xref: /openbmc/linux/net/ipv4/cipso_ipv4.c (revision b34e08d5)
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 <asm/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 struct cipso_v4_map_cache_entry {
76 	u32 hash;
77 	unsigned char *key;
78 	size_t key_len;
79 
80 	struct netlbl_lsm_cache *lsm_data;
81 
82 	u32 activity;
83 	struct list_head list;
84 };
85 static struct cipso_v4_map_cache_bkt *cipso_v4_cache = NULL;
86 
87 /* Restricted bitmap (tag #1) flags */
88 int cipso_v4_rbm_optfmt = 0;
89 int cipso_v4_rbm_strictvalid = 1;
90 
91 /*
92  * Protocol Constants
93  */
94 
95 /* Maximum size of the CIPSO IP option, derived from the fact that the maximum
96  * IPv4 header size is 60 bytes and the base IPv4 header is 20 bytes long. */
97 #define CIPSO_V4_OPT_LEN_MAX          40
98 
99 /* Length of the base CIPSO option, this includes the option type (1 byte), the
100  * option length (1 byte), and the DOI (4 bytes). */
101 #define CIPSO_V4_HDR_LEN              6
102 
103 /* Base length of the restrictive category bitmap tag (tag #1). */
104 #define CIPSO_V4_TAG_RBM_BLEN         4
105 
106 /* Base length of the enumerated category tag (tag #2). */
107 #define CIPSO_V4_TAG_ENUM_BLEN        4
108 
109 /* Base length of the ranged categories bitmap tag (tag #5). */
110 #define CIPSO_V4_TAG_RNG_BLEN         4
111 /* The maximum number of category ranges permitted in the ranged category tag
112  * (tag #5).  You may note that the IETF draft states that the maximum number
113  * of category ranges is 7, but if the low end of the last category range is
114  * zero then it is possible to fit 8 category ranges because the zero should
115  * be omitted. */
116 #define CIPSO_V4_TAG_RNG_CAT_MAX      8
117 
118 /* Base length of the local tag (non-standard tag).
119  *  Tag definition (may change between kernel versions)
120  *
121  * 0          8          16         24         32
122  * +----------+----------+----------+----------+
123  * | 10000000 | 00000110 | 32-bit secid value  |
124  * +----------+----------+----------+----------+
125  * | in (host byte order)|
126  * +----------+----------+
127  *
128  */
129 #define CIPSO_V4_TAG_LOC_BLEN         6
130 
131 /*
132  * Helper Functions
133  */
134 
135 /**
136  * cipso_v4_bitmap_walk - Walk a bitmap looking for a bit
137  * @bitmap: the bitmap
138  * @bitmap_len: length in bits
139  * @offset: starting offset
140  * @state: if non-zero, look for a set (1) bit else look for a cleared (0) bit
141  *
142  * Description:
143  * Starting at @offset, walk the bitmap from left to right until either the
144  * desired bit is found or we reach the end.  Return the bit offset, -1 if
145  * not found, or -2 if error.
146  */
147 static int cipso_v4_bitmap_walk(const unsigned char *bitmap,
148 				u32 bitmap_len,
149 				u32 offset,
150 				u8 state)
151 {
152 	u32 bit_spot;
153 	u32 byte_offset;
154 	unsigned char bitmask;
155 	unsigned char byte;
156 
157 	/* gcc always rounds to zero when doing integer division */
158 	byte_offset = offset / 8;
159 	byte = bitmap[byte_offset];
160 	bit_spot = offset;
161 	bitmask = 0x80 >> (offset % 8);
162 
163 	while (bit_spot < bitmap_len) {
164 		if ((state && (byte & bitmask) == bitmask) ||
165 		    (state == 0 && (byte & bitmask) == 0))
166 			return bit_spot;
167 
168 		bit_spot++;
169 		bitmask >>= 1;
170 		if (bitmask == 0) {
171 			byte = bitmap[++byte_offset];
172 			bitmask = 0x80;
173 		}
174 	}
175 
176 	return -1;
177 }
178 
179 /**
180  * cipso_v4_bitmap_setbit - Sets a single bit in a bitmap
181  * @bitmap: the bitmap
182  * @bit: the bit
183  * @state: if non-zero, set the bit (1) else clear the bit (0)
184  *
185  * Description:
186  * Set a single bit in the bitmask.  Returns zero on success, negative values
187  * on error.
188  */
189 static void cipso_v4_bitmap_setbit(unsigned char *bitmap,
190 				   u32 bit,
191 				   u8 state)
192 {
193 	u32 byte_spot;
194 	u8 bitmask;
195 
196 	/* gcc always rounds to zero when doing integer division */
197 	byte_spot = bit / 8;
198 	bitmask = 0x80 >> (bit % 8);
199 	if (state)
200 		bitmap[byte_spot] |= bitmask;
201 	else
202 		bitmap[byte_spot] &= ~bitmask;
203 }
204 
205 /**
206  * cipso_v4_cache_entry_free - Frees a cache entry
207  * @entry: the entry to free
208  *
209  * Description:
210  * This function frees the memory associated with a cache entry including the
211  * LSM cache data if there are no longer any users, i.e. reference count == 0.
212  *
213  */
214 static void cipso_v4_cache_entry_free(struct cipso_v4_map_cache_entry *entry)
215 {
216 	if (entry->lsm_data)
217 		netlbl_secattr_cache_free(entry->lsm_data);
218 	kfree(entry->key);
219 	kfree(entry);
220 }
221 
222 /**
223  * cipso_v4_map_cache_hash - Hashing function for the CIPSO cache
224  * @key: the hash key
225  * @key_len: the length of the key in bytes
226  *
227  * Description:
228  * The CIPSO tag hashing function.  Returns a 32-bit hash value.
229  *
230  */
231 static u32 cipso_v4_map_cache_hash(const unsigned char *key, u32 key_len)
232 {
233 	return jhash(key, key_len, 0);
234 }
235 
236 /*
237  * Label Mapping Cache Functions
238  */
239 
240 /**
241  * cipso_v4_cache_init - Initialize the CIPSO cache
242  *
243  * Description:
244  * Initializes the CIPSO label mapping cache, this function should be called
245  * before any of the other functions defined in this file.  Returns zero on
246  * success, negative values on error.
247  *
248  */
249 static int cipso_v4_cache_init(void)
250 {
251 	u32 iter;
252 
253 	cipso_v4_cache = kcalloc(CIPSO_V4_CACHE_BUCKETS,
254 				 sizeof(struct cipso_v4_map_cache_bkt),
255 				 GFP_KERNEL);
256 	if (cipso_v4_cache == NULL)
257 		return -ENOMEM;
258 
259 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
260 		spin_lock_init(&cipso_v4_cache[iter].lock);
261 		cipso_v4_cache[iter].size = 0;
262 		INIT_LIST_HEAD(&cipso_v4_cache[iter].list);
263 	}
264 
265 	return 0;
266 }
267 
268 /**
269  * cipso_v4_cache_invalidate - Invalidates the current CIPSO cache
270  *
271  * Description:
272  * Invalidates and frees any entries in the CIPSO cache.  Returns zero on
273  * success and negative values on failure.
274  *
275  */
276 void cipso_v4_cache_invalidate(void)
277 {
278 	struct cipso_v4_map_cache_entry *entry, *tmp_entry;
279 	u32 iter;
280 
281 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
282 		spin_lock_bh(&cipso_v4_cache[iter].lock);
283 		list_for_each_entry_safe(entry,
284 					 tmp_entry,
285 					 &cipso_v4_cache[iter].list, list) {
286 			list_del(&entry->list);
287 			cipso_v4_cache_entry_free(entry);
288 		}
289 		cipso_v4_cache[iter].size = 0;
290 		spin_unlock_bh(&cipso_v4_cache[iter].lock);
291 	}
292 }
293 
294 /**
295  * cipso_v4_cache_check - Check the CIPSO cache for a label mapping
296  * @key: the buffer to check
297  * @key_len: buffer length in bytes
298  * @secattr: the security attribute struct to use
299  *
300  * Description:
301  * This function checks the cache to see if a label mapping already exists for
302  * the given key.  If there is a match then the cache is adjusted and the
303  * @secattr struct is populated with the correct LSM security attributes.  The
304  * cache is adjusted in the following manner if the entry is not already the
305  * first in the cache bucket:
306  *
307  *  1. The cache entry's activity counter is incremented
308  *  2. The previous (higher ranking) entry's activity counter is decremented
309  *  3. If the difference between the two activity counters is geater than
310  *     CIPSO_V4_CACHE_REORDERLIMIT the two entries are swapped
311  *
312  * Returns zero on success, -ENOENT for a cache miss, and other negative values
313  * on error.
314  *
315  */
316 static int cipso_v4_cache_check(const unsigned char *key,
317 				u32 key_len,
318 				struct netlbl_lsm_secattr *secattr)
319 {
320 	u32 bkt;
321 	struct cipso_v4_map_cache_entry *entry;
322 	struct cipso_v4_map_cache_entry *prev_entry = NULL;
323 	u32 hash;
324 
325 	if (!cipso_v4_cache_enabled)
326 		return -ENOENT;
327 
328 	hash = cipso_v4_map_cache_hash(key, key_len);
329 	bkt = hash & (CIPSO_V4_CACHE_BUCKETS - 1);
330 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
331 	list_for_each_entry(entry, &cipso_v4_cache[bkt].list, list) {
332 		if (entry->hash == hash &&
333 		    entry->key_len == key_len &&
334 		    memcmp(entry->key, key, key_len) == 0) {
335 			entry->activity += 1;
336 			atomic_inc(&entry->lsm_data->refcount);
337 			secattr->cache = entry->lsm_data;
338 			secattr->flags |= NETLBL_SECATTR_CACHE;
339 			secattr->type = NETLBL_NLTYPE_CIPSOV4;
340 			if (prev_entry == NULL) {
341 				spin_unlock_bh(&cipso_v4_cache[bkt].lock);
342 				return 0;
343 			}
344 
345 			if (prev_entry->activity > 0)
346 				prev_entry->activity -= 1;
347 			if (entry->activity > prev_entry->activity &&
348 			    entry->activity - prev_entry->activity >
349 			    CIPSO_V4_CACHE_REORDERLIMIT) {
350 				__list_del(entry->list.prev, entry->list.next);
351 				__list_add(&entry->list,
352 					   prev_entry->list.prev,
353 					   &prev_entry->list);
354 			}
355 
356 			spin_unlock_bh(&cipso_v4_cache[bkt].lock);
357 			return 0;
358 		}
359 		prev_entry = entry;
360 	}
361 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
362 
363 	return -ENOENT;
364 }
365 
366 /**
367  * cipso_v4_cache_add - Add an entry to the CIPSO cache
368  * @skb: the packet
369  * @secattr: the packet's security attributes
370  *
371  * Description:
372  * Add a new entry into the CIPSO label mapping cache.  Add the new entry to
373  * head of the cache bucket's list, if the cache bucket is out of room remove
374  * the last entry in the list first.  It is important to note that there is
375  * currently no checking for duplicate keys.  Returns zero on success,
376  * negative values on failure.
377  *
378  */
379 int cipso_v4_cache_add(const struct sk_buff *skb,
380 		       const struct netlbl_lsm_secattr *secattr)
381 {
382 	int ret_val = -EPERM;
383 	u32 bkt;
384 	struct cipso_v4_map_cache_entry *entry = NULL;
385 	struct cipso_v4_map_cache_entry *old_entry = NULL;
386 	unsigned char *cipso_ptr;
387 	u32 cipso_ptr_len;
388 
389 	if (!cipso_v4_cache_enabled || cipso_v4_cache_bucketsize <= 0)
390 		return 0;
391 
392 	cipso_ptr = CIPSO_V4_OPTPTR(skb);
393 	cipso_ptr_len = cipso_ptr[1];
394 
395 	entry = kzalloc(sizeof(*entry), GFP_ATOMIC);
396 	if (entry == NULL)
397 		return -ENOMEM;
398 	entry->key = kmemdup(cipso_ptr, cipso_ptr_len, GFP_ATOMIC);
399 	if (entry->key == NULL) {
400 		ret_val = -ENOMEM;
401 		goto cache_add_failure;
402 	}
403 	entry->key_len = cipso_ptr_len;
404 	entry->hash = cipso_v4_map_cache_hash(cipso_ptr, cipso_ptr_len);
405 	atomic_inc(&secattr->cache->refcount);
406 	entry->lsm_data = secattr->cache;
407 
408 	bkt = entry->hash & (CIPSO_V4_CACHE_BUCKETS - 1);
409 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
410 	if (cipso_v4_cache[bkt].size < cipso_v4_cache_bucketsize) {
411 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
412 		cipso_v4_cache[bkt].size += 1;
413 	} else {
414 		old_entry = list_entry(cipso_v4_cache[bkt].list.prev,
415 				       struct cipso_v4_map_cache_entry, list);
416 		list_del(&old_entry->list);
417 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
418 		cipso_v4_cache_entry_free(old_entry);
419 	}
420 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
421 
422 	return 0;
423 
424 cache_add_failure:
425 	if (entry)
426 		cipso_v4_cache_entry_free(entry);
427 	return ret_val;
428 }
429 
430 /*
431  * DOI List Functions
432  */
433 
434 /**
435  * cipso_v4_doi_search - Searches for a DOI definition
436  * @doi: the DOI to search for
437  *
438  * Description:
439  * Search the DOI definition list for a DOI definition with a DOI value that
440  * matches @doi.  The caller is responsible for calling rcu_read_[un]lock().
441  * Returns a pointer to the DOI definition on success and NULL on failure.
442  */
443 static struct cipso_v4_doi *cipso_v4_doi_search(u32 doi)
444 {
445 	struct cipso_v4_doi *iter;
446 
447 	list_for_each_entry_rcu(iter, &cipso_v4_doi_list, list)
448 		if (iter->doi == doi && atomic_read(&iter->refcount))
449 			return iter;
450 	return NULL;
451 }
452 
453 /**
454  * cipso_v4_doi_add - Add a new DOI to the CIPSO protocol engine
455  * @doi_def: the DOI structure
456  * @audit_info: NetLabel audit information
457  *
458  * Description:
459  * The caller defines a new DOI for use by the CIPSO engine and calls this
460  * function to add it to the list of acceptable domains.  The caller must
461  * ensure that the mapping table specified in @doi_def->map meets all of the
462  * requirements of the mapping type (see cipso_ipv4.h for details).  Returns
463  * zero on success and non-zero on failure.
464  *
465  */
466 int cipso_v4_doi_add(struct cipso_v4_doi *doi_def,
467 		     struct netlbl_audit *audit_info)
468 {
469 	int ret_val = -EINVAL;
470 	u32 iter;
471 	u32 doi;
472 	u32 doi_type;
473 	struct audit_buffer *audit_buf;
474 
475 	doi = doi_def->doi;
476 	doi_type = doi_def->type;
477 
478 	if (doi_def->doi == CIPSO_V4_DOI_UNKNOWN)
479 		goto doi_add_return;
480 	for (iter = 0; iter < CIPSO_V4_TAG_MAXCNT; iter++) {
481 		switch (doi_def->tags[iter]) {
482 		case CIPSO_V4_TAG_RBITMAP:
483 			break;
484 		case CIPSO_V4_TAG_RANGE:
485 		case CIPSO_V4_TAG_ENUM:
486 			if (doi_def->type != CIPSO_V4_MAP_PASS)
487 				goto doi_add_return;
488 			break;
489 		case CIPSO_V4_TAG_LOCAL:
490 			if (doi_def->type != CIPSO_V4_MAP_LOCAL)
491 				goto doi_add_return;
492 			break;
493 		case CIPSO_V4_TAG_INVALID:
494 			if (iter == 0)
495 				goto doi_add_return;
496 			break;
497 		default:
498 			goto doi_add_return;
499 		}
500 	}
501 
502 	atomic_set(&doi_def->refcount, 1);
503 
504 	spin_lock(&cipso_v4_doi_list_lock);
505 	if (cipso_v4_doi_search(doi_def->doi) != NULL) {
506 		spin_unlock(&cipso_v4_doi_list_lock);
507 		ret_val = -EEXIST;
508 		goto doi_add_return;
509 	}
510 	list_add_tail_rcu(&doi_def->list, &cipso_v4_doi_list);
511 	spin_unlock(&cipso_v4_doi_list_lock);
512 	ret_val = 0;
513 
514 doi_add_return:
515 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_ADD, audit_info);
516 	if (audit_buf != NULL) {
517 		const char *type_str;
518 		switch (doi_type) {
519 		case CIPSO_V4_MAP_TRANS:
520 			type_str = "trans";
521 			break;
522 		case CIPSO_V4_MAP_PASS:
523 			type_str = "pass";
524 			break;
525 		case CIPSO_V4_MAP_LOCAL:
526 			type_str = "local";
527 			break;
528 		default:
529 			type_str = "(unknown)";
530 		}
531 		audit_log_format(audit_buf,
532 				 " cipso_doi=%u cipso_type=%s res=%u",
533 				 doi, type_str, ret_val == 0 ? 1 : 0);
534 		audit_log_end(audit_buf);
535 	}
536 
537 	return ret_val;
538 }
539 
540 /**
541  * cipso_v4_doi_free - Frees a DOI definition
542  * @entry: the entry's RCU field
543  *
544  * Description:
545  * This function frees all of the memory associated with a DOI definition.
546  *
547  */
548 void cipso_v4_doi_free(struct cipso_v4_doi *doi_def)
549 {
550 	if (doi_def == NULL)
551 		return;
552 
553 	switch (doi_def->type) {
554 	case CIPSO_V4_MAP_TRANS:
555 		kfree(doi_def->map.std->lvl.cipso);
556 		kfree(doi_def->map.std->lvl.local);
557 		kfree(doi_def->map.std->cat.cipso);
558 		kfree(doi_def->map.std->cat.local);
559 		break;
560 	}
561 	kfree(doi_def);
562 }
563 
564 /**
565  * cipso_v4_doi_free_rcu - Frees a DOI definition via the RCU pointer
566  * @entry: the entry's RCU field
567  *
568  * Description:
569  * This function is designed to be used as a callback to the call_rcu()
570  * function so that the memory allocated to the DOI definition can be released
571  * safely.
572  *
573  */
574 static void cipso_v4_doi_free_rcu(struct rcu_head *entry)
575 {
576 	struct cipso_v4_doi *doi_def;
577 
578 	doi_def = container_of(entry, struct cipso_v4_doi, rcu);
579 	cipso_v4_doi_free(doi_def);
580 }
581 
582 /**
583  * cipso_v4_doi_remove - Remove an existing DOI from the CIPSO protocol engine
584  * @doi: the DOI value
585  * @audit_secid: the LSM secid to use in the audit message
586  *
587  * Description:
588  * Removes a DOI definition from the CIPSO engine.  The NetLabel routines will
589  * be called to release their own LSM domain mappings as well as our own
590  * domain list.  Returns zero on success and negative values on failure.
591  *
592  */
593 int cipso_v4_doi_remove(u32 doi, struct netlbl_audit *audit_info)
594 {
595 	int ret_val;
596 	struct cipso_v4_doi *doi_def;
597 	struct audit_buffer *audit_buf;
598 
599 	spin_lock(&cipso_v4_doi_list_lock);
600 	doi_def = cipso_v4_doi_search(doi);
601 	if (doi_def == NULL) {
602 		spin_unlock(&cipso_v4_doi_list_lock);
603 		ret_val = -ENOENT;
604 		goto doi_remove_return;
605 	}
606 	if (!atomic_dec_and_test(&doi_def->refcount)) {
607 		spin_unlock(&cipso_v4_doi_list_lock);
608 		ret_val = -EBUSY;
609 		goto doi_remove_return;
610 	}
611 	list_del_rcu(&doi_def->list);
612 	spin_unlock(&cipso_v4_doi_list_lock);
613 
614 	cipso_v4_cache_invalidate();
615 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
616 	ret_val = 0;
617 
618 doi_remove_return:
619 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_DEL, audit_info);
620 	if (audit_buf != NULL) {
621 		audit_log_format(audit_buf,
622 				 " cipso_doi=%u res=%u",
623 				 doi, ret_val == 0 ? 1 : 0);
624 		audit_log_end(audit_buf);
625 	}
626 
627 	return ret_val;
628 }
629 
630 /**
631  * cipso_v4_doi_getdef - Returns a reference to a valid DOI definition
632  * @doi: the DOI value
633  *
634  * Description:
635  * Searches for a valid DOI definition and if one is found it is returned to
636  * the caller.  Otherwise NULL is returned.  The caller must ensure that
637  * rcu_read_lock() is held while accessing the returned definition and the DOI
638  * definition reference count is decremented when the caller is done.
639  *
640  */
641 struct cipso_v4_doi *cipso_v4_doi_getdef(u32 doi)
642 {
643 	struct cipso_v4_doi *doi_def;
644 
645 	rcu_read_lock();
646 	doi_def = cipso_v4_doi_search(doi);
647 	if (doi_def == NULL)
648 		goto doi_getdef_return;
649 	if (!atomic_inc_not_zero(&doi_def->refcount))
650 		doi_def = NULL;
651 
652 doi_getdef_return:
653 	rcu_read_unlock();
654 	return doi_def;
655 }
656 
657 /**
658  * cipso_v4_doi_putdef - Releases a reference for the given DOI definition
659  * @doi_def: the DOI definition
660  *
661  * Description:
662  * Releases a DOI definition reference obtained from cipso_v4_doi_getdef().
663  *
664  */
665 void cipso_v4_doi_putdef(struct cipso_v4_doi *doi_def)
666 {
667 	if (doi_def == NULL)
668 		return;
669 
670 	if (!atomic_dec_and_test(&doi_def->refcount))
671 		return;
672 	spin_lock(&cipso_v4_doi_list_lock);
673 	list_del_rcu(&doi_def->list);
674 	spin_unlock(&cipso_v4_doi_list_lock);
675 
676 	cipso_v4_cache_invalidate();
677 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
678 }
679 
680 /**
681  * cipso_v4_doi_walk - Iterate through the DOI definitions
682  * @skip_cnt: skip past this number of DOI definitions, updated
683  * @callback: callback for each DOI definition
684  * @cb_arg: argument for the callback function
685  *
686  * Description:
687  * Iterate over the DOI definition list, skipping the first @skip_cnt entries.
688  * For each entry call @callback, if @callback returns a negative value stop
689  * 'walking' through the list and return.  Updates the value in @skip_cnt upon
690  * return.  Returns zero on success, negative values on failure.
691  *
692  */
693 int cipso_v4_doi_walk(u32 *skip_cnt,
694 		     int (*callback) (struct cipso_v4_doi *doi_def, void *arg),
695 		     void *cb_arg)
696 {
697 	int ret_val = -ENOENT;
698 	u32 doi_cnt = 0;
699 	struct cipso_v4_doi *iter_doi;
700 
701 	rcu_read_lock();
702 	list_for_each_entry_rcu(iter_doi, &cipso_v4_doi_list, list)
703 		if (atomic_read(&iter_doi->refcount) > 0) {
704 			if (doi_cnt++ < *skip_cnt)
705 				continue;
706 			ret_val = callback(iter_doi, cb_arg);
707 			if (ret_val < 0) {
708 				doi_cnt--;
709 				goto doi_walk_return;
710 			}
711 		}
712 
713 doi_walk_return:
714 	rcu_read_unlock();
715 	*skip_cnt = doi_cnt;
716 	return ret_val;
717 }
718 
719 /*
720  * Label Mapping Functions
721  */
722 
723 /**
724  * cipso_v4_map_lvl_valid - Checks to see if the given level is understood
725  * @doi_def: the DOI definition
726  * @level: the level to check
727  *
728  * Description:
729  * Checks the given level against the given DOI definition and returns a
730  * negative value if the level does not have a valid mapping and a zero value
731  * if the level is defined by the DOI.
732  *
733  */
734 static int cipso_v4_map_lvl_valid(const struct cipso_v4_doi *doi_def, u8 level)
735 {
736 	switch (doi_def->type) {
737 	case CIPSO_V4_MAP_PASS:
738 		return 0;
739 	case CIPSO_V4_MAP_TRANS:
740 		if (doi_def->map.std->lvl.cipso[level] < CIPSO_V4_INV_LVL)
741 			return 0;
742 		break;
743 	}
744 
745 	return -EFAULT;
746 }
747 
748 /**
749  * cipso_v4_map_lvl_hton - Perform a level mapping from the host to the network
750  * @doi_def: the DOI definition
751  * @host_lvl: the host MLS level
752  * @net_lvl: the network/CIPSO MLS level
753  *
754  * Description:
755  * Perform a label mapping to translate a local MLS level to the correct
756  * CIPSO level using the given DOI definition.  Returns zero on success,
757  * negative values otherwise.
758  *
759  */
760 static int cipso_v4_map_lvl_hton(const struct cipso_v4_doi *doi_def,
761 				 u32 host_lvl,
762 				 u32 *net_lvl)
763 {
764 	switch (doi_def->type) {
765 	case CIPSO_V4_MAP_PASS:
766 		*net_lvl = host_lvl;
767 		return 0;
768 	case CIPSO_V4_MAP_TRANS:
769 		if (host_lvl < doi_def->map.std->lvl.local_size &&
770 		    doi_def->map.std->lvl.local[host_lvl] < CIPSO_V4_INV_LVL) {
771 			*net_lvl = doi_def->map.std->lvl.local[host_lvl];
772 			return 0;
773 		}
774 		return -EPERM;
775 	}
776 
777 	return -EINVAL;
778 }
779 
780 /**
781  * cipso_v4_map_lvl_ntoh - Perform a level mapping from the network to the host
782  * @doi_def: the DOI definition
783  * @net_lvl: the network/CIPSO MLS level
784  * @host_lvl: the host MLS level
785  *
786  * Description:
787  * Perform a label mapping to translate a CIPSO level to the correct local MLS
788  * level using the given DOI definition.  Returns zero on success, negative
789  * values otherwise.
790  *
791  */
792 static int cipso_v4_map_lvl_ntoh(const struct cipso_v4_doi *doi_def,
793 				 u32 net_lvl,
794 				 u32 *host_lvl)
795 {
796 	struct cipso_v4_std_map_tbl *map_tbl;
797 
798 	switch (doi_def->type) {
799 	case CIPSO_V4_MAP_PASS:
800 		*host_lvl = net_lvl;
801 		return 0;
802 	case CIPSO_V4_MAP_TRANS:
803 		map_tbl = doi_def->map.std;
804 		if (net_lvl < map_tbl->lvl.cipso_size &&
805 		    map_tbl->lvl.cipso[net_lvl] < CIPSO_V4_INV_LVL) {
806 			*host_lvl = doi_def->map.std->lvl.cipso[net_lvl];
807 			return 0;
808 		}
809 		return -EPERM;
810 	}
811 
812 	return -EINVAL;
813 }
814 
815 /**
816  * cipso_v4_map_cat_rbm_valid - Checks to see if the category bitmap is valid
817  * @doi_def: the DOI definition
818  * @bitmap: category bitmap
819  * @bitmap_len: bitmap length in bytes
820  *
821  * Description:
822  * Checks the given category bitmap against the given DOI definition and
823  * returns a negative value if any of the categories in the bitmap do not have
824  * a valid mapping and a zero value if all of the categories are valid.
825  *
826  */
827 static int cipso_v4_map_cat_rbm_valid(const struct cipso_v4_doi *doi_def,
828 				      const unsigned char *bitmap,
829 				      u32 bitmap_len)
830 {
831 	int cat = -1;
832 	u32 bitmap_len_bits = bitmap_len * 8;
833 	u32 cipso_cat_size;
834 	u32 *cipso_array;
835 
836 	switch (doi_def->type) {
837 	case CIPSO_V4_MAP_PASS:
838 		return 0;
839 	case CIPSO_V4_MAP_TRANS:
840 		cipso_cat_size = doi_def->map.std->cat.cipso_size;
841 		cipso_array = doi_def->map.std->cat.cipso;
842 		for (;;) {
843 			cat = cipso_v4_bitmap_walk(bitmap,
844 						   bitmap_len_bits,
845 						   cat + 1,
846 						   1);
847 			if (cat < 0)
848 				break;
849 			if (cat >= cipso_cat_size ||
850 			    cipso_array[cat] >= CIPSO_V4_INV_CAT)
851 				return -EFAULT;
852 		}
853 
854 		if (cat == -1)
855 			return 0;
856 		break;
857 	}
858 
859 	return -EFAULT;
860 }
861 
862 /**
863  * cipso_v4_map_cat_rbm_hton - Perform a category mapping from host to network
864  * @doi_def: the DOI definition
865  * @secattr: the security attributes
866  * @net_cat: the zero'd out category bitmap in network/CIPSO format
867  * @net_cat_len: the length of the CIPSO bitmap in bytes
868  *
869  * Description:
870  * Perform a label mapping to translate a local MLS category bitmap to the
871  * correct CIPSO bitmap using the given DOI definition.  Returns the minimum
872  * size in bytes of the network bitmap on success, negative values otherwise.
873  *
874  */
875 static int cipso_v4_map_cat_rbm_hton(const struct cipso_v4_doi *doi_def,
876 				     const struct netlbl_lsm_secattr *secattr,
877 				     unsigned char *net_cat,
878 				     u32 net_cat_len)
879 {
880 	int host_spot = -1;
881 	u32 net_spot = CIPSO_V4_INV_CAT;
882 	u32 net_spot_max = 0;
883 	u32 net_clen_bits = net_cat_len * 8;
884 	u32 host_cat_size = 0;
885 	u32 *host_cat_array = NULL;
886 
887 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
888 		host_cat_size = doi_def->map.std->cat.local_size;
889 		host_cat_array = doi_def->map.std->cat.local;
890 	}
891 
892 	for (;;) {
893 		host_spot = netlbl_secattr_catmap_walk(secattr->attr.mls.cat,
894 						       host_spot + 1);
895 		if (host_spot < 0)
896 			break;
897 
898 		switch (doi_def->type) {
899 		case CIPSO_V4_MAP_PASS:
900 			net_spot = host_spot;
901 			break;
902 		case CIPSO_V4_MAP_TRANS:
903 			if (host_spot >= host_cat_size)
904 				return -EPERM;
905 			net_spot = host_cat_array[host_spot];
906 			if (net_spot >= CIPSO_V4_INV_CAT)
907 				return -EPERM;
908 			break;
909 		}
910 		if (net_spot >= net_clen_bits)
911 			return -ENOSPC;
912 		cipso_v4_bitmap_setbit(net_cat, net_spot, 1);
913 
914 		if (net_spot > net_spot_max)
915 			net_spot_max = net_spot;
916 	}
917 
918 	if (++net_spot_max % 8)
919 		return net_spot_max / 8 + 1;
920 	return net_spot_max / 8;
921 }
922 
923 /**
924  * cipso_v4_map_cat_rbm_ntoh - Perform a category mapping from network to host
925  * @doi_def: the DOI definition
926  * @net_cat: the category bitmap in network/CIPSO format
927  * @net_cat_len: the length of the CIPSO bitmap in bytes
928  * @secattr: the security attributes
929  *
930  * Description:
931  * Perform a label mapping to translate a CIPSO bitmap to the correct local
932  * MLS category bitmap using the given DOI definition.  Returns zero on
933  * success, negative values on failure.
934  *
935  */
936 static int cipso_v4_map_cat_rbm_ntoh(const struct cipso_v4_doi *doi_def,
937 				     const unsigned char *net_cat,
938 				     u32 net_cat_len,
939 				     struct netlbl_lsm_secattr *secattr)
940 {
941 	int ret_val;
942 	int net_spot = -1;
943 	u32 host_spot = CIPSO_V4_INV_CAT;
944 	u32 net_clen_bits = net_cat_len * 8;
945 	u32 net_cat_size = 0;
946 	u32 *net_cat_array = NULL;
947 
948 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
949 		net_cat_size = doi_def->map.std->cat.cipso_size;
950 		net_cat_array = doi_def->map.std->cat.cipso;
951 	}
952 
953 	for (;;) {
954 		net_spot = cipso_v4_bitmap_walk(net_cat,
955 						net_clen_bits,
956 						net_spot + 1,
957 						1);
958 		if (net_spot < 0) {
959 			if (net_spot == -2)
960 				return -EFAULT;
961 			return 0;
962 		}
963 
964 		switch (doi_def->type) {
965 		case CIPSO_V4_MAP_PASS:
966 			host_spot = net_spot;
967 			break;
968 		case CIPSO_V4_MAP_TRANS:
969 			if (net_spot >= net_cat_size)
970 				return -EPERM;
971 			host_spot = net_cat_array[net_spot];
972 			if (host_spot >= CIPSO_V4_INV_CAT)
973 				return -EPERM;
974 			break;
975 		}
976 		ret_val = netlbl_secattr_catmap_setbit(secattr->attr.mls.cat,
977 						       host_spot,
978 						       GFP_ATOMIC);
979 		if (ret_val != 0)
980 			return ret_val;
981 	}
982 
983 	return -EINVAL;
984 }
985 
986 /**
987  * cipso_v4_map_cat_enum_valid - Checks to see if the categories are valid
988  * @doi_def: the DOI definition
989  * @enumcat: category list
990  * @enumcat_len: length of the category list in bytes
991  *
992  * Description:
993  * Checks the given categories against the given DOI definition and returns a
994  * negative value if any of the categories do not have a valid mapping and a
995  * zero value if all of the categories are valid.
996  *
997  */
998 static int cipso_v4_map_cat_enum_valid(const struct cipso_v4_doi *doi_def,
999 				       const unsigned char *enumcat,
1000 				       u32 enumcat_len)
1001 {
1002 	u16 cat;
1003 	int cat_prev = -1;
1004 	u32 iter;
1005 
1006 	if (doi_def->type != CIPSO_V4_MAP_PASS || enumcat_len & 0x01)
1007 		return -EFAULT;
1008 
1009 	for (iter = 0; iter < enumcat_len; iter += 2) {
1010 		cat = get_unaligned_be16(&enumcat[iter]);
1011 		if (cat <= cat_prev)
1012 			return -EFAULT;
1013 		cat_prev = cat;
1014 	}
1015 
1016 	return 0;
1017 }
1018 
1019 /**
1020  * cipso_v4_map_cat_enum_hton - Perform a category mapping from host to network
1021  * @doi_def: the DOI definition
1022  * @secattr: the security attributes
1023  * @net_cat: the zero'd out category list in network/CIPSO format
1024  * @net_cat_len: the length of the CIPSO category list in bytes
1025  *
1026  * Description:
1027  * Perform a label mapping to translate a local MLS category bitmap to the
1028  * correct CIPSO category list using the given DOI definition.   Returns the
1029  * size in bytes of the network category bitmap on success, negative values
1030  * otherwise.
1031  *
1032  */
1033 static int cipso_v4_map_cat_enum_hton(const struct cipso_v4_doi *doi_def,
1034 				      const struct netlbl_lsm_secattr *secattr,
1035 				      unsigned char *net_cat,
1036 				      u32 net_cat_len)
1037 {
1038 	int cat = -1;
1039 	u32 cat_iter = 0;
1040 
1041 	for (;;) {
1042 		cat = netlbl_secattr_catmap_walk(secattr->attr.mls.cat,
1043 						 cat + 1);
1044 		if (cat < 0)
1045 			break;
1046 		if ((cat_iter + 2) > net_cat_len)
1047 			return -ENOSPC;
1048 
1049 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
1050 		cat_iter += 2;
1051 	}
1052 
1053 	return cat_iter;
1054 }
1055 
1056 /**
1057  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
1058  * @doi_def: the DOI definition
1059  * @net_cat: the category list in network/CIPSO format
1060  * @net_cat_len: the length of the CIPSO bitmap in bytes
1061  * @secattr: the security attributes
1062  *
1063  * Description:
1064  * Perform a label mapping to translate a CIPSO category list to the correct
1065  * local MLS category bitmap using the given DOI definition.  Returns zero on
1066  * success, negative values on failure.
1067  *
1068  */
1069 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
1070 				      const unsigned char *net_cat,
1071 				      u32 net_cat_len,
1072 				      struct netlbl_lsm_secattr *secattr)
1073 {
1074 	int ret_val;
1075 	u32 iter;
1076 
1077 	for (iter = 0; iter < net_cat_len; iter += 2) {
1078 		ret_val = netlbl_secattr_catmap_setbit(secattr->attr.mls.cat,
1079 				get_unaligned_be16(&net_cat[iter]),
1080 				GFP_ATOMIC);
1081 		if (ret_val != 0)
1082 			return ret_val;
1083 	}
1084 
1085 	return 0;
1086 }
1087 
1088 /**
1089  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
1090  * @doi_def: the DOI definition
1091  * @rngcat: category list
1092  * @rngcat_len: length of the category list in bytes
1093  *
1094  * Description:
1095  * Checks the given categories against the given DOI definition and returns a
1096  * negative value if any of the categories do not have a valid mapping and a
1097  * zero value if all of the categories are valid.
1098  *
1099  */
1100 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1101 				      const unsigned char *rngcat,
1102 				      u32 rngcat_len)
1103 {
1104 	u16 cat_high;
1105 	u16 cat_low;
1106 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1107 	u32 iter;
1108 
1109 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1110 		return -EFAULT;
1111 
1112 	for (iter = 0; iter < rngcat_len; iter += 4) {
1113 		cat_high = get_unaligned_be16(&rngcat[iter]);
1114 		if ((iter + 4) <= rngcat_len)
1115 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1116 		else
1117 			cat_low = 0;
1118 
1119 		if (cat_high > cat_prev)
1120 			return -EFAULT;
1121 
1122 		cat_prev = cat_low;
1123 	}
1124 
1125 	return 0;
1126 }
1127 
1128 /**
1129  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1130  * @doi_def: the DOI definition
1131  * @secattr: the security attributes
1132  * @net_cat: the zero'd out category list in network/CIPSO format
1133  * @net_cat_len: the length of the CIPSO category list in bytes
1134  *
1135  * Description:
1136  * Perform a label mapping to translate a local MLS category bitmap to the
1137  * correct CIPSO category list using the given DOI definition.   Returns the
1138  * size in bytes of the network category bitmap on success, negative values
1139  * otherwise.
1140  *
1141  */
1142 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1143 				     const struct netlbl_lsm_secattr *secattr,
1144 				     unsigned char *net_cat,
1145 				     u32 net_cat_len)
1146 {
1147 	int iter = -1;
1148 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1149 	u32 array_cnt = 0;
1150 	u32 cat_size = 0;
1151 
1152 	/* make sure we don't overflow the 'array[]' variable */
1153 	if (net_cat_len >
1154 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1155 		return -ENOSPC;
1156 
1157 	for (;;) {
1158 		iter = netlbl_secattr_catmap_walk(secattr->attr.mls.cat,
1159 						  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_secattr_catmap_walk_rng(secattr->attr.mls.cat,
1168 						      iter);
1169 		if (iter < 0)
1170 			return -EFAULT;
1171 		cat_size += sizeof(u16);
1172 		if (cat_size > net_cat_len)
1173 			return -ENOSPC;
1174 		array[array_cnt++] = iter;
1175 	}
1176 
1177 	for (iter = 0; array_cnt > 0;) {
1178 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1179 		iter += 2;
1180 		array_cnt--;
1181 		if (array[array_cnt] != 0) {
1182 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1183 			iter += 2;
1184 		}
1185 	}
1186 
1187 	return cat_size;
1188 }
1189 
1190 /**
1191  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1192  * @doi_def: the DOI definition
1193  * @net_cat: the category list in network/CIPSO format
1194  * @net_cat_len: the length of the CIPSO bitmap in bytes
1195  * @secattr: the security attributes
1196  *
1197  * Description:
1198  * Perform a label mapping to translate a CIPSO category list to the correct
1199  * local MLS category bitmap using the given DOI definition.  Returns zero on
1200  * success, negative values on failure.
1201  *
1202  */
1203 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1204 				     const unsigned char *net_cat,
1205 				     u32 net_cat_len,
1206 				     struct netlbl_lsm_secattr *secattr)
1207 {
1208 	int ret_val;
1209 	u32 net_iter;
1210 	u16 cat_low;
1211 	u16 cat_high;
1212 
1213 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1214 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1215 		if ((net_iter + 4) <= net_cat_len)
1216 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1217 		else
1218 			cat_low = 0;
1219 
1220 		ret_val = netlbl_secattr_catmap_setrng(secattr->attr.mls.cat,
1221 						       cat_low,
1222 						       cat_high,
1223 						       GFP_ATOMIC);
1224 		if (ret_val != 0)
1225 			return ret_val;
1226 	}
1227 
1228 	return 0;
1229 }
1230 
1231 /*
1232  * Protocol Handling Functions
1233  */
1234 
1235 /**
1236  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1237  * @doi_def: the DOI definition
1238  * @len: the total tag length in bytes, not including this header
1239  * @buf: the CIPSO option buffer
1240  *
1241  * Description:
1242  * Write a CIPSO header into the beginning of @buffer.
1243  *
1244  */
1245 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1246 				unsigned char *buf,
1247 				u32 len)
1248 {
1249 	buf[0] = IPOPT_CIPSO;
1250 	buf[1] = CIPSO_V4_HDR_LEN + len;
1251 	*(__be32 *)&buf[2] = htonl(doi_def->doi);
1252 }
1253 
1254 /**
1255  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1256  * @doi_def: the DOI definition
1257  * @secattr: the security attributes
1258  * @buffer: the option buffer
1259  * @buffer_len: length of buffer in bytes
1260  *
1261  * Description:
1262  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1263  * actual buffer length may be larger than the indicated size due to
1264  * translation between host and network category bitmaps.  Returns the size of
1265  * the tag on success, negative values on failure.
1266  *
1267  */
1268 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1269 			       const struct netlbl_lsm_secattr *secattr,
1270 			       unsigned char *buffer,
1271 			       u32 buffer_len)
1272 {
1273 	int ret_val;
1274 	u32 tag_len;
1275 	u32 level;
1276 
1277 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1278 		return -EPERM;
1279 
1280 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1281 					secattr->attr.mls.lvl,
1282 					&level);
1283 	if (ret_val != 0)
1284 		return ret_val;
1285 
1286 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1287 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1288 						    secattr,
1289 						    &buffer[4],
1290 						    buffer_len - 4);
1291 		if (ret_val < 0)
1292 			return ret_val;
1293 
1294 		/* This will send packets using the "optimized" format when
1295 		 * possible as specified in  section 3.4.2.6 of the
1296 		 * CIPSO draft. */
1297 		if (cipso_v4_rbm_optfmt && ret_val > 0 && ret_val <= 10)
1298 			tag_len = 14;
1299 		else
1300 			tag_len = 4 + ret_val;
1301 	} else
1302 		tag_len = 4;
1303 
1304 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1305 	buffer[1] = tag_len;
1306 	buffer[3] = level;
1307 
1308 	return tag_len;
1309 }
1310 
1311 /**
1312  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1313  * @doi_def: the DOI definition
1314  * @tag: the CIPSO tag
1315  * @secattr: the security attributes
1316  *
1317  * Description:
1318  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1319  * attributes in @secattr.  Return zero on success, negatives values on
1320  * failure.
1321  *
1322  */
1323 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1324 				 const unsigned char *tag,
1325 				 struct netlbl_lsm_secattr *secattr)
1326 {
1327 	int ret_val;
1328 	u8 tag_len = tag[1];
1329 	u32 level;
1330 
1331 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1332 	if (ret_val != 0)
1333 		return ret_val;
1334 	secattr->attr.mls.lvl = level;
1335 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1336 
1337 	if (tag_len > 4) {
1338 		secattr->attr.mls.cat = netlbl_secattr_catmap_alloc(GFP_ATOMIC);
1339 		if (secattr->attr.mls.cat == NULL)
1340 			return -ENOMEM;
1341 
1342 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1343 						    &tag[4],
1344 						    tag_len - 4,
1345 						    secattr);
1346 		if (ret_val != 0) {
1347 			netlbl_secattr_catmap_free(secattr->attr.mls.cat);
1348 			return ret_val;
1349 		}
1350 
1351 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1352 	}
1353 
1354 	return 0;
1355 }
1356 
1357 /**
1358  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1359  * @doi_def: the DOI definition
1360  * @secattr: the security attributes
1361  * @buffer: the option buffer
1362  * @buffer_len: length of buffer in bytes
1363  *
1364  * Description:
1365  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1366  * size of the tag on success, negative values on failure.
1367  *
1368  */
1369 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1370 				const struct netlbl_lsm_secattr *secattr,
1371 				unsigned char *buffer,
1372 				u32 buffer_len)
1373 {
1374 	int ret_val;
1375 	u32 tag_len;
1376 	u32 level;
1377 
1378 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1379 		return -EPERM;
1380 
1381 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1382 					secattr->attr.mls.lvl,
1383 					&level);
1384 	if (ret_val != 0)
1385 		return ret_val;
1386 
1387 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1388 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1389 						     secattr,
1390 						     &buffer[4],
1391 						     buffer_len - 4);
1392 		if (ret_val < 0)
1393 			return ret_val;
1394 
1395 		tag_len = 4 + ret_val;
1396 	} else
1397 		tag_len = 4;
1398 
1399 	buffer[0] = CIPSO_V4_TAG_ENUM;
1400 	buffer[1] = tag_len;
1401 	buffer[3] = level;
1402 
1403 	return tag_len;
1404 }
1405 
1406 /**
1407  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1408  * @doi_def: the DOI definition
1409  * @tag: the CIPSO tag
1410  * @secattr: the security attributes
1411  *
1412  * Description:
1413  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1414  * attributes in @secattr.  Return zero on success, negatives values on
1415  * failure.
1416  *
1417  */
1418 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1419 				  const unsigned char *tag,
1420 				  struct netlbl_lsm_secattr *secattr)
1421 {
1422 	int ret_val;
1423 	u8 tag_len = tag[1];
1424 	u32 level;
1425 
1426 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1427 	if (ret_val != 0)
1428 		return ret_val;
1429 	secattr->attr.mls.lvl = level;
1430 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1431 
1432 	if (tag_len > 4) {
1433 		secattr->attr.mls.cat = netlbl_secattr_catmap_alloc(GFP_ATOMIC);
1434 		if (secattr->attr.mls.cat == NULL)
1435 			return -ENOMEM;
1436 
1437 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1438 						     &tag[4],
1439 						     tag_len - 4,
1440 						     secattr);
1441 		if (ret_val != 0) {
1442 			netlbl_secattr_catmap_free(secattr->attr.mls.cat);
1443 			return ret_val;
1444 		}
1445 
1446 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1447 	}
1448 
1449 	return 0;
1450 }
1451 
1452 /**
1453  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1454  * @doi_def: the DOI definition
1455  * @secattr: the security attributes
1456  * @buffer: the option buffer
1457  * @buffer_len: length of buffer in bytes
1458  *
1459  * Description:
1460  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1461  * size of the tag on success, negative values on failure.
1462  *
1463  */
1464 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1465 			       const struct netlbl_lsm_secattr *secattr,
1466 			       unsigned char *buffer,
1467 			       u32 buffer_len)
1468 {
1469 	int ret_val;
1470 	u32 tag_len;
1471 	u32 level;
1472 
1473 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1474 		return -EPERM;
1475 
1476 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1477 					secattr->attr.mls.lvl,
1478 					&level);
1479 	if (ret_val != 0)
1480 		return ret_val;
1481 
1482 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1483 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1484 						    secattr,
1485 						    &buffer[4],
1486 						    buffer_len - 4);
1487 		if (ret_val < 0)
1488 			return ret_val;
1489 
1490 		tag_len = 4 + ret_val;
1491 	} else
1492 		tag_len = 4;
1493 
1494 	buffer[0] = CIPSO_V4_TAG_RANGE;
1495 	buffer[1] = tag_len;
1496 	buffer[3] = level;
1497 
1498 	return tag_len;
1499 }
1500 
1501 /**
1502  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1503  * @doi_def: the DOI definition
1504  * @tag: the CIPSO tag
1505  * @secattr: the security attributes
1506  *
1507  * Description:
1508  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1509  * in @secattr.  Return zero on success, negatives values on failure.
1510  *
1511  */
1512 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1513 				 const unsigned char *tag,
1514 				 struct netlbl_lsm_secattr *secattr)
1515 {
1516 	int ret_val;
1517 	u8 tag_len = tag[1];
1518 	u32 level;
1519 
1520 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1521 	if (ret_val != 0)
1522 		return ret_val;
1523 	secattr->attr.mls.lvl = level;
1524 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1525 
1526 	if (tag_len > 4) {
1527 		secattr->attr.mls.cat = netlbl_secattr_catmap_alloc(GFP_ATOMIC);
1528 		if (secattr->attr.mls.cat == NULL)
1529 			return -ENOMEM;
1530 
1531 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1532 						    &tag[4],
1533 						    tag_len - 4,
1534 						    secattr);
1535 		if (ret_val != 0) {
1536 			netlbl_secattr_catmap_free(secattr->attr.mls.cat);
1537 			return ret_val;
1538 		}
1539 
1540 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1541 	}
1542 
1543 	return 0;
1544 }
1545 
1546 /**
1547  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1548  * @doi_def: the DOI definition
1549  * @secattr: the security attributes
1550  * @buffer: the option buffer
1551  * @buffer_len: length of buffer in bytes
1552  *
1553  * Description:
1554  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1555  * on success, negative values on failure.
1556  *
1557  */
1558 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1559 			       const struct netlbl_lsm_secattr *secattr,
1560 			       unsigned char *buffer,
1561 			       u32 buffer_len)
1562 {
1563 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1564 		return -EPERM;
1565 
1566 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1567 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1568 	*(u32 *)&buffer[2] = secattr->attr.secid;
1569 
1570 	return CIPSO_V4_TAG_LOC_BLEN;
1571 }
1572 
1573 /**
1574  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1575  * @doi_def: the DOI definition
1576  * @tag: the CIPSO tag
1577  * @secattr: the security attributes
1578  *
1579  * Description:
1580  * Parse a CIPSO local tag and return the security attributes in @secattr.
1581  * Return zero on success, negatives values on failure.
1582  *
1583  */
1584 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1585 				 const unsigned char *tag,
1586 				 struct netlbl_lsm_secattr *secattr)
1587 {
1588 	secattr->attr.secid = *(u32 *)&tag[2];
1589 	secattr->flags |= NETLBL_SECATTR_SECID;
1590 
1591 	return 0;
1592 }
1593 
1594 /**
1595  * cipso_v4_validate - Validate a CIPSO option
1596  * @option: the start of the option, on error it is set to point to the error
1597  *
1598  * Description:
1599  * This routine is called to validate a CIPSO option, it checks all of the
1600  * fields to ensure that they are at least valid, see the draft snippet below
1601  * for details.  If the option is valid then a zero value is returned and
1602  * the value of @option is unchanged.  If the option is invalid then a
1603  * non-zero value is returned and @option is adjusted to point to the
1604  * offending portion of the option.  From the IETF draft ...
1605  *
1606  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1607  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1608  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1609  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1610  *   that is unrecognized."
1611  *
1612  */
1613 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1614 {
1615 	unsigned char *opt = *option;
1616 	unsigned char *tag;
1617 	unsigned char opt_iter;
1618 	unsigned char err_offset = 0;
1619 	u8 opt_len;
1620 	u8 tag_len;
1621 	struct cipso_v4_doi *doi_def = NULL;
1622 	u32 tag_iter;
1623 
1624 	/* caller already checks for length values that are too large */
1625 	opt_len = opt[1];
1626 	if (opt_len < 8) {
1627 		err_offset = 1;
1628 		goto validate_return;
1629 	}
1630 
1631 	rcu_read_lock();
1632 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1633 	if (doi_def == NULL) {
1634 		err_offset = 2;
1635 		goto validate_return_locked;
1636 	}
1637 
1638 	opt_iter = CIPSO_V4_HDR_LEN;
1639 	tag = opt + opt_iter;
1640 	while (opt_iter < opt_len) {
1641 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1642 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1643 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1644 				err_offset = opt_iter;
1645 				goto validate_return_locked;
1646 			}
1647 
1648 		tag_len = tag[1];
1649 		if (tag_len > (opt_len - opt_iter)) {
1650 			err_offset = opt_iter + 1;
1651 			goto validate_return_locked;
1652 		}
1653 
1654 		switch (tag[0]) {
1655 		case CIPSO_V4_TAG_RBITMAP:
1656 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1657 				err_offset = opt_iter + 1;
1658 				goto validate_return_locked;
1659 			}
1660 
1661 			/* We are already going to do all the verification
1662 			 * necessary at the socket layer so from our point of
1663 			 * view it is safe to turn these checks off (and less
1664 			 * work), however, the CIPSO draft says we should do
1665 			 * all the CIPSO validations here but it doesn't
1666 			 * really specify _exactly_ what we need to validate
1667 			 * ... so, just make it a sysctl tunable. */
1668 			if (cipso_v4_rbm_strictvalid) {
1669 				if (cipso_v4_map_lvl_valid(doi_def,
1670 							   tag[3]) < 0) {
1671 					err_offset = opt_iter + 3;
1672 					goto validate_return_locked;
1673 				}
1674 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1675 				    cipso_v4_map_cat_rbm_valid(doi_def,
1676 							    &tag[4],
1677 							    tag_len - 4) < 0) {
1678 					err_offset = opt_iter + 4;
1679 					goto validate_return_locked;
1680 				}
1681 			}
1682 			break;
1683 		case CIPSO_V4_TAG_ENUM:
1684 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1685 				err_offset = opt_iter + 1;
1686 				goto validate_return_locked;
1687 			}
1688 
1689 			if (cipso_v4_map_lvl_valid(doi_def,
1690 						   tag[3]) < 0) {
1691 				err_offset = opt_iter + 3;
1692 				goto validate_return_locked;
1693 			}
1694 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1695 			    cipso_v4_map_cat_enum_valid(doi_def,
1696 							&tag[4],
1697 							tag_len - 4) < 0) {
1698 				err_offset = opt_iter + 4;
1699 				goto validate_return_locked;
1700 			}
1701 			break;
1702 		case CIPSO_V4_TAG_RANGE:
1703 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1704 				err_offset = opt_iter + 1;
1705 				goto validate_return_locked;
1706 			}
1707 
1708 			if (cipso_v4_map_lvl_valid(doi_def,
1709 						   tag[3]) < 0) {
1710 				err_offset = opt_iter + 3;
1711 				goto validate_return_locked;
1712 			}
1713 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1714 			    cipso_v4_map_cat_rng_valid(doi_def,
1715 						       &tag[4],
1716 						       tag_len - 4) < 0) {
1717 				err_offset = opt_iter + 4;
1718 				goto validate_return_locked;
1719 			}
1720 			break;
1721 		case CIPSO_V4_TAG_LOCAL:
1722 			/* This is a non-standard tag that we only allow for
1723 			 * local connections, so if the incoming interface is
1724 			 * not the loopback device drop the packet. Further,
1725 			 * there is no legitimate reason for setting this from
1726 			 * userspace so reject it if skb is NULL. */
1727 			if (skb == NULL || !(skb->dev->flags & IFF_LOOPBACK)) {
1728 				err_offset = opt_iter;
1729 				goto validate_return_locked;
1730 			}
1731 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1732 				err_offset = opt_iter + 1;
1733 				goto validate_return_locked;
1734 			}
1735 			break;
1736 		default:
1737 			err_offset = opt_iter;
1738 			goto validate_return_locked;
1739 		}
1740 
1741 		tag += tag_len;
1742 		opt_iter += tag_len;
1743 	}
1744 
1745 validate_return_locked:
1746 	rcu_read_unlock();
1747 validate_return:
1748 	*option = opt + err_offset;
1749 	return err_offset;
1750 }
1751 
1752 /**
1753  * cipso_v4_error - Send the correct response for a bad packet
1754  * @skb: the packet
1755  * @error: the error code
1756  * @gateway: CIPSO gateway flag
1757  *
1758  * Description:
1759  * Based on the error code given in @error, send an ICMP error message back to
1760  * the originating host.  From the IETF draft ...
1761  *
1762  *  "If the contents of the CIPSO [option] are valid but the security label is
1763  *   outside of the configured host or port label range, the datagram is
1764  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1765  *   returned.  The code field of the ICMP is set to 'communication with
1766  *   destination network administratively prohibited' (code 9) or to
1767  *   'communication with destination host administratively prohibited'
1768  *   (code 10).  The value of the code is dependent on whether the originator
1769  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1770  *   recipient of the ICMP message MUST be able to handle either value.  The
1771  *   same procedure is performed if a CIPSO [option] can not be added to an
1772  *   IP packet because it is too large to fit in the IP options area."
1773  *
1774  *  "If the error is triggered by receipt of an ICMP message, the message is
1775  *   discarded and no response is permitted (consistent with general ICMP
1776  *   processing rules)."
1777  *
1778  */
1779 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1780 {
1781 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1782 		return;
1783 
1784 	if (gateway)
1785 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0);
1786 	else
1787 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0);
1788 }
1789 
1790 /**
1791  * cipso_v4_genopt - Generate a CIPSO option
1792  * @buf: the option buffer
1793  * @buf_len: the size of opt_buf
1794  * @doi_def: the CIPSO DOI to use
1795  * @secattr: the security attributes
1796  *
1797  * Description:
1798  * Generate a CIPSO option using the DOI definition and security attributes
1799  * passed to the function.  Returns the length of the option on success and
1800  * negative values on failure.
1801  *
1802  */
1803 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1804 			   const struct cipso_v4_doi *doi_def,
1805 			   const struct netlbl_lsm_secattr *secattr)
1806 {
1807 	int ret_val;
1808 	u32 iter;
1809 
1810 	if (buf_len <= CIPSO_V4_HDR_LEN)
1811 		return -ENOSPC;
1812 
1813 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1814 	 * really a good assumption to make but since we only support the MAC
1815 	 * tags right now it is a safe assumption. */
1816 	iter = 0;
1817 	do {
1818 		memset(buf, 0, buf_len);
1819 		switch (doi_def->tags[iter]) {
1820 		case CIPSO_V4_TAG_RBITMAP:
1821 			ret_val = cipso_v4_gentag_rbm(doi_def,
1822 						   secattr,
1823 						   &buf[CIPSO_V4_HDR_LEN],
1824 						   buf_len - CIPSO_V4_HDR_LEN);
1825 			break;
1826 		case CIPSO_V4_TAG_ENUM:
1827 			ret_val = cipso_v4_gentag_enum(doi_def,
1828 						   secattr,
1829 						   &buf[CIPSO_V4_HDR_LEN],
1830 						   buf_len - CIPSO_V4_HDR_LEN);
1831 			break;
1832 		case CIPSO_V4_TAG_RANGE:
1833 			ret_val = cipso_v4_gentag_rng(doi_def,
1834 						   secattr,
1835 						   &buf[CIPSO_V4_HDR_LEN],
1836 						   buf_len - CIPSO_V4_HDR_LEN);
1837 			break;
1838 		case CIPSO_V4_TAG_LOCAL:
1839 			ret_val = cipso_v4_gentag_loc(doi_def,
1840 						   secattr,
1841 						   &buf[CIPSO_V4_HDR_LEN],
1842 						   buf_len - CIPSO_V4_HDR_LEN);
1843 			break;
1844 		default:
1845 			return -EPERM;
1846 		}
1847 
1848 		iter++;
1849 	} while (ret_val < 0 &&
1850 		 iter < CIPSO_V4_TAG_MAXCNT &&
1851 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1852 	if (ret_val < 0)
1853 		return ret_val;
1854 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1855 	return CIPSO_V4_HDR_LEN + ret_val;
1856 }
1857 
1858 /**
1859  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1860  * @sk: the socket
1861  * @doi_def: the CIPSO DOI to use
1862  * @secattr: the specific security attributes of the socket
1863  *
1864  * Description:
1865  * Set the CIPSO option on the given socket using the DOI definition and
1866  * security attributes passed to the function.  This function requires
1867  * exclusive access to @sk, which means it either needs to be in the
1868  * process of being created or locked.  Returns zero on success and negative
1869  * values on failure.
1870  *
1871  */
1872 int cipso_v4_sock_setattr(struct sock *sk,
1873 			  const struct cipso_v4_doi *doi_def,
1874 			  const struct netlbl_lsm_secattr *secattr)
1875 {
1876 	int ret_val = -EPERM;
1877 	unsigned char *buf = NULL;
1878 	u32 buf_len;
1879 	u32 opt_len;
1880 	struct ip_options_rcu *old, *opt = NULL;
1881 	struct inet_sock *sk_inet;
1882 	struct inet_connection_sock *sk_conn;
1883 
1884 	/* In the case of sock_create_lite(), the sock->sk field is not
1885 	 * defined yet but it is not a problem as the only users of these
1886 	 * "lite" PF_INET sockets are functions which do an accept() call
1887 	 * afterwards so we will label the socket as part of the accept(). */
1888 	if (sk == NULL)
1889 		return 0;
1890 
1891 	/* We allocate the maximum CIPSO option size here so we are probably
1892 	 * being a little wasteful, but it makes our life _much_ easier later
1893 	 * on and after all we are only talking about 40 bytes. */
1894 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1895 	buf = kmalloc(buf_len, GFP_ATOMIC);
1896 	if (buf == NULL) {
1897 		ret_val = -ENOMEM;
1898 		goto socket_setattr_failure;
1899 	}
1900 
1901 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1902 	if (ret_val < 0)
1903 		goto socket_setattr_failure;
1904 	buf_len = ret_val;
1905 
1906 	/* We can't use ip_options_get() directly because it makes a call to
1907 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1908 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1909 	 * set the IPOPT_CIPSO option. */
1910 	opt_len = (buf_len + 3) & ~3;
1911 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1912 	if (opt == NULL) {
1913 		ret_val = -ENOMEM;
1914 		goto socket_setattr_failure;
1915 	}
1916 	memcpy(opt->opt.__data, buf, buf_len);
1917 	opt->opt.optlen = opt_len;
1918 	opt->opt.cipso = sizeof(struct iphdr);
1919 	kfree(buf);
1920 	buf = NULL;
1921 
1922 	sk_inet = inet_sk(sk);
1923 
1924 	old = rcu_dereference_protected(sk_inet->inet_opt, sock_owned_by_user(sk));
1925 	if (sk_inet->is_icsk) {
1926 		sk_conn = inet_csk(sk);
1927 		if (old)
1928 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1929 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1930 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1931 	}
1932 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1933 	if (old)
1934 		kfree_rcu(old, rcu);
1935 
1936 	return 0;
1937 
1938 socket_setattr_failure:
1939 	kfree(buf);
1940 	kfree(opt);
1941 	return ret_val;
1942 }
1943 
1944 /**
1945  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1946  * @req: the connection request socket
1947  * @doi_def: the CIPSO DOI to use
1948  * @secattr: the specific security attributes of the socket
1949  *
1950  * Description:
1951  * Set the CIPSO option on the given socket using the DOI definition and
1952  * security attributes passed to the function.  Returns zero on success and
1953  * negative values on failure.
1954  *
1955  */
1956 int cipso_v4_req_setattr(struct request_sock *req,
1957 			 const struct cipso_v4_doi *doi_def,
1958 			 const struct netlbl_lsm_secattr *secattr)
1959 {
1960 	int ret_val = -EPERM;
1961 	unsigned char *buf = NULL;
1962 	u32 buf_len;
1963 	u32 opt_len;
1964 	struct ip_options_rcu *opt = NULL;
1965 	struct inet_request_sock *req_inet;
1966 
1967 	/* We allocate the maximum CIPSO option size here so we are probably
1968 	 * being a little wasteful, but it makes our life _much_ easier later
1969 	 * on and after all we are only talking about 40 bytes. */
1970 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1971 	buf = kmalloc(buf_len, GFP_ATOMIC);
1972 	if (buf == NULL) {
1973 		ret_val = -ENOMEM;
1974 		goto req_setattr_failure;
1975 	}
1976 
1977 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1978 	if (ret_val < 0)
1979 		goto req_setattr_failure;
1980 	buf_len = ret_val;
1981 
1982 	/* We can't use ip_options_get() directly because it makes a call to
1983 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1984 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1985 	 * set the IPOPT_CIPSO option. */
1986 	opt_len = (buf_len + 3) & ~3;
1987 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1988 	if (opt == NULL) {
1989 		ret_val = -ENOMEM;
1990 		goto req_setattr_failure;
1991 	}
1992 	memcpy(opt->opt.__data, buf, buf_len);
1993 	opt->opt.optlen = opt_len;
1994 	opt->opt.cipso = sizeof(struct iphdr);
1995 	kfree(buf);
1996 	buf = NULL;
1997 
1998 	req_inet = inet_rsk(req);
1999 	opt = xchg(&req_inet->opt, opt);
2000 	if (opt)
2001 		kfree_rcu(opt, rcu);
2002 
2003 	return 0;
2004 
2005 req_setattr_failure:
2006 	kfree(buf);
2007 	kfree(opt);
2008 	return ret_val;
2009 }
2010 
2011 /**
2012  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
2013  * @opt_ptr: IP option pointer
2014  *
2015  * Description:
2016  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
2017  * adjustments to the IP option structure.  Returns zero on success, negative
2018  * values on failure.
2019  *
2020  */
2021 static int cipso_v4_delopt(struct ip_options_rcu **opt_ptr)
2022 {
2023 	int hdr_delta = 0;
2024 	struct ip_options_rcu *opt = *opt_ptr;
2025 
2026 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
2027 		u8 cipso_len;
2028 		u8 cipso_off;
2029 		unsigned char *cipso_ptr;
2030 		int iter;
2031 		int optlen_new;
2032 
2033 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
2034 		cipso_ptr = &opt->opt.__data[cipso_off];
2035 		cipso_len = cipso_ptr[1];
2036 
2037 		if (opt->opt.srr > opt->opt.cipso)
2038 			opt->opt.srr -= cipso_len;
2039 		if (opt->opt.rr > opt->opt.cipso)
2040 			opt->opt.rr -= cipso_len;
2041 		if (opt->opt.ts > opt->opt.cipso)
2042 			opt->opt.ts -= cipso_len;
2043 		if (opt->opt.router_alert > opt->opt.cipso)
2044 			opt->opt.router_alert -= cipso_len;
2045 		opt->opt.cipso = 0;
2046 
2047 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2048 			opt->opt.optlen - cipso_off - cipso_len);
2049 
2050 		/* determining the new total option length is tricky because of
2051 		 * the padding necessary, the only thing i can think to do at
2052 		 * this point is walk the options one-by-one, skipping the
2053 		 * padding at the end to determine the actual option size and
2054 		 * from there we can determine the new total option length */
2055 		iter = 0;
2056 		optlen_new = 0;
2057 		while (iter < opt->opt.optlen)
2058 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2059 				iter += opt->opt.__data[iter + 1];
2060 				optlen_new = iter;
2061 			} else
2062 				iter++;
2063 		hdr_delta = opt->opt.optlen;
2064 		opt->opt.optlen = (optlen_new + 3) & ~3;
2065 		hdr_delta -= opt->opt.optlen;
2066 	} else {
2067 		/* only the cipso option was present on the socket so we can
2068 		 * remove the entire option struct */
2069 		*opt_ptr = NULL;
2070 		hdr_delta = opt->opt.optlen;
2071 		kfree_rcu(opt, rcu);
2072 	}
2073 
2074 	return hdr_delta;
2075 }
2076 
2077 /**
2078  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2079  * @sk: the socket
2080  *
2081  * Description:
2082  * Removes the CIPSO option from a socket, if present.
2083  *
2084  */
2085 void cipso_v4_sock_delattr(struct sock *sk)
2086 {
2087 	int hdr_delta;
2088 	struct ip_options_rcu *opt;
2089 	struct inet_sock *sk_inet;
2090 
2091 	sk_inet = inet_sk(sk);
2092 	opt = rcu_dereference_protected(sk_inet->inet_opt, 1);
2093 	if (opt == NULL || opt->opt.cipso == 0)
2094 		return;
2095 
2096 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2097 	if (sk_inet->is_icsk && hdr_delta > 0) {
2098 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2099 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2100 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2101 	}
2102 }
2103 
2104 /**
2105  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2106  * @reg: the request socket
2107  *
2108  * Description:
2109  * Removes the CIPSO option from a request socket, if present.
2110  *
2111  */
2112 void cipso_v4_req_delattr(struct request_sock *req)
2113 {
2114 	struct ip_options_rcu *opt;
2115 	struct inet_request_sock *req_inet;
2116 
2117 	req_inet = inet_rsk(req);
2118 	opt = req_inet->opt;
2119 	if (opt == NULL || opt->opt.cipso == 0)
2120 		return;
2121 
2122 	cipso_v4_delopt(&req_inet->opt);
2123 }
2124 
2125 /**
2126  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2127  * @cipso: the CIPSO v4 option
2128  * @secattr: the security attributes
2129  *
2130  * Description:
2131  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2132  * on success and negative values on failure.
2133  *
2134  */
2135 static int cipso_v4_getattr(const unsigned char *cipso,
2136 			    struct netlbl_lsm_secattr *secattr)
2137 {
2138 	int ret_val = -ENOMSG;
2139 	u32 doi;
2140 	struct cipso_v4_doi *doi_def;
2141 
2142 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2143 		return 0;
2144 
2145 	doi = get_unaligned_be32(&cipso[2]);
2146 	rcu_read_lock();
2147 	doi_def = cipso_v4_doi_search(doi);
2148 	if (doi_def == NULL)
2149 		goto getattr_return;
2150 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2151 	 * really a good assumption to make but since we only support the MAC
2152 	 * tags right now it is a safe assumption. */
2153 	switch (cipso[6]) {
2154 	case CIPSO_V4_TAG_RBITMAP:
2155 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2156 		break;
2157 	case CIPSO_V4_TAG_ENUM:
2158 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2159 		break;
2160 	case CIPSO_V4_TAG_RANGE:
2161 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2162 		break;
2163 	case CIPSO_V4_TAG_LOCAL:
2164 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2165 		break;
2166 	}
2167 	if (ret_val == 0)
2168 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2169 
2170 getattr_return:
2171 	rcu_read_unlock();
2172 	return ret_val;
2173 }
2174 
2175 /**
2176  * cipso_v4_sock_getattr - Get the security attributes from a sock
2177  * @sk: the sock
2178  * @secattr: the security attributes
2179  *
2180  * Description:
2181  * Query @sk to see if there is a CIPSO option attached to the sock and if
2182  * there is return the CIPSO security attributes in @secattr.  This function
2183  * requires that @sk be locked, or privately held, but it does not do any
2184  * locking itself.  Returns zero on success and negative values on failure.
2185  *
2186  */
2187 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2188 {
2189 	struct ip_options_rcu *opt;
2190 	int res = -ENOMSG;
2191 
2192 	rcu_read_lock();
2193 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2194 	if (opt && opt->opt.cipso)
2195 		res = cipso_v4_getattr(opt->opt.__data +
2196 						opt->opt.cipso -
2197 						sizeof(struct iphdr),
2198 				       secattr);
2199 	rcu_read_unlock();
2200 	return res;
2201 }
2202 
2203 /**
2204  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2205  * @skb: the packet
2206  * @secattr: the security attributes
2207  *
2208  * Description:
2209  * Set the CIPSO option on the given packet based on the security attributes.
2210  * Returns a pointer to the IP header on success and NULL on failure.
2211  *
2212  */
2213 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2214 			    const struct cipso_v4_doi *doi_def,
2215 			    const struct netlbl_lsm_secattr *secattr)
2216 {
2217 	int ret_val;
2218 	struct iphdr *iph;
2219 	struct ip_options *opt = &IPCB(skb)->opt;
2220 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2221 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2222 	u32 opt_len;
2223 	int len_delta;
2224 
2225 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2226 	if (ret_val < 0)
2227 		return ret_val;
2228 	buf_len = ret_val;
2229 	opt_len = (buf_len + 3) & ~3;
2230 
2231 	/* we overwrite any existing options to ensure that we have enough
2232 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2233 	 * that the security label is applied to the packet - we do the same
2234 	 * thing when using the socket options and it hasn't caused a problem,
2235 	 * if we need to we can always revisit this choice later */
2236 
2237 	len_delta = opt_len - opt->optlen;
2238 	/* if we don't ensure enough headroom we could panic on the skb_push()
2239 	 * call below so make sure we have enough, we are also "mangling" the
2240 	 * packet so we should probably do a copy-on-write call anyway */
2241 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2242 	if (ret_val < 0)
2243 		return ret_val;
2244 
2245 	if (len_delta > 0) {
2246 		/* we assume that the header + opt->optlen have already been
2247 		 * "pushed" in ip_options_build() or similar */
2248 		iph = ip_hdr(skb);
2249 		skb_push(skb, len_delta);
2250 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2251 		skb_reset_network_header(skb);
2252 		iph = ip_hdr(skb);
2253 	} else if (len_delta < 0) {
2254 		iph = ip_hdr(skb);
2255 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2256 	} else
2257 		iph = ip_hdr(skb);
2258 
2259 	if (opt->optlen > 0)
2260 		memset(opt, 0, sizeof(*opt));
2261 	opt->optlen = opt_len;
2262 	opt->cipso = sizeof(struct iphdr);
2263 	opt->is_changed = 1;
2264 
2265 	/* we have to do the following because we are being called from a
2266 	 * netfilter hook which means the packet already has had the header
2267 	 * fields populated and the checksum calculated - yes this means we
2268 	 * are doing more work than needed but we do it to keep the core
2269 	 * stack clean and tidy */
2270 	memcpy(iph + 1, buf, buf_len);
2271 	if (opt_len > buf_len)
2272 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2273 	if (len_delta != 0) {
2274 		iph->ihl = 5 + (opt_len >> 2);
2275 		iph->tot_len = htons(skb->len);
2276 	}
2277 	ip_send_check(iph);
2278 
2279 	return 0;
2280 }
2281 
2282 /**
2283  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2284  * @skb: the packet
2285  *
2286  * Description:
2287  * Removes any and all CIPSO options from the given packet.  Returns zero on
2288  * success, negative values on failure.
2289  *
2290  */
2291 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2292 {
2293 	int ret_val;
2294 	struct iphdr *iph;
2295 	struct ip_options *opt = &IPCB(skb)->opt;
2296 	unsigned char *cipso_ptr;
2297 
2298 	if (opt->cipso == 0)
2299 		return 0;
2300 
2301 	/* since we are changing the packet we should make a copy */
2302 	ret_val = skb_cow(skb, skb_headroom(skb));
2303 	if (ret_val < 0)
2304 		return ret_val;
2305 
2306 	/* the easiest thing to do is just replace the cipso option with noop
2307 	 * options since we don't change the size of the packet, although we
2308 	 * still need to recalculate the checksum */
2309 
2310 	iph = ip_hdr(skb);
2311 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2312 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2313 	opt->cipso = 0;
2314 	opt->is_changed = 1;
2315 
2316 	ip_send_check(iph);
2317 
2318 	return 0;
2319 }
2320 
2321 /**
2322  * cipso_v4_skbuff_getattr - Get the security attributes from the CIPSO option
2323  * @skb: the packet
2324  * @secattr: the security attributes
2325  *
2326  * Description:
2327  * Parse the given packet's CIPSO option and return the security attributes.
2328  * Returns zero on success and negative values on failure.
2329  *
2330  */
2331 int cipso_v4_skbuff_getattr(const struct sk_buff *skb,
2332 			    struct netlbl_lsm_secattr *secattr)
2333 {
2334 	return cipso_v4_getattr(CIPSO_V4_OPTPTR(skb), secattr);
2335 }
2336 
2337 /*
2338  * Setup Functions
2339  */
2340 
2341 /**
2342  * cipso_v4_init - Initialize the CIPSO module
2343  *
2344  * Description:
2345  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2346  * and negative values on failure.
2347  *
2348  */
2349 static int __init cipso_v4_init(void)
2350 {
2351 	int ret_val;
2352 
2353 	ret_val = cipso_v4_cache_init();
2354 	if (ret_val != 0)
2355 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2356 		      ret_val);
2357 
2358 	return 0;
2359 }
2360 
2361 subsys_initcall(cipso_v4_init);
2362