xref: /openbmc/linux/security/selinux/ss/sidtab.c (revision bc33f5e5)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Implementation of the SID table type.
4  *
5  * Original author: Stephen Smalley, <sds@tycho.nsa.gov>
6  * Author: Ondrej Mosnacek, <omosnacek@gmail.com>
7  *
8  * Copyright (C) 2018 Red Hat, Inc.
9  */
10 #include <linux/errno.h>
11 #include <linux/kernel.h>
12 #include <linux/list.h>
13 #include <linux/rcupdate.h>
14 #include <linux/slab.h>
15 #include <linux/sched.h>
16 #include <linux/spinlock.h>
17 #include <asm/barrier.h>
18 #include "flask.h"
19 #include "security.h"
20 #include "sidtab.h"
21 
22 struct sidtab_str_cache {
23 	struct rcu_head rcu_member;
24 	struct list_head lru_member;
25 	struct sidtab_entry *parent;
26 	u32 len;
27 	char str[];
28 };
29 
30 #define index_to_sid(index) ((index) + SECINITSID_NUM + 1)
31 #define sid_to_index(sid) ((sid) - (SECINITSID_NUM + 1))
32 
33 int sidtab_init(struct sidtab *s)
34 {
35 	u32 i;
36 
37 	memset(s->roots, 0, sizeof(s->roots));
38 
39 	for (i = 0; i < SECINITSID_NUM; i++)
40 		s->isids[i].set = 0;
41 
42 	s->frozen = false;
43 	s->count = 0;
44 	s->convert = NULL;
45 	hash_init(s->context_to_sid);
46 
47 	spin_lock_init(&s->lock);
48 
49 #if CONFIG_SECURITY_SELINUX_SID2STR_CACHE_SIZE > 0
50 	s->cache_free_slots = CONFIG_SECURITY_SELINUX_SID2STR_CACHE_SIZE;
51 	INIT_LIST_HEAD(&s->cache_lru_list);
52 	spin_lock_init(&s->cache_lock);
53 #endif
54 
55 	return 0;
56 }
57 
58 static u32 context_to_sid(struct sidtab *s, struct context *context, u32 hash)
59 {
60 	struct sidtab_entry *entry;
61 	u32 sid = 0;
62 
63 	rcu_read_lock();
64 	hash_for_each_possible_rcu(s->context_to_sid, entry, list, hash) {
65 		if (entry->hash != hash)
66 			continue;
67 		if (context_cmp(&entry->context, context)) {
68 			sid = entry->sid;
69 			break;
70 		}
71 	}
72 	rcu_read_unlock();
73 	return sid;
74 }
75 
76 int sidtab_set_initial(struct sidtab *s, u32 sid, struct context *context)
77 {
78 	struct sidtab_isid_entry *isid;
79 	u32 hash;
80 	int rc;
81 
82 	if (sid == 0 || sid > SECINITSID_NUM)
83 		return -EINVAL;
84 
85 	isid = &s->isids[sid - 1];
86 
87 	rc = context_cpy(&isid->entry.context, context);
88 	if (rc)
89 		return rc;
90 
91 #if CONFIG_SECURITY_SELINUX_SID2STR_CACHE_SIZE > 0
92 	isid->entry.cache = NULL;
93 #endif
94 	isid->set = 1;
95 
96 	hash = context_compute_hash(context);
97 
98 	/*
99 	 * Multiple initial sids may map to the same context. Check that this
100 	 * context is not already represented in the context_to_sid hashtable
101 	 * to avoid duplicate entries and long linked lists upon hash
102 	 * collision.
103 	 */
104 	if (!context_to_sid(s, context, hash)) {
105 		isid->entry.sid = sid;
106 		isid->entry.hash = hash;
107 		hash_add(s->context_to_sid, &isid->entry.list, hash);
108 	}
109 
110 	return 0;
111 }
112 
113 int sidtab_hash_stats(struct sidtab *sidtab, char *page)
114 {
115 	int i;
116 	int chain_len = 0;
117 	int slots_used = 0;
118 	int entries = 0;
119 	int max_chain_len = 0;
120 	int cur_bucket = 0;
121 	struct sidtab_entry *entry;
122 
123 	rcu_read_lock();
124 	hash_for_each_rcu(sidtab->context_to_sid, i, entry, list) {
125 		entries++;
126 		if (i == cur_bucket) {
127 			chain_len++;
128 			if (chain_len == 1)
129 				slots_used++;
130 		} else {
131 			cur_bucket = i;
132 			if (chain_len > max_chain_len)
133 				max_chain_len = chain_len;
134 			chain_len = 0;
135 		}
136 	}
137 	rcu_read_unlock();
138 
139 	if (chain_len > max_chain_len)
140 		max_chain_len = chain_len;
141 
142 	return scnprintf(page, PAGE_SIZE, "entries: %d\nbuckets used: %d/%d\n"
143 			 "longest chain: %d\n", entries,
144 			 slots_used, SIDTAB_HASH_BUCKETS, max_chain_len);
145 }
146 
147 static u32 sidtab_level_from_count(u32 count)
148 {
149 	u32 capacity = SIDTAB_LEAF_ENTRIES;
150 	u32 level = 0;
151 
152 	while (count > capacity) {
153 		capacity <<= SIDTAB_INNER_SHIFT;
154 		++level;
155 	}
156 	return level;
157 }
158 
159 static int sidtab_alloc_roots(struct sidtab *s, u32 level)
160 {
161 	u32 l;
162 
163 	if (!s->roots[0].ptr_leaf) {
164 		s->roots[0].ptr_leaf = kzalloc(SIDTAB_NODE_ALLOC_SIZE,
165 					       GFP_ATOMIC);
166 		if (!s->roots[0].ptr_leaf)
167 			return -ENOMEM;
168 	}
169 	for (l = 1; l <= level; ++l)
170 		if (!s->roots[l].ptr_inner) {
171 			s->roots[l].ptr_inner = kzalloc(SIDTAB_NODE_ALLOC_SIZE,
172 							GFP_ATOMIC);
173 			if (!s->roots[l].ptr_inner)
174 				return -ENOMEM;
175 			s->roots[l].ptr_inner->entries[0] = s->roots[l - 1];
176 		}
177 	return 0;
178 }
179 
180 static struct sidtab_entry *sidtab_do_lookup(struct sidtab *s, u32 index,
181 					     int alloc)
182 {
183 	union sidtab_entry_inner *entry;
184 	u32 level, capacity_shift, leaf_index = index / SIDTAB_LEAF_ENTRIES;
185 
186 	/* find the level of the subtree we need */
187 	level = sidtab_level_from_count(index + 1);
188 	capacity_shift = level * SIDTAB_INNER_SHIFT;
189 
190 	/* allocate roots if needed */
191 	if (alloc && sidtab_alloc_roots(s, level) != 0)
192 		return NULL;
193 
194 	/* lookup inside the subtree */
195 	entry = &s->roots[level];
196 	while (level != 0) {
197 		capacity_shift -= SIDTAB_INNER_SHIFT;
198 		--level;
199 
200 		entry = &entry->ptr_inner->entries[leaf_index >> capacity_shift];
201 		leaf_index &= ((u32)1 << capacity_shift) - 1;
202 
203 		if (!entry->ptr_inner) {
204 			if (alloc)
205 				entry->ptr_inner = kzalloc(SIDTAB_NODE_ALLOC_SIZE,
206 							   GFP_ATOMIC);
207 			if (!entry->ptr_inner)
208 				return NULL;
209 		}
210 	}
211 	if (!entry->ptr_leaf) {
212 		if (alloc)
213 			entry->ptr_leaf = kzalloc(SIDTAB_NODE_ALLOC_SIZE,
214 						  GFP_ATOMIC);
215 		if (!entry->ptr_leaf)
216 			return NULL;
217 	}
218 	return &entry->ptr_leaf->entries[index % SIDTAB_LEAF_ENTRIES];
219 }
220 
221 static struct sidtab_entry *sidtab_lookup(struct sidtab *s, u32 index)
222 {
223 	/* read entries only after reading count */
224 	u32 count = smp_load_acquire(&s->count);
225 
226 	if (index >= count)
227 		return NULL;
228 
229 	return sidtab_do_lookup(s, index, 0);
230 }
231 
232 static struct sidtab_entry *sidtab_lookup_initial(struct sidtab *s, u32 sid)
233 {
234 	return s->isids[sid - 1].set ? &s->isids[sid - 1].entry : NULL;
235 }
236 
237 static struct sidtab_entry *sidtab_search_core(struct sidtab *s, u32 sid,
238 					       int force)
239 {
240 	if (sid != 0) {
241 		struct sidtab_entry *entry;
242 
243 		if (sid > SECINITSID_NUM)
244 			entry = sidtab_lookup(s, sid_to_index(sid));
245 		else
246 			entry = sidtab_lookup_initial(s, sid);
247 		if (entry && (!entry->context.len || force))
248 			return entry;
249 	}
250 
251 	return sidtab_lookup_initial(s, SECINITSID_UNLABELED);
252 }
253 
254 struct sidtab_entry *sidtab_search_entry(struct sidtab *s, u32 sid)
255 {
256 	return sidtab_search_core(s, sid, 0);
257 }
258 
259 struct sidtab_entry *sidtab_search_entry_force(struct sidtab *s, u32 sid)
260 {
261 	return sidtab_search_core(s, sid, 1);
262 }
263 
264 int sidtab_context_to_sid(struct sidtab *s, struct context *context,
265 			  u32 *sid)
266 {
267 	unsigned long flags;
268 	u32 count, hash = context_compute_hash(context);
269 	struct sidtab_convert_params *convert;
270 	struct sidtab_entry *dst, *dst_convert;
271 	int rc;
272 
273 	*sid = context_to_sid(s, context, hash);
274 	if (*sid)
275 		return 0;
276 
277 	/* lock-free search failed: lock, re-search, and insert if not found */
278 	spin_lock_irqsave(&s->lock, flags);
279 
280 	rc = 0;
281 	*sid = context_to_sid(s, context, hash);
282 	if (*sid)
283 		goto out_unlock;
284 
285 	if (unlikely(s->frozen)) {
286 		/*
287 		 * This sidtab is now frozen - tell the caller to abort and
288 		 * get the new one.
289 		 */
290 		rc = -ESTALE;
291 		goto out_unlock;
292 	}
293 
294 	count = s->count;
295 	convert = s->convert;
296 
297 	/* bail out if we already reached max entries */
298 	rc = -EOVERFLOW;
299 	if (count >= SIDTAB_MAX)
300 		goto out_unlock;
301 
302 	/* insert context into new entry */
303 	rc = -ENOMEM;
304 	dst = sidtab_do_lookup(s, count, 1);
305 	if (!dst)
306 		goto out_unlock;
307 
308 	dst->sid = index_to_sid(count);
309 	dst->hash = hash;
310 
311 	rc = context_cpy(&dst->context, context);
312 	if (rc)
313 		goto out_unlock;
314 
315 	/*
316 	 * if we are building a new sidtab, we need to convert the context
317 	 * and insert it there as well
318 	 */
319 	if (convert) {
320 		rc = -ENOMEM;
321 		dst_convert = sidtab_do_lookup(convert->target, count, 1);
322 		if (!dst_convert) {
323 			context_destroy(&dst->context);
324 			goto out_unlock;
325 		}
326 
327 		rc = convert->func(context, &dst_convert->context,
328 				   convert->args, GFP_ATOMIC);
329 		if (rc) {
330 			context_destroy(&dst->context);
331 			goto out_unlock;
332 		}
333 		dst_convert->sid = index_to_sid(count);
334 		dst_convert->hash = context_compute_hash(&dst_convert->context);
335 		convert->target->count = count + 1;
336 
337 		hash_add_rcu(convert->target->context_to_sid,
338 			     &dst_convert->list, dst_convert->hash);
339 	}
340 
341 	if (context->len)
342 		pr_info("SELinux:  Context %s is not valid (left unmapped).\n",
343 			context->str);
344 
345 	*sid = index_to_sid(count);
346 
347 	/* write entries before updating count */
348 	smp_store_release(&s->count, count + 1);
349 	hash_add_rcu(s->context_to_sid, &dst->list, dst->hash);
350 
351 	rc = 0;
352 out_unlock:
353 	spin_unlock_irqrestore(&s->lock, flags);
354 	return rc;
355 }
356 
357 static void sidtab_convert_hashtable(struct sidtab *s, u32 count)
358 {
359 	struct sidtab_entry *entry;
360 	u32 i;
361 
362 	for (i = 0; i < count; i++) {
363 		entry = sidtab_do_lookup(s, i, 0);
364 		entry->sid = index_to_sid(i);
365 		entry->hash = context_compute_hash(&entry->context);
366 
367 		hash_add_rcu(s->context_to_sid, &entry->list, entry->hash);
368 	}
369 }
370 
371 static int sidtab_convert_tree(union sidtab_entry_inner *edst,
372 			       union sidtab_entry_inner *esrc,
373 			       u32 *pos, u32 count, u32 level,
374 			       struct sidtab_convert_params *convert)
375 {
376 	int rc;
377 	u32 i;
378 
379 	if (level != 0) {
380 		if (!edst->ptr_inner) {
381 			edst->ptr_inner = kzalloc(SIDTAB_NODE_ALLOC_SIZE,
382 						  GFP_KERNEL);
383 			if (!edst->ptr_inner)
384 				return -ENOMEM;
385 		}
386 		i = 0;
387 		while (i < SIDTAB_INNER_ENTRIES && *pos < count) {
388 			rc = sidtab_convert_tree(&edst->ptr_inner->entries[i],
389 						 &esrc->ptr_inner->entries[i],
390 						 pos, count, level - 1,
391 						 convert);
392 			if (rc)
393 				return rc;
394 			i++;
395 		}
396 	} else {
397 		if (!edst->ptr_leaf) {
398 			edst->ptr_leaf = kzalloc(SIDTAB_NODE_ALLOC_SIZE,
399 						 GFP_KERNEL);
400 			if (!edst->ptr_leaf)
401 				return -ENOMEM;
402 		}
403 		i = 0;
404 		while (i < SIDTAB_LEAF_ENTRIES && *pos < count) {
405 			rc = convert->func(&esrc->ptr_leaf->entries[i].context,
406 					   &edst->ptr_leaf->entries[i].context,
407 					   convert->args, GFP_KERNEL);
408 			if (rc)
409 				return rc;
410 			(*pos)++;
411 			i++;
412 		}
413 		cond_resched();
414 	}
415 	return 0;
416 }
417 
418 int sidtab_convert(struct sidtab *s, struct sidtab_convert_params *params)
419 {
420 	unsigned long flags;
421 	u32 count, level, pos;
422 	int rc;
423 
424 	spin_lock_irqsave(&s->lock, flags);
425 
426 	/* concurrent policy loads are not allowed */
427 	if (s->convert) {
428 		spin_unlock_irqrestore(&s->lock, flags);
429 		return -EBUSY;
430 	}
431 
432 	count = s->count;
433 	level = sidtab_level_from_count(count);
434 
435 	/* allocate last leaf in the new sidtab (to avoid race with
436 	 * live convert)
437 	 */
438 	rc = sidtab_do_lookup(params->target, count - 1, 1) ? 0 : -ENOMEM;
439 	if (rc) {
440 		spin_unlock_irqrestore(&s->lock, flags);
441 		return rc;
442 	}
443 
444 	/* set count in case no new entries are added during conversion */
445 	params->target->count = count;
446 
447 	/* enable live convert of new entries */
448 	s->convert = params;
449 
450 	/* we can safely convert the tree outside the lock */
451 	spin_unlock_irqrestore(&s->lock, flags);
452 
453 	pr_info("SELinux:  Converting %u SID table entries...\n", count);
454 
455 	/* convert all entries not covered by live convert */
456 	pos = 0;
457 	rc = sidtab_convert_tree(&params->target->roots[level],
458 				 &s->roots[level], &pos, count, level, params);
459 	if (rc) {
460 		/* we need to keep the old table - disable live convert */
461 		spin_lock_irqsave(&s->lock, flags);
462 		s->convert = NULL;
463 		spin_unlock_irqrestore(&s->lock, flags);
464 		return rc;
465 	}
466 	/*
467 	 * The hashtable can also be modified in sidtab_context_to_sid()
468 	 * so we must re-acquire the lock here.
469 	 */
470 	spin_lock_irqsave(&s->lock, flags);
471 	sidtab_convert_hashtable(params->target, count);
472 	spin_unlock_irqrestore(&s->lock, flags);
473 
474 	return 0;
475 }
476 
477 void sidtab_cancel_convert(struct sidtab *s)
478 {
479 	unsigned long flags;
480 
481 	/* cancelling policy load - disable live convert of sidtab */
482 	spin_lock_irqsave(&s->lock, flags);
483 	s->convert = NULL;
484 	spin_unlock_irqrestore(&s->lock, flags);
485 }
486 
487 void sidtab_freeze_begin(struct sidtab *s, unsigned long *flags) __acquires(&s->lock)
488 {
489 	spin_lock_irqsave(&s->lock, *flags);
490 	s->frozen = true;
491 	s->convert = NULL;
492 }
493 void sidtab_freeze_end(struct sidtab *s, unsigned long *flags) __releases(&s->lock)
494 {
495 	spin_unlock_irqrestore(&s->lock, *flags);
496 }
497 
498 static void sidtab_destroy_entry(struct sidtab_entry *entry)
499 {
500 	context_destroy(&entry->context);
501 #if CONFIG_SECURITY_SELINUX_SID2STR_CACHE_SIZE > 0
502 	kfree(rcu_dereference_raw(entry->cache));
503 #endif
504 }
505 
506 static void sidtab_destroy_tree(union sidtab_entry_inner entry, u32 level)
507 {
508 	u32 i;
509 
510 	if (level != 0) {
511 		struct sidtab_node_inner *node = entry.ptr_inner;
512 
513 		if (!node)
514 			return;
515 
516 		for (i = 0; i < SIDTAB_INNER_ENTRIES; i++)
517 			sidtab_destroy_tree(node->entries[i], level - 1);
518 		kfree(node);
519 	} else {
520 		struct sidtab_node_leaf *node = entry.ptr_leaf;
521 
522 		if (!node)
523 			return;
524 
525 		for (i = 0; i < SIDTAB_LEAF_ENTRIES; i++)
526 			sidtab_destroy_entry(&node->entries[i]);
527 		kfree(node);
528 	}
529 }
530 
531 void sidtab_destroy(struct sidtab *s)
532 {
533 	u32 i, level;
534 
535 	for (i = 0; i < SECINITSID_NUM; i++)
536 		if (s->isids[i].set)
537 			sidtab_destroy_entry(&s->isids[i].entry);
538 
539 	level = SIDTAB_MAX_LEVEL;
540 	while (level && !s->roots[level].ptr_inner)
541 		--level;
542 
543 	sidtab_destroy_tree(s->roots[level], level);
544 	/*
545 	 * The context_to_sid hashtable's objects are all shared
546 	 * with the isids array and context tree, and so don't need
547 	 * to be cleaned up here.
548 	 */
549 }
550 
551 #if CONFIG_SECURITY_SELINUX_SID2STR_CACHE_SIZE > 0
552 
553 void sidtab_sid2str_put(struct sidtab *s, struct sidtab_entry *entry,
554 			const char *str, u32 str_len)
555 {
556 	struct sidtab_str_cache *cache, *victim = NULL;
557 	unsigned long flags;
558 
559 	/* do not cache invalid contexts */
560 	if (entry->context.len)
561 		return;
562 
563 	spin_lock_irqsave(&s->cache_lock, flags);
564 
565 	cache = rcu_dereference_protected(entry->cache,
566 					  lockdep_is_held(&s->cache_lock));
567 	if (cache) {
568 		/* entry in cache - just bump to the head of LRU list */
569 		list_move(&cache->lru_member, &s->cache_lru_list);
570 		goto out_unlock;
571 	}
572 
573 	cache = kmalloc(struct_size(cache, str, str_len), GFP_ATOMIC);
574 	if (!cache)
575 		goto out_unlock;
576 
577 	if (s->cache_free_slots == 0) {
578 		/* pop a cache entry from the tail and free it */
579 		victim = container_of(s->cache_lru_list.prev,
580 				      struct sidtab_str_cache, lru_member);
581 		list_del(&victim->lru_member);
582 		rcu_assign_pointer(victim->parent->cache, NULL);
583 	} else {
584 		s->cache_free_slots--;
585 	}
586 	cache->parent = entry;
587 	cache->len = str_len;
588 	memcpy(cache->str, str, str_len);
589 	list_add(&cache->lru_member, &s->cache_lru_list);
590 
591 	rcu_assign_pointer(entry->cache, cache);
592 
593 out_unlock:
594 	spin_unlock_irqrestore(&s->cache_lock, flags);
595 	kfree_rcu(victim, rcu_member);
596 }
597 
598 int sidtab_sid2str_get(struct sidtab *s, struct sidtab_entry *entry,
599 		       char **out, u32 *out_len)
600 {
601 	struct sidtab_str_cache *cache;
602 	int rc = 0;
603 
604 	if (entry->context.len)
605 		return -ENOENT; /* do not cache invalid contexts */
606 
607 	rcu_read_lock();
608 
609 	cache = rcu_dereference(entry->cache);
610 	if (!cache) {
611 		rc = -ENOENT;
612 	} else {
613 		*out_len = cache->len;
614 		if (out) {
615 			*out = kmemdup(cache->str, cache->len, GFP_ATOMIC);
616 			if (!*out)
617 				rc = -ENOMEM;
618 		}
619 	}
620 
621 	rcu_read_unlock();
622 
623 	if (!rc && out)
624 		sidtab_sid2str_put(s, entry, *out, *out_len);
625 	return rc;
626 }
627 
628 #endif /* CONFIG_SECURITY_SELINUX_SID2STR_CACHE_SIZE > 0 */
629