projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
netns xfrm: ->get_saddr in netns
[GitHub/mt8127/android_kernel_alcatel_ttab.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
c5b3cf46 95static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
9bb182a7
YH		 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
c5b3cf46 107 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 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{
c5b3cf46 119 struct net *net = xs_net(x);
66cdb3ca
HX		 120 xfrm_address_t *saddr = &x->props.saddr;
121 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 122 struct dst_entry *dst;
123
9bb182a7 124 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 125 saddr = x->coaddr;
9bb182a7
YH		 126 daddr = prev_daddr;
127 }
128 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
129 saddr = prev_saddr;
66cdb3ca 130 daddr = x->coaddr;
9bb182a7 131 }
1da177e4 132
c5b3cf46 133 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 134
135 if (!IS_ERR(dst)) {
136 if (prev_saddr != saddr)
137 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
138 if (prev_daddr != daddr)
139 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
140 }
1da177e4 141
66cdb3ca 142 return dst;
1da177e4 143}
1da177e4 144
1da177e4
LT		 145static inline unsigned long make_jiffies(long secs)
146{
147 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
148 return MAX_SCHEDULE_TIMEOUT-1;
149 else
a716c119 150 return secs*HZ;
1da177e4
LT		 151}
152
153static void xfrm_policy_timer(unsigned long data)
154{
155 struct xfrm_policy *xp = (struct xfrm_policy*)data;
9d729f72 156 unsigned long now = get_seconds();
1da177e4
LT		 157 long next = LONG_MAX;
158 int warn = 0;
159 int dir;
160
161 read_lock(&xp->lock);
162
12a169e7 163 if (xp->walk.dead)
1da177e4
LT		 164 goto out;
165
77d8d7a6 166 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 167
168 if (xp->lft.hard_add_expires_seconds) {
169 long tmo = xp->lft.hard_add_expires_seconds +
170 xp->curlft.add_time - now;
171 if (tmo <= 0)
172 goto expired;
173 if (tmo < next)
174 next = tmo;
175 }
176 if (xp->lft.hard_use_expires_seconds) {
177 long tmo = xp->lft.hard_use_expires_seconds +
178 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
179 if (tmo <= 0)
180 goto expired;
181 if (tmo < next)
182 next = tmo;
183 }
184 if (xp->lft.soft_add_expires_seconds) {
185 long tmo = xp->lft.soft_add_expires_seconds +
186 xp->curlft.add_time - now;
187 if (tmo <= 0) {
188 warn = 1;
189 tmo = XFRM_KM_TIMEOUT;
190 }
191 if (tmo < next)
192 next = tmo;
193 }
194 if (xp->lft.soft_use_expires_seconds) {
195 long tmo = xp->lft.soft_use_expires_seconds +
196 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
197 if (tmo <= 0) {
198 warn = 1;
199 tmo = XFRM_KM_TIMEOUT;
200 }
201 if (tmo < next)
202 next = tmo;
203 }
204
205 if (warn)
6c5c8ca7 206 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 207 if (next != LONG_MAX &&
208 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
209 xfrm_pol_hold(xp);
210
211out:
212 read_unlock(&xp->lock);
213 xfrm_pol_put(xp);
214 return;
215
216expired:
217 read_unlock(&xp->lock);
4666faab 218 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 219 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 220 xfrm_pol_put(xp);
221}
222
223
224/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
225 * SPD calls.
226 */
227
0331b1f3 228struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 229{
230 struct xfrm_policy *policy;
231
0da974f4 232 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 233
234 if (policy) {
0331b1f3 235 write_pnet(&policy->xp_net, net);
12a169e7 236 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 237 INIT_HLIST_NODE(&policy->bydst);
238 INIT_HLIST_NODE(&policy->byidx);
1da177e4 239 rwlock_init(&policy->lock);
2518c7c2 240 atomic_set(&policy->refcnt, 1);
b24b8a24
PE		 241 setup_timer(&policy->timer, xfrm_policy_timer,
242 (unsigned long)policy);
1da177e4
LT		 243 }
244 return policy;
245}
246EXPORT_SYMBOL(xfrm_policy_alloc);
247
248/* Destroy xfrm_policy: descendant resources must be released to this moment. */
249
64c31b3f 250void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 251{
12a169e7 252 BUG_ON(!policy->walk.dead);
1da177e4 253
09a62660 254 BUG_ON(policy->bundles);
1da177e4
LT		 255
256 if (del_timer(&policy->timer))
257 BUG();
258
03e1ad7b 259 security_xfrm_policy_free(policy->security);
1da177e4
LT		 260 kfree(policy);
261}
64c31b3f 262EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4
LT		 263
264static void xfrm_policy_gc_kill(struct xfrm_policy *policy)
265{
266 struct dst_entry *dst;
267
268 while ((dst = policy->bundles) != NULL) {
269 policy->bundles = dst->next;
270 dst_free(dst);
271 }
272
273 if (del_timer(&policy->timer))
274 atomic_dec(&policy->refcnt);
275
276 if (atomic_read(&policy->refcnt) > 1)
277 flow_cache_flush();
278
279 xfrm_pol_put(policy);
280}
281
c4028958 282static void xfrm_policy_gc_task(struct work_struct *work)
1da177e4
LT		 283{
284 struct xfrm_policy *policy;
2518c7c2
DM		 285 struct hlist_node *entry, *tmp;
286 struct hlist_head gc_list;
1da177e4
LT		 287
288 spin_lock_bh(&xfrm_policy_gc_lock);
2518c7c2
DM		 289 gc_list.first = xfrm_policy_gc_list.first;
290 INIT_HLIST_HEAD(&xfrm_policy_gc_list);
1da177e4
LT		 291 spin_unlock_bh(&xfrm_policy_gc_lock);
292
2518c7c2 293 hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst)
1da177e4 294 xfrm_policy_gc_kill(policy);
1da177e4 295}
c9583969 296static DECLARE_WORK(xfrm_policy_gc_work, xfrm_policy_gc_task);
1da177e4
LT		 297
298/* Rule must be locked. Release descentant resources, announce
299 * entry dead. The rule must be unlinked from lists to the moment.
300 */
301
302static void xfrm_policy_kill(struct xfrm_policy *policy)
303{
304 int dead;
305
306 write_lock_bh(&policy->lock);
12a169e7
HX		 307 dead = policy->walk.dead;
308 policy->walk.dead = 1;
1da177e4
LT		 309 write_unlock_bh(&policy->lock);
310
311 if (unlikely(dead)) {
312 WARN_ON(1);
313 return;
314 }
315
bbb770e7 316 spin_lock_bh(&xfrm_policy_gc_lock);
2518c7c2 317 hlist_add_head(&policy->bydst, &xfrm_policy_gc_list);
bbb770e7 318 spin_unlock_bh(&xfrm_policy_gc_lock);
1da177e4
LT		 319
320 schedule_work(&xfrm_policy_gc_work);
321}
322
2518c7c2
DM		 323static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
324
e92303f8 325static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 326{
e92303f8 327 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 328}
329
1121994c 330static struct hlist_head *policy_hash_bysel(struct net *net, struct xfrm_selector *sel, unsigned short family, int dir)
2518c7c2 331{
1121994c 332 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 333 unsigned int hash = __sel_hash(sel, family, hmask);
334
335 return (hash == hmask + 1 ?
1121994c
AD		 336 &net->xfrm.policy_inexact[dir] :
337 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 338}
339
1121994c 340static struct hlist_head *policy_hash_direct(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
2518c7c2 341{
1121994c 342 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 343 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
344
1121994c 345 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 346}
347
2518c7c2
DM		 348static void xfrm_dst_hash_transfer(struct hlist_head *list,
349 struct hlist_head *ndsttable,
350 unsigned int nhashmask)
351{
b791160b 352 struct hlist_node *entry, *tmp, *entry0 = NULL;
2518c7c2 353 struct xfrm_policy *pol;
b791160b 354 unsigned int h0 = 0;
2518c7c2 355
b791160b 356redo:
2518c7c2
DM		 357 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
358 unsigned int h;
359
360 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
361 pol->family, nhashmask);
b791160b
YH		 362 if (!entry0) {
363 hlist_del(entry);
364 hlist_add_head(&pol->bydst, ndsttable+h);
365 h0 = h;
366 } else {
367 if (h != h0)
368 continue;
369 hlist_del(entry);
370 hlist_add_after(entry0, &pol->bydst);
371 }
372 entry0 = entry;
373 }
374 if (!hlist_empty(list)) {
375 entry0 = NULL;
376 goto redo;
2518c7c2
DM		 377 }
378}
379
380static void xfrm_idx_hash_transfer(struct hlist_head *list,
381 struct hlist_head *nidxtable,
382 unsigned int nhashmask)
383{
384 struct hlist_node *entry, *tmp;
385 struct xfrm_policy *pol;
386
387 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
388 unsigned int h;
389
390 h = __idx_hash(pol->index, nhashmask);
391 hlist_add_head(&pol->byidx, nidxtable+h);
392 }
393}
394
395static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
396{
397 return ((old_hmask + 1) << 1) - 1;
398}
399
66caf628 400static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 401{
66caf628 402 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 403 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
404 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 405 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 406 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 407 int i;
408
409 if (!ndst)
410 return;
411
412 write_lock_bh(&xfrm_policy_lock);
413
414 for (i = hmask; i >= 0; i--)
415 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
416
66caf628
AD		 417 net->xfrm.policy_bydst[dir].table = ndst;
418 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2
DM		 419
420 write_unlock_bh(&xfrm_policy_lock);
421
44e36b42 422 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 423}
424
66caf628 425static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 426{
66caf628 427 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 428 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
429 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 430 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 431 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 432 int i;
433
434 if (!nidx)
435 return;
436
437 write_lock_bh(&xfrm_policy_lock);
438
439 for (i = hmask; i >= 0; i--)
440 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
441
66caf628
AD		 442 net->xfrm.policy_byidx = nidx;
443 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2
DM		 444
445 write_unlock_bh(&xfrm_policy_lock);
446
44e36b42 447 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 448}
449
66caf628 450static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 451{
66caf628
AD		 452 unsigned int cnt = net->xfrm.policy_count[dir];
453 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 454
455 if (total)
456 *total += cnt;
457
458 if ((hmask + 1) < xfrm_policy_hashmax &&
459 cnt > hmask)
460 return 1;
461
462 return 0;
463}
464
66caf628 465static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 466{
66caf628 467 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 468
469 if ((hmask + 1) < xfrm_policy_hashmax &&
470 total > hmask)
471 return 1;
472
473 return 0;
474}
475
5a6d3416 476void xfrm_spd_getinfo(struct xfrmk_spdinfo *si)
ecfd6b18
JHS		 477{
478 read_lock_bh(&xfrm_policy_lock);
dc2caba7
AD		 479 si->incnt = init_net.xfrm.policy_count[XFRM_POLICY_IN];
480 si->outcnt = init_net.xfrm.policy_count[XFRM_POLICY_OUT];
481 si->fwdcnt = init_net.xfrm.policy_count[XFRM_POLICY_FWD];
482 si->inscnt = init_net.xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
483 si->outscnt = init_net.xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
484 si->fwdscnt = init_net.xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
8100bea7 485 si->spdhcnt = init_net.xfrm.policy_idx_hmask;
ecfd6b18
JHS		 486 si->spdhmcnt = xfrm_policy_hashmax;
487 read_unlock_bh(&xfrm_policy_lock);
488}
489EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 490
ecfd6b18 491static DEFINE_MUTEX(hash_resize_mutex);
66caf628 492static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 493{
66caf628 494 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 495 int dir, total;
496
497 mutex_lock(&hash_resize_mutex);
498
499 total = 0;
500 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 501 if (xfrm_bydst_should_resize(net, dir, &total))
502 xfrm_bydst_resize(net, dir);
2518c7c2 503 }
66caf628
AD		 504 if (xfrm_byidx_should_resize(net, total))
505 xfrm_byidx_resize(net, total);
2518c7c2
DM		 506
507 mutex_unlock(&hash_resize_mutex);
508}
509
1da177e4
LT		 510/* Generate new index... KAME seems to generate them ordered by cost
511 * of an absolute inpredictability of ordering of rules. This will not pass. */
1121994c 512static u32 xfrm_gen_index(struct net *net, int dir)
1da177e4 513{
1da177e4
LT		 514 static u32 idx_generator;
515
516 for (;;) {
2518c7c2
DM		 517 struct hlist_node *entry;
518 struct hlist_head *list;
519 struct xfrm_policy *p;
520 u32 idx;
521 int found;
522
1da177e4
LT		 523 idx = (idx_generator | dir);
524 idx_generator += 8;
525 if (idx == 0)
526 idx = 8;
1121994c 527 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2
DM		 528 found = 0;
529 hlist_for_each_entry(p, entry, list, byidx) {
530 if (p->index == idx) {
531 found = 1;
1da177e4 532 break;
2518c7c2 533 }
1da177e4 534 }
2518c7c2 535 if (!found)
1da177e4
LT		 536 return idx;
537 }
538}
539
2518c7c2
DM		 540static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
541{
542 u32 *p1 = (u32 *) s1;
543 u32 *p2 = (u32 *) s2;
544 int len = sizeof(struct xfrm_selector) / sizeof(u32);
545 int i;
546
547 for (i = 0; i < len; i++) {
548 if (p1[i] != p2[i])
549 return 1;
550 }
551
552 return 0;
553}
554
1da177e4
LT		 555int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
556{
1121994c 557 struct net *net = xp_net(policy);
2518c7c2
DM		 558 struct xfrm_policy *pol;
559 struct xfrm_policy *delpol;
560 struct hlist_head *chain;
a6c7ab55 561 struct hlist_node *entry, *newpos;
9b78a82c 562 struct dst_entry *gc_list;
1da177e4
LT		 563
564 write_lock_bh(&xfrm_policy_lock);
1121994c 565 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 566 delpol = NULL;
567 newpos = NULL;
2518c7c2 568 hlist_for_each_entry(pol, entry, chain, bydst) {
a6c7ab55 569 if (pol->type == policy->type &&
2518c7c2 570 !selector_cmp(&pol->selector, &policy->selector) &&
a6c7ab55
HX		 571 xfrm_sec_ctx_match(pol->security, policy->security) &&
572 !WARN_ON(delpol)) {
1da177e4
LT		 573 if (excl) {
574 write_unlock_bh(&xfrm_policy_lock);
575 return -EEXIST;
576 }
1da177e4
LT		 577 delpol = pol;
578 if (policy->priority > pol->priority)
579 continue;
580 } else if (policy->priority >= pol->priority) {
a6c7ab55 581 newpos = &pol->bydst;
1da177e4
LT		 582 continue;
583 }
1da177e4
LT		 584 if (delpol)
585 break;
1da177e4
LT		 586 }
587 if (newpos)
2518c7c2
DM		 588 hlist_add_after(newpos, &policy->bydst);
589 else
590 hlist_add_head(&policy->bydst, chain);
1da177e4 591 xfrm_pol_hold(policy);
1121994c 592 net->xfrm.policy_count[dir]++;
1da177e4 593 atomic_inc(&flow_cache_genid);
2518c7c2
DM		 594 if (delpol) {
595 hlist_del(&delpol->bydst);
596 hlist_del(&delpol->byidx);
12a169e7 597 list_del(&delpol->walk.all);
1121994c 598 net->xfrm.policy_count[dir]--;
2518c7c2 599 }
1121994c
AD		 600 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
601 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 602 policy->curlft.add_time = get_seconds();
1da177e4
LT		 603 policy->curlft.use_time = 0;
604 if (!mod_timer(&policy->timer, jiffies + HZ))
605 xfrm_pol_hold(policy);
1121994c 606 list_add(&policy->walk.all, &net->xfrm.policy_all);
1da177e4
LT		 607 write_unlock_bh(&xfrm_policy_lock);
608
9b78a82c 609 if (delpol)
1da177e4 610 xfrm_policy_kill(delpol);
1121994c
AD		 611 else if (xfrm_bydst_should_resize(net, dir, NULL))
612 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c
DM		 613
614 read_lock_bh(&xfrm_policy_lock);
615 gc_list = NULL;
2518c7c2
DM		 616 entry = &policy->bydst;
617 hlist_for_each_entry_continue(policy, entry, bydst) {
9b78a82c
DM		 618 struct dst_entry *dst;
619
620 write_lock(&policy->lock);
621 dst = policy->bundles;
622 if (dst) {
623 struct dst_entry *tail = dst;
624 while (tail->next)
625 tail = tail->next;
626 tail->next = gc_list;
627 gc_list = dst;
628
629 policy->bundles = NULL;
630 }
631 write_unlock(&policy->lock);
1da177e4 632 }
9b78a82c
DM		 633 read_unlock_bh(&xfrm_policy_lock);
634
635 while (gc_list) {
636 struct dst_entry *dst = gc_list;
637
638 gc_list = dst->next;
639 dst_free(dst);
640 }
641
1da177e4
LT		 642 return 0;
643}
644EXPORT_SYMBOL(xfrm_policy_insert);
645
8d1211a6 646struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u8 type, int dir,
4e81bb83 647 struct xfrm_selector *sel,
ef41aaa0
EP		 648 struct xfrm_sec_ctx *ctx, int delete,
649 int *err)
1da177e4 650{
2518c7c2
DM		 651 struct xfrm_policy *pol, *ret;
652 struct hlist_head *chain;
653 struct hlist_node *entry;
1da177e4 654
ef41aaa0 655 *err = 0;
1da177e4 656 write_lock_bh(&xfrm_policy_lock);
8d1211a6 657 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2
DM		 658 ret = NULL;
659 hlist_for_each_entry(pol, entry, chain, bydst) {
660 if (pol->type == type &&
661 !selector_cmp(sel, &pol->selector) &&
662 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 663 xfrm_pol_hold(pol);
2518c7c2 664 if (delete) {
03e1ad7b
PM		 665 *err = security_xfrm_policy_delete(
666 pol->security);
ef41aaa0
EP		 667 if (*err) {
668 write_unlock_bh(&xfrm_policy_lock);
669 return pol;
670 }
2518c7c2
DM		 671 hlist_del(&pol->bydst);
672 hlist_del(&pol->byidx);
12a169e7 673 list_del(&pol->walk.all);
8d1211a6 674 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 675 }
676 ret = pol;
1da177e4
LT		 677 break;
678 }
679 }
680 write_unlock_bh(&xfrm_policy_lock);
681
2518c7c2 682 if (ret && delete) {
1da177e4 683 atomic_inc(&flow_cache_genid);
2518c7c2 684 xfrm_policy_kill(ret);
1da177e4 685 }
2518c7c2 686 return ret;
1da177e4 687}
df71837d 688EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 689
8d1211a6
AD		 690struct xfrm_policy *xfrm_policy_byid(struct net *net, u8 type, int dir, u32 id,
691 int delete, int *err)
1da177e4 692{
2518c7c2
DM		 693 struct xfrm_policy *pol, *ret;
694 struct hlist_head *chain;
695 struct hlist_node *entry;
1da177e4 696
b5505c6e
HX		 697 *err = -ENOENT;
698 if (xfrm_policy_id2dir(id) != dir)
699 return NULL;
700
ef41aaa0 701 *err = 0;
1da177e4 702 write_lock_bh(&xfrm_policy_lock);
8d1211a6 703 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2
DM		 704 ret = NULL;
705 hlist_for_each_entry(pol, entry, chain, byidx) {
706 if (pol->type == type && pol->index == id) {
1da177e4 707 xfrm_pol_hold(pol);
2518c7c2 708 if (delete) {
03e1ad7b
PM		 709 *err = security_xfrm_policy_delete(
710 pol->security);
ef41aaa0
EP		 711 if (*err) {
712 write_unlock_bh(&xfrm_policy_lock);
713 return pol;
714 }
2518c7c2
DM		 715 hlist_del(&pol->bydst);
716 hlist_del(&pol->byidx);
12a169e7 717 list_del(&pol->walk.all);
8d1211a6 718 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 719 }
720 ret = pol;
1da177e4
LT		 721 break;
722 }
723 }
724 write_unlock_bh(&xfrm_policy_lock);
725
2518c7c2 726 if (ret && delete) {
1da177e4 727 atomic_inc(&flow_cache_genid);
2518c7c2 728 xfrm_policy_kill(ret);
1da177e4 729 }
2518c7c2 730 return ret;
1da177e4
LT		 731}
732EXPORT_SYMBOL(xfrm_policy_byid);
733
4aa2e62c
JL		 734#ifdef CONFIG_SECURITY_NETWORK_XFRM
735static inline int
33ffbbd5 736xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
1da177e4 737{
4aa2e62c
JL		 738 int dir, err = 0;
739
740 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
741 struct xfrm_policy *pol;
742 struct hlist_node *entry;
743 int i;
744
745 hlist_for_each_entry(pol, entry,
33ffbbd5 746 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 747 if (pol->type != type)
748 continue;
03e1ad7b 749 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 750 if (err) {
ab5f5e8b
JL		 751 xfrm_audit_policy_delete(pol, 0,
752 audit_info->loginuid,
2532386f 753 audit_info->sessionid,
ab5f5e8b 754 audit_info->secid);
4aa2e62c
JL		 755 return err;
756 }
7dc12d6d 757 }
33ffbbd5 758 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
4aa2e62c 759 hlist_for_each_entry(pol, entry,
33ffbbd5 760 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 761 bydst) {
762 if (pol->type != type)
763 continue;
03e1ad7b
PM		 764 err = security_xfrm_policy_delete(
765 pol->security);
4aa2e62c 766 if (err) {
ab5f5e8b
JL		 767 xfrm_audit_policy_delete(pol, 0,
768 audit_info->loginuid,
2532386f 769 audit_info->sessionid,
ab5f5e8b 770 audit_info->secid);
4aa2e62c
JL		 771 return err;
772 }
773 }
774 }
775 }
776 return err;
777}
778#else
779static inline int
33ffbbd5 780xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c
JL		 781{
782 return 0;
783}
784#endif
785
33ffbbd5 786int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c
JL		 787{
788 int dir, err = 0;
1da177e4
LT		 789
790 write_lock_bh(&xfrm_policy_lock);
4aa2e62c 791
33ffbbd5 792 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
4aa2e62c
JL		 793 if (err)
794 goto out;
795
1da177e4 796 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 797 struct xfrm_policy *pol;
798 struct hlist_node *entry;
ae8c0577 799 int i, killed;
2518c7c2 800
ae8c0577 801 killed = 0;
2518c7c2
DM		 802 again1:
803 hlist_for_each_entry(pol, entry,
33ffbbd5 804 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 805 if (pol->type != type)
806 continue;
807 hlist_del(&pol->bydst);
808 hlist_del(&pol->byidx);
1da177e4
LT		 809 write_unlock_bh(&xfrm_policy_lock);
810
ab5f5e8b 811 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
2532386f 812 audit_info->sessionid,
ab5f5e8b 813 audit_info->secid);
161a09e7 814
2518c7c2 815 xfrm_policy_kill(pol);
ae8c0577 816 killed++;
1da177e4
LT		 817
818 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 819 goto again1;
820 }
821
33ffbbd5 822 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2
DM		 823 again2:
824 hlist_for_each_entry(pol, entry,
33ffbbd5 825 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 826 bydst) {
827 if (pol->type != type)
828 continue;
829 hlist_del(&pol->bydst);
830 hlist_del(&pol->byidx);
12a169e7 831 list_del(&pol->walk.all);
2518c7c2
DM		 832 write_unlock_bh(&xfrm_policy_lock);
833
ab5f5e8b
JL		 834 xfrm_audit_policy_delete(pol, 1,
835 audit_info->loginuid,
2532386f 836 audit_info->sessionid,
ab5f5e8b 837 audit_info->secid);
2518c7c2 838 xfrm_policy_kill(pol);
ae8c0577 839 killed++;
2518c7c2
DM		 840
841 write_lock_bh(&xfrm_policy_lock);
842 goto again2;
843 }
1da177e4 844 }
2518c7c2 845
33ffbbd5 846 net->xfrm.policy_count[dir] -= killed;
1da177e4
LT		 847 }
848 atomic_inc(&flow_cache_genid);
4aa2e62c 849out:
1da177e4 850 write_unlock_bh(&xfrm_policy_lock);
4aa2e62c 851 return err;
1da177e4
LT		 852}
853EXPORT_SYMBOL(xfrm_policy_flush);
854
cdcbca7c 855int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 856 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 857 void *data)
858{
12a169e7
HX		 859 struct xfrm_policy *pol;
860 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 861 int error = 0;
862
863 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
864 walk->type != XFRM_POLICY_TYPE_ANY)
865 return -EINVAL;
1da177e4 866
12a169e7 867 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 868 return 0;
869
12a169e7
HX		 870 write_lock_bh(&xfrm_policy_lock);
871 if (list_empty(&walk->walk.all))
cdcbca7c 872 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 873 else
874 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 875 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 876 if (x->dead)
4c563f76 877 continue;
12a169e7
HX		 878 pol = container_of(x, struct xfrm_policy, walk);
879 if (walk->type != XFRM_POLICY_TYPE_ANY &&
880 walk->type != pol->type)
881 continue;
882 error = func(pol, xfrm_policy_id2dir(pol->index),
883 walk->seq, data);
884 if (error) {
885 list_move_tail(&walk->walk.all, &x->all);
886 goto out;
2518c7c2 887 }
12a169e7 888 walk->seq++;
1da177e4 889 }
12a169e7 890 if (walk->seq == 0) {
baf5d743
JHS		 891 error = -ENOENT;
892 goto out;
893 }
12a169e7 894 list_del_init(&walk->walk.all);
1da177e4 895out:
12a169e7 896 write_unlock_bh(&xfrm_policy_lock);
1da177e4
LT		 897 return error;
898}
899EXPORT_SYMBOL(xfrm_policy_walk);
900
12a169e7
HX		 901void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
902{
903 INIT_LIST_HEAD(&walk->walk.all);
904 walk->walk.dead = 1;
905 walk->type = type;
906 walk->seq = 0;
907}
908EXPORT_SYMBOL(xfrm_policy_walk_init);
909
910void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
911{
912 if (list_empty(&walk->walk.all))
913 return;
914
915 write_lock_bh(&xfrm_policy_lock);
916 list_del(&walk->walk.all);
917 write_unlock_bh(&xfrm_policy_lock);
918}
919EXPORT_SYMBOL(xfrm_policy_walk_done);
920
134b0fc5
JM		 921/*
922 * Find policy to apply to this flow.
923 *
924 * Returns 0 if policy found, else an -errno.
925 */
2518c7c2
DM		 926static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
927 u8 type, u16 family, int dir)
1da177e4 928{
2518c7c2 929 struct xfrm_selector *sel = &pol->selector;
134b0fc5 930 int match, ret = -ESRCH;
1da177e4 931
2518c7c2
DM		 932 if (pol->family != family ||
933 pol->type != type)
134b0fc5 934 return ret;
1da177e4 935
2518c7c2 936 match = xfrm_selector_match(sel, fl, family);
134b0fc5 937 if (match)
03e1ad7b
PM		 938 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
939 dir);
2518c7c2 940
134b0fc5 941 return ret;
2518c7c2 942}
1da177e4 943
52479b62
AD		 944static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
945 struct flowi *fl,
2518c7c2
DM		 946 u16 family, u8 dir)
947{
134b0fc5 948 int err;
2518c7c2
DM		 949 struct xfrm_policy *pol, *ret;
950 xfrm_address_t *daddr, *saddr;
951 struct hlist_node *entry;
952 struct hlist_head *chain;
acba48e1 953 u32 priority = ~0U;
df71837d 954
2518c7c2
DM		 955 daddr = xfrm_flowi_daddr(fl, family);
956 saddr = xfrm_flowi_saddr(fl, family);
957 if (unlikely(!daddr || !saddr))
958 return NULL;
959
960 read_lock_bh(&xfrm_policy_lock);
52479b62 961 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2
DM		 962 ret = NULL;
963 hlist_for_each_entry(pol, entry, chain, bydst) {
134b0fc5
JM		 964 err = xfrm_policy_match(pol, fl, type, family, dir);
965 if (err) {
966 if (err == -ESRCH)
967 continue;
968 else {
969 ret = ERR_PTR(err);
970 goto fail;
971 }
972 } else {
2518c7c2 973 ret = pol;
acba48e1 974 priority = ret->priority;
2518c7c2
DM		 975 break;
976 }
977 }
52479b62 978 chain = &net->xfrm.policy_inexact[dir];
acba48e1 979 hlist_for_each_entry(pol, entry, chain, bydst) {
134b0fc5
JM		 980 err = xfrm_policy_match(pol, fl, type, family, dir);
981 if (err) {
982 if (err == -ESRCH)
983 continue;
984 else {
985 ret = ERR_PTR(err);
986 goto fail;
987 }
988 } else if (pol->priority < priority) {
acba48e1
DM		 989 ret = pol;
990 break;
1da177e4
LT		 991 }
992 }
acba48e1
DM		 993 if (ret)
994 xfrm_pol_hold(ret);
134b0fc5 995fail:
1da177e4 996 read_unlock_bh(&xfrm_policy_lock);
4e81bb83 997
2518c7c2 998 return ret;
4e81bb83
MN		 999}
1000
52479b62
AD		 1001static int xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family,
1002 u8 dir, void **objp, atomic_t **obj_refp)
4e81bb83
MN		 1003{
1004 struct xfrm_policy *pol;
134b0fc5 1005 int err = 0;
4e81bb83
MN		 1006
1007#ifdef CONFIG_XFRM_SUB_POLICY
52479b62 1008 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
134b0fc5
JM		 1009 if (IS_ERR(pol)) {
1010 err = PTR_ERR(pol);
1011 pol = NULL;
1012 }
1013 if (pol || err)
4e81bb83
MN		 1014 goto end;
1015#endif
52479b62 1016 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
134b0fc5
JM		 1017 if (IS_ERR(pol)) {
1018 err = PTR_ERR(pol);
1019 pol = NULL;
1020 }
4e81bb83 1021#ifdef CONFIG_XFRM_SUB_POLICY
2518c7c2 1022end:
4e81bb83 1023#endif
1da177e4
LT		 1024 if ((*objp = (void *) pol) != NULL)
1025 *obj_refp = &pol->refcnt;
134b0fc5 1026 return err;
1da177e4
LT		 1027}
1028
df71837d
TJ		 1029static inline int policy_to_flow_dir(int dir)
1030{
1031 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1032 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1033 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1034 return dir;
1035 switch (dir) {
1036 default:
1037 case XFRM_POLICY_IN:
1038 return FLOW_DIR_IN;
1039 case XFRM_POLICY_OUT:
1040 return FLOW_DIR_OUT;
1041 case XFRM_POLICY_FWD:
1042 return FLOW_DIR_FWD;
3ff50b79 1043 }
df71837d
TJ		 1044}
1045
e0d1caa7 1046static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
1da177e4
LT		 1047{
1048 struct xfrm_policy *pol;
1049
1050 read_lock_bh(&xfrm_policy_lock);
1051 if ((pol = sk->sk_policy[dir]) != NULL) {
a716c119 1052 int match = xfrm_selector_match(&pol->selector, fl,
1da177e4 1053 sk->sk_family);
a716c119 1054 int err = 0;
df71837d 1055
3bccfbc7 1056 if (match) {
03e1ad7b
PM		 1057 err = security_xfrm_policy_lookup(pol->security,
1058 fl->secid,
1059 policy_to_flow_dir(dir));
3bccfbc7
VY		 1060 if (!err)
1061 xfrm_pol_hold(pol);
1062 else if (err == -ESRCH)
1063 pol = NULL;
1064 else
1065 pol = ERR_PTR(err);
1066 } else
1da177e4
LT		 1067 pol = NULL;
1068 }
1069 read_unlock_bh(&xfrm_policy_lock);
1070 return pol;
1071}
1072
1073static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1074{
98806f75 1075 struct net *net = xp_net(pol);
1121994c 1076 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1077 pol->family, dir);
4e81bb83 1078
98806f75 1079 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1080 hlist_add_head(&pol->bydst, chain);
e92303f8 1081 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1082 net->xfrm.policy_count[dir]++;
1da177e4 1083 xfrm_pol_hold(pol);
2518c7c2 1084
98806f75
AD		 1085 if (xfrm_bydst_should_resize(net, dir, NULL))
1086 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1087}
1088
1089static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1090 int dir)
1091{
98806f75
AD		 1092 struct net *net = xp_net(pol);
1093
2518c7c2
DM		 1094 if (hlist_unhashed(&pol->bydst))
1095 return NULL;
1da177e4 1096
2518c7c2
DM		 1097 hlist_del(&pol->bydst);
1098 hlist_del(&pol->byidx);
12a169e7 1099 list_del(&pol->walk.all);
98806f75 1100 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1101
1102 return pol;
1da177e4
LT		 1103}
1104
4666faab 1105int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4
LT		 1106{
1107 write_lock_bh(&xfrm_policy_lock);
1108 pol = __xfrm_policy_unlink(pol, dir);
1109 write_unlock_bh(&xfrm_policy_lock);
1110 if (pol) {
1111 if (dir < XFRM_POLICY_MAX)
1112 atomic_inc(&flow_cache_genid);
1113 xfrm_policy_kill(pol);
4666faab 1114 return 0;
1da177e4 1115 }
4666faab 1116 return -ENOENT;
1da177e4 1117}
a70fcb0b 1118EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1119
1120int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1121{
1121994c 1122 struct net *net = xp_net(pol);
1da177e4
LT		 1123 struct xfrm_policy *old_pol;
1124
4e81bb83
MN		 1125#ifdef CONFIG_XFRM_SUB_POLICY
1126 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1127 return -EINVAL;
1128#endif
1129
1da177e4
LT		 1130 write_lock_bh(&xfrm_policy_lock);
1131 old_pol = sk->sk_policy[dir];
1132 sk->sk_policy[dir] = pol;
1133 if (pol) {
9d729f72 1134 pol->curlft.add_time = get_seconds();
1121994c 1135 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1da177e4
LT		 1136 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1137 }
1138 if (old_pol)
1139 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1140 write_unlock_bh(&xfrm_policy_lock);
1141
1142 if (old_pol) {
1143 xfrm_policy_kill(old_pol);
1144 }
1145 return 0;
1146}
1147
1148static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1149{
0331b1f3 1150 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1da177e4
LT		 1151
1152 if (newp) {
1153 newp->selector = old->selector;
03e1ad7b
PM		 1154 if (security_xfrm_policy_clone(old->security,
1155 &newp->security)) {
df71837d
TJ		 1156 kfree(newp);
1157 return NULL; /* ENOMEM */
1158 }
1da177e4
LT		 1159 newp->lft = old->lft;
1160 newp->curlft = old->curlft;
1161 newp->action = old->action;
1162 newp->flags = old->flags;
1163 newp->xfrm_nr = old->xfrm_nr;
1164 newp->index = old->index;
4e81bb83 1165 newp->type = old->type;
1da177e4
LT		 1166 memcpy(newp->xfrm_vec, old->xfrm_vec,
1167 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1168 write_lock_bh(&xfrm_policy_lock);
1169 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1170 write_unlock_bh(&xfrm_policy_lock);
1171 xfrm_pol_put(newp);
1172 }
1173 return newp;
1174}
1175
1176int __xfrm_sk_clone_policy(struct sock *sk)
1177{
1178 struct xfrm_policy *p0 = sk->sk_policy[0],
1179 *p1 = sk->sk_policy[1];
1180
1181 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1182 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1183 return -ENOMEM;
1184 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1185 return -ENOMEM;
1186 return 0;
1187}
1188
a1e59abf 1189static int
fbda33b2 1190xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1191 unsigned short family)
1192{
1193 int err;
1194 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1195
1196 if (unlikely(afinfo == NULL))
1197 return -EINVAL;
fbda33b2 1198 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1199 xfrm_policy_put_afinfo(afinfo);
1200 return err;
1201}
1202
1da177e4
LT		 1203/* Resolve list of templates for the flow, given policy. */
1204
1205static int
4e81bb83
MN		 1206xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
1207 struct xfrm_state **xfrm,
1208 unsigned short family)
1da177e4 1209{
fbda33b2 1210 struct net *net = xp_net(policy);
1da177e4
LT		 1211 int nx;
1212 int i, error;
1213 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1214 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1215 xfrm_address_t tmp;
1da177e4
LT		 1216
1217 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1218 struct xfrm_state *x;
1219 xfrm_address_t *remote = daddr;
1220 xfrm_address_t *local = saddr;
1221 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1222
48b8d783
JK		 1223 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1224 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1225 remote = &tmpl->id.daddr;
1226 local = &tmpl->saddr;
76b3f055 1227 family = tmpl->encap_family;
a1e59abf 1228 if (xfrm_addr_any(local, family)) {
fbda33b2 1229 error = xfrm_get_saddr(net, &tmp, remote, family);
a1e59abf
PM		 1230 if (error)
1231 goto fail;
1232 local = &tmp;
1233 }
1da177e4
LT		 1234 }
1235
1236 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1237
1238 if (x && x->km.state == XFRM_STATE_VALID) {
1239 xfrm[nx++] = x;
1240 daddr = remote;
1241 saddr = local;
1242 continue;
1243 }
1244 if (x) {
1245 error = (x->km.state == XFRM_STATE_ERROR ?
1246 -EINVAL : -EAGAIN);
1247 xfrm_state_put(x);
1248 }
a4322266 1249 else if (error == -ESRCH)
1250 error = -EAGAIN;
1da177e4
LT		 1251
1252 if (!tmpl->optional)
1253 goto fail;
1254 }
1255 return nx;
1256
1257fail:
1258 for (nx--; nx>=0; nx--)
1259 xfrm_state_put(xfrm[nx]);
1260 return error;
1261}
1262
4e81bb83
MN		 1263static int
1264xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1265 struct xfrm_state **xfrm,
1266 unsigned short family)
1267{
41a49cc3
MN		 1268 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1269 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1270 int cnx = 0;
1271 int error;
1272 int ret;
1273 int i;
1274
1275 for (i = 0; i < npols; i++) {
1276 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1277 error = -ENOBUFS;
1278 goto fail;
1279 }
41a49cc3
MN		 1280
1281 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1282 if (ret < 0) {
1283 error = ret;
1284 goto fail;
1285 } else
1286 cnx += ret;
1287 }
1288
41a49cc3
MN		 1289 /* found states are sorted for outbound processing */
1290 if (npols > 1)
1291 xfrm_state_sort(xfrm, tpp, cnx, family);
1292
4e81bb83
MN		 1293 return cnx;
1294
1295 fail:
1296 for (cnx--; cnx>=0; cnx--)
41a49cc3 1297 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1298 return error;
1299
1300}
1301
1da177e4
LT		 1302/* Check that the bundle accepts the flow and its components are
1303 * still valid.
1304 */
1305
1306static struct dst_entry *
1307xfrm_find_bundle(struct flowi *fl, struct xfrm_policy *policy, unsigned short family)
1308{
1309 struct dst_entry *x;
1310 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1311 if (unlikely(afinfo == NULL))
1312 return ERR_PTR(-EINVAL);
1313 x = afinfo->find_bundle(fl, policy);
1314 xfrm_policy_put_afinfo(afinfo);
1315 return x;
1316}
1317
25ee3286
HX		 1318static inline int xfrm_get_tos(struct flowi *fl, int family)
1319{
1320 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1321 int tos;
1da177e4 1322
25ee3286
HX		 1323 if (!afinfo)
1324 return -EINVAL;
1325
1326 tos = afinfo->get_tos(fl);
1327
1328 xfrm_policy_put_afinfo(afinfo);
1329
1330 return tos;
1331}
1332
1333static inline struct xfrm_dst *xfrm_alloc_dst(int family)
1da177e4 1334{
1da177e4 1335 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
25ee3286
HX		 1336 struct xfrm_dst *xdst;
1337
1338 if (!afinfo)
1339 return ERR_PTR(-EINVAL);
1340
1341 xdst = dst_alloc(afinfo->dst_ops) ?: ERR_PTR(-ENOBUFS);
1342
1343 xfrm_policy_put_afinfo(afinfo);
1344
1345 return xdst;
1346}
1347
a1b05140
MN		 1348static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1349 int nfheader_len)
1350{
1351 struct xfrm_policy_afinfo *afinfo =
1352 xfrm_policy_get_afinfo(dst->ops->family);
1353 int err;
1354
1355 if (!afinfo)
1356 return -EINVAL;
1357
1358 err = afinfo->init_path(path, dst, nfheader_len);
1359
1360 xfrm_policy_put_afinfo(afinfo);
1361
1362 return err;
1363}
1364
25ee3286
HX		 1365static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
1366{
1367 struct xfrm_policy_afinfo *afinfo =
1368 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1369 int err;
1370
1371 if (!afinfo)
1da177e4 1372 return -EINVAL;
25ee3286
HX		 1373
1374 err = afinfo->fill_dst(xdst, dev);
1375
1da177e4 1376 xfrm_policy_put_afinfo(afinfo);
25ee3286 1377
1da177e4
LT		 1378 return err;
1379}
1380
25ee3286
HX		 1381/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1382 * all the metrics... Shortly, bundle a bundle.
1383 */
1384
1385static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1386 struct xfrm_state **xfrm, int nx,
1387 struct flowi *fl,
1388 struct dst_entry *dst)
1389{
1390 unsigned long now = jiffies;
1391 struct net_device *dev;
1392 struct dst_entry *dst_prev = NULL;
1393 struct dst_entry *dst0 = NULL;
1394 int i = 0;
1395 int err;
1396 int header_len = 0;
a1b05140 1397 int nfheader_len = 0;
25ee3286
HX		 1398 int trailer_len = 0;
1399 int tos;
1400 int family = policy->selector.family;
9bb182a7
YH		 1401 xfrm_address_t saddr, daddr;
1402
1403 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1404
1405 tos = xfrm_get_tos(fl, family);
1406 err = tos;
1407 if (tos < 0)
1408 goto put_states;
1409
1410 dst_hold(dst);
1411
1412 for (; i < nx; i++) {
1413 struct xfrm_dst *xdst = xfrm_alloc_dst(family);
1414 struct dst_entry *dst1 = &xdst->u.dst;
1415
1416 err = PTR_ERR(xdst);
1417 if (IS_ERR(xdst)) {
1418 dst_release(dst);
1419 goto put_states;
1420 }
1421
1422 if (!dst_prev)
1423 dst0 = dst1;
1424 else {
1425 dst_prev->child = dst_clone(dst1);
1426 dst1->flags |= DST_NOHASH;
1427 }
1428
1429 xdst->route = dst;
1430 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1431
1432 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1433 family = xfrm[i]->props.family;
9bb182a7
YH		 1434 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1435 family);
25ee3286
HX		 1436 err = PTR_ERR(dst);
1437 if (IS_ERR(dst))
1438 goto put_states;
1439 } else
1440 dst_hold(dst);
1441
1442 dst1->xfrm = xfrm[i];
1443 xdst->genid = xfrm[i]->genid;
1444
1445 dst1->obsolete = -1;
1446 dst1->flags |= DST_HOST;
1447 dst1->lastuse = now;
1448
1449 dst1->input = dst_discard;
1450 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1451
1452 dst1->next = dst_prev;
1453 dst_prev = dst1;
1454
1455 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1456 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1457 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1458 trailer_len += xfrm[i]->props.trailer_len;
1459 }
1460
1461 dst_prev->child = dst;
1462 dst0->path = dst;
1463
1464 err = -ENODEV;
1465 dev = dst->dev;
1466 if (!dev)
1467 goto free_dst;
1468
1469 /* Copy neighbout for reachability confirmation */
1470 dst0->neighbour = neigh_clone(dst->neighbour);
1471
a1b05140 1472 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1473 xfrm_init_pmtu(dst_prev);
1474
1475 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1476 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1477
1478 err = xfrm_fill_dst(xdst, dev);
1479 if (err)
1480 goto free_dst;
1481
1482 dst_prev->header_len = header_len;
1483 dst_prev->trailer_len = trailer_len;
1484 header_len -= xdst->u.dst.xfrm->props.header_len;
1485 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1486 }
1487
1488out:
1489 return dst0;
1490
1491put_states:
1492 for (; i < nx; i++)
1493 xfrm_state_put(xfrm[i]);
1494free_dst:
1495 if (dst0)
1496 dst_free(dst0);
1497 dst0 = ERR_PTR(err);
1498 goto out;
1499}
1500
157bfc25
MN		 1501static int inline
1502xfrm_dst_alloc_copy(void **target, void *src, int size)
1503{
1504 if (!*target) {
1505 *target = kmalloc(size, GFP_ATOMIC);
1506 if (!*target)
1507 return -ENOMEM;
1508 }
1509 memcpy(*target, src, size);
1510 return 0;
1511}
1512
1513static int inline
1514xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1515{
1516#ifdef CONFIG_XFRM_SUB_POLICY
1517 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1518 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1519 sel, sizeof(*sel));
1520#else
1521 return 0;
1522#endif
1523}
1524
1525static int inline
1526xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1527{
1528#ifdef CONFIG_XFRM_SUB_POLICY
1529 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1530 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1531#else
1532 return 0;
1533#endif
1534}
1da177e4
LT		 1535
1536static int stale_bundle(struct dst_entry *dst);
1537
1538/* Main function: finds/creates a bundle for given flow.
1539 *
1540 * At the moment we eat a raw IP route. Mostly to speed up lookups
1541 * on interfaces with disabled IPsec.
1542 */
52479b62 1543int __xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
14e50e57 1544 struct sock *sk, int flags)
1da177e4
LT		 1545{
1546 struct xfrm_policy *policy;
4e81bb83
MN		 1547 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1548 int npols;
1549 int pol_dead;
1550 int xfrm_nr;
1551 int pi;
1da177e4
LT		 1552 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1553 struct dst_entry *dst, *dst_orig = *dst_p;
1554 int nx = 0;
1555 int err;
1556 u32 genid;
42cf93cd 1557 u16 family;
df71837d 1558 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
e0d1caa7 1559
1da177e4
LT		 1560restart:
1561 genid = atomic_read(&flow_cache_genid);
1562 policy = NULL;
4e81bb83
MN		 1563 for (pi = 0; pi < ARRAY_SIZE(pols); pi++)
1564 pols[pi] = NULL;
1565 npols = 0;
1566 pol_dead = 0;
1567 xfrm_nr = 0;
1568
f7944fb1 1569 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
e0d1caa7 1570 policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
75b8c133 1571 err = PTR_ERR(policy);
0aa64774
MN		 1572 if (IS_ERR(policy)) {
1573 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
75b8c133 1574 goto dropdst;
0aa64774 1575 }
3bccfbc7 1576 }
1da177e4
LT		 1577
1578 if (!policy) {
1579 /* To accelerate a bit... */
2518c7c2 1580 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 1581 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 1582 goto nopol;
1da177e4 1583
52479b62 1584 policy = flow_cache_lookup(net, fl, dst_orig->ops->family,
42cf93cd 1585 dir, xfrm_policy_lookup);
75b8c133 1586 err = PTR_ERR(policy);
d66e37a9
MN		 1587 if (IS_ERR(policy)) {
1588 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
75b8c133 1589 goto dropdst;
d66e37a9 1590 }
1da177e4
LT		 1591 }
1592
1593 if (!policy)
8b7817f3 1594 goto nopol;
1da177e4 1595
42cf93cd 1596 family = dst_orig->ops->family;
4e81bb83
MN		 1597 pols[0] = policy;
1598 npols ++;
1599 xfrm_nr += pols[0]->xfrm_nr;
1da177e4 1600
aef21785 1601 err = -ENOENT;
8b7817f3
HX		 1602 if ((flags & XFRM_LOOKUP_ICMP) && !(policy->flags & XFRM_POLICY_ICMP))
1603 goto error;
1604
1605 policy->curlft.use_time = get_seconds();
1606
1da177e4 1607 switch (policy->action) {
5e5234ff 1608 default:
1da177e4
LT		 1609 case XFRM_POLICY_BLOCK:
1610 /* Prohibit the flow */
0aa64774 1611 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 1612 err = -EPERM;
1613 goto error;
1da177e4
LT		 1614
1615 case XFRM_POLICY_ALLOW:
4e81bb83 1616#ifndef CONFIG_XFRM_SUB_POLICY
1da177e4
LT		 1617 if (policy->xfrm_nr == 0) {
1618 /* Flow passes not transformed. */
1619 xfrm_pol_put(policy);
1620 return 0;
1621 }
4e81bb83 1622#endif
1da177e4
LT		 1623
1624 /* Try to find matching bundle.
1625 *
1626 * LATER: help from flow cache. It is optional, this
1627 * is required only for output policy.
1628 */
1629 dst = xfrm_find_bundle(fl, policy, family);
1630 if (IS_ERR(dst)) {
0aa64774 1631 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
e104411b
PM		 1632 err = PTR_ERR(dst);
1633 goto error;
1da177e4
LT		 1634 }
1635
1636 if (dst)
1637 break;
1638
4e81bb83
MN		 1639#ifdef CONFIG_XFRM_SUB_POLICY
1640 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
52479b62
AD		 1641 pols[1] = xfrm_policy_lookup_bytype(net,
1642 XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 1643 fl, family,
1644 XFRM_POLICY_OUT);
1645 if (pols[1]) {
134b0fc5 1646 if (IS_ERR(pols[1])) {
0aa64774 1647 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
134b0fc5
JM		 1648 err = PTR_ERR(pols[1]);
1649 goto error;
1650 }
4e81bb83 1651 if (pols[1]->action == XFRM_POLICY_BLOCK) {
0aa64774 1652 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK);
4e81bb83
MN		 1653 err = -EPERM;
1654 goto error;
1655 }
1656 npols ++;
1657 xfrm_nr += pols[1]->xfrm_nr;
1658 }
1659 }
1660
1661 /*
1662 * Because neither flowi nor bundle information knows about
1663 * transformation template size. On more than one policy usage
1664 * we can realize whether all of them is bypass or not after
1665 * they are searched. See above not-transformed bypass
1666 * is surrounded by non-sub policy configuration, too.
1667 */
1668 if (xfrm_nr == 0) {
1669 /* Flow passes not transformed. */
1670 xfrm_pols_put(pols, npols);
1671 return 0;
1672 }
1673
1674#endif
1675 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1da177e4
LT		 1676
1677 if (unlikely(nx<0)) {
1678 err = nx;
14e50e57
DM		 1679 if (err == -EAGAIN && sysctl_xfrm_larval_drop) {
1680 /* EREMOTE tells the caller to generate
1681 * a one-shot blackhole route.
1682 */
d66e37a9 1683 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
14e50e57
DM		 1684 xfrm_pol_put(policy);
1685 return -EREMOTE;
1686 }
815f4e57 1687 if (err == -EAGAIN && (flags & XFRM_LOOKUP_WAIT)) {
1da177e4
LT		 1688 DECLARE_WAITQUEUE(wait, current);
1689
52479b62 1690 add_wait_queue(&net->xfrm.km_waitq, &wait);
1da177e4
LT		 1691 set_current_state(TASK_INTERRUPTIBLE);
1692 schedule();
1693 set_current_state(TASK_RUNNING);
52479b62 1694 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1da177e4 1695
4e81bb83 1696 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1da177e4
LT		 1697
1698 if (nx == -EAGAIN && signal_pending(current)) {
0aa64774 1699 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1da177e4
LT		 1700 err = -ERESTART;
1701 goto error;
1702 }
1703 if (nx == -EAGAIN ||
1704 genid != atomic_read(&flow_cache_genid)) {
4e81bb83 1705 xfrm_pols_put(pols, npols);
1da177e4
LT		 1706 goto restart;
1707 }
1708 err = nx;
1709 }
0aa64774
MN		 1710 if (err < 0) {
1711 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1da177e4 1712 goto error;
0aa64774 1713 }
1da177e4
LT		 1714 }
1715 if (nx == 0) {
1716 /* Flow passes not transformed. */
4e81bb83 1717 xfrm_pols_put(pols, npols);
1da177e4
LT		 1718 return 0;
1719 }
1720
25ee3286
HX		 1721 dst = xfrm_bundle_create(policy, xfrm, nx, fl, dst_orig);
1722 err = PTR_ERR(dst);
0aa64774
MN		 1723 if (IS_ERR(dst)) {
1724 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1da177e4 1725 goto error;
0aa64774 1726 }
1da177e4 1727
4e81bb83
MN		 1728 for (pi = 0; pi < npols; pi++) {
1729 read_lock_bh(&pols[pi]->lock);
12a169e7 1730 pol_dead |= pols[pi]->walk.dead;
4e81bb83
MN		 1731 read_unlock_bh(&pols[pi]->lock);
1732 }
1733
1da177e4 1734 write_lock_bh(&policy->lock);
4e81bb83 1735 if (unlikely(pol_dead || stale_bundle(dst))) {
1da177e4
LT		 1736 /* Wow! While we worked on resolving, this
1737 * policy has gone. Retry. It is not paranoia,
1738 * we just cannot enlist new bundle to dead object.
1739 * We can't enlist stable bundles either.
1740 */
1741 write_unlock_bh(&policy->lock);
9d7d7402 1742 dst_free(dst);
00de651d 1743
0aa64774
MN		 1744 if (pol_dead)
1745 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD);
1746 else
1747 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
00de651d
HX		 1748 err = -EHOSTUNREACH;
1749 goto error;
1da177e4 1750 }
157bfc25
MN		 1751
1752 if (npols > 1)
1753 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1754 else
1755 err = xfrm_dst_update_origin(dst, fl);
1756 if (unlikely(err)) {
1757 write_unlock_bh(&policy->lock);
9d7d7402 1758 dst_free(dst);
0aa64774 1759 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
157bfc25
MN		 1760 goto error;
1761 }
1762
1da177e4
LT		 1763 dst->next = policy->bundles;
1764 policy->bundles = dst;
1765 dst_hold(dst);
1766 write_unlock_bh(&policy->lock);
1767 }
1768 *dst_p = dst;
1769 dst_release(dst_orig);
a716c119 1770 xfrm_pols_put(pols, npols);
1da177e4
LT		 1771 return 0;
1772
1773error:
4e81bb83 1774 xfrm_pols_put(pols, npols);
75b8c133
HX		 1775dropdst:
1776 dst_release(dst_orig);
1da177e4
LT		 1777 *dst_p = NULL;
1778 return err;
8b7817f3
HX		 1779
1780nopol:
aef21785 1781 err = -ENOENT;
8b7817f3
HX		 1782 if (flags & XFRM_LOOKUP_ICMP)
1783 goto dropdst;
1784 return 0;
1da177e4 1785}
14e50e57
DM		 1786EXPORT_SYMBOL(__xfrm_lookup);
1787
52479b62 1788int xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
14e50e57
DM		 1789 struct sock *sk, int flags)
1790{
52479b62 1791 int err = __xfrm_lookup(net, dst_p, fl, sk, flags);
14e50e57
DM		 1792
1793 if (err == -EREMOTE) {
1794 dst_release(*dst_p);
1795 *dst_p = NULL;
1796 err = -EAGAIN;
1797 }
1798
1799 return err;
1800}
1da177e4
LT		 1801EXPORT_SYMBOL(xfrm_lookup);
1802
df0ba92a
MN		 1803static inline int
1804xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1805{
1806 struct xfrm_state *x;
df0ba92a
MN		 1807
1808 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1809 return 0;
1810 x = skb->sp->xvec[idx];
1811 if (!x->type->reject)
1812 return 0;
1ecafede 1813 return x->type->reject(x, skb, fl);
df0ba92a
MN		 1814}
1815
1da177e4
LT		 1816/* When skb is transformed back to its "native" form, we have to
1817 * check policy restrictions. At the moment we make this in maximally
1818 * stupid way. Shame on me. :-) Of course, connected sockets must
1819 * have policy cached at them.
1820 */
1821
1822static inline int
a716c119 1823xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1da177e4
LT		 1824 unsigned short family)
1825{
1826 if (xfrm_state_kern(x))
928ba416 1827 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 1828 return x->id.proto == tmpl->id.proto &&
1829 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1830 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1831 x->props.mode == tmpl->mode &&
c5d18e98 1832 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 1833 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 1834 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1835 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 1836}
1837
df0ba92a
MN		 1838/*
1839 * 0 or more than 0 is returned when validation is succeeded (either bypass
1840 * because of optional transport mode, or next index of the mathced secpath
1841 * state with the template.
1842 * -1 is returned when no matching template is found.
1843 * Otherwise "-2 - errored_index" is returned.
1844 */
1da177e4
LT		 1845static inline int
1846xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1847 unsigned short family)
1848{
1849 int idx = start;
1850
1851 if (tmpl->optional) {
7e49e6de 1852 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 1853 return start;
1854 } else
1855 start = -1;
1856 for (; idx < sp->len; idx++) {
dbe5b4aa 1857 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 1858 return ++idx;
df0ba92a
MN		 1859 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1860 if (start == -1)
1861 start = -2-idx;
1da177e4 1862 break;
df0ba92a 1863 }
1da177e4
LT		 1864 }
1865 return start;
1866}
1867
d5422efe
HX		 1868int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1869 unsigned int family, int reverse)
1da177e4
LT		 1870{
1871 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 1872 int err;
1da177e4
LT		 1873
1874 if (unlikely(afinfo == NULL))
1875 return -EAFNOSUPPORT;
1876
d5422efe 1877 afinfo->decode_session(skb, fl, reverse);
beb8d13b 1878 err = security_xfrm_decode_session(skb, &fl->secid);
1da177e4 1879 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 1880 return err;
1da177e4 1881}
d5422efe 1882EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 1883
df0ba92a 1884static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 1885{
1886 for (; k < sp->len; k++) {
df0ba92a 1887 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 1888 *idxp = k;
1da177e4 1889 return 1;
df0ba92a 1890 }
1da177e4
LT		 1891 }
1892
1893 return 0;
1894}
1895
a716c119 1896int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 1897 unsigned short family)
1898{
f6e1e25d 1899 struct net *net = dev_net(skb->dev);
1da177e4 1900 struct xfrm_policy *pol;
4e81bb83
MN		 1901 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1902 int npols = 0;
1903 int xfrm_nr;
1904 int pi;
d5422efe 1905 int reverse;
1da177e4 1906 struct flowi fl;
d5422efe 1907 u8 fl_dir;
df0ba92a 1908 int xerr_idx = -1;
1da177e4 1909
d5422efe
HX		 1910 reverse = dir & ~XFRM_POLICY_MASK;
1911 dir &= XFRM_POLICY_MASK;
1912 fl_dir = policy_to_flow_dir(dir);
1913
0aa64774
MN		 1914 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
1915 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
1da177e4 1916 return 0;
0aa64774
MN		 1917 }
1918
eb9c7ebe 1919 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 1920
1921 /* First, check used SA against their selectors. */
1922 if (skb->sp) {
1923 int i;
1924
1925 for (i=skb->sp->len-1; i>=0; i--) {
dbe5b4aa 1926 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774
MN		 1927 if (!xfrm_selector_match(&x->sel, &fl, family)) {
1928 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 1929 return 0;
0aa64774 1930 }
1da177e4
LT		 1931 }
1932 }
1933
1934 pol = NULL;
3bccfbc7 1935 if (sk && sk->sk_policy[dir]) {
e0d1caa7 1936 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774
MN		 1937 if (IS_ERR(pol)) {
1938 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 1939 return 0;
0aa64774 1940 }
3bccfbc7 1941 }
1da177e4
LT		 1942
1943 if (!pol)
f6e1e25d 1944 pol = flow_cache_lookup(net, &fl, family, fl_dir,
1da177e4
LT		 1945 xfrm_policy_lookup);
1946
0aa64774
MN		 1947 if (IS_ERR(pol)) {
1948 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
134b0fc5 1949 return 0;
0aa64774 1950 }
134b0fc5 1951
df0ba92a 1952 if (!pol) {
d1d9facf 1953 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 1954 xfrm_secpath_reject(xerr_idx, skb, &fl);
0aa64774 1955 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 1956 return 0;
1957 }
1958 return 1;
1959 }
1da177e4 1960
9d729f72 1961 pol->curlft.use_time = get_seconds();
1da177e4 1962
4e81bb83
MN		 1963 pols[0] = pol;
1964 npols ++;
1965#ifdef CONFIG_XFRM_SUB_POLICY
1966 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 1967 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 1968 &fl, family,
1969 XFRM_POLICY_IN);
1970 if (pols[1]) {
0aa64774
MN		 1971 if (IS_ERR(pols[1])) {
1972 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
134b0fc5 1973 return 0;
0aa64774 1974 }
9d729f72 1975 pols[1]->curlft.use_time = get_seconds();
4e81bb83
MN		 1976 npols ++;
1977 }
1978 }
1979#endif
1980
1da177e4
LT		 1981 if (pol->action == XFRM_POLICY_ALLOW) {
1982 struct sec_path *sp;
1983 static struct sec_path dummy;
4e81bb83 1984 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 1985 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 1986 struct xfrm_tmpl **tpp = tp;
1987 int ti = 0;
1da177e4
LT		 1988 int i, k;
1989
1990 if ((sp = skb->sp) == NULL)
1991 sp = &dummy;
1992
4e81bb83
MN		 1993 for (pi = 0; pi < npols; pi++) {
1994 if (pols[pi] != pol &&
0aa64774
MN		 1995 pols[pi]->action != XFRM_POLICY_ALLOW) {
1996 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 1997 goto reject;
0aa64774
MN		 1998 }
1999 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2000 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 2001 goto reject_error;
0aa64774 2002 }
4e81bb83
MN		 2003 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2004 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2005 }
2006 xfrm_nr = ti;
41a49cc3
MN		 2007 if (npols > 1) {
2008 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2009 tpp = stp;
2010 }
4e81bb83 2011
1da177e4
LT		 2012 /* For each tunnel xfrm, find the first matching tmpl.
2013 * For each tmpl before that, find corresponding xfrm.
2014 * Order is _important_. Later we will implement
2015 * some barriers, but at the moment barriers
2016 * are implied between each two transformations.
2017 */
4e81bb83
MN		 2018 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2019 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 2020 if (k < 0) {
d1d9facf
JM		 2021 if (k < -1)
2022 /* "-2 - errored_index" returned */
2023 xerr_idx = -(2+k);
0aa64774 2024 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2025 goto reject;
df0ba92a 2026 }
1da177e4
LT		 2027 }
2028
0aa64774
MN		 2029 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2030 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2031 goto reject;
0aa64774 2032 }
1da177e4 2033
4e81bb83 2034 xfrm_pols_put(pols, npols);
1da177e4
LT		 2035 return 1;
2036 }
0aa64774 2037 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2038
2039reject:
df0ba92a 2040 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2041reject_error:
2042 xfrm_pols_put(pols, npols);
1da177e4
LT		 2043 return 0;
2044}
2045EXPORT_SYMBOL(__xfrm_policy_check);
2046
2047int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2048{
99a66657 2049 struct net *net = dev_net(skb->dev);
1da177e4
LT		 2050 struct flowi fl;
2051
0aa64774
MN		 2052 if (xfrm_decode_session(skb, &fl, family) < 0) {
2053 /* XXX: we should have something like FWDHDRERROR here. */
2054 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
1da177e4 2055 return 0;
0aa64774 2056 }
1da177e4 2057
99a66657 2058 return xfrm_lookup(net, &skb->dst, &fl, NULL, 0) == 0;
1da177e4
LT		 2059}
2060EXPORT_SYMBOL(__xfrm_route_forward);
2061
d49c73c7
DM		 2062/* Optimize later using cookies and generation ids. */
2063
1da177e4
LT		 2064static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2065{
d49c73c7
DM		 2066 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2067 * to "-1" to force all XFRM destinations to get validated by
2068 * dst_ops->check on every use. We do this because when a
2069 * normal route referenced by an XFRM dst is obsoleted we do
2070 * not go looking around for all parent referencing XFRM dsts
2071 * so that we can invalidate them. It is just too much work.
2072 * Instead we make the checks here on every use. For example:
2073 *
2074 * XFRM dst A --> IPv4 dst X
2075 *
2076 * X is the "xdst->route" of A (X is also the "dst->path" of A
2077 * in this example). If X is marked obsolete, "A" will not
2078 * notice. That's what we are validating here via the
2079 * stale_bundle() check.
2080 *
2081 * When a policy's bundle is pruned, we dst_free() the XFRM
2082 * dst which causes it's ->obsolete field to be set to a
2083 * positive non-zero integer. If an XFRM dst has been pruned
2084 * like this, we want to force a new route lookup.
399c180a 2085 */
d49c73c7
DM		 2086 if (dst->obsolete < 0 && !stale_bundle(dst))
2087 return dst;
2088
1da177e4
LT		 2089 return NULL;
2090}
2091
2092static int stale_bundle(struct dst_entry *dst)
2093{
5b368e61 2094 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
1da177e4
LT		 2095}
2096
aabc9761 2097void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2098{
1da177e4 2099 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2100 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2101 dev_hold(dst->dev);
1da177e4
LT		 2102 dev_put(dev);
2103 }
2104}
aabc9761 2105EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2106
2107static void xfrm_link_failure(struct sk_buff *skb)
2108{
2109 /* Impossible. Such dst must be popped before reaches point of failure. */
2110 return;
2111}
2112
2113static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2114{
2115 if (dst) {
2116 if (dst->obsolete) {
2117 dst_release(dst);
2118 dst = NULL;
2119 }
2120 }
2121 return dst;
2122}
2123
2518c7c2
DM		 2124static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p)
2125{
2126 struct dst_entry *dst, **dstp;
2127
2128 write_lock(&pol->lock);
2129 dstp = &pol->bundles;
2130 while ((dst=*dstp) != NULL) {
2131 if (func(dst)) {
2132 *dstp = dst->next;
2133 dst->next = *gc_list_p;
2134 *gc_list_p = dst;
2135 } else {
2136 dstp = &dst->next;
2137 }
2138 }
2139 write_unlock(&pol->lock);
2140}
2141
3dd0b499 2142static void xfrm_prune_bundles(struct net *net, int (*func)(struct dst_entry *))
1da177e4 2143{
2518c7c2
DM		 2144 struct dst_entry *gc_list = NULL;
2145 int dir;
1da177e4
LT		 2146
2147 read_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 2148 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2149 struct xfrm_policy *pol;
2150 struct hlist_node *entry;
2151 struct hlist_head *table;
2152 int i;
4e81bb83 2153
2518c7c2 2154 hlist_for_each_entry(pol, entry,
3dd0b499 2155 &net->xfrm.policy_inexact[dir], bydst)
2518c7c2
DM		 2156 prune_one_bundle(pol, func, &gc_list);
2157
3dd0b499
AD		 2158 table = net->xfrm.policy_bydst[dir].table;
2159 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2
DM		 2160 hlist_for_each_entry(pol, entry, table + i, bydst)
2161 prune_one_bundle(pol, func, &gc_list);
1da177e4
LT		 2162 }
2163 }
2164 read_unlock_bh(&xfrm_policy_lock);
2165
2166 while (gc_list) {
2518c7c2 2167 struct dst_entry *dst = gc_list;
1da177e4
LT		 2168 gc_list = dst->next;
2169 dst_free(dst);
2170 }
2171}
2172
2173static int unused_bundle(struct dst_entry *dst)
2174{
2175 return !atomic_read(&dst->__refcnt);
2176}
2177
ddcfd796 2178static void __xfrm_garbage_collect(struct net *net)
1da177e4 2179{
ddcfd796 2180 xfrm_prune_bundles(net, unused_bundle);
1da177e4
LT		 2181}
2182
3dd0b499 2183static int xfrm_flush_bundles(struct net *net)
1da177e4 2184{
3dd0b499 2185 xfrm_prune_bundles(net, stale_bundle);
1da177e4
LT		 2186 return 0;
2187}
2188
25ee3286 2189static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2190{
2191 do {
2192 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2193 u32 pmtu, route_mtu_cached;
2194
2195 pmtu = dst_mtu(dst->child);
2196 xdst->child_mtu_cached = pmtu;
2197
2198 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2199
2200 route_mtu_cached = dst_mtu(xdst->route);
2201 xdst->route_mtu_cached = route_mtu_cached;
2202
2203 if (pmtu > route_mtu_cached)
2204 pmtu = route_mtu_cached;
2205
2206 dst->metrics[RTAX_MTU-1] = pmtu;
2207 } while ((dst = dst->next));
2208}
2209
1da177e4
LT		 2210/* Check that the bundle accepts the flow and its components are
2211 * still valid.
2212 */
2213
5b368e61
VY		 2214int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2215 struct flowi *fl, int family, int strict)
1da177e4
LT		 2216{
2217 struct dst_entry *dst = &first->u.dst;
2218 struct xfrm_dst *last;
2219 u32 mtu;
2220
92d63dec 2221 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2222 (dst->dev && !netif_running(dst->dev)))
2223 return 0;
157bfc25
MN		 2224#ifdef CONFIG_XFRM_SUB_POLICY
2225 if (fl) {
2226 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2227 return 0;
2228 if (first->partner &&
2229 !xfrm_selector_match(first->partner, fl, family))
2230 return 0;
2231 }
2232#endif
1da177e4
LT		 2233
2234 last = NULL;
2235
2236 do {
2237 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2238
2239 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2240 return 0;
67f83cbf
VY		 2241 if (fl && pol &&
2242 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
e0d1caa7 2243 return 0;
1da177e4
LT		 2244 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2245 return 0;
9d4a706d
DM		 2246 if (xdst->genid != dst->xfrm->genid)
2247 return 0;
e53820de 2248
1bfcb10f 2249 if (strict && fl &&
13996378 2250 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
e53820de
MN		 2251 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2252 return 0;
1da177e4
LT		 2253
2254 mtu = dst_mtu(dst->child);
2255 if (xdst->child_mtu_cached != mtu) {
2256 last = xdst;
2257 xdst->child_mtu_cached = mtu;
2258 }
2259
92d63dec 2260 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2261 return 0;
2262 mtu = dst_mtu(xdst->route);
2263 if (xdst->route_mtu_cached != mtu) {
2264 last = xdst;
2265 xdst->route_mtu_cached = mtu;
2266 }
2267
2268 dst = dst->child;
2269 } while (dst->xfrm);
2270
2271 if (likely(!last))
2272 return 1;
2273
2274 mtu = last->child_mtu_cached;
2275 for (;;) {
2276 dst = &last->u.dst;
2277
2278 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2279 if (mtu > last->route_mtu_cached)
2280 mtu = last->route_mtu_cached;
2281 dst->metrics[RTAX_MTU-1] = mtu;
2282
2283 if (last == first)
2284 break;
2285
bd0bf076 2286 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2287 last->child_mtu_cached = mtu;
2288 }
2289
2290 return 1;
2291}
2292
2293EXPORT_SYMBOL(xfrm_bundle_ok);
2294
1da177e4
LT		 2295int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2296{
2297 int err = 0;
2298 if (unlikely(afinfo == NULL))
2299 return -EINVAL;
2300 if (unlikely(afinfo->family >= NPROTO))
2301 return -EAFNOSUPPORT;
e959d812 2302 write_lock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2303 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2304 err = -ENOBUFS;
2305 else {
2306 struct dst_ops *dst_ops = afinfo->dst_ops;
2307 if (likely(dst_ops->kmem_cachep == NULL))
2308 dst_ops->kmem_cachep = xfrm_dst_cache;
2309 if (likely(dst_ops->check == NULL))
2310 dst_ops->check = xfrm_dst_check;
1da177e4
LT		 2311 if (likely(dst_ops->negative_advice == NULL))
2312 dst_ops->negative_advice = xfrm_negative_advice;
2313 if (likely(dst_ops->link_failure == NULL))
2314 dst_ops->link_failure = xfrm_link_failure;
1da177e4
LT		 2315 if (likely(afinfo->garbage_collect == NULL))
2316 afinfo->garbage_collect = __xfrm_garbage_collect;
2317 xfrm_policy_afinfo[afinfo->family] = afinfo;
2318 }
e959d812 2319 write_unlock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2320 return err;
2321}
2322EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2323
2324int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2325{
2326 int err = 0;
2327 if (unlikely(afinfo == NULL))
2328 return -EINVAL;
2329 if (unlikely(afinfo->family >= NPROTO))
2330 return -EAFNOSUPPORT;
e959d812 2331 write_lock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2332 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2333 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2334 err = -EINVAL;
2335 else {
2336 struct dst_ops *dst_ops = afinfo->dst_ops;
2337 xfrm_policy_afinfo[afinfo->family] = NULL;
2338 dst_ops->kmem_cachep = NULL;
2339 dst_ops->check = NULL;
1da177e4
LT		 2340 dst_ops->negative_advice = NULL;
2341 dst_ops->link_failure = NULL;
1da177e4
LT		 2342 afinfo->garbage_collect = NULL;
2343 }
2344 }
e959d812 2345 write_unlock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2346 return err;
2347}
2348EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2349
2350static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2351{
2352 struct xfrm_policy_afinfo *afinfo;
2353 if (unlikely(family >= NPROTO))
2354 return NULL;
2355 read_lock(&xfrm_policy_afinfo_lock);
2356 afinfo = xfrm_policy_afinfo[family];
546be240
HX		 2357 if (unlikely(!afinfo))
2358 read_unlock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2359 return afinfo;
2360}
2361
2362static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2363{
546be240
HX		 2364 read_unlock(&xfrm_policy_afinfo_lock);
2365}
2366
1da177e4
LT		 2367static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2368{
e9dc8653
EB		 2369 struct net_device *dev = ptr;
2370
1da177e4
LT		 2371 switch (event) {
2372 case NETDEV_DOWN:
3dd0b499 2373 xfrm_flush_bundles(dev_net(dev));
1da177e4
LT		 2374 }
2375 return NOTIFY_DONE;
2376}
2377
2378static struct notifier_block xfrm_dev_notifier = {
d5917a35 2379 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2380};
2381
558f82ef
MN		 2382#ifdef CONFIG_XFRM_STATISTICS
2383static int __init xfrm_statistics_init(void)
2384{
2385 if (snmp_mib_init((void **)xfrm_statistics,
2386 sizeof(struct linux_xfrm_mib)) < 0)
2387 return -ENOMEM;
2388 return 0;
2389}
2390#endif
2391
d62ddc21 2392static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2393{
2518c7c2
DM		 2394 unsigned int hmask, sz;
2395 int dir;
2396
d62ddc21
AD		 2397 if (net_eq(net, &init_net))
2398 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2399 sizeof(struct xfrm_dst),
e5d679f3 2400 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2401 NULL);
1da177e4 2402
2518c7c2
DM		 2403 hmask = 8 - 1;
2404 sz = (hmask+1) * sizeof(struct hlist_head);
2405
93b851c1
AD		 2406 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2407 if (!net->xfrm.policy_byidx)
2408 goto out_byidx;
8100bea7 2409 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2410
2411 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2412 struct xfrm_policy_hash *htab;
2413
dc2caba7 2414 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2415 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2416
a35f6c5d 2417 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2418 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2419 if (!htab->table)
a35f6c5d
AD		 2420 goto out_bydst;
2421 htab->hmask = hmask;
2518c7c2
DM		 2422 }
2423
adfcf0b2 2424 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2425 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2426 if (net_eq(net, &init_net))
2427 register_netdevice_notifier(&xfrm_dev_notifier);
2428 return 0;
93b851c1 2429
a35f6c5d
AD		 2430out_bydst:
2431 for (dir--; dir >= 0; dir--) {
2432 struct xfrm_policy_hash *htab;
2433
2434 htab = &net->xfrm.policy_bydst[dir];
2435 xfrm_hash_free(htab->table, sz);
2436 }
2437 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2438out_byidx:
2439 return -ENOMEM;
d62ddc21
AD		 2440}
2441
2442static void xfrm_policy_fini(struct net *net)
2443{
93b851c1 2444 unsigned int sz;
8b18f8ea 2445 int dir;
93b851c1 2446
adfcf0b2 2447 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2448
8b18f8ea 2449 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2450 struct xfrm_policy_hash *htab;
2451
8b18f8ea 2452 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2453
2454 htab = &net->xfrm.policy_bydst[dir];
2455 sz = (htab->hmask + 1);
2456 WARN_ON(!hlist_empty(htab->table));
2457 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2458 }
2459
8100bea7 2460 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2461 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2462 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2463}
2464
d62ddc21
AD		 2465static int __net_init xfrm_net_init(struct net *net)
2466{
2467 int rv;
2468
2469 rv = xfrm_state_init(net);
2470 if (rv < 0)
2471 goto out_state;
2472 rv = xfrm_policy_init(net);
2473 if (rv < 0)
2474 goto out_policy;
2475 return 0;
2476
2477out_policy:
2478 xfrm_state_fini(net);
2479out_state:
2480 return rv;
2481}
2482
2483static void __net_exit xfrm_net_exit(struct net *net)
2484{
2485 xfrm_policy_fini(net);
2486 xfrm_state_fini(net);
2487}
2488
2489static struct pernet_operations __net_initdata xfrm_net_ops = {
2490 .init = xfrm_net_init,
2491 .exit = xfrm_net_exit,
2492};
2493
1da177e4
LT		 2494void __init xfrm_init(void)
2495{
d62ddc21 2496 register_pernet_subsys(&xfrm_net_ops);
558f82ef
MN		 2497#ifdef CONFIG_XFRM_STATISTICS
2498 xfrm_statistics_init();
2499#endif
1da177e4 2500 xfrm_input_init();
558f82ef
MN		 2501#ifdef CONFIG_XFRM_STATISTICS
2502 xfrm_proc_init();
2503#endif
1da177e4
LT		 2504}
2505
ab5f5e8b 2506#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2507static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2508 struct audit_buffer *audit_buf)
ab5f5e8b 2509{
875179fa
PM		 2510 struct xfrm_sec_ctx *ctx = xp->security;
2511 struct xfrm_selector *sel = &xp->selector;
2512
2513 if (ctx)
ab5f5e8b 2514 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2515 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2516
875179fa 2517 switch(sel->family) {
ab5f5e8b 2518 case AF_INET:
21454aaa 2519 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2520 if (sel->prefixlen_s != 32)
2521 audit_log_format(audit_buf, " src_prefixlen=%d",
2522 sel->prefixlen_s);
21454aaa 2523 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2524 if (sel->prefixlen_d != 32)
2525 audit_log_format(audit_buf, " dst_prefixlen=%d",
2526 sel->prefixlen_d);
ab5f5e8b
JL		 2527 break;
2528 case AF_INET6:
5b095d98 2529 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 2530 if (sel->prefixlen_s != 128)
2531 audit_log_format(audit_buf, " src_prefixlen=%d",
2532 sel->prefixlen_s);
5b095d98 2533 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 2534 if (sel->prefixlen_d != 128)
2535 audit_log_format(audit_buf, " dst_prefixlen=%d",
2536 sel->prefixlen_d);
ab5f5e8b
JL		 2537 break;
2538 }
2539}
2540
68277acc 2541void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2532386f 2542 uid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 2543{
2544 struct audit_buffer *audit_buf;
ab5f5e8b 2545
afeb14b4 2546 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 2547 if (audit_buf == NULL)
2548 return;
2532386f 2549 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 2550 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 2551 xfrm_audit_common_policyinfo(xp, audit_buf);
2552 audit_log_end(audit_buf);
2553}
2554EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2555
68277acc 2556void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2532386f 2557 uid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 2558{
2559 struct audit_buffer *audit_buf;
ab5f5e8b 2560
afeb14b4 2561 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 2562 if (audit_buf == NULL)
2563 return;
2532386f 2564 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 2565 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 2566 xfrm_audit_common_policyinfo(xp, audit_buf);
2567 audit_log_end(audit_buf);
2568}
2569EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2570#endif
2571
80c9abaa
SS		 2572#ifdef CONFIG_XFRM_MIGRATE
2573static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2574 struct xfrm_selector *sel_tgt)
2575{
2576 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2577 if (sel_tgt->family == sel_cmp->family &&
2578 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
a716c119 2579 sel_cmp->family) == 0 &&
80c9abaa
SS		 2580 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2581 sel_cmp->family) == 0 &&
2582 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2583 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2584 return 1;
2585 }
2586 } else {
2587 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2588 return 1;
2589 }
2590 }
2591 return 0;
2592}
2593
2594static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2595 u8 dir, u8 type)
2596{
2597 struct xfrm_policy *pol, *ret = NULL;
2598 struct hlist_node *entry;
2599 struct hlist_head *chain;
2600 u32 priority = ~0U;
2601
2602 read_lock_bh(&xfrm_policy_lock);
1121994c 2603 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
80c9abaa
SS		 2604 hlist_for_each_entry(pol, entry, chain, bydst) {
2605 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2606 pol->type == type) {
2607 ret = pol;
2608 priority = ret->priority;
2609 break;
2610 }
2611 }
8b18f8ea 2612 chain = &init_net.xfrm.policy_inexact[dir];
80c9abaa
SS		 2613 hlist_for_each_entry(pol, entry, chain, bydst) {
2614 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2615 pol->type == type &&
2616 pol->priority < priority) {
2617 ret = pol;
2618 break;
2619 }
2620 }
2621
2622 if (ret)
2623 xfrm_pol_hold(ret);
2624
2625 read_unlock_bh(&xfrm_policy_lock);
2626
2627 return ret;
2628}
2629
2630static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2631{
2632 int match = 0;
2633
2634 if (t->mode == m->mode && t->id.proto == m->proto &&
2635 (m->reqid == 0 || t->reqid == m->reqid)) {
2636 switch (t->mode) {
2637 case XFRM_MODE_TUNNEL:
2638 case XFRM_MODE_BEET:
2639 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2640 m->old_family) == 0 &&
2641 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2642 m->old_family) == 0) {
2643 match = 1;
2644 }
2645 break;
2646 case XFRM_MODE_TRANSPORT:
2647 /* in case of transport mode, template does not store
2648 any IP addresses, hence we just compare mode and
2649 protocol */
2650 match = 1;
2651 break;
2652 default:
2653 break;
2654 }
2655 }
2656 return match;
2657}
2658
2659/* update endpoint address(es) of template(s) */
2660static int xfrm_policy_migrate(struct xfrm_policy *pol,
2661 struct xfrm_migrate *m, int num_migrate)
2662{
2663 struct xfrm_migrate *mp;
2664 struct dst_entry *dst;
2665 int i, j, n = 0;
2666
2667 write_lock_bh(&pol->lock);
12a169e7 2668 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 2669 /* target policy has been deleted */
2670 write_unlock_bh(&pol->lock);
2671 return -ENOENT;
2672 }
2673
2674 for (i = 0; i < pol->xfrm_nr; i++) {
2675 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2676 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2677 continue;
2678 n++;
1bfcb10f
HX		 2679 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2680 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 2681 continue;
2682 /* update endpoints */
2683 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2684 sizeof(pol->xfrm_vec[i].id.daddr));
2685 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2686 sizeof(pol->xfrm_vec[i].saddr));
2687 pol->xfrm_vec[i].encap_family = mp->new_family;
2688 /* flush bundles */
2689 while ((dst = pol->bundles) != NULL) {
2690 pol->bundles = dst->next;
2691 dst_free(dst);
2692 }
2693 }
2694 }
2695
2696 write_unlock_bh(&pol->lock);
2697
2698 if (!n)
2699 return -ENODATA;
2700
2701 return 0;
2702}
2703
2704static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2705{
2706 int i, j;
2707
2708 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2709 return -EINVAL;
2710
2711 for (i = 0; i < num_migrate; i++) {
2712 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2713 m[i].old_family) == 0) &&
2714 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2715 m[i].old_family) == 0))
2716 return -EINVAL;
2717 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2718 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2719 return -EINVAL;
2720
2721 /* check if there is any duplicated entry */
2722 for (j = i + 1; j < num_migrate; j++) {
2723 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2724 sizeof(m[i].old_daddr)) &&
2725 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2726 sizeof(m[i].old_saddr)) &&
2727 m[i].proto == m[j].proto &&
2728 m[i].mode == m[j].mode &&
2729 m[i].reqid == m[j].reqid &&
2730 m[i].old_family == m[j].old_family)
2731 return -EINVAL;
2732 }
2733 }
2734
2735 return 0;
2736}
2737
2738int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189
AE		 2739 struct xfrm_migrate *m, int num_migrate,
2740 struct xfrm_kmaddress *k)
80c9abaa
SS		 2741{
2742 int i, err, nx_cur = 0, nx_new = 0;
2743 struct xfrm_policy *pol = NULL;
2744 struct xfrm_state *x, *xc;
2745 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2746 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2747 struct xfrm_migrate *mp;
2748
2749 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2750 goto out;
2751
2752 /* Stage 1 - find policy */
2753 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2754 err = -ENOENT;
2755 goto out;
2756 }
2757
2758 /* Stage 2 - find and update state(s) */
2759 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2760 if ((x = xfrm_migrate_state_find(mp))) {
2761 x_cur[nx_cur] = x;
2762 nx_cur++;
2763 if ((xc = xfrm_state_migrate(x, mp))) {
2764 x_new[nx_new] = xc;
2765 nx_new++;
2766 } else {
2767 err = -ENODATA;
2768 goto restore_state;
2769 }
2770 }
2771 }
2772
2773 /* Stage 3 - update policy */
2774 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2775 goto restore_state;
2776
2777 /* Stage 4 - delete old state(s) */
2778 if (nx_cur) {
2779 xfrm_states_put(x_cur, nx_cur);
2780 xfrm_states_delete(x_cur, nx_cur);
2781 }
2782
2783 /* Stage 5 - announce */
13c1d189 2784 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 2785
2786 xfrm_pol_put(pol);
2787
2788 return 0;
2789out:
2790 return err;
2791
2792restore_state:
2793 if (pol)
2794 xfrm_pol_put(pol);
2795 if (nx_cur)
2796 xfrm_states_put(x_cur, nx_cur);
2797 if (nx_new)
2798 xfrm_states_delete(x_new, nx_new);
2799
2800 return err;
2801}
e610e679 2802EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 2803#endif
kernel source for telekom puls (alcatel/mediatek)