projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
sky2: revert config space change
[linux-2.6-block.git] / net / xfrm / xfrm_policy.c
CommitLineData
a716c119 1/*
1da177e4
LT		 2 * xfrm_policy.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * Kazunori MIYAZAWA @USAGI
10 * YOSHIFUJI Hideaki
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
df71837d 13 *
1da177e4
LT		 14 */
15
66cdb3ca 16#include <linux/err.h>
1da177e4
LT		 17#include <linux/slab.h>
18#include <linux/kmod.h>
19#include <linux/list.h>
20#include <linux/spinlock.h>
21#include <linux/workqueue.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
eb9c7ebe 24#include <linux/netfilter.h>
1da177e4 25#include <linux/module.h>
2518c7c2 26#include <linux/cache.h>
68277acc 27#include <linux/audit.h>
25ee3286 28#include <net/dst.h>
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
4a3e2f71
AV		 37DEFINE_MUTEX(xfrm_cfg_mutex);
38EXPORT_SYMBOL(xfrm_cfg_mutex);
1da177e4
LT		 39
40static DEFINE_RWLOCK(xfrm_policy_lock);
41
1da177e4
LT		 42static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
43static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
44
e18b890b 45static struct kmem_cache *xfrm_dst_cache __read_mostly;
1da177e4 46
2518c7c2 47static HLIST_HEAD(xfrm_policy_gc_list);
1da177e4
LT		 48static DEFINE_SPINLOCK(xfrm_policy_gc_lock);
49
50static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
51static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
25ee3286 52static void xfrm_init_pmtu(struct dst_entry *dst);
1da177e4 53
29fa0b30
WY		 54static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
55 int dir);
56
77681021
AM		 57static inline int
58__xfrm4_selector_match(struct xfrm_selector *sel, struct flowi *fl)
59{
60 return addr_match(&fl->fl4_dst, &sel->daddr, sel->prefixlen_d) &&
61 addr_match(&fl->fl4_src, &sel->saddr, sel->prefixlen_s) &&
62 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
63 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
64 (fl->proto == sel->proto || !sel->proto) &&
65 (fl->oif == sel->ifindex || !sel->ifindex);
66}
67
68static inline int
69__xfrm6_selector_match(struct xfrm_selector *sel, struct flowi *fl)
70{
71 return addr_match(&fl->fl6_dst, &sel->daddr, sel->prefixlen_d) &&
72 addr_match(&fl->fl6_src, &sel->saddr, sel->prefixlen_s) &&
73 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
74 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
75 (fl->proto == sel->proto || !sel->proto) &&
76 (fl->oif == sel->ifindex || !sel->ifindex);
77}
78
79int xfrm_selector_match(struct xfrm_selector *sel, struct flowi *fl,
80 unsigned short family)
81{
82 switch (family) {
83 case AF_INET:
84 return __xfrm4_selector_match(sel, fl);
85 case AF_INET6:
86 return __xfrm6_selector_match(sel, fl);
87 }
88 return 0;
89}
90
c5b3cf46 91static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
9bb182a7
YH		 92 xfrm_address_t *saddr,
93 xfrm_address_t *daddr,
94 int family)
95{
96 struct xfrm_policy_afinfo *afinfo;
97 struct dst_entry *dst;
98
99 afinfo = xfrm_policy_get_afinfo(family);
100 if (unlikely(afinfo == NULL))
101 return ERR_PTR(-EAFNOSUPPORT);
102
c5b3cf46 103 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
9bb182a7
YH		 104
105 xfrm_policy_put_afinfo(afinfo);
106
107 return dst;
108}
109
25ee3286 110static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
9bb182a7
YH		 111 xfrm_address_t *prev_saddr,
112 xfrm_address_t *prev_daddr,
25ee3286 113 int family)
1da177e4 114{
c5b3cf46 115 struct net *net = xs_net(x);
66cdb3ca
HX		 116 xfrm_address_t *saddr = &x->props.saddr;
117 xfrm_address_t *daddr = &x->id.daddr;
66cdb3ca
HX		 118 struct dst_entry *dst;
119
9bb182a7 120 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
66cdb3ca 121 saddr = x->coaddr;
9bb182a7
YH		 122 daddr = prev_daddr;
123 }
124 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
125 saddr = prev_saddr;
66cdb3ca 126 daddr = x->coaddr;
9bb182a7 127 }
1da177e4 128
c5b3cf46 129 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
9bb182a7
YH		 130
131 if (!IS_ERR(dst)) {
132 if (prev_saddr != saddr)
133 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
134 if (prev_daddr != daddr)
135 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
136 }
1da177e4 137
66cdb3ca 138 return dst;
1da177e4 139}
1da177e4 140
1da177e4
LT		 141static inline unsigned long make_jiffies(long secs)
142{
143 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
144 return MAX_SCHEDULE_TIMEOUT-1;
145 else
a716c119 146 return secs*HZ;
1da177e4
LT		 147}
148
149static void xfrm_policy_timer(unsigned long data)
150{
151 struct xfrm_policy *xp = (struct xfrm_policy*)data;
9d729f72 152 unsigned long now = get_seconds();
1da177e4
LT		 153 long next = LONG_MAX;
154 int warn = 0;
155 int dir;
156
157 read_lock(&xp->lock);
158
12a169e7 159 if (xp->walk.dead)
1da177e4
LT		 160 goto out;
161
77d8d7a6 162 dir = xfrm_policy_id2dir(xp->index);
1da177e4
LT		 163
164 if (xp->lft.hard_add_expires_seconds) {
165 long tmo = xp->lft.hard_add_expires_seconds +
166 xp->curlft.add_time - now;
167 if (tmo <= 0)
168 goto expired;
169 if (tmo < next)
170 next = tmo;
171 }
172 if (xp->lft.hard_use_expires_seconds) {
173 long tmo = xp->lft.hard_use_expires_seconds +
174 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
175 if (tmo <= 0)
176 goto expired;
177 if (tmo < next)
178 next = tmo;
179 }
180 if (xp->lft.soft_add_expires_seconds) {
181 long tmo = xp->lft.soft_add_expires_seconds +
182 xp->curlft.add_time - now;
183 if (tmo <= 0) {
184 warn = 1;
185 tmo = XFRM_KM_TIMEOUT;
186 }
187 if (tmo < next)
188 next = tmo;
189 }
190 if (xp->lft.soft_use_expires_seconds) {
191 long tmo = xp->lft.soft_use_expires_seconds +
192 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
193 if (tmo <= 0) {
194 warn = 1;
195 tmo = XFRM_KM_TIMEOUT;
196 }
197 if (tmo < next)
198 next = tmo;
199 }
200
201 if (warn)
6c5c8ca7 202 km_policy_expired(xp, dir, 0, 0);
1da177e4
LT		 203 if (next != LONG_MAX &&
204 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
205 xfrm_pol_hold(xp);
206
207out:
208 read_unlock(&xp->lock);
209 xfrm_pol_put(xp);
210 return;
211
212expired:
213 read_unlock(&xp->lock);
4666faab 214 if (!xfrm_policy_delete(xp, dir))
6c5c8ca7 215 km_policy_expired(xp, dir, 1, 0);
1da177e4
LT		 216 xfrm_pol_put(xp);
217}
218
219
220/* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
221 * SPD calls.
222 */
223
0331b1f3 224struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
1da177e4
LT		 225{
226 struct xfrm_policy *policy;
227
0da974f4 228 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
1da177e4
LT		 229
230 if (policy) {
0331b1f3 231 write_pnet(&policy->xp_net, net);
12a169e7 232 INIT_LIST_HEAD(&policy->walk.all);
2518c7c2
DM		 233 INIT_HLIST_NODE(&policy->bydst);
234 INIT_HLIST_NODE(&policy->byidx);
1da177e4 235 rwlock_init(&policy->lock);
2518c7c2 236 atomic_set(&policy->refcnt, 1);
b24b8a24
PE		 237 setup_timer(&policy->timer, xfrm_policy_timer,
238 (unsigned long)policy);
1da177e4
LT		 239 }
240 return policy;
241}
242EXPORT_SYMBOL(xfrm_policy_alloc);
243
244/* Destroy xfrm_policy: descendant resources must be released to this moment. */
245
64c31b3f 246void xfrm_policy_destroy(struct xfrm_policy *policy)
1da177e4 247{
12a169e7 248 BUG_ON(!policy->walk.dead);
1da177e4 249
09a62660 250 BUG_ON(policy->bundles);
1da177e4
LT		 251
252 if (del_timer(&policy->timer))
253 BUG();
254
03e1ad7b 255 security_xfrm_policy_free(policy->security);
1da177e4
LT		 256 kfree(policy);
257}
64c31b3f 258EXPORT_SYMBOL(xfrm_policy_destroy);
1da177e4
LT		 259
260static void xfrm_policy_gc_kill(struct xfrm_policy *policy)
261{
262 struct dst_entry *dst;
263
264 while ((dst = policy->bundles) != NULL) {
265 policy->bundles = dst->next;
266 dst_free(dst);
267 }
268
269 if (del_timer(&policy->timer))
270 atomic_dec(&policy->refcnt);
271
272 if (atomic_read(&policy->refcnt) > 1)
273 flow_cache_flush();
274
275 xfrm_pol_put(policy);
276}
277
c4028958 278static void xfrm_policy_gc_task(struct work_struct *work)
1da177e4
LT		 279{
280 struct xfrm_policy *policy;
2518c7c2
DM		 281 struct hlist_node *entry, *tmp;
282 struct hlist_head gc_list;
1da177e4
LT		 283
284 spin_lock_bh(&xfrm_policy_gc_lock);
2518c7c2
DM		 285 gc_list.first = xfrm_policy_gc_list.first;
286 INIT_HLIST_HEAD(&xfrm_policy_gc_list);
1da177e4
LT		 287 spin_unlock_bh(&xfrm_policy_gc_lock);
288
2518c7c2 289 hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst)
1da177e4 290 xfrm_policy_gc_kill(policy);
1da177e4 291}
c9583969 292static DECLARE_WORK(xfrm_policy_gc_work, xfrm_policy_gc_task);
1da177e4
LT		 293
294/* Rule must be locked. Release descentant resources, announce
295 * entry dead. The rule must be unlinked from lists to the moment.
296 */
297
298static void xfrm_policy_kill(struct xfrm_policy *policy)
299{
300 int dead;
301
302 write_lock_bh(&policy->lock);
12a169e7
HX		 303 dead = policy->walk.dead;
304 policy->walk.dead = 1;
1da177e4
LT		 305 write_unlock_bh(&policy->lock);
306
307 if (unlikely(dead)) {
308 WARN_ON(1);
309 return;
310 }
311
bbb770e7 312 spin_lock_bh(&xfrm_policy_gc_lock);
2518c7c2 313 hlist_add_head(&policy->bydst, &xfrm_policy_gc_list);
bbb770e7 314 spin_unlock_bh(&xfrm_policy_gc_lock);
1da177e4
LT		 315
316 schedule_work(&xfrm_policy_gc_work);
317}
318
2518c7c2
DM		 319static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
320
e92303f8 321static inline unsigned int idx_hash(struct net *net, u32 index)
2518c7c2 322{
e92303f8 323 return __idx_hash(index, net->xfrm.policy_idx_hmask);
2518c7c2
DM		 324}
325
1121994c 326static struct hlist_head *policy_hash_bysel(struct net *net, struct xfrm_selector *sel, unsigned short family, int dir)
2518c7c2 327{
1121994c 328 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 329 unsigned int hash = __sel_hash(sel, family, hmask);
330
331 return (hash == hmask + 1 ?
1121994c
AD		 332 &net->xfrm.policy_inexact[dir] :
333 net->xfrm.policy_bydst[dir].table + hash);
2518c7c2
DM		 334}
335
1121994c 336static struct hlist_head *policy_hash_direct(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
2518c7c2 337{
1121994c 338 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 339 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
340
1121994c 341 return net->xfrm.policy_bydst[dir].table + hash;
2518c7c2
DM		 342}
343
2518c7c2
DM		 344static void xfrm_dst_hash_transfer(struct hlist_head *list,
345 struct hlist_head *ndsttable,
346 unsigned int nhashmask)
347{
b791160b 348 struct hlist_node *entry, *tmp, *entry0 = NULL;
2518c7c2 349 struct xfrm_policy *pol;
b791160b 350 unsigned int h0 = 0;
2518c7c2 351
b791160b 352redo:
2518c7c2
DM		 353 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
354 unsigned int h;
355
356 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
357 pol->family, nhashmask);
b791160b
YH		 358 if (!entry0) {
359 hlist_del(entry);
360 hlist_add_head(&pol->bydst, ndsttable+h);
361 h0 = h;
362 } else {
363 if (h != h0)
364 continue;
365 hlist_del(entry);
366 hlist_add_after(entry0, &pol->bydst);
367 }
368 entry0 = entry;
369 }
370 if (!hlist_empty(list)) {
371 entry0 = NULL;
372 goto redo;
2518c7c2
DM		 373 }
374}
375
376static void xfrm_idx_hash_transfer(struct hlist_head *list,
377 struct hlist_head *nidxtable,
378 unsigned int nhashmask)
379{
380 struct hlist_node *entry, *tmp;
381 struct xfrm_policy *pol;
382
383 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
384 unsigned int h;
385
386 h = __idx_hash(pol->index, nhashmask);
387 hlist_add_head(&pol->byidx, nidxtable+h);
388 }
389}
390
391static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
392{
393 return ((old_hmask + 1) << 1) - 1;
394}
395
66caf628 396static void xfrm_bydst_resize(struct net *net, int dir)
2518c7c2 397{
66caf628 398 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 399 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
400 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 401 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
44e36b42 402 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
2518c7c2
DM		 403 int i;
404
405 if (!ndst)
406 return;
407
408 write_lock_bh(&xfrm_policy_lock);
409
410 for (i = hmask; i >= 0; i--)
411 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
412
66caf628
AD		 413 net->xfrm.policy_bydst[dir].table = ndst;
414 net->xfrm.policy_bydst[dir].hmask = nhashmask;
2518c7c2
DM		 415
416 write_unlock_bh(&xfrm_policy_lock);
417
44e36b42 418 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 419}
420
66caf628 421static void xfrm_byidx_resize(struct net *net, int total)
2518c7c2 422{
66caf628 423 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 424 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
425 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
66caf628 426 struct hlist_head *oidx = net->xfrm.policy_byidx;
44e36b42 427 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
2518c7c2
DM		 428 int i;
429
430 if (!nidx)
431 return;
432
433 write_lock_bh(&xfrm_policy_lock);
434
435 for (i = hmask; i >= 0; i--)
436 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
437
66caf628
AD		 438 net->xfrm.policy_byidx = nidx;
439 net->xfrm.policy_idx_hmask = nhashmask;
2518c7c2
DM		 440
441 write_unlock_bh(&xfrm_policy_lock);
442
44e36b42 443 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
2518c7c2
DM		 444}
445
66caf628 446static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
2518c7c2 447{
66caf628
AD		 448 unsigned int cnt = net->xfrm.policy_count[dir];
449 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
2518c7c2
DM		 450
451 if (total)
452 *total += cnt;
453
454 if ((hmask + 1) < xfrm_policy_hashmax &&
455 cnt > hmask)
456 return 1;
457
458 return 0;
459}
460
66caf628 461static inline int xfrm_byidx_should_resize(struct net *net, int total)
2518c7c2 462{
66caf628 463 unsigned int hmask = net->xfrm.policy_idx_hmask;
2518c7c2
DM		 464
465 if ((hmask + 1) < xfrm_policy_hashmax &&
466 total > hmask)
467 return 1;
468
469 return 0;
470}
471
e071041b 472void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
ecfd6b18
JHS		 473{
474 read_lock_bh(&xfrm_policy_lock);
e071041b
AD		 475 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
476 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
477 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
478 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
479 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
480 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
481 si->spdhcnt = net->xfrm.policy_idx_hmask;
ecfd6b18
JHS		 482 si->spdhmcnt = xfrm_policy_hashmax;
483 read_unlock_bh(&xfrm_policy_lock);
484}
485EXPORT_SYMBOL(xfrm_spd_getinfo);
2518c7c2 486
ecfd6b18 487static DEFINE_MUTEX(hash_resize_mutex);
66caf628 488static void xfrm_hash_resize(struct work_struct *work)
2518c7c2 489{
66caf628 490 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
2518c7c2
DM		 491 int dir, total;
492
493 mutex_lock(&hash_resize_mutex);
494
495 total = 0;
496 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
66caf628
AD		 497 if (xfrm_bydst_should_resize(net, dir, &total))
498 xfrm_bydst_resize(net, dir);
2518c7c2 499 }
66caf628
AD		 500 if (xfrm_byidx_should_resize(net, total))
501 xfrm_byidx_resize(net, total);
2518c7c2
DM		 502
503 mutex_unlock(&hash_resize_mutex);
504}
505
1da177e4
LT		 506/* Generate new index... KAME seems to generate them ordered by cost
507 * of an absolute inpredictability of ordering of rules. This will not pass. */
1121994c 508static u32 xfrm_gen_index(struct net *net, int dir)
1da177e4 509{
1da177e4
LT		 510 static u32 idx_generator;
511
512 for (;;) {
2518c7c2
DM		 513 struct hlist_node *entry;
514 struct hlist_head *list;
515 struct xfrm_policy *p;
516 u32 idx;
517 int found;
518
1da177e4
LT		 519 idx = (idx_generator | dir);
520 idx_generator += 8;
521 if (idx == 0)
522 idx = 8;
1121994c 523 list = net->xfrm.policy_byidx + idx_hash(net, idx);
2518c7c2
DM		 524 found = 0;
525 hlist_for_each_entry(p, entry, list, byidx) {
526 if (p->index == idx) {
527 found = 1;
1da177e4 528 break;
2518c7c2 529 }
1da177e4 530 }
2518c7c2 531 if (!found)
1da177e4
LT		 532 return idx;
533 }
534}
535
2518c7c2
DM		 536static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
537{
538 u32 *p1 = (u32 *) s1;
539 u32 *p2 = (u32 *) s2;
540 int len = sizeof(struct xfrm_selector) / sizeof(u32);
541 int i;
542
543 for (i = 0; i < len; i++) {
544 if (p1[i] != p2[i])
545 return 1;
546 }
547
548 return 0;
549}
550
1da177e4
LT		 551int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
552{
1121994c 553 struct net *net = xp_net(policy);
2518c7c2
DM		 554 struct xfrm_policy *pol;
555 struct xfrm_policy *delpol;
556 struct hlist_head *chain;
a6c7ab55 557 struct hlist_node *entry, *newpos;
9b78a82c 558 struct dst_entry *gc_list;
1da177e4
LT		 559
560 write_lock_bh(&xfrm_policy_lock);
1121994c 561 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
2518c7c2
DM		 562 delpol = NULL;
563 newpos = NULL;
2518c7c2 564 hlist_for_each_entry(pol, entry, chain, bydst) {
a6c7ab55 565 if (pol->type == policy->type &&
2518c7c2 566 !selector_cmp(&pol->selector, &policy->selector) &&
a6c7ab55
HX		 567 xfrm_sec_ctx_match(pol->security, policy->security) &&
568 !WARN_ON(delpol)) {
1da177e4
LT		 569 if (excl) {
570 write_unlock_bh(&xfrm_policy_lock);
571 return -EEXIST;
572 }
1da177e4
LT		 573 delpol = pol;
574 if (policy->priority > pol->priority)
575 continue;
576 } else if (policy->priority >= pol->priority) {
a6c7ab55 577 newpos = &pol->bydst;
1da177e4
LT		 578 continue;
579 }
1da177e4
LT		 580 if (delpol)
581 break;
1da177e4
LT		 582 }
583 if (newpos)
2518c7c2
DM		 584 hlist_add_after(newpos, &policy->bydst);
585 else
586 hlist_add_head(&policy->bydst, chain);
1da177e4 587 xfrm_pol_hold(policy);
1121994c 588 net->xfrm.policy_count[dir]++;
1da177e4 589 atomic_inc(&flow_cache_genid);
29fa0b30
WY		 590 if (delpol)
591 __xfrm_policy_unlink(delpol, dir);
1121994c
AD		 592 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
593 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
9d729f72 594 policy->curlft.add_time = get_seconds();
1da177e4
LT		 595 policy->curlft.use_time = 0;
596 if (!mod_timer(&policy->timer, jiffies + HZ))
597 xfrm_pol_hold(policy);
1121994c 598 list_add(&policy->walk.all, &net->xfrm.policy_all);
1da177e4
LT		 599 write_unlock_bh(&xfrm_policy_lock);
600
9b78a82c 601 if (delpol)
1da177e4 602 xfrm_policy_kill(delpol);
1121994c
AD		 603 else if (xfrm_bydst_should_resize(net, dir, NULL))
604 schedule_work(&net->xfrm.policy_hash_work);
9b78a82c
DM		 605
606 read_lock_bh(&xfrm_policy_lock);
607 gc_list = NULL;
2518c7c2
DM		 608 entry = &policy->bydst;
609 hlist_for_each_entry_continue(policy, entry, bydst) {
9b78a82c
DM		 610 struct dst_entry *dst;
611
612 write_lock(&policy->lock);
613 dst = policy->bundles;
614 if (dst) {
615 struct dst_entry *tail = dst;
616 while (tail->next)
617 tail = tail->next;
618 tail->next = gc_list;
619 gc_list = dst;
620
621 policy->bundles = NULL;
622 }
623 write_unlock(&policy->lock);
1da177e4 624 }
9b78a82c
DM		 625 read_unlock_bh(&xfrm_policy_lock);
626
627 while (gc_list) {
628 struct dst_entry *dst = gc_list;
629
630 gc_list = dst->next;
631 dst_free(dst);
632 }
633
1da177e4
LT		 634 return 0;
635}
636EXPORT_SYMBOL(xfrm_policy_insert);
637
8d1211a6 638struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u8 type, int dir,
4e81bb83 639 struct xfrm_selector *sel,
ef41aaa0
EP		 640 struct xfrm_sec_ctx *ctx, int delete,
641 int *err)
1da177e4 642{
2518c7c2
DM		 643 struct xfrm_policy *pol, *ret;
644 struct hlist_head *chain;
645 struct hlist_node *entry;
1da177e4 646
ef41aaa0 647 *err = 0;
1da177e4 648 write_lock_bh(&xfrm_policy_lock);
8d1211a6 649 chain = policy_hash_bysel(net, sel, sel->family, dir);
2518c7c2
DM		 650 ret = NULL;
651 hlist_for_each_entry(pol, entry, chain, bydst) {
652 if (pol->type == type &&
653 !selector_cmp(sel, &pol->selector) &&
654 xfrm_sec_ctx_match(ctx, pol->security)) {
1da177e4 655 xfrm_pol_hold(pol);
2518c7c2 656 if (delete) {
03e1ad7b
PM		 657 *err = security_xfrm_policy_delete(
658 pol->security);
ef41aaa0
EP		 659 if (*err) {
660 write_unlock_bh(&xfrm_policy_lock);
661 return pol;
662 }
29fa0b30 663 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 664 }
665 ret = pol;
1da177e4
LT		 666 break;
667 }
668 }
669 write_unlock_bh(&xfrm_policy_lock);
670
2518c7c2 671 if (ret && delete) {
1da177e4 672 atomic_inc(&flow_cache_genid);
2518c7c2 673 xfrm_policy_kill(ret);
1da177e4 674 }
2518c7c2 675 return ret;
1da177e4 676}
df71837d 677EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
1da177e4 678
8d1211a6
AD		 679struct xfrm_policy *xfrm_policy_byid(struct net *net, u8 type, int dir, u32 id,
680 int delete, int *err)
1da177e4 681{
2518c7c2
DM		 682 struct xfrm_policy *pol, *ret;
683 struct hlist_head *chain;
684 struct hlist_node *entry;
1da177e4 685
b5505c6e
HX		 686 *err = -ENOENT;
687 if (xfrm_policy_id2dir(id) != dir)
688 return NULL;
689
ef41aaa0 690 *err = 0;
1da177e4 691 write_lock_bh(&xfrm_policy_lock);
8d1211a6 692 chain = net->xfrm.policy_byidx + idx_hash(net, id);
2518c7c2
DM		 693 ret = NULL;
694 hlist_for_each_entry(pol, entry, chain, byidx) {
695 if (pol->type == type && pol->index == id) {
1da177e4 696 xfrm_pol_hold(pol);
2518c7c2 697 if (delete) {
03e1ad7b
PM		 698 *err = security_xfrm_policy_delete(
699 pol->security);
ef41aaa0
EP		 700 if (*err) {
701 write_unlock_bh(&xfrm_policy_lock);
702 return pol;
703 }
29fa0b30 704 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 705 }
706 ret = pol;
1da177e4
LT		 707 break;
708 }
709 }
710 write_unlock_bh(&xfrm_policy_lock);
711
2518c7c2 712 if (ret && delete) {
1da177e4 713 atomic_inc(&flow_cache_genid);
2518c7c2 714 xfrm_policy_kill(ret);
1da177e4 715 }
2518c7c2 716 return ret;
1da177e4
LT		 717}
718EXPORT_SYMBOL(xfrm_policy_byid);
719
4aa2e62c
JL		 720#ifdef CONFIG_SECURITY_NETWORK_XFRM
721static inline int
33ffbbd5 722xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
1da177e4 723{
4aa2e62c
JL		 724 int dir, err = 0;
725
726 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
727 struct xfrm_policy *pol;
728 struct hlist_node *entry;
729 int i;
730
731 hlist_for_each_entry(pol, entry,
33ffbbd5 732 &net->xfrm.policy_inexact[dir], bydst) {
4aa2e62c
JL		 733 if (pol->type != type)
734 continue;
03e1ad7b 735 err = security_xfrm_policy_delete(pol->security);
4aa2e62c 736 if (err) {
ab5f5e8b
JL		 737 xfrm_audit_policy_delete(pol, 0,
738 audit_info->loginuid,
2532386f 739 audit_info->sessionid,
ab5f5e8b 740 audit_info->secid);
4aa2e62c
JL		 741 return err;
742 }
7dc12d6d 743 }
33ffbbd5 744 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
4aa2e62c 745 hlist_for_each_entry(pol, entry,
33ffbbd5 746 net->xfrm.policy_bydst[dir].table + i,
4aa2e62c
JL		 747 bydst) {
748 if (pol->type != type)
749 continue;
03e1ad7b
PM		 750 err = security_xfrm_policy_delete(
751 pol->security);
4aa2e62c 752 if (err) {
ab5f5e8b
JL		 753 xfrm_audit_policy_delete(pol, 0,
754 audit_info->loginuid,
2532386f 755 audit_info->sessionid,
ab5f5e8b 756 audit_info->secid);
4aa2e62c
JL		 757 return err;
758 }
759 }
760 }
761 }
762 return err;
763}
764#else
765static inline int
33ffbbd5 766xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c
JL		 767{
768 return 0;
769}
770#endif
771
33ffbbd5 772int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
4aa2e62c
JL		 773{
774 int dir, err = 0;
1da177e4
LT		 775
776 write_lock_bh(&xfrm_policy_lock);
4aa2e62c 777
33ffbbd5 778 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
4aa2e62c
JL		 779 if (err)
780 goto out;
781
1da177e4 782 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
2518c7c2
DM		 783 struct xfrm_policy *pol;
784 struct hlist_node *entry;
29fa0b30 785 int i;
2518c7c2
DM		 786
787 again1:
788 hlist_for_each_entry(pol, entry,
33ffbbd5 789 &net->xfrm.policy_inexact[dir], bydst) {
2518c7c2
DM		 790 if (pol->type != type)
791 continue;
29fa0b30 792 __xfrm_policy_unlink(pol, dir);
1da177e4
LT		 793 write_unlock_bh(&xfrm_policy_lock);
794
ab5f5e8b 795 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
2532386f 796 audit_info->sessionid,
ab5f5e8b 797 audit_info->secid);
161a09e7 798
2518c7c2 799 xfrm_policy_kill(pol);
1da177e4
LT		 800
801 write_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 802 goto again1;
803 }
804
33ffbbd5 805 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2
DM		 806 again2:
807 hlist_for_each_entry(pol, entry,
33ffbbd5 808 net->xfrm.policy_bydst[dir].table + i,
2518c7c2
DM		 809 bydst) {
810 if (pol->type != type)
811 continue;
29fa0b30 812 __xfrm_policy_unlink(pol, dir);
2518c7c2
DM		 813 write_unlock_bh(&xfrm_policy_lock);
814
ab5f5e8b
JL		 815 xfrm_audit_policy_delete(pol, 1,
816 audit_info->loginuid,
2532386f 817 audit_info->sessionid,
ab5f5e8b 818 audit_info->secid);
2518c7c2
DM		 819 xfrm_policy_kill(pol);
820
821 write_lock_bh(&xfrm_policy_lock);
822 goto again2;
823 }
1da177e4 824 }
2518c7c2 825
1da177e4
LT		 826 }
827 atomic_inc(&flow_cache_genid);
4aa2e62c 828out:
1da177e4 829 write_unlock_bh(&xfrm_policy_lock);
4aa2e62c 830 return err;
1da177e4
LT		 831}
832EXPORT_SYMBOL(xfrm_policy_flush);
833
cdcbca7c 834int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
4c563f76 835 int (*func)(struct xfrm_policy *, int, int, void*),
1da177e4
LT		 836 void *data)
837{
12a169e7
HX		 838 struct xfrm_policy *pol;
839 struct xfrm_policy_walk_entry *x;
4c563f76
TT		 840 int error = 0;
841
842 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
843 walk->type != XFRM_POLICY_TYPE_ANY)
844 return -EINVAL;
1da177e4 845
12a169e7 846 if (list_empty(&walk->walk.all) && walk->seq != 0)
4c563f76
TT		 847 return 0;
848
12a169e7
HX		 849 write_lock_bh(&xfrm_policy_lock);
850 if (list_empty(&walk->walk.all))
cdcbca7c 851 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
12a169e7
HX		 852 else
853 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
cdcbca7c 854 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
12a169e7 855 if (x->dead)
4c563f76 856 continue;
12a169e7
HX		 857 pol = container_of(x, struct xfrm_policy, walk);
858 if (walk->type != XFRM_POLICY_TYPE_ANY &&
859 walk->type != pol->type)
860 continue;
861 error = func(pol, xfrm_policy_id2dir(pol->index),
862 walk->seq, data);
863 if (error) {
864 list_move_tail(&walk->walk.all, &x->all);
865 goto out;
2518c7c2 866 }
12a169e7 867 walk->seq++;
1da177e4 868 }
12a169e7 869 if (walk->seq == 0) {
baf5d743
JHS		 870 error = -ENOENT;
871 goto out;
872 }
12a169e7 873 list_del_init(&walk->walk.all);
1da177e4 874out:
12a169e7 875 write_unlock_bh(&xfrm_policy_lock);
1da177e4
LT		 876 return error;
877}
878EXPORT_SYMBOL(xfrm_policy_walk);
879
12a169e7
HX		 880void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
881{
882 INIT_LIST_HEAD(&walk->walk.all);
883 walk->walk.dead = 1;
884 walk->type = type;
885 walk->seq = 0;
886}
887EXPORT_SYMBOL(xfrm_policy_walk_init);
888
889void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
890{
891 if (list_empty(&walk->walk.all))
892 return;
893
894 write_lock_bh(&xfrm_policy_lock);
895 list_del(&walk->walk.all);
896 write_unlock_bh(&xfrm_policy_lock);
897}
898EXPORT_SYMBOL(xfrm_policy_walk_done);
899
134b0fc5
JM		 900/*
901 * Find policy to apply to this flow.
902 *
903 * Returns 0 if policy found, else an -errno.
904 */
2518c7c2
DM		 905static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
906 u8 type, u16 family, int dir)
1da177e4 907{
2518c7c2 908 struct xfrm_selector *sel = &pol->selector;
134b0fc5 909 int match, ret = -ESRCH;
1da177e4 910
2518c7c2
DM		 911 if (pol->family != family ||
912 pol->type != type)
134b0fc5 913 return ret;
1da177e4 914
2518c7c2 915 match = xfrm_selector_match(sel, fl, family);
134b0fc5 916 if (match)
03e1ad7b
PM		 917 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
918 dir);
2518c7c2 919
134b0fc5 920 return ret;
2518c7c2 921}
1da177e4 922
52479b62
AD		 923static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
924 struct flowi *fl,
2518c7c2
DM		 925 u16 family, u8 dir)
926{
134b0fc5 927 int err;
2518c7c2
DM		 928 struct xfrm_policy *pol, *ret;
929 xfrm_address_t *daddr, *saddr;
930 struct hlist_node *entry;
931 struct hlist_head *chain;
acba48e1 932 u32 priority = ~0U;
df71837d 933
2518c7c2
DM		 934 daddr = xfrm_flowi_daddr(fl, family);
935 saddr = xfrm_flowi_saddr(fl, family);
936 if (unlikely(!daddr || !saddr))
937 return NULL;
938
939 read_lock_bh(&xfrm_policy_lock);
52479b62 940 chain = policy_hash_direct(net, daddr, saddr, family, dir);
2518c7c2
DM		 941 ret = NULL;
942 hlist_for_each_entry(pol, entry, chain, bydst) {
134b0fc5
JM		 943 err = xfrm_policy_match(pol, fl, type, family, dir);
944 if (err) {
945 if (err == -ESRCH)
946 continue;
947 else {
948 ret = ERR_PTR(err);
949 goto fail;
950 }
951 } else {
2518c7c2 952 ret = pol;
acba48e1 953 priority = ret->priority;
2518c7c2
DM		 954 break;
955 }
956 }
52479b62 957 chain = &net->xfrm.policy_inexact[dir];
acba48e1 958 hlist_for_each_entry(pol, entry, chain, bydst) {
134b0fc5
JM		 959 err = xfrm_policy_match(pol, fl, type, family, dir);
960 if (err) {
961 if (err == -ESRCH)
962 continue;
963 else {
964 ret = ERR_PTR(err);
965 goto fail;
966 }
967 } else if (pol->priority < priority) {
acba48e1
DM		 968 ret = pol;
969 break;
1da177e4
LT		 970 }
971 }
acba48e1
DM		 972 if (ret)
973 xfrm_pol_hold(ret);
134b0fc5 974fail:
1da177e4 975 read_unlock_bh(&xfrm_policy_lock);
4e81bb83 976
2518c7c2 977 return ret;
4e81bb83
MN		 978}
979
52479b62
AD		 980static int xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family,
981 u8 dir, void **objp, atomic_t **obj_refp)
4e81bb83
MN		 982{
983 struct xfrm_policy *pol;
134b0fc5 984 int err = 0;
4e81bb83
MN		 985
986#ifdef CONFIG_XFRM_SUB_POLICY
52479b62 987 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
134b0fc5
JM		 988 if (IS_ERR(pol)) {
989 err = PTR_ERR(pol);
990 pol = NULL;
991 }
992 if (pol || err)
4e81bb83
MN		 993 goto end;
994#endif
52479b62 995 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
134b0fc5
JM		 996 if (IS_ERR(pol)) {
997 err = PTR_ERR(pol);
998 pol = NULL;
999 }
4e81bb83 1000#ifdef CONFIG_XFRM_SUB_POLICY
2518c7c2 1001end:
4e81bb83 1002#endif
1da177e4
LT		 1003 if ((*objp = (void *) pol) != NULL)
1004 *obj_refp = &pol->refcnt;
134b0fc5 1005 return err;
1da177e4
LT		 1006}
1007
df71837d
TJ		 1008static inline int policy_to_flow_dir(int dir)
1009{
1010 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
a716c119
YH		 1011 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1012 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1013 return dir;
1014 switch (dir) {
1015 default:
1016 case XFRM_POLICY_IN:
1017 return FLOW_DIR_IN;
1018 case XFRM_POLICY_OUT:
1019 return FLOW_DIR_OUT;
1020 case XFRM_POLICY_FWD:
1021 return FLOW_DIR_FWD;
3ff50b79 1022 }
df71837d
TJ		 1023}
1024
e0d1caa7 1025static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
1da177e4
LT		 1026{
1027 struct xfrm_policy *pol;
1028
1029 read_lock_bh(&xfrm_policy_lock);
1030 if ((pol = sk->sk_policy[dir]) != NULL) {
a716c119 1031 int match = xfrm_selector_match(&pol->selector, fl,
1da177e4 1032 sk->sk_family);
a716c119 1033 int err = 0;
df71837d 1034
3bccfbc7 1035 if (match) {
03e1ad7b
PM		 1036 err = security_xfrm_policy_lookup(pol->security,
1037 fl->secid,
1038 policy_to_flow_dir(dir));
3bccfbc7
VY		 1039 if (!err)
1040 xfrm_pol_hold(pol);
1041 else if (err == -ESRCH)
1042 pol = NULL;
1043 else
1044 pol = ERR_PTR(err);
1045 } else
1da177e4
LT		 1046 pol = NULL;
1047 }
1048 read_unlock_bh(&xfrm_policy_lock);
1049 return pol;
1050}
1051
1052static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1053{
98806f75 1054 struct net *net = xp_net(pol);
1121994c 1055 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
2518c7c2 1056 pol->family, dir);
4e81bb83 1057
98806f75 1058 list_add(&pol->walk.all, &net->xfrm.policy_all);
2518c7c2 1059 hlist_add_head(&pol->bydst, chain);
e92303f8 1060 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
98806f75 1061 net->xfrm.policy_count[dir]++;
1da177e4 1062 xfrm_pol_hold(pol);
2518c7c2 1063
98806f75
AD		 1064 if (xfrm_bydst_should_resize(net, dir, NULL))
1065 schedule_work(&net->xfrm.policy_hash_work);
1da177e4
LT		 1066}
1067
1068static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1069 int dir)
1070{
98806f75
AD		 1071 struct net *net = xp_net(pol);
1072
2518c7c2
DM		 1073 if (hlist_unhashed(&pol->bydst))
1074 return NULL;
1da177e4 1075
2518c7c2
DM		 1076 hlist_del(&pol->bydst);
1077 hlist_del(&pol->byidx);
12a169e7 1078 list_del(&pol->walk.all);
98806f75 1079 net->xfrm.policy_count[dir]--;
2518c7c2
DM		 1080
1081 return pol;
1da177e4
LT		 1082}
1083
4666faab 1084int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1da177e4
LT		 1085{
1086 write_lock_bh(&xfrm_policy_lock);
1087 pol = __xfrm_policy_unlink(pol, dir);
1088 write_unlock_bh(&xfrm_policy_lock);
1089 if (pol) {
1090 if (dir < XFRM_POLICY_MAX)
1091 atomic_inc(&flow_cache_genid);
1092 xfrm_policy_kill(pol);
4666faab 1093 return 0;
1da177e4 1094 }
4666faab 1095 return -ENOENT;
1da177e4 1096}
a70fcb0b 1097EXPORT_SYMBOL(xfrm_policy_delete);
1da177e4
LT		 1098
1099int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1100{
1121994c 1101 struct net *net = xp_net(pol);
1da177e4
LT		 1102 struct xfrm_policy *old_pol;
1103
4e81bb83
MN		 1104#ifdef CONFIG_XFRM_SUB_POLICY
1105 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1106 return -EINVAL;
1107#endif
1108
1da177e4
LT		 1109 write_lock_bh(&xfrm_policy_lock);
1110 old_pol = sk->sk_policy[dir];
1111 sk->sk_policy[dir] = pol;
1112 if (pol) {
9d729f72 1113 pol->curlft.add_time = get_seconds();
1121994c 1114 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1da177e4
LT		 1115 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1116 }
1117 if (old_pol)
1118 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1119 write_unlock_bh(&xfrm_policy_lock);
1120
1121 if (old_pol) {
1122 xfrm_policy_kill(old_pol);
1123 }
1124 return 0;
1125}
1126
1127static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1128{
0331b1f3 1129 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1da177e4
LT		 1130
1131 if (newp) {
1132 newp->selector = old->selector;
03e1ad7b
PM		 1133 if (security_xfrm_policy_clone(old->security,
1134 &newp->security)) {
df71837d
TJ		 1135 kfree(newp);
1136 return NULL; /* ENOMEM */
1137 }
1da177e4
LT		 1138 newp->lft = old->lft;
1139 newp->curlft = old->curlft;
1140 newp->action = old->action;
1141 newp->flags = old->flags;
1142 newp->xfrm_nr = old->xfrm_nr;
1143 newp->index = old->index;
4e81bb83 1144 newp->type = old->type;
1da177e4
LT		 1145 memcpy(newp->xfrm_vec, old->xfrm_vec,
1146 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1147 write_lock_bh(&xfrm_policy_lock);
1148 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1149 write_unlock_bh(&xfrm_policy_lock);
1150 xfrm_pol_put(newp);
1151 }
1152 return newp;
1153}
1154
1155int __xfrm_sk_clone_policy(struct sock *sk)
1156{
1157 struct xfrm_policy *p0 = sk->sk_policy[0],
1158 *p1 = sk->sk_policy[1];
1159
1160 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1161 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1162 return -ENOMEM;
1163 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1164 return -ENOMEM;
1165 return 0;
1166}
1167
a1e59abf 1168static int
fbda33b2 1169xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
a1e59abf
PM		 1170 unsigned short family)
1171{
1172 int err;
1173 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1174
1175 if (unlikely(afinfo == NULL))
1176 return -EINVAL;
fbda33b2 1177 err = afinfo->get_saddr(net, local, remote);
a1e59abf
PM		 1178 xfrm_policy_put_afinfo(afinfo);
1179 return err;
1180}
1181
1da177e4
LT		 1182/* Resolve list of templates for the flow, given policy. */
1183
1184static int
4e81bb83
MN		 1185xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
1186 struct xfrm_state **xfrm,
1187 unsigned short family)
1da177e4 1188{
fbda33b2 1189 struct net *net = xp_net(policy);
1da177e4
LT		 1190 int nx;
1191 int i, error;
1192 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1193 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
a1e59abf 1194 xfrm_address_t tmp;
1da177e4
LT		 1195
1196 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1197 struct xfrm_state *x;
1198 xfrm_address_t *remote = daddr;
1199 xfrm_address_t *local = saddr;
1200 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1201
48b8d783
JK		 1202 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1203 tmpl->mode == XFRM_MODE_BEET) {
1da177e4
LT		 1204 remote = &tmpl->id.daddr;
1205 local = &tmpl->saddr;
76b3f055 1206 family = tmpl->encap_family;
a1e59abf 1207 if (xfrm_addr_any(local, family)) {
fbda33b2 1208 error = xfrm_get_saddr(net, &tmp, remote, family);
a1e59abf
PM		 1209 if (error)
1210 goto fail;
1211 local = &tmp;
1212 }
1da177e4
LT		 1213 }
1214
1215 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1216
1217 if (x && x->km.state == XFRM_STATE_VALID) {
1218 xfrm[nx++] = x;
1219 daddr = remote;
1220 saddr = local;
1221 continue;
1222 }
1223 if (x) {
1224 error = (x->km.state == XFRM_STATE_ERROR ?
1225 -EINVAL : -EAGAIN);
1226 xfrm_state_put(x);
1227 }
a4322266 1228 else if (error == -ESRCH)
1229 error = -EAGAIN;
1da177e4
LT		 1230
1231 if (!tmpl->optional)
1232 goto fail;
1233 }
1234 return nx;
1235
1236fail:
1237 for (nx--; nx>=0; nx--)
1238 xfrm_state_put(xfrm[nx]);
1239 return error;
1240}
1241
4e81bb83
MN		 1242static int
1243xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1244 struct xfrm_state **xfrm,
1245 unsigned short family)
1246{
41a49cc3
MN		 1247 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1248 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
4e81bb83
MN		 1249 int cnx = 0;
1250 int error;
1251 int ret;
1252 int i;
1253
1254 for (i = 0; i < npols; i++) {
1255 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1256 error = -ENOBUFS;
1257 goto fail;
1258 }
41a49cc3
MN		 1259
1260 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
4e81bb83
MN		 1261 if (ret < 0) {
1262 error = ret;
1263 goto fail;
1264 } else
1265 cnx += ret;
1266 }
1267
41a49cc3
MN		 1268 /* found states are sorted for outbound processing */
1269 if (npols > 1)
1270 xfrm_state_sort(xfrm, tpp, cnx, family);
1271
4e81bb83
MN		 1272 return cnx;
1273
1274 fail:
1275 for (cnx--; cnx>=0; cnx--)
41a49cc3 1276 xfrm_state_put(tpp[cnx]);
4e81bb83
MN		 1277 return error;
1278
1279}
1280
1da177e4
LT		 1281/* Check that the bundle accepts the flow and its components are
1282 * still valid.
1283 */
1284
1285static struct dst_entry *
1286xfrm_find_bundle(struct flowi *fl, struct xfrm_policy *policy, unsigned short family)
1287{
1288 struct dst_entry *x;
1289 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1290 if (unlikely(afinfo == NULL))
1291 return ERR_PTR(-EINVAL);
1292 x = afinfo->find_bundle(fl, policy);
1293 xfrm_policy_put_afinfo(afinfo);
1294 return x;
1295}
1296
25ee3286
HX		 1297static inline int xfrm_get_tos(struct flowi *fl, int family)
1298{
1299 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1300 int tos;
1da177e4 1301
25ee3286
HX		 1302 if (!afinfo)
1303 return -EINVAL;
1304
1305 tos = afinfo->get_tos(fl);
1306
1307 xfrm_policy_put_afinfo(afinfo);
1308
1309 return tos;
1310}
1311
1312static inline struct xfrm_dst *xfrm_alloc_dst(int family)
1da177e4 1313{
1da177e4 1314 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
25ee3286
HX		 1315 struct xfrm_dst *xdst;
1316
1317 if (!afinfo)
1318 return ERR_PTR(-EINVAL);
1319
1320 xdst = dst_alloc(afinfo->dst_ops) ?: ERR_PTR(-ENOBUFS);
1321
1322 xfrm_policy_put_afinfo(afinfo);
1323
1324 return xdst;
1325}
1326
a1b05140
MN		 1327static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1328 int nfheader_len)
1329{
1330 struct xfrm_policy_afinfo *afinfo =
1331 xfrm_policy_get_afinfo(dst->ops->family);
1332 int err;
1333
1334 if (!afinfo)
1335 return -EINVAL;
1336
1337 err = afinfo->init_path(path, dst, nfheader_len);
1338
1339 xfrm_policy_put_afinfo(afinfo);
1340
1341 return err;
1342}
1343
25ee3286
HX		 1344static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
1345{
1346 struct xfrm_policy_afinfo *afinfo =
1347 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1348 int err;
1349
1350 if (!afinfo)
1da177e4 1351 return -EINVAL;
25ee3286
HX		 1352
1353 err = afinfo->fill_dst(xdst, dev);
1354
1da177e4 1355 xfrm_policy_put_afinfo(afinfo);
25ee3286 1356
1da177e4
LT		 1357 return err;
1358}
1359
25ee3286
HX		 1360/* Allocate chain of dst_entry's, attach known xfrm's, calculate
1361 * all the metrics... Shortly, bundle a bundle.
1362 */
1363
1364static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1365 struct xfrm_state **xfrm, int nx,
1366 struct flowi *fl,
1367 struct dst_entry *dst)
1368{
1369 unsigned long now = jiffies;
1370 struct net_device *dev;
1371 struct dst_entry *dst_prev = NULL;
1372 struct dst_entry *dst0 = NULL;
1373 int i = 0;
1374 int err;
1375 int header_len = 0;
a1b05140 1376 int nfheader_len = 0;
25ee3286
HX		 1377 int trailer_len = 0;
1378 int tos;
1379 int family = policy->selector.family;
9bb182a7
YH		 1380 xfrm_address_t saddr, daddr;
1381
1382 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
25ee3286
HX		 1383
1384 tos = xfrm_get_tos(fl, family);
1385 err = tos;
1386 if (tos < 0)
1387 goto put_states;
1388
1389 dst_hold(dst);
1390
1391 for (; i < nx; i++) {
1392 struct xfrm_dst *xdst = xfrm_alloc_dst(family);
1393 struct dst_entry *dst1 = &xdst->u.dst;
1394
1395 err = PTR_ERR(xdst);
1396 if (IS_ERR(xdst)) {
1397 dst_release(dst);
1398 goto put_states;
1399 }
1400
1401 if (!dst_prev)
1402 dst0 = dst1;
1403 else {
1404 dst_prev->child = dst_clone(dst1);
1405 dst1->flags |= DST_NOHASH;
1406 }
1407
1408 xdst->route = dst;
1409 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1410
1411 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1412 family = xfrm[i]->props.family;
9bb182a7
YH		 1413 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1414 family);
25ee3286
HX		 1415 err = PTR_ERR(dst);
1416 if (IS_ERR(dst))
1417 goto put_states;
1418 } else
1419 dst_hold(dst);
1420
1421 dst1->xfrm = xfrm[i];
1422 xdst->genid = xfrm[i]->genid;
1423
1424 dst1->obsolete = -1;
1425 dst1->flags |= DST_HOST;
1426 dst1->lastuse = now;
1427
1428 dst1->input = dst_discard;
1429 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1430
1431 dst1->next = dst_prev;
1432 dst_prev = dst1;
1433
1434 header_len += xfrm[i]->props.header_len;
a1b05140
MN		 1435 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1436 nfheader_len += xfrm[i]->props.header_len;
25ee3286
HX		 1437 trailer_len += xfrm[i]->props.trailer_len;
1438 }
1439
1440 dst_prev->child = dst;
1441 dst0->path = dst;
1442
1443 err = -ENODEV;
1444 dev = dst->dev;
1445 if (!dev)
1446 goto free_dst;
1447
96c53401 1448 /* Copy neighbour for reachability confirmation */
25ee3286
HX		 1449 dst0->neighbour = neigh_clone(dst->neighbour);
1450
a1b05140 1451 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
25ee3286
HX		 1452 xfrm_init_pmtu(dst_prev);
1453
1454 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1455 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1456
1457 err = xfrm_fill_dst(xdst, dev);
1458 if (err)
1459 goto free_dst;
1460
1461 dst_prev->header_len = header_len;
1462 dst_prev->trailer_len = trailer_len;
1463 header_len -= xdst->u.dst.xfrm->props.header_len;
1464 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1465 }
1466
1467out:
1468 return dst0;
1469
1470put_states:
1471 for (; i < nx; i++)
1472 xfrm_state_put(xfrm[i]);
1473free_dst:
1474 if (dst0)
1475 dst_free(dst0);
1476 dst0 = ERR_PTR(err);
1477 goto out;
1478}
1479
157bfc25
MN		 1480static int inline
1481xfrm_dst_alloc_copy(void **target, void *src, int size)
1482{
1483 if (!*target) {
1484 *target = kmalloc(size, GFP_ATOMIC);
1485 if (!*target)
1486 return -ENOMEM;
1487 }
1488 memcpy(*target, src, size);
1489 return 0;
1490}
1491
1492static int inline
1493xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1494{
1495#ifdef CONFIG_XFRM_SUB_POLICY
1496 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1497 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1498 sel, sizeof(*sel));
1499#else
1500 return 0;
1501#endif
1502}
1503
1504static int inline
1505xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1506{
1507#ifdef CONFIG_XFRM_SUB_POLICY
1508 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1509 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1510#else
1511 return 0;
1512#endif
1513}
1da177e4
LT		 1514
1515static int stale_bundle(struct dst_entry *dst);
1516
1517/* Main function: finds/creates a bundle for given flow.
1518 *
1519 * At the moment we eat a raw IP route. Mostly to speed up lookups
1520 * on interfaces with disabled IPsec.
1521 */
52479b62 1522int __xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
14e50e57 1523 struct sock *sk, int flags)
1da177e4
LT		 1524{
1525 struct xfrm_policy *policy;
4e81bb83
MN		 1526 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1527 int npols;
1528 int pol_dead;
1529 int xfrm_nr;
1530 int pi;
1da177e4
LT		 1531 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1532 struct dst_entry *dst, *dst_orig = *dst_p;
1533 int nx = 0;
1534 int err;
1535 u32 genid;
42cf93cd 1536 u16 family;
df71837d 1537 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
e0d1caa7 1538
1da177e4
LT		 1539restart:
1540 genid = atomic_read(&flow_cache_genid);
1541 policy = NULL;
4e81bb83
MN		 1542 for (pi = 0; pi < ARRAY_SIZE(pols); pi++)
1543 pols[pi] = NULL;
1544 npols = 0;
1545 pol_dead = 0;
1546 xfrm_nr = 0;
1547
f7944fb1 1548 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
e0d1caa7 1549 policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
75b8c133 1550 err = PTR_ERR(policy);
0aa64774 1551 if (IS_ERR(policy)) {
59c9940e 1552 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
75b8c133 1553 goto dropdst;
0aa64774 1554 }
3bccfbc7 1555 }
1da177e4
LT		 1556
1557 if (!policy) {
1558 /* To accelerate a bit... */
2518c7c2 1559 if ((dst_orig->flags & DST_NOXFRM) ||
52479b62 1560 !net->xfrm.policy_count[XFRM_POLICY_OUT])
8b7817f3 1561 goto nopol;
1da177e4 1562
52479b62 1563 policy = flow_cache_lookup(net, fl, dst_orig->ops->family,
42cf93cd 1564 dir, xfrm_policy_lookup);
75b8c133 1565 err = PTR_ERR(policy);
d66e37a9 1566 if (IS_ERR(policy)) {
59c9940e 1567 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
75b8c133 1568 goto dropdst;
d66e37a9 1569 }
1da177e4
LT		 1570 }
1571
1572 if (!policy)
8b7817f3 1573 goto nopol;
1da177e4 1574
42cf93cd 1575 family = dst_orig->ops->family;
4e81bb83
MN		 1576 pols[0] = policy;
1577 npols ++;
1578 xfrm_nr += pols[0]->xfrm_nr;
1da177e4 1579
aef21785 1580 err = -ENOENT;
8b7817f3
HX		 1581 if ((flags & XFRM_LOOKUP_ICMP) && !(policy->flags & XFRM_POLICY_ICMP))
1582 goto error;
1583
1584 policy->curlft.use_time = get_seconds();
1585
1da177e4 1586 switch (policy->action) {
5e5234ff 1587 default:
1da177e4
LT		 1588 case XFRM_POLICY_BLOCK:
1589 /* Prohibit the flow */
59c9940e 1590 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
e104411b
PM		 1591 err = -EPERM;
1592 goto error;
1da177e4
LT		 1593
1594 case XFRM_POLICY_ALLOW:
4e81bb83 1595#ifndef CONFIG_XFRM_SUB_POLICY
1da177e4
LT		 1596 if (policy->xfrm_nr == 0) {
1597 /* Flow passes not transformed. */
1598 xfrm_pol_put(policy);
1599 return 0;
1600 }
4e81bb83 1601#endif
1da177e4
LT		 1602
1603 /* Try to find matching bundle.
1604 *
1605 * LATER: help from flow cache. It is optional, this
1606 * is required only for output policy.
1607 */
1608 dst = xfrm_find_bundle(fl, policy, family);
1609 if (IS_ERR(dst)) {
59c9940e 1610 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
e104411b
PM		 1611 err = PTR_ERR(dst);
1612 goto error;
1da177e4
LT		 1613 }
1614
1615 if (dst)
1616 break;
1617
4e81bb83
MN		 1618#ifdef CONFIG_XFRM_SUB_POLICY
1619 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
52479b62
AD		 1620 pols[1] = xfrm_policy_lookup_bytype(net,
1621 XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 1622 fl, family,
1623 XFRM_POLICY_OUT);
1624 if (pols[1]) {
134b0fc5 1625 if (IS_ERR(pols[1])) {
59c9940e 1626 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
134b0fc5
JM		 1627 err = PTR_ERR(pols[1]);
1628 goto error;
1629 }
4e81bb83 1630 if (pols[1]->action == XFRM_POLICY_BLOCK) {
59c9940e 1631 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
4e81bb83
MN		 1632 err = -EPERM;
1633 goto error;
1634 }
1635 npols ++;
1636 xfrm_nr += pols[1]->xfrm_nr;
1637 }
1638 }
1639
1640 /*
1641 * Because neither flowi nor bundle information knows about
1642 * transformation template size. On more than one policy usage
1643 * we can realize whether all of them is bypass or not after
1644 * they are searched. See above not-transformed bypass
1645 * is surrounded by non-sub policy configuration, too.
1646 */
1647 if (xfrm_nr == 0) {
1648 /* Flow passes not transformed. */
1649 xfrm_pols_put(pols, npols);
1650 return 0;
1651 }
1652
1653#endif
1654 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1da177e4
LT		 1655
1656 if (unlikely(nx<0)) {
1657 err = nx;
b27aeadb 1658 if (err == -EAGAIN && net->xfrm.sysctl_larval_drop) {
14e50e57
DM		 1659 /* EREMOTE tells the caller to generate
1660 * a one-shot blackhole route.
1661 */
59c9940e 1662 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
14e50e57
DM		 1663 xfrm_pol_put(policy);
1664 return -EREMOTE;
1665 }
815f4e57 1666 if (err == -EAGAIN && (flags & XFRM_LOOKUP_WAIT)) {
1da177e4
LT		 1667 DECLARE_WAITQUEUE(wait, current);
1668
52479b62 1669 add_wait_queue(&net->xfrm.km_waitq, &wait);
1da177e4
LT		 1670 set_current_state(TASK_INTERRUPTIBLE);
1671 schedule();
1672 set_current_state(TASK_RUNNING);
52479b62 1673 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1da177e4 1674
4e81bb83 1675 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1da177e4
LT		 1676
1677 if (nx == -EAGAIN && signal_pending(current)) {
59c9940e 1678 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1da177e4
LT		 1679 err = -ERESTART;
1680 goto error;
1681 }
1682 if (nx == -EAGAIN ||
1683 genid != atomic_read(&flow_cache_genid)) {
4e81bb83 1684 xfrm_pols_put(pols, npols);
1da177e4
LT		 1685 goto restart;
1686 }
1687 err = nx;
1688 }
0aa64774 1689 if (err < 0) {
59c9940e 1690 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1da177e4 1691 goto error;
0aa64774 1692 }
1da177e4
LT		 1693 }
1694 if (nx == 0) {
1695 /* Flow passes not transformed. */
4e81bb83 1696 xfrm_pols_put(pols, npols);
1da177e4
LT		 1697 return 0;
1698 }
1699
25ee3286
HX		 1700 dst = xfrm_bundle_create(policy, xfrm, nx, fl, dst_orig);
1701 err = PTR_ERR(dst);
0aa64774 1702 if (IS_ERR(dst)) {
59c9940e 1703 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1da177e4 1704 goto error;
0aa64774 1705 }
1da177e4 1706
4e81bb83
MN		 1707 for (pi = 0; pi < npols; pi++) {
1708 read_lock_bh(&pols[pi]->lock);
12a169e7 1709 pol_dead |= pols[pi]->walk.dead;
4e81bb83
MN		 1710 read_unlock_bh(&pols[pi]->lock);
1711 }
1712
1da177e4 1713 write_lock_bh(&policy->lock);
4e81bb83 1714 if (unlikely(pol_dead || stale_bundle(dst))) {
1da177e4
LT		 1715 /* Wow! While we worked on resolving, this
1716 * policy has gone. Retry. It is not paranoia,
1717 * we just cannot enlist new bundle to dead object.
1718 * We can't enlist stable bundles either.
1719 */
1720 write_unlock_bh(&policy->lock);
9d7d7402 1721 dst_free(dst);
00de651d 1722
0aa64774 1723 if (pol_dead)
59c9940e 1724 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLDEAD);
0aa64774 1725 else
59c9940e 1726 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
00de651d
HX		 1727 err = -EHOSTUNREACH;
1728 goto error;
1da177e4 1729 }
157bfc25
MN		 1730
1731 if (npols > 1)
1732 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1733 else
1734 err = xfrm_dst_update_origin(dst, fl);
1735 if (unlikely(err)) {
1736 write_unlock_bh(&policy->lock);
9d7d7402 1737 dst_free(dst);
59c9940e 1738 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
157bfc25
MN		 1739 goto error;
1740 }
1741
1da177e4
LT		 1742 dst->next = policy->bundles;
1743 policy->bundles = dst;
1744 dst_hold(dst);
1745 write_unlock_bh(&policy->lock);
1746 }
1747 *dst_p = dst;
1748 dst_release(dst_orig);
a716c119 1749 xfrm_pols_put(pols, npols);
1da177e4
LT		 1750 return 0;
1751
1752error:
4e81bb83 1753 xfrm_pols_put(pols, npols);
75b8c133
HX		 1754dropdst:
1755 dst_release(dst_orig);
1da177e4
LT		 1756 *dst_p = NULL;
1757 return err;
8b7817f3
HX		 1758
1759nopol:
aef21785 1760 err = -ENOENT;
8b7817f3
HX		 1761 if (flags & XFRM_LOOKUP_ICMP)
1762 goto dropdst;
1763 return 0;
1da177e4 1764}
14e50e57
DM		 1765EXPORT_SYMBOL(__xfrm_lookup);
1766
52479b62 1767int xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
14e50e57
DM		 1768 struct sock *sk, int flags)
1769{
52479b62 1770 int err = __xfrm_lookup(net, dst_p, fl, sk, flags);
14e50e57
DM		 1771
1772 if (err == -EREMOTE) {
1773 dst_release(*dst_p);
1774 *dst_p = NULL;
1775 err = -EAGAIN;
1776 }
1777
1778 return err;
1779}
1da177e4
LT		 1780EXPORT_SYMBOL(xfrm_lookup);
1781
df0ba92a
MN		 1782static inline int
1783xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1784{
1785 struct xfrm_state *x;
df0ba92a
MN		 1786
1787 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1788 return 0;
1789 x = skb->sp->xvec[idx];
1790 if (!x->type->reject)
1791 return 0;
1ecafede 1792 return x->type->reject(x, skb, fl);
df0ba92a
MN		 1793}
1794
1da177e4
LT		 1795/* When skb is transformed back to its "native" form, we have to
1796 * check policy restrictions. At the moment we make this in maximally
1797 * stupid way. Shame on me. :-) Of course, connected sockets must
1798 * have policy cached at them.
1799 */
1800
1801static inline int
a716c119 1802xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1da177e4
LT		 1803 unsigned short family)
1804{
1805 if (xfrm_state_kern(x))
928ba416 1806 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1da177e4
LT		 1807 return x->id.proto == tmpl->id.proto &&
1808 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1809 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1810 x->props.mode == tmpl->mode &&
c5d18e98 1811 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
f3bd4840 1812 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
7e49e6de
MN		 1813 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1814 xfrm_state_addr_cmp(tmpl, x, family));
1da177e4
LT		 1815}
1816
df0ba92a
MN		 1817/*
1818 * 0 or more than 0 is returned when validation is succeeded (either bypass
1819 * because of optional transport mode, or next index of the mathced secpath
1820 * state with the template.
1821 * -1 is returned when no matching template is found.
1822 * Otherwise "-2 - errored_index" is returned.
1823 */
1da177e4
LT		 1824static inline int
1825xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1826 unsigned short family)
1827{
1828 int idx = start;
1829
1830 if (tmpl->optional) {
7e49e6de 1831 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1da177e4
LT		 1832 return start;
1833 } else
1834 start = -1;
1835 for (; idx < sp->len; idx++) {
dbe5b4aa 1836 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1da177e4 1837 return ++idx;
df0ba92a
MN		 1838 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1839 if (start == -1)
1840 start = -2-idx;
1da177e4 1841 break;
df0ba92a 1842 }
1da177e4
LT		 1843 }
1844 return start;
1845}
1846
d5422efe
HX		 1847int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1848 unsigned int family, int reverse)
1da177e4
LT		 1849{
1850 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
e0d1caa7 1851 int err;
1da177e4
LT		 1852
1853 if (unlikely(afinfo == NULL))
1854 return -EAFNOSUPPORT;
1855
d5422efe 1856 afinfo->decode_session(skb, fl, reverse);
beb8d13b 1857 err = security_xfrm_decode_session(skb, &fl->secid);
1da177e4 1858 xfrm_policy_put_afinfo(afinfo);
e0d1caa7 1859 return err;
1da177e4 1860}
d5422efe 1861EXPORT_SYMBOL(__xfrm_decode_session);
1da177e4 1862
df0ba92a 1863static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1da177e4
LT		 1864{
1865 for (; k < sp->len; k++) {
df0ba92a 1866 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
d1d9facf 1867 *idxp = k;
1da177e4 1868 return 1;
df0ba92a 1869 }
1da177e4
LT		 1870 }
1871
1872 return 0;
1873}
1874
a716c119 1875int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1da177e4
LT		 1876 unsigned short family)
1877{
f6e1e25d 1878 struct net *net = dev_net(skb->dev);
1da177e4 1879 struct xfrm_policy *pol;
4e81bb83
MN		 1880 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1881 int npols = 0;
1882 int xfrm_nr;
1883 int pi;
d5422efe 1884 int reverse;
1da177e4 1885 struct flowi fl;
d5422efe 1886 u8 fl_dir;
df0ba92a 1887 int xerr_idx = -1;
1da177e4 1888
d5422efe
HX		 1889 reverse = dir & ~XFRM_POLICY_MASK;
1890 dir &= XFRM_POLICY_MASK;
1891 fl_dir = policy_to_flow_dir(dir);
1892
0aa64774 1893 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
59c9940e 1894 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 1895 return 0;
0aa64774
MN		 1896 }
1897
eb9c7ebe 1898 nf_nat_decode_session(skb, &fl, family);
1da177e4
LT		 1899
1900 /* First, check used SA against their selectors. */
1901 if (skb->sp) {
1902 int i;
1903
1904 for (i=skb->sp->len-1; i>=0; i--) {
dbe5b4aa 1905 struct xfrm_state *x = skb->sp->xvec[i];
0aa64774 1906 if (!xfrm_selector_match(&x->sel, &fl, family)) {
59c9940e 1907 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1da177e4 1908 return 0;
0aa64774 1909 }
1da177e4
LT		 1910 }
1911 }
1912
1913 pol = NULL;
3bccfbc7 1914 if (sk && sk->sk_policy[dir]) {
e0d1caa7 1915 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
0aa64774 1916 if (IS_ERR(pol)) {
59c9940e 1917 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
3bccfbc7 1918 return 0;
0aa64774 1919 }
3bccfbc7 1920 }
1da177e4
LT		 1921
1922 if (!pol)
f6e1e25d 1923 pol = flow_cache_lookup(net, &fl, family, fl_dir,
1da177e4
LT		 1924 xfrm_policy_lookup);
1925
0aa64774 1926 if (IS_ERR(pol)) {
59c9940e 1927 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 1928 return 0;
0aa64774 1929 }
134b0fc5 1930
df0ba92a 1931 if (!pol) {
d1d9facf 1932 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
df0ba92a 1933 xfrm_secpath_reject(xerr_idx, skb, &fl);
59c9940e 1934 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
df0ba92a
MN		 1935 return 0;
1936 }
1937 return 1;
1938 }
1da177e4 1939
9d729f72 1940 pol->curlft.use_time = get_seconds();
1da177e4 1941
4e81bb83
MN		 1942 pols[0] = pol;
1943 npols ++;
1944#ifdef CONFIG_XFRM_SUB_POLICY
1945 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
f6e1e25d 1946 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
4e81bb83
MN		 1947 &fl, family,
1948 XFRM_POLICY_IN);
1949 if (pols[1]) {
0aa64774 1950 if (IS_ERR(pols[1])) {
59c9940e 1951 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
134b0fc5 1952 return 0;
0aa64774 1953 }
9d729f72 1954 pols[1]->curlft.use_time = get_seconds();
4e81bb83
MN		 1955 npols ++;
1956 }
1957 }
1958#endif
1959
1da177e4
LT		 1960 if (pol->action == XFRM_POLICY_ALLOW) {
1961 struct sec_path *sp;
1962 static struct sec_path dummy;
4e81bb83 1963 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
41a49cc3 1964 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
4e81bb83
MN		 1965 struct xfrm_tmpl **tpp = tp;
1966 int ti = 0;
1da177e4
LT		 1967 int i, k;
1968
1969 if ((sp = skb->sp) == NULL)
1970 sp = &dummy;
1971
4e81bb83
MN		 1972 for (pi = 0; pi < npols; pi++) {
1973 if (pols[pi] != pol &&
0aa64774 1974 pols[pi]->action != XFRM_POLICY_ALLOW) {
59c9940e 1975 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
4e81bb83 1976 goto reject;
0aa64774
MN		 1977 }
1978 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
59c9940e 1979 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
4e81bb83 1980 goto reject_error;
0aa64774 1981 }
4e81bb83
MN		 1982 for (i = 0; i < pols[pi]->xfrm_nr; i++)
1983 tpp[ti++] = &pols[pi]->xfrm_vec[i];
1984 }
1985 xfrm_nr = ti;
41a49cc3
MN		 1986 if (npols > 1) {
1987 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
1988 tpp = stp;
1989 }
4e81bb83 1990
1da177e4
LT		 1991 /* For each tunnel xfrm, find the first matching tmpl.
1992 * For each tmpl before that, find corresponding xfrm.
1993 * Order is _important_. Later we will implement
1994 * some barriers, but at the moment barriers
1995 * are implied between each two transformations.
1996 */
4e81bb83
MN		 1997 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
1998 k = xfrm_policy_ok(tpp[i], sp, k, family);
df0ba92a 1999 if (k < 0) {
d1d9facf
JM		 2000 if (k < -1)
2001 /* "-2 - errored_index" returned */
2002 xerr_idx = -(2+k);
59c9940e 2003 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2004 goto reject;
df0ba92a 2005 }
1da177e4
LT		 2006 }
2007
0aa64774 2008 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
59c9940e 2009 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
1da177e4 2010 goto reject;
0aa64774 2011 }
1da177e4 2012
4e81bb83 2013 xfrm_pols_put(pols, npols);
1da177e4
LT		 2014 return 1;
2015 }
59c9940e 2016 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
1da177e4
LT		 2017
2018reject:
df0ba92a 2019 xfrm_secpath_reject(xerr_idx, skb, &fl);
4e81bb83
MN		 2020reject_error:
2021 xfrm_pols_put(pols, npols);
1da177e4
LT		 2022 return 0;
2023}
2024EXPORT_SYMBOL(__xfrm_policy_check);
2025
2026int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2027{
99a66657 2028 struct net *net = dev_net(skb->dev);
1da177e4 2029 struct flowi fl;
adf30907
ED		 2030 struct dst_entry *dst;
2031 int res;
1da177e4 2032
0aa64774
MN		 2033 if (xfrm_decode_session(skb, &fl, family) < 0) {
2034 /* XXX: we should have something like FWDHDRERROR here. */
59c9940e 2035 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1da177e4 2036 return 0;
0aa64774 2037 }
1da177e4 2038
adf30907
ED		 2039 dst = skb_dst(skb);
2040
2041 res = xfrm_lookup(net, &dst, &fl, NULL, 0) == 0;
2042 skb_dst_set(skb, dst);
2043 return res;
1da177e4
LT		 2044}
2045EXPORT_SYMBOL(__xfrm_route_forward);
2046
d49c73c7
DM		 2047/* Optimize later using cookies and generation ids. */
2048
1da177e4
LT		 2049static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2050{
d49c73c7
DM		 2051 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2052 * to "-1" to force all XFRM destinations to get validated by
2053 * dst_ops->check on every use. We do this because when a
2054 * normal route referenced by an XFRM dst is obsoleted we do
2055 * not go looking around for all parent referencing XFRM dsts
2056 * so that we can invalidate them. It is just too much work.
2057 * Instead we make the checks here on every use. For example:
2058 *
2059 * XFRM dst A --> IPv4 dst X
2060 *
2061 * X is the "xdst->route" of A (X is also the "dst->path" of A
2062 * in this example). If X is marked obsolete, "A" will not
2063 * notice. That's what we are validating here via the
2064 * stale_bundle() check.
2065 *
2066 * When a policy's bundle is pruned, we dst_free() the XFRM
2067 * dst which causes it's ->obsolete field to be set to a
2068 * positive non-zero integer. If an XFRM dst has been pruned
2069 * like this, we want to force a new route lookup.
399c180a 2070 */
d49c73c7
DM		 2071 if (dst->obsolete < 0 && !stale_bundle(dst))
2072 return dst;
2073
1da177e4
LT		 2074 return NULL;
2075}
2076
2077static int stale_bundle(struct dst_entry *dst)
2078{
5b368e61 2079 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
1da177e4
LT		 2080}
2081
aabc9761 2082void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
1da177e4 2083{
1da177e4 2084 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
c346dca1 2085 dst->dev = dev_net(dev)->loopback_dev;
de3cb747 2086 dev_hold(dst->dev);
1da177e4
LT		 2087 dev_put(dev);
2088 }
2089}
aabc9761 2090EXPORT_SYMBOL(xfrm_dst_ifdown);
1da177e4
LT		 2091
2092static void xfrm_link_failure(struct sk_buff *skb)
2093{
2094 /* Impossible. Such dst must be popped before reaches point of failure. */
2095 return;
2096}
2097
2098static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2099{
2100 if (dst) {
2101 if (dst->obsolete) {
2102 dst_release(dst);
2103 dst = NULL;
2104 }
2105 }
2106 return dst;
2107}
2108
2518c7c2
DM		 2109static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p)
2110{
2111 struct dst_entry *dst, **dstp;
2112
2113 write_lock(&pol->lock);
2114 dstp = &pol->bundles;
2115 while ((dst=*dstp) != NULL) {
2116 if (func(dst)) {
2117 *dstp = dst->next;
2118 dst->next = *gc_list_p;
2119 *gc_list_p = dst;
2120 } else {
2121 dstp = &dst->next;
2122 }
2123 }
2124 write_unlock(&pol->lock);
2125}
2126
3dd0b499 2127static void xfrm_prune_bundles(struct net *net, int (*func)(struct dst_entry *))
1da177e4 2128{
2518c7c2
DM		 2129 struct dst_entry *gc_list = NULL;
2130 int dir;
1da177e4
LT		 2131
2132 read_lock_bh(&xfrm_policy_lock);
2518c7c2
DM		 2133 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2134 struct xfrm_policy *pol;
2135 struct hlist_node *entry;
2136 struct hlist_head *table;
2137 int i;
4e81bb83 2138
2518c7c2 2139 hlist_for_each_entry(pol, entry,
3dd0b499 2140 &net->xfrm.policy_inexact[dir], bydst)
2518c7c2
DM		 2141 prune_one_bundle(pol, func, &gc_list);
2142
3dd0b499
AD		 2143 table = net->xfrm.policy_bydst[dir].table;
2144 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2518c7c2
DM		 2145 hlist_for_each_entry(pol, entry, table + i, bydst)
2146 prune_one_bundle(pol, func, &gc_list);
1da177e4
LT		 2147 }
2148 }
2149 read_unlock_bh(&xfrm_policy_lock);
2150
2151 while (gc_list) {
2518c7c2 2152 struct dst_entry *dst = gc_list;
1da177e4
LT		 2153 gc_list = dst->next;
2154 dst_free(dst);
2155 }
2156}
2157
2158static int unused_bundle(struct dst_entry *dst)
2159{
2160 return !atomic_read(&dst->__refcnt);
2161}
2162
ddcfd796 2163static void __xfrm_garbage_collect(struct net *net)
1da177e4 2164{
ddcfd796 2165 xfrm_prune_bundles(net, unused_bundle);
1da177e4
LT		 2166}
2167
3dd0b499 2168static int xfrm_flush_bundles(struct net *net)
1da177e4 2169{
3dd0b499 2170 xfrm_prune_bundles(net, stale_bundle);
1da177e4
LT		 2171 return 0;
2172}
2173
25ee3286 2174static void xfrm_init_pmtu(struct dst_entry *dst)
1da177e4
LT		 2175{
2176 do {
2177 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2178 u32 pmtu, route_mtu_cached;
2179
2180 pmtu = dst_mtu(dst->child);
2181 xdst->child_mtu_cached = pmtu;
2182
2183 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2184
2185 route_mtu_cached = dst_mtu(xdst->route);
2186 xdst->route_mtu_cached = route_mtu_cached;
2187
2188 if (pmtu > route_mtu_cached)
2189 pmtu = route_mtu_cached;
2190
2191 dst->metrics[RTAX_MTU-1] = pmtu;
2192 } while ((dst = dst->next));
2193}
2194
1da177e4
LT		 2195/* Check that the bundle accepts the flow and its components are
2196 * still valid.
2197 */
2198
5b368e61
VY		 2199int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2200 struct flowi *fl, int family, int strict)
1da177e4
LT		 2201{
2202 struct dst_entry *dst = &first->u.dst;
2203 struct xfrm_dst *last;
2204 u32 mtu;
2205
92d63dec 2206 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
1da177e4
LT		 2207 (dst->dev && !netif_running(dst->dev)))
2208 return 0;
157bfc25
MN		 2209#ifdef CONFIG_XFRM_SUB_POLICY
2210 if (fl) {
2211 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2212 return 0;
2213 if (first->partner &&
2214 !xfrm_selector_match(first->partner, fl, family))
2215 return 0;
2216 }
2217#endif
1da177e4
LT		 2218
2219 last = NULL;
2220
2221 do {
2222 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2223
2224 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2225 return 0;
67f83cbf
VY		 2226 if (fl && pol &&
2227 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
e0d1caa7 2228 return 0;
1da177e4
LT		 2229 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2230 return 0;
9d4a706d
DM		 2231 if (xdst->genid != dst->xfrm->genid)
2232 return 0;
e53820de 2233
1bfcb10f 2234 if (strict && fl &&
13996378 2235 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
e53820de
MN		 2236 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2237 return 0;
1da177e4
LT		 2238
2239 mtu = dst_mtu(dst->child);
2240 if (xdst->child_mtu_cached != mtu) {
2241 last = xdst;
2242 xdst->child_mtu_cached = mtu;
2243 }
2244
92d63dec 2245 if (!dst_check(xdst->route, xdst->route_cookie))
1da177e4
LT		 2246 return 0;
2247 mtu = dst_mtu(xdst->route);
2248 if (xdst->route_mtu_cached != mtu) {
2249 last = xdst;
2250 xdst->route_mtu_cached = mtu;
2251 }
2252
2253 dst = dst->child;
2254 } while (dst->xfrm);
2255
2256 if (likely(!last))
2257 return 1;
2258
2259 mtu = last->child_mtu_cached;
2260 for (;;) {
2261 dst = &last->u.dst;
2262
2263 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2264 if (mtu > last->route_mtu_cached)
2265 mtu = last->route_mtu_cached;
2266 dst->metrics[RTAX_MTU-1] = mtu;
2267
2268 if (last == first)
2269 break;
2270
bd0bf076 2271 last = (struct xfrm_dst *)last->u.dst.next;
1da177e4
LT		 2272 last->child_mtu_cached = mtu;
2273 }
2274
2275 return 1;
2276}
2277
2278EXPORT_SYMBOL(xfrm_bundle_ok);
2279
1da177e4
LT		 2280int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2281{
2282 int err = 0;
2283 if (unlikely(afinfo == NULL))
2284 return -EINVAL;
2285 if (unlikely(afinfo->family >= NPROTO))
2286 return -EAFNOSUPPORT;
e959d812 2287 write_lock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2288 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2289 err = -ENOBUFS;
2290 else {
2291 struct dst_ops *dst_ops = afinfo->dst_ops;
2292 if (likely(dst_ops->kmem_cachep == NULL))
2293 dst_ops->kmem_cachep = xfrm_dst_cache;
2294 if (likely(dst_ops->check == NULL))
2295 dst_ops->check = xfrm_dst_check;
1da177e4
LT		 2296 if (likely(dst_ops->negative_advice == NULL))
2297 dst_ops->negative_advice = xfrm_negative_advice;
2298 if (likely(dst_ops->link_failure == NULL))
2299 dst_ops->link_failure = xfrm_link_failure;
1da177e4
LT		 2300 if (likely(afinfo->garbage_collect == NULL))
2301 afinfo->garbage_collect = __xfrm_garbage_collect;
2302 xfrm_policy_afinfo[afinfo->family] = afinfo;
2303 }
e959d812 2304 write_unlock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2305 return err;
2306}
2307EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2308
2309int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2310{
2311 int err = 0;
2312 if (unlikely(afinfo == NULL))
2313 return -EINVAL;
2314 if (unlikely(afinfo->family >= NPROTO))
2315 return -EAFNOSUPPORT;
e959d812 2316 write_lock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2317 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2318 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2319 err = -EINVAL;
2320 else {
2321 struct dst_ops *dst_ops = afinfo->dst_ops;
2322 xfrm_policy_afinfo[afinfo->family] = NULL;
2323 dst_ops->kmem_cachep = NULL;
2324 dst_ops->check = NULL;
1da177e4
LT		 2325 dst_ops->negative_advice = NULL;
2326 dst_ops->link_failure = NULL;
1da177e4
LT		 2327 afinfo->garbage_collect = NULL;
2328 }
2329 }
e959d812 2330 write_unlock_bh(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2331 return err;
2332}
2333EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2334
2335static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2336{
2337 struct xfrm_policy_afinfo *afinfo;
2338 if (unlikely(family >= NPROTO))
2339 return NULL;
2340 read_lock(&xfrm_policy_afinfo_lock);
2341 afinfo = xfrm_policy_afinfo[family];
546be240
HX		 2342 if (unlikely(!afinfo))
2343 read_unlock(&xfrm_policy_afinfo_lock);
1da177e4
LT		 2344 return afinfo;
2345}
2346
2347static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2348{
546be240
HX		 2349 read_unlock(&xfrm_policy_afinfo_lock);
2350}
2351
1da177e4
LT		 2352static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2353{
e9dc8653
EB		 2354 struct net_device *dev = ptr;
2355
1da177e4
LT		 2356 switch (event) {
2357 case NETDEV_DOWN:
3dd0b499 2358 xfrm_flush_bundles(dev_net(dev));
1da177e4
LT		 2359 }
2360 return NOTIFY_DONE;
2361}
2362
2363static struct notifier_block xfrm_dev_notifier = {
d5917a35 2364 .notifier_call = xfrm_dev_event,
1da177e4
LT		 2365};
2366
558f82ef 2367#ifdef CONFIG_XFRM_STATISTICS
59c9940e 2368static int __net_init xfrm_statistics_init(struct net *net)
558f82ef 2369{
c68cd1a0
AD		 2370 int rv;
2371
59c9940e 2372 if (snmp_mib_init((void **)net->mib.xfrm_statistics,
558f82ef
MN		 2373 sizeof(struct linux_xfrm_mib)) < 0)
2374 return -ENOMEM;
c68cd1a0
AD		 2375 rv = xfrm_proc_init(net);
2376 if (rv < 0)
2377 snmp_mib_free((void **)net->mib.xfrm_statistics);
2378 return rv;
558f82ef 2379}
59c9940e
AD		 2380
2381static void xfrm_statistics_fini(struct net *net)
2382{
c68cd1a0 2383 xfrm_proc_fini(net);
59c9940e
AD		 2384 snmp_mib_free((void **)net->mib.xfrm_statistics);
2385}
2386#else
2387static int __net_init xfrm_statistics_init(struct net *net)
2388{
2389 return 0;
2390}
2391
2392static void xfrm_statistics_fini(struct net *net)
2393{
2394}
558f82ef
MN		 2395#endif
2396
d62ddc21 2397static int __net_init xfrm_policy_init(struct net *net)
1da177e4 2398{
2518c7c2
DM		 2399 unsigned int hmask, sz;
2400 int dir;
2401
d62ddc21
AD		 2402 if (net_eq(net, &init_net))
2403 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
1da177e4 2404 sizeof(struct xfrm_dst),
e5d679f3 2405 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 2406 NULL);
1da177e4 2407
2518c7c2
DM		 2408 hmask = 8 - 1;
2409 sz = (hmask+1) * sizeof(struct hlist_head);
2410
93b851c1
AD		 2411 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2412 if (!net->xfrm.policy_byidx)
2413 goto out_byidx;
8100bea7 2414 net->xfrm.policy_idx_hmask = hmask;
2518c7c2
DM		 2415
2416 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2417 struct xfrm_policy_hash *htab;
2418
dc2caba7 2419 net->xfrm.policy_count[dir] = 0;
8b18f8ea 2420 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2518c7c2 2421
a35f6c5d 2422 htab = &net->xfrm.policy_bydst[dir];
44e36b42 2423 htab->table = xfrm_hash_alloc(sz);
2518c7c2 2424 if (!htab->table)
a35f6c5d
AD		 2425 goto out_bydst;
2426 htab->hmask = hmask;
2518c7c2
DM		 2427 }
2428
adfcf0b2 2429 INIT_LIST_HEAD(&net->xfrm.policy_all);
66caf628 2430 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
d62ddc21
AD		 2431 if (net_eq(net, &init_net))
2432 register_netdevice_notifier(&xfrm_dev_notifier);
2433 return 0;
93b851c1 2434
a35f6c5d
AD		 2435out_bydst:
2436 for (dir--; dir >= 0; dir--) {
2437 struct xfrm_policy_hash *htab;
2438
2439 htab = &net->xfrm.policy_bydst[dir];
2440 xfrm_hash_free(htab->table, sz);
2441 }
2442 xfrm_hash_free(net->xfrm.policy_byidx, sz);
93b851c1
AD		 2443out_byidx:
2444 return -ENOMEM;
d62ddc21
AD		 2445}
2446
2447static void xfrm_policy_fini(struct net *net)
2448{
7c2776ee 2449 struct xfrm_audit audit_info;
93b851c1 2450 unsigned int sz;
8b18f8ea 2451 int dir;
93b851c1 2452
7c2776ee
AD		 2453 flush_work(&net->xfrm.policy_hash_work);
2454#ifdef CONFIG_XFRM_SUB_POLICY
2455 audit_info.loginuid = -1;
2456 audit_info.sessionid = -1;
2457 audit_info.secid = 0;
2458 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2459#endif
2460 audit_info.loginuid = -1;
2461 audit_info.sessionid = -1;
2462 audit_info.secid = 0;
2463 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2464 flush_work(&xfrm_policy_gc_work);
2465
adfcf0b2 2466 WARN_ON(!list_empty(&net->xfrm.policy_all));
93b851c1 2467
8b18f8ea 2468 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
a35f6c5d
AD		 2469 struct xfrm_policy_hash *htab;
2470
8b18f8ea 2471 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
a35f6c5d
AD		 2472
2473 htab = &net->xfrm.policy_bydst[dir];
2474 sz = (htab->hmask + 1);
2475 WARN_ON(!hlist_empty(htab->table));
2476 xfrm_hash_free(htab->table, sz);
8b18f8ea
AD		 2477 }
2478
8100bea7 2479 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
93b851c1
AD		 2480 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2481 xfrm_hash_free(net->xfrm.policy_byidx, sz);
1da177e4
LT		 2482}
2483
d62ddc21
AD		 2484static int __net_init xfrm_net_init(struct net *net)
2485{
2486 int rv;
2487
59c9940e
AD		 2488 rv = xfrm_statistics_init(net);
2489 if (rv < 0)
2490 goto out_statistics;
d62ddc21
AD		 2491 rv = xfrm_state_init(net);
2492 if (rv < 0)
2493 goto out_state;
2494 rv = xfrm_policy_init(net);
2495 if (rv < 0)
2496 goto out_policy;
b27aeadb
AD		 2497 rv = xfrm_sysctl_init(net);
2498 if (rv < 0)
2499 goto out_sysctl;
d62ddc21
AD		 2500 return 0;
2501
b27aeadb
AD		 2502out_sysctl:
2503 xfrm_policy_fini(net);
d62ddc21
AD		 2504out_policy:
2505 xfrm_state_fini(net);
2506out_state:
59c9940e
AD		 2507 xfrm_statistics_fini(net);
2508out_statistics:
d62ddc21
AD		 2509 return rv;
2510}
2511
2512static void __net_exit xfrm_net_exit(struct net *net)
2513{
b27aeadb 2514 xfrm_sysctl_fini(net);
d62ddc21
AD		 2515 xfrm_policy_fini(net);
2516 xfrm_state_fini(net);
59c9940e 2517 xfrm_statistics_fini(net);
d62ddc21
AD		 2518}
2519
2520static struct pernet_operations __net_initdata xfrm_net_ops = {
2521 .init = xfrm_net_init,
2522 .exit = xfrm_net_exit,
2523};
2524
1da177e4
LT		 2525void __init xfrm_init(void)
2526{
d62ddc21 2527 register_pernet_subsys(&xfrm_net_ops);
1da177e4
LT		 2528 xfrm_input_init();
2529}
2530
ab5f5e8b 2531#ifdef CONFIG_AUDITSYSCALL
1486cbd7
IJ		 2532static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2533 struct audit_buffer *audit_buf)
ab5f5e8b 2534{
875179fa
PM		 2535 struct xfrm_sec_ctx *ctx = xp->security;
2536 struct xfrm_selector *sel = &xp->selector;
2537
2538 if (ctx)
ab5f5e8b 2539 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
875179fa 2540 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
ab5f5e8b 2541
875179fa 2542 switch(sel->family) {
ab5f5e8b 2543 case AF_INET:
21454aaa 2544 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
875179fa
PM		 2545 if (sel->prefixlen_s != 32)
2546 audit_log_format(audit_buf, " src_prefixlen=%d",
2547 sel->prefixlen_s);
21454aaa 2548 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
875179fa
PM		 2549 if (sel->prefixlen_d != 32)
2550 audit_log_format(audit_buf, " dst_prefixlen=%d",
2551 sel->prefixlen_d);
ab5f5e8b
JL		 2552 break;
2553 case AF_INET6:
5b095d98 2554 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
875179fa
PM		 2555 if (sel->prefixlen_s != 128)
2556 audit_log_format(audit_buf, " src_prefixlen=%d",
2557 sel->prefixlen_s);
5b095d98 2558 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
875179fa
PM		 2559 if (sel->prefixlen_d != 128)
2560 audit_log_format(audit_buf, " dst_prefixlen=%d",
2561 sel->prefixlen_d);
ab5f5e8b
JL		 2562 break;
2563 }
2564}
2565
68277acc 2566void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2532386f 2567 uid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 2568{
2569 struct audit_buffer *audit_buf;
ab5f5e8b 2570
afeb14b4 2571 audit_buf = xfrm_audit_start("SPD-add");
ab5f5e8b
JL		 2572 if (audit_buf == NULL)
2573 return;
2532386f 2574 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 2575 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 2576 xfrm_audit_common_policyinfo(xp, audit_buf);
2577 audit_log_end(audit_buf);
2578}
2579EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2580
68277acc 2581void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2532386f 2582 uid_t auid, u32 sessionid, u32 secid)
ab5f5e8b
JL		 2583{
2584 struct audit_buffer *audit_buf;
ab5f5e8b 2585
afeb14b4 2586 audit_buf = xfrm_audit_start("SPD-delete");
ab5f5e8b
JL		 2587 if (audit_buf == NULL)
2588 return;
2532386f 2589 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
afeb14b4 2590 audit_log_format(audit_buf, " res=%u", result);
ab5f5e8b
JL		 2591 xfrm_audit_common_policyinfo(xp, audit_buf);
2592 audit_log_end(audit_buf);
2593}
2594EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2595#endif
2596
80c9abaa
SS		 2597#ifdef CONFIG_XFRM_MIGRATE
2598static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2599 struct xfrm_selector *sel_tgt)
2600{
2601 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2602 if (sel_tgt->family == sel_cmp->family &&
2603 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
a716c119 2604 sel_cmp->family) == 0 &&
80c9abaa
SS		 2605 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2606 sel_cmp->family) == 0 &&
2607 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2608 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2609 return 1;
2610 }
2611 } else {
2612 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2613 return 1;
2614 }
2615 }
2616 return 0;
2617}
2618
2619static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2620 u8 dir, u8 type)
2621{
2622 struct xfrm_policy *pol, *ret = NULL;
2623 struct hlist_node *entry;
2624 struct hlist_head *chain;
2625 u32 priority = ~0U;
2626
2627 read_lock_bh(&xfrm_policy_lock);
1121994c 2628 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
80c9abaa
SS		 2629 hlist_for_each_entry(pol, entry, chain, bydst) {
2630 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2631 pol->type == type) {
2632 ret = pol;
2633 priority = ret->priority;
2634 break;
2635 }
2636 }
8b18f8ea 2637 chain = &init_net.xfrm.policy_inexact[dir];
80c9abaa
SS		 2638 hlist_for_each_entry(pol, entry, chain, bydst) {
2639 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2640 pol->type == type &&
2641 pol->priority < priority) {
2642 ret = pol;
2643 break;
2644 }
2645 }
2646
2647 if (ret)
2648 xfrm_pol_hold(ret);
2649
2650 read_unlock_bh(&xfrm_policy_lock);
2651
2652 return ret;
2653}
2654
2655static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2656{
2657 int match = 0;
2658
2659 if (t->mode == m->mode && t->id.proto == m->proto &&
2660 (m->reqid == 0 || t->reqid == m->reqid)) {
2661 switch (t->mode) {
2662 case XFRM_MODE_TUNNEL:
2663 case XFRM_MODE_BEET:
2664 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2665 m->old_family) == 0 &&
2666 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2667 m->old_family) == 0) {
2668 match = 1;
2669 }
2670 break;
2671 case XFRM_MODE_TRANSPORT:
2672 /* in case of transport mode, template does not store
2673 any IP addresses, hence we just compare mode and
2674 protocol */
2675 match = 1;
2676 break;
2677 default:
2678 break;
2679 }
2680 }
2681 return match;
2682}
2683
2684/* update endpoint address(es) of template(s) */
2685static int xfrm_policy_migrate(struct xfrm_policy *pol,
2686 struct xfrm_migrate *m, int num_migrate)
2687{
2688 struct xfrm_migrate *mp;
2689 struct dst_entry *dst;
2690 int i, j, n = 0;
2691
2692 write_lock_bh(&pol->lock);
12a169e7 2693 if (unlikely(pol->walk.dead)) {
80c9abaa
SS		 2694 /* target policy has been deleted */
2695 write_unlock_bh(&pol->lock);
2696 return -ENOENT;
2697 }
2698
2699 for (i = 0; i < pol->xfrm_nr; i++) {
2700 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2701 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2702 continue;
2703 n++;
1bfcb10f
HX		 2704 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2705 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
80c9abaa
SS		 2706 continue;
2707 /* update endpoints */
2708 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2709 sizeof(pol->xfrm_vec[i].id.daddr));
2710 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2711 sizeof(pol->xfrm_vec[i].saddr));
2712 pol->xfrm_vec[i].encap_family = mp->new_family;
2713 /* flush bundles */
2714 while ((dst = pol->bundles) != NULL) {
2715 pol->bundles = dst->next;
2716 dst_free(dst);
2717 }
2718 }
2719 }
2720
2721 write_unlock_bh(&pol->lock);
2722
2723 if (!n)
2724 return -ENODATA;
2725
2726 return 0;
2727}
2728
2729static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2730{
2731 int i, j;
2732
2733 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2734 return -EINVAL;
2735
2736 for (i = 0; i < num_migrate; i++) {
2737 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2738 m[i].old_family) == 0) &&
2739 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2740 m[i].old_family) == 0))
2741 return -EINVAL;
2742 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2743 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2744 return -EINVAL;
2745
2746 /* check if there is any duplicated entry */
2747 for (j = i + 1; j < num_migrate; j++) {
2748 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2749 sizeof(m[i].old_daddr)) &&
2750 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2751 sizeof(m[i].old_saddr)) &&
2752 m[i].proto == m[j].proto &&
2753 m[i].mode == m[j].mode &&
2754 m[i].reqid == m[j].reqid &&
2755 m[i].old_family == m[j].old_family)
2756 return -EINVAL;
2757 }
2758 }
2759
2760 return 0;
2761}
2762
2763int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
13c1d189
AE		 2764 struct xfrm_migrate *m, int num_migrate,
2765 struct xfrm_kmaddress *k)
80c9abaa
SS		 2766{
2767 int i, err, nx_cur = 0, nx_new = 0;
2768 struct xfrm_policy *pol = NULL;
2769 struct xfrm_state *x, *xc;
2770 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2771 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2772 struct xfrm_migrate *mp;
2773
2774 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2775 goto out;
2776
2777 /* Stage 1 - find policy */
2778 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2779 err = -ENOENT;
2780 goto out;
2781 }
2782
2783 /* Stage 2 - find and update state(s) */
2784 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2785 if ((x = xfrm_migrate_state_find(mp))) {
2786 x_cur[nx_cur] = x;
2787 nx_cur++;
2788 if ((xc = xfrm_state_migrate(x, mp))) {
2789 x_new[nx_new] = xc;
2790 nx_new++;
2791 } else {
2792 err = -ENODATA;
2793 goto restore_state;
2794 }
2795 }
2796 }
2797
2798 /* Stage 3 - update policy */
2799 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2800 goto restore_state;
2801
2802 /* Stage 4 - delete old state(s) */
2803 if (nx_cur) {
2804 xfrm_states_put(x_cur, nx_cur);
2805 xfrm_states_delete(x_cur, nx_cur);
2806 }
2807
2808 /* Stage 5 - announce */
13c1d189 2809 km_migrate(sel, dir, type, m, num_migrate, k);
80c9abaa
SS		 2810
2811 xfrm_pol_put(pol);
2812
2813 return 0;
2814out:
2815 return err;
2816
2817restore_state:
2818 if (pol)
2819 xfrm_pol_put(pol);
2820 if (nx_cur)
2821 xfrm_states_put(x_cur, nx_cur);
2822 if (nx_new)
2823 xfrm_states_delete(x_new, nx_new);
2824
2825 return err;
2826}
e610e679 2827EXPORT_SYMBOL(xfrm_migrate);
80c9abaa 2828#endif
Linux 6.x block layer and io_uring tree(s)