1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * xfrm_state.c
4 *
5 * Changes:
6 * Mitsuru KANDA @USAGI
7 * Kazunori MIYAZAWA @USAGI
8 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * IPv6 support
10 * YOSHIFUJI Hideaki @USAGI
11 * Split up af-specific functions
12 * Derek Atkins <derek@ihtfp.com>
13 * Add UDP Encapsulation
14 *
15 */
16
17 #include <linux/compat.h>
18 #include <linux/workqueue.h>
19 #include <net/xfrm.h>
20 #include <linux/pfkeyv2.h>
21 #include <linux/ipsec.h>
22 #include <linux/module.h>
23 #include <linux/cache.h>
24 #include <linux/audit.h>
25 #include <linux/uaccess.h>
26 #include <linux/ktime.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/kernel.h>
30
31 #include <crypto/aead.h>
32
33 #include "xfrm_hash.h"
34
35 #define xfrm_state_deref_prot(table, net) \
36 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
37 #define xfrm_state_deref_check(table, net) \
38 rcu_dereference_check((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
39
40 static void xfrm_state_gc_task(struct work_struct *work);
41
42 /* Each xfrm_state may be linked to two tables:
43
44 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
45 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
46 destination/tunnel endpoint. (output)
47 */
48
49 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
50 static struct kmem_cache *xfrm_state_cache __ro_after_init;
51
52 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
53 static HLIST_HEAD(xfrm_state_gc_list);
54 static HLIST_HEAD(xfrm_state_dev_gc_list);
55
xfrm_state_hold_rcu(struct xfrm_state __rcu * x)56 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
57 {
58 return refcount_inc_not_zero(&x->refcnt);
59 }
60
xfrm_dst_hash(struct net * net,const xfrm_address_t * daddr,const xfrm_address_t * saddr,u32 reqid,unsigned short family)61 static inline unsigned int xfrm_dst_hash(struct net *net,
62 const xfrm_address_t *daddr,
63 const xfrm_address_t *saddr,
64 u32 reqid,
65 unsigned short family)
66 {
67 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
68
69 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
70 }
71
xfrm_src_hash(struct net * net,const xfrm_address_t * daddr,const xfrm_address_t * saddr,unsigned short family)72 static inline unsigned int xfrm_src_hash(struct net *net,
73 const xfrm_address_t *daddr,
74 const xfrm_address_t *saddr,
75 unsigned short family)
76 {
77 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
78
79 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
80 }
81
82 static inline unsigned int
xfrm_spi_hash(struct net * net,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)83 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
84 __be32 spi, u8 proto, unsigned short family)
85 {
86 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
87
88 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
89 }
90
xfrm_seq_hash(struct net * net,u32 seq)91 static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
92 {
93 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
94
95 return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
96 }
97
98 #define XFRM_STATE_INSERT(by, _n, _h, _type) \
99 { \
100 struct xfrm_state *_x = NULL; \
101 \
102 if (_type != XFRM_DEV_OFFLOAD_PACKET) { \
103 hlist_for_each_entry_rcu(_x, _h, by) { \
104 if (_x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
105 continue; \
106 break; \
107 } \
108 } \
109 \
110 if (!_x || _x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
111 /* SAD is empty or consist from HW SAs only */ \
112 hlist_add_head_rcu(_n, _h); \
113 else \
114 hlist_add_before_rcu(_n, &_x->by); \
115 }
116
xfrm_hash_transfer(struct hlist_head * list,struct hlist_head * ndsttable,struct hlist_head * nsrctable,struct hlist_head * nspitable,struct hlist_head * nseqtable,unsigned int nhashmask)117 static void xfrm_hash_transfer(struct hlist_head *list,
118 struct hlist_head *ndsttable,
119 struct hlist_head *nsrctable,
120 struct hlist_head *nspitable,
121 struct hlist_head *nseqtable,
122 unsigned int nhashmask)
123 {
124 struct hlist_node *tmp;
125 struct xfrm_state *x;
126
127 hlist_for_each_entry_safe(x, tmp, list, bydst) {
128 unsigned int h;
129
130 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
131 x->props.reqid, x->props.family,
132 nhashmask);
133 XFRM_STATE_INSERT(bydst, &x->bydst, ndsttable + h, x->xso.type);
134
135 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
136 x->props.family,
137 nhashmask);
138 XFRM_STATE_INSERT(bysrc, &x->bysrc, nsrctable + h, x->xso.type);
139
140 if (x->id.spi) {
141 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
142 x->id.proto, x->props.family,
143 nhashmask);
144 XFRM_STATE_INSERT(byspi, &x->byspi, nspitable + h,
145 x->xso.type);
146 }
147
148 if (x->km.seq) {
149 h = __xfrm_seq_hash(x->km.seq, nhashmask);
150 XFRM_STATE_INSERT(byseq, &x->byseq, nseqtable + h,
151 x->xso.type);
152 }
153 }
154 }
155
xfrm_hash_new_size(unsigned int state_hmask)156 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
157 {
158 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
159 }
160
xfrm_hash_resize(struct work_struct * work)161 static void xfrm_hash_resize(struct work_struct *work)
162 {
163 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
164 struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq;
165 unsigned long nsize, osize;
166 unsigned int nhashmask, ohashmask;
167 int i;
168
169 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
170 ndst = xfrm_hash_alloc(nsize);
171 if (!ndst)
172 return;
173 nsrc = xfrm_hash_alloc(nsize);
174 if (!nsrc) {
175 xfrm_hash_free(ndst, nsize);
176 return;
177 }
178 nspi = xfrm_hash_alloc(nsize);
179 if (!nspi) {
180 xfrm_hash_free(ndst, nsize);
181 xfrm_hash_free(nsrc, nsize);
182 return;
183 }
184 nseq = xfrm_hash_alloc(nsize);
185 if (!nseq) {
186 xfrm_hash_free(ndst, nsize);
187 xfrm_hash_free(nsrc, nsize);
188 xfrm_hash_free(nspi, nsize);
189 return;
190 }
191
192 spin_lock_bh(&net->xfrm.xfrm_state_lock);
193 write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
194
195 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
196 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
197 for (i = net->xfrm.state_hmask; i >= 0; i--)
198 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask);
199
200 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
201 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
202 oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net);
203 ohashmask = net->xfrm.state_hmask;
204
205 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
206 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
207 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
208 rcu_assign_pointer(net->xfrm.state_byseq, nseq);
209 net->xfrm.state_hmask = nhashmask;
210
211 write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
212 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
213
214 osize = (ohashmask + 1) * sizeof(struct hlist_head);
215
216 synchronize_rcu();
217
218 xfrm_hash_free(odst, osize);
219 xfrm_hash_free(osrc, osize);
220 xfrm_hash_free(ospi, osize);
221 xfrm_hash_free(oseq, osize);
222 }
223
224 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
225 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
226
227 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
228 static DEFINE_SPINLOCK(xfrm_state_dev_gc_lock);
229
230 int __xfrm_state_delete(struct xfrm_state *x);
231
232 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
233 static bool km_is_alive(const struct km_event *c);
234 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
235
xfrm_register_type(const struct xfrm_type * type,unsigned short family)236 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
237 {
238 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
239 int err = 0;
240
241 if (!afinfo)
242 return -EAFNOSUPPORT;
243
244 #define X(afi, T, name) do { \
245 WARN_ON((afi)->type_ ## name); \
246 (afi)->type_ ## name = (T); \
247 } while (0)
248
249 switch (type->proto) {
250 case IPPROTO_COMP:
251 X(afinfo, type, comp);
252 break;
253 case IPPROTO_AH:
254 X(afinfo, type, ah);
255 break;
256 case IPPROTO_ESP:
257 X(afinfo, type, esp);
258 break;
259 case IPPROTO_IPIP:
260 X(afinfo, type, ipip);
261 break;
262 case IPPROTO_DSTOPTS:
263 X(afinfo, type, dstopts);
264 break;
265 case IPPROTO_ROUTING:
266 X(afinfo, type, routing);
267 break;
268 case IPPROTO_IPV6:
269 X(afinfo, type, ipip6);
270 break;
271 default:
272 WARN_ON(1);
273 err = -EPROTONOSUPPORT;
274 break;
275 }
276 #undef X
277 rcu_read_unlock();
278 return err;
279 }
280 EXPORT_SYMBOL(xfrm_register_type);
281
xfrm_unregister_type(const struct xfrm_type * type,unsigned short family)282 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
283 {
284 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
285
286 if (unlikely(afinfo == NULL))
287 return;
288
289 #define X(afi, T, name) do { \
290 WARN_ON((afi)->type_ ## name != (T)); \
291 (afi)->type_ ## name = NULL; \
292 } while (0)
293
294 switch (type->proto) {
295 case IPPROTO_COMP:
296 X(afinfo, type, comp);
297 break;
298 case IPPROTO_AH:
299 X(afinfo, type, ah);
300 break;
301 case IPPROTO_ESP:
302 X(afinfo, type, esp);
303 break;
304 case IPPROTO_IPIP:
305 X(afinfo, type, ipip);
306 break;
307 case IPPROTO_DSTOPTS:
308 X(afinfo, type, dstopts);
309 break;
310 case IPPROTO_ROUTING:
311 X(afinfo, type, routing);
312 break;
313 case IPPROTO_IPV6:
314 X(afinfo, type, ipip6);
315 break;
316 default:
317 WARN_ON(1);
318 break;
319 }
320 #undef X
321 rcu_read_unlock();
322 }
323 EXPORT_SYMBOL(xfrm_unregister_type);
324
xfrm_get_type(u8 proto,unsigned short family)325 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
326 {
327 const struct xfrm_type *type = NULL;
328 struct xfrm_state_afinfo *afinfo;
329 int modload_attempted = 0;
330
331 retry:
332 afinfo = xfrm_state_get_afinfo(family);
333 if (unlikely(afinfo == NULL))
334 return NULL;
335
336 switch (proto) {
337 case IPPROTO_COMP:
338 type = afinfo->type_comp;
339 break;
340 case IPPROTO_AH:
341 type = afinfo->type_ah;
342 break;
343 case IPPROTO_ESP:
344 type = afinfo->type_esp;
345 break;
346 case IPPROTO_IPIP:
347 type = afinfo->type_ipip;
348 break;
349 case IPPROTO_DSTOPTS:
350 type = afinfo->type_dstopts;
351 break;
352 case IPPROTO_ROUTING:
353 type = afinfo->type_routing;
354 break;
355 case IPPROTO_IPV6:
356 type = afinfo->type_ipip6;
357 break;
358 default:
359 break;
360 }
361
362 if (unlikely(type && !try_module_get(type->owner)))
363 type = NULL;
364
365 rcu_read_unlock();
366
367 if (!type && !modload_attempted) {
368 request_module("xfrm-type-%d-%d", family, proto);
369 modload_attempted = 1;
370 goto retry;
371 }
372
373 return type;
374 }
375
xfrm_put_type(const struct xfrm_type * type)376 static void xfrm_put_type(const struct xfrm_type *type)
377 {
378 module_put(type->owner);
379 }
380
xfrm_register_type_offload(const struct xfrm_type_offload * type,unsigned short family)381 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
382 unsigned short family)
383 {
384 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
385 int err = 0;
386
387 if (unlikely(afinfo == NULL))
388 return -EAFNOSUPPORT;
389
390 switch (type->proto) {
391 case IPPROTO_ESP:
392 WARN_ON(afinfo->type_offload_esp);
393 afinfo->type_offload_esp = type;
394 break;
395 default:
396 WARN_ON(1);
397 err = -EPROTONOSUPPORT;
398 break;
399 }
400
401 rcu_read_unlock();
402 return err;
403 }
404 EXPORT_SYMBOL(xfrm_register_type_offload);
405
xfrm_unregister_type_offload(const struct xfrm_type_offload * type,unsigned short family)406 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
407 unsigned short family)
408 {
409 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
410
411 if (unlikely(afinfo == NULL))
412 return;
413
414 switch (type->proto) {
415 case IPPROTO_ESP:
416 WARN_ON(afinfo->type_offload_esp != type);
417 afinfo->type_offload_esp = NULL;
418 break;
419 default:
420 WARN_ON(1);
421 break;
422 }
423 rcu_read_unlock();
424 }
425 EXPORT_SYMBOL(xfrm_unregister_type_offload);
426
xfrm_set_type_offload(struct xfrm_state * x,bool try_load)427 void xfrm_set_type_offload(struct xfrm_state *x, bool try_load)
428 {
429 const struct xfrm_type_offload *type = NULL;
430 struct xfrm_state_afinfo *afinfo;
431
432 retry:
433 afinfo = xfrm_state_get_afinfo(x->props.family);
434 if (unlikely(afinfo == NULL))
435 goto out;
436
437 switch (x->id.proto) {
438 case IPPROTO_ESP:
439 type = afinfo->type_offload_esp;
440 break;
441 default:
442 break;
443 }
444
445 if ((type && !try_module_get(type->owner)))
446 type = NULL;
447
448 rcu_read_unlock();
449
450 if (!type && try_load) {
451 request_module("xfrm-offload-%d-%d", x->props.family,
452 x->id.proto);
453 try_load = false;
454 goto retry;
455 }
456
457 out:
458 x->type_offload = type;
459 }
460 EXPORT_SYMBOL(xfrm_set_type_offload);
461
462 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
463 [XFRM_MODE_BEET] = {
464 .encap = XFRM_MODE_BEET,
465 .flags = XFRM_MODE_FLAG_TUNNEL,
466 .family = AF_INET,
467 },
468 [XFRM_MODE_TRANSPORT] = {
469 .encap = XFRM_MODE_TRANSPORT,
470 .family = AF_INET,
471 },
472 [XFRM_MODE_TUNNEL] = {
473 .encap = XFRM_MODE_TUNNEL,
474 .flags = XFRM_MODE_FLAG_TUNNEL,
475 .family = AF_INET,
476 },
477 [XFRM_MODE_IPTFS] = {
478 .encap = XFRM_MODE_IPTFS,
479 .flags = XFRM_MODE_FLAG_TUNNEL,
480 .family = AF_INET,
481 },
482 };
483
484 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
485 [XFRM_MODE_BEET] = {
486 .encap = XFRM_MODE_BEET,
487 .flags = XFRM_MODE_FLAG_TUNNEL,
488 .family = AF_INET6,
489 },
490 [XFRM_MODE_ROUTEOPTIMIZATION] = {
491 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
492 .family = AF_INET6,
493 },
494 [XFRM_MODE_TRANSPORT] = {
495 .encap = XFRM_MODE_TRANSPORT,
496 .family = AF_INET6,
497 },
498 [XFRM_MODE_TUNNEL] = {
499 .encap = XFRM_MODE_TUNNEL,
500 .flags = XFRM_MODE_FLAG_TUNNEL,
501 .family = AF_INET6,
502 },
503 [XFRM_MODE_IPTFS] = {
504 .encap = XFRM_MODE_IPTFS,
505 .flags = XFRM_MODE_FLAG_TUNNEL,
506 .family = AF_INET6,
507 },
508 };
509
xfrm_get_mode(unsigned int encap,int family)510 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
511 {
512 const struct xfrm_mode *mode;
513
514 if (unlikely(encap >= XFRM_MODE_MAX))
515 return NULL;
516
517 switch (family) {
518 case AF_INET:
519 mode = &xfrm4_mode_map[encap];
520 if (mode->family == family)
521 return mode;
522 break;
523 case AF_INET6:
524 mode = &xfrm6_mode_map[encap];
525 if (mode->family == family)
526 return mode;
527 break;
528 default:
529 break;
530 }
531
532 return NULL;
533 }
534
535 static const struct xfrm_mode_cbs __rcu *xfrm_mode_cbs_map[XFRM_MODE_MAX];
536 static DEFINE_SPINLOCK(xfrm_mode_cbs_map_lock);
537
xfrm_register_mode_cbs(u8 mode,const struct xfrm_mode_cbs * mode_cbs)538 int xfrm_register_mode_cbs(u8 mode, const struct xfrm_mode_cbs *mode_cbs)
539 {
540 if (mode >= XFRM_MODE_MAX)
541 return -EINVAL;
542
543 spin_lock_bh(&xfrm_mode_cbs_map_lock);
544 rcu_assign_pointer(xfrm_mode_cbs_map[mode], mode_cbs);
545 spin_unlock_bh(&xfrm_mode_cbs_map_lock);
546
547 return 0;
548 }
549 EXPORT_SYMBOL(xfrm_register_mode_cbs);
550
xfrm_unregister_mode_cbs(u8 mode)551 void xfrm_unregister_mode_cbs(u8 mode)
552 {
553 if (mode >= XFRM_MODE_MAX)
554 return;
555
556 spin_lock_bh(&xfrm_mode_cbs_map_lock);
557 RCU_INIT_POINTER(xfrm_mode_cbs_map[mode], NULL);
558 spin_unlock_bh(&xfrm_mode_cbs_map_lock);
559 synchronize_rcu();
560 }
561 EXPORT_SYMBOL(xfrm_unregister_mode_cbs);
562
xfrm_get_mode_cbs(u8 mode)563 static const struct xfrm_mode_cbs *xfrm_get_mode_cbs(u8 mode)
564 {
565 const struct xfrm_mode_cbs *cbs;
566 bool try_load = true;
567
568 if (mode >= XFRM_MODE_MAX)
569 return NULL;
570
571 retry:
572 rcu_read_lock();
573
574 cbs = rcu_dereference(xfrm_mode_cbs_map[mode]);
575 if (cbs && !try_module_get(cbs->owner))
576 cbs = NULL;
577
578 rcu_read_unlock();
579
580 if (mode == XFRM_MODE_IPTFS && !cbs && try_load) {
581 request_module("xfrm-iptfs");
582 try_load = false;
583 goto retry;
584 }
585
586 return cbs;
587 }
588
xfrm_state_free(struct xfrm_state * x)589 void xfrm_state_free(struct xfrm_state *x)
590 {
591 kmem_cache_free(xfrm_state_cache, x);
592 }
593 EXPORT_SYMBOL(xfrm_state_free);
594
595 static void xfrm_state_delete_tunnel(struct xfrm_state *x);
xfrm_state_gc_destroy(struct xfrm_state * x)596 static void xfrm_state_gc_destroy(struct xfrm_state *x)
597 {
598 if (x->mode_cbs && x->mode_cbs->destroy_state)
599 x->mode_cbs->destroy_state(x);
600 hrtimer_cancel(&x->mtimer);
601 timer_delete_sync(&x->rtimer);
602 kfree_sensitive(x->aead);
603 kfree_sensitive(x->aalg);
604 kfree_sensitive(x->ealg);
605 kfree(x->calg);
606 kfree(x->encap);
607 kfree(x->coaddr);
608 kfree(x->replay_esn);
609 kfree(x->preplay_esn);
610 xfrm_unset_type_offload(x);
611 xfrm_state_delete_tunnel(x);
612 if (x->type) {
613 x->type->destructor(x);
614 xfrm_put_type(x->type);
615 }
616 if (x->xfrag.page)
617 put_page(x->xfrag.page);
618 xfrm_dev_state_free(x);
619 security_xfrm_state_free(x);
620 xfrm_state_free(x);
621 }
622
xfrm_state_gc_task(struct work_struct * work)623 static void xfrm_state_gc_task(struct work_struct *work)
624 {
625 struct xfrm_state *x;
626 struct hlist_node *tmp;
627 struct hlist_head gc_list;
628
629 spin_lock_bh(&xfrm_state_gc_lock);
630 hlist_move_list(&xfrm_state_gc_list, &gc_list);
631 spin_unlock_bh(&xfrm_state_gc_lock);
632
633 synchronize_rcu();
634
635 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
636 xfrm_state_gc_destroy(x);
637 }
638
xfrm_timer_handler(struct hrtimer * me)639 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
640 {
641 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
642 enum hrtimer_restart ret = HRTIMER_NORESTART;
643 time64_t now = ktime_get_real_seconds();
644 time64_t next = TIME64_MAX;
645 int warn = 0;
646 int err = 0;
647
648 spin_lock(&x->lock);
649 xfrm_dev_state_update_stats(x);
650
651 if (x->km.state == XFRM_STATE_DEAD)
652 goto out;
653 if (x->km.state == XFRM_STATE_EXPIRED)
654 goto expired;
655 if (x->lft.hard_add_expires_seconds) {
656 time64_t tmo = x->lft.hard_add_expires_seconds +
657 x->curlft.add_time - now;
658 if (tmo <= 0) {
659 if (x->xflags & XFRM_SOFT_EXPIRE) {
660 /* enter hard expire without soft expire first?!
661 * setting a new date could trigger this.
662 * workaround: fix x->curflt.add_time by below:
663 */
664 x->curlft.add_time = now - x->saved_tmo - 1;
665 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
666 } else
667 goto expired;
668 }
669 if (tmo < next)
670 next = tmo;
671 }
672 if (x->lft.hard_use_expires_seconds) {
673 time64_t tmo = x->lft.hard_use_expires_seconds +
674 (READ_ONCE(x->curlft.use_time) ? : now) - now;
675 if (tmo <= 0)
676 goto expired;
677 if (tmo < next)
678 next = tmo;
679 }
680 if (x->km.dying)
681 goto resched;
682 if (x->lft.soft_add_expires_seconds) {
683 time64_t tmo = x->lft.soft_add_expires_seconds +
684 x->curlft.add_time - now;
685 if (tmo <= 0) {
686 warn = 1;
687 x->xflags &= ~XFRM_SOFT_EXPIRE;
688 } else if (tmo < next) {
689 next = tmo;
690 x->xflags |= XFRM_SOFT_EXPIRE;
691 x->saved_tmo = tmo;
692 }
693 }
694 if (x->lft.soft_use_expires_seconds) {
695 time64_t tmo = x->lft.soft_use_expires_seconds +
696 (READ_ONCE(x->curlft.use_time) ? : now) - now;
697 if (tmo <= 0)
698 warn = 1;
699 else if (tmo < next)
700 next = tmo;
701 }
702
703 x->km.dying = warn;
704 if (warn)
705 km_state_expired(x, 0, 0);
706 resched:
707 if (next != TIME64_MAX) {
708 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
709 ret = HRTIMER_RESTART;
710 }
711
712 goto out;
713
714 expired:
715 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
716 x->km.state = XFRM_STATE_EXPIRED;
717
718 err = __xfrm_state_delete(x);
719 if (!err)
720 km_state_expired(x, 1, 0);
721
722 xfrm_audit_state_delete(x, err ? 0 : 1, true);
723
724 out:
725 spin_unlock(&x->lock);
726 return ret;
727 }
728
729 static void xfrm_replay_timer_handler(struct timer_list *t);
730
xfrm_state_alloc(struct net * net)731 struct xfrm_state *xfrm_state_alloc(struct net *net)
732 {
733 struct xfrm_state *x;
734
735 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
736
737 if (x) {
738 write_pnet(&x->xs_net, net);
739 refcount_set(&x->refcnt, 1);
740 atomic_set(&x->tunnel_users, 0);
741 INIT_LIST_HEAD(&x->km.all);
742 INIT_HLIST_NODE(&x->state_cache);
743 INIT_HLIST_NODE(&x->bydst);
744 INIT_HLIST_NODE(&x->bysrc);
745 INIT_HLIST_NODE(&x->byspi);
746 INIT_HLIST_NODE(&x->byseq);
747 hrtimer_setup(&x->mtimer, xfrm_timer_handler, CLOCK_BOOTTIME,
748 HRTIMER_MODE_ABS_SOFT);
749 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
750 x->curlft.add_time = ktime_get_real_seconds();
751 x->lft.soft_byte_limit = XFRM_INF;
752 x->lft.soft_packet_limit = XFRM_INF;
753 x->lft.hard_byte_limit = XFRM_INF;
754 x->lft.hard_packet_limit = XFRM_INF;
755 x->replay_maxage = 0;
756 x->replay_maxdiff = 0;
757 x->pcpu_num = UINT_MAX;
758 spin_lock_init(&x->lock);
759 x->mode_data = NULL;
760 }
761 return x;
762 }
763 EXPORT_SYMBOL(xfrm_state_alloc);
764
765 #ifdef CONFIG_XFRM_OFFLOAD
xfrm_dev_state_delete(struct xfrm_state * x)766 void xfrm_dev_state_delete(struct xfrm_state *x)
767 {
768 struct xfrm_dev_offload *xso = &x->xso;
769 struct net_device *dev = READ_ONCE(xso->dev);
770
771 if (dev) {
772 dev->xfrmdev_ops->xdo_dev_state_delete(dev, x);
773 spin_lock_bh(&xfrm_state_dev_gc_lock);
774 hlist_add_head(&x->dev_gclist, &xfrm_state_dev_gc_list);
775 spin_unlock_bh(&xfrm_state_dev_gc_lock);
776 }
777 }
778 EXPORT_SYMBOL_GPL(xfrm_dev_state_delete);
779
xfrm_dev_state_free(struct xfrm_state * x)780 void xfrm_dev_state_free(struct xfrm_state *x)
781 {
782 struct xfrm_dev_offload *xso = &x->xso;
783 struct net_device *dev = READ_ONCE(xso->dev);
784
785 if (dev && dev->xfrmdev_ops) {
786 spin_lock_bh(&xfrm_state_dev_gc_lock);
787 if (!hlist_unhashed(&x->dev_gclist))
788 hlist_del(&x->dev_gclist);
789 spin_unlock_bh(&xfrm_state_dev_gc_lock);
790
791 if (dev->xfrmdev_ops->xdo_dev_state_free)
792 dev->xfrmdev_ops->xdo_dev_state_free(dev, x);
793 WRITE_ONCE(xso->dev, NULL);
794 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
795 netdev_put(dev, &xso->dev_tracker);
796 }
797 }
798 #endif
799
__xfrm_state_destroy(struct xfrm_state * x)800 void __xfrm_state_destroy(struct xfrm_state *x)
801 {
802 WARN_ON(x->km.state != XFRM_STATE_DEAD);
803
804 spin_lock_bh(&xfrm_state_gc_lock);
805 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
806 spin_unlock_bh(&xfrm_state_gc_lock);
807 schedule_work(&xfrm_state_gc_work);
808 }
809 EXPORT_SYMBOL(__xfrm_state_destroy);
810
__xfrm_state_delete(struct xfrm_state * x)811 int __xfrm_state_delete(struct xfrm_state *x)
812 {
813 struct net *net = xs_net(x);
814 int err = -ESRCH;
815
816 if (x->km.state != XFRM_STATE_DEAD) {
817 x->km.state = XFRM_STATE_DEAD;
818
819 spin_lock(&net->xfrm.xfrm_state_lock);
820 list_del(&x->km.all);
821 hlist_del_rcu(&x->bydst);
822 hlist_del_rcu(&x->bysrc);
823 if (x->km.seq)
824 hlist_del_rcu(&x->byseq);
825 if (!hlist_unhashed(&x->state_cache))
826 hlist_del_rcu(&x->state_cache);
827 if (!hlist_unhashed(&x->state_cache_input))
828 hlist_del_rcu(&x->state_cache_input);
829
830 if (x->id.spi)
831 hlist_del_rcu(&x->byspi);
832 net->xfrm.state_num--;
833 xfrm_nat_keepalive_state_updated(x);
834 spin_unlock(&net->xfrm.xfrm_state_lock);
835
836 xfrm_dev_state_delete(x);
837
838 xfrm_state_delete_tunnel(x);
839
840 /* All xfrm_state objects are created by xfrm_state_alloc.
841 * The xfrm_state_alloc call gives a reference, and that
842 * is what we are dropping here.
843 */
844 xfrm_state_put(x);
845 err = 0;
846 }
847
848 return err;
849 }
850 EXPORT_SYMBOL(__xfrm_state_delete);
851
xfrm_state_delete(struct xfrm_state * x)852 int xfrm_state_delete(struct xfrm_state *x)
853 {
854 int err;
855
856 spin_lock_bh(&x->lock);
857 err = __xfrm_state_delete(x);
858 spin_unlock_bh(&x->lock);
859
860 return err;
861 }
862 EXPORT_SYMBOL(xfrm_state_delete);
863
864 #ifdef CONFIG_SECURITY_NETWORK_XFRM
865 static inline int
xfrm_state_flush_secctx_check(struct net * net,u8 proto,bool task_valid)866 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
867 {
868 int i, err = 0;
869
870 for (i = 0; i <= net->xfrm.state_hmask; i++) {
871 struct xfrm_state *x;
872
873 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
874 if (xfrm_id_proto_match(x->id.proto, proto) &&
875 (err = security_xfrm_state_delete(x)) != 0) {
876 xfrm_audit_state_delete(x, 0, task_valid);
877 return err;
878 }
879 }
880 }
881
882 return err;
883 }
884
885 static inline int
xfrm_dev_state_flush_secctx_check(struct net * net,struct net_device * dev,bool task_valid)886 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
887 {
888 int i, err = 0;
889
890 for (i = 0; i <= net->xfrm.state_hmask; i++) {
891 struct xfrm_state *x;
892 struct xfrm_dev_offload *xso;
893
894 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
895 xso = &x->xso;
896
897 if (xso->dev == dev &&
898 (err = security_xfrm_state_delete(x)) != 0) {
899 xfrm_audit_state_delete(x, 0, task_valid);
900 return err;
901 }
902 }
903 }
904
905 return err;
906 }
907 #else
908 static inline int
xfrm_state_flush_secctx_check(struct net * net,u8 proto,bool task_valid)909 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
910 {
911 return 0;
912 }
913
914 static inline int
xfrm_dev_state_flush_secctx_check(struct net * net,struct net_device * dev,bool task_valid)915 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
916 {
917 return 0;
918 }
919 #endif
920
xfrm_state_flush(struct net * net,u8 proto,bool task_valid)921 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
922 {
923 int i, err = 0, cnt = 0;
924
925 spin_lock_bh(&net->xfrm.xfrm_state_lock);
926 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
927 if (err)
928 goto out;
929
930 err = -ESRCH;
931 for (i = 0; i <= net->xfrm.state_hmask; i++) {
932 struct xfrm_state *x;
933 restart:
934 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
935 if (!xfrm_state_kern(x) &&
936 xfrm_id_proto_match(x->id.proto, proto)) {
937 xfrm_state_hold(x);
938 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
939
940 err = xfrm_state_delete(x);
941 xfrm_audit_state_delete(x, err ? 0 : 1,
942 task_valid);
943 xfrm_state_put(x);
944 if (!err)
945 cnt++;
946
947 spin_lock_bh(&net->xfrm.xfrm_state_lock);
948 goto restart;
949 }
950 }
951 }
952 out:
953 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
954 if (cnt)
955 err = 0;
956
957 return err;
958 }
959 EXPORT_SYMBOL(xfrm_state_flush);
960
xfrm_dev_state_flush(struct net * net,struct net_device * dev,bool task_valid)961 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
962 {
963 struct xfrm_state *x;
964 struct hlist_node *tmp;
965 struct xfrm_dev_offload *xso;
966 int i, err = 0, cnt = 0;
967
968 spin_lock_bh(&net->xfrm.xfrm_state_lock);
969 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
970 if (err)
971 goto out;
972
973 err = -ESRCH;
974 for (i = 0; i <= net->xfrm.state_hmask; i++) {
975 restart:
976 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
977 xso = &x->xso;
978
979 if (!xfrm_state_kern(x) && xso->dev == dev) {
980 xfrm_state_hold(x);
981 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
982
983 err = xfrm_state_delete(x);
984 xfrm_dev_state_free(x);
985
986 xfrm_audit_state_delete(x, err ? 0 : 1,
987 task_valid);
988 xfrm_state_put(x);
989 if (!err)
990 cnt++;
991
992 spin_lock_bh(&net->xfrm.xfrm_state_lock);
993 goto restart;
994 }
995 }
996 }
997 if (cnt)
998 err = 0;
999
1000 out:
1001 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1002
1003 spin_lock_bh(&xfrm_state_dev_gc_lock);
1004 restart_gc:
1005 hlist_for_each_entry_safe(x, tmp, &xfrm_state_dev_gc_list, dev_gclist) {
1006 xso = &x->xso;
1007
1008 if (xso->dev == dev) {
1009 spin_unlock_bh(&xfrm_state_dev_gc_lock);
1010 xfrm_dev_state_free(x);
1011 spin_lock_bh(&xfrm_state_dev_gc_lock);
1012 goto restart_gc;
1013 }
1014
1015 }
1016 spin_unlock_bh(&xfrm_state_dev_gc_lock);
1017
1018 xfrm_flush_gc();
1019
1020 return err;
1021 }
1022 EXPORT_SYMBOL(xfrm_dev_state_flush);
1023
xfrm_sad_getinfo(struct net * net,struct xfrmk_sadinfo * si)1024 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
1025 {
1026 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1027 si->sadcnt = net->xfrm.state_num;
1028 si->sadhcnt = net->xfrm.state_hmask + 1;
1029 si->sadhmcnt = xfrm_state_hashmax;
1030 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1031 }
1032 EXPORT_SYMBOL(xfrm_sad_getinfo);
1033
1034 static void
__xfrm4_init_tempsel(struct xfrm_selector * sel,const struct flowi * fl)1035 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
1036 {
1037 const struct flowi4 *fl4 = &fl->u.ip4;
1038
1039 sel->daddr.a4 = fl4->daddr;
1040 sel->saddr.a4 = fl4->saddr;
1041 sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
1042 sel->dport_mask = htons(0xffff);
1043 sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
1044 sel->sport_mask = htons(0xffff);
1045 sel->family = AF_INET;
1046 sel->prefixlen_d = 32;
1047 sel->prefixlen_s = 32;
1048 sel->proto = fl4->flowi4_proto;
1049 sel->ifindex = fl4->flowi4_oif;
1050 }
1051
1052 static void
__xfrm6_init_tempsel(struct xfrm_selector * sel,const struct flowi * fl)1053 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
1054 {
1055 const struct flowi6 *fl6 = &fl->u.ip6;
1056
1057 /* Initialize temporary selector matching only to current session. */
1058 *(struct in6_addr *)&sel->daddr = fl6->daddr;
1059 *(struct in6_addr *)&sel->saddr = fl6->saddr;
1060 sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
1061 sel->dport_mask = htons(0xffff);
1062 sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
1063 sel->sport_mask = htons(0xffff);
1064 sel->family = AF_INET6;
1065 sel->prefixlen_d = 128;
1066 sel->prefixlen_s = 128;
1067 sel->proto = fl6->flowi6_proto;
1068 sel->ifindex = fl6->flowi6_oif;
1069 }
1070
1071 static void
xfrm_init_tempstate(struct xfrm_state * x,const struct flowi * fl,const struct xfrm_tmpl * tmpl,const xfrm_address_t * daddr,const xfrm_address_t * saddr,unsigned short family)1072 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
1073 const struct xfrm_tmpl *tmpl,
1074 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1075 unsigned short family)
1076 {
1077 switch (family) {
1078 case AF_INET:
1079 __xfrm4_init_tempsel(&x->sel, fl);
1080 break;
1081 case AF_INET6:
1082 __xfrm6_init_tempsel(&x->sel, fl);
1083 break;
1084 }
1085
1086 x->id = tmpl->id;
1087
1088 switch (tmpl->encap_family) {
1089 case AF_INET:
1090 if (x->id.daddr.a4 == 0)
1091 x->id.daddr.a4 = daddr->a4;
1092 x->props.saddr = tmpl->saddr;
1093 if (x->props.saddr.a4 == 0)
1094 x->props.saddr.a4 = saddr->a4;
1095 break;
1096 case AF_INET6:
1097 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
1098 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
1099 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
1100 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
1101 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
1102 break;
1103 }
1104
1105 x->props.mode = tmpl->mode;
1106 x->props.reqid = tmpl->reqid;
1107 x->props.family = tmpl->encap_family;
1108 }
1109
1110 struct xfrm_hash_state_ptrs {
1111 const struct hlist_head *bydst;
1112 const struct hlist_head *bysrc;
1113 const struct hlist_head *byspi;
1114 unsigned int hmask;
1115 };
1116
xfrm_hash_ptrs_get(const struct net * net,struct xfrm_hash_state_ptrs * ptrs)1117 static void xfrm_hash_ptrs_get(const struct net *net, struct xfrm_hash_state_ptrs *ptrs)
1118 {
1119 unsigned int sequence;
1120
1121 do {
1122 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1123
1124 ptrs->bydst = xfrm_state_deref_check(net->xfrm.state_bydst, net);
1125 ptrs->bysrc = xfrm_state_deref_check(net->xfrm.state_bysrc, net);
1126 ptrs->byspi = xfrm_state_deref_check(net->xfrm.state_byspi, net);
1127 ptrs->hmask = net->xfrm.state_hmask;
1128 } while (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence));
1129 }
1130
__xfrm_state_lookup_all(const struct xfrm_hash_state_ptrs * state_ptrs,u32 mark,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family,struct xfrm_dev_offload * xdo)1131 static struct xfrm_state *__xfrm_state_lookup_all(const struct xfrm_hash_state_ptrs *state_ptrs,
1132 u32 mark,
1133 const xfrm_address_t *daddr,
1134 __be32 spi, u8 proto,
1135 unsigned short family,
1136 struct xfrm_dev_offload *xdo)
1137 {
1138 unsigned int h = __xfrm_spi_hash(daddr, spi, proto, family, state_ptrs->hmask);
1139 struct xfrm_state *x;
1140
1141 hlist_for_each_entry_rcu(x, state_ptrs->byspi + h, byspi) {
1142 #ifdef CONFIG_XFRM_OFFLOAD
1143 if (xdo->type == XFRM_DEV_OFFLOAD_PACKET) {
1144 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1145 /* HW states are in the head of list, there is
1146 * no need to iterate further.
1147 */
1148 break;
1149
1150 /* Packet offload: both policy and SA should
1151 * have same device.
1152 */
1153 if (xdo->dev != x->xso.dev)
1154 continue;
1155 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1156 /* Skip HW policy for SW lookups */
1157 continue;
1158 #endif
1159 if (x->props.family != family ||
1160 x->id.spi != spi ||
1161 x->id.proto != proto ||
1162 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1163 continue;
1164
1165 if ((mark & x->mark.m) != x->mark.v)
1166 continue;
1167 if (!xfrm_state_hold_rcu(x))
1168 continue;
1169 return x;
1170 }
1171
1172 return NULL;
1173 }
1174
__xfrm_state_lookup(const struct xfrm_hash_state_ptrs * state_ptrs,u32 mark,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)1175 static struct xfrm_state *__xfrm_state_lookup(const struct xfrm_hash_state_ptrs *state_ptrs,
1176 u32 mark,
1177 const xfrm_address_t *daddr,
1178 __be32 spi, u8 proto,
1179 unsigned short family)
1180 {
1181 unsigned int h = __xfrm_spi_hash(daddr, spi, proto, family, state_ptrs->hmask);
1182 struct xfrm_state *x;
1183
1184 hlist_for_each_entry_rcu(x, state_ptrs->byspi + h, byspi) {
1185 if (x->props.family != family ||
1186 x->id.spi != spi ||
1187 x->id.proto != proto ||
1188 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1189 continue;
1190
1191 if ((mark & x->mark.m) != x->mark.v)
1192 continue;
1193 if (!xfrm_state_hold_rcu(x))
1194 continue;
1195 return x;
1196 }
1197
1198 return NULL;
1199 }
1200
xfrm_input_state_lookup(struct net * net,u32 mark,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)1201 struct xfrm_state *xfrm_input_state_lookup(struct net *net, u32 mark,
1202 const xfrm_address_t *daddr,
1203 __be32 spi, u8 proto,
1204 unsigned short family)
1205 {
1206 struct xfrm_hash_state_ptrs state_ptrs;
1207 struct hlist_head *state_cache_input;
1208 struct xfrm_state *x = NULL;
1209
1210 state_cache_input = raw_cpu_ptr(net->xfrm.state_cache_input);
1211
1212 rcu_read_lock();
1213 hlist_for_each_entry_rcu(x, state_cache_input, state_cache_input) {
1214 if (x->props.family != family ||
1215 x->id.spi != spi ||
1216 x->id.proto != proto ||
1217 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1218 continue;
1219
1220 if ((mark & x->mark.m) != x->mark.v)
1221 continue;
1222 if (!xfrm_state_hold_rcu(x))
1223 continue;
1224 goto out;
1225 }
1226
1227 xfrm_hash_ptrs_get(net, &state_ptrs);
1228
1229 x = __xfrm_state_lookup(&state_ptrs, mark, daddr, spi, proto, family);
1230
1231 if (x && x->km.state == XFRM_STATE_VALID) {
1232 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1233 if (hlist_unhashed(&x->state_cache_input)) {
1234 hlist_add_head_rcu(&x->state_cache_input, state_cache_input);
1235 } else {
1236 hlist_del_rcu(&x->state_cache_input);
1237 hlist_add_head_rcu(&x->state_cache_input, state_cache_input);
1238 }
1239 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1240 }
1241
1242 out:
1243 rcu_read_unlock();
1244 return x;
1245 }
1246 EXPORT_SYMBOL(xfrm_input_state_lookup);
1247
__xfrm_state_lookup_byaddr(const struct xfrm_hash_state_ptrs * state_ptrs,u32 mark,const xfrm_address_t * daddr,const xfrm_address_t * saddr,u8 proto,unsigned short family)1248 static struct xfrm_state *__xfrm_state_lookup_byaddr(const struct xfrm_hash_state_ptrs *state_ptrs,
1249 u32 mark,
1250 const xfrm_address_t *daddr,
1251 const xfrm_address_t *saddr,
1252 u8 proto, unsigned short family)
1253 {
1254 unsigned int h = __xfrm_src_hash(daddr, saddr, family, state_ptrs->hmask);
1255 struct xfrm_state *x;
1256
1257 hlist_for_each_entry_rcu(x, state_ptrs->bysrc + h, bysrc) {
1258 if (x->props.family != family ||
1259 x->id.proto != proto ||
1260 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1261 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1262 continue;
1263
1264 if ((mark & x->mark.m) != x->mark.v)
1265 continue;
1266 if (!xfrm_state_hold_rcu(x))
1267 continue;
1268 return x;
1269 }
1270
1271 return NULL;
1272 }
1273
1274 static inline struct xfrm_state *
__xfrm_state_locate(struct xfrm_state * x,int use_spi,int family)1275 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
1276 {
1277 struct xfrm_hash_state_ptrs state_ptrs;
1278 struct net *net = xs_net(x);
1279 u32 mark = x->mark.v & x->mark.m;
1280
1281 xfrm_hash_ptrs_get(net, &state_ptrs);
1282
1283 if (use_spi)
1284 return __xfrm_state_lookup(&state_ptrs, mark, &x->id.daddr,
1285 x->id.spi, x->id.proto, family);
1286 else
1287 return __xfrm_state_lookup_byaddr(&state_ptrs, mark,
1288 &x->id.daddr,
1289 &x->props.saddr,
1290 x->id.proto, family);
1291 }
1292
xfrm_hash_grow_check(struct net * net,int have_hash_collision)1293 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
1294 {
1295 if (have_hash_collision &&
1296 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1297 net->xfrm.state_num > net->xfrm.state_hmask)
1298 schedule_work(&net->xfrm.state_hash_work);
1299 }
1300
xfrm_state_look_at(struct xfrm_policy * pol,struct xfrm_state * x,const struct flowi * fl,unsigned short family,struct xfrm_state ** best,int * acq_in_progress,int * error,unsigned int pcpu_id)1301 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1302 const struct flowi *fl, unsigned short family,
1303 struct xfrm_state **best, int *acq_in_progress,
1304 int *error, unsigned int pcpu_id)
1305 {
1306 /* Resolution logic:
1307 * 1. There is a valid state with matching selector. Done.
1308 * 2. Valid state with inappropriate selector. Skip.
1309 *
1310 * Entering area of "sysdeps".
1311 *
1312 * 3. If state is not valid, selector is temporary, it selects
1313 * only session which triggered previous resolution. Key
1314 * manager will do something to install a state with proper
1315 * selector.
1316 */
1317 if (x->km.state == XFRM_STATE_VALID) {
1318 if ((x->sel.family &&
1319 (x->sel.family != family ||
1320 !xfrm_selector_match(&x->sel, fl, family))) ||
1321 !security_xfrm_state_pol_flow_match(x, pol,
1322 &fl->u.__fl_common))
1323 return;
1324
1325 if (x->pcpu_num != UINT_MAX && x->pcpu_num != pcpu_id)
1326 return;
1327
1328 if (!*best ||
1329 ((*best)->pcpu_num == UINT_MAX && x->pcpu_num == pcpu_id) ||
1330 (*best)->km.dying > x->km.dying ||
1331 ((*best)->km.dying == x->km.dying &&
1332 (*best)->curlft.add_time < x->curlft.add_time))
1333 *best = x;
1334 } else if (x->km.state == XFRM_STATE_ACQ) {
1335 if (!*best || x->pcpu_num == pcpu_id)
1336 *acq_in_progress = 1;
1337 } else if (x->km.state == XFRM_STATE_ERROR ||
1338 x->km.state == XFRM_STATE_EXPIRED) {
1339 if ((!x->sel.family ||
1340 (x->sel.family == family &&
1341 xfrm_selector_match(&x->sel, fl, family))) &&
1342 security_xfrm_state_pol_flow_match(x, pol,
1343 &fl->u.__fl_common))
1344 *error = -ESRCH;
1345 }
1346 }
1347
1348 struct xfrm_state *
xfrm_state_find(const xfrm_address_t * daddr,const xfrm_address_t * saddr,const struct flowi * fl,struct xfrm_tmpl * tmpl,struct xfrm_policy * pol,int * err,unsigned short family,u32 if_id)1349 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1350 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1351 struct xfrm_policy *pol, int *err,
1352 unsigned short family, u32 if_id)
1353 {
1354 static xfrm_address_t saddr_wildcard = { };
1355 struct xfrm_hash_state_ptrs state_ptrs;
1356 struct net *net = xp_net(pol);
1357 unsigned int h, h_wildcard;
1358 struct xfrm_state *x, *x0, *to_put;
1359 int acquire_in_progress = 0;
1360 int error = 0;
1361 struct xfrm_state *best = NULL;
1362 u32 mark = pol->mark.v & pol->mark.m;
1363 unsigned short encap_family = tmpl->encap_family;
1364 unsigned int sequence;
1365 struct km_event c;
1366 unsigned int pcpu_id;
1367 bool cached = false;
1368
1369 /* We need the cpu id just as a lookup key,
1370 * we don't require it to be stable.
1371 */
1372 pcpu_id = raw_smp_processor_id();
1373
1374 to_put = NULL;
1375
1376 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1377
1378 rcu_read_lock();
1379 xfrm_hash_ptrs_get(net, &state_ptrs);
1380
1381 hlist_for_each_entry_rcu(x, &pol->state_cache_list, state_cache) {
1382 if (x->props.family == encap_family &&
1383 x->props.reqid == tmpl->reqid &&
1384 (mark & x->mark.m) == x->mark.v &&
1385 x->if_id == if_id &&
1386 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1387 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1388 tmpl->mode == x->props.mode &&
1389 tmpl->id.proto == x->id.proto &&
1390 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1391 xfrm_state_look_at(pol, x, fl, encap_family,
1392 &best, &acquire_in_progress, &error, pcpu_id);
1393 }
1394
1395 if (best)
1396 goto cached;
1397
1398 hlist_for_each_entry_rcu(x, &pol->state_cache_list, state_cache) {
1399 if (x->props.family == encap_family &&
1400 x->props.reqid == tmpl->reqid &&
1401 (mark & x->mark.m) == x->mark.v &&
1402 x->if_id == if_id &&
1403 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1404 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1405 tmpl->mode == x->props.mode &&
1406 tmpl->id.proto == x->id.proto &&
1407 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1408 xfrm_state_look_at(pol, x, fl, family,
1409 &best, &acquire_in_progress, &error, pcpu_id);
1410 }
1411
1412 cached:
1413 cached = true;
1414 if (best)
1415 goto found;
1416 else if (error)
1417 best = NULL;
1418 else if (acquire_in_progress) /* XXX: acquire_in_progress should not happen */
1419 WARN_ON(1);
1420
1421 h = __xfrm_dst_hash(daddr, saddr, tmpl->reqid, encap_family, state_ptrs.hmask);
1422 hlist_for_each_entry_rcu(x, state_ptrs.bydst + h, bydst) {
1423 #ifdef CONFIG_XFRM_OFFLOAD
1424 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1425 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1426 /* HW states are in the head of list, there is
1427 * no need to iterate further.
1428 */
1429 break;
1430
1431 /* Packet offload: both policy and SA should
1432 * have same device.
1433 */
1434 if (pol->xdo.dev != x->xso.dev)
1435 continue;
1436 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1437 /* Skip HW policy for SW lookups */
1438 continue;
1439 #endif
1440 if (x->props.family == encap_family &&
1441 x->props.reqid == tmpl->reqid &&
1442 (mark & x->mark.m) == x->mark.v &&
1443 x->if_id == if_id &&
1444 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1445 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1446 tmpl->mode == x->props.mode &&
1447 tmpl->id.proto == x->id.proto &&
1448 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1449 xfrm_state_look_at(pol, x, fl, family,
1450 &best, &acquire_in_progress, &error, pcpu_id);
1451 }
1452 if (best || acquire_in_progress)
1453 goto found;
1454
1455 h_wildcard = __xfrm_dst_hash(daddr, &saddr_wildcard, tmpl->reqid,
1456 encap_family, state_ptrs.hmask);
1457 hlist_for_each_entry_rcu(x, state_ptrs.bydst + h_wildcard, bydst) {
1458 #ifdef CONFIG_XFRM_OFFLOAD
1459 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1460 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1461 /* HW states are in the head of list, there is
1462 * no need to iterate further.
1463 */
1464 break;
1465
1466 /* Packet offload: both policy and SA should
1467 * have same device.
1468 */
1469 if (pol->xdo.dev != x->xso.dev)
1470 continue;
1471 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1472 /* Skip HW policy for SW lookups */
1473 continue;
1474 #endif
1475 if (x->props.family == encap_family &&
1476 x->props.reqid == tmpl->reqid &&
1477 (mark & x->mark.m) == x->mark.v &&
1478 x->if_id == if_id &&
1479 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1480 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1481 tmpl->mode == x->props.mode &&
1482 tmpl->id.proto == x->id.proto &&
1483 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1484 xfrm_state_look_at(pol, x, fl, family,
1485 &best, &acquire_in_progress, &error, pcpu_id);
1486 }
1487
1488 found:
1489 if (!(pol->flags & XFRM_POLICY_CPU_ACQUIRE) ||
1490 (best && (best->pcpu_num == pcpu_id)))
1491 x = best;
1492
1493 if (!x && !error && !acquire_in_progress) {
1494 if (tmpl->id.spi &&
1495 (x0 = __xfrm_state_lookup_all(&state_ptrs, mark, daddr,
1496 tmpl->id.spi, tmpl->id.proto,
1497 encap_family,
1498 &pol->xdo)) != NULL) {
1499 to_put = x0;
1500 error = -EEXIST;
1501 goto out;
1502 }
1503
1504 c.net = net;
1505 /* If the KMs have no listeners (yet...), avoid allocating an SA
1506 * for each and every packet - garbage collection might not
1507 * handle the flood.
1508 */
1509 if (!km_is_alive(&c)) {
1510 error = -ESRCH;
1511 goto out;
1512 }
1513
1514 x = xfrm_state_alloc(net);
1515 if (x == NULL) {
1516 error = -ENOMEM;
1517 goto out;
1518 }
1519 /* Initialize temporary state matching only
1520 * to current session. */
1521 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1522 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1523 x->if_id = if_id;
1524 if ((pol->flags & XFRM_POLICY_CPU_ACQUIRE) && best)
1525 x->pcpu_num = pcpu_id;
1526
1527 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1528 if (error) {
1529 x->km.state = XFRM_STATE_DEAD;
1530 to_put = x;
1531 x = NULL;
1532 goto out;
1533 }
1534 #ifdef CONFIG_XFRM_OFFLOAD
1535 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1536 struct xfrm_dev_offload *xdo = &pol->xdo;
1537 struct xfrm_dev_offload *xso = &x->xso;
1538 struct net_device *dev = xdo->dev;
1539
1540 xso->type = XFRM_DEV_OFFLOAD_PACKET;
1541 xso->dir = xdo->dir;
1542 xso->dev = dev;
1543 xso->flags = XFRM_DEV_OFFLOAD_FLAG_ACQ;
1544 netdev_hold(dev, &xso->dev_tracker, GFP_ATOMIC);
1545 error = dev->xfrmdev_ops->xdo_dev_state_add(dev, x,
1546 NULL);
1547 if (error) {
1548 xso->dir = 0;
1549 netdev_put(dev, &xso->dev_tracker);
1550 xso->dev = NULL;
1551 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
1552 x->km.state = XFRM_STATE_DEAD;
1553 to_put = x;
1554 x = NULL;
1555 goto out;
1556 }
1557 }
1558 #endif
1559 if (km_query(x, tmpl, pol) == 0) {
1560 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1561 x->km.state = XFRM_STATE_ACQ;
1562 x->dir = XFRM_SA_DIR_OUT;
1563 list_add(&x->km.all, &net->xfrm.state_all);
1564 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1565 XFRM_STATE_INSERT(bydst, &x->bydst,
1566 net->xfrm.state_bydst + h,
1567 x->xso.type);
1568 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1569 XFRM_STATE_INSERT(bysrc, &x->bysrc,
1570 net->xfrm.state_bysrc + h,
1571 x->xso.type);
1572 INIT_HLIST_NODE(&x->state_cache);
1573 if (x->id.spi) {
1574 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1575 XFRM_STATE_INSERT(byspi, &x->byspi,
1576 net->xfrm.state_byspi + h,
1577 x->xso.type);
1578 }
1579 if (x->km.seq) {
1580 h = xfrm_seq_hash(net, x->km.seq);
1581 XFRM_STATE_INSERT(byseq, &x->byseq,
1582 net->xfrm.state_byseq + h,
1583 x->xso.type);
1584 }
1585 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1586 hrtimer_start(&x->mtimer,
1587 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1588 HRTIMER_MODE_REL_SOFT);
1589 net->xfrm.state_num++;
1590 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1591 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1592 } else {
1593 #ifdef CONFIG_XFRM_OFFLOAD
1594 struct xfrm_dev_offload *xso = &x->xso;
1595
1596 if (xso->type == XFRM_DEV_OFFLOAD_PACKET) {
1597 xfrm_dev_state_delete(x);
1598 xfrm_dev_state_free(x);
1599 }
1600 #endif
1601 x->km.state = XFRM_STATE_DEAD;
1602 to_put = x;
1603 x = NULL;
1604 error = -ESRCH;
1605 }
1606
1607 /* Use the already installed 'fallback' while the CPU-specific
1608 * SA acquire is handled*/
1609 if (best)
1610 x = best;
1611 }
1612 out:
1613 if (x) {
1614 if (!xfrm_state_hold_rcu(x)) {
1615 *err = -EAGAIN;
1616 x = NULL;
1617 }
1618 } else {
1619 *err = acquire_in_progress ? -EAGAIN : error;
1620 }
1621
1622 if (x && x->km.state == XFRM_STATE_VALID && !cached &&
1623 (!(pol->flags & XFRM_POLICY_CPU_ACQUIRE) || x->pcpu_num == pcpu_id)) {
1624 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1625 if (hlist_unhashed(&x->state_cache))
1626 hlist_add_head_rcu(&x->state_cache, &pol->state_cache_list);
1627 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1628 }
1629
1630 rcu_read_unlock();
1631 if (to_put)
1632 xfrm_state_put(to_put);
1633
1634 if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1635 *err = -EAGAIN;
1636 if (x) {
1637 xfrm_state_put(x);
1638 x = NULL;
1639 }
1640 }
1641
1642 return x;
1643 }
1644
1645 struct xfrm_state *
xfrm_stateonly_find(struct net * net,u32 mark,u32 if_id,xfrm_address_t * daddr,xfrm_address_t * saddr,unsigned short family,u8 mode,u8 proto,u32 reqid)1646 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1647 xfrm_address_t *daddr, xfrm_address_t *saddr,
1648 unsigned short family, u8 mode, u8 proto, u32 reqid)
1649 {
1650 unsigned int h;
1651 struct xfrm_state *rx = NULL, *x = NULL;
1652
1653 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1654 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1655 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1656 if (x->props.family == family &&
1657 x->props.reqid == reqid &&
1658 (mark & x->mark.m) == x->mark.v &&
1659 x->if_id == if_id &&
1660 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1661 xfrm_state_addr_check(x, daddr, saddr, family) &&
1662 mode == x->props.mode &&
1663 proto == x->id.proto &&
1664 x->km.state == XFRM_STATE_VALID) {
1665 rx = x;
1666 break;
1667 }
1668 }
1669
1670 if (rx)
1671 xfrm_state_hold(rx);
1672 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1673
1674
1675 return rx;
1676 }
1677 EXPORT_SYMBOL(xfrm_stateonly_find);
1678
xfrm_state_lookup_byspi(struct net * net,__be32 spi,unsigned short family)1679 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1680 unsigned short family)
1681 {
1682 struct xfrm_state *x;
1683 struct xfrm_state_walk *w;
1684
1685 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1686 list_for_each_entry(w, &net->xfrm.state_all, all) {
1687 x = container_of(w, struct xfrm_state, km);
1688 if (x->props.family != family ||
1689 x->id.spi != spi)
1690 continue;
1691
1692 xfrm_state_hold(x);
1693 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1694 return x;
1695 }
1696 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1697 return NULL;
1698 }
1699 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1700
xfrm_state_lookup_spi_proto(struct net * net,__be32 spi,u8 proto)1701 static struct xfrm_state *xfrm_state_lookup_spi_proto(struct net *net, __be32 spi, u8 proto)
1702 {
1703 struct xfrm_state *x;
1704 unsigned int i;
1705
1706 rcu_read_lock();
1707 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1708 hlist_for_each_entry_rcu(x, &net->xfrm.state_byspi[i], byspi) {
1709 if (x->id.spi == spi && x->id.proto == proto) {
1710 if (!xfrm_state_hold_rcu(x))
1711 continue;
1712 rcu_read_unlock();
1713 return x;
1714 }
1715 }
1716 }
1717 rcu_read_unlock();
1718 return NULL;
1719 }
1720
__xfrm_state_insert(struct xfrm_state * x)1721 static void __xfrm_state_insert(struct xfrm_state *x)
1722 {
1723 struct net *net = xs_net(x);
1724 unsigned int h;
1725
1726 list_add(&x->km.all, &net->xfrm.state_all);
1727
1728 /* Sanitize mark before store */
1729 x->mark.v &= x->mark.m;
1730
1731 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1732 x->props.reqid, x->props.family);
1733 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1734 x->xso.type);
1735
1736 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1737 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1738 x->xso.type);
1739
1740 if (x->id.spi) {
1741 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1742 x->props.family);
1743
1744 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
1745 x->xso.type);
1746 }
1747
1748 if (x->km.seq) {
1749 h = xfrm_seq_hash(net, x->km.seq);
1750
1751 XFRM_STATE_INSERT(byseq, &x->byseq, net->xfrm.state_byseq + h,
1752 x->xso.type);
1753 }
1754
1755 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1756 if (x->replay_maxage)
1757 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1758
1759 net->xfrm.state_num++;
1760
1761 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1762 xfrm_nat_keepalive_state_updated(x);
1763 }
1764
1765 /* net->xfrm.xfrm_state_lock is held */
__xfrm_state_bump_genids(struct xfrm_state * xnew)1766 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1767 {
1768 struct net *net = xs_net(xnew);
1769 unsigned short family = xnew->props.family;
1770 u32 reqid = xnew->props.reqid;
1771 struct xfrm_state *x;
1772 unsigned int h;
1773 u32 mark = xnew->mark.v & xnew->mark.m;
1774 u32 if_id = xnew->if_id;
1775 u32 cpu_id = xnew->pcpu_num;
1776
1777 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1778 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1779 if (x->props.family == family &&
1780 x->props.reqid == reqid &&
1781 x->if_id == if_id &&
1782 x->pcpu_num == cpu_id &&
1783 (mark & x->mark.m) == x->mark.v &&
1784 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1785 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1786 x->genid++;
1787 }
1788 }
1789
xfrm_state_insert(struct xfrm_state * x)1790 void xfrm_state_insert(struct xfrm_state *x)
1791 {
1792 struct net *net = xs_net(x);
1793
1794 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1795 __xfrm_state_bump_genids(x);
1796 __xfrm_state_insert(x);
1797 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1798 }
1799 EXPORT_SYMBOL(xfrm_state_insert);
1800
1801 /* net->xfrm.xfrm_state_lock is held */
__find_acq_core(struct net * net,const struct xfrm_mark * m,unsigned short family,u8 mode,u32 reqid,u32 if_id,u32 pcpu_num,u8 proto,const xfrm_address_t * daddr,const xfrm_address_t * saddr,int create)1802 static struct xfrm_state *__find_acq_core(struct net *net,
1803 const struct xfrm_mark *m,
1804 unsigned short family, u8 mode,
1805 u32 reqid, u32 if_id, u32 pcpu_num, u8 proto,
1806 const xfrm_address_t *daddr,
1807 const xfrm_address_t *saddr,
1808 int create)
1809 {
1810 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1811 struct xfrm_state *x;
1812 u32 mark = m->v & m->m;
1813
1814 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1815 if (x->props.reqid != reqid ||
1816 x->props.mode != mode ||
1817 x->props.family != family ||
1818 x->km.state != XFRM_STATE_ACQ ||
1819 x->id.spi != 0 ||
1820 x->id.proto != proto ||
1821 (mark & x->mark.m) != x->mark.v ||
1822 x->pcpu_num != pcpu_num ||
1823 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1824 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1825 continue;
1826
1827 xfrm_state_hold(x);
1828 return x;
1829 }
1830
1831 if (!create)
1832 return NULL;
1833
1834 x = xfrm_state_alloc(net);
1835 if (likely(x)) {
1836 switch (family) {
1837 case AF_INET:
1838 x->sel.daddr.a4 = daddr->a4;
1839 x->sel.saddr.a4 = saddr->a4;
1840 x->sel.prefixlen_d = 32;
1841 x->sel.prefixlen_s = 32;
1842 x->props.saddr.a4 = saddr->a4;
1843 x->id.daddr.a4 = daddr->a4;
1844 break;
1845
1846 case AF_INET6:
1847 x->sel.daddr.in6 = daddr->in6;
1848 x->sel.saddr.in6 = saddr->in6;
1849 x->sel.prefixlen_d = 128;
1850 x->sel.prefixlen_s = 128;
1851 x->props.saddr.in6 = saddr->in6;
1852 x->id.daddr.in6 = daddr->in6;
1853 break;
1854 }
1855
1856 x->pcpu_num = pcpu_num;
1857 x->km.state = XFRM_STATE_ACQ;
1858 x->id.proto = proto;
1859 x->props.family = family;
1860 x->props.mode = mode;
1861 x->props.reqid = reqid;
1862 x->if_id = if_id;
1863 x->mark.v = m->v;
1864 x->mark.m = m->m;
1865 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1866 xfrm_state_hold(x);
1867 hrtimer_start(&x->mtimer,
1868 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1869 HRTIMER_MODE_REL_SOFT);
1870 list_add(&x->km.all, &net->xfrm.state_all);
1871 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1872 x->xso.type);
1873 h = xfrm_src_hash(net, daddr, saddr, family);
1874 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1875 x->xso.type);
1876
1877 net->xfrm.state_num++;
1878
1879 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1880 }
1881
1882 return x;
1883 }
1884
1885 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num);
1886
xfrm_state_add(struct xfrm_state * x)1887 int xfrm_state_add(struct xfrm_state *x)
1888 {
1889 struct net *net = xs_net(x);
1890 struct xfrm_state *x1, *to_put;
1891 int family;
1892 int err;
1893 u32 mark = x->mark.v & x->mark.m;
1894 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1895
1896 family = x->props.family;
1897
1898 to_put = NULL;
1899
1900 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1901
1902 x1 = __xfrm_state_locate(x, use_spi, family);
1903 if (x1) {
1904 to_put = x1;
1905 x1 = NULL;
1906 err = -EEXIST;
1907 goto out;
1908 }
1909
1910 if (use_spi && x->km.seq) {
1911 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq, x->pcpu_num);
1912 if (x1 && ((x1->id.proto != x->id.proto) ||
1913 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1914 to_put = x1;
1915 x1 = NULL;
1916 }
1917 }
1918
1919 if (use_spi && !x1)
1920 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1921 x->props.reqid, x->if_id, x->pcpu_num, x->id.proto,
1922 &x->id.daddr, &x->props.saddr, 0);
1923
1924 __xfrm_state_bump_genids(x);
1925 __xfrm_state_insert(x);
1926 err = 0;
1927
1928 out:
1929 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1930
1931 if (x1) {
1932 xfrm_state_delete(x1);
1933 xfrm_state_put(x1);
1934 }
1935
1936 if (to_put)
1937 xfrm_state_put(to_put);
1938
1939 return err;
1940 }
1941 EXPORT_SYMBOL(xfrm_state_add);
1942
1943 #ifdef CONFIG_XFRM_MIGRATE
clone_security(struct xfrm_state * x,struct xfrm_sec_ctx * security)1944 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1945 {
1946 struct xfrm_user_sec_ctx *uctx;
1947 int size = sizeof(*uctx) + security->ctx_len;
1948 int err;
1949
1950 uctx = kmalloc(size, GFP_KERNEL);
1951 if (!uctx)
1952 return -ENOMEM;
1953
1954 uctx->exttype = XFRMA_SEC_CTX;
1955 uctx->len = size;
1956 uctx->ctx_doi = security->ctx_doi;
1957 uctx->ctx_alg = security->ctx_alg;
1958 uctx->ctx_len = security->ctx_len;
1959 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1960 err = security_xfrm_state_alloc(x, uctx);
1961 kfree(uctx);
1962 if (err)
1963 return err;
1964
1965 return 0;
1966 }
1967
xfrm_state_clone_and_setup(struct xfrm_state * orig,struct xfrm_encap_tmpl * encap,struct xfrm_migrate * m)1968 static struct xfrm_state *xfrm_state_clone_and_setup(struct xfrm_state *orig,
1969 struct xfrm_encap_tmpl *encap,
1970 struct xfrm_migrate *m)
1971 {
1972 struct net *net = xs_net(orig);
1973 struct xfrm_state *x = xfrm_state_alloc(net);
1974 if (!x)
1975 goto out;
1976
1977 memcpy(&x->id, &orig->id, sizeof(x->id));
1978 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1979 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1980 x->props.mode = orig->props.mode;
1981 x->props.replay_window = orig->props.replay_window;
1982 x->props.reqid = orig->props.reqid;
1983 x->props.family = orig->props.family;
1984 x->props.saddr = orig->props.saddr;
1985
1986 if (orig->aalg) {
1987 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1988 if (!x->aalg)
1989 goto error;
1990 }
1991 x->props.aalgo = orig->props.aalgo;
1992
1993 if (orig->aead) {
1994 x->aead = xfrm_algo_aead_clone(orig->aead);
1995 x->geniv = orig->geniv;
1996 if (!x->aead)
1997 goto error;
1998 }
1999 if (orig->ealg) {
2000 x->ealg = xfrm_algo_clone(orig->ealg);
2001 if (!x->ealg)
2002 goto error;
2003 }
2004 x->props.ealgo = orig->props.ealgo;
2005
2006 if (orig->calg) {
2007 x->calg = xfrm_algo_clone(orig->calg);
2008 if (!x->calg)
2009 goto error;
2010 }
2011 x->props.calgo = orig->props.calgo;
2012
2013 if (encap || orig->encap) {
2014 if (encap)
2015 x->encap = kmemdup(encap, sizeof(*x->encap),
2016 GFP_KERNEL);
2017 else
2018 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
2019 GFP_KERNEL);
2020
2021 if (!x->encap)
2022 goto error;
2023 }
2024
2025 if (orig->security)
2026 if (clone_security(x, orig->security))
2027 goto error;
2028
2029 if (orig->coaddr) {
2030 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
2031 GFP_KERNEL);
2032 if (!x->coaddr)
2033 goto error;
2034 }
2035
2036 if (orig->replay_esn) {
2037 if (xfrm_replay_clone(x, orig))
2038 goto error;
2039 }
2040
2041 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
2042 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
2043
2044 x->props.flags = orig->props.flags;
2045 x->props.extra_flags = orig->props.extra_flags;
2046
2047 x->pcpu_num = orig->pcpu_num;
2048 x->if_id = orig->if_id;
2049 x->tfcpad = orig->tfcpad;
2050 x->replay_maxdiff = orig->replay_maxdiff;
2051 x->replay_maxage = orig->replay_maxage;
2052 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
2053 x->km.state = orig->km.state;
2054 x->km.seq = orig->km.seq;
2055 x->replay = orig->replay;
2056 x->preplay = orig->preplay;
2057 x->mapping_maxage = orig->mapping_maxage;
2058 x->lastused = orig->lastused;
2059 x->new_mapping = 0;
2060 x->new_mapping_sport = 0;
2061 x->dir = orig->dir;
2062
2063 x->mode_cbs = orig->mode_cbs;
2064 if (x->mode_cbs && x->mode_cbs->clone_state) {
2065 if (x->mode_cbs->clone_state(x, orig))
2066 goto error;
2067 }
2068
2069
2070 x->props.family = m->new_family;
2071 memcpy(&x->id.daddr, &m->new_daddr, sizeof(x->id.daddr));
2072 memcpy(&x->props.saddr, &m->new_saddr, sizeof(x->props.saddr));
2073
2074 return x;
2075
2076 error:
2077 x->km.state = XFRM_STATE_DEAD;
2078 xfrm_state_put(x);
2079 out:
2080 return NULL;
2081 }
2082
xfrm_migrate_state_find(struct xfrm_migrate * m,struct net * net,u32 if_id)2083 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net,
2084 u32 if_id)
2085 {
2086 unsigned int h;
2087 struct xfrm_state *x = NULL;
2088
2089 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2090
2091 if (m->reqid) {
2092 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
2093 m->reqid, m->old_family);
2094 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
2095 if (x->props.mode != m->mode ||
2096 x->id.proto != m->proto)
2097 continue;
2098 if (m->reqid && x->props.reqid != m->reqid)
2099 continue;
2100 if (if_id != 0 && x->if_id != if_id)
2101 continue;
2102 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
2103 m->old_family) ||
2104 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
2105 m->old_family))
2106 continue;
2107 xfrm_state_hold(x);
2108 break;
2109 }
2110 } else {
2111 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
2112 m->old_family);
2113 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
2114 if (x->props.mode != m->mode ||
2115 x->id.proto != m->proto)
2116 continue;
2117 if (if_id != 0 && x->if_id != if_id)
2118 continue;
2119 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
2120 m->old_family) ||
2121 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
2122 m->old_family))
2123 continue;
2124 xfrm_state_hold(x);
2125 break;
2126 }
2127 }
2128
2129 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2130
2131 return x;
2132 }
2133 EXPORT_SYMBOL(xfrm_migrate_state_find);
2134
xfrm_state_migrate(struct xfrm_state * x,struct xfrm_migrate * m,struct xfrm_encap_tmpl * encap,struct net * net,struct xfrm_user_offload * xuo,struct netlink_ext_ack * extack)2135 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
2136 struct xfrm_migrate *m,
2137 struct xfrm_encap_tmpl *encap,
2138 struct net *net,
2139 struct xfrm_user_offload *xuo,
2140 struct netlink_ext_ack *extack)
2141 {
2142 struct xfrm_state *xc;
2143
2144 xc = xfrm_state_clone_and_setup(x, encap, m);
2145 if (!xc)
2146 return NULL;
2147
2148 if (xfrm_init_state(xc) < 0)
2149 goto error;
2150
2151 /* configure the hardware if offload is requested */
2152 if (xuo && xfrm_dev_state_add(net, xc, xuo, extack))
2153 goto error;
2154
2155 /* add state */
2156 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
2157 /* a care is needed when the destination address of the
2158 state is to be updated as it is a part of triplet */
2159 xfrm_state_insert(xc);
2160 } else {
2161 if (xfrm_state_add(xc) < 0)
2162 goto error_add;
2163 }
2164
2165 return xc;
2166 error_add:
2167 if (xuo)
2168 xfrm_dev_state_delete(xc);
2169 error:
2170 xc->km.state = XFRM_STATE_DEAD;
2171 xfrm_state_put(xc);
2172 return NULL;
2173 }
2174 EXPORT_SYMBOL(xfrm_state_migrate);
2175 #endif
2176
xfrm_state_update(struct xfrm_state * x)2177 int xfrm_state_update(struct xfrm_state *x)
2178 {
2179 struct xfrm_state *x1, *to_put;
2180 int err;
2181 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
2182 struct net *net = xs_net(x);
2183
2184 to_put = NULL;
2185
2186 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2187 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
2188
2189 err = -ESRCH;
2190 if (!x1)
2191 goto out;
2192
2193 if (xfrm_state_kern(x1)) {
2194 to_put = x1;
2195 err = -EEXIST;
2196 goto out;
2197 }
2198
2199 if (x1->km.state == XFRM_STATE_ACQ) {
2200 if (x->dir && x1->dir != x->dir) {
2201 to_put = x1;
2202 goto out;
2203 }
2204
2205 __xfrm_state_insert(x);
2206 x = NULL;
2207 } else {
2208 if (x1->dir != x->dir) {
2209 to_put = x1;
2210 goto out;
2211 }
2212 }
2213 err = 0;
2214
2215 out:
2216 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2217
2218 if (to_put)
2219 xfrm_state_put(to_put);
2220
2221 if (err)
2222 return err;
2223
2224 if (!x) {
2225 xfrm_state_delete(x1);
2226 xfrm_state_put(x1);
2227 return 0;
2228 }
2229
2230 err = -EINVAL;
2231 spin_lock_bh(&x1->lock);
2232 if (likely(x1->km.state == XFRM_STATE_VALID)) {
2233 if (x->encap && x1->encap &&
2234 x->encap->encap_type == x1->encap->encap_type)
2235 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
2236 else if (x->encap || x1->encap)
2237 goto fail;
2238
2239 if (x->coaddr && x1->coaddr) {
2240 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
2241 }
2242 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
2243 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
2244 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
2245 x1->km.dying = 0;
2246
2247 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
2248 HRTIMER_MODE_REL_SOFT);
2249 if (READ_ONCE(x1->curlft.use_time))
2250 xfrm_state_check_expire(x1);
2251
2252 if (x->props.smark.m || x->props.smark.v || x->if_id) {
2253 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2254
2255 if (x->props.smark.m || x->props.smark.v)
2256 x1->props.smark = x->props.smark;
2257
2258 if (x->if_id)
2259 x1->if_id = x->if_id;
2260
2261 __xfrm_state_bump_genids(x1);
2262 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2263 }
2264
2265 err = 0;
2266 x->km.state = XFRM_STATE_DEAD;
2267 __xfrm_state_put(x);
2268 }
2269
2270 fail:
2271 spin_unlock_bh(&x1->lock);
2272
2273 xfrm_state_put(x1);
2274
2275 return err;
2276 }
2277 EXPORT_SYMBOL(xfrm_state_update);
2278
xfrm_state_check_expire(struct xfrm_state * x)2279 int xfrm_state_check_expire(struct xfrm_state *x)
2280 {
2281 /* All counters which are needed to decide if state is expired
2282 * are handled by SW for non-packet offload modes. Simply skip
2283 * the following update and save extra boilerplate in drivers.
2284 */
2285 if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
2286 xfrm_dev_state_update_stats(x);
2287
2288 if (!READ_ONCE(x->curlft.use_time))
2289 WRITE_ONCE(x->curlft.use_time, ktime_get_real_seconds());
2290
2291 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
2292 x->curlft.packets >= x->lft.hard_packet_limit) {
2293 x->km.state = XFRM_STATE_EXPIRED;
2294 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
2295 return -EINVAL;
2296 }
2297
2298 if (!x->km.dying &&
2299 (x->curlft.bytes >= x->lft.soft_byte_limit ||
2300 x->curlft.packets >= x->lft.soft_packet_limit)) {
2301 x->km.dying = 1;
2302 km_state_expired(x, 0, 0);
2303 }
2304 return 0;
2305 }
2306 EXPORT_SYMBOL(xfrm_state_check_expire);
2307
xfrm_state_update_stats(struct net * net)2308 void xfrm_state_update_stats(struct net *net)
2309 {
2310 struct xfrm_state *x;
2311 int i;
2312
2313 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2314 for (i = 0; i <= net->xfrm.state_hmask; i++) {
2315 hlist_for_each_entry(x, net->xfrm.state_bydst + i, bydst)
2316 xfrm_dev_state_update_stats(x);
2317 }
2318 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2319 }
2320
2321 struct xfrm_state *
xfrm_state_lookup(struct net * net,u32 mark,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)2322 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
2323 u8 proto, unsigned short family)
2324 {
2325 struct xfrm_hash_state_ptrs state_ptrs;
2326 struct xfrm_state *x;
2327
2328 rcu_read_lock();
2329 xfrm_hash_ptrs_get(net, &state_ptrs);
2330
2331 x = __xfrm_state_lookup(&state_ptrs, mark, daddr, spi, proto, family);
2332 rcu_read_unlock();
2333 return x;
2334 }
2335 EXPORT_SYMBOL(xfrm_state_lookup);
2336
2337 struct xfrm_state *
xfrm_state_lookup_byaddr(struct net * net,u32 mark,const xfrm_address_t * daddr,const xfrm_address_t * saddr,u8 proto,unsigned short family)2338 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
2339 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
2340 u8 proto, unsigned short family)
2341 {
2342 struct xfrm_hash_state_ptrs state_ptrs;
2343 struct xfrm_state *x;
2344
2345 rcu_read_lock();
2346
2347 xfrm_hash_ptrs_get(net, &state_ptrs);
2348
2349 x = __xfrm_state_lookup_byaddr(&state_ptrs, mark, daddr, saddr, proto, family);
2350 rcu_read_unlock();
2351 return x;
2352 }
2353 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
2354
2355 struct xfrm_state *
xfrm_find_acq(struct net * net,const struct xfrm_mark * mark,u8 mode,u32 reqid,u32 if_id,u32 pcpu_num,u8 proto,const xfrm_address_t * daddr,const xfrm_address_t * saddr,int create,unsigned short family)2356 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
2357 u32 if_id, u32 pcpu_num, u8 proto, const xfrm_address_t *daddr,
2358 const xfrm_address_t *saddr, int create, unsigned short family)
2359 {
2360 struct xfrm_state *x;
2361
2362 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2363 x = __find_acq_core(net, mark, family, mode, reqid, if_id, pcpu_num,
2364 proto, daddr, saddr, create);
2365 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2366
2367 return x;
2368 }
2369 EXPORT_SYMBOL(xfrm_find_acq);
2370
2371 #ifdef CONFIG_XFRM_SUB_POLICY
2372 #if IS_ENABLED(CONFIG_IPV6)
2373 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
2374 static void
__xfrm6_sort(void ** dst,void ** src,int n,int (* cmp)(const void * p),int maxclass)2375 __xfrm6_sort(void **dst, void **src, int n,
2376 int (*cmp)(const void *p), int maxclass)
2377 {
2378 int count[XFRM_MAX_DEPTH] = { };
2379 int class[XFRM_MAX_DEPTH];
2380 int i;
2381
2382 for (i = 0; i < n; i++) {
2383 int c = cmp(src[i]);
2384
2385 class[i] = c;
2386 count[c]++;
2387 }
2388
2389 for (i = 2; i < maxclass; i++)
2390 count[i] += count[i - 1];
2391
2392 for (i = 0; i < n; i++) {
2393 dst[count[class[i] - 1]++] = src[i];
2394 src[i] = NULL;
2395 }
2396 }
2397
2398 /* Rule for xfrm_state:
2399 *
2400 * rule 1: select IPsec transport except AH
2401 * rule 2: select MIPv6 RO or inbound trigger
2402 * rule 3: select IPsec transport AH
2403 * rule 4: select IPsec tunnel
2404 * rule 5: others
2405 */
__xfrm6_state_sort_cmp(const void * p)2406 static int __xfrm6_state_sort_cmp(const void *p)
2407 {
2408 const struct xfrm_state *v = p;
2409
2410 switch (v->props.mode) {
2411 case XFRM_MODE_TRANSPORT:
2412 if (v->id.proto != IPPROTO_AH)
2413 return 1;
2414 else
2415 return 3;
2416 #if IS_ENABLED(CONFIG_IPV6_MIP6)
2417 case XFRM_MODE_ROUTEOPTIMIZATION:
2418 case XFRM_MODE_IN_TRIGGER:
2419 return 2;
2420 #endif
2421 case XFRM_MODE_TUNNEL:
2422 case XFRM_MODE_BEET:
2423 case XFRM_MODE_IPTFS:
2424 return 4;
2425 }
2426 return 5;
2427 }
2428
2429 /* Rule for xfrm_tmpl:
2430 *
2431 * rule 1: select IPsec transport
2432 * rule 2: select MIPv6 RO or inbound trigger
2433 * rule 3: select IPsec tunnel
2434 * rule 4: others
2435 */
__xfrm6_tmpl_sort_cmp(const void * p)2436 static int __xfrm6_tmpl_sort_cmp(const void *p)
2437 {
2438 const struct xfrm_tmpl *v = p;
2439
2440 switch (v->mode) {
2441 case XFRM_MODE_TRANSPORT:
2442 return 1;
2443 #if IS_ENABLED(CONFIG_IPV6_MIP6)
2444 case XFRM_MODE_ROUTEOPTIMIZATION:
2445 case XFRM_MODE_IN_TRIGGER:
2446 return 2;
2447 #endif
2448 case XFRM_MODE_TUNNEL:
2449 case XFRM_MODE_BEET:
2450 case XFRM_MODE_IPTFS:
2451 return 3;
2452 }
2453 return 4;
2454 }
2455 #else
__xfrm6_state_sort_cmp(const void * p)2456 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
__xfrm6_tmpl_sort_cmp(const void * p)2457 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
2458
2459 static inline void
__xfrm6_sort(void ** dst,void ** src,int n,int (* cmp)(const void * p),int maxclass)2460 __xfrm6_sort(void **dst, void **src, int n,
2461 int (*cmp)(const void *p), int maxclass)
2462 {
2463 int i;
2464
2465 for (i = 0; i < n; i++)
2466 dst[i] = src[i];
2467 }
2468 #endif /* CONFIG_IPV6 */
2469
2470 void
xfrm_tmpl_sort(struct xfrm_tmpl ** dst,struct xfrm_tmpl ** src,int n,unsigned short family)2471 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
2472 unsigned short family)
2473 {
2474 int i;
2475
2476 if (family == AF_INET6)
2477 __xfrm6_sort((void **)dst, (void **)src, n,
2478 __xfrm6_tmpl_sort_cmp, 5);
2479 else
2480 for (i = 0; i < n; i++)
2481 dst[i] = src[i];
2482 }
2483
2484 void
xfrm_state_sort(struct xfrm_state ** dst,struct xfrm_state ** src,int n,unsigned short family)2485 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
2486 unsigned short family)
2487 {
2488 int i;
2489
2490 if (family == AF_INET6)
2491 __xfrm6_sort((void **)dst, (void **)src, n,
2492 __xfrm6_state_sort_cmp, 6);
2493 else
2494 for (i = 0; i < n; i++)
2495 dst[i] = src[i];
2496 }
2497 #endif
2498
2499 /* Silly enough, but I'm lazy to build resolution list */
2500
__xfrm_find_acq_byseq(struct net * net,u32 mark,u32 seq,u32 pcpu_num)2501 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num)
2502 {
2503 unsigned int h = xfrm_seq_hash(net, seq);
2504 struct xfrm_state *x;
2505
2506 hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
2507 if (x->km.seq == seq &&
2508 (mark & x->mark.m) == x->mark.v &&
2509 x->pcpu_num == pcpu_num &&
2510 x->km.state == XFRM_STATE_ACQ) {
2511 xfrm_state_hold(x);
2512 return x;
2513 }
2514 }
2515
2516 return NULL;
2517 }
2518
xfrm_find_acq_byseq(struct net * net,u32 mark,u32 seq,u32 pcpu_num)2519 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num)
2520 {
2521 struct xfrm_state *x;
2522
2523 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2524 x = __xfrm_find_acq_byseq(net, mark, seq, pcpu_num);
2525 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2526 return x;
2527 }
2528 EXPORT_SYMBOL(xfrm_find_acq_byseq);
2529
xfrm_get_acqseq(void)2530 u32 xfrm_get_acqseq(void)
2531 {
2532 u32 res;
2533 static atomic_t acqseq;
2534
2535 do {
2536 res = atomic_inc_return(&acqseq);
2537 } while (!res);
2538
2539 return res;
2540 }
2541 EXPORT_SYMBOL(xfrm_get_acqseq);
2542
verify_spi_info(u8 proto,u32 min,u32 max,struct netlink_ext_ack * extack)2543 int verify_spi_info(u8 proto, u32 min, u32 max, struct netlink_ext_ack *extack)
2544 {
2545 switch (proto) {
2546 case IPPROTO_AH:
2547 case IPPROTO_ESP:
2548 break;
2549
2550 case IPPROTO_COMP:
2551 /* IPCOMP spi is 16-bits. */
2552 if (max >= 0x10000) {
2553 NL_SET_ERR_MSG(extack, "IPCOMP SPI must be <= 65535");
2554 return -EINVAL;
2555 }
2556 break;
2557
2558 default:
2559 NL_SET_ERR_MSG(extack, "Invalid protocol, must be one of AH, ESP, IPCOMP");
2560 return -EINVAL;
2561 }
2562
2563 if (min > max) {
2564 NL_SET_ERR_MSG(extack, "Invalid SPI range: min > max");
2565 return -EINVAL;
2566 }
2567
2568 return 0;
2569 }
2570 EXPORT_SYMBOL(verify_spi_info);
2571
xfrm_alloc_spi(struct xfrm_state * x,u32 low,u32 high,struct netlink_ext_ack * extack)2572 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high,
2573 struct netlink_ext_ack *extack)
2574 {
2575 struct net *net = xs_net(x);
2576 unsigned int h;
2577 struct xfrm_state *x0;
2578 int err = -ENOENT;
2579 u32 range = high - low + 1;
2580 __be32 newspi = 0;
2581
2582 spin_lock_bh(&x->lock);
2583 if (x->km.state == XFRM_STATE_DEAD) {
2584 NL_SET_ERR_MSG(extack, "Target ACQUIRE is in DEAD state");
2585 goto unlock;
2586 }
2587
2588 err = 0;
2589 if (x->id.spi)
2590 goto unlock;
2591
2592 err = -ENOENT;
2593
2594 for (h = 0; h < range; h++) {
2595 u32 spi = (low == high) ? low : get_random_u32_inclusive(low, high);
2596 if (spi == 0)
2597 goto next;
2598 newspi = htonl(spi);
2599
2600 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2601 x0 = xfrm_state_lookup_spi_proto(net, newspi, x->id.proto);
2602 if (!x0) {
2603 x->id.spi = newspi;
2604 h = xfrm_spi_hash(net, &x->id.daddr, newspi, x->id.proto, x->props.family);
2605 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h, x->xso.type);
2606 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2607 err = 0;
2608 goto unlock;
2609 }
2610 xfrm_state_put(x0);
2611 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2612
2613 next:
2614 if (signal_pending(current)) {
2615 err = -ERESTARTSYS;
2616 goto unlock;
2617 }
2618
2619 if (low == high)
2620 break;
2621 }
2622
2623 if (err)
2624 NL_SET_ERR_MSG(extack, "No SPI available in the requested range");
2625
2626 unlock:
2627 spin_unlock_bh(&x->lock);
2628
2629 return err;
2630 }
2631 EXPORT_SYMBOL(xfrm_alloc_spi);
2632
__xfrm_state_filter_match(struct xfrm_state * x,struct xfrm_address_filter * filter)2633 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2634 struct xfrm_address_filter *filter)
2635 {
2636 if (filter) {
2637 if ((filter->family == AF_INET ||
2638 filter->family == AF_INET6) &&
2639 x->props.family != filter->family)
2640 return false;
2641
2642 return addr_match(&x->props.saddr, &filter->saddr,
2643 filter->splen) &&
2644 addr_match(&x->id.daddr, &filter->daddr,
2645 filter->dplen);
2646 }
2647 return true;
2648 }
2649
xfrm_state_walk(struct net * net,struct xfrm_state_walk * walk,int (* func)(struct xfrm_state *,int,void *),void * data)2650 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2651 int (*func)(struct xfrm_state *, int, void*),
2652 void *data)
2653 {
2654 struct xfrm_state *state;
2655 struct xfrm_state_walk *x;
2656 int err = 0;
2657
2658 if (walk->seq != 0 && list_empty(&walk->all))
2659 return 0;
2660
2661 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2662 if (list_empty(&walk->all))
2663 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2664 else
2665 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2666 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2667 if (x->state == XFRM_STATE_DEAD)
2668 continue;
2669 state = container_of(x, struct xfrm_state, km);
2670 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2671 continue;
2672 if (!__xfrm_state_filter_match(state, walk->filter))
2673 continue;
2674 err = func(state, walk->seq, data);
2675 if (err) {
2676 list_move_tail(&walk->all, &x->all);
2677 goto out;
2678 }
2679 walk->seq++;
2680 }
2681 if (walk->seq == 0) {
2682 err = -ENOENT;
2683 goto out;
2684 }
2685 list_del_init(&walk->all);
2686 out:
2687 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2688 return err;
2689 }
2690 EXPORT_SYMBOL(xfrm_state_walk);
2691
xfrm_state_walk_init(struct xfrm_state_walk * walk,u8 proto,struct xfrm_address_filter * filter)2692 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2693 struct xfrm_address_filter *filter)
2694 {
2695 INIT_LIST_HEAD(&walk->all);
2696 walk->proto = proto;
2697 walk->state = XFRM_STATE_DEAD;
2698 walk->seq = 0;
2699 walk->filter = filter;
2700 }
2701 EXPORT_SYMBOL(xfrm_state_walk_init);
2702
xfrm_state_walk_done(struct xfrm_state_walk * walk,struct net * net)2703 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2704 {
2705 kfree(walk->filter);
2706
2707 if (list_empty(&walk->all))
2708 return;
2709
2710 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2711 list_del(&walk->all);
2712 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2713 }
2714 EXPORT_SYMBOL(xfrm_state_walk_done);
2715
xfrm_replay_timer_handler(struct timer_list * t)2716 static void xfrm_replay_timer_handler(struct timer_list *t)
2717 {
2718 struct xfrm_state *x = timer_container_of(x, t, rtimer);
2719
2720 spin_lock(&x->lock);
2721
2722 if (x->km.state == XFRM_STATE_VALID) {
2723 if (xfrm_aevent_is_on(xs_net(x)))
2724 xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
2725 else
2726 x->xflags |= XFRM_TIME_DEFER;
2727 }
2728
2729 spin_unlock(&x->lock);
2730 }
2731
2732 static LIST_HEAD(xfrm_km_list);
2733
km_policy_notify(struct xfrm_policy * xp,int dir,const struct km_event * c)2734 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2735 {
2736 struct xfrm_mgr *km;
2737
2738 rcu_read_lock();
2739 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2740 if (km->notify_policy)
2741 km->notify_policy(xp, dir, c);
2742 rcu_read_unlock();
2743 }
2744
km_state_notify(struct xfrm_state * x,const struct km_event * c)2745 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2746 {
2747 struct xfrm_mgr *km;
2748 rcu_read_lock();
2749 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2750 if (km->notify)
2751 km->notify(x, c);
2752 rcu_read_unlock();
2753 }
2754
2755 EXPORT_SYMBOL(km_policy_notify);
2756 EXPORT_SYMBOL(km_state_notify);
2757
km_state_expired(struct xfrm_state * x,int hard,u32 portid)2758 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2759 {
2760 struct km_event c;
2761
2762 c.data.hard = hard;
2763 c.portid = portid;
2764 c.event = XFRM_MSG_EXPIRE;
2765 km_state_notify(x, &c);
2766 }
2767
2768 EXPORT_SYMBOL(km_state_expired);
2769 /*
2770 * We send to all registered managers regardless of failure
2771 * We are happy with one success
2772 */
km_query(struct xfrm_state * x,struct xfrm_tmpl * t,struct xfrm_policy * pol)2773 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2774 {
2775 int err = -EINVAL, acqret;
2776 struct xfrm_mgr *km;
2777
2778 rcu_read_lock();
2779 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2780 acqret = km->acquire(x, t, pol);
2781 if (!acqret)
2782 err = acqret;
2783 }
2784 rcu_read_unlock();
2785 return err;
2786 }
2787 EXPORT_SYMBOL(km_query);
2788
__km_new_mapping(struct xfrm_state * x,xfrm_address_t * ipaddr,__be16 sport)2789 static int __km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2790 {
2791 int err = -EINVAL;
2792 struct xfrm_mgr *km;
2793
2794 rcu_read_lock();
2795 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2796 if (km->new_mapping)
2797 err = km->new_mapping(x, ipaddr, sport);
2798 if (!err)
2799 break;
2800 }
2801 rcu_read_unlock();
2802 return err;
2803 }
2804
km_new_mapping(struct xfrm_state * x,xfrm_address_t * ipaddr,__be16 sport)2805 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2806 {
2807 int ret = 0;
2808
2809 if (x->mapping_maxage) {
2810 if ((jiffies / HZ - x->new_mapping) > x->mapping_maxage ||
2811 x->new_mapping_sport != sport) {
2812 x->new_mapping_sport = sport;
2813 x->new_mapping = jiffies / HZ;
2814 ret = __km_new_mapping(x, ipaddr, sport);
2815 }
2816 } else {
2817 ret = __km_new_mapping(x, ipaddr, sport);
2818 }
2819
2820 return ret;
2821 }
2822 EXPORT_SYMBOL(km_new_mapping);
2823
km_policy_expired(struct xfrm_policy * pol,int dir,int hard,u32 portid)2824 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2825 {
2826 struct km_event c;
2827
2828 c.data.hard = hard;
2829 c.portid = portid;
2830 c.event = XFRM_MSG_POLEXPIRE;
2831 km_policy_notify(pol, dir, &c);
2832 }
2833 EXPORT_SYMBOL(km_policy_expired);
2834
2835 #ifdef CONFIG_XFRM_MIGRATE
km_migrate(const struct xfrm_selector * sel,u8 dir,u8 type,const struct xfrm_migrate * m,int num_migrate,const struct xfrm_kmaddress * k,const struct xfrm_encap_tmpl * encap)2836 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2837 const struct xfrm_migrate *m, int num_migrate,
2838 const struct xfrm_kmaddress *k,
2839 const struct xfrm_encap_tmpl *encap)
2840 {
2841 int err = -EINVAL;
2842 int ret;
2843 struct xfrm_mgr *km;
2844
2845 rcu_read_lock();
2846 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2847 if (km->migrate) {
2848 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2849 encap);
2850 if (!ret)
2851 err = ret;
2852 }
2853 }
2854 rcu_read_unlock();
2855 return err;
2856 }
2857 EXPORT_SYMBOL(km_migrate);
2858 #endif
2859
km_report(struct net * net,u8 proto,struct xfrm_selector * sel,xfrm_address_t * addr)2860 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2861 {
2862 int err = -EINVAL;
2863 int ret;
2864 struct xfrm_mgr *km;
2865
2866 rcu_read_lock();
2867 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2868 if (km->report) {
2869 ret = km->report(net, proto, sel, addr);
2870 if (!ret)
2871 err = ret;
2872 }
2873 }
2874 rcu_read_unlock();
2875 return err;
2876 }
2877 EXPORT_SYMBOL(km_report);
2878
km_is_alive(const struct km_event * c)2879 static bool km_is_alive(const struct km_event *c)
2880 {
2881 struct xfrm_mgr *km;
2882 bool is_alive = false;
2883
2884 rcu_read_lock();
2885 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2886 if (km->is_alive && km->is_alive(c)) {
2887 is_alive = true;
2888 break;
2889 }
2890 }
2891 rcu_read_unlock();
2892
2893 return is_alive;
2894 }
2895
2896 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2897 static DEFINE_SPINLOCK(xfrm_translator_lock);
2898 static struct xfrm_translator __rcu *xfrm_translator;
2899
xfrm_get_translator(void)2900 struct xfrm_translator *xfrm_get_translator(void)
2901 {
2902 struct xfrm_translator *xtr;
2903
2904 rcu_read_lock();
2905 xtr = rcu_dereference(xfrm_translator);
2906 if (unlikely(!xtr))
2907 goto out;
2908 if (!try_module_get(xtr->owner))
2909 xtr = NULL;
2910 out:
2911 rcu_read_unlock();
2912 return xtr;
2913 }
2914 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2915
xfrm_put_translator(struct xfrm_translator * xtr)2916 void xfrm_put_translator(struct xfrm_translator *xtr)
2917 {
2918 module_put(xtr->owner);
2919 }
2920 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2921
xfrm_register_translator(struct xfrm_translator * xtr)2922 int xfrm_register_translator(struct xfrm_translator *xtr)
2923 {
2924 int err = 0;
2925
2926 spin_lock_bh(&xfrm_translator_lock);
2927 if (unlikely(xfrm_translator != NULL))
2928 err = -EEXIST;
2929 else
2930 rcu_assign_pointer(xfrm_translator, xtr);
2931 spin_unlock_bh(&xfrm_translator_lock);
2932
2933 return err;
2934 }
2935 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2936
xfrm_unregister_translator(struct xfrm_translator * xtr)2937 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2938 {
2939 int err = 0;
2940
2941 spin_lock_bh(&xfrm_translator_lock);
2942 if (likely(xfrm_translator != NULL)) {
2943 if (rcu_access_pointer(xfrm_translator) != xtr)
2944 err = -EINVAL;
2945 else
2946 RCU_INIT_POINTER(xfrm_translator, NULL);
2947 }
2948 spin_unlock_bh(&xfrm_translator_lock);
2949 synchronize_rcu();
2950
2951 return err;
2952 }
2953 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2954 #endif
2955
xfrm_user_policy(struct sock * sk,int optname,sockptr_t optval,int optlen)2956 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2957 {
2958 int err;
2959 u8 *data;
2960 struct xfrm_mgr *km;
2961 struct xfrm_policy *pol = NULL;
2962
2963 if (sockptr_is_null(optval) && !optlen) {
2964 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2965 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2966 __sk_dst_reset(sk);
2967 return 0;
2968 }
2969
2970 if (optlen <= 0 || optlen > PAGE_SIZE)
2971 return -EMSGSIZE;
2972
2973 data = memdup_sockptr(optval, optlen);
2974 if (IS_ERR(data))
2975 return PTR_ERR(data);
2976
2977 if (in_compat_syscall()) {
2978 struct xfrm_translator *xtr = xfrm_get_translator();
2979
2980 if (!xtr) {
2981 kfree(data);
2982 return -EOPNOTSUPP;
2983 }
2984
2985 err = xtr->xlate_user_policy_sockptr(&data, optlen);
2986 xfrm_put_translator(xtr);
2987 if (err) {
2988 kfree(data);
2989 return err;
2990 }
2991 }
2992
2993 err = -EINVAL;
2994 rcu_read_lock();
2995 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2996 pol = km->compile_policy(sk, optname, data,
2997 optlen, &err);
2998 if (err >= 0)
2999 break;
3000 }
3001 rcu_read_unlock();
3002
3003 if (err >= 0) {
3004 xfrm_sk_policy_insert(sk, err, pol);
3005 xfrm_pol_put(pol);
3006 __sk_dst_reset(sk);
3007 err = 0;
3008 }
3009
3010 kfree(data);
3011 return err;
3012 }
3013 EXPORT_SYMBOL(xfrm_user_policy);
3014
3015 static DEFINE_SPINLOCK(xfrm_km_lock);
3016
xfrm_register_km(struct xfrm_mgr * km)3017 void xfrm_register_km(struct xfrm_mgr *km)
3018 {
3019 spin_lock_bh(&xfrm_km_lock);
3020 list_add_tail_rcu(&km->list, &xfrm_km_list);
3021 spin_unlock_bh(&xfrm_km_lock);
3022 }
3023 EXPORT_SYMBOL(xfrm_register_km);
3024
xfrm_unregister_km(struct xfrm_mgr * km)3025 void xfrm_unregister_km(struct xfrm_mgr *km)
3026 {
3027 spin_lock_bh(&xfrm_km_lock);
3028 list_del_rcu(&km->list);
3029 spin_unlock_bh(&xfrm_km_lock);
3030 synchronize_rcu();
3031 }
3032 EXPORT_SYMBOL(xfrm_unregister_km);
3033
xfrm_state_register_afinfo(struct xfrm_state_afinfo * afinfo)3034 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
3035 {
3036 int err = 0;
3037
3038 if (WARN_ON(afinfo->family >= NPROTO))
3039 return -EAFNOSUPPORT;
3040
3041 spin_lock_bh(&xfrm_state_afinfo_lock);
3042 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
3043 err = -EEXIST;
3044 else
3045 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
3046 spin_unlock_bh(&xfrm_state_afinfo_lock);
3047 return err;
3048 }
3049 EXPORT_SYMBOL(xfrm_state_register_afinfo);
3050
xfrm_state_unregister_afinfo(struct xfrm_state_afinfo * afinfo)3051 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
3052 {
3053 int err = 0, family = afinfo->family;
3054
3055 if (WARN_ON(family >= NPROTO))
3056 return -EAFNOSUPPORT;
3057
3058 spin_lock_bh(&xfrm_state_afinfo_lock);
3059 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
3060 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
3061 err = -EINVAL;
3062 else
3063 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
3064 }
3065 spin_unlock_bh(&xfrm_state_afinfo_lock);
3066 synchronize_rcu();
3067 return err;
3068 }
3069 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
3070
xfrm_state_afinfo_get_rcu(unsigned int family)3071 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
3072 {
3073 if (unlikely(family >= NPROTO))
3074 return NULL;
3075
3076 return rcu_dereference(xfrm_state_afinfo[family]);
3077 }
3078 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
3079
xfrm_state_get_afinfo(unsigned int family)3080 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
3081 {
3082 struct xfrm_state_afinfo *afinfo;
3083 if (unlikely(family >= NPROTO))
3084 return NULL;
3085 rcu_read_lock();
3086 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
3087 if (unlikely(!afinfo))
3088 rcu_read_unlock();
3089 return afinfo;
3090 }
3091
xfrm_flush_gc(void)3092 void xfrm_flush_gc(void)
3093 {
3094 flush_work(&xfrm_state_gc_work);
3095 }
3096 EXPORT_SYMBOL(xfrm_flush_gc);
3097
xfrm_state_delete_tunnel(struct xfrm_state * x)3098 static void xfrm_state_delete_tunnel(struct xfrm_state *x)
3099 {
3100 if (x->tunnel) {
3101 struct xfrm_state *t = x->tunnel;
3102
3103 if (atomic_dec_return(&t->tunnel_users) == 1)
3104 xfrm_state_delete(t);
3105 xfrm_state_put(t);
3106 x->tunnel = NULL;
3107 }
3108 }
3109
xfrm_state_mtu(struct xfrm_state * x,int mtu)3110 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
3111 {
3112 const struct xfrm_type *type = READ_ONCE(x->type);
3113 struct crypto_aead *aead;
3114 u32 blksize, net_adj = 0;
3115
3116 if (x->km.state != XFRM_STATE_VALID ||
3117 !type || type->proto != IPPROTO_ESP)
3118 return mtu - x->props.header_len;
3119
3120 aead = x->data;
3121 blksize = ALIGN(crypto_aead_blocksize(aead), 4);
3122
3123 switch (x->props.mode) {
3124 case XFRM_MODE_TRANSPORT:
3125 case XFRM_MODE_BEET:
3126 if (x->props.family == AF_INET)
3127 net_adj = sizeof(struct iphdr);
3128 else if (x->props.family == AF_INET6)
3129 net_adj = sizeof(struct ipv6hdr);
3130 break;
3131 case XFRM_MODE_TUNNEL:
3132 break;
3133 default:
3134 if (x->mode_cbs && x->mode_cbs->get_inner_mtu)
3135 return x->mode_cbs->get_inner_mtu(x, mtu);
3136
3137 WARN_ON_ONCE(1);
3138 break;
3139 }
3140
3141 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
3142 net_adj) & ~(blksize - 1)) + net_adj - 2;
3143 }
3144 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
3145
__xfrm_init_state(struct xfrm_state * x,struct netlink_ext_ack * extack)3146 int __xfrm_init_state(struct xfrm_state *x, struct netlink_ext_ack *extack)
3147 {
3148 const struct xfrm_mode *inner_mode;
3149 const struct xfrm_mode *outer_mode;
3150 int family = x->props.family;
3151 int err;
3152
3153 if (family == AF_INET &&
3154 READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc))
3155 x->props.flags |= XFRM_STATE_NOPMTUDISC;
3156
3157 err = -EPROTONOSUPPORT;
3158
3159 if (x->sel.family != AF_UNSPEC) {
3160 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
3161 if (inner_mode == NULL) {
3162 NL_SET_ERR_MSG(extack, "Requested mode not found");
3163 goto error;
3164 }
3165
3166 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
3167 family != x->sel.family) {
3168 NL_SET_ERR_MSG(extack, "Only tunnel modes can accommodate a change of family");
3169 goto error;
3170 }
3171
3172 x->inner_mode = *inner_mode;
3173 } else {
3174 const struct xfrm_mode *inner_mode_iaf;
3175 int iafamily = AF_INET;
3176
3177 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
3178 if (inner_mode == NULL) {
3179 NL_SET_ERR_MSG(extack, "Requested mode not found");
3180 goto error;
3181 }
3182
3183 x->inner_mode = *inner_mode;
3184
3185 if (x->props.family == AF_INET)
3186 iafamily = AF_INET6;
3187
3188 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
3189 if (inner_mode_iaf) {
3190 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
3191 x->inner_mode_iaf = *inner_mode_iaf;
3192 }
3193 }
3194
3195 x->type = xfrm_get_type(x->id.proto, family);
3196 if (x->type == NULL) {
3197 NL_SET_ERR_MSG(extack, "Requested type not found");
3198 goto error;
3199 }
3200
3201 err = x->type->init_state(x, extack);
3202 if (err)
3203 goto error;
3204
3205 outer_mode = xfrm_get_mode(x->props.mode, family);
3206 if (!outer_mode) {
3207 NL_SET_ERR_MSG(extack, "Requested mode not found");
3208 err = -EPROTONOSUPPORT;
3209 goto error;
3210 }
3211
3212 x->outer_mode = *outer_mode;
3213 if (x->nat_keepalive_interval) {
3214 if (x->dir != XFRM_SA_DIR_OUT) {
3215 NL_SET_ERR_MSG(extack, "NAT keepalive is only supported for outbound SAs");
3216 err = -EINVAL;
3217 goto error;
3218 }
3219
3220 if (!x->encap || x->encap->encap_type != UDP_ENCAP_ESPINUDP) {
3221 NL_SET_ERR_MSG(extack,
3222 "NAT keepalive is only supported for UDP encapsulation");
3223 err = -EINVAL;
3224 goto error;
3225 }
3226 }
3227
3228 x->mode_cbs = xfrm_get_mode_cbs(x->props.mode);
3229 if (x->mode_cbs) {
3230 if (x->mode_cbs->init_state)
3231 err = x->mode_cbs->init_state(x);
3232 module_put(x->mode_cbs->owner);
3233 }
3234 error:
3235 return err;
3236 }
3237
3238 EXPORT_SYMBOL(__xfrm_init_state);
3239
xfrm_init_state(struct xfrm_state * x)3240 int xfrm_init_state(struct xfrm_state *x)
3241 {
3242 int err;
3243
3244 err = __xfrm_init_state(x, NULL);
3245 if (err)
3246 return err;
3247
3248 err = xfrm_init_replay(x, NULL);
3249 if (err)
3250 return err;
3251
3252 x->km.state = XFRM_STATE_VALID;
3253 return 0;
3254 }
3255
3256 EXPORT_SYMBOL(xfrm_init_state);
3257
xfrm_state_init(struct net * net)3258 int __net_init xfrm_state_init(struct net *net)
3259 {
3260 unsigned int sz;
3261
3262 if (net_eq(net, &init_net))
3263 xfrm_state_cache = KMEM_CACHE(xfrm_state,
3264 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3265
3266 INIT_LIST_HEAD(&net->xfrm.state_all);
3267
3268 sz = sizeof(struct hlist_head) * 8;
3269
3270 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
3271 if (!net->xfrm.state_bydst)
3272 goto out_bydst;
3273 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
3274 if (!net->xfrm.state_bysrc)
3275 goto out_bysrc;
3276 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
3277 if (!net->xfrm.state_byspi)
3278 goto out_byspi;
3279 net->xfrm.state_byseq = xfrm_hash_alloc(sz);
3280 if (!net->xfrm.state_byseq)
3281 goto out_byseq;
3282
3283 net->xfrm.state_cache_input = alloc_percpu(struct hlist_head);
3284 if (!net->xfrm.state_cache_input)
3285 goto out_state_cache_input;
3286
3287 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
3288
3289 net->xfrm.state_num = 0;
3290 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
3291 spin_lock_init(&net->xfrm.xfrm_state_lock);
3292 seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
3293 &net->xfrm.xfrm_state_lock);
3294 return 0;
3295
3296 out_state_cache_input:
3297 xfrm_hash_free(net->xfrm.state_byseq, sz);
3298 out_byseq:
3299 xfrm_hash_free(net->xfrm.state_byspi, sz);
3300 out_byspi:
3301 xfrm_hash_free(net->xfrm.state_bysrc, sz);
3302 out_bysrc:
3303 xfrm_hash_free(net->xfrm.state_bydst, sz);
3304 out_bydst:
3305 return -ENOMEM;
3306 }
3307
xfrm_state_fini(struct net * net)3308 void xfrm_state_fini(struct net *net)
3309 {
3310 unsigned int sz;
3311 int i;
3312
3313 flush_work(&net->xfrm.state_hash_work);
3314 xfrm_state_flush(net, 0, false);
3315 flush_work(&xfrm_state_gc_work);
3316
3317 WARN_ON(!list_empty(&net->xfrm.state_all));
3318
3319 for (i = 0; i <= net->xfrm.state_hmask; i++) {
3320 WARN_ON(!hlist_empty(net->xfrm.state_byseq + i));
3321 WARN_ON(!hlist_empty(net->xfrm.state_byspi + i));
3322 WARN_ON(!hlist_empty(net->xfrm.state_bysrc + i));
3323 WARN_ON(!hlist_empty(net->xfrm.state_bydst + i));
3324 }
3325
3326 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
3327 xfrm_hash_free(net->xfrm.state_byseq, sz);
3328 xfrm_hash_free(net->xfrm.state_byspi, sz);
3329 xfrm_hash_free(net->xfrm.state_bysrc, sz);
3330 xfrm_hash_free(net->xfrm.state_bydst, sz);
3331 free_percpu(net->xfrm.state_cache_input);
3332 }
3333
3334 #ifdef CONFIG_AUDITSYSCALL
xfrm_audit_helper_sainfo(struct xfrm_state * x,struct audit_buffer * audit_buf)3335 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
3336 struct audit_buffer *audit_buf)
3337 {
3338 struct xfrm_sec_ctx *ctx = x->security;
3339 u32 spi = ntohl(x->id.spi);
3340
3341 if (ctx)
3342 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
3343 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
3344
3345 switch (x->props.family) {
3346 case AF_INET:
3347 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
3348 &x->props.saddr.a4, &x->id.daddr.a4);
3349 break;
3350 case AF_INET6:
3351 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
3352 x->props.saddr.a6, x->id.daddr.a6);
3353 break;
3354 }
3355
3356 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
3357 }
3358
xfrm_audit_helper_pktinfo(struct sk_buff * skb,u16 family,struct audit_buffer * audit_buf)3359 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
3360 struct audit_buffer *audit_buf)
3361 {
3362 const struct iphdr *iph4;
3363 const struct ipv6hdr *iph6;
3364
3365 switch (family) {
3366 case AF_INET:
3367 iph4 = ip_hdr(skb);
3368 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
3369 &iph4->saddr, &iph4->daddr);
3370 break;
3371 case AF_INET6:
3372 iph6 = ipv6_hdr(skb);
3373 audit_log_format(audit_buf,
3374 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
3375 &iph6->saddr, &iph6->daddr,
3376 iph6->flow_lbl[0] & 0x0f,
3377 iph6->flow_lbl[1],
3378 iph6->flow_lbl[2]);
3379 break;
3380 }
3381 }
3382
xfrm_audit_state_add(struct xfrm_state * x,int result,bool task_valid)3383 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
3384 {
3385 struct audit_buffer *audit_buf;
3386
3387 audit_buf = xfrm_audit_start("SAD-add");
3388 if (audit_buf == NULL)
3389 return;
3390 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
3391 xfrm_audit_helper_sainfo(x, audit_buf);
3392 audit_log_format(audit_buf, " res=%u", result);
3393 audit_log_end(audit_buf);
3394 }
3395 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
3396
xfrm_audit_state_delete(struct xfrm_state * x,int result,bool task_valid)3397 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
3398 {
3399 struct audit_buffer *audit_buf;
3400
3401 audit_buf = xfrm_audit_start("SAD-delete");
3402 if (audit_buf == NULL)
3403 return;
3404 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
3405 xfrm_audit_helper_sainfo(x, audit_buf);
3406 audit_log_format(audit_buf, " res=%u", result);
3407 audit_log_end(audit_buf);
3408 }
3409 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
3410
xfrm_audit_state_replay_overflow(struct xfrm_state * x,struct sk_buff * skb)3411 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
3412 struct sk_buff *skb)
3413 {
3414 struct audit_buffer *audit_buf;
3415 u32 spi;
3416
3417 audit_buf = xfrm_audit_start("SA-replay-overflow");
3418 if (audit_buf == NULL)
3419 return;
3420 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3421 /* don't record the sequence number because it's inherent in this kind
3422 * of audit message */
3423 spi = ntohl(x->id.spi);
3424 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
3425 audit_log_end(audit_buf);
3426 }
3427 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
3428
xfrm_audit_state_replay(struct xfrm_state * x,struct sk_buff * skb,__be32 net_seq)3429 void xfrm_audit_state_replay(struct xfrm_state *x,
3430 struct sk_buff *skb, __be32 net_seq)
3431 {
3432 struct audit_buffer *audit_buf;
3433 u32 spi;
3434
3435 audit_buf = xfrm_audit_start("SA-replayed-pkt");
3436 if (audit_buf == NULL)
3437 return;
3438 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3439 spi = ntohl(x->id.spi);
3440 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3441 spi, spi, ntohl(net_seq));
3442 audit_log_end(audit_buf);
3443 }
3444 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
3445
xfrm_audit_state_notfound_simple(struct sk_buff * skb,u16 family)3446 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
3447 {
3448 struct audit_buffer *audit_buf;
3449
3450 audit_buf = xfrm_audit_start("SA-notfound");
3451 if (audit_buf == NULL)
3452 return;
3453 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3454 audit_log_end(audit_buf);
3455 }
3456 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
3457
xfrm_audit_state_notfound(struct sk_buff * skb,u16 family,__be32 net_spi,__be32 net_seq)3458 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
3459 __be32 net_spi, __be32 net_seq)
3460 {
3461 struct audit_buffer *audit_buf;
3462 u32 spi;
3463
3464 audit_buf = xfrm_audit_start("SA-notfound");
3465 if (audit_buf == NULL)
3466 return;
3467 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3468 spi = ntohl(net_spi);
3469 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3470 spi, spi, ntohl(net_seq));
3471 audit_log_end(audit_buf);
3472 }
3473 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
3474
xfrm_audit_state_icvfail(struct xfrm_state * x,struct sk_buff * skb,u8 proto)3475 void xfrm_audit_state_icvfail(struct xfrm_state *x,
3476 struct sk_buff *skb, u8 proto)
3477 {
3478 struct audit_buffer *audit_buf;
3479 __be32 net_spi;
3480 __be32 net_seq;
3481
3482 audit_buf = xfrm_audit_start("SA-icv-failure");
3483 if (audit_buf == NULL)
3484 return;
3485 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3486 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
3487 u32 spi = ntohl(net_spi);
3488 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3489 spi, spi, ntohl(net_seq));
3490 }
3491 audit_log_end(audit_buf);
3492 }
3493 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
3494 #endif /* CONFIG_AUDITSYSCALL */
3495