xref: /linux/net/ipv4/cipso_ipv4.c (revision 71dfa617ea9f18e4585fe78364217cd32b1fc382)
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  *   https://tools.ietf.org/id/draft-ietf-cipso-ipsecurity-01.txt
14  * The FIPS-188 specification can be found at the following URL:
15  *   https://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;
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.
191  *
192  */
193 void cipso_v4_cache_invalidate(void)
194 {
195 	struct cipso_v4_map_cache_entry *entry, *tmp_entry;
196 	u32 iter;
197 
198 	for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
199 		spin_lock_bh(&cipso_v4_cache[iter].lock);
200 		list_for_each_entry_safe(entry,
201 					 tmp_entry,
202 					 &cipso_v4_cache[iter].list, list) {
203 			list_del(&entry->list);
204 			cipso_v4_cache_entry_free(entry);
205 		}
206 		cipso_v4_cache[iter].size = 0;
207 		spin_unlock_bh(&cipso_v4_cache[iter].lock);
208 	}
209 }
210 
211 /**
212  * cipso_v4_cache_check - Check the CIPSO cache for a label mapping
213  * @key: the buffer to check
214  * @key_len: buffer length in bytes
215  * @secattr: the security attribute struct to use
216  *
217  * Description:
218  * This function checks the cache to see if a label mapping already exists for
219  * the given key.  If there is a match then the cache is adjusted and the
220  * @secattr struct is populated with the correct LSM security attributes.  The
221  * cache is adjusted in the following manner if the entry is not already the
222  * first in the cache bucket:
223  *
224  *  1. The cache entry's activity counter is incremented
225  *  2. The previous (higher ranking) entry's activity counter is decremented
226  *  3. If the difference between the two activity counters is geater than
227  *     CIPSO_V4_CACHE_REORDERLIMIT the two entries are swapped
228  *
229  * Returns zero on success, -ENOENT for a cache miss, and other negative values
230  * on error.
231  *
232  */
233 static int cipso_v4_cache_check(const unsigned char *key,
234 				u32 key_len,
235 				struct netlbl_lsm_secattr *secattr)
236 {
237 	u32 bkt;
238 	struct cipso_v4_map_cache_entry *entry;
239 	struct cipso_v4_map_cache_entry *prev_entry = NULL;
240 	u32 hash;
241 
242 	if (!READ_ONCE(cipso_v4_cache_enabled))
243 		return -ENOENT;
244 
245 	hash = cipso_v4_map_cache_hash(key, key_len);
246 	bkt = hash & (CIPSO_V4_CACHE_BUCKETS - 1);
247 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
248 	list_for_each_entry(entry, &cipso_v4_cache[bkt].list, list) {
249 		if (entry->hash == hash &&
250 		    entry->key_len == key_len &&
251 		    memcmp(entry->key, key, key_len) == 0) {
252 			entry->activity += 1;
253 			refcount_inc(&entry->lsm_data->refcount);
254 			secattr->cache = entry->lsm_data;
255 			secattr->flags |= NETLBL_SECATTR_CACHE;
256 			secattr->type = NETLBL_NLTYPE_CIPSOV4;
257 			if (!prev_entry) {
258 				spin_unlock_bh(&cipso_v4_cache[bkt].lock);
259 				return 0;
260 			}
261 
262 			if (prev_entry->activity > 0)
263 				prev_entry->activity -= 1;
264 			if (entry->activity > prev_entry->activity &&
265 			    entry->activity - prev_entry->activity >
266 			    CIPSO_V4_CACHE_REORDERLIMIT) {
267 				__list_del(entry->list.prev, entry->list.next);
268 				__list_add(&entry->list,
269 					   prev_entry->list.prev,
270 					   &prev_entry->list);
271 			}
272 
273 			spin_unlock_bh(&cipso_v4_cache[bkt].lock);
274 			return 0;
275 		}
276 		prev_entry = entry;
277 	}
278 	spin_unlock_bh(&cipso_v4_cache[bkt].lock);
279 
280 	return -ENOENT;
281 }
282 
283 /**
284  * cipso_v4_cache_add - Add an entry to the CIPSO cache
285  * @cipso_ptr: pointer to CIPSO IP option
286  * @secattr: the packet's security attributes
287  *
288  * Description:
289  * Add a new entry into the CIPSO label mapping cache.  Add the new entry to
290  * head of the cache bucket's list, if the cache bucket is out of room remove
291  * the last entry in the list first.  It is important to note that there is
292  * currently no checking for duplicate keys.  Returns zero on success,
293  * negative values on failure.
294  *
295  */
296 int cipso_v4_cache_add(const unsigned char *cipso_ptr,
297 		       const struct netlbl_lsm_secattr *secattr)
298 {
299 	int bkt_size = READ_ONCE(cipso_v4_cache_bucketsize);
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 (!READ_ONCE(cipso_v4_cache_enabled) || bkt_size <= 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 < bkt_size) {
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 		kfree(doi_def->map.std);
476 		break;
477 	}
478 	kfree(doi_def);
479 }
480 
481 /**
482  * cipso_v4_doi_free_rcu - Frees a DOI definition via the RCU pointer
483  * @entry: the entry's RCU field
484  *
485  * Description:
486  * This function is designed to be used as a callback to the call_rcu()
487  * function so that the memory allocated to the DOI definition can be released
488  * safely.
489  *
490  */
491 static void cipso_v4_doi_free_rcu(struct rcu_head *entry)
492 {
493 	struct cipso_v4_doi *doi_def;
494 
495 	doi_def = container_of(entry, struct cipso_v4_doi, rcu);
496 	cipso_v4_doi_free(doi_def);
497 }
498 
499 /**
500  * cipso_v4_doi_remove - Remove an existing DOI from the CIPSO protocol engine
501  * @doi: the DOI value
502  * @audit_info: NetLabel audit information
503  *
504  * Description:
505  * Removes a DOI definition from the CIPSO engine.  The NetLabel routines will
506  * be called to release their own LSM domain mappings as well as our own
507  * domain list.  Returns zero on success and negative values on failure.
508  *
509  */
510 int cipso_v4_doi_remove(u32 doi, struct netlbl_audit *audit_info)
511 {
512 	int ret_val;
513 	struct cipso_v4_doi *doi_def;
514 	struct audit_buffer *audit_buf;
515 
516 	spin_lock(&cipso_v4_doi_list_lock);
517 	doi_def = cipso_v4_doi_search(doi);
518 	if (!doi_def) {
519 		spin_unlock(&cipso_v4_doi_list_lock);
520 		ret_val = -ENOENT;
521 		goto doi_remove_return;
522 	}
523 	list_del_rcu(&doi_def->list);
524 	spin_unlock(&cipso_v4_doi_list_lock);
525 
526 	cipso_v4_doi_putdef(doi_def);
527 	ret_val = 0;
528 
529 doi_remove_return:
530 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_DEL, audit_info);
531 	if (audit_buf) {
532 		audit_log_format(audit_buf,
533 				 " cipso_doi=%u res=%u",
534 				 doi, ret_val == 0 ? 1 : 0);
535 		audit_log_end(audit_buf);
536 	}
537 
538 	return ret_val;
539 }
540 
541 /**
542  * cipso_v4_doi_getdef - Returns a reference to a valid DOI definition
543  * @doi: the DOI value
544  *
545  * Description:
546  * Searches for a valid DOI definition and if one is found it is returned to
547  * the caller.  Otherwise NULL is returned.  The caller must ensure that
548  * rcu_read_lock() is held while accessing the returned definition and the DOI
549  * definition reference count is decremented when the caller is done.
550  *
551  */
552 struct cipso_v4_doi *cipso_v4_doi_getdef(u32 doi)
553 {
554 	struct cipso_v4_doi *doi_def;
555 
556 	rcu_read_lock();
557 	doi_def = cipso_v4_doi_search(doi);
558 	if (!doi_def)
559 		goto doi_getdef_return;
560 	if (!refcount_inc_not_zero(&doi_def->refcount))
561 		doi_def = NULL;
562 
563 doi_getdef_return:
564 	rcu_read_unlock();
565 	return doi_def;
566 }
567 
568 /**
569  * cipso_v4_doi_putdef - Releases a reference for the given DOI definition
570  * @doi_def: the DOI definition
571  *
572  * Description:
573  * Releases a DOI definition reference obtained from cipso_v4_doi_getdef().
574  *
575  */
576 void cipso_v4_doi_putdef(struct cipso_v4_doi *doi_def)
577 {
578 	if (!doi_def)
579 		return;
580 
581 	if (!refcount_dec_and_test(&doi_def->refcount))
582 		return;
583 
584 	cipso_v4_cache_invalidate();
585 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
586 }
587 
588 /**
589  * cipso_v4_doi_walk - Iterate through the DOI definitions
590  * @skip_cnt: skip past this number of DOI definitions, updated
591  * @callback: callback for each DOI definition
592  * @cb_arg: argument for the callback function
593  *
594  * Description:
595  * Iterate over the DOI definition list, skipping the first @skip_cnt entries.
596  * For each entry call @callback, if @callback returns a negative value stop
597  * 'walking' through the list and return.  Updates the value in @skip_cnt upon
598  * return.  Returns zero on success, negative values on failure.
599  *
600  */
601 int cipso_v4_doi_walk(u32 *skip_cnt,
602 		     int (*callback) (struct cipso_v4_doi *doi_def, void *arg),
603 		     void *cb_arg)
604 {
605 	int ret_val = -ENOENT;
606 	u32 doi_cnt = 0;
607 	struct cipso_v4_doi *iter_doi;
608 
609 	rcu_read_lock();
610 	list_for_each_entry_rcu(iter_doi, &cipso_v4_doi_list, list)
611 		if (refcount_read(&iter_doi->refcount) > 0) {
612 			if (doi_cnt++ < *skip_cnt)
613 				continue;
614 			ret_val = callback(iter_doi, cb_arg);
615 			if (ret_val < 0) {
616 				doi_cnt--;
617 				goto doi_walk_return;
618 			}
619 		}
620 
621 doi_walk_return:
622 	rcu_read_unlock();
623 	*skip_cnt = doi_cnt;
624 	return ret_val;
625 }
626 
627 /*
628  * Label Mapping Functions
629  */
630 
631 /**
632  * cipso_v4_map_lvl_valid - Checks to see if the given level is understood
633  * @doi_def: the DOI definition
634  * @level: the level to check
635  *
636  * Description:
637  * Checks the given level against the given DOI definition and returns a
638  * negative value if the level does not have a valid mapping and a zero value
639  * if the level is defined by the DOI.
640  *
641  */
642 static int cipso_v4_map_lvl_valid(const struct cipso_v4_doi *doi_def, u8 level)
643 {
644 	switch (doi_def->type) {
645 	case CIPSO_V4_MAP_PASS:
646 		return 0;
647 	case CIPSO_V4_MAP_TRANS:
648 		if ((level < doi_def->map.std->lvl.cipso_size) &&
649 		    (doi_def->map.std->lvl.cipso[level] < CIPSO_V4_INV_LVL))
650 			return 0;
651 		break;
652 	}
653 
654 	return -EFAULT;
655 }
656 
657 /**
658  * cipso_v4_map_lvl_hton - Perform a level mapping from the host to the network
659  * @doi_def: the DOI definition
660  * @host_lvl: the host MLS level
661  * @net_lvl: the network/CIPSO MLS level
662  *
663  * Description:
664  * Perform a label mapping to translate a local MLS level to the correct
665  * CIPSO level using the given DOI definition.  Returns zero on success,
666  * negative values otherwise.
667  *
668  */
669 static int cipso_v4_map_lvl_hton(const struct cipso_v4_doi *doi_def,
670 				 u32 host_lvl,
671 				 u32 *net_lvl)
672 {
673 	switch (doi_def->type) {
674 	case CIPSO_V4_MAP_PASS:
675 		*net_lvl = host_lvl;
676 		return 0;
677 	case CIPSO_V4_MAP_TRANS:
678 		if (host_lvl < doi_def->map.std->lvl.local_size &&
679 		    doi_def->map.std->lvl.local[host_lvl] < CIPSO_V4_INV_LVL) {
680 			*net_lvl = doi_def->map.std->lvl.local[host_lvl];
681 			return 0;
682 		}
683 		return -EPERM;
684 	}
685 
686 	return -EINVAL;
687 }
688 
689 /**
690  * cipso_v4_map_lvl_ntoh - Perform a level mapping from the network to the host
691  * @doi_def: the DOI definition
692  * @net_lvl: the network/CIPSO MLS level
693  * @host_lvl: the host MLS level
694  *
695  * Description:
696  * Perform a label mapping to translate a CIPSO level to the correct local MLS
697  * level using the given DOI definition.  Returns zero on success, negative
698  * values otherwise.
699  *
700  */
701 static int cipso_v4_map_lvl_ntoh(const struct cipso_v4_doi *doi_def,
702 				 u32 net_lvl,
703 				 u32 *host_lvl)
704 {
705 	struct cipso_v4_std_map_tbl *map_tbl;
706 
707 	switch (doi_def->type) {
708 	case CIPSO_V4_MAP_PASS:
709 		*host_lvl = net_lvl;
710 		return 0;
711 	case CIPSO_V4_MAP_TRANS:
712 		map_tbl = doi_def->map.std;
713 		if (net_lvl < map_tbl->lvl.cipso_size &&
714 		    map_tbl->lvl.cipso[net_lvl] < CIPSO_V4_INV_LVL) {
715 			*host_lvl = doi_def->map.std->lvl.cipso[net_lvl];
716 			return 0;
717 		}
718 		return -EPERM;
719 	}
720 
721 	return -EINVAL;
722 }
723 
724 /**
725  * cipso_v4_map_cat_rbm_valid - Checks to see if the category bitmap is valid
726  * @doi_def: the DOI definition
727  * @bitmap: category bitmap
728  * @bitmap_len: bitmap length in bytes
729  *
730  * Description:
731  * Checks the given category bitmap against the given DOI definition and
732  * returns a negative value if any of the categories in the bitmap do not have
733  * a valid mapping and a zero value if all of the categories are valid.
734  *
735  */
736 static int cipso_v4_map_cat_rbm_valid(const struct cipso_v4_doi *doi_def,
737 				      const unsigned char *bitmap,
738 				      u32 bitmap_len)
739 {
740 	int cat = -1;
741 	u32 bitmap_len_bits = bitmap_len * 8;
742 	u32 cipso_cat_size;
743 	u32 *cipso_array;
744 
745 	switch (doi_def->type) {
746 	case CIPSO_V4_MAP_PASS:
747 		return 0;
748 	case CIPSO_V4_MAP_TRANS:
749 		cipso_cat_size = doi_def->map.std->cat.cipso_size;
750 		cipso_array = doi_def->map.std->cat.cipso;
751 		for (;;) {
752 			cat = netlbl_bitmap_walk(bitmap,
753 						 bitmap_len_bits,
754 						 cat + 1,
755 						 1);
756 			if (cat < 0)
757 				break;
758 			if (cat >= cipso_cat_size ||
759 			    cipso_array[cat] >= CIPSO_V4_INV_CAT)
760 				return -EFAULT;
761 		}
762 
763 		if (cat == -1)
764 			return 0;
765 		break;
766 	}
767 
768 	return -EFAULT;
769 }
770 
771 /**
772  * cipso_v4_map_cat_rbm_hton - Perform a category mapping from host to network
773  * @doi_def: the DOI definition
774  * @secattr: the security attributes
775  * @net_cat: the zero'd out category bitmap in network/CIPSO format
776  * @net_cat_len: the length of the CIPSO bitmap in bytes
777  *
778  * Description:
779  * Perform a label mapping to translate a local MLS category bitmap to the
780  * correct CIPSO bitmap using the given DOI definition.  Returns the minimum
781  * size in bytes of the network bitmap on success, negative values otherwise.
782  *
783  */
784 static int cipso_v4_map_cat_rbm_hton(const struct cipso_v4_doi *doi_def,
785 				     const struct netlbl_lsm_secattr *secattr,
786 				     unsigned char *net_cat,
787 				     u32 net_cat_len)
788 {
789 	int host_spot = -1;
790 	u32 net_spot = CIPSO_V4_INV_CAT;
791 	u32 net_spot_max = 0;
792 	u32 net_clen_bits = net_cat_len * 8;
793 	u32 host_cat_size = 0;
794 	u32 *host_cat_array = NULL;
795 
796 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
797 		host_cat_size = doi_def->map.std->cat.local_size;
798 		host_cat_array = doi_def->map.std->cat.local;
799 	}
800 
801 	for (;;) {
802 		host_spot = netlbl_catmap_walk(secattr->attr.mls.cat,
803 					       host_spot + 1);
804 		if (host_spot < 0)
805 			break;
806 
807 		switch (doi_def->type) {
808 		case CIPSO_V4_MAP_PASS:
809 			net_spot = host_spot;
810 			break;
811 		case CIPSO_V4_MAP_TRANS:
812 			if (host_spot >= host_cat_size)
813 				return -EPERM;
814 			net_spot = host_cat_array[host_spot];
815 			if (net_spot >= CIPSO_V4_INV_CAT)
816 				return -EPERM;
817 			break;
818 		}
819 		if (net_spot >= net_clen_bits)
820 			return -ENOSPC;
821 		netlbl_bitmap_setbit(net_cat, net_spot, 1);
822 
823 		if (net_spot > net_spot_max)
824 			net_spot_max = net_spot;
825 	}
826 
827 	if (++net_spot_max % 8)
828 		return net_spot_max / 8 + 1;
829 	return net_spot_max / 8;
830 }
831 
832 /**
833  * cipso_v4_map_cat_rbm_ntoh - Perform a category mapping from network to host
834  * @doi_def: the DOI definition
835  * @net_cat: the category bitmap in network/CIPSO format
836  * @net_cat_len: the length of the CIPSO bitmap in bytes
837  * @secattr: the security attributes
838  *
839  * Description:
840  * Perform a label mapping to translate a CIPSO bitmap to the correct local
841  * MLS category bitmap using the given DOI definition.  Returns zero on
842  * success, negative values on failure.
843  *
844  */
845 static int cipso_v4_map_cat_rbm_ntoh(const struct cipso_v4_doi *doi_def,
846 				     const unsigned char *net_cat,
847 				     u32 net_cat_len,
848 				     struct netlbl_lsm_secattr *secattr)
849 {
850 	int ret_val;
851 	int net_spot = -1;
852 	u32 host_spot = CIPSO_V4_INV_CAT;
853 	u32 net_clen_bits = net_cat_len * 8;
854 	u32 net_cat_size = 0;
855 	u32 *net_cat_array = NULL;
856 
857 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
858 		net_cat_size = doi_def->map.std->cat.cipso_size;
859 		net_cat_array = doi_def->map.std->cat.cipso;
860 	}
861 
862 	for (;;) {
863 		net_spot = netlbl_bitmap_walk(net_cat,
864 					      net_clen_bits,
865 					      net_spot + 1,
866 					      1);
867 		if (net_spot < 0)
868 			return 0;
869 
870 		switch (doi_def->type) {
871 		case CIPSO_V4_MAP_PASS:
872 			host_spot = net_spot;
873 			break;
874 		case CIPSO_V4_MAP_TRANS:
875 			if (net_spot >= net_cat_size)
876 				return -EPERM;
877 			host_spot = net_cat_array[net_spot];
878 			if (host_spot >= CIPSO_V4_INV_CAT)
879 				return -EPERM;
880 			break;
881 		}
882 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
883 						       host_spot,
884 						       GFP_ATOMIC);
885 		if (ret_val != 0)
886 			return ret_val;
887 	}
888 
889 	return -EINVAL;
890 }
891 
892 /**
893  * cipso_v4_map_cat_enum_valid - Checks to see if the categories are valid
894  * @doi_def: the DOI definition
895  * @enumcat: category list
896  * @enumcat_len: length of the category list in bytes
897  *
898  * Description:
899  * Checks the given categories against the given DOI definition and returns a
900  * negative value if any of the categories do not have a valid mapping and a
901  * zero value if all of the categories are valid.
902  *
903  */
904 static int cipso_v4_map_cat_enum_valid(const struct cipso_v4_doi *doi_def,
905 				       const unsigned char *enumcat,
906 				       u32 enumcat_len)
907 {
908 	u16 cat;
909 	int cat_prev = -1;
910 	u32 iter;
911 
912 	if (doi_def->type != CIPSO_V4_MAP_PASS || enumcat_len & 0x01)
913 		return -EFAULT;
914 
915 	for (iter = 0; iter < enumcat_len; iter += 2) {
916 		cat = get_unaligned_be16(&enumcat[iter]);
917 		if (cat <= cat_prev)
918 			return -EFAULT;
919 		cat_prev = cat;
920 	}
921 
922 	return 0;
923 }
924 
925 /**
926  * cipso_v4_map_cat_enum_hton - Perform a category mapping from host to network
927  * @doi_def: the DOI definition
928  * @secattr: the security attributes
929  * @net_cat: the zero'd out category list in network/CIPSO format
930  * @net_cat_len: the length of the CIPSO category list in bytes
931  *
932  * Description:
933  * Perform a label mapping to translate a local MLS category bitmap to the
934  * correct CIPSO category list using the given DOI definition.   Returns the
935  * size in bytes of the network category bitmap on success, negative values
936  * otherwise.
937  *
938  */
939 static int cipso_v4_map_cat_enum_hton(const struct cipso_v4_doi *doi_def,
940 				      const struct netlbl_lsm_secattr *secattr,
941 				      unsigned char *net_cat,
942 				      u32 net_cat_len)
943 {
944 	int cat = -1;
945 	u32 cat_iter = 0;
946 
947 	for (;;) {
948 		cat = netlbl_catmap_walk(secattr->attr.mls.cat, cat + 1);
949 		if (cat < 0)
950 			break;
951 		if ((cat_iter + 2) > net_cat_len)
952 			return -ENOSPC;
953 
954 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
955 		cat_iter += 2;
956 	}
957 
958 	return cat_iter;
959 }
960 
961 /**
962  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
963  * @doi_def: the DOI definition
964  * @net_cat: the category list in network/CIPSO format
965  * @net_cat_len: the length of the CIPSO bitmap in bytes
966  * @secattr: the security attributes
967  *
968  * Description:
969  * Perform a label mapping to translate a CIPSO category list to the correct
970  * local MLS category bitmap using the given DOI definition.  Returns zero on
971  * success, negative values on failure.
972  *
973  */
974 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
975 				      const unsigned char *net_cat,
976 				      u32 net_cat_len,
977 				      struct netlbl_lsm_secattr *secattr)
978 {
979 	int ret_val;
980 	u32 iter;
981 
982 	for (iter = 0; iter < net_cat_len; iter += 2) {
983 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
984 					     get_unaligned_be16(&net_cat[iter]),
985 					     GFP_ATOMIC);
986 		if (ret_val != 0)
987 			return ret_val;
988 	}
989 
990 	return 0;
991 }
992 
993 /**
994  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
995  * @doi_def: the DOI definition
996  * @rngcat: category list
997  * @rngcat_len: length of the category list in bytes
998  *
999  * Description:
1000  * Checks the given categories against the given DOI definition and returns a
1001  * negative value if any of the categories do not have a valid mapping and a
1002  * zero value if all of the categories are valid.
1003  *
1004  */
1005 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1006 				      const unsigned char *rngcat,
1007 				      u32 rngcat_len)
1008 {
1009 	u16 cat_high;
1010 	u16 cat_low;
1011 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1012 	u32 iter;
1013 
1014 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1015 		return -EFAULT;
1016 
1017 	for (iter = 0; iter < rngcat_len; iter += 4) {
1018 		cat_high = get_unaligned_be16(&rngcat[iter]);
1019 		if ((iter + 4) <= rngcat_len)
1020 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1021 		else
1022 			cat_low = 0;
1023 
1024 		if (cat_high > cat_prev)
1025 			return -EFAULT;
1026 
1027 		cat_prev = cat_low;
1028 	}
1029 
1030 	return 0;
1031 }
1032 
1033 /**
1034  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1035  * @doi_def: the DOI definition
1036  * @secattr: the security attributes
1037  * @net_cat: the zero'd out category list in network/CIPSO format
1038  * @net_cat_len: the length of the CIPSO category list in bytes
1039  *
1040  * Description:
1041  * Perform a label mapping to translate a local MLS category bitmap to the
1042  * correct CIPSO category list using the given DOI definition.   Returns the
1043  * size in bytes of the network category bitmap on success, negative values
1044  * otherwise.
1045  *
1046  */
1047 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1048 				     const struct netlbl_lsm_secattr *secattr,
1049 				     unsigned char *net_cat,
1050 				     u32 net_cat_len)
1051 {
1052 	int iter = -1;
1053 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1054 	u32 array_cnt = 0;
1055 	u32 cat_size = 0;
1056 
1057 	/* make sure we don't overflow the 'array[]' variable */
1058 	if (net_cat_len >
1059 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1060 		return -ENOSPC;
1061 
1062 	for (;;) {
1063 		iter = netlbl_catmap_walk(secattr->attr.mls.cat, iter + 1);
1064 		if (iter < 0)
1065 			break;
1066 		cat_size += (iter == 0 ? 0 : sizeof(u16));
1067 		if (cat_size > net_cat_len)
1068 			return -ENOSPC;
1069 		array[array_cnt++] = iter;
1070 
1071 		iter = netlbl_catmap_walkrng(secattr->attr.mls.cat, iter);
1072 		if (iter < 0)
1073 			return -EFAULT;
1074 		cat_size += sizeof(u16);
1075 		if (cat_size > net_cat_len)
1076 			return -ENOSPC;
1077 		array[array_cnt++] = iter;
1078 	}
1079 
1080 	for (iter = 0; array_cnt > 0;) {
1081 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1082 		iter += 2;
1083 		array_cnt--;
1084 		if (array[array_cnt] != 0) {
1085 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1086 			iter += 2;
1087 		}
1088 	}
1089 
1090 	return cat_size;
1091 }
1092 
1093 /**
1094  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1095  * @doi_def: the DOI definition
1096  * @net_cat: the category list in network/CIPSO format
1097  * @net_cat_len: the length of the CIPSO bitmap in bytes
1098  * @secattr: the security attributes
1099  *
1100  * Description:
1101  * Perform a label mapping to translate a CIPSO category list to the correct
1102  * local MLS category bitmap using the given DOI definition.  Returns zero on
1103  * success, negative values on failure.
1104  *
1105  */
1106 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1107 				     const unsigned char *net_cat,
1108 				     u32 net_cat_len,
1109 				     struct netlbl_lsm_secattr *secattr)
1110 {
1111 	int ret_val;
1112 	u32 net_iter;
1113 	u16 cat_low;
1114 	u16 cat_high;
1115 
1116 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1117 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1118 		if ((net_iter + 4) <= net_cat_len)
1119 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1120 		else
1121 			cat_low = 0;
1122 
1123 		ret_val = netlbl_catmap_setrng(&secattr->attr.mls.cat,
1124 					       cat_low,
1125 					       cat_high,
1126 					       GFP_ATOMIC);
1127 		if (ret_val != 0)
1128 			return ret_val;
1129 	}
1130 
1131 	return 0;
1132 }
1133 
1134 /*
1135  * Protocol Handling Functions
1136  */
1137 
1138 /**
1139  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1140  * @doi_def: the DOI definition
1141  * @len: the total tag length in bytes, not including this header
1142  * @buf: the CIPSO option buffer
1143  *
1144  * Description:
1145  * Write a CIPSO header into the beginning of @buffer.
1146  *
1147  */
1148 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1149 				unsigned char *buf,
1150 				u32 len)
1151 {
1152 	buf[0] = IPOPT_CIPSO;
1153 	buf[1] = CIPSO_V4_HDR_LEN + len;
1154 	put_unaligned_be32(doi_def->doi, &buf[2]);
1155 }
1156 
1157 /**
1158  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1159  * @doi_def: the DOI definition
1160  * @secattr: the security attributes
1161  * @buffer: the option buffer
1162  * @buffer_len: length of buffer in bytes
1163  *
1164  * Description:
1165  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1166  * actual buffer length may be larger than the indicated size due to
1167  * translation between host and network category bitmaps.  Returns the size of
1168  * the tag on success, negative values on failure.
1169  *
1170  */
1171 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1172 			       const struct netlbl_lsm_secattr *secattr,
1173 			       unsigned char *buffer,
1174 			       u32 buffer_len)
1175 {
1176 	int ret_val;
1177 	u32 tag_len;
1178 	u32 level;
1179 
1180 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1181 		return -EPERM;
1182 
1183 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1184 					secattr->attr.mls.lvl,
1185 					&level);
1186 	if (ret_val != 0)
1187 		return ret_val;
1188 
1189 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1190 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1191 						    secattr,
1192 						    &buffer[4],
1193 						    buffer_len - 4);
1194 		if (ret_val < 0)
1195 			return ret_val;
1196 
1197 		/* This will send packets using the "optimized" format when
1198 		 * possible as specified in  section 3.4.2.6 of the
1199 		 * CIPSO draft. */
1200 		if (READ_ONCE(cipso_v4_rbm_optfmt) && ret_val > 0 &&
1201 		    ret_val <= 10)
1202 			tag_len = 14;
1203 		else
1204 			tag_len = 4 + ret_val;
1205 	} else
1206 		tag_len = 4;
1207 
1208 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1209 	buffer[1] = tag_len;
1210 	buffer[3] = level;
1211 
1212 	return tag_len;
1213 }
1214 
1215 /**
1216  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1217  * @doi_def: the DOI definition
1218  * @tag: the CIPSO tag
1219  * @secattr: the security attributes
1220  *
1221  * Description:
1222  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1223  * attributes in @secattr.  Return zero on success, negatives values on
1224  * failure.
1225  *
1226  */
1227 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1228 				 const unsigned char *tag,
1229 				 struct netlbl_lsm_secattr *secattr)
1230 {
1231 	int ret_val;
1232 	u8 tag_len = tag[1];
1233 	u32 level;
1234 
1235 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1236 	if (ret_val != 0)
1237 		return ret_val;
1238 	secattr->attr.mls.lvl = level;
1239 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1240 
1241 	if (tag_len > 4) {
1242 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1243 						    &tag[4],
1244 						    tag_len - 4,
1245 						    secattr);
1246 		if (ret_val != 0) {
1247 			netlbl_catmap_free(secattr->attr.mls.cat);
1248 			return ret_val;
1249 		}
1250 
1251 		if (secattr->attr.mls.cat)
1252 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1253 	}
1254 
1255 	return 0;
1256 }
1257 
1258 /**
1259  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1260  * @doi_def: the DOI definition
1261  * @secattr: the security attributes
1262  * @buffer: the option buffer
1263  * @buffer_len: length of buffer in bytes
1264  *
1265  * Description:
1266  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1267  * size of the tag on success, negative values on failure.
1268  *
1269  */
1270 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1271 				const struct netlbl_lsm_secattr *secattr,
1272 				unsigned char *buffer,
1273 				u32 buffer_len)
1274 {
1275 	int ret_val;
1276 	u32 tag_len;
1277 	u32 level;
1278 
1279 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1280 		return -EPERM;
1281 
1282 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1283 					secattr->attr.mls.lvl,
1284 					&level);
1285 	if (ret_val != 0)
1286 		return ret_val;
1287 
1288 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1289 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1290 						     secattr,
1291 						     &buffer[4],
1292 						     buffer_len - 4);
1293 		if (ret_val < 0)
1294 			return ret_val;
1295 
1296 		tag_len = 4 + ret_val;
1297 	} else
1298 		tag_len = 4;
1299 
1300 	buffer[0] = CIPSO_V4_TAG_ENUM;
1301 	buffer[1] = tag_len;
1302 	buffer[3] = level;
1303 
1304 	return tag_len;
1305 }
1306 
1307 /**
1308  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1309  * @doi_def: the DOI definition
1310  * @tag: the CIPSO tag
1311  * @secattr: the security attributes
1312  *
1313  * Description:
1314  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1315  * attributes in @secattr.  Return zero on success, negatives values on
1316  * failure.
1317  *
1318  */
1319 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1320 				  const unsigned char *tag,
1321 				  struct netlbl_lsm_secattr *secattr)
1322 {
1323 	int ret_val;
1324 	u8 tag_len = tag[1];
1325 	u32 level;
1326 
1327 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1328 	if (ret_val != 0)
1329 		return ret_val;
1330 	secattr->attr.mls.lvl = level;
1331 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1332 
1333 	if (tag_len > 4) {
1334 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1335 						     &tag[4],
1336 						     tag_len - 4,
1337 						     secattr);
1338 		if (ret_val != 0) {
1339 			netlbl_catmap_free(secattr->attr.mls.cat);
1340 			return ret_val;
1341 		}
1342 
1343 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1344 	}
1345 
1346 	return 0;
1347 }
1348 
1349 /**
1350  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1351  * @doi_def: the DOI definition
1352  * @secattr: the security attributes
1353  * @buffer: the option buffer
1354  * @buffer_len: length of buffer in bytes
1355  *
1356  * Description:
1357  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1358  * size of the tag on success, negative values on failure.
1359  *
1360  */
1361 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1362 			       const struct netlbl_lsm_secattr *secattr,
1363 			       unsigned char *buffer,
1364 			       u32 buffer_len)
1365 {
1366 	int ret_val;
1367 	u32 tag_len;
1368 	u32 level;
1369 
1370 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1371 		return -EPERM;
1372 
1373 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1374 					secattr->attr.mls.lvl,
1375 					&level);
1376 	if (ret_val != 0)
1377 		return ret_val;
1378 
1379 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1380 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1381 						    secattr,
1382 						    &buffer[4],
1383 						    buffer_len - 4);
1384 		if (ret_val < 0)
1385 			return ret_val;
1386 
1387 		tag_len = 4 + ret_val;
1388 	} else
1389 		tag_len = 4;
1390 
1391 	buffer[0] = CIPSO_V4_TAG_RANGE;
1392 	buffer[1] = tag_len;
1393 	buffer[3] = level;
1394 
1395 	return tag_len;
1396 }
1397 
1398 /**
1399  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1400  * @doi_def: the DOI definition
1401  * @tag: the CIPSO tag
1402  * @secattr: the security attributes
1403  *
1404  * Description:
1405  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1406  * in @secattr.  Return zero on success, negatives values on failure.
1407  *
1408  */
1409 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1410 				 const unsigned char *tag,
1411 				 struct netlbl_lsm_secattr *secattr)
1412 {
1413 	int ret_val;
1414 	u8 tag_len = tag[1];
1415 	u32 level;
1416 
1417 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1418 	if (ret_val != 0)
1419 		return ret_val;
1420 	secattr->attr.mls.lvl = level;
1421 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1422 
1423 	if (tag_len > 4) {
1424 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1425 						    &tag[4],
1426 						    tag_len - 4,
1427 						    secattr);
1428 		if (ret_val != 0) {
1429 			netlbl_catmap_free(secattr->attr.mls.cat);
1430 			return ret_val;
1431 		}
1432 
1433 		if (secattr->attr.mls.cat)
1434 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1435 	}
1436 
1437 	return 0;
1438 }
1439 
1440 /**
1441  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1442  * @doi_def: the DOI definition
1443  * @secattr: the security attributes
1444  * @buffer: the option buffer
1445  * @buffer_len: length of buffer in bytes
1446  *
1447  * Description:
1448  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1449  * on success, negative values on failure.
1450  *
1451  */
1452 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1453 			       const struct netlbl_lsm_secattr *secattr,
1454 			       unsigned char *buffer,
1455 			       u32 buffer_len)
1456 {
1457 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1458 		return -EPERM;
1459 
1460 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1461 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1462 	*(u32 *)&buffer[2] = secattr->attr.secid;
1463 
1464 	return CIPSO_V4_TAG_LOC_BLEN;
1465 }
1466 
1467 /**
1468  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1469  * @doi_def: the DOI definition
1470  * @tag: the CIPSO tag
1471  * @secattr: the security attributes
1472  *
1473  * Description:
1474  * Parse a CIPSO local tag and return the security attributes in @secattr.
1475  * Return zero on success, negatives values on failure.
1476  *
1477  */
1478 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1479 				 const unsigned char *tag,
1480 				 struct netlbl_lsm_secattr *secattr)
1481 {
1482 	secattr->attr.secid = *(u32 *)&tag[2];
1483 	secattr->flags |= NETLBL_SECATTR_SECID;
1484 
1485 	return 0;
1486 }
1487 
1488 /**
1489  * cipso_v4_optptr - Find the CIPSO option in the packet
1490  * @skb: the packet
1491  *
1492  * Description:
1493  * Parse the packet's IP header looking for a CIPSO option.  Returns a pointer
1494  * to the start of the CIPSO option on success, NULL if one is not found.
1495  *
1496  */
1497 unsigned char *cipso_v4_optptr(const struct sk_buff *skb)
1498 {
1499 	const struct iphdr *iph = ip_hdr(skb);
1500 	unsigned char *optptr = (unsigned char *)&(ip_hdr(skb)[1]);
1501 	int optlen;
1502 	int taglen;
1503 
1504 	for (optlen = iph->ihl*4 - sizeof(struct iphdr); optlen > 1; ) {
1505 		switch (optptr[0]) {
1506 		case IPOPT_END:
1507 			return NULL;
1508 		case IPOPT_NOOP:
1509 			taglen = 1;
1510 			break;
1511 		default:
1512 			taglen = optptr[1];
1513 		}
1514 		if (!taglen || taglen > optlen)
1515 			return NULL;
1516 		if (optptr[0] == IPOPT_CIPSO)
1517 			return optptr;
1518 
1519 		optlen -= taglen;
1520 		optptr += taglen;
1521 	}
1522 
1523 	return NULL;
1524 }
1525 
1526 /**
1527  * cipso_v4_validate - Validate a CIPSO option
1528  * @skb: the packet
1529  * @option: the start of the option, on error it is set to point to the error
1530  *
1531  * Description:
1532  * This routine is called to validate a CIPSO option, it checks all of the
1533  * fields to ensure that they are at least valid, see the draft snippet below
1534  * for details.  If the option is valid then a zero value is returned and
1535  * the value of @option is unchanged.  If the option is invalid then a
1536  * non-zero value is returned and @option is adjusted to point to the
1537  * offending portion of the option.  From the IETF draft ...
1538  *
1539  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1540  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1541  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1542  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1543  *   that is unrecognized."
1544  *
1545  */
1546 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1547 {
1548 	unsigned char *opt = *option;
1549 	unsigned char *tag;
1550 	unsigned char opt_iter;
1551 	unsigned char err_offset = 0;
1552 	u8 opt_len;
1553 	u8 tag_len;
1554 	struct cipso_v4_doi *doi_def = NULL;
1555 	u32 tag_iter;
1556 
1557 	/* caller already checks for length values that are too large */
1558 	opt_len = opt[1];
1559 	if (opt_len < 8) {
1560 		err_offset = 1;
1561 		goto validate_return;
1562 	}
1563 
1564 	rcu_read_lock();
1565 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1566 	if (!doi_def) {
1567 		err_offset = 2;
1568 		goto validate_return_locked;
1569 	}
1570 
1571 	opt_iter = CIPSO_V4_HDR_LEN;
1572 	tag = opt + opt_iter;
1573 	while (opt_iter < opt_len) {
1574 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1575 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1576 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1577 				err_offset = opt_iter;
1578 				goto validate_return_locked;
1579 			}
1580 
1581 		if (opt_iter + 1 == opt_len) {
1582 			err_offset = opt_iter;
1583 			goto validate_return_locked;
1584 		}
1585 		tag_len = tag[1];
1586 		if (tag_len > (opt_len - opt_iter)) {
1587 			err_offset = opt_iter + 1;
1588 			goto validate_return_locked;
1589 		}
1590 
1591 		switch (tag[0]) {
1592 		case CIPSO_V4_TAG_RBITMAP:
1593 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1594 				err_offset = opt_iter + 1;
1595 				goto validate_return_locked;
1596 			}
1597 
1598 			/* We are already going to do all the verification
1599 			 * necessary at the socket layer so from our point of
1600 			 * view it is safe to turn these checks off (and less
1601 			 * work), however, the CIPSO draft says we should do
1602 			 * all the CIPSO validations here but it doesn't
1603 			 * really specify _exactly_ what we need to validate
1604 			 * ... so, just make it a sysctl tunable. */
1605 			if (READ_ONCE(cipso_v4_rbm_strictvalid)) {
1606 				if (cipso_v4_map_lvl_valid(doi_def,
1607 							   tag[3]) < 0) {
1608 					err_offset = opt_iter + 3;
1609 					goto validate_return_locked;
1610 				}
1611 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1612 				    cipso_v4_map_cat_rbm_valid(doi_def,
1613 							    &tag[4],
1614 							    tag_len - 4) < 0) {
1615 					err_offset = opt_iter + 4;
1616 					goto validate_return_locked;
1617 				}
1618 			}
1619 			break;
1620 		case CIPSO_V4_TAG_ENUM:
1621 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1622 				err_offset = opt_iter + 1;
1623 				goto validate_return_locked;
1624 			}
1625 
1626 			if (cipso_v4_map_lvl_valid(doi_def,
1627 						   tag[3]) < 0) {
1628 				err_offset = opt_iter + 3;
1629 				goto validate_return_locked;
1630 			}
1631 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1632 			    cipso_v4_map_cat_enum_valid(doi_def,
1633 							&tag[4],
1634 							tag_len - 4) < 0) {
1635 				err_offset = opt_iter + 4;
1636 				goto validate_return_locked;
1637 			}
1638 			break;
1639 		case CIPSO_V4_TAG_RANGE:
1640 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1641 				err_offset = opt_iter + 1;
1642 				goto validate_return_locked;
1643 			}
1644 
1645 			if (cipso_v4_map_lvl_valid(doi_def,
1646 						   tag[3]) < 0) {
1647 				err_offset = opt_iter + 3;
1648 				goto validate_return_locked;
1649 			}
1650 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1651 			    cipso_v4_map_cat_rng_valid(doi_def,
1652 						       &tag[4],
1653 						       tag_len - 4) < 0) {
1654 				err_offset = opt_iter + 4;
1655 				goto validate_return_locked;
1656 			}
1657 			break;
1658 		case CIPSO_V4_TAG_LOCAL:
1659 			/* This is a non-standard tag that we only allow for
1660 			 * local connections, so if the incoming interface is
1661 			 * not the loopback device drop the packet. Further,
1662 			 * there is no legitimate reason for setting this from
1663 			 * userspace so reject it if skb is NULL. */
1664 			if (!skb || !(skb->dev->flags & IFF_LOOPBACK)) {
1665 				err_offset = opt_iter;
1666 				goto validate_return_locked;
1667 			}
1668 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1669 				err_offset = opt_iter + 1;
1670 				goto validate_return_locked;
1671 			}
1672 			break;
1673 		default:
1674 			err_offset = opt_iter;
1675 			goto validate_return_locked;
1676 		}
1677 
1678 		tag += tag_len;
1679 		opt_iter += tag_len;
1680 	}
1681 
1682 validate_return_locked:
1683 	rcu_read_unlock();
1684 validate_return:
1685 	*option = opt + err_offset;
1686 	return err_offset;
1687 }
1688 
1689 /**
1690  * cipso_v4_error - Send the correct response for a bad packet
1691  * @skb: the packet
1692  * @error: the error code
1693  * @gateway: CIPSO gateway flag
1694  *
1695  * Description:
1696  * Based on the error code given in @error, send an ICMP error message back to
1697  * the originating host.  From the IETF draft ...
1698  *
1699  *  "If the contents of the CIPSO [option] are valid but the security label is
1700  *   outside of the configured host or port label range, the datagram is
1701  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1702  *   returned.  The code field of the ICMP is set to 'communication with
1703  *   destination network administratively prohibited' (code 9) or to
1704  *   'communication with destination host administratively prohibited'
1705  *   (code 10).  The value of the code is dependent on whether the originator
1706  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1707  *   recipient of the ICMP message MUST be able to handle either value.  The
1708  *   same procedure is performed if a CIPSO [option] can not be added to an
1709  *   IP packet because it is too large to fit in the IP options area."
1710  *
1711  *  "If the error is triggered by receipt of an ICMP message, the message is
1712  *   discarded and no response is permitted (consistent with general ICMP
1713  *   processing rules)."
1714  *
1715  */
1716 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1717 {
1718 	unsigned char optbuf[sizeof(struct ip_options) + 40];
1719 	struct ip_options *opt = (struct ip_options *)optbuf;
1720 	int res;
1721 
1722 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1723 		return;
1724 
1725 	/*
1726 	 * We might be called above the IP layer,
1727 	 * so we can not use icmp_send and IPCB here.
1728 	 */
1729 
1730 	memset(opt, 0, sizeof(struct ip_options));
1731 	opt->optlen = ip_hdr(skb)->ihl*4 - sizeof(struct iphdr);
1732 	rcu_read_lock();
1733 	res = __ip_options_compile(dev_net(skb->dev), opt, skb, NULL);
1734 	rcu_read_unlock();
1735 
1736 	if (res)
1737 		return;
1738 
1739 	if (gateway)
1740 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0, opt);
1741 	else
1742 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0, opt);
1743 }
1744 
1745 /**
1746  * cipso_v4_genopt - Generate a CIPSO option
1747  * @buf: the option buffer
1748  * @buf_len: the size of opt_buf
1749  * @doi_def: the CIPSO DOI to use
1750  * @secattr: the security attributes
1751  *
1752  * Description:
1753  * Generate a CIPSO option using the DOI definition and security attributes
1754  * passed to the function.  Returns the length of the option on success and
1755  * negative values on failure.
1756  *
1757  */
1758 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1759 			   const struct cipso_v4_doi *doi_def,
1760 			   const struct netlbl_lsm_secattr *secattr)
1761 {
1762 	int ret_val;
1763 	u32 iter;
1764 
1765 	if (buf_len <= CIPSO_V4_HDR_LEN)
1766 		return -ENOSPC;
1767 
1768 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1769 	 * really a good assumption to make but since we only support the MAC
1770 	 * tags right now it is a safe assumption. */
1771 	iter = 0;
1772 	do {
1773 		memset(buf, 0, buf_len);
1774 		switch (doi_def->tags[iter]) {
1775 		case CIPSO_V4_TAG_RBITMAP:
1776 			ret_val = cipso_v4_gentag_rbm(doi_def,
1777 						   secattr,
1778 						   &buf[CIPSO_V4_HDR_LEN],
1779 						   buf_len - CIPSO_V4_HDR_LEN);
1780 			break;
1781 		case CIPSO_V4_TAG_ENUM:
1782 			ret_val = cipso_v4_gentag_enum(doi_def,
1783 						   secattr,
1784 						   &buf[CIPSO_V4_HDR_LEN],
1785 						   buf_len - CIPSO_V4_HDR_LEN);
1786 			break;
1787 		case CIPSO_V4_TAG_RANGE:
1788 			ret_val = cipso_v4_gentag_rng(doi_def,
1789 						   secattr,
1790 						   &buf[CIPSO_V4_HDR_LEN],
1791 						   buf_len - CIPSO_V4_HDR_LEN);
1792 			break;
1793 		case CIPSO_V4_TAG_LOCAL:
1794 			ret_val = cipso_v4_gentag_loc(doi_def,
1795 						   secattr,
1796 						   &buf[CIPSO_V4_HDR_LEN],
1797 						   buf_len - CIPSO_V4_HDR_LEN);
1798 			break;
1799 		default:
1800 			return -EPERM;
1801 		}
1802 
1803 		iter++;
1804 	} while (ret_val < 0 &&
1805 		 iter < CIPSO_V4_TAG_MAXCNT &&
1806 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1807 	if (ret_val < 0)
1808 		return ret_val;
1809 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1810 	return CIPSO_V4_HDR_LEN + ret_val;
1811 }
1812 
1813 /**
1814  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1815  * @sk: the socket
1816  * @doi_def: the CIPSO DOI to use
1817  * @secattr: the specific security attributes of the socket
1818  *
1819  * Description:
1820  * Set the CIPSO option on the given socket using the DOI definition and
1821  * security attributes passed to the function.  This function requires
1822  * exclusive access to @sk, which means it either needs to be in the
1823  * process of being created or locked.  Returns zero on success and negative
1824  * values on failure.
1825  *
1826  */
1827 int cipso_v4_sock_setattr(struct sock *sk,
1828 			  const struct cipso_v4_doi *doi_def,
1829 			  const struct netlbl_lsm_secattr *secattr)
1830 {
1831 	int ret_val = -EPERM;
1832 	unsigned char *buf = NULL;
1833 	u32 buf_len;
1834 	u32 opt_len;
1835 	struct ip_options_rcu *old, *opt = NULL;
1836 	struct inet_sock *sk_inet;
1837 	struct inet_connection_sock *sk_conn;
1838 
1839 	/* In the case of sock_create_lite(), the sock->sk field is not
1840 	 * defined yet but it is not a problem as the only users of these
1841 	 * "lite" PF_INET sockets are functions which do an accept() call
1842 	 * afterwards so we will label the socket as part of the accept(). */
1843 	if (!sk)
1844 		return 0;
1845 
1846 	/* We allocate the maximum CIPSO option size here so we are probably
1847 	 * being a little wasteful, but it makes our life _much_ easier later
1848 	 * on and after all we are only talking about 40 bytes. */
1849 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1850 	buf = kmalloc(buf_len, GFP_ATOMIC);
1851 	if (!buf) {
1852 		ret_val = -ENOMEM;
1853 		goto socket_setattr_failure;
1854 	}
1855 
1856 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1857 	if (ret_val < 0)
1858 		goto socket_setattr_failure;
1859 	buf_len = ret_val;
1860 
1861 	/* We can't use ip_options_get() directly because it makes a call to
1862 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1863 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1864 	 * set the IPOPT_CIPSO option. */
1865 	opt_len = (buf_len + 3) & ~3;
1866 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1867 	if (!opt) {
1868 		ret_val = -ENOMEM;
1869 		goto socket_setattr_failure;
1870 	}
1871 	memcpy(opt->opt.__data, buf, buf_len);
1872 	opt->opt.optlen = opt_len;
1873 	opt->opt.cipso = sizeof(struct iphdr);
1874 	kfree(buf);
1875 	buf = NULL;
1876 
1877 	sk_inet = inet_sk(sk);
1878 
1879 	old = rcu_dereference_protected(sk_inet->inet_opt,
1880 					lockdep_sock_is_held(sk));
1881 	if (inet_test_bit(IS_ICSK, sk)) {
1882 		sk_conn = inet_csk(sk);
1883 		if (old)
1884 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1885 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1886 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1887 	}
1888 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1889 	if (old)
1890 		kfree_rcu(old, rcu);
1891 
1892 	return 0;
1893 
1894 socket_setattr_failure:
1895 	kfree(buf);
1896 	kfree(opt);
1897 	return ret_val;
1898 }
1899 
1900 /**
1901  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1902  * @req: the connection request socket
1903  * @doi_def: the CIPSO DOI to use
1904  * @secattr: the specific security attributes of the socket
1905  *
1906  * Description:
1907  * Set the CIPSO option on the given socket using the DOI definition and
1908  * security attributes passed to the function.  Returns zero on success and
1909  * negative values on failure.
1910  *
1911  */
1912 int cipso_v4_req_setattr(struct request_sock *req,
1913 			 const struct cipso_v4_doi *doi_def,
1914 			 const struct netlbl_lsm_secattr *secattr)
1915 {
1916 	int ret_val = -EPERM;
1917 	unsigned char *buf = NULL;
1918 	u32 buf_len;
1919 	u32 opt_len;
1920 	struct ip_options_rcu *opt = NULL;
1921 	struct inet_request_sock *req_inet;
1922 
1923 	/* We allocate the maximum CIPSO option size here so we are probably
1924 	 * being a little wasteful, but it makes our life _much_ easier later
1925 	 * on and after all we are only talking about 40 bytes. */
1926 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1927 	buf = kmalloc(buf_len, GFP_ATOMIC);
1928 	if (!buf) {
1929 		ret_val = -ENOMEM;
1930 		goto req_setattr_failure;
1931 	}
1932 
1933 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1934 	if (ret_val < 0)
1935 		goto req_setattr_failure;
1936 	buf_len = ret_val;
1937 
1938 	/* We can't use ip_options_get() directly because it makes a call to
1939 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1940 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1941 	 * set the IPOPT_CIPSO option. */
1942 	opt_len = (buf_len + 3) & ~3;
1943 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1944 	if (!opt) {
1945 		ret_val = -ENOMEM;
1946 		goto req_setattr_failure;
1947 	}
1948 	memcpy(opt->opt.__data, buf, buf_len);
1949 	opt->opt.optlen = opt_len;
1950 	opt->opt.cipso = sizeof(struct iphdr);
1951 	kfree(buf);
1952 	buf = NULL;
1953 
1954 	req_inet = inet_rsk(req);
1955 	opt = xchg((__force struct ip_options_rcu **)&req_inet->ireq_opt, opt);
1956 	if (opt)
1957 		kfree_rcu(opt, rcu);
1958 
1959 	return 0;
1960 
1961 req_setattr_failure:
1962 	kfree(buf);
1963 	kfree(opt);
1964 	return ret_val;
1965 }
1966 
1967 /**
1968  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
1969  * @opt_ptr: IP option pointer
1970  *
1971  * Description:
1972  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
1973  * adjustments to the IP option structure.  Returns zero on success, negative
1974  * values on failure.
1975  *
1976  */
1977 static int cipso_v4_delopt(struct ip_options_rcu __rcu **opt_ptr)
1978 {
1979 	struct ip_options_rcu *opt = rcu_dereference_protected(*opt_ptr, 1);
1980 	int hdr_delta = 0;
1981 
1982 	if (!opt || opt->opt.cipso == 0)
1983 		return 0;
1984 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
1985 		u8 cipso_len;
1986 		u8 cipso_off;
1987 		unsigned char *cipso_ptr;
1988 		int iter;
1989 		int optlen_new;
1990 
1991 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
1992 		cipso_ptr = &opt->opt.__data[cipso_off];
1993 		cipso_len = cipso_ptr[1];
1994 
1995 		if (opt->opt.srr > opt->opt.cipso)
1996 			opt->opt.srr -= cipso_len;
1997 		if (opt->opt.rr > opt->opt.cipso)
1998 			opt->opt.rr -= cipso_len;
1999 		if (opt->opt.ts > opt->opt.cipso)
2000 			opt->opt.ts -= cipso_len;
2001 		if (opt->opt.router_alert > opt->opt.cipso)
2002 			opt->opt.router_alert -= cipso_len;
2003 		opt->opt.cipso = 0;
2004 
2005 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2006 			opt->opt.optlen - cipso_off - cipso_len);
2007 
2008 		/* determining the new total option length is tricky because of
2009 		 * the padding necessary, the only thing i can think to do at
2010 		 * this point is walk the options one-by-one, skipping the
2011 		 * padding at the end to determine the actual option size and
2012 		 * from there we can determine the new total option length */
2013 		iter = 0;
2014 		optlen_new = 0;
2015 		while (iter < opt->opt.optlen)
2016 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2017 				iter += opt->opt.__data[iter + 1];
2018 				optlen_new = iter;
2019 			} else
2020 				iter++;
2021 		hdr_delta = opt->opt.optlen;
2022 		opt->opt.optlen = (optlen_new + 3) & ~3;
2023 		hdr_delta -= opt->opt.optlen;
2024 	} else {
2025 		/* only the cipso option was present on the socket so we can
2026 		 * remove the entire option struct */
2027 		*opt_ptr = NULL;
2028 		hdr_delta = opt->opt.optlen;
2029 		kfree_rcu(opt, rcu);
2030 	}
2031 
2032 	return hdr_delta;
2033 }
2034 
2035 /**
2036  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2037  * @sk: the socket
2038  *
2039  * Description:
2040  * Removes the CIPSO option from a socket, if present.
2041  *
2042  */
2043 void cipso_v4_sock_delattr(struct sock *sk)
2044 {
2045 	struct inet_sock *sk_inet;
2046 	int hdr_delta;
2047 
2048 	sk_inet = inet_sk(sk);
2049 
2050 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2051 	if (inet_test_bit(IS_ICSK, sk) && hdr_delta > 0) {
2052 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2053 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2054 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2055 	}
2056 }
2057 
2058 /**
2059  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2060  * @req: the request socket
2061  *
2062  * Description:
2063  * Removes the CIPSO option from a request socket, if present.
2064  *
2065  */
2066 void cipso_v4_req_delattr(struct request_sock *req)
2067 {
2068 	cipso_v4_delopt(&inet_rsk(req)->ireq_opt);
2069 }
2070 
2071 /**
2072  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2073  * @cipso: the CIPSO v4 option
2074  * @secattr: the security attributes
2075  *
2076  * Description:
2077  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2078  * on success and negative values on failure.
2079  *
2080  */
2081 int cipso_v4_getattr(const unsigned char *cipso,
2082 		     struct netlbl_lsm_secattr *secattr)
2083 {
2084 	int ret_val = -ENOMSG;
2085 	u32 doi;
2086 	struct cipso_v4_doi *doi_def;
2087 
2088 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2089 		return 0;
2090 
2091 	doi = get_unaligned_be32(&cipso[2]);
2092 	rcu_read_lock();
2093 	doi_def = cipso_v4_doi_search(doi);
2094 	if (!doi_def)
2095 		goto getattr_return;
2096 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2097 	 * really a good assumption to make but since we only support the MAC
2098 	 * tags right now it is a safe assumption. */
2099 	switch (cipso[6]) {
2100 	case CIPSO_V4_TAG_RBITMAP:
2101 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2102 		break;
2103 	case CIPSO_V4_TAG_ENUM:
2104 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2105 		break;
2106 	case CIPSO_V4_TAG_RANGE:
2107 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2108 		break;
2109 	case CIPSO_V4_TAG_LOCAL:
2110 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2111 		break;
2112 	}
2113 	if (ret_val == 0)
2114 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2115 
2116 getattr_return:
2117 	rcu_read_unlock();
2118 	return ret_val;
2119 }
2120 
2121 /**
2122  * cipso_v4_sock_getattr - Get the security attributes from a sock
2123  * @sk: the sock
2124  * @secattr: the security attributes
2125  *
2126  * Description:
2127  * Query @sk to see if there is a CIPSO option attached to the sock and if
2128  * there is return the CIPSO security attributes in @secattr.  This function
2129  * requires that @sk be locked, or privately held, but it does not do any
2130  * locking itself.  Returns zero on success and negative values on failure.
2131  *
2132  */
2133 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2134 {
2135 	struct ip_options_rcu *opt;
2136 	int res = -ENOMSG;
2137 
2138 	rcu_read_lock();
2139 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2140 	if (opt && opt->opt.cipso)
2141 		res = cipso_v4_getattr(opt->opt.__data +
2142 						opt->opt.cipso -
2143 						sizeof(struct iphdr),
2144 				       secattr);
2145 	rcu_read_unlock();
2146 	return res;
2147 }
2148 
2149 /**
2150  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2151  * @skb: the packet
2152  * @doi_def: the DOI structure
2153  * @secattr: the security attributes
2154  *
2155  * Description:
2156  * Set the CIPSO option on the given packet based on the security attributes.
2157  * Returns a pointer to the IP header on success and NULL on failure.
2158  *
2159  */
2160 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2161 			    const struct cipso_v4_doi *doi_def,
2162 			    const struct netlbl_lsm_secattr *secattr)
2163 {
2164 	int ret_val;
2165 	struct iphdr *iph;
2166 	struct ip_options *opt = &IPCB(skb)->opt;
2167 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2168 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2169 	u32 opt_len;
2170 	int len_delta;
2171 
2172 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2173 	if (ret_val < 0)
2174 		return ret_val;
2175 	buf_len = ret_val;
2176 	opt_len = (buf_len + 3) & ~3;
2177 
2178 	/* we overwrite any existing options to ensure that we have enough
2179 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2180 	 * that the security label is applied to the packet - we do the same
2181 	 * thing when using the socket options and it hasn't caused a problem,
2182 	 * if we need to we can always revisit this choice later */
2183 
2184 	len_delta = opt_len - opt->optlen;
2185 	/* if we don't ensure enough headroom we could panic on the skb_push()
2186 	 * call below so make sure we have enough, we are also "mangling" the
2187 	 * packet so we should probably do a copy-on-write call anyway */
2188 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2189 	if (ret_val < 0)
2190 		return ret_val;
2191 
2192 	if (len_delta > 0) {
2193 		/* we assume that the header + opt->optlen have already been
2194 		 * "pushed" in ip_options_build() or similar */
2195 		iph = ip_hdr(skb);
2196 		skb_push(skb, len_delta);
2197 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2198 		skb_reset_network_header(skb);
2199 		iph = ip_hdr(skb);
2200 	} else if (len_delta < 0) {
2201 		iph = ip_hdr(skb);
2202 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2203 	} else
2204 		iph = ip_hdr(skb);
2205 
2206 	if (opt->optlen > 0)
2207 		memset(opt, 0, sizeof(*opt));
2208 	opt->optlen = opt_len;
2209 	opt->cipso = sizeof(struct iphdr);
2210 	opt->is_changed = 1;
2211 
2212 	/* we have to do the following because we are being called from a
2213 	 * netfilter hook which means the packet already has had the header
2214 	 * fields populated and the checksum calculated - yes this means we
2215 	 * are doing more work than needed but we do it to keep the core
2216 	 * stack clean and tidy */
2217 	memcpy(iph + 1, buf, buf_len);
2218 	if (opt_len > buf_len)
2219 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2220 	if (len_delta != 0) {
2221 		iph->ihl = 5 + (opt_len >> 2);
2222 		iph_set_totlen(iph, skb->len);
2223 	}
2224 	ip_send_check(iph);
2225 
2226 	return 0;
2227 }
2228 
2229 /**
2230  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2231  * @skb: the packet
2232  *
2233  * Description:
2234  * Removes any and all CIPSO options from the given packet.  Returns zero on
2235  * success, negative values on failure.
2236  *
2237  */
2238 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2239 {
2240 	int ret_val;
2241 	struct iphdr *iph;
2242 	struct ip_options *opt = &IPCB(skb)->opt;
2243 	unsigned char *cipso_ptr;
2244 
2245 	if (opt->cipso == 0)
2246 		return 0;
2247 
2248 	/* since we are changing the packet we should make a copy */
2249 	ret_val = skb_cow(skb, skb_headroom(skb));
2250 	if (ret_val < 0)
2251 		return ret_val;
2252 
2253 	/* the easiest thing to do is just replace the cipso option with noop
2254 	 * options since we don't change the size of the packet, although we
2255 	 * still need to recalculate the checksum */
2256 
2257 	iph = ip_hdr(skb);
2258 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2259 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2260 	opt->cipso = 0;
2261 	opt->is_changed = 1;
2262 
2263 	ip_send_check(iph);
2264 
2265 	return 0;
2266 }
2267 
2268 /*
2269  * Setup Functions
2270  */
2271 
2272 /**
2273  * cipso_v4_init - Initialize the CIPSO module
2274  *
2275  * Description:
2276  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2277  * and negative values on failure.
2278  *
2279  */
2280 static int __init cipso_v4_init(void)
2281 {
2282 	int ret_val;
2283 
2284 	ret_val = cipso_v4_cache_init();
2285 	if (ret_val != 0)
2286 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2287 		      ret_val);
2288 
2289 	return 0;
2290 }
2291 
2292 subsys_initcall(cipso_v4_init);
2293