xref: /linux/net/ipv4/cipso_ipv4.c (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
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)
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) {
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 unsigned char *cipso_ptr,
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 	u32 cipso_ptr_len;
389 
390 	if (!cipso_v4_cache_enabled || cipso_v4_cache_bucketsize <= 0)
391 		return 0;
392 
393 	cipso_ptr_len = cipso_ptr[1];
394 
395 	entry = kzalloc(sizeof(*entry), GFP_ATOMIC);
396 	if (!entry)
397 		return -ENOMEM;
398 	entry->key = kmemdup(cipso_ptr, cipso_ptr_len, GFP_ATOMIC);
399 	if (!entry->key) {
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)) {
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) {
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  * @doi_def: the DOI definition
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)
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) {
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) {
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)
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)
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_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_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_catmap_walk(secattr->attr.mls.cat, cat + 1);
1043 		if (cat < 0)
1044 			break;
1045 		if ((cat_iter + 2) > net_cat_len)
1046 			return -ENOSPC;
1047 
1048 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
1049 		cat_iter += 2;
1050 	}
1051 
1052 	return cat_iter;
1053 }
1054 
1055 /**
1056  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
1057  * @doi_def: the DOI definition
1058  * @net_cat: the category list in network/CIPSO format
1059  * @net_cat_len: the length of the CIPSO bitmap in bytes
1060  * @secattr: the security attributes
1061  *
1062  * Description:
1063  * Perform a label mapping to translate a CIPSO category list to the correct
1064  * local MLS category bitmap using the given DOI definition.  Returns zero on
1065  * success, negative values on failure.
1066  *
1067  */
1068 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
1069 				      const unsigned char *net_cat,
1070 				      u32 net_cat_len,
1071 				      struct netlbl_lsm_secattr *secattr)
1072 {
1073 	int ret_val;
1074 	u32 iter;
1075 
1076 	for (iter = 0; iter < net_cat_len; iter += 2) {
1077 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
1078 					     get_unaligned_be16(&net_cat[iter]),
1079 					     GFP_ATOMIC);
1080 		if (ret_val != 0)
1081 			return ret_val;
1082 	}
1083 
1084 	return 0;
1085 }
1086 
1087 /**
1088  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
1089  * @doi_def: the DOI definition
1090  * @rngcat: category list
1091  * @rngcat_len: length of the category list in bytes
1092  *
1093  * Description:
1094  * Checks the given categories against the given DOI definition and returns a
1095  * negative value if any of the categories do not have a valid mapping and a
1096  * zero value if all of the categories are valid.
1097  *
1098  */
1099 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1100 				      const unsigned char *rngcat,
1101 				      u32 rngcat_len)
1102 {
1103 	u16 cat_high;
1104 	u16 cat_low;
1105 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1106 	u32 iter;
1107 
1108 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1109 		return -EFAULT;
1110 
1111 	for (iter = 0; iter < rngcat_len; iter += 4) {
1112 		cat_high = get_unaligned_be16(&rngcat[iter]);
1113 		if ((iter + 4) <= rngcat_len)
1114 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1115 		else
1116 			cat_low = 0;
1117 
1118 		if (cat_high > cat_prev)
1119 			return -EFAULT;
1120 
1121 		cat_prev = cat_low;
1122 	}
1123 
1124 	return 0;
1125 }
1126 
1127 /**
1128  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1129  * @doi_def: the DOI definition
1130  * @secattr: the security attributes
1131  * @net_cat: the zero'd out category list in network/CIPSO format
1132  * @net_cat_len: the length of the CIPSO category list in bytes
1133  *
1134  * Description:
1135  * Perform a label mapping to translate a local MLS category bitmap to the
1136  * correct CIPSO category list using the given DOI definition.   Returns the
1137  * size in bytes of the network category bitmap on success, negative values
1138  * otherwise.
1139  *
1140  */
1141 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1142 				     const struct netlbl_lsm_secattr *secattr,
1143 				     unsigned char *net_cat,
1144 				     u32 net_cat_len)
1145 {
1146 	int iter = -1;
1147 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1148 	u32 array_cnt = 0;
1149 	u32 cat_size = 0;
1150 
1151 	/* make sure we don't overflow the 'array[]' variable */
1152 	if (net_cat_len >
1153 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1154 		return -ENOSPC;
1155 
1156 	for (;;) {
1157 		iter = netlbl_catmap_walk(secattr->attr.mls.cat, iter + 1);
1158 		if (iter < 0)
1159 			break;
1160 		cat_size += (iter == 0 ? 0 : sizeof(u16));
1161 		if (cat_size > net_cat_len)
1162 			return -ENOSPC;
1163 		array[array_cnt++] = iter;
1164 
1165 		iter = netlbl_catmap_walkrng(secattr->attr.mls.cat, iter);
1166 		if (iter < 0)
1167 			return -EFAULT;
1168 		cat_size += sizeof(u16);
1169 		if (cat_size > net_cat_len)
1170 			return -ENOSPC;
1171 		array[array_cnt++] = iter;
1172 	}
1173 
1174 	for (iter = 0; array_cnt > 0;) {
1175 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1176 		iter += 2;
1177 		array_cnt--;
1178 		if (array[array_cnt] != 0) {
1179 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1180 			iter += 2;
1181 		}
1182 	}
1183 
1184 	return cat_size;
1185 }
1186 
1187 /**
1188  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1189  * @doi_def: the DOI definition
1190  * @net_cat: the category list in network/CIPSO format
1191  * @net_cat_len: the length of the CIPSO bitmap in bytes
1192  * @secattr: the security attributes
1193  *
1194  * Description:
1195  * Perform a label mapping to translate a CIPSO category list to the correct
1196  * local MLS category bitmap using the given DOI definition.  Returns zero on
1197  * success, negative values on failure.
1198  *
1199  */
1200 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1201 				     const unsigned char *net_cat,
1202 				     u32 net_cat_len,
1203 				     struct netlbl_lsm_secattr *secattr)
1204 {
1205 	int ret_val;
1206 	u32 net_iter;
1207 	u16 cat_low;
1208 	u16 cat_high;
1209 
1210 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1211 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1212 		if ((net_iter + 4) <= net_cat_len)
1213 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1214 		else
1215 			cat_low = 0;
1216 
1217 		ret_val = netlbl_catmap_setrng(&secattr->attr.mls.cat,
1218 					       cat_low,
1219 					       cat_high,
1220 					       GFP_ATOMIC);
1221 		if (ret_val != 0)
1222 			return ret_val;
1223 	}
1224 
1225 	return 0;
1226 }
1227 
1228 /*
1229  * Protocol Handling Functions
1230  */
1231 
1232 /**
1233  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1234  * @doi_def: the DOI definition
1235  * @len: the total tag length in bytes, not including this header
1236  * @buf: the CIPSO option buffer
1237  *
1238  * Description:
1239  * Write a CIPSO header into the beginning of @buffer.
1240  *
1241  */
1242 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1243 				unsigned char *buf,
1244 				u32 len)
1245 {
1246 	buf[0] = IPOPT_CIPSO;
1247 	buf[1] = CIPSO_V4_HDR_LEN + len;
1248 	*(__be32 *)&buf[2] = htonl(doi_def->doi);
1249 }
1250 
1251 /**
1252  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1253  * @doi_def: the DOI definition
1254  * @secattr: the security attributes
1255  * @buffer: the option buffer
1256  * @buffer_len: length of buffer in bytes
1257  *
1258  * Description:
1259  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1260  * actual buffer length may be larger than the indicated size due to
1261  * translation between host and network category bitmaps.  Returns the size of
1262  * the tag on success, negative values on failure.
1263  *
1264  */
1265 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1266 			       const struct netlbl_lsm_secattr *secattr,
1267 			       unsigned char *buffer,
1268 			       u32 buffer_len)
1269 {
1270 	int ret_val;
1271 	u32 tag_len;
1272 	u32 level;
1273 
1274 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1275 		return -EPERM;
1276 
1277 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1278 					secattr->attr.mls.lvl,
1279 					&level);
1280 	if (ret_val != 0)
1281 		return ret_val;
1282 
1283 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1284 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1285 						    secattr,
1286 						    &buffer[4],
1287 						    buffer_len - 4);
1288 		if (ret_val < 0)
1289 			return ret_val;
1290 
1291 		/* This will send packets using the "optimized" format when
1292 		 * possible as specified in  section 3.4.2.6 of the
1293 		 * CIPSO draft. */
1294 		if (cipso_v4_rbm_optfmt && ret_val > 0 && ret_val <= 10)
1295 			tag_len = 14;
1296 		else
1297 			tag_len = 4 + ret_val;
1298 	} else
1299 		tag_len = 4;
1300 
1301 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1302 	buffer[1] = tag_len;
1303 	buffer[3] = level;
1304 
1305 	return tag_len;
1306 }
1307 
1308 /**
1309  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1310  * @doi_def: the DOI definition
1311  * @tag: the CIPSO tag
1312  * @secattr: the security attributes
1313  *
1314  * Description:
1315  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1316  * attributes in @secattr.  Return zero on success, negatives values on
1317  * failure.
1318  *
1319  */
1320 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1321 				 const unsigned char *tag,
1322 				 struct netlbl_lsm_secattr *secattr)
1323 {
1324 	int ret_val;
1325 	u8 tag_len = tag[1];
1326 	u32 level;
1327 
1328 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1329 	if (ret_val != 0)
1330 		return ret_val;
1331 	secattr->attr.mls.lvl = level;
1332 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1333 
1334 	if (tag_len > 4) {
1335 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1336 						    &tag[4],
1337 						    tag_len - 4,
1338 						    secattr);
1339 		if (ret_val != 0) {
1340 			netlbl_catmap_free(secattr->attr.mls.cat);
1341 			return ret_val;
1342 		}
1343 
1344 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1345 	}
1346 
1347 	return 0;
1348 }
1349 
1350 /**
1351  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1352  * @doi_def: the DOI definition
1353  * @secattr: the security attributes
1354  * @buffer: the option buffer
1355  * @buffer_len: length of buffer in bytes
1356  *
1357  * Description:
1358  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1359  * size of the tag on success, negative values on failure.
1360  *
1361  */
1362 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1363 				const struct netlbl_lsm_secattr *secattr,
1364 				unsigned char *buffer,
1365 				u32 buffer_len)
1366 {
1367 	int ret_val;
1368 	u32 tag_len;
1369 	u32 level;
1370 
1371 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1372 		return -EPERM;
1373 
1374 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1375 					secattr->attr.mls.lvl,
1376 					&level);
1377 	if (ret_val != 0)
1378 		return ret_val;
1379 
1380 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1381 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1382 						     secattr,
1383 						     &buffer[4],
1384 						     buffer_len - 4);
1385 		if (ret_val < 0)
1386 			return ret_val;
1387 
1388 		tag_len = 4 + ret_val;
1389 	} else
1390 		tag_len = 4;
1391 
1392 	buffer[0] = CIPSO_V4_TAG_ENUM;
1393 	buffer[1] = tag_len;
1394 	buffer[3] = level;
1395 
1396 	return tag_len;
1397 }
1398 
1399 /**
1400  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1401  * @doi_def: the DOI definition
1402  * @tag: the CIPSO tag
1403  * @secattr: the security attributes
1404  *
1405  * Description:
1406  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1407  * attributes in @secattr.  Return zero on success, negatives values on
1408  * failure.
1409  *
1410  */
1411 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1412 				  const unsigned char *tag,
1413 				  struct netlbl_lsm_secattr *secattr)
1414 {
1415 	int ret_val;
1416 	u8 tag_len = tag[1];
1417 	u32 level;
1418 
1419 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1420 	if (ret_val != 0)
1421 		return ret_val;
1422 	secattr->attr.mls.lvl = level;
1423 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1424 
1425 	if (tag_len > 4) {
1426 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1427 						     &tag[4],
1428 						     tag_len - 4,
1429 						     secattr);
1430 		if (ret_val != 0) {
1431 			netlbl_catmap_free(secattr->attr.mls.cat);
1432 			return ret_val;
1433 		}
1434 
1435 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1436 	}
1437 
1438 	return 0;
1439 }
1440 
1441 /**
1442  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1443  * @doi_def: the DOI definition
1444  * @secattr: the security attributes
1445  * @buffer: the option buffer
1446  * @buffer_len: length of buffer in bytes
1447  *
1448  * Description:
1449  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1450  * size of the tag on success, negative values on failure.
1451  *
1452  */
1453 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1454 			       const struct netlbl_lsm_secattr *secattr,
1455 			       unsigned char *buffer,
1456 			       u32 buffer_len)
1457 {
1458 	int ret_val;
1459 	u32 tag_len;
1460 	u32 level;
1461 
1462 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1463 		return -EPERM;
1464 
1465 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1466 					secattr->attr.mls.lvl,
1467 					&level);
1468 	if (ret_val != 0)
1469 		return ret_val;
1470 
1471 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1472 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1473 						    secattr,
1474 						    &buffer[4],
1475 						    buffer_len - 4);
1476 		if (ret_val < 0)
1477 			return ret_val;
1478 
1479 		tag_len = 4 + ret_val;
1480 	} else
1481 		tag_len = 4;
1482 
1483 	buffer[0] = CIPSO_V4_TAG_RANGE;
1484 	buffer[1] = tag_len;
1485 	buffer[3] = level;
1486 
1487 	return tag_len;
1488 }
1489 
1490 /**
1491  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1492  * @doi_def: the DOI definition
1493  * @tag: the CIPSO tag
1494  * @secattr: the security attributes
1495  *
1496  * Description:
1497  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1498  * in @secattr.  Return zero on success, negatives values on failure.
1499  *
1500  */
1501 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1502 				 const unsigned char *tag,
1503 				 struct netlbl_lsm_secattr *secattr)
1504 {
1505 	int ret_val;
1506 	u8 tag_len = tag[1];
1507 	u32 level;
1508 
1509 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1510 	if (ret_val != 0)
1511 		return ret_val;
1512 	secattr->attr.mls.lvl = level;
1513 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1514 
1515 	if (tag_len > 4) {
1516 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1517 						    &tag[4],
1518 						    tag_len - 4,
1519 						    secattr);
1520 		if (ret_val != 0) {
1521 			netlbl_catmap_free(secattr->attr.mls.cat);
1522 			return ret_val;
1523 		}
1524 
1525 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1526 	}
1527 
1528 	return 0;
1529 }
1530 
1531 /**
1532  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1533  * @doi_def: the DOI definition
1534  * @secattr: the security attributes
1535  * @buffer: the option buffer
1536  * @buffer_len: length of buffer in bytes
1537  *
1538  * Description:
1539  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1540  * on success, negative values on failure.
1541  *
1542  */
1543 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1544 			       const struct netlbl_lsm_secattr *secattr,
1545 			       unsigned char *buffer,
1546 			       u32 buffer_len)
1547 {
1548 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1549 		return -EPERM;
1550 
1551 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1552 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1553 	*(u32 *)&buffer[2] = secattr->attr.secid;
1554 
1555 	return CIPSO_V4_TAG_LOC_BLEN;
1556 }
1557 
1558 /**
1559  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1560  * @doi_def: the DOI definition
1561  * @tag: the CIPSO tag
1562  * @secattr: the security attributes
1563  *
1564  * Description:
1565  * Parse a CIPSO local tag and return the security attributes in @secattr.
1566  * Return zero on success, negatives values on failure.
1567  *
1568  */
1569 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1570 				 const unsigned char *tag,
1571 				 struct netlbl_lsm_secattr *secattr)
1572 {
1573 	secattr->attr.secid = *(u32 *)&tag[2];
1574 	secattr->flags |= NETLBL_SECATTR_SECID;
1575 
1576 	return 0;
1577 }
1578 
1579 /**
1580  * cipso_v4_optptr - Find the CIPSO option in the packet
1581  * @skb: the packet
1582  *
1583  * Description:
1584  * Parse the packet's IP header looking for a CIPSO option.  Returns a pointer
1585  * to the start of the CIPSO option on success, NULL if one if not found.
1586  *
1587  */
1588 unsigned char *cipso_v4_optptr(const struct sk_buff *skb)
1589 {
1590 	const struct iphdr *iph = ip_hdr(skb);
1591 	unsigned char *optptr = (unsigned char *)&(ip_hdr(skb)[1]);
1592 	int optlen;
1593 	int taglen;
1594 
1595 	for (optlen = iph->ihl*4 - sizeof(struct iphdr); optlen > 0; ) {
1596 		if (optptr[0] == IPOPT_CIPSO)
1597 			return optptr;
1598 		taglen = optptr[1];
1599 		optlen -= taglen;
1600 		optptr += taglen;
1601 	}
1602 
1603 	return NULL;
1604 }
1605 
1606 /**
1607  * cipso_v4_validate - Validate a CIPSO option
1608  * @option: the start of the option, on error it is set to point to the error
1609  *
1610  * Description:
1611  * This routine is called to validate a CIPSO option, it checks all of the
1612  * fields to ensure that they are at least valid, see the draft snippet below
1613  * for details.  If the option is valid then a zero value is returned and
1614  * the value of @option is unchanged.  If the option is invalid then a
1615  * non-zero value is returned and @option is adjusted to point to the
1616  * offending portion of the option.  From the IETF draft ...
1617  *
1618  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1619  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1620  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1621  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1622  *   that is unrecognized."
1623  *
1624  */
1625 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1626 {
1627 	unsigned char *opt = *option;
1628 	unsigned char *tag;
1629 	unsigned char opt_iter;
1630 	unsigned char err_offset = 0;
1631 	u8 opt_len;
1632 	u8 tag_len;
1633 	struct cipso_v4_doi *doi_def = NULL;
1634 	u32 tag_iter;
1635 
1636 	/* caller already checks for length values that are too large */
1637 	opt_len = opt[1];
1638 	if (opt_len < 8) {
1639 		err_offset = 1;
1640 		goto validate_return;
1641 	}
1642 
1643 	rcu_read_lock();
1644 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1645 	if (!doi_def) {
1646 		err_offset = 2;
1647 		goto validate_return_locked;
1648 	}
1649 
1650 	opt_iter = CIPSO_V4_HDR_LEN;
1651 	tag = opt + opt_iter;
1652 	while (opt_iter < opt_len) {
1653 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1654 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1655 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1656 				err_offset = opt_iter;
1657 				goto validate_return_locked;
1658 			}
1659 
1660 		tag_len = tag[1];
1661 		if (tag_len > (opt_len - opt_iter)) {
1662 			err_offset = opt_iter + 1;
1663 			goto validate_return_locked;
1664 		}
1665 
1666 		switch (tag[0]) {
1667 		case CIPSO_V4_TAG_RBITMAP:
1668 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1669 				err_offset = opt_iter + 1;
1670 				goto validate_return_locked;
1671 			}
1672 
1673 			/* We are already going to do all the verification
1674 			 * necessary at the socket layer so from our point of
1675 			 * view it is safe to turn these checks off (and less
1676 			 * work), however, the CIPSO draft says we should do
1677 			 * all the CIPSO validations here but it doesn't
1678 			 * really specify _exactly_ what we need to validate
1679 			 * ... so, just make it a sysctl tunable. */
1680 			if (cipso_v4_rbm_strictvalid) {
1681 				if (cipso_v4_map_lvl_valid(doi_def,
1682 							   tag[3]) < 0) {
1683 					err_offset = opt_iter + 3;
1684 					goto validate_return_locked;
1685 				}
1686 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1687 				    cipso_v4_map_cat_rbm_valid(doi_def,
1688 							    &tag[4],
1689 							    tag_len - 4) < 0) {
1690 					err_offset = opt_iter + 4;
1691 					goto validate_return_locked;
1692 				}
1693 			}
1694 			break;
1695 		case CIPSO_V4_TAG_ENUM:
1696 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1697 				err_offset = opt_iter + 1;
1698 				goto validate_return_locked;
1699 			}
1700 
1701 			if (cipso_v4_map_lvl_valid(doi_def,
1702 						   tag[3]) < 0) {
1703 				err_offset = opt_iter + 3;
1704 				goto validate_return_locked;
1705 			}
1706 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1707 			    cipso_v4_map_cat_enum_valid(doi_def,
1708 							&tag[4],
1709 							tag_len - 4) < 0) {
1710 				err_offset = opt_iter + 4;
1711 				goto validate_return_locked;
1712 			}
1713 			break;
1714 		case CIPSO_V4_TAG_RANGE:
1715 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1716 				err_offset = opt_iter + 1;
1717 				goto validate_return_locked;
1718 			}
1719 
1720 			if (cipso_v4_map_lvl_valid(doi_def,
1721 						   tag[3]) < 0) {
1722 				err_offset = opt_iter + 3;
1723 				goto validate_return_locked;
1724 			}
1725 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1726 			    cipso_v4_map_cat_rng_valid(doi_def,
1727 						       &tag[4],
1728 						       tag_len - 4) < 0) {
1729 				err_offset = opt_iter + 4;
1730 				goto validate_return_locked;
1731 			}
1732 			break;
1733 		case CIPSO_V4_TAG_LOCAL:
1734 			/* This is a non-standard tag that we only allow for
1735 			 * local connections, so if the incoming interface is
1736 			 * not the loopback device drop the packet. Further,
1737 			 * there is no legitimate reason for setting this from
1738 			 * userspace so reject it if skb is NULL. */
1739 			if (!skb || !(skb->dev->flags & IFF_LOOPBACK)) {
1740 				err_offset = opt_iter;
1741 				goto validate_return_locked;
1742 			}
1743 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1744 				err_offset = opt_iter + 1;
1745 				goto validate_return_locked;
1746 			}
1747 			break;
1748 		default:
1749 			err_offset = opt_iter;
1750 			goto validate_return_locked;
1751 		}
1752 
1753 		tag += tag_len;
1754 		opt_iter += tag_len;
1755 	}
1756 
1757 validate_return_locked:
1758 	rcu_read_unlock();
1759 validate_return:
1760 	*option = opt + err_offset;
1761 	return err_offset;
1762 }
1763 
1764 /**
1765  * cipso_v4_error - Send the correct response for a bad packet
1766  * @skb: the packet
1767  * @error: the error code
1768  * @gateway: CIPSO gateway flag
1769  *
1770  * Description:
1771  * Based on the error code given in @error, send an ICMP error message back to
1772  * the originating host.  From the IETF draft ...
1773  *
1774  *  "If the contents of the CIPSO [option] are valid but the security label is
1775  *   outside of the configured host or port label range, the datagram is
1776  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1777  *   returned.  The code field of the ICMP is set to 'communication with
1778  *   destination network administratively prohibited' (code 9) or to
1779  *   'communication with destination host administratively prohibited'
1780  *   (code 10).  The value of the code is dependent on whether the originator
1781  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1782  *   recipient of the ICMP message MUST be able to handle either value.  The
1783  *   same procedure is performed if a CIPSO [option] can not be added to an
1784  *   IP packet because it is too large to fit in the IP options area."
1785  *
1786  *  "If the error is triggered by receipt of an ICMP message, the message is
1787  *   discarded and no response is permitted (consistent with general ICMP
1788  *   processing rules)."
1789  *
1790  */
1791 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1792 {
1793 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1794 		return;
1795 
1796 	if (gateway)
1797 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0);
1798 	else
1799 		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0);
1800 }
1801 
1802 /**
1803  * cipso_v4_genopt - Generate a CIPSO option
1804  * @buf: the option buffer
1805  * @buf_len: the size of opt_buf
1806  * @doi_def: the CIPSO DOI to use
1807  * @secattr: the security attributes
1808  *
1809  * Description:
1810  * Generate a CIPSO option using the DOI definition and security attributes
1811  * passed to the function.  Returns the length of the option on success and
1812  * negative values on failure.
1813  *
1814  */
1815 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1816 			   const struct cipso_v4_doi *doi_def,
1817 			   const struct netlbl_lsm_secattr *secattr)
1818 {
1819 	int ret_val;
1820 	u32 iter;
1821 
1822 	if (buf_len <= CIPSO_V4_HDR_LEN)
1823 		return -ENOSPC;
1824 
1825 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1826 	 * really a good assumption to make but since we only support the MAC
1827 	 * tags right now it is a safe assumption. */
1828 	iter = 0;
1829 	do {
1830 		memset(buf, 0, buf_len);
1831 		switch (doi_def->tags[iter]) {
1832 		case CIPSO_V4_TAG_RBITMAP:
1833 			ret_val = cipso_v4_gentag_rbm(doi_def,
1834 						   secattr,
1835 						   &buf[CIPSO_V4_HDR_LEN],
1836 						   buf_len - CIPSO_V4_HDR_LEN);
1837 			break;
1838 		case CIPSO_V4_TAG_ENUM:
1839 			ret_val = cipso_v4_gentag_enum(doi_def,
1840 						   secattr,
1841 						   &buf[CIPSO_V4_HDR_LEN],
1842 						   buf_len - CIPSO_V4_HDR_LEN);
1843 			break;
1844 		case CIPSO_V4_TAG_RANGE:
1845 			ret_val = cipso_v4_gentag_rng(doi_def,
1846 						   secattr,
1847 						   &buf[CIPSO_V4_HDR_LEN],
1848 						   buf_len - CIPSO_V4_HDR_LEN);
1849 			break;
1850 		case CIPSO_V4_TAG_LOCAL:
1851 			ret_val = cipso_v4_gentag_loc(doi_def,
1852 						   secattr,
1853 						   &buf[CIPSO_V4_HDR_LEN],
1854 						   buf_len - CIPSO_V4_HDR_LEN);
1855 			break;
1856 		default:
1857 			return -EPERM;
1858 		}
1859 
1860 		iter++;
1861 	} while (ret_val < 0 &&
1862 		 iter < CIPSO_V4_TAG_MAXCNT &&
1863 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1864 	if (ret_val < 0)
1865 		return ret_val;
1866 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1867 	return CIPSO_V4_HDR_LEN + ret_val;
1868 }
1869 
1870 /**
1871  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1872  * @sk: the socket
1873  * @doi_def: the CIPSO DOI to use
1874  * @secattr: the specific security attributes of the socket
1875  *
1876  * Description:
1877  * Set the CIPSO option on the given socket using the DOI definition and
1878  * security attributes passed to the function.  This function requires
1879  * exclusive access to @sk, which means it either needs to be in the
1880  * process of being created or locked.  Returns zero on success and negative
1881  * values on failure.
1882  *
1883  */
1884 int cipso_v4_sock_setattr(struct sock *sk,
1885 			  const struct cipso_v4_doi *doi_def,
1886 			  const struct netlbl_lsm_secattr *secattr)
1887 {
1888 	int ret_val = -EPERM;
1889 	unsigned char *buf = NULL;
1890 	u32 buf_len;
1891 	u32 opt_len;
1892 	struct ip_options_rcu *old, *opt = NULL;
1893 	struct inet_sock *sk_inet;
1894 	struct inet_connection_sock *sk_conn;
1895 
1896 	/* In the case of sock_create_lite(), the sock->sk field is not
1897 	 * defined yet but it is not a problem as the only users of these
1898 	 * "lite" PF_INET sockets are functions which do an accept() call
1899 	 * afterwards so we will label the socket as part of the accept(). */
1900 	if (!sk)
1901 		return 0;
1902 
1903 	/* We allocate the maximum CIPSO option size here so we are probably
1904 	 * being a little wasteful, but it makes our life _much_ easier later
1905 	 * on and after all we are only talking about 40 bytes. */
1906 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1907 	buf = kmalloc(buf_len, GFP_ATOMIC);
1908 	if (!buf) {
1909 		ret_val = -ENOMEM;
1910 		goto socket_setattr_failure;
1911 	}
1912 
1913 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1914 	if (ret_val < 0)
1915 		goto socket_setattr_failure;
1916 	buf_len = ret_val;
1917 
1918 	/* We can't use ip_options_get() directly because it makes a call to
1919 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1920 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1921 	 * set the IPOPT_CIPSO option. */
1922 	opt_len = (buf_len + 3) & ~3;
1923 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1924 	if (!opt) {
1925 		ret_val = -ENOMEM;
1926 		goto socket_setattr_failure;
1927 	}
1928 	memcpy(opt->opt.__data, buf, buf_len);
1929 	opt->opt.optlen = opt_len;
1930 	opt->opt.cipso = sizeof(struct iphdr);
1931 	kfree(buf);
1932 	buf = NULL;
1933 
1934 	sk_inet = inet_sk(sk);
1935 
1936 	old = rcu_dereference_protected(sk_inet->inet_opt, sock_owned_by_user(sk));
1937 	if (sk_inet->is_icsk) {
1938 		sk_conn = inet_csk(sk);
1939 		if (old)
1940 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1941 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1942 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1943 	}
1944 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1945 	if (old)
1946 		kfree_rcu(old, rcu);
1947 
1948 	return 0;
1949 
1950 socket_setattr_failure:
1951 	kfree(buf);
1952 	kfree(opt);
1953 	return ret_val;
1954 }
1955 
1956 /**
1957  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1958  * @req: the connection request socket
1959  * @doi_def: the CIPSO DOI to use
1960  * @secattr: the specific security attributes of the socket
1961  *
1962  * Description:
1963  * Set the CIPSO option on the given socket using the DOI definition and
1964  * security attributes passed to the function.  Returns zero on success and
1965  * negative values on failure.
1966  *
1967  */
1968 int cipso_v4_req_setattr(struct request_sock *req,
1969 			 const struct cipso_v4_doi *doi_def,
1970 			 const struct netlbl_lsm_secattr *secattr)
1971 {
1972 	int ret_val = -EPERM;
1973 	unsigned char *buf = NULL;
1974 	u32 buf_len;
1975 	u32 opt_len;
1976 	struct ip_options_rcu *opt = NULL;
1977 	struct inet_request_sock *req_inet;
1978 
1979 	/* We allocate the maximum CIPSO option size here so we are probably
1980 	 * being a little wasteful, but it makes our life _much_ easier later
1981 	 * on and after all we are only talking about 40 bytes. */
1982 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1983 	buf = kmalloc(buf_len, GFP_ATOMIC);
1984 	if (!buf) {
1985 		ret_val = -ENOMEM;
1986 		goto req_setattr_failure;
1987 	}
1988 
1989 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1990 	if (ret_val < 0)
1991 		goto req_setattr_failure;
1992 	buf_len = ret_val;
1993 
1994 	/* We can't use ip_options_get() directly because it makes a call to
1995 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1996 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1997 	 * set the IPOPT_CIPSO option. */
1998 	opt_len = (buf_len + 3) & ~3;
1999 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
2000 	if (!opt) {
2001 		ret_val = -ENOMEM;
2002 		goto req_setattr_failure;
2003 	}
2004 	memcpy(opt->opt.__data, buf, buf_len);
2005 	opt->opt.optlen = opt_len;
2006 	opt->opt.cipso = sizeof(struct iphdr);
2007 	kfree(buf);
2008 	buf = NULL;
2009 
2010 	req_inet = inet_rsk(req);
2011 	opt = xchg(&req_inet->opt, opt);
2012 	if (opt)
2013 		kfree_rcu(opt, rcu);
2014 
2015 	return 0;
2016 
2017 req_setattr_failure:
2018 	kfree(buf);
2019 	kfree(opt);
2020 	return ret_val;
2021 }
2022 
2023 /**
2024  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
2025  * @opt_ptr: IP option pointer
2026  *
2027  * Description:
2028  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
2029  * adjustments to the IP option structure.  Returns zero on success, negative
2030  * values on failure.
2031  *
2032  */
2033 static int cipso_v4_delopt(struct ip_options_rcu **opt_ptr)
2034 {
2035 	int hdr_delta = 0;
2036 	struct ip_options_rcu *opt = *opt_ptr;
2037 
2038 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
2039 		u8 cipso_len;
2040 		u8 cipso_off;
2041 		unsigned char *cipso_ptr;
2042 		int iter;
2043 		int optlen_new;
2044 
2045 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
2046 		cipso_ptr = &opt->opt.__data[cipso_off];
2047 		cipso_len = cipso_ptr[1];
2048 
2049 		if (opt->opt.srr > opt->opt.cipso)
2050 			opt->opt.srr -= cipso_len;
2051 		if (opt->opt.rr > opt->opt.cipso)
2052 			opt->opt.rr -= cipso_len;
2053 		if (opt->opt.ts > opt->opt.cipso)
2054 			opt->opt.ts -= cipso_len;
2055 		if (opt->opt.router_alert > opt->opt.cipso)
2056 			opt->opt.router_alert -= cipso_len;
2057 		opt->opt.cipso = 0;
2058 
2059 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2060 			opt->opt.optlen - cipso_off - cipso_len);
2061 
2062 		/* determining the new total option length is tricky because of
2063 		 * the padding necessary, the only thing i can think to do at
2064 		 * this point is walk the options one-by-one, skipping the
2065 		 * padding at the end to determine the actual option size and
2066 		 * from there we can determine the new total option length */
2067 		iter = 0;
2068 		optlen_new = 0;
2069 		while (iter < opt->opt.optlen)
2070 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2071 				iter += opt->opt.__data[iter + 1];
2072 				optlen_new = iter;
2073 			} else
2074 				iter++;
2075 		hdr_delta = opt->opt.optlen;
2076 		opt->opt.optlen = (optlen_new + 3) & ~3;
2077 		hdr_delta -= opt->opt.optlen;
2078 	} else {
2079 		/* only the cipso option was present on the socket so we can
2080 		 * remove the entire option struct */
2081 		*opt_ptr = NULL;
2082 		hdr_delta = opt->opt.optlen;
2083 		kfree_rcu(opt, rcu);
2084 	}
2085 
2086 	return hdr_delta;
2087 }
2088 
2089 /**
2090  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2091  * @sk: the socket
2092  *
2093  * Description:
2094  * Removes the CIPSO option from a socket, if present.
2095  *
2096  */
2097 void cipso_v4_sock_delattr(struct sock *sk)
2098 {
2099 	int hdr_delta;
2100 	struct ip_options_rcu *opt;
2101 	struct inet_sock *sk_inet;
2102 
2103 	sk_inet = inet_sk(sk);
2104 	opt = rcu_dereference_protected(sk_inet->inet_opt, 1);
2105 	if (!opt || opt->opt.cipso == 0)
2106 		return;
2107 
2108 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2109 	if (sk_inet->is_icsk && hdr_delta > 0) {
2110 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2111 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2112 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2113 	}
2114 }
2115 
2116 /**
2117  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2118  * @reg: the request socket
2119  *
2120  * Description:
2121  * Removes the CIPSO option from a request socket, if present.
2122  *
2123  */
2124 void cipso_v4_req_delattr(struct request_sock *req)
2125 {
2126 	struct ip_options_rcu *opt;
2127 	struct inet_request_sock *req_inet;
2128 
2129 	req_inet = inet_rsk(req);
2130 	opt = req_inet->opt;
2131 	if (!opt || opt->opt.cipso == 0)
2132 		return;
2133 
2134 	cipso_v4_delopt(&req_inet->opt);
2135 }
2136 
2137 /**
2138  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2139  * @cipso: the CIPSO v4 option
2140  * @secattr: the security attributes
2141  *
2142  * Description:
2143  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2144  * on success and negative values on failure.
2145  *
2146  */
2147 int cipso_v4_getattr(const unsigned char *cipso,
2148 		     struct netlbl_lsm_secattr *secattr)
2149 {
2150 	int ret_val = -ENOMSG;
2151 	u32 doi;
2152 	struct cipso_v4_doi *doi_def;
2153 
2154 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2155 		return 0;
2156 
2157 	doi = get_unaligned_be32(&cipso[2]);
2158 	rcu_read_lock();
2159 	doi_def = cipso_v4_doi_search(doi);
2160 	if (!doi_def)
2161 		goto getattr_return;
2162 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2163 	 * really a good assumption to make but since we only support the MAC
2164 	 * tags right now it is a safe assumption. */
2165 	switch (cipso[6]) {
2166 	case CIPSO_V4_TAG_RBITMAP:
2167 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2168 		break;
2169 	case CIPSO_V4_TAG_ENUM:
2170 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2171 		break;
2172 	case CIPSO_V4_TAG_RANGE:
2173 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2174 		break;
2175 	case CIPSO_V4_TAG_LOCAL:
2176 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2177 		break;
2178 	}
2179 	if (ret_val == 0)
2180 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2181 
2182 getattr_return:
2183 	rcu_read_unlock();
2184 	return ret_val;
2185 }
2186 
2187 /**
2188  * cipso_v4_sock_getattr - Get the security attributes from a sock
2189  * @sk: the sock
2190  * @secattr: the security attributes
2191  *
2192  * Description:
2193  * Query @sk to see if there is a CIPSO option attached to the sock and if
2194  * there is return the CIPSO security attributes in @secattr.  This function
2195  * requires that @sk be locked, or privately held, but it does not do any
2196  * locking itself.  Returns zero on success and negative values on failure.
2197  *
2198  */
2199 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2200 {
2201 	struct ip_options_rcu *opt;
2202 	int res = -ENOMSG;
2203 
2204 	rcu_read_lock();
2205 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2206 	if (opt && opt->opt.cipso)
2207 		res = cipso_v4_getattr(opt->opt.__data +
2208 						opt->opt.cipso -
2209 						sizeof(struct iphdr),
2210 				       secattr);
2211 	rcu_read_unlock();
2212 	return res;
2213 }
2214 
2215 /**
2216  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2217  * @skb: the packet
2218  * @secattr: the security attributes
2219  *
2220  * Description:
2221  * Set the CIPSO option on the given packet based on the security attributes.
2222  * Returns a pointer to the IP header on success and NULL on failure.
2223  *
2224  */
2225 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2226 			    const struct cipso_v4_doi *doi_def,
2227 			    const struct netlbl_lsm_secattr *secattr)
2228 {
2229 	int ret_val;
2230 	struct iphdr *iph;
2231 	struct ip_options *opt = &IPCB(skb)->opt;
2232 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2233 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2234 	u32 opt_len;
2235 	int len_delta;
2236 
2237 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2238 	if (ret_val < 0)
2239 		return ret_val;
2240 	buf_len = ret_val;
2241 	opt_len = (buf_len + 3) & ~3;
2242 
2243 	/* we overwrite any existing options to ensure that we have enough
2244 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2245 	 * that the security label is applied to the packet - we do the same
2246 	 * thing when using the socket options and it hasn't caused a problem,
2247 	 * if we need to we can always revisit this choice later */
2248 
2249 	len_delta = opt_len - opt->optlen;
2250 	/* if we don't ensure enough headroom we could panic on the skb_push()
2251 	 * call below so make sure we have enough, we are also "mangling" the
2252 	 * packet so we should probably do a copy-on-write call anyway */
2253 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2254 	if (ret_val < 0)
2255 		return ret_val;
2256 
2257 	if (len_delta > 0) {
2258 		/* we assume that the header + opt->optlen have already been
2259 		 * "pushed" in ip_options_build() or similar */
2260 		iph = ip_hdr(skb);
2261 		skb_push(skb, len_delta);
2262 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2263 		skb_reset_network_header(skb);
2264 		iph = ip_hdr(skb);
2265 	} else if (len_delta < 0) {
2266 		iph = ip_hdr(skb);
2267 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2268 	} else
2269 		iph = ip_hdr(skb);
2270 
2271 	if (opt->optlen > 0)
2272 		memset(opt, 0, sizeof(*opt));
2273 	opt->optlen = opt_len;
2274 	opt->cipso = sizeof(struct iphdr);
2275 	opt->is_changed = 1;
2276 
2277 	/* we have to do the following because we are being called from a
2278 	 * netfilter hook which means the packet already has had the header
2279 	 * fields populated and the checksum calculated - yes this means we
2280 	 * are doing more work than needed but we do it to keep the core
2281 	 * stack clean and tidy */
2282 	memcpy(iph + 1, buf, buf_len);
2283 	if (opt_len > buf_len)
2284 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2285 	if (len_delta != 0) {
2286 		iph->ihl = 5 + (opt_len >> 2);
2287 		iph->tot_len = htons(skb->len);
2288 	}
2289 	ip_send_check(iph);
2290 
2291 	return 0;
2292 }
2293 
2294 /**
2295  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2296  * @skb: the packet
2297  *
2298  * Description:
2299  * Removes any and all CIPSO options from the given packet.  Returns zero on
2300  * success, negative values on failure.
2301  *
2302  */
2303 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2304 {
2305 	int ret_val;
2306 	struct iphdr *iph;
2307 	struct ip_options *opt = &IPCB(skb)->opt;
2308 	unsigned char *cipso_ptr;
2309 
2310 	if (opt->cipso == 0)
2311 		return 0;
2312 
2313 	/* since we are changing the packet we should make a copy */
2314 	ret_val = skb_cow(skb, skb_headroom(skb));
2315 	if (ret_val < 0)
2316 		return ret_val;
2317 
2318 	/* the easiest thing to do is just replace the cipso option with noop
2319 	 * options since we don't change the size of the packet, although we
2320 	 * still need to recalculate the checksum */
2321 
2322 	iph = ip_hdr(skb);
2323 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2324 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2325 	opt->cipso = 0;
2326 	opt->is_changed = 1;
2327 
2328 	ip_send_check(iph);
2329 
2330 	return 0;
2331 }
2332 
2333 /*
2334  * Setup Functions
2335  */
2336 
2337 /**
2338  * cipso_v4_init - Initialize the CIPSO module
2339  *
2340  * Description:
2341  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2342  * and negative values on failure.
2343  *
2344  */
2345 static int __init cipso_v4_init(void)
2346 {
2347 	int ret_val;
2348 
2349 	ret_val = cipso_v4_cache_init();
2350 	if (ret_val != 0)
2351 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2352 		      ret_val);
2353 
2354 	return 0;
2355 }
2356 
2357 subsys_initcall(cipso_v4_init);
2358