projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
xfrm: hash prefixed policies based on preflen thresholds
[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
418a99ac
PJ		 42static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
43static struct xfrm_policy_afinfo __rcu *xfrm_policy_afinfo[NPROTO]
44 __read_mostly;
1da177e4 45
e18b890b 46static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 47
25ee3286 48static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 49static int stale_bundle(struct dst_entry *dst);
12fdb4d3 50static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 51static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 52
29fa0b30
WY		 53static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
54 int dir);
55
bc9b35ad 56static inline bool
200ce96e 57__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 58{
7e1dc7b6
DM		 59 const struct flowi4 *fl4 = &fl->u.ip4;
60
26bff940
AD		 61 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
62 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 63 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
64 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
65 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
66 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 67}
68
bc9b35ad 69static inline bool
200ce96e 70__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 71{
7e1dc7b6
DM		 72 const struct flowi6 *fl6 = &fl->u.ip6;
73
74 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
75 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
76 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
77 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
78 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
79 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 80}
81
bc9b35ad
DM		 82bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
83 unsigned short family)
77681021
AM		 84{
85 switch (family) {
86 case AF_INET:
87 return __xfrm4_selector_match(sel, fl);
88 case AF_INET6:
89 return __xfrm6_selector_match(sel, fl);
90 }
bc9b35ad 91 return false;
77681021
AM		 92}
93
ef8531b6
ED		 94static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
95{
96 struct xfrm_policy_afinfo *afinfo;
97
98 if (unlikely(family >= NPROTO))
99 return NULL;
100 rcu_read_lock();
101 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
102 if (unlikely(!afinfo))
103 rcu_read_unlock();
104 return afinfo;
105}
106
107static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
108{
109 rcu_read_unlock();
110}
111
c5b3cf46 112static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
6418c4e0
DM		 113 const xfrm_address_t *saddr,
114 const xfrm_address_t *daddr,
9bb182a7
YH		 115 int family)
116{
117 struct xfrm_policy_afinfo *afinfo;
118 struct dst_entry *dst;
119
120 afinfo = xfrm_policy_get_afinfo(family);
121 if (unlikely(afinfo == NULL))
122 return ERR_PTR(-EAFNOSUPPORT);
123
c5b3cf46 124 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 125
126 xfrm_policy_put_afinfo(afinfo);
127
128 return dst;
129}
130
25ee3286 131static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 132 xfrm_address_t *prev_saddr,
133 xfrm_address_t *prev_daddr,
25ee3286 134 int family)
1da177e4 135{
c5b3cf46 136 struct net *net = xs_net(x);
66cdb3ca
HX		 137 xfrm_address_t *saddr = &x->props.saddr;
138 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 139 struct dst_entry *dst;
140
9bb182a7 141 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 142 saddr = x->coaddr;
9bb182a7
YH		 143 daddr = prev_daddr;
144 }
145 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
146 saddr = prev_saddr;
66cdb3ca 147 daddr = x->coaddr;
9bb182a7 148 }
1da177e4 149
c5b3cf46 150 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 151
152 if (!IS_ERR(dst)) {
153 if (prev_saddr != saddr)
154 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
155 if (prev_daddr != daddr)
156 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
157 }
1da177e4 158
66cdb3ca 159 return dst;
1da177e4 160}
1da177e4 161
1da177e4
LT		 162static inline unsigned long make_jiffies(long secs)
163{
164 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
165 return MAX_SCHEDULE_TIMEOUT-1;
166 else
a716c119 167 return secs*HZ;
1da177e4
LT		 168}
169
170static void xfrm_policy_timer(unsigned long data)
171{
3e94c2dc 172 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 173 unsigned long now = get_seconds();
1da177e4
LT		 174 long next = LONG_MAX;
175 int warn = 0;
176 int dir;
177
178 read_lock(&xp->lock);
179
ea2dea9d 180 if (unlikely(xp->walk.dead))
1da177e4
LT		 181 goto out;
182
77d8d7a6 183 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 184
185 if (xp->lft.hard_add_expires_seconds) {
186 long tmo = xp->lft.hard_add_expires_seconds +
187 xp->curlft.add_time - now;
188 if (tmo <= 0)
189 goto expired;
190 if (tmo < next)
191 next = tmo;
192 }
193 if (xp->lft.hard_use_expires_seconds) {
194 long tmo = xp->lft.hard_use_expires_seconds +
195 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
196 if (tmo <= 0)
197 goto expired;
198 if (tmo < next)
199 next = tmo;
200 }
201 if (xp->lft.soft_add_expires_seconds) {
202 long tmo = xp->lft.soft_add_expires_seconds +
203 xp->curlft.add_time - now;
204 if (tmo <= 0) {
205 warn = 1;
206 tmo = XFRM_KM_TIMEOUT;
207 }
208 if (tmo < next)
209 next = tmo;
210 }
211 if (xp->lft.soft_use_expires_seconds) {
212 long tmo = xp->lft.soft_use_expires_seconds +
213 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
214 if (tmo <= 0) {
215 warn = 1;
216 tmo = XFRM_KM_TIMEOUT;
217 }
218 if (tmo < next)
219 next = tmo;
220 }
221
222 if (warn)
6c5c8ca7 223 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 224 if (next != LONG_MAX &&
225 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
226 xfrm_pol_hold(xp);
227
228out:
229 read_unlock(&xp->lock);
230 xfrm_pol_put(xp);
231 return;
232
233expired:
234 read_unlock(&xp->lock);
4666faab 235 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 236 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 237 xfrm_pol_put(xp);
238}
239
fe1a5f03
TT		 240static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
241{
242 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
243
244 if (unlikely(pol->walk.dead))
245 flo = NULL;
246 else
247 xfrm_pol_hold(pol);
248
249 return flo;
250}
251
252static int xfrm_policy_flo_check(struct flow_cache_object *flo)
253{
254 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
255
256 return !pol->walk.dead;
257}
258
259static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
260{
261 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
262}
263
264static const struct flow_cache_ops xfrm_policy_fc_ops = {
265 .get = xfrm_policy_flo_get,
266 .check = xfrm_policy_flo_check,
267 .delete = xfrm_policy_flo_delete,
268};
1da177e4
LT		 269
270/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
271 * SPD calls.
272 */
273
0331b1f3 274struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 275{
276 struct xfrm_policy *policy;
277
0da974f4 278 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 279
280 if (policy) {
0331b1f3 281 write_pnet(&policy->xp_net, net);
12a169e7 282 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 283 INIT_HLIST_NODE(&policy->bydst);
284 INIT_HLIST_NODE(&policy->byidx);
1da177e4 285 rwlock_init(&policy->lock);
2518c7c2 286 atomic_set(&policy->refcnt, 1);
a0073fe1 287 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 288 setup_timer(&policy->timer, xfrm_policy_timer,
289 (unsigned long)policy);
a0073fe1
SK		 290 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
291 (unsigned long)policy);
fe1a5f03 292 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 293 }
294 return policy;
295}
296EXPORT_SYMBOL(xfrm_policy_alloc);
297
298/* Destroy xfrm_policy: descendant resources must be released to this moment. */
299
64c31b3f 300void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 301{
12a169e7 302 BUG_ON(!policy->walk.dead);
1da177e4 303
0659eea9 304 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 305 BUG();
306
03e1ad7b 307 security_xfrm_policy_free(policy->security);
1da177e4
LT		 308 kfree(policy);
309}
64c31b3f 310EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 311
a0073fe1
SK		 312static void xfrm_queue_purge(struct sk_buff_head *list)
313{
314 struct sk_buff *skb;
315
302a50bc 316 while ((skb = skb_dequeue(list)) != NULL)
a0073fe1 317 kfree_skb(skb);
a0073fe1
SK		 318}
319
1da177e4
LT		 320/* Rule must be locked. Release descentant resources, announce
321 * entry dead. The rule must be unlinked from lists to the moment.
322 */
323
324static void xfrm_policy_kill(struct xfrm_policy *policy)
325{
12a169e7 326 policy->walk.dead = 1;
1da177e4 327
285ead17 328 atomic_inc(&policy->genid);
1da177e4 329
e7d8f6cb
SK		 330 if (del_timer(&policy->polq.hold_timer))
331 xfrm_pol_put(policy);
a0073fe1
SK		 332 xfrm_queue_purge(&policy->polq.hold_queue);
333
285ead17
TT		 334 if (del_timer(&policy->timer))
335 xfrm_pol_put(policy);
336
337 xfrm_pol_put(policy);
1da177e4
LT		 338}
339
2518c7c2
DM		 340static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
341
e92303f8 342static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 343{
e92303f8 344 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 345}
346
b58555f1
CG		 347/* calculate policy hash thresholds */
348static void __get_hash_thresh(struct net *net,
349 unsigned short family, int dir,
350 u8 *dbits, u8 *sbits)
351{
352 switch (family) {
353 case AF_INET:
354 *dbits = net->xfrm.policy_bydst[dir].dbits4;
355 *sbits = net->xfrm.policy_bydst[dir].sbits4;
356 break;
357
358 case AF_INET6:
359 *dbits = net->xfrm.policy_bydst[dir].dbits6;
360 *sbits = net->xfrm.policy_bydst[dir].sbits6;
361 break;
362
363 default:
364 *dbits = 0;
365 *sbits = 0;
366 }
367}
368
5f803b58
DM		 369static struct hlist_head *policy_hash_bysel(struct net *net,
370 const struct xfrm_selector *sel,
371 unsigned short family, int dir)
2518c7c2 372{
1121994c 373 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 374 unsigned int hash;
375 u8 dbits;
376 u8 sbits;
377
378 __get_hash_thresh(net, family, dir, &dbits, &sbits);
379 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2
DM		 380
381 return (hash == hmask + 1 ?
1121994c
AD		 382 &net->xfrm.policy_inexact[dir] :
383 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 384}
385
5f803b58
DM		 386static struct hlist_head *policy_hash_direct(struct net *net,
387 const xfrm_address_t *daddr,
388 const xfrm_address_t *saddr,
389 unsigned short family, int dir)
2518c7c2 390{
1121994c 391 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 392 unsigned int hash;
393 u8 dbits;
394 u8 sbits;
395
396 __get_hash_thresh(net, family, dir, &dbits, &sbits);
397 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 398
1121994c 399 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 400}
401
b58555f1
CG		 402static void xfrm_dst_hash_transfer(struct net *net,
403 struct hlist_head *list,
2518c7c2 404 struct hlist_head *ndsttable,
b58555f1
CG		 405 unsigned int nhashmask,
406 int dir)
2518c7c2 407{
b67bfe0d 408 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 409 struct xfrm_policy *pol;
b791160b 410 unsigned int h0 = 0;
b58555f1
CG		 411 u8 dbits;
412 u8 sbits;
2518c7c2 413
b791160b 414redo:
b67bfe0d 415 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 416 unsigned int h;
417
b58555f1 418 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 419 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 420 pol->family, nhashmask, dbits, sbits);
b791160b 421 if (!entry0) {
b67bfe0d 422 hlist_del(&pol->bydst);
b791160b
YH		 423 hlist_add_head(&pol->bydst, ndsttable+h);
424 h0 = h;
425 } else {
426 if (h != h0)
427 continue;
b67bfe0d 428 hlist_del(&pol->bydst);
1d023284 429 hlist_add_behind(&pol->bydst, entry0);
b791160b 430 }
b67bfe0d 431 entry0 = &pol->bydst;
b791160b
YH		 432 }
433 if (!hlist_empty(list)) {
434 entry0 = NULL;
435 goto redo;
2518c7c2
DM		 436 }
437}
438
439static void xfrm_idx_hash_transfer(struct hlist_head *list,
440 struct hlist_head *nidxtable,
441 unsigned int nhashmask)
442{
b67bfe0d 443 struct hlist_node *tmp;
2518c7c2
DM		 444 struct xfrm_policy *pol;
445
b67bfe0d 446 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 447 unsigned int h;
448
449 h = __idx_hash(pol->index, nhashmask);
450 hlist_add_head(&pol->byidx, nidxtable+h);
451 }
452}
453
454static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
455{
456 return ((old_hmask + 1) << 1) - 1;
457}
458
66caf628 459static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 460{
66caf628 461 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 462 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
463 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 464 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 465 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 466 int i;
467
468 if (!ndst)
469 return;
470
283bc9f3 471 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 472
473 for (i = hmask; i >= 0; i--)
b58555f1 474 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 475
66caf628
AD		 476 net->xfrm.policy_bydst[dir].table = ndst;
477 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 478
283bc9f3 479 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 480
44e36b42 481 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 482}
483
66caf628 484static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 485{
66caf628 486 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 487 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
488 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 489 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 490 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 491 int i;
492
493 if (!nidx)
494 return;
495
283bc9f3 496 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 497
498 for (i = hmask; i >= 0; i--)
499 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
500
66caf628
AD		 501 net->xfrm.policy_byidx = nidx;
502 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 503
283bc9f3 504 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 505
44e36b42 506 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 507}
508
66caf628 509static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 510{
66caf628
AD		 511 unsigned int cnt = net->xfrm.policy_count[dir];
512 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 513
514 if (total)
515 *total += cnt;
516
517 if ((hmask + 1) < xfrm_policy_hashmax &&
518 cnt > hmask)
519 return 1;
520
521 return 0;
522}
523
66caf628 524static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 525{
66caf628 526 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 527
528 if ((hmask + 1) < xfrm_policy_hashmax &&
529 total > hmask)
530 return 1;
531
532 return 0;
533}
534
e071041b 535void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 536{
283bc9f3 537 read_lock_bh(&net->xfrm.xfrm_policy_lock);
e071041b
AD		 538 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
539 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
540 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
541 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
542 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
543 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
544 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 545 si->spdhmcnt = xfrm_policy_hashmax;
283bc9f3 546 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
ecfd6b18
JHS		 547}
548EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 549
ecfd6b18 550static DEFINE_MUTEX(hash_resize_mutex);
66caf628 551static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 552{
66caf628 553 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 554 int dir, total;
555
556 mutex_lock(&hash_resize_mutex);
557
558 total = 0;
559 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 560 if (xfrm_bydst_should_resize(net, dir, &total))
561 xfrm_bydst_resize(net, dir);
2518c7c2 562 }
66caf628
AD		 563 if (xfrm_byidx_should_resize(net, total))
564 xfrm_byidx_resize(net, total);
2518c7c2
DM		 565
566 mutex_unlock(&hash_resize_mutex);
567}
568
1da177e4
LT		 569/* Generate new index... KAME seems to generate them ordered by cost
570 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 571static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 572{
1da177e4
LT		 573 static u32 idx_generator;
574
575 for (;;) {
2518c7c2
DM		 576 struct hlist_head *list;
577 struct xfrm_policy *p;
578 u32 idx;
579 int found;
580
e682adf0
FD		 581 if (!index) {
582 idx = (idx_generator | dir);
583 idx_generator += 8;
584 } else {
585 idx = index;
586 index = 0;
587 }
588
1da177e4
LT		 589 if (idx == 0)
590 idx = 8;
1121994c 591 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 592 found = 0;
b67bfe0d 593 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 594 if (p->index == idx) {
595 found = 1;
1da177e4 596 break;
2518c7c2 597 }
1da177e4 598 }
2518c7c2 599 if (!found)
1da177e4
LT		 600 return idx;
601 }
602}
603
2518c7c2
DM		 604static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
605{
606 u32 *p1 = (u32 *) s1;
607 u32 *p2 = (u32 *) s2;
608 int len = sizeof(struct xfrm_selector) / sizeof(u32);
609 int i;
610
611 for (i = 0; i < len; i++) {
612 if (p1[i] != p2[i])
613 return 1;
614 }
615
616 return 0;
617}
618
a0073fe1
SK		 619static void xfrm_policy_requeue(struct xfrm_policy *old,
620 struct xfrm_policy *new)
621{
622 struct xfrm_policy_queue *pq = &old->polq;
623 struct sk_buff_head list;
624
625 __skb_queue_head_init(&list);
626
627 spin_lock_bh(&pq->hold_queue.lock);
628 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 629 if (del_timer(&pq->hold_timer))
630 xfrm_pol_put(old);
a0073fe1
SK		 631 spin_unlock_bh(&pq->hold_queue.lock);
632
633 if (skb_queue_empty(&list))
634 return;
635
636 pq = &new->polq;
637
638 spin_lock_bh(&pq->hold_queue.lock);
639 skb_queue_splice(&list, &pq->hold_queue);
640 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 641 if (!mod_timer(&pq->hold_timer, jiffies))
642 xfrm_pol_hold(new);
a0073fe1
SK		 643 spin_unlock_bh(&pq->hold_queue.lock);
644}
645
7cb8a939
SK		 646static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
647 struct xfrm_policy *pol)
648{
649 u32 mark = policy->mark.v & policy->mark.m;
650
651 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
652 return true;
653
654 if ((mark & pol->mark.m) == pol->mark.v &&
655 policy->priority == pol->priority)
656 return true;
657
658 return false;
659}
660
1da177e4
LT		 661int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
662{
1121994c 663 struct net *net = xp_net(policy);
2518c7c2
DM		 664 struct xfrm_policy *pol;
665 struct xfrm_policy *delpol;
666 struct hlist_head *chain;
b67bfe0d 667 struct hlist_node *newpos;
1da177e4 668
283bc9f3 669 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 670 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 671 delpol = NULL;
672 newpos = NULL;
b67bfe0d 673 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 674 if (pol->type == policy->type &&
2518c7c2 675 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 676 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 677 xfrm_sec_ctx_match(pol->security, policy->security) &&
678 !WARN_ON(delpol)) {
1da177e4 679 if (excl) {
283bc9f3 680 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 681 return -EEXIST;
682 }
1da177e4
LT		 683 delpol = pol;
684 if (policy->priority > pol->priority)
685 continue;
686 } else if (policy->priority >= pol->priority) {
a6c7ab55 687 newpos = &pol->bydst;
1da177e4
LT		 688 continue;
689 }
1da177e4
LT		 690 if (delpol)
691 break;
1da177e4
LT		 692 }
693 if (newpos)
1d023284 694 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 695 else
696 hlist_add_head(&policy->bydst, chain);
1da177e4 697 xfrm_pol_hold(policy);
1121994c 698 net->xfrm.policy_count[dir]++;
ca925cf1 699 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 700
701 /* After previous checking, family can either be AF_INET or AF_INET6 */
702 if (policy->family == AF_INET)
703 rt_genid_bump_ipv4(net);
704 else
705 rt_genid_bump_ipv6(net);
706
a0073fe1
SK		 707 if (delpol) {
708 xfrm_policy_requeue(delpol, policy);
29fa0b30 709 __xfrm_policy_unlink(delpol, dir);
a0073fe1 710 }
e682adf0 711 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 712 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 713 policy->curlft.add_time = get_seconds();
1da177e4
LT		 714 policy->curlft.use_time = 0;
715 if (!mod_timer(&policy->timer, jiffies + HZ))
716 xfrm_pol_hold(policy);
1121994c 717 list_add(&policy->walk.all, &net->xfrm.policy_all);
283bc9f3 718 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 719
9b78a82c 720 if (delpol)
1da177e4 721 xfrm_policy_kill(delpol);
1121994c
AD		 722 else if (xfrm_bydst_should_resize(net, dir, NULL))
723 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 724
1da177e4
LT		 725 return 0;
726}
727EXPORT_SYMBOL(xfrm_policy_insert);
728
8ca2e93b
JHS		 729struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
730 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 731 struct xfrm_sec_ctx *ctx, int delete,
732 int *err)
1da177e4 733{
2518c7c2
DM		 734 struct xfrm_policy *pol, *ret;
735 struct hlist_head *chain;
1da177e4 736
ef41aaa0 737 *err = 0;
283bc9f3 738 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 739 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 740 ret = NULL;
b67bfe0d 741 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 742 if (pol->type == type &&
34f8d884 743 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 744 !selector_cmp(sel, &pol->selector) &&
745 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 746 xfrm_pol_hold(pol);
2518c7c2 747 if (delete) {
03e1ad7b
PM		 748 *err = security_xfrm_policy_delete(
749 pol->security);
ef41aaa0 750 if (*err) {
283bc9f3 751 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 752 return pol;
753 }
29fa0b30 754 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 755 }
756 ret = pol;
1da177e4
LT		 757 break;
758 }
759 }
283bc9f3 760 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 761
fe1a5f03 762 if (ret && delete)
2518c7c2 763 xfrm_policy_kill(ret);
2518c7c2 764 return ret;
1da177e4 765}
df71837d 766EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 767
8ca2e93b
JHS		 768struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
769 int dir, u32 id, int delete, int *err)
1da177e4 770{
2518c7c2
DM		 771 struct xfrm_policy *pol, *ret;
772 struct hlist_head *chain;
1da177e4 773
b5505c6e
HX		 774 *err = -ENOENT;
775 if (xfrm_policy_id2dir(id) != dir)
776 return NULL;
777
ef41aaa0 778 *err = 0;
283bc9f3 779 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 780 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 781 ret = NULL;
b67bfe0d 782 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 783 if (pol->type == type && pol->index == id &&
784 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 785 xfrm_pol_hold(pol);
2518c7c2 786 if (delete) {
03e1ad7b
PM		 787 *err = security_xfrm_policy_delete(
788 pol->security);
ef41aaa0 789 if (*err) {
283bc9f3 790 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 791 return pol;
792 }
29fa0b30 793 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 794 }
795 ret = pol;
1da177e4
LT		 796 break;
797 }
798 }
283bc9f3 799 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 800
fe1a5f03 801 if (ret && delete)
2518c7c2 802 xfrm_policy_kill(ret);
2518c7c2 803 return ret;
1da177e4
LT		 804}
805EXPORT_SYMBOL(xfrm_policy_byid);
806
4aa2e62c
JL		 807#ifdef CONFIG_SECURITY_NETWORK_XFRM
808static inline int
2e71029e 809xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 810{
4aa2e62c
JL		 811 int dir, err = 0;
812
813 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
814 struct xfrm_policy *pol;
4aa2e62c
JL		 815 int i;
816
b67bfe0d 817 hlist_for_each_entry(pol,
33ffbbd5 818 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 819 if (pol->type != type)
820 continue;
03e1ad7b 821 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 822 if (err) {
2e71029e 823 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 824 return err;
825 }
7dc12d6d 826 }
33ffbbd5 827 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 828 hlist_for_each_entry(pol,
33ffbbd5 829 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 830 bydst) {
831 if (pol->type != type)
832 continue;
03e1ad7b
PM		 833 err = security_xfrm_policy_delete(
834 pol->security);
4aa2e62c 835 if (err) {
ab5f5e8b 836 xfrm_audit_policy_delete(pol, 0,
2e71029e 837 task_valid);
4aa2e62c
JL		 838 return err;
839 }
840 }
841 }
842 }
843 return err;
844}
845#else
846static inline int
2e71029e 847xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 848{
849 return 0;
850}
851#endif
852
2e71029e 853int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 854{
2f1eb65f 855 int dir, err = 0, cnt = 0;
1da177e4 856
283bc9f3 857 write_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 858
2e71029e 859 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 860 if (err)
861 goto out;
862
1da177e4 863 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 864 struct xfrm_policy *pol;
29fa0b30 865 int i;
2518c7c2
DM		 866
867 again1:
b67bfe0d 868 hlist_for_each_entry(pol,
33ffbbd5 869 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 870 if (pol->type != type)
871 continue;
ea2dea9d 872 __xfrm_policy_unlink(pol, dir);
283bc9f3 873 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 874 cnt++;
1da177e4 875
2e71029e 876 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 877
2518c7c2 878 xfrm_policy_kill(pol);
1da177e4 879
283bc9f3 880 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 881 goto again1;
882 }
883
33ffbbd5 884 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 885 again2:
b67bfe0d 886 hlist_for_each_entry(pol,
33ffbbd5 887 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 888 bydst) {
889 if (pol->type != type)
890 continue;
ea2dea9d 891 __xfrm_policy_unlink(pol, dir);
283bc9f3 892 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 893 cnt++;
2518c7c2 894
2e71029e 895 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 896 xfrm_policy_kill(pol);
897
283bc9f3 898 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 899 goto again2;
900 }
1da177e4 901 }
2518c7c2 902
1da177e4 903 }
2f1eb65f
JHS		 904 if (!cnt)
905 err = -ESRCH;
4aa2e62c 906out:
283bc9f3 907 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 908 return err;
1da177e4
LT		 909}
910EXPORT_SYMBOL(xfrm_policy_flush);
911
cdcbca7c 912int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 913 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 914 void *data)
915{
12a169e7
HX		 916 struct xfrm_policy *pol;
917 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 918 int error = 0;
919
920 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
921 walk->type != XFRM_POLICY_TYPE_ANY)
922 return -EINVAL;
1da177e4 923
12a169e7 924 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 925 return 0;
926
283bc9f3 927 write_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 928 if (list_empty(&walk->walk.all))
cdcbca7c 929 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 930 else
931 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 932 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 933 if (x->dead)
4c563f76 934 continue;
12a169e7
HX		 935 pol = container_of(x, struct xfrm_policy, walk);
936 if (walk->type != XFRM_POLICY_TYPE_ANY &&
937 walk->type != pol->type)
938 continue;
939 error = func(pol, xfrm_policy_id2dir(pol->index),
940 walk->seq, data);
941 if (error) {
942 list_move_tail(&walk->walk.all, &x->all);
943 goto out;
2518c7c2 944 }
12a169e7 945 walk->seq++;
1da177e4 946 }
12a169e7 947 if (walk->seq == 0) {
baf5d743
JHS		 948 error = -ENOENT;
949 goto out;
950 }
12a169e7 951 list_del_init(&walk->walk.all);
1da177e4 952out:
283bc9f3 953 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 954 return error;
955}
956EXPORT_SYMBOL(xfrm_policy_walk);
957
12a169e7
HX		 958void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
959{
960 INIT_LIST_HEAD(&walk->walk.all);
961 walk->walk.dead = 1;
962 walk->type = type;
963 walk->seq = 0;
964}
965EXPORT_SYMBOL(xfrm_policy_walk_init);
966
283bc9f3 967void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 968{
969 if (list_empty(&walk->walk.all))
970 return;
971
283bc9f3 972 write_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 973 list_del(&walk->walk.all);
283bc9f3 974 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 975}
976EXPORT_SYMBOL(xfrm_policy_walk_done);
977
134b0fc5
JM		 978/*
979 * Find policy to apply to this flow.
980 *
981 * Returns 0 if policy found, else an -errno.
982 */
f299d557
DM		 983static int xfrm_policy_match(const struct xfrm_policy *pol,
984 const struct flowi *fl,
2518c7c2 985 u8 type, u16 family, int dir)
1da177e4 986{
f299d557 987 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 988 int ret = -ESRCH;
989 bool match;
1da177e4 990
2518c7c2 991 if (pol->family != family ||
1d28f42c 992 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 993 pol->type != type)
134b0fc5 994 return ret;
1da177e4 995
2518c7c2 996 match = xfrm_selector_match(sel, fl, family);
134b0fc5 997 if (match)
1d28f42c 998 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 999 dir);
2518c7c2 1000
134b0fc5 1001 return ret;
2518c7c2 1002}
1da177e4 1003
52479b62 1004static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1005 const struct flowi *fl,
2518c7c2
DM		 1006 u16 family, u8 dir)
1007{
134b0fc5 1008 int err;
2518c7c2 1009 struct xfrm_policy *pol, *ret;
0b597e7e 1010 const xfrm_address_t *daddr, *saddr;
2518c7c2 1011 struct hlist_head *chain;
acba48e1 1012 u32 priority = ~0U;
df71837d 1013
2518c7c2
DM		 1014 daddr = xfrm_flowi_daddr(fl, family);
1015 saddr = xfrm_flowi_saddr(fl, family);
1016 if (unlikely(!daddr || !saddr))
1017 return NULL;
1018
283bc9f3 1019 read_lock_bh(&net->xfrm.xfrm_policy_lock);
52479b62 1020 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 1021 ret = NULL;
b67bfe0d 1022 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1023 err = xfrm_policy_match(pol, fl, type, family, dir);
1024 if (err) {
1025 if (err == -ESRCH)
1026 continue;
1027 else {
1028 ret = ERR_PTR(err);
1029 goto fail;
1030 }
1031 } else {
2518c7c2 1032 ret = pol;
acba48e1 1033 priority = ret->priority;
2518c7c2
DM		 1034 break;
1035 }
1036 }
52479b62 1037 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1038 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1039 err = xfrm_policy_match(pol, fl, type, family, dir);
1040 if (err) {
1041 if (err == -ESRCH)
1042 continue;
1043 else {
1044 ret = ERR_PTR(err);
1045 goto fail;
1046 }
1047 } else if (pol->priority < priority) {
acba48e1
DM		 1048 ret = pol;
1049 break;
1da177e4
LT		 1050 }
1051 }
acba48e1
DM		 1052 if (ret)
1053 xfrm_pol_hold(ret);
134b0fc5 1054fail:
283bc9f3 1055 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
4e81bb83 1056
2518c7c2 1057 return ret;
4e81bb83
MN		 1058}
1059
80c802f3 1060static struct xfrm_policy *
73ff93cd 1061__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1062{
1063#ifdef CONFIG_XFRM_SUB_POLICY
1064 struct xfrm_policy *pol;
1065
1066 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1067 if (pol != NULL)
1068 return pol;
1069#endif
1070 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1071}
1072
b5fb82c4
BZ		 1073static int flow_to_policy_dir(int dir)
1074{
1075 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1076 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1077 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1078 return dir;
1079
1080 switch (dir) {
1081 default:
1082 case FLOW_DIR_IN:
1083 return XFRM_POLICY_IN;
1084 case FLOW_DIR_OUT:
1085 return XFRM_POLICY_OUT;
1086 case FLOW_DIR_FWD:
1087 return XFRM_POLICY_FWD;
1088 }
1089}
1090
fe1a5f03 1091static struct flow_cache_object *
dee9f4bc 1092xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1093 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1094{
1095 struct xfrm_policy *pol;
fe1a5f03
TT		 1096
1097 if (old_obj)
1098 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1099
b5fb82c4 1100 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1101 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1102 return ERR_CAST(pol);
fe1a5f03 1103
fe1a5f03
TT		 1104 /* Resolver returns two references:
1105 * one for cache and one for caller of flow_cache_lookup() */
1106 xfrm_pol_hold(pol);
1107
1108 return &pol->flo;
1da177e4
LT		 1109}
1110
df71837d
TJ		 1111static inline int policy_to_flow_dir(int dir)
1112{
1113 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1114 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1115 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1116 return dir;
1117 switch (dir) {
1118 default:
1119 case XFRM_POLICY_IN:
1120 return FLOW_DIR_IN;
1121 case XFRM_POLICY_OUT:
1122 return FLOW_DIR_OUT;
1123 case XFRM_POLICY_FWD:
1124 return FLOW_DIR_FWD;
3ff50b79 1125 }
df71837d
TJ		 1126}
1127
dee9f4bc
DM		 1128static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1129 const struct flowi *fl)
1da177e4
LT		 1130{
1131 struct xfrm_policy *pol;
283bc9f3 1132 struct net *net = sock_net(sk);
1da177e4 1133
283bc9f3 1134 read_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1135 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1136 bool match = xfrm_selector_match(&pol->selector, fl,
1137 sk->sk_family);
a716c119 1138 int err = 0;
df71837d 1139
3bccfbc7 1140 if (match) {
34f8d884
JHS		 1141 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1142 pol = NULL;
1143 goto out;
1144 }
03e1ad7b 1145 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1146 fl->flowi_secid,
03e1ad7b 1147 policy_to_flow_dir(dir));
3bccfbc7
VY		 1148 if (!err)
1149 xfrm_pol_hold(pol);
1150 else if (err == -ESRCH)
1151 pol = NULL;
1152 else
1153 pol = ERR_PTR(err);
1154 } else
1da177e4
LT		 1155 pol = NULL;
1156 }
34f8d884 1157out:
283bc9f3 1158 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1159 return pol;
1160}
1161
1162static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1163{
98806f75 1164 struct net *net = xp_net(pol);
1121994c 1165 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1166 pol->family, dir);
4e81bb83 1167
98806f75 1168 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1169 hlist_add_head(&pol->bydst, chain);
e92303f8 1170 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1171 net->xfrm.policy_count[dir]++;
1da177e4 1172 xfrm_pol_hold(pol);
2518c7c2 1173
98806f75
AD		 1174 if (xfrm_bydst_should_resize(net, dir, NULL))
1175 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1176}
1177
1178static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1179 int dir)
1180{
98806f75
AD		 1181 struct net *net = xp_net(pol);
1182
2518c7c2
DM		 1183 if (hlist_unhashed(&pol->bydst))
1184 return NULL;
1da177e4 1185
3a9016f9 1186 hlist_del_init(&pol->bydst);
2518c7c2 1187 hlist_del(&pol->byidx);
12a169e7 1188 list_del(&pol->walk.all);
98806f75 1189 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1190
1191 return pol;
1da177e4
LT		 1192}
1193
4666faab 1194int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1195{
283bc9f3
FD		 1196 struct net *net = xp_net(pol);
1197
1198 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1199 pol = __xfrm_policy_unlink(pol, dir);
283bc9f3 1200 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1201 if (pol) {
1da177e4 1202 xfrm_policy_kill(pol);
4666faab 1203 return 0;
1da177e4 1204 }
4666faab 1205 return -ENOENT;
1da177e4 1206}
a70fcb0b 1207EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1208
1209int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1210{
1121994c 1211 struct net *net = xp_net(pol);
1da177e4
LT		 1212 struct xfrm_policy *old_pol;
1213
4e81bb83
MN		 1214#ifdef CONFIG_XFRM_SUB_POLICY
1215 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1216 return -EINVAL;
1217#endif
1218
283bc9f3 1219 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1220 old_pol = sk->sk_policy[dir];
1221 sk->sk_policy[dir] = pol;
1222 if (pol) {
9d729f72 1223 pol->curlft.add_time = get_seconds();
e682adf0 1224 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
1da177e4
LT		 1225 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1226 }
a0073fe1
SK		 1227 if (old_pol) {
1228 if (pol)
1229 xfrm_policy_requeue(old_pol, pol);
1230
ea2dea9d
TT		 1231 /* Unlinking succeeds always. This is the only function
1232 * allowed to delete or replace socket policy.
1233 */
1da177e4 1234 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
a0073fe1 1235 }
283bc9f3 1236 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1237
1238 if (old_pol) {
1239 xfrm_policy_kill(old_pol);
1240 }
1241 return 0;
1242}
1243
d3e40a9f 1244static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1245{
0331b1f3 1246 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1247 struct net *net = xp_net(old);
1da177e4
LT		 1248
1249 if (newp) {
1250 newp->selector = old->selector;
03e1ad7b
PM		 1251 if (security_xfrm_policy_clone(old->security,
1252 &newp->security)) {
df71837d
TJ		 1253 kfree(newp);
1254 return NULL; /* ENOMEM */
1255 }
1da177e4
LT		 1256 newp->lft = old->lft;
1257 newp->curlft = old->curlft;
fb977e2c 1258 newp->mark = old->mark;
1da177e4
LT		 1259 newp->action = old->action;
1260 newp->flags = old->flags;
1261 newp->xfrm_nr = old->xfrm_nr;
1262 newp->index = old->index;
4e81bb83 1263 newp->type = old->type;
1da177e4
LT		 1264 memcpy(newp->xfrm_vec, old->xfrm_vec,
1265 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
283bc9f3 1266 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1267 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
283bc9f3 1268 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1269 xfrm_pol_put(newp);
1270 }
1271 return newp;
1272}
1273
1274int __xfrm_sk_clone_policy(struct sock *sk)
1275{
1276 struct xfrm_policy *p0 = sk->sk_policy[0],
1277 *p1 = sk->sk_policy[1];
1278
1279 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1280 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1281 return -ENOMEM;
1282 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1283 return -ENOMEM;
1284 return 0;
1285}
1286
a1e59abf 1287static int
fbda33b2 1288xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1289 unsigned short family)
1290{
1291 int err;
1292 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1293
1294 if (unlikely(afinfo == NULL))
1295 return -EINVAL;
fbda33b2 1296 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1297 xfrm_policy_put_afinfo(afinfo);
1298 return err;
1299}
1300
1da177e4
LT		 1301/* Resolve list of templates for the flow, given policy. */
1302
1303static int
a6c2e611
DM		 1304xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1305 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1306{
fbda33b2 1307 struct net *net = xp_net(policy);
1da177e4
LT		 1308 int nx;
1309 int i, error;
1310 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1311 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1312 xfrm_address_t tmp;
1da177e4 1313
9b7a787d 1314 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1315 struct xfrm_state *x;
1316 xfrm_address_t *remote = daddr;
1317 xfrm_address_t *local = saddr;
1318 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1319
48b8d783
JK		 1320 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1321 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1322 remote = &tmpl->id.daddr;
1323 local = &tmpl->saddr;
8444cf71
TE		 1324 if (xfrm_addr_any(local, tmpl->encap_family)) {
1325 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1326 if (error)
1327 goto fail;
1328 local = &tmp;
1329 }
1da177e4
LT		 1330 }
1331
1332 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1333
1334 if (x && x->km.state == XFRM_STATE_VALID) {
1335 xfrm[nx++] = x;
1336 daddr = remote;
1337 saddr = local;
1338 continue;
1339 }
1340 if (x) {
1341 error = (x->km.state == XFRM_STATE_ERROR ?
1342 -EINVAL : -EAGAIN);
1343 xfrm_state_put(x);
42054569 1344 } else if (error == -ESRCH) {
a4322266 1345 error = -EAGAIN;
42054569 1346 }
1da177e4
LT		 1347
1348 if (!tmpl->optional)
1349 goto fail;
1350 }
1351 return nx;
1352
1353fail:
9b7a787d 1354 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1355 xfrm_state_put(xfrm[nx]);
1356 return error;
1357}
1358
4e81bb83 1359static int
a6c2e611
DM		 1360xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1361 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1362{
41a49cc3
MN		 1363 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1364 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1365 int cnx = 0;
1366 int error;
1367 int ret;
1368 int i;
1369
1370 for (i = 0; i < npols; i++) {
1371 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1372 error = -ENOBUFS;
1373 goto fail;
1374 }
41a49cc3
MN		 1375
1376 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1377 if (ret < 0) {
1378 error = ret;
1379 goto fail;
1380 } else
1381 cnx += ret;
1382 }
1383
41a49cc3
MN		 1384 /* found states are sorted for outbound processing */
1385 if (npols > 1)
1386 xfrm_state_sort(xfrm, tpp, cnx, family);
1387
4e81bb83
MN		 1388 return cnx;
1389
1390 fail:
9b7a787d 1391 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1392 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1393 return error;
1394
1395}
1396
1da177e4
LT		 1397/* Check that the bundle accepts the flow and its components are
1398 * still valid.
1399 */
1400
05d84025 1401static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1402{
1403 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1404 int tos;
1da177e4 1405
25ee3286
HX		 1406 if (!afinfo)
1407 return -EINVAL;
1408
1409 tos = afinfo->get_tos(fl);
1410
1411 xfrm_policy_put_afinfo(afinfo);
1412
1413 return tos;
1414}
1415
80c802f3
TT		 1416static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1417{
1418 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1419 struct dst_entry *dst = &xdst->u.dst;
1420
1421 if (xdst->route == NULL) {
1422 /* Dummy bundle - if it has xfrms we were not
1423 * able to build bundle as template resolution failed.
1424 * It means we need to try again resolving. */
1425 if (xdst->num_xfrms > 0)
1426 return NULL;
a0073fe1
SK		 1427 } else if (dst->flags & DST_XFRM_QUEUE) {
1428 return NULL;
80c802f3
TT		 1429 } else {
1430 /* Real bundle */
1431 if (stale_bundle(dst))
1432 return NULL;
1433 }
1434
1435 dst_hold(dst);
1436 return flo;
1437}
1438
1439static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1440{
1441 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1442 struct dst_entry *dst = &xdst->u.dst;
1443
1444 if (!xdst->route)
1445 return 0;
1446 if (stale_bundle(dst))
1447 return 0;
1448
1449 return 1;
1450}
1451
1452static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1453{
1454 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1455 struct dst_entry *dst = &xdst->u.dst;
1456
1457 dst_free(dst);
1458}
1459
1460static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1461 .get = xfrm_bundle_flo_get,
1462 .check = xfrm_bundle_flo_check,
1463 .delete = xfrm_bundle_flo_delete,
1464};
1465
d7c7544c 1466static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1467{
1da177e4 1468 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1469 struct dst_ops *dst_ops;
25ee3286
HX		 1470 struct xfrm_dst *xdst;
1471
1472 if (!afinfo)
1473 return ERR_PTR(-EINVAL);
1474
d7c7544c
AD		 1475 switch (family) {
1476 case AF_INET:
1477 dst_ops = &net->xfrm.xfrm4_dst_ops;
1478 break;
dfd56b8b 1479#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1480 case AF_INET6:
1481 dst_ops = &net->xfrm.xfrm6_dst_ops;
1482 break;
1483#endif
1484 default:
1485 BUG();
1486 }
f5b0a874 1487 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1488
d4cae562 1489 if (likely(xdst)) {
141e369d
SK		 1490 struct dst_entry *dst = &xdst->u.dst;
1491
1492 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1493 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1494 if (afinfo->init_dst)
1495 afinfo->init_dst(net, xdst);
d4cae562 1496 } else
0b150932 1497 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1498
d4cae562
MB		 1499 xfrm_policy_put_afinfo(afinfo);
1500
25ee3286
HX		 1501 return xdst;
1502}
1503
a1b05140
MN		 1504static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1505 int nfheader_len)
1506{
1507 struct xfrm_policy_afinfo *afinfo =
1508 xfrm_policy_get_afinfo(dst->ops->family);
1509 int err;
1510
1511 if (!afinfo)
1512 return -EINVAL;
1513
1514 err = afinfo->init_path(path, dst, nfheader_len);
1515
1516 xfrm_policy_put_afinfo(afinfo);
1517
1518 return err;
1519}
1520
87c1e12b 1521static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1522 const struct flowi *fl)
25ee3286
HX		 1523{
1524 struct xfrm_policy_afinfo *afinfo =
1525 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1526 int err;
1527
1528 if (!afinfo)
1da177e4 1529 return -EINVAL;
25ee3286 1530
87c1e12b 1531 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1532
1da177e4 1533 xfrm_policy_put_afinfo(afinfo);
25ee3286 1534
1da177e4
LT		 1535 return err;
1536}
1537
80c802f3 1538
25ee3286
HX		 1539/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1540 * all the metrics... Shortly, bundle a bundle.
1541 */
1542
1543static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1544 struct xfrm_state **xfrm, int nx,
98313ada 1545 const struct flowi *fl,
25ee3286
HX		 1546 struct dst_entry *dst)
1547{
d7c7544c 1548 struct net *net = xp_net(policy);
25ee3286
HX		 1549 unsigned long now = jiffies;
1550 struct net_device *dev;
43a4dea4 1551 struct xfrm_mode *inner_mode;
25ee3286
HX		 1552 struct dst_entry *dst_prev = NULL;
1553 struct dst_entry *dst0 = NULL;
1554 int i = 0;
1555 int err;
1556 int header_len = 0;
a1b05140 1557 int nfheader_len = 0;
25ee3286
HX		 1558 int trailer_len = 0;
1559 int tos;
1560 int family = policy->selector.family;
9bb182a7
YH		 1561 xfrm_address_t saddr, daddr;
1562
1563 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1564
1565 tos = xfrm_get_tos(fl, family);
1566 err = tos;
1567 if (tos < 0)
1568 goto put_states;
1569
1570 dst_hold(dst);
1571
1572 for (; i < nx; i++) {
d7c7544c 1573 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1574 struct dst_entry *dst1 = &xdst->u.dst;
1575
1576 err = PTR_ERR(xdst);
1577 if (IS_ERR(xdst)) {
1578 dst_release(dst);
1579 goto put_states;
1580 }
1581
43a4dea4
SK		 1582 if (xfrm[i]->sel.family == AF_UNSPEC) {
1583 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1584 xfrm_af2proto(family));
1585 if (!inner_mode) {
1586 err = -EAFNOSUPPORT;
1587 dst_release(dst);
1588 goto put_states;
1589 }
1590 } else
1591 inner_mode = xfrm[i]->inner_mode;
1592
25ee3286
HX		 1593 if (!dst_prev)
1594 dst0 = dst1;
1595 else {
1596 dst_prev->child = dst_clone(dst1);
1597 dst1->flags |= DST_NOHASH;
1598 }
1599
1600 xdst->route = dst;
defb3519 1601 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1602
1603 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1604 family = xfrm[i]->props.family;
9bb182a7
YH		 1605 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1606 family);
25ee3286
HX		 1607 err = PTR_ERR(dst);
1608 if (IS_ERR(dst))
1609 goto put_states;
1610 } else
1611 dst_hold(dst);
1612
1613 dst1->xfrm = xfrm[i];
80c802f3 1614 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1615
f5b0a874 1616 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1617 dst1->flags |= DST_HOST;
1618 dst1->lastuse = now;
1619
1620 dst1->input = dst_discard;
43a4dea4 1621 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1622
1623 dst1->next = dst_prev;
1624 dst_prev = dst1;
1625
1626 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1627 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1628 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1629 trailer_len += xfrm[i]->props.trailer_len;
1630 }
1631
1632 dst_prev->child = dst;
1633 dst0->path = dst;
1634
1635 err = -ENODEV;
1636 dev = dst->dev;
1637 if (!dev)
1638 goto free_dst;
1639
a1b05140 1640 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1641 xfrm_init_pmtu(dst_prev);
1642
1643 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1644 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1645
87c1e12b 1646 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1647 if (err)
1648 goto free_dst;
1649
1650 dst_prev->header_len = header_len;
1651 dst_prev->trailer_len = trailer_len;
1652 header_len -= xdst->u.dst.xfrm->props.header_len;
1653 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1654 }
1655
1656out:
1657 return dst0;
1658
1659put_states:
1660 for (; i < nx; i++)
1661 xfrm_state_put(xfrm[i]);
1662free_dst:
1663 if (dst0)
1664 dst_free(dst0);
1665 dst0 = ERR_PTR(err);
1666 goto out;
1667}
1668
da7c224b 1669#ifdef CONFIG_XFRM_SUB_POLICY
be7928d2 1670static int xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1671{
1672 if (!*target) {
1673 *target = kmalloc(size, GFP_ATOMIC);
1674 if (!*target)
1675 return -ENOMEM;
1676 }
be7928d2 1677
157bfc25
MN		 1678 memcpy(*target, src, size);
1679 return 0;
1680}
da7c224b 1681#endif
157bfc25 1682
be7928d2
DB		 1683static int xfrm_dst_update_parent(struct dst_entry *dst,
1684 const struct xfrm_selector *sel)
157bfc25
MN		 1685{
1686#ifdef CONFIG_XFRM_SUB_POLICY
1687 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1688 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1689 sel, sizeof(*sel));
1690#else
1691 return 0;
1692#endif
1693}
1694
be7928d2
DB		 1695static int xfrm_dst_update_origin(struct dst_entry *dst,
1696 const struct flowi *fl)
157bfc25
MN		 1697{
1698#ifdef CONFIG_XFRM_SUB_POLICY
1699 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1700 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1701#else
1702 return 0;
1703#endif
1704}
1da177e4 1705
73ff93cd 1706static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1707 struct xfrm_policy **pols,
1708 int *num_pols, int *num_xfrms)
1709{
1710 int i;
1711
1712 if (*num_pols == 0 || !pols[0]) {
1713 *num_pols = 0;
1714 *num_xfrms = 0;
1715 return 0;
1716 }
1717 if (IS_ERR(pols[0]))
1718 return PTR_ERR(pols[0]);
1719
1720 *num_xfrms = pols[0]->xfrm_nr;
1721
1722#ifdef CONFIG_XFRM_SUB_POLICY
1723 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1724 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1725 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1726 XFRM_POLICY_TYPE_MAIN,
1727 fl, family,
1728 XFRM_POLICY_OUT);
1729 if (pols[1]) {
1730 if (IS_ERR(pols[1])) {
1731 xfrm_pols_put(pols, *num_pols);
1732 return PTR_ERR(pols[1]);
1733 }
02d0892f 1734 (*num_pols)++;
80c802f3
TT		 1735 (*num_xfrms) += pols[1]->xfrm_nr;
1736 }
1737 }
1738#endif
1739 for (i = 0; i < *num_pols; i++) {
1740 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1741 *num_xfrms = -1;
1742 break;
1743 }
1744 }
1745
1746 return 0;
1747
1748}
1749
1750static struct xfrm_dst *
1751xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1752 const struct flowi *fl, u16 family,
80c802f3
TT		 1753 struct dst_entry *dst_orig)
1754{
1755 struct net *net = xp_net(pols[0]);
1756 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1757 struct dst_entry *dst;
1758 struct xfrm_dst *xdst;
1759 int err;
1760
1761 /* Try to instantiate a bundle */
1762 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1763 if (err <= 0) {
1764 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1765 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1766 return ERR_PTR(err);
1767 }
1768
1769 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1770 if (IS_ERR(dst)) {
1771 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1772 return ERR_CAST(dst);
1773 }
1774
1775 xdst = (struct xfrm_dst *)dst;
1776 xdst->num_xfrms = err;
1777 if (num_pols > 1)
1778 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1779 else
1780 err = xfrm_dst_update_origin(dst, fl);
1781 if (unlikely(err)) {
1782 dst_free(dst);
1783 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1784 return ERR_PTR(err);
1785 }
1786
1787 xdst->num_pols = num_pols;
3e94c2dc 1788 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1789 xdst->policy_genid = atomic_read(&pols[0]->genid);
1790
1791 return xdst;
1792}
1793
a0073fe1
SK		 1794static void xfrm_policy_queue_process(unsigned long arg)
1795{
1796 int err = 0;
1797 struct sk_buff *skb;
1798 struct sock *sk;
1799 struct dst_entry *dst;
a0073fe1
SK		 1800 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1801 struct xfrm_policy_queue *pq = &pol->polq;
1802 struct flowi fl;
1803 struct sk_buff_head list;
1804
1805 spin_lock(&pq->hold_queue.lock);
1806 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1807 if (!skb) {
1808 spin_unlock(&pq->hold_queue.lock);
1809 goto out;
1810 }
a0073fe1
SK		 1811 dst = skb_dst(skb);
1812 sk = skb->sk;
1813 xfrm_decode_session(skb, &fl, dst->ops->family);
1814 spin_unlock(&pq->hold_queue.lock);
1815
1816 dst_hold(dst->path);
1817 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1818 sk, 0);
1819 if (IS_ERR(dst))
1820 goto purge_queue;
1821
1822 if (dst->flags & DST_XFRM_QUEUE) {
1823 dst_release(dst);
1824
1825 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1826 goto purge_queue;
1827
1828 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1829 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1830 xfrm_pol_hold(pol);
1831 goto out;
a0073fe1
SK		 1832 }
1833
1834 dst_release(dst);
1835
1836 __skb_queue_head_init(&list);
1837
1838 spin_lock(&pq->hold_queue.lock);
1839 pq->timeout = 0;
1840 skb_queue_splice_init(&pq->hold_queue, &list);
1841 spin_unlock(&pq->hold_queue.lock);
1842
1843 while (!skb_queue_empty(&list)) {
1844 skb = __skb_dequeue(&list);
1845
1846 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1847 dst_hold(skb_dst(skb)->path);
1848 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1849 &fl, skb->sk, 0);
1850 if (IS_ERR(dst)) {
a0073fe1
SK		 1851 kfree_skb(skb);
1852 continue;
1853 }
1854
1855 nf_reset(skb);
1856 skb_dst_drop(skb);
1857 skb_dst_set(skb, dst);
1858
a0073fe1 1859 err = dst_output(skb);
a0073fe1
SK		 1860 }
1861
e7d8f6cb
SK		 1862out:
1863 xfrm_pol_put(pol);
a0073fe1
SK		 1864 return;
1865
1866purge_queue:
1867 pq->timeout = 0;
1868 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1869 xfrm_pol_put(pol);
a0073fe1
SK		 1870}
1871
aad88724 1872static int xdst_queue_output(struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1873{
1874 unsigned long sched_next;
1875 struct dst_entry *dst = skb_dst(skb);
1876 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1877 struct xfrm_policy *pol = xdst->pols[0];
1878 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4
SK		 1879 const struct sk_buff *fclone = skb + 1;
1880
1881 if (unlikely(skb->fclone == SKB_FCLONE_ORIG &&
1882 fclone->fclone == SKB_FCLONE_CLONE)) {
1883 kfree_skb(skb);
1884 return 0;
1885 }
a0073fe1
SK		 1886
1887 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1888 kfree_skb(skb);
1889 return -EAGAIN;
1890 }
1891
1892 skb_dst_force(skb);
a0073fe1
SK		 1893
1894 spin_lock_bh(&pq->hold_queue.lock);
1895
1896 if (!pq->timeout)
1897 pq->timeout = XFRM_QUEUE_TMO_MIN;
1898
1899 sched_next = jiffies + pq->timeout;
1900
1901 if (del_timer(&pq->hold_timer)) {
1902 if (time_before(pq->hold_timer.expires, sched_next))
1903 sched_next = pq->hold_timer.expires;
e7d8f6cb 1904 xfrm_pol_put(pol);
a0073fe1
SK		 1905 }
1906
1907 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1908 if (!mod_timer(&pq->hold_timer, sched_next))
1909 xfrm_pol_hold(pol);
a0073fe1
SK		 1910
1911 spin_unlock_bh(&pq->hold_queue.lock);
1912
1913 return 0;
1914}
1915
1916static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
1917 struct dst_entry *dst,
1918 const struct flowi *fl,
1919 int num_xfrms,
1920 u16 family)
1921{
1922 int err;
1923 struct net_device *dev;
1924 struct dst_entry *dst1;
1925 struct xfrm_dst *xdst;
1926
1927 xdst = xfrm_alloc_dst(net, family);
1928 if (IS_ERR(xdst))
1929 return xdst;
1930
5b8ef341 1931 if (net->xfrm.sysctl_larval_drop || num_xfrms <= 0)
a0073fe1
SK		 1932 return xdst;
1933
1934 dst1 = &xdst->u.dst;
1935 dst_hold(dst);
1936 xdst->route = dst;
1937
1938 dst_copy_metrics(dst1, dst);
1939
1940 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
1941 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
1942 dst1->lastuse = jiffies;
1943
1944 dst1->input = dst_discard;
1945 dst1->output = xdst_queue_output;
1946
1947 dst_hold(dst);
1948 dst1->child = dst;
1949 dst1->path = dst;
1950
1951 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
1952
1953 err = -ENODEV;
1954 dev = dst->dev;
1955 if (!dev)
1956 goto free_dst;
1957
1958 err = xfrm_fill_dst(xdst, dev, fl);
1959 if (err)
1960 goto free_dst;
1961
1962out:
1963 return xdst;
1964
1965free_dst:
1966 dst_release(dst1);
1967 xdst = ERR_PTR(err);
1968 goto out;
1969}
1970
80c802f3 1971static struct flow_cache_object *
dee9f4bc 1972xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 1973 struct flow_cache_object *oldflo, void *ctx)
1974{
1975 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1976 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1977 struct xfrm_dst *xdst, *new_xdst;
1978 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1979
1980 /* Check if the policies from old bundle are usable */
1981 xdst = NULL;
1982 if (oldflo) {
1983 xdst = container_of(oldflo, struct xfrm_dst, flo);
1984 num_pols = xdst->num_pols;
1985 num_xfrms = xdst->num_xfrms;
1986 pol_dead = 0;
1987 for (i = 0; i < num_pols; i++) {
1988 pols[i] = xdst->pols[i];
1989 pol_dead |= pols[i]->walk.dead;
1990 }
1991 if (pol_dead) {
1992 dst_free(&xdst->u.dst);
1993 xdst = NULL;
1994 num_pols = 0;
1995 num_xfrms = 0;
1996 oldflo = NULL;
1997 }
1998 }
1999
2000 /* Resolve policies to use if we couldn't get them from
2001 * previous cache entry */
2002 if (xdst == NULL) {
2003 num_pols = 1;
b5fb82c4
BZ		 2004 pols[0] = __xfrm_policy_lookup(net, fl, family,
2005 flow_to_policy_dir(dir));
80c802f3
TT		 2006 err = xfrm_expand_policies(fl, family, pols,
2007 &num_pols, &num_xfrms);
2008 if (err < 0)
2009 goto inc_error;
2010 if (num_pols == 0)
2011 return NULL;
2012 if (num_xfrms <= 0)
2013 goto make_dummy_bundle;
2014 }
2015
2016 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
2017 if (IS_ERR(new_xdst)) {
2018 err = PTR_ERR(new_xdst);
2019 if (err != -EAGAIN)
2020 goto error;
2021 if (oldflo == NULL)
2022 goto make_dummy_bundle;
2023 dst_hold(&xdst->u.dst);
2024 return oldflo;
d809ec89
TT		 2025 } else if (new_xdst == NULL) {
2026 num_xfrms = 0;
2027 if (oldflo == NULL)
2028 goto make_dummy_bundle;
2029 xdst->num_xfrms = 0;
2030 dst_hold(&xdst->u.dst);
2031 return oldflo;
80c802f3
TT		 2032 }
2033
2034 /* Kill the previous bundle */
2035 if (xdst) {
2036 /* The policies were stolen for newly generated bundle */
2037 xdst->num_pols = 0;
2038 dst_free(&xdst->u.dst);
2039 }
2040
2041 /* Flow cache does not have reference, it dst_free()'s,
2042 * but we do need to return one reference for original caller */
2043 dst_hold(&new_xdst->u.dst);
2044 return &new_xdst->flo;
2045
2046make_dummy_bundle:
2047 /* We found policies, but there's no bundles to instantiate:
2048 * either because the policy blocks, has no transformations or
2049 * we could not build template (no xfrm_states).*/
a0073fe1 2050 xdst = xfrm_create_dummy_bundle(net, dst_orig, fl, num_xfrms, family);
80c802f3
TT		 2051 if (IS_ERR(xdst)) {
2052 xfrm_pols_put(pols, num_pols);
2053 return ERR_CAST(xdst);
2054 }
2055 xdst->num_pols = num_pols;
2056 xdst->num_xfrms = num_xfrms;
3e94c2dc 2057 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2058
2059 dst_hold(&xdst->u.dst);
2060 return &xdst->flo;
2061
2062inc_error:
2063 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2064error:
2065 if (xdst != NULL)
2066 dst_free(&xdst->u.dst);
2067 else
2068 xfrm_pols_put(pols, num_pols);
2069 return ERR_PTR(err);
2070}
1da177e4 2071
2774c131
DM		 2072static struct dst_entry *make_blackhole(struct net *net, u16 family,
2073 struct dst_entry *dst_orig)
2074{
2075 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2076 struct dst_entry *ret;
2077
2078 if (!afinfo) {
2079 dst_release(dst_orig);
433a1954 2080 return ERR_PTR(-EINVAL);
2774c131
DM		 2081 } else {
2082 ret = afinfo->blackhole_route(net, dst_orig);
2083 }
2084 xfrm_policy_put_afinfo(afinfo);
2085
2086 return ret;
2087}
2088
1da177e4
LT		 2089/* Main function: finds/creates a bundle for given flow.
2090 *
2091 * At the moment we eat a raw IP route. Mostly to speed up lookups
2092 * on interfaces with disabled IPsec.
2093 */
452edd59
DM		 2094struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2095 const struct flowi *fl,
2096 struct sock *sk, int flags)
1da177e4 2097{
4e81bb83 2098 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2099 struct flow_cache_object *flo;
2100 struct xfrm_dst *xdst;
452edd59 2101 struct dst_entry *dst, *route;
80c802f3 2102 u16 family = dst_orig->ops->family;
df71837d 2103 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2104 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2105
80c802f3
TT		 2106 dst = NULL;
2107 xdst = NULL;
2108 route = NULL;
4e81bb83 2109
f7944fb1 2110 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2111 num_pols = 1;
2112 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2113 err = xfrm_expand_policies(fl, family, pols,
2114 &num_pols, &num_xfrms);
2115 if (err < 0)
75b8c133 2116 goto dropdst;
80c802f3
TT		 2117
2118 if (num_pols) {
2119 if (num_xfrms <= 0) {
2120 drop_pols = num_pols;
2121 goto no_transform;
2122 }
2123
2124 xdst = xfrm_resolve_and_create_bundle(
2125 pols, num_pols, fl,
2126 family, dst_orig);
2127 if (IS_ERR(xdst)) {
2128 xfrm_pols_put(pols, num_pols);
2129 err = PTR_ERR(xdst);
2130 goto dropdst;
d809ec89
TT		 2131 } else if (xdst == NULL) {
2132 num_xfrms = 0;
2133 drop_pols = num_pols;
2134 goto no_transform;
80c802f3
TT		 2135 }
2136
b7eea454
SK		 2137 dst_hold(&xdst->u.dst);
2138 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2139 route = xdst->route;
0aa64774 2140 }
3bccfbc7 2141 }
1da177e4 2142
80c802f3 2143 if (xdst == NULL) {
1da177e4 2144 /* To accelerate a bit... */
2518c7c2 2145 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2146 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2147 goto nopol;
1da177e4 2148
80c802f3
TT		 2149 flo = flow_cache_lookup(net, fl, family, dir,
2150 xfrm_bundle_lookup, dst_orig);
2151 if (flo == NULL)
2152 goto nopol;
fe1a5f03 2153 if (IS_ERR(flo)) {
80c802f3 2154 err = PTR_ERR(flo);
75b8c133 2155 goto dropdst;
d66e37a9 2156 }
80c802f3
TT		 2157 xdst = container_of(flo, struct xfrm_dst, flo);
2158
2159 num_pols = xdst->num_pols;
2160 num_xfrms = xdst->num_xfrms;
3e94c2dc 2161 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2162 route = xdst->route;
2163 }
2164
2165 dst = &xdst->u.dst;
2166 if (route == NULL && num_xfrms > 0) {
2167 /* The only case when xfrm_bundle_lookup() returns a
2168 * bundle with null route, is when the template could
2169 * not be resolved. It means policies are there, but
2170 * bundle could not be created, since we don't yet
2171 * have the xfrm_state's. We need to wait for KM to
2172 * negotiate new SA's or bail out with error.*/
2173 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2174 dst_release(dst);
a1aa3483 2175 xfrm_pols_put(pols, drop_pols);
80c802f3 2176 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2774c131 2177
452edd59 2178 return make_blackhole(net, family, dst_orig);
80c802f3 2179 }
80c802f3 2180
5b8ef341 2181 err = -EAGAIN;
80c802f3
TT		 2182
2183 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2184 goto error;
1da177e4
LT		 2185 }
2186
80c802f3
TT		 2187no_transform:
2188 if (num_pols == 0)
8b7817f3 2189 goto nopol;
1da177e4 2190
80c802f3
TT		 2191 if ((flags & XFRM_LOOKUP_ICMP) &&
2192 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2193 err = -ENOENT;
8b7817f3 2194 goto error;
80c802f3 2195 }
8b7817f3 2196
80c802f3
TT		 2197 for (i = 0; i < num_pols; i++)
2198 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2199
80c802f3 2200 if (num_xfrms < 0) {
1da177e4 2201 /* Prohibit the flow */
59c9940e 2202 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2203 err = -EPERM;
2204 goto error;
80c802f3
TT		 2205 } else if (num_xfrms > 0) {
2206 /* Flow transformed */
80c802f3
TT		 2207 dst_release(dst_orig);
2208 } else {
2209 /* Flow passes untransformed */
2210 dst_release(dst);
452edd59 2211 dst = dst_orig;
1da177e4 2212 }
80c802f3
TT		 2213ok:
2214 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2215 if (dst && dst->xfrm &&
2216 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2217 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2218 return dst;
1da177e4 2219
80c802f3 2220nopol:
452edd59
DM		 2221 if (!(flags & XFRM_LOOKUP_ICMP)) {
2222 dst = dst_orig;
80c802f3 2223 goto ok;
452edd59 2224 }
80c802f3 2225 err = -ENOENT;
1da177e4 2226error:
80c802f3 2227 dst_release(dst);
75b8c133
HX		 2228dropdst:
2229 dst_release(dst_orig);
80c802f3 2230 xfrm_pols_put(pols, drop_pols);
452edd59 2231 return ERR_PTR(err);
1da177e4
LT		 2232}
2233EXPORT_SYMBOL(xfrm_lookup);
2234
df0ba92a 2235static inline int
8f029de2 2236xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2237{
2238 struct xfrm_state *x;
df0ba92a
MN		 2239
2240 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2241 return 0;
2242 x = skb->sp->xvec[idx];
2243 if (!x->type->reject)
2244 return 0;
1ecafede 2245 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2246}
2247
1da177e4
LT		 2248/* When skb is transformed back to its "native" form, we have to
2249 * check policy restrictions. At the moment we make this in maximally
2250 * stupid way. Shame on me. :-) Of course, connected sockets must
2251 * have policy cached at them.
2252 */
2253
2254static inline int
7db454b9 2255xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2256 unsigned short family)
2257{
2258 if (xfrm_state_kern(x))
928ba416 2259 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2260 return x->id.proto == tmpl->id.proto &&
2261 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2262 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2263 x->props.mode == tmpl->mode &&
c5d18e98 2264 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2265 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2266 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2267 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2268}
2269
df0ba92a
MN		 2270/*
2271 * 0 or more than 0 is returned when validation is succeeded (either bypass
2272 * because of optional transport mode, or next index of the mathced secpath
2273 * state with the template.
2274 * -1 is returned when no matching template is found.
2275 * Otherwise "-2 - errored_index" is returned.
2276 */
1da177e4 2277static inline int
22cccb7e 2278xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2279 unsigned short family)
2280{
2281 int idx = start;
2282
2283 if (tmpl->optional) {
7e49e6de 2284 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2285 return start;
2286 } else
2287 start = -1;
2288 for (; idx < sp->len; idx++) {
dbe5b4aa 2289 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2290 return ++idx;
df0ba92a
MN		 2291 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2292 if (start == -1)
2293 start = -2-idx;
1da177e4 2294 break;
df0ba92a 2295 }
1da177e4
LT		 2296 }
2297 return start;
2298}
2299
d5422efe
HX		 2300int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2301 unsigned int family, int reverse)
1da177e4
LT		 2302{
2303 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2304 int err;
1da177e4
LT		 2305
2306 if (unlikely(afinfo == NULL))
2307 return -EAFNOSUPPORT;
2308
d5422efe 2309 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2310 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2311 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2312 return err;
1da177e4 2313}
d5422efe 2314EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2315
9a7386ec 2316static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2317{
2318 for (; k < sp->len; k++) {
df0ba92a 2319 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2320 *idxp = k;
1da177e4 2321 return 1;
df0ba92a 2322 }
1da177e4
LT		 2323 }
2324
2325 return 0;
2326}
2327
a716c119 2328int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2329 unsigned short family)
2330{
f6e1e25d 2331 struct net *net = dev_net(skb->dev);
1da177e4 2332 struct xfrm_policy *pol;
4e81bb83
MN		 2333 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2334 int npols = 0;
2335 int xfrm_nr;
2336 int pi;
d5422efe 2337 int reverse;
1da177e4 2338 struct flowi fl;
d5422efe 2339 u8 fl_dir;
df0ba92a 2340 int xerr_idx = -1;
1da177e4 2341
d5422efe
HX		 2342 reverse = dir & ~XFRM_POLICY_MASK;
2343 dir &= XFRM_POLICY_MASK;
2344 fl_dir = policy_to_flow_dir(dir);
2345
0aa64774 2346 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2347 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2348 return 0;
0aa64774
MN		 2349 }
2350
eb9c7ebe 2351 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2352
2353 /* First, check used SA against their selectors. */
2354 if (skb->sp) {
2355 int i;
2356
9b7a787d 2357 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2358 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2359 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2360 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2361 return 0;
0aa64774 2362 }
1da177e4
LT		 2363 }
2364 }
2365
2366 pol = NULL;
3bccfbc7 2367 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2368 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2369 if (IS_ERR(pol)) {
59c9940e 2370 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2371 return 0;
0aa64774 2372 }
3bccfbc7 2373 }
1da177e4 2374
fe1a5f03
TT		 2375 if (!pol) {
2376 struct flow_cache_object *flo;
2377
2378 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2379 xfrm_policy_lookup, NULL);
2380 if (IS_ERR_OR_NULL(flo))
2381 pol = ERR_CAST(flo);
2382 else
2383 pol = container_of(flo, struct xfrm_policy, flo);
2384 }
1da177e4 2385
0aa64774 2386 if (IS_ERR(pol)) {
59c9940e 2387 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2388 return 0;
0aa64774 2389 }
134b0fc5 2390
df0ba92a 2391 if (!pol) {
d1d9facf 2392 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2393 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2394 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2395 return 0;
2396 }
2397 return 1;
2398 }
1da177e4 2399
9d729f72 2400 pol->curlft.use_time = get_seconds();
1da177e4 2401
4e81bb83 2402 pols[0] = pol;
02d0892f 2403 npols++;
4e81bb83
MN		 2404#ifdef CONFIG_XFRM_SUB_POLICY
2405 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2406 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2407 &fl, family,
2408 XFRM_POLICY_IN);
2409 if (pols[1]) {
0aa64774 2410 if (IS_ERR(pols[1])) {
59c9940e 2411 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2412 return 0;
0aa64774 2413 }
9d729f72 2414 pols[1]->curlft.use_time = get_seconds();
02d0892f 2415 npols++;
4e81bb83
MN		 2416 }
2417 }
2418#endif
2419
1da177e4
LT		 2420 if (pol->action == XFRM_POLICY_ALLOW) {
2421 struct sec_path *sp;
2422 static struct sec_path dummy;
4e81bb83 2423 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2424 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2425 struct xfrm_tmpl **tpp = tp;
2426 int ti = 0;
1da177e4
LT		 2427 int i, k;
2428
2429 if ((sp = skb->sp) == NULL)
2430 sp = &dummy;
2431
4e81bb83
MN		 2432 for (pi = 0; pi < npols; pi++) {
2433 if (pols[pi] != pol &&
0aa64774 2434 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2435 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2436 goto reject;
0aa64774
MN		 2437 }
2438 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2439 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2440 goto reject_error;
0aa64774 2441 }
4e81bb83
MN		 2442 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2443 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2444 }
2445 xfrm_nr = ti;
41a49cc3 2446 if (npols > 1) {
283bc9f3 2447 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2448 tpp = stp;
2449 }
4e81bb83 2450
1da177e4
LT		 2451 /* For each tunnel xfrm, find the first matching tmpl.
2452 * For each tmpl before that, find corresponding xfrm.
2453 * Order is _important_. Later we will implement
2454 * some barriers, but at the moment barriers
2455 * are implied between each two transformations.
2456 */
4e81bb83
MN		 2457 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2458 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2459 if (k < 0) {
d1d9facf
JM		 2460 if (k < -1)
2461 /* "-2 - errored_index" returned */
2462 xerr_idx = -(2+k);
59c9940e 2463 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2464 goto reject;
df0ba92a 2465 }
1da177e4
LT		 2466 }
2467
0aa64774 2468 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2469 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2470 goto reject;
0aa64774 2471 }
1da177e4 2472
4e81bb83 2473 xfrm_pols_put(pols, npols);
1da177e4
LT		 2474 return 1;
2475 }
59c9940e 2476 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2477
2478reject:
df0ba92a 2479 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2480reject_error:
2481 xfrm_pols_put(pols, npols);
1da177e4
LT		 2482 return 0;
2483}
2484EXPORT_SYMBOL(__xfrm_policy_check);
2485
2486int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2487{
99a66657 2488 struct net *net = dev_net(skb->dev);
1da177e4 2489 struct flowi fl;
adf30907 2490 struct dst_entry *dst;
73137147 2491 int res = 1;
1da177e4 2492
0aa64774 2493 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2494 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2495 return 0;
0aa64774 2496 }
1da177e4 2497
fafeeb6c 2498 skb_dst_force(skb);
adf30907 2499
452edd59
DM		 2500 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2501 if (IS_ERR(dst)) {
73137147 2502 res = 0;
452edd59
DM		 2503 dst = NULL;
2504 }
adf30907
ED		 2505 skb_dst_set(skb, dst);
2506 return res;
1da177e4
LT		 2507}
2508EXPORT_SYMBOL(__xfrm_route_forward);
2509
d49c73c7
DM		 2510/* Optimize later using cookies and generation ids. */
2511
1da177e4
LT		 2512static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2513{
d49c73c7 2514 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2515 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2516 * get validated by dst_ops->check on every use. We do this
2517 * because when a normal route referenced by an XFRM dst is
2518 * obsoleted we do not go looking around for all parent
2519 * referencing XFRM dsts so that we can invalidate them. It
2520 * is just too much work. Instead we make the checks here on
2521 * every use. For example:
d49c73c7
DM		 2522 *
2523 * XFRM dst A --> IPv4 dst X
2524 *
2525 * X is the "xdst->route" of A (X is also the "dst->path" of A
2526 * in this example). If X is marked obsolete, "A" will not
2527 * notice. That's what we are validating here via the
2528 * stale_bundle() check.
2529 *
2530 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2531 * dst which causes it's ->obsolete field to be set to
2532 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2533 * this, we want to force a new route lookup.
399c180a 2534 */
d49c73c7
DM		 2535 if (dst->obsolete < 0 && !stale_bundle(dst))
2536 return dst;
2537
1da177e4
LT		 2538 return NULL;
2539}
2540
2541static int stale_bundle(struct dst_entry *dst)
2542{
12fdb4d3 2543 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2544}
2545
aabc9761 2546void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2547{
1da177e4 2548 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2549 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2550 dev_hold(dst->dev);
1da177e4
LT		 2551 dev_put(dev);
2552 }
2553}
aabc9761 2554EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2555
2556static void xfrm_link_failure(struct sk_buff *skb)
2557{
2558 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2559}
2560
2561static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2562{
2563 if (dst) {
2564 if (dst->obsolete) {
2565 dst_release(dst);
2566 dst = NULL;
2567 }
2568 }
2569 return dst;
2570}
2571
e4c17216 2572void xfrm_garbage_collect(struct net *net)
c0ed1c14 2573{
ca925cf1 2574 flow_cache_flush(net);
c0ed1c14 2575}
e4c17216 2576EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2577
2578static void xfrm_garbage_collect_deferred(struct net *net)
2579{
ca925cf1 2580 flow_cache_flush_deferred(net);
c0ed1c14
SK		 2581}
2582
25ee3286 2583static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2584{
2585 do {
2586 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2587 u32 pmtu, route_mtu_cached;
2588
2589 pmtu = dst_mtu(dst->child);
2590 xdst->child_mtu_cached = pmtu;
2591
2592 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2593
2594 route_mtu_cached = dst_mtu(xdst->route);
2595 xdst->route_mtu_cached = route_mtu_cached;
2596
2597 if (pmtu > route_mtu_cached)
2598 pmtu = route_mtu_cached;
2599
defb3519 2600 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2601 } while ((dst = dst->next));
2602}
2603
1da177e4
LT		 2604/* Check that the bundle accepts the flow and its components are
2605 * still valid.
2606 */
2607
12fdb4d3 2608static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2609{
2610 struct dst_entry *dst = &first->u.dst;
2611 struct xfrm_dst *last;
2612 u32 mtu;
2613
92d63dec 2614 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2615 (dst->dev && !netif_running(dst->dev)))
2616 return 0;
2617
a0073fe1
SK		 2618 if (dst->flags & DST_XFRM_QUEUE)
2619 return 1;
2620
1da177e4
LT		 2621 last = NULL;
2622
2623 do {
2624 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2625
1da177e4
LT		 2626 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2627 return 0;
80c802f3
TT		 2628 if (xdst->xfrm_genid != dst->xfrm->genid)
2629 return 0;
b1312c89
TT		 2630 if (xdst->num_pols > 0 &&
2631 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2632 return 0;
e53820de 2633
1da177e4
LT		 2634 mtu = dst_mtu(dst->child);
2635 if (xdst->child_mtu_cached != mtu) {
2636 last = xdst;
2637 xdst->child_mtu_cached = mtu;
2638 }
2639
92d63dec 2640 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2641 return 0;
2642 mtu = dst_mtu(xdst->route);
2643 if (xdst->route_mtu_cached != mtu) {
2644 last = xdst;
2645 xdst->route_mtu_cached = mtu;
2646 }
2647
2648 dst = dst->child;
2649 } while (dst->xfrm);
2650
2651 if (likely(!last))
2652 return 1;
2653
2654 mtu = last->child_mtu_cached;
2655 for (;;) {
2656 dst = &last->u.dst;
2657
2658 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2659 if (mtu > last->route_mtu_cached)
2660 mtu = last->route_mtu_cached;
defb3519 2661 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2662
2663 if (last == first)
2664 break;
2665
bd0bf076 2666 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2667 last->child_mtu_cached = mtu;
2668 }
2669
2670 return 1;
2671}
2672
0dbaee3b
DM		 2673static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2674{
2675 return dst_metric_advmss(dst->path);
2676}
2677
ebb762f2 2678static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2679{
618f9bc7
SK		 2680 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2681
2682 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2683}
2684
f894cbf8
DM		 2685static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2686 struct sk_buff *skb,
2687 const void *daddr)
d3aaeb38 2688{
f894cbf8 2689 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2690}
2691
1da177e4
LT		 2692int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2693{
d7c7544c 2694 struct net *net;
1da177e4
LT		 2695 int err = 0;
2696 if (unlikely(afinfo == NULL))
2697 return -EINVAL;
2698 if (unlikely(afinfo->family >= NPROTO))
2699 return -EAFNOSUPPORT;
ef8531b6 2700 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2701 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2702 err = -ENOBUFS;
2703 else {
2704 struct dst_ops *dst_ops = afinfo->dst_ops;
2705 if (likely(dst_ops->kmem_cachep == NULL))
2706 dst_ops->kmem_cachep = xfrm_dst_cache;
2707 if (likely(dst_ops->check == NULL))
2708 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2709 if (likely(dst_ops->default_advmss == NULL))
2710 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2711 if (likely(dst_ops->mtu == NULL))
2712 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2713 if (likely(dst_ops->negative_advice == NULL))
2714 dst_ops->negative_advice = xfrm_negative_advice;
2715 if (likely(dst_ops->link_failure == NULL))
2716 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2717 if (likely(dst_ops->neigh_lookup == NULL))
2718 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2719 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2720 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2721 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2722 }
ef8531b6 2723 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2724
2725 rtnl_lock();
2726 for_each_net(net) {
2727 struct dst_ops *xfrm_dst_ops;
2728
2729 switch (afinfo->family) {
2730 case AF_INET:
2731 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2732 break;
dfd56b8b 2733#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2734 case AF_INET6:
2735 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2736 break;
2737#endif
2738 default:
2739 BUG();
2740 }
2741 *xfrm_dst_ops = *afinfo->dst_ops;
2742 }
2743 rtnl_unlock();
2744
1da177e4
LT		 2745 return err;
2746}
2747EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2748
2749int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2750{
2751 int err = 0;
2752 if (unlikely(afinfo == NULL))
2753 return -EINVAL;
2754 if (unlikely(afinfo->family >= NPROTO))
2755 return -EAFNOSUPPORT;
ef8531b6 2756 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2757 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2758 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2759 err = -EINVAL;
ef8531b6
ED		 2760 else
2761 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2762 NULL);
2763 }
2764 spin_unlock(&xfrm_policy_afinfo_lock);
2765 if (!err) {
2766 struct dst_ops *dst_ops = afinfo->dst_ops;
2767
2768 synchronize_rcu();
2769
2770 dst_ops->kmem_cachep = NULL;
2771 dst_ops->check = NULL;
2772 dst_ops->negative_advice = NULL;
2773 dst_ops->link_failure = NULL;
2774 afinfo->garbage_collect = NULL;
1da177e4 2775 }
1da177e4
LT		 2776 return err;
2777}
2778EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2779
d7c7544c
AD		 2780static void __net_init xfrm_dst_ops_init(struct net *net)
2781{
2782 struct xfrm_policy_afinfo *afinfo;
2783
ef8531b6
ED		 2784 rcu_read_lock();
2785 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2786 if (afinfo)
2787 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2788#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2789 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2790 if (afinfo)
2791 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2792#endif
418a99ac 2793 rcu_read_unlock();
546be240
HX		 2794}
2795
1da177e4
LT		 2796static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2797{
351638e7 2798 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2799
1da177e4
LT		 2800 switch (event) {
2801 case NETDEV_DOWN:
c0ed1c14 2802 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2803 }
2804 return NOTIFY_DONE;
2805}
2806
2807static struct notifier_block xfrm_dev_notifier = {
d5917a35 2808 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2809};
2810
558f82ef 2811#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2812static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2813{
c68cd1a0 2814 int rv;
698365fa
WC		 2815 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2816 if (!net->mib.xfrm_statistics)
558f82ef 2817 return -ENOMEM;
c68cd1a0
AD		 2818 rv = xfrm_proc_init(net);
2819 if (rv < 0)
698365fa 2820 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2821 return rv;
558f82ef 2822}
59c9940e
AD		 2823
2824static void xfrm_statistics_fini(struct net *net)
2825{
c68cd1a0 2826 xfrm_proc_fini(net);
698365fa 2827 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2828}
2829#else
2830static int __net_init xfrm_statistics_init(struct net *net)
2831{
2832 return 0;
2833}
2834
2835static void xfrm_statistics_fini(struct net *net)
2836{
2837}
558f82ef
MN		 2838#endif
2839
d62ddc21 2840static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2841{
2518c7c2
DM		 2842 unsigned int hmask, sz;
2843 int dir;
2844
d62ddc21
AD		 2845 if (net_eq(net, &init_net))
2846 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2847 sizeof(struct xfrm_dst),
e5d679f3 2848 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2849 NULL);
1da177e4 2850
2518c7c2
DM		 2851 hmask = 8 - 1;
2852 sz = (hmask+1) * sizeof(struct hlist_head);
2853
93b851c1
AD		 2854 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2855 if (!net->xfrm.policy_byidx)
2856 goto out_byidx;
8100bea7 2857 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2858
2859 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2860 struct xfrm_policy_hash *htab;
2861
dc2caba7 2862 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2863 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2864
a35f6c5d 2865 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2866 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2867 if (!htab->table)
a35f6c5d
AD		 2868 goto out_bydst;
2869 htab->hmask = hmask;
b58555f1
CG		 2870 htab->dbits4 = 32;
2871 htab->sbits4 = 32;
2872 htab->dbits6 = 128;
2873 htab->sbits6 = 128;
2518c7c2
DM		 2874 }
2875
adfcf0b2 2876 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2877 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2878 if (net_eq(net, &init_net))
2879 register_netdevice_notifier(&xfrm_dev_notifier);
2880 return 0;
93b851c1 2881
a35f6c5d
AD		 2882out_bydst:
2883 for (dir--; dir >= 0; dir--) {
2884 struct xfrm_policy_hash *htab;
2885
2886 htab = &net->xfrm.policy_bydst[dir];
2887 xfrm_hash_free(htab->table, sz);
2888 }
2889 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2890out_byidx:
2891 return -ENOMEM;
d62ddc21
AD		 2892}
2893
2894static void xfrm_policy_fini(struct net *net)
2895{
93b851c1 2896 unsigned int sz;
8b18f8ea 2897 int dir;
93b851c1 2898
7c2776ee
AD		 2899 flush_work(&net->xfrm.policy_hash_work);
2900#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 2901 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 2902#endif
2e71029e 2903 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 2904
adfcf0b2 2905 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2906
8b18f8ea 2907 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2908 struct xfrm_policy_hash *htab;
2909
8b18f8ea 2910 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2911
2912 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 2913 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 2914 WARN_ON(!hlist_empty(htab->table));
2915 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2916 }
2917
8100bea7 2918 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2919 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2920 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2921}
2922
d62ddc21
AD		 2923static int __net_init xfrm_net_init(struct net *net)
2924{
2925 int rv;
2926
59c9940e
AD		 2927 rv = xfrm_statistics_init(net);
2928 if (rv < 0)
2929 goto out_statistics;
d62ddc21
AD		 2930 rv = xfrm_state_init(net);
2931 if (rv < 0)
2932 goto out_state;
2933 rv = xfrm_policy_init(net);
2934 if (rv < 0)
2935 goto out_policy;
d7c7544c 2936 xfrm_dst_ops_init(net);
b27aeadb
AD		 2937 rv = xfrm_sysctl_init(net);
2938 if (rv < 0)
2939 goto out_sysctl;
4a93f509
SK		 2940 rv = flow_cache_init(net);
2941 if (rv < 0)
2942 goto out;
283bc9f3
FD		 2943
2944 /* Initialize the per-net locks here */
2945 spin_lock_init(&net->xfrm.xfrm_state_lock);
2946 rwlock_init(&net->xfrm.xfrm_policy_lock);
283bc9f3
FD		 2947 mutex_init(&net->xfrm.xfrm_cfg_mutex);
2948
d62ddc21
AD		 2949 return 0;
2950
4a93f509
SK		 2951out:
2952 xfrm_sysctl_fini(net);
b27aeadb
AD		 2953out_sysctl:
2954 xfrm_policy_fini(net);
d62ddc21
AD		 2955out_policy:
2956 xfrm_state_fini(net);
2957out_state:
59c9940e
AD		 2958 xfrm_statistics_fini(net);
2959out_statistics:
d62ddc21
AD		 2960 return rv;
2961}
2962
2963static void __net_exit xfrm_net_exit(struct net *net)
2964{
4a93f509 2965 flow_cache_fini(net);
b27aeadb 2966 xfrm_sysctl_fini(net);
d62ddc21
AD		 2967 xfrm_policy_fini(net);
2968 xfrm_state_fini(net);
59c9940e 2969 xfrm_statistics_fini(net);
d62ddc21
AD		 2970}
2971
2972static struct pernet_operations __net_initdata xfrm_net_ops = {
2973 .init = xfrm_net_init,
2974 .exit = xfrm_net_exit,
2975};
2976
1da177e4
LT		 2977void __init xfrm_init(void)
2978{
d62ddc21 2979 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 2980 xfrm_input_init();
2981}
2982
ab5f5e8b 2983#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2984static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2985 struct audit_buffer *audit_buf)
ab5f5e8b 2986{
875179fa
PM		 2987 struct xfrm_sec_ctx *ctx = xp->security;
2988 struct xfrm_selector *sel = &xp->selector;
2989
2990 if (ctx)
ab5f5e8b 2991 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2992 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2993
9b7a787d 2994 switch (sel->family) {
ab5f5e8b 2995 case AF_INET:
21454aaa 2996 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2997 if (sel->prefixlen_s != 32)
2998 audit_log_format(audit_buf, " src_prefixlen=%d",
2999 sel->prefixlen_s);
21454aaa 3000 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3001 if (sel->prefixlen_d != 32)
3002 audit_log_format(audit_buf, " dst_prefixlen=%d",
3003 sel->prefixlen_d);
ab5f5e8b
JL		 3004 break;
3005 case AF_INET6:
5b095d98 3006 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3007 if (sel->prefixlen_s != 128)
3008 audit_log_format(audit_buf, " src_prefixlen=%d",
3009 sel->prefixlen_s);
5b095d98 3010 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3011 if (sel->prefixlen_d != 128)
3012 audit_log_format(audit_buf, " dst_prefixlen=%d",
3013 sel->prefixlen_d);
ab5f5e8b
JL		 3014 break;
3015 }
3016}
3017
2e71029e 3018void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3019{
3020 struct audit_buffer *audit_buf;
ab5f5e8b 3021
afeb14b4 3022 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3023 if (audit_buf == NULL)
3024 return;
2e71029e 3025 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3026 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3027 xfrm_audit_common_policyinfo(xp, audit_buf);
3028 audit_log_end(audit_buf);
3029}
3030EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3031
68277acc 3032void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3033 bool task_valid)
ab5f5e8b
JL		 3034{
3035 struct audit_buffer *audit_buf;
ab5f5e8b 3036
afeb14b4 3037 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3038 if (audit_buf == NULL)
3039 return;
2e71029e 3040 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3041 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3042 xfrm_audit_common_policyinfo(xp, audit_buf);
3043 audit_log_end(audit_buf);
3044}
3045EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3046#endif
3047
80c9abaa 3048#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3049static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3050 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3051{
3052 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3053 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3054 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3055 sel_cmp->family) &&
3056 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3057 sel_cmp->family) &&
80c9abaa
SS		 3058 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3059 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3060 return true;
80c9abaa
SS		 3061 }
3062 } else {
3063 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3064 return true;
80c9abaa
SS		 3065 }
3066 }
bc9b35ad 3067 return false;
80c9abaa
SS		 3068}
3069
3e94c2dc
WC		 3070static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3071 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3072{
3073 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3074 struct hlist_head *chain;
3075 u32 priority = ~0U;
3076
283bc9f3 3077 read_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME*/
8d549c4f 3078 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, 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 ret = pol;
3083 priority = ret->priority;
3084 break;
3085 }
3086 }
8d549c4f 3087 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3088 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3089 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3090 pol->type == type &&
3091 pol->priority < priority) {
3092 ret = pol;
3093 break;
3094 }
3095 }
3096
3097 if (ret)
3098 xfrm_pol_hold(ret);
3099
283bc9f3 3100 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3101
3102 return ret;
3103}
3104
dd701754 3105static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3106{
3107 int match = 0;
3108
3109 if (t->mode == m->mode && t->id.proto == m->proto &&
3110 (m->reqid == 0 || t->reqid == m->reqid)) {
3111 switch (t->mode) {
3112 case XFRM_MODE_TUNNEL:
3113 case XFRM_MODE_BEET:
70e94e66
YH		 3114 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3115 m->old_family) &&
3116 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3117 m->old_family)) {
80c9abaa
SS		 3118 match = 1;
3119 }
3120 break;
3121 case XFRM_MODE_TRANSPORT:
3122 /* in case of transport mode, template does not store
3123 any IP addresses, hence we just compare mode and
3124 protocol */
3125 match = 1;
3126 break;
3127 default:
3128 break;
3129 }
3130 }
3131 return match;
3132}
3133
3134/* update endpoint address(es) of template(s) */
3135static int xfrm_policy_migrate(struct xfrm_policy *pol,
3136 struct xfrm_migrate *m, int num_migrate)
3137{
3138 struct xfrm_migrate *mp;
80c9abaa
SS		 3139 int i, j, n = 0;
3140
3141 write_lock_bh(&pol->lock);
12a169e7 3142 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3143 /* target policy has been deleted */
3144 write_unlock_bh(&pol->lock);
3145 return -ENOENT;
3146 }
3147
3148 for (i = 0; i < pol->xfrm_nr; i++) {
3149 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3150 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3151 continue;
3152 n++;
1bfcb10f
HX		 3153 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3154 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3155 continue;
3156 /* update endpoints */
3157 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3158 sizeof(pol->xfrm_vec[i].id.daddr));
3159 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3160 sizeof(pol->xfrm_vec[i].saddr));
3161 pol->xfrm_vec[i].encap_family = mp->new_family;
3162 /* flush bundles */
80c802f3 3163 atomic_inc(&pol->genid);
80c9abaa
SS		 3164 }
3165 }
3166
3167 write_unlock_bh(&pol->lock);
3168
3169 if (!n)
3170 return -ENODATA;
3171
3172 return 0;
3173}
3174
dd701754 3175static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3176{
3177 int i, j;
3178
3179 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3180 return -EINVAL;
3181
3182 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3183 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3184 m[i].old_family) &&
3185 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3186 m[i].old_family))
80c9abaa
SS		 3187 return -EINVAL;
3188 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3189 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3190 return -EINVAL;
3191
3192 /* check if there is any duplicated entry */
3193 for (j = i + 1; j < num_migrate; j++) {
3194 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3195 sizeof(m[i].old_daddr)) &&
3196 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3197 sizeof(m[i].old_saddr)) &&
3198 m[i].proto == m[j].proto &&
3199 m[i].mode == m[j].mode &&
3200 m[i].reqid == m[j].reqid &&
3201 m[i].old_family == m[j].old_family)
3202 return -EINVAL;
3203 }
3204 }
3205
3206 return 0;
3207}
3208
b4b7c0b3 3209int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3210 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3211 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3212{
3213 int i, err, nx_cur = 0, nx_new = 0;
3214 struct xfrm_policy *pol = NULL;
3215 struct xfrm_state *x, *xc;
3216 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3217 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3218 struct xfrm_migrate *mp;
3219
3220 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3221 goto out;
3222
3223 /* Stage 1 - find policy */
8d549c4f 3224 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3225 err = -ENOENT;
3226 goto out;
3227 }
3228
3229 /* Stage 2 - find and update state(s) */
3230 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3231 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3232 x_cur[nx_cur] = x;
3233 nx_cur++;
3234 if ((xc = xfrm_state_migrate(x, mp))) {
3235 x_new[nx_new] = xc;
3236 nx_new++;
3237 } else {
3238 err = -ENODATA;
3239 goto restore_state;
3240 }
3241 }
3242 }
3243
3244 /* Stage 3 - update policy */
3245 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3246 goto restore_state;
3247
3248 /* Stage 4 - delete old state(s) */
3249 if (nx_cur) {
3250 xfrm_states_put(x_cur, nx_cur);
3251 xfrm_states_delete(x_cur, nx_cur);
3252 }
3253
3254 /* Stage 5 - announce */
13c1d189 3255 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3256
3257 xfrm_pol_put(pol);
3258
3259 return 0;
3260out:
3261 return err;
3262
3263restore_state:
3264 if (pol)
3265 xfrm_pol_put(pol);
3266 if (nx_cur)
3267 xfrm_states_put(x_cur, nx_cur);
3268 if (nx_new)
3269 xfrm_states_delete(x_new, nx_new);
3270
3271 return err;
3272}
e610e679 3273EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3274#endif
Linux kernel - Deliverables
This page took 1.111188 seconds and 5 git commands to generate.