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