xref: /linux/net/ipv4/cipso_ipv4.c (revision 900897591b2bf01154db8f59f6ecf8d60e60f15e)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * CIPSO - Commercial IP Security Option
4  *
5  * This is an implementation of the CIPSO 2.2 protocol as specified in
6  * draft-ietf-cipso-ipsecurity-01.txt with additional tag types as found in
7  * FIPS-188.  While CIPSO never became a full IETF RFC standard many vendors
8  * have chosen to adopt the protocol and over the years it has become a
9  * de-facto standard for labeled networking.
10  *
11  * The CIPSO draft specification can be found in the kernel's Documentation
12  * directory as well as the following URL:
13  *   http://tools.ietf.org/id/draft-ietf-cipso-ipsecurity-01.txt
14  * The FIPS-188 specification can be found at the following URL:
15  *   http://www.itl.nist.gov/fipspubs/fip188.htm
16  *
17  * Author: Paul Moore <paul.moore@hp.com>
18  */
19 
20 /*
21  * (c) Copyright Hewlett-Packard Development Company, L.P., 2006, 2008
22  */
23 
24 #include <linux/init.h>
25 #include <linux/types.h>
26 #include <linux/rcupdate.h>
27 #include <linux/list.h>
28 #include <linux/spinlock.h>
29 #include <linux/string.h>
30 #include <linux/jhash.h>
31 #include <linux/audit.h>
32 #include <linux/slab.h>
33 #include <net/ip.h>
34 #include <net/icmp.h>
35 #include <net/tcp.h>
36 #include <net/netlabel.h>
37 #include <net/cipso_ipv4.h>
38 #include <linux/atomic.h>
39 #include <linux/bug.h>
40 #include <asm/unaligned.h>
41 
42 /* List of available DOI definitions */
43 /* XXX - This currently assumes a minimal number of different DOIs in use,
44  * if in practice there are a lot of different DOIs this list should
45  * probably be turned into a hash table or something similar so we
46  * can do quick lookups. */
47 static DEFINE_SPINLOCK(cipso_v4_doi_list_lock);
48 static LIST_HEAD(cipso_v4_doi_list);
49 
50 /* Label mapping cache */
51 int cipso_v4_cache_enabled = 1;
52 int cipso_v4_cache_bucketsize = 10;
53 #define CIPSO_V4_CACHE_BUCKETBITS     7
54 #define CIPSO_V4_CACHE_BUCKETS        (1 << CIPSO_V4_CACHE_BUCKETBITS)
55 #define CIPSO_V4_CACHE_REORDERLIMIT   10
56 struct cipso_v4_map_cache_bkt {
57 	spinlock_t lock;
58 	u32 size;
59 	struct list_head list;
60 };
61 
62 struct cipso_v4_map_cache_entry {
63 	u32 hash;
64 	unsigned char *key;
65 	size_t key_len;
66 
67 	struct netlbl_lsm_cache *lsm_data;
68 
69 	u32 activity;
70 	struct list_head list;
71 };
72 
73 static struct cipso_v4_map_cache_bkt *cipso_v4_cache;
74 
75 /* Restricted bitmap (tag #1) flags */
76 int cipso_v4_rbm_optfmt = 0;
77 int cipso_v4_rbm_strictvalid = 1;
78 
79 /*
80  * Protocol Constants
81  */
82 
83 /* Maximum size of the CIPSO IP option, derived from the fact that the maximum
84  * IPv4 header size is 60 bytes and the base IPv4 header is 20 bytes long. */
85 #define CIPSO_V4_OPT_LEN_MAX          40
86 
87 /* Length of the base CIPSO option, this includes the option type (1 byte), the
88  * option length (1 byte), and the DOI (4 bytes). */
89 #define CIPSO_V4_HDR_LEN              6
90 
91 /* Base length of the restrictive category bitmap tag (tag #1). */
92 #define CIPSO_V4_TAG_RBM_BLEN         4
93 
94 /* Base length of the enumerated category tag (tag #2). */
95 #define CIPSO_V4_TAG_ENUM_BLEN        4
96 
97 /* Base length of the ranged categories bitmap tag (tag #5). */
98 #define CIPSO_V4_TAG_RNG_BLEN         4
99 /* The maximum number of category ranges permitted in the ranged category tag
100  * (tag #5).  You may note that the IETF draft states that the maximum number
101  * of category ranges is 7, but if the low end of the last category range is
102  * zero then it is possible to fit 8 category ranges because the zero should
103  * be omitted. */
104 #define CIPSO_V4_TAG_RNG_CAT_MAX      8
105 
106 /* Base length of the local tag (non-standard tag).
107  *  Tag definition (may change between kernel versions)
108  *
109  * 0          8          16         24         32
110  * +----------+----------+----------+----------+
111  * | 10000000 | 00000110 | 32-bit secid value  |
112  * +----------+----------+----------+----------+
113  * | in (host byte order)|
114  * +----------+----------+
115  *
116  */
117 #define CIPSO_V4_TAG_LOC_BLEN         6
118 
119 /*
120  * Helper Functions
121  */
122 
123 /**
124  * cipso_v4_cache_entry_free - Frees a cache entry
125  * @entry: the entry to free
126  *
127  * Description:
128  * This function frees the memory associated with a cache entry including the
129  * LSM cache data if there are no longer any users, i.e. reference count == 0.
130  *
131  */
132 static void cipso_v4_cache_entry_free(struct cipso_v4_map_cache_entry *entry)
133 {
134 	if (entry->lsm_data)
135 		netlbl_secattr_cache_free(entry->lsm_data);
136 	kfree(entry->key);
137 	kfree(entry);
138 }
139 
140 /**
141  * cipso_v4_map_cache_hash - Hashing function for the CIPSO cache
142  * @key: the hash key
143  * @key_len: the length of the key in bytes
144  *
145  * Description:
146  * The CIPSO tag hashing function.  Returns a 32-bit hash value.
147  *
148  */
149 static u32 cipso_v4_map_cache_hash(const unsigned char *key, u32 key_len)
150 {
151 	return jhash(key, key_len, 0);
152 }
153 
154 /*
155  * Label Mapping Cache Functions
156  */
157 
158 /**
159  * cipso_v4_cache_init - Initialize the CIPSO cache
160  *
161  * Description:
162  * Initializes the CIPSO label mapping cache, this function should be called
163  * before any of the other functions defined in this file.  Returns zero on
164  * success, negative values on error.
165  *
166  */
167 static int __init cipso_v4_cache_init(void)
168 {
169 	u32 iter;
170 
171 	cipso_v4_cache = kcalloc(CIPSO_V4_CACHE_BUCKETS,
172 				 sizeof(struct cipso_v4_map_cache_bkt),
173 				 GFP_KERNEL);
174 	if (!cipso_v4_cache)
175 		return -ENOMEM;
176 
177 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
178 		spin_lock_init(&cipso_v4_cache[iter].lock);
179 		cipso_v4_cache[iter].size = 0;
180 		INIT_LIST_HEAD(&cipso_v4_cache[iter].list);
181 	}
182 
183 	return 0;
184 }
185 
186 /**
187  * cipso_v4_cache_invalidate - Invalidates the current CIPSO cache
188  *
189  * Description:
190  * Invalidates and frees any entries in the CIPSO cache.  Returns zero on
191  * success and negative values on failure.
192  *
193  */
194 void cipso_v4_cache_invalidate(void)
195 {
196 	struct cipso_v4_map_cache_entry *entry, *tmp_entry;
197 	u32 iter;
198 
199 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
200 		spin_lock_bh(&cipso_v4_cache[iter].lock);
201 		list_for_each_entry_safe(entry,
202 					 tmp_entry,
203 					 &cipso_v4_cache[iter].list, list) {
204 			list_del(&entry->list);
205 			cipso_v4_cache_entry_free(entry);
206 		}
207 		cipso_v4_cache[iter].size = 0;
208 		spin_unlock_bh(&cipso_v4_cache[iter].lock);
209 	}
210 }
211 
212 /**
213  * cipso_v4_cache_check - Check the CIPSO cache for a label mapping
214  * @key: the buffer to check
215  * @key_len: buffer length in bytes
216  * @secattr: the security attribute struct to use
217  *
218  * Description:
219  * This function checks the cache to see if a label mapping already exists for
220  * the given key.  If there is a match then the cache is adjusted and the
221  * @secattr struct is populated with the correct LSM security attributes.  The
222  * cache is adjusted in the following manner if the entry is not already the
223  * first in the cache bucket:
224  *
225  *  1. The cache entry's activity counter is incremented
226  *  2. The previous (higher ranking) entry's activity counter is decremented
227  *  3. If the difference between the two activity counters is geater than
228  *     CIPSO_V4_CACHE_REORDERLIMIT the two entries are swapped
229  *
230  * Returns zero on success, -ENOENT for a cache miss, and other negative values
231  * on error.
232  *
233  */
234 static int cipso_v4_cache_check(const unsigned char *key,
235 				u32 key_len,
236 				struct netlbl_lsm_secattr *secattr)
237 {
238 	u32 bkt;
239 	struct cipso_v4_map_cache_entry *entry;
240 	struct cipso_v4_map_cache_entry *prev_entry = NULL;
241 	u32 hash;
242 
243 	if (!cipso_v4_cache_enabled)
244 		return -ENOENT;
245 
246 	hash = cipso_v4_map_cache_hash(key, key_len);
247 	bkt = hash & (CIPSO_V4_CACHE_BUCKETS - 1);
248 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
249 	list_for_each_entry(entry, &cipso_v4_cache[bkt].list, list) {
250 		if (entry->hash == hash &&
251 		    entry->key_len == key_len &&
252 		    memcmp(entry->key, key, key_len) == 0) {
253 			entry->activity += 1;
254 			refcount_inc(&entry->lsm_data->refcount);
255 			secattr->cache = entry->lsm_data;
256 			secattr->flags |= NETLBL_SECATTR_CACHE;
257 			secattr->type = NETLBL_NLTYPE_CIPSOV4;
258 			if (!prev_entry) {
259 				spin_unlock_bh(&cipso_v4_cache[bkt].lock);
260 				return 0;
261 			}
262 
263 			if (prev_entry->activity > 0)
264 				prev_entry->activity -= 1;
265 			if (entry->activity > prev_entry->activity &&
266 			    entry->activity - prev_entry->activity >
267 			    CIPSO_V4_CACHE_REORDERLIMIT) {
268 				__list_del(entry->list.prev, entry->list.next);
269 				__list_add(&entry->list,
270 					   prev_entry->list.prev,
271 					   &prev_entry->list);
272 			}
273 
274 			spin_unlock_bh(&cipso_v4_cache[bkt].lock);
275 			return 0;
276 		}
277 		prev_entry = entry;
278 	}
279 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
280 
281 	return -ENOENT;
282 }
283 
284 /**
285  * cipso_v4_cache_add - Add an entry to the CIPSO cache
286  * @skb: the packet
287  * @secattr: the packet's security attributes
288  *
289  * Description:
290  * Add a new entry into the CIPSO label mapping cache.  Add the new entry to
291  * head of the cache bucket's list, if the cache bucket is out of room remove
292  * the last entry in the list first.  It is important to note that there is
293  * currently no checking for duplicate keys.  Returns zero on success,
294  * negative values on failure.
295  *
296  */
297 int cipso_v4_cache_add(const unsigned char *cipso_ptr,
298 		       const struct netlbl_lsm_secattr *secattr)
299 {
300 	int ret_val = -EPERM;
301 	u32 bkt;
302 	struct cipso_v4_map_cache_entry *entry = NULL;
303 	struct cipso_v4_map_cache_entry *old_entry = NULL;
304 	u32 cipso_ptr_len;
305 
306 	if (!cipso_v4_cache_enabled || cipso_v4_cache_bucketsize <= 0)
307 		return 0;
308 
309 	cipso_ptr_len = cipso_ptr[1];
310 
311 	entry = kzalloc(sizeof(*entry), GFP_ATOMIC);
312 	if (!entry)
313 		return -ENOMEM;
314 	entry->key = kmemdup(cipso_ptr, cipso_ptr_len, GFP_ATOMIC);
315 	if (!entry->key) {
316 		ret_val = -ENOMEM;
317 		goto cache_add_failure;
318 	}
319 	entry->key_len = cipso_ptr_len;
320 	entry->hash = cipso_v4_map_cache_hash(cipso_ptr, cipso_ptr_len);
321 	refcount_inc(&secattr->cache->refcount);
322 	entry->lsm_data = secattr->cache;
323 
324 	bkt = entry->hash & (CIPSO_V4_CACHE_BUCKETS - 1);
325 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
326 	if (cipso_v4_cache[bkt].size < cipso_v4_cache_bucketsize) {
327 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
328 		cipso_v4_cache[bkt].size += 1;
329 	} else {
330 		old_entry = list_entry(cipso_v4_cache[bkt].list.prev,
331 				       struct cipso_v4_map_cache_entry, list);
332 		list_del(&old_entry->list);
333 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
334 		cipso_v4_cache_entry_free(old_entry);
335 	}
336 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
337 
338 	return 0;
339 
340 cache_add_failure:
341 	if (entry)
342 		cipso_v4_cache_entry_free(entry);
343 	return ret_val;
344 }
345 
346 /*
347  * DOI List Functions
348  */
349 
350 /**
351  * cipso_v4_doi_search - Searches for a DOI definition
352  * @doi: the DOI to search for
353  *
354  * Description:
355  * Search the DOI definition list for a DOI definition with a DOI value that
356  * matches @doi.  The caller is responsible for calling rcu_read_[un]lock().
357  * Returns a pointer to the DOI definition on success and NULL on failure.
358  */
359 static struct cipso_v4_doi *cipso_v4_doi_search(u32 doi)
360 {
361 	struct cipso_v4_doi *iter;
362 
363 	list_for_each_entry_rcu(iter, &cipso_v4_doi_list, list)
364 		if (iter->doi == doi && refcount_read(&iter->refcount))
365 			return iter;
366 	return NULL;
367 }
368 
369 /**
370  * cipso_v4_doi_add - Add a new DOI to the CIPSO protocol engine
371  * @doi_def: the DOI structure
372  * @audit_info: NetLabel audit information
373  *
374  * Description:
375  * The caller defines a new DOI for use by the CIPSO engine and calls this
376  * function to add it to the list of acceptable domains.  The caller must
377  * ensure that the mapping table specified in @doi_def->map meets all of the
378  * requirements of the mapping type (see cipso_ipv4.h for details).  Returns
379  * zero on success and non-zero on failure.
380  *
381  */
382 int cipso_v4_doi_add(struct cipso_v4_doi *doi_def,
383 		     struct netlbl_audit *audit_info)
384 {
385 	int ret_val = -EINVAL;
386 	u32 iter;
387 	u32 doi;
388 	u32 doi_type;
389 	struct audit_buffer *audit_buf;
390 
391 	doi = doi_def->doi;
392 	doi_type = doi_def->type;
393 
394 	if (doi_def->doi == CIPSO_V4_DOI_UNKNOWN)
395 		goto doi_add_return;
396 	for (iter = 0; iter < CIPSO_V4_TAG_MAXCNT; iter++) {
397 		switch (doi_def->tags[iter]) {
398 		case CIPSO_V4_TAG_RBITMAP:
399 			break;
400 		case CIPSO_V4_TAG_RANGE:
401 		case CIPSO_V4_TAG_ENUM:
402 			if (doi_def->type != CIPSO_V4_MAP_PASS)
403 				goto doi_add_return;
404 			break;
405 		case CIPSO_V4_TAG_LOCAL:
406 			if (doi_def->type != CIPSO_V4_MAP_LOCAL)
407 				goto doi_add_return;
408 			break;
409 		case CIPSO_V4_TAG_INVALID:
410 			if (iter == 0)
411 				goto doi_add_return;
412 			break;
413 		default:
414 			goto doi_add_return;
415 		}
416 	}
417 
418 	refcount_set(&doi_def->refcount, 1);
419 
420 	spin_lock(&cipso_v4_doi_list_lock);
421 	if (cipso_v4_doi_search(doi_def->doi)) {
422 		spin_unlock(&cipso_v4_doi_list_lock);
423 		ret_val = -EEXIST;
424 		goto doi_add_return;
425 	}
426 	list_add_tail_rcu(&doi_def->list, &cipso_v4_doi_list);
427 	spin_unlock(&cipso_v4_doi_list_lock);
428 	ret_val = 0;
429 
430 doi_add_return:
431 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_ADD, audit_info);
432 	if (audit_buf) {
433 		const char *type_str;
434 		switch (doi_type) {
435 		case CIPSO_V4_MAP_TRANS:
436 			type_str = "trans";
437 			break;
438 		case CIPSO_V4_MAP_PASS:
439 			type_str = "pass";
440 			break;
441 		case CIPSO_V4_MAP_LOCAL:
442 			type_str = "local";
443 			break;
444 		default:
445 			type_str = "(unknown)";
446 		}
447 		audit_log_format(audit_buf,
448 				 " cipso_doi=%u cipso_type=%s res=%u",
449 				 doi, type_str, ret_val == 0 ? 1 : 0);
450 		audit_log_end(audit_buf);
451 	}
452 
453 	return ret_val;
454 }
455 
456 /**
457  * cipso_v4_doi_free - Frees a DOI definition
458  * @doi_def: the DOI definition
459  *
460  * Description:
461  * This function frees all of the memory associated with a DOI definition.
462  *
463  */
464 void cipso_v4_doi_free(struct cipso_v4_doi *doi_def)
465 {
466 	if (!doi_def)
467 		return;
468 
469 	switch (doi_def->type) {
470 	case CIPSO_V4_MAP_TRANS:
471 		kfree(doi_def->map.std->lvl.cipso);
472 		kfree(doi_def->map.std->lvl.local);
473 		kfree(doi_def->map.std->cat.cipso);
474 		kfree(doi_def->map.std->cat.local);
475 		break;
476 	}
477 	kfree(doi_def);
478 }
479 
480 /**
481  * cipso_v4_doi_free_rcu - Frees a DOI definition via the RCU pointer
482  * @entry: the entry's RCU field
483  *
484  * Description:
485  * This function is designed to be used as a callback to the call_rcu()
486  * function so that the memory allocated to the DOI definition can be released
487  * safely.
488  *
489  */
490 static void cipso_v4_doi_free_rcu(struct rcu_head *entry)
491 {
492 	struct cipso_v4_doi *doi_def;
493 
494 	doi_def = container_of(entry, struct cipso_v4_doi, rcu);
495 	cipso_v4_doi_free(doi_def);
496 }
497 
498 /**
499  * cipso_v4_doi_remove - Remove an existing DOI from the CIPSO protocol engine
500  * @doi: the DOI value
501  * @audit_secid: the LSM secid to use in the audit message
502  *
503  * Description:
504  * Removes a DOI definition from the CIPSO engine.  The NetLabel routines will
505  * be called to release their own LSM domain mappings as well as our own
506  * domain list.  Returns zero on success and negative values on failure.
507  *
508  */
509 int cipso_v4_doi_remove(u32 doi, struct netlbl_audit *audit_info)
510 {
511 	int ret_val;
512 	struct cipso_v4_doi *doi_def;
513 	struct audit_buffer *audit_buf;
514 
515 	spin_lock(&cipso_v4_doi_list_lock);
516 	doi_def = cipso_v4_doi_search(doi);
517 	if (!doi_def) {
518 		spin_unlock(&cipso_v4_doi_list_lock);
519 		ret_val = -ENOENT;
520 		goto doi_remove_return;
521 	}
522 	if (!refcount_dec_and_test(&doi_def->refcount)) {
523 		spin_unlock(&cipso_v4_doi_list_lock);
524 		ret_val = -EBUSY;
525 		goto doi_remove_return;
526 	}
527 	list_del_rcu(&doi_def->list);
528 	spin_unlock(&cipso_v4_doi_list_lock);
529 
530 	cipso_v4_cache_invalidate();
531 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
532 	ret_val = 0;
533 
534 doi_remove_return:
535 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_DEL, audit_info);
536 	if (audit_buf) {
537 		audit_log_format(audit_buf,
538 				 " cipso_doi=%u res=%u",
539 				 doi, ret_val == 0 ? 1 : 0);
540 		audit_log_end(audit_buf);
541 	}
542 
543 	return ret_val;
544 }
545 
546 /**
547  * cipso_v4_doi_getdef - Returns a reference to a valid DOI definition
548  * @doi: the DOI value
549  *
550  * Description:
551  * Searches for a valid DOI definition and if one is found it is returned to
552  * the caller.  Otherwise NULL is returned.  The caller must ensure that
553  * rcu_read_lock() is held while accessing the returned definition and the DOI
554  * definition reference count is decremented when the caller is done.
555  *
556  */
557 struct cipso_v4_doi *cipso_v4_doi_getdef(u32 doi)
558 {
559 	struct cipso_v4_doi *doi_def;
560 
561 	rcu_read_lock();
562 	doi_def = cipso_v4_doi_search(doi);
563 	if (!doi_def)
564 		goto doi_getdef_return;
565 	if (!refcount_inc_not_zero(&doi_def->refcount))
566 		doi_def = NULL;
567 
568 doi_getdef_return:
569 	rcu_read_unlock();
570 	return doi_def;
571 }
572 
573 /**
574  * cipso_v4_doi_putdef - Releases a reference for the given DOI definition
575  * @doi_def: the DOI definition
576  *
577  * Description:
578  * Releases a DOI definition reference obtained from cipso_v4_doi_getdef().
579  *
580  */
581 void cipso_v4_doi_putdef(struct cipso_v4_doi *doi_def)
582 {
583 	if (!doi_def)
584 		return;
585 
586 	if (!refcount_dec_and_test(&doi_def->refcount))
587 		return;
588 	spin_lock(&cipso_v4_doi_list_lock);
589 	list_del_rcu(&doi_def->list);
590 	spin_unlock(&cipso_v4_doi_list_lock);
591 
592 	cipso_v4_cache_invalidate();
593 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
594 }
595 
596 /**
597  * cipso_v4_doi_walk - Iterate through the DOI definitions
598  * @skip_cnt: skip past this number of DOI definitions, updated
599  * @callback: callback for each DOI definition
600  * @cb_arg: argument for the callback function
601  *
602  * Description:
603  * Iterate over the DOI definition list, skipping the first @skip_cnt entries.
604  * For each entry call @callback, if @callback returns a negative value stop
605  * 'walking' through the list and return.  Updates the value in @skip_cnt upon
606  * return.  Returns zero on success, negative values on failure.
607  *
608  */
609 int cipso_v4_doi_walk(u32 *skip_cnt,
610 		     int (*callback) (struct cipso_v4_doi *doi_def, void *arg),
611 		     void *cb_arg)
612 {
613 	int ret_val = -ENOENT;
614 	u32 doi_cnt = 0;
615 	struct cipso_v4_doi *iter_doi;
616 
617 	rcu_read_lock();
618 	list_for_each_entry_rcu(iter_doi, &cipso_v4_doi_list, list)
619 		if (refcount_read(&iter_doi->refcount) > 0) {
620 			if (doi_cnt++ < *skip_cnt)
621 				continue;
622 			ret_val = callback(iter_doi, cb_arg);
623 			if (ret_val < 0) {
624 				doi_cnt--;
625 				goto doi_walk_return;
626 			}
627 		}
628 
629 doi_walk_return:
630 	rcu_read_unlock();
631 	*skip_cnt = doi_cnt;
632 	return ret_val;
633 }
634 
635 /*
636  * Label Mapping Functions
637  */
638 
639 /**
640  * cipso_v4_map_lvl_valid - Checks to see if the given level is understood
641  * @doi_def: the DOI definition
642  * @level: the level to check
643  *
644  * Description:
645  * Checks the given level against the given DOI definition and returns a
646  * negative value if the level does not have a valid mapping and a zero value
647  * if the level is defined by the DOI.
648  *
649  */
650 static int cipso_v4_map_lvl_valid(const struct cipso_v4_doi *doi_def, u8 level)
651 {
652 	switch (doi_def->type) {
653 	case CIPSO_V4_MAP_PASS:
654 		return 0;
655 	case CIPSO_V4_MAP_TRANS:
656 		if ((level < doi_def->map.std->lvl.cipso_size) &&
657 		    (doi_def->map.std->lvl.cipso[level] < CIPSO_V4_INV_LVL))
658 			return 0;
659 		break;
660 	}
661 
662 	return -EFAULT;
663 }
664 
665 /**
666  * cipso_v4_map_lvl_hton - Perform a level mapping from the host to the network
667  * @doi_def: the DOI definition
668  * @host_lvl: the host MLS level
669  * @net_lvl: the network/CIPSO MLS level
670  *
671  * Description:
672  * Perform a label mapping to translate a local MLS level to the correct
673  * CIPSO level using the given DOI definition.  Returns zero on success,
674  * negative values otherwise.
675  *
676  */
677 static int cipso_v4_map_lvl_hton(const struct cipso_v4_doi *doi_def,
678 				 u32 host_lvl,
679 				 u32 *net_lvl)
680 {
681 	switch (doi_def->type) {
682 	case CIPSO_V4_MAP_PASS:
683 		*net_lvl = host_lvl;
684 		return 0;
685 	case CIPSO_V4_MAP_TRANS:
686 		if (host_lvl < doi_def->map.std->lvl.local_size &&
687 		    doi_def->map.std->lvl.local[host_lvl] < CIPSO_V4_INV_LVL) {
688 			*net_lvl = doi_def->map.std->lvl.local[host_lvl];
689 			return 0;
690 		}
691 		return -EPERM;
692 	}
693 
694 	return -EINVAL;
695 }
696 
697 /**
698  * cipso_v4_map_lvl_ntoh - Perform a level mapping from the network to the host
699  * @doi_def: the DOI definition
700  * @net_lvl: the network/CIPSO MLS level
701  * @host_lvl: the host MLS level
702  *
703  * Description:
704  * Perform a label mapping to translate a CIPSO level to the correct local MLS
705  * level using the given DOI definition.  Returns zero on success, negative
706  * values otherwise.
707  *
708  */
709 static int cipso_v4_map_lvl_ntoh(const struct cipso_v4_doi *doi_def,
710 				 u32 net_lvl,
711 				 u32 *host_lvl)
712 {
713 	struct cipso_v4_std_map_tbl *map_tbl;
714 
715 	switch (doi_def->type) {
716 	case CIPSO_V4_MAP_PASS:
717 		*host_lvl = net_lvl;
718 		return 0;
719 	case CIPSO_V4_MAP_TRANS:
720 		map_tbl = doi_def->map.std;
721 		if (net_lvl < map_tbl->lvl.cipso_size &&
722 		    map_tbl->lvl.cipso[net_lvl] < CIPSO_V4_INV_LVL) {
723 			*host_lvl = doi_def->map.std->lvl.cipso[net_lvl];
724 			return 0;
725 		}
726 		return -EPERM;
727 	}
728 
729 	return -EINVAL;
730 }
731 
732 /**
733  * cipso_v4_map_cat_rbm_valid - Checks to see if the category bitmap is valid
734  * @doi_def: the DOI definition
735  * @bitmap: category bitmap
736  * @bitmap_len: bitmap length in bytes
737  *
738  * Description:
739  * Checks the given category bitmap against the given DOI definition and
740  * returns a negative value if any of the categories in the bitmap do not have
741  * a valid mapping and a zero value if all of the categories are valid.
742  *
743  */
744 static int cipso_v4_map_cat_rbm_valid(const struct cipso_v4_doi *doi_def,
745 				      const unsigned char *bitmap,
746 				      u32 bitmap_len)
747 {
748 	int cat = -1;
749 	u32 bitmap_len_bits = bitmap_len * 8;
750 	u32 cipso_cat_size;
751 	u32 *cipso_array;
752 
753 	switch (doi_def->type) {
754 	case CIPSO_V4_MAP_PASS:
755 		return 0;
756 	case CIPSO_V4_MAP_TRANS:
757 		cipso_cat_size = doi_def->map.std->cat.cipso_size;
758 		cipso_array = doi_def->map.std->cat.cipso;
759 		for (;;) {
760 			cat = netlbl_bitmap_walk(bitmap,
761 						 bitmap_len_bits,
762 						 cat + 1,
763 						 1);
764 			if (cat < 0)
765 				break;
766 			if (cat >= cipso_cat_size ||
767 			    cipso_array[cat] >= CIPSO_V4_INV_CAT)
768 				return -EFAULT;
769 		}
770 
771 		if (cat == -1)
772 			return 0;
773 		break;
774 	}
775 
776 	return -EFAULT;
777 }
778 
779 /**
780  * cipso_v4_map_cat_rbm_hton - Perform a category mapping from host to network
781  * @doi_def: the DOI definition
782  * @secattr: the security attributes
783  * @net_cat: the zero'd out category bitmap in network/CIPSO format
784  * @net_cat_len: the length of the CIPSO bitmap in bytes
785  *
786  * Description:
787  * Perform a label mapping to translate a local MLS category bitmap to the
788  * correct CIPSO bitmap using the given DOI definition.  Returns the minimum
789  * size in bytes of the network bitmap on success, negative values otherwise.
790  *
791  */
792 static int cipso_v4_map_cat_rbm_hton(const struct cipso_v4_doi *doi_def,
793 				     const struct netlbl_lsm_secattr *secattr,
794 				     unsigned char *net_cat,
795 				     u32 net_cat_len)
796 {
797 	int host_spot = -1;
798 	u32 net_spot = CIPSO_V4_INV_CAT;
799 	u32 net_spot_max = 0;
800 	u32 net_clen_bits = net_cat_len * 8;
801 	u32 host_cat_size = 0;
802 	u32 *host_cat_array = NULL;
803 
804 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
805 		host_cat_size = doi_def->map.std->cat.local_size;
806 		host_cat_array = doi_def->map.std->cat.local;
807 	}
808 
809 	for (;;) {
810 		host_spot = netlbl_catmap_walk(secattr->attr.mls.cat,
811 					       host_spot + 1);
812 		if (host_spot < 0)
813 			break;
814 
815 		switch (doi_def->type) {
816 		case CIPSO_V4_MAP_PASS:
817 			net_spot = host_spot;
818 			break;
819 		case CIPSO_V4_MAP_TRANS:
820 			if (host_spot >= host_cat_size)
821 				return -EPERM;
822 			net_spot = host_cat_array[host_spot];
823 			if (net_spot >= CIPSO_V4_INV_CAT)
824 				return -EPERM;
825 			break;
826 		}
827 		if (net_spot >= net_clen_bits)
828 			return -ENOSPC;
829 		netlbl_bitmap_setbit(net_cat, net_spot, 1);
830 
831 		if (net_spot > net_spot_max)
832 			net_spot_max = net_spot;
833 	}
834 
835 	if (++net_spot_max % 8)
836 		return net_spot_max / 8 + 1;
837 	return net_spot_max / 8;
838 }
839 
840 /**
841  * cipso_v4_map_cat_rbm_ntoh - Perform a category mapping from network to host
842  * @doi_def: the DOI definition
843  * @net_cat: the category bitmap in network/CIPSO format
844  * @net_cat_len: the length of the CIPSO bitmap in bytes
845  * @secattr: the security attributes
846  *
847  * Description:
848  * Perform a label mapping to translate a CIPSO bitmap to the correct local
849  * MLS category bitmap using the given DOI definition.  Returns zero on
850  * success, negative values on failure.
851  *
852  */
853 static int cipso_v4_map_cat_rbm_ntoh(const struct cipso_v4_doi *doi_def,
854 				     const unsigned char *net_cat,
855 				     u32 net_cat_len,
856 				     struct netlbl_lsm_secattr *secattr)
857 {
858 	int ret_val;
859 	int net_spot = -1;
860 	u32 host_spot = CIPSO_V4_INV_CAT;
861 	u32 net_clen_bits = net_cat_len * 8;
862 	u32 net_cat_size = 0;
863 	u32 *net_cat_array = NULL;
864 
865 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
866 		net_cat_size = doi_def->map.std->cat.cipso_size;
867 		net_cat_array = doi_def->map.std->cat.cipso;
868 	}
869 
870 	for (;;) {
871 		net_spot = netlbl_bitmap_walk(net_cat,
872 					      net_clen_bits,
873 					      net_spot + 1,
874 					      1);
875 		if (net_spot < 0) {
876 			if (net_spot == -2)
877 				return -EFAULT;
878 			return 0;
879 		}
880 
881 		switch (doi_def->type) {
882 		case CIPSO_V4_MAP_PASS:
883 			host_spot = net_spot;
884 			break;
885 		case CIPSO_V4_MAP_TRANS:
886 			if (net_spot >= net_cat_size)
887 				return -EPERM;
888 			host_spot = net_cat_array[net_spot];
889 			if (host_spot >= CIPSO_V4_INV_CAT)
890 				return -EPERM;
891 			break;
892 		}
893 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
894 						       host_spot,
895 						       GFP_ATOMIC);
896 		if (ret_val != 0)
897 			return ret_val;
898 	}
899 
900 	return -EINVAL;
901 }
902 
903 /**
904  * cipso_v4_map_cat_enum_valid - Checks to see if the categories are valid
905  * @doi_def: the DOI definition
906  * @enumcat: category list
907  * @enumcat_len: length of the category list in bytes
908  *
909  * Description:
910  * Checks the given categories against the given DOI definition and returns a
911  * negative value if any of the categories do not have a valid mapping and a
912  * zero value if all of the categories are valid.
913  *
914  */
915 static int cipso_v4_map_cat_enum_valid(const struct cipso_v4_doi *doi_def,
916 				       const unsigned char *enumcat,
917 				       u32 enumcat_len)
918 {
919 	u16 cat;
920 	int cat_prev = -1;
921 	u32 iter;
922 
923 	if (doi_def->type != CIPSO_V4_MAP_PASS || enumcat_len & 0x01)
924 		return -EFAULT;
925 
926 	for (iter = 0; iter < enumcat_len; iter += 2) {
927 		cat = get_unaligned_be16(&enumcat[iter]);
928 		if (cat <= cat_prev)
929 			return -EFAULT;
930 		cat_prev = cat;
931 	}
932 
933 	return 0;
934 }
935 
936 /**
937  * cipso_v4_map_cat_enum_hton - Perform a category mapping from host to network
938  * @doi_def: the DOI definition
939  * @secattr: the security attributes
940  * @net_cat: the zero'd out category list in network/CIPSO format
941  * @net_cat_len: the length of the CIPSO category list in bytes
942  *
943  * Description:
944  * Perform a label mapping to translate a local MLS category bitmap to the
945  * correct CIPSO category list using the given DOI definition.   Returns the
946  * size in bytes of the network category bitmap on success, negative values
947  * otherwise.
948  *
949  */
950 static int cipso_v4_map_cat_enum_hton(const struct cipso_v4_doi *doi_def,
951 				      const struct netlbl_lsm_secattr *secattr,
952 				      unsigned char *net_cat,
953 				      u32 net_cat_len)
954 {
955 	int cat = -1;
956 	u32 cat_iter = 0;
957 
958 	for (;;) {
959 		cat = netlbl_catmap_walk(secattr->attr.mls.cat, cat + 1);
960 		if (cat < 0)
961 			break;
962 		if ((cat_iter + 2) > net_cat_len)
963 			return -ENOSPC;
964 
965 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
966 		cat_iter += 2;
967 	}
968 
969 	return cat_iter;
970 }
971 
972 /**
973  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
974  * @doi_def: the DOI definition
975  * @net_cat: the category list in network/CIPSO format
976  * @net_cat_len: the length of the CIPSO bitmap in bytes
977  * @secattr: the security attributes
978  *
979  * Description:
980  * Perform a label mapping to translate a CIPSO category list to the correct
981  * local MLS category bitmap using the given DOI definition.  Returns zero on
982  * success, negative values on failure.
983  *
984  */
985 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
986 				      const unsigned char *net_cat,
987 				      u32 net_cat_len,
988 				      struct netlbl_lsm_secattr *secattr)
989 {
990 	int ret_val;
991 	u32 iter;
992 
993 	for (iter = 0; iter < net_cat_len; iter += 2) {
994 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
995 					     get_unaligned_be16(&net_cat[iter]),
996 					     GFP_ATOMIC);
997 		if (ret_val != 0)
998 			return ret_val;
999 	}
1000 
1001 	return 0;
1002 }
1003 
1004 /**
1005  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
1006  * @doi_def: the DOI definition
1007  * @rngcat: category list
1008  * @rngcat_len: length of the category list in bytes
1009  *
1010  * Description:
1011  * Checks the given categories against the given DOI definition and returns a
1012  * negative value if any of the categories do not have a valid mapping and a
1013  * zero value if all of the categories are valid.
1014  *
1015  */
1016 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1017 				      const unsigned char *rngcat,
1018 				      u32 rngcat_len)
1019 {
1020 	u16 cat_high;
1021 	u16 cat_low;
1022 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1023 	u32 iter;
1024 
1025 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1026 		return -EFAULT;
1027 
1028 	for (iter = 0; iter < rngcat_len; iter += 4) {
1029 		cat_high = get_unaligned_be16(&rngcat[iter]);
1030 		if ((iter + 4) <= rngcat_len)
1031 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1032 		else
1033 			cat_low = 0;
1034 
1035 		if (cat_high > cat_prev)
1036 			return -EFAULT;
1037 
1038 		cat_prev = cat_low;
1039 	}
1040 
1041 	return 0;
1042 }
1043 
1044 /**
1045  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1046  * @doi_def: the DOI definition
1047  * @secattr: the security attributes
1048  * @net_cat: the zero'd out category list in network/CIPSO format
1049  * @net_cat_len: the length of the CIPSO category list in bytes
1050  *
1051  * Description:
1052  * Perform a label mapping to translate a local MLS category bitmap to the
1053  * correct CIPSO category list using the given DOI definition.   Returns the
1054  * size in bytes of the network category bitmap on success, negative values
1055  * otherwise.
1056  *
1057  */
1058 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1059 				     const struct netlbl_lsm_secattr *secattr,
1060 				     unsigned char *net_cat,
1061 				     u32 net_cat_len)
1062 {
1063 	int iter = -1;
1064 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1065 	u32 array_cnt = 0;
1066 	u32 cat_size = 0;
1067 
1068 	/* make sure we don't overflow the 'array[]' variable */
1069 	if (net_cat_len >
1070 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1071 		return -ENOSPC;
1072 
1073 	for (;;) {
1074 		iter = netlbl_catmap_walk(secattr->attr.mls.cat, iter + 1);
1075 		if (iter < 0)
1076 			break;
1077 		cat_size += (iter == 0 ? 0 : sizeof(u16));
1078 		if (cat_size > net_cat_len)
1079 			return -ENOSPC;
1080 		array[array_cnt++] = iter;
1081 
1082 		iter = netlbl_catmap_walkrng(secattr->attr.mls.cat, iter);
1083 		if (iter < 0)
1084 			return -EFAULT;
1085 		cat_size += sizeof(u16);
1086 		if (cat_size > net_cat_len)
1087 			return -ENOSPC;
1088 		array[array_cnt++] = iter;
1089 	}
1090 
1091 	for (iter = 0; array_cnt > 0;) {
1092 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1093 		iter += 2;
1094 		array_cnt--;
1095 		if (array[array_cnt] != 0) {
1096 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1097 			iter += 2;
1098 		}
1099 	}
1100 
1101 	return cat_size;
1102 }
1103 
1104 /**
1105  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1106  * @doi_def: the DOI definition
1107  * @net_cat: the category list in network/CIPSO format
1108  * @net_cat_len: the length of the CIPSO bitmap in bytes
1109  * @secattr: the security attributes
1110  *
1111  * Description:
1112  * Perform a label mapping to translate a CIPSO category list to the correct
1113  * local MLS category bitmap using the given DOI definition.  Returns zero on
1114  * success, negative values on failure.
1115  *
1116  */
1117 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1118 				     const unsigned char *net_cat,
1119 				     u32 net_cat_len,
1120 				     struct netlbl_lsm_secattr *secattr)
1121 {
1122 	int ret_val;
1123 	u32 net_iter;
1124 	u16 cat_low;
1125 	u16 cat_high;
1126 
1127 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1128 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1129 		if ((net_iter + 4) <= net_cat_len)
1130 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1131 		else
1132 			cat_low = 0;
1133 
1134 		ret_val = netlbl_catmap_setrng(&secattr->attr.mls.cat,
1135 					       cat_low,
1136 					       cat_high,
1137 					       GFP_ATOMIC);
1138 		if (ret_val != 0)
1139 			return ret_val;
1140 	}
1141 
1142 	return 0;
1143 }
1144 
1145 /*
1146  * Protocol Handling Functions
1147  */
1148 
1149 /**
1150  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1151  * @doi_def: the DOI definition
1152  * @len: the total tag length in bytes, not including this header
1153  * @buf: the CIPSO option buffer
1154  *
1155  * Description:
1156  * Write a CIPSO header into the beginning of @buffer.
1157  *
1158  */
1159 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1160 				unsigned char *buf,
1161 				u32 len)
1162 {
1163 	buf[0] = IPOPT_CIPSO;
1164 	buf[1] = CIPSO_V4_HDR_LEN + len;
1165 	*(__be32 *)&buf[2] = htonl(doi_def->doi);
1166 }
1167 
1168 /**
1169  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1170  * @doi_def: the DOI definition
1171  * @secattr: the security attributes
1172  * @buffer: the option buffer
1173  * @buffer_len: length of buffer in bytes
1174  *
1175  * Description:
1176  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1177  * actual buffer length may be larger than the indicated size due to
1178  * translation between host and network category bitmaps.  Returns the size of
1179  * the tag on success, negative values on failure.
1180  *
1181  */
1182 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1183 			       const struct netlbl_lsm_secattr *secattr,
1184 			       unsigned char *buffer,
1185 			       u32 buffer_len)
1186 {
1187 	int ret_val;
1188 	u32 tag_len;
1189 	u32 level;
1190 
1191 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1192 		return -EPERM;
1193 
1194 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1195 					secattr->attr.mls.lvl,
1196 					&level);
1197 	if (ret_val != 0)
1198 		return ret_val;
1199 
1200 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1201 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1202 						    secattr,
1203 						    &buffer[4],
1204 						    buffer_len - 4);
1205 		if (ret_val < 0)
1206 			return ret_val;
1207 
1208 		/* This will send packets using the "optimized" format when
1209 		 * possible as specified in  section 3.4.2.6 of the
1210 		 * CIPSO draft. */
1211 		if (cipso_v4_rbm_optfmt && ret_val > 0 && ret_val <= 10)
1212 			tag_len = 14;
1213 		else
1214 			tag_len = 4 + ret_val;
1215 	} else
1216 		tag_len = 4;
1217 
1218 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1219 	buffer[1] = tag_len;
1220 	buffer[3] = level;
1221 
1222 	return tag_len;
1223 }
1224 
1225 /**
1226  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1227  * @doi_def: the DOI definition
1228  * @tag: the CIPSO tag
1229  * @secattr: the security attributes
1230  *
1231  * Description:
1232  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1233  * attributes in @secattr.  Return zero on success, negatives values on
1234  * failure.
1235  *
1236  */
1237 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1238 				 const unsigned char *tag,
1239 				 struct netlbl_lsm_secattr *secattr)
1240 {
1241 	int ret_val;
1242 	u8 tag_len = tag[1];
1243 	u32 level;
1244 
1245 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1246 	if (ret_val != 0)
1247 		return ret_val;
1248 	secattr->attr.mls.lvl = level;
1249 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1250 
1251 	if (tag_len > 4) {
1252 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1253 						    &tag[4],
1254 						    tag_len - 4,
1255 						    secattr);
1256 		if (ret_val != 0) {
1257 			netlbl_catmap_free(secattr->attr.mls.cat);
1258 			return ret_val;
1259 		}
1260 
1261 		if (secattr->attr.mls.cat)
1262 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1263 	}
1264 
1265 	return 0;
1266 }
1267 
1268 /**
1269  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1270  * @doi_def: the DOI definition
1271  * @secattr: the security attributes
1272  * @buffer: the option buffer
1273  * @buffer_len: length of buffer in bytes
1274  *
1275  * Description:
1276  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1277  * size of the tag on success, negative values on failure.
1278  *
1279  */
1280 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1281 				const struct netlbl_lsm_secattr *secattr,
1282 				unsigned char *buffer,
1283 				u32 buffer_len)
1284 {
1285 	int ret_val;
1286 	u32 tag_len;
1287 	u32 level;
1288 
1289 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1290 		return -EPERM;
1291 
1292 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1293 					secattr->attr.mls.lvl,
1294 					&level);
1295 	if (ret_val != 0)
1296 		return ret_val;
1297 
1298 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1299 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1300 						     secattr,
1301 						     &buffer[4],
1302 						     buffer_len - 4);
1303 		if (ret_val < 0)
1304 			return ret_val;
1305 
1306 		tag_len = 4 + ret_val;
1307 	} else
1308 		tag_len = 4;
1309 
1310 	buffer[0] = CIPSO_V4_TAG_ENUM;
1311 	buffer[1] = tag_len;
1312 	buffer[3] = level;
1313 
1314 	return tag_len;
1315 }
1316 
1317 /**
1318  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1319  * @doi_def: the DOI definition
1320  * @tag: the CIPSO tag
1321  * @secattr: the security attributes
1322  *
1323  * Description:
1324  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1325  * attributes in @secattr.  Return zero on success, negatives values on
1326  * failure.
1327  *
1328  */
1329 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1330 				  const unsigned char *tag,
1331 				  struct netlbl_lsm_secattr *secattr)
1332 {
1333 	int ret_val;
1334 	u8 tag_len = tag[1];
1335 	u32 level;
1336 
1337 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1338 	if (ret_val != 0)
1339 		return ret_val;
1340 	secattr->attr.mls.lvl = level;
1341 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1342 
1343 	if (tag_len > 4) {
1344 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1345 						     &tag[4],
1346 						     tag_len - 4,
1347 						     secattr);
1348 		if (ret_val != 0) {
1349 			netlbl_catmap_free(secattr->attr.mls.cat);
1350 			return ret_val;
1351 		}
1352 
1353 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1354 	}
1355 
1356 	return 0;
1357 }
1358 
1359 /**
1360  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1361  * @doi_def: the DOI definition
1362  * @secattr: the security attributes
1363  * @buffer: the option buffer
1364  * @buffer_len: length of buffer in bytes
1365  *
1366  * Description:
1367  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1368  * size of the tag on success, negative values on failure.
1369  *
1370  */
1371 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1372 			       const struct netlbl_lsm_secattr *secattr,
1373 			       unsigned char *buffer,
1374 			       u32 buffer_len)
1375 {
1376 	int ret_val;
1377 	u32 tag_len;
1378 	u32 level;
1379 
1380 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1381 		return -EPERM;
1382 
1383 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1384 					secattr->attr.mls.lvl,
1385 					&level);
1386 	if (ret_val != 0)
1387 		return ret_val;
1388 
1389 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1390 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1391 						    secattr,
1392 						    &buffer[4],
1393 						    buffer_len - 4);
1394 		if (ret_val < 0)
1395 			return ret_val;
1396 
1397 		tag_len = 4 + ret_val;
1398 	} else
1399 		tag_len = 4;
1400 
1401 	buffer[0] = CIPSO_V4_TAG_RANGE;
1402 	buffer[1] = tag_len;
1403 	buffer[3] = level;
1404 
1405 	return tag_len;
1406 }
1407 
1408 /**
1409  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1410  * @doi_def: the DOI definition
1411  * @tag: the CIPSO tag
1412  * @secattr: the security attributes
1413  *
1414  * Description:
1415  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1416  * in @secattr.  Return zero on success, negatives values on failure.
1417  *
1418  */
1419 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1420 				 const unsigned char *tag,
1421 				 struct netlbl_lsm_secattr *secattr)
1422 {
1423 	int ret_val;
1424 	u8 tag_len = tag[1];
1425 	u32 level;
1426 
1427 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1428 	if (ret_val != 0)
1429 		return ret_val;
1430 	secattr->attr.mls.lvl = level;
1431 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1432 
1433 	if (tag_len > 4) {
1434 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1435 						    &tag[4],
1436 						    tag_len - 4,
1437 						    secattr);
1438 		if (ret_val != 0) {
1439 			netlbl_catmap_free(secattr->attr.mls.cat);
1440 			return ret_val;
1441 		}
1442 
1443 		if (secattr->attr.mls.cat)
1444 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1445 	}
1446 
1447 	return 0;
1448 }
1449 
1450 /**
1451  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1452  * @doi_def: the DOI definition
1453  * @secattr: the security attributes
1454  * @buffer: the option buffer
1455  * @buffer_len: length of buffer in bytes
1456  *
1457  * Description:
1458  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1459  * on success, negative values on failure.
1460  *
1461  */
1462 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1463 			       const struct netlbl_lsm_secattr *secattr,
1464 			       unsigned char *buffer,
1465 			       u32 buffer_len)
1466 {
1467 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1468 		return -EPERM;
1469 
1470 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1471 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1472 	*(u32 *)&buffer[2] = secattr->attr.secid;
1473 
1474 	return CIPSO_V4_TAG_LOC_BLEN;
1475 }
1476 
1477 /**
1478  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1479  * @doi_def: the DOI definition
1480  * @tag: the CIPSO tag
1481  * @secattr: the security attributes
1482  *
1483  * Description:
1484  * Parse a CIPSO local tag and return the security attributes in @secattr.
1485  * Return zero on success, negatives values on failure.
1486  *
1487  */
1488 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1489 				 const unsigned char *tag,
1490 				 struct netlbl_lsm_secattr *secattr)
1491 {
1492 	secattr->attr.secid = *(u32 *)&tag[2];
1493 	secattr->flags |= NETLBL_SECATTR_SECID;
1494 
1495 	return 0;
1496 }
1497 
1498 /**
1499  * cipso_v4_optptr - Find the CIPSO option in the packet
1500  * @skb: the packet
1501  *
1502  * Description:
1503  * Parse the packet's IP header looking for a CIPSO option.  Returns a pointer
1504  * to the start of the CIPSO option on success, NULL if one is not found.
1505  *
1506  */
1507 unsigned char *cipso_v4_optptr(const struct sk_buff *skb)
1508 {
1509 	const struct iphdr *iph = ip_hdr(skb);
1510 	unsigned char *optptr = (unsigned char *)&(ip_hdr(skb)[1]);
1511 	int optlen;
1512 	int taglen;
1513 
1514 	for (optlen = iph->ihl*4 - sizeof(struct iphdr); optlen > 1; ) {
1515 		switch (optptr[0]) {
1516 		case IPOPT_END:
1517 			return NULL;
1518 		case IPOPT_NOOP:
1519 			taglen = 1;
1520 			break;
1521 		default:
1522 			taglen = optptr[1];
1523 		}
1524 		if (!taglen || taglen > optlen)
1525 			return NULL;
1526 		if (optptr[0] == IPOPT_CIPSO)
1527 			return optptr;
1528 
1529 		optlen -= taglen;
1530 		optptr += taglen;
1531 	}
1532 
1533 	return NULL;
1534 }
1535 
1536 /**
1537  * cipso_v4_validate - Validate a CIPSO option
1538  * @option: the start of the option, on error it is set to point to the error
1539  *
1540  * Description:
1541  * This routine is called to validate a CIPSO option, it checks all of the
1542  * fields to ensure that they are at least valid, see the draft snippet below
1543  * for details.  If the option is valid then a zero value is returned and
1544  * the value of @option is unchanged.  If the option is invalid then a
1545  * non-zero value is returned and @option is adjusted to point to the
1546  * offending portion of the option.  From the IETF draft ...
1547  *
1548  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1549  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1550  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1551  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1552  *   that is unrecognized."
1553  *
1554  */
1555 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1556 {
1557 	unsigned char *opt = *option;
1558 	unsigned char *tag;
1559 	unsigned char opt_iter;
1560 	unsigned char err_offset = 0;
1561 	u8 opt_len;
1562 	u8 tag_len;
1563 	struct cipso_v4_doi *doi_def = NULL;
1564 	u32 tag_iter;
1565 
1566 	/* caller already checks for length values that are too large */
1567 	opt_len = opt[1];
1568 	if (opt_len < 8) {
1569 		err_offset = 1;
1570 		goto validate_return;
1571 	}
1572 
1573 	rcu_read_lock();
1574 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1575 	if (!doi_def) {
1576 		err_offset = 2;
1577 		goto validate_return_locked;
1578 	}
1579 
1580 	opt_iter = CIPSO_V4_HDR_LEN;
1581 	tag = opt + opt_iter;
1582 	while (opt_iter < opt_len) {
1583 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1584 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1585 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1586 				err_offset = opt_iter;
1587 				goto validate_return_locked;
1588 			}
1589 
1590 		if (opt_iter + 1 == opt_len) {
1591 			err_offset = opt_iter;
1592 			goto validate_return_locked;
1593 		}
1594 		tag_len = tag[1];
1595 		if (tag_len > (opt_len - opt_iter)) {
1596 			err_offset = opt_iter + 1;
1597 			goto validate_return_locked;
1598 		}
1599 
1600 		switch (tag[0]) {
1601 		case CIPSO_V4_TAG_RBITMAP:
1602 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1603 				err_offset = opt_iter + 1;
1604 				goto validate_return_locked;
1605 			}
1606 
1607 			/* We are already going to do all the verification
1608 			 * necessary at the socket layer so from our point of
1609 			 * view it is safe to turn these checks off (and less
1610 			 * work), however, the CIPSO draft says we should do
1611 			 * all the CIPSO validations here but it doesn't
1612 			 * really specify _exactly_ what we need to validate
1613 			 * ... so, just make it a sysctl tunable. */
1614 			if (cipso_v4_rbm_strictvalid) {
1615 				if (cipso_v4_map_lvl_valid(doi_def,
1616 							   tag[3]) < 0) {
1617 					err_offset = opt_iter + 3;
1618 					goto validate_return_locked;
1619 				}
1620 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1621 				    cipso_v4_map_cat_rbm_valid(doi_def,
1622 							    &tag[4],
1623 							    tag_len - 4) < 0) {
1624 					err_offset = opt_iter + 4;
1625 					goto validate_return_locked;
1626 				}
1627 			}
1628 			break;
1629 		case CIPSO_V4_TAG_ENUM:
1630 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1631 				err_offset = opt_iter + 1;
1632 				goto validate_return_locked;
1633 			}
1634 
1635 			if (cipso_v4_map_lvl_valid(doi_def,
1636 						   tag[3]) < 0) {
1637 				err_offset = opt_iter + 3;
1638 				goto validate_return_locked;
1639 			}
1640 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1641 			    cipso_v4_map_cat_enum_valid(doi_def,
1642 							&tag[4],
1643 							tag_len - 4) < 0) {
1644 				err_offset = opt_iter + 4;
1645 				goto validate_return_locked;
1646 			}
1647 			break;
1648 		case CIPSO_V4_TAG_RANGE:
1649 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1650 				err_offset = opt_iter + 1;
1651 				goto validate_return_locked;
1652 			}
1653 
1654 			if (cipso_v4_map_lvl_valid(doi_def,
1655 						   tag[3]) < 0) {
1656 				err_offset = opt_iter + 3;
1657 				goto validate_return_locked;
1658 			}
1659 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1660 			    cipso_v4_map_cat_rng_valid(doi_def,
1661 						       &tag[4],
1662 						       tag_len - 4) < 0) {
1663 				err_offset = opt_iter + 4;
1664 				goto validate_return_locked;
1665 			}
1666 			break;
1667 		case CIPSO_V4_TAG_LOCAL:
1668 			/* This is a non-standard tag that we only allow for
1669 			 * local connections, so if the incoming interface is
1670 			 * not the loopback device drop the packet. Further,
1671 			 * there is no legitimate reason for setting this from
1672 			 * userspace so reject it if skb is NULL. */
1673 			if (!skb || !(skb->dev->flags & IFF_LOOPBACK)) {
1674 				err_offset = opt_iter;
1675 				goto validate_return_locked;
1676 			}
1677 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1678 				err_offset = opt_iter + 1;
1679 				goto validate_return_locked;
1680 			}
1681 			break;
1682 		default:
1683 			err_offset = opt_iter;
1684 			goto validate_return_locked;
1685 		}
1686 
1687 		tag += tag_len;
1688 		opt_iter += tag_len;
1689 	}
1690 
1691 validate_return_locked:
1692 	rcu_read_unlock();
1693 validate_return:
1694 	*option = opt + err_offset;
1695 	return err_offset;
1696 }
1697 
1698 /**
1699  * cipso_v4_error - Send the correct response for a bad packet
1700  * @skb: the packet
1701  * @error: the error code
1702  * @gateway: CIPSO gateway flag
1703  *
1704  * Description:
1705  * Based on the error code given in @error, send an ICMP error message back to
1706  * the originating host.  From the IETF draft ...
1707  *
1708  *  "If the contents of the CIPSO [option] are valid but the security label is
1709  *   outside of the configured host or port label range, the datagram is
1710  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1711  *   returned.  The code field of the ICMP is set to 'communication with
1712  *   destination network administratively prohibited' (code 9) or to
1713  *   'communication with destination host administratively prohibited'
1714  *   (code 10).  The value of the code is dependent on whether the originator
1715  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1716  *   recipient of the ICMP message MUST be able to handle either value.  The
1717  *   same procedure is performed if a CIPSO [option] can not be added to an
1718  *   IP packet because it is too large to fit in the IP options area."
1719  *
1720  *  "If the error is triggered by receipt of an ICMP message, the message is
1721  *   discarded and no response is permitted (consistent with general ICMP
1722  *   processing rules)."
1723  *
1724  */
1725 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1726 {
1727 	unsigned char optbuf[sizeof(struct ip_options) + 40];
1728 	struct ip_options *opt = (struct ip_options *)optbuf;
1729 	int res;
1730 
1731 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1732 		return;
1733 
1734 	/*
1735 	 * We might be called above the IP layer,
1736 	 * so we can not use icmp_send and IPCB here.
1737 	 */
1738 
1739 	memset(opt, 0, sizeof(struct ip_options));
1740 	opt->optlen = ip_hdr(skb)->ihl*4 - sizeof(struct iphdr);
1741 	rcu_read_lock();
1742 	res = __ip_options_compile(dev_net(skb->dev), opt, skb, NULL);
1743 	rcu_read_unlock();
1744 
1745 	if (res)
1746 		return;
1747 
1748 	if (gateway)
1749 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0, opt);
1750 	else
1751 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0, opt);
1752 }
1753 
1754 /**
1755  * cipso_v4_genopt - Generate a CIPSO option
1756  * @buf: the option buffer
1757  * @buf_len: the size of opt_buf
1758  * @doi_def: the CIPSO DOI to use
1759  * @secattr: the security attributes
1760  *
1761  * Description:
1762  * Generate a CIPSO option using the DOI definition and security attributes
1763  * passed to the function.  Returns the length of the option on success and
1764  * negative values on failure.
1765  *
1766  */
1767 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1768 			   const struct cipso_v4_doi *doi_def,
1769 			   const struct netlbl_lsm_secattr *secattr)
1770 {
1771 	int ret_val;
1772 	u32 iter;
1773 
1774 	if (buf_len <= CIPSO_V4_HDR_LEN)
1775 		return -ENOSPC;
1776 
1777 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1778 	 * really a good assumption to make but since we only support the MAC
1779 	 * tags right now it is a safe assumption. */
1780 	iter = 0;
1781 	do {
1782 		memset(buf, 0, buf_len);
1783 		switch (doi_def->tags[iter]) {
1784 		case CIPSO_V4_TAG_RBITMAP:
1785 			ret_val = cipso_v4_gentag_rbm(doi_def,
1786 						   secattr,
1787 						   &buf[CIPSO_V4_HDR_LEN],
1788 						   buf_len - CIPSO_V4_HDR_LEN);
1789 			break;
1790 		case CIPSO_V4_TAG_ENUM:
1791 			ret_val = cipso_v4_gentag_enum(doi_def,
1792 						   secattr,
1793 						   &buf[CIPSO_V4_HDR_LEN],
1794 						   buf_len - CIPSO_V4_HDR_LEN);
1795 			break;
1796 		case CIPSO_V4_TAG_RANGE:
1797 			ret_val = cipso_v4_gentag_rng(doi_def,
1798 						   secattr,
1799 						   &buf[CIPSO_V4_HDR_LEN],
1800 						   buf_len - CIPSO_V4_HDR_LEN);
1801 			break;
1802 		case CIPSO_V4_TAG_LOCAL:
1803 			ret_val = cipso_v4_gentag_loc(doi_def,
1804 						   secattr,
1805 						   &buf[CIPSO_V4_HDR_LEN],
1806 						   buf_len - CIPSO_V4_HDR_LEN);
1807 			break;
1808 		default:
1809 			return -EPERM;
1810 		}
1811 
1812 		iter++;
1813 	} while (ret_val < 0 &&
1814 		 iter < CIPSO_V4_TAG_MAXCNT &&
1815 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1816 	if (ret_val < 0)
1817 		return ret_val;
1818 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1819 	return CIPSO_V4_HDR_LEN + ret_val;
1820 }
1821 
1822 /**
1823  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1824  * @sk: the socket
1825  * @doi_def: the CIPSO DOI to use
1826  * @secattr: the specific security attributes of the socket
1827  *
1828  * Description:
1829  * Set the CIPSO option on the given socket using the DOI definition and
1830  * security attributes passed to the function.  This function requires
1831  * exclusive access to @sk, which means it either needs to be in the
1832  * process of being created or locked.  Returns zero on success and negative
1833  * values on failure.
1834  *
1835  */
1836 int cipso_v4_sock_setattr(struct sock *sk,
1837 			  const struct cipso_v4_doi *doi_def,
1838 			  const struct netlbl_lsm_secattr *secattr)
1839 {
1840 	int ret_val = -EPERM;
1841 	unsigned char *buf = NULL;
1842 	u32 buf_len;
1843 	u32 opt_len;
1844 	struct ip_options_rcu *old, *opt = NULL;
1845 	struct inet_sock *sk_inet;
1846 	struct inet_connection_sock *sk_conn;
1847 
1848 	/* In the case of sock_create_lite(), the sock->sk field is not
1849 	 * defined yet but it is not a problem as the only users of these
1850 	 * "lite" PF_INET sockets are functions which do an accept() call
1851 	 * afterwards so we will label the socket as part of the accept(). */
1852 	if (!sk)
1853 		return 0;
1854 
1855 	/* We allocate the maximum CIPSO option size here so we are probably
1856 	 * being a little wasteful, but it makes our life _much_ easier later
1857 	 * on and after all we are only talking about 40 bytes. */
1858 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1859 	buf = kmalloc(buf_len, GFP_ATOMIC);
1860 	if (!buf) {
1861 		ret_val = -ENOMEM;
1862 		goto socket_setattr_failure;
1863 	}
1864 
1865 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1866 	if (ret_val < 0)
1867 		goto socket_setattr_failure;
1868 	buf_len = ret_val;
1869 
1870 	/* We can't use ip_options_get() directly because it makes a call to
1871 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1872 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1873 	 * set the IPOPT_CIPSO option. */
1874 	opt_len = (buf_len + 3) & ~3;
1875 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1876 	if (!opt) {
1877 		ret_val = -ENOMEM;
1878 		goto socket_setattr_failure;
1879 	}
1880 	memcpy(opt->opt.__data, buf, buf_len);
1881 	opt->opt.optlen = opt_len;
1882 	opt->opt.cipso = sizeof(struct iphdr);
1883 	kfree(buf);
1884 	buf = NULL;
1885 
1886 	sk_inet = inet_sk(sk);
1887 
1888 	old = rcu_dereference_protected(sk_inet->inet_opt,
1889 					lockdep_sock_is_held(sk));
1890 	if (sk_inet->is_icsk) {
1891 		sk_conn = inet_csk(sk);
1892 		if (old)
1893 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1894 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1895 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1896 	}
1897 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1898 	if (old)
1899 		kfree_rcu(old, rcu);
1900 
1901 	return 0;
1902 
1903 socket_setattr_failure:
1904 	kfree(buf);
1905 	kfree(opt);
1906 	return ret_val;
1907 }
1908 
1909 /**
1910  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1911  * @req: the connection request socket
1912  * @doi_def: the CIPSO DOI to use
1913  * @secattr: the specific security attributes of the socket
1914  *
1915  * Description:
1916  * Set the CIPSO option on the given socket using the DOI definition and
1917  * security attributes passed to the function.  Returns zero on success and
1918  * negative values on failure.
1919  *
1920  */
1921 int cipso_v4_req_setattr(struct request_sock *req,
1922 			 const struct cipso_v4_doi *doi_def,
1923 			 const struct netlbl_lsm_secattr *secattr)
1924 {
1925 	int ret_val = -EPERM;
1926 	unsigned char *buf = NULL;
1927 	u32 buf_len;
1928 	u32 opt_len;
1929 	struct ip_options_rcu *opt = NULL;
1930 	struct inet_request_sock *req_inet;
1931 
1932 	/* We allocate the maximum CIPSO option size here so we are probably
1933 	 * being a little wasteful, but it makes our life _much_ easier later
1934 	 * on and after all we are only talking about 40 bytes. */
1935 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1936 	buf = kmalloc(buf_len, GFP_ATOMIC);
1937 	if (!buf) {
1938 		ret_val = -ENOMEM;
1939 		goto req_setattr_failure;
1940 	}
1941 
1942 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1943 	if (ret_val < 0)
1944 		goto req_setattr_failure;
1945 	buf_len = ret_val;
1946 
1947 	/* We can't use ip_options_get() directly because it makes a call to
1948 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1949 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1950 	 * set the IPOPT_CIPSO option. */
1951 	opt_len = (buf_len + 3) & ~3;
1952 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1953 	if (!opt) {
1954 		ret_val = -ENOMEM;
1955 		goto req_setattr_failure;
1956 	}
1957 	memcpy(opt->opt.__data, buf, buf_len);
1958 	opt->opt.optlen = opt_len;
1959 	opt->opt.cipso = sizeof(struct iphdr);
1960 	kfree(buf);
1961 	buf = NULL;
1962 
1963 	req_inet = inet_rsk(req);
1964 	opt = xchg((__force struct ip_options_rcu **)&req_inet->ireq_opt, opt);
1965 	if (opt)
1966 		kfree_rcu(opt, rcu);
1967 
1968 	return 0;
1969 
1970 req_setattr_failure:
1971 	kfree(buf);
1972 	kfree(opt);
1973 	return ret_val;
1974 }
1975 
1976 /**
1977  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
1978  * @opt_ptr: IP option pointer
1979  *
1980  * Description:
1981  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
1982  * adjustments to the IP option structure.  Returns zero on success, negative
1983  * values on failure.
1984  *
1985  */
1986 static int cipso_v4_delopt(struct ip_options_rcu __rcu **opt_ptr)
1987 {
1988 	struct ip_options_rcu *opt = rcu_dereference_protected(*opt_ptr, 1);
1989 	int hdr_delta = 0;
1990 
1991 	if (!opt || opt->opt.cipso == 0)
1992 		return 0;
1993 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
1994 		u8 cipso_len;
1995 		u8 cipso_off;
1996 		unsigned char *cipso_ptr;
1997 		int iter;
1998 		int optlen_new;
1999 
2000 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
2001 		cipso_ptr = &opt->opt.__data[cipso_off];
2002 		cipso_len = cipso_ptr[1];
2003 
2004 		if (opt->opt.srr > opt->opt.cipso)
2005 			opt->opt.srr -= cipso_len;
2006 		if (opt->opt.rr > opt->opt.cipso)
2007 			opt->opt.rr -= cipso_len;
2008 		if (opt->opt.ts > opt->opt.cipso)
2009 			opt->opt.ts -= cipso_len;
2010 		if (opt->opt.router_alert > opt->opt.cipso)
2011 			opt->opt.router_alert -= cipso_len;
2012 		opt->opt.cipso = 0;
2013 
2014 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2015 			opt->opt.optlen - cipso_off - cipso_len);
2016 
2017 		/* determining the new total option length is tricky because of
2018 		 * the padding necessary, the only thing i can think to do at
2019 		 * this point is walk the options one-by-one, skipping the
2020 		 * padding at the end to determine the actual option size and
2021 		 * from there we can determine the new total option length */
2022 		iter = 0;
2023 		optlen_new = 0;
2024 		while (iter < opt->opt.optlen)
2025 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2026 				iter += opt->opt.__data[iter + 1];
2027 				optlen_new = iter;
2028 			} else
2029 				iter++;
2030 		hdr_delta = opt->opt.optlen;
2031 		opt->opt.optlen = (optlen_new + 3) & ~3;
2032 		hdr_delta -= opt->opt.optlen;
2033 	} else {
2034 		/* only the cipso option was present on the socket so we can
2035 		 * remove the entire option struct */
2036 		*opt_ptr = NULL;
2037 		hdr_delta = opt->opt.optlen;
2038 		kfree_rcu(opt, rcu);
2039 	}
2040 
2041 	return hdr_delta;
2042 }
2043 
2044 /**
2045  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2046  * @sk: the socket
2047  *
2048  * Description:
2049  * Removes the CIPSO option from a socket, if present.
2050  *
2051  */
2052 void cipso_v4_sock_delattr(struct sock *sk)
2053 {
2054 	struct inet_sock *sk_inet;
2055 	int hdr_delta;
2056 
2057 	sk_inet = inet_sk(sk);
2058 
2059 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2060 	if (sk_inet->is_icsk && hdr_delta > 0) {
2061 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2062 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2063 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2064 	}
2065 }
2066 
2067 /**
2068  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2069  * @reg: the request socket
2070  *
2071  * Description:
2072  * Removes the CIPSO option from a request socket, if present.
2073  *
2074  */
2075 void cipso_v4_req_delattr(struct request_sock *req)
2076 {
2077 	cipso_v4_delopt(&inet_rsk(req)->ireq_opt);
2078 }
2079 
2080 /**
2081  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2082  * @cipso: the CIPSO v4 option
2083  * @secattr: the security attributes
2084  *
2085  * Description:
2086  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2087  * on success and negative values on failure.
2088  *
2089  */
2090 int cipso_v4_getattr(const unsigned char *cipso,
2091 		     struct netlbl_lsm_secattr *secattr)
2092 {
2093 	int ret_val = -ENOMSG;
2094 	u32 doi;
2095 	struct cipso_v4_doi *doi_def;
2096 
2097 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2098 		return 0;
2099 
2100 	doi = get_unaligned_be32(&cipso[2]);
2101 	rcu_read_lock();
2102 	doi_def = cipso_v4_doi_search(doi);
2103 	if (!doi_def)
2104 		goto getattr_return;
2105 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2106 	 * really a good assumption to make but since we only support the MAC
2107 	 * tags right now it is a safe assumption. */
2108 	switch (cipso[6]) {
2109 	case CIPSO_V4_TAG_RBITMAP:
2110 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2111 		break;
2112 	case CIPSO_V4_TAG_ENUM:
2113 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2114 		break;
2115 	case CIPSO_V4_TAG_RANGE:
2116 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2117 		break;
2118 	case CIPSO_V4_TAG_LOCAL:
2119 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2120 		break;
2121 	}
2122 	if (ret_val == 0)
2123 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2124 
2125 getattr_return:
2126 	rcu_read_unlock();
2127 	return ret_val;
2128 }
2129 
2130 /**
2131  * cipso_v4_sock_getattr - Get the security attributes from a sock
2132  * @sk: the sock
2133  * @secattr: the security attributes
2134  *
2135  * Description:
2136  * Query @sk to see if there is a CIPSO option attached to the sock and if
2137  * there is return the CIPSO security attributes in @secattr.  This function
2138  * requires that @sk be locked, or privately held, but it does not do any
2139  * locking itself.  Returns zero on success and negative values on failure.
2140  *
2141  */
2142 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2143 {
2144 	struct ip_options_rcu *opt;
2145 	int res = -ENOMSG;
2146 
2147 	rcu_read_lock();
2148 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2149 	if (opt && opt->opt.cipso)
2150 		res = cipso_v4_getattr(opt->opt.__data +
2151 						opt->opt.cipso -
2152 						sizeof(struct iphdr),
2153 				       secattr);
2154 	rcu_read_unlock();
2155 	return res;
2156 }
2157 
2158 /**
2159  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2160  * @skb: the packet
2161  * @secattr: the security attributes
2162  *
2163  * Description:
2164  * Set the CIPSO option on the given packet based on the security attributes.
2165  * Returns a pointer to the IP header on success and NULL on failure.
2166  *
2167  */
2168 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2169 			    const struct cipso_v4_doi *doi_def,
2170 			    const struct netlbl_lsm_secattr *secattr)
2171 {
2172 	int ret_val;
2173 	struct iphdr *iph;
2174 	struct ip_options *opt = &IPCB(skb)->opt;
2175 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2176 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2177 	u32 opt_len;
2178 	int len_delta;
2179 
2180 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2181 	if (ret_val < 0)
2182 		return ret_val;
2183 	buf_len = ret_val;
2184 	opt_len = (buf_len + 3) & ~3;
2185 
2186 	/* we overwrite any existing options to ensure that we have enough
2187 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2188 	 * that the security label is applied to the packet - we do the same
2189 	 * thing when using the socket options and it hasn't caused a problem,
2190 	 * if we need to we can always revisit this choice later */
2191 
2192 	len_delta = opt_len - opt->optlen;
2193 	/* if we don't ensure enough headroom we could panic on the skb_push()
2194 	 * call below so make sure we have enough, we are also "mangling" the
2195 	 * packet so we should probably do a copy-on-write call anyway */
2196 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2197 	if (ret_val < 0)
2198 		return ret_val;
2199 
2200 	if (len_delta > 0) {
2201 		/* we assume that the header + opt->optlen have already been
2202 		 * "pushed" in ip_options_build() or similar */
2203 		iph = ip_hdr(skb);
2204 		skb_push(skb, len_delta);
2205 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2206 		skb_reset_network_header(skb);
2207 		iph = ip_hdr(skb);
2208 	} else if (len_delta < 0) {
2209 		iph = ip_hdr(skb);
2210 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2211 	} else
2212 		iph = ip_hdr(skb);
2213 
2214 	if (opt->optlen > 0)
2215 		memset(opt, 0, sizeof(*opt));
2216 	opt->optlen = opt_len;
2217 	opt->cipso = sizeof(struct iphdr);
2218 	opt->is_changed = 1;
2219 
2220 	/* we have to do the following because we are being called from a
2221 	 * netfilter hook which means the packet already has had the header
2222 	 * fields populated and the checksum calculated - yes this means we
2223 	 * are doing more work than needed but we do it to keep the core
2224 	 * stack clean and tidy */
2225 	memcpy(iph + 1, buf, buf_len);
2226 	if (opt_len > buf_len)
2227 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2228 	if (len_delta != 0) {
2229 		iph->ihl = 5 + (opt_len >> 2);
2230 		iph->tot_len = htons(skb->len);
2231 	}
2232 	ip_send_check(iph);
2233 
2234 	return 0;
2235 }
2236 
2237 /**
2238  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2239  * @skb: the packet
2240  *
2241  * Description:
2242  * Removes any and all CIPSO options from the given packet.  Returns zero on
2243  * success, negative values on failure.
2244  *
2245  */
2246 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2247 {
2248 	int ret_val;
2249 	struct iphdr *iph;
2250 	struct ip_options *opt = &IPCB(skb)->opt;
2251 	unsigned char *cipso_ptr;
2252 
2253 	if (opt->cipso == 0)
2254 		return 0;
2255 
2256 	/* since we are changing the packet we should make a copy */
2257 	ret_val = skb_cow(skb, skb_headroom(skb));
2258 	if (ret_val < 0)
2259 		return ret_val;
2260 
2261 	/* the easiest thing to do is just replace the cipso option with noop
2262 	 * options since we don't change the size of the packet, although we
2263 	 * still need to recalculate the checksum */
2264 
2265 	iph = ip_hdr(skb);
2266 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2267 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2268 	opt->cipso = 0;
2269 	opt->is_changed = 1;
2270 
2271 	ip_send_check(iph);
2272 
2273 	return 0;
2274 }
2275 
2276 /*
2277  * Setup Functions
2278  */
2279 
2280 /**
2281  * cipso_v4_init - Initialize the CIPSO module
2282  *
2283  * Description:
2284  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2285  * and negative values on failure.
2286  *
2287  */
2288 static int __init cipso_v4_init(void)
2289 {
2290 	int ret_val;
2291 
2292 	ret_val = cipso_v4_cache_init();
2293 	if (ret_val != 0)
2294 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2295 		      ret_val);
2296 
2297 	return 0;
2298 }
2299 
2300 subsys_initcall(cipso_v4_init);
2301