projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
netns xfrm: propagate netns into policy byidx hash
[deliverable/linux.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT		 2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
68277acc 27#include <linux/audit.h>
25ee3286 28#include <net/dst.h>
1da177e4
LT		 29#include <net/xfrm.h>
30#include <net/ip.h>
558f82ef
MN		 31#ifdef CONFIG_XFRM_STATISTICS
32#include <net/snmp.h>
33#endif
1da177e4 34
44e36b42
DM		 35#include "xfrm_hash.h"
36
28faa979 37int sysctl_xfrm_larval_drop __read_mostly = 1;
14e50e57 38
558f82ef
MN		 39#ifdef CONFIG_XFRM_STATISTICS
40DEFINE_SNMP_STAT(struct linux_xfrm_mib, xfrm_statistics) __read_mostly;
41EXPORT_SYMBOL(xfrm_statistics);
42#endif
43
4a3e2f71
AV		 44DEFINE_MUTEX(xfrm_cfg_mutex);
45EXPORT_SYMBOL(xfrm_cfg_mutex);
1da177e4
LT		 46
47static DEFINE_RWLOCK(xfrm_policy_lock);
48
1da177e4
LT		 49static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
50static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
51
e18b890b 52static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 53
2518c7c2 54static HLIST_HEAD(xfrm_policy_gc_list);
1da177e4
LT		 55static DEFINE_SPINLOCK(xfrm_policy_gc_lock);
56
57static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
58static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
25ee3286 59static void xfrm_init_pmtu(struct dst_entry *dst);
1da177e4 60
77681021
AM		 61static inline int
62__xfrm4_selector_match(struct xfrm_selector *sel, struct flowi *fl)
63{
64 return addr_match(&fl->fl4_dst, &sel->daddr, sel->prefixlen_d) &&
65 addr_match(&fl->fl4_src, &sel->saddr, sel->prefixlen_s) &&
66 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
67 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
68 (fl->proto == sel->proto || !sel->proto) &&
69 (fl->oif == sel->ifindex || !sel->ifindex);
70}
71
72static inline int
73__xfrm6_selector_match(struct xfrm_selector *sel, struct flowi *fl)
74{
75 return addr_match(&fl->fl6_dst, &sel->daddr, sel->prefixlen_d) &&
76 addr_match(&fl->fl6_src, &sel->saddr, sel->prefixlen_s) &&
77 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
78 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
79 (fl->proto == sel->proto || !sel->proto) &&
80 (fl->oif == sel->ifindex || !sel->ifindex);
81}
82
83int xfrm_selector_match(struct xfrm_selector *sel, struct flowi *fl,
84 unsigned short family)
85{
86 switch (family) {
87 case AF_INET:
88 return __xfrm4_selector_match(sel, fl);
89 case AF_INET6:
90 return __xfrm6_selector_match(sel, fl);
91 }
92 return 0;
93}
94
9bb182a7
YH		 95static inline struct dst_entry *__xfrm_dst_lookup(int tos,
96 xfrm_address_t *saddr,
97 xfrm_address_t *daddr,
98 int family)
99{
100 struct xfrm_policy_afinfo *afinfo;
101 struct dst_entry *dst;
102
103 afinfo = xfrm_policy_get_afinfo(family);
104 if (unlikely(afinfo == NULL))
105 return ERR_PTR(-EAFNOSUPPORT);
106
107 dst = afinfo->dst_lookup(tos, saddr, daddr);
108
109 xfrm_policy_put_afinfo(afinfo);
110
111 return dst;
112}
113
25ee3286 114static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 115 xfrm_address_t *prev_saddr,
116 xfrm_address_t *prev_daddr,
25ee3286 117 int family)
1da177e4 118{
66cdb3ca
HX		 119 xfrm_address_t *saddr = &x->props.saddr;
120 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 121 struct dst_entry *dst;
122
9bb182a7 123 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 124 saddr = x->coaddr;
9bb182a7
YH		 125 daddr = prev_daddr;
126 }
127 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
128 saddr = prev_saddr;
66cdb3ca 129 daddr = x->coaddr;
9bb182a7 130 }
1da177e4 131
9bb182a7
YH		 132 dst = __xfrm_dst_lookup(tos, saddr, daddr, family);
133
134 if (!IS_ERR(dst)) {
135 if (prev_saddr != saddr)
136 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
137 if (prev_daddr != daddr)
138 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
139 }
1da177e4 140
66cdb3ca 141 return dst;
1da177e4 142}
1da177e4 143
1da177e4
LT		 144static inline unsigned long make_jiffies(long secs)
145{
146 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
147 return MAX_SCHEDULE_TIMEOUT-1;
148 else
a716c119 149 return secs*HZ;
1da177e4
LT		 150}
151
152static void xfrm_policy_timer(unsigned long data)
153{
154 struct xfrm_policy *xp = (struct xfrm_policy*)data;
9d729f72 155 unsigned long now = get_seconds();
1da177e4
LT		 156 long next = LONG_MAX;
157 int warn = 0;
158 int dir;
159
160 read_lock(&xp->lock);
161
12a169e7 162 if (xp->walk.dead)
1da177e4
LT		 163 goto out;
164
77d8d7a6 165 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 166
167 if (xp->lft.hard_add_expires_seconds) {
168 long tmo = xp->lft.hard_add_expires_seconds +
169 xp->curlft.add_time - now;
170 if (tmo <= 0)
171 goto expired;
172 if (tmo < next)
173 next = tmo;
174 }
175 if (xp->lft.hard_use_expires_seconds) {
176 long tmo = xp->lft.hard_use_expires_seconds +
177 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
178 if (tmo <= 0)
179 goto expired;
180 if (tmo < next)
181 next = tmo;
182 }
183 if (xp->lft.soft_add_expires_seconds) {
184 long tmo = xp->lft.soft_add_expires_seconds +
185 xp->curlft.add_time - now;
186 if (tmo <= 0) {
187 warn = 1;
188 tmo = XFRM_KM_TIMEOUT;
189 }
190 if (tmo < next)
191 next = tmo;
192 }
193 if (xp->lft.soft_use_expires_seconds) {
194 long tmo = xp->lft.soft_use_expires_seconds +
195 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
196 if (tmo <= 0) {
197 warn = 1;
198 tmo = XFRM_KM_TIMEOUT;
199 }
200 if (tmo < next)
201 next = tmo;
202 }
203
204 if (warn)
6c5c8ca7 205 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 206 if (next != LONG_MAX &&
207 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
208 xfrm_pol_hold(xp);
209
210out:
211 read_unlock(&xp->lock);
212 xfrm_pol_put(xp);
213 return;
214
215expired:
216 read_unlock(&xp->lock);
4666faab 217 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 218 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 219 xfrm_pol_put(xp);
220}
221
222
223/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
224 * SPD calls.
225 */
226
0331b1f3 227struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 228{
229 struct xfrm_policy *policy;
230
0da974f4 231 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 232
233 if (policy) {
0331b1f3 234 write_pnet(&policy->xp_net, net);
12a169e7 235 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 236 INIT_HLIST_NODE(&policy->bydst);
237 INIT_HLIST_NODE(&policy->byidx);
1da177e4 238 rwlock_init(&policy->lock);
2518c7c2 239 atomic_set(&policy->refcnt, 1);
b24b8a24
PE		 240 setup_timer(&policy->timer, xfrm_policy_timer,
241 (unsigned long)policy);
1da177e4
LT		 242 }
243 return policy;
244}
245EXPORT_SYMBOL(xfrm_policy_alloc);
246
247/* Destroy xfrm_policy: descendant resources must be released to this moment. */
248
64c31b3f 249void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 250{
12a169e7 251 BUG_ON(!policy->walk.dead);
1da177e4 252
09a62660 253 BUG_ON(policy->bundles);
1da177e4
LT		 254
255 if (del_timer(&policy->timer))
256 BUG();
257
03e1ad7b 258 security_xfrm_policy_free(policy->security);
1da177e4
LT		 259 kfree(policy);
260}
64c31b3f 261EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4
LT		 262
263static void xfrm_policy_gc_kill(struct xfrm_policy *policy)
264{
265 struct dst_entry *dst;
266
267 while ((dst = policy->bundles) != NULL) {
268 policy->bundles = dst->next;
269 dst_free(dst);
270 }
271
272 if (del_timer(&policy->timer))
273 atomic_dec(&policy->refcnt);
274
275 if (atomic_read(&policy->refcnt) > 1)
276 flow_cache_flush();
277
278 xfrm_pol_put(policy);
279}
280
c4028958 281static void xfrm_policy_gc_task(struct work_struct *work)
1da177e4
LT		 282{
283 struct xfrm_policy *policy;
2518c7c2
DM		 284 struct hlist_node *entry, *tmp;
285 struct hlist_head gc_list;
1da177e4
LT		 286
287 spin_lock_bh(&xfrm_policy_gc_lock);
2518c7c2
DM		 288 gc_list.first = xfrm_policy_gc_list.first;
289 INIT_HLIST_HEAD(&xfrm_policy_gc_list);
1da177e4
LT		 290 spin_unlock_bh(&xfrm_policy_gc_lock);
291
2518c7c2 292 hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst)
1da177e4 293 xfrm_policy_gc_kill(policy);
1da177e4 294}
c9583969 295static DECLARE_WORK(xfrm_policy_gc_work, xfrm_policy_gc_task);
1da177e4
LT		 296
297/* Rule must be locked. Release descentant resources, announce
298 * entry dead. The rule must be unlinked from lists to the moment.
299 */
300
301static void xfrm_policy_kill(struct xfrm_policy *policy)
302{
303 int dead;
304
305 write_lock_bh(&policy->lock);
12a169e7
HX		 306 dead = policy->walk.dead;
307 policy->walk.dead = 1;
1da177e4
LT		 308 write_unlock_bh(&policy->lock);
309
310 if (unlikely(dead)) {
311 WARN_ON(1);
312 return;
313 }
314
bbb770e7 315 spin_lock_bh(&xfrm_policy_gc_lock);
2518c7c2 316 hlist_add_head(&policy->bydst, &xfrm_policy_gc_list);
bbb770e7 317 spin_unlock_bh(&xfrm_policy_gc_lock);
1da177e4
LT		 318
319 schedule_work(&xfrm_policy_gc_work);
320}
321
2518c7c2
DM		 322static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
323
e92303f8 324static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 325{
e92303f8 326 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 327}
328
2518c7c2
DM		 329static struct hlist_head *policy_hash_bysel(struct xfrm_selector *sel, unsigned short family, int dir)
330{
a35f6c5d 331 unsigned int hmask = init_net.xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 332 unsigned int hash = __sel_hash(sel, family, hmask);
333
334 return (hash == hmask + 1 ?
8b18f8ea 335 &init_net.xfrm.policy_inexact[dir] :
a35f6c5d 336 init_net.xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 337}
338
339static struct hlist_head *policy_hash_direct(xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
340{
a35f6c5d 341 unsigned int hmask = init_net.xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 342 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
343
a35f6c5d 344 return init_net.xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 345}
346
2518c7c2
DM		 347static void xfrm_dst_hash_transfer(struct hlist_head *list,
348 struct hlist_head *ndsttable,
349 unsigned int nhashmask)
350{
b791160b 351 struct hlist_node *entry, *tmp, *entry0 = NULL;
2518c7c2 352 struct xfrm_policy *pol;
b791160b 353 unsigned int h0 = 0;
2518c7c2 354
b791160b 355redo:
2518c7c2
DM		 356 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
357 unsigned int h;
358
359 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
360 pol->family, nhashmask);
b791160b
YH		 361 if (!entry0) {
362 hlist_del(entry);
363 hlist_add_head(&pol->bydst, ndsttable+h);
364 h0 = h;
365 } else {
366 if (h != h0)
367 continue;
368 hlist_del(entry);
369 hlist_add_after(entry0, &pol->bydst);
370 }
371 entry0 = entry;
372 }
373 if (!hlist_empty(list)) {
374 entry0 = NULL;
375 goto redo;
2518c7c2
DM		 376 }
377}
378
379static void xfrm_idx_hash_transfer(struct hlist_head *list,
380 struct hlist_head *nidxtable,
381 unsigned int nhashmask)
382{
383 struct hlist_node *entry, *tmp;
384 struct xfrm_policy *pol;
385
386 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
387 unsigned int h;
388
389 h = __idx_hash(pol->index, nhashmask);
390 hlist_add_head(&pol->byidx, nidxtable+h);
391 }
392}
393
394static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
395{
396 return ((old_hmask + 1) << 1) - 1;
397}
398
66caf628 399static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 400{
66caf628 401 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 402 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
403 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 404 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 405 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 406 int i;
407
408 if (!ndst)
409 return;
410
411 write_lock_bh(&xfrm_policy_lock);
412
413 for (i = hmask; i >= 0; i--)
414 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
415
66caf628
AD		 416 net->xfrm.policy_bydst[dir].table = ndst;
417 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2
DM		 418
419 write_unlock_bh(&xfrm_policy_lock);
420
44e36b42 421 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 422}
423
66caf628 424static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 425{
66caf628 426 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 427 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
428 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 429 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 430 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 431 int i;
432
433 if (!nidx)
434 return;
435
436 write_lock_bh(&xfrm_policy_lock);
437
438 for (i = hmask; i >= 0; i--)
439 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
440
66caf628
AD		 441 net->xfrm.policy_byidx = nidx;
442 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2
DM		 443
444 write_unlock_bh(&xfrm_policy_lock);
445
44e36b42 446 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 447}
448
66caf628 449static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 450{
66caf628
AD		 451 unsigned int cnt = net->xfrm.policy_count[dir];
452 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 453
454 if (total)
455 *total += cnt;
456
457 if ((hmask + 1) < xfrm_policy_hashmax &&
458 cnt > hmask)
459 return 1;
460
461 return 0;
462}
463
66caf628 464static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 465{
66caf628 466 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 467
468 if ((hmask + 1) < xfrm_policy_hashmax &&
469 total > hmask)
470 return 1;
471
472 return 0;
473}
474
5a6d3416 475void xfrm_spd_getinfo(struct xfrmk_spdinfo *si)
ecfd6b18
JHS		 476{
477 read_lock_bh(&xfrm_policy_lock);
dc2caba7
AD		 478 si->incnt = init_net.xfrm.policy_count[XFRM_POLICY_IN];
479 si->outcnt = init_net.xfrm.policy_count[XFRM_POLICY_OUT];
480 si->fwdcnt = init_net.xfrm.policy_count[XFRM_POLICY_FWD];
481 si->inscnt = init_net.xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
482 si->outscnt = init_net.xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
483 si->fwdscnt = init_net.xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
8100bea7 484 si->spdhcnt = init_net.xfrm.policy_idx_hmask;
ecfd6b18
JHS		 485 si->spdhmcnt = xfrm_policy_hashmax;
486 read_unlock_bh(&xfrm_policy_lock);
487}
488EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 489
ecfd6b18 490static DEFINE_MUTEX(hash_resize_mutex);
66caf628 491static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 492{
66caf628 493 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 494 int dir, total;
495
496 mutex_lock(&hash_resize_mutex);
497
498 total = 0;
499 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 500 if (xfrm_bydst_should_resize(net, dir, &total))
501 xfrm_bydst_resize(net, dir);
2518c7c2 502 }
66caf628
AD		 503 if (xfrm_byidx_should_resize(net, total))
504 xfrm_byidx_resize(net, total);
2518c7c2
DM		 505
506 mutex_unlock(&hash_resize_mutex);
507}
508
1da177e4
LT		 509/* Generate new index... KAME seems to generate them ordered by cost
510 * of an absolute inpredictability of ordering of rules. This will not pass. */
7a12122c 511static u32 xfrm_gen_index(int dir)
1da177e4 512{
1da177e4
LT		 513 static u32 idx_generator;
514
515 for (;;) {
2518c7c2
DM		 516 struct hlist_node *entry;
517 struct hlist_head *list;
518 struct xfrm_policy *p;
519 u32 idx;
520 int found;
521
1da177e4
LT		 522 idx = (idx_generator | dir);
523 idx_generator += 8;
524 if (idx == 0)
525 idx = 8;
e92303f8 526 list = init_net.xfrm.policy_byidx + idx_hash(&init_net, idx);
2518c7c2
DM		 527 found = 0;
528 hlist_for_each_entry(p, entry, list, byidx) {
529 if (p->index == idx) {
530 found = 1;
1da177e4 531 break;
2518c7c2 532 }
1da177e4 533 }
2518c7c2 534 if (!found)
1da177e4
LT		 535 return idx;
536 }
537}
538
2518c7c2
DM		 539static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
540{
541 u32 *p1 = (u32 *) s1;
542 u32 *p2 = (u32 *) s2;
543 int len = sizeof(struct xfrm_selector) / sizeof(u32);
544 int i;
545
546 for (i = 0; i < len; i++) {
547 if (p1[i] != p2[i])
548 return 1;
549 }
550
551 return 0;
552}
553
1da177e4
LT		 554int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
555{
2518c7c2
DM		 556 struct xfrm_policy *pol;
557 struct xfrm_policy *delpol;
558 struct hlist_head *chain;
a6c7ab55 559 struct hlist_node *entry, *newpos;
9b78a82c 560 struct dst_entry *gc_list;
1da177e4
LT		 561
562 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 563 chain = policy_hash_bysel(&policy->selector, policy->family, dir);
564 delpol = NULL;
565 newpos = NULL;
2518c7c2 566 hlist_for_each_entry(pol, entry, chain, bydst) {
a6c7ab55 567 if (pol->type == policy->type &&
2518c7c2 568 !selector_cmp(&pol->selector, &policy->selector) &&
a6c7ab55
HX		 569 xfrm_sec_ctx_match(pol->security, policy->security) &&
570 !WARN_ON(delpol)) {
1da177e4
LT		 571 if (excl) {
572 write_unlock_bh(&xfrm_policy_lock);
573 return -EEXIST;
574 }
1da177e4
LT		 575 delpol = pol;
576 if (policy->priority > pol->priority)
577 continue;
578 } else if (policy->priority >= pol->priority) {
a6c7ab55 579 newpos = &pol->bydst;
1da177e4
LT		 580 continue;
581 }
1da177e4
LT		 582 if (delpol)
583 break;
1da177e4
LT		 584 }
585 if (newpos)
2518c7c2
DM		 586 hlist_add_after(newpos, &policy->bydst);
587 else
588 hlist_add_head(&policy->bydst, chain);
1da177e4 589 xfrm_pol_hold(policy);
dc2caba7 590 init_net.xfrm.policy_count[dir]++;
1da177e4 591 atomic_inc(&flow_cache_genid);
2518c7c2
DM		 592 if (delpol) {
593 hlist_del(&delpol->bydst);
594 hlist_del(&delpol->byidx);
12a169e7 595 list_del(&delpol->walk.all);
dc2caba7 596 init_net.xfrm.policy_count[dir]--;
2518c7c2 597 }
7a12122c 598 policy->index = delpol ? delpol->index : xfrm_gen_index(dir);
e92303f8 599 hlist_add_head(&policy->byidx, init_net.xfrm.policy_byidx+idx_hash(&init_net, policy->index));
9d729f72 600 policy->curlft.add_time = get_seconds();
1da177e4
LT		 601 policy->curlft.use_time = 0;
602 if (!mod_timer(&policy->timer, jiffies + HZ))
603 xfrm_pol_hold(policy);
adfcf0b2 604 list_add(&policy->walk.all, &init_net.xfrm.policy_all);
1da177e4
LT		 605 write_unlock_bh(&xfrm_policy_lock);
606
9b78a82c 607 if (delpol)
1da177e4 608 xfrm_policy_kill(delpol);
66caf628
AD		 609 else if (xfrm_bydst_should_resize(&init_net, dir, NULL))
610 schedule_work(&init_net.xfrm.policy_hash_work);
9b78a82c
DM		 611
612 read_lock_bh(&xfrm_policy_lock);
613 gc_list = NULL;
2518c7c2
DM		 614 entry = &policy->bydst;
615 hlist_for_each_entry_continue(policy, entry, bydst) {
9b78a82c
DM		 616 struct dst_entry *dst;
617
618 write_lock(&policy->lock);
619 dst = policy->bundles;
620 if (dst) {
621 struct dst_entry *tail = dst;
622 while (tail->next)
623 tail = tail->next;
624 tail->next = gc_list;
625 gc_list = dst;
626
627 policy->bundles = NULL;
628 }
629 write_unlock(&policy->lock);
1da177e4 630 }
9b78a82c
DM		 631 read_unlock_bh(&xfrm_policy_lock);
632
633 while (gc_list) {
634 struct dst_entry *dst = gc_list;
635
636 gc_list = dst->next;
637 dst_free(dst);
638 }
639
1da177e4
LT		 640 return 0;
641}
642EXPORT_SYMBOL(xfrm_policy_insert);
643
4e81bb83
MN		 644struct xfrm_policy *xfrm_policy_bysel_ctx(u8 type, int dir,
645 struct xfrm_selector *sel,
ef41aaa0
EP		 646 struct xfrm_sec_ctx *ctx, int delete,
647 int *err)
1da177e4 648{
2518c7c2
DM		 649 struct xfrm_policy *pol, *ret;
650 struct hlist_head *chain;
651 struct hlist_node *entry;
1da177e4 652
ef41aaa0 653 *err = 0;
1da177e4 654 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 655 chain = policy_hash_bysel(sel, sel->family, dir);
656 ret = NULL;
657 hlist_for_each_entry(pol, entry, chain, bydst) {
658 if (pol->type == type &&
659 !selector_cmp(sel, &pol->selector) &&
660 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 661 xfrm_pol_hold(pol);
2518c7c2 662 if (delete) {
03e1ad7b
PM		 663 *err = security_xfrm_policy_delete(
664 pol->security);
ef41aaa0
EP		 665 if (*err) {
666 write_unlock_bh(&xfrm_policy_lock);
667 return pol;
668 }
2518c7c2
DM		 669 hlist_del(&pol->bydst);
670 hlist_del(&pol->byidx);
12a169e7 671 list_del(&pol->walk.all);
dc2caba7 672 init_net.xfrm.policy_count[dir]--;
2518c7c2
DM		 673 }
674 ret = pol;
1da177e4
LT		 675 break;
676 }
677 }
678 write_unlock_bh(&xfrm_policy_lock);
679
2518c7c2 680 if (ret && delete) {
1da177e4 681 atomic_inc(&flow_cache_genid);
2518c7c2 682 xfrm_policy_kill(ret);
1da177e4 683 }
2518c7c2 684 return ret;
1da177e4 685}
df71837d 686EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 687
ef41aaa0
EP		 688struct xfrm_policy *xfrm_policy_byid(u8 type, int dir, u32 id, int delete,
689 int *err)
1da177e4 690{
2518c7c2
DM		 691 struct xfrm_policy *pol, *ret;
692 struct hlist_head *chain;
693 struct hlist_node *entry;
1da177e4 694
b5505c6e
HX		 695 *err = -ENOENT;
696 if (xfrm_policy_id2dir(id) != dir)
697 return NULL;
698
ef41aaa0 699 *err = 0;
1da177e4 700 write_lock_bh(&xfrm_policy_lock);
e92303f8 701 chain = init_net.xfrm.policy_byidx + idx_hash(&init_net, id);
2518c7c2
DM		 702 ret = NULL;
703 hlist_for_each_entry(pol, entry, chain, byidx) {
704 if (pol->type == type && pol->index == id) {
1da177e4 705 xfrm_pol_hold(pol);
2518c7c2 706 if (delete) {
03e1ad7b
PM		 707 *err = security_xfrm_policy_delete(
708 pol->security);
ef41aaa0
EP		 709 if (*err) {
710 write_unlock_bh(&xfrm_policy_lock);
711 return pol;
712 }
2518c7c2
DM		 713 hlist_del(&pol->bydst);
714 hlist_del(&pol->byidx);
12a169e7 715 list_del(&pol->walk.all);
dc2caba7 716 init_net.xfrm.policy_count[dir]--;
2518c7c2
DM		 717 }
718 ret = pol;
1da177e4
LT		 719 break;
720 }
721 }
722 write_unlock_bh(&xfrm_policy_lock);
723
2518c7c2 724 if (ret && delete) {
1da177e4 725 atomic_inc(&flow_cache_genid);
2518c7c2 726 xfrm_policy_kill(ret);
1da177e4 727 }
2518c7c2 728 return ret;
1da177e4
LT		 729}
730EXPORT_SYMBOL(xfrm_policy_byid);
731
4aa2e62c
JL		 732#ifdef CONFIG_SECURITY_NETWORK_XFRM
733static inline int
734xfrm_policy_flush_secctx_check(u8 type, struct xfrm_audit *audit_info)
1da177e4 735{
4aa2e62c
JL		 736 int dir, err = 0;
737
738 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
739 struct xfrm_policy *pol;
740 struct hlist_node *entry;
741 int i;
742
743 hlist_for_each_entry(pol, entry,
8b18f8ea 744 &init_net.xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 745 if (pol->type != type)
746 continue;
03e1ad7b 747 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 748 if (err) {
ab5f5e8b
JL		 749 xfrm_audit_policy_delete(pol, 0,
750 audit_info->loginuid,
2532386f 751 audit_info->sessionid,
ab5f5e8b 752 audit_info->secid);
4aa2e62c
JL		 753 return err;
754 }
7dc12d6d 755 }
a35f6c5d 756 for (i = init_net.xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
4aa2e62c 757 hlist_for_each_entry(pol, entry,
a35f6c5d 758 init_net.xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 759 bydst) {
760 if (pol->type != type)
761 continue;
03e1ad7b
PM		 762 err = security_xfrm_policy_delete(
763 pol->security);
4aa2e62c 764 if (err) {
ab5f5e8b
JL		 765 xfrm_audit_policy_delete(pol, 0,
766 audit_info->loginuid,
2532386f 767 audit_info->sessionid,
ab5f5e8b 768 audit_info->secid);
4aa2e62c
JL		 769 return err;
770 }
771 }
772 }
773 }
774 return err;
775}
776#else
777static inline int
778xfrm_policy_flush_secctx_check(u8 type, struct xfrm_audit *audit_info)
779{
780 return 0;
781}
782#endif
783
784int xfrm_policy_flush(u8 type, struct xfrm_audit *audit_info)
785{
786 int dir, err = 0;
1da177e4
LT		 787
788 write_lock_bh(&xfrm_policy_lock);
4aa2e62c
JL		 789
790 err = xfrm_policy_flush_secctx_check(type, audit_info);
791 if (err)
792 goto out;
793
1da177e4 794 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 795 struct xfrm_policy *pol;
796 struct hlist_node *entry;
ae8c0577 797 int i, killed;
2518c7c2 798
ae8c0577 799 killed = 0;
2518c7c2
DM		 800 again1:
801 hlist_for_each_entry(pol, entry,
8b18f8ea 802 &init_net.xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 803 if (pol->type != type)
804 continue;
805 hlist_del(&pol->bydst);
806 hlist_del(&pol->byidx);
1da177e4
LT		 807 write_unlock_bh(&xfrm_policy_lock);
808
ab5f5e8b 809 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
2532386f 810 audit_info->sessionid,
ab5f5e8b 811 audit_info->secid);
161a09e7 812
2518c7c2 813 xfrm_policy_kill(pol);
ae8c0577 814 killed++;
1da177e4
LT		 815
816 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 817 goto again1;
818 }
819
a35f6c5d 820 for (i = init_net.xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2
DM		 821 again2:
822 hlist_for_each_entry(pol, entry,
a35f6c5d 823 init_net.xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 824 bydst) {
825 if (pol->type != type)
826 continue;
827 hlist_del(&pol->bydst);
828 hlist_del(&pol->byidx);
12a169e7 829 list_del(&pol->walk.all);
2518c7c2
DM		 830 write_unlock_bh(&xfrm_policy_lock);
831
ab5f5e8b
JL		 832 xfrm_audit_policy_delete(pol, 1,
833 audit_info->loginuid,
2532386f 834 audit_info->sessionid,
ab5f5e8b 835 audit_info->secid);
2518c7c2 836 xfrm_policy_kill(pol);
ae8c0577 837 killed++;
2518c7c2
DM		 838
839 write_lock_bh(&xfrm_policy_lock);
840 goto again2;
841 }
1da177e4 842 }
2518c7c2 843
dc2caba7 844 init_net.xfrm.policy_count[dir] -= killed;
1da177e4
LT		 845 }
846 atomic_inc(&flow_cache_genid);
4aa2e62c 847out:
1da177e4 848 write_unlock_bh(&xfrm_policy_lock);
4aa2e62c 849 return err;
1da177e4
LT		 850}
851EXPORT_SYMBOL(xfrm_policy_flush);
852
4c563f76
TT		 853int xfrm_policy_walk(struct xfrm_policy_walk *walk,
854 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 855 void *data)
856{
12a169e7
HX		 857 struct xfrm_policy *pol;
858 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 859 int error = 0;
860
861 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
862 walk->type != XFRM_POLICY_TYPE_ANY)
863 return -EINVAL;
1da177e4 864
12a169e7 865 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 866 return 0;
867
12a169e7
HX		 868 write_lock_bh(&xfrm_policy_lock);
869 if (list_empty(&walk->walk.all))
adfcf0b2 870 x = list_first_entry(&init_net.xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 871 else
872 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
adfcf0b2 873 list_for_each_entry_from(x, &init_net.xfrm.policy_all, all) {
12a169e7 874 if (x->dead)
4c563f76 875 continue;
12a169e7
HX		 876 pol = container_of(x, struct xfrm_policy, walk);
877 if (walk->type != XFRM_POLICY_TYPE_ANY &&
878 walk->type != pol->type)
879 continue;
880 error = func(pol, xfrm_policy_id2dir(pol->index),
881 walk->seq, data);
882 if (error) {
883 list_move_tail(&walk->walk.all, &x->all);
884 goto out;
2518c7c2 885 }
12a169e7 886 walk->seq++;
1da177e4 887 }
12a169e7 888 if (walk->seq == 0) {
baf5d743
JHS		 889 error = -ENOENT;
890 goto out;
891 }
12a169e7 892 list_del_init(&walk->walk.all);
1da177e4 893out:
12a169e7 894 write_unlock_bh(&xfrm_policy_lock);
1da177e4
LT		 895 return error;
896}
897EXPORT_SYMBOL(xfrm_policy_walk);
898
12a169e7
HX		 899void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
900{
901 INIT_LIST_HEAD(&walk->walk.all);
902 walk->walk.dead = 1;
903 walk->type = type;
904 walk->seq = 0;
905}
906EXPORT_SYMBOL(xfrm_policy_walk_init);
907
908void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
909{
910 if (list_empty(&walk->walk.all))
911 return;
912
913 write_lock_bh(&xfrm_policy_lock);
914 list_del(&walk->walk.all);
915 write_unlock_bh(&xfrm_policy_lock);
916}
917EXPORT_SYMBOL(xfrm_policy_walk_done);
918
134b0fc5
JM		 919/*
920 * Find policy to apply to this flow.
921 *
922 * Returns 0 if policy found, else an -errno.
923 */
2518c7c2
DM		 924static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
925 u8 type, u16 family, int dir)
1da177e4 926{
2518c7c2 927 struct xfrm_selector *sel = &pol->selector;
134b0fc5 928 int match, ret = -ESRCH;
1da177e4 929
2518c7c2
DM		 930 if (pol->family != family ||
931 pol->type != type)
134b0fc5 932 return ret;
1da177e4 933
2518c7c2 934 match = xfrm_selector_match(sel, fl, family);
134b0fc5 935 if (match)
03e1ad7b
PM		 936 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
937 dir);
2518c7c2 938
134b0fc5 939 return ret;
2518c7c2 940}
1da177e4 941
2518c7c2
DM		 942static struct xfrm_policy *xfrm_policy_lookup_bytype(u8 type, struct flowi *fl,
943 u16 family, u8 dir)
944{
134b0fc5 945 int err;
2518c7c2
DM		 946 struct xfrm_policy *pol, *ret;
947 xfrm_address_t *daddr, *saddr;
948 struct hlist_node *entry;
949 struct hlist_head *chain;
acba48e1 950 u32 priority = ~0U;
df71837d 951
2518c7c2
DM		 952 daddr = xfrm_flowi_daddr(fl, family);
953 saddr = xfrm_flowi_saddr(fl, family);
954 if (unlikely(!daddr || !saddr))
955 return NULL;
956
957 read_lock_bh(&xfrm_policy_lock);
958 chain = policy_hash_direct(daddr, saddr, family, dir);
959 ret = NULL;
960 hlist_for_each_entry(pol, entry, chain, bydst) {
134b0fc5
JM		 961 err = xfrm_policy_match(pol, fl, type, family, dir);
962 if (err) {
963 if (err == -ESRCH)
964 continue;
965 else {
966 ret = ERR_PTR(err);
967 goto fail;
968 }
969 } else {
2518c7c2 970 ret = pol;
acba48e1 971 priority = ret->priority;
2518c7c2
DM		 972 break;
973 }
974 }
8b18f8ea 975 chain = &init_net.xfrm.policy_inexact[dir];
acba48e1 976 hlist_for_each_entry(pol, entry, chain, bydst) {
134b0fc5
JM		 977 err = xfrm_policy_match(pol, fl, type, family, dir);
978 if (err) {
979 if (err == -ESRCH)
980 continue;
981 else {
982 ret = ERR_PTR(err);
983 goto fail;
984 }
985 } else if (pol->priority < priority) {
acba48e1
DM		 986 ret = pol;
987 break;
1da177e4
LT		 988 }
989 }
acba48e1
DM		 990 if (ret)
991 xfrm_pol_hold(ret);
134b0fc5 992fail:
1da177e4 993 read_unlock_bh(&xfrm_policy_lock);
4e81bb83 994
2518c7c2 995 return ret;
4e81bb83
MN		 996}
997
134b0fc5 998static int xfrm_policy_lookup(struct flowi *fl, u16 family, u8 dir,
4e81bb83
MN		 999 void **objp, atomic_t **obj_refp)
1000{
1001 struct xfrm_policy *pol;
134b0fc5 1002 int err = 0;
4e81bb83
MN		 1003
1004#ifdef CONFIG_XFRM_SUB_POLICY
1005 pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB, fl, family, dir);
134b0fc5
JM		 1006 if (IS_ERR(pol)) {
1007 err = PTR_ERR(pol);
1008 pol = NULL;
1009 }
1010 if (pol || err)
4e81bb83
MN		 1011 goto end;
1012#endif
1013 pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN, fl, family, dir);
134b0fc5
JM		 1014 if (IS_ERR(pol)) {
1015 err = PTR_ERR(pol);
1016 pol = NULL;
1017 }
4e81bb83 1018#ifdef CONFIG_XFRM_SUB_POLICY
2518c7c2 1019end:
4e81bb83 1020#endif
1da177e4
LT		 1021 if ((*objp = (void *) pol) != NULL)
1022 *obj_refp = &pol->refcnt;
134b0fc5 1023 return err;
1da177e4
LT		 1024}
1025
df71837d
TJ		 1026static inline int policy_to_flow_dir(int dir)
1027{
1028 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1029 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1030 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1031 return dir;
1032 switch (dir) {
1033 default:
1034 case XFRM_POLICY_IN:
1035 return FLOW_DIR_IN;
1036 case XFRM_POLICY_OUT:
1037 return FLOW_DIR_OUT;
1038 case XFRM_POLICY_FWD:
1039 return FLOW_DIR_FWD;
3ff50b79 1040 }
df71837d
TJ		 1041}
1042
e0d1caa7 1043static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
1da177e4
LT		 1044{
1045 struct xfrm_policy *pol;
1046
1047 read_lock_bh(&xfrm_policy_lock);
1048 if ((pol = sk->sk_policy[dir]) != NULL) {
a716c119 1049 int match = xfrm_selector_match(&pol->selector, fl,
1da177e4 1050 sk->sk_family);
a716c119 1051 int err = 0;
df71837d 1052
3bccfbc7 1053 if (match) {
03e1ad7b
PM		 1054 err = security_xfrm_policy_lookup(pol->security,
1055 fl->secid,
1056 policy_to_flow_dir(dir));
3bccfbc7
VY		 1057 if (!err)
1058 xfrm_pol_hold(pol);
1059 else if (err == -ESRCH)
1060 pol = NULL;
1061 else
1062 pol = ERR_PTR(err);
1063 } else
1da177e4
LT		 1064 pol = NULL;
1065 }
1066 read_unlock_bh(&xfrm_policy_lock);
1067 return pol;
1068}
1069
1070static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1071{
98806f75 1072 struct net *net = xp_net(pol);
2518c7c2
DM		 1073 struct hlist_head *chain = policy_hash_bysel(&pol->selector,
1074 pol->family, dir);
4e81bb83 1075
98806f75 1076 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1077 hlist_add_head(&pol->bydst, chain);
e92303f8 1078 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1079 net->xfrm.policy_count[dir]++;
1da177e4 1080 xfrm_pol_hold(pol);
2518c7c2 1081
98806f75
AD		 1082 if (xfrm_bydst_should_resize(net, dir, NULL))
1083 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1084}
1085
1086static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1087 int dir)
1088{
98806f75
AD		 1089 struct net *net = xp_net(pol);
1090
2518c7c2
DM		 1091 if (hlist_unhashed(&pol->bydst))
1092 return NULL;
1da177e4 1093
2518c7c2
DM		 1094 hlist_del(&pol->bydst);
1095 hlist_del(&pol->byidx);
12a169e7 1096 list_del(&pol->walk.all);
98806f75 1097 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1098
1099 return pol;
1da177e4
LT		 1100}
1101
4666faab 1102int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4
LT		 1103{
1104 write_lock_bh(&xfrm_policy_lock);
1105 pol = __xfrm_policy_unlink(pol, dir);
1106 write_unlock_bh(&xfrm_policy_lock);
1107 if (pol) {
1108 if (dir < XFRM_POLICY_MAX)
1109 atomic_inc(&flow_cache_genid);
1110 xfrm_policy_kill(pol);
4666faab 1111 return 0;
1da177e4 1112 }
4666faab 1113 return -ENOENT;
1da177e4 1114}
a70fcb0b 1115EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1116
1117int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1118{
1119 struct xfrm_policy *old_pol;
1120
4e81bb83
MN		 1121#ifdef CONFIG_XFRM_SUB_POLICY
1122 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1123 return -EINVAL;
1124#endif
1125
1da177e4
LT		 1126 write_lock_bh(&xfrm_policy_lock);
1127 old_pol = sk->sk_policy[dir];
1128 sk->sk_policy[dir] = pol;
1129 if (pol) {
9d729f72 1130 pol->curlft.add_time = get_seconds();
7a12122c 1131 pol->index = xfrm_gen_index(XFRM_POLICY_MAX+dir);
1da177e4
LT		 1132 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1133 }
1134 if (old_pol)
1135 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1136 write_unlock_bh(&xfrm_policy_lock);
1137
1138 if (old_pol) {
1139 xfrm_policy_kill(old_pol);
1140 }
1141 return 0;
1142}
1143
1144static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1145{
0331b1f3 1146 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1da177e4
LT		 1147
1148 if (newp) {
1149 newp->selector = old->selector;
03e1ad7b
PM		 1150 if (security_xfrm_policy_clone(old->security,
1151 &newp->security)) {
df71837d
TJ		 1152 kfree(newp);
1153 return NULL; /* ENOMEM */
1154 }
1da177e4
LT		 1155 newp->lft = old->lft;
1156 newp->curlft = old->curlft;
1157 newp->action = old->action;
1158 newp->flags = old->flags;
1159 newp->xfrm_nr = old->xfrm_nr;
1160 newp->index = old->index;
4e81bb83 1161 newp->type = old->type;
1da177e4
LT		 1162 memcpy(newp->xfrm_vec, old->xfrm_vec,
1163 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1164 write_lock_bh(&xfrm_policy_lock);
1165 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1166 write_unlock_bh(&xfrm_policy_lock);
1167 xfrm_pol_put(newp);
1168 }
1169 return newp;
1170}
1171
1172int __xfrm_sk_clone_policy(struct sock *sk)
1173{
1174 struct xfrm_policy *p0 = sk->sk_policy[0],
1175 *p1 = sk->sk_policy[1];
1176
1177 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1178 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1179 return -ENOMEM;
1180 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1181 return -ENOMEM;
1182 return 0;
1183}
1184
a1e59abf
PM		 1185static int
1186xfrm_get_saddr(xfrm_address_t *local, xfrm_address_t *remote,
1187 unsigned short family)
1188{
1189 int err;
1190 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1191
1192 if (unlikely(afinfo == NULL))
1193 return -EINVAL;
1194 err = afinfo->get_saddr(local, remote);
1195 xfrm_policy_put_afinfo(afinfo);
1196 return err;
1197}
1198
1da177e4
LT		 1199/* Resolve list of templates for the flow, given policy. */
1200
1201static int
4e81bb83
MN		 1202xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
1203 struct xfrm_state **xfrm,
1204 unsigned short family)
1da177e4
LT		 1205{
1206 int nx;
1207 int i, error;
1208 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1209 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1210 xfrm_address_t tmp;
1da177e4
LT		 1211
1212 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1213 struct xfrm_state *x;
1214 xfrm_address_t *remote = daddr;
1215 xfrm_address_t *local = saddr;
1216 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1217
48b8d783
JK		 1218 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1219 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1220 remote = &tmpl->id.daddr;
1221 local = &tmpl->saddr;
76b3f055 1222 family = tmpl->encap_family;
a1e59abf
PM		 1223 if (xfrm_addr_any(local, family)) {
1224 error = xfrm_get_saddr(&tmp, remote, family);
1225 if (error)
1226 goto fail;
1227 local = &tmp;
1228 }
1da177e4
LT		 1229 }
1230
1231 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1232
1233 if (x && x->km.state == XFRM_STATE_VALID) {
1234 xfrm[nx++] = x;
1235 daddr = remote;
1236 saddr = local;
1237 continue;
1238 }
1239 if (x) {
1240 error = (x->km.state == XFRM_STATE_ERROR ?
1241 -EINVAL : -EAGAIN);
1242 xfrm_state_put(x);
1243 }
a4322266 1244 else if (error == -ESRCH)
1245 error = -EAGAIN;
1da177e4
LT		 1246
1247 if (!tmpl->optional)
1248 goto fail;
1249 }
1250 return nx;
1251
1252fail:
1253 for (nx--; nx>=0; nx--)
1254 xfrm_state_put(xfrm[nx]);
1255 return error;
1256}
1257
4e81bb83
MN		 1258static int
1259xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1260 struct xfrm_state **xfrm,
1261 unsigned short family)
1262{
41a49cc3
MN		 1263 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1264 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1265 int cnx = 0;
1266 int error;
1267 int ret;
1268 int i;
1269
1270 for (i = 0; i < npols; i++) {
1271 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1272 error = -ENOBUFS;
1273 goto fail;
1274 }
41a49cc3
MN		 1275
1276 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1277 if (ret < 0) {
1278 error = ret;
1279 goto fail;
1280 } else
1281 cnx += ret;
1282 }
1283
41a49cc3
MN		 1284 /* found states are sorted for outbound processing */
1285 if (npols > 1)
1286 xfrm_state_sort(xfrm, tpp, cnx, family);
1287
4e81bb83
MN		 1288 return cnx;
1289
1290 fail:
1291 for (cnx--; cnx>=0; cnx--)
41a49cc3 1292 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1293 return error;
1294
1295}
1296
1da177e4
LT		 1297/* Check that the bundle accepts the flow and its components are
1298 * still valid.
1299 */
1300
1301static struct dst_entry *
1302xfrm_find_bundle(struct flowi *fl, struct xfrm_policy *policy, unsigned short family)
1303{
1304 struct dst_entry *x;
1305 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1306 if (unlikely(afinfo == NULL))
1307 return ERR_PTR(-EINVAL);
1308 x = afinfo->find_bundle(fl, policy);
1309 xfrm_policy_put_afinfo(afinfo);
1310 return x;
1311}
1312
25ee3286
HX		 1313static inline int xfrm_get_tos(struct flowi *fl, int family)
1314{
1315 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1316 int tos;
1da177e4 1317
25ee3286
HX		 1318 if (!afinfo)
1319 return -EINVAL;
1320
1321 tos = afinfo->get_tos(fl);
1322
1323 xfrm_policy_put_afinfo(afinfo);
1324
1325 return tos;
1326}
1327
1328static inline struct xfrm_dst *xfrm_alloc_dst(int family)
1da177e4 1329{
1da177e4 1330 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
25ee3286
HX		 1331 struct xfrm_dst *xdst;
1332
1333 if (!afinfo)
1334 return ERR_PTR(-EINVAL);
1335
1336 xdst = dst_alloc(afinfo->dst_ops) ?: ERR_PTR(-ENOBUFS);
1337
1338 xfrm_policy_put_afinfo(afinfo);
1339
1340 return xdst;
1341}
1342
a1b05140
MN		 1343static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1344 int nfheader_len)
1345{
1346 struct xfrm_policy_afinfo *afinfo =
1347 xfrm_policy_get_afinfo(dst->ops->family);
1348 int err;
1349
1350 if (!afinfo)
1351 return -EINVAL;
1352
1353 err = afinfo->init_path(path, dst, nfheader_len);
1354
1355 xfrm_policy_put_afinfo(afinfo);
1356
1357 return err;
1358}
1359
25ee3286
HX		 1360static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
1361{
1362 struct xfrm_policy_afinfo *afinfo =
1363 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1364 int err;
1365
1366 if (!afinfo)
1da177e4 1367 return -EINVAL;
25ee3286
HX		 1368
1369 err = afinfo->fill_dst(xdst, dev);
1370
1da177e4 1371 xfrm_policy_put_afinfo(afinfo);
25ee3286 1372
1da177e4
LT		 1373 return err;
1374}
1375
25ee3286
HX		 1376/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1377 * all the metrics... Shortly, bundle a bundle.
1378 */
1379
1380static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1381 struct xfrm_state **xfrm, int nx,
1382 struct flowi *fl,
1383 struct dst_entry *dst)
1384{
1385 unsigned long now = jiffies;
1386 struct net_device *dev;
1387 struct dst_entry *dst_prev = NULL;
1388 struct dst_entry *dst0 = NULL;
1389 int i = 0;
1390 int err;
1391 int header_len = 0;
a1b05140 1392 int nfheader_len = 0;
25ee3286
HX		 1393 int trailer_len = 0;
1394 int tos;
1395 int family = policy->selector.family;
9bb182a7
YH		 1396 xfrm_address_t saddr, daddr;
1397
1398 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1399
1400 tos = xfrm_get_tos(fl, family);
1401 err = tos;
1402 if (tos < 0)
1403 goto put_states;
1404
1405 dst_hold(dst);
1406
1407 for (; i < nx; i++) {
1408 struct xfrm_dst *xdst = xfrm_alloc_dst(family);
1409 struct dst_entry *dst1 = &xdst->u.dst;
1410
1411 err = PTR_ERR(xdst);
1412 if (IS_ERR(xdst)) {
1413 dst_release(dst);
1414 goto put_states;
1415 }
1416
1417 if (!dst_prev)
1418 dst0 = dst1;
1419 else {
1420 dst_prev->child = dst_clone(dst1);
1421 dst1->flags |= DST_NOHASH;
1422 }
1423
1424 xdst->route = dst;
1425 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1426
1427 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1428 family = xfrm[i]->props.family;
9bb182a7
YH		 1429 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1430 family);
25ee3286
HX		 1431 err = PTR_ERR(dst);
1432 if (IS_ERR(dst))
1433 goto put_states;
1434 } else
1435 dst_hold(dst);
1436
1437 dst1->xfrm = xfrm[i];
1438 xdst->genid = xfrm[i]->genid;
1439
1440 dst1->obsolete = -1;
1441 dst1->flags |= DST_HOST;
1442 dst1->lastuse = now;
1443
1444 dst1->input = dst_discard;
1445 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1446
1447 dst1->next = dst_prev;
1448 dst_prev = dst1;
1449
1450 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1451 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1452 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1453 trailer_len += xfrm[i]->props.trailer_len;
1454 }
1455
1456 dst_prev->child = dst;
1457 dst0->path = dst;
1458
1459 err = -ENODEV;
1460 dev = dst->dev;
1461 if (!dev)
1462 goto free_dst;
1463
1464 /* Copy neighbout for reachability confirmation */
1465 dst0->neighbour = neigh_clone(dst->neighbour);
1466
a1b05140 1467 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1468 xfrm_init_pmtu(dst_prev);
1469
1470 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1471 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1472
1473 err = xfrm_fill_dst(xdst, dev);
1474 if (err)
1475 goto free_dst;
1476
1477 dst_prev->header_len = header_len;
1478 dst_prev->trailer_len = trailer_len;
1479 header_len -= xdst->u.dst.xfrm->props.header_len;
1480 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1481 }
1482
1483out:
1484 return dst0;
1485
1486put_states:
1487 for (; i < nx; i++)
1488 xfrm_state_put(xfrm[i]);
1489free_dst:
1490 if (dst0)
1491 dst_free(dst0);
1492 dst0 = ERR_PTR(err);
1493 goto out;
1494}
1495
157bfc25
MN		 1496static int inline
1497xfrm_dst_alloc_copy(void **target, void *src, int size)
1498{
1499 if (!*target) {
1500 *target = kmalloc(size, GFP_ATOMIC);
1501 if (!*target)
1502 return -ENOMEM;
1503 }
1504 memcpy(*target, src, size);
1505 return 0;
1506}
1507
1508static int inline
1509xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1510{
1511#ifdef CONFIG_XFRM_SUB_POLICY
1512 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1513 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1514 sel, sizeof(*sel));
1515#else
1516 return 0;
1517#endif
1518}
1519
1520static int inline
1521xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1522{
1523#ifdef CONFIG_XFRM_SUB_POLICY
1524 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1525 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1526#else
1527 return 0;
1528#endif
1529}
1da177e4
LT		 1530
1531static int stale_bundle(struct dst_entry *dst);
1532
1533/* Main function: finds/creates a bundle for given flow.
1534 *
1535 * At the moment we eat a raw IP route. Mostly to speed up lookups
1536 * on interfaces with disabled IPsec.
1537 */
14e50e57
DM		 1538int __xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl,
1539 struct sock *sk, int flags)
1da177e4
LT		 1540{
1541 struct xfrm_policy *policy;
4e81bb83
MN		 1542 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1543 int npols;
1544 int pol_dead;
1545 int xfrm_nr;
1546 int pi;
1da177e4
LT		 1547 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1548 struct dst_entry *dst, *dst_orig = *dst_p;
1549 int nx = 0;
1550 int err;
1551 u32 genid;
42cf93cd 1552 u16 family;
df71837d 1553 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
e0d1caa7 1554
1da177e4
LT		 1555restart:
1556 genid = atomic_read(&flow_cache_genid);
1557 policy = NULL;
4e81bb83
MN		 1558 for (pi = 0; pi < ARRAY_SIZE(pols); pi++)
1559 pols[pi] = NULL;
1560 npols = 0;
1561 pol_dead = 0;
1562 xfrm_nr = 0;
1563
f7944fb1 1564 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
e0d1caa7 1565 policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
75b8c133 1566 err = PTR_ERR(policy);
0aa64774
MN		 1567 if (IS_ERR(policy)) {
1568 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
75b8c133 1569 goto dropdst;
0aa64774 1570 }
3bccfbc7 1571 }
1da177e4
LT		 1572
1573 if (!policy) {
1574 /* To accelerate a bit... */
2518c7c2 1575 if ((dst_orig->flags & DST_NOXFRM) ||
dc2caba7 1576 !init_net.xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 1577 goto nopol;
1da177e4 1578
e0d1caa7 1579 policy = flow_cache_lookup(fl, dst_orig->ops->family,
42cf93cd 1580 dir, xfrm_policy_lookup);
75b8c133 1581 err = PTR_ERR(policy);
d66e37a9
MN		 1582 if (IS_ERR(policy)) {
1583 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
75b8c133 1584 goto dropdst;
d66e37a9 1585 }
1da177e4
LT		 1586 }
1587
1588 if (!policy)
8b7817f3 1589 goto nopol;
1da177e4 1590
42cf93cd 1591 family = dst_orig->ops->family;
4e81bb83
MN		 1592 pols[0] = policy;
1593 npols ++;
1594 xfrm_nr += pols[0]->xfrm_nr;
1da177e4 1595
aef21785 1596 err = -ENOENT;
8b7817f3
HX		 1597 if ((flags & XFRM_LOOKUP_ICMP) && !(policy->flags & XFRM_POLICY_ICMP))
1598 goto error;
1599
1600 policy->curlft.use_time = get_seconds();
1601
1da177e4 1602 switch (policy->action) {
5e5234ff 1603 default:
1da177e4
LT		 1604 case XFRM_POLICY_BLOCK:
1605 /* Prohibit the flow */
0aa64774 1606 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 1607 err = -EPERM;
1608 goto error;
1da177e4
LT		 1609
1610 case XFRM_POLICY_ALLOW:
4e81bb83 1611#ifndef CONFIG_XFRM_SUB_POLICY
1da177e4
LT		 1612 if (policy->xfrm_nr == 0) {
1613 /* Flow passes not transformed. */
1614 xfrm_pol_put(policy);
1615 return 0;
1616 }
4e81bb83 1617#endif
1da177e4
LT		 1618
1619 /* Try to find matching bundle.
1620 *
1621 * LATER: help from flow cache. It is optional, this
1622 * is required only for output policy.
1623 */
1624 dst = xfrm_find_bundle(fl, policy, family);
1625 if (IS_ERR(dst)) {
0aa64774 1626 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
e104411b
PM		 1627 err = PTR_ERR(dst);
1628 goto error;
1da177e4
LT		 1629 }
1630
1631 if (dst)
1632 break;
1633
4e81bb83
MN		 1634#ifdef CONFIG_XFRM_SUB_POLICY
1635 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1636 pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN,
1637 fl, family,
1638 XFRM_POLICY_OUT);
1639 if (pols[1]) {
134b0fc5 1640 if (IS_ERR(pols[1])) {
0aa64774 1641 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
134b0fc5
JM		 1642 err = PTR_ERR(pols[1]);
1643 goto error;
1644 }
4e81bb83 1645 if (pols[1]->action == XFRM_POLICY_BLOCK) {
0aa64774 1646 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK);
4e81bb83
MN		 1647 err = -EPERM;
1648 goto error;
1649 }
1650 npols ++;
1651 xfrm_nr += pols[1]->xfrm_nr;
1652 }
1653 }
1654
1655 /*
1656 * Because neither flowi nor bundle information knows about
1657 * transformation template size. On more than one policy usage
1658 * we can realize whether all of them is bypass or not after
1659 * they are searched. See above not-transformed bypass
1660 * is surrounded by non-sub policy configuration, too.
1661 */
1662 if (xfrm_nr == 0) {
1663 /* Flow passes not transformed. */
1664 xfrm_pols_put(pols, npols);
1665 return 0;
1666 }
1667
1668#endif
1669 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1da177e4
LT		 1670
1671 if (unlikely(nx<0)) {
1672 err = nx;
14e50e57
DM		 1673 if (err == -EAGAIN && sysctl_xfrm_larval_drop) {
1674 /* EREMOTE tells the caller to generate
1675 * a one-shot blackhole route.
1676 */
d66e37a9 1677 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
14e50e57
DM		 1678 xfrm_pol_put(policy);
1679 return -EREMOTE;
1680 }
815f4e57 1681 if (err == -EAGAIN && (flags & XFRM_LOOKUP_WAIT)) {
1da177e4
LT		 1682 DECLARE_WAITQUEUE(wait, current);
1683
50a30657 1684 add_wait_queue(&init_net.xfrm.km_waitq, &wait);
1da177e4
LT		 1685 set_current_state(TASK_INTERRUPTIBLE);
1686 schedule();
1687 set_current_state(TASK_RUNNING);
50a30657 1688 remove_wait_queue(&init_net.xfrm.km_waitq, &wait);
1da177e4 1689
4e81bb83 1690 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1da177e4
LT		 1691
1692 if (nx == -EAGAIN && signal_pending(current)) {
0aa64774 1693 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1da177e4
LT		 1694 err = -ERESTART;
1695 goto error;
1696 }
1697 if (nx == -EAGAIN ||
1698 genid != atomic_read(&flow_cache_genid)) {
4e81bb83 1699 xfrm_pols_put(pols, npols);
1da177e4
LT		 1700 goto restart;
1701 }
1702 err = nx;
1703 }
0aa64774
MN		 1704 if (err < 0) {
1705 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1da177e4 1706 goto error;
0aa64774 1707 }
1da177e4
LT		 1708 }
1709 if (nx == 0) {
1710 /* Flow passes not transformed. */
4e81bb83 1711 xfrm_pols_put(pols, npols);
1da177e4
LT		 1712 return 0;
1713 }
1714
25ee3286
HX		 1715 dst = xfrm_bundle_create(policy, xfrm, nx, fl, dst_orig);
1716 err = PTR_ERR(dst);
0aa64774
MN		 1717 if (IS_ERR(dst)) {
1718 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1da177e4 1719 goto error;
0aa64774 1720 }
1da177e4 1721
4e81bb83
MN		 1722 for (pi = 0; pi < npols; pi++) {
1723 read_lock_bh(&pols[pi]->lock);
12a169e7 1724 pol_dead |= pols[pi]->walk.dead;
4e81bb83
MN		 1725 read_unlock_bh(&pols[pi]->lock);
1726 }
1727
1da177e4 1728 write_lock_bh(&policy->lock);
4e81bb83 1729 if (unlikely(pol_dead || stale_bundle(dst))) {
1da177e4
LT		 1730 /* Wow! While we worked on resolving, this
1731 * policy has gone. Retry. It is not paranoia,
1732 * we just cannot enlist new bundle to dead object.
1733 * We can't enlist stable bundles either.
1734 */
1735 write_unlock_bh(&policy->lock);
9d7d7402 1736 dst_free(dst);
00de651d 1737
0aa64774
MN		 1738 if (pol_dead)
1739 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD);
1740 else
1741 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
00de651d
HX		 1742 err = -EHOSTUNREACH;
1743 goto error;
1da177e4 1744 }
157bfc25
MN		 1745
1746 if (npols > 1)
1747 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1748 else
1749 err = xfrm_dst_update_origin(dst, fl);
1750 if (unlikely(err)) {
1751 write_unlock_bh(&policy->lock);
9d7d7402 1752 dst_free(dst);
0aa64774 1753 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
157bfc25
MN		 1754 goto error;
1755 }
1756
1da177e4
LT		 1757 dst->next = policy->bundles;
1758 policy->bundles = dst;
1759 dst_hold(dst);
1760 write_unlock_bh(&policy->lock);
1761 }
1762 *dst_p = dst;
1763 dst_release(dst_orig);
a716c119 1764 xfrm_pols_put(pols, npols);
1da177e4
LT		 1765 return 0;
1766
1767error:
4e81bb83 1768 xfrm_pols_put(pols, npols);
75b8c133
HX		 1769dropdst:
1770 dst_release(dst_orig);
1da177e4
LT		 1771 *dst_p = NULL;
1772 return err;
8b7817f3
HX		 1773
1774nopol:
aef21785 1775 err = -ENOENT;
8b7817f3
HX		 1776 if (flags & XFRM_LOOKUP_ICMP)
1777 goto dropdst;
1778 return 0;
1da177e4 1779}
14e50e57
DM		 1780EXPORT_SYMBOL(__xfrm_lookup);
1781
1782int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl,
1783 struct sock *sk, int flags)
1784{
1785 int err = __xfrm_lookup(dst_p, fl, sk, flags);
1786
1787 if (err == -EREMOTE) {
1788 dst_release(*dst_p);
1789 *dst_p = NULL;
1790 err = -EAGAIN;
1791 }
1792
1793 return err;
1794}
1da177e4
LT		 1795EXPORT_SYMBOL(xfrm_lookup);
1796
df0ba92a
MN		 1797static inline int
1798xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1799{
1800 struct xfrm_state *x;
df0ba92a
MN		 1801
1802 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1803 return 0;
1804 x = skb->sp->xvec[idx];
1805 if (!x->type->reject)
1806 return 0;
1ecafede 1807 return x->type->reject(x, skb, fl);
df0ba92a
MN		 1808}
1809
1da177e4
LT		 1810/* When skb is transformed back to its "native" form, we have to
1811 * check policy restrictions. At the moment we make this in maximally
1812 * stupid way. Shame on me. :-) Of course, connected sockets must
1813 * have policy cached at them.
1814 */
1815
1816static inline int
a716c119 1817xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1da177e4
LT		 1818 unsigned short family)
1819{
1820 if (xfrm_state_kern(x))
928ba416 1821 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 1822 return x->id.proto == tmpl->id.proto &&
1823 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1824 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1825 x->props.mode == tmpl->mode &&
c5d18e98 1826 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 1827 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 1828 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1829 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 1830}
1831
df0ba92a
MN		 1832/*
1833 * 0 or more than 0 is returned when validation is succeeded (either bypass
1834 * because of optional transport mode, or next index of the mathced secpath
1835 * state with the template.
1836 * -1 is returned when no matching template is found.
1837 * Otherwise "-2 - errored_index" is returned.
1838 */
1da177e4
LT		 1839static inline int
1840xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1841 unsigned short family)
1842{
1843 int idx = start;
1844
1845 if (tmpl->optional) {
7e49e6de 1846 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 1847 return start;
1848 } else
1849 start = -1;
1850 for (; idx < sp->len; idx++) {
dbe5b4aa 1851 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 1852 return ++idx;
df0ba92a
MN		 1853 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1854 if (start == -1)
1855 start = -2-idx;
1da177e4 1856 break;
df0ba92a 1857 }
1da177e4
LT		 1858 }
1859 return start;
1860}
1861
d5422efe
HX		 1862int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1863 unsigned int family, int reverse)
1da177e4
LT		 1864{
1865 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 1866 int err;
1da177e4
LT		 1867
1868 if (unlikely(afinfo == NULL))
1869 return -EAFNOSUPPORT;
1870
d5422efe 1871 afinfo->decode_session(skb, fl, reverse);
beb8d13b 1872 err = security_xfrm_decode_session(skb, &fl->secid);
1da177e4 1873 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 1874 return err;
1da177e4 1875}
d5422efe 1876EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 1877
df0ba92a 1878static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 1879{
1880 for (; k < sp->len; k++) {
df0ba92a 1881 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 1882 *idxp = k;
1da177e4 1883 return 1;
df0ba92a 1884 }
1da177e4
LT		 1885 }
1886
1887 return 0;
1888}
1889
a716c119 1890int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 1891 unsigned short family)
1892{
1893 struct xfrm_policy *pol;
4e81bb83
MN		 1894 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1895 int npols = 0;
1896 int xfrm_nr;
1897 int pi;
d5422efe 1898 int reverse;
1da177e4 1899 struct flowi fl;
d5422efe 1900 u8 fl_dir;
df0ba92a 1901 int xerr_idx = -1;
1da177e4 1902
d5422efe
HX		 1903 reverse = dir & ~XFRM_POLICY_MASK;
1904 dir &= XFRM_POLICY_MASK;
1905 fl_dir = policy_to_flow_dir(dir);
1906
0aa64774
MN		 1907 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
1908 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
1da177e4 1909 return 0;
0aa64774
MN		 1910 }
1911
eb9c7ebe 1912 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 1913
1914 /* First, check used SA against their selectors. */
1915 if (skb->sp) {
1916 int i;
1917
1918 for (i=skb->sp->len-1; i>=0; i--) {
dbe5b4aa 1919 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774
MN		 1920 if (!xfrm_selector_match(&x->sel, &fl, family)) {
1921 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 1922 return 0;
0aa64774 1923 }
1da177e4
LT		 1924 }
1925 }
1926
1927 pol = NULL;
3bccfbc7 1928 if (sk && sk->sk_policy[dir]) {
e0d1caa7 1929 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774
MN		 1930 if (IS_ERR(pol)) {
1931 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 1932 return 0;
0aa64774 1933 }
3bccfbc7 1934 }
1da177e4
LT		 1935
1936 if (!pol)
e0d1caa7 1937 pol = flow_cache_lookup(&fl, family, fl_dir,
1da177e4
LT		 1938 xfrm_policy_lookup);
1939
0aa64774
MN		 1940 if (IS_ERR(pol)) {
1941 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
134b0fc5 1942 return 0;
0aa64774 1943 }
134b0fc5 1944
df0ba92a 1945 if (!pol) {
d1d9facf 1946 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 1947 xfrm_secpath_reject(xerr_idx, skb, &fl);
0aa64774 1948 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 1949 return 0;
1950 }
1951 return 1;
1952 }
1da177e4 1953
9d729f72 1954 pol->curlft.use_time = get_seconds();
1da177e4 1955
4e81bb83
MN		 1956 pols[0] = pol;
1957 npols ++;
1958#ifdef CONFIG_XFRM_SUB_POLICY
1959 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1960 pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN,
1961 &fl, family,
1962 XFRM_POLICY_IN);
1963 if (pols[1]) {
0aa64774
MN		 1964 if (IS_ERR(pols[1])) {
1965 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
134b0fc5 1966 return 0;
0aa64774 1967 }
9d729f72 1968 pols[1]->curlft.use_time = get_seconds();
4e81bb83
MN		 1969 npols ++;
1970 }
1971 }
1972#endif
1973
1da177e4
LT		 1974 if (pol->action == XFRM_POLICY_ALLOW) {
1975 struct sec_path *sp;
1976 static struct sec_path dummy;
4e81bb83 1977 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 1978 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 1979 struct xfrm_tmpl **tpp = tp;
1980 int ti = 0;
1da177e4
LT		 1981 int i, k;
1982
1983 if ((sp = skb->sp) == NULL)
1984 sp = &dummy;
1985
4e81bb83
MN		 1986 for (pi = 0; pi < npols; pi++) {
1987 if (pols[pi] != pol &&
0aa64774
MN		 1988 pols[pi]->action != XFRM_POLICY_ALLOW) {
1989 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 1990 goto reject;
0aa64774
MN		 1991 }
1992 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
1993 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 1994 goto reject_error;
0aa64774 1995 }
4e81bb83
MN		 1996 for (i = 0; i < pols[pi]->xfrm_nr; i++)
1997 tpp[ti++] = &pols[pi]->xfrm_vec[i];
1998 }
1999 xfrm_nr = ti;
41a49cc3
MN		 2000 if (npols > 1) {
2001 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2002 tpp = stp;
2003 }
4e81bb83 2004
1da177e4
LT		 2005 /* For each tunnel xfrm, find the first matching tmpl.
2006 * For each tmpl before that, find corresponding xfrm.
2007 * Order is _important_. Later we will implement
2008 * some barriers, but at the moment barriers
2009 * are implied between each two transformations.
2010 */
4e81bb83
MN		 2011 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2012 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2013 if (k < 0) {
d1d9facf
JM		 2014 if (k < -1)
2015 /* "-2 - errored_index" returned */
2016 xerr_idx = -(2+k);
0aa64774 2017 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2018 goto reject;
df0ba92a 2019 }
1da177e4
LT		 2020 }
2021
0aa64774
MN		 2022 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2023 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2024 goto reject;
0aa64774 2025 }
1da177e4 2026
4e81bb83 2027 xfrm_pols_put(pols, npols);
1da177e4
LT		 2028 return 1;
2029 }
0aa64774 2030 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2031
2032reject:
df0ba92a 2033 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2034reject_error:
2035 xfrm_pols_put(pols, npols);
1da177e4
LT		 2036 return 0;
2037}
2038EXPORT_SYMBOL(__xfrm_policy_check);
2039
2040int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2041{
2042 struct flowi fl;
2043
0aa64774
MN		 2044 if (xfrm_decode_session(skb, &fl, family) < 0) {
2045 /* XXX: we should have something like FWDHDRERROR here. */
2046 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
1da177e4 2047 return 0;
0aa64774 2048 }
1da177e4
LT		 2049
2050 return xfrm_lookup(&skb->dst, &fl, NULL, 0) == 0;
2051}
2052EXPORT_SYMBOL(__xfrm_route_forward);
2053
d49c73c7
DM		 2054/* Optimize later using cookies and generation ids. */
2055
1da177e4
LT		 2056static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2057{
d49c73c7
DM		 2058 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2059 * to "-1" to force all XFRM destinations to get validated by
2060 * dst_ops->check on every use. We do this because when a
2061 * normal route referenced by an XFRM dst is obsoleted we do
2062 * not go looking around for all parent referencing XFRM dsts
2063 * so that we can invalidate them. It is just too much work.
2064 * Instead we make the checks here on every use. For example:
2065 *
2066 * XFRM dst A --> IPv4 dst X
2067 *
2068 * X is the "xdst->route" of A (X is also the "dst->path" of A
2069 * in this example). If X is marked obsolete, "A" will not
2070 * notice. That's what we are validating here via the
2071 * stale_bundle() check.
2072 *
2073 * When a policy's bundle is pruned, we dst_free() the XFRM
2074 * dst which causes it's ->obsolete field to be set to a
2075 * positive non-zero integer. If an XFRM dst has been pruned
2076 * like this, we want to force a new route lookup.
399c180a 2077 */
d49c73c7
DM		 2078 if (dst->obsolete < 0 && !stale_bundle(dst))
2079 return dst;
2080
1da177e4
LT		 2081 return NULL;
2082}
2083
2084static int stale_bundle(struct dst_entry *dst)
2085{
5b368e61 2086 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
1da177e4
LT		 2087}
2088
aabc9761 2089void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2090{
1da177e4 2091 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2092 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2093 dev_hold(dst->dev);
1da177e4
LT		 2094 dev_put(dev);
2095 }
2096}
aabc9761 2097EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2098
2099static void xfrm_link_failure(struct sk_buff *skb)
2100{
2101 /* Impossible. Such dst must be popped before reaches point of failure. */
2102 return;
2103}
2104
2105static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2106{
2107 if (dst) {
2108 if (dst->obsolete) {
2109 dst_release(dst);
2110 dst = NULL;
2111 }
2112 }
2113 return dst;
2114}
2115
2518c7c2
DM		 2116static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p)
2117{
2118 struct dst_entry *dst, **dstp;
2119
2120 write_lock(&pol->lock);
2121 dstp = &pol->bundles;
2122 while ((dst=*dstp) != NULL) {
2123 if (func(dst)) {
2124 *dstp = dst->next;
2125 dst->next = *gc_list_p;
2126 *gc_list_p = dst;
2127 } else {
2128 dstp = &dst->next;
2129 }
2130 }
2131 write_unlock(&pol->lock);
2132}
2133
1da177e4
LT		 2134static void xfrm_prune_bundles(int (*func)(struct dst_entry *))
2135{
2518c7c2
DM		 2136 struct dst_entry *gc_list = NULL;
2137 int dir;
1da177e4
LT		 2138
2139 read_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 2140 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2141 struct xfrm_policy *pol;
2142 struct hlist_node *entry;
2143 struct hlist_head *table;
2144 int i;
4e81bb83 2145
2518c7c2 2146 hlist_for_each_entry(pol, entry,
8b18f8ea 2147 &init_net.xfrm.policy_inexact[dir], bydst)
2518c7c2
DM		 2148 prune_one_bundle(pol, func, &gc_list);
2149
a35f6c5d
AD		 2150 table = init_net.xfrm.policy_bydst[dir].table;
2151 for (i = init_net.xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2
DM		 2152 hlist_for_each_entry(pol, entry, table + i, bydst)
2153 prune_one_bundle(pol, func, &gc_list);
1da177e4
LT		 2154 }
2155 }
2156 read_unlock_bh(&xfrm_policy_lock);
2157
2158 while (gc_list) {
2518c7c2 2159 struct dst_entry *dst = gc_list;
1da177e4
LT		 2160 gc_list = dst->next;
2161 dst_free(dst);
2162 }
2163}
2164
2165static int unused_bundle(struct dst_entry *dst)
2166{
2167 return !atomic_read(&dst->__refcnt);
2168}
2169
2170static void __xfrm_garbage_collect(void)
2171{
2172 xfrm_prune_bundles(unused_bundle);
2173}
2174
1c095399 2175static int xfrm_flush_bundles(void)
1da177e4
LT		 2176{
2177 xfrm_prune_bundles(stale_bundle);
2178 return 0;
2179}
2180
25ee3286 2181static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2182{
2183 do {
2184 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2185 u32 pmtu, route_mtu_cached;
2186
2187 pmtu = dst_mtu(dst->child);
2188 xdst->child_mtu_cached = pmtu;
2189
2190 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2191
2192 route_mtu_cached = dst_mtu(xdst->route);
2193 xdst->route_mtu_cached = route_mtu_cached;
2194
2195 if (pmtu > route_mtu_cached)
2196 pmtu = route_mtu_cached;
2197
2198 dst->metrics[RTAX_MTU-1] = pmtu;
2199 } while ((dst = dst->next));
2200}
2201
1da177e4
LT		 2202/* Check that the bundle accepts the flow and its components are
2203 * still valid.
2204 */
2205
5b368e61
VY		 2206int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2207 struct flowi *fl, int family, int strict)
1da177e4
LT		 2208{
2209 struct dst_entry *dst = &first->u.dst;
2210 struct xfrm_dst *last;
2211 u32 mtu;
2212
92d63dec 2213 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2214 (dst->dev && !netif_running(dst->dev)))
2215 return 0;
157bfc25
MN		 2216#ifdef CONFIG_XFRM_SUB_POLICY
2217 if (fl) {
2218 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2219 return 0;
2220 if (first->partner &&
2221 !xfrm_selector_match(first->partner, fl, family))
2222 return 0;
2223 }
2224#endif
1da177e4
LT		 2225
2226 last = NULL;
2227
2228 do {
2229 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2230
2231 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2232 return 0;
67f83cbf
VY		 2233 if (fl && pol &&
2234 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
e0d1caa7 2235 return 0;
1da177e4
LT		 2236 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2237 return 0;
9d4a706d
DM		 2238 if (xdst->genid != dst->xfrm->genid)
2239 return 0;
e53820de 2240
1bfcb10f 2241 if (strict && fl &&
13996378 2242 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
e53820de
MN		 2243 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2244 return 0;
1da177e4
LT		 2245
2246 mtu = dst_mtu(dst->child);
2247 if (xdst->child_mtu_cached != mtu) {
2248 last = xdst;
2249 xdst->child_mtu_cached = mtu;
2250 }
2251
92d63dec 2252 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2253 return 0;
2254 mtu = dst_mtu(xdst->route);
2255 if (xdst->route_mtu_cached != mtu) {
2256 last = xdst;
2257 xdst->route_mtu_cached = mtu;
2258 }
2259
2260 dst = dst->child;
2261 } while (dst->xfrm);
2262
2263 if (likely(!last))
2264 return 1;
2265
2266 mtu = last->child_mtu_cached;
2267 for (;;) {
2268 dst = &last->u.dst;
2269
2270 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2271 if (mtu > last->route_mtu_cached)
2272 mtu = last->route_mtu_cached;
2273 dst->metrics[RTAX_MTU-1] = mtu;
2274
2275 if (last == first)
2276 break;
2277
bd0bf076 2278 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2279 last->child_mtu_cached = mtu;
2280 }
2281
2282 return 1;
2283}
2284
2285EXPORT_SYMBOL(xfrm_bundle_ok);
2286
1da177e4
LT		 2287int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2288{
2289 int err = 0;
2290 if (unlikely(afinfo == NULL))
2291 return -EINVAL;
2292 if (unlikely(afinfo->family >= NPROTO))
2293 return -EAFNOSUPPORT;
e959d812 2294 write_lock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2295 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2296 err = -ENOBUFS;
2297 else {
2298 struct dst_ops *dst_ops = afinfo->dst_ops;
2299 if (likely(dst_ops->kmem_cachep == NULL))
2300 dst_ops->kmem_cachep = xfrm_dst_cache;
2301 if (likely(dst_ops->check == NULL))
2302 dst_ops->check = xfrm_dst_check;
1da177e4
LT		 2303 if (likely(dst_ops->negative_advice == NULL))
2304 dst_ops->negative_advice = xfrm_negative_advice;
2305 if (likely(dst_ops->link_failure == NULL))
2306 dst_ops->link_failure = xfrm_link_failure;
1da177e4
LT		 2307 if (likely(afinfo->garbage_collect == NULL))
2308 afinfo->garbage_collect = __xfrm_garbage_collect;
2309 xfrm_policy_afinfo[afinfo->family] = afinfo;
2310 }
e959d812 2311 write_unlock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2312 return err;
2313}
2314EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2315
2316int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2317{
2318 int err = 0;
2319 if (unlikely(afinfo == NULL))
2320 return -EINVAL;
2321 if (unlikely(afinfo->family >= NPROTO))
2322 return -EAFNOSUPPORT;
e959d812 2323 write_lock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2324 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2325 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2326 err = -EINVAL;
2327 else {
2328 struct dst_ops *dst_ops = afinfo->dst_ops;
2329 xfrm_policy_afinfo[afinfo->family] = NULL;
2330 dst_ops->kmem_cachep = NULL;
2331 dst_ops->check = NULL;
1da177e4
LT		 2332 dst_ops->negative_advice = NULL;
2333 dst_ops->link_failure = NULL;
1da177e4
LT		 2334 afinfo->garbage_collect = NULL;
2335 }
2336 }
e959d812 2337 write_unlock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2338 return err;
2339}
2340EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2341
2342static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2343{
2344 struct xfrm_policy_afinfo *afinfo;
2345 if (unlikely(family >= NPROTO))
2346 return NULL;
2347 read_lock(&xfrm_policy_afinfo_lock);
2348 afinfo = xfrm_policy_afinfo[family];
546be240
HX		 2349 if (unlikely(!afinfo))
2350 read_unlock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2351 return afinfo;
2352}
2353
2354static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2355{
546be240
HX		 2356 read_unlock(&xfrm_policy_afinfo_lock);
2357}
2358
1da177e4
LT		 2359static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2360{
e9dc8653
EB		 2361 struct net_device *dev = ptr;
2362
721499e8 2363 if (!net_eq(dev_net(dev), &init_net))
e9dc8653
EB		 2364 return NOTIFY_DONE;
2365
1da177e4
LT		 2366 switch (event) {
2367 case NETDEV_DOWN:
2368 xfrm_flush_bundles();
2369 }
2370 return NOTIFY_DONE;
2371}
2372
2373static struct notifier_block xfrm_dev_notifier = {
d5917a35 2374 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2375};
2376
558f82ef
MN		 2377#ifdef CONFIG_XFRM_STATISTICS
2378static int __init xfrm_statistics_init(void)
2379{
2380 if (snmp_mib_init((void **)xfrm_statistics,
2381 sizeof(struct linux_xfrm_mib)) < 0)
2382 return -ENOMEM;
2383 return 0;
2384}
2385#endif
2386
d62ddc21 2387static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2388{
2518c7c2
DM		 2389 unsigned int hmask, sz;
2390 int dir;
2391
d62ddc21
AD		 2392 if (net_eq(net, &init_net))
2393 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2394 sizeof(struct xfrm_dst),
e5d679f3 2395 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2396 NULL);
1da177e4 2397
2518c7c2
DM		 2398 hmask = 8 - 1;
2399 sz = (hmask+1) * sizeof(struct hlist_head);
2400
93b851c1
AD		 2401 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2402 if (!net->xfrm.policy_byidx)
2403 goto out_byidx;
8100bea7 2404 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2405
2406 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2407 struct xfrm_policy_hash *htab;
2408
dc2caba7 2409 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2410 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2411
a35f6c5d 2412 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2413 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2414 if (!htab->table)
a35f6c5d
AD		 2415 goto out_bydst;
2416 htab->hmask = hmask;
2518c7c2
DM		 2417 }
2418
adfcf0b2 2419 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2420 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2421 if (net_eq(net, &init_net))
2422 register_netdevice_notifier(&xfrm_dev_notifier);
2423 return 0;
93b851c1 2424
a35f6c5d
AD		 2425out_bydst:
2426 for (dir--; dir >= 0; dir--) {
2427 struct xfrm_policy_hash *htab;
2428
2429 htab = &net->xfrm.policy_bydst[dir];
2430 xfrm_hash_free(htab->table, sz);
2431 }
2432 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2433out_byidx:
2434 return -ENOMEM;
d62ddc21
AD		 2435}
2436
2437static void xfrm_policy_fini(struct net *net)
2438{
93b851c1 2439 unsigned int sz;
8b18f8ea 2440 int dir;
93b851c1 2441
adfcf0b2 2442 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2443
8b18f8ea 2444 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2445 struct xfrm_policy_hash *htab;
2446
8b18f8ea 2447 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2448
2449 htab = &net->xfrm.policy_bydst[dir];
2450 sz = (htab->hmask + 1);
2451 WARN_ON(!hlist_empty(htab->table));
2452 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2453 }
2454
8100bea7 2455 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2456 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2457 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2458}
2459
d62ddc21
AD		 2460static int __net_init xfrm_net_init(struct net *net)
2461{
2462 int rv;
2463
2464 rv = xfrm_state_init(net);
2465 if (rv < 0)
2466 goto out_state;
2467 rv = xfrm_policy_init(net);
2468 if (rv < 0)
2469 goto out_policy;
2470 return 0;
2471
2472out_policy:
2473 xfrm_state_fini(net);
2474out_state:
2475 return rv;
2476}
2477
2478static void __net_exit xfrm_net_exit(struct net *net)
2479{
2480 xfrm_policy_fini(net);
2481 xfrm_state_fini(net);
2482}
2483
2484static struct pernet_operations __net_initdata xfrm_net_ops = {
2485 .init = xfrm_net_init,
2486 .exit = xfrm_net_exit,
2487};
2488
1da177e4
LT		 2489void __init xfrm_init(void)
2490{
d62ddc21 2491 register_pernet_subsys(&xfrm_net_ops);
558f82ef
MN		 2492#ifdef CONFIG_XFRM_STATISTICS
2493 xfrm_statistics_init();
2494#endif
1da177e4 2495 xfrm_input_init();
558f82ef
MN		 2496#ifdef CONFIG_XFRM_STATISTICS
2497 xfrm_proc_init();
2498#endif
1da177e4
LT		 2499}
2500
ab5f5e8b 2501#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2502static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2503 struct audit_buffer *audit_buf)
ab5f5e8b 2504{
875179fa
PM		 2505 struct xfrm_sec_ctx *ctx = xp->security;
2506 struct xfrm_selector *sel = &xp->selector;
2507
2508 if (ctx)
ab5f5e8b 2509 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2510 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2511
875179fa 2512 switch(sel->family) {
ab5f5e8b 2513 case AF_INET:
21454aaa 2514 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2515 if (sel->prefixlen_s != 32)
2516 audit_log_format(audit_buf, " src_prefixlen=%d",
2517 sel->prefixlen_s);
21454aaa 2518 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2519 if (sel->prefixlen_d != 32)
2520 audit_log_format(audit_buf, " dst_prefixlen=%d",
2521 sel->prefixlen_d);
ab5f5e8b
JL		 2522 break;
2523 case AF_INET6:
5b095d98 2524 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 2525 if (sel->prefixlen_s != 128)
2526 audit_log_format(audit_buf, " src_prefixlen=%d",
2527 sel->prefixlen_s);
5b095d98 2528 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 2529 if (sel->prefixlen_d != 128)
2530 audit_log_format(audit_buf, " dst_prefixlen=%d",
2531 sel->prefixlen_d);
ab5f5e8b
JL		 2532 break;
2533 }
2534}
2535
68277acc 2536void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2532386f 2537 uid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 2538{
2539 struct audit_buffer *audit_buf;
ab5f5e8b 2540
afeb14b4 2541 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 2542 if (audit_buf == NULL)
2543 return;
2532386f 2544 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 2545 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 2546 xfrm_audit_common_policyinfo(xp, audit_buf);
2547 audit_log_end(audit_buf);
2548}
2549EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2550
68277acc 2551void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2532386f 2552 uid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 2553{
2554 struct audit_buffer *audit_buf;
ab5f5e8b 2555
afeb14b4 2556 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 2557 if (audit_buf == NULL)
2558 return;
2532386f 2559 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 2560 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 2561 xfrm_audit_common_policyinfo(xp, audit_buf);
2562 audit_log_end(audit_buf);
2563}
2564EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2565#endif
2566
80c9abaa
SS		 2567#ifdef CONFIG_XFRM_MIGRATE
2568static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2569 struct xfrm_selector *sel_tgt)
2570{
2571 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2572 if (sel_tgt->family == sel_cmp->family &&
2573 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
a716c119 2574 sel_cmp->family) == 0 &&
80c9abaa
SS		 2575 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2576 sel_cmp->family) == 0 &&
2577 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2578 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2579 return 1;
2580 }
2581 } else {
2582 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2583 return 1;
2584 }
2585 }
2586 return 0;
2587}
2588
2589static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2590 u8 dir, u8 type)
2591{
2592 struct xfrm_policy *pol, *ret = NULL;
2593 struct hlist_node *entry;
2594 struct hlist_head *chain;
2595 u32 priority = ~0U;
2596
2597 read_lock_bh(&xfrm_policy_lock);
2598 chain = policy_hash_direct(&sel->daddr, &sel->saddr, sel->family, dir);
2599 hlist_for_each_entry(pol, entry, chain, bydst) {
2600 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2601 pol->type == type) {
2602 ret = pol;
2603 priority = ret->priority;
2604 break;
2605 }
2606 }
8b18f8ea 2607 chain = &init_net.xfrm.policy_inexact[dir];
80c9abaa
SS		 2608 hlist_for_each_entry(pol, entry, chain, bydst) {
2609 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2610 pol->type == type &&
2611 pol->priority < priority) {
2612 ret = pol;
2613 break;
2614 }
2615 }
2616
2617 if (ret)
2618 xfrm_pol_hold(ret);
2619
2620 read_unlock_bh(&xfrm_policy_lock);
2621
2622 return ret;
2623}
2624
2625static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2626{
2627 int match = 0;
2628
2629 if (t->mode == m->mode && t->id.proto == m->proto &&
2630 (m->reqid == 0 || t->reqid == m->reqid)) {
2631 switch (t->mode) {
2632 case XFRM_MODE_TUNNEL:
2633 case XFRM_MODE_BEET:
2634 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2635 m->old_family) == 0 &&
2636 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2637 m->old_family) == 0) {
2638 match = 1;
2639 }
2640 break;
2641 case XFRM_MODE_TRANSPORT:
2642 /* in case of transport mode, template does not store
2643 any IP addresses, hence we just compare mode and
2644 protocol */
2645 match = 1;
2646 break;
2647 default:
2648 break;
2649 }
2650 }
2651 return match;
2652}
2653
2654/* update endpoint address(es) of template(s) */
2655static int xfrm_policy_migrate(struct xfrm_policy *pol,
2656 struct xfrm_migrate *m, int num_migrate)
2657{
2658 struct xfrm_migrate *mp;
2659 struct dst_entry *dst;
2660 int i, j, n = 0;
2661
2662 write_lock_bh(&pol->lock);
12a169e7 2663 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 2664 /* target policy has been deleted */
2665 write_unlock_bh(&pol->lock);
2666 return -ENOENT;
2667 }
2668
2669 for (i = 0; i < pol->xfrm_nr; i++) {
2670 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2671 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2672 continue;
2673 n++;
1bfcb10f
HX		 2674 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2675 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 2676 continue;
2677 /* update endpoints */
2678 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2679 sizeof(pol->xfrm_vec[i].id.daddr));
2680 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2681 sizeof(pol->xfrm_vec[i].saddr));
2682 pol->xfrm_vec[i].encap_family = mp->new_family;
2683 /* flush bundles */
2684 while ((dst = pol->bundles) != NULL) {
2685 pol->bundles = dst->next;
2686 dst_free(dst);
2687 }
2688 }
2689 }
2690
2691 write_unlock_bh(&pol->lock);
2692
2693 if (!n)
2694 return -ENODATA;
2695
2696 return 0;
2697}
2698
2699static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2700{
2701 int i, j;
2702
2703 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2704 return -EINVAL;
2705
2706 for (i = 0; i < num_migrate; i++) {
2707 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2708 m[i].old_family) == 0) &&
2709 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2710 m[i].old_family) == 0))
2711 return -EINVAL;
2712 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2713 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2714 return -EINVAL;
2715
2716 /* check if there is any duplicated entry */
2717 for (j = i + 1; j < num_migrate; j++) {
2718 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2719 sizeof(m[i].old_daddr)) &&
2720 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2721 sizeof(m[i].old_saddr)) &&
2722 m[i].proto == m[j].proto &&
2723 m[i].mode == m[j].mode &&
2724 m[i].reqid == m[j].reqid &&
2725 m[i].old_family == m[j].old_family)
2726 return -EINVAL;
2727 }
2728 }
2729
2730 return 0;
2731}
2732
2733int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189
AE		 2734 struct xfrm_migrate *m, int num_migrate,
2735 struct xfrm_kmaddress *k)
80c9abaa
SS		 2736{
2737 int i, err, nx_cur = 0, nx_new = 0;
2738 struct xfrm_policy *pol = NULL;
2739 struct xfrm_state *x, *xc;
2740 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2741 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2742 struct xfrm_migrate *mp;
2743
2744 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2745 goto out;
2746
2747 /* Stage 1 - find policy */
2748 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2749 err = -ENOENT;
2750 goto out;
2751 }
2752
2753 /* Stage 2 - find and update state(s) */
2754 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2755 if ((x = xfrm_migrate_state_find(mp))) {
2756 x_cur[nx_cur] = x;
2757 nx_cur++;
2758 if ((xc = xfrm_state_migrate(x, mp))) {
2759 x_new[nx_new] = xc;
2760 nx_new++;
2761 } else {
2762 err = -ENODATA;
2763 goto restore_state;
2764 }
2765 }
2766 }
2767
2768 /* Stage 3 - update policy */
2769 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2770 goto restore_state;
2771
2772 /* Stage 4 - delete old state(s) */
2773 if (nx_cur) {
2774 xfrm_states_put(x_cur, nx_cur);
2775 xfrm_states_delete(x_cur, nx_cur);
2776 }
2777
2778 /* Stage 5 - announce */
13c1d189 2779 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 2780
2781 xfrm_pol_put(pol);
2782
2783 return 0;
2784out:
2785 return err;
2786
2787restore_state:
2788 if (pol)
2789 xfrm_pol_put(pol);
2790 if (nx_cur)
2791 xfrm_states_put(x_cur, nx_cur);
2792 if (nx_new)
2793 xfrm_states_delete(x_new, nx_new);
2794
2795 return err;
2796}
e610e679 2797EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 2798#endif
Linux kernel - Deliverables
This page took 0.808613 seconds and 5 git commands to generate.