
4e81bb83
MN		 1356 return cnx;
1357
1358 fail:
9b7a787d 1359 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1360 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1361 return error;
1362
1363}
1364
1da177e4
LT		 1365/* Check that the bundle accepts the flow and its components are
1366 * still valid.
1367 */
1368
05d84025 1369static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1370{
1371 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1372 int tos;
1da177e4 1373
25ee3286
HX		 1374 if (!afinfo)
1375 return -EINVAL;
1376
1377 tos = afinfo->get_tos(fl);
1378
1379 xfrm_policy_put_afinfo(afinfo);
1380
1381 return tos;
1382}
1383
80c802f3
TT		 1384static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1385{
1386 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1387 struct dst_entry *dst = &xdst->u.dst;
1388
1389 if (xdst->route == NULL) {
1390 /* Dummy bundle - if it has xfrms we were not
1391 * able to build bundle as template resolution failed.
1392 * It means we need to try again resolving. */
1393 if (xdst->num_xfrms > 0)
1394 return NULL;
a0073fe1
SK		 1395 } else if (dst->flags & DST_XFRM_QUEUE) {
1396 return NULL;
80c802f3
TT		 1397 } else {
1398 /* Real bundle */
1399 if (stale_bundle(dst))
1400 return NULL;
1401 }
1402
1403 dst_hold(dst);
1404 return flo;
1405}
1406
1407static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1408{
1409 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1410 struct dst_entry *dst = &xdst->u.dst;
1411
1412 if (!xdst->route)
1413 return 0;
1414 if (stale_bundle(dst))
1415 return 0;
1416
1417 return 1;
1418}
1419
1420static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1421{
1422 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1423 struct dst_entry *dst = &xdst->u.dst;
1424
1425 dst_free(dst);
1426}
1427
1428static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1429 .get = xfrm_bundle_flo_get,
1430 .check = xfrm_bundle_flo_check,
1431 .delete = xfrm_bundle_flo_delete,
1432};
1433
d7c7544c 1434static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1435{
1da177e4 1436 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1437 struct dst_ops *dst_ops;
25ee3286
HX		 1438 struct xfrm_dst *xdst;
1439
1440 if (!afinfo)
1441 return ERR_PTR(-EINVAL);
1442
d7c7544c
AD		 1443 switch (family) {
1444 case AF_INET:
1445 dst_ops = &net->xfrm.xfrm4_dst_ops;
1446 break;
dfd56b8b 1447#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1448 case AF_INET6:
1449 dst_ops = &net->xfrm.xfrm6_dst_ops;
1450 break;
1451#endif
1452 default:
1453 BUG();
1454 }
f5b0a874 1455 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1456
d4cae562 1457 if (likely(xdst)) {
141e369d
SK		 1458 struct dst_entry *dst = &xdst->u.dst;
1459
1460 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1461 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1462 if (afinfo->init_dst)
1463 afinfo->init_dst(net, xdst);
d4cae562 1464 } else
0b150932 1465 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1466
d4cae562
MB		 1467 xfrm_policy_put_afinfo(afinfo);
1468
25ee3286
HX		 1469 return xdst;
1470}
1471
a1b05140
MN		 1472static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1473 int nfheader_len)
1474{
1475 struct xfrm_policy_afinfo *afinfo =
1476 xfrm_policy_get_afinfo(dst->ops->family);
1477 int err;
1478
1479 if (!afinfo)
1480 return -EINVAL;
1481
1482 err = afinfo->init_path(path, dst, nfheader_len);
1483
1484 xfrm_policy_put_afinfo(afinfo);
1485
1486 return err;
1487}
1488
87c1e12b 1489static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1490 const struct flowi *fl)
25ee3286
HX		 1491{
1492 struct xfrm_policy_afinfo *afinfo =
1493 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1494 int err;
1495
1496 if (!afinfo)
1da177e4 1497 return -EINVAL;
25ee3286 1498
87c1e12b 1499 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1500
1da177e4 1501 xfrm_policy_put_afinfo(afinfo);
25ee3286 1502
1da177e4
LT		 1503 return err;
1504}
1505
80c802f3 1506
25ee3286
HX		 1507/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1508 * all the metrics... Shortly, bundle a bundle.
1509 */
1510
1511static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1512 struct xfrm_state **xfrm, int nx,
98313ada 1513 const struct flowi *fl,
25ee3286
HX		 1514 struct dst_entry *dst)
1515{
d7c7544c 1516 struct net *net = xp_net(policy);
25ee3286
HX		 1517 unsigned long now = jiffies;
1518 struct net_device *dev;
43a4dea4 1519 struct xfrm_mode *inner_mode;
25ee3286
HX		 1520 struct dst_entry *dst_prev = NULL;
1521 struct dst_entry *dst0 = NULL;
1522 int i = 0;
1523 int err;
1524 int header_len = 0;
a1b05140 1525 int nfheader_len = 0;
25ee3286
HX		 1526 int trailer_len = 0;
1527 int tos;
1528 int family = policy->selector.family;
9bb182a7
YH		 1529 xfrm_address_t saddr, daddr;
1530
1531 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1532
1533 tos = xfrm_get_tos(fl, family);
1534 err = tos;
1535 if (tos < 0)
1536 goto put_states;
1537
1538 dst_hold(dst);
1539
1540 for (; i < nx; i++) {
d7c7544c 1541 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1542 struct dst_entry *dst1 = &xdst->u.dst;
1543
1544 err = PTR_ERR(xdst);
1545 if (IS_ERR(xdst)) {
1546 dst_release(dst);
1547 goto put_states;
1548 }
1549
43a4dea4
SK		 1550 if (xfrm[i]->sel.family == AF_UNSPEC) {
1551 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1552 xfrm_af2proto(family));
1553 if (!inner_mode) {
1554 err = -EAFNOSUPPORT;
1555 dst_release(dst);
1556 goto put_states;
1557 }
1558 } else
1559 inner_mode = xfrm[i]->inner_mode;
1560
25ee3286
HX		 1561 if (!dst_prev)
1562 dst0 = dst1;
1563 else {
1564 dst_prev->child = dst_clone(dst1);
1565 dst1->flags |= DST_NOHASH;
1566 }
1567
1568 xdst->route = dst;
defb3519 1569 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1570
1571 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1572 family = xfrm[i]->props.family;
9bb182a7
YH		 1573 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1574 family);
25ee3286
HX		 1575 err = PTR_ERR(dst);
1576 if (IS_ERR(dst))
1577 goto put_states;
1578 } else
1579 dst_hold(dst);
1580
1581 dst1->xfrm = xfrm[i];
80c802f3 1582 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1583
f5b0a874 1584 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1585 dst1->flags |= DST_HOST;
1586 dst1->lastuse = now;
1587
1588 dst1->input = dst_discard;
43a4dea4 1589 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1590
1591 dst1->next = dst_prev;
1592 dst_prev = dst1;
1593
1594 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1595 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1596 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1597 trailer_len += xfrm[i]->props.trailer_len;
1598 }
1599
1600 dst_prev->child = dst;
1601 dst0->path = dst;
1602
1603 err = -ENODEV;
1604 dev = dst->dev;
1605 if (!dev)
1606 goto free_dst;
1607
a1b05140 1608 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1609 xfrm_init_pmtu(dst_prev);
1610
1611 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1612 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1613
87c1e12b 1614 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1615 if (err)
1616 goto free_dst;
1617
1618 dst_prev->header_len = header_len;
1619 dst_prev->trailer_len = trailer_len;
1620 header_len -= xdst->u.dst.xfrm->props.header_len;
1621 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1622 }
1623
1624out:
1625 return dst0;
1626
1627put_states:
1628 for (; i < nx; i++)
1629 xfrm_state_put(xfrm[i]);
1630free_dst:
1631 if (dst0)
1632 dst_free(dst0);
1633 dst0 = ERR_PTR(err);
1634 goto out;
1635}
1636
da7c224b 1637#ifdef CONFIG_XFRM_SUB_POLICY
be7928d2 1638static int xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1639{
1640 if (!*target) {
1641 *target = kmalloc(size, GFP_ATOMIC);
1642 if (!*target)
1643 return -ENOMEM;
1644 }
be7928d2 1645
157bfc25
MN		 1646 memcpy(*target, src, size);
1647 return 0;
1648}
da7c224b 1649#endif
157bfc25 1650
be7928d2
DB		 1651static int xfrm_dst_update_parent(struct dst_entry *dst,
1652 const struct xfrm_selector *sel)
157bfc25
MN		 1653{
1654#ifdef CONFIG_XFRM_SUB_POLICY
1655 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1656 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1657 sel, sizeof(*sel));
1658#else
1659 return 0;
1660#endif
1661}
1662
be7928d2
DB		 1663static int xfrm_dst_update_origin(struct dst_entry *dst,
1664 const struct flowi *fl)
157bfc25
MN		 1665{
1666#ifdef CONFIG_XFRM_SUB_POLICY
1667 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1668 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1669#else
1670 return 0;
1671#endif
1672}
1da177e4 1673
73ff93cd 1674static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1675 struct xfrm_policy **pols,
1676 int *num_pols, int *num_xfrms)
1677{
1678 int i;
1679
1680 if (*num_pols == 0 || !pols[0]) {
1681 *num_pols = 0;
1682 *num_xfrms = 0;
1683 return 0;
1684 }
1685 if (IS_ERR(pols[0]))
1686 return PTR_ERR(pols[0]);
1687
1688 *num_xfrms = pols[0]->xfrm_nr;
1689
1690#ifdef CONFIG_XFRM_SUB_POLICY
1691 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1692 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1693 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1694 XFRM_POLICY_TYPE_MAIN,
1695 fl, family,
1696 XFRM_POLICY_OUT);
1697 if (pols[1]) {
1698 if (IS_ERR(pols[1])) {
1699 xfrm_pols_put(pols, *num_pols);
1700 return PTR_ERR(pols[1]);
1701 }
02d0892f 1702 (*num_pols)++;
80c802f3
TT		 1703 (*num_xfrms) += pols[1]->xfrm_nr;
1704 }
1705 }
1706#endif
1707 for (i = 0; i < *num_pols; i++) {
1708 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1709 *num_xfrms = -1;
1710 break;
1711 }
1712 }
1713
1714 return 0;
1715
1716}
1717
1718static struct xfrm_dst *
1719xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1720 const struct flowi *fl, u16 family,
80c802f3
TT		 1721 struct dst_entry *dst_orig)
1722{
1723 struct net *net = xp_net(pols[0]);
1724 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1725 struct dst_entry *dst;
1726 struct xfrm_dst *xdst;
1727 int err;
1728
1729 /* Try to instantiate a bundle */
1730 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1731 if (err <= 0) {
1732 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1733 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1734 return ERR_PTR(err);
1735 }
1736
1737 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1738 if (IS_ERR(dst)) {
1739 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1740 return ERR_CAST(dst);
1741 }
1742
1743 xdst = (struct xfrm_dst *)dst;
1744 xdst->num_xfrms = err;
1745 if (num_pols > 1)
1746 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1747 else
1748 err = xfrm_dst_update_origin(dst, fl);
1749 if (unlikely(err)) {
1750 dst_free(dst);
1751 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1752 return ERR_PTR(err);
1753 }
1754
1755 xdst->num_pols = num_pols;
3e94c2dc 1756 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1757 xdst->policy_genid = atomic_read(&pols[0]->genid);
1758
1759 return xdst;
1760}
1761
a0073fe1
SK		 1762static void xfrm_policy_queue_process(unsigned long arg)
1763{
1764 int err = 0;
1765 struct sk_buff *skb;
1766 struct sock *sk;
1767 struct dst_entry *dst;
a0073fe1
SK		 1768 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1769 struct xfrm_policy_queue *pq = &pol->polq;
1770 struct flowi fl;
1771 struct sk_buff_head list;
1772
1773 spin_lock(&pq->hold_queue.lock);
1774 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1775 if (!skb) {
1776 spin_unlock(&pq->hold_queue.lock);
1777 goto out;
1778 }
a0073fe1
SK		 1779 dst = skb_dst(skb);
1780 sk = skb->sk;
1781 xfrm_decode_session(skb, &fl, dst->ops->family);
1782 spin_unlock(&pq->hold_queue.lock);
1783
1784 dst_hold(dst->path);
1785 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1786 sk, 0);
1787 if (IS_ERR(dst))
1788 goto purge_queue;
1789
1790 if (dst->flags & DST_XFRM_QUEUE) {
1791 dst_release(dst);
1792
1793 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1794 goto purge_queue;
1795
1796 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1797 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1798 xfrm_pol_hold(pol);
1799 goto out;
a0073fe1
SK		 1800 }
1801
1802 dst_release(dst);
1803
1804 __skb_queue_head_init(&list);
1805
1806 spin_lock(&pq->hold_queue.lock);
1807 pq->timeout = 0;
1808 skb_queue_splice_init(&pq->hold_queue, &list);
1809 spin_unlock(&pq->hold_queue.lock);
1810
1811 while (!skb_queue_empty(&list)) {
1812 skb = __skb_dequeue(&list);
1813
1814 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1815 dst_hold(skb_dst(skb)->path);
1816 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1817 &fl, skb->sk, 0);
1818 if (IS_ERR(dst)) {
a0073fe1
SK		 1819 kfree_skb(skb);
1820 continue;
1821 }
1822
1823 nf_reset(skb);
1824 skb_dst_drop(skb);
1825 skb_dst_set(skb, dst);
1826
a0073fe1 1827 err = dst_output(skb);
a0073fe1
SK		 1828 }
1829
e7d8f6cb
SK		 1830out:
1831 xfrm_pol_put(pol);
a0073fe1
SK		 1832 return;
1833
1834purge_queue:
1835 pq->timeout = 0;
1836 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1837 xfrm_pol_put(pol);
a0073fe1
SK		 1838}
1839
aad88724 1840static int xdst_queue_output(struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1841{
1842 unsigned long sched_next;
1843 struct dst_entry *dst = skb_dst(skb);
1844 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1845 struct xfrm_policy *pol = xdst->pols[0];
1846 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4
SK		 1847 const struct sk_buff *fclone = skb + 1;
1848
1849 if (unlikely(skb->fclone == SKB_FCLONE_ORIG &&
1850 fclone->fclone == SKB_FCLONE_CLONE)) {
1851 kfree_skb(skb);
1852 return 0;
1853 }
a0073fe1
SK		 1854
1855 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1856 kfree_skb(skb);
1857 return -EAGAIN;
1858 }
1859
1860 skb_dst_force(skb);
a0073fe1
SK		 1861
1862 spin_lock_bh(&pq->hold_queue.lock);
1863
1864 if (!pq->timeout)
1865 pq->timeout = XFRM_QUEUE_TMO_MIN;
1866
1867 sched_next = jiffies + pq->timeout;
1868
1869 if (del_timer(&pq->hold_timer)) {
1870 if (time_before(pq->hold_timer.expires, sched_next))
1871 sched_next = pq->hold_timer.expires;
e7d8f6cb 1872 xfrm_pol_put(pol);
a0073fe1
SK		 1873 }
1874
1875 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1876 if (!mod_timer(&pq->hold_timer, sched_next))
1877 xfrm_pol_hold(pol);
a0073fe1
SK		 1878
1879 spin_unlock_bh(&pq->hold_queue.lock);
1880
1881 return 0;
1882}
1883
1884static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 1885 struct xfrm_flo *xflo,
a0073fe1
SK		 1886 const struct flowi *fl,
1887 int num_xfrms,
1888 u16 family)
1889{
1890 int err;
1891 struct net_device *dev;
b8c203b2 1892 struct dst_entry *dst;
a0073fe1
SK		 1893 struct dst_entry *dst1;
1894 struct xfrm_dst *xdst;
1895
1896 xdst = xfrm_alloc_dst(net, family);
1897 if (IS_ERR(xdst))
1898 return xdst;
1899
b8c203b2
SK		 1900 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
1901 net->xfrm.sysctl_larval_drop ||
1902 num_xfrms <= 0)
a0073fe1
SK		 1903 return xdst;
1904
b8c203b2 1905 dst = xflo->dst_orig;
a0073fe1
SK		 1906 dst1 = &xdst->u.dst;
1907 dst_hold(dst);
1908 xdst->route = dst;
1909
1910 dst_copy_metrics(dst1, dst);
1911
1912 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
1913 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
1914 dst1->lastuse = jiffies;
1915
1916 dst1->input = dst_discard;
1917 dst1->output = xdst_queue_output;
1918
1919 dst_hold(dst);
1920 dst1->child = dst;
1921 dst1->path = dst;
1922
1923 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
1924
1925 err = -ENODEV;
1926 dev = dst->dev;
1927 if (!dev)
1928 goto free_dst;
1929
1930 err = xfrm_fill_dst(xdst, dev, fl);
1931 if (err)
1932 goto free_dst;
1933
1934out:
1935 return xdst;
1936
1937free_dst:
1938 dst_release(dst1);
1939 xdst = ERR_PTR(err);
1940 goto out;
1941}
1942
80c802f3 1943static struct flow_cache_object *
dee9f4bc 1944xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 1945 struct flow_cache_object *oldflo, void *ctx)
1946{
b8c203b2 1947 struct xfrm_flo *xflo = (struct xfrm_flo *)ctx;
80c802f3
TT		 1948 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1949 struct xfrm_dst *xdst, *new_xdst;
1950 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1951
1952 /* Check if the policies from old bundle are usable */
1953 xdst = NULL;
1954 if (oldflo) {
1955 xdst = container_of(oldflo, struct xfrm_dst, flo);
1956 num_pols = xdst->num_pols;
1957 num_xfrms = xdst->num_xfrms;
1958 pol_dead = 0;
1959 for (i = 0; i < num_pols; i++) {
1960 pols[i] = xdst->pols[i];
1961 pol_dead |= pols[i]->walk.dead;
1962 }
1963 if (pol_dead) {
1964 dst_free(&xdst->u.dst);
1965 xdst = NULL;
1966 num_pols = 0;
1967 num_xfrms = 0;
1968 oldflo = NULL;
1969 }
1970 }
1971
1972 /* Resolve policies to use if we couldn't get them from
1973 * previous cache entry */
1974 if (xdst == NULL) {
1975 num_pols = 1;
b5fb82c4
BZ		 1976 pols[0] = __xfrm_policy_lookup(net, fl, family,
1977 flow_to_policy_dir(dir));
80c802f3
TT		 1978 err = xfrm_expand_policies(fl, family, pols,
1979 &num_pols, &num_xfrms);
1980 if (err < 0)
1981 goto inc_error;
1982 if (num_pols == 0)
1983 return NULL;
1984 if (num_xfrms <= 0)
1985 goto make_dummy_bundle;
1986 }
1987
b8c203b2
SK		 1988 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
1989 xflo->dst_orig);
80c802f3
TT		 1990 if (IS_ERR(new_xdst)) {
1991 err = PTR_ERR(new_xdst);
1992 if (err != -EAGAIN)
1993 goto error;
1994 if (oldflo == NULL)
1995 goto make_dummy_bundle;
1996 dst_hold(&xdst->u.dst);
1997 return oldflo;
d809ec89
TT		 1998 } else if (new_xdst == NULL) {
1999 num_xfrms = 0;
2000 if (oldflo == NULL)
2001 goto make_dummy_bundle;
2002 xdst->num_xfrms = 0;
2003 dst_hold(&xdst->u.dst);
2004 return oldflo;
80c802f3
TT		 2005 }
2006
2007 /* Kill the previous bundle */
2008 if (xdst) {
2009 /* The policies were stolen for newly generated bundle */
2010 xdst->num_pols = 0;
2011 dst_free(&xdst->u.dst);
2012 }
2013
2014 /* Flow cache does not have reference, it dst_free()'s,
2015 * but we do need to return one reference for original caller */
2016 dst_hold(&new_xdst->u.dst);
2017 return &new_xdst->flo;
2018
2019make_dummy_bundle:
2020 /* We found policies, but there's no bundles to instantiate:
2021 * either because the policy blocks, has no transformations or
2022 * we could not build template (no xfrm_states).*/
b8c203b2 2023 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2024 if (IS_ERR(xdst)) {
2025 xfrm_pols_put(pols, num_pols);
2026 return ERR_CAST(xdst);
2027 }
2028 xdst->num_pols = num_pols;
2029 xdst->num_xfrms = num_xfrms;
3e94c2dc 2030 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2031
2032 dst_hold(&xdst->u.dst);
2033 return &xdst->flo;
2034
2035inc_error:
2036 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2037error:
2038 if (xdst != NULL)
2039 dst_free(&xdst->u.dst);
2040 else
2041 xfrm_pols_put(pols, num_pols);
2042 return ERR_PTR(err);
2043}
1da177e4 2044
2774c131
DM		 2045static struct dst_entry *make_blackhole(struct net *net, u16 family,
2046 struct dst_entry *dst_orig)
2047{
2048 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2049 struct dst_entry *ret;
2050
2051 if (!afinfo) {
2052 dst_release(dst_orig);
433a1954 2053 return ERR_PTR(-EINVAL);
2774c131
DM		 2054 } else {
2055 ret = afinfo->blackhole_route(net, dst_orig);
2056 }
2057 xfrm_policy_put_afinfo(afinfo);
2058
2059 return ret;
2060}
2061
1da177e4
LT		 2062/* Main function: finds/creates a bundle for given flow.
2063 *
2064 * At the moment we eat a raw IP route. Mostly to speed up lookups
2065 * on interfaces with disabled IPsec.
2066 */
452edd59
DM		 2067struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2068 const struct flowi *fl,
2069 struct sock *sk, int flags)
1da177e4 2070{
4e81bb83 2071 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2072 struct flow_cache_object *flo;
2073 struct xfrm_dst *xdst;
452edd59 2074 struct dst_entry *dst, *route;
80c802f3 2075 u16 family = dst_orig->ops->family;
df71837d 2076 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2077 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2078
80c802f3
TT		 2079 dst = NULL;
2080 xdst = NULL;
2081 route = NULL;
4e81bb83 2082
f7944fb1 2083 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2084 num_pols = 1;
2085 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2086 err = xfrm_expand_policies(fl, family, pols,
2087 &num_pols, &num_xfrms);
2088 if (err < 0)
75b8c133 2089 goto dropdst;
80c802f3
TT		 2090
2091 if (num_pols) {
2092 if (num_xfrms <= 0) {
2093 drop_pols = num_pols;
2094 goto no_transform;
2095 }
2096
2097 xdst = xfrm_resolve_and_create_bundle(
2098 pols, num_pols, fl,
2099 family, dst_orig);
2100 if (IS_ERR(xdst)) {
2101 xfrm_pols_put(pols, num_pols);
2102 err = PTR_ERR(xdst);
2103 goto dropdst;
d809ec89
TT		 2104 } else if (xdst == NULL) {
2105 num_xfrms = 0;
2106 drop_pols = num_pols;
2107 goto no_transform;
80c802f3
TT		 2108 }
2109
b7eea454
SK		 2110 dst_hold(&xdst->u.dst);
2111 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2112 route = xdst->route;
0aa64774 2113 }
3bccfbc7 2114 }
1da177e4 2115
80c802f3 2116 if (xdst == NULL) {
b8c203b2
SK		 2117 struct xfrm_flo xflo;
2118
2119 xflo.dst_orig = dst_orig;
2120 xflo.flags = flags;
2121
1da177e4 2122 /* To accelerate a bit... */
2518c7c2 2123 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2124 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2125 goto nopol;
1da177e4 2126
80c802f3 2127 flo = flow_cache_lookup(net, fl, family, dir,
b8c203b2 2128 xfrm_bundle_lookup, &xflo);
80c802f3
TT		 2129 if (flo == NULL)
2130 goto nopol;
fe1a5f03 2131 if (IS_ERR(flo)) {
80c802f3 2132 err = PTR_ERR(flo);
75b8c133 2133 goto dropdst;
d66e37a9 2134 }
80c802f3
TT		 2135 xdst = container_of(flo, struct xfrm_dst, flo);
2136
2137 num_pols = xdst->num_pols;
2138 num_xfrms = xdst->num_xfrms;
3e94c2dc 2139 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2140 route = xdst->route;
2141 }
2142
2143 dst = &xdst->u.dst;
2144 if (route == NULL && num_xfrms > 0) {
2145 /* The only case when xfrm_bundle_lookup() returns a
2146 * bundle with null route, is when the template could
2147 * not be resolved. It means policies are there, but
2148 * bundle could not be created, since we don't yet
2149 * have the xfrm_state's. We need to wait for KM to
2150 * negotiate new SA's or bail out with error.*/
2151 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2152 dst_release(dst);
a1aa3483 2153 xfrm_pols_put(pols, drop_pols);
80c802f3 2154 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2774c131 2155
f92ee619 2156 return ERR_PTR(-EREMOTE);
80c802f3 2157 }
80c802f3 2158
5b8ef341 2159 err = -EAGAIN;
80c802f3
TT		 2160
2161 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2162 goto error;
1da177e4
LT		 2163 }
2164
80c802f3
TT		 2165no_transform:
2166 if (num_pols == 0)
8b7817f3 2167 goto nopol;
1da177e4 2168
80c802f3
TT		 2169 if ((flags & XFRM_LOOKUP_ICMP) &&
2170 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2171 err = -ENOENT;
8b7817f3 2172 goto error;
80c802f3 2173 }
8b7817f3 2174
80c802f3
TT		 2175 for (i = 0; i < num_pols; i++)
2176 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2177
80c802f3 2178 if (num_xfrms < 0) {
1da177e4 2179 /* Prohibit the flow */
59c9940e 2180 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2181 err = -EPERM;
2182 goto error;
80c802f3
TT		 2183 } else if (num_xfrms > 0) {
2184 /* Flow transformed */
80c802f3
TT		 2185 dst_release(dst_orig);
2186 } else {
2187 /* Flow passes untransformed */
2188 dst_release(dst);
452edd59 2189 dst = dst_orig;
1da177e4 2190 }
80c802f3
TT		 2191ok:
2192 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2193 if (dst && dst->xfrm &&
2194 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2195 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2196 return dst;
1da177e4 2197
80c802f3 2198nopol:
452edd59
DM		 2199 if (!(flags & XFRM_LOOKUP_ICMP)) {
2200 dst = dst_orig;
80c802f3 2201 goto ok;
452edd59 2202 }
80c802f3 2203 err = -ENOENT;
1da177e4 2204error:
80c802f3 2205 dst_release(dst);
75b8c133
HX		 2206dropdst:
2207 dst_release(dst_orig);
80c802f3 2208 xfrm_pols_put(pols, drop_pols);
452edd59 2209 return ERR_PTR(err);
1da177e4
LT		 2210}
2211EXPORT_SYMBOL(xfrm_lookup);
2212
f92ee619
SK		 2213/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2214 * Otherwise we may send out blackholed packets.
2215 */
2216struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2217 const struct flowi *fl,
2218 struct sock *sk, int flags)
2219{
b8c203b2
SK		 2220 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
2221 flags | XFRM_LOOKUP_QUEUE);
f92ee619
SK		 2222
2223 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2224 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2225
2226 return dst;
2227}
2228EXPORT_SYMBOL(xfrm_lookup_route);
2229
df0ba92a 2230static inline int
8f029de2 2231xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2232{
2233 struct xfrm_state *x;
df0ba92a
MN		 2234
2235 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2236 return 0;
2237 x = skb->sp->xvec[idx];
2238 if (!x->type->reject)
2239 return 0;
1ecafede 2240 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2241}
2242
1da177e4
LT		 2243/* When skb is transformed back to its "native" form, we have to
2244 * check policy restrictions. At the moment we make this in maximally
2245 * stupid way. Shame on me. :-) Of course, connected sockets must
2246 * have policy cached at them.
2247 */
2248
2249static inline int
7db454b9 2250xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2251 unsigned short family)
2252{
2253 if (xfrm_state_kern(x))
928ba416 2254 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2255 return x->id.proto == tmpl->id.proto &&
2256 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2257 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2258 x->props.mode == tmpl->mode &&
c5d18e98 2259 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2260 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2261 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2262 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2263}
2264
df0ba92a
MN		 2265/*
2266 * 0 or more than 0 is returned when validation is succeeded (either bypass
2267 * because of optional transport mode, or next index of the mathced secpath
2268 * state with the template.
2269 * -1 is returned when no matching template is found.
2270 * Otherwise "-2 - errored_index" is returned.
2271 */
1da177e4 2272static inline int
22cccb7e 2273xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2274 unsigned short family)
2275{
2276 int idx = start;
2277
2278 if (tmpl->optional) {
7e49e6de 2279 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2280 return start;
2281 } else
2282 start = -1;
2283 for (; idx < sp->len; idx++) {
dbe5b4aa 2284 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2285 return ++idx;
df0ba92a
MN		 2286 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2287 if (start == -1)
2288 start = -2-idx;
1da177e4 2289 break;
df0ba92a 2290 }
1da177e4
LT		 2291 }
2292 return start;
2293}
2294
d5422efe
HX		 2295int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2296 unsigned int family, int reverse)
1da177e4
LT		 2297{
2298 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2299 int err;
1da177e4
LT		 2300
2301 if (unlikely(afinfo == NULL))
2302 return -EAFNOSUPPORT;
2303
d5422efe 2304 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2305 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2306 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2307 return err;
1da177e4 2308}
d5422efe 2309EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2310
9a7386ec 2311static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2312{
2313 for (; k < sp->len; k++) {
df0ba92a 2314 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2315 *idxp = k;
1da177e4 2316 return 1;
df0ba92a 2317 }
1da177e4
LT		 2318 }
2319
2320 return 0;
2321}
2322
a716c119 2323int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2324 unsigned short family)
2325{
f6e1e25d 2326 struct net *net = dev_net(skb->dev);
1da177e4 2327 struct xfrm_policy *pol;
4e81bb83
MN		 2328 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2329 int npols = 0;
2330 int xfrm_nr;
2331 int pi;
d5422efe 2332 int reverse;
1da177e4 2333 struct flowi fl;
d5422efe 2334 u8 fl_dir;
df0ba92a 2335 int xerr_idx = -1;
1da177e4 2336
d5422efe
HX		 2337 reverse = dir & ~XFRM_POLICY_MASK;
2338 dir &= XFRM_POLICY_MASK;
2339 fl_dir = policy_to_flow_dir(dir);
2340
0aa64774 2341 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2342 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2343 return 0;
0aa64774
MN		 2344 }
2345
eb9c7ebe 2346 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2347
2348 /* First, check used SA against their selectors. */
2349 if (skb->sp) {
2350 int i;
2351
9b7a787d 2352 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2353 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2354 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2355 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2356 return 0;
0aa64774 2357 }
1da177e4
LT		 2358 }
2359 }
2360
2361 pol = NULL;
3bccfbc7 2362 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2363 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2364 if (IS_ERR(pol)) {
59c9940e 2365 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2366 return 0;
0aa64774 2367 }
3bccfbc7 2368 }
1da177e4 2369
fe1a5f03
TT		 2370 if (!pol) {
2371 struct flow_cache_object *flo;
2372
2373 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2374 xfrm_policy_lookup, NULL);
2375 if (IS_ERR_OR_NULL(flo))
2376 pol = ERR_CAST(flo);
2377 else
2378 pol = container_of(flo, struct xfrm_policy, flo);
2379 }
1da177e4 2380
0aa64774 2381 if (IS_ERR(pol)) {
59c9940e 2382 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2383 return 0;
0aa64774 2384 }
134b0fc5 2385
df0ba92a 2386 if (!pol) {
d1d9facf 2387 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2388 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2389 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2390 return 0;
2391 }
2392 return 1;
2393 }
1da177e4 2394
9d729f72 2395 pol->curlft.use_time = get_seconds();
1da177e4 2396
4e81bb83 2397 pols[0] = pol;
02d0892f 2398 npols++;
4e81bb83
MN		 2399#ifdef CONFIG_XFRM_SUB_POLICY
2400 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2401 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2402 &fl, family,
2403 XFRM_POLICY_IN);
2404 if (pols[1]) {
0aa64774 2405 if (IS_ERR(pols[1])) {
59c9940e 2406 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2407 return 0;
0aa64774 2408 }
9d729f72 2409 pols[1]->curlft.use_time = get_seconds();
02d0892f 2410 npols++;
4e81bb83
MN		 2411 }
2412 }
2413#endif
2414
1da177e4
LT		 2415 if (pol->action == XFRM_POLICY_ALLOW) {
2416 struct sec_path *sp;
2417 static struct sec_path dummy;
4e81bb83 2418 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2419 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2420 struct xfrm_tmpl **tpp = tp;
2421 int ti = 0;
1da177e4
LT		 2422 int i, k;
2423
2424 if ((sp = skb->sp) == NULL)
2425 sp = &dummy;
2426
4e81bb83
MN		 2427 for (pi = 0; pi < npols; pi++) {
2428 if (pols[pi] != pol &&
0aa64774 2429 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2430 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2431 goto reject;
0aa64774
MN		 2432 }
2433 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2434 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2435 goto reject_error;
0aa64774 2436 }
4e81bb83
MN		 2437 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2438 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2439 }
2440 xfrm_nr = ti;
41a49cc3 2441 if (npols > 1) {
283bc9f3 2442 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2443 tpp = stp;
2444 }
4e81bb83 2445
1da177e4
LT		 2446 /* For each tunnel xfrm, find the first matching tmpl.
2447 * For each tmpl before that, find corresponding xfrm.
2448 * Order is _important_. Later we will implement
2449 * some barriers, but at the moment barriers
2450 * are implied between each two transformations.
2451 */
4e81bb83
MN		 2452 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2453 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2454 if (k < 0) {
d1d9facf
JM		 2455 if (k < -1)
2456 /* "-2 - errored_index" returned */
2457 xerr_idx = -(2+k);
59c9940e 2458 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2459 goto reject;
df0ba92a 2460 }
1da177e4
LT		 2461 }
2462
0aa64774 2463 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2464 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2465 goto reject;
0aa64774 2466 }
1da177e4 2467
4e81bb83 2468 xfrm_pols_put(pols, npols);
1da177e4
LT		 2469 return 1;
2470 }
59c9940e 2471 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2472
2473reject:
df0ba92a 2474 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2475reject_error:
2476 xfrm_pols_put(pols, npols);
1da177e4
LT		 2477 return 0;
2478}
2479EXPORT_SYMBOL(__xfrm_policy_check);
2480
2481int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2482{
99a66657 2483 struct net *net = dev_net(skb->dev);
1da177e4 2484 struct flowi fl;
adf30907 2485 struct dst_entry *dst;
73137147 2486 int res = 1;
1da177e4 2487
0aa64774 2488 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2489 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2490 return 0;
0aa64774 2491 }
1da177e4 2492
fafeeb6c 2493 skb_dst_force(skb);
adf30907 2494
b8c203b2 2495 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2496 if (IS_ERR(dst)) {
73137147 2497 res = 0;
452edd59
DM		 2498 dst = NULL;
2499 }
adf30907
ED		 2500 skb_dst_set(skb, dst);
2501 return res;
1da177e4
LT		 2502}
2503EXPORT_SYMBOL(__xfrm_route_forward);
2504
d49c73c7
DM		 2505/* Optimize later using cookies and generation ids. */
2506
1da177e4
LT		 2507static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2508{
d49c73c7 2509 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2510 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2511 * get validated by dst_ops->check on every use. We do this
2512 * because when a normal route referenced by an XFRM dst is
2513 * obsoleted we do not go looking around for all parent
2514 * referencing XFRM dsts so that we can invalidate them. It
2515 * is just too much work. Instead we make the checks here on
2516 * every use. For example:
d49c73c7
DM		 2517 *
2518 * XFRM dst A --> IPv4 dst X
2519 *
2520 * X is the "xdst->route" of A (X is also the "dst->path" of A
2521 * in this example). If X is marked obsolete, "A" will not
2522 * notice. That's what we are validating here via the
2523 * stale_bundle() check.
2524 *
2525 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2526 * dst which causes it's ->obsolete field to be set to
2527 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2528 * this, we want to force a new route lookup.
399c180a 2529 */
d49c73c7
DM		 2530 if (dst->obsolete < 0 && !stale_bundle(dst))
2531 return dst;
2532
1da177e4
LT		 2533 return NULL;
2534}
2535
2536static int stale_bundle(struct dst_entry *dst)
2537{
12fdb4d3 2538 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2539}
2540
aabc9761 2541void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2542{
1da177e4 2543 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2544 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2545 dev_hold(dst->dev);
1da177e4
LT		 2546 dev_put(dev);
2547 }
2548}
aabc9761 2549EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2550
2551static void xfrm_link_failure(struct sk_buff *skb)
2552{
2553 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2554}
2555
2556static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2557{
2558 if (dst) {
2559 if (dst->obsolete) {
2560 dst_release(dst);
2561 dst = NULL;
2562 }
2563 }
2564 return dst;
2565}
2566
e4c17216 2567void xfrm_garbage_collect(struct net *net)
c0ed1c14 2568{
ca925cf1 2569 flow_cache_flush(net);
c0ed1c14 2570}
e4c17216 2571EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2572
2573static void xfrm_garbage_collect_deferred(struct net *net)
2574{
ca925cf1 2575 flow_cache_flush_deferred(net);
c0ed1c14
SK		 2576}
2577
25ee3286 2578static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2579{
2580 do {
2581 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2582 u32 pmtu, route_mtu_cached;
2583
2584 pmtu = dst_mtu(dst->child);
2585 xdst->child_mtu_cached = pmtu;
2586
2587 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2588
2589 route_mtu_cached = dst_mtu(xdst->route);
2590 xdst->route_mtu_cached = route_mtu_cached;
2591
2592 if (pmtu > route_mtu_cached)
2593 pmtu = route_mtu_cached;
2594
defb3519 2595 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2596 } while ((dst = dst->next));
2597}
2598
1da177e4
LT		 2599/* Check that the bundle accepts the flow and its components are
2600 * still valid.
2601 */
2602
12fdb4d3 2603static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2604{
2605 struct dst_entry *dst = &first->u.dst;
2606 struct xfrm_dst *last;
2607 u32 mtu;
2608
92d63dec 2609 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2610 (dst->dev && !netif_running(dst->dev)))
2611 return 0;
2612
a0073fe1
SK		 2613 if (dst->flags & DST_XFRM_QUEUE)
2614 return 1;
2615
1da177e4
LT		 2616 last = NULL;
2617
2618 do {
2619 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2620
1da177e4
LT		 2621 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2622 return 0;
80c802f3
TT		 2623 if (xdst->xfrm_genid != dst->xfrm->genid)
2624 return 0;
b1312c89
TT		 2625 if (xdst->num_pols > 0 &&
2626 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2627 return 0;
e53820de 2628
1da177e4
LT		 2629 mtu = dst_mtu(dst->child);
2630 if (xdst->child_mtu_cached != mtu) {
2631 last = xdst;
2632 xdst->child_mtu_cached = mtu;
2633 }
2634
92d63dec 2635 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2636 return 0;
2637 mtu = dst_mtu(xdst->route);
2638 if (xdst->route_mtu_cached != mtu) {
2639 last = xdst;
2640 xdst->route_mtu_cached = mtu;
2641 }
2642
2643 dst = dst->child;
2644 } while (dst->xfrm);
2645
2646 if (likely(!last))
2647 return 1;
2648
2649 mtu = last->child_mtu_cached;
2650 for (;;) {
2651 dst = &last->u.dst;
2652
2653 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2654 if (mtu > last->route_mtu_cached)
2655 mtu = last->route_mtu_cached;
defb3519 2656 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2657
2658 if (last == first)
2659 break;
2660
bd0bf076 2661 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2662 last->child_mtu_cached = mtu;
2663 }
2664
2665 return 1;
2666}
2667
0dbaee3b
DM		 2668static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2669{
2670 return dst_metric_advmss(dst->path);
2671}
2672
ebb762f2 2673static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2674{
618f9bc7
SK		 2675 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2676
2677 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2678}
2679
f894cbf8
DM		 2680static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2681 struct sk_buff *skb,
2682 const void *daddr)
d3aaeb38 2683{
f894cbf8 2684 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2685}
2686
1da177e4
LT		 2687int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2688{
d7c7544c 2689 struct net *net;
1da177e4
LT		 2690 int err = 0;
2691 if (unlikely(afinfo == NULL))
2692 return -EINVAL;
2693 if (unlikely(afinfo->family >= NPROTO))
2694 return -EAFNOSUPPORT;
ef8531b6 2695 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2696 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2697 err = -ENOBUFS;
2698 else {
2699 struct dst_ops *dst_ops = afinfo->dst_ops;
2700 if (likely(dst_ops->kmem_cachep == NULL))
2701 dst_ops->kmem_cachep = xfrm_dst_cache;
2702 if (likely(dst_ops->check == NULL))
2703 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2704 if (likely(dst_ops->default_advmss == NULL))
2705 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2706 if (likely(dst_ops->mtu == NULL))
2707 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2708 if (likely(dst_ops->negative_advice == NULL))
2709 dst_ops->negative_advice = xfrm_negative_advice;
2710 if (likely(dst_ops->link_failure == NULL))
2711 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2712 if (likely(dst_ops->neigh_lookup == NULL))
2713 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2714 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2715 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2716 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2717 }
ef8531b6 2718 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2719
2720 rtnl_lock();
2721 for_each_net(net) {
2722 struct dst_ops *xfrm_dst_ops;
2723
2724 switch (afinfo->family) {
2725 case AF_INET:
2726 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2727 break;
dfd56b8b 2728#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2729 case AF_INET6:
2730 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2731 break;
2732#endif
2733 default:
2734 BUG();
2735 }
2736 *xfrm_dst_ops = *afinfo->dst_ops;
2737 }
2738 rtnl_unlock();
2739
1da177e4
LT		 2740 return err;
2741}
2742EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2743
2744int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2745{
2746 int err = 0;
2747 if (unlikely(afinfo == NULL))
2748 return -EINVAL;
2749 if (unlikely(afinfo->family >= NPROTO))
2750 return -EAFNOSUPPORT;
ef8531b6 2751 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2752 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2753 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2754 err = -EINVAL;
ef8531b6
ED		 2755 else
2756 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2757 NULL);
2758 }
2759 spin_unlock(&xfrm_policy_afinfo_lock);
2760 if (!err) {
2761 struct dst_ops *dst_ops = afinfo->dst_ops;
2762
2763 synchronize_rcu();
2764
2765 dst_ops->kmem_cachep = NULL;
2766 dst_ops->check = NULL;
2767 dst_ops->negative_advice = NULL;
2768 dst_ops->link_failure = NULL;
2769 afinfo->garbage_collect = NULL;
1da177e4 2770 }
1da177e4
LT		 2771 return err;
2772}
2773EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2774
d7c7544c
AD		 2775static void __net_init xfrm_dst_ops_init(struct net *net)
2776{
2777 struct xfrm_policy_afinfo *afinfo;
2778
ef8531b6
ED		 2779 rcu_read_lock();
2780 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2781 if (afinfo)
2782 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2783#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2784 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2785 if (afinfo)
2786 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2787#endif
418a99ac 2788 rcu_read_unlock();
546be240
HX		 2789}
2790
1da177e4
LT		 2791static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2792{
351638e7 2793 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2794
1da177e4
LT		 2795 switch (event) {
2796 case NETDEV_DOWN:
c0ed1c14 2797 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2798 }
2799 return NOTIFY_DONE;
2800}
2801
2802static struct notifier_block xfrm_dev_notifier = {
d5917a35 2803 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2804};
2805
558f82ef 2806#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2807static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2808{
c68cd1a0 2809 int rv;
698365fa
WC		 2810 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2811 if (!net->mib.xfrm_statistics)
558f82ef 2812 return -ENOMEM;
c68cd1a0
AD		 2813 rv = xfrm_proc_init(net);
2814 if (rv < 0)
698365fa 2815 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2816 return rv;
558f82ef 2817}
59c9940e
AD		 2818
2819static void xfrm_statistics_fini(struct net *net)
2820{
c68cd1a0 2821 xfrm_proc_fini(net);
698365fa 2822 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2823}
2824#else
2825static int __net_init xfrm_statistics_init(struct net *net)
2826{
2827 return 0;
2828}
2829
2830static void xfrm_statistics_fini(struct net *net)
2831{
2832}
558f82ef
MN		 2833#endif
2834
d62ddc21 2835static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2836{
2518c7c2
DM		 2837 unsigned int hmask, sz;
2838 int dir;
2839
d62ddc21
AD		 2840 if (net_eq(net, &init_net))
2841 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2842 sizeof(struct xfrm_dst),
e5d679f3 2843 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2844 NULL);
1da177e4 2845
2518c7c2
DM		 2846 hmask = 8 - 1;
2847 sz = (hmask+1) * sizeof(struct hlist_head);
2848
93b851c1
AD		 2849 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2850 if (!net->xfrm.policy_byidx)
2851 goto out_byidx;
8100bea7 2852 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2853
2854 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2855 struct xfrm_policy_hash *htab;
2856
dc2caba7 2857 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2858 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2859
a35f6c5d 2860 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2861 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2862 if (!htab->table)
a35f6c5d
AD		 2863 goto out_bydst;
2864 htab->hmask = hmask;
2518c7c2
DM		 2865 }
2866
adfcf0b2 2867 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2868 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2869 if (net_eq(net, &init_net))
2870 register_netdevice_notifier(&xfrm_dev_notifier);
2871 return 0;
93b851c1 2872
a35f6c5d
AD		 2873out_bydst:
2874 for (dir--; dir >= 0; dir--) {
2875 struct xfrm_policy_hash *htab;
2876
2877 htab = &net->xfrm.policy_bydst[dir];
2878 xfrm_hash_free(htab->table, sz);
2879 }
2880 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2881out_byidx:
2882 return -ENOMEM;
d62ddc21
AD		 2883}
2884
2885static void xfrm_policy_fini(struct net *net)
2886{
93b851c1 2887 unsigned int sz;
8b18f8ea 2888 int dir;
93b851c1 2889
7c2776ee
AD		 2890 flush_work(&net->xfrm.policy_hash_work);
2891#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2892 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 2893#endif
2e71029e 2894 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 2895
adfcf0b2 2896 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2897
8b18f8ea 2898 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2899 struct xfrm_policy_hash *htab;
2900
8b18f8ea 2901 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2902
2903 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2904 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2905 WARN_ON(!hlist_empty(htab->table));
2906 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2907 }
2908
8100bea7 2909 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2910 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2911 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2912}
2913
d62ddc21
AD		 2914static int __net_init xfrm_net_init(struct net *net)
2915{
2916 int rv;
2917
59c9940e
AD		 2918 rv = xfrm_statistics_init(net);
2919 if (rv < 0)
2920 goto out_statistics;
d62ddc21
AD		 2921 rv = xfrm_state_init(net);
2922 if (rv < 0)
2923 goto out_state;
2924 rv = xfrm_policy_init(net);
2925 if (rv < 0)
2926 goto out_policy;
d7c7544c 2927 xfrm_dst_ops_init(net);
b27aeadb
AD		 2928 rv = xfrm_sysctl_init(net);
2929 if (rv < 0)
2930 goto out_sysctl;
4a93f509
SK		 2931 rv = flow_cache_init(net);
2932 if (rv < 0)
2933 goto out;
283bc9f3
FD		 2934
2935 /* Initialize the per-net locks here */
2936 spin_lock_init(&net->xfrm.xfrm_state_lock);
2937 rwlock_init(&net->xfrm.xfrm_policy_lock);
283bc9f3
FD		 2938 mutex_init(&net->xfrm.xfrm_cfg_mutex);
2939
d62ddc21
AD		 2940 return 0;
2941
4a93f509
SK		 2942out:
2943 xfrm_sysctl_fini(net);
b27aeadb
AD		 2944out_sysctl:
2945 xfrm_policy_fini(net);
d62ddc21
AD		 2946out_policy:
2947 xfrm_state_fini(net);
2948out_state:
59c9940e
AD		 2949 xfrm_statistics_fini(net);
2950out_statistics:
d62ddc21
AD		 2951 return rv;
2952}
2953
2954static void __net_exit xfrm_net_exit(struct net *net)
2955{
4a93f509 2956 flow_cache_fini(net);
b27aeadb 2957 xfrm_sysctl_fini(net);
d62ddc21
AD		 2958 xfrm_policy_fini(net);
2959 xfrm_state_fini(net);
59c9940e 2960 xfrm_statistics_fini(net);
d62ddc21
AD		 2961}
2962
2963static struct pernet_operations __net_initdata xfrm_net_ops = {
2964 .init = xfrm_net_init,
2965 .exit = xfrm_net_exit,
2966};
2967
1da177e4
LT		 2968void __init xfrm_init(void)
2969{
d62ddc21 2970 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 2971 xfrm_input_init();
2972}
2973
ab5f5e8b 2974#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2975static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2976 struct audit_buffer *audit_buf)
ab5f5e8b 2977{
875179fa
PM		 2978 struct xfrm_sec_ctx *ctx = xp->security;
2979 struct xfrm_selector *sel = &xp->selector;
2980
2981 if (ctx)
ab5f5e8b 2982 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2983 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2984
9b7a787d 2985 switch (sel->family) {
ab5f5e8b 2986 case AF_INET:
21454aaa 2987 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2988 if (sel->prefixlen_s != 32)
2989 audit_log_format(audit_buf, " src_prefixlen=%d",
2990 sel->prefixlen_s);
21454aaa 2991 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2992 if (sel->prefixlen_d != 32)
2993 audit_log_format(audit_buf, " dst_prefixlen=%d",
2994 sel->prefixlen_d);
ab5f5e8b
JL		 2995 break;
2996 case AF_INET6:
5b095d98 2997 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 2998 if (sel->prefixlen_s != 128)
2999 audit_log_format(audit_buf, " src_prefixlen=%d",
3000 sel->prefixlen_s);
5b095d98 3001 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3002 if (sel->prefixlen_d != 128)
3003 audit_log_format(audit_buf, " dst_prefixlen=%d",
3004 sel->prefixlen_d);
ab5f5e8b
JL		 3005 break;
3006 }
3007}
3008
2e71029e 3009void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3010{
3011 struct audit_buffer *audit_buf;
ab5f5e8b 3012
afeb14b4 3013 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3014 if (audit_buf == NULL)
3015 return;
2e71029e 3016 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3017 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3018 xfrm_audit_common_policyinfo(xp, audit_buf);
3019 audit_log_end(audit_buf);
3020}
3021EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3022
68277acc 3023void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3024 bool task_valid)
ab5f5e8b
JL		 3025{
3026 struct audit_buffer *audit_buf;
ab5f5e8b 3027
afeb14b4 3028 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3029 if (audit_buf == NULL)
3030 return;
2e71029e 3031 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3032 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3033 xfrm_audit_common_policyinfo(xp, audit_buf);
3034 audit_log_end(audit_buf);
3035}
3036EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3037#endif
3038
80c9abaa 3039#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3040static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3041 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3042{
3043 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3044 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3045 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3046 sel_cmp->family) &&
3047 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3048 sel_cmp->family) &&
80c9abaa
SS		 3049 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3050 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3051 return true;
80c9abaa
SS		 3052 }
3053 } else {
3054 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3055 return true;
80c9abaa
SS		 3056 }
3057 }
bc9b35ad 3058 return false;
80c9abaa
SS		 3059}
3060
3e94c2dc
WC		 3061static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3062 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3063{
3064 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3065 struct hlist_head *chain;
3066 u32 priority = ~0U;
3067
283bc9f3 3068 read_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME*/
8d549c4f 3069 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3070 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3071 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3072 pol->type == type) {
3073 ret = pol;
3074 priority = ret->priority;
3075 break;
3076 }
3077 }
8d549c4f 3078 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3079 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3080 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3081 pol->type == type &&
3082 pol->priority < priority) {
3083 ret = pol;
3084 break;
3085 }
3086 }
3087
3088 if (ret)
3089 xfrm_pol_hold(ret);
3090
283bc9f3 3091 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3092
3093 return ret;
3094}
3095
dd701754 3096static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3097{
3098 int match = 0;
3099
3100 if (t->mode == m->mode && t->id.proto == m->proto &&
3101 (m->reqid == 0 || t->reqid == m->reqid)) {
3102 switch (t->mode) {
3103 case XFRM_MODE_TUNNEL:
3104 case XFRM_MODE_BEET:
70e94e66
YH		 3105 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3106 m->old_family) &&
3107 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3108 m->old_family)) {
80c9abaa
SS		 3109 match = 1;
3110 }
3111 break;
3112 case XFRM_MODE_TRANSPORT:
3113 /* in case of transport mode, template does not store
3114 any IP addresses, hence we just compare mode and
3115 protocol */
3116 match = 1;
3117 break;
3118 default:
3119 break;
3120 }
3121 }
3122 return match;
3123}
3124
3125/* update endpoint address(es) of template(s) */
3126static int xfrm_policy_migrate(struct xfrm_policy *pol,
3127 struct xfrm_migrate *m, int num_migrate)
3128{
3129 struct xfrm_migrate *mp;
80c9abaa
SS		 3130 int i, j, n = 0;
3131
3132 write_lock_bh(&pol->lock);
12a169e7 3133 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3134 /* target policy has been deleted */
3135 write_unlock_bh(&pol->lock);
3136 return -ENOENT;
3137 }
3138
3139 for (i = 0; i < pol->xfrm_nr; i++) {
3140 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3141 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3142 continue;
3143 n++;
1bfcb10f
HX		 3144 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3145 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3146 continue;
3147 /* update endpoints */
3148 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3149 sizeof(pol->xfrm_vec[i].id.daddr));
3150 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3151 sizeof(pol->xfrm_vec[i].saddr));
3152 pol->xfrm_vec[i].encap_family = mp->new_family;
3153 /* flush bundles */
80c802f3 3154 atomic_inc(&pol->genid);
80c9abaa
SS		 3155 }
3156 }
3157
3158 write_unlock_bh(&pol->lock);
3159
3160 if (!n)
3161 return -ENODATA;
3162
3163 return 0;
3164}
3165
dd701754 3166static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3167{
3168 int i, j;
3169
3170 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3171 return -EINVAL;
3172
3173 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3174 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3175 m[i].old_family) &&
3176 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3177 m[i].old_family))
80c9abaa
SS		 3178 return -EINVAL;
3179 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3180 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3181 return -EINVAL;
3182
3183 /* check if there is any duplicated entry */
3184 for (j = i + 1; j < num_migrate; j++) {
3185 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3186 sizeof(m[i].old_daddr)) &&
3187 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3188 sizeof(m[i].old_saddr)) &&
3189 m[i].proto == m[j].proto &&
3190 m[i].mode == m[j].mode &&
3191 m[i].reqid == m[j].reqid &&
3192 m[i].old_family == m[j].old_family)
3193 return -EINVAL;
3194 }
3195 }
3196
3197 return 0;
3198}
3199
b4b7c0b3 3200int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3201 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3202 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3203{
3204 int i, err, nx_cur = 0, nx_new = 0;
3205 struct xfrm_policy *pol = NULL;
3206 struct xfrm_state *x, *xc;
3207 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3208 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3209 struct xfrm_migrate *mp;
3210
3211 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3212 goto out;
3213
3214 /* Stage 1 - find policy */
8d549c4f 3215 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3216 err = -ENOENT;
3217 goto out;
3218 }
3219
3220 /* Stage 2 - find and update state(s) */
3221 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3222 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3223 x_cur[nx_cur] = x;
3224 nx_cur++;
3225 if ((xc = xfrm_state_migrate(x, mp))) {
3226 x_new[nx_new] = xc;
3227 nx_new++;
3228 } else {
3229 err = -ENODATA;
3230 goto restore_state;
3231 }
3232 }
3233 }
3234
3235 /* Stage 3 - update policy */
3236 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3237 goto restore_state;
3238
3239 /* Stage 4 - delete old state(s) */
3240 if (nx_cur) {
3241 xfrm_states_put(x_cur, nx_cur);
3242 xfrm_states_delete(x_cur, nx_cur);
3243 }
3244
3245 /* Stage 5 - announce */
13c1d189 3246 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3247
3248 xfrm_pol_put(pol);
3249
3250 return 0;
3251out:
3252 return err;
3253
3254restore_state:
3255 if (pol)
3256 xfrm_pol_put(pol);
3257 if (nx_cur)
3258 xfrm_states_put(x_cur, nx_cur);
3259 if (nx_new)
3260 xfrm_states_delete(x_new, nx_new);
3261
3262 return err;
3263}
e610e679 3264EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3265#endif
Linux kernel - Deliverables
This page took 1.006347 seconds and 5 git commands to generate.