projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
gre: Use inner mac length when computing tunnel length
[linux-2.6-block.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT		 2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
68277acc 27#include <linux/audit.h>
25ee3286 28#include <net/dst.h>
6ce74ec7 29#include <net/flow.h>
1da177e4
LT		 30#include <net/xfrm.h>
31#include <net/ip.h>
558f82ef
MN		 32#ifdef CONFIG_XFRM_STATISTICS
33#include <net/snmp.h>
34#endif
1da177e4 35
44e36b42
DM		 36#include "xfrm_hash.h"
37
a0073fe1
SK		 38#define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
39#define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
40#define XFRM_MAX_QUEUE_LEN 100
41
b8c203b2
SK		 42struct xfrm_flo {
43 struct dst_entry *dst_orig;
44 u8 flags;
45};
46
418a99ac
PJ		 47static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock);
48static struct xfrm_policy_afinfo __rcu *xfrm_policy_afinfo[NPROTO]
49 __read_mostly;
1da177e4 50
e18b890b 51static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 52
25ee3286 53static void xfrm_init_pmtu(struct dst_entry *dst);
80c802f3 54static int stale_bundle(struct dst_entry *dst);
12fdb4d3 55static int xfrm_bundle_ok(struct xfrm_dst *xdst);
a0073fe1 56static void xfrm_policy_queue_process(unsigned long arg);
1da177e4 57
29fa0b30
WY		 58static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
59 int dir);
60
bc9b35ad 61static inline bool
200ce96e 62__xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 63{
7e1dc7b6
DM		 64 const struct flowi4 *fl4 = &fl->u.ip4;
65
26bff940
AD		 66 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
67 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) &&
7e1dc7b6
DM		 68 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
69 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
70 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
71 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 72}
73
bc9b35ad 74static inline bool
200ce96e 75__xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
77681021 76{
7e1dc7b6
DM		 77 const struct flowi6 *fl6 = &fl->u.ip6;
78
79 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
80 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
81 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
82 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
83 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
84 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
77681021
AM		 85}
86
bc9b35ad
DM		 87bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
88 unsigned short family)
77681021
AM		 89{
90 switch (family) {
91 case AF_INET:
92 return __xfrm4_selector_match(sel, fl);
93 case AF_INET6:
94 return __xfrm6_selector_match(sel, fl);
95 }
bc9b35ad 96 return false;
77681021
AM		 97}
98
ef8531b6
ED		 99static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
100{
101 struct xfrm_policy_afinfo *afinfo;
102
103 if (unlikely(family >= NPROTO))
104 return NULL;
105 rcu_read_lock();
106 afinfo = rcu_dereference(xfrm_policy_afinfo[family]);
107 if (unlikely(!afinfo))
108 rcu_read_unlock();
109 return afinfo;
110}
111
112static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
113{
114 rcu_read_unlock();
115}
116
c5b3cf46 117static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
6418c4e0
DM		 118 const xfrm_address_t *saddr,
119 const xfrm_address_t *daddr,
9bb182a7
YH		 120 int family)
121{
122 struct xfrm_policy_afinfo *afinfo;
123 struct dst_entry *dst;
124
125 afinfo = xfrm_policy_get_afinfo(family);
126 if (unlikely(afinfo == NULL))
127 return ERR_PTR(-EAFNOSUPPORT);
128
c5b3cf46 129 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 130
131 xfrm_policy_put_afinfo(afinfo);
132
133 return dst;
134}
135
25ee3286 136static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 137 xfrm_address_t *prev_saddr,
138 xfrm_address_t *prev_daddr,
25ee3286 139 int family)
1da177e4 140{
c5b3cf46 141 struct net *net = xs_net(x);
66cdb3ca
HX		 142 xfrm_address_t *saddr = &x->props.saddr;
143 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 144 struct dst_entry *dst;
145
9bb182a7 146 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 147 saddr = x->coaddr;
9bb182a7
YH		 148 daddr = prev_daddr;
149 }
150 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
151 saddr = prev_saddr;
66cdb3ca 152 daddr = x->coaddr;
9bb182a7 153 }
1da177e4 154
c5b3cf46 155 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 156
157 if (!IS_ERR(dst)) {
158 if (prev_saddr != saddr)
159 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
160 if (prev_daddr != daddr)
161 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
162 }
1da177e4 163
66cdb3ca 164 return dst;
1da177e4 165}
1da177e4 166
1da177e4
LT		 167static inline unsigned long make_jiffies(long secs)
168{
169 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
170 return MAX_SCHEDULE_TIMEOUT-1;
171 else
a716c119 172 return secs*HZ;
1da177e4
LT		 173}
174
175static void xfrm_policy_timer(unsigned long data)
176{
3e94c2dc 177 struct xfrm_policy *xp = (struct xfrm_policy *)data;
9d729f72 178 unsigned long now = get_seconds();
1da177e4
LT		 179 long next = LONG_MAX;
180 int warn = 0;
181 int dir;
182
183 read_lock(&xp->lock);
184
ea2dea9d 185 if (unlikely(xp->walk.dead))
1da177e4
LT		 186 goto out;
187
77d8d7a6 188 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 189
190 if (xp->lft.hard_add_expires_seconds) {
191 long tmo = xp->lft.hard_add_expires_seconds +
192 xp->curlft.add_time - now;
193 if (tmo <= 0)
194 goto expired;
195 if (tmo < next)
196 next = tmo;
197 }
198 if (xp->lft.hard_use_expires_seconds) {
199 long tmo = xp->lft.hard_use_expires_seconds +
200 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
201 if (tmo <= 0)
202 goto expired;
203 if (tmo < next)
204 next = tmo;
205 }
206 if (xp->lft.soft_add_expires_seconds) {
207 long tmo = xp->lft.soft_add_expires_seconds +
208 xp->curlft.add_time - now;
209 if (tmo <= 0) {
210 warn = 1;
211 tmo = XFRM_KM_TIMEOUT;
212 }
213 if (tmo < next)
214 next = tmo;
215 }
216 if (xp->lft.soft_use_expires_seconds) {
217 long tmo = xp->lft.soft_use_expires_seconds +
218 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
219 if (tmo <= 0) {
220 warn = 1;
221 tmo = XFRM_KM_TIMEOUT;
222 }
223 if (tmo < next)
224 next = tmo;
225 }
226
227 if (warn)
6c5c8ca7 228 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 229 if (next != LONG_MAX &&
230 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
231 xfrm_pol_hold(xp);
232
233out:
234 read_unlock(&xp->lock);
235 xfrm_pol_put(xp);
236 return;
237
238expired:
239 read_unlock(&xp->lock);
4666faab 240 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 241 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 242 xfrm_pol_put(xp);
243}
244
fe1a5f03
TT		 245static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
246{
247 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
248
249 if (unlikely(pol->walk.dead))
250 flo = NULL;
251 else
252 xfrm_pol_hold(pol);
253
254 return flo;
255}
256
257static int xfrm_policy_flo_check(struct flow_cache_object *flo)
258{
259 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
260
261 return !pol->walk.dead;
262}
263
264static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
265{
266 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
267}
268
269static const struct flow_cache_ops xfrm_policy_fc_ops = {
270 .get = xfrm_policy_flo_get,
271 .check = xfrm_policy_flo_check,
272 .delete = xfrm_policy_flo_delete,
273};
1da177e4
LT		 274
275/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
276 * SPD calls.
277 */
278
0331b1f3 279struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 280{
281 struct xfrm_policy *policy;
282
0da974f4 283 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 284
285 if (policy) {
0331b1f3 286 write_pnet(&policy->xp_net, net);
12a169e7 287 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 288 INIT_HLIST_NODE(&policy->bydst);
289 INIT_HLIST_NODE(&policy->byidx);
1da177e4 290 rwlock_init(&policy->lock);
2518c7c2 291 atomic_set(&policy->refcnt, 1);
a0073fe1 292 skb_queue_head_init(&policy->polq.hold_queue);
b24b8a24
PE		 293 setup_timer(&policy->timer, xfrm_policy_timer,
294 (unsigned long)policy);
a0073fe1
SK		 295 setup_timer(&policy->polq.hold_timer, xfrm_policy_queue_process,
296 (unsigned long)policy);
fe1a5f03 297 policy->flo.ops = &xfrm_policy_fc_ops;
1da177e4
LT		 298 }
299 return policy;
300}
301EXPORT_SYMBOL(xfrm_policy_alloc);
302
303/* Destroy xfrm_policy: descendant resources must be released to this moment. */
304
64c31b3f 305void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 306{
12a169e7 307 BUG_ON(!policy->walk.dead);
1da177e4 308
0659eea9 309 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer))
1da177e4
LT		 310 BUG();
311
03e1ad7b 312 security_xfrm_policy_free(policy->security);
1da177e4
LT		 313 kfree(policy);
314}
64c31b3f 315EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4 316
a0073fe1
SK		 317static void xfrm_queue_purge(struct sk_buff_head *list)
318{
319 struct sk_buff *skb;
320
302a50bc 321 while ((skb = skb_dequeue(list)) != NULL)
a0073fe1 322 kfree_skb(skb);
a0073fe1
SK		 323}
324
1da177e4
LT		 325/* Rule must be locked. Release descentant resources, announce
326 * entry dead. The rule must be unlinked from lists to the moment.
327 */
328
329static void xfrm_policy_kill(struct xfrm_policy *policy)
330{
12a169e7 331 policy->walk.dead = 1;
1da177e4 332
285ead17 333 atomic_inc(&policy->genid);
1da177e4 334
e7d8f6cb
SK		 335 if (del_timer(&policy->polq.hold_timer))
336 xfrm_pol_put(policy);
a0073fe1
SK		 337 xfrm_queue_purge(&policy->polq.hold_queue);
338
285ead17
TT		 339 if (del_timer(&policy->timer))
340 xfrm_pol_put(policy);
341
342 xfrm_pol_put(policy);
1da177e4
LT		 343}
344
2518c7c2
DM		 345static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
346
e92303f8 347static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 348{
e92303f8 349 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 350}
351
b58555f1
CG		 352/* calculate policy hash thresholds */
353static void __get_hash_thresh(struct net *net,
354 unsigned short family, int dir,
355 u8 *dbits, u8 *sbits)
356{
357 switch (family) {
358 case AF_INET:
359 *dbits = net->xfrm.policy_bydst[dir].dbits4;
360 *sbits = net->xfrm.policy_bydst[dir].sbits4;
361 break;
362
363 case AF_INET6:
364 *dbits = net->xfrm.policy_bydst[dir].dbits6;
365 *sbits = net->xfrm.policy_bydst[dir].sbits6;
366 break;
367
368 default:
369 *dbits = 0;
370 *sbits = 0;
371 }
372}
373
5f803b58
DM		 374static struct hlist_head *policy_hash_bysel(struct net *net,
375 const struct xfrm_selector *sel,
376 unsigned short family, int dir)
2518c7c2 377{
1121994c 378 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 379 unsigned int hash;
380 u8 dbits;
381 u8 sbits;
382
383 __get_hash_thresh(net, family, dir, &dbits, &sbits);
384 hash = __sel_hash(sel, family, hmask, dbits, sbits);
2518c7c2
DM		 385
386 return (hash == hmask + 1 ?
1121994c
AD		 387 &net->xfrm.policy_inexact[dir] :
388 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 389}
390
5f803b58
DM		 391static struct hlist_head *policy_hash_direct(struct net *net,
392 const xfrm_address_t *daddr,
393 const xfrm_address_t *saddr,
394 unsigned short family, int dir)
2518c7c2 395{
1121994c 396 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
b58555f1
CG		 397 unsigned int hash;
398 u8 dbits;
399 u8 sbits;
400
401 __get_hash_thresh(net, family, dir, &dbits, &sbits);
402 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits);
2518c7c2 403
1121994c 404 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 405}
406
b58555f1
CG		 407static void xfrm_dst_hash_transfer(struct net *net,
408 struct hlist_head *list,
2518c7c2 409 struct hlist_head *ndsttable,
b58555f1
CG		 410 unsigned int nhashmask,
411 int dir)
2518c7c2 412{
b67bfe0d 413 struct hlist_node *tmp, *entry0 = NULL;
2518c7c2 414 struct xfrm_policy *pol;
b791160b 415 unsigned int h0 = 0;
b58555f1
CG		 416 u8 dbits;
417 u8 sbits;
2518c7c2 418
b791160b 419redo:
b67bfe0d 420 hlist_for_each_entry_safe(pol, tmp, list, bydst) {
2518c7c2
DM		 421 unsigned int h;
422
b58555f1 423 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits);
2518c7c2 424 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
b58555f1 425 pol->family, nhashmask, dbits, sbits);
b791160b 426 if (!entry0) {
b67bfe0d 427 hlist_del(&pol->bydst);
b791160b
YH		 428 hlist_add_head(&pol->bydst, ndsttable+h);
429 h0 = h;
430 } else {
431 if (h != h0)
432 continue;
b67bfe0d 433 hlist_del(&pol->bydst);
1d023284 434 hlist_add_behind(&pol->bydst, entry0);
b791160b 435 }
b67bfe0d 436 entry0 = &pol->bydst;
b791160b
YH		 437 }
438 if (!hlist_empty(list)) {
439 entry0 = NULL;
440 goto redo;
2518c7c2
DM		 441 }
442}
443
444static void xfrm_idx_hash_transfer(struct hlist_head *list,
445 struct hlist_head *nidxtable,
446 unsigned int nhashmask)
447{
b67bfe0d 448 struct hlist_node *tmp;
2518c7c2
DM		 449 struct xfrm_policy *pol;
450
b67bfe0d 451 hlist_for_each_entry_safe(pol, tmp, list, byidx) {
2518c7c2
DM		 452 unsigned int h;
453
454 h = __idx_hash(pol->index, nhashmask);
455 hlist_add_head(&pol->byidx, nidxtable+h);
456 }
457}
458
459static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
460{
461 return ((old_hmask + 1) << 1) - 1;
462}
463
66caf628 464static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 465{
66caf628 466 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 467 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
468 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 469 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 470 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 471 int i;
472
473 if (!ndst)
474 return;
475
283bc9f3 476 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 477
478 for (i = hmask; i >= 0; i--)
b58555f1 479 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir);
2518c7c2 480
66caf628
AD		 481 net->xfrm.policy_bydst[dir].table = ndst;
482 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2 483
283bc9f3 484 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 485
44e36b42 486 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 487}
488
66caf628 489static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 490{
66caf628 491 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 492 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
493 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 494 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 495 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 496 int i;
497
498 if (!nidx)
499 return;
500
283bc9f3 501 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 502
503 for (i = hmask; i >= 0; i--)
504 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
505
66caf628
AD		 506 net->xfrm.policy_byidx = nidx;
507 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2 508
283bc9f3 509 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2 510
44e36b42 511 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 512}
513
66caf628 514static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 515{
66caf628
AD		 516 unsigned int cnt = net->xfrm.policy_count[dir];
517 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 518
519 if (total)
520 *total += cnt;
521
522 if ((hmask + 1) < xfrm_policy_hashmax &&
523 cnt > hmask)
524 return 1;
525
526 return 0;
527}
528
66caf628 529static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 530{
66caf628 531 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 532
533 if ((hmask + 1) < xfrm_policy_hashmax &&
534 total > hmask)
535 return 1;
536
537 return 0;
538}
539
e071041b 540void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18 541{
283bc9f3 542 read_lock_bh(&net->xfrm.xfrm_policy_lock);
e071041b
AD		 543 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
544 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
545 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
546 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
547 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
548 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
549 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18 550 si->spdhmcnt = xfrm_policy_hashmax;
283bc9f3 551 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
ecfd6b18
JHS		 552}
553EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 554
ecfd6b18 555static DEFINE_MUTEX(hash_resize_mutex);
66caf628 556static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 557{
66caf628 558 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 559 int dir, total;
560
561 mutex_lock(&hash_resize_mutex);
562
563 total = 0;
564 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 565 if (xfrm_bydst_should_resize(net, dir, &total))
566 xfrm_bydst_resize(net, dir);
2518c7c2 567 }
66caf628
AD		 568 if (xfrm_byidx_should_resize(net, total))
569 xfrm_byidx_resize(net, total);
2518c7c2
DM		 570
571 mutex_unlock(&hash_resize_mutex);
572}
573
880a6fab
CG		 574static void xfrm_hash_rebuild(struct work_struct *work)
575{
576 struct net *net = container_of(work, struct net,
577 xfrm.policy_hthresh.work);
578 unsigned int hmask;
579 struct xfrm_policy *pol;
580 struct xfrm_policy *policy;
581 struct hlist_head *chain;
582 struct hlist_head *odst;
583 struct hlist_node *newpos;
584 int i;
585 int dir;
586 unsigned seq;
587 u8 lbits4, rbits4, lbits6, rbits6;
588
589 mutex_lock(&hash_resize_mutex);
590
591 /* read selector prefixlen thresholds */
592 do {
593 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock);
594
595 lbits4 = net->xfrm.policy_hthresh.lbits4;
596 rbits4 = net->xfrm.policy_hthresh.rbits4;
597 lbits6 = net->xfrm.policy_hthresh.lbits6;
598 rbits6 = net->xfrm.policy_hthresh.rbits6;
599 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq));
600
601 write_lock_bh(&net->xfrm.xfrm_policy_lock);
602
603 /* reset the bydst and inexact table in all directions */
604 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
605 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
606 hmask = net->xfrm.policy_bydst[dir].hmask;
607 odst = net->xfrm.policy_bydst[dir].table;
608 for (i = hmask; i >= 0; i--)
609 INIT_HLIST_HEAD(odst + i);
610 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
611 /* dir out => dst = remote, src = local */
612 net->xfrm.policy_bydst[dir].dbits4 = rbits4;
613 net->xfrm.policy_bydst[dir].sbits4 = lbits4;
614 net->xfrm.policy_bydst[dir].dbits6 = rbits6;
615 net->xfrm.policy_bydst[dir].sbits6 = lbits6;
616 } else {
617 /* dir in/fwd => dst = local, src = remote */
618 net->xfrm.policy_bydst[dir].dbits4 = lbits4;
619 net->xfrm.policy_bydst[dir].sbits4 = rbits4;
620 net->xfrm.policy_bydst[dir].dbits6 = lbits6;
621 net->xfrm.policy_bydst[dir].sbits6 = rbits6;
622 }
623 }
624
625 /* re-insert all policies by order of creation */
626 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
627 newpos = NULL;
628 chain = policy_hash_bysel(net, &policy->selector,
629 policy->family,
630 xfrm_policy_id2dir(policy->index));
631 hlist_for_each_entry(pol, chain, bydst) {
632 if (policy->priority >= pol->priority)
633 newpos = &pol->bydst;
634 else
635 break;
636 }
637 if (newpos)
638 hlist_add_behind(&policy->bydst, newpos);
639 else
640 hlist_add_head(&policy->bydst, chain);
641 }
642
643 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
644
645 mutex_unlock(&hash_resize_mutex);
646}
647
648void xfrm_policy_hash_rebuild(struct net *net)
649{
650 schedule_work(&net->xfrm.policy_hthresh.work);
651}
652EXPORT_SYMBOL(xfrm_policy_hash_rebuild);
653
1da177e4
LT		 654/* Generate new index... KAME seems to generate them ordered by cost
655 * of an absolute inpredictability of ordering of rules. This will not pass. */
e682adf0 656static u32 xfrm_gen_index(struct net *net, int dir, u32 index)
1da177e4 657{
1da177e4
LT		 658 static u32 idx_generator;
659
660 for (;;) {
2518c7c2
DM		 661 struct hlist_head *list;
662 struct xfrm_policy *p;
663 u32 idx;
664 int found;
665
e682adf0
FD		 666 if (!index) {
667 idx = (idx_generator | dir);
668 idx_generator += 8;
669 } else {
670 idx = index;
671 index = 0;
672 }
673
1da177e4
LT		 674 if (idx == 0)
675 idx = 8;
1121994c 676 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2 677 found = 0;
b67bfe0d 678 hlist_for_each_entry(p, list, byidx) {
2518c7c2
DM		 679 if (p->index == idx) {
680 found = 1;
1da177e4 681 break;
2518c7c2 682 }
1da177e4 683 }
2518c7c2 684 if (!found)
1da177e4
LT		 685 return idx;
686 }
687}
688
2518c7c2
DM		 689static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
690{
691 u32 *p1 = (u32 *) s1;
692 u32 *p2 = (u32 *) s2;
693 int len = sizeof(struct xfrm_selector) / sizeof(u32);
694 int i;
695
696 for (i = 0; i < len; i++) {
697 if (p1[i] != p2[i])
698 return 1;
699 }
700
701 return 0;
702}
703
a0073fe1
SK		 704static void xfrm_policy_requeue(struct xfrm_policy *old,
705 struct xfrm_policy *new)
706{
707 struct xfrm_policy_queue *pq = &old->polq;
708 struct sk_buff_head list;
709
710 __skb_queue_head_init(&list);
711
712 spin_lock_bh(&pq->hold_queue.lock);
713 skb_queue_splice_init(&pq->hold_queue, &list);
e7d8f6cb
SK		 714 if (del_timer(&pq->hold_timer))
715 xfrm_pol_put(old);
a0073fe1
SK		 716 spin_unlock_bh(&pq->hold_queue.lock);
717
718 if (skb_queue_empty(&list))
719 return;
720
721 pq = &new->polq;
722
723 spin_lock_bh(&pq->hold_queue.lock);
724 skb_queue_splice(&list, &pq->hold_queue);
725 pq->timeout = XFRM_QUEUE_TMO_MIN;
e7d8f6cb
SK		 726 if (!mod_timer(&pq->hold_timer, jiffies))
727 xfrm_pol_hold(new);
a0073fe1
SK		 728 spin_unlock_bh(&pq->hold_queue.lock);
729}
730
7cb8a939
SK		 731static bool xfrm_policy_mark_match(struct xfrm_policy *policy,
732 struct xfrm_policy *pol)
733{
734 u32 mark = policy->mark.v & policy->mark.m;
735
736 if (policy->mark.v == pol->mark.v && policy->mark.m == pol->mark.m)
737 return true;
738
739 if ((mark & pol->mark.m) == pol->mark.v &&
740 policy->priority == pol->priority)
741 return true;
742
743 return false;
744}
745
1da177e4
LT		 746int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
747{
1121994c 748 struct net *net = xp_net(policy);
2518c7c2
DM		 749 struct xfrm_policy *pol;
750 struct xfrm_policy *delpol;
751 struct hlist_head *chain;
b67bfe0d 752 struct hlist_node *newpos;
1da177e4 753
283bc9f3 754 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1121994c 755 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 756 delpol = NULL;
757 newpos = NULL;
b67bfe0d 758 hlist_for_each_entry(pol, chain, bydst) {
a6c7ab55 759 if (pol->type == policy->type &&
2518c7c2 760 !selector_cmp(&pol->selector, &policy->selector) &&
7cb8a939 761 xfrm_policy_mark_match(policy, pol) &&
a6c7ab55
HX		 762 xfrm_sec_ctx_match(pol->security, policy->security) &&
763 !WARN_ON(delpol)) {
1da177e4 764 if (excl) {
283bc9f3 765 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 766 return -EEXIST;
767 }
1da177e4
LT		 768 delpol = pol;
769 if (policy->priority > pol->priority)
770 continue;
771 } else if (policy->priority >= pol->priority) {
a6c7ab55 772 newpos = &pol->bydst;
1da177e4
LT		 773 continue;
774 }
1da177e4
LT		 775 if (delpol)
776 break;
1da177e4
LT		 777 }
778 if (newpos)
1d023284 779 hlist_add_behind(&policy->bydst, newpos);
2518c7c2
DM		 780 else
781 hlist_add_head(&policy->bydst, chain);
1da177e4 782 xfrm_pol_hold(policy);
1121994c 783 net->xfrm.policy_count[dir]++;
ca925cf1 784 atomic_inc(&net->xfrm.flow_cache_genid);
ca4c3fc2 785
786 /* After previous checking, family can either be AF_INET or AF_INET6 */
787 if (policy->family == AF_INET)
788 rt_genid_bump_ipv4(net);
789 else
790 rt_genid_bump_ipv6(net);
791
a0073fe1
SK		 792 if (delpol) {
793 xfrm_policy_requeue(delpol, policy);
29fa0b30 794 __xfrm_policy_unlink(delpol, dir);
a0073fe1 795 }
e682adf0 796 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index);
1121994c 797 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 798 policy->curlft.add_time = get_seconds();
1da177e4
LT		 799 policy->curlft.use_time = 0;
800 if (!mod_timer(&policy->timer, jiffies + HZ))
801 xfrm_pol_hold(policy);
1121994c 802 list_add(&policy->walk.all, &net->xfrm.policy_all);
283bc9f3 803 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 804
9b78a82c 805 if (delpol)
1da177e4 806 xfrm_policy_kill(delpol);
1121994c
AD		 807 else if (xfrm_bydst_should_resize(net, dir, NULL))
808 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c 809
1da177e4
LT		 810 return 0;
811}
812EXPORT_SYMBOL(xfrm_policy_insert);
813
8ca2e93b
JHS		 814struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
815 int dir, struct xfrm_selector *sel,
ef41aaa0
EP		 816 struct xfrm_sec_ctx *ctx, int delete,
817 int *err)
1da177e4 818{
2518c7c2
DM		 819 struct xfrm_policy *pol, *ret;
820 struct hlist_head *chain;
1da177e4 821
ef41aaa0 822 *err = 0;
283bc9f3 823 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 824 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2 825 ret = NULL;
b67bfe0d 826 hlist_for_each_entry(pol, chain, bydst) {
2518c7c2 827 if (pol->type == type &&
34f8d884 828 (mark & pol->mark.m) == pol->mark.v &&
2518c7c2
DM		 829 !selector_cmp(sel, &pol->selector) &&
830 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 831 xfrm_pol_hold(pol);
2518c7c2 832 if (delete) {
03e1ad7b
PM		 833 *err = security_xfrm_policy_delete(
834 pol->security);
ef41aaa0 835 if (*err) {
283bc9f3 836 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 837 return pol;
838 }
29fa0b30 839 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 840 }
841 ret = pol;
1da177e4
LT		 842 break;
843 }
844 }
283bc9f3 845 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 846
fe1a5f03 847 if (ret && delete)
2518c7c2 848 xfrm_policy_kill(ret);
2518c7c2 849 return ret;
1da177e4 850}
df71837d 851EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 852
8ca2e93b
JHS		 853struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
854 int dir, u32 id, int delete, int *err)
1da177e4 855{
2518c7c2
DM		 856 struct xfrm_policy *pol, *ret;
857 struct hlist_head *chain;
1da177e4 858
b5505c6e
HX		 859 *err = -ENOENT;
860 if (xfrm_policy_id2dir(id) != dir)
861 return NULL;
862
ef41aaa0 863 *err = 0;
283bc9f3 864 write_lock_bh(&net->xfrm.xfrm_policy_lock);
8d1211a6 865 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2 866 ret = NULL;
b67bfe0d 867 hlist_for_each_entry(pol, chain, byidx) {
34f8d884
JHS		 868 if (pol->type == type && pol->index == id &&
869 (mark & pol->mark.m) == pol->mark.v) {
1da177e4 870 xfrm_pol_hold(pol);
2518c7c2 871 if (delete) {
03e1ad7b
PM		 872 *err = security_xfrm_policy_delete(
873 pol->security);
ef41aaa0 874 if (*err) {
283bc9f3 875 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ef41aaa0
EP		 876 return pol;
877 }
29fa0b30 878 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 879 }
880 ret = pol;
1da177e4
LT		 881 break;
882 }
883 }
283bc9f3 884 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 885
fe1a5f03 886 if (ret && delete)
2518c7c2 887 xfrm_policy_kill(ret);
2518c7c2 888 return ret;
1da177e4
LT		 889}
890EXPORT_SYMBOL(xfrm_policy_byid);
891
4aa2e62c
JL		 892#ifdef CONFIG_SECURITY_NETWORK_XFRM
893static inline int
2e71029e 894xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
1da177e4 895{
4aa2e62c
JL		 896 int dir, err = 0;
897
898 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
899 struct xfrm_policy *pol;
4aa2e62c
JL		 900 int i;
901
b67bfe0d 902 hlist_for_each_entry(pol,
33ffbbd5 903 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 904 if (pol->type != type)
905 continue;
03e1ad7b 906 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 907 if (err) {
2e71029e 908 xfrm_audit_policy_delete(pol, 0, task_valid);
4aa2e62c
JL		 909 return err;
910 }
7dc12d6d 911 }
33ffbbd5 912 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
b67bfe0d 913 hlist_for_each_entry(pol,
33ffbbd5 914 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 915 bydst) {
916 if (pol->type != type)
917 continue;
03e1ad7b
PM		 918 err = security_xfrm_policy_delete(
919 pol->security);
4aa2e62c 920 if (err) {
ab5f5e8b 921 xfrm_audit_policy_delete(pol, 0,
2e71029e 922 task_valid);
4aa2e62c
JL		 923 return err;
924 }
925 }
926 }
927 }
928 return err;
929}
930#else
931static inline int
2e71029e 932xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid)
4aa2e62c
JL		 933{
934 return 0;
935}
936#endif
937
2e71029e 938int xfrm_policy_flush(struct net *net, u8 type, bool task_valid)
4aa2e62c 939{
2f1eb65f 940 int dir, err = 0, cnt = 0;
1da177e4 941
283bc9f3 942 write_lock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 943
2e71029e 944 err = xfrm_policy_flush_secctx_check(net, type, task_valid);
4aa2e62c
JL		 945 if (err)
946 goto out;
947
1da177e4 948 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2 949 struct xfrm_policy *pol;
29fa0b30 950 int i;
2518c7c2
DM		 951
952 again1:
b67bfe0d 953 hlist_for_each_entry(pol,
33ffbbd5 954 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 955 if (pol->type != type)
956 continue;
ea2dea9d 957 __xfrm_policy_unlink(pol, dir);
283bc9f3 958 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 959 cnt++;
1da177e4 960
2e71029e 961 xfrm_audit_policy_delete(pol, 1, task_valid);
161a09e7 962
2518c7c2 963 xfrm_policy_kill(pol);
1da177e4 964
283bc9f3 965 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 966 goto again1;
967 }
968
33ffbbd5 969 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2 970 again2:
b67bfe0d 971 hlist_for_each_entry(pol,
33ffbbd5 972 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 973 bydst) {
974 if (pol->type != type)
975 continue;
ea2dea9d 976 __xfrm_policy_unlink(pol, dir);
283bc9f3 977 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
ea2dea9d 978 cnt++;
2518c7c2 979
2e71029e 980 xfrm_audit_policy_delete(pol, 1, task_valid);
2518c7c2
DM		 981 xfrm_policy_kill(pol);
982
283bc9f3 983 write_lock_bh(&net->xfrm.xfrm_policy_lock);
2518c7c2
DM		 984 goto again2;
985 }
1da177e4 986 }
2518c7c2 987
1da177e4 988 }
2f1eb65f
JHS		 989 if (!cnt)
990 err = -ESRCH;
4aa2e62c 991out:
283bc9f3 992 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
4aa2e62c 993 return err;
1da177e4
LT		 994}
995EXPORT_SYMBOL(xfrm_policy_flush);
996
cdcbca7c 997int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 998 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 999 void *data)
1000{
12a169e7
HX		 1001 struct xfrm_policy *pol;
1002 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 1003 int error = 0;
1004
1005 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
1006 walk->type != XFRM_POLICY_TYPE_ANY)
1007 return -EINVAL;
1da177e4 1008
12a169e7 1009 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 1010 return 0;
1011
283bc9f3 1012 write_lock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7 1013 if (list_empty(&walk->walk.all))
cdcbca7c 1014 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 1015 else
1016 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 1017 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 1018 if (x->dead)
4c563f76 1019 continue;
12a169e7
HX		 1020 pol = container_of(x, struct xfrm_policy, walk);
1021 if (walk->type != XFRM_POLICY_TYPE_ANY &&
1022 walk->type != pol->type)
1023 continue;
1024 error = func(pol, xfrm_policy_id2dir(pol->index),
1025 walk->seq, data);
1026 if (error) {
1027 list_move_tail(&walk->walk.all, &x->all);
1028 goto out;
2518c7c2 1029 }
12a169e7 1030 walk->seq++;
1da177e4 1031 }
12a169e7 1032 if (walk->seq == 0) {
baf5d743
JHS		 1033 error = -ENOENT;
1034 goto out;
1035 }
12a169e7 1036 list_del_init(&walk->walk.all);
1da177e4 1037out:
283bc9f3 1038 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1039 return error;
1040}
1041EXPORT_SYMBOL(xfrm_policy_walk);
1042
12a169e7
HX		 1043void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
1044{
1045 INIT_LIST_HEAD(&walk->walk.all);
1046 walk->walk.dead = 1;
1047 walk->type = type;
1048 walk->seq = 0;
1049}
1050EXPORT_SYMBOL(xfrm_policy_walk_init);
1051
283bc9f3 1052void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net)
12a169e7
HX		 1053{
1054 if (list_empty(&walk->walk.all))
1055 return;
1056
283bc9f3 1057 write_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */
12a169e7 1058 list_del(&walk->walk.all);
283bc9f3 1059 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
12a169e7
HX		 1060}
1061EXPORT_SYMBOL(xfrm_policy_walk_done);
1062
134b0fc5
JM		 1063/*
1064 * Find policy to apply to this flow.
1065 *
1066 * Returns 0 if policy found, else an -errno.
1067 */
f299d557
DM		 1068static int xfrm_policy_match(const struct xfrm_policy *pol,
1069 const struct flowi *fl,
2518c7c2 1070 u8 type, u16 family, int dir)
1da177e4 1071{
f299d557 1072 const struct xfrm_selector *sel = &pol->selector;
bc9b35ad
DM		 1073 int ret = -ESRCH;
1074 bool match;
1da177e4 1075
2518c7c2 1076 if (pol->family != family ||
1d28f42c 1077 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
2518c7c2 1078 pol->type != type)
134b0fc5 1079 return ret;
1da177e4 1080
2518c7c2 1081 match = xfrm_selector_match(sel, fl, family);
134b0fc5 1082 if (match)
1d28f42c 1083 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
03e1ad7b 1084 dir);
2518c7c2 1085
134b0fc5 1086 return ret;
2518c7c2 1087}
1da177e4 1088
52479b62 1089static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
062cdb43 1090 const struct flowi *fl,
2518c7c2
DM		 1091 u16 family, u8 dir)
1092{
134b0fc5 1093 int err;
2518c7c2 1094 struct xfrm_policy *pol, *ret;
0b597e7e 1095 const xfrm_address_t *daddr, *saddr;
2518c7c2 1096 struct hlist_head *chain;
acba48e1 1097 u32 priority = ~0U;
df71837d 1098
2518c7c2
DM		 1099 daddr = xfrm_flowi_daddr(fl, family);
1100 saddr = xfrm_flowi_saddr(fl, family);
1101 if (unlikely(!daddr || !saddr))
1102 return NULL;
1103
283bc9f3 1104 read_lock_bh(&net->xfrm.xfrm_policy_lock);
52479b62 1105 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2 1106 ret = NULL;
b67bfe0d 1107 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1108 err = xfrm_policy_match(pol, fl, type, family, dir);
1109 if (err) {
1110 if (err == -ESRCH)
1111 continue;
1112 else {
1113 ret = ERR_PTR(err);
1114 goto fail;
1115 }
1116 } else {
2518c7c2 1117 ret = pol;
acba48e1 1118 priority = ret->priority;
2518c7c2
DM		 1119 break;
1120 }
1121 }
52479b62 1122 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 1123 hlist_for_each_entry(pol, chain, bydst) {
134b0fc5
JM		 1124 err = xfrm_policy_match(pol, fl, type, family, dir);
1125 if (err) {
1126 if (err == -ESRCH)
1127 continue;
1128 else {
1129 ret = ERR_PTR(err);
1130 goto fail;
1131 }
1132 } else if (pol->priority < priority) {
acba48e1
DM		 1133 ret = pol;
1134 break;
1da177e4
LT		 1135 }
1136 }
acba48e1
DM		 1137 if (ret)
1138 xfrm_pol_hold(ret);
134b0fc5 1139fail:
283bc9f3 1140 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
4e81bb83 1141
2518c7c2 1142 return ret;
4e81bb83
MN		 1143}
1144
80c802f3 1145static struct xfrm_policy *
73ff93cd 1146__xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
80c802f3
TT		 1147{
1148#ifdef CONFIG_XFRM_SUB_POLICY
1149 struct xfrm_policy *pol;
1150
1151 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1152 if (pol != NULL)
1153 return pol;
1154#endif
1155 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1156}
1157
b5fb82c4
BZ		 1158static int flow_to_policy_dir(int dir)
1159{
1160 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1161 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1162 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1163 return dir;
1164
1165 switch (dir) {
1166 default:
1167 case FLOW_DIR_IN:
1168 return XFRM_POLICY_IN;
1169 case FLOW_DIR_OUT:
1170 return XFRM_POLICY_OUT;
1171 case FLOW_DIR_FWD:
1172 return XFRM_POLICY_FWD;
1173 }
1174}
1175
fe1a5f03 1176static struct flow_cache_object *
dee9f4bc 1177xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
fe1a5f03 1178 u8 dir, struct flow_cache_object *old_obj, void *ctx)
4e81bb83
MN		 1179{
1180 struct xfrm_policy *pol;
fe1a5f03
TT		 1181
1182 if (old_obj)
1183 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
4e81bb83 1184
b5fb82c4 1185 pol = __xfrm_policy_lookup(net, fl, family, flow_to_policy_dir(dir));
80c802f3 1186 if (IS_ERR_OR_NULL(pol))
fe1a5f03 1187 return ERR_CAST(pol);
fe1a5f03 1188
fe1a5f03
TT		 1189 /* Resolver returns two references:
1190 * one for cache and one for caller of flow_cache_lookup() */
1191 xfrm_pol_hold(pol);
1192
1193 return &pol->flo;
1da177e4
LT		 1194}
1195
df71837d
TJ		 1196static inline int policy_to_flow_dir(int dir)
1197{
1198 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1199 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1200 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1201 return dir;
1202 switch (dir) {
1203 default:
1204 case XFRM_POLICY_IN:
1205 return FLOW_DIR_IN;
1206 case XFRM_POLICY_OUT:
1207 return FLOW_DIR_OUT;
1208 case XFRM_POLICY_FWD:
1209 return FLOW_DIR_FWD;
3ff50b79 1210 }
df71837d
TJ		 1211}
1212
dee9f4bc
DM		 1213static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1214 const struct flowi *fl)
1da177e4
LT		 1215{
1216 struct xfrm_policy *pol;
283bc9f3 1217 struct net *net = sock_net(sk);
1da177e4 1218
283bc9f3 1219 read_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1220 if ((pol = sk->sk_policy[dir]) != NULL) {
bc9b35ad
DM		 1221 bool match = xfrm_selector_match(&pol->selector, fl,
1222 sk->sk_family);
a716c119 1223 int err = 0;
df71837d 1224
3bccfbc7 1225 if (match) {
34f8d884
JHS		 1226 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1227 pol = NULL;
1228 goto out;
1229 }
03e1ad7b 1230 err = security_xfrm_policy_lookup(pol->security,
1d28f42c 1231 fl->flowi_secid,
03e1ad7b 1232 policy_to_flow_dir(dir));
3bccfbc7
VY		 1233 if (!err)
1234 xfrm_pol_hold(pol);
1235 else if (err == -ESRCH)
1236 pol = NULL;
1237 else
1238 pol = ERR_PTR(err);
1239 } else
1da177e4
LT		 1240 pol = NULL;
1241 }
34f8d884 1242out:
283bc9f3 1243 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1244 return pol;
1245}
1246
1247static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1248{
98806f75 1249 struct net *net = xp_net(pol);
1121994c 1250 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1251 pol->family, dir);
4e81bb83 1252
98806f75 1253 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1254 hlist_add_head(&pol->bydst, chain);
e92303f8 1255 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1256 net->xfrm.policy_count[dir]++;
1da177e4 1257 xfrm_pol_hold(pol);
2518c7c2 1258
98806f75
AD		 1259 if (xfrm_bydst_should_resize(net, dir, NULL))
1260 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1261}
1262
1263static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1264 int dir)
1265{
98806f75
AD		 1266 struct net *net = xp_net(pol);
1267
2518c7c2
DM		 1268 if (hlist_unhashed(&pol->bydst))
1269 return NULL;
1da177e4 1270
3a9016f9 1271 hlist_del_init(&pol->bydst);
2518c7c2 1272 hlist_del(&pol->byidx);
12a169e7 1273 list_del(&pol->walk.all);
98806f75 1274 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1275
1276 return pol;
1da177e4
LT		 1277}
1278
4666faab 1279int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4 1280{
283bc9f3
FD		 1281 struct net *net = xp_net(pol);
1282
1283 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1284 pol = __xfrm_policy_unlink(pol, dir);
283bc9f3 1285 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1286 if (pol) {
1da177e4 1287 xfrm_policy_kill(pol);
4666faab 1288 return 0;
1da177e4 1289 }
4666faab 1290 return -ENOENT;
1da177e4 1291}
a70fcb0b 1292EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1293
1294int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1295{
1121994c 1296 struct net *net = xp_net(pol);
1da177e4
LT		 1297 struct xfrm_policy *old_pol;
1298
4e81bb83
MN		 1299#ifdef CONFIG_XFRM_SUB_POLICY
1300 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1301 return -EINVAL;
1302#endif
1303
283bc9f3 1304 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1305 old_pol = sk->sk_policy[dir];
1306 sk->sk_policy[dir] = pol;
1307 if (pol) {
9d729f72 1308 pol->curlft.add_time = get_seconds();
e682adf0 1309 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0);
1da177e4
LT		 1310 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1311 }
a0073fe1
SK		 1312 if (old_pol) {
1313 if (pol)
1314 xfrm_policy_requeue(old_pol, pol);
1315
ea2dea9d
TT		 1316 /* Unlinking succeeds always. This is the only function
1317 * allowed to delete or replace socket policy.
1318 */
1da177e4 1319 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
a0073fe1 1320 }
283bc9f3 1321 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1322
1323 if (old_pol) {
1324 xfrm_policy_kill(old_pol);
1325 }
1326 return 0;
1327}
1328
d3e40a9f 1329static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1da177e4 1330{
0331b1f3 1331 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
283bc9f3 1332 struct net *net = xp_net(old);
1da177e4
LT		 1333
1334 if (newp) {
1335 newp->selector = old->selector;
03e1ad7b
PM		 1336 if (security_xfrm_policy_clone(old->security,
1337 &newp->security)) {
df71837d
TJ		 1338 kfree(newp);
1339 return NULL; /* ENOMEM */
1340 }
1da177e4
LT		 1341 newp->lft = old->lft;
1342 newp->curlft = old->curlft;
fb977e2c 1343 newp->mark = old->mark;
1da177e4
LT		 1344 newp->action = old->action;
1345 newp->flags = old->flags;
1346 newp->xfrm_nr = old->xfrm_nr;
1347 newp->index = old->index;
4e81bb83 1348 newp->type = old->type;
1da177e4
LT		 1349 memcpy(newp->xfrm_vec, old->xfrm_vec,
1350 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
283bc9f3 1351 write_lock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4 1352 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
283bc9f3 1353 write_unlock_bh(&net->xfrm.xfrm_policy_lock);
1da177e4
LT		 1354 xfrm_pol_put(newp);
1355 }
1356 return newp;
1357}
1358
1359int __xfrm_sk_clone_policy(struct sock *sk)
1360{
1361 struct xfrm_policy *p0 = sk->sk_policy[0],
1362 *p1 = sk->sk_policy[1];
1363
1364 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1365 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1366 return -ENOMEM;
1367 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1368 return -ENOMEM;
1369 return 0;
1370}
1371
a1e59abf 1372static int
fbda33b2 1373xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1374 unsigned short family)
1375{
1376 int err;
1377 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1378
1379 if (unlikely(afinfo == NULL))
1380 return -EINVAL;
fbda33b2 1381 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1382 xfrm_policy_put_afinfo(afinfo);
1383 return err;
1384}
1385
1da177e4
LT		 1386/* Resolve list of templates for the flow, given policy. */
1387
1388static int
a6c2e611
DM		 1389xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1390 struct xfrm_state **xfrm, unsigned short family)
1da177e4 1391{
fbda33b2 1392 struct net *net = xp_net(policy);
1da177e4
LT		 1393 int nx;
1394 int i, error;
1395 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1396 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1397 xfrm_address_t tmp;
1da177e4 1398
9b7a787d 1399 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) {
1da177e4
LT		 1400 struct xfrm_state *x;
1401 xfrm_address_t *remote = daddr;
1402 xfrm_address_t *local = saddr;
1403 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1404
48b8d783
JK		 1405 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1406 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1407 remote = &tmpl->id.daddr;
1408 local = &tmpl->saddr;
8444cf71
TE		 1409 if (xfrm_addr_any(local, tmpl->encap_family)) {
1410 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
a1e59abf
PM		 1411 if (error)
1412 goto fail;
1413 local = &tmp;
1414 }
1da177e4
LT		 1415 }
1416
1417 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1418
1419 if (x && x->km.state == XFRM_STATE_VALID) {
1420 xfrm[nx++] = x;
1421 daddr = remote;
1422 saddr = local;
1423 continue;
1424 }
1425 if (x) {
1426 error = (x->km.state == XFRM_STATE_ERROR ?
1427 -EINVAL : -EAGAIN);
1428 xfrm_state_put(x);
42054569 1429 } else if (error == -ESRCH) {
a4322266 1430 error = -EAGAIN;
42054569 1431 }
1da177e4
LT		 1432
1433 if (!tmpl->optional)
1434 goto fail;
1435 }
1436 return nx;
1437
1438fail:
9b7a787d 1439 for (nx--; nx >= 0; nx--)
1da177e4
LT		 1440 xfrm_state_put(xfrm[nx]);
1441 return error;
1442}
1443
4e81bb83 1444static int
a6c2e611
DM		 1445xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1446 struct xfrm_state **xfrm, unsigned short family)
4e81bb83 1447{
41a49cc3
MN		 1448 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1449 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1450 int cnx = 0;
1451 int error;
1452 int ret;
1453 int i;
1454
1455 for (i = 0; i < npols; i++) {
1456 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1457 error = -ENOBUFS;
1458 goto fail;
1459 }
41a49cc3
MN		 1460
1461 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1462 if (ret < 0) {
1463 error = ret;
1464 goto fail;
1465 } else
1466 cnx += ret;
1467 }
1468
41a49cc3
MN		 1469 /* found states are sorted for outbound processing */
1470 if (npols > 1)
1471 xfrm_state_sort(xfrm, tpp, cnx, family);
1472
4e81bb83
MN		 1473 return cnx;
1474
1475 fail:
9b7a787d 1476 for (cnx--; cnx >= 0; cnx--)
41a49cc3 1477 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1478 return error;
1479
1480}
1481
1da177e4
LT		 1482/* Check that the bundle accepts the flow and its components are
1483 * still valid.
1484 */
1485
05d84025 1486static inline int xfrm_get_tos(const struct flowi *fl, int family)
25ee3286
HX		 1487{
1488 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1489 int tos;
1da177e4 1490
25ee3286
HX		 1491 if (!afinfo)
1492 return -EINVAL;
1493
1494 tos = afinfo->get_tos(fl);
1495
1496 xfrm_policy_put_afinfo(afinfo);
1497
1498 return tos;
1499}
1500
80c802f3
TT		 1501static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1502{
1503 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1504 struct dst_entry *dst = &xdst->u.dst;
1505
1506 if (xdst->route == NULL) {
1507 /* Dummy bundle - if it has xfrms we were not
1508 * able to build bundle as template resolution failed.
1509 * It means we need to try again resolving. */
1510 if (xdst->num_xfrms > 0)
1511 return NULL;
a0073fe1
SK		 1512 } else if (dst->flags & DST_XFRM_QUEUE) {
1513 return NULL;
80c802f3
TT		 1514 } else {
1515 /* Real bundle */
1516 if (stale_bundle(dst))
1517 return NULL;
1518 }
1519
1520 dst_hold(dst);
1521 return flo;
1522}
1523
1524static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1525{
1526 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1527 struct dst_entry *dst = &xdst->u.dst;
1528
1529 if (!xdst->route)
1530 return 0;
1531 if (stale_bundle(dst))
1532 return 0;
1533
1534 return 1;
1535}
1536
1537static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1538{
1539 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1540 struct dst_entry *dst = &xdst->u.dst;
1541
1542 dst_free(dst);
1543}
1544
1545static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1546 .get = xfrm_bundle_flo_get,
1547 .check = xfrm_bundle_flo_check,
1548 .delete = xfrm_bundle_flo_delete,
1549};
1550
d7c7544c 1551static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1da177e4 1552{
1da177e4 1553 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
d7c7544c 1554 struct dst_ops *dst_ops;
25ee3286
HX		 1555 struct xfrm_dst *xdst;
1556
1557 if (!afinfo)
1558 return ERR_PTR(-EINVAL);
1559
d7c7544c
AD		 1560 switch (family) {
1561 case AF_INET:
1562 dst_ops = &net->xfrm.xfrm4_dst_ops;
1563 break;
dfd56b8b 1564#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 1565 case AF_INET6:
1566 dst_ops = &net->xfrm.xfrm6_dst_ops;
1567 break;
1568#endif
1569 default:
1570 BUG();
1571 }
f5b0a874 1572 xdst = dst_alloc(dst_ops, NULL, 0, DST_OBSOLETE_NONE, 0);
25ee3286 1573
d4cae562 1574 if (likely(xdst)) {
141e369d
SK		 1575 struct dst_entry *dst = &xdst->u.dst;
1576
1577 memset(dst + 1, 0, sizeof(*xdst) - sizeof(*dst));
0b150932 1578 xdst->flo.ops = &xfrm_bundle_fc_ops;
9d7b0fc1
PM		 1579 if (afinfo->init_dst)
1580 afinfo->init_dst(net, xdst);
d4cae562 1581 } else
0b150932 1582 xdst = ERR_PTR(-ENOBUFS);
80c802f3 1583
d4cae562
MB		 1584 xfrm_policy_put_afinfo(afinfo);
1585
25ee3286
HX		 1586 return xdst;
1587}
1588
a1b05140
MN		 1589static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1590 int nfheader_len)
1591{
1592 struct xfrm_policy_afinfo *afinfo =
1593 xfrm_policy_get_afinfo(dst->ops->family);
1594 int err;
1595
1596 if (!afinfo)
1597 return -EINVAL;
1598
1599 err = afinfo->init_path(path, dst, nfheader_len);
1600
1601 xfrm_policy_put_afinfo(afinfo);
1602
1603 return err;
1604}
1605
87c1e12b 1606static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
0c7b3eef 1607 const struct flowi *fl)
25ee3286
HX		 1608{
1609 struct xfrm_policy_afinfo *afinfo =
1610 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1611 int err;
1612
1613 if (!afinfo)
1da177e4 1614 return -EINVAL;
25ee3286 1615
87c1e12b 1616 err = afinfo->fill_dst(xdst, dev, fl);
25ee3286 1617
1da177e4 1618 xfrm_policy_put_afinfo(afinfo);
25ee3286 1619
1da177e4
LT		 1620 return err;
1621}
1622
80c802f3 1623
25ee3286
HX		 1624/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1625 * all the metrics... Shortly, bundle a bundle.
1626 */
1627
1628static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1629 struct xfrm_state **xfrm, int nx,
98313ada 1630 const struct flowi *fl,
25ee3286
HX		 1631 struct dst_entry *dst)
1632{
d7c7544c 1633 struct net *net = xp_net(policy);
25ee3286
HX		 1634 unsigned long now = jiffies;
1635 struct net_device *dev;
43a4dea4 1636 struct xfrm_mode *inner_mode;
25ee3286
HX		 1637 struct dst_entry *dst_prev = NULL;
1638 struct dst_entry *dst0 = NULL;
1639 int i = 0;
1640 int err;
1641 int header_len = 0;
a1b05140 1642 int nfheader_len = 0;
25ee3286
HX		 1643 int trailer_len = 0;
1644 int tos;
1645 int family = policy->selector.family;
9bb182a7
YH		 1646 xfrm_address_t saddr, daddr;
1647
1648 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1649
1650 tos = xfrm_get_tos(fl, family);
1651 err = tos;
1652 if (tos < 0)
1653 goto put_states;
1654
1655 dst_hold(dst);
1656
1657 for (; i < nx; i++) {
d7c7544c 1658 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
25ee3286
HX		 1659 struct dst_entry *dst1 = &xdst->u.dst;
1660
1661 err = PTR_ERR(xdst);
1662 if (IS_ERR(xdst)) {
1663 dst_release(dst);
1664 goto put_states;
1665 }
1666
43a4dea4
SK		 1667 if (xfrm[i]->sel.family == AF_UNSPEC) {
1668 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1669 xfrm_af2proto(family));
1670 if (!inner_mode) {
1671 err = -EAFNOSUPPORT;
1672 dst_release(dst);
1673 goto put_states;
1674 }
1675 } else
1676 inner_mode = xfrm[i]->inner_mode;
1677
25ee3286
HX		 1678 if (!dst_prev)
1679 dst0 = dst1;
1680 else {
1681 dst_prev->child = dst_clone(dst1);
1682 dst1->flags |= DST_NOHASH;
1683 }
1684
1685 xdst->route = dst;
defb3519 1686 dst_copy_metrics(dst1, dst);
25ee3286
HX		 1687
1688 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1689 family = xfrm[i]->props.family;
9bb182a7
YH		 1690 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1691 family);
25ee3286
HX		 1692 err = PTR_ERR(dst);
1693 if (IS_ERR(dst))
1694 goto put_states;
1695 } else
1696 dst_hold(dst);
1697
1698 dst1->xfrm = xfrm[i];
80c802f3 1699 xdst->xfrm_genid = xfrm[i]->genid;
25ee3286 1700
f5b0a874 1701 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
25ee3286
HX		 1702 dst1->flags |= DST_HOST;
1703 dst1->lastuse = now;
1704
1705 dst1->input = dst_discard;
43a4dea4 1706 dst1->output = inner_mode->afinfo->output;
25ee3286
HX		 1707
1708 dst1->next = dst_prev;
1709 dst_prev = dst1;
1710
1711 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1712 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1713 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1714 trailer_len += xfrm[i]->props.trailer_len;
1715 }
1716
1717 dst_prev->child = dst;
1718 dst0->path = dst;
1719
1720 err = -ENODEV;
1721 dev = dst->dev;
1722 if (!dev)
1723 goto free_dst;
1724
a1b05140 1725 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1726 xfrm_init_pmtu(dst_prev);
1727
1728 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1729 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1730
87c1e12b 1731 err = xfrm_fill_dst(xdst, dev, fl);
25ee3286
HX		 1732 if (err)
1733 goto free_dst;
1734
1735 dst_prev->header_len = header_len;
1736 dst_prev->trailer_len = trailer_len;
1737 header_len -= xdst->u.dst.xfrm->props.header_len;
1738 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1739 }
1740
1741out:
1742 return dst0;
1743
1744put_states:
1745 for (; i < nx; i++)
1746 xfrm_state_put(xfrm[i]);
1747free_dst:
1748 if (dst0)
1749 dst_free(dst0);
1750 dst0 = ERR_PTR(err);
1751 goto out;
1752}
1753
da7c224b 1754#ifdef CONFIG_XFRM_SUB_POLICY
be7928d2 1755static int xfrm_dst_alloc_copy(void **target, const void *src, int size)
157bfc25
MN		 1756{
1757 if (!*target) {
1758 *target = kmalloc(size, GFP_ATOMIC);
1759 if (!*target)
1760 return -ENOMEM;
1761 }
be7928d2 1762
157bfc25
MN		 1763 memcpy(*target, src, size);
1764 return 0;
1765}
da7c224b 1766#endif
157bfc25 1767
be7928d2
DB		 1768static int xfrm_dst_update_parent(struct dst_entry *dst,
1769 const struct xfrm_selector *sel)
157bfc25
MN		 1770{
1771#ifdef CONFIG_XFRM_SUB_POLICY
1772 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1773 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1774 sel, sizeof(*sel));
1775#else
1776 return 0;
1777#endif
1778}
1779
be7928d2
DB		 1780static int xfrm_dst_update_origin(struct dst_entry *dst,
1781 const struct flowi *fl)
157bfc25
MN		 1782{
1783#ifdef CONFIG_XFRM_SUB_POLICY
1784 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1785 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1786#else
1787 return 0;
1788#endif
1789}
1da177e4 1790
73ff93cd 1791static int xfrm_expand_policies(const struct flowi *fl, u16 family,
80c802f3
TT		 1792 struct xfrm_policy **pols,
1793 int *num_pols, int *num_xfrms)
1794{
1795 int i;
1796
1797 if (*num_pols == 0 || !pols[0]) {
1798 *num_pols = 0;
1799 *num_xfrms = 0;
1800 return 0;
1801 }
1802 if (IS_ERR(pols[0]))
1803 return PTR_ERR(pols[0]);
1804
1805 *num_xfrms = pols[0]->xfrm_nr;
1806
1807#ifdef CONFIG_XFRM_SUB_POLICY
1808 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1809 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1810 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1811 XFRM_POLICY_TYPE_MAIN,
1812 fl, family,
1813 XFRM_POLICY_OUT);
1814 if (pols[1]) {
1815 if (IS_ERR(pols[1])) {
1816 xfrm_pols_put(pols, *num_pols);
1817 return PTR_ERR(pols[1]);
1818 }
02d0892f 1819 (*num_pols)++;
80c802f3
TT		 1820 (*num_xfrms) += pols[1]->xfrm_nr;
1821 }
1822 }
1823#endif
1824 for (i = 0; i < *num_pols; i++) {
1825 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1826 *num_xfrms = -1;
1827 break;
1828 }
1829 }
1830
1831 return 0;
1832
1833}
1834
1835static struct xfrm_dst *
1836xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
4ca2e685 1837 const struct flowi *fl, u16 family,
80c802f3
TT		 1838 struct dst_entry *dst_orig)
1839{
1840 struct net *net = xp_net(pols[0]);
1841 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1842 struct dst_entry *dst;
1843 struct xfrm_dst *xdst;
1844 int err;
1845
1846 /* Try to instantiate a bundle */
1847 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
d809ec89
TT		 1848 if (err <= 0) {
1849 if (err != 0 && err != -EAGAIN)
80c802f3
TT		 1850 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1851 return ERR_PTR(err);
1852 }
1853
1854 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1855 if (IS_ERR(dst)) {
1856 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1857 return ERR_CAST(dst);
1858 }
1859
1860 xdst = (struct xfrm_dst *)dst;
1861 xdst->num_xfrms = err;
1862 if (num_pols > 1)
1863 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1864 else
1865 err = xfrm_dst_update_origin(dst, fl);
1866 if (unlikely(err)) {
1867 dst_free(dst);
1868 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1869 return ERR_PTR(err);
1870 }
1871
1872 xdst->num_pols = num_pols;
3e94c2dc 1873 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 1874 xdst->policy_genid = atomic_read(&pols[0]->genid);
1875
1876 return xdst;
1877}
1878
a0073fe1
SK		 1879static void xfrm_policy_queue_process(unsigned long arg)
1880{
1881 int err = 0;
1882 struct sk_buff *skb;
1883 struct sock *sk;
1884 struct dst_entry *dst;
a0073fe1
SK		 1885 struct xfrm_policy *pol = (struct xfrm_policy *)arg;
1886 struct xfrm_policy_queue *pq = &pol->polq;
1887 struct flowi fl;
1888 struct sk_buff_head list;
1889
1890 spin_lock(&pq->hold_queue.lock);
1891 skb = skb_peek(&pq->hold_queue);
2bb53e25
SK		 1892 if (!skb) {
1893 spin_unlock(&pq->hold_queue.lock);
1894 goto out;
1895 }
a0073fe1
SK		 1896 dst = skb_dst(skb);
1897 sk = skb->sk;
1898 xfrm_decode_session(skb, &fl, dst->ops->family);
1899 spin_unlock(&pq->hold_queue.lock);
1900
1901 dst_hold(dst->path);
1902 dst = xfrm_lookup(xp_net(pol), dst->path, &fl,
1903 sk, 0);
1904 if (IS_ERR(dst))
1905 goto purge_queue;
1906
1907 if (dst->flags & DST_XFRM_QUEUE) {
1908 dst_release(dst);
1909
1910 if (pq->timeout >= XFRM_QUEUE_TMO_MAX)
1911 goto purge_queue;
1912
1913 pq->timeout = pq->timeout << 1;
e7d8f6cb
SK		 1914 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout))
1915 xfrm_pol_hold(pol);
1916 goto out;
a0073fe1
SK		 1917 }
1918
1919 dst_release(dst);
1920
1921 __skb_queue_head_init(&list);
1922
1923 spin_lock(&pq->hold_queue.lock);
1924 pq->timeout = 0;
1925 skb_queue_splice_init(&pq->hold_queue, &list);
1926 spin_unlock(&pq->hold_queue.lock);
1927
1928 while (!skb_queue_empty(&list)) {
1929 skb = __skb_dequeue(&list);
1930
1931 xfrm_decode_session(skb, &fl, skb_dst(skb)->ops->family);
1932 dst_hold(skb_dst(skb)->path);
1933 dst = xfrm_lookup(xp_net(pol), skb_dst(skb)->path,
1934 &fl, skb->sk, 0);
1935 if (IS_ERR(dst)) {
a0073fe1
SK		 1936 kfree_skb(skb);
1937 continue;
1938 }
1939
1940 nf_reset(skb);
1941 skb_dst_drop(skb);
1942 skb_dst_set(skb, dst);
1943
a0073fe1 1944 err = dst_output(skb);
a0073fe1
SK		 1945 }
1946
e7d8f6cb
SK		 1947out:
1948 xfrm_pol_put(pol);
a0073fe1
SK		 1949 return;
1950
1951purge_queue:
1952 pq->timeout = 0;
1953 xfrm_queue_purge(&pq->hold_queue);
e7d8f6cb 1954 xfrm_pol_put(pol);
a0073fe1
SK		 1955}
1956
aad88724 1957static int xdst_queue_output(struct sock *sk, struct sk_buff *skb)
a0073fe1
SK		 1958{
1959 unsigned long sched_next;
1960 struct dst_entry *dst = skb_dst(skb);
1961 struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
e7d8f6cb
SK		 1962 struct xfrm_policy *pol = xdst->pols[0];
1963 struct xfrm_policy_queue *pq = &pol->polq;
4d53eff4 1964
d0bf4a9e 1965 if (unlikely(skb_fclone_busy(skb))) {
4d53eff4
SK		 1966 kfree_skb(skb);
1967 return 0;
1968 }
a0073fe1
SK		 1969
1970 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) {
1971 kfree_skb(skb);
1972 return -EAGAIN;
1973 }
1974
1975 skb_dst_force(skb);
a0073fe1
SK		 1976
1977 spin_lock_bh(&pq->hold_queue.lock);
1978
1979 if (!pq->timeout)
1980 pq->timeout = XFRM_QUEUE_TMO_MIN;
1981
1982 sched_next = jiffies + pq->timeout;
1983
1984 if (del_timer(&pq->hold_timer)) {
1985 if (time_before(pq->hold_timer.expires, sched_next))
1986 sched_next = pq->hold_timer.expires;
e7d8f6cb 1987 xfrm_pol_put(pol);
a0073fe1
SK		 1988 }
1989
1990 __skb_queue_tail(&pq->hold_queue, skb);
e7d8f6cb
SK		 1991 if (!mod_timer(&pq->hold_timer, sched_next))
1992 xfrm_pol_hold(pol);
a0073fe1
SK		 1993
1994 spin_unlock_bh(&pq->hold_queue.lock);
1995
1996 return 0;
1997}
1998
1999static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net,
b8c203b2 2000 struct xfrm_flo *xflo,
a0073fe1
SK		 2001 const struct flowi *fl,
2002 int num_xfrms,
2003 u16 family)
2004{
2005 int err;
2006 struct net_device *dev;
b8c203b2 2007 struct dst_entry *dst;
a0073fe1
SK		 2008 struct dst_entry *dst1;
2009 struct xfrm_dst *xdst;
2010
2011 xdst = xfrm_alloc_dst(net, family);
2012 if (IS_ERR(xdst))
2013 return xdst;
2014
b8c203b2
SK		 2015 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) ||
2016 net->xfrm.sysctl_larval_drop ||
2017 num_xfrms <= 0)
a0073fe1
SK		 2018 return xdst;
2019
b8c203b2 2020 dst = xflo->dst_orig;
a0073fe1
SK		 2021 dst1 = &xdst->u.dst;
2022 dst_hold(dst);
2023 xdst->route = dst;
2024
2025 dst_copy_metrics(dst1, dst);
2026
2027 dst1->obsolete = DST_OBSOLETE_FORCE_CHK;
2028 dst1->flags |= DST_HOST | DST_XFRM_QUEUE;
2029 dst1->lastuse = jiffies;
2030
2031 dst1->input = dst_discard;
2032 dst1->output = xdst_queue_output;
2033
2034 dst_hold(dst);
2035 dst1->child = dst;
2036 dst1->path = dst;
2037
2038 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0);
2039
2040 err = -ENODEV;
2041 dev = dst->dev;
2042 if (!dev)
2043 goto free_dst;
2044
2045 err = xfrm_fill_dst(xdst, dev, fl);
2046 if (err)
2047 goto free_dst;
2048
2049out:
2050 return xdst;
2051
2052free_dst:
2053 dst_release(dst1);
2054 xdst = ERR_PTR(err);
2055 goto out;
2056}
2057
80c802f3 2058static struct flow_cache_object *
dee9f4bc 2059xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
80c802f3
TT		 2060 struct flow_cache_object *oldflo, void *ctx)
2061{
b8c203b2 2062 struct xfrm_flo *xflo = (struct xfrm_flo *)ctx;
80c802f3
TT		 2063 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2064 struct xfrm_dst *xdst, *new_xdst;
2065 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
2066
2067 /* Check if the policies from old bundle are usable */
2068 xdst = NULL;
2069 if (oldflo) {
2070 xdst = container_of(oldflo, struct xfrm_dst, flo);
2071 num_pols = xdst->num_pols;
2072 num_xfrms = xdst->num_xfrms;
2073 pol_dead = 0;
2074 for (i = 0; i < num_pols; i++) {
2075 pols[i] = xdst->pols[i];
2076 pol_dead |= pols[i]->walk.dead;
2077 }
2078 if (pol_dead) {
2079 dst_free(&xdst->u.dst);
2080 xdst = NULL;
2081 num_pols = 0;
2082 num_xfrms = 0;
2083 oldflo = NULL;
2084 }
2085 }
2086
2087 /* Resolve policies to use if we couldn't get them from
2088 * previous cache entry */
2089 if (xdst == NULL) {
2090 num_pols = 1;
b5fb82c4
BZ		 2091 pols[0] = __xfrm_policy_lookup(net, fl, family,
2092 flow_to_policy_dir(dir));
80c802f3
TT		 2093 err = xfrm_expand_policies(fl, family, pols,
2094 &num_pols, &num_xfrms);
2095 if (err < 0)
2096 goto inc_error;
2097 if (num_pols == 0)
2098 return NULL;
2099 if (num_xfrms <= 0)
2100 goto make_dummy_bundle;
2101 }
2102
b8c203b2
SK		 2103 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family,
2104 xflo->dst_orig);
80c802f3
TT		 2105 if (IS_ERR(new_xdst)) {
2106 err = PTR_ERR(new_xdst);
2107 if (err != -EAGAIN)
2108 goto error;
2109 if (oldflo == NULL)
2110 goto make_dummy_bundle;
2111 dst_hold(&xdst->u.dst);
2112 return oldflo;
d809ec89
TT		 2113 } else if (new_xdst == NULL) {
2114 num_xfrms = 0;
2115 if (oldflo == NULL)
2116 goto make_dummy_bundle;
2117 xdst->num_xfrms = 0;
2118 dst_hold(&xdst->u.dst);
2119 return oldflo;
80c802f3
TT		 2120 }
2121
2122 /* Kill the previous bundle */
2123 if (xdst) {
2124 /* The policies were stolen for newly generated bundle */
2125 xdst->num_pols = 0;
2126 dst_free(&xdst->u.dst);
2127 }
2128
2129 /* Flow cache does not have reference, it dst_free()'s,
2130 * but we do need to return one reference for original caller */
2131 dst_hold(&new_xdst->u.dst);
2132 return &new_xdst->flo;
2133
2134make_dummy_bundle:
2135 /* We found policies, but there's no bundles to instantiate:
2136 * either because the policy blocks, has no transformations or
2137 * we could not build template (no xfrm_states).*/
b8c203b2 2138 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family);
80c802f3
TT		 2139 if (IS_ERR(xdst)) {
2140 xfrm_pols_put(pols, num_pols);
2141 return ERR_CAST(xdst);
2142 }
2143 xdst->num_pols = num_pols;
2144 xdst->num_xfrms = num_xfrms;
3e94c2dc 2145 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2146
2147 dst_hold(&xdst->u.dst);
2148 return &xdst->flo;
2149
2150inc_error:
2151 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
2152error:
2153 if (xdst != NULL)
2154 dst_free(&xdst->u.dst);
2155 else
2156 xfrm_pols_put(pols, num_pols);
2157 return ERR_PTR(err);
2158}
1da177e4 2159
2774c131
DM		 2160static struct dst_entry *make_blackhole(struct net *net, u16 family,
2161 struct dst_entry *dst_orig)
2162{
2163 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2164 struct dst_entry *ret;
2165
2166 if (!afinfo) {
2167 dst_release(dst_orig);
433a1954 2168 return ERR_PTR(-EINVAL);
2774c131
DM		 2169 } else {
2170 ret = afinfo->blackhole_route(net, dst_orig);
2171 }
2172 xfrm_policy_put_afinfo(afinfo);
2173
2174 return ret;
2175}
2176
1da177e4
LT		 2177/* Main function: finds/creates a bundle for given flow.
2178 *
2179 * At the moment we eat a raw IP route. Mostly to speed up lookups
2180 * on interfaces with disabled IPsec.
2181 */
452edd59
DM		 2182struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
2183 const struct flowi *fl,
2184 struct sock *sk, int flags)
1da177e4 2185{
4e81bb83 2186 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
80c802f3
TT		 2187 struct flow_cache_object *flo;
2188 struct xfrm_dst *xdst;
452edd59 2189 struct dst_entry *dst, *route;
80c802f3 2190 u16 family = dst_orig->ops->family;
df71837d 2191 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
4b021628 2192 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
e0d1caa7 2193
80c802f3
TT		 2194 dst = NULL;
2195 xdst = NULL;
2196 route = NULL;
4e81bb83 2197
f7944fb1 2198 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
80c802f3
TT		 2199 num_pols = 1;
2200 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
2201 err = xfrm_expand_policies(fl, family, pols,
2202 &num_pols, &num_xfrms);
2203 if (err < 0)
75b8c133 2204 goto dropdst;
80c802f3
TT		 2205
2206 if (num_pols) {
2207 if (num_xfrms <= 0) {
2208 drop_pols = num_pols;
2209 goto no_transform;
2210 }
2211
2212 xdst = xfrm_resolve_and_create_bundle(
2213 pols, num_pols, fl,
2214 family, dst_orig);
2215 if (IS_ERR(xdst)) {
2216 xfrm_pols_put(pols, num_pols);
2217 err = PTR_ERR(xdst);
2218 goto dropdst;
d809ec89
TT		 2219 } else if (xdst == NULL) {
2220 num_xfrms = 0;
2221 drop_pols = num_pols;
2222 goto no_transform;
80c802f3
TT		 2223 }
2224
b7eea454
SK		 2225 dst_hold(&xdst->u.dst);
2226 xdst->u.dst.flags |= DST_NOCACHE;
80c802f3 2227 route = xdst->route;
0aa64774 2228 }
3bccfbc7 2229 }
1da177e4 2230
80c802f3 2231 if (xdst == NULL) {
b8c203b2
SK		 2232 struct xfrm_flo xflo;
2233
2234 xflo.dst_orig = dst_orig;
2235 xflo.flags = flags;
2236
1da177e4 2237 /* To accelerate a bit... */
2518c7c2 2238 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 2239 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 2240 goto nopol;
1da177e4 2241
80c802f3 2242 flo = flow_cache_lookup(net, fl, family, dir,
b8c203b2 2243 xfrm_bundle_lookup, &xflo);
80c802f3
TT		 2244 if (flo == NULL)
2245 goto nopol;
fe1a5f03 2246 if (IS_ERR(flo)) {
80c802f3 2247 err = PTR_ERR(flo);
75b8c133 2248 goto dropdst;
d66e37a9 2249 }
80c802f3
TT		 2250 xdst = container_of(flo, struct xfrm_dst, flo);
2251
2252 num_pols = xdst->num_pols;
2253 num_xfrms = xdst->num_xfrms;
3e94c2dc 2254 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols);
80c802f3
TT		 2255 route = xdst->route;
2256 }
2257
2258 dst = &xdst->u.dst;
2259 if (route == NULL && num_xfrms > 0) {
2260 /* The only case when xfrm_bundle_lookup() returns a
2261 * bundle with null route, is when the template could
2262 * not be resolved. It means policies are there, but
2263 * bundle could not be created, since we don't yet
2264 * have the xfrm_state's. We need to wait for KM to
2265 * negotiate new SA's or bail out with error.*/
2266 if (net->xfrm.sysctl_larval_drop) {
80c802f3 2267 dst_release(dst);
a1aa3483 2268 xfrm_pols_put(pols, drop_pols);
80c802f3 2269 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2774c131 2270
f92ee619 2271 return ERR_PTR(-EREMOTE);
80c802f3 2272 }
80c802f3 2273
5b8ef341 2274 err = -EAGAIN;
80c802f3
TT		 2275
2276 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
2277 goto error;
1da177e4
LT		 2278 }
2279
80c802f3
TT		 2280no_transform:
2281 if (num_pols == 0)
8b7817f3 2282 goto nopol;
1da177e4 2283
80c802f3
TT		 2284 if ((flags & XFRM_LOOKUP_ICMP) &&
2285 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
2286 err = -ENOENT;
8b7817f3 2287 goto error;
80c802f3 2288 }
8b7817f3 2289
80c802f3
TT		 2290 for (i = 0; i < num_pols; i++)
2291 pols[i]->curlft.use_time = get_seconds();
8b7817f3 2292
80c802f3 2293 if (num_xfrms < 0) {
1da177e4 2294 /* Prohibit the flow */
59c9940e 2295 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 2296 err = -EPERM;
2297 goto error;
80c802f3
TT		 2298 } else if (num_xfrms > 0) {
2299 /* Flow transformed */
80c802f3
TT		 2300 dst_release(dst_orig);
2301 } else {
2302 /* Flow passes untransformed */
2303 dst_release(dst);
452edd59 2304 dst = dst_orig;
1da177e4 2305 }
80c802f3
TT		 2306ok:
2307 xfrm_pols_put(pols, drop_pols);
0c183379
G		 2308 if (dst && dst->xfrm &&
2309 dst->xfrm->props.mode == XFRM_MODE_TUNNEL)
2310 dst->flags |= DST_XFRM_TUNNEL;
452edd59 2311 return dst;
1da177e4 2312
80c802f3 2313nopol:
452edd59
DM		 2314 if (!(flags & XFRM_LOOKUP_ICMP)) {
2315 dst = dst_orig;
80c802f3 2316 goto ok;
452edd59 2317 }
80c802f3 2318 err = -ENOENT;
1da177e4 2319error:
80c802f3 2320 dst_release(dst);
75b8c133
HX		 2321dropdst:
2322 dst_release(dst_orig);
80c802f3 2323 xfrm_pols_put(pols, drop_pols);
452edd59 2324 return ERR_PTR(err);
1da177e4
LT		 2325}
2326EXPORT_SYMBOL(xfrm_lookup);
2327
f92ee619
SK		 2328/* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2329 * Otherwise we may send out blackholed packets.
2330 */
2331struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
2332 const struct flowi *fl,
2333 struct sock *sk, int flags)
2334{
b8c203b2
SK		 2335 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk,
2336 flags | XFRM_LOOKUP_QUEUE);
f92ee619
SK		 2337
2338 if (IS_ERR(dst) && PTR_ERR(dst) == -EREMOTE)
2339 return make_blackhole(net, dst_orig->ops->family, dst_orig);
2340
2341 return dst;
2342}
2343EXPORT_SYMBOL(xfrm_lookup_route);
2344
df0ba92a 2345static inline int
8f029de2 2346xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
df0ba92a
MN		 2347{
2348 struct xfrm_state *x;
df0ba92a
MN		 2349
2350 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
2351 return 0;
2352 x = skb->sp->xvec[idx];
2353 if (!x->type->reject)
2354 return 0;
1ecafede 2355 return x->type->reject(x, skb, fl);
df0ba92a
MN		 2356}
2357
1da177e4
LT		 2358/* When skb is transformed back to its "native" form, we have to
2359 * check policy restrictions. At the moment we make this in maximally
2360 * stupid way. Shame on me. :-) Of course, connected sockets must
2361 * have policy cached at them.
2362 */
2363
2364static inline int
7db454b9 2365xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1da177e4
LT		 2366 unsigned short family)
2367{
2368 if (xfrm_state_kern(x))
928ba416 2369 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 2370 return x->id.proto == tmpl->id.proto &&
2371 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
2372 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
2373 x->props.mode == tmpl->mode &&
c5d18e98 2374 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 2375 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 2376 !(x->props.mode != XFRM_MODE_TRANSPORT &&
2377 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 2378}
2379
df0ba92a
MN		 2380/*
2381 * 0 or more than 0 is returned when validation is succeeded (either bypass
2382 * because of optional transport mode, or next index of the mathced secpath
2383 * state with the template.
2384 * -1 is returned when no matching template is found.
2385 * Otherwise "-2 - errored_index" is returned.
2386 */
1da177e4 2387static inline int
22cccb7e 2388xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1da177e4
LT		 2389 unsigned short family)
2390{
2391 int idx = start;
2392
2393 if (tmpl->optional) {
7e49e6de 2394 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 2395 return start;
2396 } else
2397 start = -1;
2398 for (; idx < sp->len; idx++) {
dbe5b4aa 2399 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 2400 return ++idx;
df0ba92a
MN		 2401 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2402 if (start == -1)
2403 start = -2-idx;
1da177e4 2404 break;
df0ba92a 2405 }
1da177e4
LT		 2406 }
2407 return start;
2408}
2409
d5422efe
HX		 2410int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2411 unsigned int family, int reverse)
1da177e4
LT		 2412{
2413 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 2414 int err;
1da177e4
LT		 2415
2416 if (unlikely(afinfo == NULL))
2417 return -EAFNOSUPPORT;
2418
d5422efe 2419 afinfo->decode_session(skb, fl, reverse);
1d28f42c 2420 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
1da177e4 2421 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 2422 return err;
1da177e4 2423}
d5422efe 2424EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 2425
9a7386ec 2426static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 2427{
2428 for (; k < sp->len; k++) {
df0ba92a 2429 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 2430 *idxp = k;
1da177e4 2431 return 1;
df0ba92a 2432 }
1da177e4
LT		 2433 }
2434
2435 return 0;
2436}
2437
a716c119 2438int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 2439 unsigned short family)
2440{
f6e1e25d 2441 struct net *net = dev_net(skb->dev);
1da177e4 2442 struct xfrm_policy *pol;
4e81bb83
MN		 2443 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2444 int npols = 0;
2445 int xfrm_nr;
2446 int pi;
d5422efe 2447 int reverse;
1da177e4 2448 struct flowi fl;
d5422efe 2449 u8 fl_dir;
df0ba92a 2450 int xerr_idx = -1;
1da177e4 2451
d5422efe
HX		 2452 reverse = dir & ~XFRM_POLICY_MASK;
2453 dir &= XFRM_POLICY_MASK;
2454 fl_dir = policy_to_flow_dir(dir);
2455
0aa64774 2456 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 2457 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2458 return 0;
0aa64774
MN		 2459 }
2460
eb9c7ebe 2461 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 2462
2463 /* First, check used SA against their selectors. */
2464 if (skb->sp) {
2465 int i;
2466
9b7a787d 2467 for (i = skb->sp->len-1; i >= 0; i--) {
dbe5b4aa 2468 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 2469 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 2470 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 2471 return 0;
0aa64774 2472 }
1da177e4
LT		 2473 }
2474 }
2475
2476 pol = NULL;
3bccfbc7 2477 if (sk && sk->sk_policy[dir]) {
e0d1caa7 2478 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 2479 if (IS_ERR(pol)) {
59c9940e 2480 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 2481 return 0;
0aa64774 2482 }
3bccfbc7 2483 }
1da177e4 2484
fe1a5f03
TT		 2485 if (!pol) {
2486 struct flow_cache_object *flo;
2487
2488 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2489 xfrm_policy_lookup, NULL);
2490 if (IS_ERR_OR_NULL(flo))
2491 pol = ERR_CAST(flo);
2492 else
2493 pol = container_of(flo, struct xfrm_policy, flo);
2494 }
1da177e4 2495
0aa64774 2496 if (IS_ERR(pol)) {
59c9940e 2497 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2498 return 0;
0aa64774 2499 }
134b0fc5 2500
df0ba92a 2501 if (!pol) {
d1d9facf 2502 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 2503 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 2504 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 2505 return 0;
2506 }
2507 return 1;
2508 }
1da177e4 2509
9d729f72 2510 pol->curlft.use_time = get_seconds();
1da177e4 2511
4e81bb83 2512 pols[0] = pol;
02d0892f 2513 npols++;
4e81bb83
MN		 2514#ifdef CONFIG_XFRM_SUB_POLICY
2515 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 2516 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 2517 &fl, family,
2518 XFRM_POLICY_IN);
2519 if (pols[1]) {
0aa64774 2520 if (IS_ERR(pols[1])) {
59c9940e 2521 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 2522 return 0;
0aa64774 2523 }
9d729f72 2524 pols[1]->curlft.use_time = get_seconds();
02d0892f 2525 npols++;
4e81bb83
MN		 2526 }
2527 }
2528#endif
2529
1da177e4
LT		 2530 if (pol->action == XFRM_POLICY_ALLOW) {
2531 struct sec_path *sp;
2532 static struct sec_path dummy;
4e81bb83 2533 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 2534 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 2535 struct xfrm_tmpl **tpp = tp;
2536 int ti = 0;
1da177e4
LT		 2537 int i, k;
2538
2539 if ((sp = skb->sp) == NULL)
2540 sp = &dummy;
2541
4e81bb83
MN		 2542 for (pi = 0; pi < npols; pi++) {
2543 if (pols[pi] != pol &&
0aa64774 2544 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 2545 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 2546 goto reject;
0aa64774
MN		 2547 }
2548 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 2549 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2550 goto reject_error;
0aa64774 2551 }
4e81bb83
MN		 2552 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2553 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2554 }
2555 xfrm_nr = ti;
41a49cc3 2556 if (npols > 1) {
283bc9f3 2557 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family, net);
41a49cc3
MN		 2558 tpp = stp;
2559 }
4e81bb83 2560
1da177e4
LT		 2561 /* For each tunnel xfrm, find the first matching tmpl.
2562 * For each tmpl before that, find corresponding xfrm.
2563 * Order is _important_. Later we will implement
2564 * some barriers, but at the moment barriers
2565 * are implied between each two transformations.
2566 */
4e81bb83
MN		 2567 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2568 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2569 if (k < 0) {
d1d9facf
JM		 2570 if (k < -1)
2571 /* "-2 - errored_index" returned */
2572 xerr_idx = -(2+k);
59c9940e 2573 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2574 goto reject;
df0ba92a 2575 }
1da177e4
LT		 2576 }
2577
0aa64774 2578 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2579 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2580 goto reject;
0aa64774 2581 }
1da177e4 2582
4e81bb83 2583 xfrm_pols_put(pols, npols);
1da177e4
LT		 2584 return 1;
2585 }
59c9940e 2586 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2587
2588reject:
df0ba92a 2589 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2590reject_error:
2591 xfrm_pols_put(pols, npols);
1da177e4
LT		 2592 return 0;
2593}
2594EXPORT_SYMBOL(__xfrm_policy_check);
2595
2596int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2597{
99a66657 2598 struct net *net = dev_net(skb->dev);
1da177e4 2599 struct flowi fl;
adf30907 2600 struct dst_entry *dst;
73137147 2601 int res = 1;
1da177e4 2602
0aa64774 2603 if (xfrm_decode_session(skb, &fl, family) < 0) {
72032fdb 2604 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
1da177e4 2605 return 0;
0aa64774 2606 }
1da177e4 2607
fafeeb6c 2608 skb_dst_force(skb);
adf30907 2609
b8c203b2 2610 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE);
452edd59 2611 if (IS_ERR(dst)) {
73137147 2612 res = 0;
452edd59
DM		 2613 dst = NULL;
2614 }
adf30907
ED		 2615 skb_dst_set(skb, dst);
2616 return res;
1da177e4
LT		 2617}
2618EXPORT_SYMBOL(__xfrm_route_forward);
2619
d49c73c7
DM		 2620/* Optimize later using cookies and generation ids. */
2621
1da177e4
LT		 2622static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2623{
d49c73c7 2624 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
f5b0a874
DM		 2625 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2626 * get validated by dst_ops->check on every use. We do this
2627 * because when a normal route referenced by an XFRM dst is
2628 * obsoleted we do not go looking around for all parent
2629 * referencing XFRM dsts so that we can invalidate them. It
2630 * is just too much work. Instead we make the checks here on
2631 * every use. For example:
d49c73c7
DM		 2632 *
2633 * XFRM dst A --> IPv4 dst X
2634 *
2635 * X is the "xdst->route" of A (X is also the "dst->path" of A
2636 * in this example). If X is marked obsolete, "A" will not
2637 * notice. That's what we are validating here via the
2638 * stale_bundle() check.
2639 *
2640 * When a policy's bundle is pruned, we dst_free() the XFRM
f5b0a874
DM		 2641 * dst which causes it's ->obsolete field to be set to
2642 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2643 * this, we want to force a new route lookup.
399c180a 2644 */
d49c73c7
DM		 2645 if (dst->obsolete < 0 && !stale_bundle(dst))
2646 return dst;
2647
1da177e4
LT		 2648 return NULL;
2649}
2650
2651static int stale_bundle(struct dst_entry *dst)
2652{
12fdb4d3 2653 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
1da177e4
LT		 2654}
2655
aabc9761 2656void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2657{
1da177e4 2658 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2659 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2660 dev_hold(dst->dev);
1da177e4
LT		 2661 dev_put(dev);
2662 }
2663}
aabc9761 2664EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2665
2666static void xfrm_link_failure(struct sk_buff *skb)
2667{
2668 /* Impossible. Such dst must be popped before reaches point of failure. */
1da177e4
LT		 2669}
2670
2671static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2672{
2673 if (dst) {
2674 if (dst->obsolete) {
2675 dst_release(dst);
2676 dst = NULL;
2677 }
2678 }
2679 return dst;
2680}
2681
e4c17216 2682void xfrm_garbage_collect(struct net *net)
c0ed1c14 2683{
ca925cf1 2684 flow_cache_flush(net);
c0ed1c14 2685}
e4c17216 2686EXPORT_SYMBOL(xfrm_garbage_collect);
c0ed1c14
SK		 2687
2688static void xfrm_garbage_collect_deferred(struct net *net)
2689{
ca925cf1 2690 flow_cache_flush_deferred(net);
c0ed1c14
SK		 2691}
2692
25ee3286 2693static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2694{
2695 do {
2696 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2697 u32 pmtu, route_mtu_cached;
2698
2699 pmtu = dst_mtu(dst->child);
2700 xdst->child_mtu_cached = pmtu;
2701
2702 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2703
2704 route_mtu_cached = dst_mtu(xdst->route);
2705 xdst->route_mtu_cached = route_mtu_cached;
2706
2707 if (pmtu > route_mtu_cached)
2708 pmtu = route_mtu_cached;
2709
defb3519 2710 dst_metric_set(dst, RTAX_MTU, pmtu);
1da177e4
LT		 2711 } while ((dst = dst->next));
2712}
2713
1da177e4
LT		 2714/* Check that the bundle accepts the flow and its components are
2715 * still valid.
2716 */
2717
12fdb4d3 2718static int xfrm_bundle_ok(struct xfrm_dst *first)
1da177e4
LT		 2719{
2720 struct dst_entry *dst = &first->u.dst;
2721 struct xfrm_dst *last;
2722 u32 mtu;
2723
92d63dec 2724 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2725 (dst->dev && !netif_running(dst->dev)))
2726 return 0;
2727
a0073fe1
SK		 2728 if (dst->flags & DST_XFRM_QUEUE)
2729 return 1;
2730
1da177e4
LT		 2731 last = NULL;
2732
2733 do {
2734 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2735
1da177e4
LT		 2736 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2737 return 0;
80c802f3
TT		 2738 if (xdst->xfrm_genid != dst->xfrm->genid)
2739 return 0;
b1312c89
TT		 2740 if (xdst->num_pols > 0 &&
2741 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
9d4a706d 2742 return 0;
e53820de 2743
1da177e4
LT		 2744 mtu = dst_mtu(dst->child);
2745 if (xdst->child_mtu_cached != mtu) {
2746 last = xdst;
2747 xdst->child_mtu_cached = mtu;
2748 }
2749
92d63dec 2750 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2751 return 0;
2752 mtu = dst_mtu(xdst->route);
2753 if (xdst->route_mtu_cached != mtu) {
2754 last = xdst;
2755 xdst->route_mtu_cached = mtu;
2756 }
2757
2758 dst = dst->child;
2759 } while (dst->xfrm);
2760
2761 if (likely(!last))
2762 return 1;
2763
2764 mtu = last->child_mtu_cached;
2765 for (;;) {
2766 dst = &last->u.dst;
2767
2768 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2769 if (mtu > last->route_mtu_cached)
2770 mtu = last->route_mtu_cached;
defb3519 2771 dst_metric_set(dst, RTAX_MTU, mtu);
1da177e4
LT		 2772
2773 if (last == first)
2774 break;
2775
bd0bf076 2776 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2777 last->child_mtu_cached = mtu;
2778 }
2779
2780 return 1;
2781}
2782
0dbaee3b
DM		 2783static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2784{
2785 return dst_metric_advmss(dst->path);
2786}
2787
ebb762f2 2788static unsigned int xfrm_mtu(const struct dst_entry *dst)
d33e4553 2789{
618f9bc7
SK		 2790 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2791
2792 return mtu ? : dst_mtu(dst->path);
d33e4553
DM		 2793}
2794
f894cbf8
DM		 2795static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst,
2796 struct sk_buff *skb,
2797 const void *daddr)
d3aaeb38 2798{
f894cbf8 2799 return dst->path->ops->neigh_lookup(dst, skb, daddr);
d3aaeb38
DM		 2800}
2801
1da177e4
LT		 2802int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2803{
d7c7544c 2804 struct net *net;
1da177e4
LT		 2805 int err = 0;
2806 if (unlikely(afinfo == NULL))
2807 return -EINVAL;
2808 if (unlikely(afinfo->family >= NPROTO))
2809 return -EAFNOSUPPORT;
ef8531b6 2810 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2811 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2812 err = -ENOBUFS;
2813 else {
2814 struct dst_ops *dst_ops = afinfo->dst_ops;
2815 if (likely(dst_ops->kmem_cachep == NULL))
2816 dst_ops->kmem_cachep = xfrm_dst_cache;
2817 if (likely(dst_ops->check == NULL))
2818 dst_ops->check = xfrm_dst_check;
0dbaee3b
DM		 2819 if (likely(dst_ops->default_advmss == NULL))
2820 dst_ops->default_advmss = xfrm_default_advmss;
ebb762f2
SK		 2821 if (likely(dst_ops->mtu == NULL))
2822 dst_ops->mtu = xfrm_mtu;
1da177e4
LT		 2823 if (likely(dst_ops->negative_advice == NULL))
2824 dst_ops->negative_advice = xfrm_negative_advice;
2825 if (likely(dst_ops->link_failure == NULL))
2826 dst_ops->link_failure = xfrm_link_failure;
d3aaeb38
DM		 2827 if (likely(dst_ops->neigh_lookup == NULL))
2828 dst_ops->neigh_lookup = xfrm_neigh_lookup;
1da177e4 2829 if (likely(afinfo->garbage_collect == NULL))
c0ed1c14 2830 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
418a99ac 2831 rcu_assign_pointer(xfrm_policy_afinfo[afinfo->family], afinfo);
1da177e4 2832 }
ef8531b6 2833 spin_unlock(&xfrm_policy_afinfo_lock);
d7c7544c
AD		 2834
2835 rtnl_lock();
2836 for_each_net(net) {
2837 struct dst_ops *xfrm_dst_ops;
2838
2839 switch (afinfo->family) {
2840 case AF_INET:
2841 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2842 break;
dfd56b8b 2843#if IS_ENABLED(CONFIG_IPV6)
d7c7544c
AD		 2844 case AF_INET6:
2845 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2846 break;
2847#endif
2848 default:
2849 BUG();
2850 }
2851 *xfrm_dst_ops = *afinfo->dst_ops;
2852 }
2853 rtnl_unlock();
2854
1da177e4
LT		 2855 return err;
2856}
2857EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2858
2859int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2860{
2861 int err = 0;
2862 if (unlikely(afinfo == NULL))
2863 return -EINVAL;
2864 if (unlikely(afinfo->family >= NPROTO))
2865 return -EAFNOSUPPORT;
ef8531b6 2866 spin_lock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2867 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2868 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2869 err = -EINVAL;
ef8531b6
ED		 2870 else
2871 RCU_INIT_POINTER(xfrm_policy_afinfo[afinfo->family],
2872 NULL);
2873 }
2874 spin_unlock(&xfrm_policy_afinfo_lock);
2875 if (!err) {
2876 struct dst_ops *dst_ops = afinfo->dst_ops;
2877
2878 synchronize_rcu();
2879
2880 dst_ops->kmem_cachep = NULL;
2881 dst_ops->check = NULL;
2882 dst_ops->negative_advice = NULL;
2883 dst_ops->link_failure = NULL;
2884 afinfo->garbage_collect = NULL;
1da177e4 2885 }
1da177e4
LT		 2886 return err;
2887}
2888EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2889
d7c7544c
AD		 2890static void __net_init xfrm_dst_ops_init(struct net *net)
2891{
2892 struct xfrm_policy_afinfo *afinfo;
2893
ef8531b6
ED		 2894 rcu_read_lock();
2895 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET]);
d7c7544c
AD		 2896 if (afinfo)
2897 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
dfd56b8b 2898#if IS_ENABLED(CONFIG_IPV6)
ef8531b6 2899 afinfo = rcu_dereference(xfrm_policy_afinfo[AF_INET6]);
d7c7544c
AD		 2900 if (afinfo)
2901 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2902#endif
418a99ac 2903 rcu_read_unlock();
546be240
HX		 2904}
2905
1da177e4
LT		 2906static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2907{
351638e7 2908 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
e9dc8653 2909
1da177e4
LT		 2910 switch (event) {
2911 case NETDEV_DOWN:
c0ed1c14 2912 xfrm_garbage_collect(dev_net(dev));
1da177e4
LT		 2913 }
2914 return NOTIFY_DONE;
2915}
2916
2917static struct notifier_block xfrm_dev_notifier = {
d5917a35 2918 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2919};
2920
558f82ef 2921#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2922static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2923{
c68cd1a0 2924 int rv;
698365fa
WC		 2925 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib);
2926 if (!net->mib.xfrm_statistics)
558f82ef 2927 return -ENOMEM;
c68cd1a0
AD		 2928 rv = xfrm_proc_init(net);
2929 if (rv < 0)
698365fa 2930 free_percpu(net->mib.xfrm_statistics);
c68cd1a0 2931 return rv;
558f82ef 2932}
59c9940e
AD		 2933
2934static void xfrm_statistics_fini(struct net *net)
2935{
c68cd1a0 2936 xfrm_proc_fini(net);
698365fa 2937 free_percpu(net->mib.xfrm_statistics);
59c9940e
AD		 2938}
2939#else
2940static int __net_init xfrm_statistics_init(struct net *net)
2941{
2942 return 0;
2943}
2944
2945static void xfrm_statistics_fini(struct net *net)
2946{
2947}
558f82ef
MN		 2948#endif
2949
d62ddc21 2950static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2951{
2518c7c2
DM		 2952 unsigned int hmask, sz;
2953 int dir;
2954
d62ddc21
AD		 2955 if (net_eq(net, &init_net))
2956 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2957 sizeof(struct xfrm_dst),
e5d679f3 2958 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2959 NULL);
1da177e4 2960
2518c7c2
DM		 2961 hmask = 8 - 1;
2962 sz = (hmask+1) * sizeof(struct hlist_head);
2963
93b851c1
AD		 2964 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2965 if (!net->xfrm.policy_byidx)
2966 goto out_byidx;
8100bea7 2967 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2968
2969 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2970 struct xfrm_policy_hash *htab;
2971
dc2caba7 2972 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2973 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2974
a35f6c5d 2975 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2976 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2977 if (!htab->table)
a35f6c5d
AD		 2978 goto out_bydst;
2979 htab->hmask = hmask;
b58555f1
CG		 2980 htab->dbits4 = 32;
2981 htab->sbits4 = 32;
2982 htab->dbits6 = 128;
2983 htab->sbits6 = 128;
2518c7c2 2984 }
880a6fab
CG		 2985 net->xfrm.policy_hthresh.lbits4 = 32;
2986 net->xfrm.policy_hthresh.rbits4 = 32;
2987 net->xfrm.policy_hthresh.lbits6 = 128;
2988 net->xfrm.policy_hthresh.rbits6 = 128;
2989
2990 seqlock_init(&net->xfrm.policy_hthresh.lock);
2518c7c2 2991
adfcf0b2 2992 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2993 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
880a6fab 2994 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
d62ddc21
AD		 2995 if (net_eq(net, &init_net))
2996 register_netdevice_notifier(&xfrm_dev_notifier);
2997 return 0;
93b851c1 2998
a35f6c5d
AD		 2999out_bydst:
3000 for (dir--; dir >= 0; dir--) {
3001 struct xfrm_policy_hash *htab;
3002
3003 htab = &net->xfrm.policy_bydst[dir];
3004 xfrm_hash_free(htab->table, sz);
3005 }
3006 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 3007out_byidx:
3008 return -ENOMEM;
d62ddc21
AD		 3009}
3010
3011static void xfrm_policy_fini(struct net *net)
3012{
93b851c1 3013 unsigned int sz;
8b18f8ea 3014 int dir;
93b851c1 3015
7c2776ee
AD		 3016 flush_work(&net->xfrm.policy_hash_work);
3017#ifdef CONFIG_XFRM_SUB_POLICY
2e71029e 3018 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false);
7c2776ee 3019#endif
2e71029e 3020 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false);
7c2776ee 3021
adfcf0b2 3022 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 3023
8b18f8ea 3024 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 3025 struct xfrm_policy_hash *htab;
3026
8b18f8ea 3027 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 3028
3029 htab = &net->xfrm.policy_bydst[dir];
5b653b2a 3030 sz = (htab->hmask + 1) * sizeof(struct hlist_head);
a35f6c5d
AD		 3031 WARN_ON(!hlist_empty(htab->table));
3032 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 3033 }
3034
8100bea7 3035 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 3036 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
3037 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 3038}
3039
d62ddc21
AD		 3040static int __net_init xfrm_net_init(struct net *net)
3041{
3042 int rv;
3043
59c9940e
AD		 3044 rv = xfrm_statistics_init(net);
3045 if (rv < 0)
3046 goto out_statistics;
d62ddc21
AD		 3047 rv = xfrm_state_init(net);
3048 if (rv < 0)
3049 goto out_state;
3050 rv = xfrm_policy_init(net);
3051 if (rv < 0)
3052 goto out_policy;
d7c7544c 3053 xfrm_dst_ops_init(net);
b27aeadb
AD		 3054 rv = xfrm_sysctl_init(net);
3055 if (rv < 0)
3056 goto out_sysctl;
4a93f509
SK		 3057 rv = flow_cache_init(net);
3058 if (rv < 0)
3059 goto out;
283bc9f3
FD		 3060
3061 /* Initialize the per-net locks here */
3062 spin_lock_init(&net->xfrm.xfrm_state_lock);
3063 rwlock_init(&net->xfrm.xfrm_policy_lock);
283bc9f3
FD		 3064 mutex_init(&net->xfrm.xfrm_cfg_mutex);
3065
d62ddc21
AD		 3066 return 0;
3067
4a93f509
SK		 3068out:
3069 xfrm_sysctl_fini(net);
b27aeadb
AD		 3070out_sysctl:
3071 xfrm_policy_fini(net);
d62ddc21
AD		 3072out_policy:
3073 xfrm_state_fini(net);
3074out_state:
59c9940e
AD		 3075 xfrm_statistics_fini(net);
3076out_statistics:
d62ddc21
AD		 3077 return rv;
3078}
3079
3080static void __net_exit xfrm_net_exit(struct net *net)
3081{
4a93f509 3082 flow_cache_fini(net);
b27aeadb 3083 xfrm_sysctl_fini(net);
d62ddc21
AD		 3084 xfrm_policy_fini(net);
3085 xfrm_state_fini(net);
59c9940e 3086 xfrm_statistics_fini(net);
d62ddc21
AD		 3087}
3088
3089static struct pernet_operations __net_initdata xfrm_net_ops = {
3090 .init = xfrm_net_init,
3091 .exit = xfrm_net_exit,
3092};
3093
1da177e4
LT		 3094void __init xfrm_init(void)
3095{
d62ddc21 3096 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 3097 xfrm_input_init();
3098}
3099
ab5f5e8b 3100#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 3101static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
3102 struct audit_buffer *audit_buf)
ab5f5e8b 3103{
875179fa
PM		 3104 struct xfrm_sec_ctx *ctx = xp->security;
3105 struct xfrm_selector *sel = &xp->selector;
3106
3107 if (ctx)
ab5f5e8b 3108 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 3109 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 3110
9b7a787d 3111 switch (sel->family) {
ab5f5e8b 3112 case AF_INET:
21454aaa 3113 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 3114 if (sel->prefixlen_s != 32)
3115 audit_log_format(audit_buf, " src_prefixlen=%d",
3116 sel->prefixlen_s);
21454aaa 3117 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 3118 if (sel->prefixlen_d != 32)
3119 audit_log_format(audit_buf, " dst_prefixlen=%d",
3120 sel->prefixlen_d);
ab5f5e8b
JL		 3121 break;
3122 case AF_INET6:
5b095d98 3123 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 3124 if (sel->prefixlen_s != 128)
3125 audit_log_format(audit_buf, " src_prefixlen=%d",
3126 sel->prefixlen_s);
5b095d98 3127 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 3128 if (sel->prefixlen_d != 128)
3129 audit_log_format(audit_buf, " dst_prefixlen=%d",
3130 sel->prefixlen_d);
ab5f5e8b
JL		 3131 break;
3132 }
3133}
3134
2e71029e 3135void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid)
ab5f5e8b
JL		 3136{
3137 struct audit_buffer *audit_buf;
ab5f5e8b 3138
afeb14b4 3139 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 3140 if (audit_buf == NULL)
3141 return;
2e71029e 3142 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3143 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3144 xfrm_audit_common_policyinfo(xp, audit_buf);
3145 audit_log_end(audit_buf);
3146}
3147EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
3148
68277acc 3149void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2e71029e 3150 bool task_valid)
ab5f5e8b
JL		 3151{
3152 struct audit_buffer *audit_buf;
ab5f5e8b 3153
afeb14b4 3154 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 3155 if (audit_buf == NULL)
3156 return;
2e71029e 3157 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
afeb14b4 3158 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 3159 xfrm_audit_common_policyinfo(xp, audit_buf);
3160 audit_log_end(audit_buf);
3161}
3162EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
3163#endif
3164
80c9abaa 3165#ifdef CONFIG_XFRM_MIGRATE
bc9b35ad
DM		 3166static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
3167 const struct xfrm_selector *sel_tgt)
80c9abaa
SS		 3168{
3169 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
3170 if (sel_tgt->family == sel_cmp->family &&
70e94e66
YH		 3171 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr,
3172 sel_cmp->family) &&
3173 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr,
3174 sel_cmp->family) &&
80c9abaa
SS		 3175 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
3176 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
bc9b35ad 3177 return true;
80c9abaa
SS		 3178 }
3179 } else {
3180 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
bc9b35ad 3181 return true;
80c9abaa
SS		 3182 }
3183 }
bc9b35ad 3184 return false;
80c9abaa
SS		 3185}
3186
3e94c2dc
WC		 3187static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel,
3188 u8 dir, u8 type, struct net *net)
80c9abaa
SS		 3189{
3190 struct xfrm_policy *pol, *ret = NULL;
80c9abaa
SS		 3191 struct hlist_head *chain;
3192 u32 priority = ~0U;
3193
283bc9f3 3194 read_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME*/
8d549c4f 3195 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir);
b67bfe0d 3196 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3197 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3198 pol->type == type) {
3199 ret = pol;
3200 priority = ret->priority;
3201 break;
3202 }
3203 }
8d549c4f 3204 chain = &net->xfrm.policy_inexact[dir];
b67bfe0d 3205 hlist_for_each_entry(pol, chain, bydst) {
80c9abaa
SS		 3206 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
3207 pol->type == type &&
3208 pol->priority < priority) {
3209 ret = pol;
3210 break;
3211 }
3212 }
3213
3214 if (ret)
3215 xfrm_pol_hold(ret);
3216
283bc9f3 3217 read_unlock_bh(&net->xfrm.xfrm_policy_lock);
80c9abaa
SS		 3218
3219 return ret;
3220}
3221
dd701754 3222static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
80c9abaa
SS		 3223{
3224 int match = 0;
3225
3226 if (t->mode == m->mode && t->id.proto == m->proto &&
3227 (m->reqid == 0 || t->reqid == m->reqid)) {
3228 switch (t->mode) {
3229 case XFRM_MODE_TUNNEL:
3230 case XFRM_MODE_BEET:
70e94e66
YH		 3231 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr,
3232 m->old_family) &&
3233 xfrm_addr_equal(&t->saddr, &m->old_saddr,
3234 m->old_family)) {
80c9abaa
SS		 3235 match = 1;
3236 }
3237 break;
3238 case XFRM_MODE_TRANSPORT:
3239 /* in case of transport mode, template does not store
3240 any IP addresses, hence we just compare mode and
3241 protocol */
3242 match = 1;
3243 break;
3244 default:
3245 break;
3246 }
3247 }
3248 return match;
3249}
3250
3251/* update endpoint address(es) of template(s) */
3252static int xfrm_policy_migrate(struct xfrm_policy *pol,
3253 struct xfrm_migrate *m, int num_migrate)
3254{
3255 struct xfrm_migrate *mp;
80c9abaa
SS		 3256 int i, j, n = 0;
3257
3258 write_lock_bh(&pol->lock);
12a169e7 3259 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 3260 /* target policy has been deleted */
3261 write_unlock_bh(&pol->lock);
3262 return -ENOENT;
3263 }
3264
3265 for (i = 0; i < pol->xfrm_nr; i++) {
3266 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
3267 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
3268 continue;
3269 n++;
1bfcb10f
HX		 3270 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
3271 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 3272 continue;
3273 /* update endpoints */
3274 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
3275 sizeof(pol->xfrm_vec[i].id.daddr));
3276 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
3277 sizeof(pol->xfrm_vec[i].saddr));
3278 pol->xfrm_vec[i].encap_family = mp->new_family;
3279 /* flush bundles */
80c802f3 3280 atomic_inc(&pol->genid);
80c9abaa
SS		 3281 }
3282 }
3283
3284 write_unlock_bh(&pol->lock);
3285
3286 if (!n)
3287 return -ENODATA;
3288
3289 return 0;
3290}
3291
dd701754 3292static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
80c9abaa
SS		 3293{
3294 int i, j;
3295
3296 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
3297 return -EINVAL;
3298
3299 for (i = 0; i < num_migrate; i++) {
70e94e66
YH		 3300 if (xfrm_addr_equal(&m[i].old_daddr, &m[i].new_daddr,
3301 m[i].old_family) &&
3302 xfrm_addr_equal(&m[i].old_saddr, &m[i].new_saddr,
3303 m[i].old_family))
80c9abaa
SS		 3304 return -EINVAL;
3305 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
3306 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
3307 return -EINVAL;
3308
3309 /* check if there is any duplicated entry */
3310 for (j = i + 1; j < num_migrate; j++) {
3311 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
3312 sizeof(m[i].old_daddr)) &&
3313 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
3314 sizeof(m[i].old_saddr)) &&
3315 m[i].proto == m[j].proto &&
3316 m[i].mode == m[j].mode &&
3317 m[i].reqid == m[j].reqid &&
3318 m[i].old_family == m[j].old_family)
3319 return -EINVAL;
3320 }
3321 }
3322
3323 return 0;
3324}
3325
b4b7c0b3 3326int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189 3327 struct xfrm_migrate *m, int num_migrate,
8d549c4f 3328 struct xfrm_kmaddress *k, struct net *net)
80c9abaa
SS		 3329{
3330 int i, err, nx_cur = 0, nx_new = 0;
3331 struct xfrm_policy *pol = NULL;
3332 struct xfrm_state *x, *xc;
3333 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
3334 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
3335 struct xfrm_migrate *mp;
3336
3337 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
3338 goto out;
3339
3340 /* Stage 1 - find policy */
8d549c4f 3341 if ((pol = xfrm_migrate_policy_find(sel, dir, type, net)) == NULL) {
80c9abaa
SS		 3342 err = -ENOENT;
3343 goto out;
3344 }
3345
3346 /* Stage 2 - find and update state(s) */
3347 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
283bc9f3 3348 if ((x = xfrm_migrate_state_find(mp, net))) {
80c9abaa
SS		 3349 x_cur[nx_cur] = x;
3350 nx_cur++;
3351 if ((xc = xfrm_state_migrate(x, mp))) {
3352 x_new[nx_new] = xc;
3353 nx_new++;
3354 } else {
3355 err = -ENODATA;
3356 goto restore_state;
3357 }
3358 }
3359 }
3360
3361 /* Stage 3 - update policy */
3362 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
3363 goto restore_state;
3364
3365 /* Stage 4 - delete old state(s) */
3366 if (nx_cur) {
3367 xfrm_states_put(x_cur, nx_cur);
3368 xfrm_states_delete(x_cur, nx_cur);
3369 }
3370
3371 /* Stage 5 - announce */
13c1d189 3372 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 3373
3374 xfrm_pol_put(pol);
3375
3376 return 0;
3377out:
3378 return err;
3379
3380restore_state:
3381 if (pol)
3382 xfrm_pol_put(pol);
3383 if (nx_cur)
3384 xfrm_states_put(x_cur, nx_cur);
3385 if (nx_new)
3386 xfrm_states_delete(x_new, nx_new);
3387
3388 return err;
3389}
e610e679 3390EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 3391#endif
Linux 6.x block layer and io_uring tree(s)