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