xref: /openbmc/linux/net/ipv4/cipso_ipv4.c (revision 8684014d)
1 /*
2  * CIPSO - Commercial IP Security Option
3  *
4  * This is an implementation of the CIPSO 2.2 protocol as specified in
5  * draft-ietf-cipso-ipsecurity-01.txt with additional tag types as found in
6  * FIPS-188.  While CIPSO never became a full IETF RFC standard many vendors
7  * have chosen to adopt the protocol and over the years it has become a
8  * de-facto standard for labeled networking.
9  *
10  * The CIPSO draft specification can be found in the kernel's Documentation
11  * directory as well as the following URL:
12  *   http://tools.ietf.org/id/draft-ietf-cipso-ipsecurity-01.txt
13  * The FIPS-188 specification can be found at the following URL:
14  *   http://www.itl.nist.gov/fipspubs/fip188.htm
15  *
16  * Author: Paul Moore <paul.moore@hp.com>
17  *
18  */
19 
20 /*
21  * (c) Copyright Hewlett-Packard Development Company, L.P., 2006, 2008
22  *
23  * This program is free software;  you can redistribute it and/or modify
24  * it under the terms of the GNU General Public License as published by
25  * the Free Software Foundation; either version 2 of the License, or
26  * (at your option) any later version.
27  *
28  * This program is distributed in the hope that it will be useful,
29  * but WITHOUT ANY WARRANTY;  without even the implied warranty of
30  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
31  * the GNU General Public License for more details.
32  *
33  * You should have received a copy of the GNU General Public License
34  * along with this program;  if not, see <http://www.gnu.org/licenses/>.
35  *
36  */
37 
38 #include <linux/init.h>
39 #include <linux/types.h>
40 #include <linux/rcupdate.h>
41 #include <linux/list.h>
42 #include <linux/spinlock.h>
43 #include <linux/string.h>
44 #include <linux/jhash.h>
45 #include <linux/audit.h>
46 #include <linux/slab.h>
47 #include <net/ip.h>
48 #include <net/icmp.h>
49 #include <net/tcp.h>
50 #include <net/netlabel.h>
51 #include <net/cipso_ipv4.h>
52 #include <linux/atomic.h>
53 #include <linux/bug.h>
54 #include <asm/unaligned.h>
55 
56 /* List of available DOI definitions */
57 /* XXX - This currently assumes a minimal number of different DOIs in use,
58  * if in practice there are a lot of different DOIs this list should
59  * probably be turned into a hash table or something similar so we
60  * can do quick lookups. */
61 static DEFINE_SPINLOCK(cipso_v4_doi_list_lock);
62 static LIST_HEAD(cipso_v4_doi_list);
63 
64 /* Label mapping cache */
65 int cipso_v4_cache_enabled = 1;
66 int cipso_v4_cache_bucketsize = 10;
67 #define CIPSO_V4_CACHE_BUCKETBITS     7
68 #define CIPSO_V4_CACHE_BUCKETS        (1 << CIPSO_V4_CACHE_BUCKETBITS)
69 #define CIPSO_V4_CACHE_REORDERLIMIT   10
70 struct cipso_v4_map_cache_bkt {
71 	spinlock_t lock;
72 	u32 size;
73 	struct list_head list;
74 };
75 
76 struct cipso_v4_map_cache_entry {
77 	u32 hash;
78 	unsigned char *key;
79 	size_t key_len;
80 
81 	struct netlbl_lsm_cache *lsm_data;
82 
83 	u32 activity;
84 	struct list_head list;
85 };
86 
87 static struct cipso_v4_map_cache_bkt *cipso_v4_cache;
88 
89 /* Restricted bitmap (tag #1) flags */
90 int cipso_v4_rbm_optfmt = 0;
91 int cipso_v4_rbm_strictvalid = 1;
92 
93 /*
94  * Protocol Constants
95  */
96 
97 /* Maximum size of the CIPSO IP option, derived from the fact that the maximum
98  * IPv4 header size is 60 bytes and the base IPv4 header is 20 bytes long. */
99 #define CIPSO_V4_OPT_LEN_MAX          40
100 
101 /* Length of the base CIPSO option, this includes the option type (1 byte), the
102  * option length (1 byte), and the DOI (4 bytes). */
103 #define CIPSO_V4_HDR_LEN              6
104 
105 /* Base length of the restrictive category bitmap tag (tag #1). */
106 #define CIPSO_V4_TAG_RBM_BLEN         4
107 
108 /* Base length of the enumerated category tag (tag #2). */
109 #define CIPSO_V4_TAG_ENUM_BLEN        4
110 
111 /* Base length of the ranged categories bitmap tag (tag #5). */
112 #define CIPSO_V4_TAG_RNG_BLEN         4
113 /* The maximum number of category ranges permitted in the ranged category tag
114  * (tag #5).  You may note that the IETF draft states that the maximum number
115  * of category ranges is 7, but if the low end of the last category range is
116  * zero then it is possible to fit 8 category ranges because the zero should
117  * be omitted. */
118 #define CIPSO_V4_TAG_RNG_CAT_MAX      8
119 
120 /* Base length of the local tag (non-standard tag).
121  *  Tag definition (may change between kernel versions)
122  *
123  * 0          8          16         24         32
124  * +----------+----------+----------+----------+
125  * | 10000000 | 00000110 | 32-bit secid value  |
126  * +----------+----------+----------+----------+
127  * | in (host byte order)|
128  * +----------+----------+
129  *
130  */
131 #define CIPSO_V4_TAG_LOC_BLEN         6
132 
133 /*
134  * Helper Functions
135  */
136 
137 /**
138  * cipso_v4_bitmap_walk - Walk a bitmap looking for a bit
139  * @bitmap: the bitmap
140  * @bitmap_len: length in bits
141  * @offset: starting offset
142  * @state: if non-zero, look for a set (1) bit else look for a cleared (0) bit
143  *
144  * Description:
145  * Starting at @offset, walk the bitmap from left to right until either the
146  * desired bit is found or we reach the end.  Return the bit offset, -1 if
147  * not found, or -2 if error.
148  */
149 static int cipso_v4_bitmap_walk(const unsigned char *bitmap,
150 				u32 bitmap_len,
151 				u32 offset,
152 				u8 state)
153 {
154 	u32 bit_spot;
155 	u32 byte_offset;
156 	unsigned char bitmask;
157 	unsigned char byte;
158 
159 	/* gcc always rounds to zero when doing integer division */
160 	byte_offset = offset / 8;
161 	byte = bitmap[byte_offset];
162 	bit_spot = offset;
163 	bitmask = 0x80 >> (offset % 8);
164 
165 	while (bit_spot < bitmap_len) {
166 		if ((state && (byte & bitmask) == bitmask) ||
167 		    (state == 0 && (byte & bitmask) == 0))
168 			return bit_spot;
169 
170 		bit_spot++;
171 		bitmask >>= 1;
172 		if (bitmask == 0) {
173 			byte = bitmap[++byte_offset];
174 			bitmask = 0x80;
175 		}
176 	}
177 
178 	return -1;
179 }
180 
181 /**
182  * cipso_v4_bitmap_setbit - Sets a single bit in a bitmap
183  * @bitmap: the bitmap
184  * @bit: the bit
185  * @state: if non-zero, set the bit (1) else clear the bit (0)
186  *
187  * Description:
188  * Set a single bit in the bitmask.  Returns zero on success, negative values
189  * on error.
190  */
191 static void cipso_v4_bitmap_setbit(unsigned char *bitmap,
192 				   u32 bit,
193 				   u8 state)
194 {
195 	u32 byte_spot;
196 	u8 bitmask;
197 
198 	/* gcc always rounds to zero when doing integer division */
199 	byte_spot = bit / 8;
200 	bitmask = 0x80 >> (bit % 8);
201 	if (state)
202 		bitmap[byte_spot] |= bitmask;
203 	else
204 		bitmap[byte_spot] &= ~bitmask;
205 }
206 
207 /**
208  * cipso_v4_cache_entry_free - Frees a cache entry
209  * @entry: the entry to free
210  *
211  * Description:
212  * This function frees the memory associated with a cache entry including the
213  * LSM cache data if there are no longer any users, i.e. reference count == 0.
214  *
215  */
216 static void cipso_v4_cache_entry_free(struct cipso_v4_map_cache_entry *entry)
217 {
218 	if (entry->lsm_data)
219 		netlbl_secattr_cache_free(entry->lsm_data);
220 	kfree(entry->key);
221 	kfree(entry);
222 }
223 
224 /**
225  * cipso_v4_map_cache_hash - Hashing function for the CIPSO cache
226  * @key: the hash key
227  * @key_len: the length of the key in bytes
228  *
229  * Description:
230  * The CIPSO tag hashing function.  Returns a 32-bit hash value.
231  *
232  */
233 static u32 cipso_v4_map_cache_hash(const unsigned char *key, u32 key_len)
234 {
235 	return jhash(key, key_len, 0);
236 }
237 
238 /*
239  * Label Mapping Cache Functions
240  */
241 
242 /**
243  * cipso_v4_cache_init - Initialize the CIPSO cache
244  *
245  * Description:
246  * Initializes the CIPSO label mapping cache, this function should be called
247  * before any of the other functions defined in this file.  Returns zero on
248  * success, negative values on error.
249  *
250  */
251 static int __init cipso_v4_cache_init(void)
252 {
253 	u32 iter;
254 
255 	cipso_v4_cache = kcalloc(CIPSO_V4_CACHE_BUCKETS,
256 				 sizeof(struct cipso_v4_map_cache_bkt),
257 				 GFP_KERNEL);
258 	if (cipso_v4_cache == NULL)
259 		return -ENOMEM;
260 
261 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
262 		spin_lock_init(&cipso_v4_cache[iter].lock);
263 		cipso_v4_cache[iter].size = 0;
264 		INIT_LIST_HEAD(&cipso_v4_cache[iter].list);
265 	}
266 
267 	return 0;
268 }
269 
270 /**
271  * cipso_v4_cache_invalidate - Invalidates the current CIPSO cache
272  *
273  * Description:
274  * Invalidates and frees any entries in the CIPSO cache.  Returns zero on
275  * success and negative values on failure.
276  *
277  */
278 void cipso_v4_cache_invalidate(void)
279 {
280 	struct cipso_v4_map_cache_entry *entry, *tmp_entry;
281 	u32 iter;
282 
283 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
284 		spin_lock_bh(&cipso_v4_cache[iter].lock);
285 		list_for_each_entry_safe(entry,
286 					 tmp_entry,
287 					 &cipso_v4_cache[iter].list, list) {
288 			list_del(&entry->list);
289 			cipso_v4_cache_entry_free(entry);
290 		}
291 		cipso_v4_cache[iter].size = 0;
292 		spin_unlock_bh(&cipso_v4_cache[iter].lock);
293 	}
294 }
295 
296 /**
297  * cipso_v4_cache_check - Check the CIPSO cache for a label mapping
298  * @key: the buffer to check
299  * @key_len: buffer length in bytes
300  * @secattr: the security attribute struct to use
301  *
302  * Description:
303  * This function checks the cache to see if a label mapping already exists for
304  * the given key.  If there is a match then the cache is adjusted and the
305  * @secattr struct is populated with the correct LSM security attributes.  The
306  * cache is adjusted in the following manner if the entry is not already the
307  * first in the cache bucket:
308  *
309  *  1. The cache entry's activity counter is incremented
310  *  2. The previous (higher ranking) entry's activity counter is decremented
311  *  3. If the difference between the two activity counters is geater than
312  *     CIPSO_V4_CACHE_REORDERLIMIT the two entries are swapped
313  *
314  * Returns zero on success, -ENOENT for a cache miss, and other negative values
315  * on error.
316  *
317  */
318 static int cipso_v4_cache_check(const unsigned char *key,
319 				u32 key_len,
320 				struct netlbl_lsm_secattr *secattr)
321 {
322 	u32 bkt;
323 	struct cipso_v4_map_cache_entry *entry;
324 	struct cipso_v4_map_cache_entry *prev_entry = NULL;
325 	u32 hash;
326 
327 	if (!cipso_v4_cache_enabled)
328 		return -ENOENT;
329 
330 	hash = cipso_v4_map_cache_hash(key, key_len);
331 	bkt = hash & (CIPSO_V4_CACHE_BUCKETS - 1);
332 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
333 	list_for_each_entry(entry, &cipso_v4_cache[bkt].list, list) {
334 		if (entry->hash == hash &&
335 		    entry->key_len == key_len &&
336 		    memcmp(entry->key, key, key_len) == 0) {
337 			entry->activity += 1;
338 			atomic_inc(&entry->lsm_data->refcount);
339 			secattr->cache = entry->lsm_data;
340 			secattr->flags |= NETLBL_SECATTR_CACHE;
341 			secattr->type = NETLBL_NLTYPE_CIPSOV4;
342 			if (prev_entry == NULL) {
343 				spin_unlock_bh(&cipso_v4_cache[bkt].lock);
344 				return 0;
345 			}
346 
347 			if (prev_entry->activity > 0)
348 				prev_entry->activity -= 1;
349 			if (entry->activity > prev_entry->activity &&
350 			    entry->activity - prev_entry->activity >
351 			    CIPSO_V4_CACHE_REORDERLIMIT) {
352 				__list_del(entry->list.prev, entry->list.next);
353 				__list_add(&entry->list,
354 					   prev_entry->list.prev,
355 					   &prev_entry->list);
356 			}
357 
358 			spin_unlock_bh(&cipso_v4_cache[bkt].lock);
359 			return 0;
360 		}
361 		prev_entry = entry;
362 	}
363 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
364 
365 	return -ENOENT;
366 }
367 
368 /**
369  * cipso_v4_cache_add - Add an entry to the CIPSO cache
370  * @skb: the packet
371  * @secattr: the packet's security attributes
372  *
373  * Description:
374  * Add a new entry into the CIPSO label mapping cache.  Add the new entry to
375  * head of the cache bucket's list, if the cache bucket is out of room remove
376  * the last entry in the list first.  It is important to note that there is
377  * currently no checking for duplicate keys.  Returns zero on success,
378  * negative values on failure.
379  *
380  */
381 int cipso_v4_cache_add(const struct sk_buff *skb,
382 		       const struct netlbl_lsm_secattr *secattr)
383 {
384 	int ret_val = -EPERM;
385 	u32 bkt;
386 	struct cipso_v4_map_cache_entry *entry = NULL;
387 	struct cipso_v4_map_cache_entry *old_entry = NULL;
388 	unsigned char *cipso_ptr;
389 	u32 cipso_ptr_len;
390 
391 	if (!cipso_v4_cache_enabled || cipso_v4_cache_bucketsize <= 0)
392 		return 0;
393 
394 	cipso_ptr = CIPSO_V4_OPTPTR(skb);
395 	cipso_ptr_len = cipso_ptr[1];
396 
397 	entry = kzalloc(sizeof(*entry), GFP_ATOMIC);
398 	if (entry == NULL)
399 		return -ENOMEM;
400 	entry->key = kmemdup(cipso_ptr, cipso_ptr_len, GFP_ATOMIC);
401 	if (entry->key == NULL) {
402 		ret_val = -ENOMEM;
403 		goto cache_add_failure;
404 	}
405 	entry->key_len = cipso_ptr_len;
406 	entry->hash = cipso_v4_map_cache_hash(cipso_ptr, cipso_ptr_len);
407 	atomic_inc(&secattr->cache->refcount);
408 	entry->lsm_data = secattr->cache;
409 
410 	bkt = entry->hash & (CIPSO_V4_CACHE_BUCKETS - 1);
411 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
412 	if (cipso_v4_cache[bkt].size < cipso_v4_cache_bucketsize) {
413 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
414 		cipso_v4_cache[bkt].size += 1;
415 	} else {
416 		old_entry = list_entry(cipso_v4_cache[bkt].list.prev,
417 				       struct cipso_v4_map_cache_entry, list);
418 		list_del(&old_entry->list);
419 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
420 		cipso_v4_cache_entry_free(old_entry);
421 	}
422 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
423 
424 	return 0;
425 
426 cache_add_failure:
427 	if (entry)
428 		cipso_v4_cache_entry_free(entry);
429 	return ret_val;
430 }
431 
432 /*
433  * DOI List Functions
434  */
435 
436 /**
437  * cipso_v4_doi_search - Searches for a DOI definition
438  * @doi: the DOI to search for
439  *
440  * Description:
441  * Search the DOI definition list for a DOI definition with a DOI value that
442  * matches @doi.  The caller is responsible for calling rcu_read_[un]lock().
443  * Returns a pointer to the DOI definition on success and NULL on failure.
444  */
445 static struct cipso_v4_doi *cipso_v4_doi_search(u32 doi)
446 {
447 	struct cipso_v4_doi *iter;
448 
449 	list_for_each_entry_rcu(iter, &cipso_v4_doi_list, list)
450 		if (iter->doi == doi && atomic_read(&iter->refcount))
451 			return iter;
452 	return NULL;
453 }
454 
455 /**
456  * cipso_v4_doi_add - Add a new DOI to the CIPSO protocol engine
457  * @doi_def: the DOI structure
458  * @audit_info: NetLabel audit information
459  *
460  * Description:
461  * The caller defines a new DOI for use by the CIPSO engine and calls this
462  * function to add it to the list of acceptable domains.  The caller must
463  * ensure that the mapping table specified in @doi_def->map meets all of the
464  * requirements of the mapping type (see cipso_ipv4.h for details).  Returns
465  * zero on success and non-zero on failure.
466  *
467  */
468 int cipso_v4_doi_add(struct cipso_v4_doi *doi_def,
469 		     struct netlbl_audit *audit_info)
470 {
471 	int ret_val = -EINVAL;
472 	u32 iter;
473 	u32 doi;
474 	u32 doi_type;
475 	struct audit_buffer *audit_buf;
476 
477 	doi = doi_def->doi;
478 	doi_type = doi_def->type;
479 
480 	if (doi_def->doi == CIPSO_V4_DOI_UNKNOWN)
481 		goto doi_add_return;
482 	for (iter = 0; iter < CIPSO_V4_TAG_MAXCNT; iter++) {
483 		switch (doi_def->tags[iter]) {
484 		case CIPSO_V4_TAG_RBITMAP:
485 			break;
486 		case CIPSO_V4_TAG_RANGE:
487 		case CIPSO_V4_TAG_ENUM:
488 			if (doi_def->type != CIPSO_V4_MAP_PASS)
489 				goto doi_add_return;
490 			break;
491 		case CIPSO_V4_TAG_LOCAL:
492 			if (doi_def->type != CIPSO_V4_MAP_LOCAL)
493 				goto doi_add_return;
494 			break;
495 		case CIPSO_V4_TAG_INVALID:
496 			if (iter == 0)
497 				goto doi_add_return;
498 			break;
499 		default:
500 			goto doi_add_return;
501 		}
502 	}
503 
504 	atomic_set(&doi_def->refcount, 1);
505 
506 	spin_lock(&cipso_v4_doi_list_lock);
507 	if (cipso_v4_doi_search(doi_def->doi) != NULL) {
508 		spin_unlock(&cipso_v4_doi_list_lock);
509 		ret_val = -EEXIST;
510 		goto doi_add_return;
511 	}
512 	list_add_tail_rcu(&doi_def->list, &cipso_v4_doi_list);
513 	spin_unlock(&cipso_v4_doi_list_lock);
514 	ret_val = 0;
515 
516 doi_add_return:
517 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_ADD, audit_info);
518 	if (audit_buf != NULL) {
519 		const char *type_str;
520 		switch (doi_type) {
521 		case CIPSO_V4_MAP_TRANS:
522 			type_str = "trans";
523 			break;
524 		case CIPSO_V4_MAP_PASS:
525 			type_str = "pass";
526 			break;
527 		case CIPSO_V4_MAP_LOCAL:
528 			type_str = "local";
529 			break;
530 		default:
531 			type_str = "(unknown)";
532 		}
533 		audit_log_format(audit_buf,
534 				 " cipso_doi=%u cipso_type=%s res=%u",
535 				 doi, type_str, ret_val == 0 ? 1 : 0);
536 		audit_log_end(audit_buf);
537 	}
538 
539 	return ret_val;
540 }
541 
542 /**
543  * cipso_v4_doi_free - Frees a DOI definition
544  * @doi_def: the DOI definition
545  *
546  * Description:
547  * This function frees all of the memory associated with a DOI definition.
548  *
549  */
550 void cipso_v4_doi_free(struct cipso_v4_doi *doi_def)
551 {
552 	if (doi_def == NULL)
553 		return;
554 
555 	switch (doi_def->type) {
556 	case CIPSO_V4_MAP_TRANS:
557 		kfree(doi_def->map.std->lvl.cipso);
558 		kfree(doi_def->map.std->lvl.local);
559 		kfree(doi_def->map.std->cat.cipso);
560 		kfree(doi_def->map.std->cat.local);
561 		break;
562 	}
563 	kfree(doi_def);
564 }
565 
566 /**
567  * cipso_v4_doi_free_rcu - Frees a DOI definition via the RCU pointer
568  * @entry: the entry's RCU field
569  *
570  * Description:
571  * This function is designed to be used as a callback to the call_rcu()
572  * function so that the memory allocated to the DOI definition can be released
573  * safely.
574  *
575  */
576 static void cipso_v4_doi_free_rcu(struct rcu_head *entry)
577 {
578 	struct cipso_v4_doi *doi_def;
579 
580 	doi_def = container_of(entry, struct cipso_v4_doi, rcu);
581 	cipso_v4_doi_free(doi_def);
582 }
583 
584 /**
585  * cipso_v4_doi_remove - Remove an existing DOI from the CIPSO protocol engine
586  * @doi: the DOI value
587  * @audit_secid: the LSM secid to use in the audit message
588  *
589  * Description:
590  * Removes a DOI definition from the CIPSO engine.  The NetLabel routines will
591  * be called to release their own LSM domain mappings as well as our own
592  * domain list.  Returns zero on success and negative values on failure.
593  *
594  */
595 int cipso_v4_doi_remove(u32 doi, struct netlbl_audit *audit_info)
596 {
597 	int ret_val;
598 	struct cipso_v4_doi *doi_def;
599 	struct audit_buffer *audit_buf;
600 
601 	spin_lock(&cipso_v4_doi_list_lock);
602 	doi_def = cipso_v4_doi_search(doi);
603 	if (doi_def == NULL) {
604 		spin_unlock(&cipso_v4_doi_list_lock);
605 		ret_val = -ENOENT;
606 		goto doi_remove_return;
607 	}
608 	if (!atomic_dec_and_test(&doi_def->refcount)) {
609 		spin_unlock(&cipso_v4_doi_list_lock);
610 		ret_val = -EBUSY;
611 		goto doi_remove_return;
612 	}
613 	list_del_rcu(&doi_def->list);
614 	spin_unlock(&cipso_v4_doi_list_lock);
615 
616 	cipso_v4_cache_invalidate();
617 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
618 	ret_val = 0;
619 
620 doi_remove_return:
621 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_DEL, audit_info);
622 	if (audit_buf != NULL) {
623 		audit_log_format(audit_buf,
624 				 " cipso_doi=%u res=%u",
625 				 doi, ret_val == 0 ? 1 : 0);
626 		audit_log_end(audit_buf);
627 	}
628 
629 	return ret_val;
630 }
631 
632 /**
633  * cipso_v4_doi_getdef - Returns a reference to a valid DOI definition
634  * @doi: the DOI value
635  *
636  * Description:
637  * Searches for a valid DOI definition and if one is found it is returned to
638  * the caller.  Otherwise NULL is returned.  The caller must ensure that
639  * rcu_read_lock() is held while accessing the returned definition and the DOI
640  * definition reference count is decremented when the caller is done.
641  *
642  */
643 struct cipso_v4_doi *cipso_v4_doi_getdef(u32 doi)
644 {
645 	struct cipso_v4_doi *doi_def;
646 
647 	rcu_read_lock();
648 	doi_def = cipso_v4_doi_search(doi);
649 	if (doi_def == NULL)
650 		goto doi_getdef_return;
651 	if (!atomic_inc_not_zero(&doi_def->refcount))
652 		doi_def = NULL;
653 
654 doi_getdef_return:
655 	rcu_read_unlock();
656 	return doi_def;
657 }
658 
659 /**
660  * cipso_v4_doi_putdef - Releases a reference for the given DOI definition
661  * @doi_def: the DOI definition
662  *
663  * Description:
664  * Releases a DOI definition reference obtained from cipso_v4_doi_getdef().
665  *
666  */
667 void cipso_v4_doi_putdef(struct cipso_v4_doi *doi_def)
668 {
669 	if (doi_def == NULL)
670 		return;
671 
672 	if (!atomic_dec_and_test(&doi_def->refcount))
673 		return;
674 	spin_lock(&cipso_v4_doi_list_lock);
675 	list_del_rcu(&doi_def->list);
676 	spin_unlock(&cipso_v4_doi_list_lock);
677 
678 	cipso_v4_cache_invalidate();
679 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
680 }
681 
682 /**
683  * cipso_v4_doi_walk - Iterate through the DOI definitions
684  * @skip_cnt: skip past this number of DOI definitions, updated
685  * @callback: callback for each DOI definition
686  * @cb_arg: argument for the callback function
687  *
688  * Description:
689  * Iterate over the DOI definition list, skipping the first @skip_cnt entries.
690  * For each entry call @callback, if @callback returns a negative value stop
691  * 'walking' through the list and return.  Updates the value in @skip_cnt upon
692  * return.  Returns zero on success, negative values on failure.
693  *
694  */
695 int cipso_v4_doi_walk(u32 *skip_cnt,
696 		     int (*callback) (struct cipso_v4_doi *doi_def, void *arg),
697 		     void *cb_arg)
698 {
699 	int ret_val = -ENOENT;
700 	u32 doi_cnt = 0;
701 	struct cipso_v4_doi *iter_doi;
702 
703 	rcu_read_lock();
704 	list_for_each_entry_rcu(iter_doi, &cipso_v4_doi_list, list)
705 		if (atomic_read(&iter_doi->refcount) > 0) {
706 			if (doi_cnt++ < *skip_cnt)
707 				continue;
708 			ret_val = callback(iter_doi, cb_arg);
709 			if (ret_val < 0) {
710 				doi_cnt--;
711 				goto doi_walk_return;
712 			}
713 		}
714 
715 doi_walk_return:
716 	rcu_read_unlock();
717 	*skip_cnt = doi_cnt;
718 	return ret_val;
719 }
720 
721 /*
722  * Label Mapping Functions
723  */
724 
725 /**
726  * cipso_v4_map_lvl_valid - Checks to see if the given level is understood
727  * @doi_def: the DOI definition
728  * @level: the level to check
729  *
730  * Description:
731  * Checks the given level against the given DOI definition and returns a
732  * negative value if the level does not have a valid mapping and a zero value
733  * if the level is defined by the DOI.
734  *
735  */
736 static int cipso_v4_map_lvl_valid(const struct cipso_v4_doi *doi_def, u8 level)
737 {
738 	switch (doi_def->type) {
739 	case CIPSO_V4_MAP_PASS:
740 		return 0;
741 	case CIPSO_V4_MAP_TRANS:
742 		if (doi_def->map.std->lvl.cipso[level] < CIPSO_V4_INV_LVL)
743 			return 0;
744 		break;
745 	}
746 
747 	return -EFAULT;
748 }
749 
750 /**
751  * cipso_v4_map_lvl_hton - Perform a level mapping from the host to the network
752  * @doi_def: the DOI definition
753  * @host_lvl: the host MLS level
754  * @net_lvl: the network/CIPSO MLS level
755  *
756  * Description:
757  * Perform a label mapping to translate a local MLS level to the correct
758  * CIPSO level using the given DOI definition.  Returns zero on success,
759  * negative values otherwise.
760  *
761  */
762 static int cipso_v4_map_lvl_hton(const struct cipso_v4_doi *doi_def,
763 				 u32 host_lvl,
764 				 u32 *net_lvl)
765 {
766 	switch (doi_def->type) {
767 	case CIPSO_V4_MAP_PASS:
768 		*net_lvl = host_lvl;
769 		return 0;
770 	case CIPSO_V4_MAP_TRANS:
771 		if (host_lvl < doi_def->map.std->lvl.local_size &&
772 		    doi_def->map.std->lvl.local[host_lvl] < CIPSO_V4_INV_LVL) {
773 			*net_lvl = doi_def->map.std->lvl.local[host_lvl];
774 			return 0;
775 		}
776 		return -EPERM;
777 	}
778 
779 	return -EINVAL;
780 }
781 
782 /**
783  * cipso_v4_map_lvl_ntoh - Perform a level mapping from the network to the host
784  * @doi_def: the DOI definition
785  * @net_lvl: the network/CIPSO MLS level
786  * @host_lvl: the host MLS level
787  *
788  * Description:
789  * Perform a label mapping to translate a CIPSO level to the correct local MLS
790  * level using the given DOI definition.  Returns zero on success, negative
791  * values otherwise.
792  *
793  */
794 static int cipso_v4_map_lvl_ntoh(const struct cipso_v4_doi *doi_def,
795 				 u32 net_lvl,
796 				 u32 *host_lvl)
797 {
798 	struct cipso_v4_std_map_tbl *map_tbl;
799 
800 	switch (doi_def->type) {
801 	case CIPSO_V4_MAP_PASS:
802 		*host_lvl = net_lvl;
803 		return 0;
804 	case CIPSO_V4_MAP_TRANS:
805 		map_tbl = doi_def->map.std;
806 		if (net_lvl < map_tbl->lvl.cipso_size &&
807 		    map_tbl->lvl.cipso[net_lvl] < CIPSO_V4_INV_LVL) {
808 			*host_lvl = doi_def->map.std->lvl.cipso[net_lvl];
809 			return 0;
810 		}
811 		return -EPERM;
812 	}
813 
814 	return -EINVAL;
815 }
816 
817 /**
818  * cipso_v4_map_cat_rbm_valid - Checks to see if the category bitmap is valid
819  * @doi_def: the DOI definition
820  * @bitmap: category bitmap
821  * @bitmap_len: bitmap length in bytes
822  *
823  * Description:
824  * Checks the given category bitmap against the given DOI definition and
825  * returns a negative value if any of the categories in the bitmap do not have
826  * a valid mapping and a zero value if all of the categories are valid.
827  *
828  */
829 static int cipso_v4_map_cat_rbm_valid(const struct cipso_v4_doi *doi_def,
830 				      const unsigned char *bitmap,
831 				      u32 bitmap_len)
832 {
833 	int cat = -1;
834 	u32 bitmap_len_bits = bitmap_len * 8;
835 	u32 cipso_cat_size;
836 	u32 *cipso_array;
837 
838 	switch (doi_def->type) {
839 	case CIPSO_V4_MAP_PASS:
840 		return 0;
841 	case CIPSO_V4_MAP_TRANS:
842 		cipso_cat_size = doi_def->map.std->cat.cipso_size;
843 		cipso_array = doi_def->map.std->cat.cipso;
844 		for (;;) {
845 			cat = cipso_v4_bitmap_walk(bitmap,
846 						   bitmap_len_bits,
847 						   cat + 1,
848 						   1);
849 			if (cat < 0)
850 				break;
851 			if (cat >= cipso_cat_size ||
852 			    cipso_array[cat] >= CIPSO_V4_INV_CAT)
853 				return -EFAULT;
854 		}
855 
856 		if (cat == -1)
857 			return 0;
858 		break;
859 	}
860 
861 	return -EFAULT;
862 }
863 
864 /**
865  * cipso_v4_map_cat_rbm_hton - Perform a category mapping from host to network
866  * @doi_def: the DOI definition
867  * @secattr: the security attributes
868  * @net_cat: the zero'd out category bitmap in network/CIPSO format
869  * @net_cat_len: the length of the CIPSO bitmap in bytes
870  *
871  * Description:
872  * Perform a label mapping to translate a local MLS category bitmap to the
873  * correct CIPSO bitmap using the given DOI definition.  Returns the minimum
874  * size in bytes of the network bitmap on success, negative values otherwise.
875  *
876  */
877 static int cipso_v4_map_cat_rbm_hton(const struct cipso_v4_doi *doi_def,
878 				     const struct netlbl_lsm_secattr *secattr,
879 				     unsigned char *net_cat,
880 				     u32 net_cat_len)
881 {
882 	int host_spot = -1;
883 	u32 net_spot = CIPSO_V4_INV_CAT;
884 	u32 net_spot_max = 0;
885 	u32 net_clen_bits = net_cat_len * 8;
886 	u32 host_cat_size = 0;
887 	u32 *host_cat_array = NULL;
888 
889 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
890 		host_cat_size = doi_def->map.std->cat.local_size;
891 		host_cat_array = doi_def->map.std->cat.local;
892 	}
893 
894 	for (;;) {
895 		host_spot = netlbl_catmap_walk(secattr->attr.mls.cat,
896 					       host_spot + 1);
897 		if (host_spot < 0)
898 			break;
899 
900 		switch (doi_def->type) {
901 		case CIPSO_V4_MAP_PASS:
902 			net_spot = host_spot;
903 			break;
904 		case CIPSO_V4_MAP_TRANS:
905 			if (host_spot >= host_cat_size)
906 				return -EPERM;
907 			net_spot = host_cat_array[host_spot];
908 			if (net_spot >= CIPSO_V4_INV_CAT)
909 				return -EPERM;
910 			break;
911 		}
912 		if (net_spot >= net_clen_bits)
913 			return -ENOSPC;
914 		cipso_v4_bitmap_setbit(net_cat, net_spot, 1);
915 
916 		if (net_spot > net_spot_max)
917 			net_spot_max = net_spot;
918 	}
919 
920 	if (++net_spot_max % 8)
921 		return net_spot_max / 8 + 1;
922 	return net_spot_max / 8;
923 }
924 
925 /**
926  * cipso_v4_map_cat_rbm_ntoh - Perform a category mapping from network to host
927  * @doi_def: the DOI definition
928  * @net_cat: the category bitmap in network/CIPSO format
929  * @net_cat_len: the length of the CIPSO bitmap in bytes
930  * @secattr: the security attributes
931  *
932  * Description:
933  * Perform a label mapping to translate a CIPSO bitmap to the correct local
934  * MLS category bitmap using the given DOI definition.  Returns zero on
935  * success, negative values on failure.
936  *
937  */
938 static int cipso_v4_map_cat_rbm_ntoh(const struct cipso_v4_doi *doi_def,
939 				     const unsigned char *net_cat,
940 				     u32 net_cat_len,
941 				     struct netlbl_lsm_secattr *secattr)
942 {
943 	int ret_val;
944 	int net_spot = -1;
945 	u32 host_spot = CIPSO_V4_INV_CAT;
946 	u32 net_clen_bits = net_cat_len * 8;
947 	u32 net_cat_size = 0;
948 	u32 *net_cat_array = NULL;
949 
950 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
951 		net_cat_size = doi_def->map.std->cat.cipso_size;
952 		net_cat_array = doi_def->map.std->cat.cipso;
953 	}
954 
955 	for (;;) {
956 		net_spot = cipso_v4_bitmap_walk(net_cat,
957 						net_clen_bits,
958 						net_spot + 1,
959 						1);
960 		if (net_spot < 0) {
961 			if (net_spot == -2)
962 				return -EFAULT;
963 			return 0;
964 		}
965 
966 		switch (doi_def->type) {
967 		case CIPSO_V4_MAP_PASS:
968 			host_spot = net_spot;
969 			break;
970 		case CIPSO_V4_MAP_TRANS:
971 			if (net_spot >= net_cat_size)
972 				return -EPERM;
973 			host_spot = net_cat_array[net_spot];
974 			if (host_spot >= CIPSO_V4_INV_CAT)
975 				return -EPERM;
976 			break;
977 		}
978 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
979 						       host_spot,
980 						       GFP_ATOMIC);
981 		if (ret_val != 0)
982 			return ret_val;
983 	}
984 
985 	return -EINVAL;
986 }
987 
988 /**
989  * cipso_v4_map_cat_enum_valid - Checks to see if the categories are valid
990  * @doi_def: the DOI definition
991  * @enumcat: category list
992  * @enumcat_len: length of the category list in bytes
993  *
994  * Description:
995  * Checks the given categories against the given DOI definition and returns a
996  * negative value if any of the categories do not have a valid mapping and a
997  * zero value if all of the categories are valid.
998  *
999  */
1000 static int cipso_v4_map_cat_enum_valid(const struct cipso_v4_doi *doi_def,
1001 				       const unsigned char *enumcat,
1002 				       u32 enumcat_len)
1003 {
1004 	u16 cat;
1005 	int cat_prev = -1;
1006 	u32 iter;
1007 
1008 	if (doi_def->type != CIPSO_V4_MAP_PASS || enumcat_len & 0x01)
1009 		return -EFAULT;
1010 
1011 	for (iter = 0; iter < enumcat_len; iter += 2) {
1012 		cat = get_unaligned_be16(&enumcat[iter]);
1013 		if (cat <= cat_prev)
1014 			return -EFAULT;
1015 		cat_prev = cat;
1016 	}
1017 
1018 	return 0;
1019 }
1020 
1021 /**
1022  * cipso_v4_map_cat_enum_hton - Perform a category mapping from host to network
1023  * @doi_def: the DOI definition
1024  * @secattr: the security attributes
1025  * @net_cat: the zero'd out category list in network/CIPSO format
1026  * @net_cat_len: the length of the CIPSO category list in bytes
1027  *
1028  * Description:
1029  * Perform a label mapping to translate a local MLS category bitmap to the
1030  * correct CIPSO category list using the given DOI definition.   Returns the
1031  * size in bytes of the network category bitmap on success, negative values
1032  * otherwise.
1033  *
1034  */
1035 static int cipso_v4_map_cat_enum_hton(const struct cipso_v4_doi *doi_def,
1036 				      const struct netlbl_lsm_secattr *secattr,
1037 				      unsigned char *net_cat,
1038 				      u32 net_cat_len)
1039 {
1040 	int cat = -1;
1041 	u32 cat_iter = 0;
1042 
1043 	for (;;) {
1044 		cat = netlbl_catmap_walk(secattr->attr.mls.cat, cat + 1);
1045 		if (cat < 0)
1046 			break;
1047 		if ((cat_iter + 2) > net_cat_len)
1048 			return -ENOSPC;
1049 
1050 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
1051 		cat_iter += 2;
1052 	}
1053 
1054 	return cat_iter;
1055 }
1056 
1057 /**
1058  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
1059  * @doi_def: the DOI definition
1060  * @net_cat: the category list in network/CIPSO format
1061  * @net_cat_len: the length of the CIPSO bitmap in bytes
1062  * @secattr: the security attributes
1063  *
1064  * Description:
1065  * Perform a label mapping to translate a CIPSO category list to the correct
1066  * local MLS category bitmap using the given DOI definition.  Returns zero on
1067  * success, negative values on failure.
1068  *
1069  */
1070 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
1071 				      const unsigned char *net_cat,
1072 				      u32 net_cat_len,
1073 				      struct netlbl_lsm_secattr *secattr)
1074 {
1075 	int ret_val;
1076 	u32 iter;
1077 
1078 	for (iter = 0; iter < net_cat_len; iter += 2) {
1079 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
1080 					     get_unaligned_be16(&net_cat[iter]),
1081 					     GFP_ATOMIC);
1082 		if (ret_val != 0)
1083 			return ret_val;
1084 	}
1085 
1086 	return 0;
1087 }
1088 
1089 /**
1090  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
1091  * @doi_def: the DOI definition
1092  * @rngcat: category list
1093  * @rngcat_len: length of the category list in bytes
1094  *
1095  * Description:
1096  * Checks the given categories against the given DOI definition and returns a
1097  * negative value if any of the categories do not have a valid mapping and a
1098  * zero value if all of the categories are valid.
1099  *
1100  */
1101 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1102 				      const unsigned char *rngcat,
1103 				      u32 rngcat_len)
1104 {
1105 	u16 cat_high;
1106 	u16 cat_low;
1107 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1108 	u32 iter;
1109 
1110 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1111 		return -EFAULT;
1112 
1113 	for (iter = 0; iter < rngcat_len; iter += 4) {
1114 		cat_high = get_unaligned_be16(&rngcat[iter]);
1115 		if ((iter + 4) <= rngcat_len)
1116 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1117 		else
1118 			cat_low = 0;
1119 
1120 		if (cat_high > cat_prev)
1121 			return -EFAULT;
1122 
1123 		cat_prev = cat_low;
1124 	}
1125 
1126 	return 0;
1127 }
1128 
1129 /**
1130  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1131  * @doi_def: the DOI definition
1132  * @secattr: the security attributes
1133  * @net_cat: the zero'd out category list in network/CIPSO format
1134  * @net_cat_len: the length of the CIPSO category list in bytes
1135  *
1136  * Description:
1137  * Perform a label mapping to translate a local MLS category bitmap to the
1138  * correct CIPSO category list using the given DOI definition.   Returns the
1139  * size in bytes of the network category bitmap on success, negative values
1140  * otherwise.
1141  *
1142  */
1143 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1144 				     const struct netlbl_lsm_secattr *secattr,
1145 				     unsigned char *net_cat,
1146 				     u32 net_cat_len)
1147 {
1148 	int iter = -1;
1149 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1150 	u32 array_cnt = 0;
1151 	u32 cat_size = 0;
1152 
1153 	/* make sure we don't overflow the 'array[]' variable */
1154 	if (net_cat_len >
1155 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1156 		return -ENOSPC;
1157 
1158 	for (;;) {
1159 		iter = netlbl_catmap_walk(secattr->attr.mls.cat, iter + 1);
1160 		if (iter < 0)
1161 			break;
1162 		cat_size += (iter == 0 ? 0 : sizeof(u16));
1163 		if (cat_size > net_cat_len)
1164 			return -ENOSPC;
1165 		array[array_cnt++] = iter;
1166 
1167 		iter = netlbl_catmap_walkrng(secattr->attr.mls.cat, iter);
1168 		if (iter < 0)
1169 			return -EFAULT;
1170 		cat_size += sizeof(u16);
1171 		if (cat_size > net_cat_len)
1172 			return -ENOSPC;
1173 		array[array_cnt++] = iter;
1174 	}
1175 
1176 	for (iter = 0; array_cnt > 0;) {
1177 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1178 		iter += 2;
1179 		array_cnt--;
1180 		if (array[array_cnt] != 0) {
1181 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1182 			iter += 2;
1183 		}
1184 	}
1185 
1186 	return cat_size;
1187 }
1188 
1189 /**
1190  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1191  * @doi_def: the DOI definition
1192  * @net_cat: the category list in network/CIPSO format
1193  * @net_cat_len: the length of the CIPSO bitmap in bytes
1194  * @secattr: the security attributes
1195  *
1196  * Description:
1197  * Perform a label mapping to translate a CIPSO category list to the correct
1198  * local MLS category bitmap using the given DOI definition.  Returns zero on
1199  * success, negative values on failure.
1200  *
1201  */
1202 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1203 				     const unsigned char *net_cat,
1204 				     u32 net_cat_len,
1205 				     struct netlbl_lsm_secattr *secattr)
1206 {
1207 	int ret_val;
1208 	u32 net_iter;
1209 	u16 cat_low;
1210 	u16 cat_high;
1211 
1212 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1213 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1214 		if ((net_iter + 4) <= net_cat_len)
1215 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1216 		else
1217 			cat_low = 0;
1218 
1219 		ret_val = netlbl_catmap_setrng(&secattr->attr.mls.cat,
1220 					       cat_low,
1221 					       cat_high,
1222 					       GFP_ATOMIC);
1223 		if (ret_val != 0)
1224 			return ret_val;
1225 	}
1226 
1227 	return 0;
1228 }
1229 
1230 /*
1231  * Protocol Handling Functions
1232  */
1233 
1234 /**
1235  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1236  * @doi_def: the DOI definition
1237  * @len: the total tag length in bytes, not including this header
1238  * @buf: the CIPSO option buffer
1239  *
1240  * Description:
1241  * Write a CIPSO header into the beginning of @buffer.
1242  *
1243  */
1244 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1245 				unsigned char *buf,
1246 				u32 len)
1247 {
1248 	buf[0] = IPOPT_CIPSO;
1249 	buf[1] = CIPSO_V4_HDR_LEN + len;
1250 	*(__be32 *)&buf[2] = htonl(doi_def->doi);
1251 }
1252 
1253 /**
1254  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1255  * @doi_def: the DOI definition
1256  * @secattr: the security attributes
1257  * @buffer: the option buffer
1258  * @buffer_len: length of buffer in bytes
1259  *
1260  * Description:
1261  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1262  * actual buffer length may be larger than the indicated size due to
1263  * translation between host and network category bitmaps.  Returns the size of
1264  * the tag on success, negative values on failure.
1265  *
1266  */
1267 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1268 			       const struct netlbl_lsm_secattr *secattr,
1269 			       unsigned char *buffer,
1270 			       u32 buffer_len)
1271 {
1272 	int ret_val;
1273 	u32 tag_len;
1274 	u32 level;
1275 
1276 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1277 		return -EPERM;
1278 
1279 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1280 					secattr->attr.mls.lvl,
1281 					&level);
1282 	if (ret_val != 0)
1283 		return ret_val;
1284 
1285 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1286 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1287 						    secattr,
1288 						    &buffer[4],
1289 						    buffer_len - 4);
1290 		if (ret_val < 0)
1291 			return ret_val;
1292 
1293 		/* This will send packets using the "optimized" format when
1294 		 * possible as specified in  section 3.4.2.6 of the
1295 		 * CIPSO draft. */
1296 		if (cipso_v4_rbm_optfmt && ret_val > 0 && ret_val <= 10)
1297 			tag_len = 14;
1298 		else
1299 			tag_len = 4 + ret_val;
1300 	} else
1301 		tag_len = 4;
1302 
1303 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1304 	buffer[1] = tag_len;
1305 	buffer[3] = level;
1306 
1307 	return tag_len;
1308 }
1309 
1310 /**
1311  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1312  * @doi_def: the DOI definition
1313  * @tag: the CIPSO tag
1314  * @secattr: the security attributes
1315  *
1316  * Description:
1317  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1318  * attributes in @secattr.  Return zero on success, negatives values on
1319  * failure.
1320  *
1321  */
1322 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1323 				 const unsigned char *tag,
1324 				 struct netlbl_lsm_secattr *secattr)
1325 {
1326 	int ret_val;
1327 	u8 tag_len = tag[1];
1328 	u32 level;
1329 
1330 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1331 	if (ret_val != 0)
1332 		return ret_val;
1333 	secattr->attr.mls.lvl = level;
1334 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1335 
1336 	if (tag_len > 4) {
1337 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1338 						    &tag[4],
1339 						    tag_len - 4,
1340 						    secattr);
1341 		if (ret_val != 0) {
1342 			netlbl_catmap_free(secattr->attr.mls.cat);
1343 			return ret_val;
1344 		}
1345 
1346 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1347 	}
1348 
1349 	return 0;
1350 }
1351 
1352 /**
1353  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1354  * @doi_def: the DOI definition
1355  * @secattr: the security attributes
1356  * @buffer: the option buffer
1357  * @buffer_len: length of buffer in bytes
1358  *
1359  * Description:
1360  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1361  * size of the tag on success, negative values on failure.
1362  *
1363  */
1364 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1365 				const struct netlbl_lsm_secattr *secattr,
1366 				unsigned char *buffer,
1367 				u32 buffer_len)
1368 {
1369 	int ret_val;
1370 	u32 tag_len;
1371 	u32 level;
1372 
1373 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1374 		return -EPERM;
1375 
1376 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1377 					secattr->attr.mls.lvl,
1378 					&level);
1379 	if (ret_val != 0)
1380 		return ret_val;
1381 
1382 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1383 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1384 						     secattr,
1385 						     &buffer[4],
1386 						     buffer_len - 4);
1387 		if (ret_val < 0)
1388 			return ret_val;
1389 
1390 		tag_len = 4 + ret_val;
1391 	} else
1392 		tag_len = 4;
1393 
1394 	buffer[0] = CIPSO_V4_TAG_ENUM;
1395 	buffer[1] = tag_len;
1396 	buffer[3] = level;
1397 
1398 	return tag_len;
1399 }
1400 
1401 /**
1402  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1403  * @doi_def: the DOI definition
1404  * @tag: the CIPSO tag
1405  * @secattr: the security attributes
1406  *
1407  * Description:
1408  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1409  * attributes in @secattr.  Return zero on success, negatives values on
1410  * failure.
1411  *
1412  */
1413 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1414 				  const unsigned char *tag,
1415 				  struct netlbl_lsm_secattr *secattr)
1416 {
1417 	int ret_val;
1418 	u8 tag_len = tag[1];
1419 	u32 level;
1420 
1421 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1422 	if (ret_val != 0)
1423 		return ret_val;
1424 	secattr->attr.mls.lvl = level;
1425 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1426 
1427 	if (tag_len > 4) {
1428 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1429 						     &tag[4],
1430 						     tag_len - 4,
1431 						     secattr);
1432 		if (ret_val != 0) {
1433 			netlbl_catmap_free(secattr->attr.mls.cat);
1434 			return ret_val;
1435 		}
1436 
1437 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1438 	}
1439 
1440 	return 0;
1441 }
1442 
1443 /**
1444  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1445  * @doi_def: the DOI definition
1446  * @secattr: the security attributes
1447  * @buffer: the option buffer
1448  * @buffer_len: length of buffer in bytes
1449  *
1450  * Description:
1451  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1452  * size of the tag on success, negative values on failure.
1453  *
1454  */
1455 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1456 			       const struct netlbl_lsm_secattr *secattr,
1457 			       unsigned char *buffer,
1458 			       u32 buffer_len)
1459 {
1460 	int ret_val;
1461 	u32 tag_len;
1462 	u32 level;
1463 
1464 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1465 		return -EPERM;
1466 
1467 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1468 					secattr->attr.mls.lvl,
1469 					&level);
1470 	if (ret_val != 0)
1471 		return ret_val;
1472 
1473 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1474 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1475 						    secattr,
1476 						    &buffer[4],
1477 						    buffer_len - 4);
1478 		if (ret_val < 0)
1479 			return ret_val;
1480 
1481 		tag_len = 4 + ret_val;
1482 	} else
1483 		tag_len = 4;
1484 
1485 	buffer[0] = CIPSO_V4_TAG_RANGE;
1486 	buffer[1] = tag_len;
1487 	buffer[3] = level;
1488 
1489 	return tag_len;
1490 }
1491 
1492 /**
1493  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1494  * @doi_def: the DOI definition
1495  * @tag: the CIPSO tag
1496  * @secattr: the security attributes
1497  *
1498  * Description:
1499  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1500  * in @secattr.  Return zero on success, negatives values on failure.
1501  *
1502  */
1503 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1504 				 const unsigned char *tag,
1505 				 struct netlbl_lsm_secattr *secattr)
1506 {
1507 	int ret_val;
1508 	u8 tag_len = tag[1];
1509 	u32 level;
1510 
1511 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1512 	if (ret_val != 0)
1513 		return ret_val;
1514 	secattr->attr.mls.lvl = level;
1515 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1516 
1517 	if (tag_len > 4) {
1518 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1519 						    &tag[4],
1520 						    tag_len - 4,
1521 						    secattr);
1522 		if (ret_val != 0) {
1523 			netlbl_catmap_free(secattr->attr.mls.cat);
1524 			return ret_val;
1525 		}
1526 
1527 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1528 	}
1529 
1530 	return 0;
1531 }
1532 
1533 /**
1534  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1535  * @doi_def: the DOI definition
1536  * @secattr: the security attributes
1537  * @buffer: the option buffer
1538  * @buffer_len: length of buffer in bytes
1539  *
1540  * Description:
1541  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1542  * on success, negative values on failure.
1543  *
1544  */
1545 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1546 			       const struct netlbl_lsm_secattr *secattr,
1547 			       unsigned char *buffer,
1548 			       u32 buffer_len)
1549 {
1550 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1551 		return -EPERM;
1552 
1553 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1554 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1555 	*(u32 *)&buffer[2] = secattr->attr.secid;
1556 
1557 	return CIPSO_V4_TAG_LOC_BLEN;
1558 }
1559 
1560 /**
1561  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1562  * @doi_def: the DOI definition
1563  * @tag: the CIPSO tag
1564  * @secattr: the security attributes
1565  *
1566  * Description:
1567  * Parse a CIPSO local tag and return the security attributes in @secattr.
1568  * Return zero on success, negatives values on failure.
1569  *
1570  */
1571 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1572 				 const unsigned char *tag,
1573 				 struct netlbl_lsm_secattr *secattr)
1574 {
1575 	secattr->attr.secid = *(u32 *)&tag[2];
1576 	secattr->flags |= NETLBL_SECATTR_SECID;
1577 
1578 	return 0;
1579 }
1580 
1581 /**
1582  * cipso_v4_validate - Validate a CIPSO option
1583  * @option: the start of the option, on error it is set to point to the error
1584  *
1585  * Description:
1586  * This routine is called to validate a CIPSO option, it checks all of the
1587  * fields to ensure that they are at least valid, see the draft snippet below
1588  * for details.  If the option is valid then a zero value is returned and
1589  * the value of @option is unchanged.  If the option is invalid then a
1590  * non-zero value is returned and @option is adjusted to point to the
1591  * offending portion of the option.  From the IETF draft ...
1592  *
1593  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1594  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1595  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1596  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1597  *   that is unrecognized."
1598  *
1599  */
1600 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1601 {
1602 	unsigned char *opt = *option;
1603 	unsigned char *tag;
1604 	unsigned char opt_iter;
1605 	unsigned char err_offset = 0;
1606 	u8 opt_len;
1607 	u8 tag_len;
1608 	struct cipso_v4_doi *doi_def = NULL;
1609 	u32 tag_iter;
1610 
1611 	/* caller already checks for length values that are too large */
1612 	opt_len = opt[1];
1613 	if (opt_len < 8) {
1614 		err_offset = 1;
1615 		goto validate_return;
1616 	}
1617 
1618 	rcu_read_lock();
1619 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1620 	if (doi_def == NULL) {
1621 		err_offset = 2;
1622 		goto validate_return_locked;
1623 	}
1624 
1625 	opt_iter = CIPSO_V4_HDR_LEN;
1626 	tag = opt + opt_iter;
1627 	while (opt_iter < opt_len) {
1628 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1629 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1630 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1631 				err_offset = opt_iter;
1632 				goto validate_return_locked;
1633 			}
1634 
1635 		tag_len = tag[1];
1636 		if (tag_len > (opt_len - opt_iter)) {
1637 			err_offset = opt_iter + 1;
1638 			goto validate_return_locked;
1639 		}
1640 
1641 		switch (tag[0]) {
1642 		case CIPSO_V4_TAG_RBITMAP:
1643 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1644 				err_offset = opt_iter + 1;
1645 				goto validate_return_locked;
1646 			}
1647 
1648 			/* We are already going to do all the verification
1649 			 * necessary at the socket layer so from our point of
1650 			 * view it is safe to turn these checks off (and less
1651 			 * work), however, the CIPSO draft says we should do
1652 			 * all the CIPSO validations here but it doesn't
1653 			 * really specify _exactly_ what we need to validate
1654 			 * ... so, just make it a sysctl tunable. */
1655 			if (cipso_v4_rbm_strictvalid) {
1656 				if (cipso_v4_map_lvl_valid(doi_def,
1657 							   tag[3]) < 0) {
1658 					err_offset = opt_iter + 3;
1659 					goto validate_return_locked;
1660 				}
1661 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1662 				    cipso_v4_map_cat_rbm_valid(doi_def,
1663 							    &tag[4],
1664 							    tag_len - 4) < 0) {
1665 					err_offset = opt_iter + 4;
1666 					goto validate_return_locked;
1667 				}
1668 			}
1669 			break;
1670 		case CIPSO_V4_TAG_ENUM:
1671 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1672 				err_offset = opt_iter + 1;
1673 				goto validate_return_locked;
1674 			}
1675 
1676 			if (cipso_v4_map_lvl_valid(doi_def,
1677 						   tag[3]) < 0) {
1678 				err_offset = opt_iter + 3;
1679 				goto validate_return_locked;
1680 			}
1681 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1682 			    cipso_v4_map_cat_enum_valid(doi_def,
1683 							&tag[4],
1684 							tag_len - 4) < 0) {
1685 				err_offset = opt_iter + 4;
1686 				goto validate_return_locked;
1687 			}
1688 			break;
1689 		case CIPSO_V4_TAG_RANGE:
1690 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1691 				err_offset = opt_iter + 1;
1692 				goto validate_return_locked;
1693 			}
1694 
1695 			if (cipso_v4_map_lvl_valid(doi_def,
1696 						   tag[3]) < 0) {
1697 				err_offset = opt_iter + 3;
1698 				goto validate_return_locked;
1699 			}
1700 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1701 			    cipso_v4_map_cat_rng_valid(doi_def,
1702 						       &tag[4],
1703 						       tag_len - 4) < 0) {
1704 				err_offset = opt_iter + 4;
1705 				goto validate_return_locked;
1706 			}
1707 			break;
1708 		case CIPSO_V4_TAG_LOCAL:
1709 			/* This is a non-standard tag that we only allow for
1710 			 * local connections, so if the incoming interface is
1711 			 * not the loopback device drop the packet. Further,
1712 			 * there is no legitimate reason for setting this from
1713 			 * userspace so reject it if skb is NULL. */
1714 			if (skb == NULL || !(skb->dev->flags & IFF_LOOPBACK)) {
1715 				err_offset = opt_iter;
1716 				goto validate_return_locked;
1717 			}
1718 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1719 				err_offset = opt_iter + 1;
1720 				goto validate_return_locked;
1721 			}
1722 			break;
1723 		default:
1724 			err_offset = opt_iter;
1725 			goto validate_return_locked;
1726 		}
1727 
1728 		tag += tag_len;
1729 		opt_iter += tag_len;
1730 	}
1731 
1732 validate_return_locked:
1733 	rcu_read_unlock();
1734 validate_return:
1735 	*option = opt + err_offset;
1736 	return err_offset;
1737 }
1738 
1739 /**
1740  * cipso_v4_error - Send the correct response for a bad packet
1741  * @skb: the packet
1742  * @error: the error code
1743  * @gateway: CIPSO gateway flag
1744  *
1745  * Description:
1746  * Based on the error code given in @error, send an ICMP error message back to
1747  * the originating host.  From the IETF draft ...
1748  *
1749  *  "If the contents of the CIPSO [option] are valid but the security label is
1750  *   outside of the configured host or port label range, the datagram is
1751  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1752  *   returned.  The code field of the ICMP is set to 'communication with
1753  *   destination network administratively prohibited' (code 9) or to
1754  *   'communication with destination host administratively prohibited'
1755  *   (code 10).  The value of the code is dependent on whether the originator
1756  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1757  *   recipient of the ICMP message MUST be able to handle either value.  The
1758  *   same procedure is performed if a CIPSO [option] can not be added to an
1759  *   IP packet because it is too large to fit in the IP options area."
1760  *
1761  *  "If the error is triggered by receipt of an ICMP message, the message is
1762  *   discarded and no response is permitted (consistent with general ICMP
1763  *   processing rules)."
1764  *
1765  */
1766 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1767 {
1768 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1769 		return;
1770 
1771 	if (gateway)
1772 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0);
1773 	else
1774 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0);
1775 }
1776 
1777 /**
1778  * cipso_v4_genopt - Generate a CIPSO option
1779  * @buf: the option buffer
1780  * @buf_len: the size of opt_buf
1781  * @doi_def: the CIPSO DOI to use
1782  * @secattr: the security attributes
1783  *
1784  * Description:
1785  * Generate a CIPSO option using the DOI definition and security attributes
1786  * passed to the function.  Returns the length of the option on success and
1787  * negative values on failure.
1788  *
1789  */
1790 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1791 			   const struct cipso_v4_doi *doi_def,
1792 			   const struct netlbl_lsm_secattr *secattr)
1793 {
1794 	int ret_val;
1795 	u32 iter;
1796 
1797 	if (buf_len <= CIPSO_V4_HDR_LEN)
1798 		return -ENOSPC;
1799 
1800 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1801 	 * really a good assumption to make but since we only support the MAC
1802 	 * tags right now it is a safe assumption. */
1803 	iter = 0;
1804 	do {
1805 		memset(buf, 0, buf_len);
1806 		switch (doi_def->tags[iter]) {
1807 		case CIPSO_V4_TAG_RBITMAP:
1808 			ret_val = cipso_v4_gentag_rbm(doi_def,
1809 						   secattr,
1810 						   &buf[CIPSO_V4_HDR_LEN],
1811 						   buf_len - CIPSO_V4_HDR_LEN);
1812 			break;
1813 		case CIPSO_V4_TAG_ENUM:
1814 			ret_val = cipso_v4_gentag_enum(doi_def,
1815 						   secattr,
1816 						   &buf[CIPSO_V4_HDR_LEN],
1817 						   buf_len - CIPSO_V4_HDR_LEN);
1818 			break;
1819 		case CIPSO_V4_TAG_RANGE:
1820 			ret_val = cipso_v4_gentag_rng(doi_def,
1821 						   secattr,
1822 						   &buf[CIPSO_V4_HDR_LEN],
1823 						   buf_len - CIPSO_V4_HDR_LEN);
1824 			break;
1825 		case CIPSO_V4_TAG_LOCAL:
1826 			ret_val = cipso_v4_gentag_loc(doi_def,
1827 						   secattr,
1828 						   &buf[CIPSO_V4_HDR_LEN],
1829 						   buf_len - CIPSO_V4_HDR_LEN);
1830 			break;
1831 		default:
1832 			return -EPERM;
1833 		}
1834 
1835 		iter++;
1836 	} while (ret_val < 0 &&
1837 		 iter < CIPSO_V4_TAG_MAXCNT &&
1838 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1839 	if (ret_val < 0)
1840 		return ret_val;
1841 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1842 	return CIPSO_V4_HDR_LEN + ret_val;
1843 }
1844 
1845 /**
1846  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1847  * @sk: the socket
1848  * @doi_def: the CIPSO DOI to use
1849  * @secattr: the specific security attributes of the socket
1850  *
1851  * Description:
1852  * Set the CIPSO option on the given socket using the DOI definition and
1853  * security attributes passed to the function.  This function requires
1854  * exclusive access to @sk, which means it either needs to be in the
1855  * process of being created or locked.  Returns zero on success and negative
1856  * values on failure.
1857  *
1858  */
1859 int cipso_v4_sock_setattr(struct sock *sk,
1860 			  const struct cipso_v4_doi *doi_def,
1861 			  const struct netlbl_lsm_secattr *secattr)
1862 {
1863 	int ret_val = -EPERM;
1864 	unsigned char *buf = NULL;
1865 	u32 buf_len;
1866 	u32 opt_len;
1867 	struct ip_options_rcu *old, *opt = NULL;
1868 	struct inet_sock *sk_inet;
1869 	struct inet_connection_sock *sk_conn;
1870 
1871 	/* In the case of sock_create_lite(), the sock->sk field is not
1872 	 * defined yet but it is not a problem as the only users of these
1873 	 * "lite" PF_INET sockets are functions which do an accept() call
1874 	 * afterwards so we will label the socket as part of the accept(). */
1875 	if (sk == NULL)
1876 		return 0;
1877 
1878 	/* We allocate the maximum CIPSO option size here so we are probably
1879 	 * being a little wasteful, but it makes our life _much_ easier later
1880 	 * on and after all we are only talking about 40 bytes. */
1881 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1882 	buf = kmalloc(buf_len, GFP_ATOMIC);
1883 	if (buf == NULL) {
1884 		ret_val = -ENOMEM;
1885 		goto socket_setattr_failure;
1886 	}
1887 
1888 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1889 	if (ret_val < 0)
1890 		goto socket_setattr_failure;
1891 	buf_len = ret_val;
1892 
1893 	/* We can't use ip_options_get() directly because it makes a call to
1894 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1895 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1896 	 * set the IPOPT_CIPSO option. */
1897 	opt_len = (buf_len + 3) & ~3;
1898 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1899 	if (opt == NULL) {
1900 		ret_val = -ENOMEM;
1901 		goto socket_setattr_failure;
1902 	}
1903 	memcpy(opt->opt.__data, buf, buf_len);
1904 	opt->opt.optlen = opt_len;
1905 	opt->opt.cipso = sizeof(struct iphdr);
1906 	kfree(buf);
1907 	buf = NULL;
1908 
1909 	sk_inet = inet_sk(sk);
1910 
1911 	old = rcu_dereference_protected(sk_inet->inet_opt, sock_owned_by_user(sk));
1912 	if (sk_inet->is_icsk) {
1913 		sk_conn = inet_csk(sk);
1914 		if (old)
1915 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1916 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1917 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1918 	}
1919 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1920 	if (old)
1921 		kfree_rcu(old, rcu);
1922 
1923 	return 0;
1924 
1925 socket_setattr_failure:
1926 	kfree(buf);
1927 	kfree(opt);
1928 	return ret_val;
1929 }
1930 
1931 /**
1932  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1933  * @req: the connection request socket
1934  * @doi_def: the CIPSO DOI to use
1935  * @secattr: the specific security attributes of the socket
1936  *
1937  * Description:
1938  * Set the CIPSO option on the given socket using the DOI definition and
1939  * security attributes passed to the function.  Returns zero on success and
1940  * negative values on failure.
1941  *
1942  */
1943 int cipso_v4_req_setattr(struct request_sock *req,
1944 			 const struct cipso_v4_doi *doi_def,
1945 			 const struct netlbl_lsm_secattr *secattr)
1946 {
1947 	int ret_val = -EPERM;
1948 	unsigned char *buf = NULL;
1949 	u32 buf_len;
1950 	u32 opt_len;
1951 	struct ip_options_rcu *opt = NULL;
1952 	struct inet_request_sock *req_inet;
1953 
1954 	/* We allocate the maximum CIPSO option size here so we are probably
1955 	 * being a little wasteful, but it makes our life _much_ easier later
1956 	 * on and after all we are only talking about 40 bytes. */
1957 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1958 	buf = kmalloc(buf_len, GFP_ATOMIC);
1959 	if (buf == NULL) {
1960 		ret_val = -ENOMEM;
1961 		goto req_setattr_failure;
1962 	}
1963 
1964 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1965 	if (ret_val < 0)
1966 		goto req_setattr_failure;
1967 	buf_len = ret_val;
1968 
1969 	/* We can't use ip_options_get() directly because it makes a call to
1970 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1971 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1972 	 * set the IPOPT_CIPSO option. */
1973 	opt_len = (buf_len + 3) & ~3;
1974 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1975 	if (opt == NULL) {
1976 		ret_val = -ENOMEM;
1977 		goto req_setattr_failure;
1978 	}
1979 	memcpy(opt->opt.__data, buf, buf_len);
1980 	opt->opt.optlen = opt_len;
1981 	opt->opt.cipso = sizeof(struct iphdr);
1982 	kfree(buf);
1983 	buf = NULL;
1984 
1985 	req_inet = inet_rsk(req);
1986 	opt = xchg(&req_inet->opt, opt);
1987 	if (opt)
1988 		kfree_rcu(opt, rcu);
1989 
1990 	return 0;
1991 
1992 req_setattr_failure:
1993 	kfree(buf);
1994 	kfree(opt);
1995 	return ret_val;
1996 }
1997 
1998 /**
1999  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
2000  * @opt_ptr: IP option pointer
2001  *
2002  * Description:
2003  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
2004  * adjustments to the IP option structure.  Returns zero on success, negative
2005  * values on failure.
2006  *
2007  */
2008 static int cipso_v4_delopt(struct ip_options_rcu **opt_ptr)
2009 {
2010 	int hdr_delta = 0;
2011 	struct ip_options_rcu *opt = *opt_ptr;
2012 
2013 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
2014 		u8 cipso_len;
2015 		u8 cipso_off;
2016 		unsigned char *cipso_ptr;
2017 		int iter;
2018 		int optlen_new;
2019 
2020 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
2021 		cipso_ptr = &opt->opt.__data[cipso_off];
2022 		cipso_len = cipso_ptr[1];
2023 
2024 		if (opt->opt.srr > opt->opt.cipso)
2025 			opt->opt.srr -= cipso_len;
2026 		if (opt->opt.rr > opt->opt.cipso)
2027 			opt->opt.rr -= cipso_len;
2028 		if (opt->opt.ts > opt->opt.cipso)
2029 			opt->opt.ts -= cipso_len;
2030 		if (opt->opt.router_alert > opt->opt.cipso)
2031 			opt->opt.router_alert -= cipso_len;
2032 		opt->opt.cipso = 0;
2033 
2034 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2035 			opt->opt.optlen - cipso_off - cipso_len);
2036 
2037 		/* determining the new total option length is tricky because of
2038 		 * the padding necessary, the only thing i can think to do at
2039 		 * this point is walk the options one-by-one, skipping the
2040 		 * padding at the end to determine the actual option size and
2041 		 * from there we can determine the new total option length */
2042 		iter = 0;
2043 		optlen_new = 0;
2044 		while (iter < opt->opt.optlen)
2045 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2046 				iter += opt->opt.__data[iter + 1];
2047 				optlen_new = iter;
2048 			} else
2049 				iter++;
2050 		hdr_delta = opt->opt.optlen;
2051 		opt->opt.optlen = (optlen_new + 3) & ~3;
2052 		hdr_delta -= opt->opt.optlen;
2053 	} else {
2054 		/* only the cipso option was present on the socket so we can
2055 		 * remove the entire option struct */
2056 		*opt_ptr = NULL;
2057 		hdr_delta = opt->opt.optlen;
2058 		kfree_rcu(opt, rcu);
2059 	}
2060 
2061 	return hdr_delta;
2062 }
2063 
2064 /**
2065  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2066  * @sk: the socket
2067  *
2068  * Description:
2069  * Removes the CIPSO option from a socket, if present.
2070  *
2071  */
2072 void cipso_v4_sock_delattr(struct sock *sk)
2073 {
2074 	int hdr_delta;
2075 	struct ip_options_rcu *opt;
2076 	struct inet_sock *sk_inet;
2077 
2078 	sk_inet = inet_sk(sk);
2079 	opt = rcu_dereference_protected(sk_inet->inet_opt, 1);
2080 	if (opt == NULL || opt->opt.cipso == 0)
2081 		return;
2082 
2083 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2084 	if (sk_inet->is_icsk && hdr_delta > 0) {
2085 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2086 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2087 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2088 	}
2089 }
2090 
2091 /**
2092  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2093  * @reg: the request socket
2094  *
2095  * Description:
2096  * Removes the CIPSO option from a request socket, if present.
2097  *
2098  */
2099 void cipso_v4_req_delattr(struct request_sock *req)
2100 {
2101 	struct ip_options_rcu *opt;
2102 	struct inet_request_sock *req_inet;
2103 
2104 	req_inet = inet_rsk(req);
2105 	opt = req_inet->opt;
2106 	if (opt == NULL || opt->opt.cipso == 0)
2107 		return;
2108 
2109 	cipso_v4_delopt(&req_inet->opt);
2110 }
2111 
2112 /**
2113  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2114  * @cipso: the CIPSO v4 option
2115  * @secattr: the security attributes
2116  *
2117  * Description:
2118  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2119  * on success and negative values on failure.
2120  *
2121  */
2122 static int cipso_v4_getattr(const unsigned char *cipso,
2123 			    struct netlbl_lsm_secattr *secattr)
2124 {
2125 	int ret_val = -ENOMSG;
2126 	u32 doi;
2127 	struct cipso_v4_doi *doi_def;
2128 
2129 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2130 		return 0;
2131 
2132 	doi = get_unaligned_be32(&cipso[2]);
2133 	rcu_read_lock();
2134 	doi_def = cipso_v4_doi_search(doi);
2135 	if (doi_def == NULL)
2136 		goto getattr_return;
2137 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2138 	 * really a good assumption to make but since we only support the MAC
2139 	 * tags right now it is a safe assumption. */
2140 	switch (cipso[6]) {
2141 	case CIPSO_V4_TAG_RBITMAP:
2142 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2143 		break;
2144 	case CIPSO_V4_TAG_ENUM:
2145 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2146 		break;
2147 	case CIPSO_V4_TAG_RANGE:
2148 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2149 		break;
2150 	case CIPSO_V4_TAG_LOCAL:
2151 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2152 		break;
2153 	}
2154 	if (ret_val == 0)
2155 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2156 
2157 getattr_return:
2158 	rcu_read_unlock();
2159 	return ret_val;
2160 }
2161 
2162 /**
2163  * cipso_v4_sock_getattr - Get the security attributes from a sock
2164  * @sk: the sock
2165  * @secattr: the security attributes
2166  *
2167  * Description:
2168  * Query @sk to see if there is a CIPSO option attached to the sock and if
2169  * there is return the CIPSO security attributes in @secattr.  This function
2170  * requires that @sk be locked, or privately held, but it does not do any
2171  * locking itself.  Returns zero on success and negative values on failure.
2172  *
2173  */
2174 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2175 {
2176 	struct ip_options_rcu *opt;
2177 	int res = -ENOMSG;
2178 
2179 	rcu_read_lock();
2180 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2181 	if (opt && opt->opt.cipso)
2182 		res = cipso_v4_getattr(opt->opt.__data +
2183 						opt->opt.cipso -
2184 						sizeof(struct iphdr),
2185 				       secattr);
2186 	rcu_read_unlock();
2187 	return res;
2188 }
2189 
2190 /**
2191  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2192  * @skb: the packet
2193  * @secattr: the security attributes
2194  *
2195  * Description:
2196  * Set the CIPSO option on the given packet based on the security attributes.
2197  * Returns a pointer to the IP header on success and NULL on failure.
2198  *
2199  */
2200 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2201 			    const struct cipso_v4_doi *doi_def,
2202 			    const struct netlbl_lsm_secattr *secattr)
2203 {
2204 	int ret_val;
2205 	struct iphdr *iph;
2206 	struct ip_options *opt = &IPCB(skb)->opt;
2207 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2208 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2209 	u32 opt_len;
2210 	int len_delta;
2211 
2212 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2213 	if (ret_val < 0)
2214 		return ret_val;
2215 	buf_len = ret_val;
2216 	opt_len = (buf_len + 3) & ~3;
2217 
2218 	/* we overwrite any existing options to ensure that we have enough
2219 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2220 	 * that the security label is applied to the packet - we do the same
2221 	 * thing when using the socket options and it hasn't caused a problem,
2222 	 * if we need to we can always revisit this choice later */
2223 
2224 	len_delta = opt_len - opt->optlen;
2225 	/* if we don't ensure enough headroom we could panic on the skb_push()
2226 	 * call below so make sure we have enough, we are also "mangling" the
2227 	 * packet so we should probably do a copy-on-write call anyway */
2228 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2229 	if (ret_val < 0)
2230 		return ret_val;
2231 
2232 	if (len_delta > 0) {
2233 		/* we assume that the header + opt->optlen have already been
2234 		 * "pushed" in ip_options_build() or similar */
2235 		iph = ip_hdr(skb);
2236 		skb_push(skb, len_delta);
2237 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2238 		skb_reset_network_header(skb);
2239 		iph = ip_hdr(skb);
2240 	} else if (len_delta < 0) {
2241 		iph = ip_hdr(skb);
2242 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2243 	} else
2244 		iph = ip_hdr(skb);
2245 
2246 	if (opt->optlen > 0)
2247 		memset(opt, 0, sizeof(*opt));
2248 	opt->optlen = opt_len;
2249 	opt->cipso = sizeof(struct iphdr);
2250 	opt->is_changed = 1;
2251 
2252 	/* we have to do the following because we are being called from a
2253 	 * netfilter hook which means the packet already has had the header
2254 	 * fields populated and the checksum calculated - yes this means we
2255 	 * are doing more work than needed but we do it to keep the core
2256 	 * stack clean and tidy */
2257 	memcpy(iph + 1, buf, buf_len);
2258 	if (opt_len > buf_len)
2259 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2260 	if (len_delta != 0) {
2261 		iph->ihl = 5 + (opt_len >> 2);
2262 		iph->tot_len = htons(skb->len);
2263 	}
2264 	ip_send_check(iph);
2265 
2266 	return 0;
2267 }
2268 
2269 /**
2270  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2271  * @skb: the packet
2272  *
2273  * Description:
2274  * Removes any and all CIPSO options from the given packet.  Returns zero on
2275  * success, negative values on failure.
2276  *
2277  */
2278 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2279 {
2280 	int ret_val;
2281 	struct iphdr *iph;
2282 	struct ip_options *opt = &IPCB(skb)->opt;
2283 	unsigned char *cipso_ptr;
2284 
2285 	if (opt->cipso == 0)
2286 		return 0;
2287 
2288 	/* since we are changing the packet we should make a copy */
2289 	ret_val = skb_cow(skb, skb_headroom(skb));
2290 	if (ret_val < 0)
2291 		return ret_val;
2292 
2293 	/* the easiest thing to do is just replace the cipso option with noop
2294 	 * options since we don't change the size of the packet, although we
2295 	 * still need to recalculate the checksum */
2296 
2297 	iph = ip_hdr(skb);
2298 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2299 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2300 	opt->cipso = 0;
2301 	opt->is_changed = 1;
2302 
2303 	ip_send_check(iph);
2304 
2305 	return 0;
2306 }
2307 
2308 /**
2309  * cipso_v4_skbuff_getattr - Get the security attributes from the CIPSO option
2310  * @skb: the packet
2311  * @secattr: the security attributes
2312  *
2313  * Description:
2314  * Parse the given packet's CIPSO option and return the security attributes.
2315  * Returns zero on success and negative values on failure.
2316  *
2317  */
2318 int cipso_v4_skbuff_getattr(const struct sk_buff *skb,
2319 			    struct netlbl_lsm_secattr *secattr)
2320 {
2321 	return cipso_v4_getattr(CIPSO_V4_OPTPTR(skb), secattr);
2322 }
2323 
2324 /*
2325  * Setup Functions
2326  */
2327 
2328 /**
2329  * cipso_v4_init - Initialize the CIPSO module
2330  *
2331  * Description:
2332  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2333  * and negative values on failure.
2334  *
2335  */
2336 static int __init cipso_v4_init(void)
2337 {
2338 	int ret_val;
2339 
2340 	ret_val = cipso_v4_cache_init();
2341 	if (ret_val != 0)
2342 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2343 		      ret_val);
2344 
2345 	return 0;
2346 }
2347 
2348 subsys_initcall(cipso_v4_init);
2349